The R136S mutation in the APOE3 gene confers resilience against tau pathology via inhibition of the cGAS-STING-IFN pathway

Summary:

The Christchurch mutation (R136S) in the APOE3 (E3S/S) gene is associated with attenuated tau load and cognitive decline despite the presence of a causal PSEN1 mutation and high amyloid burden in the carrier. However, the molecular mechanisms enabling the E3S/S mutation to mitigate tau-induced neurodegeneration remain unclear. Here, we replaced mouse Apoe with wild-type human APOE3 or APOE3S/S on a tauopathy background. The R136S mutation decreased tau load and protected against tau-induced synaptic loss, myelin loss, and reduction in hippocampal theta and gamma powers. Additionally, the R136S mutation reduced interferon responses to tau pathology in both mouse and human microglia, suppressing cGAS-STING pathway activation. Treating E3 tauopathy mice with a cGAS inhibitor protected against tau-induced synaptic loss and induced transcriptomic alterations similar to the R136S mutation across brain cell types. Thus, suppression of the microglial cGAS-STING-IFN pathway plays a central role in mediating the protective effects of R136S against tauopathy.

Graphical Abstract

eTOC

The APOE3 Christchurch mutation (R136S) reduces tau pathology despite PSEN1 mutation and high amyloid, though its protective mechanism remains unclear. Naguib et al. show that R136S suppresses microglial cGAS-STING-IFN signaling in tauopathy mice. Inhibiting cGAS preserves synapses and mimics R136S-linked gene expression, suggesting that cGAS suppression is a key part of the underlying protective effects of R136S.

Introduction

Alzheimer’s disease (AD) is the most prevalent form of dementia, characterized by pathological amyloid beta (Aβ) plaques and neurofibrillary tau tangles¹. Amyloid pathology typically develops first followed by tau pathology, which correlates with AD progression and cognitive impairment^2–4. Most AD patients have sporadic late-onset AD (LOAD), in which the apolipoprotein E (APOE) gene contains a predominant risk variant^5,6. APOE is a lipid transporter expressed in astrocytes and microglia and has been shown to affect microglial functions including neuroinflammation and phagocytosis^7–9. The three major APOE variants, APOE2, APOE3, and APOE4, differ at two amino acid positions, 112 and 158. E3 is the most common APOE variant and does not alter AD risk, while E4 elevates AD risk and E2 reduces AD risk^12,13. Other variants in APOE have been discovered, such as the APOE3-Jacksonville mutation, which reduces amyloid toxicity¹⁰.

A recent study revealed that a patient carrying the familial PSEN1 E280A mutation, which causes autosomal dominant AD as early as 30–40 years of age¹¹, was protected against cognitive decline until her 70’s, a benefit attributed to homozygosity for the R136S or Christchurch mutation on APOE3 background (E3S/S)¹². Despite high amyloid burden, the carrier exhibited low tau pathology in AD-vulnerable regions, frontal and occipital cortices¹³, suggesting protective ability of the R136S mutation against tau-induced neurodegeneration. In mouse models on the risk-enhancing E4 background, the R136S mutation protected against tau pathology, neurodegeneration, and neuroinflammation¹⁴. Additionally, E3S/S also protected against toxicity induced by inoculation of AD-tau in a transgenic amyloid mouse model¹⁵. Thus, it appears the protective effects of R136S mutation are likely relevant to both E3 and E4 carriers. Furthermore, identifying the mechanisms underlying R136S-mediated protection may elucidate potential therapies applicable to other tauopathies.

Our current study aims to dissect the mechanisms underlying the protective function of E3S/S against tauopathy. We generated human E3 and E3S/S knockin models by replacing mouse Apoe alleles with human E3 and E3S/S cDNA, followed by crossing with the well-established P301S mouse model¹⁶. We characterized the effects of E3S/S on tau pathology, synaptic loss, network dysfunction, and demyelination. We performed snRNA-seq to identify cell type-specific transcriptomic changes, and discovered that E3S/S reduced cGAS-STING-IFN signaling in microglia without affecting disease-associated microglial signatures (DAMs)¹⁷ or microglial neurodegenerative phenotype (MGnD)¹⁸. Downregulation of cGAS-STING-IFN signaling in microglia was confirmed in both mouse microglia and human iPSC-derived microglial-like cells (iMGLs) following tau treatment. We also treated E3/P301S mice with a cGAS inhibitor to determine whether protective function of the R136S mutation can be attributed to cGAS inhibition; thus, we compared cGAS inhibitor profiles with transcriptional signatures induced by the R136S mutation across several cell types. Our study revealed a central role of dampened cGAS-STING-IFN activation in R136S-induced resilience.

Results

APOE3S/S reduces tau inclusions and protects against tau-induced loss of synapses.

To directly assess the effects of the R136S mutation on human E3 background, we used CRISPR/Cas9 strategies to insert human E3 cDNA and APOE3R136S cDNA (E3S/S) into the exon of the mouse Apoe locus, resulting in replacement of mouse Apoe with human E3 or E3S/S (Fig. 1A). PCR confirmed the correct recombination and insertion of human E3 or E3S/S cDNA at the mouse Apoe locus (Fig. S1 A–C). Mouse lines without nonspecific integration for both E3 and E3S/S were expanded, and Sanger sequencing was performed to confirm the replacement accuracy (Fig. S1 B, C). We then crossed E3 or E3S/S mice with mice expressing the human MAPT transgene with P301S mutation (P301S), to study effects of E3S/Son tau pathology in the absence of amyloid, hereafter referred to as E3/P301S or E3S/S/P301S mice¹⁶ (Fig. 1A). E3S/S did not affect APOE expression in the frontal cortex or plasma in males (Fig. 1B, 1C, S1D). However, E3S/Sled to a modest increase in frontal cortex APOE expression in females (Fig. S1E–F).

Figure 1. ApoE3S/S mutation reduces tau inclusions and protects against tau-induced loss of synapses.

(A) Schematic illustrating knockin of human APOE3 or APOE3R136S on the mouse Apoe locus, and cross with P301S mice. See also Figure S1.

(B) Representative western blot for APOE in frontal cortex lysates with 4–5 animals/group, 1 independent experiment. APOE knockout (KO) frontal cortex was used as a control. See also Figure S1.

(C) Quantification of APOE western blot of frontal cortex lysates, showing no differences between groups. Data are reported as mean ± SEM. Data were analyzed by one-way ANOVA with Tukey’s multiple comparison test.

(D) Representative immunofluorescence images of whole hippocampus of E3/P301S and E3S/S/P301S mice labeled with MC-1 antibody (red); scale bar, 500um.

(E) Quantification of MC-1 immunofluorescence intensities, showing decreased MC-1 throughout E3S/S/P301S mice. Results are presented as average intensity measures from 3–4 sections per animal with 8 animals/experimental group, 1 independent experiment. Data are reported as mean ± SEM. **p=0.0053, unpaired student’s t-test.

(F) Representative immunofluorescence images of CA3 and CA1 subregions of hippocampus of E3/P301S and E3S/S/P301S mice respectively labeled with MC-1 (red); scale bar, 50um.

(G) Quantification of MC-1 immunofluorescence intensities in CA1 and CA3. Results are presented as average intensity measures from 3–4 sections per animal with 8 animals/experimental group, 1 independent experiment. Data are reported as mean ± SEM. **p= 0.0013 and **p=0.0011 for CA1 and CA3 respectively. Data were analyzed by unpaired student’s t-test.

(H) Representative immunofluorescence images of CA1 subregion of hippocampus of E3/P301S and E3S/S/P301S mice respectively labeled with PSD-95 (red); scale bar, 50um.

(I) Quantification of PSD-95 immunofluorescence intensities in CA1. Results are presented as average intensity measures from 3–4 sections per animal with 8 animals/experimental group, 1 independent experiment. Data are reported as mean ± SEM. ****p<0.0001; *p=0.0291. n.s, not significant. Data were analyzed by two-way ANOVA with mixed-effects model.

(J) Representative western blot for synaptophysin and GAPDH in frontal cortex lysate from all four experimental groups.

(K) Quantification of synaptophysin expression normalized to GAPDH in western blot of frontal cortex lysates with 4–5 animals/group, 1 independent experiment. Data are reported as mean ± SEM. **p = 0.0095 for E3/P301S compared to E3S/S/P301S. Data were analyzed by one-way ANOVA with Tukey’s multiple comparison test.

We next determined the effect of R136S on tau pathology in 9–10-month-old male E3/P301S and E3S/S/P301S mice. Immunostaining for MC-1, a conformation-specific antibody for aggregated tau¹⁹, indicated a marked decrease in tau load across the hippocampus in E3S/S/P301S mice, consistent with previous studies (Fig. 1D–G)¹⁴. We measured synaptic integrity and found higher PSD95 immunofluorescence in E3S/S/P301S mice compared to E3/P301S mice (Fig 1. H, I). R136S also mitigated tau-induced reduction of synaptophysin in the frontal cortex (Fig. 1J–K).

APOE3S/Sprotects against tau-induced loss of theta and gamma power in hippocampus and cortex.

To dissect how E3S/S affects network activity in tauopathy mice, we performed 16 channel local field potential (LFP) recordings in awake 6–7-month-old male mice exploring an open field chamber (Fig. 2A). Animals’ overall speed did not differ across experimental groups (Fig. 2B). LFP recording measures amplitudes of oscillations across 0.5–250HZ (Fig. 2C). Previous studies have shown that tau pathology can interfere with neuronal circuits, resulting in theta and gamma oscillation deficits long before neurons die^16,20–23. We quantified theta and gamma power in the somatosensory cortex, visual cortex, hippocampal CA1, and hippocampal dentate gyrus (DG) regions. The average amplitudes of theta power across all brain regions were reduced by tau pathology in mice with the E3 background, but not in those with the E3S/S background, particularly in the DG (Fig. 2D–F). In contrast, tau pathology decreased average gamma power across all brain regions in both E3 and E3S/S mice (Fig. 2G). However, E3S/S mice did not exhibit tau-induced reduction of gamma power in the DG (Fig. 2H, I).

Figure 2: ApoE3S/S protects against tau-induced loss of theta and gamma power in hippocampus.

(A) Schematic of local field potential (LFP) recordings in male 6–7 month old mice in open field chamber. See also Figure S2.

(B) Maximum speed (mm/s) calculated during LFP recordings from all experimental groups. N= 5 for E3, n=6 for E3/P301S, n= 5 for E3S/S, n= 7 E3S/S/P301S, 1 independent experiment for all LFP data. Data were analyzed by one-way ANOVA.

(C) Representative traces of LFP band, theta band and gamma band respectively for all 16 channels in four brain regions: somatosensory cortex (S1), CA1, DG of hippocampus and visual cortex (V1). Scale bar represents both amplitude and time (1mV/0.1sec).

(D) Quantification of average theta power in all brain regions showing tau-induced reduction in E3/P301S mice. n= 5 for E3, n=6 for E3/P301S, n= 5 for E3S/S, n= 7 E3S/S/P301S. *p=0.0163.

(E) Average theta power from DG in all four experimental groups.

(F) Quantification of average theta power in DG showing tau-induced reduction in E3/P301S mice. n= 5 for E3, n=6 for E3/P301S, n= 5 for E3S/S, n= 7 E3S/S/P301S. **p=0.0042.

(G) Quantification of average gamma power in all brain regions across all four groups. n= 5 for E3, n=6 for E3/P301S, n= 5 for E3S/S, n= 7 E3S/S/P301S. **p=0.0016, *p=0.0307.

(H) Average gamma power from DG.

(I) Quantification of gamma power in DG. n= 5 for E3, n=6 for E3/P301S, n= 5 for E3S/S, n= 7 E3S/S/P301S. **p=0.0061, ****p<0.0001. Data were analyzed by two-way ANOVA, Tukey’s multiple comparison test (D, F, G, I).

