Estrogen Lowers Alzheimer β-Amyloid Generation by Stimulatingtrans-Golgi Network Vesicle Biogenesis*

Estrogen reduces the risk of Alzheimer's disease in post-menopausal women, β-amyloid (Aβ) burden in animal models of Alzheimer's disease, and secretion of Aβ from neuronal cultures. The biological basis for these effects remains unknown. Here, utilizing cell-free systems derived from both neuroblastoma cells and primary neurons, we demonstrate that 17β-estradiol (17β-E2) stimulates formation of vesicles containing the β-amyloid precursor protein (βAPP) from the trans-Golgi network (TGN). Accelerated βAPP trafficking precludes maximal Aβ generation within the TGN. 17β-E2 appears to modulate TGN phospholipid levels, particularly those of phosphatidylinositol, and to recruit soluble trafficking factors, such as Rab11, to the TGN. Together, these results suggest that estrogen may exert its anti-Aβ effects by regulating βAPP trafficking within the late secretory pathway. These results suggest a novel mechanism through which 17β-E2 may act in estrogen-responsive tissues and illustrate how altering the kinetics of the transport of a protein can influence its metabolic fate.

Reports that post-menopausal estrogen replacement therapy (ERT) 1 is clinically efficacious in delaying the onset of Alzheimer's disease (AD) (1) or improving cognition (2) in post-menopausal women have been corroborated by numerous studies (3)(4)(5). Reports questioning the validity of these retrospective studies (6) demonstrated that women with clinically defined AD have no improvement on specific cognitive tasks following short (2-15-month) trials of estrogen treatment. However, the neuroprotective effects of estrogen are hypothesized to safeguard against the development of AD and not to aid in recovering lost function. The biological mechanism(s) through which estrogen exerts its neuroprotective effects remain largely un-known; hypotheses have included modulation of basal forebrain cholinergic activity, dendritic plasticity, N-methyl-D-aspartic acid (NMDA) receptor density, and regulation of neurotrophin signaling pathways (7).
ERT may slow AD progression by reducing the release of A␤, the primary constituent of amyloid plaques, into brain parenchyma. This phenomenon was first described in non-neuronal (8) and neuronal (9) cell culture. It has now received support from animal models demonstrating that ovariectomized guinea pigs (10) and transgenic mutant ␤APP/PS1-expressing mice (11) develop increased levels of parenchymal A␤ compared with intact littermates. This A␤ burden is reversible with postsurgical estrogen replacement. In addition, a recent study (12) demonstrates an inverse relationship between the levels of 17␤-E2 and A␤42, the more amyloidogenic A␤ variant, in the cerebrospinal fluid of female patients with AD. In summary, results from retrospective clinical trials, recent and on-going animal studies, and corroborating cell culture data all support the hypothesis that post-menopausal cessation of estrogen production may facilitate A␤ deposition and that the clinical efficacy of ERT may be due, in part, to a direct A␤-lowering effect.
A␤ is proteolytically derived from the ␤-amyloid precursor protein (␤APP) within the secretory pathway by distinct enzymatic activities known as ␤and ␥-secretase (reviewed in Refs. 13 and 14). Aggregated A␤ peptides are found predominantly within the extraneuronal space and are believed to initiate toxic and inflammatory cascades leading to neuronal death. The major population of secreted A␤ peptides is generated within the trans-Golgi network (TGN) (15)(16)(17), also the major site of ␤APP residence in neurons at steady state. Full-length ␤APP is transported in TGN-derived secretory vesicles to the cell surface if not first proteolyzed to A␤ or an intermediate metabolite. At the plasma membrane ␤APP is either cleaved to produce a soluble molecule ␤APP (18) or, alternatively, reinternalized within clathrin-coated vesicles to an endosomal/lysosomal degradation pathway (19,20). Thus, the distribution of ␤APP between the TGN and cell surface has a direct influence upon the relative generation of soluble ␤APP versus A␤. This phenomenon makes delineation of the mechanisms responsible for regulating ␤APP trafficking from the TGN biologically relevant.
The formation of TGN vesicles is initiated by the coordinated assembly of coat protein complexes on the cytoplasmic face of donor biological membranes (reviewed in Ref. 21). This complex of cytosolic coat proteins with phospholipid bilayers is necessary to generate coated buds and vesicles. Among the lipids involved in bud formation, phosphatidylinositol (PI) plays a unique role. Its inositol head group can be phosphorylated at single or multiple sites to give rise to a variety of phosphoinositides that can be substrates for enzymes including PI kinases, phospholipase C, and PI-directed phospholipase D (22). These enzymes produce soluble inositol polyphosphates that in turn act as second messengers. These reactions also alter the levels of specific phosphoinositides. Together these effects organize vesicular traffic both spatially and temporally.
Estrogen has also been suggested to alter late secretory pathway vesicle formation. Specifically, electron microscopy in neuroendocrine cells demonstrated an increase in secretory vesicle formation after exposure to estrogen (23). In addition, other steroid molecules such as glucocorticoids have been suggested to regulate trafficking of vesicles from the TGN to the plasma membrane (24,25). Despite this evidence, a direct link between estrogen and known mechanisms involved in vesicle formation has not been shown. However, studies (26 -28) have suggested promising directions for further exploring the link between estrogen and secretory pathway trafficking. For example, transcriptions of methylating enzymes involved in the conversion of phospholipids, and Rab11, a small GTPase involved in the targeting of transport vesicles to the plasma membrane, are up-regulated in response to estrogen (26 -28).
