Humanized APOE genotypes influence lifespan independently of tau aggregation in the P301S mouse model of tauopathy

Mouse husbandry and experimental procedures were performed in accordance with the protocols and policies approved by the Institutional Animal Care and Use Committee at the University of Florida. All mice were maintained under a 12-h light/dark cycle and had access to water and food ad libitum. The mice were maintained under specific pathogen-free conditions. PS19 mice were obtained from Jackson Labs and maintained on a B6/C3H background as heterozygotes for P301S tau transgene, developing age-progressive hindlimb paralysis at 9–12 months of age [12]. APOE targeted replacement (TR) mice were obtained from Duke University and were maintained as homozygotes on C57BL6 background [13–15]. PS19 mice were mated with APOE homozygote mice to produce N1 cohorts that were heterozygous for APOE. N2 cohorts of PS19 mice that are heterozygous or homozygous for APOE were generated by back-crossing the N1 generation mice with corresponding APOE homozygous mice. APOE and tau levels in these N1 and N2 cohorts were described in a previous study [11]. Cohorts were set up to be aged until the animals reached hindlimb paralysis. At endpoint, all euthanasia was performed with intracardiac perfusion of cold phosphate buffered saline (PBS) containing heparin and the brains were fixed in 10% normal buffered formalin (Fisher Scientific). The fixed brains were then sliced sagitally and processed for paraffin embedding.

All brains were obtained from the University of Florida Neuromedicine Human Brain and Tissue Bank. Information on individual patients are summarized in Additional file: Table S1. 6

For APOE genotyping, DNA was extracted from formalin fixed paraffin embedded (FFPE) sections or frozen brain tissue using DNA extraction kits (GeneRead DNA FFPE Kit, Qiagen; and QIAamp DNA Mini Kit, Qiagen). Two single nucleotide polymorphisms (SNPs) within the APOE gene, rs7412 and rs429358, were detected using the Human TaqMan™ SNP Genotyping Assay (Applied Biosystems, Cat. no. 4351376) and TaqMan™ Genotyping Master Mix (Applied Biosystems, Cat. no. 4371355). The relative fluorescent units (RFU) were measured by performing qRT-PCR. The frequencies of APOE alleles and corresponding genotypes were obtained by calculating the ratio and combination of RFUs for both SNPs.

FFPE slides with mouse brain sections were deparaffinized and probed with primary antibodies as described before [11]. For antigen retrieval, slides receiving AT8 (1:1000; Invitrogen MN1020), AT100 (1:1000; Invitrogen MN1060), AT270 (1:1000; Invitrogen MN1050), AT180 (1:1000; Invitrogen MN1040) and glial fibrillary acidic protein (GFAP) (1:3000; Cell Signaling) antibodies were steamed in water at high pressure for 15 min, while Iba-1 (1:2000; Wako) antibody slides were steamed in citrate buffer pH 6.0 (Target Retrieval Solution, Dako). Slides were incubated in 3% hydrogen peroxide (Fisher Scientific) for 20 min to block endogenous peroxidase activity and then washed three times in PBS for 5 min each. Slides were then blocked in 2% fetal bovine serum (FBS) (Hyclone, GE) for 45 min before incubating in primary antibody diluted in block solution overnight at 4 °C. The following day, slides were washed and appropriate secondary antibody (ImmPRESS Polymer Reagent, Vector Labs) was applied for 30 min at room temperature. Following PBS washes, color was developed using 3,3’-diaminobenzidine (Vector DAB, Vector Labs) and slides were counterstained with haematoxylin (Vector Labs). Next, brain sections were dehydrated in a series of ethanol, cleared in xylene, mounted in Cytoseal-60 media (Fisher Scientific) and coverslipped.

Immunostained images were captured using a Scanscope XT image scanner (Aperio, Vista, CA). Percent immunoreactivity was computed using the Aperio Positive Pixel Count program (Aperio, Vista, CA). The data is shown as the average %immunoreactivity ± S.E.M. per group. Statistical comparisons were conducted using 1-way ANOVA (GraphPad Prism 7).

