Altered estradiol-dependent cellular Ca2+ homeostasis and endoplasmic reticulum stress response in Premenstrual Dysphoric Disorder

Abstract

Premenstrual Dysphoric Disorder (PMDD) is characterized by debilitating mood symptoms in the luteal phase of the menstrual cycle. Prior studies of affected women have implicated a differential response to ovarian steroids. However, the molecular basis of these patients' differential response to hormone remains poorly understood. We performed transcriptomic analyses of lymphoblastoid cell lines (LCLs) derived from women with PMDD and asymptomatic controls cultured under untreated (steroid-free), estradiol-treated (E2), and progesterone-treated (P4) conditions. Weighted gene correlation network analysis (WGCNA) of transcriptomes identified four gene modules with significant diagnosis x hormone interactions, including one enriched for neuronal functions. Next, in a gene-level analysis comparing transcriptional response to hormone across diagnoses, a generalized linear model identified 1522 genes differentially responsive to E2 (E2-DRGs). Among the top 10 E2-DRGs was a physically interacting network (NUCB1, DST, GCC2, GOLGB1) involved in endoplasmic reticulum (ER)-Golgi function. qRT-PCR validation reproduced a diagnosis x E2 interaction (F(1,24)=7.01, p = 0.014) for NUCB1, a regulator of cellular Ca2+ and ER stress. Finally, we used a thapsigargin (Tg) challenge assay to test whether E2 induces differences in Ca2+ homeostasis and ER stress response in PMDD. PMDD LCLs had a 1.36-fold decrease in Tg-induced XBP1 splicing response compared to controls, and a 1.62-fold decreased response (p = 0.005), with a diagnosis x treatment interaction (F(3,33)=3.51, p = 0.026) in the E2-exposed condition. Altered hormone-dependent in cellular Ca2+ dynamics and ER stress may contribute to the pathophysiology of PMDD.

Trial registration: ClinicalTrials.gov NCT00001259 NCT00001322.

Conflict of interest statement

DRR serves on the Clinical Advisory Board of Sage Therapeutics, and reports consulting fees and travel reimbursement from Sage Therapeutics.

© 2021. This is a U.S. government work and not under copyright protection in the U.S.; foreign copyright protection may apply.

Figures

Fig. 1. Functional analysis of gene modules from hormone-exposed PMDD LCLs.

a Cluster dendrogram of gene modules. In WGCNA analysis of 46 input transcriptomes, 21,746 unique gene features (“genes”) were clustered into 65 co-expression modules using gene–gene correlation patterns across all input transcriptomes. b Module eigengenes showing significant diagnosis x hormone interactions. Eigengenes (defined as the first principal component of a given module) of four modules (Brown, Pink, Turquoise, and Magenta) showed differential regulation by hormone (ANOVA interaction P < 0.05). Closed circles denote controls. Open circles denote PMDD. Note: An extreme outlier in the E2-treated control group was excluded in the eigengene plot of the Magenta module for scaling purposes. Hormone x diagnosis interaction was significant both before and after exclusion of this outlier. P-values are not adjusted for multiple comparisons. c Module membership of ESC/E(Z) complex genes. Nearly half (6/13) of ESC/E(Z) complex genes were assigned to the Turquoise module. Remaining ESC/E(Z) genes were clustered into the Blue, AntiqueWhite4 (aw4), DarkOliveGreen (dog), Cyan (cy), Green (g), and SkyBlue2 (sb2) modules. d ESC/E(Z) complex genes in the Turquoise module. Genes plotted according to module membership and strength of association with diagnosis. ESC/E(Z) complex genes (SUZ12, EED, AEBP2, HDAC2, MTF2, SIRT2) are more likely to have high module membership within the Turquoise module. e Word cloud representations of GSEA results. Top 20 GSEA results were used to generate word clouds for each differentially regulated module. Word size is proportional to frequency of word among GO terms identified by GSEA results. Modules were enriched for terms relating to MAP Kinase and GTPase signaling (Brown, top left), chromatin remodeling and proteasome function (Turquoise, top right), mitochondrial function (Pink, bottom left), and neuronal and synaptic function (Magenta, bottom right). Full, quantitative GSEA results are tabulated in Supplemental Table S3. Word clouds generated via worditout.com.

Fig. 2. Top E2-DRGs implicate ER–Golgi function and Ca2+ homeostasis in PMDD.

a GeneMANIA visualization of select E2-DRGs. Key genes relating to ER–Golgi function and intracellular calcium homeostasis (NUCB1, DST, GCC2, GOLGB1) were selected from among the top 10 E2-DRGs and included as inputs for GeneMANIA network visualization (genemania.org, accessed January 2019). The resulting output revealed a structural network of Golgi–ER-associated genes, with physical interactions comprising 67.6% of the weighted network. b qRT-PCR validation of select E2-DRGs. NUCB1, DHCR7, PPP2R5D, and GOLGB1 were selected for qRT-PCR validation in an independent replication cohort of 13 cases and 13 controls. NUCB1 qRT-PCR results revealed a significant main effect of diagnosis (F(1,24)=4.61, p = 0.042) and interaction of diagnosis and hormone (F(1,24)=7.01, p = 0.014). In post-hoc testing, E2-treated PMDD samples had significantly reduced NUCB1 expression compared to control samples (p = 0.003). qRT-PCR results for DHCR7 did not reveal significant effects of diagnosis, hormone, or their interaction, but in multiple comparisons testing, DHCR7 expression was significantly reduced in E2-treated PMDD samples compared to E2-treated control samples (p = 0.040). No significant differences for PPP2R5D or GOLGB1 were found in qRT-PCR validation.

Fig. 3. Thapsigargin (Tg) challenge differentiates PMDD and Control LCLs.

a Experimental design of Tg challenge assays. PMDD and control LCLs were cultured in phenol red-free RPMI for 5–6 days and treated in four arms: (1) no treatment (vehicle), (2) E2 (100 nM) for 24 h, (3) Tg (0.2 uM) and CaCl2 (0.8 mM) for 4 h, and (4) E2 (100 nM) for 24 h with Tg (0.2 uM) and CaCl2 (0.8 mM) for the last 4 h. Cells were harvested and assayed for the spliced-to-unspliced ratio of XBP1 by qRT-PCR. b XBP1 S/US ratios in Tg challenge assays (by diagnosis group). Control LCLs showed an increase in XBP1 S/US ratio after Tg treatment. Pre-treatment with E2 potentiated this response. Both of these effects were blunted or absent in PMDD LCLs. A significant diagnosis x treatment interaction (ANOVA-RM F(3,33) = 3.508, p = 0.026) in XBP1 splicing response to Tg was observed. Post-hoc analysis revealed a significant difference between control and PMDD LCLs only when cells were treated with the combination of E2 + Tg (p = 0.005). E2 had a significant modulatory effect on Tg response in control cells (p = 0.007) but not in PMDD cells (p = 0.693) (see Supplemental Table S6). c S/US XBP1 ratios in Tg challenge assays (by individual cell line). Examining S/US XBP1 ratios of individual cell lines across treatment conditions shows qualitatively distinct trajectories between control and PMDD groups, most notably, the loss of E2-mediated potentiation of XBP1 splicing in PMDD LCLs. However, significant within-group heterogeneity among cell lines is also appreciated.