In the present study we demonstrate using a cell-free budding assay that estrogen stimulates the budding of ␤APPcontaining vesicles from the TGN. This alteration in ␤APP trafficking leads to a significant decrease in A␤ generation. In addition, we examine the cell biological mechanism underlying this clinically relevant observation. First, we demonstrate that estrogen exposure initiates the recruitment of Rab11 from the cytosol to the TGN. In addition, expressing a C-terminally truncated mutant Rab11 that does not interact with membranes leads to increased A␤ formation, an effect attenuated by 17␤-E2. These results suggest a mechanism in which increased Rab11 recruitment to TGN membranes helps initiate budding of vesicles containing ␤APP. A separate series of experiments revealed that estrogen also increases the level of PI in the TGN. Furthermore, cell-free assays performed with yeast cytosol containing mutant PITP, a potent PI regulator, corroborate that PI levels are regulated by estrogen and are integral to the budding of ␤APP. Taken together, the effects of estrogen on Rab11 localization and PI levels contribute to increased TGN budding and decreased ␤APP in the TGN. These changes result in decreased A␤ generation and may explain the diminished recovery of A␤ seen in response to estrogen treatment in cell culture and AD animal models.
Neuronal Cultures-Primary neuronal cultures were derived from the cerebral cortices of embryonic day 17 (E17) rat or E14 mouse embryos obtained from timed pregnant animals (Charles River Breeding Laboratories) as described previously (9,19). Brains were removed, and cortices and meninges were excised from the remaining brain. Cortices were triturated in glass pipettes until cells were dissociated. Cells were counted in a hemocytometer and plated in equal amounts in serum-free Neurobasal media with N2 supplement (Invitrogen), 25 mM glutamate, and 0.5 mM L-glutamine on poly-D-lysine-treated (0.1 mg/ml; Sigma) 100-mm dishes (Fisher) (ϳ8 -10 ϫ 10 6 cells per plate) for biochemical analyses. For human mixed brain cultures, cerebrocortical tissue was isolated from gestational age 8 -12-week aborted fetuses (Karolinska Hospital, Sweden). Cultures were maintained in the absence or presence of 200 nM 17␤-E2 for 7 days. Media were replaced every 2 days. Over 95% of cells in the preparations were neurons (9,19).
Antibodies-Antibody 4G8 was purchased from Senetek; FCA3340 recognizing A␤40 but not A␤42 and FCA 3542 recognizing A␤42 but not A␤40 were kindly provided by F. Checler (9). Polyclonal Rab11 antibody and polyclonal GDP dissociation inhibitor antibody were purchased from Zymed Laboratories Inc., South San Francisco, CA.
Enzymes and Drugs-Complete TM protease inhibitor mixture tablets, ATP, GTP, creatine phosphate, and creatine phosphokinase were obtained from Roche Molecular Biochemicals.
Cell Surface Biotinylation-Biotinylation was performed on confluent monolayer N2a cells overexpressing ␤APP695 by using sulfo-NHS-LC-biotin (sulfosuccinimidyl-6-(biotinamido)-hexanoate; Pierce). The reagent was dissolved in PBS with calcium and magnesium, pH 7.2, at 0.5 mg/ml and added twice to the cultures for 20 min at 4°C. After thorough washing, the cells were lysed with 3% SDS as described above. ␤APP was immunoprecipitated by using 369 antiserum (9,19) and was analyzed by Western blot. Biotinylated ␤APP was detected by using horseradish peroxidase-conjugated streptavidin and reaction with a chemiluminescent substrate (PerkinElmer Life Sciences).
Sucrose Gradients-To separate and enrich TGN and ER membranes, cells were homogenized using a stainless steel ball-bearing homogenizer in 0.25 M sucrose, 10 mM Tris-HCl, pH 7.4, 1 mM MgAc 2 , and a protease inhibitor mixture in a final concentration of 1 volume of cell pellet per 5 volumes of homogenizing medium. The homogenate was loaded on top of a step gradient composed of 1 ml of 2 M sucrose, 4 ml of 1.3 M sucrose, 3.5 ml of 1.16 M sucrose, and 2.0 ml of 0.8 M sucrose. All solutions contained 10 mM Tris-HCl, pH 7.4, and 1 mM MgAc 2 . The gradients were centrifuged for 2.5 h at 100,000 ϫ g in a Beckman SW41Ti rotor. Twelve 1-ml fractions were collected from the top of each gradient and assayed for total protein by the method of Bradford. A␤ was assayed by running fractions on 4 -12% SDS-PAGE, transferring the proteins onto polyvinylidene difluoride membranes, and performing Western blot analysis with 4G8, FCA3440, and FCA3642.
Preparation of Permeabilized N2a Cells-It has been well established that incubation of cells at 20°C leads to an accumulation of membrane and secretory proteins in the TGN (15,17). To assay A␤ generation, cells were pulse-labeled with [ 35 S]methionine (500 Ci/ml) for 15 min at 37°C, washed with PBS (prewarmed to 20°C), and chased for 2 h at 20°C in complete media prewarmed to 20°C. Cells were incubated at 4°C in "swelling buffer" (10 mM KCl, 10 mM Hepes, pH 7.2) for 10 min. The buffer was aspirated and replaced with 1 ml of "breaking buffer" (90 mM KCl, 10 mM Hepes, pH 7.2), after which the cells were broken by scraping with a rubber policeman. The cells were centrifuged at 800 ϫ g for 5 min, washed in 3-5 ml of breaking buffer, and resuspended in 5 volumes of breaking buffer. This procedure resulted in Ͼ95% cell breakage evaluated by staining with trypan blue. Broken cells (ϳ2 ϫ 10 6 cells) were incubated in a final volume of 300 l containing 2.5 mM MgCl 2 , 0.5 mM CaCl 2 , 110 mM KCl, and an energy-regenerating system (ERS) consisting of 1 mM ATP, 0.02 mM GTP, 10 mM creatine phosphate, 80 g/ml creatine phosphokinase, and a protease inhibitor mixture. Incubations were carried out at 20 or 37°C as indicated. Each experiment was performed at least 3 times.
