Estrogen Mobilizes a Subset of Estrogen Receptor-α-Immunoreactive Vesicles in Inhibitory Presynaptic Boutons in Hippocampal CA1

Abstract

Although the classical mechanism of estrogen action involves activation of nuclear transcription factor receptors, estrogen also has acute effects on neuronal signaling that occur too rapidly to involve gene expression. These rapid effects are likely to be mediated by extranuclear estrogen receptors associated with the plasma membrane and/or cytoplasmic organelles. Here we used a combination of serial-section electron microscopic immunocytochemistry, immunofluorescence, and Western blotting to show that estrogen receptor-α is associated with clusters of vesicles in perisomatic inhibitory boutons in hippocampal CA1 and that estrogen treatment mobilizes these vesicle clusters toward synapses. Estrogen receptor-α is present in approximately one-third of perisomatic inhibitory boutons, and specifically in those that express cholecystokinin, not parvalbumin. We also found a high degree of extranuclear estrogen receptor-α colocalization with neuropeptide Y. Our results suggest a novel mode of estrogen action in which a subset of vesicles within a specific population of inhibitory boutons responds directly to estrogen by moving toward synapses. The mobilization of these vesicles may influence acute effects of estrogen mediated by estrogen receptor-α signaling at inhibitory synapses.

Introduction

The classical mechanism of estrogen action involves activation of nuclear transcription factor receptors, estrogen receptor-α (ERα) (Greene et al., 1986; Koike et al., 1987) and/or ERβ (Kuiper et al., 1996), and subsequent regulation of gene expression. However, estrogens also exert acute effects on synaptic physiology that occur too rapidly to involve changes in gene expression (Teyler et al., 1980; Wong and Moss, 1992; Rudick and Woolley, 2003). These rapid effects have been attributed variously to extranuclear ERα/ERβ or to novel ERs (Qiu et al., 2006). The ER(s) mediating the rapid effects of estrogen could be located at the plasma membrane and/or in cytoplasmic compartments. Some studies have reported extranuclear immunoreactivity for ERα (Blaustein et al., 1992; Milner et al., 2001) or ERβ (Mitra et al., 2003) in dendrites, axons, and glia in the hypothalamus and hippocampus. However, these studies are primarily qualitative, and little is known about how extranuclear ERs might influence neuronal function.

The dorsal CA1 region of the hippocampus is a likely site for extranuclear ER action. Estrogen has profound effects on synaptic structure and function in CA1, yet few neurons in this region express a nuclear ER (Hart et al., 2001; Blurton-Jones and Tuszynski, 2002; Mitra et al., 2003). Estrogen increases dendritic spine and excitatory synapse numbers in CA1 (Woolley and McEwen, 1992; Adams et al., 2001), increases the sensitivity of CA1 pyramidal cells to excitatory synaptic input (Woolley et al., 1997; Rudick and Woolley, 2001), and increases dorsal hippocampal seizure susceptibility (Terasawa and Timiras, 1968). Both in vitro (Murphy et al., 1998) and in vivo (Rudick and Woolley, 2001; Rudick et al., 2003) studies suggest that estrogen regulates excitatory synaptic connectivity through an initial suppression of inhibitory GABAergic synaptic transmission. Consistent with a direct effect of estrogen on GABAergic neurons, nearly all of the neurons in dorsal CA1 that express a nuclear ER are GABAergic (Hart et al., 2001). These cells express ERα; although some CA1 cells express ERβ mRNA (Shughrue et al., 1997), there is little evidence for ERβ protein in CA1 (Blurton-Jones and Tuszynski, 2002; Mitra et al., 2003). However, despite the fact that nuclear ERα in dorsal CA1 is limited to GABAergic neurons, only 5–14% of these neurons express ERα (Hart et al., 2001). Thus, it appears that too few cells express a nuclear ER to account entirely for the widespread effects of estrogen on CA1 synapses.

In addition to the few ERα-positive nuclei in CA1, there is substantial ERα immunoreactivity (IR) in the neuropil of this region. Because extranuclear ERα-IR is particularly distinct in the cell body layer, we hypothesized that ERα might be present in axonal boutons of GABAergic basket cells that form perisomatic synapses with CA1 pyramidal cells. Here we used serial-section electron microscopic immunocytochemistry and immunofluorescence to show that clusters of vesicles within a specific neurochemical subpopulation of inhibitory boutons are ERα immunoreactive and that estrogen treatment shifts the location of ERα-immunoreactive vesicle clusters toward synapses. These findings demonstrate a novel mode of estrogen action on presynaptic vesicles.

Materials and Methods

Animals.

All animal procedures were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the Northwestern University Institutional Animal Care and Use Committee. Twenty adult, female Sprague Dawley rats [∼250 g; 10 for bright-field microscopy and electron microscopy (EM), and 10 for immunofluorescence; Harlan, Indianapolis, IN] were ovariectomized under ketamine (85 mg/kg)/xylazine (13 mg/kg, i.p.) anesthesia and 3 d later were given subcutaneous injections of 10 μg of 17β-estradiol benzoate or oil vehicle as described previously (Rudick and Woolley, 2001; Ledoux and Woolley, 2005). Twenty-four hours after injection, animals were deeply anesthetized with sodium pentobarbital (80 mg/kg, i.p.) and perfused either with 2% paraformaldehyde/2% glutaraldehyde in 0.1 m phosphate buffer (PB) for immunolabeling with diaminobenzidine (DAB; Sigma, St. Louis, MO) or with 4% paraformaldehyde in PB for immunofluorescence. Additional adult female rats (∼250 g; Harlan) were perfused with 2% paraformaldehyde/2% glutaraldehyde in PB for immunolabeling with silver-enhanced nanogold particles (Nanoprobes, Yaphank, NY) or with buffer containing protease inhibitors for Western blots. The efficacy of estradiol treatment was verified by visual assessment of the uterus at the time of perfusion.

Immunolabeling for bright-field microscopy and EM.

