Loss of ovarian hormones is detrimental in early disease stages of mouse models of Alzheimer’s disease and multi-etiology dementia

All experiments were approved by the Albany Medical College Animal Care and Use Committee and in compliance with the ARRIVE guidelines. App^NL−F (App^tm2.1Tcs/App^tm2.1Tcs) mice [38] were acquired congenic on a C57BL/6 J background from Dr. John Cirrito at Washington University following MTA from Riken, Japan. These mice were bred in house and were fed a standard chow diet (Purina Lab Diet 5P76) for the duration of this study. Female mice of about 5 months of age were housed (3–4 per cage) at 21 °C, 30–70% humidity, with a 12 h light/dark cycle. To model surgical menopause, which usually happens in younger women compared to natural menopause which occurs in middle aged ones, cages of mice were randomly selected to receive either an ovariectomy (OVX) or a sham surgery (Intact). Mice were allowed to rest for 3 weeks. Within each group half the cages were subjected to a unilateral carotid artery occlusion surgery (UCCAO), a model of VCID, inducing chronic cerebral hypoperfusion in the right hemisphere to model MED. Control animals (AD only model) received sham surgery. One month later, mice were subjected to a battery of behavioral tests for 2 weeks. At the end of the study (mice ~ 7.5 months old), blood flow measurement was collected using laser speckle imaging as described previously [35, 39, 40]. Mice were then euthanized by deep anesthesia followed by cardiac puncture to collect plasma then perfusion with ice cold heparinized saline. Brains were collected and either post-fixed for IHC or regionally dissected and flash frozen for molecular biology. One mouse in the Intact Sham group presented with a growth on its ovary during tissue collection and was removed from the study. Figure 1A shows the experimental groups and timeline.

Bilateral ovariectomy was performed under isoflurane anesthesia as described previously [41]. Control mice received a sham surgery and were left gonadally intact (Intact group).

UCCAO or sham surgeries were performed as described previously [35, 39, 40] 3 weeks after OVX surgeries to allow enough recovery time. Animals were allowed to recover for 4 weeks before behavior testing started.

Locomotor activity was evaluated using an open field test, as previously described [35, 36, 39–41]. Mice were placed in a square arena (45 × 45 cm) and allowed to explore freely for 10 min, then removed and placed in a “recovery cage” so as not to expose naïve cage mates to them. Videos were analyzed using ANY-maze software (Stoelting, Wood Dale, Illinois, USA). Total distance traveled (m) was measured as a proxy for locomotor activity. Intact Sham n = 13 mice, Intact VCID n = 15 mice, OVX Sham n = 13 mice, OVX VCID n = 14 mice.

Spatial recognition and object recognition memory were tested back-to-back using the combined test as described before [36, 41]. A spatial cue (a horizontal piece of black tape) placed on the wall of the open field arena where objects were initially placed. The test consisted of three 10 min trials: a “training” trial, the object place recognition test (OPRT) trial, and the novel object recognition test (NORT) trial, with an inter-trial interval of one and a half hours. Mice that did not interact with both objects during “training” trial, or those that didn’t interact with the objects for at least 20 s during the 10 min testing were excluded from analysis. For OPRT: Intact Sham n = 9 mice, Intact VCID n = 12 mice, OVX Sham n = 10 mice, OVX VCID n = 9 mice. For NORT: Intact Sham n = 11 mice, Intact VCID n = 13 mice, OVX Sham n = 10 mice, OVX VCID n = 10 mice.

The Barnes maze test, used to evaluate hippocampal-dependent spatial learning and memory, was performed using as previously described protocol [35, 36]. Mice were trained to find an escape pod located under one of the 20 holes in the brightly lit maze (Stoelting, Catalog # 60,170), using spatial cues in the room. Mice were given two training trials a day (3 min each or until the mouse found the escape pod) and 3 h between trials during the same day for 4 consecutive days to test spatial learning. Spatial memory was tested on the sixth day (1 day break) by allowing the mice to explore the maze for 2 min. Mice that failed to learn the position of the escape hole were excluded from the memory test analysis (one mouse excluded in each group). Video recordings for the probe trial of some mice were lost due to a computer malfunction resulting in an exclusion of these mice from the memory test (but not learning) analysis. This resulted in the following n numbers: for spatial learning: Intact Sham n = 12 mice, Intact VCID n = 15 mice, OVX Sham n = 13 mice, OVX VCID n = 14 mice; for spatial memory: Intact Sham n = 8 mice, Intact VCID n = 10 mice, OVX Sham n = 9 mice, OVX VCID n = 10 mice.

