APOE-TR^+/+/5xFAD^+/− (EFAD) mice express 5 familial AD mutations (APP K670N/M671L, I716V, V717I, PS1 M146L, and L286V) under control of the neuron-specific mouse Thy-1 promoter, and express human APOE3 or APOE4 (Youmans et al., 2012; Tai et al., 2017). For this study, female mice were either homozygous for human APOE3 (E3FAD) or heterozygous for APOE3 and APOE4 (E3/4FAD). Mice homozygous for APOE4 (E4FAD) were not used because EFAD mice of this genotype are insensitive to the beneficial effects of E₂ on object recognition and spatial memory or CA1 dendritic spine density (Balu et al., 2023; Taxier et al., 2022a). Given the extensive research establishing validity of the EFAD model of AD relative to wild-type conditions (Lewandowski et al., 2020) and our previous work demonstrating the efficacy of EGX358 in young wild-type female mice (Hanson et al., 2018; Fleischer et al., 2021), wild-type mice were not used in the current study. EFAD mice were bred, weaned, and genotyped at the University of Illinois at Chicago (UIC) and shipped to the University of Wisconsin-Milwaukee at 2 months of age, where they were aged to 5 months before the start of OVX surgery. Although mice of this age are still considered young adults, amyloid accumulation in EFAD mice is already striking by 4 months of age and quite significant by 6 months of age (Youmans et al., 2012; Tai et al., 2017). Our previous work with 6–7 month E3FAD and E4FAD males and females indicated that E4FAD mice of both sexes already exhibited widespread dendritic spine loss in the hippocampus and prefrontal cortex, with concomitant deficits in multiple memory tasks (Taxier et al., 2022a,b), so we sought here to intervene with EGX358 at an earlier age to model early-stage AD.

Prior to surgery, mice were housed in groups of up to 5 per cage but then were singly housed following OVX for the remainder of the study to enable treatment dosing based on each mouse’s hydrogel intake and weight. Mice were treated and behaviorally tested in 4 separate cohorts, each including approximately even numbers of E3FAD and E3/4FAD mice. Mice were maintained on a 12 h light/dark cycle (lights on a 07:00) with all procedures conducted during the light portion of the cycle. Mice received ad libitum access to food (Teklad Rodent Diet 8,604, Inotiv) for the duration of the study. Although typical isoflavone concentrations (daidzein + genistein aglycone equivalents) in this food range from 575 to 870 mg/kg, human diets naturally contain some soy, so this diet is more translatable than soy-free chow. Importantly, all mice received the same diet, so this food is unlikely to have contributed to observed treatment or genotype effects. Mice were provided with ad libitum access to water up until 2 weeks after surgery, at which point water bottles were removed and all hydration was provided via hydrogel (ClearH20) containing vehicle or EGX358 (see below for dosing information). Hydrogel consumption was measured daily and replenished as needed. Protocols and procedures followed the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the University of Wisconsin-Milwaukee Institutional Animal Care and Use Committee.

A schematic of the study’s overall experimental design is shown in Figure 1. As described in section 2.1, mice were bred at UIC and sent to UWM at 2 months of age where they were matured to 5 months of age to allow for the development of amyloid deposition, elevated phospho-tau, gliosis, and synaptic degeneration (Youmans et al., 2012; Tai et al., 2014; Liu et al., 2015; Cacciottolo et al., 2016; Balu et al., 2023). At 5 months of age, mice were OVX and 1 week later began oral treatment with vehicle (dimethylsulfoxide, DMSO) or EGX358 (10 mg/kg/day) mixed into hydrogel water gel for animal hydration. Following 4 weeks of initial treatment, mice began behavioral testing, through which daily treatments continued. The month-long pre-training treatment provided ample time for EGX358 to exert long-term effects on brain function prior to behavioral assessment; treatment continued through the study until tissue collection to assess effects of chronic treatment on behavioral and physiological outcomes. After initial pre-treatment, mice were tested in the open field to assess locomotor activity and anxiety-like behaviors, followed 24 h later by the start of object placement and object recognition testing. Object placement and object recognition tests were conducted on separate days 3–4 days apart for an individual mouse, the order of which was counterbalanced within and between groups. Together, open field and testing in both object tasks lasted about 2 weeks. Four days after the conclusion of memory testing, anxiety-like behaviors were evaluated in the elevated plus maze, followed 2 days later by assessment of vasomotor reaction to a single drug-induced hot flash. Together, plus maze and hot flash testing lasted about 1 week. Mice were euthanized 3 weeks later for measurement of uterine weights.