Immunostaining for ERα was as described previously (Hart et al., 2001). Briefly, perfusion-fixed brains were removed, blocked to contain the hippocampus, postfixed overnight at 4°C, cryoprotected with 30% sucrose, and sectioned (50 or 100 μm), using an SM 2000 R freezing microtome (Leica, Bannockburn, IL), into coronal sections spanning the dorsal hippocampus. To increase antigenicity, tissue was preincubated in 10% sodium borohydride (Sigma) before processing (Leenen et al., 1985). For visualization with DAB, alternating sections were labeled with rabbit polyclonal MC-20 (0.5 μg/ml; Santa Cruz Biotechnology, Santa Cruz, CA) or mouse monoclonal 6F11 (1:50; Novocastra, Newcastle, UK), followed by species-specific biotinylated IgG (1:800; Vector Laboratories, Burlingame, CA), and visualized with DAB using an ABC kit (Vector Laboratories). Neither MC-20 (Azcoitia et al., 1999) nor 6F11 (Bevitt et al., 1997) cross-reacts with ERβ. Some sections were then dehydrated, cleared, and coverslipped under Eukitt for bright-field evaluation of nuclear and exatranuclear labeling. The remaining sections were stained with osmium tetroxide and flat embedded in Eponate resin (Ted Pella, Redding, CA) for EM. For each brain, 200–230 serial thin sections (∼70 nm) were cut on a Reichert Ultracut S ultramicrotome (Leica), collected on formvar-coated slot grids, and stained with uranyl acetate and lead citrate (Ted Pella). Brains were coded before sectioning so that the experimenter was blind to the treatment condition during all phases of image collection, three-dimensional reconstruction, and analysis.

Additional animals were used for MC-20 immunolabeling visualized with silver-enhanced nanogold particles. Tissue was processed identically to that above, except that the primary incubation was followed by nanogold-coupled IgG (1:50) and HQ-silver reagent (Nanoprobes). Sections were then stained with osmium tetroxide and flat embedded in Eponate resin (Ted Pella) for thin sectioning. Short series of thin sections (∼70 nm) were cut and, as above, collected on formvar-coated slot grids and stained with uranyl acetate and lead citrate.

Quantification of ERα-immunoreactive nuclei.

The numbers of ERα-immunoreactive nuclei in the dorsal subiculum, CA1, CA3, and dentate gyrus subregions were estimated using the optical disector/fractionator (West et al., 1991) as described by Hart et al. (2001). Tissue was visualized on an Olympus (Tokyo, Japan) BX60 microscope equipped with a Dage (Michigan City, IN) DC330 camera and ImagePro Plus software (Media Cybernetics, Silver Spring, MD). Briefly, a grid consisting of squares (353 μm²) was superimposed random-systematically over a low-magnification image of every fifth section through the dorsal hippocampus. Each square contained an identically placed counting frame that was 1/16th the area of the square. When any part of the hippocampus was contained within the counting frame, the number of labeled nuclei within the frame was counted. The most intensely labeled nuclei generally were observed in conjunction with lighter cytoplasmic labeling in the soma and most proximal dendrites. Counts were made with a 100× oil-immersion lens starting 3 μm below the surface and continuing in 0.5 μm steps though a depth of 13 μm. The total number of cells was then estimated using the following formula: # cells = ΣQ⁻ (t/h)(1/asf)(1/ssf), where Q⁻ is the total number of cells counted, t is the section thickness (50 μm), h is the height of the disector (10 μm), asf is the area sampling fraction (1/16), and ssf is the section sampling fraction (1/5).

Three-dimensional EM reconstruction.

The dorsal CA1 pyramidal cell layer was imaged using a JEOL (Peabody, MA) 100CX electron microscope. In tissue labeled with MC-20 visualized with DAB, five areas, each containing an ERα-immunoreactive perisomatic bouton, were identified in a section near the middle of each set of serial thin sections. Negatives of these five areas were taken at 10,000× from the middle section and from 40 to 50 serial sections on both sides (five stacks of 80–100 serial negatives per brain). Negatives were scanned at 1600 dpi and aligned with SEM Align software (courtesy of K. M. Harris, University of Texas, Austin, TX, and J. C. Fiala, Boston University, Boston, MA). The plasma membrane, synaptic densities, and mitochondria were traced and reconstructed using IGL Trace software (courtesy of K. M. Harris and J. C. Fiala). The location of each vesicle in a bouton was also marked. Measurements of volume, synaptic area, and numbers of vesicles for each bouton were generated using IGL Trace. The distance between each vesicle and synapse in a bouton was measured using Reconstruct software (courtesy of K. M. Harris and J. C. Fiala) and converted into a measurement of relative distance, with 1.0 being the maximum distance from the synapse that a vesicle could be. For boutons with more than one synapse, the synapse nearest to the ERα-immunoreactive vesicle cluster was used for the relative distance measure.

Western blots.

Western blots were used to confirm MC-20 specificity in dorsal CA1. Buffer-perfused brains were removed, blocked to contain the hippocampus, and sectioned (300 μm), using an OTS 4000 oscillating tissue slicer (EMS, Hatfield, PA), into coronal sections spanning the rostrocaudal extent of the dorsal hippocampus. The CA1 region was dissected from these slices on ice and homogenized in ice-cold 20 mm PBS, pH 7.4, with 2 mm EGTA and 1 mm phenylmethylsulfonyl fluoride (PMSF). The homogenate was lysed with 1.5% Triton X-100, incubated on ice for 30 min, and centrifuged at 15,000 × g for 20 min. The supernatant was collected as the whole-cell fraction and kept frozen at −80°C until use.