The nest building test was used to evaluate performance of activities of daily living as previously described [35, 40, 41]. Mice were provided with intact nestlet material and individually housed overnight with food and water. The next day (16 h later), mice were carefully removed from the cages and returned to their group home cage. The nests were scored by three independent scorers, blinded to the mouse ID and experimental group, using a scale of 1–5 in 0.5-point increments [42], with 1 being the lowest possible score and 5 being the highest. For each mouse the final value was the average of the three scorers. Intact Sham n = 13 mice, Intact VCID n = 15 mice, OVX Sham n = 13 mice, OVX VCID n = 14 mice.

As previously described, frozen hippocampi were homogenized in 200 µL of T-Per buffer (78,510, ThermoFisher Scientific) supplemented with protease and phosphatase inhibitor cocktail (HALT, 1,861,284 ThermoFisher Scientific) and spun at 21,000 g for 20 min at 4◦C. The supernatant containing soluble proteins was removed and used to quantify total soluble protein levels using a Bicinchoninic Acid (BCA) protein assay according to manufacturer’s instructions (23,227, ThermoFisher Scientific). The pellet containing insoluble proteins was resuspended in 100 µL of 70% formic acid and stored at −80◦C. On the day of the enzyme-linked immunosorbent assay (ELISA), the samples were thawed, and 2 mL (20x) of neutralizing solution (1 M Tris base, 0.5 M Na2HPO4, 0.05% NaN3) were added. The neutralized sample solution was processed for Aβ quantification using the Human Aβ40 ELISA (KHB3481, ThermoFisher Scientific) and Human Aβ42 ELISA (KHB3441, ThermoFisher Scientific) kits according to the manufacturer’s instructions. Outliers were identified using Grubb’s test and excluded if affecting normality. Exclusions for Aβ40: 1 mouse in the Intact Sham and 1 in the OVX Sham group. Exclusions for Aβ42: 1 mouse in the Intact Sham and 1 in the Intact VCID group.

Brains used for immunofluorescent labeling were fixed overnight in a 4% paraformaldehyde (PFA) solution, then cryoprotected in 30% sucrose, then frozen in optimal cutting temperature (O.C.T.) solution (23–730-571, ThermoFisher Scientific) and stored at −80 °C until further processing. Using a cryostat (Cm1950, Leica), 35 μm-thick sections were obtained and stored at 4 °C. A 25G needle was used to mark the left side of the brain contralateral to the hypoperfusion side. Immunofluorescent labeling was done as previously described [39, 43, 44]. Briefly, slices were permeabilized and blocked, using a 0.3% PBS with triton (TPBS) and 5% donkey serum solution, for one hour at room temperature. Primary antibodies were applied overnight at 4 °C in blocking solution. After 3 washes in PBS, the corresponding secondary antibodies and DAPI (1:1000; D1306, Thermo Fisher Scientific) were added in blocking buffer for 1 h at room temperature before mounting between slides and coverslip using Fluoromount-G™ mounting medium (0100–01; Southerbiotech). Images of brain slices were obtained at 10 × magnification using the Axio Observer Fluorescent Microscope (Carl Zeiss Microscopy, Oberkochen, Germany).

Antibodies used for the microglia panel: rabbit anti goat anti-Iba-1 (1:1000; PA5-18,039, lot # ZD4305091, Invitrogen) and rat anti-CD68 (1:1000; MCA1957, lot #1708, Bio-Rad). Rhodamine Red-X donkey anti-rat (1:1000; 712–295-150 Jackson ImmunoResearch) and Alexa Fluor 647 donkey anti-goat (1:1000; 705–605-147, Jackson ImmunoResearch). To label astrocytes, we used a rat-GFAP (Glial Fibrillary Acidic Protein) antibody (1:2500, 13–0300, lot #T1275011, Invitrogen) and Rhodamine Red-X donkey anti-rat (1:1000; 712–295-150 Jackson ImmunoResearch).

All measurements were performed by experimenters blinded to treatment conditions using ImageJ (NIH) software. For each animal about 4 brain sections were analyzed and averaged. Regions of interest (ROIs) were drawn on the right hemisphere around the retrosplenial cortex (Rsp Ctx), which is one of the first areas to show functional impairment in AD and plays a key role in memory and spatial learning [45]; the area around the entorhinal cortex (Ent Ctx) which is often the earliest area to show histological alterations in AD and is involved in processing spatial and non-spatial information [46]; and around the hippocampus (Hipc), specifically the cornu ammonis regions 1 (CA1) and 3 (CA3) areas which are is known to be severely affected in AD and are essential for memory and spatial navigation [47].