Mice were single-housed and provided with ¼ of a carprofen (Rimadyl, Bioserv) tablet (2 mg/full tablet) in their cage the day prior to surgery. At the start of surgery, mice received a 5 mg/kg subcutaneous dose of Rimadyl for pain management. They were then anesthetized with isoflurane in 100% oxygen (5% for induction, 2–3% for maintenance). Briefly, two dorsal incisions were first made in the skin over the pelvic bones and underlying muscle walls. The ovaries and oviducts were isolated, ligated at the tip of the uterine horn, tied off with nylon monofilament (Monomid), and removed. The muscle walls were then sutured with chromic gut, and incision sites closed with wound clips, which were removed 6–8 days after surgery (Lewis et al., 2008). Mice received additional ¼ carprofen tablets 12 and 36 h after surgery for postsurgical analgesia and were given 1 week to recover prior to the start of drug administration.

Long-term oral vehicle and EGX358 treatments were dissolved in 10% DMSO and delivered via HydroGel (ClearH₂O), which served as the sole source of hydration. Hydrogels were prepared separately for each mouse as per manufacturer instructions. Briefly, to make an aliquot for one mouse, 50 g of gel was melted in a 50 mL Falcon tube in a 60°C water bath for 10 min and then vehicle or EGX358 solution was injected into the gel using aseptic technique, and the tube shaken by hand and then vortexed vigorously for up to a minute. Our pilot testing with a blue dye injected into the melted gel at the same volume as dissolved compound filled the gel in a completely homogenous manner. Prepared gel was placed in a plastic cup affixed to the wall of the cage and mice were allowed to freely consume the gel which was replenished daily between 8 am and 12 pm. Gel was made fresh for each mouse every 2–3 days, based on the mouse’s weight and average gel consumption across the previous week. Mice were first acclimated to plain hydrogel consumption without treatment for 3 days. Mice quickly learned to consume the gel and readily ate the entire gel presented to them, resulting in normal levels of hydration. Mean consumption (g/day) was calculated and used to determine the concentration of drug required to achieve 10 mg/kg/day EGX358 dissolved in 10% DMSO. This dose was based on our previously published work in which EGX358 was administered via oral gavage at a dose of 0.5 mg/kg/day (Hanson et al., 2018; Fleischer et al., 2021) and unpublished observations from our lab that estrogen compound dosing should be increased by approximately 20x for HydroGel administration. As a control, vehicle gels were mixed with the same amount of 10% DMSO relative to volume of gel consumption. Mice were weighed weekly (g), and gel consumption was recorded daily, both of which were used to adjust the amount of EGX358 in DMSO to achieve the desired 10 mg/kg/day dose.

To chemically induce a rapid and transient rise in tail skin temperature meant to mimic a hot flash, the neurokinin-3 tachykinin receptor (NK3R) agonist senktide (Tocris Biosciences) was injected just once after a 10 min baseline recording period. Senktide was dissolved in 0.9% saline and injected subcutaneously at a dose of 0.5 mg/kg (5 mL/kg) (Fleischer et al., 2021; Krull et al., 2017).

Following 4 weeks of daily treatment with hydrogel, mice were habituated to an open field for 5 min/day for 2 days to prepare them for subsequent object recognition and object placement training and testing (Fleischer et al., 2021). Each day, mice were placed into the lower center of an empty white box (60 cm × 60 cm × 47 cm) divided into a 5 × 5 square grid. During the first day of habituation, exploratory behaviors were recorded and quantified automatically by Any-maze software (Stoelting), which tracked the center of the mouse’s body (Hanson et al., 2018; Fleischer et al., 2021). Time (s) spent in the center zone (innermost 9 squares) and outer zone (outermost 16 squares) were measured as indications of exploration and anxiety-like behavior (more center time indicates less anxiety-like behavior) (La-Vu et al., 2020). Path length (m) and average speed (m/s) were recorded to assess locomotor function and measure potential baseline effects of treatment or genotype on performance factors that might influence object memory results. Our previous work showed no differences between OVX E3FAD and E4FAD mice on these measures (Taxier et al., 2022c), suggesting no effect of APOE4 homozygosity on anxiety-like or explorative behaviors in the open field.