FFPE tissues were sectioned sagittally, starting 1 mm from the midline. Initial slides were stained with haematoxylin and eosin, their location identified using Paxinos and Franklin’s mouse brain atlas and further slides were prepared according to desired bregma locations. 5 µm sections were collected spaced at regular intervals of 25 µm followed by an interval of 250 µm. This pattern was repeated until reaching the target area, which was approximately 3.2 mm from the midline bregma. Slides were stained with haematoxylin and eosin at target area. Assessments of ventricle volume were made at two distances from bregma. One measurement of ventricle volume was made at 2.8–2.76 mm from bregma, and a second measurement was made at 3.2 mm from bregma. Using an Aperio Image Scope application to view the images, the entire perimeter of the lateral ventricles that were visible in the images was outlined and the area of the marked areas were recorded. In more medial sections where the ventricles were discontinuous, the total area of the ventricles was summed.

Frozen hemibrains were cryo-pulverized and used for tau extraction as detailed earlier [16]. Briefly, brain fractions S1 and S3 containing tau with differential soluble properties were obtained by treating the brain homogenates with a sequential series of buffers with increasing detergent concentrations—Tris buffered saline, high salt buffer containing 10% sucrose and finally 1% sarkosyl solution. The S1 fraction refers to clarified homogenate prepared in Tris buffered saline while S3 fraction refers to the pellet obtained following 1% sarkosyl incubation. The S3 pellet was further solubilized in 4 M urea and 2% SDS and used in western blotting. All steps in the protocol included addition of protease and phosphatase inhibitors (Pierce Protease and Phosphatase Inhibitor Mini Tablets, Thermo Scientific). Immunoblotting methods were followed as described earlier [16]. Membranes were incubated with primary antibodies [Iba-1 (1:250; Novus Biologicals), GFAP (1:1000; Cell Signaling), GAPDH (1:1000; Abcam), AT8 (1:1000; Invitrogen MN1020), AT100 (1:1000; Invitrogen MN1060), AT270 (1:1000; Invitrogen MN1050), AT180 (1:1000; Invitrogen MN1040), PHF1 (1:1000; Peter Davies), total tau (1:1000; Abcam ab254256] and signals detected using multiplex Li-Cor Odyssey Infrared Imaging system (Li-Cor Biosciences, Lincoln, NE). Relative band intensity was quantified using ImageJ software (NIH).

Human brain tissue sections were received from the UF Neuromedicine Human Brain and Tissue Bank (UF HBTB) following protocols approved by the Institutional Review Board of the University of Florida. Staging of AD neuropathologic changes and associated co-pathologies was conducted per guidelines published by the National Institute on Aging-Alzheimer’s Association [17] by a board-certified neuropathologist (SP). 5 μm thick tissue sections of FFPE brain tissue specimens were deparaffinized and rehydrated. Heat-induced epitope retrieval was performed in a pressure cooker (Tintoretriever, Bio SB) for 15 min at high pressure in a 0.05% Tween-20 solution. Endogenous peroxidase was quenched by incubation of sections in 1.5% hydrogen peroxide/0.005% Triton-X-100 diluted in sterile PBS (Invitrogen) for 20 min, following multiple washes in tap water and subsequently, 0.1 M Tris, pH 7.6. Non-specific antibody binding was minimized with sections incubated in 2% FBS/0.1 M Tris, pH 7.6. Primary antibodies were diluted in 2% FBS/0.1 M Tris, pH 7.6 (anti-Aβ antibody Ab5 [Todd E. Golde], 1:1000; anti-tau (AT8), 1:5000, invitrogen MN1020; anti-α-synuclein (94-3A10) [Benoit I. Giasson], 1:10,000). Sections were incubated with primary antibody over night at 4 °C, washed one time in 0.1 M Tris, pH 7.6, followed by 2% FBS/0.1 M Tris, pH 7.6 for 5 min, incubated in goat anti-rabbit IgG horseradish peroxidase (HRP) conjugated secondary antibody (Millipore Sigma) for 1 h, additionally washed one time in 0.1 M Tris, pH 7.6, followed by 2% FBS/0.1 M Tris, pH 7.6 for 5 min, and incubated in VectaStain ABC Peroxidase HRP Kit (diluted in 2% FBS/0.1 M Tris, pH 7.6 at 1:1000) for 1 h. After a final wash in 0.1 M Tris, pH 7.6 for 5 min, immunocomplexes were visualized using 3,3′-diaminobenzidine (DAB). Tissue sections were counterstained with hematoxylin (Sigma Aldrich, St. Louis) for 2 min, dehydrated and cover slipped using Cytoseal 60 mounting medium (Thermo Scientific).