Formation of Nascent Secretory Vesicles in Permeabilized Cells-Following incubation of broken cells, vesicle and membrane fractions were separated by centrifugation at 14,000 rpm for 15 s at 4°C in a Brinkman centrifuge. Vesicle (supernatant) and membrane (pellet) fractions were extracted with a cell lysis buffer containing 0.5% Nonidet P-40 and 0.5% deoxylcholate. In some experiments, membrane fractions were further extracted with 70% formic acid and neutralized with 2 M Tris-HCl, pH 8.3.
Immunoprecipitation-Extracted proteins from the various fractions were brought to 0.5% SDS and heated for 3 min at 75°C. Samples were treated with IP buffer (10 mM sodium phosphate, pH 7.4, 100 mM sodium chloride, 1% Triton X-100) and the appropriate antibody added. After incubating overnight, samples were treated with protein A-Sepharose, and immunoprecipitable material was analyzed by SDS-PAGE using 10 -20% Tricine gels (for A␤ species) or 4 -12% Tris glycine gels (for full-length ␤APP).
Neuronal TGN Budding-Neurons were labeled with [ 35 S]methionine (500 Ci/ml) for 2 h at 37°C and washed with PBS. To isolate TGN membranes, cells were homogenized using a stainless steel ballbearing homogenizer in 0.25 M sucrose, 10 mM Tris-HCl, pH 7.4, 1 mM MgAc 2 , and a protease inhibitor mixture in a final concentration of 1 volume of cell pellet per 5 volumes of homogenizing medium. The homogenate was loaded on top of a step gradient composed of 1 ml of 2 M sucrose, 4 ml of 1.3 M sucrose, 3.5 ml of 1.16 M sucrose, and 2.0 ml of 0.8 M sucrose. All solutions contained 10 mM Tris-HCl, pH 7.4, and 1 mM MgAc 2 . The gradients were centrifuged for 2.5 h at 100,000 ϫ g in a Beckman SW41Ti rotor. Twelve 1-ml fractions were collected and assayed for the presence of TGN-38 by Western blot to determine the fraction of TGN membrane enrichment. This fraction was removed from the gradient, and membranes were collected by centrifugation at 15,000 rpm in 0.25 M sucrose for 1 h at 4°C. Membranes were resuspended in 300 l containing 2.5 mM MgCl 2 , 0.5 mM CaCl 2 , 110 mM KCl, and an ERS consisting of 1 mM ATP, 0.02 mM GTP, 10 mM creatine phosphate, 80 g/ml creatine phosphokinase, and a protease inhibitor mixture. Incubations were carried out at 20 or 37°C as indicated. Vesicles were isolated and proteins analyzed as described above. Each experiment was performed at least 3 times.
Preparation of Cytosol Fraction from Yeast Secretory Mutants-Cells were grown at 24°C in yeast extract, peptone, and dextrose (YPD) medium to early log phase. 200 A 600 units were centrifuged for 5 min at 2500 rpm. Cells were washed twice by dilution in reaction buffer (25 mM Hepes-KOH, pH 7.2, 125 mM KOAc) and once in homogenization buffer (reaction buffer plus, 0.5 mM phenylmethylsulfonyl fluoride and 1 mM DTT). 400 l of acid-washed glass beads were added, and cells were lysed for 8 30-s periods of agitation in a vortex at full speed. The samples were mixed with 600 l of homogenization buffer and vortexed four times, each for 30 s at full speed, followed by centrifugation at 3000 ϫ g for 5 min at 4°C. After transferring to Eppendorf tubes, the supernatant was further centrifuged at 100,000 ϫ g for 1 h at 4°C. 50-l aliquots were frozen in liquid nitrogen after protein was measured by the Bradford assay and stored at Ϫ80°C.
Preparation of Mammalian Cytosol Fraction-Ten plates of cells were collected in their media and washed with TEA buffer (10 mM triethanolamine, 140 mM KOAc, pH 7.2) followed by a wash in homogenization buffer (25 mM Hepes-KOH, pH 7.2, 125 mM KOAc, protease inhibitor mixture) and resuspension in 1 volume of pellet/5 volumes of homogenization buffer. Cells were broken in a Ballet homogenizer (clearance 18 m) and a post-nuclear supernatant created by centrifuging at 800 ϫ g for 5 min at 4°C. The post-nuclear supernatant was then centrifuged at 100,000 ϫ g for 1 h at 4°C. 50-l aliquots were frozen in liquid nitrogen after protein was measured by the Bradford assay and stored at Ϫ80°C.
