Rapid signaling of estrogen to WAVE1 and moesin controls neuronal spine formation via the actin cytoskeleton

Abstract

Estrogens are important regulators of neuronal cell morphology, and this is thought to be critical for gender-specific differences in brain function and dysfunction. Dendritic spine formation is dependent on actin remodeling by the WASP-family verprolin homologous (WAVE1) protein, which controls actin polymerization through the actin-related protein (Arp)-2/3 complex. Emerging evidence indicates that estrogens are effective regulators of the actin cytoskeleton in various cell types via rapid, extranuclear signaling mechanisms. We here show that 17beta-estradiol (E2) administration to rat cortical neurons leads to phosphorylation of WAVE1 on the serine residues 310, 397, and 441 and to WAVE1 redistribution toward the cell membrane at sites of dendritic spine formation. WAVE1 phosphorylation is found to be triggered by a Galpha(i)/Gbeta protein-dependent, rapid extranuclear signaling of estrogen receptor alpha to c-Src and to the small GTPase Rac1. Rac1 recruits the cyclin-dependent kinase (Cdk5) that directly phosphorylates WAVE1 on the three serine residues. After WAVE1 phosphorylation by E2, the Arp-2/3 complex concentrates at sites of spine formation, where it triggers the local reorganization of actin fibers. In parallel, E2 recruits a Galpha(13)-dependent pathway to RhoA and ROCK-2, leading to activation of actin remodeling via the actin-binding protein, moesin. Silencing of WAVE1 or of moesin abrogates the increase in dendritic spines induced by E2 in cortical neurons. In conclusion, our findings indicate that the control of actin polymerization and branching via moesin or WAVE1 is a key function of estrogen receptor alpha in neurons, which may be particularly relevant for the regulation of dendritic spines.

Fig. 1.

E2 induces dendritic spine formation and dynamic actin remodeling in cortical neurons. A, Neuronal cells were treated for different times with E2 (10 nm). Actin fibers were stained with phalloidin linked to Texas Red (red labeling), and nuclei were counterstained with 4′,6-diamidino-2-phenylindole (blue labeling). The insets show dendritic spines at higher magnification from the white boxes. The yellow boxes on the cells indicate a sample cellular area that is analyzed in the colored graphs. In these graphs, the longitudinal axis represents the gray level, and the horizontal axis indicates pixels. Blue, red, and yellow areas indicate the extracellular, plasma membrane, and cytoplasmic fractions, respectively. B and C, Mean intensity ratio of actin staining in the membrane/cytoplasm in the same experiment. The results are derived from the sampling of five areas of the cell membrane of 40 different random cells. Areas of minimum and maximum cell membrane thickness were always included. The results are expressed as the mean ± sd of the measurements. *, P < 0.05 vs. 0 min. D, Mean number (±sd) of dendritic protrusions per 10 μm dendrite length. *, P < 0.05 vs. 0 min. CON, Control.

Fig. 2.

Estrogen activates actin remodeling in neurons via WAVE1 and moesin. A, Protein extracts from neuronal cells treated with 10 nm E2 for 0–120 min were assayed with Western analysis for their overall content of WAVE1, actin, or P-Ser³¹⁰-, P-Ser³⁹⁷-, and P-Ser⁴⁴¹-WAVE1. B, Densitometric analysis of the P-WAVE1 bands in A. C, Western analysis for wild-type and Thr⁵⁵⁸-P-moesin of neuronal protein extracts treated with E2 (10 nm) for 0–60 min. D, Neuronal cells were treated for different times with E2 (10 nm). Actin fibers were stained with phalloidin linked to Texas Red (red labeling), WAVE1 was stained with fluorescein isothiocyanate (green labeling), and nuclei were counterstained with 4′,6-diamidino-2-phenylindole (blue labeling). E, Mean ± sd thickness of the cell membrane and mean number ± sd of membrane-localized WAVE1, estimated as WAVE1 spots at the cell membrane per cell, indicated as the mean of the measurement of 40 different cells per condition. F, Mean intensity ratio of actin staining in the membrane/cytoplasm in the same experiment. The results are derived from the sampling of five areas of the cell membrane of 40 different random cells. Areas of minimum and maximum cell membrane thickness were always included. *, P < 0.05 vs. 0 min. G, Neuronal cells were transfected with siRNA vs. WAVE1 or with antisense or sense PONs toward moesin for 48 h. WAVE1 expression and cellular localization status was checked by staining for WAVE1 (fluorescein isothiocyanate). Actin fibers were stained with phalloidin-Texas Red, and nuclei were counterstained with 4′,6-diamidino-2-phenylindole. H, Mean ± sd thickness of the cell membrane as well as the intensity of actin staining in the cytoplasm and membrane in the same experiment. Technical details are as in E and F. D and G, The white boxes on the cells indicate a sample cellular area that is analyzed in the colored graphs. In these graphs, the longitudinal axis represents the gray level, and the horizontal axis indicates pixels. Blue, red, and yellow areas indicate the extracellular, plasma membrane, and cytoplasmic fractions, respectively. CON, Control.

