Sex specific effects of a high fat diet on metabolism, cognition, and pathology in the Tg-SwDI mouse model of Alzheimer’s disease

Abstract

Background

Alzheimer’s disease (AD) is the leading cause of dementia in the US, with over 80% of affected individuals experiencing comorbid metabolic disease. Along with age and sex, metabolic syndrome and prediabetes are known risk factors for developing dementia and AD, highlighting the complex nature of the disease. How these risk factors affect cerebral amyloid angiopathy (CAA) is less well studied. As such, we examined the effect of diet-induced metabolic syndrome and sex on cognition, neuroinflammation, and pathology in the Tg-SwDI mouse model of AD and CAA.

Methods

Male and female Tg-SwDI and WT mice were fed a low fat (LFD; 10% fat) or high fat (HFD; 60% fat) diet from 3 to 10 months of age. Metabolic, cognitive, and neuropathology outcomes were assessed.

Results

All HFD-fed mice gained weight and exhibited impaired glucose tolerance. Metabolic disturbances were most severe in AD females receiving HFD. In both males and females, HFD-fed AD mice showed increased anxiety-like behavior, decreased locomotor activity, and impaired recognition memory in the open field and novel object recognition tests, respectively. HFD-fed AD females specifically exhibited spatial memory deficits in the Barnes maze. Hippocampal microgliosis, activated microglia, and astrogliosis were more severe in AD mice, HFD decreased hippocampal microgliosis and astrogliosis but increased cytokine and chemokine expression in AD females. HFD-fed AD females had greater β-amyloid plaques and CAA in the thalamus compared to LFD-fed AD controls. All metrics of neuroinflammation significantly correlated with CAA pathology in the thalamus.

Conclusion

AD females experienced greater metabolic, cognitive, and pathologic effects in response to a HFD compared to AD males and WT controls. These observations provide a better understanding of how metabolic disease may differentially affect the development of dementia in men and women.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-026-03719-0.

Background

Alzheimer’s disease (AD) is the most common cause of dementia, with over 55 million people affected worldwide [1]. Risk factors such as age, sex, cardiovascular disease, and genetics highlight the complex nature of this disease [2]. Currently, over 80% of individuals with AD have comorbid metabolic disease [3–5]. Further, there is ample and growing evidence that prediabetes and metabolic syndrome increase the risk of developing dementia [2, 5–7]. Conservative estimates suggest the prevalence of these metabolic conditions is around 10% globally [8, 9]. Characterized by insulin resistance, hyperglycemia, hyperlipidemia, hypertension, obesity, or impaired glucose tolerance, these metabolic conditions greatly increase the risk for developing diabetes and cardiovascular disease [9]. Impaired metabolism can cause oxidative stress and persistent inflammation, which are pathological hallmarks shared with AD [7].

In both humans and animal models, metabolic disease has been associated with cognitive impairment [2]. Following long-term access to a high-fat diet (HFD), mice exhibit significant weight gain, insulin resistance, glucose tolerance, and ultimately cognitive decline [10–13]. Specifically, our laboratory and others have shown that HFD-induced metabolic disease causes cognitive deficits, reduces cerebral blood flow, and increases inflammation in wild type (WT) mice and that these deficits can be improved by reversing dietary fat or administering healthy plasma [11, 14, 15]. In both humans and animal models, studies show that when both HFD and β-amyloid pathology are present cognitive impairment and inflammation are more severe [16–18]. Specifically, extensive research shows that comorbid HFD-induced metabolic disease and dementia result in worsening systemic inflammation, neuroinflammation, β-amyloid burden, and recognition and spatial memory [16, 17, 19–30]. We have also shown that these effects are sex-specific, such that pathology and behavioral deficits exacerbated by HFD are worse in females [19, 31–34].

Due to the complexity of this disease, there is a large, ongoing effort to characterize AD models to better understand pathophysiology, as well as identify models suitable for investigating potential therapeutics. Often those with AD have both parenchymal β-amyloid plaques and β-amyloid accumulation within cerebral vessels, known as cerebral amyloid angiopathy (CAA) [35]. The Tg-SwDl mice are a transgenic mouse model with both parenchymal β-amyloid plaques and CAA pathology [36]. While the effect of HFD on β-amyloid pathology across sexes has been well studied in other models of AD and dementia, similar investigations have not been performed in this mouse model with CAA pathology. We hypothesized that the Tg-SwDl mouse model will have worse cognitive impairment than wild type control mice and that metabolic disease will exacerbate these impairments more so in females than males. This study is the first to characterize sex differences in the impact of a high fat diet on metabolism, cognition, and pathology in the Tg-SwDl transgenic model of AD with CAA. This study was ultimately designed to further characterize this AD mouse model and highlight sex and metabolic health as key biological variables.

Methods

Animals and experimental design

This study was conducted in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals, and protocols (protocol number 23-12002) were approved by the Institutional Animal Care and Use Committee at Albany Medical College (Albany, NY, USA). In the animal facility, temperature and humidity were set at 72 °F, 30–70% humidity, with a 12-h light/dark cycle (7 a.m. on/7 p.m. off). Mice were fed a standard chow diet (Purina Lab Diet 5P76) until three months of age. They were group housed in Allentown cages. Environmental enrichment (Nestlets and Shepherd Shacks) was provided, and mice were group housed at all times. Male and female wild type (WT; N = 73; C57BL/6J; #000664) and Tg-SwDl transgenic mice (Tg-SwDl; N = 66, C57BL/6-Tg(Thy1-APPSwDutIowa)BWevn/Mmjax; #034843; MMRRC_034843-JAX) were purchased from Jackson Laboratories (Bar Harbor, ME). Mice from the Tg-SwDI strain were obtained from the Mutant Mouse Resource and Research Center at the Jackson Laboratory, an NIH-funded strain repository, and were donated to the center by William Van Nostrand, Ph.D., Stony Brook University. Tg-SwDl mice were then used to maintain a colony at Albany Medical Center’s Animal Resource Facility. These transgenic mice express the human APP gene containing the Swedish K670N/M671L, Dutch E693Q, and Iowa D694N mutations regulated by the Thy1 promoter [25]. Starting at approximately three months of age, the mice develop Aβ deposition in the parenchyma and learning and memory deficits [36–38]. At around six months of age β-amyloid begins to accumulate in cerebral vessels. At 3 months of age, mice were placed on either a HFD (60% fat from lard, 20% protein, 20% carbohydrate, 5.24 kcal/g; D12492, Research Diets, New Brunswick, NJ, Additional File 1) or a low-fat control diet (LFD) (10% fat from lard, 20% protein, 70% carbohydrate, 3.82 kcal/g; D12450J, Research Diets) until the end of the study. The 60% fat diet was selected for the HFD groups because this diet is well characterized to induce obesity and prediabetes, with some studies suggesting higher fat percentages result in greater metabolic dysfunction [39, 40]. The LFD, control diet was selected to ensure protein, fiber, vitamin, and mineral contents, as well as fat source, were consistent to reduce variability between diets (Additional File 1). Mice were weighed every 4 weeks throughout the diet. At 9 months, mice underwent a glucose tolerance test (GTT), followed by a 2-week rest period and behavioral testing. After testing, animals were euthanized, and tissue was collected at 10 months of age. Experiments were conducted in cohorts of up to 25 mice with a total of 139 mice. In total, 7 mice were excluded due to premature death or the presence of other major health exclusions (hydrocephaly, large fighting wounds, tumors). The remaining 132 brains were post-fixed for immunohistochemistry. Blinding to diet and sex was not possible during in vivo experiments due to mouse appearance. During analysis, experimenters were blinded to sex, diet, and genotype.