Lipid Extraction-Fractions from the sucrose gradients were diluted to 1-ml total volume if necessary, transferred to 13 ϫ 100 borosilicate test tubes, and mixed with 3.25 ml of chloroform/methanol 1:2.2 (29). The samples were vortexed and centrifuged to pellet proteins. Supernatants were transferred to clean tubes and phase separated by adding 1 ml each of chloroform and 20 mM acetic acid. After aspirating the upper phases, lower phases were reduced in volume by evaporation under a nitrogen stream and dried in a Speed-Vac concentrator.
Phospholipid Analysis-For phospholipid analysis, samples were diluted with 30 l of chloroform/methanol (2:1), and 10-l aliquots from each were spotted onto a TLC plate. Plates were developed in chloroform/ethanol/triethylamine/water (30:35:35:7) and visualized by spraying with 3% copper(II) acetate in 8% phosphoric acid followed by charring. Quantitation for phospholipids was performed on a Storm 860 PhosphorImager in the blue fluorescence mode.
Inositol Lipid Extraction and Analysis-The method of Pike and Eakes (30) was used with minor modifications. 1-ml fractions were transferred to 13 ϫ 100-mm borosilicate tubes containing 1 ml of methanol/concentrated HCl (10:1, v/v) and vortexed with 2 ml of chloroform. The phases were separated and the aqueous phase discarded. The organic phase was re-extracted with 1 ml of methanol, 1 M HCl (1:1), phase separated, and the upper phase discarded. The lower organic phase was evaporated under a nitrogen stream and then dried in a Speed-Vac concentrator. Samples were dissolved in a minimal amount of chloroform/methanol/water (5:5:1) and spotted onto TLC plates, impregnated with 1% potassium oxalate and developed in chloroform/ methanol/water/ammonium hydroxide (17:13.2:2.8:1). Tritiated phosphatidylinositols were detected by autoradiography, and their identity was determined by co-migration with authentic standards.
Densitometry-Band intensities were analyzed and quantified using NIH ImageQuant software, version 1.52.

RESULTS
Estrogen Accelerates ␤APP Transport from trans-Golgi Network-Prior studies demonstrated a significant decrease in levels of secreted A␤ from cells pre-treated with 200 nM 17␤-E2 for 1 week (9). By utilizing a cell-free reconstitution system derived from ␤APP-expressing neuroblastoma (N2a) cells, we examined whether these decreases were secondary to alterations in ␤APP trafficking. The cell-free system has been used extensively as a vehicle within which ␤APP trafficking and A␤ generation have been studied (15,17). In addition, it has been characterized within other cell types (31)(32)(33) where electron microscopy has demonstrated the integrity of TGN stacks and TGN budding (31).
N2a cells treated without or with 200 nM 17␤-E2 for 1 week were labeled with [ 35 S]methionine and incubated at 20°C to accumulate ␤APP in the TGN. Cells were permeabilized and returned to 37°C. Cell-free reconstitution of ␤APP-containing vesicle budding was followed by centrifugation at 4°C to halt further ␤APP trafficking and processing and to separate nascent vesicles from donor membranes (see "Materials and Methods" for details).
␤APP was isolated by membrane solubilization, immunoprecipitation using a C-terminal-specific antibody (369), and separation using SDS-PAGE. A minor ␤APP band is sometimes detected, which may be attributable to incomplete maturation of ␤APP. At 20°C, it is possible that glycosylation is delayed or inhibited. To test this, samples were treated with endo-␤-Nacetylglucosaminidase H and were found to be largely resistant to proteolysis (data not shown) suggesting that immature ␤APP accounts for only a small percentage of total ␤APP. In addition, it is possible that this band is partially accounted for by endogenous mouse ␤APP. ␤APP recovered from TGN-derived vesicles after a 90-min in vitro incubation represents 14% of the original pool of labeled ␤APP within the TGN. Performing the assay at 20°C, or omitting the ERS reduces the percentage of ␤APP recovered from TGN-derived vesicles to 3.5 and 7%, respectively, of total labeled ␤APP (Fig. 1, A and B). When cell-free preparations are created from cells treated for 1 week with 200 nM 17␤-E2, ␤APP recovered from the TGN vesicles represents 30% of the labeled ␤APP protein within the TGN (Fig. 1, A and B). This increase in budding suggests that 17␤-E2 exposure alters cellular trafficking machinery to increase the basal rate of TGN vesicle formation. This increase in TGN vesicle biogenesis could be explained by either activation of integral TGN membrane proteins or changes in the phospholipid composition of the TGN membrane itself. The change cannot, however, be attributed to regulation or activation of cytosolic trafficking proteins because of their removal using extensive washes prior to in vitro incubations.
By stimulating basal TGN budding machinery, 17␤-E2 should increase not only the kinetics of ␤APP trafficking but that of other proteins within the same subcellular distribution. To assess the kinetics of estrogen-enhanced ␤APP budding, [ 35 S]methionine-labeled ␤APP-containing TGN membranes, derived from untreated and estrogen-treated cells, were incubated at 37°C for 30 -120 min. The budding of TGN-derived vesicles was increased by prior estrogen exposure at all time points, reaching a maximum at 90 min (Fig. 1, C and D). Identical experiments were performed studying the trafficking of APLP2, a ␤APP-related protein, and TrkB, an unrelated integral membrane protein. TGN budding of both proteins was similarly increased after estrogen exposure for 7 days (data not shown). These results suggest that prior exposure to estrogen results in a consistent and nonspecific increase in the formation of TGN vesicles.
