Expression of Liver X Receptor Target Genes Decreases Cellular Amyloid (cid:1) Peptide Secretion*

A hallmark of Alzheimer’s disease is the deposition of plaques of amyloid (cid:1) peptide (A (cid:1) ) in the brain. A (cid:1) is thought to be formed from the amyloid precursor protein (APP) in cholesterol-enriched membrane rafts, and cellular cholesterol depletion decreases A (cid:1) formation. The liver X receptors (LXR) play a key role in regulating genes that control cellular cholesterol efflux and membrane composition and are widely expressed in cells of the central nervous system. We show that treatment of APP-expressing cells with LXR activators reduces the formation of A (cid:1) . LXR activation resulted in increased levels of the ATP-binding cassette transporter A1 (ABCA1) and stearoyl CoA desaturase, and expression of these genes individually decreased formation of A (cid:1) . Expression of ABCA1 led to both decreased (cid:1) -cleavage product of APP Sw ( i.e. C99 peptide) and reduced (cid:2) -secre-tase-cleavage of C99 peptide. Remarkably, these effects of ABCA1 on APP processing were independent of cellular lipid efflux. LXR and ABCA1-induced changes in membrane lipid organization had favorable effects on processing of APP, suggesting a new approach to the treatment of Alzheimer’s disease. The deposition of plaques of amyloid (cid:1) peptide (A (cid:1) ) 1 in the brain is the hallmark Analysis— The significance ABCA1-mediated cleavage. suggests of (cid:2) altered to the site of (cid:2) and represents a new to link alteration in and A (cid:1)

The deposition of plaques of amyloid ␤ peptide (A␤) 1 in the brain is the hallmark of Alzheimer's disease (AD). The generation of A␤ requires sequential cleavage of the type I integral membrane amyloid precursor protein (APP) by ␤then ␥-secretase. ␤-Secretase cleaves APP extracellularly, leaving a 99residue C-terminal fragment (C99) that remains membranebound. ␥-Secretase then mediates an intramembranous cleavage, yielding the A␤ peptide (for review see Refs. 1 and 2). An alternative initial cleavage of APP by ␣-secretase precludes subsequent A␤ formation. Simons and Ehehalt (3) have proposed that the initial cleavage of APP by ␤-secretase occurs in cholesterol-rich liquid ordered domains of the plasma membrane known as "rafts." The basis of this hypothesis was the finding that depletion of cellular cholesterol by cyclodextrin treatment decreases raft formation and markedly reduces A␤ formation (4). Interestingly, presenilins, important components of ␥-secretase, have also been found in cholesterol-rich domains (5).
Cellular lipid homeostasis is controlled by sterol regulator element-binding proteins, transcription factors regulating cholesterol and fatty acid synthesis pathways, and by liver X receptors (LXRs), oxysterol-activated nuclear receptors that induce a battery of genes involved in cellular lipid efflux and transport (6,7). The two forms of LXR, ␣ and ␤, are both expressed in the brain. LXR␤ is broadly expressed in the developing and adult brain and is present in both neurons and glial cells (8). Recent studies show an essential role for LXRs in brain structure and function as aging LXR␣/␤ knockout mice develop cellular lipid inclusions, abnormalities of the choroid plexus, and closure of the lateral ventricles (8). Although this pathology is different from that of AD, LXRs could potentially have a role in modulating the course of chronic neurodegenerative diseases.
An important target of LXRs is the ATP-binding cassette transporter A1 (ABCA1) (9). ABCA1, the defective molecule in Tangier disease, mediates efflux of cellular phospholipids and cholesterol to lipid-poor apolipoprotein, including apolipoprotein A-I (apoA-I) and apoE, which are present in the cerebrospinal fluid (10). Treatment of mice with LXR agonists resulted in increased expression of LXR target genes in the brain, especially ABCA1 (8), and LXR activation induces lipid efflux from glial cells (11). These observations led us to hypothesize that activation of LXRs in neurons might induce ABCA1, resulting in cholesterol efflux and reduced secretion of A␤.

