Long-term ovarian hormone deprivation alters functional connectivity, brain neurochemical profile and white matter integrity in the Tg2576 amyloid mouse model of Alzheimer’s disease

Abstract

Premenopausal bilateral ovariectomy is considered to be one of the risk factors of Alzheimer’s disease (AD). However, the underlying mechanisms remain unclear. Here, we aimed to investigate long-term neurological consequences of ovariectomy in a rodent AD model, TG2576 (TG), and wild-type mice (WT) that underwent an ovariectomy or sham-operation, using in vivo MRI biomarkers. An increase in osmoregulation and energy metabolism biomarkers in the hypothalamus, a decrease in white matter integrity, and a decrease in the resting-state functional connectivity was observed in ovariectomized TG mice compared to sham-operated TG mice. In addition, we observed an increase in functional connectivity in ovariectomized WT mice compared to sham-operated WT mice. Furthermore, genotype (TG vs. WT) effects on imaging markers and GFAP immunoreactivity levels were observed, but there was no effect of interaction (Genotype × Surgery) on amyloid-beta-and GFAP immunoreactivity levels. Taken together, our results indicated that both genotype and ovariectomy alters imaging biomarkers associated with AD.

1. Introduction

Women are at higher risk for Alzheimer’s disease (AD) than men, but the underlying biological mechanisms are not fully understood (Beam et al., 2018). Studying sex-specific risk factors such as ovariectomy might increase our understanding on sex differences in the risk of AD. Bilateral ovariectomy (i.e., bilateral ovarian removal) is a widely used surgical procedure in women to prevent ovarian and breast cancer. Epidemiological and molecular studies have provided evidence that premenopausal ovariectomy is associated with an increased risk of accelerated aging, multimorbidity, and dementia (Levine et al., 2016; Rocca et al., 2007; Rocca et al., 2016). However, the long-term neurological consequences of ovariectomy on the progression of AD pathology remain to be investigated. Therefore, in this preclinical in vivo neuroimaging study we investigated how bilateral ovariectomy impacts neurological correlates of AD pathophysiology.

Women undergoing ovariectomy, which leads disruption of feedback mechanisms of the hypothalamic-pituitary-gonadal (HPG) axis, demonstrate physiologically lower levels of ovarian hormones such as estrogen than age-matched postmenopausal women (Korse et al., 2009). Estrogen has neuromodulatory and neuroprotective functions in the brain (Markowska and Savonenko, 2002). For example, estrogen demonstrates anti-inflammatory activity (Vegeto et al., 2001), regulates of toxic amyloid-beta (Aβ) clearance (Jayaraman et al., 2012), regulates synaptic integrity (Li et al., 2004), and exerts antioxidant effects (Moosmann and Behl, 1999). While these studies demonstrated that estrogen is important for neuroprotection, the effects of long-term estrogen deprivation following ovariectomy on brain integrity are unclear. Preclinical animal models may shed light into effects of ovariectomy on brain function, metabolism and structure.

Bilateral ovariectomy, performed in cycling young adult rodents, results in a decline in bone mass and activation of compensatory neuroendocrine responses, such as increased luteinizing hormone in response to decreased levels of estrogen, which mimics physiological responses occurring after premenopausal ovariectomy in humans (Koebele and Bimonte-Nelson, 2016). This model has been widely used to increase our understanding of the behavioral, cellular and molecular alterations associated with ovarian hormone loss and HPG axis dysregulation (Agca et al., 2020).

Current evidence suggests that the loss of ovarian function after ovariectomy may alter white matter integrity, energy metabolism, inflammatory response, structural, and functional integrity (Anckaerts et al., 2019b; Benedusi et al., 2012; Luo et al., 2016; Zeydan et al., 2019). Detecting these changes in vivo is challenging. Only two studies using magnetic resonance techniques reported long term structural and functional alterations after ovariectomy in female wild-type (C57BL/6J) mice and women (Anckaerts et al., 2019b; Zeydan et al., 2019). In vivo imaging studies investigating the long-term effects of loss of ovarian function on brain integrity in transgenic mouse models that mimic a range of AD-related pathologies are missing and may clarify the mechanistic underpinnings of these outcomes. In this study, we employed magnetic resonance spectroscopy (MRS), resting state functional MRI (rsfMRI) and diffusion tensor imaging (DTI) to respectively unravel metabolic, functional and structural alterations associated with ovarian hormone loss and its interaction with AD pathophysiology (Biswal et al., 1995; Kantarci et al., 2017; Kara et al., 2011). The hippocampus, cortex, and hypothalamus brain regions were selected to represent parts of the brain regions known to be affected during the AD process and after estrogen depletion (Morrison et al., 2006). These regions have been verified to be implicated in estrogen signaling, as the expression of estrogen receptors have been demonstrated in these regions (Mitra et al., 2003).

The present study was devised to clarify the impact of long-term loss of ovarian hormones on in vivo neuroimaging biomarkers, amyloid load and neuroinflammatory markers in Tg2576 and wild-type mice. Although no animal model can reproduce the full human symptoms of AD, transgenic mouse models developing one or more AD related pathology increase our understanding about the disease. The Tg2576 mice, which develop progressive amyloid pathology, exhibit learning and memory impairment, elevated level amyloid peptides and deposits (Aβ40/42), and reactive gliosis which are in line with the biochemical and epidemiological findings of AD (Hsiao et al., 1996). Our hypothesis was that ovariectomy would accelerate the neuroinflammatory markers and amyloid pathology in Tg2576 mice and correlate with related neurological alterations.

