Linear and inverted U-shaped dose-response functions describe estrogen effects on hippocampal activity in young women

Abstract

In animals, 17-beta-estradiol (E2) enhances hippocampal plasticity in a dose-dependent, monotonically increasing manner, but this relationship can also exhibit an inverted U-shaped function. To investigate E2's dose-response function in the human hippocampus, we pharmacologically increased E2 levels in 125 naturally cycling women (who were in their low-hormone menstruation phase) to physiological (equivalent to menstrual cycle peak) and supraphysiological (equivalent to levels during early pregnancy) concentrations in a placebo-controlled design. Twenty-four hours after first E2 intake, we measured brain activity during encoding of neutral and negative pictures and then tested recognition memory 24 h after encoding. Here we report that E2 exhibits both a monotonically increasing relationship with hippocampal activity as well as an inverted U-shaped relationship, depending on the hippocampal region. Hippocampal activity exhibiting a U-shaped relationship inflects at supraphysiological E2 levels, suggesting that while E2 within physiological ranges stimulates hippocampal activity, supraphysiological ranges show opposite effects.

Fig. 1

Study design. a Volunteers participated in the study on three consecutive days. On the evening of Day 1, baseline hormone concentrations and mood were assessed, followed by the first dose of either 17-beta-estradiol (E2; 2, 4, 6 or 12 mg) or placebo, depending on assigned group (double-blind). On the morning of Day 2, volunteers took the second dose on their own. In the afternoon of Day 2, when E2 levels were expected to peak, volunteers encoded emotional pictures inside the scanner (b, upper red box). On the evening of Day 3, volunteers performed a recognition test (b, lower red box) and rated the encoded pictures for arousal. See Supplementary Methods for more details on the timeline of the study

Fig. 2

17-beta-estradiol (E2) levels in saliva of the experimental groups (N = 125). a Absolute levels. Error bars represent standard errors of the mean. Confirming the effectiveness of the pharmacological manipulation, the increases in salivary and serum E2 levels from baseline (Day 1) to the expected peak (Day 2) differed as intended between experimental groups (saliva: F(4, 47.98) = 102.84, p < .001; serum: F(4,31.24) = 87.54, p < .001; see Supplementary Note 2 for tests of variance homogeneity). b Robust linear regression analyses between dose of E2 valerate given per kg bodyweight and salivary E2 increase (p < .001). A low Bayesian Information Criterion difference score (ΔBIC; linear—quadratic model) of −1.36 suggests that the linear and quadratic model fit equally well. Colors represent experimental groups

Fig. 3

Regression between memory and E2 increases in salivary 17-beta-estradiol (E2; N = 123). Neither hippocampus-dependent memory (recollection; a) nor hippocampus-independent memory (familiarity; b) shows significant relationships to salivary Day 1 to Day 2 E2 increase. Colors represent experimental groups (all p_corr’s > .2)

Fig. 4

Linear regression between E2 increases and mean arousal ratings. Negative pictures are represented by asterisks, neutral pictures by circles. Colors represent experimental groups (N = 123). A significant positive linear relationship was observed between salivary E2 increase and arousal ratings for neutral pictures (solid line; p_corr = .015) but not negative pictures (dashed line; p_corr = .524)

Fig. 5

Relationship between 17-beta-estradiol (E2) increase and the remember >know contrast. a Statistical t-map of the positive linear relationship between salivary Day 1 to Day 2 increase in E2 and the remember > know contrast (N = 118). Statistical maps are thresholded at FWE-corrected p < .05 (warm color scale) and p_unc < .005 (cold color scale) for visualization purposes. A cluster in the right hippocampus showed a positive linear relationship with salivary E2 increase (FWE-corrected p = .043; t-test). b Robust regression analysis between contrast estimates (y-axis) and salivary E2 increase (x-axis) confirmed the positive linear relationship. Colors represent experimental groups. c Statistical t-map of the negative quadratic relationship between salivary Day 1 to Day 2 E2 increases and the remember > know contrast. A cluster in the left hippocampus showed an inverted U-shaped relationship with salivary E2 (FWE-corrected p = .006; t-test). d Robust regression analysis between contrast estimates (y-axis) and salivary E2 increase (x-axis) confirmed the inverted U-shaped relationship. A Bayesian Information Criterion difference (ΔBIC) of 8.5 between linear and quadratic models strongly evidences a better model fit of the quadratic compared to the linear model

Fig. 6

Relationship between E2 increase and the negative > neutral contrast. a Statistical t-map of the negative linear relationship between salivary Day 1 to Day 2 increase in 17-beta-estradiol (E2) and the negative > neutral contrast (N = 121). The statistical map is thresholded at FWE-corrected p < .05 (warm color scale) and p_unc < .005 (cold color scale) for visualization purposes. Activity in a cluster in the right precuneus [circled in red; FWE-corrected p = .034; t-test] showed a negative linear relationship with salivary E2 increase. A negative linear relationship in the brainstem was not significant ([FWE-corrected p = .056; t-test)). A psycho-physiological interaction (PPI) analysis (seed: precuneus cluster, region of interest: brainstem cluster) revealed a negative linear relationship between E2 levels and precuneus-brainstem connectivity. b Robust regression analysis between contrast estimates (y-axis), extracted from the peak of the precuneus cluster, and salivary E2 levels (x-axis) confirmed the negative linear relationship. Colors represent experimental groups

Fig. 7

Relationship between E2 increase and the emotional enhancement of memory contrast. a Statistical t-map of the negative quadratic relationship between salivary Day 1 to Day 2 increase in 17-beta-estradiol (E2) and the emotional enhancement of memory (EEM; hit > miss × negative > neutral) contrast in the right insula (N = 121; FWE-corrected p = .015; t-test). The statistical map is thresholded at FWE-corrected p < .05 (warm color scale) and p_unc < .005 (cold color scale) for visualization purposes. b Robust regression analysis between contrast estimates (y-axis) and salivary E2 increase (x-axis) confirmed the inverted U-shaped relationship. Difference in Bayesian Information Criterion (ΔBIC) between the linear and quadratic model was 16.9, strongly evidencing a better model fit of the quadratic compared to the linear model. Colors represent experimental groups