Iba1 and CD68 positive cells were counted using ImageJ (NIH) software and classified based on ramified morphology and CD68 expression as described previously [37, 48, 49]. For astrocytes, in addition to the area covered, GFAP fluorescent staining intensity (IntDen) was measured in all ROIs using the raw unmodified/non- thresholded images, and GFAP positive cells were counted. For the CA1 and CA3 regions of the hippocampus, an r-score for astrocyte reactivity was calculated for each animal as described in Luijerink et al. [50]. The average score of two independent experimenters that were blinded to treatment group was used for statistical analysis. R-score could not be calculated in the cortical areas as cortical astrocytes express very low, even undetectable, levels GFAP under physiological conditions and only significantly increases only under reactive or pathological conditions [51–54]. Outliers were identified using Grubb’s test and excluded if affecting normality. For GFAP in the IntDen in the entorhinal cortex 1 mouse in the Intact VCID group, for the %area in the CA1: 1 mouse in the OVX VCID group; for the entorhinal cortex: 1 mouse in the Intact Sham and 1 in the Intact VCID group.

Statistical analyses were completed using GraphPad Prism (GraphPad Software, San Diego, CA, USA). Data were evaluated for normal distribution and statistical outliers were identified using Grubbs’ test and removed if they affected normality which is indicated in the methods above. Data were analyzed using a 2-way ANOVA followed by Tukey’s post hoc test or 3-way repeated measures ANOVAs were used for measure tracked over time. Statistical significance was set at p < 0.05. Data are expressed as mean + SEM. Only main effects are presented on graphs unless an interaction effect is observed.

We have previously shown that loss of estrogens in a mouse model of menopause leads to exacerbated cognitive decline caused by VCID in WT female mice [35]. We sought to test whether this also occurs in AD and MED mouse models. We used App^NL−F mice that were either gonadally intact controls or OVX. Mice also received either sham surgery (AD only) or VCID-inducing surgery (MED). One month later, mice underwent a battery of behavioral tests (Fig. 1A). We validated our MED model by assessing decreased blood flow in the right side of the brain at the end of the experimental timeline. Using laser speckle imaging, we found that the VCID inducing surgery led to a sustained decrease in blood flow regardless of OVX ~ 7 weeks post-surgery (Fig. 1B, main effect of VCID p = 0.0376, F (1, 33) = 4.691). We next assessed the behavioral effects of OVX in both dementia models. We did not find any significant differences in locomotor activity using the open field test (Fig. 2A), or in spatial recognition memory using the novel object place test (Fig. 2B), or in object recognition memory using the novel object recognition test (Fig. 2C). However, we found changes in spatial learning and memory as tested by the Barnes maze test (Fig. 2D-I). We found that the loss of ovarian hormones led to impaired spatial learning, as evidenced by OVX mice having a higher number of errors (Fig. 2D, main effect of OVX p = 0.0324, F (1, 50) = 4.841), and a worse path efficiency to 1st entry (Fig. 2E, main effect of OVX p = 0.0031, F (1, 50) = 9.663). Cerebral hypoperfusion, resulting in MED, impaired spatial learning as well, which is supported by a longer latency of MED mice to find the escape hole (Fig. 2F, main effect of VCID p = 0.0446, F (1, 50) = 4.247) and a trend in having a worse path efficiency during the hidden trials (Fig. 2E, main effect of VCID p = 0.0556, F (1, 50) = 3.841). Spatial memory was impaired by loss of ovarian hormones as evident by a main effect of OVX during the probe trial of Barnes maze on latency to first entry (Fig. 2G, main effect of OVX p = 0.0302, F (1, 33) = 5.131), and the percentage of time spent in the target cone (Fig. 2H, main effect of OVX p = 0.0279, F (1, 32) = 5.306). Some OVX effects were dependent on dementia type as we found a significant OVX x VCID interaction on the percentage of time spent in the target cone (Fig. 2H, interaction p = 0.0090, F (1, 32) = 7.747) and the percentage of errors made throughout the probe trial (Fig. 2I, interaction p = 0.0155, F (1, 32) = 6.535). In both cases, MED OVX mice performed worse than AD OVX mice as indicated by Tukey’s post hoc test (Intact:VCID vs. OVX:VCID p = 0.0042 Fig. 2H and p = 0.0167 Fig. 2I respectively). The negative effects of the loss of ovarian hormones extended to the activities of daily living. This was supported by a main effect of OVX to reduce nest scores (Fig. 2J, main effect of OVX p = 0.0300, F (1, 51) = 4.985) in the nest building test (an assessment of activities of daily living). These results show that the loss of ovarian hormones impairs spatial learning and memory as well as activities of daily living in both AD and MED models.

