Genetic and Pharmacological Targeting of Hsp72 Overexpression/Induction in an Alzheimer's Disease Mouse Model With Combined Amyloid and Tau Pathology

ABSTRACT

Heat shock protein 72 (Hsp72), a stress‐inducible member of the Hsp70 family, has been implicated in mitigating the pathogenic effects of amyloid‐beta (Aβ) and tau pathology; both key factors associated with the development and progression of Alzheimer's disease (AD). However, no in vivo study has been conducted to investigate the ability of Hsp72 to provide benefit when both Aβ and tau pathology are present concurrently. This study set out to utilize a genetic approach (transgenic overexpression) and a pharmacological validation approach (BGP‐15) to increase Hsp72 expression in mice exhibiting Aβ and tau pathology (5xFADxTg30 model). Our findings indicate that genetic Hsp72 overexpression did not significantly alter Aβ or tau burden, deficits in motor performance, or alterations to body composition in this model, but modestly improved survival rates (driven by findings in male mice) and altered anxiety‐like behavior (driven by findings in female mice). Similarly, BGP‐15 treatment was unable to rescue behavioral traits, motor deficits, body composition, or alleviate Aβ or tau accumulation, although this treatment also showed a trend towards enhancing survival (again in male mice) and altered anxiety‐like behavior (specifically in female mice). While the 5xFADxTg30 mice demonstrated seizure susceptibility compared to WT mice, BGP‐15 was unable to alleviate this trait. These findings were independent of any measurable increase in Hsp72 abundance in the brain in response to BGP‐15. Together, these findings suggest a limited impact of genetic overexpression and pharmacological targeting of Hsp72 protein abundance in this combined Aβ and tau model, although some sex‐specific differences are noted.

Significance Statement

Alzheimer's disease involves the build‐up of two harmful proteins in the brain, amyloid‐β (Aβ) and tau. As the protective protein Hsp72 can reduce Aβ and tau in laboratory studies, it has been considered a potential treatment target. This study tested whether attempting to increase Hsp72 abundance either (1) genetically or (2) with the drug BGP‐15 could improve symptoms in a mouse model that develops both Aβ and tau pathology. Despite thorough testing, neither approach reduced Aβ or tau or improved body composition, movement, behavior, or metabolism, though both genetic and pharmacological treatments tended to slightly improve survival in male mice and anxiety‐like behavior in female mice. These findings suggest that targeting Hsp72 expression levels alone is unlikely to be an effective treatment strategy for Alzheimer's disease, but the sex‐specific differences identified warrant further investigation.

We investigated targeting Hsp72 protein expression genetically and via a drug, in a model of combined amyloid and tau pathology. Neither approach reduced amyloid or tau or improved body composition or movement ability. Both treatments tended to slightly improve survival in male mice and anxiety‐like behavior in female mice.

1. Introduction

While a multifaceted condition, Alzheimer's disease (AD) is characterized by the accumulation of amyloid‐beta (Aβ) plaques and neurofibrillary tangles (NFTs) composed of hyperphosphorylated tau protein. These pathological hallmarks are central to disease progression and are associated with neuronal dysfunction and eventually neuronal loss (Hampel et al. 2021). In recent years, several drugs targeting associated pathways such as Aducanumab (Sevigny et al. 2016), Lecanemab (van Dyck et al. 2023), and Donanemab (Mintun et al. 2021) have gained FDA approval for use in AD patients. However, these treatments are far from optimal, and further research into how Aβ and tau associated pathology contribute to the development and progression of AD and identification of potential modulators of these pathways is vital.

The heat shock protein 70 (Hsp70) family of proteins constitute a group of highly conserved proteins that act as molecular chaperones to maintain cellular proteostasis by playing an important role in the assembly, folding and degradation of proteins; thus preventing aggregates of misfolded proteins (Zhang et al. 2014; Lu et al. 2014). Heat shock protein 72 (Hsp72, often referred to interchangeably in the literature as Hsp70) is a stress‐inducible member of the Hsp70 family and its expression levels have been found to be elevated in both the circulation (Marino Gammazza et al. 2022) and in the brains of patients with AD (Hamos et al. 1991; Koren 3rd et al. 2009; Perez et al. 1991; Yoo et al. 1999). Interestingly, in the study by Hamos and colleagues, the elevated Hsp72 levels observed in the brain were localized exclusively in neuritic plaques and NFTs, suggestive of a compensatory attempt by Hsp72 to mitigate the accumulation of Aβ and tau pathology observed in the brain of these patients (Hamos et al. 1991).

In vitro and pre‐clinical studies have provided experimental evidence that Hsp70/Hsp72 (referred herein to as Hsp72) interacts with both Aβ and hyperphosphorylated tau. Recombinant Hsp72 inhibits early stage Aβ aggregation in vitro (Evans et al. 2006), and Hsp72 overexpression rescues neurons from the toxic effects of intracellular Aβ accumulation (Magrane et al. 2004). Additionally, Hsp72‐induced microglial activation may serve to facilitate Aβ clearance (Kakimura et al. 2002). Transgenic mouse studies have shown promise for Hsp72 being protective in pre‐clinical models of AD. When using transgenic mice expressing a mutant form of the amyloid precursor protein (APP), reduced cognitive impairment was observed when crossed with transgenic mice overexpressing Hsp72 (Hoshino et al. 2011). Transgenic mice expressing Hsp72 also displayed lower levels of Aβ, Aβ plaque deposition, and neuronal and synaptic loss than control mice (Hoshino et al. 2011). In an alternative model, intranasally administered recombinant human Hsp72 was found to reduce Aβ plaque formation in a mouse model of AD, which coincided with a normalization of neuron density in the hippocampus and a protection of spatial memory (Bobkova et al. 2014).

In terms of tau pathology, induction of Hsp72 decreases tau levels and selectively decreases hyperphosphorylated tau species (Petrucelli et al. 2004). Inhibition of tau aggregation by Hsp72 occurs by preventing the formation of mature fibrils and oligomeric intermediates (Patterson et al. 2011). Additionally, Hsp72 is also able to recruit the co‐chaperone ubiquitin ligase CHIP, which causes the ubiquitination of tau (Jinwal et al. 2013). Together, these findings have led to the hypothesis that targeting Hsp72 expression levels through genetic means or via small molecule activation could provide an environment that promotes the removal of excess tau (Jinwal et al. 2013). However, pre‐clinical studies in this space are currently lacking. Further, given Aβ and tau pathology occur concurrently in AD, investigating the effect of Hsp72 induction in a model with both Aβ and tau elevation in tandem would be more reflective of the human pathology and provide more translational insights.

Multi‐target drug candidate BGP‐15 [O‐(3‐piperidino‐2‐hydroxy‐1‐propyl)‐nicotinic acid amidoxime dihydrochloride], a hydroximic acid derivative, is known to induce Hsp72 expression under certain pathological conditions (Chung et al. 2008; Gehrig et al. 2012; Wu et al. 2015). Aside from its Hsp72 induction properties, BGP‐15 has also been shown to inhibit Poly(ADP‐ribose) polymerase‐1 (PARP‐1) (Racz et al. 2002), alleviate ER stress (Boslem et al. 2023), reduce oxidative stress (Gonzalez et al. 2024), decrease GM3 gangliosides (Sapra et al. 2014), dampen inflammation (Boslem et al. 2023), improve mitochondrial function (Gyongyosi et al. 2023; Horvath et al. 2021), promote neuronal survival (Eroglu et al. 2014), and increase IGF1 expression (Sapra et al. 2014), all of which are pathological factors associated with AD development and/or progression and therefore potential therapeutic targets for AD in their own right (Ajoolabady et al. 2022; Chan et al. 2012; Dhapola et al. 2024; Martire et al. 2013; Kinney et al. 2018; Bhatia et al. 2022; Goel et al. 2022; Miao et al. 2024).

Consequently, we set out to investigate in a pre‐clinical model exhibiting concurrent Aβ and tau pathology (the 5xFADxTg30mouse model) (Heraud et al. 2014; Marshall et al. 2024), whether genetically elevating Hsp72 expression (through generation of a triple transgenic mouse, 5xFADxTg30xHsp72Tg), could protect against the deleterious effects of Aβ and tau. Secondly, we evaluated a pharmacological approach to test whether treatment with the drug BGP‐15 is efficacious in increasing brain Hsp72 abundance and improving the phenotype of this 5xFADxTg30 model.

2. Materials and Methods

2.1. Animal Studies

Ethics for these studies was approved by the Alfred Medical Research Education Precinct Animal Ethics Committee (AMREP AEC) and conducted according to the guidelines of the National Health and Medical Research Council of Australia for animal experimentation. Hemizygous 5xFAD and Tg30 mice were crossed to produce the double transgenic, 5xFADxTg30 as previously described (Heraud et al. 2014; Marshall et al. 2024). For the triple transgenic model, 5xFADxTg30 mice were subsequently crossed with hemizygous Hsp72Tg mice which overexpress Hsp72 in cardiac, skeletal muscle, and brain tissue (Henstridge et al. 2015; Henstridge et al. 2014; Marber et al. 1995; Marshall et al. 2018) to create the triple transgenic model, 5xFADxTg30xHsp72Tg. The 5xFADxTg30xHsp72Tg mice were compared with wildtype littermate control mice (WT) and the double transgenic 5xFADxTg30 mice. Both male and female mice were utilized in the studies to ensure representation of both sexes. All mice were on a C57BL/6J genetic background (> 10 backcrosses) and were genotyped via ear clip samples performed by Transnetyx (TN, USA) using real‐time PCR. Mice were fed a normal chow diet (NC) (14.0 MJ/kg, 75.2% kJ from carbohydrate, 4.8% from fat, 20% from protein; Specialty Feeds, Glen Forrest, Australia) across their lifespan (study endpoint was 10 months of age to meet animal welfare ethical requirements). During the experiment, mice had free access to food and water (except for in fasting periods before a glucose tolerance test or at endpoint) and were housed at 22°C ± 1°C and 40%–70% humidity on a 12 h light/dark cycle. Enrichment including nesting material such as shredded paper, gnawing objects, and plastic mouse domes were provided to each cage to prevent boredom. For the BGP‐15 studies, 5xFADxTg30 mice and littermate WT mice were administered BGP‐15 (N‐Gene Research Laboratories Inc., Hungary) in drinking water at (100 mg/kg) or water alone (vehicle control). The dose of BGP‐15 was based on a bioavailability study (Boslem et al. 2023). On the morning of tissue collection, BGP‐15 containing water bottles were removed and replaced with non‐BGP‐15 containing water bottles so that mice had a period of at least 3 h since their last opportunity to consume BGP‐15. Mice were administered sodium pentobarbitone via intra‐peritoneal injection and cardiac puncture was utilized to withdraw blood before the animals were perfused with chilled phosphate buffered saline via a perfusion pump with a catheter inserted in the left ventricle. The investigators conducting the experiments were blinded to the genotype and treatment of each mouse until endpoint of the study besides for the seizure study where treatment was not blinded. In the pharmacological experiments, mouse cages were randomized to receive vehicle or drug using the Microsoft Excel randomization tool. Animal numbers utilized and any exclusions are indicated in the respective figure legends. A figure summarizing the timeline and sequence of behavioral tests for the respective studies is provided in Figure S8 as a reference.

