Studies on the role of 17β-estradiol (E2) in the hippocampus have mainly focused on CA1 and CA3 regions, whereas in dentate gyrus (DG), its role is largely unknown. Here, we examined potential functions of E2 in DG, particularly during development. Immunohistochemistry and in situ hybridization revealed abundance of estrogen receptor (ER)α, but not ERβ, expression in DG. Similar to CA1, analysis of synapse densities revealed a reduction in spine synapse number in DG molecular layer of immature rats and adult mice after inhibition of estradiol synthesis using letrozole. Interestingly, strong expression of ERα was found in Cajal–Retzius (CR) cells, which regulate neuronal migration and synaptogenesis via the extracellular matrix protein reelin. Immunoreactivity of aromatase, the final enzyme of estradiol synthesis, was strongest in mature granule cells. In hippocampal slice cultures, exogenous application of E2 caused an increase in reelin expression in CR cells, which was abolished after blockade of ERs using ICI182,780. Vice versa, inhibition of aromatase activity by letrozole resulted in reduced reelin expression, suggesting that E2 deriving from hippocampal sources contributes to the regulation of reelin as well as to the maintenance of spine synapses in DG. E2 further regulated Notch1, a signaling protein involved in neuronal differentiation.
It is increasingly recognized that 17β-estradiol (E2) has important neuromodulatory functions in the hippocampus, including the regulation of spine and spine synapse density (Woolley and McEwen 1992; Segal and Murphy 2001; Kretz et al. 2004; Rune et al. 2006; Mukai et al. 2007), synaptic protein expression (Kretz et al. 2004; Prange-Kiel et al. 2006; Jelks et al. 2007), long-term plasticity (Warren et al. 1995; Foy et al. 1999; Smith and McMahon 2005; Mukai et al. 2007), and granule cell neurogenesis (Tanapat et al. 1999; Fester et al. 2006). Originally, these effects have been attributed to gonadal estradiol that may reach the hippocampus via the blood–brain barrier (Woolley and McEwen 1992). However, the demonstration that hippocampal neurons express aromatase, the enzyme that converts testosterone to 17β-estradiol (Wehrenberg et al. 2001; Hojo et al. 2004), and are capable of producing E2 de novo (Prange-Kiel et al. 2003; Amateau et al. 2004; Hojo et al. 2004; Kretz et al. 2004; Fester et al. 2009) suggests that many of the effects previously attributed to estradiol of gonadal origin may actually be caused by estradiol that is cyclically produced and released within the hippocampus via regulation of estradiol synthesis by gonadotropin-releasing hormone (Prange-Kiel et al. 2008).
In this context, it is interesting that estrogen-induced spine formation was shown in CA1, but not in CA3 (Spencer et al. 2008), whereas the dentate gyrus (DG) was not considered yet. Region-specific regulation of synaptic proteins in response to endogenously synthesized estradiol and region-specific patterns of aromatase immunoreactivity, however, were shown in the hippocampus (Prange-Kiel et al. 2006). Furthermore, knowledge about the hormone's roles during development is comparably sparse. Evidence suggests that these roles may involve functions that are unique to the developmental period, including the stimulation of axonal outgrowth (von Schassen et al. 2006) and the promotion of ion channel maturation (Wojtowicz et al. 2008; Nunez and McCarthy 2009). Considerable expression of estrogen receptors (ERs) in developing hippocampus (O'Keefe et al. 1995; Orikasa et al. 2000; Solum and Handa 2001; González et al. 2007) further strengthens the notion that the hormone may be important for developmental processes. In addition, expression of aromatase in hippocampus commences early (MacLusky et al. 1994; Amateau et al. 2004; Kretz et al. 2004; Zhao et al. 2007), suggesting that estradiol produced within the hippocampus is a potential regulatory factor during hippocampal development.
In this study, we addressed potential roles of E2 in the DG. Specifically, we 1) carried out a mapping of genomic ER (α and β) expression in DG using immunohistochemistry (IHC) and in situ hybridization (ISH). 2) We further used antisera against aromatase to determine the enzyme's developmental pattern of expression in DG. 3) We determined whether reduction of hippocampal estradiol synthesis due to inhibition of aromatase activity affects synaptogenesis in DG. 4) Prompted by the finding of a unique expression of ERα in Cajal–Retzius (CR) cells, we studied whether E2 influences expression of the extracellular matrix protein reelin that is characteristic for these cells (Rakic and Caviness 1995; Förster et al. 2006). 5) Finally, we examined whether Notch1, a signaling protein involved in neuronal differentiation and acting downstream of reelin (Hashimoto-Torii et al. 2008; Sibbe et al. 2009), is regulated by E2 levels in hippocampal tissue. Our results strongly support the notion that estradiol—including estradiol that is produced within DG—plays an important role for maturational processes in DG.
