In this study, we used male and female mice from 5xFAD transgenic mouse line (B6.Cg-Tg(APPSwFlLon, PSEN1*M146L*L286V)6799Vas/Mmjax with C57BL6J genetic background from the Jackson Laboratory). This mouse line was maintained by crossing of WT and heterozygous mice. We used 5 adult wild-type (WT) mice as control group, 6–10 month-old (it is equivalent to 30–40 year-old in humans (Flurkey et al., 2007)); 5 adult 5xFAD mice, 6–10 month-old; and 4 aged WT mice, 18–22 month-old (it is equivalent to 56–66 year-old in humans). All mice were acquired from the DZNE animal facility. The mice were housed under standard conditions of temperature (22 ± 2°C), humidity (60 ± 10%), and a 12-h light/dark cycle. The mice were provided with food and water ad libitum.

All treatments and behavioral procedures were conducted in accordance with animal research ethics standards defined by German law and approved by the Ethical Committee on Animal Health and Care of the State of Saxony-Anhalt. To obtain the required samples, the mice were subjected to euthanasia. Anesthetized mice (ketamine/xylazine; 90 mg/kg and 18 mg/kg body weight in 0.9% NaCl solution) were transcardially perfused with ice-cold PBS followed by 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer, pH 7.4 (PB). After mice were subjected to cervical dislocation, the brains were carefully dissected and fixed in 4% PFA in PBS for a period of 24 h, after which they were transferred to a 30% sucrose in PBS for cryoprotection. The brains were then sectioned at a thickness of 40 μm using a cryostat microtome (Leica, 1950) and collected in an ice-cold cryoprotectant solution (ethylene glycol-based: 30% ethylene glycol, 30% glycerol, 10% 0.2 M sodium phosphate buffer pH 7.4, in ddH₂O), ensuring optimal preservation for subsequent analysis.

Slices were washed in PBS with 0.5% Triton X-100, and blocked in 10% normal goat serum in PBS. Sections were incubated with primary antibodies 2 nights (for SST it’s 4 nights) at 4°C, incubated in secondary antibodies for 4 h at room temperature. Concentrations of antibodies were as follows: chicken anti-VGLUT1 (dilution 1:500, Synaptic Systems Cat# 135 316, RRID:AB_2619822), chicken anti-VGLUT2 (dilution 1:500, Synaptic Systems Cat# 135 416, RRID:AB_2619824), guinea pig anti-Homer1 (dilution 1:500, Synaptic Systems Cat# 160 004, RRID:AB_10549720), guinea pig anti-VGAT (dilution 1:500, Synaptic Systems Cat# 131 004, RRID:AB_887873), mouse anti-gephyrin (dilution 1:500, Synaptic Systems Cat# 147 011, RRID:AB_887717), rat anti-SST (Somatostatin) (dilution 1:200, Millipore Cat# MAB354, RRID:AB_2255365), rabbit anti- VIP (Vasoactive Intestinal Peptide) (dilution 1:500, ImmunoStar Cat# 20077, RRID:AB_572270), chicken anti-PV (Parvalbumin) (dilution 1:1,000, Synaptic Systems Cat# 195 006, RRID:AB_2619887), goat anti-chicken Alexa Fluor 488 (dilution 1:500, Invitrogen, A11039), goat anti-chicken Alexa Fluor 647 (dilution 1:1,000, Invitrogen, A21449), goat anti-rabbit Alexa Fluor 546 (dilution 1:500, Invitrogen, A11035), goat anti-rabbit Alexa Fluor 647 (dilution 1:500, Invitrogen, A21245), goat anti-rat Alexa Fluor 488 (dilution 1:200, Invitrogen, A11006), goat anti-rat Alexa Fluor 546 (dilution 1:200, Invitrogen, A11081), goat anti-mouse Alexa Fluor 488 (dilution 1:500, Invitrogen, A11029), goat anti-guinea pig Alexa Fluor 546 (dilution 1:500, Invitrogen, A11074). Finally, nuclei were counterstained with DAPI (dilution 1:400, Invitrogen, D1306), and slides were mounted with permanent mounting medium (Fluoromount™ Aqueous Mounting Medium) for microscopic evaluation.

Research regions are shown in Figure 1. Images were captured by confocal laser scanning microscopy (Zeiss LSM 700) and a Zeiss Axio Imager 2. For synapses, imaging was done with the EC Plan-Apochromat, 63 × /1.40 oil M27 objective (8-bit, 10 optical sections, 0.30 μm intervals between sections, 512 × 512 pixels, the pixel size of 0.07μm, and the scan zoom of 3.0). For SST, VIP, and PV cells, imaging was done with the EC Plan-Neofluar, 20 × /0.50 M27 objective (8-bit, 20 optical sections, 1.40 μm intervals between sections, 1024 × 1024 pixels, the pixel size of 0.63 μm, and the scan zoom of 0.50).

