Estrogen Modulates Neuronal Bcl-xl Expression and β-Amyloid-Induced Apoptosis

Relevance to Alzheimer’s Disease


  • Christian J. Pike

    1. Institute for Brain Aging and Dementia, Gillespie Neuroscience Research Facility, University of California-Irvine, Irvine, California, U.S.A.
    Search for more papers by this author

  • Lippincott Williams & Wilkins, Inc., Philadelphia

  • Abbreviations used: Aβ, β-amyloid; AD, Alzheimer’s disease; BDP, breakdown products; DIV, day(s) in vitro; DMEM, Dulbecco’s modified Eagle’s medium; ER, estrogen receptor; ERE, estrogen-response element; ERT, estrogen replacement therapy; PHF, paired helical filament; SDS, sodium dodecyl sulfate.

Address correspondence and reprint requests to Dr. C.J. Pike at Institute for Brain Aging and Dementia, 1113 Gillespie Neuroscience Research Facility, University of California-Irvine, Irvine, CA 92697-4540, U.S.A.


Abstract: Recent findings indicate that estrogen is neuroprotective, a cellular effect that may contribute to its clinical benefits in delaying the development of Alzheimer’s disease. In this report, we identify a novel neuronal action of estrogen that may contribute to its neuroprotective mechanism(s). Specifically, we report that estrogen significantly increases the expression of the antiapoptotic protin Bcl-XL in cultured hippocampal neurons. This effect presumably reflects classic estrogen transcriptional regulation, as we identified a putative estrogen response element in the bcl-X gene. Estrogen-induced enhancement of Bcl-XL is associated with a reduction in measures of β-amyloid-induced apoptosis, including inhibition of both caspase-mediated proteolysis and neurotoxicity. A similar relationship between estrogen, Bcl-XL expression, and resistance to degeneration was also observed in human hippocampus. We report neuronal colocalization of estrogen receptor and Bcl-XL immunoreactivities that is most prominent in hippocampal subfield CA3, a region that shows relatively little immunoreactivity to paired helical filament-1, a marker of Alzheimer’s disease neurodegeneration. These data suggest a novel mechanism of estrogen neuroprotection that may be relevant to estrogen’s suggested ability to modulate neuronal viability across the life span, from neural sexual differentiation and development through age-related neurodegenerative conditions.

Accumulating data suggest a significant modulatory role for estrogen in the pathogenesis of Alzheimer’s disease (AD) (for review, see Henderson, 1997). Depletion of estrogen after menopause is thought to increase the susceptibility of females to AD, a position supported by findings that women exhibit an increased prevalence of AD (Zhang et al., 1990; Bachman et al., 1992) and more severe AD-related cognitive deficits (Buckwalter et al., 1993; Henderson and Buckwalter, 1994) relative to men. Further, a significant association between estrogen replacement therapy (ERT) and a reduced risk of developing AD has been reported in several (Henderson et al., 1994; Paganini-Hill and Henderson, 1994; Tang et al., 1996; Kawas et al., 1997) but not all (Brenner et al., 1994) studies. In addition to this preventive role, ERT also may have therapeutic actions against AD, improving cognitive function and/or delaying disease progression (Fillit et al., 1986; Honjo et al., 1989; Ohkura et al., 1994; Schneider et al., 1996; Doraiswamy et al., 1997). Unknown, however, are the cellular and molecular mechanisms of estrogen’s actions that underlie its inhibition of AD pathology.

Estrogen induces multiple cellular effects that either alone or in combination may be relevant to delaying AD onset and preserving cognitive abilities (for reviews, see Birge, 1997; McEwen et al., 1997). Considering the extensive neurodegeneration in AD, one particularly intriguing action of estrogen is neuroprotection. Cell culture studies have noted that estrogen can improve neuronal viability (Faivre-Bauman et al., 1981; Arimatsu and Hatanaka, 1986) and attenuate cell death induced by AD-related insults (Behl et al., 1995, 1997; Goodman et al., 1996; Singer et al., 1996; Green et al., 1997; Mattson et al., 1997). Currently, the mechanism underlying estrogen neuroprotection is unclear, with supportive evidence for both genomic and nongenomic estrogen pathways.

As neurodegeneration in AD appears to involve apoptosis (for review, see Cotman and Anderson, 1995), one possible mechanism of estrogen neuroprotection involves modulation of apoptosis. A role for estrogen in apoptosis regulation has been suggested in nonneural hormone-responsive tissues. For example, in both breast (Bhargava et al., 1994; Doglioni et al., 1994) and uterus (Gompel et al., 1994; Otsuki et al., 1994), estrogen appears to modulate cellular viability by estrogen receptor (ER)-dependent regulation of the antiapoptotic protein Bcl-2. The well-established responsiveness of specific brain regions to estrogen suggests the possibility that estrogen may modulate neuronal vulnerability to apoptosis by acting as an endogenous regulator of apoptosis-related factors. Given the apparent involvement of apoptosis in AD neurodegeneration, the therapeutic benefits of ERT may reflect in part an inhibition of apoptotic pathways in estrogen-responsive neuronal populations. In this report, we begin to examine this novel mechanistic theory of estrogen neuroprotection with a specific focus on Bcl-XL, an antiapoptotic member of the Bcl family closely related to Bcl-2 (Boise et al., 1993) that exhibits relatively high constitutive levels of neuronal expression in adult brain (Yachnis et al., 1997).