Material and Methods
All experiments were performed in accordance with the German law on the use of laboratory animals. For IHC or ISH, Wistar rats of different ages (postnatal days [P]5, 8, 10, 26, and 60; n = 4 each [2 male, 2 female]) were deeply anesthetized using a ketamine–xylazine mixture (ketamine 12 mg/mL, xylazine 0.16% in saline, intraperitoneally [i.p.]) and then transcardially perfused with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS). Brains were removed, postfixed for 4 h in PFA, cryoprotected (25% sucrose in PBS) for 24–48 h, and deeply frozen in isopentane (−50 °C). For electron microscopy (EM), immature rats or adult mice (see below) were anesthetized as described above and perfused with 3% glutaraldehyde (in PBS).
In Vivo Experiments
Ten immature (P9) female rats were injected every morning with either the aromatase inhibitor letrozole (n = 5; 4 μg/g body weight, i.p.; letrozole was kindly provided by Novartis, was dissolved in 96% ethanol and prediluted 1:1 with 0.9% NaCl prior to injection) or with vehicle only (n = 5) for 7 days, and then perfused and the hippocampi processed for EM. Furthermore, adult (12-week old) female mice were injected for 7 days with letrozole (10 μg/g body weight, i.p.; n = 3) or vehicle (n = 4) and then treated as described above. In this experiment, an additional experimental group—female mice that had been ovariectomized but not injected (n = 3)—was also included (Zhou et al., 2010).
Organotypic Slice Culture
Hippocampal slice culture preparation was performed according to Stoppini et al. (1991). Briefly, P5 rats (a total of ca. 50) were decapitated, the hippocampus was dissected out, gently placed on the platform of a McIlwain tissue chopper and sliced perpendicular to its longitudinal axis (400 μm). Slices were collected in preparation solution (minimal essential medium [MEM] supplemented with 2 mM Glutamine, pH 7.3) and then transferred to moistened membrane inserts (Millicell CM, 0.4 μm culture plate inserts, 30 mm diameter; Millipore) and maintained in vitro in a 37 °C 95%/5% CO2 humidified incubator. In order to minimize culture variability, slices were distributed such that adjacent slices (sister cultures) were alternately placed into inserts that were intended for different experimental conditions (for IHC analysis: one slice of a pair was subjected to experimental manipulation—e.g., E2, letrozole, or ICI182,780 treatment—whereas the other served as control; for western blot analysis: culture triples were distributed into E2-, letrozole-, or control groups). Incubation medium consisted of 50% MEM, 25% Hank's balanced salt solution, 25% heat-inactivated horse serum, supplemented with 2 mM glutamine, 30 mM glucose, 0.044% NaHCO3, 100 units/mL penicillin, and 100 μg/mL streptomycine (all tissue culture reagents were obtained from Invitrogen). Medium was changed every second day. Following 4 days of incubation, the medium was supplemented for 7 days with either letrozole (10−7 M), E2 (water soluble, Sigma; 10−7 M), or a combination of E2 (10−7 M) and ICI182,780 (Tocris; Cologne; 10−7 M). At the end of the culturing period, slices were fixed with 4% PFA (in PBS) for 2 h, cryoprotected (25% sucrose/PBS, 2 h), and frozen on dry ice, for IHC. For western blots, cultures from the same hippocampus were pooled (5–8 per experimental group, cultured together on one membrane) and immediately deep-frozen until further treatment.
