The estrous cycle modulates hippocampal spine dynamics, dendritic processing, and spatial coding

Abstract

Histological evidence suggests that the estrous cycle exerts a powerful effect on CA1 neurons in mammalian hippocampus. Decades have passed since this landmark observation, yet how the estrous cycle shapes dendritic spine dynamics and hippocampal spatial coding in vivo remains a mystery. Here, we used a custom hippocampal microperiscope and two-photon calcium imaging to track CA1 pyramidal neurons in female mice over multiple cycles. Estrous cycle stage had a potent effect on spine dynamics, with heightened density during periods of greater estradiol (proestrus). These morphological changes were accompanied by greater somatodendritic coupling and increased infiltration of back-propagating action potentials into the apical dendrite. Finally, tracking CA1 response properties during navigation revealed enhanced place field stability during proestrus, evident at the single-cell and population level. These results establish the estrous cycle as a driver of large-scale structural and functional plasticity in hippocampal circuits essential for learning and memory.

Figure 1.

Dendritic spine density is longitudinally modulated by estrous cycle stage. A. Experimental pipeline schematic. Measurements were taken from female mice once every 12 hours across two consecutive estrous cycles (8–10 days). Prior to each session, vaginal lavage samples were taken and stained using Shorr stain (top). Estrous stage in upper left hand corner (D = diestrus, P = proestrus, E = estrus, M = metestrus). Scale bar = 1 mm. Each estrous stage is representative of a unique hormonal profile. Representation of the relative concentrations of ovarian hormones across the four estrous stages are shown, including 17β-estradiol (E2, navy), progesterone (P4, blue), luteinizing hormone (LH, yellow), and follicle-stimulating hormone (FSH, pink). Classifications were performed using EstrousNet, with transfer learning ResNet50 architecture shown here (top right), including four convolutional modules converging on a pooling and softmax classification layer. We then used two-photon imaging to track the structure or functional responses of hippocampal neurons (bottom) to determine changes across sessions. A schematic illustrating microperiscope implantation and light path for hippocampal imaging in a Thy1-GFP-M mouse demonstrates the imaging technique for dendritic experiments. Representative images are shown for dendritic spines (scale bar = 10 μm), whole dendrites (scale bar = 50 μm), and CA1 population imaging (scale bar = 100 μm). Finally, imaging data is matched with estrous stage to analyze changes in neuronal structure and function. B. Example average projection of the transverse imaging plane of CA1 through the microperiscope in a mouse sparsely expressing Thy1-GFP-M. C. Weighted projection (see Methods) of the apical dendrite shown in the dashed box of (B). D. Filtered and binarized image (see Methods) of the dendrite shown in (C), to allow for the identification and classification of individual dendritic spines, and to mask extraneous projections. E. Binarized thresholded projections of a gaussian-averaged dendrite from apical CA1, taken 24 hours apart in the four archetypal stages of estrous. Circles indicate spine addition, triangles indicate spine subtraction, both relative to diestrus. Color indicates spine type (yellow = filopodium, red = thin, blue = stubby, navy = mushroom), with legend representative of the classic morphological features of each type. F. Percent spine density change from baseline (averaged across stages) for each recorded dendrite across consecutive recordings taken at 12 hour intervals. Stage lengths are interpolated to 24 hrs to allow plotting of all dendritic segments (see Methods; D = diestrus, P = proestrus, E = estrus, M = metestrus). Blue lines indicate individual dendritic segment time courses, while the bolded navy line indicates mean spine density ± standard error. Wilcoxon signed-rank test against a grand mean of all dendrites included at each timepoint. **p < 0.01, ***p < 0.001, ****p < 0.0001. G. The number of spines of each type (filopodium (yellow), thin (red), stubby (blue), and mushroom (navy)) per 10 μm section of CA1 apical dendritic segment for each estrous stage, mean ± standard error. Bars indicate significant modulation across all stages for thin and stubby spines (linear mixed effects model). *p < 0.1, **p < 0.01.

Figure 2.

Dendritic spine properties dynamically shift across the estrous cycle. A. Survival fraction of all spines present on recording n that were present on recording 1 (at 12 hr intervals) for all spines that were present on the first recording session (light blue), as well as spines that spontaneously appeared during proestrus (navy), where recording 1 is considered the time point during proestrus at which the spine was first observed. Mean ± standard error. B. The proportion of classical spine types across the entire dendritic spine population. Yellow = filopodium, red = thin, blue = stubby, navy = mushroom. C. The proportion of spine types for spines that appeared during proestrus but were immediately pruned in the recording following proestrus (transient spines). D. The proportion of spine types for spines that appear during proestrus and are maintained throughout the entire next cycle (stable spines). E. Percent turnover by session of all spines, analyzed respective to spine type (yellow = filopodium, red = thin, blue = stubby, navy = mushroom). Mean ± standard error. F. Transition matrix for all spines that were present in diestrus and recorded until proestrus (D>P), and spines that were present in proestrus and recorded until estrus (P>E). Spines are classified in order of least to most stable (NS = no spine, F = filopodium, T = thin, S = stubby, and M = mushroom), as established by the survival fraction shown in (E). Transition probability is shown by brightness, where completely dark is 0% transition probability and completely bright is 100% transition probability. Transitions to more stable spine types are pseudocolored in red, persistence within the same type is shown in white, and transitions to less stable types are shown in blue.

Figure 3.