Western blots were also used to confirm ERα in cytoplasmic, synaptosomal, and synaptic vesicle fractions isolated from the dorsal hippocampus by differential centrifugation (Huttner et al., 1983). For these experiments, rats were perfused with ice-cold homogenization buffer (320 mm sucrose, 4 mm HEPES-NaOH buffer, pH 7.4, 2 mm EGTA, 1 mm sodium orthovanadate, 0.1 mm PMSF, 50 mm sodium fluoride, 10 mm sodium pyrophosphate, 20 mm glycerophosphate, 1 μg/ml leupeptin, and 1 μg/ml aprotinin). The brains were rapidly removed, and all subsequent steps were performed at 4°C. Dorsal hippocampi were dissected and homogenized, followed by centrifugation at 1000 × g for 10 min to remove large cell fragments and nuclear material. The supernatant was centrifuged at 17,000 × g for 15 min to yield cytoplasmic proteins in the supernatant. The pellet from this spin was resuspended in homogenization buffer and centrifuged at 17,000 × g for an additional 15 min to yield washed synaptosomes. The synaptosomal fraction then was hypo-osmotically lysed and centrifuged at 25,000 × g for 20 min to yield synaptosomal plasma membranes and crude synaptic vesicles. Crude synaptic vesicles then were centrifuged at 160,000 × g for 2 h to pellet the synaptic vesicle fraction. Cytoplasmic proteins, washed synaptosomes, and synaptic vesicle fractions were analyzed. For all Western blots, protein concentration was determined using the Bio-Rad (Hercules, CA) protein assay. The protein sample was mixed with Laemmli sample buffer (62.5 mm Tris-HCl, pH 6.8, 25% glycerol, 2% SDS, 0.01% bromophenol blue, and 5% β-mercaptoethanol), boiled for 5 min, and separated on a 10% SDS-PAGE gel. After electrophoresis, proteins were transferred to polyvinylidene difluoride membranes (Millipore, Billerica, MA). The membrane was blocked in 5% nonfat milk, probed for synaptophysin with MAB5258 (1:1,000,000; Chemicon, Temecula, CA) and/or ERα with MC-20 (0.1 μg/ml), followed by horseradish peroxidase-coupled anti-mouse (1:2000) or anti-rabbit (1:1000) IgG (Vector Laboratories), and visualized using enhanced chemiluminescence (ECL Plus; GE Healthcare, Piscataway, NJ).

Immunofluorescence.

Perfusion-fixed brains were removed, blocked to contain the hippocampus, postfixed overnight at 4°C, cryoprotected with 30% sucrose, and cut on a freezing microtome (Leica) into coronal sections (40 μm) spanning the rostrocaudal extent of the dorsal hippocampus. Sections were systematically distributed into four groups double labeled for ERα (0.1 μg/ml, MC-20) and one of the following: mouse monoclonal glutamic acid decarboxylase 65 (GAD65) (0.3 μg/ml, MAB351; Chemicon), mouse monoclonal parvalbumin (PV) (1:8000, P3088; Sigma), rabbit polyclonal cholecystokinin (CCK) (1:1000, PC206L; Calbiochem, La Jolla, CA), or rabbit polyclonal neuropeptide Y (NPY) (1:4000, N9528; Sigma). Tissue processing was as described previously (Hart et al., 2001), except for visualization with species-specific IgG directly coupled to a fluorochrome. For ERα colocalization, ERα-IR was visualized with Alexa Fluor 568, whereas GAD-, PV-, CCK-, and NPY-IRs were visualized with Alexa Fluor 488 (each 2.5 μg/ml; Invitrogen, San Diego, CA). For NPY colocalization with GAD, NPY-IR was visualized with Alexa Fluor 568, whereas GAD-IR was visualized with Alexa Fluor 488. Incubations were simultaneous in cases in which primaries were raised in different species (ERα and GAD or PV); controls included omission of primary antiserum or antibody. Incubations were sequential in cases in which primaries were raised in the same species (ERα and CCK or NPY). In these cases, a second blocking step was included after the first secondary incubation, and some tissue was processed with the second primary antiserum omitted as an additional control. In these controls, no labeling of the first primary antiserum with the second secondary antiserum was detected. All sections were mounted in order on subbed slides, dehydrated, cleared, rehydrated, coverslipped under Vectashield (Vector Laboratories), and sealed with nail polish.

Slides were coded before image collection, so the experimenter was blind to the treatment condition of each brain during all phases of image collection and analysis. Tissue was imaged using a spinning disc laser confocal system (PerkinElmer, Wellesley, MA) with a 100× oil objective. For each brain, three image stacks in each hemisphere were collected from three sections in each brain, for a total of 18 stacks per brain. The volume of each stack was 21,600 μm³ (ERα and GAD or PV) or 10,800 μm³ (ERα and CCK or NPY) and consisted of a 70 × 70 μm image field taken at 0.2 μm z-steps with the CA1 pyramidal cell layer in the center of the field. Quantification of immunofluorescence and colocalization were performed using Volocity software (Improvision, Lexington, MA).

Statistics.

Cumulative histograms of relative distances were generated for oil and estradiol treatment groups and compared using the Kolmogorov–Smirnov (K-S) test (D_max = 0.163 for p < 0.01; n = 100). The mean relative distance of ERα-immunoreactive vesicles and means for measures of basic structural characteristics of boutons as well as fluorescence and colocalization were calculated for each brain, and treatment groups were compared using Student's t test (unpaired, two-tailed; n = 4) using SPSS (Chicago, IL) software.

Data display.

All figures were prepared with Photoshop (Adobe Systems, San Jose, CA). Graphs and histograms were plotted using SigmaPlot (SPSS). Bouton reconstruction files were imported into and rendered with 3D StudioMax (Autodesk, San Raphael, CA).

Results

Nuclear and extranuclear ERα-IR

We confirmed nuclear and extranuclear ERα-IR in the dorsal CA1 cell body layer of adult female rats using MC-20, a polyclonal antiserum raised against the C terminus of mouse ERα visualized with DAB. As described previously (Hart et al., 2001), nuclear ERα-IR was present in a relatively small proportion of cells located primarily in the dendritic layers; pyramidal cell nuclei were unlabeled (Fig. 1A). Extranuclear ERα-immunoreactive puncta were observed throughout CA1 but were particularly distinct in the cell body layer and occasionally appeared to ring unlabeled pyramidal cell somata. This suggested that ERα-IR might be contained within perisomatic axonal boutons of inhibitory basket cells, a question we addressed subsequently with EM. Both nuclear and extranuclear ERα labeling were absent when the primary antiserum was eliminated (Fig. 1B) or when the antiserum was preadsorbed with blocking peptide (Hart et al., 2001). Western blots of whole-cell extracts from dorsal CA1 confirmed that the MC-20 antiserum recognized a single protein band at ∼67 kDa (Fig. 1C), the predicted size of ERα in the rat (Koike et al., 1987). Qualitatively similar punctate ERα-IR was also observed in tissue labeled with 6F11, a monoclonal antibody raised against full-length ERα (data not shown).