Fig. 2.

Estrogen activates actin remodeling in neurons via WAVE1 and moesin. A, Protein extracts from neuronal cells treated with 10 nm E2 for 0–120 min were assayed with Western analysis for their overall content of WAVE1, actin, or P-Ser³¹⁰-, P-Ser³⁹⁷-, and P-Ser⁴⁴¹-WAVE1. B, Densitometric analysis of the P-WAVE1 bands in A. C, Western analysis for wild-type and Thr⁵⁵⁸-P-moesin of neuronal protein extracts treated with E2 (10 nm) for 0–60 min. D, Neuronal cells were treated for different times with E2 (10 nm). Actin fibers were stained with phalloidin linked to Texas Red (red labeling), WAVE1 was stained with fluorescein isothiocyanate (green labeling), and nuclei were counterstained with 4′,6-diamidino-2-phenylindole (blue labeling). E, Mean ± sd thickness of the cell membrane and mean number ± sd of membrane-localized WAVE1, estimated as WAVE1 spots at the cell membrane per cell, indicated as the mean of the measurement of 40 different cells per condition. F, Mean intensity ratio of actin staining in the membrane/cytoplasm in the same experiment. The results are derived from the sampling of five areas of the cell membrane of 40 different random cells. Areas of minimum and maximum cell membrane thickness were always included. *, P < 0.05 vs. 0 min. G, Neuronal cells were transfected with siRNA vs. WAVE1 or with antisense or sense PONs toward moesin for 48 h. WAVE1 expression and cellular localization status was checked by staining for WAVE1 (fluorescein isothiocyanate). Actin fibers were stained with phalloidin-Texas Red, and nuclei were counterstained with 4′,6-diamidino-2-phenylindole. H, Mean ± sd thickness of the cell membrane as well as the intensity of actin staining in the cytoplasm and membrane in the same experiment. Technical details are as in E and F. D and G, The white boxes on the cells indicate a sample cellular area that is analyzed in the colored graphs. In these graphs, the longitudinal axis represents the gray level, and the horizontal axis indicates pixels. Blue, red, and yellow areas indicate the extracellular, plasma membrane, and cytoplasmic fractions, respectively. CON, Control.

Fig. 3.

Estrogen signals to WAVE1 via ERα. A, Cortical neurons were exposed for 20 min to 10 nm E2 in the presence or absence of the pure ER antagonist ICI 182,780 (ICI; 100 nm) or with the ERα-selective ligand PPT (10 nm) or the ERβ-selective agonist DPN (10 nm). Total phosphorylation of WAVE1 was assayed with Western analysis as a gel migration shift in the band. B, Cortical neurons were transfected with siRNA vs. ERα (siRNA ERα) or with vehicle, and protein analysis for ERα, ERβ, actin, or wild-type (WAVE1) or P-Ser³¹⁰-, P-Ser³⁹⁷-, and P-Ser⁴⁴¹-WAVE1 was performed on cell lysates after treatment for 20 min with 10 nm E2. C, The graph displays the quantitative analysis of the intensity of the bands in B, obtained as number of photons measured by the ChemiDoc digital imaging system and evaluated with the Quantity One Software (Bio-Rad, Hercules, CA). CON, Control.

Fig. 4.

ERα signals to WAVE1 via Gαi, Gβ, c-Src, and Rac1. A, Cortical neuronal cells were exposed for 20 min to 10 nm E2, in the presence or absence of the G protein inhibitor PTX (100 ng/ml) or the c-Src inhibitor PP2 (0.2 μm) or after transfection with dominant-negative Gα13 or Gαi constructs or siRNAs vs. Gβ1. Wild-type c-Src or P-Tyr⁴¹⁶-c-Src were assayed in cell extracts. B–D, Neuronal cell protein extracts were immunoprecipitated with antibodies toward the indicated G proteins and co-immunoprecipitation of ERα, ERβ, and Gαi (C) was tested by Western analysis. E, Cortical neurons were treated for 20 min with 10 nm E2, with or without transfection of a dominant-negative Rac1 for 48 h. Rac1, wild-type WAVE1, or P-Ser³¹⁰-, P-Ser³⁹⁷-, and P-Ser⁴⁴¹-WAVE1 were assayed in cell extracts. CON, Control.