Estrogen (17β-estradiol) was purchased from Sigma (St. Louis, MO, U.S.A.) and Research Biochemicals International (Natick, MA, U.S.A.). Droloxifene (3-hydroxytamoxifen citrate) was purchased from Research Biochemicals International. Synthetic β-amyloid (Aβ) peptides were generously provided by Dr. Charles Glabe (University of California-Irvine, Irvine, CA, U.S.A.). Monoclonal and polyclonal Bcl-X antibodies were purchased from Transduction Laboratories (B22620 and B22630; Lexington, KY, U.S.A.), Calbiochem (PC67; San Diego, CA, U.S.A.), and Santa Cruz Biotechnology (sc-634; Santa Cruz, CA, U.S.A.). Monoclonal ER antibodies (specific for the ER α subtype by western blot analysis, according to the manufacturers) were purchased from Alexis Corporation (clone 33; San Diego, CA, U.S.A.), Affinity BioReagents (clone 33; Golden, CO, U.S.A.), and Novocastra Laboratories (clone 6F 11; Newcastle upon Tyne, U.K.). Monoclonal β-tubulin antibody was purchased from Boehringer Mannheim Biochemicals (clone KMX-1; Indianapolis, IN, U.S.A.). Rabbit antisera against caspase-generated actin breakdown products (BDPs) (Yang et al., 1998) were generously provided by Dr. Greg Cole (University of California-Los Angeles, CA, U.S.A.).

Neuronal cell culture

Primary cultures of hippocampal neurons were prepared from embryonic (gestational day 18) Sprague-Dawley rat pups, with minor modifications of a previously described protocol (Pike et al., 1993). In brief, dissected hippocampi were incubated 5 min in 0.125% trypsin at 37°C followed by trypsin quenching with 1 volume of Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal calf serum and 5% horse serum (sera were dextran-charcoal stripped). Cell suspensions were centrifuged (5 min at 200 g), resuspended in serum-free DMEM, mechanically dissociated by repeated passage through a Pasteur pipette, and then filtered through sterile 40-μm nylon mesh (Falcon, Franklin Lakes, NJ, U.S.A.). Cells were plated on poly-L-lysine (0.05 mg/ml)-coated multiwell plates (Nunc, Naperville, IL, U.S.A.) at either 2.5 × 104 cells/cm2 (viability experiments) or 1.5 × 105 cells/cm2 (western blots) in serum-free, phenol red-free DMEM buffered by 26 mM bicarbonate and 20 mM HEPES and supplemented with 100 μg/ml transferrin, 5 mg/ml insulin, 100 μM putrescine, and 30 nM selenium. Some experiments used long-term neuronal cultures that were prepared as described above except that they were plated on a confluent monolayer of rat cerebrocortical astrocytes (Pike et al., 1994). Cultures were maintained in a humidified incubator with 5% CO2 at 37°C.

Experimental treatment of cultures

Cultures were treated with freshly prepared estrogen (solubilized in 100% ethanol, then diluted in DMEM to a final ethanol concentration of ≤0.03%) at 1 day in vitro (DIV) for short-term cultures and at 10 DIV for long-term cultures. After 48-h estrogen exposure, cultures were either processed for western blots (see below) or immunocytochemistry (see below), or treated with toxic insults. Insult exposure was maintained for 18-24 h after which cultures were either processed for western blot or analyzed for cell viability. Neuronal viability was determined by standard cell-counting procedures previously described (Pike et al., 1993, 1996). In brief, viable cells (positive labeling with the dye calcein acetoxymethyl ester; Molecular Probes, Eugene, OR, U.S.A.) were counted in three to four fields per well, three wells per condition. Raw data were statistically examined by ANOVA followed by pairwise comparisons using the Fisher LSD test. Each experiment was repeated in at least three independent experiments. For graphical presentation, data were normalized as a percentage of viability observed in the untreated control condition.

Western blots

Neuronal cultures were processed for western blots with minor modifications of a standard protocol previously described (Pike et al., 1996). In brief, cell lysates were collected in boiling extraction buffer [1% sodium dodecyl sulfate (SDS), 10 mM Tris, pH 7.3], boiled at 100°C for 5 min, sheared by repeated passage through a 26-gauge needle, centrifuged at 16,000 g for 5 min, and quantified for protein content in the SDS-soluble fraction by the bicinchoninic acid method (Pierce, Rockford, IL, U.S.A.). Lysates were then diluted into reducing sample buffer (62.5 mM Tris-HCl, 1% SDS, 2.5% glycerol, and 0.5% 2-β-mercaptoethanol). In experiments involving exposure to toxic insults, cell lysates were collected directly into equal volumes of reducing sample buffer. Equal amounts of samples were electrophoresed on 12% polyacrylamide gels under constant voltage (125 V), and then transferred to 0.45 μm polyvinylidene difluoride membrane (Millipore, Medford, MA, U.S.A.). Membranes were processed using standard techniques with various primary antibodies and appropriate horseradish peroxidase-conjugated anti-mouse (Jackson ImmunoResearch Laboratories, West Grove, PA, U.S.A.) or anti-rabbit (Pierce) secondary antibodies. Immunoblots were visualized on Hyperfilm ECL (Amersham, Arlington Heights, UL, U.S.A.) after exposure to enhanced chemiluminescence reagents (Amersham). In some experiments, individual immunoblots were processed serially with more than one primary antibody by using a stripping procedure (30-min exposure at 60°C to buffer containing 62.5 mM Tris, 2% SDS, and 0.7% 2-β-mercaptoethanol, pH 6.7) between antibody labels. To provide semiquantitative analysis of band intensity, band densitometry was determined from scanned images of nonsaturated immunoblot films, using NIH Image, version 1.59b software. All studies were repeated in at least three independent culture preparations and combined data were statistically compared by matched t test.