For analysis of stains, slides were scanned on an Aperio AT2 scanner (Leica biosystems) at 20 × magnification. Digital slides were analyzed using the QuPath platform (version 0.3.2) [18]. Cortex and white matter were annotated for regional analysis. After exclusion of tissue and staining artifacts, the ‘Positive Pixel Detection’ tool (e.g., Downsample factor 2, Gaussian sigma 1 μm, Hematoxylin threshold (‘Negative’) 1.5 OD units, DAB threshold (‘Positive’) 0.2 OD units) was used to determine the percentage of area covered by pathological staining. ‘Positive Pixel Detection’ thresholds were adjusted for individual stains.

The number of samples were 3 mice each for parental E2H, E3H and E4H cohorts and 4 mice each for PS/E2H, PS/E3H and PS/E4H cohorts. All mice were male, except 1 female for E4H, 1 female for PS/E2H and 2 females for PS/E4H cohort. RNA was extracted from frozen, pulverized forebrains using TRIzol reagent (Invitrogen). Extracted RNA was cleaned using RNeasy mini extraction kit with an on-column DNase treatment (Qiagen). RNA quality was quantified with the Qubit RNA HS assay (Invitrogen) and transferred to LC Sciences (Houston, TX) for library construction and Rnaseq. Poly(A) RNA sequencing library was prepared following Illumina’s TruSeq-stranded-mRNA sample preparation protocol. RNA integrity was checked with Agilent Technologies 2100 Bioanalyzer. Poly(A) tail-containing mRNAs were purified using oligo-(dT) magnetic beads with two rounds of purification. After purification, poly(A) RNA was fragmented using divalent cation buffer in elevated temperature. Quality control analysis and quantification of the sequencing library were performed using Agilent Technologies 2100 Bioanalyzer High Sensitivity DNA Chip. Paired-ended sequencing was performed on Illumina’s NovaSeq 6000 sequencing system. Firstly, Cutadapt [19] and perl scripts in house were used to remove the reads that contained adaptor contamination, low quality bases and undetermined bases. Then sequence quality was verified using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/↗). We used HISAT2 [20] to map reads to the genome of ftp://ftp.ensembl.org/pub/release-101/fasta/mus_musculus/dna/↗. The mapped reads of each sample were assembled using StringTie [21]. Then, all transcriptomes were merged to reconstruct a comprehensive transcriptome using perl scripts and gffcompare. After the final transcriptome was generated, StringTie [21]] and ballgown (http://www.bioconductor.org/packages/release/bioc/html/ballgown.html↗) was used to estimate the expression levels of all transcripts. StringTie [21] was used to perform expression level for mRNAs by calculating FPKM (fragments per kilobase of exon per million mapped fragments). mRNAs differential expression analysis were performed by R package DESeq2 [22] between two different groups (and by R package edgeR [23] between two samples). The mRNAs with the parameter of false discovery rate (FDR) below 0.05 and absolute fold change ≥ 2 were considered differentially expressed mRNAs. Gene Set Enrichment Analysis (GSEA) with GO and KEGG databases were done using gene sets with |NES|≥ 1 & NOM p-value < 0.05 & FDR q-value < 0.25, where NES = normalized enrichment score, NOM = NOM and FDR = false discovery rate.