Brains and slice cultures were cut on a cryotome and sections collected in PBS. Subsequently, brain sections (30 μm) were processed “free-floating” (Bender et al. 2007), whereas slice culture sections (20 μm) were mounted to glass slides (always corresponding control and experimental sections on the same slide) and dried, before being processed. For IHC, both brain and culture sections were preincubated with 3% normal goat serum (in PBS) for 1 h at room temperature (RT) and then primary antibodies were applied for 48 h at 4 °C. The following antibodies were used: polyclonal rabbit-anti-ERα C-terminus (Santa Cruz Biotechnology, HC-20, directed against the C-terminus of human ERα; 1:100), polyclonal rabbit-anti-ERα N-terminus (Santa Cruz, H-184, directed against AA 2-185 of human ERα; 1:100), polyclonal rabbit-anti-ERβ (Dianova, PAT-311, directed against AA 376-388 of mouse ERβ; 1:100; antibody showed distinct immunolabeling in paraventricular nucleus of hypothalamus, where ERβ is robustly expressed; Orikasa et al. 2000; data not shown), polyclonal goat-anti-Notch1 (Santa Cruz, sc-6015, 1:500; directed against the C-terminal fragment of Notch1), polyclonal rabbit-anti-aromatase (Yague et al. 2006; directed against AA 488-503 of mouse aromatase; 1:1200; kind gift of Dr I. Azcoitia, Madrid), monoclonal mouse-anti-aromatase (Acris Antibodies, SM2222P; directed against AA 376-390 of human aromatase; 1:60), monoclonal mouse-anti-reelin (Millipore, 1:500), and monoclonal mouse-anti-NeuN (Millipore, 1:200). After primary antibody incubation, sections were washed twice for 15 min in PBS, before secondary antibodies were applied for 3 h at RT: goat-anti-rabbit, anti-mouse, or anti-guinea pig IgGs, coupled with either Cy3 (Dianova; 1:500) or Alexa-488 for immunofluorescence (Molecular Probes; 1:500), or with biotin for light microscopy (1:250; Vector Laboratories). For immunofluorescence, sections were washed again, treated for 1 min with 4,6-diamidino-2-phenylindol (Sigma), then mounted on glass slides, embedded with fluorescent mounting medium (Dako), and coverslipped. For light microscopy, sections were subjected after washes to biotin–avidin–peroxidase complex solution (ABC-Kit, Vector Laboratories). Antibody binding was visualized by incubating sections in a solution containing 0.04% 3,3′-diaminobenzidine, 0.01% H2O2, and 0.01% NiCl2. Sections were then dehydrated in a graded ethanol series and coverslipped with Entellan (Merck). Control experiments included treatment of sections as above, but with primary antibody omitted. No immunoreactivity was observed under these conditions. Results were viewed and photographed using a Leica Axiophot fluorescence microscope or a confocal laser scanning microscope (SP2; Leica).
Deep-frozen cultures were thawed at 4 °C and then manually homogenized (15% wt/vol) with a buffer containing 50 mM Tris–HCl (pH 7.4), 150 mM NaCl, 1% Nonidet P-40, 0.1% sodium dodecyl sulfate (SDS), 0.5% sodium deoxycholate, 5 mM ethylenediaminetetraacetic acid, and a mixture of proteinase inhibitors. The homogenates were centrifuged at 20 000 × g at 4 °C for 20 min, and the supernatant was collected and frozen until further use. For analysis, extracts were run on 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis. Samples were boiled for 5 min, briefly cooled on ice, and then separated at a voltage that prevented excessive heat. Proteins were blotted on nitrocellulose membranes, and blots were treated with 5% milk powder solution (in PBS + 0.3% Triton-X100) and incubated with monoclonal mouse-anti-reelin (1:1000), polyclonal goat-anti-Notch1 (1:500), or monoclonal rabbit anti-Notch1-intracellular cleaved domain (NICD, Cell Signaling Technology, 1:150) antibodies for 24 h at 4 °C. Antibody binding was detected using the ECL-Plus kit (Amersham Pharmacia) and quantified by densitometry using the software “Image J” (Abramoff et al. 2004). Glycerinaldehyde-3-phosphate-dehydrogenase (GAPDH) expression, which was not affected by either E2 or letrozole treatment, was determined as an internal standard (using monoclonal mouse-anti-GADPH from Ambion; 1:10.000).
In Situ Hybridization
ISH was performed as described previously, using antisense and sense-probes generated from transcription vectors containing ERα or ERβ cDNA from marmoset (Rune et al. 2002). Probes were labeled with digoxigenin (3.5:6.5 digoxigenin-uridine triphosphate (UTP)/UTP), using T7- respectively T3-polymerase, according to the manufacturer's instructions (Roche Diagnostics). Sections were processed free-floating, that is, first washed in 2× saline sodium citrate (SSC: 0.3 M sodium chloride, 0.03 M sodium citrate) for 30 min, then subjected to 30-min incubation in a solution composed of 2× SSC/prehybridization solution (1:1). Prehybridization solution consisted of: 50% formamide, 4× SSC, 1× Denhardt's solution, 5% dextrane sulfate, 250 μg/mL yeast transfer RNA, and 100 μg/mL salmon sperm DNA (all components from Sigma). Prehybridization was carried out at 45 °C for 1 h. For hybridization, digoxigenin-labeled RNA probes were added (final concentration: 10 ng/mL prehybridization solution) and sections were incubated for at least 12 h at 45 °C. For all steps, RNAse-free solutions and sterile 6-well plates were used. Following hybridization, sections were subjected to washes with increasing stringency including 2× SSC at RT (twice 20 min), 50% formamide/2× SSX at 55 °C (20 min), 50% formamide/0.1× SSC at 55 °C (20 min), and 0.1× SSC at 55 °C. Hybrid molecules were detected using an anti-digoxigenin antibody tagged with alkaline phosphatase (Roche). Staining was carried out using 4-nitroblue terazolium chloride and 5-bromo-4-chloro-3-indolyl-phosphate (Roche) as chromogens. Color reaction was stopped when distinct, blue-brownish cytoplasmic signal was recognizable. Subsequently, sections were mounted on glass slides, dehydrated, and embedded with Entellan. For ISH/IHC double labeling, sections were first processed for ISH as described above and then subjected to IHC for reelin. To facilitate distinction of ISH- and IHC-signal, antibody binding to reelin was detected using Alexa-488-conjugated secondary antibodies in these experiments.
