Cholinergic Modification of Neurogenesis and Gliosis Improves the Memory of AβPPswe/PSEN1dE9 Alzheimer’s Disease Model Mice Fed a High-Fat Diet

Animals

AβPPswe/PSEN1dE9 (AβPP/PS1) transgenic mice were used. The double-transgenic (Tg) mice were originally obtained from Jackson Laboratory (strain name B6C3-Tg (AβPPswe, PSEN1dE9) 85Dbo/J; stock number 004462), and were bred and raised in our laboratory for experiments. These B6C3-based Tg mice harbor two AD-related genes, one encoding a chimeric mouse/human amyloid-β protein precursor containing the K595N/M596L Swedish mutation (AβPPswe) and the other a mutant human presenilin 1 carrying a deletion of exon 9 (PSEN1dE9). The transgenes are controlled separately by mouse prion protein promoter elements, and are expressed predominantly in neurons in the central nervous system [56, 57]. The AβPP/PS1 Tg mice were maintained as double hemizygotes by interbreeding with wild-type (Wt) mice (B6C3F1/J background strain) purchased from Jackson Laboratory. In this study, we used male and female hemizygous Tg mice and their Wt littermates. Males and females were included approximately equally in each experimental group. The mice were genotyped from tail tissue as previously described [48]. Up to three littermates were kept in one plastic cage. The cages were kept at a temperature of 23–25°C with a 12-h light-dark cycle. All animal procedures and experiments in this study were approved by our institution’s ethics committee and were conducted according to the guidelines for animal experimentation required by the University of Tokyo.

Diet, food consumption, and weight

To create the T2DM-AD model mice, 4-month-old male and female AβPP/PS1 Tg mice were fed a high-fat diet (HFD) (HFD-32, CLEA Japan; its composition is shown in Supplementary Table 1) for 8 weeks from 4 months of age, which induces early-onset memory decline in the Tg mice at 6 months, as reported previously [48]. In agreement with our previous research [48], several studies have shown that AβPP/PS1 mice fed a normal diet (ND) begin forming major amyloid deposits in the brain at 6 months of age [57–59] and develop memory problems by 9–12 months of age [52, 61]. Therefore, as a pre-onset control, we used 6-month-old Tg mice continuously fed the ND (MF Rations, Oriental Yeast, Japan; its composition is shown in Supplementary Table 1). As a normal aging control, Wt littermates maintained continuously on the ND were used. Until 4 months of age, all of the mice were maintained on the ND. The mice were given free access to water and food. During the HFD feeding period, we changed the food in the cage every 2 days and determined the consumption per cage by measuring the difference in food pellet weight. All of the mice were weighed every other week during the ND or HFD experimental period.

Cholinesterase inhibitor (ChEI) treatment

In a clinical setting, rivastigmine can be administered subcutaneously to AD patients [62]. In a pilot experiment, we administered donepezil intraperitoneally (2 mg/kg/day) [32] as well as rivastigmine by subcutaneous infusion (2 mg/kg/day), into 6-month-old Tg-HFD mice for a week. Donepezil did not show any ameliorative effects, but rivastigmine tended to ameliorate the neuroinflammation and hippocampal neurogenesis compared to non-treated Tg-HFD mice. Therefore, we used rivastigmine as a cholinesterase inhibitor in this study. An Alzet Osmotic Pump 2006 (Alzet, Durect) was implanted subcutaneously into HFD-fed AβPP/PS1 Tg mice at 4.5 months of age. Each mouse was deeply anesthetized with an intramuscular injection of ketamine (50 mg/kg) and xylazine (5 mg/kg), the back was minimally incised under general anesthesia, and a subdermal space was expanded for pump insertion. An Alzet Osmotic Pump 2006 was filled with a rivastigmine solution (2 mg/kg/day) or saline (vehicle) (Otsuka, Japan), the pump was immediately implanted subcutaneously using a sterile instrument, and the cut was sutured. The wound was thoroughly disinfected, and the mouse was warmed on a heating plate (37°C) to maintain body temperature until it recovered from the anesthesia. The implanted pump remained in place until the end of the experiments, and continuously released the solution under the skin for 6 weeks. The solution containers were filled and the osmotic pump was implanted according to the manufacturer’s instructions.