Gene lists identifying cell-types within the brain were described previously [24–30]. The geometric mean of the FPKM for genes identified for each cell or cellular sub-type was calculated from each animal. Group means were calculated, and between group significance values determined by 1-way ANOVA. Outliers were removed from the group if their value fell outside of 1.5 times the inter-quartile range (IQR).

Immunohistochemical data was analyzed using 1-way ANOVA unless otherwise indicated in figure legend. Outliers, if applicable, were removed using Rout’s test with Q = 1%. The neuropathological scoring was done in a blinded manner. All data was assembled using Adobe Photoshop Elements.

As described earlier [11], mice carrying humanized APOE alleles [13–15] were bred with P301S tau transgenic Line PS19 mice [12] to generate tau transgenic mice that are heterozygous for APOE (PS/E2h, PS/E3h and PS/E4h) or homozygous (PS/E2H, PS/E3H and PS/E4H) for human APOE. PS19 mice show progressive paralysis as a result of accumulation of tau pathology in the spinal cord and brain [12]. In our colony, the median age to humane endpoint of bilateral hindlimb paralysis for PS19 mice on the B6/C3H background, carrying mouse Apoe, is 308 days (n = 7 males, 9 females) (Fig. 1A). We examined whether the presence of specific human APOE or the gene dosage of APOE influenced lifespan in PS19 mice (Fig. 1A). The median number of days to humane endpoint of bilateral hindlimb paralysis for PS/E2H, PS/E3H and PS/E4H mice was 332, 381 and 318 days, respectively (Fig. 1A). The longer average survival for PS/E3H mice was statistically different from PS/E2H mice (p = 0.015), PS/E4H mice (p = 0.008) and PS19 mice (p = 0.003). There was no statistical difference in median age to paralysis between PS/E2H, PS/E4H and PS19 cohorts. Previous studies, using coronal brain sections of PS19 crossed with humanized APOE mice, showed that APOE4 exacerbates ventricular enlargement [7]. In our study, we used sagittal sections from two selected brain areas, one a medial and another a lateral location with respect to midline to explore changes in ventricular volume. In our cohorts of PS19xAPOE mice, we observed high variability in ventricular volumes and did find any statistically significant differences in ventricular volume between the three genotypes (Additional file 1: Fig. S1A, B).

We investigated the levels of tau and phospho-tau in brains of these paralyzed mouse cohorts (Fig. 1, Additional file 2: Fig. S2). Using an antibody against total tau, we observed that all of our mouse lines showed comparable tau levels in both the detergent soluble (S1) and detergent insoluble (S3) fractions (Fig. 1B). Using AT8 (pSer202/Thr205) antibody staining in brain sections, we observed lower levels of phospho-tau immunoreactivity in PS/E2H mice compared to PS/E3H (p < 0.05) and PS/E4H (p < 0.01) mice in the hippocampus (Fig. 1C). Cortical AT8 immunoreactivity was somewhat lower in PS/E2H mice compared to PS/E3H and PS/E4H mice, though this did not reach statistical significance due to high variability (Fig. 1C). Immunoblotting of soluble and detergent-insoluble fractions from hemibrains showed that AT8-immunopositive detergent-soluble tau was equivalent between the three APOE genotypes (Fig. 1D). Notably, we failed to detect insoluble tau showing AT8-site phosphorylation by immunoblotting in these mice (Additional file 2: Fig. S2G). These findings suggest that most of the AT8 immunoreactivity seen in the cortex of PS19 mice is from phosphorylated soluble tau.