Twenty-four hours after the second habituation session, mice were trained in object recognition (OR) or object placement (OP) tasks, the order of which were counterbalanced within each group. OR and OP were chosen to measure object recognition and spatial memory, respectively, as both depend on the integrity of the dorsal hippocampus and mPFC (Eichenbaum, 2017; Tuscher et al., 2018; Sauer et al., 2022). Further, our laboratory has shown repeatedly that OR and OP are sensitive to the memory-enhancing effects of E₂ (Koss et al., 2018; Kim et al., 2019; Gross et al., 2021) and ERβ agonism (Boulware et al., 2013; Hanson et al., 2018; Fleischer et al., 2021) in OVX wild-type and the memory-enhancing effects of E2 in OVX EFAD mice (Taxier et al., 2022a). In the training phase, mice were first accustomed to the empty open field box for 2 min and then returned to their home cage, during which time two identical objects were placed near the northeast and northwest corners of the box. Mice were then returned to the box and allowed to freely explore the objects until they had accumulated 30 s of object exploration (defined by nose contact with the objects) or until 20 min had elapsed. Those that did not reach this criterion were not advanced to testing right away, but rather were given a second training session with different object a few days later, as our experience indicates that some mice overcome a hesitancy to explore the objects when given a second chance to do so. As described below in section 2.9, all mice that underwent this second training session met the 30 s criterion so were advanced to testing.

Testing occurred 24 h later for OR and 4 h later for OP, which are timepoints at which acute E₂ treatment previously enhanced memory relative to vehicle treatment in OVX E3FAD and E3/4FAD mice (Taxier et al., 2022a). During testing, one training object was replaced with a novel object (OR) or moved to the southwest or southeast corner (OP), and mice were again allowed to accumulate 30 s of total exploration. The time spent with novel or moved object, respectively, as well as time to accumulate 30 s of exploration, were quantified. Because mice have an inherent preference for novelty, those who remember the identity and location of the training objects should spend significantly more time than chance (15 s) with the novel and moved objects. Interactions with the objects were manually scored in ANY-maze (Stoelting Co.) by researchers blinded to genotype and treatment groups. To assess non-mnemonic performance factors during training, distance traveled (m) and velocity (m/s) were also analyzed. Between each mouse, the open field was cleaned with 70% ethanol.

As an additional measure of anxiety-like behaviors, elevated plus maze (EPM) testing was performed as described previously (Fleischer et al., 2021). Briefly, the EPM apparatus consisted of two open arms (30 cm x 5 cm) constructed from white Plexiglass with a clear lip (0.5 cm) attached to the sides to prevent mice from falling, two closed arms (30 cm x 5 cm x 15 cm) made of black Plexiglass walls, and a gray Plexiglass floor. The four arms converged in the center zone. Mice were placed in the center of the maze facing the open arm furthest from the experimenter and allowed to freely explore for 10 min. Entry into the open or closed arms was defined as the presence of all four paws within the arm; mice were considered to be in the center zone when at least one paw was outside of an arm. Number of entries into and time spent in the center zone and in each arm were quantified manually in Any-maze by a researcher blind to treatment and genotype. The EPM apparatus was cleaned with mild soap and water between each test.

A primary indicator of hot flashes is vasodilation, which can be measured in rodents by infrared imaging of tail skin temperature (T_Skin), which increases to release excess body heat. Injection of the NK3R agonist senktide induces rapid and transient neurokinin B-mediated vasodilation and is used to study the neural mechanisms underlying hot flashes (Krull et al., 2017; Krajewski-Hall et al., 2017; Padilla et al., 2018). We previously reported that long-term daily oral gavage with EGX358 reduced T_Skin in response to acute subcutaneous senktide injection in OVX wild type mice (Fleischer et al., 2021). Here, on the day of acute senktide injection (0.5 mg/kg), mice were placed in their home cage in a secondary container (10 in x 18 in x 10 in) beneath a thermal camera (E8, FLIR, Wilsonville, OR) for 1 h to acclimate to the novel environment and testing room. After acclimation, the home cage lid was removed and continuous thermal imaging was initiated. T_Skin was recorded for the next 10 min to establish a baseline and to minimize effects of stress or altered behavior due to cage top removal. Next, a researcher removed the mouse from its home cage and administered a subcutaneous injection of senktide. Upon returning the mouse to its cage, T_Skin was recorded for another 20 min. The cage top was then replaced and the mouse returned to the colony room.

