Consistent with previous findings (Milner et al. 2005; Mitra et al. 2003; Shughrue and Merchenthaler 2001), ERβ-ir in the adult dentate gyrus was found primarily in the granule cell layer and subgranular hilus and occasionally in hilar interneurons (Fig. 1A). At higher magnification, the majority of neuronal ERβ-labeling in the granule cell layer and subgranular layer appeared to be extranuclear (Fig. 1B). In agreement with previous studies in adult rats (Abrous et al. 2005; Rao and Shetty 2004), somata with DCX-ir were found in the subgranular zone of the dentate gyrus (Fig. 1C). Processes from DCX-labeled somata penetrated through the granule cell layer into the molecular layer (Fig. 1D) and DCX-labeled dendritic processes rarely extended into the hilus. At PND 7 and 14, immunoreactivities for ERβ and DCX distribution were largely similar to those observed in the adult. However, at both juvenile ages, cells containing both markers were more abundant (Figs. 1E–H) and DCX-labeled somata extended more processes into the hilus compared to adults (Figs. 1E, H). Moreover, at PND 7, DCX-labeled somata were additionally found throughout the molecular layer and hilus (Fig. 1H).#

Electron microscopic examination of the subgranular hilus confirmed previous observations (Rao and Shetty, 2004) that DCX-immunoreactive somata resembled neurons (Fig. 2). In adult rats, DCX-containing cells often had morphological features that distinguished them from mature neurons. Most DCX-labeled cells resembled granule cells (Commons and Milner, 1995): they were small (5–10 μm in diameter) and contained scant cytoplasm and smooth nuclei with clumps of heterochromatin near the nuclear membrane (Figs. 2A–C and 3A). Many (12 out of 19) DCX-labeled cells were elongated (Figs. 2A, B) whereas some (5 out of 19) were more rounded (Fig. 2C). A few (2 out of 19) DCX-labeled cells had invaginated nuclei and thus resembled interneurons (Drake et al., 1999). DCX-labeled somata were closely apposed to the somata of mature granule cells (Figs. 2B, C). Some (4 out of 19) DCX-labeled neurons had darkened nuclei (Fig. 2A), identifying them as immature (Seri et al., 2001). Immunogold labeling for DCX was found throughout the cytoplasm and did not appear to be affiliated with any particular organelles. DCX-labeled somata often received synapses from unlabeled terminals (Fig. 4). In several instances, more than one terminal synapsed on a DCX-labeled soma (Fig. 4B). Synapses between terminals and DCX-labeled somata were usually difficult to classify as asymmetric or symmetric because aggregations of electron-dense material obscured the pre- and post-synaptic plasma membranes; however, in some cases symmetric synapses between unlabeled terminals and DCX-labeled somata could be distinguished (Fig. 4C).#

As in adults, in immature rats DCX-labeled cells resembled neurons (Fig. 7A). DCX-labeled cells ranged from 4 to 9 μm in diameter (6.9±0.6 μm; n=7), were either elongated or ovoid and contained dark nuclei with clumps of heterochromatin. Unlike in adults, DCX-labeled neurons in the PND 14 rats had abundant cytoplasm and ruffled plasma membranes. Occasionally processes emanating from the DCX-labeled somata projected into the central hilus (Fig. 7A).#

In adult rats, DCX-labeled somata and dendritic processes frequently possessed ERβ-ir in the cytoplasm (Figs. 2C D; 3; 4; 5). Consistent with our previous observations (Milner et al., 2005), ERβ-ir was associated occasionally with the plasma membrane of DCX-containing somata (Figs. 2C, D). ERβ-ir frequently was associated with mitochondria (Figs. 3A, B, C, E) and endomembranes (Figs. 3B, D, E and 4A, C) in DCX-labeled somata.#

In DCX-containing dendrites, ERβ-ir was often seen near the plasma membrane (Fig. 5). Regions of the dendritic plasma membrane containing ERβ-ir were apposed to non-DCX-containing neuronal profiles (Fig. 5A) or glial profiles (Fig. 5C). Like somata, dendritic profiles contained ERβ-ir adjacent to mitochondria and endoplasmic reticula, and not all ERβ-labeled dendritic profiles contained DCX-ir (Fig. 6B).#