Expression of c-Fos, an immediate early-gene marker, is a well-established indicator of mouse brain activity²⁴. To examine whether E3S/S modulates neuronal activity, we performed 3D c-Fos mapping. 8-month-old male mice from all experimental groups were exposed to an unfamiliar open field for 10 minutes and returned to their home cages for 45 minutes, followed by collection for whole-brain c-Fos labeling, clearing, and light-sheet imaging (Fig. S2A). Tauopathy increased c-Fos-based brain activity on the E3 background, but this was reversed on the E3S/S background (Fig. S2B–C). Detailed mapping revealed that key regions affected by E3S/S include the CA1 pyramidal layer, CA1 stratum oriens, dorsal subiculum, postsubiculum, parasubiculum, presubiculum, and entorhinal cortex (Fig. S2D). These critical regions for memory retention and retrieval are highly vulnerable to Aβ and tau pathology^25,26,27. Protection from the R136S mutation against tau-induced network dysfunction and hyperexcitability is consistent with delayed cognitive impairment observed in the patient¹.

APOE3S/S induces cell type-specific transcriptomic changes in hippocampi of tauopathy mice

We next examined the cell type-specific effects underlying R136S-mediated protection. We performed snRNA-seq of hippocampi from male mice in all four genotypes (E3, E3/P301S, E3S/S, and E3S/S/P301S) at 9–10 months of age. 88,605 nuclei passed stringent quality control (QC) (Fig. S3). All major cell types were similarly represented across genotypes (Fig. S3D). The R136S mutation in tauopathy mice led to transcriptomic alterations in multiple cell types, including excitatory neurons (EN), inhibitory neurons (IN), astrocytes (AST), oligodendrocytes (OL), microglia (MG), and choroid plexus cells (CHOR) (Fig. S3H, Table S1). Given that only 1,602 CHOR cells were captured, we focused primarily on other affected cell types including EN, IN, AST, OL, and MG (Fig. S3H, Table S1).

APOE is expressed by both astrocytes and microglia, but primarily astrocytes. Immunostaining for GFAP revealed that tau-induced astrogliosis was not affected by R136S mutation (Fig. S4A, B). Transcriptomic analyses revealed that the R136S mutation downregulated Thioesterase Superfamily Member 4 (Them4), which modulates mitochondrial fatty acid metabolism and is involved in apoptosis (Fig. S4C, Table S1). Them4 has been shown to regulate PI3K-AKT1 activity, which is involved in toxicity associated with AD risk allele TREM2^28–30. Another top downregulated gene in astrocytes was metabotropic glutamate receptor 5 (Grm5) (Fig. S4C, Table S1). Grm5 blockade in human astrocytes reduced pro-inflammatory responses to TNF-α³¹. Among the top upregulated genes are prominent AD risk genes Apoe and Clu³² (Fig. S4C, Table S1). Astrocytic Clu overexpression rescued synaptic deficits in an amyloid mouse model³³. Other upregulated genes, including a glial high-affinity glutamate transporter (Slc1a2), play a role in synaptic clearance of glutamate, thereby preventing excitotoxicity³⁴ (Fig. S4C, Table S1).

In ENs, the top DEGs were involved in spliceosome assembly and function pathways, including Srsf2, Srsf7, and Srsf11, Rbm25, Rbm39, Zranb2, Rsrp1, Luc7l3 & Luc7l2 (Fig. S4D, Table S1). Using gene set enrichment analysis (GSEA), we found that upregulated genes in ENs also were involved in regulation of membrane potential, inorganic ion transmembrane transport, and RNA processing (Pnn, Pnisr) (Fig. S4E). The primary downregulated pathways were related to neuronal development and organization (Fig. S4F). Although INs exhibited fewer DEGs, the top pathways were also associated with RNA processing (Fig. S4G, Table S1). Comparing the top 50 DEGs between E3S/S/P301S and E3/P301S across excitatory and inhibitory neurons showed that splicing-related genes were shared and predominantly upregulated in both cell types (Fig. S4H, Table S1). Within these shared genes, STRING analysis identified processing of capped intron-containing pre-mRNA, including RBM25, as an enriched pathway in Reactome (14/271 genes in pathway, enrichment score: 2.81081, FDR = 0.0062) (Fig. S4I). To validate increased expression of Rbm25 in E3S/S/P301S in protein, we co-immunolabeled for RBM25 and NeuN, confirming that E3S/S/P301S neurons had higher expression of RBM25 compared to E3/P301S neurons (Fig. S4J–K). These findings are consistent with previous studies showing that RBM25 is downregulated in tangle-bearing AT8+ neurons in AD brains³⁵.

APOE3S/S induces transcriptomic changes in oligodendrocyte, lipid biosynthesis and protects against tau-induced myelin loss.

In oligodendrocytes, pseudobulk analysis identified upregulated expression of myelination-associated markers, including Myelin Oligodendrocyte Glycoprotein (Mog) (1.36-fold increase, p = 2.11E-73) and Opalin (1.28-fold increase, p = 3.55E-16), in E3S/S/P301S hippocampi compared with E3/P301S counterparts (Fig. 3A, Table S1). Subclustering analyses revealed OL1–OL5 as primary subclusters, with few cells in subclusters OL6 and OL7 (Fig. 3C). Subcluster 2 (OL2) was reduced by tau in E3 but preserved in E3S/Soligodendrocytes (Fig. 3C–D, Table S2). OL2 upregulated several genes associated with myelination including Pprd, Mobp, and Adgrv1³⁶ (Fig. 3E). Pathways upregulated by this subcluster were involved in synthesis of several lipid subtypes, including glycosphingolipids and sphingolipids³⁷ (Fig. 3F). To determine how R136S affects global lipid synthesis in tauopathy, we conducted lipid profiling on E3S/S/P301S and E3/P301S hippocampi (Fig. 3G–I). E3S/S/P301S hippocampi showed increases in total glycosphingolipids (Hex1Cer, Hex2Cer) and myelin (SM, ST) lipids compared with E3/P301S brains. In contrast, expression of complex gangliosides GD1a and GT1 found to be increased in progranulin deficient model of frontotemporal dementia (FTD)³⁸, were reduced in E3S/S/P301S hippocampi (Fig. 3H). E3S/S/P301S mice also exhibit elevated alkyl-phosphatidylcholine (PC), phosphatidylethanolamine (PE), and phosphatidylserine (PS) (Fig. 3I). Immunohistochemical analysis of myelin basic protein (MBP) confirmed myelin integrity preservation, with E3S/S/P301S mice exhibiting greater MBP+ area in hippocampal white matter compared with E3/P301S mice (p < 0.001, Fig. 3J–K).

Figure 3. ApoeE3S/S induces transcriptomic changes in oligodendrocyte, alters lipid biosynthesis, and protects against tau-induced myelin loss.

(A) Volcano plot of DEGs in oligodendrocytes between E3S/S/P301S and E3/P301S. Dashed lines represent log2foldchange threshold of +/− 0.1 and adjusted p value threshold of 0.05. n=3 hippocampi/group, 1 independent experiment for all snRNAseq data. See also Figures S3 and S4 and Tables S1 and S2.

(B) UMAP of oligodendrocyte subclusters split by genotype.

(C) Quantification of cell ratios within each subcluster, genotype, and sample.

(D) Quantification of average cell ratios in each subcluster by genotype.

(E) Volcano plot of oligodendrocyte cluster 2 (OL2) markers. Dashed lines represent log2foldchange threshold of +/− 0.1 and adjusted p value threshold of 0.05.

(F) Gene ontology pathway analysis of upregulated DEGs from oligodendrocyte subcluster 2 in E3S/S/P301S vs. E3/P301S.

(G–I) Quantification of glycosphingolipids and myelin lipids (I), ganglioside species (J), and lipid composition (K) using LCMS lipidomic analyses. The lipid classes are presented as mean ± SEM. N=6 hippocampi/group, 1 independent experiment for all lipidomic data. *p < 0.05, **p < 0.01, ***p < 0.001. Data were analyzed by multiple non-parametric Welch t-tests.

(G) E3S/S/P301S mice exhibit higher glycosphingolipids and myelin lipids relative to E3/P301S. Ceramide (Cer), Hexosylceramide (Hex1Cer), Dihexosylceramide (Hex2Cer), Sulfatide (ST), Sphingomyelin (SM), and Sphingosine (SPH).

(H) E3S/S/P301S mice exhibit significant reduction in the fold change of total GD1a and GQ1b. GM3:monosialodihexosylganglioside, GM2:monosialoganglioside, GM1:monosialotetrahexosylganglioside, GD2:disialoganglioside, GQ1:tetrasialotetrahexosylganglioside, and GT1:trisialotetrahexosylganglioside.

(I) E3S/S/P301S mice exhibit elevated PC-O, PE, and PC-P. cCa(acylcarnitine), BMP: bis(monoacylglycerol)phosphate, CE(cholesterol ester), DAG(diacylglycerol), LPC: lysophosphatidylcholine, LPE:lysophosphatidylethanolamine, LPI:lysophosphatidylinositol, LPS: lysophosphatidylserine, PA:phosphatidic acid, PC:phosphatidylcholine, PC-O:alkyl-phosphatidylcholine, PE:phosphatidylethanolamine, PE(O): alkyl-phosphatidylethanolamines, PG: phosphatidylglycerol, PI: phosphatidylinositol, PS: phosphatidylserine, TAG: triacylglycerol.

(J) Representative images of immunofluorescence for MBP across each genotype.

(K) Quantification of MBP signal across genotypes, n=8 mice per genotype with 3–4 sections per mouse, 1 independent experiment. *p=0.0484, **p<0.01, ****, p<0.0001. Data shown as mean ± SEM and analyzed via two-way ANOVA with Sidak’s post-hoc test.

APOE3S/S diminishes tau-induced morphological and IFN responses in microglia.

Previous studies showed striking APOE isoform-specific effects on microglia activation³⁹, and microglia play critical roles in mediating tau toxicity^9,40. The E3S/S mutation did not alter tau-induced microgliosis (Fig. 4A–B). Using IMARIS 3D reconstruction of microglial morphology, we found that tau increased soma volume in E3/P301S microglia, while the E3S/S mutation prevented this increase (Fig. 4C–D). In addition, E3S/S mitigated tau-induced decrease in microglial branching, resulting in a more ramified morphology compared to E3/P301S counterparts (Fig. 4C, E), consistent with attenuation of activation.

Figure 4. ApoES/S rescues tau-induced morphological changes and suppresses IFN responses in microglia.

(A) Representative images of IBA1 immunofluorescence from CA1 region of hippocampus in all four experimental groups. Scale bar represents 50um.

(B) Quantification of total IBA1 intensity in CA1 region. ***p=0.0003, ****p<0.0001. Results are presented as average intensity measures from 3–4 sections per animal with 8 animals/experimental group, 1 independent experiment. Data are reported as mean ± SEM and were analyzed by two-way ANOVA with mixed-effects model.

(C) Representative IMARIS 3D reconstructed microglia from all experimental groups. Scale bar represents 10um.

(D) Quantification of soma volume size using IMARIS 3D reconstruction in all experimental groups, showing tau-induced increase in soma volume in E3 microglia that is not present with R136S mutation. Each dot represents one microglia. N=197 for E3, n=418 for E3/P301S, n=262 for E3S/S, n=376 for E3S/S/P301S, from 8 mice/genotype and 4 hippocampal sections/mouse. *p<0.05. Data are reported as mean ± SEM and were analyzed by linear mixed model.

(E) Quantification of number of microglial branches using IMARIS 3D reconstruction in all experimental groups. Each dot represents one microglia. N=197 for E3, n=418 for E3/P301S, n=262 for E3S/S, n=376 microglia for E3S/S/P301S, from 8 mice /genotype and 4 hippocampal sections/mouse. *p<0.05. Data are reported as mean ± SEM and were analyzed by linear mixed model.

(F) Quantification of cell ratios within each microglial subcluster and sample. See also Figure S3 and S4 and Tables S1 and S3. n=3 hippocampi/group, 1 independent experiment for all snRNAseq data.

(G) Quantification of average cell ratios for each microglial subcluster by genotype.

(H) Correlation between microglia cluster 2 gene expression and published DAM gene signature from Keren-Shaul et al. 2017

(I) Dotplot DAM gene expression across four genotypes.

(J) Volcano plot of DEGs in excitatory neurons between E3S/S/P301S and E3/P301S.

(K) Dotplot of interferon gene expression by genotype.

(L) Upstream regulators identified by Ingenuity Pathway Analysis for both up- and downregulated genes by E3S/S

(M) Representative immunofluorescence images of CA1 subregion of hippocampus of E3, E3/P301S, E3S/S, E3S/S/P301S groups co-labeled with IBA1 (green) and STING (red); scale bar, 50um. The arrows indicate colocalization.