2. Material and methods

2.1. Animal care and ethical statement

All mice used in this experiment were housed in groups, and had free access to water, and fed ad libitum with standard rodent diet. Following the ovariectomy and sham operations, animals were shifted to an estrogen-free diet (i.e., low phytoestrogen content Sniff R/M-H rodent diet, www.ssniff.de, Germany). All procedures were performed in strict accordance with the European Directive 2010/63/EU on the protection of animals used for scientific purposes. The protocols were approved by the Committee on Animal Care and Use at the University of Antwerp, Belgium (permit numbers 2014-54 and 2014-86) and all efforts were made to ensure animal well-being.

2.2. Animals

In this study transgenic Tg2576 (TG) mice, which were developed and described by Hsiao et al. (1996). TG mice are on C57BL/6xSJL background and express a human amyloid precursor protein 695 (APP₆₉₅₎ with the Swedish double mutation (K670N, M671L) driven by the hamster prion protein promoter resulting in elevated levels of Aβ and ultimately amyloid plaques (Hsiao et al., 1996). C57BL/6xSJL mice were used as wild-type (WT) controls. Founder mice were provided by K. Hsiao-Ashe. At the age of four weeks, transgenicity was identified by polymerase chain reaction of tail DNA as described elsewhere (Hsiao et al., 1996).

2.3. Experimental design

The summary of the experimental set up is depicted in Fig. 1. Prior to operation, WT mice (3-month-old, N = 27) were used to extract reference resting state brain networks (see Supplementary Methods (1.6.4. rsfMRI data analysis) for further details). TG (N = 24) and WT (N = 27) mice were subjected to ovariectomy or sham-surgery at 3.5 months of age (14 weeks ± 1 week) as described earlier (Sophocleous and Idris, 2014) (see Supplementary Methods 1.1. Ovariectomy/sham-operation). At 18 months of age (± 2 weeks), ovariectomized or sham- operated WT and TG mice were scanned using in vivo MR imaging (DTI, rsfMRI) and MR spectroscopy. After in vivo scans and blood collection, all mice were euthanized for collecting brain tissues. Some animals were excluded from the study/analysis either due to the presence of imaging artifacts or the occurrence of death prior to end-point measurements. The definitive information on the number of animals included in the data analysis is depicted under relevant figure legends or tables.

Fig. 1.

Presurgery groups (TG and WT) were either ovariectomized (OVX) or sham-operated, at age 3.5 months. The 18-month-old postsurgery groups (TGOVX, N = 7; TGSHAM, N = 6; WTOVX, N = 11; WTSHAM, N = 9) were scanned with MRI, and a blood sample was acquired. After the MRI scan, mice were euthanized for tissue collection for immunohistochemistry. N, number of animals; DTI, diffusion tensor imaging; 1H MRS, proton magnetic resonance spectroscopy; rsfMRI, resting state functional magnetic resonance imaging; M, Month; TG, Tg2576 mice; WT, wild-type mice (C57BL/6 × SJL); TGOVX, transgenic ovariectomized; TGSHAM, transgenic sham-operated; WTOVX, wild-type ovariectomized; WTSHAM, wild-type sham-operated; LH, luteinizing hormone.

2.4. MRI procedures

2.4.1. Animal handling

While DTI and rsfMRI data acquisitions were performed on the same day, MRS data were acquired on another day in order to reduce the time of exposure of mice to isoflurane. For all MR experiments mice were initially anesthetized with 3.5% isoflurane (IsoFlo, Abbott, Illinois, USA) in 66% nitrogen (N₂) (400cc/min) and 33% oxygen (O₂) (200 cc/min) in an induction box. For the MRS acquisition all mice were maintained at 1.5–2% isoflurane in a mixture of 66% N₂ and 33% O₂ (Orije et al., 2015). During the rsfMRI measurements, sedation of animals was performed using a combination of isoflurane and medetomidine (Domitor). Before initiating the rsfMRI experiments, mice received a bolus injection of medetomidine (0.3 mg Domitor/kg BW, Pfizer, Karlsruhe, Germany) via the subcutaneous route (Shah et al., 2015; Shah D. et al., 2016). Following medetomidine administration, the isoflurane level was decreased immediately to 1%. Fifteen minutes and twenty after bolus injection the level of isoflurane was adjusted to 0.8% and 0.6%, respectively. Ten minutes before the rsfMRI measurements, isoflurane was decreased to 0.4%. RsfMRI scans were consistently acquired 40 minutes after the bolus injection, and isoflurane level was maintained at 0.4% during the rsfMRI scans (Shah et al., 2015; Shah D. et al., 2016). After the rsfMRI procedures, isoflurane levels were gradually increased from 0.4% to 1.5–2% to make sure that the animals remained stable for the consecutive DTI and 3D acquisitions as described earlier (Anckaerts et al., 2019a). Additional details on animal handling and monitoring of physiological parameters are provided in Supplementary Methods (1.5. Animal handling).

2.4.2. Image acquisition

All imaging experiments were performed on a 9.4 T MR scanner (Bruker Biospin MRI, Ettlingen, Germany).

2.4.2.1. Magnetic resonance localized spectroscopy.

MRS experiments were performed using a mouse-head cryogenic receiver RF coil (Bruker Biospin MRI, Ettlingen, Germany). Localized in vivo 1H-spectra were acquired by the Localization by Adiabatic Selective Refocusing (LASER) sequence. For single voxel 1H MRS, the volume of interest was placed on the hippocampus, cerebral cortex and hypothalamus (Supplementary Figure 2). A single-slice multi-echo spin echo sequence was used for T₂ relaxation mapping to acquire tissue water relaxation time. The detailed description is available in Supplementary Methods (1.6.1. Magnetic resonance imaging and spectroscopy).