T_Skin was quantified using FLIR Tools+ software according to previous work (Fleischer et al., 2021; Krull et al., 2017) by a researcher blind to treatment and genotype. T_Skin was measured by averaging tail temperature in a 1 cm line beginning 2 cm from the tail base. Baseline T_Skin measurements were taken immediately following the removal of the cage top and then 7.5, 5, and 2.5 min prior to injection. Following senktide injection, T_Skin was measured immediately after injection and every minute for 20 min. Change in T_Skin (ΔT_Skin) due to injection was analyzed, which was calculated as T_Skin Raw – T_Skin Baseline, where T_Skin Raw was the T_Skin quantification at a given time point, and T_Skin Baseline was the T_Skin at 2.5 min prior to injection, which allowed changes in temperature to be normalized to each mouse’s own pre-injection temperature (Krull et al., 2017; Krajewski-Hall et al., 2017).

Approximately 3 weeks after vasomotor testing, mice were weighed for one final time and euthanized via cervical dislocation. The uterus was immediately dissected and weighed (g).

All statistical analyses were conducted using GraphPad Prism 10 software (La Jolla, CA) as per our previously published methods (Fleischer et al., 2021; Taxier et al., 2022a). To assess within-group learning for OR and OP, one-sample t-tests were used to determine whether the time spent with each object during testing significantly differed from chance (15 s). Between-group differences in memory, all EPM measures, and uterine weights were assessed using two-way ANOVAs with treatment and genotype as between-subject variables. T_Skin and body weight measures were analyzed using three-way ANOVAs with the same two between-subject variables and time as the within-subject variable. Significant main effects were followed by Tukey’s post hoc tests for OR, OP, open field, and EPM data. For all measures except those dependent on time, outliers were defined as values more than two standard deviations from the mean were excluded from analysis. Statistical significance was set at p ≤ 0.05 for all statistical tests, and trends were considered as 0.05 ≤ p ≤ 0.10.

During OR training, only 3 mice failed to reach the 30 s exploration criterion (one E3FAD treated with vehicle, one E3FAD treated with EGX358, and one E3/4FAD mouse treated with vehicle) and only two mice did so during OP training (the same two E3FAD mice as in OR). When retrained, each mouse reached the 30 s exploration criterion so was advanced to testing and was included in all OR and OP analyses.

We previously showed that OVX E3FAD and E3/4FAD females did not differ in open field measurements of anxiety-like behavior or locomotor activity (Taxier et al., 2022c) and that EGX358 treatment did not affect anxiety-like behaviors in the open field or EPM in OVX wild-type mice (Fleischer et al., 2021). Nevertheless, to measure possible effects of EGX358 in E3FAD and E3/4FAD on anxiety or activity, we used the open field and EPM to measure anxiety-like and exploratory behaviors.

Because we previously found that long-term oral EGX358 administration reduced the magnitude of change in T_Skin in response to the NK3R agonist senktide in OVX wild-type C57BL/6 mice (Fleischer et al., 2021), we next assessed the extent to which APOE genotype might affect the ability of EGX358 to alleviate this drug-induced hot flash. Senktide caused a rapid and transient increase in T_Skin, such that T_Skin in all groups was significantly increased within 5 min and then returned to baseline within 15 min (F_{(20, 1,071)} = 106.4, p < 0.0001; Figure 4). This increase was especially pronounced in E3/4FAD females compared to E3FAD females (F_{(1, 1,071)} = 92.11, p < 0.0001) and in vehicle-treated mice relative to those treated with EGX358 (F_{(1, 1,071)} = 32.47, p < 0.0001). As illustrated in Figure 4, EGX358 significantly reduced the magnitude of the increase in T_Skin in E3FAD females but not in E3/4FAD females (Genotype x Treatment; F_{(1, 1,071)} = 5.466, p = 0.0196).