Glucose tolerance test

As previously described, mice were given a GTT to assess metabolic status at 9 months of age [15, 20, 32]. The mice were fasted overnight, and their fasting blood glucose levels (t = 0) were measured from their tail vein the next morning using a glucometer (Verio IQ, OneTouch, Sunnyvale, CA, USA). Following an intraperitoneal injection of 2 g/kg of glucose, blood glucose levels were measured at 15, 30, 60, 90, and 120 min post-injection to assess metabolic response to a glucose challenge.

Behavioral testing

Following a 2-week recovery period, mice underwent testing for exploratory activity and anxiety-like behavior in the open field (day 1), recognition memory in the novel object recognition test (NORT; day 2), and spatial learning and memory in the Barnes maze (days 8-13). For each assessment, mice were acclimated in their home cage to the lighting and room for 1 hour prior to testing. Between each animal, 70% ethanol was used to clean the apparatus to remove olfactory cues. During each test, videos were recorded and independently analyzed using automated retracking software (ANY-maze 7.0, Stoelting, Wood Dale, IL).

Open field

Mice were placed in the testing apparatus (495 x 495 mm box) for 10 minutes in total. General locomotor activity was assessed using distance traveled, and the percent of time spent in the corners of the apparatus was used to determine anxiety-like behavior. Two mice were excluded from this behavioral test for being a statistical outlier via Grubb’s outlier test, resulting in a group size of 15-20/group.

Novel Object Recognition Test

Two five-minute trials performed in the same open field apparatus within 1.5-2.5 hours of each other constituted the NORT. For the first trial, mice were placed in the box and allowed to explore two identical objects (rubber ducks). During the second trial, the right object was replaced with a novel object (saltshaker), and the animal was allowed to freely explore. The percentage of time the animal spent with the novel object relative to the total amount of time with both objects during the second trial was used to assess recognition memory. The intertrial interval was ~2 hours. Due to an interruption during testing for some cohorts, 36 mice were excluded because the intertrial interval was inconsistent with other cohorts. Further, one animal that spent less than 2 seconds exploring the objects was excluded from analysis (9-16/group).

Barnes maze

Hippocampus-dependent spatial learning and memory were assessed using an seven-day variation of the Barnes Maze test. This protocol has been previously described in detail [33, 41]. On the first day, 1 beaker trial was performed in which the mice were guided to the target hole using a clear beaker. Following the beaker trial on the first day and twice a day for the next three days, the mice performed learning trials to learn the target hole in which they were given three minutes to find the target hole, allowing them to escape upon finding the target. On the seventh day, the mice were subjected to a two-minute probe trial in which there was no escape after finding the target. During all trials, visual cues were displayed on the wall.

Immunofluorescence

Mice were perfused with ice-cold 0.9% heparin saline. Brains were extracted and cut into right and left hemispheres. One half of the brain was post-fixed in 4% paraformaldehyde for 24 hours (side was determined by a flip of a coin), followed by submersion in 30% sucrose for at least 48 hours. Brains were then snap frozen in OCT (Thermo Fisher, 23-730-571) and stored at -80ºC until sectioning. Brains were sectioned coronally at 35 microns on a Leica CM1950 cryostat into 7 series. A series’ sections were washed with phosphate buffered saline (PBS) for 5 minutes 3 times. Slices were then transferred to a blocking and permeabilization buffer containing 0.3% Triton X-100 (Millipore, T9284) PBS (TPBS) solution with 10% donkey serum for one hour at room temperature. Following blocking, slices were incubated in primary antibody solution in 0.3% TPBS overnight in a cold room, containing 1:1000 goat Iba1 (Thermo Fisher, PA5-18039), 1:1000 rat CD68 (BIO-RAD, MCA1957), and 1:1000 rabbit GFAP (EMD Millipore, AB5804). Tissue was washed in PBS for 10 minutes 3 times before being incubated in secondary antibody solution in 0.3% TPBS for one hour at room temperature: 1:500 anti-goat 647 (Jackson Immuno Research, 705-605-147), 1:500 anti-rat Rhodamine Red-X (Jackson Immuno Research, 712-295-150), 1:500 anti-rabbit 488 (Jackson Immuno Research, 705-545-147), and 1:1000 DAPI (Thermo Fisher, D1306). Following treatment with secondary antibodies, tissue was washed twice in PBS for 10 minutes and once in PBS with 0.01% sodium azide (Krackeler, 45-71289-50G) for 15 minutes. Sections from another series were washed once with PBS for 10 minutes before being placed in a permeabilization buffer of 0.5% TPBS for 1 hour at room temperature. Then, slices were placed in a blocking buffer consisting of 0.5% TPBS and 4% donkey serum for 2 hours at room temperature. Following blocking, all sections were incubated in primary antibody solution, containing 1:500 rabbit human β-amyloid (polyclonal antibody derived from the 1-43 amino acid β-amyloid peptide, Thermo Fisher, 71-5800) in 0.5% TPBS and 4% donkey serum, for 24 hours in a warm room. Slices were then washed for 10 minutes 3 times before incubation in secondary antibody solution for 2 hours in a warm room: 1:500 rabbit 488, 1:100 lectin 649 (Vector Laboratories, DL-1178-1), and 1:1000 DAPI in 0.5% TPBS with 4% donkey serum. After 2 hours, all sections were washed with PBS for 10 minutes twice and once with PBS with 0.1% sodium azide. All sections were mounted from anterior to posterior and cover slipped with 120uL of fluoromount-G. Slides were allowed to dry overnight before being stored at 4ºC and imaged. Using the Axio Observer Fluorescent Microscope (Carl Zeiss Microscopy, Oberkochen, German), images of brain slices were obtained at 10x magnification using the same exposure times for each stain/labeling across all animals.

Quantification of β-amyloid plaques and CAA

Using ImageJ (NIH), image brightness in each channel was adjusted to the same threshold for all animals. Total β-amyloid plaques were quantified by measuring the percent area covered within each region of interest (ROI) by a blinded experimenter: the area of the retrosplenial cortex (rspCTX), stratum oriens of cornu ammonis 1 (CA1so), and ventral posterior thalamus (VP thal). In the same regions, CAA was assessed by measuring the percent area covered by pixels where β-amyloid and lectin colocalized. Parenchymal β-amyloid was assessed by subtracting the area of the region containing CAA from the area of the ROI containing total β-amyloid. The value for each animal is representative of an average of two to three ROIs from sections containing the anterior, dorsal hippocampus. These regions are associated with memory, spatial learning, spatial processing, sensorimotor integration, and are known to be affected in AD and/or in the Tg-SwDI mouse model [42–45].

Quantification of glia-related metrics

Similarly to previously described, ImageJ was utilized to set image brightness thresholds in each channel for all animals to assess microgliosis (Iba1), activated microglia (colocalization of Iba1 and CD68), and astrogliosis (GFAP). ROIs were drawn around the rspCTX, CA1so, polymorphic layer of the dentate gyrus (DGpo), VP thal, ventromedial hypothalamus (VMH), lateral hypothalamus (LH), and basolateral amygdala (BLA). Similarly, percent area covered averaged between one to three sections was utilized to assess the three modalities of neuroinflammation.