Retrovirus-mediated labeling of adult-born neurons

Plat-GP cells and constructs for retroviral vectors that express GFP in new neurons were kindly donated by Dr. K. Inokuchi’s laboratory at the University of Toyama, and we produced the retrovirus vector in our laboratory. The virus was prepared essentially as described previously [63]. Briefly, the two constructs, pMXs-SIN-CAG-GFP and pVSV-G, were co-transfected into Plat-GP cells by the calcium phosphate method, and the medium was replaced 12–16 h after transfection to minimize cell damage. After 2-3 days, the culture medium containing the retrovirus was collected, and the retrovirus was concentrated by two consecutive ultracentrifugations, at 50,000×g for 90 min and 50,000×g for 90 min, at 4°C. The retroviral pellet was resuspended in phosphate-buffered saline (PBS) to 0.05% of the original volume. The retrovirus was thusly cultured and concentrated for each injection. The virus solution was kept on ice until just before use, when it was injected into the dentate gyrus of 5-month-old mice. For the injection, the mouse was deeply anesthetized by intramuscular injection of a mixed solution of ketamine/xylazine (50/5 mg/kg), and was placed in a stereotactic device. A concentrated GFP-expressing retroviral solution (more than 10⁸ infectious units/mL) was stereotactically injected into a single site in the hippocampal dentate gyrus (AP: +2.0 mm posterior; ML: ±1.5 mm lateral; DV: –2.0 mm ventral 1.5 μL/site) at a constant speed (0.5 μL/min). The mouse was placed on a heating plate (37°C) while the wound was sutured, and was kept on the plate to maintain body temperature until it recovered from the anesthesia. The integration of the retroviral vector into the host genome only occurs in dividing cells, and the cells are permanently labeled with GFP. The mice were sacrificed and examined 4 weeks after the retrovirus injection.

BrdU administration and cell counts

To label newly born cells in 5-month-old mice, bromodeoxyuridine (BrdU, Wako, Japan) was injected into mice intraperitoneally (100 mg/kg in 0.9% saline) for 5 consecutive days (24-h intervals). The animals were sacrificed 4 weeks after the last injection, and the brain was dissected out and fixed. For BrdU immunohistochemistry, a series of every 12th coronal section (480-μm apart) throughout the entire dentate gyrus was processed. All of the BrdU-positive cells in the dentate gyrus (the granule cell layer, including the subgranular zone) were counted under a 100× objective lens by an observer blinded to the experimental conditions. We analyzed 4 sections per mouse and multiplied the number of BrdU-positive cells by 12 to obtain a stereological estimate of the total number of these cells per hippocampal dentate gyrus [32].

Morris water maze (MWM) test

Rivastigmine treatment was stopped when the mice were 6 months old. Nine days before the end of rivastigmine treatment, we assessed spatial learning and memory by subjecting the mice to MWM tests for 6 consecutive days, and to contextual fear conditioning (CFC) tests on the next 3 days. To minimize the confounding effects of anxiety on learning, the mice were handled extensively for 2 min per day for 7 consecutive days prior to beginning the MWM testing [64]. The MWM test was conducted as described previously with some modifications [65]. Briefly, on days 1–5 of MWM testing, the mice were placed in a dimly lit circular pool (120-cm diameter, 35-cm deep) filled with opaque water (23°C±1°C) and surrounded by abundant extra-maze distal and visual cues, and were required to arrive at a hidden platform submerged 1 cm below the surface of the water within 60 s. A total of 30 trials (6 trials per day for 5 consecutive days) were performed in the training sessions. In the spatial task, the mice were carefully placed afloat at the edge of the pool, with their face turned toward the pool wall. The starting points were semi-randomly chosen from north (N), east (E), south (S), or west (W), and the escape platform was located in the middle of the SW quadrant. All mice took the test in the same order. In each trial, mice that failed to find the platform within 60 s were guided to the platform and allowed to sit on the platform for 20 s before being picked up. After the initial training, the inter-trial interval was approximately 30 min. Twenty-four hours after the training session, in the last trial (day 6), the escape platform was removed and the mice were allowed to swim for 60 s in the pool starting from the N side (the single-probe trial). The time spent in the target quadrant where the platform was previously located during a single 60-s trial was measured. Just before the first trial each day, the mouse cages were placed in the same condition as the behavioral test for at least 30 min to allow the mice to acclimatize to the experimental field. Mice were warmed on a heating plate (37°C) between trials and after the last trial to prevent hypothermia. All data were recorded and analyzed with the SMART video tracking system (Panlab).