The primary goal of this study was to determine the extent to which the novel highly selective ERβ agonist EGX358 could enhance object recognition and spatial memory in EFAD mice homozygous for human APOE3 or heterozygous for human APOE3 and APOE4. We focused on E3FAD and E3/4FAD mice due to previous work from the Frick lab showing that an acute dorsal hippocampal infusion of E₂ increases CA1 dendritic spine density and enhances memory consolidation in the object recognition and object placement tasks among OVX E3FAD and E3/4FAD, but not E4FAD, mice (Taxier et al., 2022a). Accordingly, subsequent work from the LaDu lab showed that spatial memory in the Morris water maze was also improved by long-term oral administration of E₂ in OVX E3FAD and E3/4FAD, but not E4FAD, mice (Balu et al., 2024). Thus, we hypothesized that only E3FAD and E3/4FAD females would be responsive to an ERβ agonist like EGX358 in measures of spatial and object recognition memories.

In OVX female wild-type mice and rats, acute systemic injection or dorsal hippocampal infusion of commercial ERβ agonists like diarylpropionitrile (DPN) consistently enhance memory in tasks like object recognition and object placement (Boulware et al., 2013; Tuscher et al., 2015). Moreover, dorsal hippocampal infusion of an ERβ antagonist blocks memory consolidation in both tasks among OVX wild-type mice (Kim and Frick, 2017), suggesting that activation of ERβ in the dorsal hippocampus is both sufficient and necessary for object recognition and spatial memory formation. Consistent with these effects, we previously showed that long-term daily oral administration of EGX358 to OVX wild-type mice enhanced memory in both object tasks and reduced the magnitude of a senktide-induced hot flash to a similar extent as E₂ and DPN, without affecting anxiety- or depression-like behaviors or body weight (Fleischer et al., 2021). Here, we found that long-term daily oral treatment with EGX358 improved object recognition memory in OVX E3FAD and E3/4FAD mice relative to vehicle controls, which is consistent with our data from wild-type mice. In light of the amyloid burden and synaptic degeneration previously observed in 6 month-old female E3FAD and E3/4FAD mice (Youmans et al., 2012; Tai et al., 2014; Liu et al., 2015; Cacciottolo et al., 2016; Balu et al., 2023), these findings suggest the exciting possibility that ERβ agonist treatment may not only be beneficial for recognition memory in menopausal women without dementia, but also in APOE3/3+ and APOE3/4+ women with symptoms of cognitive decline.

In contrast to our previous wild-type findings, EGX358 did not enhance memory in the object placement test relative to chance or to vehicle controls. This null effect could be explained, in part, due to the unexpectedly strong memory shown by the vehicle-treated E3FAD group; in our prior studies, vehicle-treated OVX and gonadally-intact female E3FAD mice did not show evidence of learning in either object task (Taxier et al., 2022a,b). However, comparisons with chance indicate that EGX358 did not significantly increase the time spent with the moved object in either E3FAD or E3/4FAD females, suggesting no benefit to learning within each genotype. This finding contrasts with previously beneficial effects on memory of acute or chronic E₂ treatments in these genotypes (Taxier et al., 2022a; Balu et al., 2024). The most obvious explanation is that spatial memory as assessed by object placement is not sensitive to ERβ agonism in mice with established AD-like pathologies including high levels of Aβ42, oligomeric Aβ, amyloid deposition, phospho-tau, gliosis, and synaptic degeneration. However, that conclusion is premature, as the effects of treatment could not truly be assessed here due to the performance of vehicle-treated E3FAD mice. In addition, methodological factors like drug dose and method of delivery (hydrogel here vs. gavage in our previous work) may also play a role. Future studies should re-examine effects of EGX358 on OP memory in EFAD mice and address whether spatial memory in OVX E3FAD and E3/4FAD mice might be enhanced by higher doses of EGX358 or a more controlled method of administration (e.g., gavage).

As in our previous work (Fleischer et al., 2021), we modeled a hot flash in this study using the selective neurokinin-3 tachykinin receptor agonist senktide. Senktide mimics the hormone neurokinin B, which is released by kisspeptin, neurokinin B, and dynorphin (KNDy) neurons in the hypothalamus. It causes vasodilation by binding neurokinin-3 tachykinin receptors and activating warm-sensitive neurons in the ventromedial preoptic area, resulting in rapid and transient increases in tail skin temperature in OVX mice and gonadally-intact mice of both sexes (Krull et al., 2017; Krajewski-Hall et al., 2017). In our hands, subcutaneous injection of 0.5 mg/kg senktide in OVX wild-type mice produces a T_Skin increase of >3^o within 5 min that returns to baseline within about 15 min, and the magnitude of this increase is reduced similarly by E₂, DPN, and EGX358 (Fleischer et al., 2021). In this study, senktide injection caused a rapid and transient increase in T_Skin among mice of both genotypes, although the increase was generally of a somewhat higher magnitude (>4^o) and lasted longer compared to wild-type mice. Importantly, however, EGX358 treatment reduced the magnitude of the change in T_Skin among E3FAD, but not E3/4FAD, mice. These findings suggest that presence of even a single APOE4 allele is sufficient to block the beneficial effects of long-term ERβ agonism in a preclinical model of menopause-related hot flashes. This result is consistent with clinical data showing that phytoSERM compounds – plant-derived selective estrogen receptor modulators – that preferentially bind to ERβ, including genistein, daidzein, and S-equol, reduced the frequency of hot flashes in APOE3+, but not APOE3/4+, women (Wang et al., 2020).