For EM, brains were dissected, hippocampi were isolated and cut into tissue blocks of ∼2 mm thickness (one brain from a vehicle-injected rat had to be discarded because of insufficuent perfusion). These blocks were postfixed in 1% OsO4, dehydrated in graded ethanol + propylene oxide and then embedded in glycid ether (Serva). Following embedding, blocks were trimmed to contain only DG. Pairs of consecutive serial ultrathin sections, containing DG molecular layer (ML), were then cut using a Reichert-Jung OmU3 ultramicrotome and collected on grids. Sections were stained with uranyl acetate, followed by lead citrate. Electron micrographs were taken of representative fields at ×6600 magnification. Areas occupied by interfering structures such as large dendrites or blood vessels were avoided.
For semiquantitative analysis of reelin expression in organotypic slice cultures, sections were screened for reelin-immunoreactive cells in DG ML using a confocal microscope (SP2; Leica). Ten cells per culture were then randomly selected, and digitized pictures were taken at ×630 magnification by an observer blinded to the experimental condition. Once conditions for taking pictures were optimized (see below), the chosen parameters were kept constant for the documentation of the entire experiment. For subsequent analysis of the digitized pictures, the cell imaging software Openlab (Openlab 2.3.1; Improvision) was used. First, background levels were determined using control sections that had undergone immunostaining with the primary antibody omitted. A gray value that was slightly higher than the background level was then chosen as the appropriate threshold applied to every image under analysis. Only cells in which the nucleus was visible as an unlabeled field in the center of the cell were included. For each individual cell, 5 squares (2 × 2 μm) were superimposed over the labeled area in a way that all regions of the cytoplasm were equally represented. Intensity of staining (indicated value on a gray scale) was then determined for each square and summated. The data from the 10 selected cells per culture were also summated, and the mean was calculated as an index for “average reelin expression/CR cell” in a defined slice culture (resulting in one value per culture that equaled n = 1). Finally, results from experimental and corresponding control slices were compared and analyzed using paired t-test. Data are presented as percentage of control values (mean ± standard error, standard error of the mean [SEM]). Significance levels were set at P < 0.05.
For western blot analysis, relative expression levels of reelin, Notch1, or NICD versus GADPH were calculated. Data were statistically analyzed using paired t-tests (always experimentally treated cultures were compared with corresponding controls) and presented as percentage of control values (SEM; n = 1 refers here to pooled cultures from one hippocampus).
For the analysis of spine synapse density, EM prints covering corresponding neuropil fields in pairs of consecutive serial ultrathin sections were analyzed using the disector technique (Sterio 1984). Photographs were generally taken from the medial aspect of DG ML (the termination zone of medial perforant path). A reference grid was superimposed on the prints, and spine synapses that were present on the reference section, but not on the look-up section, were counted. Only synapses for which presynaptic vesicles, a synaptic cleft and a postsynaptic density were clearly identifiable were included. The disector volume was calculated by the distance (0.1 μm) between the reference and the lookup section. At least 20 neuropil fields (=10 pairs) were analyzed per animal by an observer blinded to the experimental condition. Synapse densities were calculated for each animal (spine synapses/μm3), and data were statistically analyzed using unpaired Student's t-test (significance level: P < 0.05).