Figure 1.

Extranuclear ERα-IR in the dorsal CA1 cell body layer. A, ERα-IR visualized with bright-field microscopy. The dashed line delineates the boundary of the cell body layer (above) and stratum radiatum (below). Extranuclear ERα-immunoreactive puncta (small dark dots, some of which are indicated by arrowheads) are numerous and occasionally appear to ring unlabeled somata (asterisk). A cell with interneuron-like morphology located in the proximal stratum radiatum contains nuclear ERα-IR (arrow). B, High-magnification view of the same area from tissue processed with the primary antiserum omitted. No labeling is observed. C, Western blot probed with MC-20, the antiserum used for quantification of ERα-IR, showing a single band at ∼67 kDa. D, Electron micrograph of one of two consecutive sections showing MC-20 labeling for ERα on a portion of vesicles (arrowhead) located in a presynaptic bouton forming a symmetric synapse (arrow) with a CA1 pyramidal cell soma. Note that ERα-immunoreactive vesicles appear to be clustered. E, Higher magnification of the same section as in D shows ERα-IR associated with vesicles (large arrowhead) and nearby on the bouton plasma membrane (small arrowheads). F, The section adjacent to that in D shows the same ERα-immunoreactive vesicle cluster. G, ERα labeling with 6F11 is qualitatively similar to labeling with MC-20, with ERα-IR associated with vesicles and the bouton plasma membrane (arrowhead). Scale bars: A, B, 10 μm; D, F, G, 200 nm; E, 200 nm.

We confirmed that, as reported previously for 72 h treatment (Hart et al., 2001), estrogen treatment for 24 h in vivo had no significant effect on the number of ERα-immunoreactive nuclei in any hippocampal subregion. We quantified ERα-immunoreactive nuclei in the dorsal hippocampus of six ovariectomized oil-treated (OVX+O) and six ovariectomized, estrogen-treated (OVX+E) adult female rats using the optical disector/fractionator. The total numbers of ERα-immunoreactive nuclei were as follows: in subiculum: 1850 ± 507 (OVX+O) and 1100 ± 73 (OVX+E); in CA1: 1566 ± 286 (OVX+O) and 917 ± 70 (OVX+E); in CA3: 3567 ± 724 (OVX+O) and 2733 ± 279 (OVX+E); in dentate gyrus: 2000 ± 392 (OVX+O) and 1767 ± 196 (OVX+E). All p values were >0.07. Thus, in contrast to the effect of estrogen on hippocampal slice cultures, in which estrogen treatment for 8 d upregulates ERα-IR (Rune et al., 2002), shorter treatments with estrogen in vivo do not appear to influence nuclear ERα expression in the hippocampus.

Next, we used pre-embedding EM immunocytochemistry to investigate whether inhibitory axonal boutons in the CA1 cell body layer contain ERα-IR. We were especially interested in the possibility that ERα is expressed in inhibitory boutons because we have shown previously that estrogen suppresses inhibitory GABAergic synaptic transmission in CA1 (Rudick and Woolley, 2001; Rudick et al., 2003). Inhibitory boutons were recognized as those that formed synapses with symmetric presynaptic and postsynaptic densities, >95% of which are GABA immunoreactive in the CA1 cell body layer (Ledoux and Woolley, 2005). Electron imaging of tissue labeled for ERα using either the MC-20 antiserum (Fig. 1D–F) or the 6F11 antibody (Fig. 1G) revealed that ERα-IR within inhibitory boutons was most often found associated with vesicles that appeared to be clustered. Occasionally, we also observed labeling on patches of the plasma membrane near ERα-immunoreactive vesicles. Additionally, small clumps of ERα-IR were present within pyramidal cell bodies, excitatory boutons, dendrites, and glia. The pattern of immunoreactivity was identical in tissue labeled with MC-20 compared with 6F11.

We performed two additional analyses to confirm ERα associated with synaptic vesicles: EM immunocytochemistry with immunogold visualization and Western blot analysis of synaptic vesicle fractions. Evaluation of extranuclear ERα-IR in the CA1 cell body layer visualized with immunogold showed the same pattern of labeling as observed with DAB. Gold particles were observed most frequently associated with vesicles in axonal boutons (1–5 particles) (Fig. 2A), dendritic spines (1–6 particles), and glia (1–10 particles); particles also were observed in neuronal nuclei and associated with somatic organelles. In some series, gold particles were observed in the same locations spanning consecutive sections.

Figure 2.

Immunogold and Western blot confirmation of ERα-IR association with synaptic vesicles. A, Electron micrograph of ERα-IR visualized with immunogold (arrow) in a presynaptic bouton in the CA1 cell body layer. Scale bar, 500 nm. B, A Western blot of cytoplasmic, synaptosomal, and synaptic vesicle proteins probed for synaptophysin (Syn) and ERα shows that ERα is present in both cytoplasmic and synaptic vesicle fractions. Cytoplasmic fractions show a single band at ∼67 kDa, whereas synaptic vesicle fractions show a major band at ∼67 kDa and two additional minor bands (see Results).

Western blots of cell fractions probed with MC-20 (Fig. 2B) also confirmed that ERα is associated with synaptic vesicles. Consistent with Western blots of whole-cell fractions, cytoplasmic protein fractions contained a single band at ∼67 kDa. Synaptophysin, a presynaptic vesicle protein, was present in relatively low levels in cytoplasmic fractions. Synaptophysin was concentrated in synaptosomes, as expected, but ERα did not represent a sufficient proportion of total synaptosomal protein to be detectable. However, ERα was detectable in synaptic vesicle fractions derived from synaptosomes. Interestingly, ERα labeling in synaptic vesicle fractions consistently showed three bands. The major band appeared at the expected ∼67 kDa, along with a slightly higher molecular weight band and a faint band at a lower molecular weight. The higher molecular weight band could represent ERα that has been post-translationally modified by phosphorylation or palmitoylation, which has been shown to promote its association with membranes (Acconcia et al., 2005). The identity of the lower molecular weight band is unknown.