Sedimentation assay

The proportion of pelletable Aβ peptide was determined by using our previously described technique (Pike et al., 1995b). In brief, peptide stock solutions were diluted to 25 μM in 20 mM 3-(N-morpholino)propanesulfonic acid (MOPS) (pH 7.3), incubated for 48 h in the presence or absence of 30 nM 17β-estradiol, then ultracentrifuged for 1 h at 1 × 105g. The ratio of protein concentrations, as determined by fluorescamine assay, of the supernatant relative to noncentrifuged peptide was used to calculate the levels of sedimentable peptide. Data represent the mean values of triplicate samples statistically compared by matched t test.

Human case selection

Samples of postmortem human brain from AD (n = 11) and nonpathological control cases (n = 12) were examined by immunohistochemistry (see below). All cases were obtained from the Institute for Brain Aging Tissue Repository and had been examined by a licensed neuropathologist. AD cases satisfied CERAD (Consortium to Establish a Registry for Alzheimer’s Disease) guidelines for diagnosis of AD and did not contain neuropathological evidence of mixed dementias or AD subtypes, including Lewy body and cerebrovascular amyloidosis variations. Control cases exhibited little or no evidence of AD pathology or other forms of neural injury or disease. All female cases had a negative history of ERT. Brain tissues from all cases were fixed and processed as previously described (Cummings et al., 1996) with postmortem delays that ranged from 0.5 to 8 h and averaged 4.9 ± 0.4 h.


Immunocytochemistry of cultured neurons was performed with monoclonal Bcl-X antibody (Transduction Laboratories) using the Vector (Burlingame, CA) ABC (avidin-biotin complex) Elite reagents, as previously described (Pike and Cotman, 1993). Immunohistochemistry of paraformaldehyde-fixed human brain sections was conducted with noted modifications of a previously described protocol (Pike et al., 1994, 1995a). It is noteworthy that sections were exposed to an antigen-unmasking protocol before immunolabeling, a procedure critical for the detection of many antigens including ER and Bcl-X (for review, see Shi et al., 1995). All sections were processed serially in one of two batches to minimize interexperiment variability and facilitate comparisons across cases. Free-floating 40-μm sections were boiled (microwave or conventional heating) 2 × 5 min in 10 mM citrate, pH 6.0. After gradually cooling to room temperature, sections were treated for 20 min with 10% methanol, 1.5% hydrogen peroxide in TBS (0.85% NaCl, 100 mM Tris-HCl, pH 7.4) to quench endogenous peroxidase activity, and then processed for single- or double-immunolabeling using ABC reagents (Vector), as previously described (Pike et al., 1994, 1995a). Immunolabels were visualized with either precipitating diaminobenzidine substrate (Vector) or the red fluorophore Cy3 (Jackson ImmunoResearch Laboratories). Control studies that alternately excluded either primary or secondary antibodies from the protocol did not reveal detectable levels of immunoreactivity. In addition, preabsorption of ER antibody (clone 33) with 10-fold (wt/vol) excess of the immunizing peptide (Affinity BioReagents) robustly attenuated ER immunoreactivity. To provide insight into the variations in immunoreactivity observed both regionally and according to case classifications, relative levels of immunoreactivity were rated as previously described (Pike et al., 1994) according to the following scale: -, absent; ±, faint; +, light; + +, moderate; and + + +, strong immunoreactivity.


Estrogen increases Bcl-XL expression in cultured neurons

To begin evaluating the hypothesis that estrogen modulates both expression of Bcl-XL and vulnerability to apoptosis in neurons, we examined neuronal levels of Bcl-XL in the presence and absence of estrogen. Longterm (≥10 DIV) mixed cultures consisting of differentiated rat hippocampal neurons maintained on a confluent layer of astrocytes were treated for 48 h with 0 or 30 nM estrogen, then processed for immunocytochemistry using a monoclonal Bcl-X antibody. In this culture system, neurons exhibited moderate basal levels of Bcl-X immunoreactivity whereas astrocytes showed low to undetectable levels (Fig. 1A). Examination of treated cultures revealed that estrogen significantly increased Bcl-X immunoreactivity in neurons with no apparent effect in astrocytes (Fig. 1B). Western blot analysis of these cultures showed that estrogen induces a moderate and significant increase in Bcl-XL protein levels (Fig. 1C). Note that the size of the observed doublet of Bcl-X bands is consistent with the 29- and 31 -kDa Bcl-XL bands observed in brain lysates (Mizuguchi et al., 1996) and inconsistent with the ∼20-kDa size of proapoptotic Bcl-XS (Krajewski et al., 1994). Reprobing immunoblots with an antibody directed against β-tubulin revealed no estrogen-induced alterations in expression of this protein, a finding that confirms the estrogen-induced increase in Bcl-XL is not due to either a generalized estrogen effect on protein expression or unequal loading of protein samples in the gel.

Figure 1.

Estrogen increases neuronal Bcl-xL levels in long-term cultures of rat hippocampal neurons. Mixed neuron—glia cultures were maintained for at least 10 DIV before a 48-h exposure to 30 nM 17β-estradiol, and then processed for immunocytochemistry or western blot, using a monoclonal Bcl-x antibody. A: Untreated cultures exhibit low levels of Bcl-x immunoreactivity that is confined largely to neurons. B: After estrogen exposure, increased Bcl-x immunoreactivity is apparent in neurons but not in underlying astrocytes. C: Serial western blotting of a single immunoblot reveals an estrogen-induced increase in Bcl-xL but not in β-tubulin. D: The mouse bcl-x gene contains a putative ERE with only a single nucleotide mismatch relative to the consensus ERE (conserved nucleotide in boldface), suggesting that bcl-x is a target gene for estrogen regulation. This putative ERE is located at positions 3,175-3,187 according to the sequence reported by Grillot et al. (1997).

