Estrogen receptor beta in astrocytes modulates cognitive function in mid-age female mice

Abstract

Menopause is associated with cognitive deficits and brain atrophy, but the brain region and cell-specific mechanisms are not fully understood. Here, we identify a sex hormone by age interaction whereby loss of ovarian hormones in female mice at midlife, but not young age, induced hippocampal-dependent cognitive impairment, dorsal hippocampal atrophy, and astrocyte and microglia activation with synaptic loss. Selective deletion of estrogen receptor beta (ERβ) in astrocytes, but not neurons, in gonadally intact female mice induced the same brain effects. RNA sequencing and pathway analyses of gene expression in hippocampal astrocytes from midlife female astrocyte-ERβ conditional knock out (cKO) mice revealed Gluconeogenesis I and Glycolysis I as the most differentially expressed pathways. Enolase 1 gene expression was increased in hippocampi from both astrocyte-ERβ cKO female mice at midlife and from postmenopausal women. Gain of function studies showed that ERβ ligand treatment of midlife female mice reversed dorsal hippocampal neuropathology.

Fig. 1. Female mice show abrupt brain substructure volume loss after midlife compared to males, and ovariectomized females have smaller dorsal hippocampal volumes than sham treated females at midlife.

a Schematic showing timing of gonadectomy (GDX) or sham surgery, cognitive assessment by Morris Water Maze (MWM), in vivo MRI, and pathology (PATH) at young, midlife, and old ages. b In vivo MRI was collected from young, midlife, and old female and male mice. Substructure volumes visualized on the mean template (dorsal hippocampus = red, ventral hippocampus = green, cortex = blue, striatum = purple). Sham-treated female (red) and male (blue) volumes are expressed as a percentage of intercranial volume (ICV) for (c) frontal cortex, (d) striatum, and (e) dorsal hippocampus over the lifespan. Males showed gradual atrophy in frontal cortex and striatum from young to midlife to old age. In contrast, females maintained volumes through midlife, followed by atrophy from midlife to old age. Dorsal hippocampus showed atrophy in females from midlife to old age, while males did not have atrophy. Two-way ANOVA indicated a significant interaction between sex and age in dorsal hippocampus (p = 0.0059). n = 12 for all groups (c–e). p values were calculated by two-sided Welch’s t-test. Female sham-treated (solid) and GDX (diagonal lines) substructure volumes, assessed by MRI, expressed as a percentage of intercranial volume (ICV) are shown for (f) whole hippocampus, (g) dorsal hippocampus, and (h) ventral hippocampus. GDX females showed smaller dorsal hippocampus than sham females at midlife (p = 0.014). Female midlife sham n = 6 and midlife GDX n = 8, old sham n = 5 and old GDX n = 8. p values were calculated by two-sided Welch’s t-test. All box plots with center lines showing the medians, boxes indicating the interquartile range, and whiskers indicating a maximum of 1.5 times the interquartile range beyond the box.

Fig. 2. Ovariectomy of female mice worsens cognition at midlife, but not young age.