In PND 14 rats, DCX-containing somata and dendrites expressed ERβ-ir at locations similar to those observed in adults (Figs. 7B, C, D). In both DCX-labeled somata and dendrites, ERβ-ir was adjacent to mitochondria (Fig. 7B), endomembranes (Fig. 7C) and the plasma membrane (Fig. 7D). Sometimes, cell profiles that contained ERβ-ir apposed DCX-labeled dendrites (Fig. 7B).#

The present study employed DCX-ir to identify newly born cells in the adult and neonatal rat dentate gyrus. DCX is expressed in proliferating progenitor cells and newly generated neuroblasts, with >90% of bromodeoxyuridine (BrdU)-labeled cells (4 days after BrdU injections) containing DCX-ir (Abrous et al. 2005; Brown et al. 2003). Within developing neurons, the DCX protein is associated with microtubules (Moores et al., 2004); this association is required for proper neuronal migration and axonal wiring (Koizumi et al., 2006).#

Our previous study demonstrated that ERβ-ir is found in discrete patches within neuronal profiles (Milner et al., 2005), so absence of the peroxidase reaction product in a particular DCX-labeled profile does not preclude the possibility that ERβ-ir is in some other portion of the neuron. Peroxidase labeling for ERβ was chosen because it is highly sensitive and maximizes antigen detection. However, because the DAB reaction product can diffuse, peroxidase can result in less precise subcellular localization of immunoreactivity that immunogold labeling. In our previous study (Milner et al., 2005), the subcellular localization of ERβ-ir was consistent using peroxidase and immunogold methods. The restricted distribution of ERβ-immunoperoxidase labeling may be related to apparent low abundance of immunoreactive protein and to carefully selected labeling conditions. The antiserum to ERβ utilized in this study has been extensively characterized (Milner et al. 2005; Mitra et al. 2003). However, the possibility exists that the antibody may recognize peptide sequences contained in other proteins. Thus, the labeling described herein should be interpreted as ERβ-“like” immunoreactivity.#

The ERβ and DCX antisera work optimally under different labeling conditions: optimal labeling of ERβ is achieved in the absence of membrane penetrating procedures (i.e., freeze–thaw), whereas DCX labeling requires detergents (e.g., Triton X-100) or membrane penetrating procedures. The present study employed the labeling conditions required to label for DCX. Thus, together with the observation that ERβ-ir is found in discrete patches, the detection of ERβ in DCX-labeled somata and processes in this study was likely underestimated.#

Within both the adult and neonatal rat dentate gyrus, DCX-labeling was found in cells that resembled neurons. The cells with highest DCX labeling appeared to possess a distinct morphology: elongated somata and nuclei. These cells may have been migrating to their final destination in the dentate gyrus. It is not clear whether these DCX-labeled cells are neurons or belong to the DCX-expressing neuronal precursors classified as “type-3 putative neuronal-restricted progenitors” (Jessberger et al., 2005). Previous studies indicate that newly born cells in adults first extend their axons and then pull their somata up into the correct cell layer (Alvarez-Buylla et al., 2002). The immaturity of these newly generated cells is evident in the asymmetrical and non-spherical appearance of the DCX-labeled somata and their nuclei.#

Most DCX-labeled cells resembled granule cells; however, some DCX-labeled cells had morphologies that resembled interneurons. This finding is in agreement with a previous study demonstrating that about 14% of newly generated neurons in the dentate gyrus of adult rats are GABAergic basket cells (Liu et al., 2003). Moreover, mu opioid receptors, which are almost exclusively on GABAergic interneurons in the rat dentate gyrus (Drake and Milner 1999), are found on some newly generated cells (Eisch, 2002). The existence of MORs on newly born neurons may underlie the profound inhibitory effect of opiates on adult neurogenesis (Eisch et al., 2000). Our finding that some DCX-containing cells resembling interneurons contained ERβ-ir also is in agreement with studies showing ERβ-ir on parvalbumin-labeled neurons in other regions of the hippocampal formation (Blurton-Jones and Tuszynski, 2002).#

In both adult and neonatal rats, DCX-labeled somata and their proximal dendrites were contacted by unlabeled terminals. Some of these terminals formed distinguishable symmetrical synapses suggesting that they are inhibitory (Peters et al., 1991). These findings support physiological studies demonstrating that newborn granule cells in the adult rat dentate gyrus receive only GABAergic synapses (Wadiche et al., 2005). Moreover, GABAergic synaptic events appear before glutamatergic activity in CA1 pyramidal cells during neonatal development (Tyzio et al., 1999). Because GABAA receptor-mediated inhibitory post-synaptic potentials of immature granule cells are comparable to mature granule cells, others have suggested that the development of GABAergic synapses could play a role in maturational processes such as cell death (Karten et al., 2006).#