FIG. 1.

The classic mechanism of estrogen-induced protein regulation involves a genomic pathway in which activated ER acts as a transcription factor after binding the DNA of target genes at specific regions called estrogenresponse elements (EREs). The consensus ERE, originally identified in the Xenopus vitellogenin A2 gene (Klein-Hitpass et al., 1988; Kumar and Chambon, 1988), is a 13-nucleotide palindrome containing two five-nucleotide half-sites separated by three nucleotides of unconstrained sequence. If estrogen’s regulation of Bcl-XL expression is directly mediated by a classic genomic pathway, then the bcl-x gene would be expected to contain an ERE. Examination of the mouse bcl-x sequence (Grillot et al., 1997) reveals the presence of an imperfect ERE that matches the consensus ERE in nine of 10 nucleotides (Fig. 1D) and the functional pS2 ERE in all 10 nucleotides (Berry et al., 1989). The putative bcl-x ERE is located in the 3′ untranslated region ∼1 kb downstream of the stop codon (Grillot et al., 1997).

To simplify interpretation of our culture studies, we next examined estrogen effects in neuron-enriched short-term hippocampal cultures, which are nearly devoid of glia (Pike et al., 1993). Consistent with our findings in the long-term, mixed-culture paradigm, we found that 48-h exposure to either 1 or 30 nM 17β-estradiol induced a significant increase in Bcl-XL expression in neuronenriched cultures. Densitometry measures indicated that the magnitude of this estrogen-mediated effect is ∼50% above basal levels (Fig. 2).

Figure 2.

Estrogen exposure significantly increases Bcl-xL expression in neuron-enriched hippocampal cultures. A: A representative western blot shows increased Bcl-xL levels induced by 48-h exposures to 1 and 30 nM 17β-estradiol. Cells were cultured for 1 day before estrogen treatment. Reprobing this blot with β-tubulin antibody confirms the specificity of the Bcl-xL effect. B: Densitometry analyses of western blots from four independent experiments show that the mean (± SEM) estrogen-induced increases in Bcl-xL expression are ∼50% above basal levels in untreated control cultures. *p < 0.05, by matched t test relative to 0 nM 17β-estradiol condition.

FIG. 2.

Estrogen attenuates Aβ-induced apoptosis in cultured neurons

As the established functional role of Bcl-XL is to negatively regulate progression through apoptosis (Boise et al., 1993; Gonzalez-Garcia et al., 1995), our finding that estrogen significantly increases neuronal Bcl-XL expression predicts that estrogen treatment should also be associated with inhibition of neuronal apoptotic pathways. Recent evidence indicates that antiapoptotic actions of Bcl proteins (i.e., Bcl-2 and Bcl-XL) are linked to decreased activation of caspases (Chinnaiyan et al., 1996; Srinivasan et al., 1996), a family of proteases activated during apoptosis that promote cell death by degrading numerous cellular proteins (Villa et al., 1997). One cellular target of caspases is actin (Mashima et al., 1997). Recent data demonstrate caspase-generated actin BDPs in degenerating neurons and microglia of AD brain (Yang et al., 1998). Using a highly specific antibody against actin BDPs in our short-term, neuron-enriched culture system, we have found that apoptotic but not necrotic insults are associated with significant increases in actin BDPs, an effect that is completely blocked by caspase inhibitors (unpublished observations). To evaluate the possibility that estrogen inhibits caspase-mediated proteolytic events, we investigated whether estrogen treatment affects the generation of actin BDPs in neurons exposed to Aβ, an apoptotic insult (Forloni et al., 1993; Loo et al., 1993) implicated in AD neurodegeneration (Yankner, 1996). We pretreated short-term neuronal cultures for 48 h with 30 nM 17β-estradiol followed by an 18-24-h exposure to a neurotoxic dose (25 μM) of aggregated Aβ peptide, and then probed western blots produced from these experiments with actin BDP antibody. We observed that Aβ neurotoxicity is associated with a significant, approximately fivefold increase in the expression of the ∼30-kDa actin BDP. It is important that in cultures pretreated with estrogen, the Aβ-induced increase in actin BDP expression was significantly reduced by ∼30% (Fig. 3).

Figure 3.

Estrogen pretreatment significantly reduces caspasemediated proteolysis of actin induced by Aβ exposure. A: A representative western blot shows that 24-h exposure of neuron-enriched hippocampal cultures to 25 μM Aβ25-35 causes a robust increase in the ∼30-kDa actin BDP over low baseline levels. Cultures pretreated for 48 h with 30 nM 17β-estradiol exhibit a significant decrease in the Aβ-induced generation of actin BDP. B: Analysis of immunoblots from three separate experiments by band densitometry reveals that the magnitude of the protective estrogen effect is an ∼30% reduction of the Aβ-induced signal. *p < 0.05, relative to the control condition; #p < 0.05, relative to the Aβ condition.

FIG. 3.