(N) Quantification of co-localized integrated density of STING in IBA1+ cells. Results are presented as average intensity measures from 3–4 sections per animal with 8 animals/genotype, 1 independent experiment. *p=0.0157 for E3 compared to E3/P301S. Data were analyzed by two-way ANOVA with mixed-effects model. Data are reported as mean ± SEM.

SnRNA-seq revealed that microglia in E3/P301S and E3S/S/P301S mice exhibited similar loss of microglial subcluster 1 (MG1), enriched in homeostatic markers, and similar induction of subcluster 2 (MG2) (Fig. 4F–G, Table S3). MG2 strongly correlated with DAM markers (Fig. 4H). Expression of DAM genes elevated by tauopathy did not differ between E3/P301S and E3S/S/P301S microglia (Fig. 4I). Pseudobulk analyses revealed numerous interferon genes downregulated by E3S/S/P301S microglia, including Trim30a, Stat1 and Ifi204 (Fig. 4J, Table S1). Indeed, E3/P301S microglia demonstrated a robust type I interferon response to tau, which was markedly blunted in E3S/S/P301S counterparts (Fig 4K, Table S1). Indeed, top inhibited upstream regulators in E3S/S/P301S microglia included antiviral activators as cGAS and IFNA, and top activated upstream regulators included antiviral suppressors TREX1 and RNASEH2B (Fig. 4L, Table S3). The cGAS-STING pathway was previously shown to play a role in tau-induced neurodegeneration⁴¹. Immunostaining for STING and IBA1 confirmed that STING induction by tau occurred solely in E3 microglia (Fig. 4M–N), consistent with the R136S mutation dampening cGAS-STING-associated interferon response to tauopathy.

APOE3S/S accelerates tau processing and suppresses tau- and Aβ-induced cGAS-STING-IFN response in primary microglia.

Our data thus far indicated decreased tau load and suppression of microglia cGAS-STING-IFN response in E3S/S/P301S mice. To dissect how the R136S mutation affects microglial tau processing, we isolated microglia from brains of E3 mice and performed a tau pulse-chase assay. Microglia were treated with 0N4R tau fibrils and analyzed at two timepoints: 2 hours post-treatment and 24 hours after extracellular tau supply was removed (Fig. 5A). At 2 hours, E3S/S microglia contained more intracellular tau than E3 microglia; by 24 hours, E3S/S microglia exhibited less intracellular tau, indicating enhanced tau processing and/or degradation (Fig. 5B–C). These findings suggest that E3S/S microglia show increased tau uptake and clearance, consistent with prior studies^15,42.

Figure 5. ApoE3S/S accelerates tau processing and suppresses tau and Aß induced cGAS-STING-IFN response in primary microglia.

(A) Schematic for tau chasing assay experiment.

(B) Representative immunofluorescence images of E3 and E3S/S microglia at 2 hours and 24 hours following tau exposure. Red: anti-tau antibody. Scale bar represents 50um.

(C) Quantification of tau immunolabeling in red fluorescence normalized to number of nuclei/image in E3 and E3S/S microglia at 2 hours and 24 hours. Each point on graph represents one well of a chamber slide, with 4–5 images taken per slide and 2 chamber slides/condition. 3 independent biological replicates were completed (separate microglia dissections) with 8 wells/group each dissection, with n= 122 microglia for E3 2 hours, n=90 microglia for E3 24 hours, n=117 microglia for E3S/S 2 hours, and n=114 microglia for E3S/S 24 hours. *p=0.0228 for E3 compared to E3S/S at 2 hours, ****p<0.0001 for E3S/S compared to 2 hours compared to 24 hours, and ****p<0.0001 for E3 compared to E3S/S at 24 hours. Data is reported as mean +/− SEM, and analyzed with two-way ANOVA with mixed effects model.

(D) Schematic of treatment plan for primary microglia isolated from either E3 or E3S/S pups cultured with basal media, treated with 1 mg/ml 0N4R tau fibrils or 1 mM Aβ42 respectively for 24 hours.

(E) Volcano plot of DEGs in primary microglia between E3S/S+ tau and E3 + tau thresholded by log2foldchange +/−0.1 and p value threshold of 0.05. n=4–6 wells/experimental group, 2 independent biological replicates (separate microglia dissections). See also Figure S5 and Table S4.

(F) Top downregulated pathways in pathways in E3S/S versus E3 primary microglia under tau stimulation using GSEA. Top 6 pathways were shown after inputting top 500 upregulated genes into GSEA.

(G) Top predicted downregulated pathways in pathways in E3S/S versus E3 primary microglia under Aβ stimulation using GSEA. Top 6 pathways were shown after inputting top 500 downregulated genes into GSEA.

(H) Top predicted upstream regulators inhibited by E3S/S primary microglia compared to E3 under tau stimulation, including STING and cGAS.

(I) Representative confocal images of E3S/S and E3 primary microglia labeled with STING (green) and GM130 (red) with merge shown in yellow following 24 hours of tau stimulation. Scale bar represents 20um.

(J) Quantification of STING and GM130 colocalization area in E3S/S and E3 primary microglia following 24 hours of tau stimulation. n=12 images/condition obtained from 4 wells/group and 2 independent experiments. **p<0.01. Data is reported as mean +/− SEM and analyzed by two-way ANOVA.

We next analyzed transcriptomic responses of E3S/S microglia to tau or Aβ treatment (Fig. 5D, Fig. S5). Bulk RNA sequencing revealed broad downregulation of interferon genes (e.g. Ifi214, Irf7, Isg15, USP18, Ifit3) in tau-treated E3S/S microglia compared to E3 counterparts (Fig. 5E, Table S4). GSEA identified IFNα and IFNγ signaling as top downregulated pathways (Fig. 5F). Similarly, Aβ-treated E3S/S microglia downregulated interferon genes (Ifi211, Ifi204, Cxcl13, Ifi47), with IFN signaling, cytokine signaling, and innate immune response among the top downregulated pathways (Fig. 5F–G, Fig. S5E, Table S4). Conversely, top upregulated pathways diverged. (Fig. S5E–G, Table S4). Tau-treated E3S/S microglia showed enrichment in mRNA processing and transcription pathways (Fig. S5F, Table S4), while Aβ-treated E3S/S microglia upregulated Rho-GTPases and collagen synthesis pathways (Fig. S5G, Table S4). Ingenuity Pathway Analysis identified STING1 and cGAS as predicted upstream regulators inhibited in tau-treated E3S/S microglia (Fig. 5H). To confirm E3S/S mitigation of the cGAS-STING-IFN pathway in protein, we visualized activated STING in E3 and E3S/S microglia treated with tau (Fig. 5I–J). Following tau treatment, E3 microglia showed greater STING/GM130 colocalization compared to E3S/S microglia (Fig. 5I–J). These findings demonstrate that E3S/S microglia downregulate cGAS-STING-IFN activation in response to tau in vivo and in vitro.

R136S mutation lowers extracellular tau and suppresses tau-induced cGAS-STING-IFN response in human iPSC-derived microglia.

We next investigated how the R136S mutation alters responses to tau in human microglia. Human E3 and E3S/S iPSCs were differentiated into microglia-like cells (iMGLs)⁴³ (Fig. 6A), confirmed by the expression of microglia markers, IBA1 and TMEM119 (Fig. 6B). Both E3 and E3S/S iMGLs were treated with recombinant 0N4R tau fibrils, and intracellular tau was assessed after 24 hours. E3S/S iMGLs exhibited increased intracellular tau compared to E3 iMGLs, while media from E3S/S cells contained less extracellular tau (Fig. 6C–E). The tau-induced interferon response, measured with ELISA, revealed that E3S/S prevented the induction of IP10, an interferon-stimulated factor, seen in E3 iMGLs (Fig. 6F). Immunocytochemistry showed that tau treatment substantially increased pSTING and pTBK1 expression in E3 iMGLs, but activation was abolished in E3S/S iMGLs, indicating that R136S suppressed tau-induced activation of the cGAS-STING-IFN pathway (Fig. 6G–J). In summary, the R136S mutation enhances tau uptake and processing while dampening downstream cGAS-STING-IFN activation in mouse and human microglia.

Figure 6. R136S mutation suppresses tau-induced cGAS-STING-IFN response in human iPSC-derived microglia.

(A) Schematic for human iPSC-derived microglia differentiation.

(B) Representative immunofluorescence images of human iPSC-derived microglia expressing classical microglial markers, IBA1 (green, top) and TMEM119 (green, bottom). Scale bar represents 20um.

(C) Representative immunofluorescence images of E3S/S and E3 human iPSC-derived microglia following 24 hour recombinant tau stimulation, labeled with HT-7 tau antibody (green). Scale bar represents 5um. (D–E) Quantification of intracellular (D), extracellular (E) tau after 24 hours of tau uptake in E3S/S and E3 human iPSC-derived microglia. Outliers were excluded based on ROUT outlier test. n=20 from 20 individual wells performed in 2 independent experiments. *p<0.05. **p<0.01. Data are reported as mean +/− SEM and analyzed with unpaired t-test.

(F) ELISA measure of IP-10 in media after 24 hours of tau stimulation in E3S/S and E3 human iPSC-derived microglia. ELISA of IP-10 was normalized to protein concentration. n=10 from 10 individual wells performed in 2 independent experiments. ****p<0.0001. Data shown mean +/− SEM and analyzed by two-way ANOVA with mixed-effects model.

(G) Representative immunofluorescence images of pSTING (green) in E3S/S and E3 human iPSC-derived microglia at baseline and after 24 hours of tau stimulation. Scale bar represents 20um.

(H) Quantification of pSTING (green) in E3S/S and E3 human iPSC-derived microglia at baseline and after 24 hours of tau stimulation. Results are presented as average intensity from three images per well, n=7–10 wells from 3 independent experiments. Outliers were excluded based on ROUT outlier test. **p=0.002 for E3 compared to E3/tau, **p=0.0093 for E3S/S/tau compared to E3/tau. Data are reported as mean +/− SEM and analyzed by two-way ANOVA with mixed-effects model.

(I) Representative immunofluorescence images of pTBK1 (green) in E3S/Sand E3 human iPSC-derived microglia at baseline and after 24 hours of tau stimulation. Scale bar represents 20um.

(J) Quantification of pTBK1 (green) in E3S/Sand E3 human iPSC-derived microglia at baseline and after 24 hours of tau stimulation. Results are presented as average intensity from three images per well, n=10 wells from 3 independent experiments. ****p<0.0001 for all comparisons shown. Data are reported as mean +/− SEM and analyzed by two-way ANOVA with mixed-effects model.

cGAS inhibition ameliorates tau spread and mimics transcriptomic and protective effects of R136S mutation against synaptic loss induced by tauopathy.

Because cGAS-STING-IFN was among the top downregulated pathways induced by the R136S mutation in microglia both in vivo and in vitro, we investigated whether pharmacological inhibition of the interferon response phenocopies the R136S mutation. Our previous work established the efficacy of a brain permeable cGAS inhibitor (TDI6570, also known as G097, hereafter referred to as cGASi) formulated in chow in protecting against tau-induced deficits in P301S mice^41,44. To confirm target engagement, P301S mice were treated with 25 mg//kg cGASi via oral gavage and perfused at 2- or 6-hours post-treatment. Western blot demonstrated that cGASi reduced STING and downstream pTBK1 activation in bulk frontal cortex (Fig. S6A–D). Immunohistochemistry showed further reduction in STING-GM130 colocalization in microglia within 6 hours of treatment, indicating robust inhibition of the cGAS-STING-IFN pathway by cGASi in vivo (Fig. S6E–F).

The R136S mutation reduces tau load in human carriers and E3/P301S mice (Fig. 1). To investigate the role of cGAS inhibition in tau spread, we conducted a tau seeding assay in 3–4-month-old P301S mice. K18 tau fibrils were injected into one side of the hippocampus to induce contralateral tau propagation. Mice were treated daily with cGAS inhibitor (cGASi, 25 mg/kg, mixed in almond butter) or vehicle control for one-month post-seeding, then perfused. Immunolabeling demonstrated a significant reduction in MC-1-positive tau spread in cGASi-treated mice compared to controls (Fig. 7A–C), suggesting enhanced microglial processing and tau degradation, like those observed with the R136S mutation.

Fig 7. cGAS inhibition ameliorates tau spread and mimics transcriptomic and protective effects of R136S mutation against synaptic loss induced by tauopathy.