To determine whether the effect of 17␤-E2 on ␤APP trafficking was reflected in the number of ␤APP molecules on the plasma membrane, we labeled cells with biotin for 4 h at 4°C. Prior estrogen treatment resulted in an increase in ␤APP molecules at the cell surface of ϳ75% (Fig. 1E). Further experiments are underway to assess if estrogen has any effect on the endocytosis of ␤APP at the cell surface. It is well accepted that a population of A␤ is generated from full-length ␤APP, which has been internalized from the cell surface (34), and future experiments will assess the contribution of estrogen to regulating this segment of ␤APP transport. Regardless of the role of 17␤-E2 in endocytosis, estrogen-stimulated TGN budding is particularly relevant to Alzheimer's disease pathogenesis, because of the localization of ␤APP and A␤ generation within the TGN. These findings raised the possibility that changing ␤APP trafficking may result in a direct alteration in A␤ generation.
Stimulation of ␤APP-containing vesicle formation from the TGN by estrogen was concentration-dependent, with a minimal effective concentration of 20 nM and a half-maximal effect at 30 -50 nM ( Fig. 2A). To obtain a maximal effect, 200 nM 17␤-E2 was used in the subsequent studies. Interestingly, the effect of 17␤-E2 on ␤APP trafficking was largely reversed by 50 nM tamoxifen, the nonsteroidal anti-estrogen that blocks estrogen binding to it receptors, although 50 nM tamoxifen alone was able to partially stimulate ␤APP budding from the TGN. In addition, 17␣-estradial (17␣-E2) also mimics the effect of 17␤-E2, however, to a lesser extent (Fig. 2B). Together, these data suggest that the specific effect of estrogen on ␤APP trafficking is probably mediated by estrogen receptors although non-ERmediated mechanisms may not be excluded, especially given the 17␣-E2 effect. Further investigations are necessary for the elucidation of detailed mechanisms underlying the effect of estrogen on protein trafficking.
Estrogen Reduces A␤ Generation in the trans-Golgi Network-A␤ peptides are generated from ␤APP with variable N and C termini. The ϳ4-kDa peptide can end at either amino acid 40 (A␤-(x-40)) or amino acid 42 (A␤-(x-42)). Although the endoplasmic reticulum has recently been implicated in the generation of specific insoluble and non-secreted A␤ peptides, the major site of generation for secreted A␤ is the TGN (17).
To examine whether estrogen's stimulation of ␤APP trafficking from the TGN altered A␤ generation, ␤APP-transfected N2a cells were treated without or with 200 nM 17␤-E2 for 1 week, labeled with [ 35 S]methionine for 3 h, and homogenized in 0.25 M sucrose. Cellular extracts were separated by centrifuga-tion using a discontinuous sucrose gradient, and A␤ peptides were immunoprecipitated from sequential fractions using antibody 4G8. The gradient distribution of A␤ is consistent with previous experiments showing A␤ concentrated within TGN fractions (17). There was a significant and consistent diminution in A␤ levels in all A␤-containing TGN fractions following estrogen treatment, but no apparent redistribution (Fig. 3A), demonstrating that in intact cells the subcellular localization of A␤ peptides is unaffected by estrogen treatment, whereas its generation appears to be significantly reduced.
To examine the effect of estrogen treatment upon the de novo generation of A␤ within the TGN, cell-free assays were performed. A␤ peptides were immunoprecipitated with antibodies specific for the two major C termini (40/42). In the presence of an ERS at 37°C, A␤40 and A␤42 were both detected within the TGN at 37°C (Fig. 3B, lanes 1 and 5). When cell-free TGN preparations were generated from cells treated for 1 week with estrogen there was a 37% reduction in A␤40 formation and an 83% reduction in A␤42 formation (Fig. 3B, lanes 2 and 6). Within post-TGN vesicles, 17% of total A␤40 and 14% of total A␤42 were recovered (Fig. 3B, lanes 3 and 7). Those levels were further reduced following 17␤-E2 treatment (Fig. 3B, lanes 4  and 8).
Despite quantitatively less A␤42 production in the estrogentreated cells lines, both A␤40 and A␤42 were transported at equal rates (22 versus 19%), suggesting that the major effect of estrogen on A␤ metabolism is not on A␤ trafficking but rather on de novo TGN A␤ generation. That estrogen might regulate the trafficking of ␤APP, but not that of its metabolite A␤, is plausible; ␤APP is an integral plasma membrane protein whose cell surface residence and therefore its TGN packaging and transport should be closely regulated. A␤ variants, as by-products of aberrant ␤APP proteolysis, are more likely pack- Endocytosis is not excluded as a relevant pathway that may contribute to A␤ generation. Data from cell-surface biotinylation demonstrates that ␤APP increases with estrogen. To determine whether full-length cell-surface ␤APP contributes to A␤ generation would require budding assays using purified membrane fractions. Although this question is exceedingly interesting, it is well beyond the scope of the present studies and represents a logical continuation of the project.
Cytosol from Estrogen-treated Cells Stimulates Cell-free TGN Budding-The importance of cytosolic trafficking proteins in the genesis and budding of TGN vesicles is well established (reviewed in Ref. 35). It is not known, however, whether these trafficking factors are regulated by estrogen. To study the effects of estrogen upon cytosolic regulation of TGN vesicle biogenesis, we utilized the cell-free trafficking assay.