Our findings of significant estrogen-mediated increases in neuronal Bcl-XL and decreases in neuronal caspase-mediated proteolysis predict that an additional consequence of estrogen exposure should be decreased vulnerability to Aβ-induced apoptotic death. We evaluated this possibility using an experimental design parallel to that described above. In brief, short-term neuron cultures were pretreated for 48 h with 1 or 30 nM 17β-estradiol, exposed for 18-24 h with various doses of aggregated Aβ peptides, then analyzed for cell viability. We observed that increasing doses of active Aβ fragment Aβ25-35 caused dose-dependent neuronal loss consistent with our previous reports (Pike et al., 1993, 1995b). However, estrogen pretreatment resulted in moderate and significant levels of neuroprotection against toxic doses of Aβ25-35 (Fig. 4A). Similar estrogen protective effects were also observed in response to neurotoxicity induced by the full-length Aβ peptide Aβ1-42 (Fig. 4B). It is noteworthy that comparable levels of neuroprotection were induced by 1 and 30 nM 17β-estradiol, demonstrating the potential physiological relevance of this protective pathway (Fig. 4C). To confirm that this neuroprotection was due to a cellular estrogen effect rather than inhibition of Aβ aggregation, as has been observed for some Aβ protective agents (Lorenzo and Yankner, 1994; Tomiyama et al., 1996; Zhang et al., 1996), we verified that sedimentation (a measure of Aβ aggregation; Pike et al., 1995b) of 25 μM Aβ25-25 was not significantly altered by 48-h incubation with 30 nM 17β-estradiol (Aβ = 21.3 ± 0.4% sedimentation; Aβ + estrogen = 22.2 ± 0.3% sedimentation; p = 0.14).

Figure 4.

Estrogen protects against Aβ-induced neurotoxicity in cultured hippocampal neurons. A: Neuronal cell loss induced over 24 h by increasing doses of aggregated Aβ25-35 is significantly attenuated by 48-h pretreatment with 30 nM 17β-estradiol (solid columns) relative to the untreated condition (hatched columns). Data show mean viability (± SEM) values from three independent experiments. B: In parallel experiments, similar neuroprotective estrogen effects are also observed in response to toxicity induced by 25 μM Aβ1-42. Data show mean viability (± SEM) values from three independent experiments. C: Neuroprotection against 25 μM Aβ25-35 is observed at significant and comparable levels by 48-h pretreatment with both 1 and 30 nM 17β-estradiol. Data show mean viability (± SEM) values from one experiment representative of three independent experiments. D: The antiestrogen droloxifene inhibits estrogen neuroprotection. Hippocampal neurons (1 DIV) were pretreated for 48 h with 0 nM (light columns) or 10 nM (dark columns) 17β-estradiol in the presence (solid columns) or absence (hatched columns) of 0.1 μM droloxifene, and then exposed to 0 or 15 μM Aβ25-35. After 24 h, viability was determined by cell counting. Data show mean viability (± SEM) values from one experiment representative of three independent experiments. *p < 0.05, relative to matched Aβ condition (A, B, and D) or 0 nM 17β-estradiol condition (C).

FIG. 4.

As hippocampus contains functional ERs (Maggi et al., 1989; Bettini et al., 1992), the estrogen neuroprotection observed in our hippocampal culture system could conceivably proceed by either nongenomic estrogen effects or classic ER-dependent genomic pathways that result from estrogen-mediated gene regulation. Our demonstration of estrogen-induced increases in Bcl-xL expression are suggestive of an ER-dependent genomic pathway. If estrogen neuroprotection against Aβ involves ER-dependent effects, then protection should be inhibited by ER antagonists. To investigate this possibility, we conducted estrogen-protection experiments in the presence or absence of a 10 M excess of droloxifene, an antiestrogen metabolite of tamoxifen that shows strong affinity for ER. Consistent with the notion of an ER-dependent effect, the ability of estrogen to attenuate Aβ neurotoxicity was almost completely inhibited by droloxifene (Fig. 4D); similar but less robust effects were also observed with tamoxifen (data not shown). Note that droloxifene alone provides mild, nonsignificant levels of protection consistent with its established actions as a partial agonist with a relatively high ratio of antagonist/agonist activities (Eppenberger et al., 1991).

ER and Bcl-xL are colocalized in human hippocampus

The cell culture findings presented above support a relationship between estrogen, Bcl-xL expression, and vulnerability to apoptosis that we hypothesize may contribute to neuroprotective actions of estrogen. If such a relationship operates in human brain, then several predictions can be made. First, consistent with our in vitro observations that estrogen positively regulates Bcl-xL levels, we would predict a regulator/product relationship between ER and Bcl-xL. To begin evaluating this possibility, we examined patterns of ER and Bcl-x immunoreactivities in aged, nonpathological human hippocampus from both males and females. We noted differential distribution of ER immunoreactivity with relatively strong labeling in CA3/CA4 and comparatively weak labeling in CA1. Immunoreactivity was observed predominantly within pyramidal neurons, although immunoreactivity was detected in some cells with glial morphologies. In addition, stratum oriens interneurons in hippocampal subfields CA1, CA2, and CA3 exhibited strong ER immunoreactivity. Appreciable levels of neuronal ER immunoreactivity were also observed in hilus and to varying extents in dentate gyrus (Fig. 5A). At the cellular level, ER immunoreactivity was typically cytoplasmic, in agreement with some previous findings (Blaustein, 1993), although the more classic nuclear localization was also observed (Fig. 5E). Patterns and intensities of ER immunoreactivity were comparable between males and females (Table 1). These findings were observed with two different monoclonal ER antibodies.

Figure 5.

