PS cDKO mice were generated and genotyped as described previously (Feng et al., 2004). Briefly, PS cDKO mice were obtained by crossing the forebrain specific PS1 heterozygous knockout mice with conventional PS2 knockout (PS2^–/–) mice on B6/CBA genetic background. Mice with the Cre transgene, fPS1/fPS1, and PS2^–/– served as PS cDKO mice, their littermates (without Cre, fPS1/⁺ and PS2^+/+) served as wild-type (WT) controls. Both male and female PS cDKO mice were used in the experiments. Mice had ad libitum access to water and chow and were housed at 22 ± 2 °C in 12 h/12 h light/dark schedule and 50–60% humidity. The experiment protocols were approved by the Animal Ethics Committee of East China Normal University.

Cortical and hippocampal tissue were obtained from 2-, 3-, 4-month-old mice, experiments were carried out as described previously with minor modifications (Cao et al., 2018). Briefly, mice were decapitated and the head was immersed in liquid nitrogen for 4 s, then the cortex and hippocampus were quickly dissected. Total protein was prepared from both tissues as described previously (Cao et al., 2018) and the protein concentration was measured with a BCA kit (Beyotime, China). 60 μg of proteins was boiled with SDS loading buffer and separated on 5–15% SDS-PAGE gels and transferred onto a PVDF membrane. The membrane was blocked with 5% skim milk in TBST for 1 h at room temperature, incubated with primary antibodies in 5% skim milk/TBST at 4 °C overnight, and then rinsed 4 times with TBST. The membrane was then incubated with secondary antibodies for 1 h at room temperature, and images were captured by Odyssey LI-COR Imager. The primary antibodies used include rabbit-anti-GFAP (1:10,000, abcam); mouse-anti-GAPDH (1:5,000, bioworld); rabbit-anti-IBA1 (1:1,000, abcam). The secondary antibodies used include anti-Mouse IRDye 800CW (1:15,000, Li-Cor Biosciences, United Kingdom) and anti-Rabbit IRDye 800CW (1:15,000, Li-Cor Biosciences, United Kingdom). The immunoreactive bands were quantified by the ImageJ software. Band intensities were normalized to loading control GAPDH (anti-GAPDH antibody, 1:5,000, Bioworld, United States).

Mice were deeply anesthetized and perfused transcardially with ice-cold 4% paraformaldehyde (PFA) in PBS and brains were fixed in 4% PFA at 4 °C overnight. Frozen brain blocks were sliced into 20 μm sections using a cryostat (Leica Microsystems, Germany). The sections were subsequently washed for 15 min with 1.2% Triton X-100 in PBS, blocked in blocking buffer (5% normal goat serum, 2% BSA and 0.2% Triton X-100) for 1 h, and incubated with primary antibodies at 4 °C overnight. After washing with 0.1% Tween 20 in PBS, the sections were incubated with secondary antibodies for 2 h at room temperature. The sections were mounted with antifade mounting medium (Beyotime, China) and imaged using Zeiss Axio Imager A1 microscope (Carl Zeiss Microscopy GmbH, Germany). The primary antibodies used: rabbit-anti-GFAP (1:1,000, abcam, ab7260); rabbit-anti-IBA1 (1:1,000, abcam, ab178847). The secondary antibodies used: Alexa Fluor^® 488 Goat Anti-Rabbit IgG (H + L) Antibody (1:600, Invitrogen). The number and positive area of GFAP⁺ or IBA1⁺ cells were measured using the ImageJ analysis software.

For RNA Sequencing (RNA-Seq), cortex samples from two mice were pooled to generate one sample (n = 3). Total RNA was extracted by Trizol Reagent (Invitrogen Life Technologies, United States), and RNA quality was determined using NanoDrop spectrophotometer (Thermo Scientific, United States). After RNA purification, the sequencing library was constructed and then sequenced on a Navaseq 6000 sequncing platform (Illumina, United States) by Personalbio Co., Ltd. (Shanghai, China).