Contextual fear conditioning (CFC) test

On the day following the final day of the MWM tests, the mice were subjected to CFC to evaluate hippocampal-dependent contextual memory and non-hippocampal (amygdala)-dependent cued memory. All equipment and software for the CFC tests were purchased from Med Associates. The testing procedures have been described in detail [48, 66]. In brief, on the first day of CFC testing, mice were placed in a conditioning chamber (30 cm wide×26 cm deep×24 cm high). The chamber consisted of a white-lit rectangular room with a stainless steel grid floor (2-mm diameter wire, spaced 5 mm apart) and aluminum and acrylic walls. The chambers were scented with 70% isopropanol. In the acquisition (training) phase of the CFC test, the mice underwent three conditioning trials during a 370-s session under the same conditions, after which an electric shock (1 s, 0.75 mA) was delivered through the floor accompanied by an aural tone (white noise). The tone sounded at 128, 212, and 296 s into the session, lasting for 10 s each time, and the footshock was delivered during the last 1 s of the tone.

Twenty-four hours later, during the second day of CFC, the mice were placed in the same environmental conditions as on the first day for 512 s, and the durations of mouse freezing behavior were measured throughout the session to test context-elicited fear (contextual memory). Neither tones nor shocks were presented.

Twenty-four hours after the contextual memory test, on the third day of CFC testing, the mice were placed into a novel context, and the same tone as on the first day was sounded for 512 s. Freezing times were analyzed throughout the session to test for tone-elicited fear (cued memory) without electroshock. The new context had a red-lit triangular room with a white plastic floor and an acrylic black triangle-frame roof, and was scented with 70% ethanol. The tone began sounding 128 s after the start of the session.

Before the first session each day, the mouse cages were placed in a low-lit, quiet room for at least 30 min to soothe anxiety. All of the sessions were recorded by a video camera attached to the chamber, and the freezing time was automatically calculated with Video Freeze software (Med Associates).

Collection of blood samples and measurement of serum insulin levels

Twenty-four hours after the cognitive experiments were finished, the mice were deeply anesthetized, and their blood was collected from a cardiac vessel immediately prior to perfusion. The blood samples were kept at room temperature for 2 h to promote clotting. The samples were then centrifuged at 4°C for 15 min at 2,320×g. The obtained supernatant serum was carefully aliquoted into several sterile tubes. The samples were stored at –20°C until just before use. The serum insulin levels were measured using a mouse-specific insulin ELISA kit (Morinaga, Japan) following the manufacturer’sinstructions.

Sandwich ELISA assay

Mice were deeply anesthetized and decapitated, and the hippocampus was immediately dissected out, frozen with liquid nitrogen, and kept at –80°C until homogenization. For homogenization, the frozen hippocampus was thawed and homogenized in ice-cold RIPA buffer containing a cocktail of protease inhibitors (Santa Cruz) using an ultrasonic homogenizer (Tomy Seiko, Japan). The homogenates were centrifuged (100,000×g, 1 h, 4°C), and the aliquoted supernatants were stored at –80°C until their analysis by ELISA (soluble fraction). The remaining pellet was further solubilized in a 5 M guanidine HCl/50 mM Tris HCl solution (pH 8.0) by incubating for 3 h at room temperature, and centrifuged (100,000×g, 1 h, 4°C). The supernatants were aliquoted and kept at –80°C until use (insoluble fraction, guanidine soluble). Protein concentrations for the soluble and insoluble extracts were determined using the BCA protein assay kit (TaKaRa Bio, Japan) according to the manufacturer’s instructions. The levels of amyloid-β (Aβ)₄₀ and Aβ₄₂ in the hippocampus were determined by a sandwich ELISA with Human/Rat (Mouse) Aβ₄₀ and Aβ₄₂ ELISA Kits (Wako, Japan). The Aβ oligomer levels were measured with the Human Amyloid-β Oligomers (82E1-specific) Assay Kit (IBL, Japan). The levels of IL-1β and TNF-α in the hippocampal supernatants were assayed using mouse IL-1β and TNF-α ELISA Kits (R&D Systems). The hippocampal extracts were appropriately diluted with the diluent buffer to fall within the standard range and applied to the ELISA plates. All ELISA procedures were conducted according to the manufacturer’s instructions. The standards and samples were assayed in duplicate at appropriate dilutions. The optical density was analyzed at 450 nm within 30 min of chromogen mixing on a microplate spectrophotometer. Concentrations were calculated according to a standard curve. Data obtained from the hippocampal homogenates are expressed as picomoles of Aβ species per gram of total protein (pmol/g) or as picograms of cytokines per gram of total protein (pg/g), taking into account the dilution factor.