β-amyloid enzyme-linked immunosorbent assay (ELISA)

ELISAs were performed to quantify levels of soluble and insoluble β-amyloid 40 and 42 in the hippocampus of Tg-SwDI mice, according to the manufacturer’s instructions (ThermoFisher KHB3442; R&D Systems Inc, DAB140B and DAB142). The hippocampus was dissected from one hemisphere of the brain following perfusion with 0.9% heparinized saline at tissue collection. Insoluble samples were prepared using 70% formic acid after homogenization and 20X neutralizing solution immediately before performing the ELISA. For all ELISAs, samples were diluted 1:5 for soluble levels of β-amyloid 42, 1:10 for soluble levels of β-amyloid40, and 1:1000 for all insoluble levels. All β-amyloid levels were normalized to the relative level of soluble proteins in the sample, using a Pierce BCA protein assay (Thermo Fisher, 23225).

Hippocampal cytokines

Cytokine levels in the hippocampus were assessed in Tg-SwDI and WT mice using the Bio-Plex Pro Mouse Cytokine 23-Plex Assay according to the manufacturer’s instructions (Bio-Rad, Carlsbad, CA, USA, M60009RDPD). Hippocampal tissue was utilized from regional dissections during tissue collection, following perfusion with 0.9% heparinized saline. Cytokine levels were normalized to total soluble proteins in the hippocampal section using the Pierce BCA protein assay (Thermo Fisher, 23225).

Statistics

Statistical analyses were performed using Prism 10 (GraphPad Software, San Diego, CA, USA), with statistical significance set at p < 0.05. All data are shown as mean ± SEM. Statistical outliers were assessed with Grubbs’ test, after which a 2-way ANOVA was performed with Fisher’s least significant difference in data segregated by sex, where strain (WT vs. Tg-SwDI) and diet (LFD vs. HFD) were the independent variables. Fisher’s least significant difference multiple comparisons were established a priori to more thoroughly investigate the effect of diet and strain within and between groups. To assess for sex differences, a separate 3-way ANOVA was performed without post-hoc analyses for all tests except for GTT, such that strain, diet, and sex were the independent variables. For all metabolic data, a ROUT test was performed prior to analysis to assess for statistical outliers. Further, one-sample t-tests, or Wilcoxon tests for data sets not normally distributed, were performed to assess individual group performance relative to chance (50% in NORT, 15% in Barnes Maze). For all ANOVAs, the Geisser-Greenhouse correction was utilized to compensate for unequal variances across groups. Correlations were assessed using Spearman’s correlation matrices for the appropriate data sets. F statistics for 2-way and 3-way ANOVAs are supplied in Additional File 2 and Additional File 3, respectively.

Results

HFD causes greater metabolic disturbances in AD and WT females compared to males

Previously, our lab has shown that HFD-induced metabolic syndrome is more severe in females in animal models of AD and vascular contributions to cognitive impairment and dementia (VCID) [19, 20, 33]. In order to investigate these features in the Tg-SwDI model, WT and AD mice received either HFD or LFD from 3 months of age onward, and the GTT was used at 9 months to assess metabolic status (Fig. 1A). We demonstrate here that these sex differences are consistent in the Tg-SwDI AD model. For each sex, there was a main effect of diet on weight gain following 6 months of dietary intervention (p < 0.0001, Fig. 1B-C), on percent of visceral fat relative to total body weight (p < 0.001, Fig. 1D-E.), and on area under the blood glucose curve (AUC) during GTT (p < 0.0001, Fig. 1F-I). Further, there was a main effect of strain in both sexes on body weight at GTT (p < 0.01, Fig. 1B-C) and GTT AUC (p < 0.001, Fig. 1H-I), such that AD mice weighed more and had greater GTT AUC. Notably, prolonged exposure to a HFD induced a metabolic syndrome phenotype consistent with our previous findings, with elevated blood glucose levels after fasting (Fig. 1F-G at time = 0) and in response to a glucose challenge (Fig. 1H-I) [46, 47]. The effects of strain and diet are sex dependent, demonstrated by post-hoc tests that show AD HFD females have greater metabolic impairment compared to WT HFD females (p < 0.01 for weight, % visceral fat, and GTT AUC) but AD LFD males have less body weight, % visceral fat, and GTT AUC compared to WT LFD males (p < 0.05 for all). Assessment of sex differences through a 3-way ANOVA (Additional File 2.) showed a significant effect of sex, diet, interaction between strain and sex, interaction between strain and diet, and interaction between sex and diet for all metabolic outcomes (p < 0.05 for all). Altogether, these findings suggest AD females experience greater metabolic impairment in response to a HFD compared to WT females or WT or AD males.

Fig. 1.

Diet affects weight, visceral adiposity, and glucose tolerance in WT and Tg-SwDI males and females. A The experimental timeline is shown. B, C Body weight prior to the GTT and following 6 months of dietary intervention was used to assess weight gain. D, E Wet weight (g) of visceral fat was assessed at the end of the experiment and analyzed relative to percent of total body weight at that time. F-I GTT was performed to assess metabolism and diabetic state. F, G After overnight fasting (t = 0), blood glucose levels were measured, and mice were injected with a glucose challenge during which blood glucose was measured at 15, 30, 60, 90, and 120 min. Each group was compared to the control WT LFD group at each time point, such that significantly different blood glucose compared to control is denoted as * p < 0.05 WT LFD vs. WT HFD, # p < 0.05 WT LFD vs. TgSwDI LFD, and ^ p < 0.05 WT LFD vs. TgSwDI HFD. H, I Area under the curve (AUC) from GTT was assessed as a measure of responsiveness to the glucose challenge. Higher AUC is indicative of worse metabolic disease. Error bars represent SEM. * p < 0.05; ** p < 0.01; **** p < 0.0001. Two-way ANOVA (n = 11–20/group). Main effects are demonstrated above the graph when significant, and post-hoc comparisons are demonstrated as significance bars where appropriate

HFD worsens cognitive deficits in AD mice in a sex-specific manner

Our previous research shows that diet differentially affects cognition in females in models of AD, VCID, and mixed dementia [19, 20, 32]. Here, we show that these observations hold true in the Tg-SwDI model of AD and CAA. Percentage of time in corners and in center during the open field test were used as an assessment of anxiety-like behavior. There was a significant main effect of diet in both males and females in the % of time in corners (p < 0.05, Fig. 2A-B) and in the males in % of time in center (p < 0.05, Fig. 2C), such that mice on HFD showed increased anxiety-like behavior. Post-hoc tests demonstrated significantly greater anxiety-like behavior in both WT and AD HFD females compared to their LFD-fed female controls (p < 0.05, Fig. 2B). Total distance traveled during the open field test was used as an assessment of general locomotor activity. In both males and females, there was a main effect of strain (p < 0.05, Fig. 2E-F) and diet (p < 0.0001, Fig. 2E-F), with HFD-fed and AD mice independently traveling less. Post-hoc tests highlighted this effect, showing that all HFD-fed groups traveled smaller distances compared to the LFD-fed mice (p < 0.05, Fig. 2E-F). Together, this demonstrates that HFD AD females specifically show increased anxiety-like behavior and HFD-fed mice traveled less.

Fig. 2.