2.4.2.2. RsfMRI and DTI.

RsfMRI and DTI scans were acquired using a mouse-head quadrature surface coil for detection (Bruker Biospin MRI, Ettlingen, Germany) and a quadrature volume RF coil for transmission as described earlier (Shah D. et al., 2016). The anatomical position of the brain regions selected for rsfMRI and DTI data collection is presented in Supplementary Figures 3 and 6. RsfMRI signals were measured by a T₂*-weighted single shot echo-planar-imaging (EPI) sequence (TR 20 0 0 ms-150 repetitions). In vivo DTI scans were acquired using a coronal diffusion-weighted four-shot spin echo-EPI sequence. The detailed description is available in Supplementary Methods (1.6.2. rsfMRI and DTI).

2.4.3. MR image and spectroscopy data analysis

2.4.3.1. 1H MRS data analysis.

MRS data analysis was performed using LC model (Provencher 1993) as described previously (see Supplementary Methods (1.6.3. H-MRS spectra processing and quantification) for further details) (Öz et al., 2010; Tkáč et al., 2004). Briefly the contributions of individual metabolites to the MRS spectra were calculated relative to unsuppressed internal tissue water signal acquired from the same voxel of interest. Partial volume effects (e.g., relative proportions of gray, white and cerebral spinal fluid) and tissue relaxation times were taken into consideration to measure metabolic concentration (μmol/g).

2.4.3.2. RsfMRI and DTI data analysis.

Resting state functional MRI preprocessing was performed as described earlier (Shah, D. et al., 2016). Briefly, region of interest correlation and seed-based analysis were performed in 18-month-old mice (see Supplementary Methods (1.6.4. rsfMRI data analysis) for detailed description). Briefly, for each subject, correlation coefficients (i.e., Fisher’s z-transformed Pearson’s R-value), are calculated between the blood-oxygen-level-dependent (BOLD) time series of each pair of ROIs to generate functional connectivity (FC) matrices. The correlation score between each pair of ROIs in the FC matrices refers to the FC strength. Additionally, for seed-based analysis, individual z-transformed FC-maps were acquired for all 18-month-old mice with the right anterior cingulate as seed.

DTI preprocessing was performed as described earlier (for detailed description see Supplementary Methods 1.6.5. DTI data analysis) (Anckaerts et al., 2019a; Hamaide et al., 2017). Briefly, the ROI based analysis was used to detect diffusion parameter alterations in an a priori hypothesis driven manner. A semi-automatic in-house pipeline was used to extract mean diffusion parameters from each ROI. Briefly, in vivo diffusion tensor images of each mouse brain were registered to the mouse brain atlas (developed in-house). The structural segmentations of white matter structures in the atlas were back-transformed to native space of diffusion maps from each animal and refined by manual segmentation. For exploratory purposes, diffusion parameters of gray matter were also processed. However, these results were only presented in Supplementary Table 5.

2.5. Tissue processing and histology

Mice were brought under deep anaesthesia using an intraperitoneal injection of 60 mg pentobarbital/kg body weight (Nembutal; Ceva Sante Animal, Brussels, Belgium), followed by transcardial perfusion with ice cold saline (i.e., 0.9% sodium chloride) solution as described earlier (Praet et al., 2015). After decapitation and brain dissection, the brain was cut into two pieces from the mid-line. Half of the brain was snap-frozen in liquid nitrogen vapour and stored at −80 °C until further processing for ELISA analysis. Half of the brain was postfixed in 4% paraformaldehyde for 3 hours (Merck Millipore, Merck KGaA, Darmstadt, Germany). Fixed brains were freeze protected via a sucrose gradient (sucrose, Sigma Aldrich): 2 h at 5%, 2 h at 10%, and overnight at 20%. Thereafter, brain tissues were frozen in liquid nitrogen vapour and stored at −80 °C. The brains were cryo-sectioned in a coronal plane at 12 μm thickness. The approximate positions of the slices were Bregma −1, 34 (~±0.5) mm and Bregma 1.1 (~±0.5) mm (Supplementary Figure 1). Immunofluorescence analyses were performed using anti-GFAP and anti-Iba1 antibody, and Aβ plaques were stained using thioflavin-S according to a previously described protocol (Praet et al., 2012; Shah, D. et al., 2016). Detailed descriptions of the procedures can be found in Supplementary methods (1.2. Tissue processing and histology).

2.5.1. Image analysis of brain sections

The stained coronal images were quantified by determining the relative optical density of thioflavin-S positive amyloid plaques, and GFAP positive cells. To quantify the percentage (%) stained area, ImageJ software was used (Schneider et al., 2012) (see Supplementary Methods [1.3. Image analysis of brain sections] for further details).

2.6. ELISA

A solid phase sandwich enzyme-linked immunosorbent assay (ELISA) kit was used according to the manufacture’s guidelines to determine tissue Aβ_1–42 levels within brain tissue of sham-operated and ovariectomized 18-month-old TG mice. The detailed description is available in Supplementary Methods (1.4. ELISA).

2.7. Plasma hormone analysis

Blood was collected from nonanesthetized mice as previously described (Golde et al., 2005). In brief, blood was drawn from a facial vein using a lancet and collected in prechilled EDTA-coated commercial tubes. The average blood volume collected per mouse was up to 100–250 μL. After collection of blood, fluid replacement (sterile 0.9% saline solution), equal to the amount of blood collected, was provided via subcutaneous injection. Blood was centrifuged at 6600 rpm for 15 minutes at 4 °C. The supernatant (plasma) was used to assess luteinizing hormone levels by the ELISA assay as described previously (Anckaerts et al., 2019b).