Next, we investigated whether other tau phosphorylation epitopes showed any differential association with APOE genotypes. Immunohistochemistry analysis of pThr212/Ser214 (AT100), pThr181 (AT270) and pThr231 (AT180) burden showed equivalent levels between PS/E2H, PS/E3H and PS/E4H mice (Additional file 2: Fig. S2A-B). Immunoblotting of soluble and detergent-insoluble fractions with these antibodies showed variable patterns of immunoreactivity (Additional file 2: Fig. S2C–G). Consistent with the immunohistological staining, the level of AT100 immunostaining in soluble fractions was low (Additional file 2: Fig. S2C). In the limited samples we analyzed here, soluble AT100 levels were somewhat lower in PS/E2H mice as compared to other genotypes (Additional file 2: Fig. S2C; p < 0.05 relative to PS/E3H and PS/E4H). Robust levels of AT270 and AT180 immunoreactivity were detected in soluble fractions from all genotypes (Additional file 2: Fig. S2D, E). We also analyzed biochemical levels of PHF1 (Ser396/404), finding robust staining in soluble fractions as well as detectable amounts in detergent-insoluble fractions (Additional file 2: Fig. S2F). In the detergent-insoluble fractions, AT100-, AT180- and AT8-positive phospho-tau were undetectable (Additional file 2: Fig. S2C, E, G). Presence of total tau on these same blots confirmed that these phospho-tau epitopes were specifically not induced. Levels of silver-positive Gallyas staining showed high variability in forebrains of these mice and thus were not considered for analysis (data not shown). Collectively, these data suggest that most of the cortical phospho-tau present in the brains of these mice was present as detergent soluble tau. Since our data shows that PS/E3H and PS/E4H mice had equivalent phospho-tau pathology, but PS/E4H mice reached end-stage (hindlimb paralysis) earlier than PS/E3H mice, this suggests that the E3 allele imparts some resilience to tau-mediated pathogenesis.

We examined hippocampal and cortical microgliosis and astrogliosis in PS19 mice carrying different APOE alleles, using Iba-1 and GFAP antibodies respectively. Hippocampal burden of Iba-1 immunoreactivity was somewhat lower in PS/E2H mice compared to PS/E3H and PS/E4H mice (PS/E2H vs PS/E3H: p < 0.05), with the latter two genotypes showing equivalent levels (Fig. 2A). There was no difference in cortical microglia staining in the three cohorts (Fig. 2A). Immunoblotting analyses of Iba-1 levels from the whole forebrain showed equivalent levels among the three APOE genotypes of PS19 mice (Additional file 3: Fig. S3A), indicating region-specific microgliosis in these mice.

We observed lower hippocampal burden of GFAP in PS/E2H mice compared to PS/E3H mice (p < 0.01), with no difference observed between PS/E3H and PS/E4H mice (Fig. 2B). There was no difference in cortical astroglia staining among the three cohorts of bigenic mice (Fig. 2B). GFAP immunoblotting showed similar levels in PS19 mice bearing mouse Apoe or human APOE2, APOE3 or APOE4 (Additional file 3: Fig. S3B). Notably, the PS/E3H bigenic mice showed higher astrogliosis compared to the corresponding parental E3H mice (Additional file 3: Fig. S3B; p < 0.05), while the other group comparisons (PS/E2H vs E2H and PS/E4H vs E4H) did not reach statistical significance. Our results indicate that presence of the tau transgene induces variable levels of astrogliosis in all parental APOE genotypes, and most consistently in E3H mice.

We investigated the lifespan of PS19 mice carrying one human APOE allele and one endogenous Apoe allele. In these heterozygous APOE mice, we did not observe any differential time to paralysis based on APOE genotype. The median ages of survival of the PS/E2h, PS/E3h and PS/E4h mice were 358, 325 and 365 days respectively (Additional file 4: Fig. S4A, B). Interestingly, the PS/E2h mice had a longer lifespan compared to its APOE homozygous counterpart, PS/E2H mice (p = 0.028) (Additional file 4: Fig. S4B), showing the detrimental influence of increased APOE2 gene dose on age to paralysis. Likewise, the PS/E4h mice had a longer lifespan compared to PS/E4H mice (p = 0.013) (Additional file 4: Fig. S4B). Although the median age to paralysis for PS/E3h mice was substantially shorter than that of PS/E3H mice (325 vs 381 days), the difference was not statistically significant. Within the APOE heterozygous PS19 cohorts, we did not observe any statistically significant changes in the cortical or hippocampal burden of phospho-tau or Iba-1 (Additional file 4: Fig. S4C, D). Cortical (but not hippocampal) levels of GFAP was lower in PS/E2h mice compared to PS/E3h mice (p < 0.05), with no difference noted between PS/E3h and PS/E4h mice (Additional file 4: Fig. S4E). Whether this difference is biologically meaningful is uncertain, given the high variability in reactivity within these cohorts.