(A) Representative fluorescence images from a tau spreading assay in the hippocampi of P301S mice treated daily with a cGAS inhibitor or control via almond butter at a dose of 15 mg/kg body weight. Scale bar represents 500um. See also Figure S6.

(B) Higher magnification of CA3 regions of P301S hippocampi in tau spreading assay with seeding side on the left and spreading side on the right, MC-1 pathogenic tau labeled in red. Scale bar represents 100um.

(C) Quantification of MC-1 positive tau labeling on contralateral side in control and cGAS-i treated P301S mice, showing decreased tau spreading following cGAS-i treatment. Results are presented as average intensity measures from 3–4 sections per animal, 1 independent experiment. N=6 mice for CTL; N=7 mice for cGASi. Data are reported as mean ± SEM. **p<0.01 cGASi vs. control. Data were analyzed by unpaired t-test.

(D) Schematic plan for 6-month-old E3 and E3S/S mice received either a control diet or a cGAS inhibitor diet for 3 months. N=3 hippocampi/group, 1 independent experiment.

(E) Venn diagram showing effect of cGASi (left) in comparison to effect of R136S mutation (right) in microglia as well as overlapping genes between the effects (middle, in green area). See also Figure S7 and Table S5.

(F) Correlation analysis for expression of the overlapping genes in (G).

(G) Top upregulated pathways for overlapping genes in (G).

(H) Venn diagram showing effect of cGASi (left) in comparison to effect of R136S mutation (right) in excitatory neurons as well as overlapping genes between the effects (middle, highlighted in green).

(I) Correlation analysis for expression of the overlapping genes in (J).

(J) Top upregulated pathways for overlapping genes in (J).

(K) Venn diagram showing effect of cGASi (left) in comparison to effect of R136S mutation (right) in inhibitory neurons as well as overlapping genes between the effects (middle, highlighted in green).

(L) Correlation of gene expression for overlapping genes in inhibitory neurons in (M).

(M) Representative immunofluorescence images of PSD95 staining in CA1 subregion of hippocampus of E3 mice treated with control diet, E3/P301S mice treated with control diet, E3/P301S mice treated with cGAS inhibitor diet, and E3S/S/P301S mice treated with control diet.

(N) Quantification of PSD95 immunofluorescence intensity for the corresponding conditions in (M), n=11 animals for E3/control, n=7 animals for E3/P301S/control, n=8 animals for E3/P301S/cGASi, n=8 animals for E3S/S/P301S/control with 3–4 sections per mouse, 1 independent experiment. Data are reported as mean ± SEM. Data were analyzed by one-way ANOVA with Tukey’s multiple comparison, *p=0.025; **p=0.0028.

We then assessed to what extent cGAS inhibition mimics the protective effects of the R136S mutation post disease onset by treating 6-month-old E3/P301S mice with control or cGASi diet for 3 months, followed by snRNA-seq and immunohistochemistry (Fig. 7D). A total of 205,400 nuclei were analyzed after stringent quality control, with all major cell types similarly represented across genotypes (Fig. S7). We compared cell type-specific transcriptomic changes induced by cGASi in E3/P301S mice to those observed in E3S/S mice, both in the presence of tau pathology. In microglia, cGASi treatment induced 180 DEGs overlapping with those altered by the R136S mutation, with highly correlated expression (R=0.91, Fig. 7E–F, Table S5). Pathway analysis of overlapping DEGs revealed that both cGASi and the R136S mutation upregulated pathways related to cell motility, endocytosis, migration, and metabolic processes (Fig. 7G). Similarly, significant overlaps and correlations in expression patterns were observed in ENs (R=0.93, Fig. 7H–I, Table S5) and INs (R=0.97, Fig. 7K–L, S6E, Table S5). In ENs, shared pathways included cell morphogenesis, synaptic transmission, cell migration and communication, and regulation of response to stimuli (Fig. 7J). INs were not included in pathway analysis due to a low number of upregulated DEGs (<30). These findings strongly support that cGAS inhibition mimics the R136S transcriptome in neurons and microglia. Furthermore, cGASi treatment ameliorated PSD95 loss in E3/P301S tauopathy mice, phenocopying what we observed in E3S/S/P301S mice (Fig. 7M–N). However, cGASi did not alter hippocampal myelin expression in E3/P301S mice, suggesting that the protection against demyelination in E3S/Smice may involve a cGAS-independent mechanism, potentially regulated by TLR7⁴⁵ (Fig. S7 H, I). In summary, cGAS inhibition reduced tau spread, and treatment after disease onset recapitulated key aspects of R136S -mediated resilience to tau toxicity, including transcriptomic profiles in neurons and microglia and preservation of synaptic integrity.

DISCUSSION

Our study combines functional assays, transcriptomic analyses, and pharmacological intervention to shed light on mechanisms enabling E3S/S to protect against tauopathy. We established APOE3 and APOE3S/S knockin mice and demonstrated that E3S/S protects against tau-induced theta and gamma oscillation declines. E3S/S mice exhibited decreased tau load and tau-induced synaptic loss. The E3S/S mutation affects multiple cell types and downregulates cGAS-STING-IFN in microglia. In primary mouse microglia, the E3S/S mutation enhanced tau processing and suppressed tau-induced cGAS-STING-IFN pathway activation. This finding was recapitulated in human iPSC-derived microglia. Ultimately, treating E3/P301S mice with cGASi after network dysfunction onset phenocopied the E3S/S mutation on microglial and neuronal transcriptomes and protected against tau-induced synaptic loss.

Our model replaces the mouse Apoe gene with human E3 or E3S/S cDNA. Consistent with a previous study replacing mouse Apoe with human APOE3, the R136S mutation did not affect brain APOE expression in males¹⁵, while females showed increased APOE in E3S/S/P301S cortex. Unlike Chen et al¹⁵, which reported elevated plasma APOE expression in male E3S/S mice, we observed similar APOE expression in E3 and E3S/S mice in male plasma. APOE expression was not affected by R136S on the APOE4 background in the Nelson et al study¹⁴. Exactly how the R136S mutation affects APOE3 metabolism in brains vs. plasma and sex-dependent effects remain unclear. Nevertheless, comparable APOE in our male E3 and E3S/S brains and plasma allow us to investigate the specific effects induced by R136S mutation without confounds from different APOE isoforms.

The E3S/S patient had lower tau load compared to other individuals with the same PSEN1 mutation and high amyloid pathology¹². E3S/S/P301S mice demonstrated less tau pathology across the hippocampus, like previous observations on the APOE4 background¹⁴. Previous studies elucidated robust protective effects of R136S in an E3 amyloid model induced by tau aggregates isolated from AD brains (AD-tau)¹⁵. Thus, while protective effects of the R136S mutation were discovered in an E3 carrier with amyloid pathology, it is likely that protective mechanisms from the R136S mutation also apply to FTD-tau and primary tauopathies.

How does the R136S mutation lower tau load? Previous studies suggest neuronal or microglial mechanisms. All current E3S/S knockin models are whole-body; conditional knockin models are needed to solidify cell-specific mechanisms. We observed more intracellular tau in E3S/S primary microglia than in E3 at 2 hours, indicating E3S/S primary microglia exhibit more efficient tau uptake in early timepoints than E3. After external tau removal, we found E3S/S microglia had less intracellular tau than E3 microglia—indicating that R136S increased tau processing. Our E3S/S human-derived iPSC microglia also phagocytose tau more efficiently than E3 microglia: after 24 hours of tau stimulation, we found increased intracellular tau and less extracellular tau. Our data align with recent studies using E3S/S microglia co-cultured with PSEN1 neurons or Aβ−42, in which the R136S mutation increased phagocytosis and reduced phosphorylated tau load to protect neurons⁴². Previous studies have shown that type I interferon signaling was found to exacerbate tau seeding⁴⁶. Thus, it is plausible that the attenuated interferon response in E3S/S iMGLs may facilitate more efficient tau uptake and processing.

While the exact mechanism of R136S-dependent tau reduction remains unknown, we propose a role for the cGAS-STING-IFN pathway, as treatment with cGASi reduced tau spreading. Genetic deletion of cGAS did not reduce tau inclusions in 9-month-old P301S mice⁴¹, suggesting a ‘window of opportunity’ for cGAS inhibition to reduce tau spread. This supports a prominent role of microglial activation in tau spread, consistent with previous findings on microglial-derived exosomes⁴⁷ and NF-κB signaling⁴⁰. Although we observed a reduction in interferon pathways following amyloid treatment in primary microglia, further studies are needed to assess Aβ spread following cGAS inhibition. Additionally, since many of our observations occurred in mouse microglia further studies are needed to detail mechanistic overlap with human microglia.

LFP recording of network activities showed that E3S/S mice are protected against tau-induced deficits in theta and gamma oscillations at 6–7 months of age before tau-induced neurodegeneration at 9-months of age, suggesting that network dysfunction represents early pathogenetic mechanisms in AD⁴⁸. Previous studies show that amyloid pathology induces abnormal theta and gamma oscillations, likely due to deficient inhibitory interneuron activity, not examined here^20,49,50. Evidence that FTD tau pathology leads to similar network dysfunction suggests converging network dysfunction in AD and primary tauopathies^51–53. Indeed, gamma frequency modulation was beneficial in models with amyloid and/or tau pathology⁵⁴. Using 3D c-FOS mapping, we observed protection against hyperexcitability in E3S/S tauopathy mice compared to E3/P301S mice. It is conceivable that synaptic loss in E3/P301S mice disrupts network stability, while compensatory mechanisms, such as homeostatic scaling and inhibitory neuron loss in tauopathy mice⁵⁵, can drive pathological hyperactivity. This aligns with evidence that tau-induced hyperexcitability exacerbates synaptic and cognitive decline in a model of amyloid toxicity⁵⁶. Further studies are needed to delineate molecular events downstream of cGAS-STING-IFN in tau-induced synaptic loss, hyperexcitability and network dysfunction.

Our snRNA-seq analyses of excitatory and inhibitory neurons revealed significant upregulation of RNA-splicing pathways by the R136S mutation. Among RNA-splicing factors, RBM25 is downregulated in E3 tauopathy neurons, but rescued by R136S. Consistent with protective effects of RBM25, previous snRNA-seq profiling revealed lower RBM25 expression in AT8+ vs. AT8- neurons of AD brains³⁵. Moreover, screening of 453 splicing factors showed RBM25 activates splicing in >90% of targeted exons across contexts, including MAPT exon 10⁵⁷. Among the downregulated pathways in E3S/S/P301S hippocampus, neurogenesis is known to be elevated following hyperexcitability^58,59, observed in E3/P301S mice using cFOS activity mapping. The downregulation of cFOS activity by the R136S mutation could explain downregulated neurogenesis in the EN transcriptome. The R136S mutation also distinctly affected the astrocyte transcriptome, upregulating protective genes such as EAAT2, the main glutamate transporter that exerts protective function against excitotoxity. EAAT2 expression diminishes with age⁶⁰ and correlates with signs of neuronal death in postmortem brains of sporadic AD patients⁶¹. Because EAAT2 agonism is neuroprotective⁶², dissecting how R136S enhances EAAT2 expression may be therapeutic.

AD patients exhibit myelin loss, and higher myelin expression is associated with tau pathology protection in AD^43,44. Our findings suggest that the R136S mutation alters oligodendrocyte transcriptomes and protects against tau-induced myelin loss. Notably, APOE4 has also been linked to demyelination, even without tau pathology⁴⁵. Additionally, transcriptomic changes have been observed in oligodendrocyte subpopulations with the R136S mutation on an E4 background¹⁴. Our lipidomic study revealed elevated Hex1Cer, Hex2Cer, and SM (sphingomyelin) in E3S/S/P301S, suggesting enhanced glycosphingolipid and sphingolipid synthesis. This elevation likely contributes to increased myelination. We also observed reduced complex glycosphingolipids, GD1a and GT1b gangliosides and GALNT7, a gene involved in ganglioside synthesis, in E3S/S/P301S. These gangliosides are known to form toxic aggregates commonly associated with neurodegenerative diseases including FTD³⁸ and Gaucher’s disease⁶³. The increase in APOE3S/S mouse ceramides may reflect an adaptive response to tau pathology, supported by upregulated sphingolipid synthesis in oligodendrocytes. While ceramides are typically linked to neurodegeneration, their role here may be context-dependent, potentially contributing to neuroprotection. Conversely, the reduction in gangliosides, such as GD1a and GQ1b, suggests improved lysosomal function³⁸, which could mitigate tau-induced toxicity. Future studies should define how R136S-driven lipid remodeling enhances lysosomal function and myelination to counteract tau toxicity.