2.7.2. Statistical analysis

Two-tail student’s t-test and two-way analysis of variance (ANOVA) were performed where appropriate. Two-way ANOVA was used to test for the following model: the main effects (i.e., independent variables) of genotype (G) and surgery (S) and the interaction between two factors, namely genotype and surgery (GxS). Genotype included two levels (TG, WT) and surgery consisted of two levels (sham-operation and ovariectomy). If interaction between genotype and surgery was not significant, the interaction was left out of the model (Engqvist, 2005). If the interaction was significant, we performed post-hoc unpaired t-tests. The data were normally distributed and had equal variance. The normality was confirmed by visual inspection of Q-Q (Quantile-Quantile) plots. JMP PRO13 was used for two-way ANOVA and Graphpad Prism 6.0 was used for Student’s t-test analysis. Statistical significance was accepted at p < 0.05. Significant p-values for main effects and post-hoc Student’s t-test (from the significant interaction) were reported after correcting for multiple comparisons using false discovery rate. The adjusted critical p-value (p < 0.05) was considered significant.

3. Results

In the present work, we modeled the ovarian hormone deprivation (e.g., estrogen) in premenopausal women following ovariectomy and determined the effects of ovariectomy on brain integrity. We applied in vivo MR imaging and spectroscopy tools to measure changes in white matter integrity, resting state brain connectivity and brain metabolism in brain regions of ovariectomized or sham-operated TG mice at the age of 18 months.

3.1. Magnetic resonance spectroscopy

MRS provides a wealth of information regarding in vivo brain neurochemistry, including neuronal health, gliosis, osmoregulation, and energy metabolism (Kara et al., 2011). To investigate whether ovariectomy differentially altered metabolic profile of the hypothalamus, hippocampus and cortex in WT and TG mice, we performed MR spectroscopy at 14.5 months after surgery (see Fig. 2). We found statistically significant changes in concentration of following brain metabolites such as taurine (osmolar/glial marker), glutathione (cellular oxidative stress marker), creatine and phospho creatine (energy metabolism markers).

Fig. 2.

Mean metabolite concentrations (μmol/g) from the hypothalamus, cerebral cortex, and hippocampus of 18-month-old TGOVX (filled bars with red color), TGSHAM (stripped bars with red color), WTSHAM (stripped bar with black color), WTOVX (filled bars with gray color) mice. Two-way ANOVA revealed significant main effects (genotype (horizontal line with red color), surgery (horizontal line with blue color) and interaction (genotype × surgery, horizontal line with green color). For p-values and descriptive statistics (number of animals, mean and standard deviations) see Supplementary Table 1, 2 and 3. Abbreviations: Cr, creatine; GABA, γ-aminobutyric acid; Glu, glutamate; Gln, glutamine; GSH, glutathione; PCr, phosphocreatine; TGOVX, transgenic ovariectomized; TGSHAM, transgenic sham-operated; WTOVX, wild-type ovariectomized; WTSHAM, wild-type sham-operated; Glx, GPC (glycerophosphorylcholine) + PCh (phosphorylcholine); Ins, myo-inositol; tNAA, NAA (N-acetylaspartate) + NAAG (N-acetylaspartylglutamate); Error bar indicates standard deviation.

For the hypothalamus, two-way ANOVA yielded statistically significant interaction (i.e., genotype × surgery) effect on glutathione (GSH) (F(1,26) = 3.04, p = 0.009), taurine (F(1,26) = 5, p = 0.036) and total creatine (tCr) (F(1,26) = 9.51, p = 0.005) (Fig. 2 and Supplementary Figure 2). The post-hoc Student’s t test revealed that concentrations of taurine (p = 0.025), and tCr (p = 0.025) were higher in TGOVX compared to TGSHAM. Furthermore, for taurine and tCr, TGOVX mice had a higher metabolic concentration as compared to WTOVX mice (p = 0.005). WTSHAM mice had statistically higher concentration of GSH compared to TGSHAM (p = 0.042) (Supplementary Figure 2 and Supplementary Table 1). Two-way ANOVA revealed that there was statistically significant effect of genotype on taurine, Cr, and tCr in the cerebral cortex and the hippocampus regions such that taurine, Cr, and tCr levels were higher in TG mice compared to WT mice (p < 0.05) (Fig. 2). There was a statistically significant effect of surgery on phosphocreatine (PCr) in the cerebral cortex such that the ovariectomized mice had higher PCr levels compared to sham-operated mice (Fig. 2). The results are demonstrated in Supplementary Figures 2B and 2C (See Supplementary Tables 2 and 3 for F values).

3.2. Diffusion tensor imaging

DTI is an effective means of quantifying parameters of microstructural white matter architecture and its integrity in vivo by assessing the three-dimensional diffusion profile of water in the brain (Kantarci et al., 2017). To investigate white matter microarchitectural changes in the brain of WT and TG mice at 14.5 months after surgery, we measured diffusion parameters (e.g., fractional anisotropy, FA; radial diffusivity, RD; axial diffusivity, AxD) of white matter regions using DTI. These diffusion parameters are sensitive to demyelination and axonal loss (Winklewski et al., 2018).