Fig. 5.

Estrogen signals to WAVE1 via a Gαi-Gβ/c-Src/Rac1/Cdk5 cascade, whereas signaling to moesin requires a Gα13/RhoA/ROCK2 pathway. A–C, Cortical neurons were treated with E2 in the presence or absence of the specific Cdk5 inhibitor roscovitine (Rosc; 50 μm) or the c-Src inhibitor PP2 (0.2 μm) or after transfection with a dominant-negative Rac1 construct. A and B, Whole protein extracts were immunoprecipitated with an antibody against Cdk5. The IPs were used in a kinase assay to phosphorylate histone H1 (substrate of Cdk5). Phosphorylated histone H1 was detected after Western analysis with a phosphospecific antibody. C, Neuronal cell content of Cdk5, WAVE1, and P-Ser³¹⁰-, P-Ser³⁹⁷-, and P-Ser⁴⁴¹-WAVE1 were assayed with Western analysis. D–F, Cortical neurons were exposed to 10 nm E2 for 20 min with or without transfection of WAVE1 constructs with inactivating mutations of the Ser³¹⁰, Ser³⁹⁷, and Ser⁴⁴¹ Cdk5 phosphorylation sites. D, Western analysis of wild-type or P-Ser³¹⁰-, P-Ser³⁹⁷-, or P-Ser⁴⁴¹-WAVE1 was performed. E, Dendritic spine morphology and number were measured with immunofluorescence after staining actin fibers with phalloidin/Texas Red. Scale bar, 5 μm. F, The graph shows the quantitative analysis of spine density calculated as the number of spines per 10 μm dendrite length, normalized vs. nontransfected, vehicle-treated neurons and expressed as a percentage. The results are expressed as the mean ± sd. *, P < 0.05 vs. E2 in nontransfected cells. G, Cortical neurons were treated with E2 in the presence or absence of the G protein inhibitor PTX (100 ng/ml), the c-Src inhibitor PP2 (0.2 μm), the Cdk5 inhibitor roscovitine (Rosc; 50 μm), or the Rho-associated kinase inhibitor (ROCK-2) Y-27632 (Y, 10 μm) or after transfection with dominant-negative Gα13, Gαi, Rac1, or RhoA constructs or with siRNAs vs. Gβ1. Western analysis for actin, wild-type, and Thr⁵⁵⁸-P-moesin, wild-type, and P-Ser³¹⁰-, or P-Ser³⁹⁷-, or P-Ser⁴⁴¹-WAVE1 was performed. CON, Control.

Fig. 6.

Estrogen signaling to WAVE1 turns into membrane localization of the Arp-2/3 complex. Cortical neurons were treated for 20 min with 10 nm E2 in the presence or absence of the Cdk5 inhibitor roscovitine (Rosc; 20 μm) or the c-Src inhibitor PP2 (0.2 μm) or after transfection with a dominant-negative Rac1 construct. A, Cells were stained with an antibody against Arp-2 (fluorescein isothiocyanate), actin fibers were stained with phalloidin linked to Texas Red, and nuclei were counterstained with 4′,6-diamidino-2-phenylindole. Yellow arrows indicate membrane-localized Arp-2. B, Quantification of the membrane-localized Arp-2 complexes in the different conditions. Results are expressed as percent vs. control cells (mean ± sd). *, P < 0.05 vs. control. Membrane-localized Arp-2 complexes were counted in 40 different cells. CON, Control.

Fig. 7.

Estrogen increases dendritic spine density via WAVE1 and moesin. Cortical neurons were incubated in the presence of 10 nm E2 for 20 min in baseline conditions or after silencing of WAVE1 with specific siRNAs or of moesin with antisense PONs Sense PONs for moesin served as control. A, Dendritic spine morphology and number were measured with immunofluorescence after staining actin fibers with phalloidin/Texas Red. Scale bar, 5 μm. B, The graph shows the quantitative analysis of spine density expressed as the number of spines per 10 μm dendrite length. The results are expressed as the mean ± sd. *, P < 0.05 vs. control. CON, Control.

Fig. 8.

Signaling cascades of ERα to WAVE1 and moesin and dendritic spine formation in cortical neurons.