The raw data obtained by sequencing were filtered by removing adaptor-related and low-quality reads, and then the clean reads were mapped to the mouse reference genome. The expression level of each gene was calculated and expressed by the FPKM value (Fragments Per Kilobase of exon per Million fragments mapped). DESeq was used for analysis of differentially expressed genes (DEGs), and genes with a | log₂ (fold change) | > 1 and p-value < 0.05 were considered differentially expressed. The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of DEGs were conducted as described previously, respectively (Ashburner et al., 2000; Kanehisa et al., 2004). GO and Pathways were considered significantly enriched with p < 0.05. The Protein-Protein Interaction (PPI) network was analyzed using the online STRING database,¹ which provides physical (direct) interactions and functional (indirect) associations between proteins. Based on the DEGs, the PPI pairs with score > 0.95 were identified from the STRING database. After the PPI analysis by STRING, the interaction network was plotted and visualized by Cytoscape software.

Real time polymerase chain reaction (RT-PCR) was performed as described previously (Cao et al., 2018). Briefly, cortex and hippocampus of cDKO mice and WT mice were dissected and immediately frozen in liquid nitrogen and stored in a freezer at -80 °C prior to use. RNA was extracted using Trizol (Invitrogen, United States) and cDNA was generated using moloney murine leukemia virus (MMLV) reverse transcriptase (Invitrogen, United States). Diluted cDNA was used as a template for the TB Green (Takara, RR420) RT-PCR analysis using CFX97 real time system (Bio-Rad, United States). The primer sequences for RT-PCR are listed in Supplementary Table 2. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as the reference gene. The gene expression levels were calculated and described as 2^{–Δ ΔCt} values.

All data were presented as mean ± SEM. To compare data between two groups, unpaired two-tailed t-test was used. To examine significance among more than two groups, one-way ANOVA followed by Tukey’s multiple comparisons test were used. All statistical analyses were performed using GraphPad Prism. Statistical significance was set at p < 0.05 for all tests.

To further investigate the dynamic patterns of glial cells in different brain regions of cDKO mice with age, we first used Western blotting to determine the protein levels of astrogliosis marker glial fibrillary acidic protein (GFAP) and microgliosis marker ionized calcium-binding adapter molecule 1 (IBA1) in the cortex and hippocampus of cDKO mice at 2, 3, and 4 months. At 2 months of age we found that both GFAP and IBA1 did not change their protein expression level in the cortex and hippocampus (Figures 1A–F), there was a significant increase in GFAP expression in both cortex and hippocampus of 3- and 4-month-old cDKO mice (Figures 1A–D), whereas we only detected a higher IBA1 expression in the cortex without a different expression in the hippocampus of 3- and 4-month-old cDKO mice (Figures 1A,B,E,F). We further performed real time-PCR (RT-PCR) to probe the expression changes of GFAP and triggering receptor expressed on myeloid cells 2 (TREM2). As shown in Figure 1G, significantly increased GFAP expression was found at 3 and 4 months, revealing age-dependent and progressive upregulation in the cortex of cDKO mice. Similarly, we also found a significant increase in the expression of TREM2, a microglial surface immunoreceptor, in the cortex of cDKO mice (Figure 1G). However, in the hippocampus, the expression of GFAP was comparable to controls at all ages examined (Figure 1H), but a significant increase in TREM2 expression was observed in cDKO mice at 3 months (Figure 1H). Thus, cDKO mice exhibited age-related, progressive increase of gliosis with significantly more GFAP expression in the cortex.