An overview of the two-way ANOVA results is depicted in Supplementary Table 5 (see Supplementary Figure 3 and Supplementary Table 4 for the anatomical position of ROIs). The two-way ANOVA revealed a statistically significant interaction [genotype × surgery] effect in the corpus callosum (splenium, AxD: F(1,27) = 7.6, p = 0.010; FA: F(1,27) = 9.5, p = 0.005), fimbria (AxD: F(1,27) = 7.9, p = 0.009; MD: F(1,27) = 4.8, p = 0.038), and external capsule (RD: F(1,27) = 6.5, p = 0.017, FA: F(1,27) = 4.6, p = 0.042, MD: F(1,27) = 4.3, p = 0.049). The post-hoc t test revealed that TGOVX mice had significantly lower FA in the corpus callosum (splenium) region than WTOVX and TGSHAM (p = 0.019 and 0.037 respectively) (Fig. 3A). Similarly, TGOVX mice had lower AxD compared to WTOVX (p = 0.017) in the same region (Fig. 3A). In the fimbria region, TGOVX mice had lower AxD compared to both WTOVX and TGSHAM (p = 0.046 and 0.049 resp.) (Fig. 3B). TGOVX had lower MD values than WTOVX (p = 0.042) in the external capsule (Fig. 3C). In the same region, RD values were also lower in TGOVX compared to WTOVX (p = 0.027) (Fig. 3C). Two-way ANOVA yielded a statistically significant main effect of genotype on AxD in anterior commissure (F(1,28) = 8.6, p = 0.025) with WT mice having higher values than TG mice, regardless of the surgery (Fig. 3D).

Fig. 3.

Mean DTI indices of white matter structures (A (sCC), B (FI), C (EC), D (AC)) of sham-operated and OVX WT and TG mice. Significant effects are indicated as (Surgery, S; Genotype, G) and interaction (GxS). Black lines indicate significant results based on post hoc t tests (FDR corrected p values). Abbreviations: AC, anterior commissure; AD, axial diffusivity; AxD, axial diffusivity; EC, external capsule; FA, fractional anisotropy is a ratio quantity and has no unit; ; FI, fimbria; MD, mean diffusivity; NS, nonsignificant; OVX, ovariectomized; RD, radial diffusivity; sCC, splenium of corpus callosum; SHAM, sham-operated; TG, transgenic; WT, wild type. Values are mean ± standard deviation.

3.3. Resting state fMRI

To determine whether ovariectomy altered FC between brain regions in TG and WT mice, we performed RsfMRI at 14.5 months after surgery. We measured FC between different brain regions by assessing temporal coherence between their spontaneous low frequency blood-oxygen level dependent (BOLD) fluctuations (Biswal et al., 1995).

BOLD FC was assessed in 18-month-old mouse groups (i.e., TGOVX, TGSHAM, WTOVX, and WTSHAM). For each group, OVX and sham-operated TG and WT mice, mean z-transformed FC matrices are calculated (Fig. 4A). The ROIs that were included into the study are presented in Supplementary Figure 6. Figure 4A shows FC matrices of 18-month-old mice groups (TGOVX vs. TGSHAM and WTOVX vs. WTSHAM). Fig. 4B depicts for which pair of regions a FC statistically significant (FDR corrected p < 0.05) genotype, surgery or interaction is observed. This outcome is visually represented on a binary matrix (color boxes indicate statistically significant results, p < 0.05). The results of two-way ANOVA analysis of pairs of regions for which the FC demonstrates a statistically significant genotype, surgery and interaction effect are depicted in Fig. 4B (see Supplementary Table 7 for further details). In general FC between ROIs was lower for the TG group than for the WT group across the entire brain which was also confirmed by seed-based analysis (Fig. 4C). No significant interaction or surgery effect was observed with seed-based analysis using the right cingulate as the seed when the threshold was selected as p = 0.001 (uncorrected).

Fig. 4.

(A) Resting-state functional connection (FC) based correlation matrices of 18-month-old ovariectomized (OVX) (Top-right) or sham (Bottom-left) operated TG and WT groups. The colours within of the FC matrices represent z-transformed correlation of the BOLD time series between each pair of region of interests (ROIs). (B) Statistical analyses of FC is shown as a binary matrix, indicating statistically significant effect (after FDR correction) of genotype (green box) (Top-right), surgery (gray box) (Bottom-left) and interaction (blue box) (Top-right) for each ROI–ROI connection. (C) Group averaged statistical FC maps obtained from a seed-based analysis at 18 months with the right anterior cingulate cortex as seed (represented by the blue arrow) are presented for WT and TG. (D) Post-hoc t test were performed based on significant interaction effect (p < 0.05, significant after multiple comparison correction). Black lines indicate significant differences between groups (i.e., TGOVX, TGSHAM, WTOVX, WTSHAM) p value of each significant interaction (Genotype (G) × Surgery (S)) are depicted under each graph (Figure D). Bar graphs indicate the mean values, and the error bars indicates standard error of the mean. Number of animals per group: WTSHAM/OVX: N = 7/8; TGSHAM/OVX: N = 6/7. Abbreviations: Cg_a1, cingulate cortex area 1; Cg_a2, cingulate cortex area 2; CpU, caudate putamen; dHC, dorsal hippocampus; HT, hypothalamus; L, left; MC, motor cortex; PFC, prefrontal cortex; Pta, parietal association area; R, right; RSC, retrosplenial cortex; TH, thalamus; TG, transgenic mice; vHC, ventral hippocampus; WT, wild-type mice; correlations, z- transformed functional connectivity scores.

Fig. 4B and D demonstrate significant interactions results for the FC of 18-month-old mice. The post-hoc Student’s t test revealed that TGSHAM group had significantly (p < 0.05) higher correlation compared to TGOVX group between dorsal hippocampus (left) and prefrontal cortex (left), dorsal hippocampus (left) and ventral hippocampus (right). Additionally, post-hoc Student’s t test revealed that WTOVX had significantly (p < 0.05) higher correlation scores compared to WT sham between cingulate (a1) (right) and posterior frontal cortex (left), cingulate (a2) (right) and prefrontal cortex (right), cingulate (a2) (left) and prefrontal cortex (right), dorsal hippocampus (left) and prefrontal cortex (left), motor cortex (right) and prefrontal cortex (left).