Cytosol prepared from N2a cells stimulated TGN vesicle budding between 1 ng of protein/ml and 1 mg of protein/ml with maximal stimulation being reached at 100 ng of protein/ml (Fig. 4A). When compared with reactions without exogenous cytosol, reactions with cytosol at 100 ng of protein/ml stimulated ␤APP budding by ϳ25% (Fig. 4B, lanes 3 versus 5). Cytosol prepared from cells that had been treated with estrogen (estrogen-primed cytosol) stimulated budding nearly 2ϫ when compared with reactions using cytosol prepared from control cells (estrogen-naive cytosol) (Fig. 4B, lanes 5 versus 7) and nearly 2.5ϫ when compared with identical reactions without cytosol (Fig. 4B, lanes 3 versus 7).
We also examined whether estrogen-primed cytosol would have any effect on TGN budding in cells already exposed to estrogen. Cell-free assay systems prepared from 17␤-E2treated cells were incubated at 37°C without cytosol, with estrogen-naive cytosol, or with estrogen-primed cytosol. In the absence of any exogenous cytosol, TGN budding was stimulated 2ϫ in cells that had been pre-treated with 17␤-E2 (Fig. 4B,  lanes 3 versus 4). Addition of estrogen-naive cytosol only slightly increased TGN budding in both cell conditions (Fig. 4B,  lanes 3 and 4 versus 5 and 6). Addition of estrogen-primed cytosol to untreated cells significantly increased ␤APP budding when compared with estrogen-naive cytosol (Fig. 4B, lanes 5  versus 7) but did not have an additive effect upon the membranes derived from 17␤-E2-pretreated cells (Fig. 4B, lanes 6  versus 8). These results suggest that treating cells with 17␤-E2 for 1 week generated a maximally active secretory pathway that could not be further stimulated.
Estrogen Stimulates TGN Budding in Neuronal Cell-free Systems-We examined whether this stimulatory effect was limited to cytosol derived from N2a cells. We prepared cytosol from human, rat, and mouse primary fetal cortical neurons that had been untreated or treated with 200 nM 17␤-E2 for 1 week. A greater than 2-fold increase in TGN budding was demonstrated in all three species examined, comparing estro- gen-naive versus estrogen-primed cytosol (Fig. 4C). That human, rat, and mouse estrogen-primed cytosol could stimulate TGN budding suggested that neuronal cytosol could be used to identify estrogen-responsive factors and that mammalian neurons may also be used directly for in vitro assays.
To explore this possibility, we established a cell-free reconstitution system using primary neuronal cultures (see "Materials and Methods"). Budding of ␤APP-containing vesicles was increased from TGN membranes derived from estrogen-treated rat and mouse neurons when compared with TGN membranes derived from untreated neuronal cultures (Fig. 4D). Thus primary neuronal cultures appear to be influenced by chronic estrogen administration in a fashion similar to that of N2a cells, suggesting that common trafficking mechanisms are employed in neurons and neuronal cell lines.
Estrogen Recruits Rab11 to TGN Membranes-Rab11, a cytosolic GTPase, is important for both formation of TGN-derived vesicles and for vesicle formation during endocytosis (36,37). We studied Rab11 because of a report suggesting that Rab11 is induced in response to estrogen administration (26). We found that Rab11 protein levels were not significantly changed in mouse N2a cells following 1 week of estrogen exposure (Fig.  5A). Cytosolic Rab11 protein levels were slightly decreased following estrogen treatment in both mouse N2a cells (Fig. 5B) and human primary neuronal cultures (data not shown).
When protein levels were assayed following subcellular fractionation, Rab11 was highly enriched in the TGN and endosomal membrane fractions derived from estrogen-treated cells (Fig. 5D). These results suggest that estrogen stimulates the recruitment of cytosolic Rab11 to the TGN membranes. Although the change of Rab11 localization in the presence of estrogen may be significant, it seems likely that Rab11 will be only one of many proteins, working in a complex, to initiate changes in vesicle traffic originating from the TGN. Interest-ingly, it has been reported that Rab11 forms an insoluble complex with presenilin 1 (PS1) (38), a protein mutation of which causes the majority of early onset forms of AD. Future studies will be aimed at determining whether PS1 mutations FIG. 5. Rab11 is recruited to TGN membranes in response to 17␤-E2. N2a cells were incubated in the absence or presence of 17␤-E2 for 1 week. Cells were lysed and analyzed by Western blot for total Rab11 (A), cytosolic Rab11 (B), or cytosolic GDP dissociation inhibitor-1 (GDI-1) (C). Additional cells, prepared identically, were subjected to subcellular fractionation as described in the legend to Fig. 3. Samples representing vesicle fractions (lanes 1 and 2) Golgi/TGN (lanes 3 and 4), and ER/PM/Mitochondria (lanes 5-8) were analyzed for Rab11 by Western blot (D).

FIG. 4. Cytosolic factors are up-regulated in response to 17␤-E2.
A, cell-free ␤APP budding assays in the presence or absence of cytosol prepared from cells incubated with 17␤-E2. Titration of cytosol demonstrates an increase in ␤APP budding at all concentrations when using estrogen-primed cytosol versus estrogen-naive cytosol, with a maximum effect at 100 ng of protein/ml. B, ␤APP budding assays using no cytosol (Cyt.) (lanes 3 and 4), estrogen-naive cytosol (lanes 5 and 6), and estrogen-primed cytosol (lanes 7 and 8) in untreated (odd lanes) or estrogen-treated (even lanes) cells. C, cytosol derived from primary human (lanes 1 and 2), rat (lanes 3 and 4), and mouse (lanes 5 and 6) neurons after incubation either in the absence (odd lanes) or presence (even lanes) of estrogen was used to stimulate ␤APP budding in N2a cell-free assays. D, cell-free assays were performed in rat (left panel) and mouse (right panel) primary neurons. In each cell type, cells were incubated in either the absence or presence of 200 nM 17␤-E2 for 1 week before cell-free assays were performed. The budding of ␤APP was assayed using 369 as described in the legend to Fig.  1A. Experiments performed at 20°C provided negative control. alter late secretory pathway traffic either dependent on, or independent of, Rab11 involvement.