For synaptic density analysis, we conducted high-resolution confocal microscopy to examine excitatory synapses defined by co-localized VGlut1/VGlut2 and Homer1 immunopositive puncta, and inhibitory synapses defined by co-localized VGAT (vesicular GABA transporter) and gephyrin puncta (Figure 2). We utilized Fiji ImageJ (NIH) and Synapse Counter plug-in for ImageJ (Dzyubenko et al., 2016) to collect synapse data. Four brain slices were observed for each of the three mouse groups (one brain slice from each mouse and four mice representing each group). Four values were obtained from each brain region in each slice and then averaged. Four values were obtained from each group of mice.

Regarding the VIP, SST, and PV cells, four slices were analyzed for each of the three mouse groups (one brain slice from each mouse and four mice from each group). Two regions of interest (ROIs) from each brain region were analyzed and the derived values were averaged. Four values were obtained from each mouse group, except for the S1BF-L4 region of VIP where only 2 or 3 data were obtained due to imperfect staining. The three adjacent layers of 2.8μm vertical axis with the largest cell counts were projected onto a surface, and the total number of interneurons on the surface was calculated and compared. ZEN 3.5 (blue edition) was used to manually quantify the cells in each image.

Statistical significance was assessed by analyzing the numerical differences in several brain regions across different groups of mice.

Graphpad Prism 8.0 was used for plotting and executing statistical analysis. Error bars in Figures 3–5, 6, 7, 8 represent the mean ± SEM values. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 indicate statistically significant differences. A RM One-Way ANOVA was employed to evaluate the summary p value on synapse density, I/E ratios, or interneuron density across several brain areas in each group of mice. A RM Two-Way ANOVA with Holm-Sidak post-hoc analysis was used to compare data between pre- and postsynapses, to compare synapse density between excitatory and inhibitory synapses, or to compare data between three groups of mice. Mixed-effects One-Way or Two-Way ANOVA with Holm-Sidak post-hoc analysis were used to evaluate data containing missing values. A Three-Way ANOVA with Holm-Sidak post-hoc analysis was used to compare the mean synapse densities of both excitatory and inhibitory synapses across all studied regions between different groups of mice.

Pearson’s correlation coefficient (r) is used to study the correlation between two variables. The greater the absolute value of r (| r|), the stronger the correlation. With the used number of samples, a correlation coefficient 0 means no correlation, between 0 and 0.5 suggests a weak correlation, between 0.5 and 0.8 suggests a moderate correlation, and ≥ 0.8 indicates a strong correlation. p < 0.05 or p < 0.01 indicates the significant linear correlation between the two variables.

Normally, Aβ is only expressed in small amounts in the brain, where it has neurotrophic effects on immature, undifferentiated neurons. In the brain of AD patients, the expression of Aβ is abnormal and greatly increased, and it has toxic effects on differentiated and mature neurons (Gouras et al., 2015). The amyloid hypothesis suggests that overexpressed Aβ is the trigger for all pathophysiological responses to AD (Glenner and Wong, 1984, Selkoe and Hardy, 2016). Oligo-Aβ42 is the most toxic among all Aβ species (Hu et al., 2014). 5xFAD mice are widely used because they have large amounts of Aβ42 that accumulate abnormally.

For three groups of mice, the correlations among their excitatory synapse densities are weak across brain regions, whereas their inhibitory synapses are closely related (Figure 9A). This suggests a higher region-specific susceptibility of excitatory synapses to aging and amyloidosis. Previous studies implied that Aβ oligomers contribute to synaptic shifts (Palop et al., 2007, Li and Selkoe, 2020, Walsh and Selkoe, 2020). Aβ42 oligomers induce overexpression of Vti1a, leading to high frequency spontaneous vesicular exocytosis associated with Glu, which is further accompanied by downregulation of VGLUT1, and eventually cause depletion of Glu content in vesicles and significantly reduced Glu release. Ultimately, this sequence of events contributes to neurodegeneration in the later stages of AD (Zott et al., 2019, Yang X.-K. et al., 2023).

The sites of the most excitatory synapses are L1/5 in adult WT, are PirL3, BMA, DM in adult 5xFAD and aged WT mice. These regions are closely related to cognitive function (Joyce et al., 2024), olfaction (Blazing and Franks, 2020), emotional response (Patin and Pause, 2015), endocrine system (Ishii and Iadecola, 2015), etc.

Statistical variances in synaptic density within the same brain region among different groups of mice suggest that amyloidosis impacts synaptic density in S1BF and BMA, while aging affects synaptic density in S1BF, Pir, BMA, and DM in the mouse brain (Figures 3, 4). Our findings support the idea that amyloidosis and senescence might remodel synaptic connections and synaptic distribution in mouse brains.