2.2. Behavioral Tests

A series of behavioral tests were conducted to assess both genetic and drug interventions.

2.2.1. Y‐Maze

To assess exploratory behavior and short‐term memory, time spent in a novel arm was recorded using a Y‐Maze protocol as previously described (Daglas et al. 2023; Marshall et al. 2024). Briefly, mice were placed in a two‐arm Y‐Maze and left to explore for 10 min. After a 2‐h period, mice were placed back into the Y‐Maze for a second time for 5 min with access to three arms and the time spent in the novel arm recorded. Arms were marked by a unique cue at its end to differentiate it from the other arms.

2.2.2. Novel Object Recognition

To assess non‐spatial memory in an open field, a novel object recognition test was performed as previously described (Marshall et al. 2024). Briefly, mice were placed in a field and left for a 10‐min habituation period. An object (one of two 50 mL falcon tubes wrapped in aluminum foil) was replaced with a novel object (an old soda can), and the mouse was again allowed to explore for 10 min. All objects were cleaned between each trial. The time spent exploring the novel object was recorded and analyzed.

2.2.3. Elevated Plus Maze

Anxiety‐like behavior was also assessed via an elevated plus maze (Thergarajan et al. 2022). The elevated plus maze is comprised of two open arms and two closed arms that extend from a common central platform elevated to a height of 40 cm. The closed arms provide protection via a 15 cm high wall with a passageway 4.5 cm wide. The open arms have no walls and are the same width. Mice are placed on the center square, facing an open arm and allowed to freely explore the apparatus for a 10‐min period. Trials were tracked and analyzed using video tracking software (Topscan CleverSys, USA).

2.2.4. Cheeseboard Maze

To assess spatial memory in the genetic study, we utilized a Cheeseboard maze. The Cheeseboard Maze uses the reward of food as its end goal. The maze was constructed from a circular table (94 cm diameter) that was painted white. Caps (Corning 50 mL polypropyl tube caps, 3.25 cm diameter, 1 cm depth) were painted the same color and were superglued to the table to be used as wells. Thirty‐two wells were spread evenly in a radial pattern of eight lines of four wells. The innermost well of each line was 14 cm from the center of the maze and the outermost well was 5 cm from the edge. Along each line, every well was separated from the next by 5 cm. A semitransparent cylindrical start cage (28 cm diameter) was attached to a pulley system and placed in the centre of the maze. High contrast spatial cues (A3 size) were attached to the interior of the faraday cage (“North,” “South,” “East,” “West”). Baseline weights of mice were obtained by taking the average weight over 3 days before commencement of food restriction. Food restriction was introduced gradually over 3 days prior to testing. Mice were fed approximately 80% of free feeding amount (determined in a preliminary test on a separate cohort of mice) to obtain total weight loss of 10%. The weights of the mice were recorded every day during the Cheeseboard task and individually weighed food pellets supplied to maintain their weight at 90%. Mice were habituated to strawberry milk as a food reward from 3 days prior to the testing, during the acclimatization days. During acclimatization, mice were given 5 min to roam the Cheeseboard freely, then were given the opportunity to lick a cotton tip which had been dipped in strawberry milk. At this time the Maze was not baited. The four corners of the maze were brushed with strawberry milk prior to testing to minimize the effect of odor as a cue. The maze was cleaned with diluted ethanol (10%) after each trial. During the testing phase, two 60 s trials with an inter trial time of 20 min were conducted on each day for 6 days. One well was baited with a cap full of strawberry milk as a reward. The location of the baited well was conserved across trials and days for each mouse. On the 7th day, the probe trial occurred where no well was baited, and the duration spent in the previously baited well was recorded.

2.2.5. Morris Water Maze

To assess spatial memory in the pharmacological study, a Morris Water Maze was conducted over a 7‐day training period as previously described (Daglas et al. 2023; Marshall et al. 2024). Briefly, mice were provided 60 s to find a hidden platform, and if they were unable to do so in this time, was gently guided onto the platform. The mouse remained on the platform for 15 s before being removed to its cage. Over 6 days, the animals were given four trials per day, each from one of four starting positions, and allowed 120 s to find the submerged platform. On Day 7, the platform was removed and a probe trial conducted to assess the amount of time spent in the correct quadrant of the maze as an indicator of recall memory.

2.3. Motor Function and Movement Testing

2.3.1. Rotarod

Motor function was assessed via Rotarod testing. Briefly, mice underwent a training session that consisted of three 5‐min trials on a spinning rod with a constant speed of 4 rpm. The test consisted of three trials with the rotating rod progressively accelerating from 4 to 40 rpm over 300 s. Latency to fall off was recorded. Any animals still on the rod at the end of the test were removed and a latency of 300 s was recorded.

2.3.2. Hangwire

Forelimb strength and muscular endurance were assessed via a hangwire test. A wire rod was suspended horizontally between two vertical beams, ~45 cm from a padded surface. Mice were encouraged to grasp the rod with their forepaws. Once the forepaws were firmly grasping the rod, mice were allowed to hang freely without the help of their hind legs. Latency to fall was recorded (three times) with 5 min rest between each trial.

2.3.3. Gait Analysis (DigiGait)

To assess gait, a DigiGait treadmill and analysis system was utilized (Mouse Specifics Inc., USA). DigiGait analysis provides quantitative data for gait parameters for each limb independently (left and right forelimbs and hindlimbs). Each mouse was placed in a perspex chamber fitted onto a transparent treadmill and allowed to acclimatize. The treadmill was then run at 10 cm/s for a short period to capture video footage of a period of uniform walking of between 3 and 8 s. Images were collected through the transparent treadmill at 157 frames per second and stored in audio video interleaved format. Acceptable footage was achieved when all four paws remained within frame for the duration and the mouse did not hit the back bumper. Mice that were unable to walk at 10 cm/s without hitting the back bumper for three attempts were excluded from this analysis. Data taken from the video analysis were analyzed using the data analysis software.

2.3.4. Activity

24‐h physical activity levels were measured and divided into 12‐h light and 12‐h dark periods. For the genetic study, a Promethion metabolic cage system (Sable Systems International, USA) was utilized and for the BGP‐15 study, a Comprehensive Laboratory Animal Monitoring System (CLAMS, Columbus Instruments, USA) was used to measure total movement (beam breaks) over 48 h. The first 24 h served as an acclimatization period and the 24–48 h period served as the period where the data were analyzed (Lancaster and Henstridge 2018).

2.4. Seizure/Epileptogenesis Susceptibility

2.4.1. Electrode Implantation

At 6 months of age, to facilitate amygdala kindling, mice underwent surgical implantation of recording and stimulating electrodes, as previously described with minor modifications (Tan et al. 2012; Li et al. 2018). Briefly, mice were anesthetized (Isoflurane) and a midline incision was made through the skin to expose the scalp. 1 mm holes were drilled through the skull, and using stereotaxic guidance, a bipolar electrode (P1 Technologies) was implanted into the left amygdala (Bregma coordinates: ML: −3.10 mm; AP: −0.8 mm; DV: −4.2 mm), and a unipolar electrode was implanted in the right dorsal hippocampus (ML: 1.3 mm; AP: −1.8 mm; DV: −1.3 mm). Two additional screw electrodes were implanted over the cerebellum (ground and reference). Electrodes were connected to an electrical socket and the exposed area of the skull and wires were covered in dental cement (Vertex).

2.4.2. Kindling

The electrode assembly was connected to an Accupulser Pulse Generator/Stimulator (A310, World Precision Instruments) using cables to stimulate the brain, and electrophysiological activity visualized and recorded using LabChart software (AD Instruments). Electrical stimulations were delivered using a 200 μA stimulus of 1 ms biphasic square wave pulses (frequency 60 Hz were graded for behavioral severity using the Racine scale; Chan et al. 2015; Racine 1972), and electrographic duration measured from the EEG. The process was repeated 5 days a week for 3 weeks (total 15 stimulations). As the number of stimulations increased, mice experienced seizures which increased in severity and duration, a process known as kindling. During each rest period, mice were continuously monitored for any spontaneous seizures which may occur or other epileptiform abnormalities in the EEG. If mice experienced five consecutive convulsive seizures as a result of the stimulation, they have reached the end‐stage of kindling and receive no more stimulations. Alternatively, once the 15 stimulations had occurred the procedure was complete.

2.4.3. Video‐EEG Monitoring

During the experimental period, mice were video recorded to document any spontaneously occurring seizures (as opposed to those triggered by electrical stimulation) which are determined by offline review of the recordings. This procedure was conducted for 5 consecutive weeks—1 week prior to the onset of kindling, then continued for the 3‐week period of kindling whenever the mice were not being stimulated, followed by another week period occurring after completion of kindling.

2.5. Metabolic Testing

Body weight was measured using laboratory scales (Mettler Toledo, Switzerland). Body composition (total fat mass and total lean mass) was measured using a 4‐in‐1 EchoMRI machine (EchoMRI, USA) as previously described (Lancaster and Henstridge 2018). Oral glucose tolerance tests (oGTT) were performed on mice that were fasted for 5–6 h at 5 months of age. Mice received an oral gavage of 2 g glucose/kg lean body mass and blood glucose concentration was measured via a glucometer (AccuCheck, Roche Diabetes Care, Australia) at the indicated times on the blood collected from the tail.