ERα Expression in the DG
To determine the expression patterns of ERs in the DG, we applied antisera directed against ERα or ERβ to hippocampal sections from rats of different ages. These studies revealed a robust expression of ERα (Fig. 1A), but only weak expression of ERβ (Fig. 1N), in DG as early as P5. ERα immunoreactivity was detected throughout DG (including granule cells and CA3 pyramidal neurons). However, a particularly strong signal was observed in DG ML, in cells aligning to the hippocampal fissure (Fig. 1A,C, arrows). Closer inspection of these cells revealed a morphology characteristic of CR cells: one prominent dendrite extending from the soma and running nearly parallel to the hippocampal fissure (Fig. 1E,F, arrowheads), whereas fine appendages radiated away from it perpendicularly (Fig. 1E, arrows). When we compared expression patterns in rats of different ages, we found that ERα immunoreactivity was strongly reduced in ML of older rats (Fig. 1B), which is to be expected if these cells are CR cells, because—in rodents—they largely disappear with maturation (Edmunds and Parnavelas 1982; Del Rio et al. 1995). Another characteristic feature of CR cells is expression of the extracellular matrix protein reelin (D'Arcangelo et al. 1995; Ogawa et al. 1995). Therefore, we carried out double-labeling studies using antisera against ERα and reelin (Fig. 1C,D,G–I, antiserum directed against ERα C-terminus). These studies revealed that a large proportion (∼70%), but not all, of ERα-positive cells in ML coexpressed reelin, confirming that they represent CR cells.
It was recently noted that available antisera against ERs are not always 100% selective for their targets but may cross-react with other proteins (Mukai et al. 2007). We therefore carried out several control experiments in order to corroborate our data (Lörincz and Nusser 2008): 1) in addition to detecting ERα C-terminus, we used an antiserum directed against the N-terminus, in order to determine ERα expression. As shown in Figure 1K–M, this antiserum revealed a very similar pattern of expression in ML, and 2) we applied ISH using digoxigenin-labeled riboprobes to determine ERα messenger RNA (mRNA) expression. Comparable with the IHC-results, these studies revealed robust expression of ERα in early postnatal ML (Fig. 2A,D). As with the antisera, the majority of ERα mRNA-postive cells coexpressed reelin (Fig. 2B,E), indicating that these cells are CR cells. Vice versa, virtually all of the reelin-expressing cells coexpressed ERα mRNA (Fig. 2B,C,E,F), suggesting that ER signaling may be important for CR cell function. Notably, in none of these analyses a difference between sexes was evident.
Distinct Pattern of Aromatase Expression in Developing DG
Previous data have indicated a paracrine mode of action for E2 that is synthesized by hippocampal neurons (Prange-Kiel et al. 2003, 2006). Therefore, we next asked whether potential sources of developmentally active E2, close to the CR cells, exist. With established antisera against rat aromatase (Yague et al. 2006), we observed robust immunoreactivity in developing DG. Remarkably, immunosignal was specifically found in the outer parts of the granule cell layer (GCL), where mature granule cells, expressing NeuN, are located (Fig. 3A,C–E). In contrast, the inner part of GCL that mainly contains immature granule cells (Bender et al. 2001) was devoid of immunoreactivity. Aromatase expression was further found in hilar neurons (Fig. 3A), but no immunolabeling of astrocytes or other types of glial cells was observed. In DG of mature rats, aromatase immunolabeling was still predominantly seen in the outer part of GCL, although expression appeared to be decreased and mainly localized to dendrites (Fig. 3B). These data suggest that specifically the granule cells, which are maturationally advanced and may have already established functional synaptic contacts (Vendrell et al. 1998), have the capacity to synthesize and release E2.
Spine Synapse Density in ML Is Reduced after In Vivo Treatment with Letrozole
Established roles of estrogen in hippocampus involve the regulation of synaptic densities (Woolley and McEwen 1992; Segal and Murphy 2001; Kretz et al. 2004). We reasoned that this role could also apply to synaptogenesis during development (which in ML occurs mainly postnatally) and that endogenous E2-production, if significant, should influence this process. We tested this hypothesis in vivo, using immature female rats that were injected for 1 week daily (P9–P16) with letrozole. The brains were then harvested, and spine synapse density in ML was determined as a measure of synaptogenesis. As shown in Figure 4, synaptic density was significantly reduced in letrozole-injected rats as compared with rats that had received only vehicle injections (vehicle: 1.98 ± 0.02, n = 4; letrozole: 1.77 ± 0.07 synapses per μm3, n = 5, P = 0.04), suggesting that the blockade of aromatase had negatively affected synaptogenesis in ML.
Because letrozole, injected i.p., inhibits aromatase function not only in the brain but also in the ovaries, we further extended this analysis, taking advantage of a set of adult experimental animals that had been produced for a parallel study (Zhou et al., 2010). In this study, adult (12-week old) female mice were injected for 1 week with letrozole and compared with vehicle-injected controls. In addition, mice that were ovariectomized but not letrozole injected were included as a third experimental group. Analysis of synapse densities in ML again revealed a significant reduction in mice that had been injected with letrozole compared with vehicle-injected controls (vehicle: 1.97 ± 0.16, n = 4, letrozole: 1.29 ± 0.06 synapses per μm3, n = 3, P = 0.02; Fig. 5A–C). Interestingly, in mice that had been ovariectomized but not letrozole injected, synapse density was not reduced compared with the controls (1.8 ± 0.29 synapses per μm3, n = 3, P > 0.05; Fig. 5D), suggesting that reduction of gonadal E2 alone is not sufficient to significantly affect synaptogenesis in DG. Thus, other—and as we presume—local sources of E2 must play a major role.
