ERβ mediates sex-specific protection in the App-NL-G-F mouse model of Alzheimer’s disease

Male and female APP^NL−G−F knock-in mice (carrying the Swedish [NL], Arctic [G] and Iberian [F] mutations in the humanized Aβ peptide [13]) were obtained from local breeding using the C57/BL6 J strain background. At 2.5 months of age, female mice were selected randomly for bilateral ovariectomy or sham surgery. Similarly, at 3 months of age male and female mice were randomly selected for LY500307↗ (0.35 mg/kg/day, Santa Cruz Biotechnology, Dallas, TX, USA), dissolved in vehicle solution (40% Captisol, [Cydex pharmaceuticals, Lawrence, KS, USA], 1% ethanol, and 59% 0.1 M PBS), or vehicle treatment (vehicle solution) through oral gavage administration. The treatment regimen was daily delivery over 7 days, followed by 7 days of rest. The resting period was included to avoid hormone-induced downregulation of ERβ gene expression [16]. This was repeated twice after which animals were subjected to behavior studies [2 days after last treatment and 2 days of rest between tests] and sacrificed at 5 months of age. For brain dissection, animals were deeply anesthetized with isoflurane followed by intracardial ice-cold 0.1 M PBS perfusion. Half brain was fixed in cold 4% paraformaldehyde and the hippocampus and cerebral cortex of the other half were snap-frozen for biochemical assays. All procedures were performed in accordance with approved ethical permits (ethical approval ID 407 and ID 2199–2021, Linköping’s animal ethical board).

4 µm thick paraffin-embedded sagittal mouse brain sections were fixed on glass slides, hydrated, followed by heat-induced antigen retrieval in a pressure steamer at 121 °C for 20 min, followed by 15 min permeabilization with 0.5% Triton-X 100 (Millipore, Burlington, MA, USA) and blocking using 10% Horse Serum (ThermoFisher Scientific, Waltham, MA, USA), 0.1% Tween-20 (Millipore) in 0.1 M PBS for 1 h at 37 °C. Following blocking the slides were immunostained over-night at 4 °C with antibodies specific to Aβ (1:2000 dilution, 82E1, IBL-Tecan, Männedorf, Switzerland), Iba1 (1:300, ab178846; and 1:300 ab225260, both from Abcam, Cambridge, UK), GFAP (1:300, GA5 Alexa Fluor₄₈₈-labeled, Millipore), CD68 (1:300, Ab283654↗, Abcam), and/or ERβ (1:5000, PP-PPZ0506-00, R&D Systems, Minneapolis, MN, USA) (Supplemental Table 1). For antibodies raised in mice, we used 1 × mouse-on-mouse IgG blocking solution (ThermoFisher Scientific) prior to antibody incubation. Secondary antibodies were Alexa Fluor₄₈₈, Alexa Fluor₅₆₈ (both from ThemoFisher Scientific). To reduce autofluorescence, the sections were incubated in 1 mM CuSO₄ diluted in 50 mM ammonium acetate for 15 min. Nuclear staining was with 300 nM DAPI (ThermoFischer Scientific) for 10 min, prior to mounting. To visualize amyloid plaques, we used 1 × AmyloGlo stain (Biosensis, Thebarton, Australia) supplemented to the secondary antibody solution. ABC-HRP kit and Impact-DAB (both from Vector Laboratories, Newark, CA, USA) were used for immunohistochemical staining according to manufacturer’s recommendations. Immunofluorescence images were captured using an AxioPlan-2 fluorescent microscope (Carl Zeiss, Oberkochen, Germany) and the Zeiss AxioVision 4.0 software (Carl Zeiss). Image analysis was performed on at least 3 sections per mouse using the ImageJ software (NIH, Bethesda, MD, USA) and setting image threshold and counting was as described previously [18]. For each acquired image, the image lookup table (LUT) was kept linear and covered the whole image data. Association of microglia to plaques were quantified by counting number of microglia within 20 µm radius of plaque edge.