That estrogen influenced both ␤APP and Rab11 localization suggested a possible link between the trafficking factor and the transport of ␤APP. Utilizing doubly transfected cells expressing ␤APP 695 and either wild-type Rab11 (Rab11WT) or a Cterminally truncated Rab11 (Rab11⌬C) which cannot interact with membranes, we examined if a mutated Rab11 would disrupt ␤APP trafficking and therefore TGN A␤ production.
Despite equal levels of ␤APP expression in the two cells lines (Fig. 6A), ␤APP/Rab11⌬C cells produced significantly more TGN A␤ and secreted more A␤ than control cells expressing ␤APP/Rab11WT (Fig. 6B, lane 3 versus 1), suggesting that ␤APP trafficking was likely altered. Prior treatment of these cell lines with estrogen attenuated these alterations in A␤ production (Fig. 6B, lanes 1 and 3 versus 2 and 4). These results suggest that Rab11⌬C limits basal trafficking by restricting TGN vesicle formation. This would reduce budding of ␤APP and lead to increased A␤ formation.
Estrogen may be acting to increase basal budding through multiple routes. One possibility is through recruiting endogenous WT Rab11. Transfected Rab11⌬C may compete with WT Rab11 for binding with other necessary trafficking factors to influence budding. Alterations in membrane lipid composition also could contribute to the demonstrated change in budding. Data in support of the second model are presented below.
Estrogen Alters TGN Phospholipid Composition-Protein trafficking within the secretory pathway is believed to be due to a combination of events. The recruitment of cytosolic proteins to the membranes and inherent changes in the lipid composition are believed to initiate the curvature of the membranes (39). That TGN membranes were shown to have a higher rate of vesicle formation after estrogen treatment, in the absence of cytosol, suggested that inherent membrane rearrangements must, in part, contribute to this enhanced vesicle biogenesis. To examine if estrogen influences TGN membrane composition, we measured TGN phospholipid levels using subcellular fractionation and TLC.
The TLC analysis revealed a slight loss of phosphatidylethanolamine, phosphatidylserine, and phosphatidylcholine in TGN membranes after estrogen treatment. At the same time, an increase in these lipids is seen in the post-TGN vesicles with estrogen treatment (Fig. 7, A and B). Because the regulation of phosphatidylinositol (PI) levels by PI-transfer protein (PITP) is essential in membrane vesiculation (40), we examined PI levels in TGN and post-TGN vesicles following estrogen exposure. PI incorporation into TGN membranes and subsequently formed TGN vesicles is increased with estrogen (Fig. 7, C and D). More importantly, the PI shifted from the TGN to the vesicle fraction in estrogen-treated cells. These data suggest that one mechanism through which estrogen stimulates vesicle biogenesis and ␤APP trafficking is through a cytosolic factor-independent redistribution of PI from TGN membranes into nascent vesicles.
The increased phospholipid mass in vesicular fractions following estrogen treatment is likely a result of increased vesicle numbers, not an alteration of vesicle composition; these lipids are most likely structural components of the vesicles, not necessarily part of the budding machinery. Increasing TGN vesicle formation does not necessarily require an increase in total lipid synthesis. Increased vesiculation could be caused by a net shift of lipid mass from the TGN to transport vesicles resulting in an increase in the number of vesicles without altering the phospholipid composition of these vesicles.
Mutations in PITP Inhibit TGN Budding-To study the effects of altering phospholipid levels on TGN budding, we utilized the fact that mammalian neurons and the budding yeast Saccharomyces cerevisiae are highly homologous in many of the proteins regulating secretion. Temperature-sensitive mutants of S. cerevisiae block specific points in the secretory pathway at non-permissive temperatures (reviewed in Ref. 41). Cytosol from WT or mutant yeast strains was used to study TGN budding in the N2a cell-free budding assay. Cytosol from WT, Sec1 (yeast homologue of mammalian Munc-18), and Sec18 (yeast homologue of mammalian NSF) mutant cells supported budding (Fig. 8, lanes 2-4 and 6). In contrast, cytosol from Sec14 (yeast homologue of mammalian PITP) reduced budding to 18% of control (Fig. 8, lane 5).
The inability of cytosol containing mutant Sec14 to support TGN budding is consistent with the function of its mammalian homologue PITP. Mutations in Sec14 or PITP severely restrict the ability of cells to form new secretory vesicles from the TGN (42). These results support the role of phosphatidylinositol in the budding of nascent vesicles from the TGN, a phenomenon demonstrated to be estrogen-responsive. Future studies will be needed to determine which other TGN budding proteins in addition to Rab11 may be responsive to estrogen. DISCUSSION The ability of estrogen replacement therapy to delay the onset of AD in post-menopausal women raised speculation as to the mechanism through which it exerted these protective effects. Estrogen has long been viewed as a neuroprotective molecule. Thus it was unclear whether the ability of estrogen to delay AD onset was due to specific amelioration of an ADassociated pathology or a nonspecific neuroprotective response. The demonstration by Xu et al. (9) that the secretion of ␤-amyloid peptides, a central and unvarying component of AD pathology, was reduced in cells incubated in the presence of estrogen, suggested a direct influence of estrogen upon ␤APP metabolism.