2.6. Laboratory Analysis

2.6.1. Western Blotting

For blots in Figure 1, samples were homogenized in protein lysis buffer (50 mM HEPES, 2 mM EDTA, 50 mM β‐glycerophosphate, 1 mM DL‐Dithiothreitol (DTT), 1 mM Na₃VO₄, 1.0% Triton X‐100, 10% glycerol, 1 mM NaF, 1 mM PMSF, 0.05% protease inhibitor cocktail (P8340; Sigma, USA) and 0.05% phosphate inhibitor cocktail (P5726, Sigma, USA)). The protein concentration was determined using a Pierce BCA Protein Assay Kit (Life Technologies, USA) and equal amounts of protein solubilized in 4X Laemmeli's buffer (40% glycerol, 8.2% SDS, 50% 0.5 M Tris–HCL (pH 6.8), 0.5 mL 1% Bromophenol blue, 10% H₂O and 20 mM DTT). Samples were run on precast 12% bis tris gel (Nupage, Life Technologies, USA) and proteins transferred onto a membrane using the iBlot/PVDF stack (Life Technologies, USA). Membranes were blocked in TBST with 5% skim milk powder and incubated with appropriate primary antibody overnight at 4°C. Membranes were incubated at room temperature for 30 min with LiCor fluorescent secondary antibody in TBST +0.01% SDS and visualized using a LI‐COR Odyssey Scanner (LICORbio, USA). Blots in Figures 3, 4 and 7 were obtained as previously described (Marshall et al. 2024). Protein was obtained using RIPA buffer (Thermo Scientific, USA) with protease and phosphatase inhibitors (Thermo Scientific, USA). A BCA protein assay (Thermo Scientific, USA) was performed to determine protein. Gel electrophoresis was performed in a Mini‐PROTEAN Tetra Cell (Bio‐Rad, Australia) using 4%–15% gradient polyacrylamide gels (Bio‐Rad, Gladesville, Australia). Transfer was performed in the Trans‐Blot Turbo Semi‐Dry Transfer System (Bio‐Rad, Australia) according to manufacturer instructions (except when probing for MOAB‐2 which was conducted via wet transfer). Transfer quality was visually inspected via Ponceau staining (Sigma‐Aldrich, Australia). Bands were visualized electrochemically using the SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific, Rockford, IL, USA) on a Fujifilm LAS‐3000 Imager (Fuji Photo Film Co. Ltd., Japan). Band intensities for western blot analysis were quantified and normalized in ImageJ software (National Institutes of Health, USA) and normalized to the internal loading control (GAPDH). Each GAPDH band was acquired from the same membrane as the protein of interest. Original source western blots are provided in Figure S10 with indicated molecular weight markers. A list of the primary antibodies used in this analysis is provided below in Table 1.

FIGURE 1.

Validation of model, body composition analysis and survival with Hsp72 transgenic overexpression. Representative western blots, demonstrating overexpression of Hsp72 in the 5xFADxTg30 mouse model. Brain was harvested at 10 months and dissected into (A) hippocampus and (B) cerebellum. Quantification performed using GAPDH as loading control ####p < 0.0001 WT vs. 5xFADxTg30xHsp72Tg, ^^^^p < 0.0001 5xFADxTg30 vs. 5xFADxTg30xHsp72Tg, one‐way ANOVA with Tukey post hoc test, n = 5–6 per group, F values and degrees of freedom, F (DFn, DFd): (A) hippocampus, genotype, F (2, 14) = 121.3; (B) cerebellum, genotype, F (2, 14) = 58.28. Source blots provided in Figure S10, one WT sample removed from analysis as an outlier identified in GraphPad Prism. Body composition analysis: (C) body weight, (D) lean mass, (E) fat mass, (F) body fat percentage (%) *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 WT vs. 5xFADxTg30, #p < 0.05, ##p < 0.01, ###p < 0.001, ####p < 0.0001 WT vs. 5xFADxTg30xHsp72Tg, ^p < 0.05 5xFADxTg30 vs. 5xFADxTg30xHsp72Tg, Mixed‐effects analysis, n = 24–28 per group. F values and degrees of freedom, F (DFn, DFd): (C) body weight, time, F (2.248, 163.2) = 70.44, genotype, F (2, 76) = 32.98, interaction, F (16, 581) = 37.52; (D) lean mass, time, F (3.343, 242.8) = 60.31, genotype, F (2, 76) = 17.92, interaction, F (16, 581) = 14.74; (E) fat mass, time, F (1.684, 122.3) = 71.09, genotype, F (2, 76) =29.8, interaction, F (16, 581) = 28.84; (F) body fat percentage, time, F (3.29, 239) = 67.52, genotype, F (2, 76) = 20.13, interaction, F (16, 581) = 17.66. (G) Brain weight (n = 12–16) and (H) Tibialis Anterior (n = 9–12) weight, one‐way ANOVA with Tukey post hoc, *p < 0.05, ##p < 0.01, F values and degrees of freedom, F (DFn, DFd): (G) brain, genotype, F (2, 40) = 3.359, p = 0.045; (H) tibialis anterior, genotype, F (2, 28) = 7.818, p = 0.002. (I) Kaplan–Meier survival curve, Log‐rank (Mantel‐Cox) test, n = 24–28 per group. Male and female mice combined. Graphs indicate mean ± SD.

FIGURE 3.

Physical activity, motor control, muscular endurance and glucose tolerance with Hsp72 transgenic overexpression. (A) Physical activity levels over a 24‐h period as measured by beam breaks divided into light and dark cycles, one‐way ANOVA, n = 11–13, F values and degrees of freedom, F (DFn, DFd): Genotype, F (2, 3) = 0.045, p = 0.956. (B) Latency to fall off rotarod between 7 and 10 months of age, mixed effects analysis with Tukey's post hoc test, n = 21–26. F values and degrees of freedom, F (DFn, DFd): time × genotype F (6, 205) = 3.808, p = 0.0013, time F (2.703, 184.7) = 7.177, p = 0.0002, genotype F (2, 74) = 112.1. (C) Latency to fall from hangwire at 8 and 10 months of age, mixed effects analysis with Tukey's post hoc test, n = 20–25. F values and degrees of freedom, F (DFn, DFd): time × genotype F (2, 66) = 0.052, p = 0.949, time F (1, 66) = 8.089, p = 0.006, genotype F (2, 72) = 58,77. (D) Glucose tolerance excursion curves at 5 months of age, two‐way RM ANOVA with Tukey's post hoc testing and (E) corresponding incremental area under the curve (AUC) and one‐way ANOVA, *p < 0.05, **p < 0.01, WT vs. 5xFADxTg30, #p < 0.05, WT vs. 5xFADxTg30xHsp72Tg, n = 24–26, F values and degrees of freedom, F (DFn, DFd): (D) excursion curves, time × genotype, F (12, 432) = 1.419, p = 0.154 time, F (2.858, 205.8) = 300.6, genotype, F (2, 72) = 7.35, p = 0.0012; (E) AUC, genotype, F (2, 72) = 2.85, p = 0.064. (F) Representative western blot images (from female samples) of hippocampus tissue from endpoint at 10 months of age with corresponding quantification for (G) phosphorylated Tau (H) Tau and (I) Aβ *p < 0.05, **p < 0.01, ****p < 0.0001 WT vs. 5xFADxTg30, ##p < 0.01, ###p < 0.001, ####p < 0.0001 WT vs. 5xFADxTg30xHsp72Tg, one‐way ANOVA with Tukey post hoc test, n = 9–11, F values and degrees of freedom, F (DFn, DFd): (G) phosphorylated Tau, genotype, F (2, 27) = 13.42; (H) Tau, genotype, F (2, 27) = 16.99 (I) Aβ, genotype, F (2, 27) = 7.78, p = 0.021. Male and female mice combined. Graphs indicate mean ± SD.

FIGURE 4.

Characterization of model, body composition analysis and survival with BGP‐15 treatment. (A) Mass spectrometry analysis of BGP‐15 concentration in the plasma and brain of male mice at endpoint, n = 6 per group. Representative western blots (from male samples) of Hsp72 following BGP‐15 treatment. Brain was harvested at 10 months and dissected into (B) hippocampus and (C) cerebellum. Quantification performed using GAPDH as loading control, two‐way ANOVA, n = 6 per group, F values and degrees of freedom, F (DFn, DFd): (B) hippocampus, interaction F (1, 20) = 2.216, p = 0.152, genotype F (1, 20) = 0.841, p = 0.3701, treatment, F (1, 20) = 0.635, p = 0.435; (C) cerebellum, interaction F (1, 20) = 0.998 p = 0.330, genotype F (1, 20) = 0.019, p = 0.892, treatment, F (1, 20) = 0.029, p = 0.596. Body composition analysis: (D) body weight, (E) lean mass, (F) fat mass, and (G) body fat percentage (%) *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 WT CON vs. 5xFADxTg30 CON, ##p < 0.01, ###p < 0.001, ####p < 0.0001 WT CON vs. 5xFADxTg30 BGP‐15, ^p < 0.05 ^^p < 0.01, ^^^p < 0.001, ^^^^p < 0.0001 WT BGP‐15 vs. 5xFADxTg30 CON, &p < 0.05, &&p < 0.01, &&&p < 0.001, &&&&p < 0.0001, WT BGP‐15 vs. 5xFAD BGP‐15, mixed‐effects analysis, n = 24–30 per group. F values and degrees of freedom, F (DFn, DFd): (D) body weight, time, F (3.158, 303.6) = 285.5, genotype, F (3, 101) = 35.59, interaction, F (9.475, 303.6) = 56.15; (E) lean mass, time, F (2.662, 261.6) = 127.2, genotype, F (3, 101) = 15.25, interaction, F (7.987, 261.6) = 15.56; (F) fat mass, time, F (1.999, 196.4) = 256.2, genotype, F (3, 101) = 24.13, interaction, F (5.996, 196.4) = 27.24; (G) body fat percentage, time, F (2.735, 268.7) = 274, genotype, F (3, 101) = 13.87, interaction, F (8.204, 268.7) = 16.74. (H) Tibialis Anterior weight (n = 9–12), two‐way ANOVA with Tukey post hoc test, F values and degrees of freedom, F (DFn, DFd): interaction F (1, 40) = 2.139, p = 0.151, treatment, F (1, 40) = 2.78e‐005, p 0.996, genotype F (1, 40) = 144. (I) Kaplan–Meier survival curve, **p < 0.01 WT CON vs. 5xFADxTg30 CON, ^^p < 0.01 WT BGP‐15 vs. 5xFAD CON, Log‐rank (Mantel‐Cox) test, n = 24–40 per group. Male and female mice combined. Graphs indicate mean ± SD.