3.4. Histology and plasma hormone levels

Fig. 5A shows a representative double immunofluorescent staining of Iba1 (red color) and GFAP (green color) in the Tg2576 cortex. Region of interests are depicted on Nissl stained reference images (from Allen Mouse Brain Atlas). These regions were used for estimation of quantitative percent (%) stained area of GFAP immunofluorescence staining. Two-way ANOVA revealed significant effect of genotype (TG > WT) on level of GFAP immunoreactivity in the cingulate, hippocampus, hypothalamus and corpus callosum (FDR adjusted p < 0.001, p = 0.001, p = 0.047 and p = 0.002 respectively). Fig. 6A depicts representative thioflavin-S images. The quantification of thioflavin-S positive Aβ aggregation (% area) in the cortex, hippocampus and white matter revealed no appreciable difference between OVX and sham treated TG mice (Fig. 6B).

Fig. 5.

A representative GFAP and Iba1 merged image showing activated glia cells (see white arrow, green color is for GFAP, red color is for Iba1) in the cortex (A). A representative Nissl staining of brain slides (from the Allen Mouse Brain Atlas) is showing location of regions of interest used for quantitative analysis (B). Quantification of immunofluorescence staining for GFAP in gray matter regions (cingulate, hippocampus, hypothalamus) and white matter regions (corpus callosum and external capsule) in 18-month-old SHAM or OVX treated wild-type (WT) and TG2576 (TG) mice (TGOVX, N = 8; TGSHAM, N = 5; WTOVX, N = 5;WTSHAM, N = 6) (A). (C). Two-way ANOVA was performed to investigate the main effects (Genotype (G), Surgery (S)) and interaction (GxT). The FDR adjusted p values are given in boxes on each graph. Abbreviations: CC, corpus callosum; Cg, cingulate; EC, external capsule; GFAP, glial fibrillary acidic protein staining astrocytes; HC, hippocampus; HT, hypothalamus; IBA1, allograft inflammatory factor 1 staining microglia; OVX, ovariectomized; SHAM, sham-operated; TG, TG2576 mice; WT, wild-type mice. Bar graphs indicate group mean and error bars indicate standard deviation.

Fig. 6.

Fluorescence microscopy showing a representative brain section of ovariectomized (TGOVX) and sham-operated (TGSHAM) 18-month-old Tg2576 mice. Aβ aggregates appear light green on the sections (A). Quantification of Aβ aggregates (thioflavin-S fluorescence staining) in the cortex, hippocampus and white matter (corpus callosum and external capsule) (TGOVX, N = 6–7; TGSHAM, N = 4–6)) (B). Tissue Aβ42 (ELISA) was quantified from the whole brain tissue of 18- month-old SHAM or OVX treated TG mice (p < 0.05, two tailed Student’s t-test) (TGOVX, N = 9; TGSHAM, N = 4) (C). Two-way ANOVA indicates a statistically significant (p < 0.001) effect of surgery(S) on Luteinizing hormone level (LH) whereby ovariectomized mice had higher level of LH compared to sham-operated mice (Age of mice is between 16- and 18-mont-old, N = 9–11 per group). Bar graphs indicate the mean values and the error bars indicate standard deviation. Abbreviations: CX, cortex; GFAP, glial fibrillary acidic protein staining astrocytes; HC, hippocampus; IBA-1, allograft inflammatory factor 1 staining microglia; OVX, ovariectomized; SHAM, sham-operated. Bar graphs: Mean ± SD.

Quantification of the tissue Aβ42 levels using an ELISA kit (Fig. 6C) did not reveal any appreciable difference in Aβ42 levels between TGSHAM mice and TGOVX mice. Two-way ANOVA revealed statistically significant effect of surgery on luteinizing hormone levels such that ovariectomized mice had significantly higher luteinizing hormone level compared to luteinizing hormone level of sham-operated mice (F(1,32) = 80, p < 0.001) (Fig. 6D).

4. Discussion

The results of this study support previous findings that metabolic, functional and structural alterations occur in the brain of Tg2576 mice due to progression of AD pathology, and extend those earlier observations that some of these pathological alterations may be accelerated by ovariectomy.

4.1. Magnetic resonance spectroscopy

Our study demonstrated an increase in taurine levels in ovariectomized TG mice compared to sham-operated TG mice and ovariectomized WT mice in the hypothalamus. Taurine, an organic acid, is synthesized from cysteine by astrocytes but not by neurons (Brand et al., 1993). Earlier preclinical MRS studies have demonstrated that alterations in taurine levels might be associated with glial cell activity and/or osmoregulation (Choi et al., 2007). Our histological evaluation depicted that the increase in taurine levels in ovariectomized TG mice compared to sham-operated TG mice in the hypothalamus was not due to increased amyloid-beta (Aβ₄₂ and total Aβ) and GFAP-positive astrocyte levels. A recent study reported an inhibitory role of high estrogen on taurine biosynthesis, suggesting that taurine might be produced at higher levels under low estrogen conditions (Ma et al., 2015). Our findings showed that increased levels in ovariectomized TG mice might be due to loss of ovarian function as well as genotype related pathophysiological changes.