Our results also align with a previous study of metabolic profiles in male mice expressing human APOE3 or APOE4. APOE4 mice had increased lipid metabolism, increased body temperature, and more metabolically active brown adipose tissue (Arbones-Mainar et al., 2016). A handful of studies have examined neurokinin signaling in rodent models of dementia and found that senktide reversed spatial memory deficits in male Wistar rats infused with Aβ1-42 (Koca et al., 2023), and reduced dentate gyrus circuit disorders in socially-isolated male 3x Tg-AD mice (Huang et al., 2023), possibly due to the promotion of forebrain cholinergic activity by neurokinin 3 signaling (de Souza Silva et al., 2013). However, to the best of our knowledge, this is the first study to examine whether APOE genotype affects the vasomotor response to senktide or any NK3 receptor agonist in male or female rodent models of AD. In a small observational study in middle-aged women, hot flashes were linked to worse verbal memory, and the degree of impairment may have been affected by secondary symptoms, such as worsening sleep quality (Maki et al., 2008). More recently, higher levels of objectively assessed vasomotor symptoms, particularly at night, were associated with higher plasma biomarkers of amyloid plaques (Thurston et al., 2024). Thus, the APOE genotype differences observed here in response to senktide injection suggest a critical need to more fully examine how symptoms associated with the menopausal transition, such as hot flashes, may contribute to cognitive decline and cognitive response to estrogen therapy.

We also examined anxiety-like behaviors in the open field and elevated plus maze, given a previous report from the Frick lab that EFAD females with two copies of APOE4 (E4FAD) exhibited greater levels of anxiety-like behavior in the open field (less time in the center) relative to E3FAD females (Taxier et al., 2022c). Thus, we thought it possible that E3/4FAD females might be more anxious than E3FAD females. Although ERβ agonism has been reported to lower anxiety-related phenotypes in mice and rats (Rocha et al., 2005; Walf and Frye, 2005; Walf and Frye, 2007), other studies report no effects of ERβ agonism on anxiety- and depression-like behaviors in female mice (Eid et al., 2020). Likewise, long-term oral gavage of EGX358 did not alter exploratory behaviors in the open field or EPM tests among wild-type mice (Fleischer et al., 2021), so we did not expect treatment effects in either genotype. Consistent with this hypothesis, EGX358 did not affect any measure of open field or elevated plus maze behavior. However, we surprisingly found that E3/4FAD mice spent more time in the center of the open field than their E3FAD counterparts, irrespective of treatment, suggesting less anxiety-like behavior in mice with one copy of APOE4. This result is somewhat perplexing given that two copies of APOE4 previously reduced center time in the open field (Taxier et al., 2022c). Interestingly, neither treatment nor genotype affected activity in the elevated plus maze, another measure of anxiety-like behavior, so the extent to which APOE genotype more generally modulates anxiety-like and exploratory behaviors is unclear.

It should be noted that we took great care in our treatment and handling protocols to minimize chronic stress in our mice, which perhaps partially explains the modest anxiety-related results. In human studies, APOE4 carriers are more likely to have anxiety than non-carriers (Holmes et al., 2016; Xu et al., 2023), so our results are unexpected in this light. Regarding EGX358, the lack of effects in the open field and plus maze could indicate that ERβ is not involved in regulating anxiety-like behaviors among individuals with established brain amyloid, phospho-tau burden, and AD-related neurodegeneration, and may not be an effective therapy for mood-related symptoms in AD patients. However, considerably more work will be needed to fully understand the clinical applicability of these compounds for anxiety- and depression-related behaviors in AD patients. Critically, EGX358 did not produce any adverse effects on anxiety-like behaviors in either of our tests, suggesting that it could be used as a therapeutic for memory and hot flashes without negatively impacting anxiety.