Estrogen Stimulates Reelin Expression in CR Cells
Many of the known functions of CR cells involve expression and release of reelin (Förster et al. 2006). Therefore, we studied next whether reelin expression in CR cells is regulated by estrogen. For this purpose, we used organotypic slice cultures of early postnatal hippocampus, which—after a brief recovery period (4 days in vitro)—were treated for 7 days with either E2 (1), a combination of E2 and the ER-blocker ICI182,780 (2), or the aromatase inhibitor letrozole (3), and intensity of reelin immunoreactivity in CR cells was then determined. These studies revealed that: 1) Supplementing exogenous E2 to the culture medium caused an increase of reelin expression in CR cells (to 143 ± 35% of control values; n = 16, P = 0.019; Fig. 6A–C). 2) This increase was abolished when ICI182,780 was coapplied (E2 + ICI182,780: 94 ± 14% of control values; n = 12, P > 0.05), suggesting that the effect was mediated via ERs in the classical genomic way (Fig. 6B). 3) In contrast, blocking of endogenous E2 synthesis with letrozole resulted in a significant downregulation of reelin expression in CR cells (to 82 ± 7% of control values; n = 25, P = 0.048; Fig. 6D), suggesting that E2 deriving from hippocampal sources contributes to this regulation. By counting cells, we ruled out that differing survival rates of CR cells under experimental conditions account for these results: Numbers of reelin-expressing cells in ML were not altered in the experimental compared with the control groups (data not shown).
In order to corroborate these findings, we carried out western blot analyses of cultures that were treated with E2 or letrozole. Consistent with previous findings (Lambert de Rouvroit et al. 1999; Biamonte et al. 2009), western blots revealed 3 reelin-positive bands in the range of 180, 300, and 400 kDa for each of the experimental groups (Fig. 6E). Quantitative analysis of all bands combined revealed significantly enhanced levels of reelin in the cultures that were treated with E2 (158 ± 25% compared with controls; n = 5, P = 0.044; Fig. 6E,F). Letrozole treatment, on the other side, resulted in a slight but nonsignificant decrease of expression (90 ± 9% compared with controls; n = 5, P > 0.05; Fig. 6F). When individual bands were analyzed, differences between the groups were similar, suggesting that the expression and not the processing of the protein was affected by E2. Taken together, these data indicate that E2 exerts a stimulatory effect on reelin expression in CR cells. Furthermore, the reduction of reelin expression after letrozole treatment suggests that E2 from hippocampal sources might contribute to this regulation.
E2 Treatment Reduces Intracellular Cleavage of Notch1
The data presented above, indicating a role of E2 in the regulation of reelin expression and synaptogenesis in developing DG, raise the question whether these 2 processes could be connected—for instance, by proteins acting downstream of both E2 and reelin in a signal cascade that eventually regulates synapse formation. The transmembrane receptor and transcription activator Notch1 could be such a mediator, because it is regulated by both reelin (Hashimoto-Torii et al. 2008; Sibbe et al. 2009) and E2 (Rizzo et al. 2008), and is further known to contribute to differentiation processes in neurons that involve synaptic maturation (Sestan et al. 1999).
To examine whether Notch1 could play a role in E2/reelin-mediated processes, we first determined its localization in developing DG (P10), using an antibody against the 120 kDa, membrane-tethered, intracellular portion of Notch1 (Tokunaga et al. 2004; Hashimoto-Torii et al. 2008). Consistent with the data of others (Sibbe et al. 2009), our studies revealed distinct immunosignal on granule cells, which extended into the dendritic field (Fig. 7A, arrows), suggesting Notch1 expression on the membrane of granule cell somata and dendrites.
We next examined whether E2 or letrozole treatment affects levels of Notch1 in the organotypic slice cultures of P5 rats. Here, we did not observe any effect of either E2 or letrozole when we probed the tissue for membrane-tethered Notch1 (Fig. 7B,C). However, when we used an antibody that only recognizes the intracellularly cleaved and transcriptionally active domain of Notch1 (NICD), we found that levels were significantly reduced after E2 treatment (to 72 ± 8% of control levels, n = 6, P = 0.02; Fig. 7B,D). No effect of letrozole treatment was detectable under these conditions. These findings suggest that estradiol does not regulate expression levels of Notch1 but may influence the cleavage (and thus activation) of the Notch1 receptors.