Diet and sex interact to affect anxiety-like behavior, general activity, and recognition memory. A-D Anxiety-like behavior was assessed through percent of time spent in the corners (A, B) and in the center (C, D) during the open field test. Percent of time spent in the corners was increased in WT and Tg-SwDI animals receiving HFD across sexes, and similarly percent of time in the center was decreased in WT and TgSwDI animals receiving HFD. E, F Distance traveled during the open field test was used to assess general locomotor activity. HFD and AD independently decreased the total distance traveled in both males (E) and females (F). G, H Recognition memory was assessed using the novel object recognition test (NORT). Percent of time spent with the novel object relative to total time spent with objects was calculated, such that performance significantly greater than 50% chance is indicative of intact memory. All AD males and HFD AD females did not perform significantly greater than chance. Pink and blue line = chance. Error bars represent SEM. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, # p < 0.05 vs. chance, ## p < 0.01 vs. chance, ### p < 0.001 vs. chance. Two-way ANOVA and one sample t-test (n = 13–20/group). Main effects are demonstrated above the graph when significant, and post-hoc comparisons are demonstrated as significance bars where appropriate

Recognition memory was assessed in the novel object recognition test (NORT). Preference for the novel object in the NORT is measured as the recognition index: percent of time spent with the novel object relative to total time spent with objects. Performance not greater than 50% chance indicates impairment in recognition memory. Assessed individually within groups, all AD males (LFD-fed males, p > 0.05, t = 1.804, df = 15, HFD-fed males, p > 0.05, t = 1.204, df = 12, Fig. 2G) and HFD-fed AD females (p > 0.05, t = 0.7974, df 8, Fig. 2H) did not perform significantly greater than chance, indicating impaired recognition memory. Comparisons between groups demonstrated a main effect of strain in the females (p < 0.01, Fig. 2H), such that AD females spent significantly less time exploring the novel object, indicating more severely impaired memory. Together, this suggests that HFD-fed AD mice show impairment in recognition memory regardless of sex.

The Barnes maze test was used to assess spatial learning and memory via the hidden trials and probe trial, respectively. The percentage of time spent in the portion of the maze between the center and holes directly adjacent to the target hole (target cone) and percentage of incorrect hole entries were used to assess performance. In the probe trial, to assess spatial memory, percent of time in the target cone was assessed in each group independently to compare performance to chance, which is 15% of time spent in the target cone. All AD females and LFD-fed AD males did not perform significantly different to chance (AD F LFD, p > 0.05, t = 0.07613, dt = 15, AD F HFD, p > 0.05, t = 0.7334, df = 13, AD M LFD, p > 0.05, t = 0.4706, df = 15, Fig. 3A-B), indicating impairment in spatial memory. Comparisons between groups revealed a main effect of diet in the males (p < 0.05, Fig. 3A) and a main effect of strain in the females (p < 0.01, Fig. 3B), such that HFD-fed males spent more time in the target cone and AD females spent less time in the target cone. Further, there was a main effect of strain on percent errors in the probe trial in females (p < 0.001, Fig. 3D), such that AD females made greater errors. Post-hoc comparisons showed that AD HFD-fed females made significantly more errors than WT HFD-fed females (p < 0.01). Interestingly, during spatial learning, there was a significant effect of diet in the males and a significant effect of strain in the females (p < 0.05, Additional File 4 C-D), such that HFD and WT strain independently resulted in less errors during learning in males and females, respectively. Taken together, these findings demonstrate that AD mice receiving HFD show impairment in spatial memory and that these observations are particularly strong in HFD-fed AD females.

Fig. 3.

HFD impairs spatial memory in female Tg-SwDI animals. Spatial learning and memory were assessed using the Barnes maze. Spatial memory was assessed through the probe trial (A-D) through percent of time in the cone adjacent to the target, such that performance above 15% chance was indicative of greater learning,, B) and percent of entries in the incorrect holes relative to the correct hole (C, D). TgSwDI females spent less time in the target cone (F) and had greater percentage of erroneous entries (H) during the probe trial compared to WT females. Red line = chance. Error bars represent SEM. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, # p < 0.05 vs. chance, ## p < 0.01 vs. chance, ### p < 0.001 vs. chance. Two-way ANOVA and one sample t-test or Wilcoxon test if not normally distributed (n = 14–20/group). Main effects are demonstrated above the graph when significant, and post-hoc comparisons are demonstrated as significance bars where appropriate

Given that diet-driven weight gain can affect movement and parameters influenced by speed and distance, we assessed how all cognitive metrics were related to distance traveled in the open field apparatus. Time in corners (r²=-0.37, p < 0.0001, Additional File 5) and time in center (r² = 0.24, p < 0.001) during the open field tests negatively and positively correlated with distance traveled, respectively. There were no significant correlations with cognitive measures. These results indicate that anxiety-like behavior is influenced by general locomotor activity but that all other cognitive findings are independent of this metric.

AD-induced glial activation is altered by HFD in females

Previously, our lab has investigated glial activation in other models of AD with and without comorbid metabolic dysfunction [19, 20, 32, 41]. Here, we sought to expand upon these observations in the Tg-SwDI model that contains CAA pathology. Activated microglia was assessed as percent of the region of interest covered by cells immunolabeled with both Iba1 and CD68, due to CD68 being an indication of active phagocytosis. Similarly, microgliosis and astrogliosis were quantified by percent of the region of interest covered by Iba1-immunolabeled or GFAP-immunolabeled cells. Given previous findings that suggest hippocampal glia are involved in β-amyloid pathology, we first examined these measures of glial activation in the stratum oriens layer of cornu ammonis 1 (CA1so) and the polymorphic layer of the dentate gyrus (DGpo, Fig. 4A) [48]. In both regions and in males and females, there was a main effect of strain on microgliosis (p < 0.0001, Fig. 4B-E), such that AD males and females showed increased area of microglia regardless of diet. Further, in both regions AD females receiving HFD showed less microgliosis compared to AD females receiving LFD (p < 0.01, Fig. 4C and E). Interestingly, there was similarly a main effect of strain on activated microglia and astrogliosis across sexes (p < 0.01, Fig. 4F-M), resulting in increased area covered by activated microglia and astroglia in hippocampal tissue from AD mice. In CA1so, post-hoc tests reveal that HFD-fed AD females show significantly less astrogliosis compared to LFD-fed AD females (p < 0.01, Fig. 4K). Further, a 3-way ANOVA was performed on each glial metric to assess for sex differences. In CA1so, there was a main effect of sex and an interaction between strain and sex on activated microglia, as well as an interaction between sex and diet and interaction between strain, sex, and diet on microgliosis (p < 0.05, Additional File 3). This resulted in greater overall glial activation in female AD mice compared to male AD mice. Overall, these results suggest that the Tg-SwDI mice have more hippocampal glial activation than WT mice and that HFD decreases glia in some regions in AD females.

Fig. 4.