To illuminate the potential mechanisms by which APOE could influence the neuropathology in PS19 mice, we conducted bulk RNAseq analysis of the forebrains of PS19 mice homozygous for human APOE and corresponding parental APOE homozygous cohorts (Additional file 6: Table S2). Comparison of the three parental APOE homozygous cohorts showed clear APOE genotype-based differences in gene expression (Fig. 3; Additional file 6: Table S2A-C). Relative to E3H mice, the chemokine genes, Ccl21b and Ccl21d, were downregulated and the cysteine protease, Capn11, was upregulated in E2H mice. Gene Ontology (GO) enrichment analysis showed that multiple GO terms associated with immune function (chemokine receptor binding, chemotaxis and cellular response to cytokines) were differentially altered in E2H mice. Analysis of KEGG pathways also showed consistent enrichment in immune pathways in E2H mice relative to E3H mice (Fig. 3B–D). Fewer differences were observed between E4H and E3H mice at this age. Notable genes that were upregulated in E4H mice (relative to E3H mice) were Serpina3n (cysteine protease inhibitor) and Eps8l1 (signaling adaptor). Enrichment of GO terms and KEGG pathways showed few significant events – most of which were related to metabolic pathways and cellular response to DNA damage (Fig. 3F–G). Comparison of E4H to E2H revealed several immune genes upregulated in E4H mice, such as Ccl121b and Ccl21d (Fig. 3H). The GO term and KEGG pathway profiles in E4H vs E2H comparison were similar to E2H vs E3H comparison, with immune pathways identified as enriched in both the analyses (Fig. 3I, J; C, D). Notably, while immune genes were downregulated in E2H mice (relative to E3H mice), these genes were upregulated in E4H mice (relative to E2H mice).

There were substantial differences in PS19 mice carrying the three APOE alleles (Additional file 5: Fig. S5A). Genes that were differentially upregulated in the comparisons of parental APOE lines were also found in the PS19 tau transgenics carrying the corresponding APOE allele (examples include Eps8l1 in E2H vs E3H, Eps8l1 in E4H vs E3H and Ccl21d in E4H vs E2H comparisons) (Fig. 3B, E, H vs Additional file 5: Fig. S5 B, C, D; Additional file 6: Table S2D-F). We normalized the within-group differences to gain insights into the differential gene expression (DEG) patterns that are determined by APOE genotype or presence of tau transgene (Additional file 5: Fig. S5E–H). The common DEG list showed few genes that are altered and few common pathways were found to be enriched when we investigated differences between E2H vs E3H mice or E4H vs E3H mice, after allowing for the effect of PS19 genotype to be normalized (Additional file 5: Fig. S5E, G). For example, relative to APOE3, NADP + transhydrogenase activity and myeloid cell chemotaxis were downregulated in APOE2 genotypes, whereas no upregulated pathways were identified in this comparison. On the other hand, relative to APOE3, dUMP metabolic process was down regulated and translational fidelity pathway was upregulated in APOE4 mice. We further tested for the effect of the PS19 genotype, after normalizing for the differential contribution of APOE. The presence of tau transgene led to upregulation of a larger number of genes in these mice, with 127 genes commonly upregulated in E2H vs E3H comparison and 252 genes upregulated in E4H vs E3H comparison (Additional file 5: Fig. S5F–H). Expression of tau transgene leads to a robust immune response and stress response that are upregulated in both APOE2 (relative to APOE3) and APOE4 (relative to APOE3) genotypes (Additional file 5: Fig. S5F–H). This suggests that tau expression imparts an overwhelming immune phenotype in these APOE homozygous mice. There were fewer genes that were commonly downregulated in these comparisons (Additional file 5: Fig. S5F–H), one being the keratinocyte activation pathway that was downregulated in both these DEGs (E2H vs E3H and E4H vs E3H).