The excitatory synaptic ratio of RSA-L2/3 to RSG-L2/3 is 1.35, 1.22, and 1.12 in adult WT, adult 5xFAD, and aged WT mice, respectively. Meanwhile, the inhibitory synaptic ratio of RSA-L2/3 to RSG-L2/3 is 1.21, 1.14, and 1.14 in adult WT, adult FAD, and aged WT mice, respectively. Previous studies have shown that pyramidal neurons in layer 4/5 of RSG send lateral rami of ipsilateral and contralateral axons to layer 2/3 of RSA (Sugar et al., 2011). From our data, we can see that the number of synapses at the RSA-L2/3 site is relatively reduced in the adult 5xFAD and aged WT mice.

There are more excitatory synapses than inhibitory synapses, which aligns with a prior study indicating that there are more glutamatergic neurons than GABAergic neurons in the cerebral cortex (Rubenstein and Merzenich, 2003).

Through a comparative analysis of pre- and post-synaptic densities within each mouse groups, it is shown that all synapses exhibit a moderate to high level of correlation, with the exception of the excitatory synapses in adult 5xFAD mice (Figure 9B). This finding aligns with a previous study that reported a match between pre- and postsynaptic structures (Kay et al., 2011). Importantly, we found that amyloidosis may disrupt the alignment of excitatory pre- and postsynaptic quantities. Aβ fibril may stimulate glial cells to overrelease excitatory amino acids, such as glutamate. At the same time, Aβ causes the mislocalization and internalization of the fast excitatory amino acid transporter (EAAT) and changes the clearance of glutamate (Scimemi et al., 2013). Glutamate is the main excitatory neurotransmitter in the brain (Orrego and Villanueva, 1993), and excitatory toxicity hypothesis refers to the frequent stimulation of neurons by excessive glutamate leading to neuronal dysfunction (Belov Kirdajova et al., 2020). This excitotoxicity eventually leads to neuronal calcium overload and has been linked to neurodegenerative diseases (Lipton and Rosenberg, 1994). There is a moderate or strong correlation between the densities of inhibitory pre- and post-synapses (Figure 9B, I-pre and I-post), and there are only a few regions where there are significant differences between them (Figure 4B, S1BF-L4, BMA and DM). In the RSC and Pir, there is no statistically significant difference between the pre- and post-synapses of excitatory or inhibitory synapses (Figures 4A,B).

We found moderate correlations for I/E ratios among adult WT, adult 5xFAD, and aged WT mice (Figure 9A). Although the levels of cortical activities may fluctuate, the single pyramidal cell still maintains a balance of synaptic excitation and inhibition (E/I balance) (Okun and Lampl, 2008, Atallah and Scanziani, 2009). The mechanism behind this might be that as excitation increases, inhibition will increase correspondingly. E/I balance contributes indispensably to the cognitive functions like memory, learning, etc. (Froemke, 2015).

Here we can see that the regions with the highest I/E values are located in L1–L4 of S1BF, RSC, and BMA of the three groups of mice. Our synaptic studies in mouse models of AD, aging as well as WT controls had shown that brain regions S1BF, RSC, and BMA are of highest relative inhibition (Figure 5).

AD, schizophrenia, and autism are accompanied by an increased E/I ratio and a relative increase in excitability. However, the E/I balance of major depression shifted toward inhibition (Yang B. et al., 2023). In patients with depression, connectivity changes are frequently reported in several brain regions, including the amygdala, frontal cortex (Howard et al., 2019), anterior cingulate cortex, and ventral striatum (Pruessner et al., 2010, Otte et al., 2016, Roddy et al., 2021, Chai et al., 2023). Amygdala is involved in the processing of fear memory (Zhou et al., 2018, Chai et al., 2023). It can be seen that the regions with the strongest relative inhibition in 5xFAD and aged WT mice are highly coincident with the main lesion area of depression. So far, the correlation between depression and dementia is in dispute. Some literatures suggest that the two are related (Lutz et al., 2020, Elser et al., 2023), whilst others do not (Huang et al., 2020).

Inhibitory variation leads to the E/I (excitation/inhibition) balance disorder (Francis, 2005; Le Roux et al., 2008). AD and aging are characterized by an E/I imbalance (Busche and Konnerth, 2016), a relative glutamatergic overexcitability (Lipton and Rosenberg, 1994; Reisberg et al., 2003; Francis, 2005), and a decrease in GABAergic activity (Busche and Konnerth, 2016). Non-neocortical brain regions contribute the most to the fluctuation in I/E ratio. Pir, BMA and DM in aged WT mice exhibit the largest drop in the I/E ratio together (Figure 5). The piriform cortex is an important part of the olfactory pathway, and its damage leads to olfactory disorders (Neville and Haberly, 2004; Attems et al., 2015). Researches have shown that olfactory dysfunction appears as an early symptom in normal elderly people (Schubert et al., 2012). Hypothalamic abnormalities are associated with memory impairment (Hamani et al., 2008).