Although estrogen has myriad effects within cells, a direct influence upon discrete components of secretory pathway machinery has not been reported. Several studies reporting effects upon the morphology of individual compartments within the secretory pathway suggested that protein transit though those compartments could be influenced by estrogen. In particular, the TGN appears to vesiculate in response to long term exposure to estrogen. That ␤APP trafficking through the TGN may be accelerated by estrogen suggests a direct mechanism by which estrogen alters levels of A␤ secretion; by decreasing the substrate pool of ␤APP from which A␤ is generated, estrogen could be indirectly responsible for the reduced production and, therefore, reduced secretion of A␤.
We sought to examine the validity of this hypothesis by utilizing a cell-free assay that allowed quantitative study of the kinetics of TGN-specific ␤APP vesicle budding. By taking advantage of the fact that secretory pathway transit can be restricted at specific points, temperature blocks were enforced, and then trafficking was re-initiated by removing those blocks. By using these assays we demonstrated a significant elevation in the rate with which TGN vesicles were derived from donor TGN membranes following estrogen exposure. Budding of vesicles carrying ␤APP as cargo diminishes the substrate pool from which A␤ peptides can be derived.
The production of A␤42, believed to be more highly pathogenic than A␤40, was reduced to a greater extent than A␤40 in the cell-free experiments; however, the mechanisms explaining this phenomenon are not evident. It is possible that estrogen affects the regulation or localization of the ␥-secretase(s). However, it is not yet feasible to study that possibility. Irrespective of the differential effects upon A␤40 and A␤42, these results are a direct demonstration that the metabolism of a protein can be significantly modified by altering its subcellular transport pathway or, more interestingly, simply by influencing its kinetics of transit through specific compartments.
The pharmacological studies of the estrogen concentration dependence were necessary to assess if the effect of estrogen on ␤APP metabolism and trafficking was physiologically relevant.
Previous in vivo animal studies demonstrated estrogen concentration dependence of A␤ formation similar to studies in whole cells. Here we demonstrate that cell-free A␤ generation/secretion is also dose-dependent, consistent with those studies in whole cells and animal models.
To elucidate which pathway estrogen is exerting these ␤APP trafficking and A␤-lowering effects, we studied the effects of 17␣-E2, as well as those of tamoxifen in identical cell-free ␤APP trafficking assays to those performed with 17␤-E2. Tamoxifen did support ␤APP trafficking to a lesser degree but more importantly inhibited the robust effect of 17␤-E2 when the two were co-administered. Also, 17␣-E2 was surprisingly effective at stimulating ␤APP trafficking, although again not quite as efficient as 17␤-E2. These results showing that different estrogen analogues support vesicle budding with varying degrees of success suggest that in addition to a likely direct genomic effect of 17␤-E2, the indirect genomic effects involving estrogen receptor/mitogen-activated protein kinase/extracellular signal-regulated kinase pathways also probably play a role (7,43).
There is a significant barrier to a complete understanding of steroid neurochemistry in the brain; this occurs because many actions of estrogen are mediated via different intracellular mechanisms with the same compound acting as an agonist in one type of neuron and an antagonist in a neighboring cell. In addition to different receptor-mediated pathways, a large number of possible non-genomic effects of estrogen, e.g. reducing oxidative insults, cannot be ruled out. It is clear that many additional studies will need to be performed in receptor knockout or kinase-deficient cell lines, if the field is to begin to understand these competing pathways adequately.
The central findings derived from these experiments suggest several immediate productive areas of investigation. The major conclusions include the following: 1) estrogen influences the kinetics of protein transport through the TGN; 2) this alteration significantly influences the metabolism of ␤APP resulting in diminished A␤ production, and enhancement of TGN vesicle biogenesis which occurs from recruitment of Rab11 to the TGN membranes from the cytosol and an alteration of PI within the TGN. It is likely that lipids and proteins other than PI and Rab11 are redistributed following estrogen treatment. Never- FIG. 7. Phosphatidylinositol is redistributed from TGN to post-TGN vesicles in response to 17␤-E2. A, vesicle and TGN fractions were prepared using subcellular fractionation of N2a cells incubated in either the absence or presence of 17␤-E2. Phospholipids were extracted from these fractions and analyzed using TLC as described under "Materials and Methods." B, quantification of TLC represented as % change in TGN or vesicle fractions after 17␤-E2 treatment. C, vesicle and TGN fractions were prepared as described above, except that N2a cells were labeled first with tritiated inositol as described under "Materials and Methods." D, quantification of tritiated phosphatidylinositols before and after 17␤-E2 treatment in vesicle and TGN fractions. PE, phosphatidylethanolamine; PS, phosphatidylserine; PC, phosphatidylcholine; PI, phosphatidylinositol; PIP, phosphatidylinositol phosphate; PIP 2 , phosphatidylinositol bisphosphate. theless, the finding that it is possible to affect the rate of ␤APP metabolism by altering its trafficking within neurons by modulating the levels of an endogenously produced steroid provides a plausible mechanism by which estrogen is clinically efficacious at delaying or preventing AD. In addition, this basic conceptual framework raises the possibility that other errors of protein metabolism or transport that occur as a result of somatic disease or simply as a consequence of decreased estrogen production late in life may be similarly rectified by resetting the rate of protein transport within the secretory pathway with estrogen administration.