AD, diet, and sex interact to affect microgliosis, activated microglia, and astrogliosis in the hippocampus. (Ai) Hippocampal regions of interest examined included the stratum oriens layer of the cornu ammonis 1 (CA1so) and the polymorphic layer of the dentate gyrus (DGpo) in anterior coronal sections. (Aii) Representative images of Iba1, CD68, GFAP, and DAPI fluorescence in the hippocampus are shown. B-E Microgliosis was assessed through Iba1 immunofluorescence by percent of area covered in the region of interest. F-I Activated microglia was assessed by colocalizing Iba1 and CD68 immunofluorescence and quantifying the percent of colocalized fluorescence in the region of interest. J-M Astrogliosis was examined through GFAP immunofluorescence via percent of area covered in the region of interest. Error bars represent SEM. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Two-way ANOVA (n = 5–6/group). Main effects are demonstrated above the graph when significant, and post-hoc comparisons are demonstrated as significance bars where appropriate

These trends were consistent in the ventral posterior thalamus (VP thal), which was an area of interest due to its implication in CAA pathology in this model [37, 45, 49]. Specifically, post-hoc tests showed that microgliosis and activated microglia were significantly decreased in the VP thal of HFD-fed AD females compared to control AD females (p < 0.05, Additional File 6 C and Additional File 6E).

We also examined these measures of glial activation in the hypothalamus (ventromedial hypothalamus, VMH and lateral hypothalamus, LH) and amygdala (basolateral amygdala, BLA) due to these regions being involved in hunger/satiety and anxiety-like behavior, respectively [50, 51]. In the females, there was a main effect of strain on microgliosis (p < 0.05, Fig. 5E) and activated microglia (p < 0.01, Fig. 5K) in the LH, such that inflammation was greater in AD mice. Further there was an interaction between strain and diet on microgliosis in the LH (p < 0.05, Fig. 5E), where HFD decreased microglia in AD females. Additionally, post-hoc tests revealed that HFD AD females had less microglia in the LH compared to LFD AD females (p < 0.05, Fig. 5E). Interestingly, in the males, there was a main effect of strain (p < 0.001, Fig. 5J) and an interaction between strain and diet (p < 0.05) on activated microglia in the LH such that HFD increased activated microglia in AD males. Further, post-hoc tests demonstrated that HFD AD males had greater activated microglia in the LH (p < 0.01, Fig. 5J) and microgliosis in the VMH (p < 0.05, Fig. 5B) compared to LFD AD mice. Similar observations persisted in the BLA, such that there were greater activated microglia in the BLA of male and female AD mice, resulting in a main effect of strain in both sexes (p < 0.05, Fig. 5L-M). Additional post hoc comparisons showed that HFD-fed AD females had less astrogliosis compared to LFD AD females (p < 0.05, Fig. 5S). Interestingly, in the males, there was a main effect of strain on BLA astrogliosis (p < 0.05, Fig. 5R), such that AD males had decreased astrogliosis compared to WT males regardless of diet. Altogether, these results demonstrate a similar decrease in AD-induced glial activation in the hypothalamus and amygdala of HFD-fed AD females.

Fig. 5.

AD, diet, and sex interact to affect microgliosis, activated microglia, and astrogliosis in the hypothalamus and amygdala. (Ai) The regions of interest examined are demonstrated in a ventral view of an anterior coronal section, including the ventromedial hypothalamus (VMH), lateral hypothalamus (LH), and basolateral amygdala (BLA). (Aii) Representative images are shown in the LH for all treatment groups. B-G Microgliosis was assessed through Iba1 immunofluorescence by percent of area covered in the region of interest. H-M Activated microglia was assessed by colocalizing Iba1 and CD68 immunofluorescence and quantifying the percent of colocalized fluorescence in the region of interest. N-S Astrogliosis was examined through GFAP immunofluorescence via percent of area covered in the region of interest. Error bars represent SEM. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Two-way ANOVA (n = 5–6/group). Main effects are demonstrated above the graph when significant, and post-hoc comparisons are demonstrated as significance bars where appropriate

HFD increases neuroinflammation in female AD mice

Previously, we have demonstrated that diet-induced obesity and metabolic disease affect cytokine, chemokine, and interleukin (IL) expression in the brain and periphery in other models of AD and VCID [20, 32]. Here, we aimed to better elucidate local inflammatory responses in the hippocampus of Tg-SwDI mice in response to diet-induced metabolic dysfunction. All cytokines and chemokines were normalized to total mass of soluble proteins in the hippocampal sample. There was a main effect of strain on IL-1α levels in both males and females (p < 0.001, Fig. 6A-B), such that there were greater levels of IL-1α in AD mice compared to WT mice. In the females, there was additionally a main effect of diet (p < 0.05, Fig. 6B) and an interaction between diet and strain (p < 0.05), where HFD resulted in a greater increase in IL-1α expression in AD females. Interestingly, for IL-17, similar results were observed in the females, where HFD-fed AD females had greater levels of IL-17 compared to WT or AD controls, such that there was a main effect of strain, diet, and interaction between the two (p < 0.05, Fig. 6D). However, in the males, there was a significant interaction between strain and diet (p < 0.05, Fig. 6C), such that HFD resulted in greater hippocampal IL-17 in AD males only. Similarly, in males and females, there was a main effect of strain on monocyte chemoattractant protein-1 (MCP-1), MIP-1b, and eotaxin levels (p < 0.05, Fig. 6E-J), such that chemokine levels were increased in AD mice except for MCP-1 in males where levels were decreased compared to WT mice. In the females, there was a main effect of diet and an interaction between strain and diet on MCP-1 (p < 0.05, Fig. 6F) and eotaxin levels (p < 0.05, Fig. 6J), such that a HFD increased these levels and this trend was exacerbated in AD females. For both RANTES and tumor necrosis factor alpha (TNF-α), there were significant main effects of strain, diet, and interaction between strain and diet only in the females (p < 0.05, Fig. 6L and N), where the levels of RANTES and TNF-α were increased in AD females and exacerbated in females receiving a HFD. Taken together, these results demonstrate that most hippocampal cytokines and chemokines are increased occasionally in AD males and more frequently in AD females, where diet-induced metabolic dysfunction exacerbates cytokine levels.

Fig. 6.

AD, diet, and sex interact to affect hippocampal cytokine expression. Cytokine and chemokine levels in the hippocampus were normalized to the total amount of soluble protein in the sample. A-D Interleukin values were assessed in males and females. E-N Chemokine and cytokine levels were similarly assessed in males and females. Error bars represent SEM. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Two-way ANOVA (n = 7–8/group). Main effects are demonstrated above the graph when significant, and post-hoc comparisons are demonstrated as significance bars where appropriate