Brain-tissue perfusion

Twenty-four hours after the last day of CFC testing, the mice were deeply anesthetized, and transcardially perfused with ice-cold PBS (pH 7.4) followed by 4% paraformaldehyde (PFA) (pH 7.4). The brain was removed and postfixed in fresh 4% PFA solution at 4°C for 24 h, and then equilibrated in 30% sucrose/PBS solution at 4°C until it sank (3 days). The dehydrated brain was then embedded in Tissue-Tek OCT Compound (Sakura Finetek, Japan) and frozen at –80°C. The frozen brain samples were sliced into 40-μm coronal sections using an HM525 cryostat (Microm, Germany). Brain slices were taken from the entire hippocampus. All slices were submerged in cryoprotectant solution and stored at –20°C until use.

X-34 synthesis and histochemical staining

We used a previously described protocol to organically synthesize X-34 and use it to stain senile plaque [48, 67]. X-34 is a highly fluorescent dye that is derived from Congo Red and specifically binds amyloid aggregates. Briefly, X-34 was synthesized from tetraethyl p-xylylenediphosphonate (Tokyo Chemical Industry, Japan), 5-formylsalicylic acid (Sigma-Aldrich), and potassium tert-butoxide (Sigma-Aldrich) in our laboratory, and was used in solution. Brain sections were washed 3 times in Tris-buffered saline (TBS), incubated with X-34 solution for 30 min, gently washed by dipping in TBS, and incubated in a 0.2% NaOH/80% ethanol solution for 2 min. The samples were washed again with TBS and incubated with TOTO-3 iodide (Molecular Probes) in TBS to stain the cell nuclei. The samples were then washed with TBS and mounted on microscope slides. Three sections were treated for each animal.

Immunohistochemistry

Brain sections were washed 3 times in TBS or TBS containing 0.3% Triton-X (0.3% TBS-X), then blocked with 3% normal donkey serum (NDS) diluted in 0.3% TBS-X at room temperature for 90 min.

For immunofluorescent BrdU staining, the sections were washed twice and placed in 10 mM citric acid (pH 6.0) at 90°C for 5 min in a water bath for antigen activation. The sections were then cooled at room temperature, rinsed with TBS, incubated in 1 N HCl at 37°C for 30 min, and incubated in boric acid buffer (pH 8.4) at room temperature for 10 min. The sections were then rinsed twice in TBS and incubated in blocking solution at room temperature for 60 min. After blocking, the free-floating sections were incubated with primary antibodies at 4°C for 3 days, washed 3 times with TBS, and incubated with secondary antibodies conjugated with various fluorescent dyes at room temperature for 2 h in light-resistant containers.

Next, the brain sections were washed once with TBS and incubated with 4′6-diamidino-2-phenylindole (DAPI) (1:10000, Sigma) in TBS at room temperature for 5 min. Excess DAPI was removed by washing with TBS, and the sections were mounted on microscope slides with Immu-Mount (Thermo Scientific) and visualized using a confocal microscope.

The primary antibodies used in this study were anti-BrdU rat IgG (1:200, Oxford Biotech), anti-NeuN mouse IgG (1:1000, Millipore), anti-GFAP mouse IgG (1:500, Sigma), anti-ionized calcium-binding adaptor molecule 1 (Iba1) rabbit IgG (1:1000, Wako, Japan), anti-doublecortin (DCX) guinea pig IgG (1:2000, Millipore), and anti-GFP rat IgG (1:1000, Nacalai Tesque), as reported previously [49]. The secondary antibodies used were anti-rat IgG donkey IgG Alexa488 (1:1000, Molecular Probes), anti-mouse IgG donkey IgG Alexa647 (1:1000, Molecular Probes), anti-rabbit IgG donkey IgG Alexa568 (1:1000, Molecular Probes), anti-rabbit donkey IgG Alexa488 (1:1000, Molecular Probes), and anti-guinea pig donkey IgG Cy5 (1:200, Jackson ImmunoResearch). All antibodies were diluted in 0.3% TBS-X containing 3% NDS. For DCX immunostaining, three sections were measured for each mouse.