Total Cr plays a role in cognitive performance and is considered a marker for cell energy metabolism (Ferguson et al., 2002). When brain regions become energy deprived, Cr transporters, may not work effectively causing an increase in tCr concentration in the extracellular matrix (Gallant et al., 2006). An increase in tCr in ovariectomized TG mice compared to sham-operated TG mice in the hypothalamus might be attributed to disturbed energy metabolism resulting from a prolonged deprivation of estrogen (Xu and López, 2018). The absence of any change in tCr levels between WT mice groups (sham-operated vs. ovariectomized) suggests that low estrogen levels alone were not sufficient to cause any change in tCr when Alzheimer pathology was not present. The underlying cellular and molecular mechanisms that caused significant differences in taurine and tCr levels between ovariectomized TG mice and sham-operated TG mice need further investigation. We observed a significant main effect of genotype on Cr and tCr levels acquired from the hippocampus and the cerebral cortex, such that TG mice had a higher concentration of Cr and tCr than WT mice. The increase in tCr and Cr observed in this study is consistent with earlier studies (Burklen et al., 2006; Gallant et al., 2006). These changes have been associated with an increase in AD related pathology (Burklen et al., 2006; Gallant et al., 2006).

4.2. Diffusion tensor imaging

Two interesting findings of our study were the significant decrease in FA in the splenium of the corpus callosum and AxD in the fimbria of the ovariectomized TG mice compared with sham-operated TG mice. A decrease in FA is considered to be an un-specific marker for axonal and myelin damage (Aung et al., 2013; Filippi and Agosta, 2011). A link between diffusivity parameters (e.g., FA and/or RD) and myelin integrity was demonstrated by animal and human studies (Kantarci et al., 2014; Lee et al., 2009; Song et al., 2004; Song et al., 2002). Some of these studies suggest that a decrease in white matter FA is correlated with progression of dementia (Kantarci et al., 2014; Lee et al., 2009) and AD-related tau but not Aβ pathology (Kantarci et al., 2017). In our study long-term ovariectomy might have accelerated white matter degeneration in ovariectomized TG mice through deprivation of ovarian estrogen. A protective role of estrogen on white matter integrity of ovariectomized middle aged female rats has been demonstrated (Luo et al., 2016). AD related pathology, such as inflammation and Aβ deposition, may also cause white matter degeneration (Mitew et al., 2010; Schmued et al., 2013). In our study, the levels of Aβ₄₂ and Aβ deposits were not significantly different between ovariectomized and sham-operated TG mice. These results suggest that, regardless of Aβ₄₂ and Aβ levels, ovariectomy-related physiological alterations might have contributed to a decrease in white matter FA in TG mice. This could be due to a lack of protective effects of estrogen along with the presence of amyloid pathology in ovariectomized TG mice. The decrease in AxD may indicate a disruption in the diffusivity of water parallel to white matter due to axonal injury (Loy et al., 2007; Song et al., 2003; Sun et al., 2007). The observed decrease in AxD in the anterior commissure of TG mice compared to WT mice is in line with earlier studies (Sun et al., 2005; Zerbi et al., 2013) and may indicate a possible axonal and myelin injury in TG mice compared to WT mice. RD is also considered as a marker for demyelination (Song et al., 2002). We observed a decrease in RD in the external capsule of TG mice compared with WT mice (both mice were OVX). Our results are in line with earlier studies that reported a decrease in RD in TG mice compared to WT mice (Sun et al., 2005; Zerbi et al., 2013). Taken together, these results suggest both genotype and ovariectomy influence axonal and myelin integrity.

4.3. Resting state fMRI

RsfMRI analysis depicted that long-term deprivation of estrogen had an effect of brain FC connectivity both in ovariectomized TG and WT mice compared with sham-operated TG and WT mice. The interpretation of perturbation in FC is challenging. An increase in FC in ovariectomized WT mice compared with sham-operated WT mice may reflect overcompensation mechanisms, which was activated due to ovarian hormone loss (Anckaerts et al., 2019b). The decrease in FC in TGOVX mice compared with TGSHAM mice might be due to an interaction between AD pathology and endocrine disruption (i.e., low estrogen). Dumas and colleagues (Dumas et al., 2013) reported that postmenopausal women with cognitive complaints had increased cortical activity responses (on BOLD fMRI) during memory performance tests as compared to women without cognitive complains. Their results suggested that increased activation may be related to either the activation of a compensation mechanism or the less- efficient use of neuronal resources (Dumas et al., 2013). Vega and colleagues (Vega et al., 2016) reported that higher cognitive complaints were associated with higher functional connectivity (measured with resting state fMRI) in the executive control network. The authors speculated that the increased resting state connectivity observed in the study may reflect increased cognitive work to sustain sufficient performance during the normal (or early pathological) cognitive aging process (Vega et al., 2016). Other studies also reported enhanced FC strength in the early postmenopausal women compared to premenopausal controls (Zhang et al., 2018). Changes in the FC pattern from high to low over the course of the AD have been demonstrated in the default mode like network (Schultz et al., 2017; Shah, D. et al., 2016). Our results suggest that lower FC in ovariectomized TG mice compared to sham-operated TG mice could be due to accelerated amyloid pathology. We did not investigate the effect of ovariectomy on vascular pathology in this study. However, accelerated vascular pathology in ovariectomized TG mice might have contributed to FC alterations. The main effect of genotype on FC is partially in line with earlier studies (Belloy et al., 2018; Shah, D. et al., 2016). These studies associated FC changes with the presence of amyloid pathology in TG mice. In the current study, the most prominently observed genotype effect on FC was associated with the cingulate network. Consistent with our study, decreased resting-state activity in the cingulate cortex and default-mode network regions has been reported in AD subjects and mouse models of AD (Belloy et al., 2018; Greicius et al., 2004; Hedden et al., 2009; Mevel et al., 2011; Shah, D. et al., 2016; Sheline and Raichle, 2013; Sperling et al., 2009).