In this study, we demonstrate expression of ERα in hippocampal CR and a stimulatory role of E2 on the expression of reelin, which is mediated via the classical genomic way of ER signaling. Inhibition of hippocampal aromatase using letrozole reduced the expression of reelin in hippocampal slice cultures, suggesting that endogenously produced E2 contributes to reelin regulation. We further show that blockade of aromatase affects spine synapse density in the ML of immature rats and mature mice. These data suggest that E2 has a significant influence on synaptic plasticity in DG, which already commences early in development and may, at least in part, be mediated via the regulation of reelin. This notion is further supported by our finding that E2 affects Notch1 signaling, which acts downstream of reelin and is critical for neuronal—including synaptic—differentiation.
Aromatase—the enzyme that converts testosterone into estrogen—is expressed in the brain (Naftolin et al. 1975; Roselli et al. 1985; Balthazart 1991), but its neuronal expression has long been thought to be associated with the neuroendocrine regulation of reproductive behavior. This view has changed, when it was shown that brain-derived estrogen is involved in functions such as the regulation of synaptic plasticity, neurogenesis, or neuroprotection, which go much beyond the “classical” role of the hormone (for review, see Balthazart and Ball 2006; Prange-Kiel and Rune 2006; Rune et al. 2006; Garcia-Segura 2008). In addition, cerebrally synthesized estrogen often exerts its effects rapidly and only locally (Prange-Kiel et al. 2003, 2006). This has led to the view that neuron-derived estrogen should be considered a paracrine/autocrine neuromodulator—rather than an endocrine acting hormone—that can influence neuronal function in networks highly specifically and effectively (Balthazart and Ball 2006). In the mature hippocampus, estrogen modulation involves functions such as synapse density (Kretz et al. 2004; Mukai et al. 2007), synaptic protein expression (Kretz et al. 2004; Jelks et al. 2007), or long-term plasticity (Foy et al. 1999; Smith and McMahon 2005; Mukai et al. 2007).
The findings of our study add another facet to this story by suggesting that during development E2—generated within the hippocampus—can act as a differentiation-promoting factor which stimulates synaptogenesis and may exert some of its effects via regulation of reelin expression in CR cells. Our conclusion that the hippocampus itself is a source of developmentally active E2 is based on the following findings: 1) Inhibition of aromatase in hippocampal slice cultures by letrozole leads to a small but significant reduction of reelin signal intensity in CR cells as determined by image analysis and to a nonsignificant reduction of reelin expression as determined by western blot analysis (which may be nonsignificant due to resolution limitations of the method). 2) The immature hippocampus—including DG (Fig. 3)—expresses aromatase, the final enzyme in estrogen biosynthesis (Amateau et al. 2004; Kretz et al. 2004; Zhao et al. 2007). 3) Production of estrogen in the ovaries—a potential alternative source—is very low in rats during the first 2 postnatal weeks (Lamprecht et al. 1976; Bakker and Baum 2008), and it is therefore unlikely that gonadal E2 plays a significant role for the developmental processes that are discussed here. 4) Letrozole treatment, but not ovariectomy, affected synaptogenesis in ML (Figs. 4 and 5). Although more pronounced in mature brain, this effect was also detectable in immature brain, suggesting that the underlying mechanisms are established early. 5) Our conclusion is further supported by earlier notions that hippocampal-derived E2 promotes neuronal differentiation in primary neuronal culture systems (Fester et al. 2006; von Schassen et al. 2006) and by observations suggesting a role of endogenous E2 production in developing neocortex (Martinez-Cerdeno et al. 2007). In addition, recent findings showing an E2-mediated upregulation of reelin expression in developing cerebellum (Biamonte et al. 2009) are fully concordant with the findings of our study. In hippocampus, this upregulation is likely mediated by the ERα subtype, which in the rat—as in humans (González et al. 2007)—appears to be specifically expressed on CR cells.
What could be the role of the estrogen-mediated regulation of reelin? Reelin is a glycoprotein that is secreted by CR cells (d'Arcangelo et al. 1995; Ogawa et al. 1995), the first neurons that differentiate in the developing cortex (Edmunds and Parnavelas 1982; Förster et al. 2006). Reelin secretion by CR cells is crucial for the correct lamination of cortical neurons, including the GCL (Förster et al. 2006). A reduced reelin expression as it was found in our studies after letrozole treatment might thus be expected to result in disturbances of DG granule cell layering. However, no such effect was observed in slice cultures even after long-term treatment (>4 weeks) with letrozole (Lanowski J-S, Bender RA, personal observations). In light of recent findings, this may not be too surprising because a reduction of reelin expression to 50% or more is required to seriously affect neuronal migration, as shown by Niu et al. (2008). Reduction of reelin expression to 80–90% of control levels, as it was found after letrozole in our model, might thus be too subtle to detectably alter granule cell lamination.
