Hormonal milieu influences whole-brain structural dynamics across the menstrual cycle using dense sampling in multiple individuals

Abstract

Gonadal hormone receptors are widely distributed across the brain, yet their influence on brain structure remains understudied. Here, using precision imaging, we examined four females, including one with endometriosis and one using oral contraceptives (OC), across a monthly period. Whole-brain analyses revealed spatiotemporal patterns of brain volume changes, with substantial variations across the monthly period. In typical cycles, spatiotemporal patterns were associated with serum progesterone levels, while in cycles with endometriosis and during OC intake, patterns were associated with serum estradiol levels. The volume changes were widely distributed rather than region-specific, suggesting a widespread but coordinated influence of hormonal fluctuations. These findings underscore the importance of considering diverse hormonal milieus beyond typical menstrual cycles in understanding structural brain dynamics and suggest that hormonal rhythms may drive widespread structural brain changes.

Fig. 1. Hormone concentrations of estradiol, progesterone and the progesterone-to-estradiol ratio for female participants (n = 4).

a, Hormones concentrations across the test sessions for female participants (n = 4). Solid lines and colored shaded areas represent hormonal levels; gray shading indicates menses in typical cycles and the endometriosis cycle, and inactive pill phase in the oral contraceptives (OC) cycle; dashed lines indicate ovulation. Hormone levels indicate a typical hormonal balance in the typical and 28andMe (typical) cycles, while hormone levels in the endometriosis and OC cycles suggest estradiol dominance. b, To test whether hormonal profiles differed among the four individuals, a one-way MANOVA was conducted, followed by a post hoc ANOVAs, and two-sided post hoc t-tests. The box-and-whisker plots show the median (centerline), upper and lower quartiles (box), minimum and maximum values (whiskers); individual points are shown. Asterisks indicate significance level (***P < 0.001, **P < 0.01, *P < 0.05) based on two-sided post hoc t-tests with Bonferroni correction for multiple comparisons. For exact P values, see Main. Graphs were created with GraphPad Prism (version 10). NS, nonsignificant; MANOVA, multivariate analysis of variance; ANOVAs, analyses of variance. Source data

Fig. 2. Schematic illustration of data assessment, processing workflow and data reduction.

T1w images were assessed over a 4–5-week period for each participant. Images were then preprocessed using the longitudinal pipeline approach in CAT12. Next, SVD was applied to decompose the preprocessed images into spatial and temporal components. Spatial components represent changes in brain volumes and cortical thickness across different regions, while temporal components reflect how these spatial components evolve over time. Warm and cool colors in the spatial component represent positive (warm colors) and negative (cool colors) associations between spatial components and temporal patterns. This suggests that regions marked in warm colors increase as the associated temporal pattern increases, while those in cool colors decrease. Note that the spatial and temporal components shown are examples and do not represent actual results. Graphs were created with GraphPad Prism (version 10). SVD, singular value decomposition; TPM, tissue probability maps.

Fig. 3. Volumetric and cortical thickness spatial patterns that explained at least 10% of the variance across the female participants (n = 4).

a, The spatial patterns illustrate the volumetric patterns of involved brain regions that change over time across the female participants (n = 4; the endometriosis, oral contraceptives (OC), typical and 28andMe (typical) cycle). b, The spatial patterns illustrate the cortical thickness patterns of involved brain regions that change over time across the female participants (n = 4; the endometriosis, OC, typical and 28andMe (typical) cycle). For a and b, volumetric and cortical thickness spatial patterns were derived using SVD. Spatial weights were thresholded, retaining only values within the ranges of −0.1 to −0.01 and 0.01 to 0.1, while excluding values between −0.01 and 0.01 that indicate minimal contribution to the respective spatial pattern (color bar). Source data

Fig. 4. VSTPs across the different female cycles (n = 4).