Colocalization of extranuclear ERα-IR with GAD-IR

Having shown that ERα-IR is contained within inhibitory axonal boutons where it is associated with vesicles, we next asked how many inhibitory boutons contain ERα-IR. To investigate this question, we double labeled hippocampal tissue for ERα and GAD, the rate-limiting enzyme in GABA synthesis and a marker for inhibitory neurons; GAD antibodies label inhibitory axonal boutons intensely. We evaluated GAD-IR and ERα-IR in five OVX+O and five OVX+E adult female rats. Projected images from stacks of optical sections showed characteristic GAD-IR in the cell body layer, with labeled varicosities arranged in rings around unlabeled pyramidal cell somata (Fig. 3A). For both GAD-IR and ERα-IR, measurements of fluorescence intensity, volume, number of labeled objects, and percentage of colocalization were unaffected by estrogen, so data from both treatment groups were combined (n = 10). The lack of estrogen effect on extranuclear ERα-IR mirrored the lack of effect on nuclear ERα-IR. Quantitative analysis of 18 image stacks per brain (21,600 μm³ per stack) showed that 32.4 ± 0.6% of the 3815 ± 141 GAD-immunoreactive varicosities per stack contained ERα-immunoreactive puncta (Fig. 3B,C), and 24.0 ± 0.5% of the 5549 ± 187 ERα-immunoreactive puncta per stack were located in GAD-immunoreactive varicosities (Fig. 3C). This analysis indicated that extranuclear ERα-IR is found in approximately one-third of GABAergic varicosities in the cell body layer but that a substantial fraction of punctate ERα-IR is located in non-GABAergic structures as well. The observation that not all ERα-immunoreactive puncta colocalize with GAD-immunoreactive varicosities is consistent with previous qualitative EM observations of ERα-IR in multiple extranuclear sites, such as dendrites, glia, excitatory axonal boutons, and somatic organelles (Milner et al., 2001).

Figure 3.

Colocalization of GAD-immunoreactive varicosities and ERα-immunoreactive puncta. A, A 2-μm-thick image stack shows GAD-immunoreactive varicosities surrounding unlabeled somata (asterisks), presumably of pyramidal cells. B, Higher-magnification views of the boxed area in A through 0.4 μm of tissue. Portions of three GAD-immunoreactive varicosities (green) are shown, one of which contains punctate ERα-IR (red). C, Quantification of GAD-immunoreactive and ERα-immunoreactive colocalization. The entire height of each bar represents the average total number of labeled varicosities (GAD-IR) or puncta (ERα-IR) per 4 μm stack of optical sections (21,600 μm³ volume); the colored portion of each bar represents the average number and percentage of varicosities or puncta that are double labeled. Error bars indicate SEM. Scale bars: A, 10 μm; B, 1 μm.

Serial EM reconstruction of axonal boutons containing ERα-IR

Next, we used serial-section EM to generate three-dimensional reconstructions of axon segments containing ERα-immunoreactive vesicles to address three questions: (1) Does a single axon contain boutons both with and without ERα-immunoreactive vesicles? (2) Are boutons with ERα-immunoreactive vesicles structurally different from those without labeled vesicles? (3) Does estrogen treatment affect ERα-immunoreactive vesicles? We reconstructed axon segments containing ERα-IR visualized with DAB in tissue from each of eight rats, four OVX+O and four OVX+E (three to seven segments per rat). A total of 48 perisomatic inhibitory boutons from ERα-immunoreactive-containing axon segments were completely reconstructed and analyzed, as were 32 neighboring inhibitory boutons that synapsed with the same CA1 pyramidal cells. Examination of reconstructed boutons confirmed that, when present, ERα-immunoreactive vesicles were clustered but that clusters of labeled vesicles occur in only a subset of boutons on any single axon (Fig. 4). Thirty-three of the 80 boutons that we reconstructed contained ERα-immunoreactive vesicle clusters, roughly consistent with the GAD/ERα immunofluorescence colocalization analysis. In boutons containing labeled clusters, ERα-immunoreactive vesicles accounted for 10.7 ± 0.8% of all presynaptic vesicles, and the number of vesicles per cluster did not differ between OVX+O and OVX+E boutons. It should be noted that because the DAB reaction product is diffusible up to ∼100 nm (Courtoy et al., 1983), it is possible that our measurement of 10.7% vesicles per bouton overestimates the true number of ERα-containing vesicles. Interestingly, patches of ERα-IR on the bouton plasma membrane were strongly associated with the presence of ERα-immunoreactive vesicles; we found only one bouton that lacked labeled vesicles and showed any detectable ERα-IR at the plasma membrane. None of the basic structural features of individual boutons (volume, presynaptic density area, single vs multiple synapses, presence of mitochondria, total vesicle number or vesicle density) was different between boutons that contained ERα-immunoreactive clusters and those that did not, and none of these parameters was affected by estrogen treatment (Table 1).

Figure 4.

Three-dimensional reconstruction of an axon segment containing ERα-immunoreactive vesicles. An axon segment from an estradiol-treated animal reconstructed through 80 serial sections shows two boutons that form perisomatic inhibitory synapses (yellow) with a CA1 pyramidal cell. Only one of the boutons contains a cluster of ERα-immunoreactive vesicles (red). Green, Non-ERα-immunoreactive vesicles; gray, bouton plasma membrane.

Table 1.