DCX-labeled cells contained ERβ in their somata. The majority of dual-labeled cells had granule cell-like morphology. In the present study as well as in our previous study of ERβ under optimal labeling conditions (Milner et al., 2005), all ERβ-ir was in the cytoplasm rather than in the nucleus, favoring the theory that ERβ is acting in a non-classical fashion. It is possible that cytoplasmic ERβ could be translocated to the nucleus with hormone stimulation, although the absence of nuclear ERβ (Milner et al., 2005) argues against this interpretation. As with mature granule cells (Milner et al., 2005), ERβ-ir in DCX-labeled neurons was often associated with the plasma membrane, consistent with the possibility of non-genomic activation by extracellular ligands (Cahill et al., 2001). As in our previous study (Milner et al., 2005), the plasmalemmal ERβ labeling was all on the cytoplasmic side of the membrane, consistent with the antibody labeling the intracellular portion of the receptor. In agreement with other studies (Hrabovszky et al. 2004; Milner et al. 2005; Cammarata et al. 2004; Yang et al. 2004), ERβ-ir in DCX-labeled cells was associated with cytoplasmic organelles, particularly endoplasmic reticula, mitochondria and endomembranes near mitochondria. Our previous study in the hippocampal formation demonstrated that ERβ-ir is primarily associated with the endomembranes adjacent to mitochondria (Milner et al., 2005). Endoplasmic reticula interconnected with the mitochondria are thought to generate Ca⁺⁺ signals that control processes such as synaptic plasticity and neuronal excitability (Berridge 2002; Mironov et al. 2005), and it is possible that ERβ affects this Ca⁺⁺ signaling. ERβ also may have a more direct effect on mitochondrial functions such as neuroprotection, mitochondrial transcription or cell differentiation (Felty and Roy 2005; Lobaton et al. 2005; Yang et al. 2004; Psarra et al. 2006; Chen et al. 2005). Our findings are also consistent with the possibility that estrogen actions on mitochondria are involved in disease progression as well as normal function. Mutations of mitochondrial DNA are associated with diabetes, Alzheimer's disease and Parkinson's disease (Simpkins et al., 2005). These conditions all show sex differences (Dluzen and Horstink 2003; Brookmeyer et al. 1998).#

The finding that DCX-labeled cells contain ERβ-ir in both adults and neonates suggests that ERβ may play a role in the genesis and/or maturation of newly born neurons throughout life. Previous studies show that estrogen can regulate neurogenesis in both adults (Tanapat et al., 1999) and neonates (Brannvall et al., 2002) and that ERβ activation is more potent than ERα activation at increasing hippocampal neurogenesis (Mazzucco et al., 2006). Several other lines of evidence suggest that ERβ activation is involved in neurogenesis. ERβ can activate the MAP kinase pathway (Abraham et al. 2004; Singer et al. 1999), which is important for cascade signaling and synaptic plasticity (Thomas and Huganir, 2004). Estrogen can also increase brain derived neurotrophic factor (BDNF) levels in dentate granule cells (Scharfman et al., 2003). BDNF administered directly into the dentate gyrus leads to increased neurogenesis of granule cells and to the development of ectopic granule cells (Scharfman et al., 2005). Like ERβ, TrkB, the receptor for BDNF, is associated with mitochondria in hippocampal neurons (Drake et al., 1999).#

In some instances, DCX-labeled somata were apposed to profiles with glial morphology that contained ERβ. ERβ-labeling has been shown in astrocytes (identified by the presence of glial fibrillary acidic protein) in adult rat brains (Azcoitia et al., 1999). In vitro studies suggest that ERβs in glia are involved in the neuroprotective effects of estrogen and selective estrogen receptor modulators, possibly by decreasing cytokine levels (Carswell et al. 2004; Dhandapani and Brann 2002 2003).#