Maladaptive microglia are central in pathological tau spread and toxicity^40,47. While the exact toxic pathways remain elusive, recent studies provide compelling evidence of interferon pathway damage when tau pathology is present. Our previous studies suggest that the source of IFN signaling could be partially attributed to hyperactive cGAS-STING in tauopathy mice and human AD brains⁴¹. Cgas deletion mitigated tau-induced microglial IFN-I and protected against synapse loss, decreased synaptic plasticity, and cognitive deficits⁴¹. Moreover, APOE4 exacerbates tau-induced neurodegeneration, partially attributed to a critical immune hub that involves IFN-driven microglial activation and interaction with cytotoxic T cells^64,65. Previous findings showed that R136S mutation enhances microglial response around amyloid plaques on E3 background¹⁵, but downregulates disease-associated gene profiles on E4 background¹⁴. Our current findings reveal that the resilient E3S/S allele suppresses tau-induced cGAS-STING-IFN signaling in microglia while maintaining DAM responses, protectively balancing immune activation and resilience pathways. Upon tau stimulation, primary microglia from E3S/S mice exhibit reduced expression of interferon-stimulated genes, indicating the R136S mutation downregulates microglial cGAS-STING-IFN in vivo and in vitro. Notably, the dampening effect of the E3S/S mutation on tau-induced cGAS-STING activation replicates in human iPSC-derived microglia, suggesting that suppression of the interferon response to tau is conserved in humans.

Previous research demonstrated that the R136S mutation impairs APOE3’s interaction with heparan sulfate proteoglycans (HSPGs). Marino et al. developed an APOE3 Christchurch-inspired antibody to disrupt APOE4-HSPG interactions, resulting in reduced tau pathology in P301S mouse retinas and brains⁶⁶. In macrophages, HSPGs modulate inflammation through interferon-beta (IFN-β) via IFN-β sequestration, enabling proteoglycan sulfation to influence inflammatory responses⁶⁷. The mechanistic link between impaired APOE3 R136S–HSPG interaction and attenuation of the interferon response following tau stimulation remains to be elucidated, particularly in the context of microglial signaling and neuroinflammatory modulation.

Pharmacological cGAS inhibition in E3/P301S mice after network dysfunction onset was employed to compare the protective effects of chronic inhibition with effects conferred by R136S on the E3 background. Consistent with previous studies in tauopathy mice expressing mouse Apoe, cGAS inhibition mitigated tau-induced synaptic loss⁴¹. The significant overlap and correlation of transcriptomes across cell types suggest that resilience observed in the human R136S carrier can be partially replicated through pharmacological cGAS inhibition. cGAS inhibition, however, did not prevent myelin loss, indicating the R136S mutation may confer cGAS-independent protection or that earlier intervention is required. Future studies elucidating the relationship between APOE isoforms and cGAS activation may identify shared mechanisms and therapeutic potential of cGAS inhibitors in AD and related dementia. Collectively, this study elucidates the impact of the R136S mutation on tau-induced network activity and function, reveals extensive transcriptomic effects on neurons and glial cells, and identifies a potential therapeutic pathway.

Limitations of the study.

Our results were obtained from full-body knockin models, limiting our ability to attribute the observed effects to microglial-derived APOE3S/S in vivo. Future studies utilizing conditional models of APOE3 and APOE3S/S knockin are required. Secondly, while cGASi administration corrected transcriptomes across cell types and protected synapses, whether cGAS inhibition rescues circuit dysfunction or alterations in neuronal splicing factors remains unknown. Additionally, the exact downstream events linking cGAS-STING inhibition to neuronal activity changes and tau pathology remain unclear. Additional experiments are also needed to parse mtDNA release, cGAS-STING activation, and neurodegeneration. These limitations underscore the need for studies to fully elucidate the mechanisms by which the APOE3-R136S mutation protects against tau pathology.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Li Gan (lig2033@med.cornell.edu).

Materials Availability

The plasmids and cell lines generated in this study are available on request upon the completion of a Material Transfer Agreement (MTA).

Data and Code Availability

• Bulk and single-cell RNA-seq data have been deposited at GEO and are publicly available as of the date of publication. Accession numbers are listed in the Key Resources Table.

• All custom code used for bulk RNA-seq and snRNA-seq data analysis are publicly available. Original code has been deposited at GitHub and is publicly available as of the date of publication. The DOI is listed in the Key Resources Table.

• Any additional information required to re-analyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Experimental models: Organisms/strains
C57BL/6J	The Jackson Laboratory	RRID:IMSR_JAX:000664
Human APOE3 or APOE3S/S (C57BL/6J)	Gan Lab, Weill Cornell, this paper	N/A
Oligonucleotides
Apoe-L-Arm-F2 5’ end: CGGCAAGGGGAGGTAAA CAGAC	Gan Lab, Weill Cornell, this paper	N/A
Apoe-WT-F3 3’ end: CGGGTGCTCTGTTTTGGA GATG	Gan Lab, Weill Cornell, this paper	N/A
ApoeWT-5-R2 5’ end: CGGCCCACAGAGCCTTCA TCTTC	Gan Lab, Weill Cornell, this paper	N/A
Apoe-R-Arm-R2 3’ end: CTCAACCCACCAGCAGAA GAGATC	Gan Lab, Weill Cornell, this paper	N/A
hAPOE-5R1 5’ end: CGCAGGTAATCCCAAAAG CGACC	Gan Lab, Weill Cornell, this paper	N/A
SV-polyA-F1 3’ end: CACACCTCCCCCTGAACC TGAAAC	Gan Lab, Weill Cornell, this paper	N/A
PS19-F GGT ATT AGC CTA TGG GGG ACA C	Udeochu, J.C., Amin, S., Huang, Y et al., 202341	N/A
PS19-R GGC ATC TCA GCA ATG TCT CC	Udeochu, J.C., Amin, S., Huang, Y et al., 202341	N/A
Antibodies
Rabbit anti-PSD95 (mouse tissue)	Millipore	Cat# MAB1596
Goat anti-Iba1 (mouse tissue)	Abcam	Cat# ab5076
Rabbit anti-pSTING (mouse tissue)	Cell Signaling	Cat# 72971S
Mouse anti-MC-1 (mouse tissue)	Gift from Peter Davies	N/A
Rabbit anti-GFAP (mouse tissue)	Abcam	Cat# ab7260
Mouse anti-NeuN (mouse tissue)	Millipore	Cat# MAB377
Goat anti- APOE (mouse tissue)	Calbiochem	Cat# 178479
Rabbit anti-RBM-25 (mouse tissue)	Proteintech	Cat#25297-1-AP
Mouse anti-GAPDH (mouse tissue)	GeneTex	Cat#GTX627408
Rabbit anti-MBP (mouse tissue)	ThermoFisher	Cat#PA1-46447
Rabbit anti-Synaptophysin (mouse tissue)	Abcam	Cat#ab23754
Mouse anti-GM130 (mouse tissue and primary mouse microglia cells)	BD Bioscience	Cat#610822
Rabbit anti-pSTING (iPSCs)	Cell Signaling Technology	Cat#19781S
Rabbit anti-STING	Cell Signaling Technology	Cat#13647S
Rabbit anti-TBK1	Cell Signaling Technology	Cat#3504S
Rabbit anti-pTBK1 (iPSCs)	Abcam	Cat#ab109272
Rabbit anti-IBA1 (iPSCs)	Wako	Cat#019-19741
Rabbit anti-TMEM119 (iPSCs)	Sigma Aldrich	Cat#HPA051870
Mouse anti-Tau (HT-7) (iPSCs)	Thermo Fisher Scientific	Cat#MN1000
Rabbit-anti-cFOS, clone 9F6 (cFOS mapping, mouse tissue)	Cell Signaling Technology	Cat#2250S, clone 9F6
Alexa Fluor 488 Donkey Anti-Goat IgG	Thermo Fisher Scientific	Cat. # A-11055; RRID:AB_2534102
Alexa Fluor 488 Donkey Anti-Rabbit IgG	Thermo Fisher Scientific	Cat. # A-21206; RRID:AB_2535792
Alexa Fluor 568 Donkey Anti-Mouse IgG	Thermo Fisher Scientific	Cat. # A10037; RRID:AB_11180865
Alexa Fluor 568 Donkey Anti-Rabbit IgG	Thermo Fisher Scientific	Cat. # A10042; RRID:AB_2534017
Alexa Fluor 488 Goat Anti-Rabbit IgG	Thermo Fisher Scientific	Cat. # A-11008; RRID:AB_143165
Alexa Fluor 647 donkey-anti-rabbit	Jackson ImmunoResearch	Cat#711-607-003
Chemicals, peptides, and recombinant proteins
DMEM/F-12	Thermo Fisher Scientific	Cat#11320033
GlutaMAX Supplement	Thermo Fisher Scientific	Cat#35050061
MEM Non-Essential Amino Acids Solution (100X)	Thermo Fisher Scientific	Cat#11140050
Fetal Bovine Serum	Thermo Fisher Scientific	Cat#10439001
Trypsin-EDTA (0.05%), phenol red	Thermo Fisher Scientific	Cat#25300054
Real-time PCR was performed using SsoAdvanced Universal SYBR® Green Supermix	Bio-Rad	Cat#1725270
Reveal Decloaker	Biocare Medical	Cat#RV1000
Cas9 protein	Integrated DNA technologies	Cat. # 1081058
RNase-free DNase	Qiagen	Cat. # 79254
Bovine serum albumin	Sigma-Aldrich	Cat. # A3294
DL-1,4-Dithiothreitol (DTT)	Thermo Fisher Scientific	Cat. # 426380500
Paraformaldehyde	Electron Microscopy Services	Cat. # 15710-S
PBS	Gibco	Cat. # 70011-044
Triton X-100	Sigma-Aldrich	Cat. # T9284
TBST	Thermo Fisher Scientific	Cat. # 28360
Donkey serum	Jackson ImmunoResearc h	Cat. # 017-000-121
Goat serum	Jackson ImmunoResearc h	Cat. # 005-000-121
Penicillin-streptomycin	Gibco	Cat. # 15070063
Poly-D-lysine	Thermo Fisher Scientific	Cat. # A389401
Recombinant mouse GM-CSF, carrier free	R&D Systems	Cat. # 415-ML-020-CF
0N4R tau fibrils	Dr. Sue-Ann Mok, University of Alberta	N/A
Amyloid β fibrils	Dr. Yueming Li, Memorial Sloan Kettering	N/A
mTeSR Plus	STEMCELL	Cat#100-0276
hESC-Qualified Matrigel	Corning	Cat#354277
Y-27632 (ROCK inhibitor)	StemCELL	Cat#72308
STEMdiff Hematopoietic kit	STEMCELL	Cat#5310
B-27™ Supplement (50X), serum free	Gibco	Cat#17504044
N-2 Supplement (100x)	Gibco	Cat#17502048
GlutaMAX™ Supplement	Gibco	Cat#35050061
MEM Non-essential Amino Acid Solution (100x)	Sigma Aldrich	Cat#M7145
Monothioglycerol	Sigma Aldrich	Cat#M6145
insulin-transferrin-selenite	Gibco	Cat#41400-045
Insulin solution human	Sigma Aldrich	Cat#I9278
Recombinant Human IL-34	Peprotech	Cat#200-34
Human Recombinant TGF-beta 1	STEMCELL	Cat#78067.3
Recombinant Human M-CSF Protein	R&D Systems	Cat#216-MC-010/CF
Human Recombinant Fractalkine (CX3CL1)	STEMCELL	Cat#78051.2
Recombinant Human CD200 (C-His)	Novo protein	Cat#C311
2,2’-Thiodiethanol (TDE)	Sigma-Aldrich	Cat# 166782-500G
Dichloromethane (DCM)	Sigma-Aldrich	Cat# 270997-1L
Dimethyl Sulfoxide (DMSO)	Sigma-Aldrich	Cat# D2650-100ML
Glycine	Sigma-Aldrich	Cat# G7126-1KG
Heparin Sodium	Acros Organics	Cat# AC411210010
30% Hydrogen Peroxide (H2O2) Solution	Sigma-Aldrich	Cat# 216763-100ML
Iohexol (Histodenz)	Axis-Shield	Cat# AXS-1002424
Methanol (MeOH)	Sigma-Aldrich	Cat# 34860-4X4L-R
32% Paraformaldehyde (PFA) Solution, EM Grade	Electron Microscopy Sciences	Cat# 100496-496
Potassium Chloride (KCl)	Sigma-Aldrich	Cat# P4504-1KG
Potassium Phosphate Monobasic (KH2PO4)	Fisher Chemical	Cat# P285-500
Sodium Chloride (NaCl)	Sigma-Aldrich	Cat# S9625-5KG
Sodium Phosphate Dibasic (Na2HPO4)	Sigma-Aldrich	Cat# S9763-5KG
Sodium Phosphate Momobasic (NaH2PO4)	Sigma-Aldrich	Cat# RDD007-1KG
Triton X-100	Sigma-Aldrich	Cat# X100-1L
Tween-20	Sigma-Aldrich	Cat# P1379-500ML
Halt phosphatase inhibitor cocktail	Thermo Fisher Scientific	Cat. # 78420
cOmplete protease inhibitor cocktail	Roche	Cat. # 11697498001
NuPAGE MES SDS running buffer	Invitrogen	Cat. # NP0002
Clarity Western ECL substrate	Bio-Rad	Cat. # 1705060
TDI6570 (cGAS inhibitor)	Synthesized by Gan Lab at Wuxi AppTec (Tianjin) Co., Ltd.	Lama et al.44
RIPA lysis and extraction buffer	Thermo Fisher Scientific	Cat. # 89901
Dimethyl sulfoxide (DMSO)	MP Biomedicals	Cat. # 02196055
RNase-free DNase	Qiagen	Cat. # 79254
DNase I	Millipore Sigma	Cat. # 260913
cOmplete protease inhibitor cocktail	Roche	Cat. # 11697498001
SOLA SPE plates	ThermoFisher	Cat #60509-001
Acclaim™ C30 HPLC Columns	ThermoFisher	Cat# 075718
OE240 Exactive Orbitrap MS	ThermoFisher	N/A
Vanquish UHPLC system	ThermoFisher	N/A
Orbitrap Exploris 240 mass spectrometer	ThermoFisher	Cat#BRE725535
Kinetex HILIC column	Phenomenex	Cat#00D-4461-AN
Critical commercial assays
Pierce BCA Protein Assay Kit	Thermo Fisher Scientific	Cat#23225
QuickRNA MicroPrep Kit	Zymo	Cat#R1051
Human Apolipoprotein E ELISA Kit	Invitrogen	Cat#EHAPOE
Human IP-10 (CXCL10) ELISA kit	Invitrogen	Cat# KAC2361
Human Total Tau ELISA kit	Invitrogen	Cat # KHB0041
Illumina Stranded mRNA Prep	Illumina	Cat#20040532
Chromium Single Cell 3’ Reagent Kits v3.1	10x Genomics	Cat#PN-1000268
Qiagen Plasmid Maxi kit	Qiagen	Cat#12963
VECTASHIELD® Antifade Mounting Medium with DAPI	Vector Laboratories	Cat#H-1200-10
MOM Basic Immunodetection Kit	Vector Laboratories	Cat. # BMK-2202
Deposited data
Immunoblot data	This paper	N/A
Raw and processed data (bulk RNA sequencing data)	This paper	GEO: GSE297743
Raw and processed data (single nuclei RNA sequencing data)	This paper	GEO: GSE297721
Software and Algorithms
Original code	This paper	https://github.com/lifan36/Naguib-APOE3CC-2025
R 4.2.2	The R project	https://www.r-project.org/
RStudio 2022.07.2	RStudio: Integrated Development for R. RStudio	https://rstudio.com
Noldus Ethovision XT v.16	N/A	https://www.noldus.com/ethovision-
Gene set enrichment analysis (GSEA)	Subramanian et al., 2005	https://www.gsea-msigdb.org/gsea/msigdb/index.jsp
Adobe Illustrator	Illustrator v26.5.2	https://www.adobe.com/products/illustrator.html
GraphPad Prism 6	Prism v9.2	https://www.graphpad.com
Cell Ranger- 6.1.2	10x Genomics	https://support.10xgenomics.com/single-cell-gene-expression/software/pipelines/latest/what-is-cell-ranger
NIS Elements AR 5.21.03 64-bit	Nikon Instruments Inc.	https://www.microscope.healthcare.nikon.com/products/software/nis-elements
MATLAB	MATLAB R2022a	https://www.mathworks.com/products/matlab.html
Seurat	Seurat v4.0	https://satijalab.org/seurat/
Fiji/ImageJ	Fiji v2.8	RRID: SCR_002285
OpenEphys GUI	OpenEphys	https://open-ephys.org/
Biorender	Biorender	https://www.biorender.com/
Lipidcruncher	N/A	In house software
IMARIS	IMARIS 10.2	https://imaris.oxinst.com/

