Functional brain network topology across the menstrual cycle is estradiol dependent and correlates with individual well-being

Abstract

The menstrual cycle (MC) is a sex hormone-related phenomenon that repeats itself cyclically during the woman's reproductive life. In this explorative study, we hypothesized that coordinated variations of multiple sex hormones may affect the large-scale organization of the brain functional network and that, in turn, such changes might have psychological correlates, even in the absence of overt clinical signs of anxiety and/or depression. To test our hypothesis, we investigated longitudinally, across the MC, the relationship between the sex hormones and both brain network and psychological changes. We enrolled 24 naturally cycling women and, at the early-follicular, peri-ovulatory, and mid-luteal phases of the MC, we performed: (a) sex hormone dosage, (b) magnetoencephalography recording to study the brain network topology, and (c) psychological questionnaires to quantify anxiety, depression, self-esteem, and well-being. We showed that during the peri-ovulatory phase, in the alpha band, the leaf fraction and the tree hierarchy of the brain network were reduced, while the betweenness centrality (BC) of the right posterior cingulate gyrus (rPCG) was increased. Furthermore, the increase in BC was predicted by estradiol levels. Moreover, during the luteal phase, the variation of estradiol correlated positively with the variations of both the topological change and environmental mastery dimension of the well-being test, which, in turn, was related to the increase in the BC of rPCG. Our results highlight the effects of sex hormones on the large-scale brain network organization as well as on their possible relationship with the psychological state across the MC. Moreover, the fact that physiological changes in the brain topology occur throughout the MC has widespread implications for neuroimaging studies.

Keywords: behavior; emotional stimuli; magnetoencephalography; posterior cingulate gyrus; premenstrual dysphoric disorder; sex hormones.

FIGURE 1

Data analysis pipeline. (a) Neuronal activity recorded in the sensors space using a magnetoencephalography (MEG). Activity in the Alpha frequency (8–13 Hz) is represented. (b) Raw MEG signals recorded by 163 sensors (a subset is shown here). (c) Cardiac (upper) and blinking (lower) artifacts as estimated by the independent component analysis (ICA). (d) Sensors signals after pre‐processing and cleaning. (e) Structural magnetic resonance image (MRI) of a subject. (f) The structural MRI, the MEG sensors and the head of the subject are co‐registered in the same coordinate system. (g) Through a Beamformer algorithm, source time series are estimated in regions of interest within the brain, according to a parcellation based on the ALL atlas. (h) Functional connectivity matrix estimated using the phase linearity measurement (PLM) which is calculated between each pair of 90 brain regions. In the matrix, rows and columns are the regions of interest, while the entries are the estimated values of the PLM. (i) Brain topology representation based on the minimum spanning tree (MST) reconstruction, where brain regions are represented by red dots and edge are represented as lines. (j) Once a frequency‐specific MST has been obtained, it is characterized by topological parameters

FIGURE 2

Brain topology comparison. The box plots refer to the comparison of topological parameters in alpha band, estimated in 24 women (N = 24) during the MC. From left to right, (a) the BC of the right posterior cingulate gyrus (rPCG), (b) the leaf fraction (Lf), and (c) the tree hierarchy (Th), respectively. In each box plot, the values are shown at early follicular (T1), peri‐ovulatory (T2) and mid‐luteal (T3) phase. The upper and lower bound of the rectangles refer to the 25th to 75th percentiles, the median value is represented by a horizontal line inside each box, the whiskers extend to the 10th and 90th percentiles, and further data are considered as outliers and represented by the filled circles. From left to right, the box plots show the significantly higher BC in the rPCG during the peri‐ovulatory phase, as compared to the early follicular (p = 0.0003) and mid‐luteal (p = 0.0055) phases, and the reduction in the network integration in peri‐ovulatory phase as compared to both the early follicular (Lf p = 0.016; Th p = 0.032) and the mid‐luteal (Lf p = 0.004; Th p = 0.006) phases. Significance p value: *p < 0.05, **p < 0.01, ***p < 0.001

FIGURE 3

Multilinear model with leave‐one‐out cross‐validation (LOOCV). The model aims to predict the topological variation of the brain network expressed by the BC changes during the MC (Δ T2–T1 and Δ T3–T2) of the right posterior cingulate gyrus (rPCG). (a) Explained variance of the additive model composed of four nuisance variables (repeated measures, age, education, cycle length), and four predictors (progesterone, luteinizing hormone (LH), follicle‐stimulating hormone (FSH), estradiol). Significant predictor in underlined text; positive coefficient indicated with β+. (b) Scatter plot of the Observed topological values versus the topological values predicted by the model with LOOCV. (c) Scatter plot of the standardized residuals (standardization of the difference between observed and predicted (LOOCV) values). The distribution results symmetrical with respect to the 0, with a standard deviation lower than 2.5

FIGURE 4

Correlation between topological data and hormone blood levels. Spearman's correlation between the Δ values (here expressed as the difference between the mid‐luteal (T3) and peri‐ovulatory (T2) phases of the MC) of the BC of the right posterior cingulate gyrus (rPCG) and the Δ values of the estradiol levels along the MC (p = 0.007, p _FDR = 0.028)

FIGURE 5

Correlation between topological data and psychological dimensions of well‐being test. Spearman's correlation between the Δ values (here expressed as the difference between the mid‐luteal (T3) and peri‐ovulatory (T2) phases of the MC) of the BC of the right posterior cingulate gyrus (rPCG) and the Δ values of the psychological dimension of well‐being test (Environmental mastery scores) along the MC (p = 0.007, p _FDR = 0.043)

FIGURE 6

Correlation between hormone blood levels and psychological dimensions of well‐being test. Spearman's correlation between the Δ values (here expressed as the difference between the mid‐luteal (T3) and peri‐ovulatory (T2) phases of the MC) of estradiol and the Δ values of the psychological dimension of well‐being test (Environmental mastery scores) along the MC (p < 0.001, p _FDR < 0.001)