This figure depicts VSTPs across the endometriosis cycle, the oral contraceptives (OC) cycle, the typical cycle and the 28andMe (typical) cycle. a, VSTP1 shows spatial distribution of brain regions involved in component 1 (left) and the associated temporal dynamics (right). Warm colors in the spatial map indicate regions with positive associations to the temporal pattern (indicating regional volume increases as the temporal pattern increases). Cool colors in the spatial map indicate negative associations to the temporal pattern (reflecting regional volume decreases as the temporal pattern increases). b, VSTP2 shows spatial distribution of brain regions involved in component 2 (left) and the associated temporal dynamics (right). Warm colors in the spatial map indicate regions with positive associations to the temporal pattern (indicating regional volume increases as the temporal pattern increases). Cool colors in the spatial map indicate negative associations to the temporal pattern (reflecting regional volume decreases as the temporal pattern increases). c, VSTP3 shows spatial distribution of brain regions involved in component 3 (left) and the associated temporal dynamics (right). Warm colors in the spatial map indicate regions with positive associations to the temporal pattern (indicating regional volume increases as the temporal pattern increases). Cool colors in the spatial map indicate negative associations to the temporal pattern (reflecting regional volume decreases as the temporal pattern increases). For a–c, volumetric and cortical thickness spatial patterns were derived using SVD. Spatial weights were thresholded, retaining only values within the ranges of −0.1 to −0.01 and 0.01 to 0.1, while excluding values between −0.01 and 0.01 that indicate minimal contribution to the respective spatial pattern (color bar). Solid black lines represent standardized eigenvectors (temporal pattern); dashed colored lines represent square-rooted and standardized hormonal values; gray shading indicates menses in typical cycles and the endometriosis cycle, and inactive pill phase in the OC cycle; dashed lines indicate ovulation. Asterisks indicate significant time-series regressions between hormone levels and the spatiotemporal patterns after FDR correction for multiple comparisons was performed. For exact P values, see main text. Plots were created with GraphPad Prism (version 10). VSTPs, volumetric spatiotemporal patterns. Source data

Fig. 5. CSTPs across the different female cycles (n = 4).

This figure depicts CSTPs across the endometriosis cycle, the OC cycle, the typical cycle and the 28andMe (typical) cycle. a, CSTP1 shows spatial distribution of brain regions involved in component 1 (left) and the associated temporal dynamics (right). Warm colors in the spatial map indicate regions with positive associations to the temporal pattern (indicating regional cortical thickness increases as the temporal pattern increases). Cool colors in the spatial map indicate negative associations to the temporal pattern (reflecting regional cortical thickness decreases as the temporal pattern increases). b, CSTP2 shows spatial distribution of brain regions involved in component 2 (left) and the associated temporal dynamics (right). Warm colors in the spatial map indicate regions with positive associations to the temporal pattern (indicating regional cortical thickness increases as the temporal pattern increases). Cool colors in the spatial map indicate negative associations to the temporal pattern (reflecting regional cortical thickness decreases as the temporal pattern increases). For a and b, volumetric and cortical thickness spatial patterns were derived using SVD. Spatial weights were thresholded, retaining only values within the ranges of −0.1 to −0.01 and 0.01 to 0.1, while excluding values between −0.01 and 0.01 that indicate minimal contribution to the respective spatial pattern (color bar). Solid black lines represent standardized eigenvectors (temporal pattern); dashed colored lines represent square-rooted and standardized hormonal values; gray shading indicates menses in typical cycles and the endometriosis cycle, and inactive pill phase in the OC cycle; dashed lines indicate ovulation. Asterisks indicate significant time-series regressions between hormone levels and the spatiotemporal patterns after FDR correction for multiple comparisons was performed. For exact P values, see main text. Plots were created with GraphPad Prism (version 10). CSTPs, cortical thickness spatiotemporal patterns. Source data

Fig. 6. Significant voxels associated with hormone concentrations in the female participants (n = 4).

a, The significant voxel-wise associations across all four cycles (n = 4; endometriosis cycle, oral contraceptives (OC) cycle, typical cycle and 28andMe (typical) cycle). b, The presentation of the significant voxels for each cycle separately (endometriosis cycle, n = 1; OC cycle, n = 1; typical cycle, n = 1; and 28andMe (typical) cycle, n = 1). For a and b, GLMs were used for vertex-wise analysis with the TFCE method that controls for multiple comparisons by applying an FWE correction. Hormone concentrations were square-rooted. Positive associations are displayed in red, negative associations are displayed in blue, with P values ranging from 0.01 to 0.0001 (color bar). GLMs, general linear models; TFCE, threshold-free cluster enhancement; FWE, family-wise error; Ratio, progesterone-to-estradiol ratio. Source data

Fig. 7. Significant vertices associated with hormone concentrations in the female participants (n = 4).

a, The significant vertex-wise associations across all four cycles (n = 4; endometriosis cycle, oral contraceptives (OC) cycle, typical cycle and 28andMe (typical) cycle). b, The presentation of the significant vertices for each cycle separately (endometriosis cycle, n = 1; OC cycle, n = 1; typical cycle, n = 1; and 28andMe (typical) cycle, n = 1). For a and b, GLMs were used for vertex-wise analysis with the TFCE method that controls for multiple comparisons by applying an FWE correction. Hormone concentrations were square-rooted. Only positive associations were observed, with P values ranging from 0.01 to 0.0001 (color bar). Ratio, progesterone-to-estradiol ratio. Source data

Extended Data Fig. 1. Timeline of the data collection for the female participants (n = 4).