STAR★Methods

EXPERIMENTAL MODELS AND STUDY PARTICIPANT DETAILS

Generating Mouse Lines

To introduce a human APOE3 and APOE3Christchurch (APOE3-R136S, E3S/S) cDNA into mouse Apoe genetic locus, a donor plasmid was designed and made. The human APOE3 or APOE3S/S cDNA and a poly A tail were flanked by left and right homology arms which were designed to insert human APOE3 or APOE3S/S right before the start codon of mouse Apoe gene. Upon recombination, the mouse Apoe promoter and regulatory elements would drive expression of the inserted human APOE3 or APOE3S/S cDNA, whereas the expression of mouse Apoe gene would be inactivated. A single guide RNA (sgRNA) was selected. RNP containing sgRNA and Cas9 protein was prepared. Pronuclear injection of RNP and the donor plasmid was performed at the Rockefeller University transgenic core facility. Genomic DNA from founders was isolated from tail lysate. To screen the specific knockin mice, two PCR amplifications were performed with primers flanking the outside of the homology arms and on the human APOE3 transgene. The PCR products were sequenced to validate the correct insertion and the locus integrity. Non-specific integration of the donor DNA was characterized by two simple PCR amplifications using two sets of primers on the backbone of the plasmids.

To distinguish homozygous from heterozygous mice, tail DNA from F2 offspring was further characterized by PCR with three primers, two primers on the targeting genomic sequences and one primer on the transgene. Three independent APOE3 or APOE3S/S lines without non-specific integration were selected and maintained. One APOE3 or APOE3S/S line was used to cross with a transgenic mouse model overexpressing human tau harboring the P301S mutation (The Jackson Laboratory, 008169) to evaluate the effect of APOE3 Christchurch on tau pathology. Mice were housed no more than five per cage, given ad libitum access to food and water and housed in a pathogen-free barrier facility at 21–23 °C with 30–70% humidity on a 12-h light/12-h dark cycle. All in vivo experiments, unless otherwise noted, were completed in male mice 9–10 months. All mouse protocols were approved by the Institutional Animal Care and Use Committee at Weill Cornell Medicine. Specific primers for genotyping can be found in the STAR methods of this manuscript.

Generating Primary Mouse Microglia

Primary microglial cells were collected from male and female mouse pups at postnatal days 1–3. Briefly, the brain cortices were isolated and minced. Tissues were dissociated in 0.25% Trypsin-EDTA for 10 minutes at 37 °C and agitated every 5 minutes. Two hundred microliters of DNAse I (Millipore) was then added. Trypsin was neutralized with complete medium (DMEM; Thermo Fisher) supplemented with 10% heat-inactivated FBS (Hyclone), and tissues were filtered through 70-μm cell strainers (BD Falcon) and pelleted by centrifugation at 250g. Mixed glial cultures were maintained in growth medium at 37 °C and 5% CO₂ for 7–10 days in vitro. Once bright, round cells began to appear in the mixed glial cultures, recombinant mouse granulocyte–macrophage colony-stimulating factor (1 ng ml⁻¹; Life Technologies) was added to promote microglia proliferation. Primary microglial cells were collected by mechanical agitation after 48–72 hours and plated on poly-D-lysine-coated 24-well plates (Corning) in growth medium. Microglia were maintained in DMEM supplemented with 10% FBS, 100 U ml⁻¹ penicillin and 100 μg ml⁻¹ streptomycin. At 24 hours after plating, microglia were treated with 1ug/ml of 0N4R recombinant tau obtained (from Dr. Sue-Ann Mok’s group, University of Alberta, Canada) or Aβ monomers at a concentration of 4.2ug/mL (from Dr. Yueming Li’s lab, MSK, New York) for 24 hours.

Generating Human iPSC-derived Microglia

Male human APOE3 and its isogenic APOE3 Christchurch iPSCs were obtained by Jackson Laboratory. These iPSCs were cultured with mTeSR plus media (STEMCELL) in hESC-Qualified Matrigel (Corning)-coated 6-well plates. At 70–80% confluence, the iPSCs were dissociated with ReLeSR (STEMCELL) solution. The dissociated iPSCs were seeded into fresh Matrigell-coated-6-Well plate with mTeSR plus media containing 10 μM ROCK inhibitor (Tocris). For generating the human HPCs, we used the STEMdiff Hematopoietic kit (STEMCELL). Briefly, dissociated iPSCs with ReLeSR were seeded in Matrigel-coated-6-well plate with mTeSR plus containing 10 μM ROCK inhibitor. At Day 0, mTeSR plus media was replaced with Basal media containing Sup A. At Day 3, the media was replaced with Basal media containing Sup B. Half of the media was changed with Sup B containing media every other day. At Day 13, we used HPCs for Microglia like-cell differentiation. To generate iMGLs, we used Microglia Basal media [DMEM/F12 (Gibco), B27 (Gibco), N2 (Gibco), GlutaMAX (Gibco), NEAA (Sigma Aldrich), Monothioglycerol (Sigma Aldrich), insulin-transferrin-selenite (Gibco), and insulin (Sigma Aldrich)] 50,000 of HPCs were seeded in Matrigel-coated 6-well plate with Microglia Basal media containing 100 ng/ml of IL-34 (Peprotech), 50 ng/ml of TGF-β1 (STEMCELL), and 25 ng/ml or M-CSF (R&D Systems). 1ml of the Basal media containing 3 cytokines was added every other day until Day 24. On Day 25, the media was replaced with the media additionally containing 2 more cytokines [100 ng/ml of CD200 (Novo protein) and 100ng/ml of Cx3CL1 (STEMCELL)]. At Day 28, iMGLs were used for experiments.

METHOD DETAILS

Electrophysiological recording in freely moving mice

All experimental procedures were approved by the Weill Cornell Medical College Animal Care and Use Committee following National Institutes of Health guideline. Male mice from all experimental groups (6–7 months) were initially anesthetized with 3.5 % isoflurane mounted on a stereotaxic frame and maintained under ~1.2% isoflurane. Body temperature was maintained at 37 °C with a regulated heating blanket (World Precision Instruments, Sarasota, FL). A craniotomy was drilled for electrode insertion. Animals were implanted with a custom 16-ch 3D electrode array (Kedou Brain-Computer Technology, Suzhou, China) that contains 4 tetrodes over somatosensory (S1), the CA1 of hippocampus, dentate gyrus (1), and primary visual cortex (V1) (S1 : antero-posterior (AP) −1.8mm, mediolateral (ML) 2.3mm, dorsoventral (DV) 1.0mm, CA1: AP −1.8mm, ML 1.5mm, DG 1.4 mm, DG: AP −1.8mm, ML 1.0mm, DV 2.0mm, V1: AP − 3.0mm, ML 2.3mm, DV 0.6mm. Fig 2A). Stainless steel screws were implanted into the skull to provide electrical ground and mechanical stability for drives and the whole construct was bonded to the skull using C&B-Metabond luting cement (Parkell, Edgewood, NY). Two weeks after implantation, animals were put in a circular open field chamber (8228, Pinnacle Technology, Lawrence, KS) to record spontaneous activity continuously for 30 min. Electrophysiological data were acquired using an Intan RHD eadstage (Intan Technologies, LA, CA) and the Open Ephys acquisition board and software (OEPS, Alges, Portugal) sampled at 30kHz⁶⁸. Locomotion was simultaneously acquired with a FLIR camera (Teledyne FLIR, North Billerica, MA) and the open-source Bonsia software⁶⁹ recording at 50 Hz.