ER and Bcl-x are colocalized in neurons from human hippocampus. A, C, and E show ER immunoreactivity, whereas B, D, and F show Bcl-x immunoreactivity. A and B: Low-magnification images taken from serial sections of hippocampus from a nonpathological control case that demonstrate similar distributions of ER and Bcl-x immunoreactivities within the pyramidal layer of CA3. C and D: Images from a single hippocampal section double-immunolabeled for ER (C, fluorescent Cy3 label) and Bcl-x (D, diaminobenzidine label) that illustrate the high level of cellular colocalization (arrows) of these two labels in pyramidal CA3 neurons. E: High-magnification image of ER immunoreactivity in CA3 pyramidal neurons reveals both cytoplasmic and nuclear localization, although the former was observed more frequently. Preabsorption of the ER antibody with 10-fold (wt/vol) excess of immunizing peptide largely attenuated the observed immunostaining (data not shown). F: High-magnification image of Bcl-x immunoreactivity in CA3 pyramidal neurons shows that Bcl-x exhibits a cytoplasmic, nonnuclear localization that usually extends into the proximal portions of neurites.

Table 1. Comparison of ER and Bcl-x immunoreactivities in hippocampus CA1 and CA3 across sex and AD diagnosis
   ER immunoreactivityBcl-x immunoreactivity
DiagnosisSexAge (yr)CA1CA3CA1CA3
  1. The relative intensities of ER and Bcl-x immunoreactivities in pathologically normal (Control) and AD cases were rated according to the following scale: -, absent; ±, faint; +, light; ++, moderate; and +++, strong immunoreactivity.


FIG. 5.


Immunolabeling of human hippocampus with Bcl-x antibodies revealed a differential staining pattern across subfields that was striking in its similarity to the distribution of ER immunoreactivity (Fig. 5B and Table 1). As noted with ER antibodies, Bcl-x immunolabeling was most prominent in CA3/CA4, observed at significant levels within hilus and dentate gyrus, and present at relatively low levels in CA1. Staining was primarily neuronal, although cerebrovascular cells also exhibited strong Bcl-x immunoreactivity and some glial immunoreactivity was observed. At the cellular level, Bcl-x labeling of neurons was observed predominantly in the cytoplasm and proximal portions of neuritic processes (Fig. 5F). In comparing Bcl-x immunolabeling across sex, the regional distribution of Bcl-x was conserved but there appeared to be relatively lower levels of Bcl-x immunoreactivity in females (Table 1). These findings were observed using four different monoclonal and polyclonal Bcl-x antibodies.

The regional parallels between ER and Bcl-x immunoreactivities suggested a colocalization of these factors in select neuronal populations of hippocampus consistent with the putative regulator/product relationship. To confirm colocalization at the cellular level, single hippocampal sections were serially processed with Bcl-x and ER antibodies. Most ER-immunolabeled neurons also exhibited positive Bcl-x immunolabeling, indicating a strong concordance between ER and Bcl-x immunoreactivities within individual neurons (Fig. 5C and D).

ER and Bcl-xL-immunoreactive neurons exhibit resistance to AD pathology

If a regulator/product relationship between ER and Bcl-xL expression operates in human hippocampal neurons as suggested above, then one predicted functional consequence of this relationship would be relative protection against neurodegeneration. To begin examining this possibility and its potential relevance to AD, we examined patterns of ER and Bcl-x immunoreactivity in AD hippocampus, then compared these with the labeling pattern of paired helical filament-1 (PHF-1) antibody (Greenberg and Davies, 1990), an established marker of AD neurodegeneration that labels hyperphosphorylated tau localized to dystrophic neurites, neuropil threads, and neurofibrillary tangles. Consistent with observations in normal aged hippocampus, ER (Fig. 6A and B) and Bcl-x (Fig. 6C and D) antibodies immunolabeled hippocampal subfields from AD cases in the following rank order: CA3 > hilus ≥ dentate gyrus > CA1. There was not an obvious increase in the intensity of either ER or Bcl-x immunoreactivity in the AD cases relative to controls, although examination of additional cases with varying disease severity would be required to confirm this point. It is noteworthy that PHF-1 immunoreactivity across hippocampal subfields exhibited a pattern inverse to that of ER and Bcl-x immunoreactivities; PHF-1 labeling was prominent in CA1 and relatively weak in CA3, hilus, and dentate gyrus (Fig. 6E and F). Thus, hippocampal subfields exhibiting relatively high levels of ER and Bcl-x immunoreactivities showed low levels of neurodegeneration as visualized by PHF-1 immunolabeling.

Figure 6.

Regions of AD hippocampus that exhibit colocalization of ER and Bcl-x show relatively low levels of neurodegeneration as visualized with PHF-1 immunoreactivity. Three adjacent sections of AD hippocampus were single-immunolabeled with ER (A and B), Bcl-x (C and D), or PHF-1 (E and F). As labeled, A, C, and E show low magnification of hippocampal fields CA1, dentate gyrus (DG), and hilus, whereas B, D, and F show matched low magnification of hippocampal fields CA3, dentate gyrus (DG), and hilus from the same sections. Hippocampal subfields CA3, dentate gyrus, and hilus exhibit relatively high levels of ER and Bcl-x labeling and low immunoreactivity for PHF-1. In contrast, CA1 exhibits low levels or ER and Bcl-x immunoreactivities and high levels of PHF-1 immunolabeling. Note that not only is PHF-1 immunolabeling confined largely to the pyramidal layer (neurofibrillary tangles, dystrophic neurites, and neuropil threads) but is also observed in the form of neuropil threads within stratum radiatum. The locations of pyramidal cell layers correspond to the labels CA1 and CA3, respectively.