ERβ in astrocytic profiles apposing DCX-labeled cells could also be important for growth regulation of new neurons by nerve growth factor (NGF). NGF synthesis by astrocytes is thought to support regeneration of neuronal processes and affect synaptic reorganization (Althaus and Richter-Landsberg, 2000). We previously found that estrogens affect the levels of TrkA (receptors for NGF) in hippocampal astrocytic profiles in vivo and that TrkA-immunoreactive astrocytes often are positioned next to dendrites (McCarthy et al., 2002).#

Adult female Sprague–Dawley rats (225–275 g) from Taconic Farms (Chatham, NY) were housed in groups of two with ad libitum access to food and water. The stage of the estrus cycle was determined by taking vaginal smears (Turner and Bagnara, 1971). All rats were in the proestrus phase when the tissue was harvested, which is characterized by high levels of circulating estrogen and maximal expression of DCX (Tanapat et al., 1999). Neonatal female rats were housed with the dam and male siblings from Taconic Farms (Chatham, NY) with ad libitum access to food and water, until the tissue was harvested at PND 7 or PND 14. All methods of housing and tissue collection were approved by the Institutional Animal Care and Use Committees of Weill-Cornell Medical College and Rockefeller University and conform to National Institutes of Health guidelines.#

Six adult female rats in proestrus were deeply anesthetized with sodium pentobarbital (150 mg/kg i.p.) and perfused through the ascending aorta with: (1) 10–15 ml saline (0.9%) containing 1,000 U of heparin; (2) 50 ml of 3.75% acrolein (Polysciences) and 2% paraformaldehyde in 0.1 M phosphate buffer (PB; pH 7.4); and (3) 200 ml of 2% paraformaldehyde in PB. Three PND 7 and three PND 14 female rats were deeply anesthetized with sodium pentobarbital (150 mg/kg i.p.) and perfused through ascending aorta with (1) 5–7 ml saline (0.9%) containing 1,000 U of heparin and (2) 30 ml (PND 7) or 40 ml (PND 14) of 3.75% acrolein and 2% paraformaldehyde in PB. All brains were removed from the skull, cut into 4–5 mm coronal blocks using a rodent brain matrix (Activational) and postfixed for 30 min in 2% paraformaldehyde in PB. Coronal sections, 40 μm thick, were cut through the hippocampal formation on a Vibratome, collected in PB and treated with 1% sodium borohydride in PB for 30 min.#

The ERβ antibody is a polyclonal antibody generated in rabbit (485; Merck Research Laboratories, Rahway, NJ) against a conserved sequence (rat aa 64–82) of the mouse, human and rat ERβ that is located within the A/B domain of ERβ (exons 2–3) and is not present in ERα (Mitra et al., 2003). Previous Western blot analyses showed that the ERβ 485 antibody recognizes a protein that migrates at about 60 kDa on SF9 cell blots, 55 kDa on human ovary and testes blots and about 70 kDa on whole tissue extracts of rat and mouse brain (Mitra et al., 2003). The antibody labels nuclei in cos-7 cells transfected with ERβ and is eliminated in these cells and in rodent hippocampal tissue by preadsorption with the antigenic peptide (Milner et al. 2005; Mitra et al. 2003). Light microscopic analysis of mouse hippocampal sections fixed with 3.75% acrolein and 2% paraformaldehyde and labeled for the ERβ 485 antiserum preadsorbed with the antigenic peptide in the presence of 0.1% Triton-X 100 revealed no detectable nuclear or extranuclear labeling (Mitra et al., 2003). Electron microscopic analysis of hippocampal tissue sections fixed with 3.75% acrolein and 2% paraformaldehyde and processed with the same labeling conditions as those used in the present study, but omitting primary ERβ antiserum, revealed no labeling in neuronal profiles; only one or two immunoperoxidase-labeled glial process per 3025 μm² field were found (Milner et al., 2005). At least 5 splice variants of ERβ have been identified (Price et al., 2000). The selectivity of this antiserum for individual splice variants has not been determined.#

The polyclonal antibody against DCX was generated in a goat (C-18; SC-8066; Santa Cruz Biotechnology, Santa Cruz, CA) against the C-terminus of a DCX peptide of human origin. Western blot analyses of SK-N-SH whole cell lysates, mouse embryo and rat hippocampal extracts show that this antiserum recognizes a protein about 40 kDa, consistent with the molecular weight of the DCX protein (Santa Cruz Biotechnology; Brown et al., 2003). The DCX protein has been used as a marker of recently differentiated neurons in the central nervous system (Abrous et al., 2005) and is associated with microtubules in developing neurons (Moores et al., 2004). The C-18 antibody to DCX is found in neuroblast-like late progenitor cells (Jessberger et al., 2005). Nearly all DCX-labeled cells colocalize with early neuronal antigens (as well as BrDU), but not with antigens specific to glia, undifferentiated cells or apoptotic cells (Rao and Shetty, 2004).#