Local field potential analysis

The offline LFP analysis was performed using custom-written Matlab script (MathWorks, Natick, MA). Briefly, LFP data will be preprocessed by first applying an anti-aliasing lowpass (<400Hz) zero-phase 8-order Chebyshev Type I filter then downsampling to 1000Hz. The power of theta (4–10 Hz) and gamma (30–90 Hz) were calculated then averaged by each tetrode to obtain the low and high band oscillation activity from each brain region. The normalized power was converted from the above value by dividing the average 10 second baseline value at resting state when mice didn’t show any movement in the behavior video. The averaged power for the entire open field session was also calculated for the group comparison. Markless pose estimation for locomotion measurement was performed by deep learning based DeepLabCut from the behavioral video data set⁷⁰.

cFOS Mapping Experiment

8–9 month old male mice were habituated in single-house cages for 4 hours prior to start of experiment. Then, mice were individually placed in the aforementioned circular open field chamber, where their behavior was recorded from a camera underneath them. After 10 minutes of free-moving behavior recording, mice were placed back in their home cages for 45 minutes where they were promptly transcardially perfused and whole brains were collected for further 3D analysis.

cGAS-inhibitor target engagement studies

To assess target engagement of TDI-6570 acutely, we administered TDI-6570 at a dose of 25mg/kg dissolved in DMSO via oral gavage to 8–9 month old mice. At 2 and 6 hours post-gavage, transcardial perfusion was performed and we collected tissue for both biochemical and immunohistochemical analyses.

Tau seeding and spreading assay

To assess if the cGASi affects tau seeding and spreading, we injected 3.5–4.5 month old male and female P301S mice with 2uL of 1.8mg/ml K18 tau fibrils into their one hippocampi (anterior-posterior −2.0, medial-lateral +/− 1.3, dorsal-ventral −2.1). Mice were treated daily with cGASi (25 mg/kg body weight) or vehicle control mixed with almond butter Monday -Friday, and 150 mg/Kg chow diet over the weekend. Mouse weights and food consumed per c. age were collected throughout the study. After one month, mice were collected using transcardial perfusion (see methods below) and whole brains were collected for immunohistochemistry (see methods below) of MC-1 on the ipsilateral (seeding side) and contralateral (spreading side).

Chronic cGAS inhibitor treatment

TDI-6570 were prepared using the methods as described previously⁴⁴. Food pellets containing TDI-6570 (300 mg of drug per kg diet) were prepared by Research Diet (RDI) according to previously published methods, and control diet pellets were prepared similarly without TDI-6570⁴¹. The diet pellets were used before the expiration dates tentatively set for 6 months after the date of diet preparation. 6–7 month old male mice from all experimental conditions were treated with control or TDI-6570 pellets for 3 months before transcardial perfusion. Mouse weights and food consumed per cage were collected throughout the study.

Tissue collection

Mice were euthanized using FatalPlus and perfused with cold PBS following cardiac puncture for blood collection. Blood was collected in EDTA-coated tubes and spun down to isolate plasma. After complete perfusion, left hemibrain was fixed in 4% paraformaldehyde for 24–48 hours followed by storage in 30% sucrose. Right hemibrain was microdissected for hippocampus and frontal cortex, which were frozen in dry ice and stored at −80°C until further processing.

Immunohistochemistry

Fixed hemibrain was sectioned in 30 um-thick sections via microtome and stored in cryoprotectant. 3–5 hippocampal sections were picked per animal and following rinses with PBST, underwent antigen retrieval using Reveal Decloaker (Biocare Medical). Sections were incubated overnight at 4°C in primary antibodies: PSD-95 (Millipore, 1:400), IBA-1 (Abcam, 1:250), STING (Cell Signaling, 1:200), NeuN (Millipore, 1:200), MC-1, a conformational specific antibody for aggregated tau, (Gift from Peter Davies lab, 1:2500), GFAP (Abcam, 1:500), GM-130 (1:200), MBP (1:400). Following Alexafluor secondary antibody incubation at 1:500, sections were rinsed with PBST and mounted using Prolong Gold Anti-Fade reagent with DAPI. Slides were sealed with clear nail polish and kept at 4°C. Immunofluorescence was visualized using either LSM880 laser scanning confocal microscope and 3–4 images per animal were captured at 40X in CA1 and CA3 regions of hippocampus. Whole hippocampal images were taken at 10X with 11–13 planes and 2×2 images were stitched. After image capture, images were separated by color channel and underwent background subtraction and thresholding. Signal intensity was measured via ImageJ. Intensity of 1–2 control sections stained only with secondary antibody was averaged and subtracted from all sections. Signal intensity was normalized to area, averaged across sections for each animal, and plotted in Prism. Experimenters performing imaging and quantification were blinded.

IMARIS 3D Reconstruction

For the 3D reconstruction of microglia, we used Z-stack images containing 25 stacks at 40X magnification in the CA1 region of the hippocampus on the Zeiss LSM 880 confocal. Raw czi files were used for further analysis during IMARIS software (version 9.31, Oxford Instruments). First, IMARIS was used to reconstruct the microglial surface using filaments. We then graphed the measurements: segment terminal points (for branch length) and soma volume. We used linear mixed model analysis to account for multiple cells in each image and multiple hippocampal sections/animal.

Whole brain clearing and imaging for cFos Mapping

Fixed mouse brains from aged AD genetic cohorts were delipidated with a modified Adipo-Clear protocol^71,72. Briefly, perfusion fixed brain samples were washed with B1n buffer (H2O/0.1% Triton X-100/0.3 M glycine, pH 7), then transferred to a methanol gradient series (20%, 40%, 60%, 80%) in B1n buffer, 4 mL for each brain, 1 h for each step; then 100% methanol for 1 h; then overnight incubation in 2:1 mixture of DCM:methanol and a 1.5 h incubation in 100% DCM the following day; then 100% methanol for 1 h three times, and reverse methanol gradient series (80%, 60%, 40%, 20%) in B1n buffer, 30 minutes for each step. Samples were then washed in B1n buffer for 1 h and overnight. The above procedures were done at room temperature with rocking to complete delipidation. The delipidated samples were then blocked in PTxwH buffer (PBS/0.1% Triton X100/0.05% Tween 20) with 5% DMSO and 0.3M glycine for 3 h and overnight at 37°C, then washed with PTxwH for 1 h, 2 h, and overnight at room temperature. For staining, brain samples were incubated in primary antibody (monoclonal rabbit-anti-cFOS, clone 9F6, Cell Signaling 2250S, Lot#11, 1:200) diluted in PTxwH for 2 weeks at 37°C. After primary antibody incubation, samples were washed in PTxwH for 1 h, 2 h, 4 h, overnight, then 1 d, and then incubated in secondary antibody (Jackson ImmunoResearch 711-607-003, Alexa Fluor 647 donkey-anti-rabbit, Lot#149379, 1:200) diluted in PTxwH at room temperature for 2 weeks. Samples were then washed in PTxwH for 1 h, 2 h, 4 h, overnight, then 1 d. Samples were further fixed in 1% PFA at 4°C overnight, washed in PTxwH at RT (1h, 2h, 4h, overnight), then blocked in B1n at RT overnight, and washed in PTxwH at RT (1h, 2h, 4h). Samples were then bleached in 0.3% H2O2 at 4°C overnight, washed 3 times in 20mM PB (16mM Na2HPO4, 4mM NaH2PO4 in H2O) at RT for 2 hours. For clearing, samples were dehydrated with methanol gradient with water (20%, 40%, 60%, 80%) 1h each step, then 100% methanol for 1h, DCM/methanol mixture overnight, and 100% DCM for 1h twice the next day, then 100% methanol for 1h twice and methanol gradient with water (80%, 60%, 40%, 20%) 1h each step. Next, brains were washed twice (1h, overnight) in 20mM PB buffer and finally twice in PTS solution (25% 2,2’-thiodiethanol/10mM PB) (3h, overnight), then equilibrated with 75% histodenz buffer (Cosmo Bio USA AXS-1002424) with refractive index adjusted to 1.53 using 2,2’-thiodiethanol. Samples were stored at −20°C until acquisition. The cleared brain samples were imaged horizontally with tiling using the LifeCanvas SmartSPIM lightsheet microscope. 647 nm lasers were used for the GFP viral labeling IHC imaging with the 3.6×/0.2 detection lens. Lightsheet illumination is focused with NA 0.2 lens, and axially scanned with electrically tunable lens coupled to the camera (Hamamatsu Orca Back-Thin Fusion) in slit mode. Camera exposure was set at fast mode (2 ms) with 16b imaging. The X/Y sampling rate was 1.866 μm and Z step set at 2 μm.

Registration for Whole-brain imaging

Whole-brain 3D datasets were registered using the Common Coordinate Framework (CCF) reference brain, as previously described^73,74. After downsampling full-resolution datasets, registration was performed in two steps: a 3D affine transformation followed by a 3D B-spline. Advanced Mattes Mutual Information was used to compute the similarity for both affine and B-spline transformations. The optimizer used affine transformation was Adaptive Stochastic Gradient Descent, and Standard Gradient Descent for B-spline transformation. After registration, the output transformations were used to transform the annotation atlas. Registration was performed using Elastix registration toolbox⁷⁵.

c-Fos cell segmentation and cell counting

The whole-brain signal distribution of c-Fos cells was automatically detected using custom Python scripts. First, the signal background was reduced using a rolling ball algorithm, and the signal threshold was set using adaptive thresholding. The centroids were calculated using 3D-connected components, and area-based counting was performed using the ARA atlas, as previously described. Statistical analysis was performed using ANOVA, with animal genotype as the independent variable and area counts as the dependent variables, followed by Tukey post hoc analysis.

Single nuclei RNA sequencing

Nuclei isolation from frozen mouse hippocampi was adapted from a previous study [Habib 2017], with modifications. All procedures were performed on ice or at 4°C. In brief, postmortem brain tissue was placed in 1,500 μl of Sigma nuclei PURE lysis buffer (Sigma, NUC201–1KT) and homogenized with a Dounce tissue grinder (Sigma, D8938–1SET) with 15 strokes with pestle A and 15 strokes with pestle B. The homogenized tissue was filtered through a 35-μm cell strainer, centrifuged at 600g for 5 minutes at 4 °C and washed three times with 1 ml of PBS containing 1% bovine serum albumin (BSA, Thermo Fisher Scientific, 37525), 20 mM DTT (Thermo Fisher Scientific, 426380500) and 0.2 U μl⁻¹ recombinant RNase inhibitor (Ambion, AM2684). Nuclei were then centrifuged at 600g for 5 minutes at 4 °C and resuspended in 350 μl of PBS containing 0.04% BSA and 1× DAPI, followed by fluorescence-activated cell sorting to remove cell debris. The sorted suspension of DAPI-stained nuclei was counted and diluted to a concentration of 1,000 nuclei per μl in PBS containing 0.04% BSA.