Western blotting results revealed that the protein expression levels of GFAP and IBA1 increased with age in the cortex and hippocampus at 3 and 4 months of age. Notably, PS cDKO mice exhibited impaired post-synaptic responses as early as 6 weeks old prior to the onset of neurodegeneration (Zhang et al., 2010). The subregions of cortex reported to be affected in the very early stage of AD include somatosensory cortex (SS) and retrosplenial cortex (RS) (Pengas et al., 2010; Stephen et al., 2010). Given that Western blotting examined the changes in the whole cortex or hippocampus, we performed immunofluorescence experiments to gain insight into whether in the subregions of the cortex and hippocampus astrocyte and microglial activation occurred in the early stage before the onset of neurodegeneration. We measured GFAP and IBA1 immunoreactivity in 6-week- and 2-month-old cDKO mice. In the SS of cDKO mice, GFAP reactivity increased at both 6 weeks and 2 months of age (Figures 2A–C). However, there was no significant change in number or positive area (%) of microglia between cDKO mice and age-matched controls (Figures 2A,D,E). Thus, the occurrence of astrogliosis appeared to start earlier than that of microgliosis in the SS of cDKO mice.

Given that the volume loss and atrophy of retroslenial cortex (RS) was observed at the earliest clinical stage of AD (Pengas et al., 2010), we further examined GFAP and IBA1 immunoreactivity in the RS of cDKO mice at 6 weeks and 2 months of age. At 6 weeks, no detectable difference was found in GFAP and IBA1 reactivity of the RS in the cDKO mice relative to control (Figures 3A–E). Comparison of stained brain sections and quantitative analysis revealed a striking increase in the number and positive area of astrocyte and microglia in the RS of cDKO mice at 2 months of age, whereas we could easily see age-dependent and progressive increases of both astrogliosis and microgliosis in the RS of cDKO mice (Figures 3A–E).

We next investigated whether astrocyte and microglia reactivity changed in the hippocampus of cDKO mice. At 6 weeks and 2 months, for all subregions of the hippocampus, including CA1, CA3 and dentate gyrus (DG), there were no statistically significant differences in GFAP reactivity between cDKO and WT mice (Supplementary Figures 1A–C). Similarly, IBA1 reactivity was also unaltered in cDKO mice relative to WT mice at both 6 weeks and 2 months of age (Supplementary Figures 2A–C).

To characterize what changes in gene expression are potentially associated with the gliosis in the cortex of cDKO mice, we performed transcriptome sequencing analysis using RNA from the cortex of 2-month-old cDKO mice to identify DEGs. According to the fold change (> 2) and p-value (<0.05) criteria, a total of 224 DEGs were differentially regulated, of which 159 DEGs were up-regulated and 65 DEGs were down-regulated (Figure 4A and Supplementary Table 1). Notably, we found that glial and inflammatory related genes include astrocyte marker GFAP, inflammatory chemokines Ccl3, Ccl4, Ccl6, and Ccl9 and complement C1qb were all upregulated in 2-month-old of cDKO mice (Supplementary Table 1). Intriguingly, RNA-seq results of 2-month-old cDKO mice have both confirmed and further expanded previous microarray and RNA-seq results of cDKO mice at 3, 5, 6, 10 months of age (Beglopoulos et al., 2004; Dong et al., 2007; Cao et al., 2018), suggesting an early involvement of gliosis and inflammation in early neurodegeneration due to presenilins deficiency.

We next used gene ontology (GO) enrichment analysis to determine cellular processes and pathways affected in the cDKO mice and identified top 20 enriched GO terms (Figure 4B). Enriched biological process GO terms were related to immune system process and regulation, defense response, inflammatory related process, chemokine production and its regulation (Figure 4C). We found that downregulated genes were mostly related to generalized cellular functions. In contrast, upregulated genes were more associated with inflammation related process.

To obtain a deeper insight into the functions of upregulated DEGs, KEGG pathway analysis was performed on the 159 DEGs. The top-one in KEGG pathway analysis was phagosome, which is involved in homeostasis, innate and adaptive immune responses (Russell et al., 2009; Figure 4D). Interestingly, enriched KEGG pathway also included chemokine signaling pathway, cytokine-cytokine receptor interaction, complement and coagulation cascades (Figure 4D). Collectively, KEGG enrichment analysis of upregulated DEGs revealed an association of presenilins deficiency with phagocytosis and inflammatory cytokines and chemokines. KEGG enrichment analysis was also consistent with GO term analysis that revealed immune/inflammatory related processes, cytokine production and its regulation as the top-two enriched biological process GO term (Figures 4B,C).