HFD exacerbates pathology in the thalamus of AD females

While we have previously examined the effect of HFD on neuropathology in other AD and dementia models, this study is the first to examine how comorbid metabolic dysfunction affects CAA pathology in the Tg-SwDI model [19, 20, 41]. Extensive research demonstrates the hippocampus and cortex are heavily burdened by β-amyloid pathology and additional findings show the cortex and thalamus also contain significant CAA in the Tg-SwDI model [48, 49]. As such, we examined pathology in the CA1so, retrosplenial cortex (rspCTX), and VP thal. We measured CAA by quantifying the area of a region of interest containing β-amyloid plaques colocalized with blood vessels, using lectin staining. Similarly, we also assessed total β-amyloid deposition, parenchymal β-amyloid, and vessel density using the percent of the region containing β-amyloid -tagged plaques or lectin staining. Unsurprisingly, across all regions and in both males and females, there was a main effect of strain on total β-amyloid deposition (p < 0.05, Fig. 8.B-E, Additional File 7), such that tissue from WT mice contained no β-amyloid. Interestingly, post-hoc tests revealed a significant increase in total β-amyloid in the VP thal of HFD-fed AD females compared to LFD-fed AD females (p < 0.001, Fig. 8E). In all regions, there was also a main effect of strain on blood vessel density (p < 0.05, Fig. 8F-I, Additional File 7), such that AD males and females had increased vascular density compared to WT controls. In the VP thal and rspCTX, there was a main effect of strain on CAA in both males and females (p < 0.05, Fig. 8L-M, Additional File 7), indicating CAA is only observed in tissue from AD mice (Additional File 8). Further, post-hoc tests show that AD females receiving HFD have greater CAA in the VP thal compared to AD females receiving LFD (p < 0.001, Fig. 8M). Similarly, when examining parenchymal β-amyloid deposition, there was a main effect of strain in all regions in both males and females (p < 0.05, Fig. 8N-Q, Additional File 7), such that there was no β-amyloid in the parenchyma of WT mice. Further, there was a main effect of diet and an interaction between diet and strain on parenchymal β-amyloid in the VP thaL (p < 0.05, Fig. 8Q), where HFD-fed AD females had more parenchymal β-amyloid compared to LFD-fed AD females (p < 0.001). Additionally, β-amyloid burden was further interrogated by assessing four species (40 S, 40I, 42 S, 42I) within the hippocampus of AD mice, given the region’s involvement in cognition and β- β-amyloid pathology. Interestingly, diet did not affect relative levels of β-amyloid species in either males or females. There was a main effect of species in both sexes (p < 0.0001, Fig. 7A and C), such that the mean mass (pg) of β-amyloid relative to total soluble proteins (µg) was significantly different across species, and an interaction between diet and β-amyloid species in the males (p < 0.05, Fig. 7A). Further, a 3-way ANOVA performed for sex differences demonstrated a main effect of sex (p < 0.05, Additional File 3) and an interaction between species and sex (p < 0.05), such that β-amyloid levels were consistently elevated in AD females and sex differentially altered β-amyloid levels within β-amyloid species. Further, the ratio of β-amyloid 40 and β-amyloid 42 was assessed across β-amyloid species and sex. Diet did not affect this ratio, but there was similarly a main effect of β-amyloid species in males and females (p < 0.0001, Fig. 7B and D). Taken together, these findings show that exposure to HFD increases total β-amyloid deposition, CAA, and parenchymal β-amyloid in the thalamus in AD females and AD increases vascularity across several brain regions in both sexes.

Fig. 8.

High fat diet does not affect β-amyloid species in the hippocampus. Insoluble and soluble species for β-amyloid 40 and 42 were assessed in the hippocampus of TgSwDI animals. A, C There was an effect of β-amyloid species in males and females (p<0.0001), such that mass (pg) of β-amyloid normalized to total soluble proteins was significantly different across β-amyloid species. B, D Similarly, there was an effect of β-amyloid species in both males and females when the ratio of β-amyloid 40β/42β was assessed across insoluble and soluble β-amyloid species. Error bars represent SEM. Two-way ANOVA (n=8-9/group). Main effects are demonstrated above the graph when significant, and post-hoc comparisons are demonstrated as significance bars where appropriate

Fig. 7.

Sex and diet interact to affect CAA pathology and blood vessel density. (Ai) Regions of interest in the hippocampus (CA1so) and thalamus (ventral posterior thalamus, VP thal) are shown in a dorsal view of an anterior coronal section. (Aii) Representative images of β-amyloid, lectin, and DAPI fluorescence are shown in the VP thal. B-E Total β-amyloid pathology was assessed via percent of area of interest containing β-amyloid immunofluorescence. F-I Blood vessel density was assessed through lectin fluorescence by percent of area covered in the region of interest. J-M CAA pathology was measured via colocalization of β-amyloid and lectin fluorescence and assessed by the percent of the region of interest covered by the colocalized fluorescence. N-Q Parenchymal β-amyloid was measured by subtracting the β-amyloid colocalized with lectin from total β-amyloid in each region. The human β-amyloid antibody was derived from the full length 1–43 β-amyloid peptide. Error bars represent SEM. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Two-way ANOVA (n = 5–6/group). Main effects are demonstrated above the graph when significant, and post-hoc comparisons are demonstrated as significance bars where appropriate

Worse metabolic outcomes correlate with CAA pathology in AD females

To assess how HFD-induced obesity related to CAA pathology across sexes, we performed correlation matrices between metabolic measures, weight at the end of study and AUC during the GTT, and CAA pathology first in all AD mice and then separately in AD males and females. When data were pooled across sexes, there was a significant positive correlation between thalamic CAA pathology and percent visceral fat (r² = 0.461, p < 0.05, n = 24, Fig. 9A). However, when assessed separately, only AD females demonstrated significant positive correlations between metabolic outcomes and CAA pathology. Specifically, thalamic CAA pathology positively correlated with percent visceral fat (r² = 0.643, p < 0.05, n = 12, Fig. 9C), and hippocampal CAA pathology tended to positively correlate with AUC during the GTT (r² = 0.510, p = 0.094, n = 12) and percent visceral fat (r² = 0.517, p = 0.089, n = 12), while not reaching statistical significance. Interestingly, in AD males, hippocampal CAA negatively correlated with AUC during the GTT (r²=-0.762, p < 0.05, n = 12, Fig. 9B). These results suggest that HFD exacerbates vascular β-amyloid pathology, specifically in females.

Fig. 9.

CAA pathology in the thalamus and hippocampus correlates with metabolic metrics in AD females. A correlation matrix was constructed between metabolic (endpoint weight, GTT AUC, and % visceral fat) metrics and CAA in the VP thalamus, CA1so, and rspCTX in all AD animals (A), AD males (B), and AD females (C). When analyzing all AD animals together, there was a significant correlation between thalamic CAA and visceral fat (A). When sex was considered independently, it was revealed that this positive correlation was consistent in the thalamus and hippocampus of AD females (C) but not males (B). Number and color gradient in box reflects spearman r coefficient. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. (n = 12–24)

Discussion

This study sought to better understand the effect of endocrine risk factors, such as sex and metabolic disease, on AD and VCID by examining metabolism, cognition, and neuropathology. Evidence shows that women are more likely to develop AD, likely due to underlying differences in metabolism, phagocytosis, and immune response that change during aging [52]. Further, metabolic risk factors such as metabolic syndrome, prediabetes, and obesity are known to increase the risk of developing VCID, with these comorbidities confounding the risk for women [2, 3, 53]. Previous studies in our lab have shown how diet-induced metabolic dysfunction worsens cognition, inflammation, and pathology in other AD and VCID models in a sex-specific manner [19, 20, 31, 32, 34]. How exactly sex and metabolic syndrome interact to affect cognition and pathology is unknown in the Tg-SwDI model of AD and VCID (CAA). We used chronic HFD administration to model obesity and metabolic syndrome. While all animals developed impaired glucose tolerance, female mice experienced greater metabolic disturbances. Further, these sex differences were exacerbated in AD mice. Similarly, metabolic syndrome resulted in greater cognitive impairments in AD females compared to males. While HFD increased anxiety-like behavior across sexes in AD animals, females also experienced disturbances in spatial memory. Interestingly, HFD decreased hippocampal microgliosis and astrogliosis but increased cytokine and chemokine expression in AD females. Further, HFD increased thalamic β-amyloid plaques and CAA pathology in AD females, with little effect on neuropathology in the males. Together, these results suggest that females with AD are more vulnerable to metabolic, cognitive, and pathologic effects of diet-induced metabolic syndrome. This, coupled with previous findings, support the idea that metabolic disease may differentially increase the risk of developing dementia and alter the disease process in women compared to men [2].