a, b Cognitive assessment by the Morris water maze. Three different ages (Young, Midlife, and Old) of sham-treated C57BL/6 wild type (a) females or (b) males were tested using Morris Water Maze (MWM) behavioral testing. A probe trial was conducted 2 h after the 5-day platform hidden training (without escape platform). % of time in the target quadrant of the maze (TQ) indicated memory of the platform location in preference to the average of other 3 quadrants (OQ). a Significant preference for the TQ compared to the OQ for all ages was observed in sham treated females (p < 0.0001, TQ vs. OQ in all ages), indicating intact reference memory in each group with no between group differences. Blue line indicates the null hypothesis (25% in TQ). Female young n = 24, midlife n = 30, and old n = 16. b Significant preference for the TQ compared to the OQ for all ages was observed in sham treated males indicating intact reference memory in each group. Male young n = 7, midlife n = 14, and old n = 7. c Percent (%) time in target quadrant (TQ) and the average of other quadrants (OQ) among females (red) and males (blue), at young and midlife, all GDX (diagonal lines). Intact cognition (significant preference for the TQ compared to OQ) was observed in GDX females young and in all GDX males (young and midlife), while GDX females at midlife had impairment. Between groups differences showed a significant decrease of % time in TQ in GDX females at midlife compared to GDX males at midlife (p = 0.0093). There was also a significant decrease of % time in TQ in GDX females at midlife compared to sham females at midlife (c vs. a, p = 0.0085). Female GDX young n = 17 and midlife n = 8, Male GDX young n = 7 and midlife n = 7. d Significant working memory impairment, assessed by Y maze, was observed in midlife GDX females (p = 0.0011, vs. midlife sham; p = 0.0024, vs. young GDX). Female young sham n = 14 and GDX n = 16, midlife sham n = 13 and GDX n = 16. Percent (%) time in TQ correlated with (e) dorsal hippocampus volume (r = 0.28; p = 0.011), but not with (f) ventral hippocampus volume (r = 0.027; p = 0.809). p values were calculated by two-sided Mann–Whitney U test, except for two-sided Pearson correlation analyses (e, f). All box plots with center lines showing the medians, boxes indicating the interquartile range, and whiskers indicating a maximum of 1.5 times the interquartile range beyond the box.

Fig. 3. Gonadectomy induces glial activation and synaptic loss in dorsal hippocampus in female mice at midlife, but not young age.

a Representative ×40 images of LCN2 (red), GFAP (green), merged images for colocalization (yellow, white arrowheads) showing astrocyte reactivity; c MHCII (green), IBA1(red), merged images for colocalization (yellow, white arrowheads) showing activated microglia; e CLEC7A (green), P2RY12 (red), merged image for colocalization (yellow, white arrowheads) showing disease-associated microglia (DAM); g SYN1(red), PSD95 (green), merged images for colocalization (yellow, white arrowheads) showing pre- and post-synaptic staining. Inset: Magnification at ×100. Nuclei were counterstained with DAPI (blue). Bar = 20 um. Quantitative analysis of (b) LCN2⁺GFAP⁺ area fraction, (d) MHCII⁺IBA1⁺ area fraction, (f) CLEC7A⁺P2RY12⁺ area fraction, and (h) SYN1⁺PSD95⁺ area fraction in dorsal hippocampal CA1 region from GDX and sham females at young and midlife ages. Midlife GDX females showed a significant increase in reactive astrocytes (p = 0.0221), activated microglia (p = 0.0289), DAM (p = 0.0028), and synaptic loss (p = 0.009) each as compared to midlife sham females. Female young sham n = 7, 6, 5, 5, and GDX n = 8, 7, 5, 5, midlife sham n = 6, 7, 10, 10 and GDX n = 7, 8, 12, 12 (b, d, f, h, respectively). p values were calculated by two-sided Mann–Whitney U test. All box plots with center lines showing the medians, boxes indicating the interquartile range, and whiskers indicating from the minimum and to the maximum values.

Fig. 4. Validation of Cre expression specificity in mGFAP-Cre 77.6 mice and ERβ selective deletion in the astrocyte ERβ cKO.

a Breeding scheme to generate mGFAP-Cre; RiboTag mice. b Quantification of HA-colocalization with GFAP⁺ and ALDH1L1⁺ for astrocytes, NEUN⁺ cells for neurons, IBA1⁺ and P2RY12⁺ for microglia, and CC1⁺ cells for oligodendrocytes. N = 4 for GFAP, ALDH1L1, and P2RY12, and n = 3 for IBA1, NEUN, and CC1. Error bars represent SEM. p values were calculated by two-sided one sample t-test. c Representative ×40 images of HA (red), GFAP (green, top), and ALDH1L1 (green, bottom) with merged images for colocalization (yellow, arrowheads) in dorsal hippocampal CA1 region from mGFAP-Cre(77.6); RiboTag mice. Bar = 20 um. Experiments were repeated independently at least twice. d Breeding scheme to generate mGFAP-Cre(77.6); ERβ^fl/fl mice. e Representative ×40 image of ERβ (red) and GFAP (green) with colocalization (yellow) in dorsal hippocampus CA1 region from WT littermates (WT, left), mGFAP-Cre(77.6); ERβ^fl/fl (astrocyte ERβ cKO, right) mice. Bar = 10 um. f Quantitative analysis of area fraction of colocalized ERβ⁺ GFAP⁺ in dorsal hippocampal CA1 region from WT littermates (WT, red), astrocyte ERβ cKO (green), and rNSE-Cre; ERβ^fl/fl (neuron ERβ cKO, orange) mice. A significant decrease of ERβ⁺ GFAP⁺ colocalized area was observed in astrocyte ERβ cKO, compared to WT littermates (p = 0.0002) and compared to neuron ERβ cKO (p = 0.0019). n = 8 per group. p values were calculated by two-sided Mann–Whitney U test. Box plots with center lines showing the medians, boxes indicating the interquartile range, and whiskers indicating from the minimum and to the maximum values.