Estrogens such as E₂ regulate energy balance and homeostasis, and their loss at menopause and after OVX in rodents is associated with increased fat mass and body weight (Mauvais-Jarvis et al., 2013; Palacios et al., 2024; Roesch, 2006). Thus, we assessed the effects of EGX358 treatment on body weights throughout the experiment. As in our previous work with long-term oral EGX358 treatment in wild-type female mice (Fleischer et al., 2021), EGX358 in this study did not influence body weights throughout the 10 weeks of treatment. This null effect is consistent with studies showing that neither long-term ERβ agonist treatment nor knockout of ERβ prevent OVX-induced increases in body weights in rodents (Roesch, 2006; Opas et al., 2009; Van Pelt et al., 2015). Indeed, studies of knockout mice and ERα agonists indicate that the effects of estrogens on body weight are largely driven by ERα (Roesch, 2006; Van Pelt et al., 2015; Wegorzewska et al., 2008), so the inability of EGX358 to counter the lipogenic effects of OVX are not surprising. Nevertheless, the present data are an important confirmation of our earlier work that long-term oral EGX358 treatment has no adverse effects on body weight, which suggests that EGX358 treatment in humans would not likely cause undue weight gain or loss.

In contrast to treatment, APOE genotype significantly affected body weights. Although weekly body weights increased throughout the study in all groups, as would be expected after OVX, E3/4FAD mice in both treatment groups gained significantly more weight during the course of the experiment than both E3FAD groups. All groups started at approximately the same mean body weight, but weights increased nearly two grams in both E3/4FAD groups during week 2 and this genotype difference continued for the remainder of the study. A recent study in young adult female E3FAD and E4FAD mice found that E4FADs actually gained slightly less weight over the course of the 12 week study, but had a higher percentage of visceral body fat (Christensen and Pike, 2023). Additionally, E4FADs had higher blood levels of insulin, leptin, and total cholesterol, which are associated with metabolic impairments (Christensen and Pike, 2023). Another study in APOE knock-in mice found no effect of genotype in females on weight gain over the 12 week study, both in high-fat and low-fat diets (Jones et al., 2019). Because all mice used in our study were ovariectomized, APOE4 may have amplified the effects of ovarian hormone loss on metabolism and adiposity, causing OVX E3/4FADs to gain more weight than E3FADs. In humans, hypothalamic atrophy in AD patients is thought to lead to insulin and leptin resistance, ultimately leading to weight loss as the disease progresses (Ishii and Iadecola, 2015; Vercruysse et al., 2018). This loss occurs earlier and to a greater extent in AD patients that express at least one copy of APOE4, particularly women APOE4 carriers (Ukraintseva et al., 2024; Ando et al., 2022; Vanhanen et al., 2001). Thus, the weight gain observed in E3/4FAD mice is not consistent with clinical findings. Although OVX is an imperfect model of menopause, the abrupt cessation of circulating ovarian hormones following surgery has repeatedly been shown to mimic the menopause-associated gain in weight, and estrogen treatments consistently ameliorate this weight gain both in women and OVX female rodents (Witte et al., 2010; Mamounis et al., 2014). However, widespread brain alterations in AD must impact the interplay between metabolic function and gonadal hormones in ways that differ from aging or menopause alone, so future studies will be needed to better understand the disparate effects of OVX on body weight in E3/4FAD mice compared to menopause in APOE4+ women with AD.

Finally, we measured effects of EGX358 treatment on uterine weights given the well-known uterotrophic effects of estrogen treatments (Modder et al., 2004; Zhu and Pollard, 2007). Although a significant main effect of treatment was observed, such that EGX358 treatment was associated with increased uterine weights, all weights remained considerably less than those of gonadally-intact C57BL/6 mice (typically ~150 mg). In one study in which uterine weights of sham-operated C57BL/6 females averaged 150 mg, OVX reduced uterine weights to a mean of 50 mg and 60 days of E₂ treatment dose-dependently increased weights to between 150–200 mg (Modder et al., 2004). As such, the EGX358-induced increase here, although statistically significant, is not reflective of uterine hypertrophy. Indeed, this increase somewhat mitigates the atrophic state of the uterus after ovary removal, which may be beneficial for uterine health.