FIG. 6.


In this report, we present data demonstrating that in cultured hippocampal neurons estrogen increases Bcl-xL expression and decreases both the caspase-mediated proteolysis and cell death induced by Aβ. Complementary immunohistochemical findings in human hippocampus demonstrate colocalization of ER and Bcl-x in neuronal populations that exhibit relative resistance to AD neurodegeneration. On the basis of these data, we propose a novel mechanism of estrogen neuroprotection that we theorize may contribute to the protective actions of estrogen against AD. We suggest that in estrogen-responsive neuronal populations, estrogen can enhance neuronal resilience to apoptotic insults by inducing an ER-dependent genomic pathway that involves increased expression of the antiapoptotic protein Bcl-xL and subsequent inhibition of downstream apoptotic steps including caspase-mediated proteolysis (Fig. 7).

Figure 7.

Proposed mechanistic model of estrogen neuroprotection based on the presented cell culture and immunohistochemical findings. We hypothesize that estrogen neuroprotection is mediated in part by a classic genomic pathway involving ER binding and translocation of this complex to the nucleus where it acts as a transcription factor. Further, we theorize that activation of this pathway mediates increased expression of Bcl-xL, an antiapoptotic protein that can inhibit downstream caspases. As a result of Bcl-xL-mediated attenuation of apoptosis pathways, neurons should exhibit increased resistance to apoptotic insults (e.g., Aβ) and thus be afforded the opportunity to undergo repair and/or regenerative processes.

FIG. 7.

Our first key finding is that estrogen significantly increases Bcl-xL levels in cultured hippocampal neurons. These novel data indicate that estrogen can directly regulate neuronal expression of an apoptosis-related factor, a conclusion consistent with a recent in vivo observation, suggesting a similar estrogen-mediated modulation of Bcl-2 (Garcia-Segura et al., 1998). These data are supported by observations in nonneural hormone-responsive tissues where one normal estrogen function is regulating the expression of antiapoptotic members of the Bcl family. For example, uterine levels of both ER and the antiapoptotic protein Bcl-2 rise during proliferative phases of the menstrual cycle and fall during apoptotic secretory stages (Gompel et al., 1994; Otsuki et al., 1994). This estrogen pathway may be activated pathologically in breast cancer as suggested by the high degree of cellular colocalization for ER and Bcl-2 in breast tumors (Bhargava et al., 1994; Doglioni et al., 1994). Consistent with a role in tumorigenesis, estrogen both increases Bcl-2 expression and decreases vulnerability to apoptosis by an ER-dependent mechanism in breast carcinoma cell lines (Teixeira et al., 1995; Wang and Phang, 1995).

Although the definitive mechanism of estrogen’s regulation of Bcl-xL expression remains undetermined, our identification of a putative ERE in the bcl-x gene suggests that this effect is mediated by a classic genomic pathway. An imperfect ERE with the same sequence as the putative bcl-x ERE was identified in the pS2 gene and demonstrated to mediate the estrogen-induced regulation of this gene (Berry et al., 1989). It is interesting that our observations of estrogen-induced Bcl-xL regulation and neuroprotection were moderate in intensity, findings that parallel the moderate activation of the imperfect pS2 ERE by estrogen (Berry et al., 1989). The location of the putative bcl-x ERE in the downstream 3′ region is consistent with the position of EREs from some other estrogen-regulated genes, including c-fos (Hyder et al., 1992). These data demonstrate that, consistent with our theorized model, bcl-x likely serves as a target for transcriptional regulation by activated estrogen receptor.

In this investigation we have focused specifically on Bcl-xL, although it is likely that estrogen may regulate neuronal expression of other Bcl proteins including Bcl-2. Our emphasis on Bcl-xL rather than Bcl-2 stems from their differential patterns of constitutive expression in brain; whereas both Bcl-2 and Bcl-xL are expressed at high levels during neural development, only Bcl-xL retains strong basal levels of expression in the adult human brain (Yachnis et al., 1997). The retained constitutive expression of Bcl-xL suggests that it may serve primarily as a neuronal maintenance factor in adult brain that promotes the continued viability necessary for the longterm survival of nondividing neurons. Thus, preserving the balance of endogenous Bcl-xL regulators such as estrogen may be crucial to the health of specific neuronal populations.

Alteration of protein levels by transcriptional regulation is the classic mechanism of cellular action for estrogen and other steroid hormones (Carson et al., 1990). Thus, proteins identified as targets of estrogen regulation likely have functional roles in one or more estrogenmediated effects. As the primary functional consequences of increased Bcl-xL expression are enhanced viability and protection from apoptosis (Gonzalez-Garcia et al., 1995), we predicted that estrogen-induced increases in neuronal Bcl-xL should be associated with corresponding increases in resilience to injury. Consistent with this possibility, we found that estrogen reduces both the levels of caspase-mediated actin proteolysis and cell death in cultured neurons challenged with Aβ. As noted above, a similar functional relationship between estrogen induction of Bcl-2 and resistance to apoptosis was observed in breast carcinoma cells (Teixeira et al., 1995; Wang and Phang, 1995; Kandouz et al., 1996; Huang et al., 1997). It is noteworthy that although these observations suggest a contributory role Bcl-xL in estrogen neuroprotection, additional studies are required to confirm a causal relationship.