Structural characteristics of reconstructed perisomatic inhibitory boutons in the dorsal CA1 cell body layer

Bouton structural parameter	Oil treated		Estradiol treated
	ERα+ (n = 4)	ERα− (n = 4)	ERα+ (n = 4)	ERα− (n = 4)
Volume (μm3)	0.44 ± 0.11	0.45 ± 0.08	0.49 ± 0.06	0.46 ± 0.07
Presynaptic density area (μm2)	0.08 ± 0.02	0.09 ± 0.02	0.08 ± 0.01	0.09 ± 0.01
Boutons with multiple synapses (%)	56 ± 21	64 ± 8	51 ± 11	63 ± 5
Boutons containing mitochondria (%)	100 ± 0	89 ± 7	88 ± 7	87 ± 4
Total number of vesicles	1891 ± 144	2245 ± 144	2325 ± 324	2950 ± 729
Vesicle density (#/μm3)	4882 ± 685	5676 ± 764	5104 ± 611	6298 ± 780

The mean (±SEM) for each parameter is shown both for boutons with (ERα+) and without (ERα−) ERα-immunoreactive vesicle clusters from oil- and estradiol-treated animals. Neither ERα-IR content nor estrogen treatment significantly affected any parameter measured.

Although the basic structural features of ERα-immunoreactive-containing boutons did not differ between OVX+O and OVX+E animals, plotting the location of presynaptic vesicles within each bouton revealed a striking effect of estrogen to shift the location of ERα-immunoreactive vesicle clusters closer to synapses (Fig. 5A,B). Because axonal boutons vary widely by size, we determined the proximity of presynaptic vesicles to synapses using a relative distance measurement. For each bouton, the distance between each presynaptic vesicle and the nearest synapse was measured and converted to a value between 0 and 1.0, with 1.0 being the maximum distance that a vesicle could be from the synapse. Comparison of histograms of relative distances in the 33 completely reconstructed boutons that contained ERα-IR (Fig. 5C,D, filled bars) showed that labeled vesicle clusters were located significantly closer to the nearest synapse in OVX+E compared with OVX+O boutons (n = 13 OVX+O; n = 20 OVX+E; K-S test, p < 0.01) (Fig. 5D, inset). Comparison of the mean relative distances on a per animal basis confirmed that estrogen decreased the distance between ERα-immunoreactive vesicles and the nearest synapse by approximately one-half (n = 4; unpaired, two-tailed, Student's t test, p = 0.03). Additionally, although initial inspection of the histograms showing relative distances of unlabeled vesicles in OVX+O and OVX+E boutons (Fig. 5C,D, open bars) suggested that they also might be affected by estrogen, no significant difference or trend was detected (K-S test, p > 0.20). Thus, the effect of estrogen to shift the location of presynaptic vesicles is specific for those that are ERα immunoreactive. The observation that estrogen regulates the location of labeled vesicle clusters relative to synapses argues strongly that ERα-IR on vesicles is either ERα itself or a very closely related protein. These appear to be a specific subset of presynaptic vesicles that, because they contain ERα, are regulated by estrogen separately from other presynaptic vesicles in the same bouton.

Figure 5.

Three-dimensional reconstructions and vesicle distances in perisomatic inhibitory boutons containing ERα-immunoreactive vesicle clusters. A, B, Side and end views (rotated by 90°) of reconstructed boutons from an oil-treated (A) and an estradiol-treated (B) animal showing that ERα-immunoreactive vesicle clusters (red) are located closer to the synapse (yellow) in the bouton from an estrogen-treated animal (green, non-ERα-immunoreactive vesicles; gray, bouton plasma membrane). C, D, Histograms showing the relative distance from the synapse of vesicles in reconstructed boutons containing ERα-immunoreactive clusters in oil-treated (C) and estradiol-treated (D) animals. Distributions of ERα-immunoreactive vesicles are represented with filled bars (gray, oil; black, estradiol), whereas the distribution of non-ERα-immunoreactive vesicles are represented by open bars. The inset in D shows cumulative histograms of the relative distances of ERα-immunoreactive vesicles in oil (gray) and estradiol (black) animals demonstrating that labeled vesicles are located significantly closer to synapses after estrogen treatment (K-S test, p < 0.01).

Colocalization of extranuclear ERα-IR with PV- or CCK-IR

The perisomatic inhibitory boutons we reconstructed arise from basket cells, a population of GABAergic neurons that can be divided into two subtypes based on PV or CCK immunolabeling, each of which has distinct physiological roles in the hippocampus (Freund and Buzsaki, 1996; Freund, 2003). To determine whether ERα-IR is located specifically in one class of basket cell, we used double-label immunofluorescence to quantify colocalization of ERα-IR with PV- or CCK-IR in the cell body layer in five OVX+O and five OVX+E animals. Estrogen treatment did not affect any parameter of PV-IR, CCK-IR, or colocalization, so data from both groups were combined (n = 10). Tissue double labeled for PV and ERα showed the characteristic “beads on a string” PV staining of axons and axonal varicosities, but there was no colocalization of ERα-IR and PV-IR (Fig. 6A). In contrast, ERα-IR did colocalize with CCK-IR, which labels in a more punctate pattern (Morales and Bloom, 1997) than GAD or PV (Fig. 6B). Quantitative analysis of 18 image stacks per brain (10,800 μm³ per stack) showed that 25.2 ± 0.6% of 3817 ± 177 CCK-immunoreactive puncta per stack colocalized with ERα-IR and 36.2 ± 0.8% of 2762 ± 72 ERα-immunoreactive puncta per stack colocalized with CCK-IR (Fig. 6E). Thus, the axonal boutons containing ERα-immunoreactive vesicle clusters that we analyzed with serial-section EM likely belong to CCK basket cells and not PV basket cells.

Figure 6.

Colocalization of PV-, CCK-, and NPY-IR with extranuclear ERα-IR. A, ERα-immunoreactive puncta (red) did not colocalize with PV-immunoreactive structures (green). B, Some ERα-immunoreactive puncta (red) did colocalize (yellow) with CCK-immunoreactive puncta (green). C, Some ERα-immunoreactive puncta (red) also colocalized (yellow) with NPY-IR (green). D, The majority (∼73%) of NPY-immunoreactive puncta (red) are colocalized (yellow) with GAD-IR (green). E, Quantification of CCK- and NPY-IR with ERα-IR. The entire height of each bar represents the average total number of labeled puncta per 2 μm stack of optical sections (10,800 μm³ volume); the colored portion of each bar represents the average number and percentage of puncta that were double labeled. Error bars indicate SEM. Scale bar, 1 μm.