FIGURE 7.

Glucose tolerance and seizure susceptibility with BGP‐15 treatment. (A) Glucose tolerance excursions curves at 5 months of age with two‐way RM ANOVA with Tukey's post hoc testing, &p ≤ 0.05, &&p ≤ 0.01 WT BGP‐15 vs. 5xFAD BGP‐15, ^p < 0.05 ^^p < 0.01, WT BGP‐15 vs. 5xFADxTg30 CON, F values and degrees of freedom, F (DFn, DFd): time × genotype F (18, 582) = 1.627, time, F (2.688, 260.8) = 285.8, genotype F (3, 97) = 4.758. (B) Corresponding incremental area under the curve (AUC), one‐way ANOVA, * = main effect with WT vs. 5xFADxTg30 groups, n = 24–28, F values and degrees of freedom, F (DFn, DFd): interaction F (1, 97) = 0.5529, treatment, F (1, 97) = 0.457, genotype F (1, 97) = 5.1954. (C) Seizure susceptibility indicated by stimulations to first Class V seizure, (D) average seizure length and (E) average seizure class at 7 months of age, one‐way ANOVA, *p ≤ 0.05, **p ≤ 0.01 WT‐CON vs. 5xFADxTg30‐CON, #p ≤ 0.05, ##p ≤ 0.01 WT‐CON vs. 5xFADxTg30‐BGP‐15, n = 4–7, F values and degrees of freedom, F (DFn, DFd): (C) stimulations, genotype, F (2, 14) = 14.72, (D) length, genotype, F (2, 14) = 22.13, (E) class, genotype, F (2, 14) = 34.00. (F) Representative western blot images (from female samples) of hippocampus tissue from endpoint at 10 months of age with corresponding quantification for (G) phosphorylated Tau (H) Tau and (I) Aβ ***p < 0.001, ****p < 0.0001 WT‐CON vs. 5xFADxTg30‐CON, ##p < 0.01, ####p < 0.0001 WT vs. 5xFADxTg30‐BGP‐15, one‐way ANOVA with Tukey post hoc test, n = 15–16, F values and degrees of freedom, F (DFn, DFd): (G) phosphorylated Tau, genotype, F (2, 43) = 9.716, p = 0.0003; (H) Tau, genotype, F (2, 43) = 43.02, (I) Aβ, genotype, F (2, 43) = 15.58. One WT sample removed from analysis as an outlier identified in GraphPad Prism. Male and female mice combined. Graphs indicate mean ± SD.

TABLE 1.

Antibodies used in the study.

Antibody name	Dilution	Vendor and catalog #, RRID #	Clonality	Host
Anti‐GAPDH	1:2000	Cell Signaling Technology, cat #5174, RRID:AB_10622025	Recombinant monoclonal	Rabbit
Anti‐Tau	1:1000	Dako, cat #A0024, RRID:AB_10013724	Polyclonal	Rabbit
Anti‐Phospho‐Tau (Ser202, Thr205)	1:1000	Thermo Fisher Scientific, cat #MN1020, RRID:AB_223647	Monoclonal	Mouse
Anti‐Beta Amyloid (MOAB‐2)	1:2000	Novus Biologicals, cat #NBP2‐13075, RRID:AB_3260692	Monoclonal	Mouse
Anti‐Hsp72	1:1000	Enzo Life Sciences (formerly StressGen), cat #ADI‐SPA‐810, RRID:AB_2039260	Monoclonal	Mouse
IRDye 800CW goat anti‐mouse IgG	1:10,000	LI‐COR Biosciences, cat #926‐32210, RRID:AB_621842	Polyclonal	Goat
IRDye 680RD goat anti‐rabbit IgG	1:10,000	LI‐COR Biosciences, cat #926‐68071, RRID:AB_10956166	Polyclonal	Goat
Anti‐rabbit IgG	1:5000	Cell Signaling Technology, cat #7074, RRID:AB_2099233	Polyclonal	Goat
Anti‐Mouse IgG, HRP‐linked	1:5000	Cell Signaling Technology, cat #7076, RRID:AB_330924	Polyclonal	Human

2.6.2. BGP‐15 Analysis

Liquid chromatography–mass spectrometry (LC–MS) analysis was performed on a Shimadzu 8030 triple quadrupole instrument coupled with a Shimadzu Nexera UHPLC. MS analysis was conducted in positive mode electrospray ionization and quantitation of the analytes performed in MRM mode. The column employed was a Phenomenex Kinetex EVO C18 column (2.6 μm particle size, 50 × 2.1 mm i.d.) equipped with a Phenomenex Security Guard column with C18 packing material. The column was maintained at 40°C. The mobile phase consisted of 10 mM ammonium acetate in water (pH adjusted to 9.5) (Mobile Phase A) and 100% acetonitrile (Mobile Phase B). Separations were conducted using a flow rate of 0.4 mL/min and an injection volume of 3 μL. Samples were maintained in the autosampler at a temperature of 15°C. Extraction of the BGP‐15 from plasma was conducted using protein precipitation with acetonitrile (3‐fold volume ratio). Plasma standards were freshly prepared on the day of analysis and a calibration curve established. Solution standards were diluted from the stock solution (5 mg/mL in methanol) with 50% acetonitrile in water. Plasma standards were prepared by spiking blank mouse plasma (20 μL) with solution standards (4 μL) before protein precipitation by the addition of 100% acetonitrile (containing internal standard; diazepam, 0.025 μg/mL). Standards were then vortexed and centrifuged (10,000 rpm, 3 min). The supernatant was subsequently separated, transferred to a HPLC vial and injected for LC–MS analysis. Plasma samples were similarly prepared except blank 50% acetonitrile in water (4 μL) was added instead of the solution standard. Brain tissue was homogenized using a MP biomedical FastPrep 24G tissue homogenizer at 6.0 m/s for 40 s and 40 μL of the homogenate aliquoted into an eppendorf tube. This was spiked with BGP‐15 spiking solution and proteins precipitated by the addition of 120 μL cold acetonitrile containing internal standard (diazepam; 0.5 μg/mL). The mixture was vortexed for 3 × 10 s followed by centrifugation at 10,000 rpm for 5 min at 21°C. An aliquot of the supernatant (70 μL) was then transferred to polypropylene HPLC vials for LCMS analysis.

2.7. Statistical Analysis

Data was assessed with GraphPad PRISM v10 (San Diego, California, USA, RRID SCR_002798). One‐way ANOVA was used when comparing a single factor across three groups (e.g., genetic study outcomes), while two‐way ANOVA was used when assessing the effects of two independent variables (e.g., genotype and drug treatment in the pharmacological studies). Repeated measures ANOVA was used for longitudinal outcomes measured over multiple time points. The 5xFADxTg30 model exhibits premature mortality during disease progression, resulting in missing observations at later time points. As repeated measures ANOVA requires complete datasets, GraphPad Prism automatically fitted a mixed‐effects model using Restricted Maximum Likelihood (REML) estimation when missing values were present. This approach permits inclusion of all available measurements from each animal rather than excluding animals with incomplete longitudinal datasets. For longitudinal analysis, mouse ID was treated as the repeated measures subject factor to account for within‐animal correlations over time. In the genetic studies, genotype and time were included as fixed effects. In the pharmacological studies, genotype, treatment and time were included as fixed effects. Raw, untransformed data were used for all analyses. Group sizes were generally balanced across experimental conditions, and no substantial assumption violations were identified that required data transformation or alternative statistical approaches. Tukey's multiple comparison post hoc testing was conducted following significant main or interaction effects from the ANOVA or REML analysis. Kaplan–Meier analysis was conducted for the survival analysis. Power analysis was based on data in a previous description of this model (Heraud et al. 2014), and calculations were performed using α = 0.05, power = 0.80 and assumed a partial rescue of the phenotype by the interventions. This analysis suggested the requirement of ~23 mice per group. No interim analyses or stopping rules were used, and animals were followed (and data utilized) until study completion or humane endpoint criteria specified in the animal ethics approval. Some secondary assessments and analysis were conducted on sub‐cohorts of mice due to study timing, feasibility and resource availability. All data points are biological replicates (individual mice or tissue samples from individual mice) not technical replicates. Data was excluded only when technical issues prevented collection of a specific outcome measure or if outliers were detected in GraphPad Prism via the Robust egression and Outlier removal (ROUT) function. Any statistical removal of data points is listed in the respective figure legend. Statistics are indicated in each of the figure legends with p < 0.05 considered significant. In the Hsp72Tg genetic studies, “*” is indicating a WT vs. 5xFADxTg30 effect, “#” is indicating a WT vs. 5xFADxTg30xHsp72Tg effect, “^” is indicating a 5xFADxTg30 vs. 5xFADxTg30xHsp72Tg effect and “&” is indicating a main effect for time. For the BGP‐15 drug studies, “*” indicates a WT CON vs. 5xFADxTg30 CON effect, “#” indicates a WT CON vs. 5xFADxTg30 BGP‐15 effect, “^” indicates a WT BGP‐15 vs. 5xFADxTg30 CON effect, “&” indicates a WT BGP‐15 vs. 5xFADxTg30 BGP‐15 effect, “$”; indicates a 5xFADxTg30‐CON vs. 5xFADxTg30 BGP‐15 effect and “+” indicates a WT‐CON vs. WT‐BGP‐15 effect. Data is displayed graphically as scatterplots with mean ± SD. In the graphs, the significant symbols above specific timepoints refer to post hoc analysis results at that timepoint. In the main figures of the manuscript, male and female mouse data is combined to increase statistical power while the Supporting Information figures display the disaggregated data separated by sex for identification of any sex‐specific effect. Given the number of measures and number of comparisons between groups, a summary table of all results can be found in Figure S9 for clarity.