Starting in early AD stages, insoluble β-amyloid peptides start to accumulate and agglomerate leading to plaque formation [55]. We used ELISA to quantify Aβ40 and Aβ42 peptides in the insoluble fraction of proteins extracted from the hippocampus and the cortex. In the hippocampus, we observed an OVX x VCID interaction [p = 0.0120, F (1, 22) = 7.506] wherein either VCID on its own compared to intact sham mice (Tukey’s post hoc test, Intact:Sham vs. Intact:VCID p = 0.0188), or OVX on its own compared to intact sham mice (Tukey’s post hoc test, Intact:Sham vs. OVX:Sham p = 0.0396) led to an increase in insoluble Aβ40 levels (Fig. 3A). For Aβ42 (Fig. 3B), we observed a trend in the main effect of cerebral hypoperfusion towards increased levels [main effect of VCID p = 0.0657, F (1, 22) = 3.751]. These results show that the loss of ovarian hormones seems to have selective effects on Aβ peptides.

In both AD patients and rodent models of AD, astrocytes can become reactive even before the formation of amyloid plaques [56–59]. Astrocyte reactivity, termed astrogliosis, is characterized by morphological changes (hypertrophy, increased volume and branching) and upregulation of GFAP expression [57, 58]. Using immunofluorescent labelling, we quantified the intensity (InDen) and coverage of the GFAP labeling to assess astrogliosis in several ROIs of the brain (Fig. 4A-B). We found an increase in both GFAP labeling intensity (IntDen) and area covered (%) in the CA3 region of the hippocampus of OVX mice regardless of dementia type [Fig. 4C, main effect of OVX p = 0.0252, F (1, 21) = 5.805, and Fig. 4D p = 0.0324, F (1, 21) = 5.248, respectively]. No changes were found in the other regions studied i.e. the CA1, EntCtx, and RspCtx (Fig. 4C-D). Changes in GFAP cell density and R-scores did not reach statistical significance levels in any region (Sup. Figure 1). The increases in GFAP labeling intensity and area indicate a potential increase in astrogliosis due to loss of ovarian hormones limited to the CA3 area.

Microglia, a major part of the innate immune system of the CNS, are emerging as potential therapeutic targets for AD [60, 61]. Findings using AD mouse models show that microglia reactivity in the early stages of disease is a complex process that can elicit both protective and detrimental effects [61–65]. We assessed microglia reactivity in the CA1 and CA3 of the hippocampus, the entorhinal cortex and the retrosplenial cortex using immunolabeling for Iba1 and CD68 (Fig. 5A-B). In the CA1, VCID led to an increase in Iba1 positive cell density [Fig. 5C CA1, main effect of VCID, p = 0.0436, F (1, 20) = 4.640]; no changes were observed in the percentage of activated microglia (non-ramified Iba1 +/CD68 + cells, Fig. 5D CA1). In the CA3, we did not detect changes in Iba1 positive cell density (Fig. 5C CA3); however, OVX led to an increase in the percentage of activated microglia [Fig. 5D CA3, main effect of OVX, p = 0.0318, F (1, 19) = 5.373]. In the EntCtx, VCID led to a decrease in Iba1 positive cell density [Fig. 5C EntCtx, main effect of VCID, p-0.0276, F (1, 20) = 5.646]; OVX led to a trend in decrease in the percentage of activated microglia (Fig. 5D EntCtx, effect of OVX, p = 0.0697). In the RspCtx, we observed an interaction in the effect of the two variables on Iba1 cell density [Fig. 5C RspCtx, OVX x VCID interaction, p = 0.0019, F (1, 20) = 12.85], wherein a combination of VCID and OVX led to a higher Iba1 cell density in the RspCtx of OVX:VCID animals compared to those that received OVX or VCID on their own (Tukey’s post hoc test, Intact:VCID vs. OVX:VCID p = 0.0301; OVX:Sham vs. OVX:VCID p = 0.0234). OVX led to a decrease in the percentage of activated microglia [Fig. 5D RspCtx, effect of OVX, p = 0.0184, F (1, 20) = 6.590]. All together these results indicate that in female App^NL−F mice, either cerebral hypoperfusion and/or loss of ovarian hormones lead to impaired microglia reactivity with regions specific effects.

Not applicable.

Not applicable.

The authors declare no competing interests.

The datasets acquired and/or analyzed during the current study are available from the corresponding author on reasonable request.