We wanted to compare the individual DEG in PS19 mice within each APOE genotype to understand how tau transgene expression is related to eventual tau-mediated neurodegeneration phenotype. Comparing the PS/E4H to E4H mice, we noted that 216 genes were upregulated while 40 genes were down regulated (Fig. 4A; Additional file 6: Table S2G). Genes corresponding to glial activation, such as Gfap, C1qa, C1qb and Mpeg1, were highly upregulated (Fig. 4A). Notable genes that have been reported to be part of the APOE-associated pathogenic signature in AD, Trem2 (↑3.7x, p = 1.46E−22) and Clec7A (↑36.2x, p = 3.1E−25), were upregulated in PS/E4H mice compared to E4H mice. APOE, Trem1 and Treml1 remained unchanged, but Treml2 was upregulated (↑4.7x, p = 4.22E−05). We also examined the cell types most perturbed in these mice. There was a slight decrease in neuronal gene expression (p = 0.052) and robust increase in glial genes (astrocyte, p = 0.0081; microglia, p < 0.0001; oligodendrocyte, p = 0.001) (Fig. 4B). We further examined whether a specific AD-associated astrocyte profile could be correlated with the early death phenotype. We found that pan-reactive astrocyte profile was upregulated (p = 0.0053), consistent with Gfap upregulation, but neither A1 or A2 astrocyte profiles were specifically induced. Among other aging/AD-associated gene expression signatures identified from human brains, we could identify plaque induced gene (PIG, p = 0.0342) and the H2M (p < 0.0001) homeostatic microglia signatures in the PS/E4H mice (Fig. 4B). The upward trends in microglia neurodegenerative signatures (MGnD and DAM) did not survive multiple testing correction (Fig. 4B). We performed pathway analysis using enriched gene sets. Using the GO and KEGG databases, we identified immune processes as the main pathways emerging in the PS/E4H vs E4H comparison. Examples of enriched pathways in PS/E4H mice include Il-6 and Ifn-γ signaling, antigen processing and presentation, Toll-like receptor and chemokine signaling pathway (Fig. 4C, D).

Compared to age-matched E3H mice, the PS/E3H mice showed 155 transcripts upregulated and 22 transcripts downregulated (Fig. 5A, Additional file 6: Table S2H). Mostly immune system mediators, such as Itgax, Fcrls, Lag3 and Gfap were upregulated in the PS/E3H mice relative to age-matched E3H mice (Fig. 5A). Similar to changes observed in PS/E4H cohorts, Trem2 (↑2.8x, p = 6.32E−13) and Clec7A (↑19.5x, p = 7.4E−18) were also upregulated in PS/E3H mice compared to E3H mice, though the strength of association (based on fold change and p values) was lower in the in PS/E3H mice. APOE, Trem1 and Treml1 levels also remained similar between PS/E3H and E3H mice, while Treml2 was upregulated (↑4.1x, p = 8.7E−04). The cell types that were most perturbed in the PS/E3H mice were microglia (p = 0.0025) and oligodendrocytes (p = 0.0102) (Fig. 5B). Similar to PS/E4H vs E4H comparison, we observed global changes in astrocytic gene expression (p = 0.0295) in PS/E3H mice, though the A1/A2 markers were not changed (Fig. 5B). Among the aging/AD-associated gene expression profiles examined in the PS/E3H vs E3H comparison, only the H2M signature was upregulated (p < 0.0001), with no changes in ARM, DAM, MGnD and PIG (Fig. 5B).

As there was a distinct difference in lifespan of the PS/E4H and PS/E3H mice, we were curious whether DEG analysis between these two cohorts could provide insights into tau-mediated pathological effects in E4H vs E3H background. A total of 45 genes were significantly altered, with 14 genes upregulated in PS/E4H mice (Additional file 5: Fig. S5C; Additional file 6: Table S2E). Analysis of the enriched gene sets revealed gene ontogeny and pathways related to pyrimidine metabolism, prolactin signaling, histone demethylation and immunoglobulin pathways most enriched in PS/E4H relative to PS/E3H mice (Additional file 5: Fig. S5I–J). Most of these pathway changes involved only a few genes that were enriched in the comparison analysis, ranging from 1 to 3 genes and the strength of the association (p values and gene numbers enriched in each pathway) were also weak. While our expectation was that the pathogenic interaction of APOE-TREM2 would be exacerbated in PS/E4H mice, surprisingly, we did not note any differential regulation (measured by fold change and p-value of significance) of Trem family members in the PS/E4H mice compared to PS/E3H mice (Additional file 5: Fig. S5K).