3. Results

3.1. Genetic Model

We first assessed the level of overexpression of Hsp72 to validate the model. Western blotting of the hippocampus (Figure 1A) and cerebellum (Figure 1B) demonstrated a significant upregulation of Hsp72 protein expression in both regions in 5xFADxTg30xHsp72Tg mice compared to 5xFADxTg30 and WT mice. Body composition was assessed monthly across the study via EchoMRI to determine whether Hsp72 overexpression could rescue the reduced body weight, lean mass, and fat mass previously observed in the 5xFADxTg30 mice (Marshall et al. 2024). Body weight and lean mass were significantly different between WT mice and 5xFADxTg30 mice from 2 months of age, with 5xFADxTg30xHsp72Tg not differing from 5xFADxTg30 mice (Figure 1C,D). Fat mass and body fat percentage were significantly different between WT mice and 5xFADxTg30 mice from 4 and 5 months of age respectively (Figure 1E,F). Hsp72 overexpression was able to partially rescue the fat loss phenotype in the 5xFADxTg30 mice at one specific time point (7 months of age) (Figure 1E,F). There was no sex effect for the body composition measures (Figure S1A–D). Assessment of brain weight identified a small but significant decrease in the 5xFADxTg30 mice compared to WT but not in the 5xFADxTg30xHsp72Tg mice (Figure 1G). No particular sex drove this phenotype (Figure S1E). Muscle weight was significantly reduced in both 5xFADxTg30 and 5xFADxTg30xHsp72Tg mice (Figure 1H), a finding which was driven more by the female mice than the male mice (Figure S1F). Survival curve analysis demonstrated that 89% of 5xFADxTg30xHsp72Tg mice remained alive at endpoint of the study compared to 79% of 5xFADxTg30 mice. Statistical analysis revealed an overall p value of 0.06 between groups; however, comparison of 5xFADxTg30 and 5xFADxTg30xHsp72Tg groups to WT mice in isolation demonstrated that 5xFADxTg30 survival was significantly reduced compared to WT mice, whereas 5xFADxTg30xHsp72Tg mice did not statistically differ from WT mice, suggestive of a modest protective effect of Hsp72 overexpression on survival (Figure 1I). Although not statistically significant, this modest protective effect was almost entirely driven by Hsp72‐induced protection in the male mice, where separation of the survival curves is observed (Figure S1G).

A battery of behavior tests was conducted to assess the effect of Hsp72 overexpression in this model. Spatial learning and memory were assessed via the Cheeseboard maze (male mice only). The latency to find the baited well improved in all three groups over the 6 days of trials (Figure 2A). While WT mice tended to perform better than both 5xFADxTg30 and 5xFADxTg30xHsp72Tg groups, no statistically significant differences were noted, and no difference was observed between 5xFADxTg30 and 5xFADxTg30xHsp72Tg groups on any day (Figure 2A,B). When the baited well was removed, both WT and 5xFADxTg30xHsp72Tg groups tended to spend more time in the baited well areas, but this did not reach statistical significance (Figure 2C). Memory function was assessed via Y‐Maze performance which did not demonstrate any significant differences between genotypes at 7 or 9 months of age (Figure 2D,E). Separating for sex, female 5xFADxTg30xHsp72Tg mice spent less time in the novel arm than WT mice at the 7‐month timepoint only (Figure S2A,B). Similarly, no difference was observed between genotypes with a novel object recognition maze (Figure 2F) with no sex effect apparent (Figure S2C). At 9 months of age, anxiety‐like behavior was assessed using an elevated plus maze. 5xFADxTg30 mice displayed more risk‐taking behavior as they spent significantly longer in the open arm and less time in the closed arm than the WT mice (Figure 2G). This behavior was partially rescued by Hsp72 overexpression (p = 0.055 and p = 0.057 5xFADxTg30 vs. 5xFADxTg30xHsp72Tg, for open and closed arm respectively Figure 2G). This protective effect was driven by female 5xFADxTg30xHsp72Tg mice with significant differences observed in the female cohorts when data was divided by sex (Figure S2D).

FIGURE 2.

Behavioral testing with Hsp72 transgenic overexpression. (A) Latency to baited well and (B) corresponding area under the curve over 6 days for Cheeseboard maze testing, &&&p < 0.05, main effect of time, two‐way RM ANOVA, (A) latency, F values and degrees of freedom, F (DFn, DFd): time × genotype, F (10, 55) = 0.4123, p = 0.9350; time, F (3.044, 33.49) = 8.687, p = 0.0002, genotype, F (2, 11) = 3.145, p = 0.083: (B) area under curve, genotype, F (2, 11) = 3.194, p = 0.081. (C) On the final day of testing the reward was removed and the duration of time spent in the previously baited well recorded, n = 4–5 (males only conducted at 8 months of age), F values and degrees of freedom, F (DFn, DFd): Genotype, F (2, 11) = 0.215, p = 0.810. Y‐Maze testing with time spent in the novel arm at (D) 7 months and (E) 9 months of age, one‐way ANOVA, n = 23–26, F values and degrees of freedom, F (DFn, DFd): (D) 7 months, genotype, F (2, 72) = 2.548, p = 0.085, (E) 9 months, genotype, F (2, 70) = 0.259, p = 0.772. (F) Number of bouts of investigative behavior of the novel object at 9 months of age, one‐way ANOVA, n = 24–25, F values and degrees of freedom, F (DFn, DFd), genotype F (2, 71) = 0.283, p = 0.7542. (G) Anxiety‐like behavior assessed at 9 months of age via an elevated plus maze, duration of overall time spent in each arm. #p < 0.05, WT vs. 5xFADxTg30xHsp72Tg, ****p < 0.0001 WT vs. 5xFADxTg30, two‐way ANOVA with Tukey's post hoc testing, n = 23–25, F values and degrees of freedom, F (DFn, DFd): interaction, F (2, 138) = 24.55, arm F (1, 138) = 23.78, genotype, F (2, 138) = 0.0002, p = 0.999. Male and female mice combined. Graphs indicate mean ± SD.

Previously it was demonstrated that the 5xFADxTg30 mice exhibit a decreased motor performance (Heraud et al. 2014; Marshall et al. 2024). As Hsp72 overexpression has been shown to improve running performance (Henstridge et al. 2014) and preserve muscle function and slow progression in a muscular dystrophy model (Gehrig et al. 2012), we assessed the physical capabilities of the mice. General physical activity measured over 24‐h was not different between the genotypes (Figure 3A) with no sex‐specific effect noted (Figure S3A). However, when mice undertook a rotarod test, the WT mice demonstrated a steady performance from 7 to 10 months of age, whereas the 5xFADxTg30 mice had reduced performance from 7 months of age which worsened with increasing age (Figure 3B). Hsp72 overexpression was not able to reverse this defect in motor control. These findings were reproduced in both male and female mice (Figure S3B). Next, we assessed muscular strength and endurance through use of a hangwire test. Hangwire performance was markedly reduced in the 5xFADxTg30 mice of which Hsp72 overexpression could not protect against (Figure 3C), with no difference between male and female cohorts (Figure S3C). Metabolic control was also assessed given previous observations of altered metabolic control in both the 5xFADxTg30 model (Marshall et al. 2024) and with Hsp72 overexpression (Chung et al. 2008; Henstridge et al. 2014). 5xFADxTg30 mice had reduced fasting glucose levels and glucose excursions following a glucose bolus compared to WT mice (Figure 3D), driven by more obvious differences in the female mice (Figure S3D). Again, no difference was observed between 5xFADxTg30 and 5xFADxTg30xHsp72Tg mice (Figure 3D,E, Figure S3D,E).

At endpoint, to assess whether the Hsp72 overexpression was able to reduce the burden of tau and amyloid in the brain, we performed western blotting on tissue derived from the hippocampus. 5xFADxTg30 mice exhibited high levels of phosphorylated and total tau and had accumulated Aβ (Figure 3F–I). 5xFADxTg30xHsp72Tg mice exhibited the same levels of these proteins, indicating Hsp72 overexpression was not able to limit the burden of these brain aggregates in this model (Figure 3F–I). These effects were observed in both male and female mice to a similar degree (Figure S3F).

3.2. Pharmacological Model

Next, we attempted a pharmacological validation study of the genetic findings by treating mice with the Hsp72 co‐inducing agent BGP‐15. At endpoint, plasma and brain tissue were assessed for BGP‐15 levels. Plasma and brain concentration of BGP‐15 was detected in both WT and 5xFADxTg30 mice but was elevated to a greater degree in the WT mice compared to the 5xFADxTg30 mice. However, the levels were not statistically significant between the groups (Figure 4A). To assess if long term administration of BGP‐15 and subsequent brain exposure to the BGP‐15 levels demonstrated in Figure 4A resulted in an increase in Hsp72 protein abundances in the brain, both the hippocampus and cerebellum were probed. Neither brain region displayed an increase in Hsp72 (Figure 4B,C) in either sex (Figure S4A,B), suggesting that chronic exposure to BGP‐15 at this dose did not shift the expression profile of this protein. Consequently, any findings with this intervention would be considered independent of an increase in brain Hsp72 expression at 10 months of age. Body composition analysis revealed no alterations to body weight, lean mass, fat mass or fat mass percentage with BGP‐15 treatment in either WT or 5xFADxTg30 groups (Figure 4D–G), while individual muscle weight was also unchanged by BGP‐15 treatment (Figure 4H). No sex difference for body composition analysis or muscle weight was observed for BGP‐15 treatment (Figure S4C–G). Survival curve analysis demonstrated that 88% of 5xFADxTg30 mice treated with BGP‐15 remained alive at endpoint of the study compared to 70% of 5xFADxTg30 mice treated with vehicle control. Reflecting this, 5xFADxTg30 survival was significantly reduced compared to WT mice, whereas 5xFADxTg30 mice treated with BGP‐15 did not statistically differ from WT mice, suggestive of a modest protective effect of BGP‐15 treatment on survival (Figure 4I). Like in the genetic model, this protective effect was driven by the protection offered to the male mice (Figure S4H).