Because the number of CCK-immunoreactive puncta was greater than the number of GAD-immunoreactive varicosities in the same volume, either there are multiple CCK-immunoreactive puncta per varicosity and/or some CCK-immunoreactive puncta are located in structures other than GABAergic varicosities. Consistent with the second possibility, CCK labeling has been shown previously in dendrites in CA1 (Harris et al., 1985; Morales and Bloom, 1997). The number of ERα/CCK-immunoreactive structures per stack was 902 ± 42, which is similar to the number of ERα/GAD-immunoreactive structures calculated for the same volume, 624 ± 31. Thus, most of the puncta that colocalize ERα-IR and CCK-IR are likely to be in GABAergic boutons.

Colocalization of extranuclear ERα-IR with NPY-IR

A subset of GABAergic neurons in CA1 also expresses NPY (Freund and Buzsaki, 1996), and some NPY neurons express nuclear ERα (Sar et al., 1990), suggesting that extranuclear ERα-IR also might be found in NPY neurons. To investigate this, we analyzed colocalization of ERα-IR and NPY-IR in the same tissue used for PV and CCK analyses. No measure of immunofluorescence or colocalization was affected by estrogen treatment, so data from both groups were combined (n = 10). Quantitative analysis of ERα-IR and NPY-IR in the cell body layer (Fig. 6C) showed that 30.3 ± 0.6% of 4165 ± 170 NPY-immunoreactive puncta colocalized with ERα-IR and 49.9 ± 2.2% of 2662 ± 44 ERα-immunoreactive puncta colocalized with NPY-IR (Fig. 6E).

The number of ERα/NPY-immunoreactive objects was greater than the number of ERα/GAD-immunoreactive objects in the same volume, which suggested that a substantial fraction of ERα/NPY-IR might not be GABAergic. Thus, we performed an additional experiment to quantify NPY/GAD colocalization (Fig. 6D) and found that ∼73% of NPY-immunoreactive puncta in the cell body layer colocalized with GAD (data not shown). The remaining ∼27% of NPY-immunoreactive puncta likely belong to dendrites, excitatory boutons, or cytoplasmic organelles (Milner and Veznedaroglu, 1992; V. A. Ledoux and C. S. Woolley unpublished observations). Interestingly, the number of ERα-immunoreactive puncta per stack that colocalized with NPY-IR (∼1250) is very similar to the number of NPY-immunoreactive puncta that were negative for GAD (∼1125). It is therefore possible that many ERα/NPY-immunoreactive puncta are not located in GABAergic boutons.

Discussion

We used serial-section EM immunocytochemistry, cell fractionation, and immunofluorescence to show that ERα is associated with clusters of vesicles in inhibitory presynaptic boutons in the hippocampal CA1 cell body layer and that estrogen treatment shifts the location of these vesicles toward synapses. ERα-immunoreactive vesicle clusters are abundant, occurring in approximately one-third of perisomatic GABAergic boutons, but only in a subset of those on a single axon. Extranuclear ERα-IR colocalizes with CCK and NPY, but not with PV, indicating that different neurochemical classes of GABA neurons are likely to be differentially sensitive to estrogen through ERα located in presynaptic boutons. Our findings point to a novel mode of estrogen action in the brain in which estrogen acts directly on a subset of ERα-containing vesicles to mobilize them toward synapses.

ERα-immunoreactive vesicles

ERα-immunoreactive vesicles may contain GABA and undergo exocytosis to release neurotransmitter. Labeled vesicles in perisomatic boutons were small clear vesicles that were morphologically indistinguishable from unlabeled ones. We did occasionally observe ERα-IR associated with presynaptic densities, suggesting fusion at synaptic active zones. If ERα-immunoreactive vesicles function to release neurotransmitter, then our results indicate a specialized subpopulation of GABAergic vesicles within individual presynaptic boutons that is regulated independently of the others (i.e., by estrogen). If this is the case, then the simplest interpretation of the estrogen-induced mobilization of labeled vesicles is that it reflects their movement toward synaptic release sites. Mobilization of ERα-immunoreactive vesicles toward synapses may influence the sensitivity of hippocampal synapses to acute estrogen action.

It is important to note that this effect to mobilize ERα-immunoreactive vesicles toward synapses is distinct from our previous observations that 24 h estrogen treatment suppresses GABA release (Rudick et al., 2003) and decreases the number of vesicles docked at GABAergic synapses (Ledoux and Woolley, 2005). Our previous studies did not distinguish between boutons and/or vesicles that contain ERα and those that do not and focused specifically on docked vesicles that are much closer to presynaptic release sites than the vast majority of ERα-immunoreactive vesicles. Thus, the observation that estrogen mobilizes specifically ERα-immunoreactive vesicles toward synapses is further evidence that ERα confers specialized estrogen sensitivity to a distinct subset of vesicles in inhibitory boutons.

In addition to the possibility that ERα-immunoreactive vesicles are neurotransmitter vesicles, an alternative is that ERα-immunoreactive vesicles are specialized endosomes that deliver ERα to, and/or retrieve it from, the bouton plasma membrane. Patches of ERα-IR at the bouton membrane, usually near labeled vesicle clusters, are consistent with this idea. Studies in a variety of cell types have demonstrated ERα at the plasma membrane where it functions as a G-protein-coupled receptor (Razandi et al., 1999; Wyckoff et al., 2001) and can influence secretion of neuropeptides (Navarro et al., 2003). If ERα-immunoreactive vesicles are involved in regulating membrane-associated ERα, then estrogen effects on ERα-immunoreactive vesicle location could result from exocytosis and/or endocytosis at different sites relative to synapses. Interestingly, however, we found that the number of vesicles per ERα-immunoreactive cluster was not affected by estrogen, arguing against the idea that estrogen specifically stimulates ERα delivery to or retrieval from the plasma membrane.