Notably, aging affects PV cells in the RSG and CA1 region of hippocampus apparently, but amyloidosis doesn’t show this effect (Figure 7). We can deduce that aging damages retrosplenial cortex and hippocampus of the mouse brain. RSC dysfunction is correlated with learning deterioration, and RSA neurons encode task-related dimensions across learning in mice (Sun et al., 2021). Furthermore, RSG (Sigwald et al., 2016; Tsai et al., 2022) or RSA (Pan et al., 2022) may play a predominant role in the recall of contextual fear memory. This discrepancy is also reflected in the average number of PV cells in the three types of mice. The average quantity of PV cells in aged WT mice is lower than that in adult WT controls, although the difference between adult WT controls and adult 5xFAD mice is not significant (Figure 7). This implies that aging affects the distribution of PV interneurons in the mouse brain, while amyloidosis does not. It is in line with the previous study in humans, which suggested that PV interneurons exhibited relative resistance to degeneration in AD patients (Ferrer et al., 1991).

Compared with adult WT controls, the number of SST in PoDG is significantly decreased in aged WT mice (Figure 8). We can infer that aging damages the hippocampus of the mouse brain. The average quantity of SST cells in adult 5xFAD mice and aged WT mice is lower than that in adult WT controls. It indicates that both amyloidosis and aging affect the distribution of SST interneurons in the mouse brain. Previous studies have shown that the decline of SST in AD patients and AD mouse models is related to factors such as Aβ, ptau, APOE4 genes, etc. (Xu et al., 2020). However, decreased SST in elderly patients and mice may cause Aβ formation (Watamura et al., 2022).

Inhibitory interneurons mainly comprise PV, VIP, SST, and a few other interneurons (Kelsom and Lu, 2013; Yavorska and Wehr, 2016). Aβ causes inhibitory interneurons dysfunction (Zhang et al., 2017; Ranasinghe et al., 2022). Previous studies have shown that SST, VIP, and PV interneuronal markers have no co-localization in the cerebral cortex (Suzuki and Bekkers, 2010; Rudy et al., 2011), which echoes our study that PV, SST, VIP interneurons are scattered in different areas in S1BF (Figure 6).

In the RSG/RSA region, PV cells are mainly distributed in L5 (Figures 7, 11), while SST cells are evenly distributed in L2/3 and L5 (Figures 8, 12). At the hippocampus, PV cells are mainly distributed in PyCA and GrDG, while SST cells are mainly distributed in CAOr and PoDG. PV and SST interneurons are scattered in different areas in the retrosplenial cortex and hippocampus, and these two populations are not overlapping in the hippocampus. In the S1BF region, layers 2/3 and 6 have the lowest distribution of PV cells but the highest distribution of SST cells, and they also have the most similar density of PV and SST cells (Figure 13). L4 and L5 have the highest PV cell density and the lowest SST cell density, with the largest difference in PV and SST cell densities among the six layers of S1BF.

We observed that there are no PV or SST cells in the hypothalamus in all three groups of mice.

Previous investigations on E/I balance were carried out in the neocortex and hippocampus (Kvitsiani et al., 2013, Moore and Wehr, 2013, Xue et al., 2014, Ferguson and Gao, 2018, Sohal and Rubenstein, 2019, Terstege and Epp, 2023). However, the limbic system is the first and foremost site for the progression of AD. Different from previous E/I balance studies, our study used AD and aging mouse models to explore the linear correlation between PV, SST, and E/I balance in brain regions including the S1BF, RSC, Pir, and BMA (Figure 14). Previous studies have concluded that PV rather than SST cells inhibit excitatory neurons (Xue et al., 2014). PV directly inhibits excitatory neurons and regulates E/I balance (Song et al., 2021), while SST also inhibits PV cells (Viollet et al., 2008) and has more complex effect on the network activity. Our experimental results show that PV cell density changes and I/E changes are moderately related between adult and aged WT mice across brain regions (Figure 14C), whereas there is little correlation between SST cell density changes and PV cell density changes, or between SST cell density changes and I/E changes. Thus, PV interneurons may more prominently regulate E/I balance.

The region of greatest relative inhibitory upregulation is S1BF-L5 in both adult 5xFAD and aged WT mice (Figures 14A,B, I/E changes). Meaningfully, in both adult 5xFAD and aged WT mice, S1BF-L5 is also the region with the highest growth rate of PV cells (Figures 14A,B, PV density changes). Our data imply that S1BF-L5 plays an important role in the E/I imbalance (Le Roux et al., 2008).