Fig. 5. Selective deletion of ERβ in astrocytes induces cognitive deficits, dorsal hippocampal atrophy and neuropathology therein at midlife in female mice.

a Gonadally intact wild type (WT, red) females along with conditional knock-outs of ERβ in astrocytes (astrocyte ERβ cKO, green) or in neurons (neuron ERβ cKO, orange) were assessed at midlife for spatial reference memory by MWM. WT and neuron ERβ cKO showed preference for the TQ over the OQ (p < 0.0001), while astrocyte ERβ cKO did not. Blue line indicates the null hypothesis (25% in TQ). Between groups differences showed a significant decrease in % time in TQ in gonadally intact females with ERβ deleted in astrocytes (p = 0.0077 vs. WT; p = 0.0021 vs. neuron ERβ cKO). WT n = 23, astrocyte ERβ cKO n = 14, and neuron ERβ cKO n = 16. Substructure volumes, assessed by MRI, taken as a percentage of intercranial volume (ICV) are shown for (b) hippocampus, (c) dorsal hippocampus, and (d) ventral hippocampus. Astrocyte ERβ cKO females showed worse atrophy compared to WT and compared to neuron ERβ cKO at midlife in whole hippocampus (p = 0.0048, vs. WT; p = 0.004973, vs. neuron ERβ cKO) and dorsal hippocampus (p = 0.0055 vs. WT; p = 0.0016 vs. neuron ERβ cKO). WT n = 15, astrocyte ERβ cKO n = 17, and neuron ERβ cKO n = 18. e–l Neuropathology. Representative ×40 images of (e) LCN2 (red), GFAP (green), merged images for colocalization (yellow, white arrowheads) showing astrocyte reactivity, (g) MHCII (green), IBA1(red), merged images for colocalization (yellow, white arrowheads) showing activated microglia, (i) CLEC7A (green), P2RY12 (red), merged image for colocalization (yellow, white arrowheads) showing disease-associated microglia (DAM), (k) SYN1 (red), PSD95 (green), merged images for colocalization (yellow, white arrowheads) showing pre- and post-synaptic staining. Inset: Magnification at ×100. Nuclei were counterstained with DAPI (blue). Bar = 20 um. Quantitative analysis of (f) LCN2⁺GFAP⁺ area fraction for reactive astrocytes, (h) MHCII⁺IBA1⁺ area fraction for activated microglia, (j) CLEC7A⁺P2RY12⁺ area fraction for disease associated microglia (DAM), and (l) SYN1⁺PSD95⁺ area fraction for synapses in dorsal hippocampal CA1 region among WT (red), astrocyte ERβ cKO (green), and neuron ERβ cKO (orange) females at midlife. Astrocyte ERβ cKO showed a significant increase of reactive astrocytes (p = 0.0093, vs. WT; p = 0.0019, vs. neuron ERβ cKO), activated microglia (p = 0.014, vs. WT; p = 0.0379, vs. neuron ERβ cKO), and DAM (p = 0.0289, vs. WT), as well as synaptic loss (p = 0.0289, vs. WT). WT n = 7, astrocyte ERβ cKO n = 8, and neuron ERβ cKO n = 8. p values were calculated by either two-sided Mann–Whitney U test (a, f, h, j, l) or two-sided Welch’s t-test (b–d). All box plots with center lines showing the medians, boxes indicating the interquartile range, and whiskers indicating either a maximum of 1.5 times the interquartile range beyond the box (a–d) or from the minimum and to the maximum values (f, h, j, l).