ERα contains neither a known membrane targeting sequence nor stretches of hydrophobic residues characteristic of an integral membrane protein; however, in breast cancer cells and Chinese hamster ovary cells, ERα has been shown to associate with caveolin-1, which facilitates its transport to the membrane (Razandi et al., 2003). In these cells, membrane ERα forms a signaling complex with other proteins through which estrogen can rapidly increase cAMP and activate mitogen-activated protein kinase (MAPK) and phosphatidylinositol-3 kinase (PI3K) (for review, see Levin, 2005). This requires dimerization-competent ERα and involves the ligand-binding and AF2 domains (Razandi et al., 2004). Similar rapid signaling through MAPK and PI3K also occurs in neurons (Bryant et al., 2005; Mannella and Brinton, 2006). Mobilization of ERα-containing vesicles could influence synaptic function by bringing an estrogen-sensitive rapid signaling complex nearer to synaptic release sites.

The ability of membrane-associated ERα to activate MAPK could provide a link between rapid estrogen signaling and our previous observation that estrogen decreases the number of vesicles docked at GABAergic synapses. Estrogen acutely activates MAPK in hippocampal neurons (Bi et al., 2000), possibly through a membrane ER (Kuroki et al., 2000). One substrate for MAPK in presynaptic boutons is synapsin-I, which anchors synaptic vesicles to the actin cytoskeleton in a phosphorylation-dependent manner (Jovanovic et al., 1996). Phosphorylation of synapsin-I at MAPK sites is involved in mobilization of synaptic vesicles for release (Chi et al., 2003), and at inhibitory synapses, synapsins are critical for maintaining the readily releasable pool of vesicles (Gitler et al., 2004). As mentioned above, we have shown previously that estrogen suppresses GABA release in CA1, at least in part by decreasing the number of vesicles docked at inhibitory synapses. Therefore, it is tempting to speculate that ERα signaling in inhibitory boutons could mediate the estrogen-induced decrease in docked vesicle number through regulation of MAPK and synapsin phosphorylation.

Colocalization of ERα and CCK

The finding that extranuclear ERα-IR colocalizes with CCK, but not PV, indicates that direct effects of estrogen on presynaptic boutons likely occur only in the CCK subtype of the basket cell. This is significant because PV and CCK cells differ in their electrophysiological and anatomical properties and their influence on pyramidal cells. Most PV basket cells are capable of firing at high frequencies without accommodation and release GABA synchronously; CCK cells fire at lower frequencies, do show accommodation, and release GABA asynchronously (Pawelzik et al., 2002; Hefft and Jonas, 2005). These and other findings (Klausberger et al., 2003; Vreugdenhil et al., 2003) have been interpreted to indicate that PV cells play a greater role than CCK cells in entraining rhythmic activity of pyramidal cells, particularly at gamma frequency (Freund, 2003).

Additionally, anatomical studies show that PV cells receive, on average, ∼15,000 excitatory and ∼1000 inhibitory inputs (Gulyas et al., 1999), whereas CCK cells receive only ∼5000 excitatory but ∼3000 inhibitory inputs (Matyas et al., 2004). CCK cells also receive GABAergic input from a specialized, interneuron-driven population of calretinin-positive neurons (Gulyas et al., 1996) and serotonergic input from the median raphe, neither of which innervate PV cells (Miettinen and Freund, 1992; Morales and Bloom, 1997). Thus, PV and CCK cells appear to monitor the activities of different populations of neurons, and CCK cells are in a better position to respond to modulatory GABAergic and serotonergic inputs. These differences have led to the hypothesis that relatively nonplastic ensembles of PV basket cells are responsible for synchronizing the activity of pyramidal cells, whereas more modifiable groups of CCK basket cells integrate local and subcortical inputs to fine-tune pyramidal cell activity (Freund, 2003). The colocalization of ERα-IR with CCK- but not PV-IR suggests that estrogen could act through ERα in CCK basket cell axonal boutons to influence such fine-tuning of hippocampal activity.

Colocalization of ERα and NPY

The colocalization of ERα-IR with NPY-IR provides an additional clue to the role of extranuclear ERα in the hippocampus. In CA1, NPY is expressed primarily in GABAergic interneurons (Freund and Buzsaki, 1996) but also is expressed within excitatory neurons (Milner and Veznedaroglu, 1992). NPY inhibits presynaptic glutamate release in the hippocampus (Colmers et al., 1988), and in cortical synaptosomes, this occurs through a Y1 receptor-dependent reduction in Ca²⁺ influx through N- and/or P/Q-type Ca²⁺ channels (Wang, 2005). Although it is likely that ERα/NPY sites in the CA1 cell body layer are located too far from most glutamatergic inputs to influence them, it is possible that a similar mechanism could regulate GABA release at perisomatic synapses. Calcium-dependent GABA release in CA1 also depends on N- and/or P/Q-type Ca²⁺ channels (Doze et al., 1995). Additionally, NPY has been shown to suppress GABAergic synaptic transmission in the spinal cord (Moran et al., 2004) and paraventricular nucleus (Pronchuk et al., 2002), and at least some GABAergic neurons express Y1 receptors (Oberto et al., 2001). Thus, sites at which ERα and NPY are colocalized, whether in inhibitory boutons or other structures, could provide an additional means by which estrogen suppresses GABAergic synaptic transmission, in this case by modulating the release of NPY.

In summary, quantitative analysis of ERα in perisomatic inhibitory boutons demonstrates a novel mode of estrogen action in the brain, to mobilize a specific subset of vesicles that contain ERα. In the CA1 cell body layer, these vesicles occur selectively in one class of basket cell, CCK cells, where they may be a subset of neurotransmitter vesicles and/or may be involved in regulating ERα at the bouton plasma membrane. Estrogen previously has been shown to have profound effects on inhibitory synapses in CA1 that are likely to be causally related to a subsequent increase in excitatory synapse number. The current study represents an initial step toward understanding how estrogen could act through extranuclear receptors associated with a specific subcellular compartment, presynaptic vesicles, to influence neuronal function. Direct effects of estrogen on presynaptic boutons may provide a link between rapid estrogen signaling (e.g., through modulation of protein kinase activity) and the longer-term effects of estrogen on synaptic structure and function.