Confocal microscopy and cell quantification

Sections stained with X-34 were analyzed with a confocal microscope (TCS SP2; Leica, Germany) using a 40× oil-immersion objective lens. Fluorescence immunostaining was quantified using a confocal microscope with a 20× (GFAP/Iba1), 40× (DCX), or 100× (BrdU/Iba1 and BrdU/NeuN) oil-immersion objective lens. Images were displayed in the maximum-intensity projection of a z-series stack acquired at 1.0-μm intervals throughout 40-μm-thick sections. For the confocal imaging of dendritic spines, we used a 63× oil-immersion objective lens with a 2.0× optical zoom at 0.5-μm step intervals (z-stack).

To count GFAP+ or Iba1+ cells in the hippocampus, immunofluorescence images were taken from three consecutive coronal sections per mouse, from Bregma –1.70 to –2.10 mm, spaced 0.2 mm apart. GFAP+ and Iba+ cells in the hippocampus were manually counted from digitalized confocal images in the hippocampal areas, and were averaged per section.

For the analyses of dendritic spine density and the maturity of newborn neurons, GFP-labeled spines were divided into four subtypes according to their morphology. During the process of synaptogenesis, a mature spine (mushroom and thin spine), which has a neck and head, is known to be derived from immature finger-like protrusions (filopodia) and stubby spines [43, 69]. Thus, in this study, all of the GFP+ spines were categorized and quantified manually as immature type or mature type based on these criteria.

Dendritic growth of DCX+ neurons

The maturity of doublecortin (DCX)-immunopositive granule cells was evaluated as previously described [70, 71]. In the morphological analysis, all of the DCX-positive (+) cells in the dentate gyrus were manually classified as immature or mature according to their dendritic growth pattern. As newborn DCX+ granule cells mature, they gradually project dendrites toward the molecular layer, and the dendrites develop more and more complex branches.

In our analysis, DCX+ cells located adjacent to the subgranular zone (SGZ) and having bipolar neurites parallel to the SGZ, or having primary dendrites that did not extend beyond the middle of the granule cell layer, were considered immature. DCX-reactive cells in the SGZ with long dendrites projecting toward and crossing the molecular layer were considered to be mature. Only DCX+ cells with sharply defined cell bodies were assessed for maturity.

Image processing

Images of X-34 staining were obtained by confocal microscopy and analyzed with ImageJ software as described previously [48]. Briefly, the outer edge of the hippocampal area was manually designated into regions-of-interest, and the percentage of the area containing senile plaque, which was labeled with X-34 fluorescence, was automatically calculated using a plug-in. This process was applied to each section to determine the proportion of amyloid load in the hippocampus.

Statistical analyses

Data are presented as means±SEM of the number of biological replicates (as noted in the figure legends). All data were analyzed for significant differences by one-way ANOVA followed by Dunnett’s or Williams’s multiple-comparison post hoc test (one group served as a control), by a Chi-squared test, and by a non-paired two-tailed Student’s t-test for two groups (placebo group, AD-Vehicle mice versus treatment group, AD-Rivastigmine mice). We considered p < 0.05 to be statistically significant.

Lack of influence of rivastigmine on the diabetic pathology of T2DM-AD model mice

In this study, we used T2DM-AD model mice fed a high-fat diet (HFD) for 2 months (Tg-HFD) [48]. Our previous study [48] revealed that the Tg-HFD mice show early-onset amnesia, so we used the Tg-HFD mice as AD model mice. As a normal aging control and pre-onset control, we used wild-type (Wt) littermate and Tg mice fed a normal diet (ND), respectively. At the end of the experimental period, despite similar food consumption in the two groups (Fig. 1A, B; ND-fed group versus HFD-fed group), the HFD-fed mice weighed about 1.5 times more than the ND-fed mice (Fig. 1C). Rivastigmine treatment had no significant effect on the food intake or body weight. The HFD-fed mice gradually accumulated fat and became obese, but had no other behavioral abnormalities in their cages. Rivastigmine is reported to carry a risk of appetite loss for some AD patients [15]; however, we did not observe any significant difference in feeding behavior in the rivastigmine-treated mice. Even after the surgery to implant the osmotic pump, the mice quickly resumed their normal dietary habits.