Frozen cortical and hippocampal tissues were thawed and homogenized in ice-cold TBS buffer (50 mM Tris–HCl, pH 7.6, 150 mM NaCl, and protease inhibitor cocktail (Roche, Basel, Switzerland)). The homogenates were centrifuged at 24 000 × g for 45 min at + 4 °C, yielding a soluble fraction (supernatant) and an insoluble fraction (pellet). The pellets were solubilized by resuspension in 6 M Guanidine-HCl and sonication using a water-bath sonicator (Bioruptor, 5 min max output (Diagenode, Denville, NJ, USA)). Soluble pellets were centrifuged at 24 000 × g for 45 min at + 4 °C and the supernatant (defined as insoluble fraction) was diluted in TBS to yield 0.5 M Guanidine-HCl. Similarly, Guanidine-HCl was added to the soluble fractions to yield a concentration of 0.5 M Guanidine-HCl. Total protein concentration was determined using the BCA Protein Assay (ThermoFisher Scientific) or the Coomassie Protein Assay reagent (Sigma-Aldrich, St Louis, MO, USA). Quantification of Aβ_1–40 and Aβ_1–42 in soluble and insoluble fractions was performed using the EZHS-SET ELISA kit, following manufacturer’s instructions (Millipore) and read on a Tecan plate spectrophotometer. Proinflammatory cytokine profiling on soluble fractions were performed using the V-PLEX proinflammatory panel 1 (mouse) kit (Mesoscale Discovery, Rockville MD, USA) on soluble brain fractions according to manufacturer’s instructions. The kit allows multiplex quantification of IFN-γ, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-10, CXCL1 (KC/GRO, keratinocyte-derived chemokine/growth-related oncogene), IL12p70, and TNF-α. Samples were read on the MESO QuickPlex SQ120 reader and data were analyzed using the Discovery Workbench 4.0 software (both from Mesoscale Discovery). The concentration of each cytokine in the tissue lysates was normalized with the total protein concentration of the respective sample.

Cortical and hippocampal tissue were homogenized in ice-cold 4 × PIPES buffer pH 6.8 (40 mM Piperazine-1,4-bis(2-ethanesulfonic acid), 1.2 M Sucrose, 0.4 M NaCl, 27 mM MgCl₂, and 1 × protease inhibitor cocktail, all from Sigma). Cell debris were pelleted and supernatants were centrifuged at 24 000 × g for 45 min at + 4 °C. The pellet was resuspended in a low volume of 4 × PIPIES buffer, and protein concentration was measured and adjusted to 2.5 mg/ml. 100 µg protein was incubated at 37 °C for 30 min followed by chloroform–methanol protein precipitation. In brief, 600 µl of cholorform:methanol in ratio 2:1 was added to the protein mixture and incubated for 30 min at room temperature (RT) under agitation followed by centrifugation and phase separation at 24,000×g for 15 min at RT. The intermediate was isolated, resuspended in 600 µl cholorform:methanol 1:2 and incubated for 60 min at RT under agitation. The protein was precipitated by centrifugation at 24,000×g for 15 min at RT, supernatant was removed, and pellet was let to dry. The protein pellet was resuspended in SDS-loading buffer to yield 3 mg/ml. In brief, 10–30 µg of protein were loaded on 4–20% gradient SDS-PAGE gels and proteins were transferred to a PVDF membrane. After blocking the membrane was subjected to antibody against Aβ1–16 (6E10, BioLegend), APP N-terminus (22 C11, Millipore), APP C-terminus (A8717, Sigma-Aldrich) and antibody against β-Actin (AC-15, Millipore). Detection was performed using ECL substrate (ThermoFisher) and exposure to light-sensitive films or CCD camera. Quantification of bands was performed using ImageJ software (NIH). All blots were processed in parallel.

Total RNA from cells or tissue was extracted using the RNeasy plus mini kit, RNeasy plus micro kit or Allprep DNA/RNA kit (Qiagen) according to manufacturer’s instructions, and RNA concentrations and quality were determined with NanoDrop (ThermoFisher Scientific). Complementary DNA was synthesized using SuperScript IV VILO Master Mix cDNA synthesis kit (ThermoFisher Scientific). The qPCR reaction contained 5 or 10 ng of cDNA, exon-exon spanning primers (500 nM), and KAPA SYBR Fast qPCR master mix (Sigma-Aldrich) or using TaqMan assays (Supplemental Table 2) and TaqMan Fast Advanced Master Mix (Applied Biosystems) and was performed on an ABI 7500 fast thermal cycler (Applied Biosystems) according to manufacturer's instructions. Expression relative to housekeeping gene was calculated using the ΔCt method.

Results are expressed as means ± SD. The statistical analyses were performed using GraphPad Prism 9.02 software (GraphPad Software, San Diego, USA). Data were tested for equal variance by F-tests. Unpaired two-tailed Student’s t-tests were used to compare between two groups. Unless stated otherwise, multiple group analyses were performed by two-way or three-way analysis of variance (ANOVA), followed by uncorrected Fisher’s LSD test or corrected post-hoc tests for multiple comparisons as indicated in figure legends. Significance level was set at < 0.05 (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). All analyses are based on at least 3 biological replicates.