Coronal sections through the hippocampal formation were processed for the dual immunocytochemical localization of ERβ and DCX antisera. Some sections were processed for the light microscopic identification of DCX or ERβ using immunoperoxidase using the relevant portions of the dual labeling procedure detailed below. Antisera were diluted 1:3000 ERβ (immunoperoxidase) and 1:2000 (immunogold) or 1:4000 (immunoperoxidase) for DCX. For dual labeling for EM, avidin–biotin complex (ABC) localization of ERβ was combined with the immunogold-silver localization of DCX (Chan et al., 1990). Briefly, tissue sections were incubated in: (1) 0.5% BSA in 0.1 M Tris–saline (pH 7.6; TS) for 30 min; and (2) anti-ERβ and anti-DCX in 0.1% BSA in TS for 1 day at 23 °C followed by 4 additional days at 4 °C. Sections prepared for light microscopy contained 0.25% Triton-X 100 in the primary antisera. Sections processed for EM were subjected to a freeze–thaw step prior to incubation in primary antisera. For this, the tissue was placed into a container with a mesh bottom that was submerged in a cryoprotectant solution (composed of 30% sucrose and 30% ethylene glycol in PB) for 15 min. The container then was rapidly immersed in succession in liquid Freon, liquid nitrogen and PB.#

Sections were processed for immunoperoxidase labeling of ERβ by incubating in (1) biotinylated donkey anti-rabbit IgG in 0.1% BSA in TS (1:400, Amersham Laboratories, Piscataway, NJ) for 30 min; (2) peroxidase–avidin complex for 30 min; and (3) diaminobenzidine (Aldrich, Milwaukee, WI) and H2O2 for 6 min. After washing in TS and then PB, sections were processed for immunogold labeling of DCX by incubating in donkey anti-goat IgG conjugated to 1 nm gold particles (AuroProbe One; Amersham, Arlington Heights, IL) in 0.1% gelatin and 0.08% BSA in PBS, pH 7.4, for 2 h at room temperature. Sections were rinsed in PBS, postfixed in 2% glutaraldehyde in PBS for 10 min and rinsed in PBS and 0.2 M sodium citrate (pH 7.4). The gold particles were enhanced with a silver solution (EMS, Fort Washington, PA) for 7–8 min.#

Sections for light microscopy were mounted on gelatin-coated slides, dehydrated and coverslipped with DPX mounting medium (Aldrich). Sections were analyzed and photographed with a Nikon Eclipse 80i light microscope equipped with bright-field and differential interference contrast optics and a Micropublisher digital camera (Q Imaging, Barnaby, British Columbia).#

For EM, sections were postfixed for 1 h in 2% osmium tetroxide, dehydrated with alcohols and propylene oxide and embedded between two sheets of plastic in EMbed 812 (Electron Microscopy Sciences, Fort Washington, PA). Ultrathin sections (70 nm thick) were collected from the midseptotemporal level of the dorsal hippocampal formation and cut on an Ultracut UCT ultratome (Leica, Wien, Austria). Sections were counterstained with Reynold's lead citrate and uranyl acetate and final preparations were analyzed on an FEI Tecnai Biotwin transmission electron microscope. Images were acquired with a digital camera system (Advanced Microscopy techniques, v. 3.2). Digital images were adjusted for brightness and contrast using Adobe Photoshop 6.0 on a Macintosh G5 computer. Final images were assembled using Quark X-Press 4.1.#

DCX and ERβ immunolabeled profiles were classified according to the nomenclature of Peters et al. (1991). Somata were identified by the presence of a nucleus. Dendrites contained regular microtubule arrays and mitochondria and were usually postsynaptic to axon terminal profiles. Terminals had numerous small synaptic vesicles, often contacted other neuronal profiles and had minimal diameter greater than 0.2 μm. Astrocytic profiles were recognized by their tendency to conform to the boundaries of surrounding profiles, by the presence of glial filaments and/or by the absence of microtubules.#