As shown in Fig. 1D, HFD-fed mice had significantly higher serum insulin levels, an important indicator of type 2 diabetes mellitus (T2DM), compared with ND-fed mice (one-way ANOVA, F _(3,31)  = 11.0, p < 0.01; AD-Vehicle mice, n = 8, Dunnett’s post hoc test, compared with AD-Vehicle: Wt-ND, n = 8, ^**p < 0.01; Tg-ND, n = 8, ^**p < 0.01). Rivastigmine did not noticeably affect these high insulin levels. Thus, all of the HFD-fed mice developed the T2DM pathophysiology, regardless of whether they were treated with rivastigmine or the control solution.

Rivastigmine suppresses glial neuroinflammatory responses in T2DM-AD model mice

We evaluated neuroinflammatory reactions in the brain of the T2DM-AD model mice fed the HFD for 2 months, by immunohistochemistry and ELISA. Figure 2 shows the results of immunostaining for astrocytes (GFAP) and microglia (Iba1) in the hippocampus of each group. This analysis showed that the inflammation-related glial cells were activated in the hippocampus of the T2DM-AD model mice (AD-Vehicle), as indicated by their intense staining and high cell density, whereas normal control mice (Wt-ND) showed more sparse astrocytes and microglia (Fig. 2A). Rivastigmine treatment suppressed the glial inflammatory responses compared to the AD-Vehicle group. Notably, rivastigmine effectively inhibited the high concentration and hypertrophy of astrocytes and microglia around microaggregates composed of enlarged glial cells, which surrounded senile plaques (Supplementary Figure 1), in these T2DM-AD model mice.

Quantitative analyses revealed that there were significantly more astrocytes and microglia, including reactive hypertrophied cells, in the hippocampus of the AD-Vehicle group compared to that of the Wt control (Fig. 2B, C; GFAP+ cells, one-way ANOVA, F _(3,15)  = 16.0, p < 0.01; AD-Vehicle mice, n = 4, Dunnett’s post hoc test, compared with AD-Vehicle: Wt-ND, n = 4, ^**p < 0.01; Tg-ND, n = 4, ^**p < 0.01; Iba1+ cells, one-way ANOVA, F _(3,15)  = 5.14, p < 0.05; AD-Vehicle mice, n = 4, Dunnett’s post hoc test, compared with AD-Vehicle: Wt-ND, n = 4, ^**p < 0.01). Neuroinflammation was not induced by the HFD alone, as we did not detect a significant increase in the number of activated glial cells in the Wt-HFD versus the Wt-ND hippocampus (Supplementary Figure 2).

We found that rivastigmine treatment suppressed neuroinflammation in these T2DM-AD model mice. Importantly, rivastigmine treatment suppressed the increased number of GFAP+ astrocytes in the hippocampus of the AD-Vehicle mice (Fig. 2B, AD-Rivastigmine, n = 4, ^**p < 0.01). Rivastigmine treatment also caused a significant decrease in the number of Iba1+ microglia (^*p < 0.05) (Fig. 2C). To clarify the effect of rivastigmine on the activation of microglial cells, we also quantified the number of newly born microglia in the hippocampus [72, 73]. As shown in Fig. 3, we detected a significant decrease in the number of double-positive (BrdU+/Iba1+) cells in the AD-Rivastigmine group, compared with that in the AD-Vehicle group (Fig. 3, ^*p < 0.05). Collectively, these results indicated that severe inflammation, which appeared as strong microglial generation, occurred in the AD-Vehicle mice, and that rivastigmine significantly inhibited this cellular inflammatory response.

Chronic neuroinflammation is mediated by proinflammatory cytokines. ELISAs revealed that the levels of IL-1β and TNF-α were significantly increased in the hippocampus of AD-Vehicle mice, and that rivastigmine substantially decreased these levels (Fig. 4) (IL-1β, one-way ANOVA, F _(3,15)  = 7.60, p < 0.01; AD-Vehicle mice, n = 4, Dunnett’s post hoc test, compared with AD-Vehicle: Wt-ND, n = 4, ^*p < 0.05; AD-Rivastigmine, n = 4, ^**p < 0.01; TNF-α, one-way ANOVA, F _(3,15)  = 6.64, p < 0.01; AD-Vehicle mice, n = 4, Dunnett’s post hoc test, compared with AD-Vehicle: Wt-ND, n = 4, ^**p < 0.01; AD-Rivastigmine, n = 4, ^**p < 0.01). These results indicated that AβPPswe/PSEN1dE9 AD model mice fed the HFD for 2 months (AD-Vehicle group) develop elevated proinflammatory cytokine levels, and that rivastigmine treatment affects not only cellular, but also molecular inflammatory responses.