In this study, we found that AD females had more severe metabolic impairment in response to chronic HFD administration. Specifically, HFD resulted in greater weight gain, visceral fat accumulation, and glucose intolerance in AD females compared to males. This is consistent with previous findings in our lab and others in other models of AD and VCID, which have demonstrated increased weight gain, worsening glucose intolerance, and increased subcutaneous and visceral fat following a chronic HFD [19, 20, 31, 46, 54]. Interestingly, when HFD is started at a younger age (< 9 weeks), males gain more weight and exhibit greater metabolic impairment than female mice [15, 55–58]. Notably, this sex difference is reversed when the diet is initiated at an older age [15, 59]. When animals are introduced to HFD at an older age, as was done in this study, females exhibit greater metabolic impairment, including increased weight gain and adiposity and worse glucose tolerance [10, 15, 57]. An additional consideration when discussing diet-induced metabolic dysfunction is differences in food consumption. Ongoing research in humans and animals suggests hormones may have an effect on vulnerability to HFD but that their effect on feeding behavior is not well understood [60]. A previous study in our lab in a different AD model showed that HFD-fed AD females exhibited greater food intake compared to HFD-fed WT females or HFD-fed AD males, indicating a potential effect of feeding behavior on metabolic sex differences [20]. While this study was performed in a different AD model, we did not measure food consumption in the current study and cannot definitively determine the effect of feeding behavior on our observed sex difference. This would be an interesting parameter to investigate in future investigations and may provide more insight into whether sex differences in weight and metabolic factors are affected by alterations in feeding behavior in this AD model.

Previously, we have shown that a similar diet regimen resulted in changes in the periphery in AD males and females: hepatic fibrosis, steatosis, and increases in circulating leptin. However, in the hypothalamus, levels of GFAP and IL-1β were greater in females and associated with their increased weight gain in response to a HFD, suggesting neuroinflammation in this region may contribute to metabolic sex differences [20]. These observations are consistent with trends in AD patients, where women are thought to be more susceptible to metabolic disease and diabetes [3]. Interestingly, we also found sex differences in our LFD-fed, control AD males and females, such that the AD males weighed less, had less visceral fat, and had greater glucose tolerance than WT males. Conversely, LFD-fed AD females weighed more and had worse glucose tolerance compared to WT females. This is also consistent with human data, which shows that in mid-adulthood underweight men and overweight females have greater risk of developing AD [61]. Additionally, previous research has demonstrated that mutant β-amyloid precursor protein (APP) differentially alters lipid metabolism in the periphery in control vs. obesogenic settings in AD and CAA models, suggesting a mechanism for these metabolic differences [62]. Further, recent findings demonstrated that glucagon-like peptide-1 therapy conjugated with estradiol can improve HFD-induced metabolic, cognitive, and pathologic deficits via distinct sex-specific mechanisms [63]. Together, these findings suggest that underlying sex differences in metabolism may be exacerbated by AD and vascular dementia, necessitating the need to better understand how these differences may be modulated to target sex-specific effects of disease.

We also demonstrated diet-driven sex differences in cognition in this model. Specifically, we showed that HFD increased anxiety-like behavior and decreased exploration in males and females, but that recognition and spatial memory were impaired only in AD females. These cognitive sex differences are similar to what we have shown previously in other models of AD and VCID [19, 20, 32, 41]. Interestingly, in both sexes, HFD was sufficient to induce increased anxiety-like behavior and decreased mobility, and these trends were exacerbated in AD animals. This is consistent with previous findings that model obesity-driven changes in affect [64]. Although, decreased mobility did correlate with anxiety-like behavior, which presents a limitation in this behavioral assay to distinguish physical restrictions due to weight and anxiety-like behavior. Notably, decreased mobility may affect movement in the NORT and barnes maze, distance traveled did not correlate with recognition and spatial memory metrics, suggesting diet-induced physical restrictions are not major confounding variables for these tests. In assessments of memory, AD and/or HFD were required to induce impairments. For recognition memory, all AD males showed impairment, but in females only AD HFD animals were impaired. Although, AD females on control diet did demonstrate worse recognition memory compared to WT controls. Interestingly, men and women are known to have differing strengths in performance in metrics of memory [65]. One study found that women with family history of AD consistently performed better on recognition memory tasks compared to men with a positive family history [66]. In assessments of spatial memory, HFD AD females experienced significant impairment, whereas HFD AD males did not perform differently than control AD males. Recent findings in another model of AD showed similar pronounced spatial deficits in AD females compared to males [67]. Our findings supplement these observations of sex-specific cognitive deficits by demonstrating that many of these features are exacerbated by HFD. Additionally, these data suggest that the combined effect of metabolic disease and AD often result in cognitive impairments more severe than each individual insult.

When assessing for glial activation, we found that microgliosis, activated microglia, and astrogliosis were all consistently elevated in AD animals. Interestingly, HFD appeared to temper the activation of glia in AD females, while HFD had little effect on AD males. Previous work in our lab modeling metabolic disease in AD models has shown differences in glial response between males and females [19, 20]. This and ongoing aging research support the notion that sex differences may contribute to underlying differences in metabolism, immune reactivity, and autophagy [52, 68, 69]. However, our results suggest that both diet-induced metabolic disease and VCID together, specifically in females, may impair immune responsivity in the brain. Recent studies have demonstrated a similar phenomenon of immune exhaustion in AD, showing that some T cells and microglia become exhausted by pathology and that these conditions result in worsening cognitive impairment [70–72]. Additionally, some research suggests that immune exhaustion in AD and other diseases is modulated by diet-induced obesity and intrinsic, molecular sex differences [71, 73]. Further, imaging studies in patients suggest that prediabetic women experience cerebral hypometabolism to a greater extent than men [53]. It is possible that this holds true in mice as well, suggesting that metabolic disease may impair metabolism in the brain, which could affect mobilization of neuroinflammatory cells. While microglia specifically have been heavily implicated in β-amyloid pathology, AD models that lack microglia are shown to have greater β-amyloid and CAA pathology and early lethality [74]. These findings, coupled with our observations of decreased microgliosis and astrogliosis in AD HFD females, suggest that metabolic disease may worsen pathology by impairing appropriate glia immunoreactivity in a sex-specific manner. Our assessment of cytokines and chemokines in the hippocampus provides further context for these findings.