Behavior tests were conducted to assess the effect of BGP‐15 treatment in this model. In this study, spatial memory was assessed via Morris Water Maze. From Day 1, differences were noted between WT and 5xFADxTg30 groups irrespective of BGP‐15 treatment. Although both groups improved their latency and a learning curve was observed over the 6 days, the difference between WT and 5xFADxTg30 cohorts remained (Figure 5A,B). This finding was similar between male and female mice (Figure S5A,B). A genotype but not a treatment effect was also observed for time spent in the quadrant area on the probe trial day with 5xFADxTg30 mice spending less time in the quadrant area (Figure 5C), an effect that was more obvious in male mice (Figure S5C). While this can indicate a decrease in spatial memory, it could also indicate a motor phenotype causing a slower swim speed to the platform. Consequently, swim velocity was measured and was indeed significantly reduced in 5xFADxTg30 mice (Figure 5D) but not affected by BGP‐15, with statistical significance identified for female but not male mice (Figure S5D). Y‐Maze testing revealed that the 5xFADxTg30 mice spent less time in the novel arm than WT mice, but this was not altered with BGP‐15 treatment (Figure 5E). This phenotype was driven by the female mice (Figure S5E). No genotype or drug effect was identified with novel object recognition testing (Figure 5F, Figure S5F). Given the previously noted difference in anxiety‐related behavior in the 5xFADxTg30 mice (Figure 2G), we investigated this once more in an elevated plus maze at 9 months of age. 5xFADxTg30 mice spent a greater proportion of time in the open arm and less time in the closed arm (Figure 5G), again an effect equally observed in male and female mice (Figure S5G). Similarly to what was seen with Hsp72 genetic overexpression, BGP‐15 treatment did decrease time spent in the open arm specifically in female 5xFADxTg30 mice (Figure S5G); however, when sexes were combined BGP‐15 treatment did not elicit an effect (Figure 5G).

FIGURE 5.

Behavioral testing with BGP‐15 treatment. (A) Latency to platform (mixed effects analysis). (B) Corresponding area under the curve over 6 days for the Morris water maze testing, two‐way ANOVA, F values and degrees of freedom, F (DFn, DFd): interaction F (1, 28) = 0.475, p = 0.496, treatment, F (1, 28) = 0.066, p = 0.799, genotype F (1, 28) = 52.57. (C) Probe trial on Day 7 of Morris water maze testing. Time spent in platform quadrant when mice were 8 months of age, two‐way ANOVA, n = 6–9 and conducted at 8 months of age, F values and degrees of freedom, F (DFn, DFd): interaction F (1, 28) = 0.336, p = 0.567, treatment, F (1, 28) = 0.915, p = 0.347, genotype F (1, 28) = 8.94, p = 0.006. (D) Average swim velocity, two‐way ANOVA, n = 20–24, F values and degrees of freedom, F (DFn, DFd): interaction F (1, 85) = 4.917, p = 0.029, treatment, F (1, 85) = 0.0007, p = 0.979, genotype F (1, 85) = 11.42, p = 0.001. (E) Y‐Maze testing with time spent in the novel arm at 7 months of age, two‐way ANOVA, n = 23–24, F values and degrees of freedom, F (DFn, DFd): interaction F (1, 90) = 0.7025, p = 0.404, treatment, F (1, 90) = 1.220, p = 0.272, genotype F (1, 90) = 15.91, p = 0.0001. (F) Number of bouts of investigative behavior of the novel object at 7 months of age, two‐way ANOVA, n = 8–9, F values and degrees of freedom, F (DFn, DFd): interaction F (1, 30) = 0.2437, p = 0.6252 treatment, F (1, 30) = 1.013, p = 0.322, genotype F (1, 40) = 0.002, p = 0.968. (G) Anxiety‐like behavior assessed via an elevated plus maze, at 9 months of age with duration of overall time spent in each arm, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 WT CON vs. 5xFADxTg30 CON, #p ≤ 0.05, ##p < 0.01, ###p < 0.001, WT CON vs. 5xFADxTg30 BGP‐15, ^p < 0.05 ^^p < 0.01, ^^^^p < 0.0001 WT BGP‐15 vs. 5xFADxTg30 CON, &p < 0.05, &&&p < 0.001, &&&&p < 0.0001, WT BGP‐15 vs. 5xFAD BGP‐15, two‐way ANOVA with Tukey post hoc, n = 7–9, F values and degrees of freedom, F (DFn, DFd): Open, interaction F (1, 29) = 2.479, p = 0.126, treatment, F (1, 29) = 1.504, p = 0.230, genotype F (1, 29) = 58.44, closed, interaction F (1, 29) = 3.554, p = 0.070 treatment, F (1, 29) = 1.562, p = 0.221, genotype F (1, 29) = 63.68. Male and female mice combined. Graphs indicate mean ± SD.

Assessment of physical activity levels revealed a decrease in activity in the 5xFADxTg30 mice in both light and dark cycles, which was not rescued by BGP‐15 administration (Figure 6A). The effect in the light cycle was more obvious in female mice and the effect in the dark cycle more obvious in the male mice (Figure S6A). Next, an indication of gross motor strength and control was assessed through rotarod performance and hangwire testing. As before (Figure 3B,C), 5xFADxTg30 mice exhibited deficits in performance for both tests that worsened with age, with BGP‐15 treatment unable to provide any improvement (Figure 6B,C). Effects were similar in male and female mice (Figure S6B,C). Given the marked differences in rotarod performance and hangwire ability in the genetic overexpression study, assessment of gait was made in male mice via DigiGait analysis prior to the timepoint in which differences were noted for rotarod performance. Analysis of swing duration and propulsion duration demonstrated a decrease in the 5xFADxTg30 mice in all four limbs suggestive of muscle weakness, which was unable to be improved by BGP‐15 (Figure 6D–G). Further, stride frequency was increased in the 5xFADxTg30 mice as a likely compensatory mechanism for weakness. Again, this was unaltered by BGP‐15 treatment (Figure 6D–G). Consequently, alterations to gait are detectable in the 5xFADxTg30 model prior to decrements in rotarod performance. In relation to metabolic control, assessment of glucose tolerance again revealed a decrease in the glucose excursions in the 5xFADxTg30 mice, which BGP‐15 treatment did not alter (Figure 7A,B). Female mice contributed to the effect on glucose excursions more than male mice (Figure S7A,B).

FIGURE 6.

Physical activity, motor control, muscular endurance and gait with BGP‐15 treatment. (A) Physical activity levels over a 24‐h period as measured by beam breaks divided into light and dark cycles, two‐way ANOVA with Tukey's post hoc test, n = 24–28 conducted at 5 months of age, F values and degrees of freedom, F (DFn, DFd): light, interaction F (1, 98) = 2.95, treatment, F (1, 98) = 0.093, genotype F (1, 98) = 19.94, dark, interaction F (1, 98) = 0.3295, treatment, F (1, 98) = 0.045, genotype F (1, 98) = 7.118. (B) Latency to fall off rotarod between 7 and 10 months of age, mixed effects analysis with Tukey's post hoc test, n = 22–27. (C) Latency to fall from hangwire at 8 and 10 months of age, mixed effects analysis with Tukey's post hoc test, n = 24–26. DigiGait analysis: (D) left forelimb, (E) right forelimb, (F) left hindlimb, and (G) right hindlimb swing duration, propulsion duration and stride frequency at 5 months of age, two‐Way ANOVA with Tukey's post hoc test, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 WT CON vs. 5xFADxTg30 CON, ##p < 0.01, ###p < 0.001, ####p < 0.0001 WT CON vs. 5xFADxTg30 BGP‐15, ^p < 0.05 ^^p < 0.01, ^^^p < 0.001, ^^^^p < 0.0001 WT BGP‐15 vs. 5xFADxTg30 CON, &p < 0.05, &&p < 0.01, &&&p < 0.001, &&&&p < 0.0001, WT BGP‐15 vs. 5xFAD BGP‐15 n = 12–21, F values and degrees of freedom, F (DFn, DFd), (D) Left forelimb swing, interaction, F (1, 64) = 0.483, treatment, F (1, 64) = 0.014, genotype, F (1, 64) = 52.00, left forelimb propulsion, interaction, F (1, 64) = 0.0012, treatment, F (1, 64) = 0.230, genotype, F (1, 64) = 24.36, left forelimb stride, interaction, F (1, 64) = 0.292, treatment, F (1, 64) = 0.163, genotype F (1, 64) = 43.95. (E) Right forelimb swing, interaction, F (1, 64) = 0.003, treatment, F (1, 64) = 2.019, genotype, F (1, 64) = 65.13, right forelimb propulsion, interaction, F (1, 64) = 1.285, treatment, F (1, 64) = 0.676 genotype, F (1, 64) = 44.08, right forelimb stride, interaction, F (1, 64) = 0.6204, treatment, F (1, 64) = 1.527, genotype F (1, 64) = 61.07. (F) Left hindlimb swing, interaction, F (1, 64) = 2.687, treatment, F (1, 64) = 0.024, genotype, F (1, 64) = 37.65, left hindlimb propulsion, interaction, F (1, 64) = 0.2508, treatment, F (1, 64) = 0.888, genotype, F (1, 64) = 56.62, left hindlimb stride, interaction, F (1, 64) = 1.060, treatment, F (1, 64) = 0.418, genotype F (1, 64) = 56.76. (G) Right hindlimb swing, interaction, F (1, 64) = 0.006, treatment, F (1, 64) = 0.459, genotype, F (1, 64) = 42.89, right hindlimb propulsion, interaction, F (1, 64) = 0.120, treatment, F (1, 64) = 1.259 genotype, F (1, 64) = 59.25, right hindlimb stride, interaction, F (1, 64) = 0.124, treatment, F (1, 64) = 1.50, genotype F (1, 64) = 52.93. Male and female mice combined. Graphs indicate mean ± SD.