Rivastigmine does not prevent the amyloid pathology of T2DM-AD model mice

We also analyzed the amyloid pathology in each group of mice. First, we measured the senile plaque load in the hippocampus by histochemical labeling with X-34. Figure 5A shows representative images of X-34 staining in each group of mice. The percent area (Fig. 5B) and number of senile plaques (Fig. 5C) in both the vehicle-treated (AD-Vehicle) and rivastigmine-treated (AD-Rivastigmine) AD model mice was comparable to that in the pre-onset control mice (Tg-ND). These results indicated that ingesting large amounts high-fat food does not alter amyloid deposition during the experimental food period, and that rivastigmine does not beneficially influence the amount of amyloid plaque accumulation.

Second, we investigated by ELISA the levels of three types of soluble Aβ peptides (Aβ₄₀, Aβ₄₂, and Aβ oligomers), which are more toxic than insoluble peptides, in the hippocampus. Compared to normal control mice (Wt-ND), the levels of all of the Aβ peptide types were markedly elevated in all three groups of AβPP/PS1 mice, and rivastigmine did not significantly decrease them (Fig. 5D-F). Interestingly, AD-Vehicle mice produced significantly more Aβ₄₂ than Tg-ND mice (Fig. 5E; one-way ANOVA, F _(3,15)  = 13.5, p < 0.01; AD-Vehicle mice, n = 4, Dunnett’s post hoc test, compared with AD-Vehicle: Tg-ND, n = 4, ^*p < 0.05). We also performed ELISAs for the insoluble Aβ species, which showed no significant differences among the three groups of Tg mice (Fig. 5G-I). These results suggested that long-term HFD feeding increases the soluble Aβ₄₂ in the hippocampus of the Tg mice, but this increase is not blocked by rivastigmine treatment.

Rivastigmine ameliorates the memory decline in T2DM-AD model mice

To assess rivastigmine’s effect on cognitive decline in the T2DM-AD model mice, we subjected the mice to MWM and CFC tests, both of which involve hippocampal-dependent memory tasks (Figs. 6 and 7). We used four groups of mice in these behavioral tests; in addition to the AD-Vehicle mice (placebo group) and AD-Rivastigmine mice (treatment group), their littermate Wt and Tg mice fed ND were used as control groups.

In the MWM tests, spatial learning and memory was evaluated in both an acquisition phase (days 1–5) and a probe trial (day 6). In the acquisition phase of the MWM test, we observed significant differences in memory function between the normal control (Wt-ND) and the AD-Vehicle group during the training period, but did not observe any significant difference between the AD-Vehicle and the AD-Rivastigmine mice (Fig. 6A, B). In the single-probe trial on day 6, the memory performance was significantly impaired in the AD-Vehicle mice compared with the normal control (Wt-ND) mice, and this cognitive decline was averted in the AD-Rivastigmine mice (Fig. 6C, D; one-way ANOVA, F _(3,31)  = 7.50, p < 0.01; AD-Vehicle mice, n = 8, Dunnett’s post hoc test, compared with AD-Vehicle: Wt-ND, n = 8, ^**p < 0.01; AD-Rivastigmine, n = 8, ^**p < 0.01). In the single-probe test, most of the mice in the placebo group failed to enter the target quadrant, indicating that they did not retain the spatial memory. In contrast, mice from the rivastigmine-treated group often entered the target quadrant where the hidden platform had been placed during the training sessions.

The same sets of mice were next subjected to CFC tests. On the first day, the mice were trained with a conditioning task. On the second day, we evaluated the hippocampal-dependent memoryperformance in each group by measuring their freezing time. We found that rivastigmine treatment significantly improved the ability of the AD model mice to perform this hippocampal-dependent memory task (Fig. 7A, B; one-way ANOVA, F _(3,31)  = 9.00, p < 0.01; AD-Vehicle mice, n = 8, Dunnett’s post hoc test, compared with AD-Vehicle: Wt-ND, n = 8, ^**p < 0.01; AD-Rivastigmine, n = 8, ^**p < 0.01). We also measured the freezing time of the mice in the tone fear-conditioning test on the following day, which showed that the cued memory was impaired in the AD-Vehicle mice. Rivastigmine had no detectable effect on this impairment (Fig. 7C, D). These results indicated that rivastigmine effectively prevents the decline in the hippocampal-dependent, but not amygdala-dependent, memory performance of the 6-month-old T2DM-AD model mice.