In our study rsfMRI signal was acquired using an optimized medetomidine bolus and low dose of isoflurane anesthesia protocol (Shah, Disha et al., 2016). Although anesthesia is required to minimize stress and animal movement, type and/or dosage of anesthesia may alter neural activity and cerebrovascular tone which effect FC patterns. The anesthesia protocol, used in this study, is a well-established protocol (Grandjean et al., 2014; Shah Disha et al., 2016; Shah D. et al., 2016). It has been reported that lower anesthetic depts acquired with this anesthetic protocol provided stable physiological parameters, and higher FC values compared to other anesthetic protocols (Grandjean et al., 2020; Grandjean et al., 2014; Shah, Disha et al., 2016). More detailed information regarding effects of different type of anesthetic protocols on FC signal can be found somewhere else (Bukhari et al., 2017; Grandjean et al., 2014; Jonckers et al., 2014). There might be an interaction effect of ovarian hormones and anesthesia on FC signal. Although there is no literature available about the interaction effect of ovarian hormones and anesthesia on FC signal, this effect might be insignificant due to low dose of anesthesia used in this study.

4.4. Histology and plasma hormone levels

Neuroinflammation and Aβ accumulation (soluble/insoluble) are pathological features of AD. Therefore, we investigated the impact of ovariectomy on these markers. In this study, an increased GFAP immunoreactivity in the gray and white matter regions of TG2576 mice compared to wild-type mice was quantified. A strong genotype effect was observed. These results are in line with earlier studies (Kara et al., 2015; Shah D. et al., 2016). The level of GFAP positive astrocytes have been associated with Aβ deposits in the brain of rodent models of AD (Matsuoka et al., 2001) and AD patients (Nagele et al., 2003). In our study ovariectomy did not cause any increase in Aβ levels in ovariectomized TG mice compared to sham-operated TG mice, which may explain why reactive astrocyte levels were not different between TG mouse groups (ovariectomized vs. sham-operated). Our results are in line with earlier studies that showed that Aβ (soluble and/or insoluble forms) accumulation was not mediated by ovariectomy (Green et al., 2005; Heikkinen et al., 2004; Levin-Allerhand et al., 2002; Tschiffely et al., 2016; Yue et al., 2005; Zeydan et al., 2019). However, our results are not in line with earlier studies suggesting modulation of Aβ accumulation by ovariectomy (Ding et al., 2013; Jenna and Emily, 2012; Levin-Allerhand and Smith, 2002; Palm et al., 2014; Yao et al., 2012; Yao et al., 2018; Zheng et al., 2002). Among other factors, the controversy between these results may be due to the time gap between the ovariectomy operation and Aβ quantification, the background strain, and transgene variations, which influence AD pathology (Jack et al., 2007). In contrast to our study, a current human autopsy study associated surgical menopause with increased burden of AD-related plaques (Bove et al., 2014). However, one of the drawbacks of the study was that the type of surgery (hysterectomy or unilateral or bilateral oophorectomy) was not specified. Therefore, it was not clear whether the increase in brain Aβ levels in the autopsied patients was associated with bilateral oophorectomy or not. A recent in vivo human study measured cortical Aβ levels using position emission tomography (Zeydan et al., 2019). There was no significant difference in cortical Aβ levels between controls and bilateral oophorectomy group.

Following ovariectomy sequential physiological changes occur including depletion of estrogen and an increase in gonadotropins, such as luteinizing hormone. Elevated plasma luteinizing concentration levels in the ovariectomized mice in our study validate that the ovariectomies were successfully performed. Furthermore, our results suggest there was no significant effect of genotype on luteinizing hormone levels.

In this study, the survival rate in TG group (54%) was smaller compared to WT mice (74%). These results are in line with earlier reports (King and Arendash, 2002; King et al., 1999; Lewis et al., 2004). Earlier studies speculated that cerebral bleeds and ischemic brain damage as well as over expression of APP gene in TG2576 mice might be an underlying reason for decreased survival rate in TG mice compared to WT mice with the same genetic background. However, the exact reasons need further investigation.

A few limitations of the current investigation should be acknowledged. First, as the study included only one time point, interpretations about causality need further investigation. We partially overcome this limitation by using appropriate control groups. Second, bilateral ovariectomy generates a cascade of events and makes it difficult to depict the ovarian hormones and ovarian functions that may have had direct effects on the alterations that we observed in our neuroimaging study (Rocca et al., 2017; Rocca et al., 2014; Rocca et al., 2009). In this study, we used a sham surgery to create control groups. Although a sham surgery is one of the most common methods to create control groups, it creates less tissue destruction relative to ovariectomy. Alternatively, ovariectomized mice with estrogen pellets might also serve as control mice. Future studies are needed to address which type of control group should be used to study the impact of ovariectomy on brain aging.

In conclusion, these results suggest that the osmoregulatory response was elevated, energy metabolism was disrupted, and white matter and network integrity decreased after ovariectomy in TG2576 mice compared to sham-operated TG2576 mice. These MRI results support data from other human and animal studies, suggesting that abrupt hormonal changes associated with ovariectomy may accelerate the disease progression by causing abnormalities in brain structure, function, and metabolism. Furthermore, our findings demonstrated that imaging methods could pick up neurologically relevant changes associated with ovariectomy while histopathology could not. There is need for longitudinal studies to investigate the progression of ovariectomy related changes in the brain using in vivo imaging methods.