Estrous modulates somatodendritic coupling and action potential propagation. A. Schematic showing microinjection of dilute CaMKII-Cre AAV into TIT2L-GCaMP6s transgenic mouse for sparse, stable dendritic expression prior to implantation of hippocampal microperiscope. B. Average projection of sparse CA1 neurons expressing GCaMP6s throughout the somatodendritic axis, as viewed through the microperiscope. C. Confocal image of a TIT2L-GCaMP6s mouse with viral injection of dilute CaMKII-Cre into CA1 in a coronal slice from a mouse implanted with a microperiscope (boundaries of implantation shown in dotted box). D. Schematic of the air-lifted carbon fiber floating track that the mice explored during imaging. Average projection of an example place cell showing somatic and apical dendritic ROI selection, from which ΔF/F is extracted. E. Average projection of an example pyramidal cell in CA1 showing soma and apical dendrite ROI selection. F. Somatic and apical dendritic ΔF/F across a 40 second interval as the animal traverses a floating chamber. Dendritic ΔF/F (cyan) is normalized to somatic ΔF/F (blue). G. Correlations between somatic and dendritic ΔF/F from the entire recording across estrous stages. Significance between correlations across stages given by pairwise linear mixed effect models. Mean ± standard error. **p < 0.01, ****p < 0.0001. H. Average projection of an example pyramidal cell in CA1 showing somatic and subROI selection, with three highlighted 6 μm-wide subROIs ~50 μm apart along the dendritic segment. I. Sub-ROI ΔF/F at increasingly distal points along the main apical dendritic branch (orange = most proximal, green = medial, yellow = most distal), normalized to somatic signal (top, blue). J. Example pyramidal cell response from a recording taken in proestrus. Forty seconds of normalized somatic ΔF/F (blue, top) is shown, aligned with a logical trace (red, top) set to true during frames when a bAP occurred. A heatmap shows the response of each subROI over the same 40 second epoch. The subROI response during each bAP is normalized to mean somatic ΔF/F during the bAP, then all bAP responses are averaged to create a representative bAP curve, which is fit to an exponential to determine the length constant (right). K. Example pyramidal cell response from a recording taken in estrus. Somatic signal, subROI signal, and bAP fitting is shown as described in J. L. Correlations between soma and increasingly distal subROIs along the dendritic branch across estrous stages. Bar indicates significant modulation across all stages (linear mixed effects model). Mean ± standard error. ****p < 0.0001. M. Distributions of the average length constant from all CA1 pyramidal cells analyzed across estrous stages. Pairwise linear mixed effect model. Mean ± standard error. *p < 0.1, ****p < 0.0001.

Figure 4.

Changes in environmental cues induce remapping in CA1 place cells. A. Three-dimensional schematic of a plug implant for top-down two-photon calcium imaging of CA1 cell bodies in CaMKIIa-Cre × TIT2L-GCaMP6s mice. B. Average projection of CA1 somata through the glass plug implant, scale bar = 100 μm. Inset, ROI selection from the region shown in dashed box, with randomized color scheme. C. Schematic of floating chamber experimental design, where the mouse is head-fixed for two-photon imaging. The walls of the chamber are lined with fluorescent local visual cues and the base of the chamber includes interchangeable textural cues. D. Experimental design schematic, where an open-top box represents a freely moving mouse during the acclimation period, and a closed box represents a head-fixed mouse during the recording period. Mice are allowed to run 10 laps, motivated by water reward, before being moved to the resting cage and subsequently introduced to the next environment. E. Angular position output of the floating chamber tracking system aligned with ΔF/F transients of a single CA1 place cell across three environments : A, B, and A’. ΔF/F transients are shown in red in the angular ΔF/F trace for visualization. F. Smoothed ΔF/F transients from six example cells across laps around the floating chamber (top), as well as lap-averaged one-dimensional tuning curves (bottom; mean ± sem). G. Single-cell angular ΔF/F from six example cells averaged across laps as the mouse moves from environment A (top left) to environment B (top center), and back to environment A (A’; top right). A = blue, B = orange, A’ = purple. Note similar place fields in environment A and A’, but shifted place field location in environment B. H. Place cell responses in CA1 from an example recording across the three environments: A (blue), B (orange), A’ (purple). Average responses (normalized ΔF/F) of all place cells are ordered by peak position in environment A for all three environments. Responses in environment A are cross validated by determining peak position in odd trials and plotting even trials. Note that place fields are in similar positions for environment A and A’, but remap in environment B.

Figure 5.

The stability of place fields is modulated by estrous cycle stage. A. Six example cells showing place fields averaged across laps (mean ± standard error) in the proestrus (P) and estrus (E) stages. Positional tuning curves for environment A shown in blue, B in orange, and A’ in purple. Note that neurons imaged during proestrus exhibit more stable place fields between environments A and A’ than neurons imaged during estrus. B. Cross-validated average responses (normalized ΔF/F) of all place cells ordered by peak position in environment A in each estrous stage. D = diestrus, P = proestrus, E = estrus, M = metestrus. C. Distribution of the circular distance between place cell peaks across environments (either A to A’, i.e., stability, or A to B, i.e., remapping) in each estrous stage. Box plots show mean (horizontal line), first and third quartile (box), and maximum and minimum (whiskers). A dotted line at zero is shown for reference. Pairwise linear mixed effects model. ***p < 0.001, **p < 0.01. D. Population vector correlation distributions across estrous stages in the same environment A (stability; A>A’) and different environments (remapping; A>B). Distributions were created by bootstrapping data from n = 6 mice and n = 3679 cells over 100 iterations, sampling with replacement. A dotted line at zero is shown for reference. Overlay indicates mean ± bootstrapped 95% CI. Asterisks denote significance. Significance was defined as lower 5% CI higher than higher 95% CI. E. Probability density plot of the average prediction accuracy from decoders predicting position in environment A’ (A>A’, top) compared to environment B (A>B, bottom). Responses for each of the four estrous stages are averaged across one recording per stage for each of n = 6 mice.