For droplet-based snRNA-seq, libraries were prepared with Chromium Single Cell 3′ Reagent Kits v3 (10x Genomics, PN-1000075) according to the manufacturer’s protocol. The snRNA-seq libraries were sequenced on a NovaSeq 6000 sequencer (Illumina) with 100 cycles. Gene counts were obtained by aligning reads to the mm10 genome with Cell Ranger software (v.3.1.0; 10x Genomics). To account for unspliced nuclear transcripts, reads mapping to pre-mRNA were counted. Cell Ranger 3.1.0 default parameters were used to call cell barcodes. We further removed genes expressed in no more than three cells, cells with individual gene counts over 4,000 or less than 300, cells with UMI counts over 20,000 and cells with a high fraction of mitochondrial reads (>5%). Potential doublet cells were predicted using DoubletFinder for each sample separately, with high-confidence doublets removed. Normalization and clustering were done with the Seurat package v3.2.2 (Stuart, Butler 2019). In brief, counts for all nuclei were scaled by the total library size multiplied by a scale factor (10,000) and transformed to log space. A set of 2,000 highly variable genes was identified with SCTransform from the sctransform R packageFindVariableFeatures function with vst method. This returned a corrected individual molecular identifier count matrix, a log-transformed data matrix and Pearson residuals from the regularized negative binomial regression model. Principal-component analysis was done on all genes, and t-distributed stochastic neighbor embedding was run on the top 15 principal components. Cell clusters were identified with the Seurat functions FindNeighbors (using the top 15 principal components) and FindClusters (resolution = 0.1). In this analysis, the neighborhood size parameter pK was estimated using the mean variance-normalized bimodality coefficient (BCmvn) approach, with 15 principal components used and pN set as 0.25 by default. Sample integration was performed using FindIntegrationAnchors and IntegrateData functions in Seurat. For the baseline E3, E3P301S, E3S/Sand E3S/S/P301S cohort snRNAseq, we sequenced 3 hippocampi/group. For the cGASi-treated cohort, we also sequenced 3 hippocampi/group for the following groups: E3/control diet, E3S/S/control diet, E3P301S/control diet, E3S/S/P301S/control diet, E3/cGASi, E3S/S/cGASi, E3P301S/cGASi. For each cluster, we assigned a cell-type label using statistical enrichment for sets of marker genes and manual evaluation of gene expression for small sets of known marker genes. Differential gene expression analysis was done using the FindMarkers function and MAST (Finak 2015). To identify gene ontology and pathways enriched in the DEGs, DEGS were analyzed using the MSigDB gene annotation database (Subramanian 2005, Liberzon 2011). To control for multiple testing, we used the Benjamini–Hochberg approach to constrain the FDR.

Western blot

25 ug frontal cortex or hippocampal lysate prepared above were boiled for 5 minutes and run on 26-well 4–12% Bis-Tris gels (Invitrogen) using MES buffer (Invitrogen) for 1 hour. Proteins were transferred from gel onto PVDF membrane for 2 hours. Membranes were washed three times for 10 minutes each in TBS with 0.01% triton X-100 (TBST). Membranes were blocked for 30 minutes in 5% milk in TBST and incubated with APOE (1:800, CalBiochem), STING (1:1,000), pTBK (1:1,000), TBK (1:1,000) and GAPDH (1:5,000, GeneTex) antibodies overnight in cold room. The following day, membranes were washed three times for 10 minutes each in TBST and incubated in appropriate HRP secondary for 1 hour, rinsed, and followed by ECL development and imaging using Bio-Rad imager.

Lipidomic analyses: Lipid extraction

For lipidomic analysis, hippocampal samples from 6 mixed sex mice (aged 9–10 months old) were homogenized in with ice-cold milli-Q water containing a cOmplete, Mini Protease Inhibitor Cocktail using a bead mill homogenizer (VWR). Protein concentration in the lysates was quantified using the BCA assay, and approximately 50 μg of tissue lysate was transferred to Pyrex glass tubes with PTFE-lined caps. Lipid extraction followed the Folch method⁷⁶: 6 mL of ice-cold chloroform and methanol (2:1 v/v) and 1.5 mL of water were added to each sample. The tubes were thoroughly vortexed to ensure homogeneous mixing of polar and non-polar solvents. SPLAH internal standards were added before extraction. The samples were centrifuged at 1000 rpm for 20 minutes at 4°C to separate the organic and aqueous phases. The lower organic phase was carefully pipetted into a glass tube using a sterile glass pipette, avoiding the intermediate layer containing cell debris and precipitated proteins. The organic phase was then dried under a nitrogen stream until all solvents evaporated. Finally, the samples were reconstituted in 150 μL of chloroform:methanol (2:1) and stored at −80°C until mass spectrometry (MS) analysis.

Gangliosides extraction

The aqueous layer containing gangliosides was collected and dried under a gentle stream of nitrogen. The dried aqueous phase was reconstituted in 1 mL pure water and subsequently desalted by using Sola HRP SPE 30mg/2mL 96-well plate 1EA (Thermo Scientific #60509–001). Initially, the cartridges were cleaned 3 times with 1 mL of MeOH and equilibrated 3 times with water. Samples were loaded onto the column, washed 3 times with water and, finally, the gangliosides were eluted by three times 1 mL of MeOH. The eluate was dried under nitrogen flow and reconstituted in MeOH/H₂O/CHCl₃ (60:9:120, v/v/v).

Lipidomic analyses: LC-MS/MS analyssis of lipidomics

Lipids were separated using ultra-high-performance liquid chromatography (UHPLC) coupled with tandem mass spectrometry (MS/MS). UHPLC analysis was conducted on a C30 reverse-phase column (Thermo Acclaim C30, 2.1 × 150 mm, 2.6 μm) maintained at 50°C and connected to a Vanquish Horizon UHPLC system, along with an OE240 Exactive Orbitrap MS (Thermo Fisher Scientific) equipped with a heated electrospray ionization probe. Each sample (2 μL) was analyzed in both positive and negative ionization modes. The mobile phase included 60:40 water:acetonitrile with 10 mM ammonium formate and 0.1% formic acid, while mobile phase B consisted of 90:10 isopropanol:acetonitrile with the same additives. The chromatographic gradient involved: Initial isocratic elution at 30% B from −3 to 0 minutes, followed by a linear increase to 43% B (0–2 minutes), then 55% B (2–2.1 minutes), 65% B (2.1–12 minutes), 85% B (12–18 minutes), and 100% B (18–20 minutes). Holding at 100% B from 20–25 minutes, a linear decrease to 30% B by 25.1 minutes, and holding from 25.1–28 minutes. Flow rate of 0.26 mL/minute, injection volume of 2 μL, and column temperature of 55°C. Mass spectrometer settings included an ion transfer tube temperature of 300°C, vaporizer temperature of 275°C, Orbitrap resolution of 120,000 for MS1 and 30,000 for MS2, RF lens at 70%, with a maximum injection time of 50 ms for MS1 and 54 ms for MS2. Positive and negative ion voltages were set at 3250 V and 2500 V, respectively. Gas flow rates included auxiliary gas at 10 units, sheath gas at 40 units, and sweep gas at 1 unit. High-energy collision dissociation (HCD) fragmentation was stepped at 15%, 25%, and 35%, and data-dependent tandem MS (ddMS2) ran with a cycle time of 1.5 s, an isolation window of 1 m/z, an intensity threshold of 1.0e4, and a dynamic exclusion time of 2.5 s. Full-scan mode with ddMS² was performed over an m/z range of 250–1700, with EASYICTM used for internal calibration. The raw data were processed and aligned with LipidSearch 5.1, using a precursor tolerance of 5 ppm and a product tolerance of 8 ppm. Further filtering and normalization were conducted using an in-house app, Lipidcruncher. Semi-targeted quantification was performed by normalizing the area under the curve (AUC) to the AUC of internal standards and further normalized with the total quantified protein.

For LC-MS/MS analysis of gangliosides

Samples were analyzed using a Vanquish UHPLC system (Thermo Scientific) coupled to an Orbitrap Exploris 240 mass spectrometer (Thermo Scientific, #BRE725535). Separation was achieved on a Kinetex HILIC column (Phenomenex, #00D-4461-AN; 2.6 μm, 100 × 2.1 mm). The mobile phase consisted of solvent A (acetonitrile with 0.2% v/v acetic acid) and solvent B (water containing 10 mM ammonium acetate, pH 6.1, adjusted with acetic acid). The column temperature was maintained at 50 °C. A gradient elution was employed at a constant flow rate of 0.6 mL/minutes: 12.3% B at 0 minutes, a linear increase to 22.1% B from 1 to 15 minutes, followed by column equilibration at 12.3% B for 5 minutes.

Mass spectrometry analysis was performed in heated electrospray ionization (HESI) mode under the following conditions: spray voltage at −4.5 kV, heated capillary temperature at 300 °C, and vaporizer temperature at 250 °C. The gas settings were as follows: sheath gas at 40 units, auxiliary gas at 5 units, and sweep gas at 1 unit. The ion transfer tube temperature was maintained at 300 °C. For MS1 analysis, the Orbitrap resolution was set to 120,000 with a scan range of 700–1800 m/z, RF lens at 60%, and an AGC target set to standard. For MS2, the Orbitrap resolution was set to 30,000. Internal calibration was achieved using EASY-IC^™. LipidSearch 5.1, using a precursor tolerance of 5 ppm and a product tolerance of 8 ppm. Further filtering and normalization were conducted using an in-house app, Lipidcruncher. Quantification was achieved by normalizing the area under the curve to GM3-d5 standards, followed by further normalization to the amount of protein used in the preparation.

Bulk sequencing analysis

Primary microglia isolated from E3 and E3S/Spups were isolated and treated with tau or amyloid as described above and collected 24 hours later. Total RNA was extracted from the samples using QuickRNA MicroPrep Kit (Zymo Research). After RNA isolation, total RNA integrity was checked using a 2100 Bioanalyzer (Agilent Technologies), and concentrations were measured by Nanodrop (Thermo Fisher. RNA was isolated from microglia using the Quick-RNA MicroPrep Kit (Zymo Research, R1051). RNA was shipped to Novogene for library preparation and bulk RNA sequencing. Differential gene expression was analyzed with the DESeq2 1.38.3 package⁷⁷. Counts were normalized using the median of ratios method. Genes with <15 counts across all samples were excluded from analysis. To control for multiple testing, we used the Benjamini–Hochberg approach to constrain the FDR. Pathway analysis was done using the MSigDB gene annotation database. Ingenuity Pathway Analysis (QIAGEN, Inc.) was used to identify gene activation networks and upstream regulators. Raw read counts per gene were extracted using HTSeq-count v0.11.2⁷⁸.

Tau pulse-chase experiment

For pulse-chase experiments, primary microglia were plated after shaking onto 8-well chamber slides at a density of 1 × 10⁵ cells/well. 24 hours after plating, media with 1ug/ml of 0N4R tau fibrils from Dr. Sue-Ann Mok’s lab was added to each well or fresh media was changed for control wells. After 2 hours, media was collected from both conditions and ½ of the wells were washed 3 times with PBS and fixed for 15 minutes with 4% PFA and sucrose. After 15 minutes, cells were washed 3 times with PBS. The remaining wells were exchanged with fresh media for another 22 hours. At 24 hours post-initial tau treatment, media was collected and remaining cells were fixed as previously described. Fixed chamber wells were stored at 4C until immunocytochemistry for tau was performed. Cells were washed with PBS and blocked for 1 hour at room temperature with 10% BSA in TBST. After washing with PBS again, cells were incubated with HT-7 tau antibody (1:600) diluted in 1% BSA in TBST) for 2 hours at room temperature, then washed 3 times, and incubated with Alexa 568 secondary antibody (1:500) for 1 hour at room temperature. Coverslips were placed with anti-fade reagent containing DAPI and sealed with nail polish.

iMGL Tau Uptake Assay

iMGLs were seeded to a Poly-D lysine-coated 6-well plate at a density of 1.5 × 10⁵/well. After 48 hours of seeding the cells, 1 μg/ml of 0N4R tau fibrils from Dr. Sue-Ann Mok’s lab was treated to iMGLs. After 24 hours, the expression of tau in the media was measured using a human total tau ELISA kit (Invitrogen). ELISA tau was normalized with protein concentration.

IP-10 ELISA

To measure the secreted IP-10 from iMGLs after tau treatment, we used human IP-10 ELISA kit (Invitrogen). iMGLs were seeded in a Poly-D lysine-coated 6-well plate at a density of 1.5 × 10⁵/well. After 48 hours, 1 μg/ml of tau was treated to iMGLs. After 24 hours of incubation, the human IP-10 from iMGLs was measured using a human IP-10 kit. ELISA IP-10 was normalized with protein concentration.

Supplementary Material

2

Supplemental Table 1. DEGs of all cell types, related to Figures 3–5.

3

Supplemental Table 2. Oligodendrocyte cell markers, related to Figure 3.

4

Supplemental Table 3. Microglia cell markers, related to Figure 4.

5

Supplemental Table 4. Primary microglia quality control and supplemental data, related to Figure 5.

6

Supplemental Table 5. DEGs of inhibitory neurons, excitatory neurons and microglia treated with cGAS-inhibitor, related to Figure 7.

Highlights.

• Christchurch R136S mutation in APOE3 markedly reduces tau load in tauopathy mice

• R136S improves synapses, network function, and myelin in APOE3 tauopathy mice.

• R136S dampens tau-induced cGAS-STING-IFN activation in mouse and human microglia.

• cGAS inhibitor preserves synapses and mimics R136S transcriptomics in tauopathy mice.