Smaller changes in reelin expression are required to affect synaptogenesis, another process, in which reelin emerges as a major player (Borrell et al. 1999; Herz and Chen 2006; Niu et al. 2008). Because E2, as shown in this study, promotes both reelin expression and synaptogenesis, a mutual pathway with the hormone acting upstream to reelin in the signaling cascade is an attractive hypothesis. E2 and reelin already share several signaling pathways. ERs convey estrogen signaling not only via genomic routes (Kushner et al. 2000; Saville et al. 2000; Levin 2005) but further use nongenomic pathways such as signaling via phosphatidyl-inositol-3-kinase (PI3), protein kinase B (Akt), glykogen synthase kinase 3 (GSK-3), and LIM kinase (Bi et al. 2000; Simoncini et al. 2000; Medunjanin et al. 2005; Spencer et al. 2008). These signaling cascades are also stimulated by reelin in order to regulate processes such as the radial migration of neurons and the growth and branching of dendrites during development (Jossin and Goffinet 2007; Chai et al. 2009). E2 and reelin could thus interact in DG in order to coordinate developmental processes such as synaptogenesis, and this interaction could involve an E2-mediated stimulation of the transcription of reelin (Biamonte et al. 2009; present study).
E2 in concert with reelin, or E2 alone, could further influence synaptogenesis by modulating the function of regulatory proteins such as Notch1, which is substantially expressed on the dendrites of granule cells in the immature DG (Fig. 7). Notch1 is a transmembrane receptor that is cleaved upon ligand binding, releasing the intracellular portion (NICD) that translocates to the nucleus where it induces transcription of multiple target genes (Schroeter et al. 1998; Artavanis-Tsakonas et al. 1999). In cortex, Notch1 signaling is involved in many developmental processes, including radial neuronal migration (Hashimoto-Torii et al. 2008; Sibbe et al. 2009) and dendritic differentiation (Berezovska et al. 1999; Sestan et al. 1999; Redmond et al. 2000; Breunig et al. 2007). In maturing cortical neurons, synaptic contacts induce Notch1 activation, which then initiates a restriction of dendritic growth and subsequent arrest in maturity (Sestan et al. 1999). Notch1 may thus play a critical role in controlling the number of synaptic contacts and in maintaining the stability of neurites and connections. Interestingly, both E2 and reelin have the capacity to regulate the function of Notch1. In breast cancer cells, E2 has been shown to reduce Notch's transcriptional activity via an ERα-mediated inhibition of Notch cleavage by γ-secretase (Rizzo et al. 2008). If such a regulation would also occur in brain, a loosening of the Notch-induced maturational arrest could be the consequence and additional synapse formation could result. Our data (Fig. 7) provide first evidence that such a mechanism could indeed play a role in brain, with E2 possibly acting as an auto- or paracrine signal that primes the network for new synapse formation (Srivastava et al. 2008). Experiments are in work in our laboratory to test this hypothesis. Reelin, on the other side, has been shown to increase Notch function by inhibiting the degradation of NICD (Hashimoto-Torii et al. 2008; Sibbe et al. 2009). This mechanism is active in very immature neurons that are still migrating (Hashimoto-Torii et al. 2008) and in radial glia cells of the DG (Sibbe et al. 2009), whereas its relevance for synaptogenesis remains to be determined. Thus, the ultimate outcome of E2 actions in DG could be complex and could depend on type and maturational state of the cells that are targeted.
In summary, our data demonstrate that in DG—as in CA1—E2 is involved in regulating synaptic plasticity. They further suggest specific developmental roles of E2, which appear not to be related to the regulation of reproductive behavior, but indicate a capacity of E2 to exert differentiation-promoting effects. Thus, what is well accepted for many other hormones seems also to be true for the estrogens: In brain, their capacities can be channeled into functions that diverge from the functions they are usually known for—and this not only in adult but also in developing neuronal networks.
Deutsche Forschungsgemeinschaft (Rune 436/4-4)
We wish to thank Dr I. Azcoitia from University Madrid for providing us with anti-aromatase antiserum, and B. Kruck, S. Asmus, E. Schäfer, and H. Herbort for excellent technical assistance. Conflict of Interest: None declared.