Since estrogen (E2) has been ascribed neuroprotective properties [1–3], we sought to explore if selective activation of the estrogen receptor beta (ERβ, Esr2 gene product), a more clinically relevant target than ERα, can be protective against amyloid-related pathology in the App^NL−G−F mouse model of AD. Since the expression of ERβ in the brain has been questioned due to poor antibody specificities to ERβ, we first performed immunohistochemical analysis using a validated ERβ antibody in wild-type (WT) and ERβ knockout (Esr2-KO) mouse brains. This revealed scattered expression with both cytoplasmic and nuclear localization in several brain regions affected in AD, including the frontotemporal, primary motor, somatosensory, and visual cortices, as well as in the granule layers of CA2 and dentate gyrus (DG) of the hippocampus (Supplemental Fig. 1 A, B). Highest number of ERβ positive cells were seen in frontal and primary motor cortex, as well as in the hippocampus (Supplemental Fig. 1B). ERα (Esr1) and ERβ (Esr2) had similar expression between male and female mice in cortex and hippocampus although ERα expression was about 5–tenfold higher than ERβ in both brain regions, and ERα expression did not change upon ERβ loss (Supplemental Fig. 1C–F).

To study the effect of ERβ activation on AD pathology we treated App^NL−G−F male and female mice with the selective ERβ agonist LY500307↗ daily through oral gavage (0.35 mg/kg) every other week over 5 weeks, starting at 3 months of age. A subset of female mice was ovariectomized (OVX) 2 weeks prior to treatment to study the influence of loss of circulating E2 (Fig. 1A). At the end of the treatment, the mice were subjected to memory tests, where the mice treated with LY500307↗ (LY) performed better than vehicle-treated mice in the Y-maze spatial memory test (Fig. 1B, C). Interestingly, we did not observe any adverse effect of OVX on spatial memory, rather a better performance in combination with LY (Fig. 1C). There were no effects on total arm entries (Fig. 1D). Associative memory was tested using the cued fear conditioning (FC) paradigm (Fig. 1E). Mice given LY performed better than vehicle-treated mice in the FC test, with longer episodes of freezing both during the contextual (Fig. 1F) and cued tests (Fig. 1G, H). Interestingly, LY-treated females also displayed increased freezing at baseline, even in the new context, suggestive of an overall better memory performance after LY treatment (Fig. 1H). Again, OVX did not have a negative effect on memory performance, oppositely, it slightly improved memory in the contextual FC test, although less significant compared to LY treatment (Fig. 1F). Overall, these results suggest that ERβ activation may indeed act neuroprotective in the App^NL−G−F model.

Next, we analyzed the effect of LY on amyloid pathology. LY-treated mice had generally lower number and smaller size of amyloid plaques in different cortical regions and in hippocampus in both male (Fig. 2A–C) and female (Fig. 2D–F) App^NL−G−F mice. When taking number of ERβ positive cells into account in those regions, we could observe that the largest effect size of LY treatment in lowering Aβ plaques overlapped with highest levels of ERβ positive cells in those regions (Fig. 2G, Supplemental Fig. 1B). OVX did not have any effect on number of plaques, but slightly (but significantly) increased existing plaque area in visual and somatosensory (Vis/Ss) cortex (Fig. 2E, F). In line with these results, the levels of soluble and insoluble neurotoxic amyloid beta (Aβ₄₂) were overall lower in cortex and hippocampus in both male (Fig. 2H, I) and female (Fig. 2J, K) App^NL−G−F mice after LY treatment, although it did not reach statistical significance for hippocampal soluble and cortical insoluble Aβ₄₂ in male mice and no statistical significance for cortical insoluble Aβ₄₂ levels in females. Interestingly, OVX increased Aβ₄₂ levels in the cortex, while having no effects on Aβ₄₂ levels in other brain areas. Aβ₄₀ levels were similar to Aβ₄₂ levels (Supplemental Fig. 2 A–D) and Aβ₄₂/Aβ₄₀ ratio did not differ with LY treatment, although there was an overall increase in soluble Aβ₄₂/Aβ₄₀ ratio in female cortex upon OVX (Supplemental Fig. 2F). Furthermore, female mice had generally higher levels of soluble Aβ₄₂ levels compared to male mice (Supplemental Fig. 2G, H). These data suggest that ERβ activation reduces Aβ levels and plaque load in both male and female App^NL−G−F mice, but sex differences exist, and that OVX can worsen amyloid pathology although differently in in different brain regions.