Relative to age-matched E2H mice, the PS/E2H showed 82 upregulated and 11 down regulated genes (Fig. 6, Additional file 6: Table S2I). Genes that were most perturbed were microglia or astrocyte-specific, such as C1qa, Grn, Itgax and Gfap (Fig. 6A), reiterating the commonalities of tau overexpression paradigm observed previously in PS/E3H and PS/E4H mice (Fig. 4 and 5). PS/E2H mice showed upregulated transcripts of Trem2 (↑2.2x, p = 3.7E−07), Clec7a (↑16.6x, p = 2.4E−15) and Treml2 (↑2.9x, p = 0.018), whereas APOE, Trem1 and Treml1 were unchanged. We observed only modest upregulation in microglial cell profile (p = 0.019), with other cell types (neurons, astrocytes and oligodendrocytes) remaining unaffected (Fig. 6B). The pan-reactive astrocyte gene expression profile (p = 0.0038), but not the A1/A2, was also identified (Fig. 6B). Among the aging/AD-associated transcriptional signatures, we found robust upregulation of H2M (p < 0.001) and modestly increased PIG (p = 0.0533) signature (Fig. 6B). Other AD-associated profiles, such as ARM, DAM or MGnD were not affected (Fig. 6B). Pathway analysis revealed immune system processes and defense response being enriched overall, with specific involvement of antigen processing, complement cascades, and NOD-like receptor signaling cascades in PS/E2H mice (Fig. 6C, D).

In order to understand the relationship between tau phosphorylation and APOE genotype, we next turned our attention to human AD cases that were stratified for APOE4 or non-APOE4 genotypes (Fig. 7; Additional file 6: Table S1). All of the AD cases had a primary neuropathological diagnosis of AD [17]. Some of the AD patients also exhibited co-pathology in form of α-synuclein positive Lewy bodies alongside canonical phospho-tau and amyloid β (Aβ) deposits (Additional file 6: Table S1). These patients were classified as having ‘mixed’ pathologies (open diamonds, Fig. 7). The age-matched healthy controls with no history of dementia were all non-E4 genotype (Additional file 6: Table S1). Immunostaining with pan Aβ and AT8 antibodies demonstrated that AD patients of both E4 and non-E4 genotypes showed the expected robust patterns of Aβ deposits and intracellular tau tangles relative to age-matched controls (****p < 0.0001 for Aβ; **p < 0.01 for phospho-tau) (Fig. 7A, B). We did not observe any obvious differences in the Aβ and tau burden between the E4 and non-E4 individuals within the neuropathologically-characterized AD group (Fig. 7A, B). Interestingly, we found that AD patients with E4 allele had slightly higher α-synuclein burden relative to non-E4 AD patients (p = 0.030) and controls (p = 0.0082) (Fig. 7C). We then wanted to investigate whether there was a correlation between tau and Aβ or tau and α-synuclein based on APOE4 genotype within AD patients. Phospho-tau burden was correlated to Aβ burden in the E4 AD group (p = 0.019) but not in the non-E4 AD group (Fig. 7D). On the other hand, tau and α-synuclein burden showed no correlation in either non-E4 or E4 AD cohorts (Fig. 7E).

All rodent studies were approved by the University of Florida IACUC. All human autopsy materials were handled with the approval of the Institutional Review Board (IRB) at University of Florida. All patients provided written informed consent.

All authors have approved the manuscript and agree with its submission.

The authors declare no competing interests.

All data and code are available from the corresponding author upon reasonable request following publication. The RNAseq data is available from GEO database (GSE226093).