(a) The timeline of data collection for the typical cycle (n = 1). (b) The timeline of data collection for the 28andMe (typical) cycle (n = 1). (c) The timeline of data collection for the endometriosis cycle (n = 1). (d) The timeline of data collection for the oral contraceptives cycle (n = 1). For (a–d) MRI (MRI symbol), blood draw for hormonal assessments (syringe symbol), and mood questionnaires (paper symbol) were acquired simultaneously on each test day. Purple timeline bars represent the endometriosis cycle, the oral contraceptives cycle, and the typical cycle acquired in Jena, Germany. The turquoise timeline bar represents the 28andMe (typical) cycle acquired in Santa Barbara, California, USA.

Extended Data Fig. 2. Timeline of data collection and hormonal levels for the male participant (n = 1).

(a) The timeline of data collection for the male participant (n = 1). MRI (MRI symbol), blood draw for hormonal assessments (syringe symbol), and mood questionnaires (paper symbol) were acquired simultaneously on each test day. Data was acquired in Jena, Germany. (b) The hormonal levels of estradiol, progesterone, and the progesterone-to-estradiol ratio for the male participant across the five-week period. Solid lines and colored shaded areas represent hormonal levels. The male participant presented with anticipated low hormonal concentrations of estradiol, progesterone, and a low progesterone-to-estradiol ratio. Source data

Extended Data Fig. 3. Volumetric spatiotemporal patterns that significantly fluctuated in the male participant (n = 1) across the five-week period.

This figure depicts volumetric spatiotemporal patterns in the male participant that explained at least 10% of the variance. Spatial distribution of brain regions (top) and the associated temporal dynamics (bottom) of volumetric spatiotemporal pattern 1 (VSTP1), volumetric spatiotemporal pattern 2 (VSTP2) and volumetric spatiotemporal pattern 3 (VSTP3) are shown. Warm colors in the spatial maps indicate regions with positive associations to the temporal pattern (indicating regional volume increases as the temporal pattern increases). Cool colors in the spatial maps indicate negative associations to the temporal pattern (reflecting regional volume decreases as the temporal pattern increases). Spatial weights were thresholded, retaining only values within the ranges of −0.05 to −0.01 and 0.01 to 0.05, while excluding values between −0.01 and 0.01 that indicate minimal contribution to the respective spatial pattern (color bar). Solid black lines represent standardized eigenvectors (temporal pattern); dashed colored lines represent square-rooted and standardized hormonal values. Generalized additive models (GAMs) revealed that the volumetric spatiotemporal patterns fluctuated significantly across time. Time-series regressions revealed that the volumetric temporal patterns were not associated with hormonal levels in the male. For exact p-values, see Supplementary Table 10 and 11. Graphs were created using GraphPad Prism (version 10). Source data

Extended Data Fig. 4. Cortical thickness spatiotemporal patterns in the male participant (n = 1) across the five-week period.

This figure depicts cortical thickness spatiotemporal patterns in the male participant that explained at least 10% of the variance. Spatial distribution of brain regions (top) and the associated temporal dynamics (bottom) of cortical thickness spatiotemporal pattern 1 (CSTP1), cortical thickness spatiotemporal pattern 2 (CSTP2) and cortical thickness spatiotemporal pattern 3 (CSTP3) are shown. Warm colors in the spatial maps indicate regions with positive associations to the temporal pattern (indicating regional cortical thickness increases as the temporal pattern increases). Cool colors in the spatial maps indicate negative associations to the temporal pattern (reflecting regional cortical thickness decreases as the temporal pattern increases). Spatial weights were thresholded, retaining only those within the ranges of -0.02 to -0.01 and 0.01 to 0.02, while excluding values between -0.01 and 0.01 that indicate minimal contribution to the respective spatial pattern (color bar). Solid black lines represent standardized eigenvectors (temporal pattern); dashed colored lines represent square-rooted and standardized hormonal values. Generalized additive models (GAMs) revealed that the cortical thickness temporal patterns did not significantly fluctuate across the five-week period and time-series regressions revealed that they were not associated with hormonal levels in the male. For exact p-values, see Supplementary Table 10 and 11. Graphs were created using GraphPad Prism (version 10). Source data

Extended Data Fig. 5. Results from the voxel- and vertex-wise analyses in the male participant (n = 1) across the five-week period.

To directly link hormonal fluctuations to structural brain measures, complementary voxel- and vertex-wise analyses were also conducted using general linear models (GLMs) in the male participant as a sensitivity check with the Threshold-Free Cluster Enhancement (TFCE) method which controls for multiple comparisons by applying a family-wise error (FWE) correction. No significant associations were found in the voxel- and vertex-wise analysis between structural brain measures (volume, cortical thickness) and hormone levels in males. Estradiol, square-rooted estradiol levels; progesterone, square-rooted progesterone levels; ratio, square-rooted progesterone-to-estradiol ratio. Source data