To explore if reduced Aβ₄₂ levels in LY-treated mice was a consequence of lower APP levels or a shift from amyloidogenic β-secretase processing to non-amyloidogenic α-secretase processing, we analyzed the levels of full-length APP (FL-APP) and processed APP fragments. Western blot analysis revealed that LY-treatment of male mice had no effect on FL-APP levels in cortex nor in hippocampus (Fig. 3A–D). However, FL-APP was significantly increased in female cortex upon OVX and decreased upon LY-treatment of OVX females (Fig. 3A–C). LY treatment did not result in any difference in C-terminal fragment β-CTF levels relative to FL-APP (Fig. 3D), but β-CTF was increased upon OVX relative to β-actin levels (Fig. 3E), suggesting α- or β-secretase activities are not altered by ERβ activation (Fig. 3A, B, D). In addition, the expression of App or processing enzymes (Bace1, Psen1, and Adam10) were not altered by LY treatment or OVX (Supplemental Fig. 3C–J), further suggesting that ERβ does not modulate APP processing, although OVX increased APP protein levels in female cortex (Fig, 3 A, C). Finally, we again observed lower total Aβ levels upon LY treatment in both male and female cortex and hippocampus and an interesting increase in cortex upon OVX (Fig. 3A, B, F), similar to what is seen in Fig. 2J. These data suggest that ERβ does not directly modulate APP processing but may rather be involved in the clearance of Aβ.

Astrocytes and microglia take active part in amyloid pathogenesis, including Aβ clearance [19]. Therefore, we sought to investigate the impact of ERβ activation on astrocytic and microglial response. We could not detect any major effects on astrogliosis in male and female mice treated with LY or in OVX females (Supplemental Fig. 4 A–D). Similarly, we did not see any difference in microglia numbers upon LY treatment in male App^NL−G−F mice (Fig. 4A, B, Supplemental Fig. 5A). However, LY treatment significantly reduced number of activated CD68 + microglia, especially in the male hippocampus (Fig. 4A, C), and we also observed that LY treatment promoted microglia association to amyloid plaques in male frontal/motor cortex and in hippocampus (Fig. 4D). In female App^NL−G−F mice, LY treatment had less effect on microglia (Fig. 4E–H, Supplemental Fig. 5B) compared to male mice, and we could detect a slight, but significant, effect of LY on lowering the number of activated CD68 + microglia in frontal/motor cortex and in hippocampus (Fig. 4G), but no effect on microglia association to plaques (Fig. 4H). We could also not detect any significant effect of OVX on microglia numbers or activation. However, female mice had an overall increased number of CD68 + microglia compared to male mice, and overall increased number of plaque-associated microglia, but less response to LY (Supplemental Fig. 5C, D).

We next analyzed the expression of ERβ in microglia. To avoid false positives, we used ERβ knockout (Esr2-KO) mice as staining control (Fig. 5A). Although some microglia showed ERβ positive staining, most microglia were ERβ negative (Fig. 5A). In App^NL−G−F mice, plaque-associated microglia were both ERβ positive and negative and LY treatment had no overall effect on ERβ-Iba1 co-expression in any brain region examined (Fig. 5B, C). However, there were slightly but significantly more ERβ + microglia in male compared to female App^NL−G−F brains (Fig. 5D), which could possibly explain the higher responsiveness of male microglia to LY-treatment (in Fig. 4).

Studying the expression of two specific microglial markers associated with anti-inflammatory and pro-resolving responses, Trem2 and Cx3cr1, we could not detect any effect of LY in either male or female hippocampus (Fig. 5E–H). However, OVX lowered expression of these markers. No difference in these markers was observed in cortex (data not shown). Using a detection-panel of proinflammatory cytokines we observed significantly lower levels of CXCL1 (KC/GRO) in the hippocampus of LY-treated male mice (Fig. 5H) and of IL-12p70 in both male and female LY-treated mice (Fig. 5H). Additionally, the level of the danger signal IL-10 was lower both male and female LY-treated mice (Fig. 5I), which may suggest less hyperinflammation [20, 21] upon LY treatment. Of these markers, IL-12p70 and IL-10 levels were sustained in LY-treated OVX females compared to vehicle-treated mice (Fig. 5H, I). Thus, these data suggest less neuroinflammation upon ERβ activation in both male and female App^NL−G−F mice and that OVX can bypass this effect of ERβ. In addition, it suggests that ERβ exhibit its neuroprotective effects likely through additional cell types than microglia, especially in females.

All procedures were performed in accordance with approved ethical permits (ethical approval ID 407 and ID 2199–2021, Linköping’s animal ethical board).

Not applicable.

The authors declare that they have no competing interests.

No datasets were generated or analysed during the current study.