Rivastigmine reverses the retarded hippocampal neurogenesis in T2DM-AD model mice

Next, we compared the level of adult hippocampal neurogenesis among the four groups; Wt-ND, Tg-ND, AD (Tg-HFD)-vehicle, and AD (Tg-HFD)-rivastigmine. To count the number of new neurons in the hippocampus, we injected BrdU into these mice (n = 5 in each group) 5 times at 5 months of age, and examined their brains 4 weeks later. Figure 8 shows the quantification of BrdU-positive and NeuN-positive (BrdU+/NeuN+) cells in the dentate gurus of the four groups. Surprisingly, immunostaining showed that BrdU+/NeuN+ cells were markedly increased in the two groups of T2DM-AD mice (AD-Vehicle and AD-Rivastigmine) compared with normal control (Wt-ND) mice (one-way ANOVA, F _(3,19)  = 4.16, p < 0.05; Williams’ post hoc test, compared with Wt-ND: AD-Rivastigmine and AD-Vehicle, ^*p < 0.05).

To evaluate the maturation of the new neurons in the two T2DM-AD mouse groups (AD-Vehicle and AD-Rivastigmine), we performed immunohistochemistry for doublecortin (DCX) [70, 71]. Figure 9A shows the qualitative analysis of DCX-positive (DCX+) cells, which are newly born neuroblasts and immature neurons. For the morphological analysis, DCX+ cells in the dentate gyrus were classified into immature cells or mature cells according to their dendritic pattern (Fig. 9B), and the percentage of mature DCX+ cells in each group was quantified (Fig. 9C). Statistical analysis showed a clear decrease in the percentage of mature DCX+ cells in the AD-Vehicle group, whereas the AD-Rivastigmine group maintained the same percentage of mature DCX+ cells as that seen in the normal control group (Wt-ND) (one-way ANOVA, F _(3,11)  = 6.49, p < 0.05; Dunnett’s post hoc test, compared with AD-Vehicle: Wt-ND, ^*p < 0.05; AD-Rivastigmine, ^*p < 0.05). These results suggested that rivastigmine effectively reverses the retarded maturation of new neurons in the AD model mice.

To further assess the effect of rivastigmine treatment to reverse the retarded maturation of new neurons in the AD model mice, we analyzed the maturity of dendritic spines using a retroviral GFP labeling method. In this experiment, we labeled the hippocampal spines in 5-month-old mice by a retrovirus, and examined their brains 4 weeks later (Fig. 10A). The GFP-labeled spines of 4-week-old neurons were classified as immature type (filopodia and stubby) and mature type (thin or mushroom-shaped) (Fig. 10B). Confocal observation revealed that the development of dendrite complexity and the formation of spines in newly born neurons were significantly impaired in the AD-Vehicle mice compared with normal control mice (Wt-ND) (Fig. 10C, D). Intriguingly, rivastigmine treatment ameliorated this aberrant spine formation. In addition, quantitative analysis showed that there were significantly fewer GFP+ spines in the AD-Vehicle mice than in Wt-ND mice, and that rivastigmine restored the spine density (Fig. 10E; Total spines/10 μm: one-way ANOVA, F _(3,32)  = 39.04, p < 0.01; Dunnett’s post hoc test, compared with AD-Vehicle: Wt-ND, ^**p < 0.01; AD-Rivastigmine, ^**p < 0.01). Classification of these GFP+ spines revealed that there was a statistically lower percentage of mature spines in the new neurons of AD-Vehicle mice compared with Wt-ND and AD-Rivastigmine mice (Fig. 10E, ^**p < 0.01). These findings suggested that the development of dendritic spines in the AD-Vehicle mice was severely damaged by inflammatory responses, and that rivastigmine treatment noticeably prevented this neuroinflammation-related dysfunction in the maturation of newborn neurons.

These results pertaining to adult hippocampal neurogenesis suggested that in the T2DM-AD model mice, neuroinflammation strongly retards the maturation of newly born granule cells with respect to both dendrite and spine morphogenesis, and in response to this maturation defect, the number of newborn neurons markedly increases, mainly to compensate for the absence of structurally and functionally mature newly born neurons. Moreover, our results indicate that rivastigmine recovers the maturation, but not the proliferative activity, of newborn neurons predominantly by reducing the severe neuroinflammatory responses in the T2DM-AD model mice.