Consistent with our glial findings in this study, and previous work in other animal models of AD [32, 34], we found that most cytokine levels were increased consistently in female AD mice compared to WT. Interestingly, we also demonstrated that AD females showed even greater levels of cytokines when receiving a HFD compared to a LFD. IL-1α and IL-17 were both found to be increased in the hippocampus of males and females receiving HFD, and both of these cytokines have been implicated in plaque formation and accelerated aging in AD [75–77]. Similarly, HFD exacerbated levels of MCP-1, eotaxin, MIP-1b, TNF-α, and RANTES in females, all of which have been shown to be associated with worsening pathology and cognitive impairment [78–80]. Furthermore, randomized control trials have demonstrated the potential for TNF-α inhibitors to be utilized to improve cognition in patients with AD [81]. Interestingly, previous findings in our lab have found that dementia and chronic HFD modulate MIP-1b and TNF-α levels in the periphery of different dementia models [20, 32]. Further, additional preclinical and clinical studies have demonstrated marked increases in RANTES, TNF-α, MCP-1, eotaxin, and IL-17 [76, 82–85]. Taken together with our findings, this evidence demonstrates a detrimental effect of local cytokine accumulation on pathology and cognition. Notably, we found that levels of hippocampal MIP-1b were elevated in both male and female AD mice, and nearly undetectable in WT mice. Recent research has shown that β-amyloid plaque-associated microglia with proinflammatory phenotypes express MIP-1b [79]. Despite the elevated cytokine levels in HFD AD females, our immunohistochemistry analysis showed consistently decreased microgliosis and astrogliosis across hippocampal, cortical, hypothalamic, and thalamic regions. While the precise mechanism for this inflammatory dichotomy is not well understood, one study found that exhausted, β-amyloid -associated microglia demonstrate pro-inflammatory gene expression and that genetically engaging immune-checkpoint modulators within microglia can increase plaque phagocytosis, decrease pro-inflammatory cytokines, an improve cognitive impairment in mouse models of AD [72].This evidence suggests a potential landscape for simultaneous cytokine release with immune cell exhaustion.

Our assessment of pathology demonstrated that HFD exacerbates total brain β-amyloid beta accumulation and CAA pathology in the thalamus in females but not males. Interestingly, our investigation into four β-amyloid species within the hippocampus revealed that HFD did not specifically worsen β-amyloid burden and even decreased the burden in some species. Further, we showed that poor metabolic outcomes correlated with thalamic CAA pathology again in females but not males. Previous research in this Tg-SwDI model has shown prominent plaque accumulation in the cortical parenchyma, as well as in the vasculature in the thalamus [38]. Additionally, accumulation of activated microglia, reactive astrocytes, and complement proteins were found adjacent to the CAA pathology [37, 48, 49]. These findings suggest neuroinflammation and vasculature pathology coexist and exacerbate each other. Recent work in both clinical and animal models has shown that vascular β-amyloid interacts with monocytes to promote complement-mediated blood-brain barrier injury [86]. This indicates that inflammatory components may actually directly contribute to the spread of pathology, specifically within the vasculature. As mentioned above, in AD models that lack microglia, CAA pathology is exacerbated, suggesting that existence of some microglia is required to mitigate pathology [74]. This study is the first to demonstrate that diet-induced metabolic syndrome may minimize glial activation while exacerbating cytokine levels and CAA pathology in females, further complicating the relationship between local inflammation and CAA. Additionally, we found that our AD mice, regardless of diet, had consistently greater blood vessel density in our areas of interest (hippocampus, cortex, and thalamus) compared to WT animals. Interestingly, the most potent vascular permeability factor, vascular endothelial growth factor (VEGF), has previously been implicated in AD and CAA pathology, such that selectively inhibiting three of its receptors resulted in decreased β-amyloid deposition in vessels and decreased glial reactivity [87]. Together with our results, this suggests that HFD may enhance pathology-related blood vessel growth in a sex-specific manner. Future investigations examining changes to the cerebral microvasculature in response to diet and dementia would provide mechanistic insight into these sex-based vascular changes. Altogether, our results show that diet-induced metabolic disease worsens disease pathology by both increasing β-amyloid and blood vessel density, ultimately increasing the amount of vascular β-amyloid.

The Tg-SwDI model is a widely used model of CAA and AD which shows parenchymal β-amyloid accumulation and cognitive deficits at 3 months of age and CAA pathology at 6 months of age [36–38]. In this study, we performed behavioral testing and ended the study at ten months of age to avoid floor effects in our behavioral tests. Despite the pronounced, early pathology of this model, future studies investigating the effect of diet-induced metabolic dysfunction in older mice would be beneficial to further characterize this animal model.

In this study we utilized a 60% HFD, a well characterized intervention for causing prediabetes and obesity in rodents [19, 20, 39, 40]. This diet which equates to 60% fat from lard, is equivalent to an extremely high fat western diet. While this intervention is extreme and may not be representative of western diets in their entirety, this model of diet induced obesity has wide ranging effects on AD pathologies and is of clinical importance with the rising obesity epidemic [19, 20, 31, 34]. Other diets with distinct fat and macronutrient sources are likely to influence effects on CAA pathology differently to what we have reported here. A ketogenic diet for example, which has a high fat content from unsaturated sources and involves the consumption of limited carbohydrates (typically < 5%), has emerged as a potential therapeutic for neurodegenerative disease [88, 89]. Key components of the mediterranean diet (another diet with a high unsaturated fat content), such as extra virgin olive oil, have also been highlighted in its capacity to improve CAA pathology [90]. It is therefore likely that different fat sources will differentially impact CAA pathology, which we believe is a topic of interest in future investigations in this mouse model.

Conclusion

To our knowledge, this is the first study performed in the Tg-SwDI model to investigate interactions between diet-induced metabolic disease and sex on AD and VCID. We demonstrated that across metabolic, cognitive, and pathologic findings, AD females were consistently more vulnerable to HFD-induced deficits. Further, we showed for the first time in this model that diet aggravates cytokine levels and β-amyloid /CAA pathology in multiple ways, while attenuating glial activation, in females. Our findings add pertinent detail about metabolic and sex risk factors in VCID that is similar to previous findings in AD. This work supports the importance of understanding how women may be at higher risk of metabolic disease and comorbid dementia. Future studies further elucidating pathophysiology in this unique array of comorbidities is necessary to identify ideal therapeutic and preventative tools for dementia.

Abbreviations

AD

Alzheimer’s disease

CAA

Cerebral amyloid angiopathy

LFD

Low fat diet

HFD

High fat diet

WT

Wild type

Tg-SwDI

C57BL/6-Tg(Thy1-APPSwDutIowa)BWevn/Mmjax

GTT

Glucose tolerance test

NORT

Novel object recognition test

TPBS

Triton X-100 PBS

ROI

Region of interest

rspCTX

Retrosplenial cortex

CA1so

Stratum oriens of cornu ammonis 1

VP thal

Ventral posterior thalamus

DGpo

Dentate gyrus

VMH

Ventromedial hypothalamus

LH

Lateral hypothalamus

BLA

Basolateral amygdala

ELISA

Enzyme-linked immunosorbent assay

AUC

Area under the curve

IL

Interleukin

TNF-α

Tumor necrosis factor alpha

Authors’ contributions

KLZ, KP, and ST obtained funding for and designed the experiments. RDK established and maintained colony. CAT, RMS, RR, RDK, CAT and KBM performed the animal work. SS, CAT, RMS, KBM, BT, RR, AES, CAG, MB, EAG performed the experiments. SS, RMS, KBM, JJL, CC, and RR analyzed the data. OJG, RG, and KCM provided consultation, training, and data interpretation for cytokine experiments. SS prepared the figures. SS, RMS, and KBM prepared the manuscript. KLZ, KP, and CC edited the manuscript. All authors read and approved the final manuscript.

Funding

This work was funded by NIA U01 AG072464 (KLZ, KP, ST), NINDS R01 NS110749 (KLZ, ST), Alzheimer’s Association AARG-21-849204 (KLZ), NIA R21 AG089534 (KLZ), Bright Focus Foundation A2022001F (CAG); American Heart Association 908879 (AES).

Data availability

The datasets used and analyzed during this study are available from the corresponding author upon request.

Declarations

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Tissue or data from humans was not utilized in this study. All animal work was approved by IACUC.