Despite this in‐depth phenotypic analysis of the 5xFADxTg30 model over multiple cohorts, we did not have a definitive cause of death for the mice that died before endpoint and the trend for BGP‐15 to alleviate death (see Figure 4I). Anecdotal observations identified rare periods of shaking or twitching behavior, which could be indicative of a seizure phenotype, and promoting seizures as a possible cause of death. Subsequently, in a separate cohort of male mice (given the different response between sexes refer to Figure S4H), we assessed seizure susceptibility in the model using the electrical kindling model. 5xFADxTg30 mice took fewer stimulations to trigger a Class V (convulsive) seizure, had longer average seizure duration, and a higher average seizure class than WT mice (Figure 7C–E), consistent evidence of seizure susceptibility in 5xFADxTg30 mice. Despite this, BGP‐15 treated 5xFADxTg30 mice did not demonstrate any improvement related to seizure susceptibility (Figure 7C–E). Finally, as with the genetic model, we assessed whether the BGP‐15 treatment was able to reduce the burden of tau and amyloid in the brain via western blotting. 5xFADxTg30 mice exhibited high levels of phosphorylated and total tau and had accumulated Aβ (Figure 7F–I). 5xFADxTg30 mice treated with BGP‐15 exhibited the same levels of these proteins, indicating that the compound was unable to limit the burden of these aggregates in the brain (Figure 7F–I). Like in the genetic model, no differences were observed for this measure between male and female mice (Figure S7C,D).

4. Discussion

The present study assessed the ability of Hsp72 genetic overexpression and attempted pharmacological targeting of Hsp72 protein abundance to improve physiological and behavioral parameters in the 5xFADxTg30 AD mouse model. Interest in the utility of Hsp72 overexpression and/or BGP‐15 treatment to mitigate defects in this model stemmed primarily from the body of work suggesting that Hsp72 protein elevation had the potential to limit Aβ and/or tau burden. In our hands, we did not find any changes to the levels of these aggregates with these interventions in this model. While neither intervention altered Aβ and/or tau, minor phenotypic improvements were observed in terms of survival rates in male mice with both genetic and BGP‐15 interventions and altered fear recognition in female mice in response to both genetic and BGP‐15 interventions. Further, our in‐depth phenotypic analysis provides new insights into the presentation of the 5xFADxTg30 model, particularly in terms of our novel findings regarding gait alterations and seizure susceptibility.

4.1. Genetic Model

Despite the high levels of Hsp72 overexpression that were achieved in the brain (Figure 1A,B), no alteration in levels of Aβ and tau was observed. While, to the best of our knowledge, no other in vivo study has investigated the effect of Hsp72 elevation in a tau model, laboratory‐based studies had indicated that Hsp72 could be efficacious in doing so. The inability of Hsp72 overexpression to effect tau in our model could stem from the added complexity found within living organisms and the interactions between multiple cell and organ types and peripheral factors. Alternatively, the combination of both concurrent Aβ and tau pathology may have presented a scenario where cellular homeostasis was compromised to such a degree that it prevented Hsp72 from acting on tau compared to when tau pathology is present in isolation. While studies such as those from Hoshino and colleagues in Aβ‐only models (Hoshino et al. 2011) have demonstrated that Hsp72 transgenic overexpression decreases Aβ levels, several differences exist between studies that may explain the discrepancy. Firstly, the model used in the current study carries three APP mutations (Swedish, Florida, and London) and two presenilin mutations compared to just the Swedish mutation in the previous work, which leads to phenotypes of differing severity. For example, in terms of plaque area, the 5xFAD model reaches close to maximal plaque area at 4 months of age (Forner et al. 2021), compared to models with the single Swedish mutation where the pathology is slower to evolve and rapidly increases after 12 months of age (Callahan et al. 2001). Furthermore, concurrent Aβ and tau pathology in our model may have presented an environment where Hsp72 could not act on Aβ compared to when Aβ pathology has been exclusively present in previous studies. Given Aβ and tau lesions often appear simultaneously in AD and interact (Busche and Hyman 2020), this underscores the importance of assessing potential interventions in a model where both pathologies are observed.

A major difference observed between WT and 5xFADxTg30 mice in the behavior tasks was in the elevated plus maze where 5xFADxTg30 mice spent significantly longer periods in the open arm and less time in the closed arms, suggestive of a potential reduction in levels of anxiety and/or possible increased risk‐taking behavior and/or impaired threat detection. There was a trend for Hsp72 overexpression to reduce open‐arm exploration, shifting behavior towards WT levels (p ≤ 0.06, male and female data combined, Figure 2G) which was driven by alterations in the female mice (p ≤ 0.05; Figure S2D). Hsp72 overexpression may be contributing to the mice becoming more cautious and avoiding risk (the natural response to the specific experimental conditions that is lost in 5xFADxTg30 mice). Why Hsp72 may be contributing to the normalization of the elevated plus maze results specifically in females cannot be inferred from this study but warrants future investigation.

Survival also trended to be higher with Hsp72 overexpression with double the number of 5xFADxTg30 mice not reaching study endpoint compared to 5xFADxTg30xHsp72Tg mice (Figure 1I). Disaggregated data for sex indicated this was due more to alterations in the male survival with a clear deviation in the survival curve identified in male but not female mice (Figure S1G). There is a limitation to the survival analysis as it is not true lifespan assessment that was measured as it was necessary to cease the study at 10 months of age due to the deteriorating health condition and therefore welfare of the 5xFADxTg30 mice. It is possible that if true lifespan was assessed (i.e., death as an endpoint in all mice) that a greater difference may have been observed. As we are yet unsure of the precise cause of death in the model, it is difficult to speculate on the potential protective mechanism of Hsp72 overexpression and why it is more efficacious in the male mice.

Various investigations have demonstrated that lower muscle mass and strength metrics are associated with the impairment of cognition. Indeed, low appendicular lean soft tissue mass is independently associated with faster 3‐year cognitive decline in executive functions with aging (Tessier et al. 2022). The Hsp72 transgenic model overexpresses Hsp72 not only in the brain but also in muscle tissue. Hsp72 overexpression had been previously shown to improve exercise performance (Henstridge et al. 2014), improve rotarod and foot fault performance in a cerebral ischemic model (Xu et al. 2011), and preserve muscle strength and partially ameliorate the dystrophic pathology in a mouse model of muscular dystrophy (Gehrig et al. 2012). Given the muscle‐related phenotype of the 5xFADxTg30 model (decreased muscle weight, impaired performance in rotarod and hangwire testing), it was important to study whether Hsp72 could rescue these deficits potentially from direct effects of Hsp72 overexpression in the muscle. Our findings do not provide any evidence that Hsp72 overexpression can offset these phenotypic traits despite overexpression both centrally (brain) and peripherally (muscle) where either site (or both sites in combination) could have provided beneficial effects.

4.2. Pharmacological Model

While genetic Hsp72 overexpression had modest effects in rescuing phenotypes associated with the 5xFADxTg30 model, it was still of interest to examine the ability of BGP‐15 to increase Hsp72 abundance in the brain and whether it could elicit any improvement due to its multi‐modal nature. As discussed in the introduction section, in addition to its Hsp72 induction properties, BGP‐15 has also been shown to display beneficial properties towards a variety of other cellular processes implicated in AD development or pathology. Our assessment of brain regions following 8 months of chronic treatment with BGP‐15 did not detect any change in Hsp72 expression levels, bringing into question the compound's ability to increase Hsp72 in this setting. Further, BGP‐15 did not lead to any significant changes in body composition, many behavioral measures, metabolic measures, movement measures, or Aβ/tau levels. However, like the Hsp72 overexpression model, there was a trend for increased survival (driven by male mice) and a decrease in open‐arm exploration in female mice. Despite a lack of increase in Hsp72 expression in the brain at endpoint, the convergent findings from the genetic and pharmacological interventions do suggest a potential shared mechanism of action targeting the same pathway. This parallel, sex‐specific pattern across two independent modalities makes a chance finding unlikely and instead supports the interpretation that both interventions modulate a common biological mechanism, with downstream effects that manifest in a sex‐dependent manner that modestly enhance survival in males and alter anxiety‐related circuitry in females. Thus, we cannot rule out that BGP‐15 may have increased Hsp72 expression levels at earlier timepoints in the study, in different parts of the brain than were tested or expression changes are cyclical that were missed by chance due to the timing of endpoint dissection.

Given the survival finding, our attention turned to attempting to determine the underlying cause of decreased survival in the 5xFADxTg30 mice. Observations by investigators suggested that at stages the 5xFADxTg30 mice may have been displaying tremor‐like behavior, presenting the possibility they may be having some type of seizure. The prevalence of epilepsy in patients with AD is 10‐fold higher than the age‐matched general population and up to 64% of AD patients will develop at least one unprovoked seizure (Hauser et al. 1986; Risse et al. 1990). However, there is little known about the biological mechanisms associated with AD which generate the vulnerability towards epilepsy. Assessment of seizure‐related outcomes did indeed demonstrate increased seizure susceptibility in this model, aligning with other models of AD including the hAPP (Palop et al. 2007) and APdE9 (Minkeviciene et al. 2009) models driven by Aβ, as well as 3xTg (Vande Vyver et al. 2022) and Tg4510 (Liu et al. 2017) models of tau. However, BGP‐15 treatment did not offer any improvement against this pronounced phenotype in our hands. Our findings suggest that the 5xFADxTg30 model may prove useful for investigators wishing to study the combined effects of Aβ and tau aggregates on seizure activity.

Apart from BGP‐15, other compounds have attempted to induce Hsp72 expression in various models such as Geranylgeranylacetone (GGA), an anti‐ulcer agent. Oral administration of GGA in APP23 transgenic mice (Aβ model) demonstrated an improvement in cognitive function and a reduction in Aβ plaque deposition and synaptic loss (Hoshino et al. 2013). However, this appeared to be independent of Hsp72 as no observable change in Hsp72 expression was observed in brain tissue with this intervention in this model (Hoshino et al. 2013). Treatment with GGA also demonstrated an improved cognitive deficit and lowered the levels of Aβ peptide in APP/PS1 mice which express a chimeric mouse/human APP and a mutant human presenilin 1 (PS1) (Sun et al. 2017). However, it must be noted in this study that no data indicating an elevation in Hsp72 was provided by the authors. Furthermore, when the same compound was tested in a randomized, double‐blind, placebo‐controlled study in AD patients, improvements were observed in individuals if administered prior to atrophy progression in the medial temporal areas (Yokoyama et al. 2019) but whether GGA increased Hsp72 in these patients was again not measured. It remains to be determined whether GGA's beneficial effects in models of AD or patients with AD are due to Hsp72 induction or by an alternative mechanism. It also highlights the need for further drug development of bona fide Hsp72 expression activators for the brain.