Fig. 6. RNA sequencing and pathway analysis of hippocampal astrocytes from female astrocyte-ERβ cKO mice at midlife.

a Heatmaps of top seven downregulated genes (FDR < 0.1, log2 fold change is >0.5) in hippocampal astrocytes from astrocyte ERβ cKO mice and corresponding genes from GDX mice, each compared to that observed in WT sham mice. Asterisks indicate genes previously reported as estrogen responsive^,. b Top two most differentially expressed pathways in dorsal hippocampal astrocytes of ERβ cKO versus WT mice were the Gluconeogenesis I and Glycolysis I pathways. Right-tailed Fisher’s exact test. c The top differentially expressed gene in the Gluconeogenesis I and Glycolysis I pathways was Enolase 1 (Eno1) in hippocampal astrocytes of ERβ cKO versus WT mice. (WT, n = 3; cKO, n = 4, FDR = 0.00036), two-sided Mann–Whitney U test. For the box-and-whisker plots, the box indicates the median and 25–75th percentile range, and the whiskers extend to a maximum of 1.5 times the interquartile range beyond the box.

Fig. 7. Upregulation of human Enolase 1 gene in hippocampus in women during aging.

a Analysis of human hippocampus microarray data of in women (GSE11882, ages: 26–99 years old, n = 21). Higher gene expression levels of Enolase 1 (ENO1) correlated with older ages (r = 0.526; p = 0.014, blue line). b Human ENO1 gene expression was increased in females at menopausal ages (>60 years, n = 13) compared to younger ages (<50 years, n = 8), two-sided Mann–Whitney U test. For the box-and-whisker plots, the box indicates the median and 25–75th percentile range, and the whiskers extend to a maximum of 1.5 times the interquartile range beyond the box.

Fig. 8. ERβ ligand treatment reverses dorsal hippocampal pathology in female mice at midlife.

a Schematic showing timing of gonadectomy (GDX) or sham (Sham) surgery, ERβ ligand (ERβ L) or vehicle (V) treatment, MWM cognitive assessment, and neuropathology (PATH). b Gonadally intact vehicle-treated midlife females (Sham + V, solid), gonadectomized vehicle-treated midlife females (GDX + V, diagonal lines) or gonadectomized ERβ ligand-treated midlife females (GDX + ERβ L, crossed lines) were assessed for spatial reference memory by MWM. Gonadally intact vehicle-treated mice showed preference for the TQ over the OQ (p < 0.0001), while gonadectomized vehicle-treated did not. In contrast, gonadectomized ERβ ligand-treated showed preference for the TQ over the OQ (p = 0.007). Blue line indicates the null hypothesis (25% in TQ). Sham + V, n = 10; GDX + V, n = 6; GDX + ERβ L, n = 7. Quantitative analysis of (c) LCN2⁺GFAP⁺ area fraction for astrocyte reactivity, (d) MHCII⁺IBA1⁺ area fraction for activated microglia, (e) CLEC7A⁺P2RY12⁺ area fraction for disease-associated microglia (DAM), and (f) SYN1⁺PSD95⁺ area fraction for synaptic loss, each in dorsal hippocampus CA1 region in sham + V (solid), GDX + V (diagonal lines), and GDX + ERβ L (crossed lines). GDX + V showed increased reactive astrocytes (p < 0.0001), activated microglia (p = 0.0003), and disease-associated microglia (DAM) (p = 0.0027), as well as synaptic loss (p = 0.0034), each as compared to the Sham + V group. In contrast, there was a decrease in reactive astrocytes (p = 0.0046), activated microglia (p = 0.0007), and disease-associated microglia (DAM) (p = 0.0346), with increased synaptic staining (p = 0.0445) in GDX + ERβ L treated compared to GDX + V treated. Sham + V, n = 8; GDX + V, n = 14; GDX + ERβ L, n = 4. Two-sided Mann–Whitney U test. Box plots with center lines showing the medians, boxes indicating the interquartile range, and whiskers indicating from the minimum and to the maximum values.