Our data suggest that estrogen neuroprotection against Aβ may involve an ER-dependent genomic pathway. This conclusion is supported in part by our observation that estrogen neuroprotection was blocked by the ER antagonist droloxifene, a finding that confirms the involvement of ER activation. This interpretation is consistent with a report by Singer and colleagues (1996) demonstrating that estrogen neuroprotection against glutamate excitotoxicity is attenuated by the related ER antagonist tamoxifen. However, several other studies have suggested nongenomic actions in estrogen neuroprotection. For example, Green and colleagues (1997) report that the relatively weak ER agonist 17α-estradiol can exhibit neuroprotective activity comparable with 17β-estradiol and that the protective actions of both agents are only partially inhibited by tamoxifen. Others have suggested that the antioxidant activity of estrogen (Mooradian, 1993) contributes to its mechanism of neuroprotection (Behl et al., 1995, 1997; Goodman et al., 1996; Keller et al., 1997). Although antioxidant properties likely contribute to estrogen-mediated cellular protection, the involvement of an antioxidant role of estrogen is unlikely in the present context, as antioxidants do not inhibit Aβ neurotoxicity in our culture paradigm (Pike et al., 1997). An alternative to an antioxidant mechanism of estrogen protection is also suggested by the recent report that estrogen significantly inhibits neurotoxicity induced by Aβ but not by the oxidative insult iron (Mook-Jung et al., 1997). Another theory is that estrogen neuroprotection may be mediated by nongenomic modulation of glutamate receptors (Weaver et al., 1997). Although activation of glutamate receptors can potentiate Aβ-induced degeneration (Koh et al., 1990; Mattson et al., 1992), glutamate receptor antagonists do not attenuate direct Aβ neurotoxicity (Busciglio et al., 1993). One consistent theme in this literature is that estrogen has many cellular actions that may be involved in neuroprotection. We suggest that a new and potentially significant neuroprotective mechanism of estrogen is the positive regulation of Bcl-xL expression.

To determine whether and how estrogen neuroprotection is mediated in humans, it will be necessary to evaluate putative protective mechanisms in human brain tissue. Toward this goal, we began investigating whether our cell culture findings extrapolate to human brain. Immunohistochemical observations in normal aged and AD hippocampus confirm two key predictions generated from our culture data. First, we observed cellular colocalization of ER and Bcl-x immunoreactivities in select neuronal populations of human hippocampus, a finding consistent with the proposed regulator/product relationship between these two proteins. The preferential distributions of ER and Bcl-x immunoreactivities in the CA3 region of aged human hippocampus relative to CA1 are themselves novel observations. Immunocytochemical localization of ER to specific hippocampal subfields in primate brain has not been thoroughly investigated; however, recent findings do suggest that ER mRNA in expressed at high levels in both human (Tohgi et al., 1995) and monkey (Register et al., 1998) hippocampus. In contrast, overall levels of hippocampal ER immunoreactivity in rodent are relatively low (Weiland et al., 1997), perhaps suggesting an important species difference. Nonetheless, the regional pattern of hippocampal ER expression in rodents may be consistent with our immunocytochemical findings in human, namely, that ER levels appear higher in CA3 relative to CA1 (Simerly et al., 1990; O’Keefe et al., 1995). We observed a similar CA3 predominance with Bcl-xL immunoreactivity. It is interesting that comparison of hippocampal Bcl-xL immunoreactivity across sex suggested a trend of relatively lower levels in aged women. This apparent depletion of Bcl-xL in older females may be related to greatly reduced estrogen levels in postmenopausal women. Consistent with this interpretation, recent findings show that ovariectomy of rodents is associated with reduced hypothalamic expression of Bcl-2 (Garcia-Segura et al., 1998), an antiapoptotic protein closely related to Bcl-xL. More subtle age-related disruptions in hormonal regulation of Bcl proteins might be expected in men, as aging males typically exhibit only mild to moderate reductions in testosterone (Kaiser and Morley, 1995), which is locally converted to estrogen by aromatase. If these data are correct in identifying a positive modulatory role of estrogen in maintaining neuronal expression of Bcl-xL, then they may provide mechanistic insight into the identification of female sex as a risk factor for the development of AD.

A second cell culture prediction that also may extrapolate to human brain is estrogen neuroprotection. We observed in cultured hippocampal neurons an ER-dependent neuroprotection against the AD insult Aβ. In AD brain, we noted that hippocampal subfields exhibiting ER and Bcl-x colocalization were relatively resistant to neurodegeneration. The preferential neurodegenerative involvement of CA1 versus CA3 is a well-recognized feature of AD neuropathology, but a firm understanding of this disease characteristic has remained elusive. We hypothesize that one component of CA3 resistance is its neuronal expression of ER, which on activation may yield increased expression of Bcl-xL and a consequent decreased vulnerability to degeneration. The degree to which this putative protective relationship applies to other ER-containing neuronal populations affected by AD (e.g., entorhinal cortex) and how it may vary with hormone status (e.g., gender and ERT) are the topics of ongoing investigations.

In summary, we have demonstrated a relationship between estrogen, Bcl-xL expression, and resistance to degeneration in both cultured hippocampal neurons and in neuronal populations of human hippocampus. On the basis of these data, we hypothesize that estrogen may act as an endogenous regulator of Bcl-xL expression and, as a consequence, function as a modulator of neuronal apoptosis. Additional study of this model may generate new insight into the ability of estrogen to modulate neuronal viability throughout life, from neural development through age-related neurodegenerative disorders.


This study was supported by AG00538, AG15961, and a fellowship from the John Douglas French Foundation. I thank Nima Ramezan-Arab for excellent technical assistance, Dr. Greg Cole for providing fractin antibody, and the Institute for Brain Aging Tissue Repository for supplying human brain tissue.