To further identify the pivotal genes involved in the onset and progression of early neurodegeneration, DEGs were used to construct the PPI. Importantly, we identified two key PPI networks, an immune response and microglia activation related gene network (Tyrobp, Fcer1g, Cd68, C3ar1, Tlr2, Itgax, Itgb2, Ptprc, Cd22, Ptpn6, Btk, and Clec5a) and a cytokine/chemokine related gene network including Ccl4, Ccl6, Ccl9, Cxcl13, and Gpr183 (Figure 4E). Some enriched genes including Tyrobp and C3ar1 were selected for RT-PCR analysis to validate the RNA-seq data. RT-PCR results confirmed these genes were upregulated, as observed from the RNA-seq data (Supplementary Figure 3). Consistent with biological process GO term enrichment results, proteins encoded by these genes in these two networks are highly related to inflammation (Figure 4C).

Given the gliosis observed in the early neurodegeneration of cDKO mice, we further determined whether increased gliosis was accompanied by the corresponding changes in the expression of inflammatory factors with age in different brain regions of cDKO mice. We performed RT-PCR to probe the expression changes of several representative DEGs chose based on RNA-seq results and previous microarray and RNA-seq data in the cortex and the hippocampus of 6-week-, 2-, 3-, and 4-month-old cDKO mice. These included five inflammatory chemokines (Ccl2-Ccl6), four complement genes (C1qa, C1qb, C3, and C4) and four members of the cathepsin (Cath) family (Cath B, D, S, and Z).

Chemokines, which are proinflammatory or chemotactic cytokines, are important mediators in inflammation and have been potentially implicated in contributing to AD pathogenesis (Lalli et al., 2015; Marciniak et al., 2015). Here, we detected an age-dependent increase of Ccl3, Ccl4, and Ccl6 expression with no alterations in Ccl2 and Ccl5 expression within both the cortex and hippocampus of cDKO mice (Figures 5A,B). Interestingly, we found that Ccl4 mRNA level was highly increased in the cortex of cDKO mice compared with littermate controls at 2 months (Figure 5A), appearing much earlier than the upregulation of other Ccl genes, suggesting that Ccl4 upregulation may accompany or precede astrocyte and microglia activation and is likely an early event anticipating the onset and progression of neurodegeneration.

Since the classical complement cascade including C1q, C3, and C4 can be produced by glia and is involved in AD attributable to neuroinflammation (Dalakas et al., 2020), we next examined the expression of these complement genes changes with age. Indeed, striking increases in complement genes C1qa, C3, and C4 were seen in the cortex of cDKO mice at 4 months but not at 2 and 3 months (Figure 6A). Interestingly, no increase was found in C1qa, C3, and C4 expression in the hippocampus of cDKO mice, relative to controls (Figure 6B), whereas significant increases of C1qb expression were found in both the cortex and hippocampus of cDKO mice at 3 and 4 months of age, compared to the control (Figures 6A,B). Notably, despite the large increases in C1qb expression in both the cortex and hippocampus of cDKO mice, the increase of C1qb expression in cDKO mice was much more prominent in the cortex (Figure 6A).

We next assessed whether members of the cathepsin protease family (Cath B, D, S and Z), which are endo-lysosomal proteases and have been implicated in AD pathogenesis (Lemere et al., 1995; Di Domenico et al., 2016), showed age-dependent and progressive changes in expression in cDKO mice. Significant increases of Cath S and Cath Z expression were also seen in the cortex of cDKO mice at 4 months but not at 2 and 3 months compared with controls (Figure 6C). The increase of these cathepsin genes in the cDKO hippocampus was not significant relative to controls (Figure 6D), suggesting that cathepsin dysfunction may be a later event in the onset and progression of neurodegeneration in cDKO mice.