Limitations of our study include the necessity to cease the study at 10 months of age due to ethical requirements meaning that our mice would not be considered “old”. Greater aging may have revealed further age‐specific phenotypes and allowed for greater delineation between groups. This is especially true for the survival curve with BGP‐15 treatment as the curves began to separate only after 8 months of age (Figure 4I). Also, only one dose of the BGP‐15 was assessed. While this dose was based on previous studies and was sufficient to lead to brain BGP‐15 exposure, it is possible higher doses may be more efficacious. Moreover, although immunohistochemical staining for amyloid‐β and tau was part of the original study design to provide complementary information on regional pathology distribution, technical issues compromised the integrity of the samples precluding reliable analysis. Despite this, our biochemical findings were clear that there was no effect with each intervention and histological assessment could be included as part of any future work. Finally, as is the case with all animal models of human disease, translating behavioral change in mouse models to cognitive or behavioral changes in humans is challenging and the use of mouse models as tools is limited to answering specific targeted questions.

One important caveat to our findings is that these studies specifically evaluated whether increasing Hsp72 abundance, either genetically or through pharmacological induction, is sufficient to modify disease progression in the model. It must be noted that Hsp72 function is not solely determined by protein abundance alone. Indeed, its chaperone activity can be influenced by such factors as subcellular localization, nucleotide‐binding state, interactions with co‐chaperones and client proteins, and post‐translational modifications to name a few. Rather, our findings indicate that increasing Hsp72 abundance alone was insufficient to meaningfully alter pathological burden or rescue most disease‐associated phenotypes in the 5xFADxTg30 model. It remains possible that alternative approaches that modify Hsp72 activity, client interactions, chaperone network function, or other related factors or mechanisms may yield different outcomes and warrant further investigation.

In conclusion, the 5xFADxTg30 model, which is characterized by both Aβ and tau accumulation, displays multiple deficits in motor/muscular performance, alterations to body composition, decreased survival, heightened seizure susceptibility, and changes to anxiety and risk‐taking behavior. While genetic targeting of Hsp72 expression levels and attempted pharmacological verification with BGP‐15 in this model exhibited some minor protective effects in a sex‐dependent manner, they do not fully rescue any phenotypic defects observed in the model or reduce tau or Aβ levels. Consequently, the ability of Hsp72 overexpression/induction to mitigate tau and Aβ‐associated phenotypes in vivo when both are present in combination is not supported by our thorough investigations.

Author Contributions

J. P. S. Marshall: investigation, formal analysis, writing – original draft, writing – review and editing, visualization, funding acquisition. A. C. Calkin: resources, writing – review and editing, supervision. B. G. Drew: resources, writing – review and editing, supervision. J. Morizzi: methodology, investigation, writing – review and editing. A. Harutyunyan: methodology, investigation, formal analysis, writing – review and editing. D. C. Henstridge: conceptualization, formal analysis, resources, writing – original draft, writing – review and editing, visualization, supervision, Funding acquisition. P. Wright: methodology, investigation, writing – review and editing. N. C. Jones: methodology, resources, writing – review and editing. J. M. Collins: resources, writing – review and editing. O. M. West: investigation, writing – review and editing. P. A. Adlard: conceptualization, methodology, resources, writing – review and editing, supervision. M. A. Febbraio: conceptualization, resources, writing – review and editing, supervision, funding acquisition. A. E. King: resources, writing – review and editing. J. Ng: investigation, formal analysis, writing – review and editing, visualization.

Funding

This work was supported by the Yulgibar Foundation, the Dementia Australia Research Foundation, and the Shine of Foundation.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Disaggregated data relating to Figure 1. (A) Female and male body weight, (B) lean mass, (C) fat mass, (D) and body fat percentage, females: n = 14–17, males: n = 10–14. (E) Brain weight, females: n = 6–10, males: n = 6. (F) Tibialis Anterior weight females: n = 3–7, males: n = 5–6. (G) Survival analysis females: n = 14–17, males: n = 10–15. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 WT vs. 5xFADxTg30, #p < 0.05, ##p < 0.01, ###p < 0.001, ####p < 0.0001 WT vs. 5xFADxTg30xHsp72Tg. Graphs indicate mean ± SD.

Figure S2: Disaggregated data relating to Figure 2. (A) Female and male Y‐Maze performance at 7 months of age females: n = 12–15, male n = 10–13 and (B) at 9 months of age females: n = 13–14, males n = 9–13. (C) Number of bouts of investigative behavior during a novel object recognition test females: n = 13–15, males 9–12, (D) Anxiety‐like behavior assessed via an elevated plus maze, females: n = 11–15, males n = 9–12. *p < 0.05, ****p < 0.0001 WT vs. 5xFADxTg30, #p < 0.05, WT vs. 5xFADxTg30xHsp72Tg, ^p ≤ 0.05 5xFADxTg30 vs. 5xFADxTg30xHsp72Tg. Graphs indicate mean ± SD.

Figure S3: Disaggregated data relating to Figure 3. (A) Female and male physical activity (beam breaks): females n = 5–7, males n = 6–7. (B) Female and male rotarod performance: females n = 11–15, males = 9–13. (C) Female and male Hangwire performance: females n = 11–15, males n = 9–13. (D) Female and male glucose tolerance curves: females n = 13–14, males n = 10–12. (E) Area under the curve (AUC) females n = 13–14, males n = 10–12. (F) Representative western blot images from female and male hippocampus tissue from endpoint at 10 months of age with corresponding quantification for pTau, Tau and Aβ, females: n = 6, males: n = 3–5. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 WT vs. 5xFADxTg30, #p < 0.05, ##p < 0.01, ###p < 0.001, ####p < 0.0001 WT vs. 5xFADxTg30xHsp72Tg. Graphs indicate mean ± SD.

Figure S4: Disaggregated data relating to Figure 4. (A) Female and male hippocampus Hsp72 expression levels and (B) cerebellum Hsp72 expression levels, n = 3. (C) Female and male body weight, (D) lean mass, (E) fat mass, (F) and body fat percentage, females n = 9–14, males 10–16. (G) Tibialis Anterior weight, females n = 7–8, males 2–4. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 WT CON vs. 5xFADxTg30 CON, ##p < 0.01, ###p < 0.001, ####p < 0.0001 WT CON vs. 5xFADxTg30 BGP‐15, ^p < 0.05 ^^p < 0.01, ^^^p < 0.001, ^^^^p < 0.0001 WT BGP‐15 vs. 5xFADxTg30 CON, &p < 0.05, &&p < 0.01, &&&p < 0.001, &&&&p < 0.0001, WT BGP‐15 vs. 5xFAD BGP‐15 (H) Kaplan–Meier survival curve, Log‐rank (Mantel‐Cox) test females n = 9–14, males = 10–25. *p < 0.05 WT CON vs. 5xFADxTg30 CON, ^^p < 0.01, WT BGP‐15 vs. 5xFADxTg30 CON. Graphs indicate mean ± SD.

Figure S5: Disaggregated data relating to Figure 5. (A) Female and male latency to platform (B) Area under the curve (C) Time spent in platform quadrant, females n = 3–4, males n = 3–5. (D) Swim velocity, females n = 9–14, males n = 8–15. (E) Y‐Maze performance, time spent in novel arm, females n = 9–14, males n = 10–15. (F) Number of bouts of investigative behavior of the novel object, females n = 3–4, males = 4–5. (G) Duration spent in each arm of elevated plus maze females 3–4, male 3–5. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 WT CON vs. 5xFADxTg30 CON, #p < 0.05, ##p < 0.01, ###p < 0.001 WT CON vs. 5xFADxTg30 BGP‐15, ^p < 0.05 ^^p < 0.01, ^^^p < 0.001 WT BGP‐15 vs. 5xFADxTg30 CON, &p < 0.05, &&p < 0.01, WT BGP‐15 vs. 5xFAD BGP‐15. + = p < 0.05 WT‐CON vs. WT‐BGP‐15, $ p ≤ 0.05 5xFADxTg30‐CON vs. 5xFADxTg30 BGP‐15. Graphs indicate mean ± SD.

Figure S6: Disaggregated data relating to Figure 6. (A) Physical activity over a 24‐h period (beam breaks): females n = 9–14, males n = 10–15. (B) Latency to fall off rotarod from 7 to 10 months: females n = 9–14, males n = 10–15. (C) Time on hangwire at 8 and 10 months of age: females n = 9–14, males n = 10–15. **p < 0.01, ***p < 0.001, ****p < 0.0001 WT CON vs. 5xFADxTg30 CON, #p < 0.05, ##p < 0.01, ###p < 0.001, ####p < 0.0001 WT CON vs. 5xFADxTg30 BGP‐15, ^^p < 0.01, ^^^p < 0.001, ^^^^p < 0.0001 WT BGP‐15 vs. 5xFADxTg30 CON, &p < 0.05, &&p < 0.01, &&&&p < 0.0001, WT BGP‐15 vs. 5xFAD BGP‐15. + = p < 0.05 WT‐CON vs. WT‐BGP‐15, $ p ≤ 0.05 5xFADxTg30‐CON vs. 5xFADxTg30 BGP‐15. Graphs indicate mean ± SD.

Figure S7: Disaggregated data relating to Figure 7. (A) Glucose tolerance excursion curves: female n = 9–14, male n = 10–15. (B) AUC: females n = 9–14, males n = 10–15. (C) Western blots from hippocampus tissue for females n = 8, and (D) males n = 7–8. *p < 0.05, ***p < 0.001, ****p < 0.0001 WT CON vs. 5xFADxTg30 CON. #p < 0.05, ##p < 0.01, ####p < 0.0001 WT CON vs. 5xFADxTg30 BGP‐15. ^p < 0.05, ^^p < 0.01, WT BGP‐15 vs. 5xFADxTg30 CON. $ p < 0.05 5xFADxTg30‐CON vs. 5xFADxTg30 BGP‐15. + p < 0.05 WT‐CON vs. WT‐BGP‐15. Graphs indicate mean ± SD.

Figure S8: Timeline and sequence of behavioral tests for (A) genetic study, (B) pharmacological study, and (C) seizure study as a separate component of the Pharmacological studies.

Figure S9: Summary table of findings for the genetic and pharmacological studies.

Figure S10: Original immunoblot images through which images in manuscript were sourced. Yellow band indicates sections displayed in main manuscript figures.

Data S1: Transparent Science Questionnaire.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.