MATERIALS AND METHODS
Cell Culture-Mouse neuroblastoma Neuro2a cells stably expressing Swedish APP695 were previously described (12). Neuro2a-APP Sw cells were grown in Dulbecco's modified Eagle's medium (DMEM)/OptiMEM supplemented with 5% fetal bovine serum at 37°C in a humidified 5% CO 2 incubator. Tissue culture reagents were from Invitrogen. Transient and stable transfections were performed with LipofectAMINE 2000 (Invitrogen). 22(R)-Hydroxycholesterol and the synthetic LXR activator TO901317 were purchased from Sigma. 9-cis-Retinoic acid (9-cis-RA) was from Biomol, apoA-I was from Biodesign, and the apoE isoforms were from Calbiochem.
Detection of Secreted A␤-In experiments with LXR activation, the cells were first induced with activators for 24 h in DMEM, 5% lipoprotein-deficient serum. The cells were then incubated in fresh medium for 4 h for A␤ measurements. In experiments with ABCA1 transient transfections, 24 h after transfection, the cells were incubated with DMEM, 1% lipoprotein-deficient serum with or without apolipoproteins for 6 h for A␤ measurements. After the indicated treatments, conditioned medium was collected on ice and centrifuged at 6000 ϫ g for 10 min to remove cell debris. Immunoprecipitation of A␤ or C99 was performed with monoclonal antibody 4G8 (Signet) and protein A/G-conjugated agarose (Santa Cruz). A␤ and C99 were then extracted in NuPAGE sample buffer (Invitrogen) and fractionated in 4 -12% NuPAGE Bis-Tris Gel (Invitrogen). Fractionated proteins were then transferred to polyvinylidene difluoride membrane (Bio-Rad) and blotted with monoclonal antibody 6E10 (Signet) after boiling. The immunoblots were developed using the ECL system (Pierce) scanned and quantified by ImageQuant (Molecular dynamics). Quantitation of A␤40 and A␤42 was performed using a commercial A␤ enzyme-linked immunosorbent assay kit (BIOSOURCE). Cellular Cholesterol Efflux-The efflux experiments were performed as described (13). Briefly, Neuro2a cells were labeled with 1 Ci/ml The cells were incubated with 10 g/ml purified human apoA-I or apoEs in DMEM, 1% lipoprotein-deficient serum for 6 h, and the medium was collected and centrifuged at 6000 ϫ g for 10 min to remove cell debris and cholesterol crystals. The cells were lysed in 0.1 M sodium hydroxide, 0.1% SDS, and radioactivity was determined by liquid scintillation counting. Efflux was expressed as the percentage of radioactivity in the medium relative to the total radioactivity in cells and medium.
Statistical Analysis-The significance of differences between groups was assessed by Student's t test

RESULTS
To evaluate the effects of LXR activators on APP processing, we employed neuron-derived Neuro2a cells, stably transfected with human APP Sw containing a mutation that increases the formation of total A␤ (the Swedish mutation) (14). The cells were treated with increasing doses of the synthetic LXR␣/␤ activator TO901317 (TO), and A␤ secretion into medium was measured by immunoprecipitation and Western blotting (Fig.   1A). This revealed a decrease in the secretion of A␤, with an approximate 50% reduction at 1 M TO. The amount of the soluble form of APP (APPs␣) in medium was not significantly changed by administration of the LXR activator (not shown). LXR acts in a heterodimeric complex with retinoid X receptor (RXR), and the response of genes to LXR activators is increased in the presence of RXR activators, such as 9-cis-retinoic acid (15). When cells were treated with 1 M TO901317 plus 1 M 9-cis-retinoic acid, there was a further reduction in A␤ secretion to 20% of control (Fig. 1B). A␤ formation was reduced when the cells were treated with the natural LXR activator, 22(R)-OH cholesterol, together with 9-cis-retinoic acid (Fig.  1C). LXR/RXR activation also decreased the secretion of A␤ formed from endogenous APP in 293 cells (data not shown), indicating that these effects were not dependent on overexpression of mutant APP. A␤ is secreted in several different forms. Although A␤40 (40 amino acids) is the predominant species, A␤42 (42 amino acids) is a minor but more amyloidogenic form. Whereas A␤42 is formed predominantly in the endoplasmic reticulum and trans-Golgi, A␤40 is made in plasma membrane, endocytic compartments, and trans-Golgi (16,17). Measurement of both forms of A␤ in medium of Neuro2a cells by enzyme-linked immunosorbent assay revealed that LXR/RXR activation (1 M TO, 1 M 9-cis-RA) decreased the secretion of A␤40 by about 70%, whereas A␤42 was more moderately reduced (Fig. 2). Because the concentration of A␤40 in medium was ϳ8-fold more than that of A␤42, the marked decrease in A␤ signal in Fig. 1 primarily reflects a reduction in A␤40.
Neuro2a cells treated with LXR/RXR activators (1 M TO and 1 M 9-cis-RA) showed a marked induction of ABCA1 protein (Fig. 3A). To determine whether induction of ABCA1 might be responsible for decreased secretion of A␤, Neuro2a cells were transfected with ABCA1 and incubated with or without apoA-I. Transfection of ABCA1 resulted in levels of ABCA1 protein comparable with that induced by LXR/RXR activators (Fig. 3A). Expression of ABCA1 decreased the formation of A␤ without affecting cellular APP levels (Fig. 3B). Surprisingly, the major effect of ABCA1 expression was observed without the addition of the extracellular acceptor apoA-I, and there was only a slight further decrease in A␤ when apoA-I was added to medium (Fig. 3B). Measurement of cellular cholesterol efflux showed a small but significant increase in efflux (1.0 Ϯ 0.1% of total cellular cholesterol) with expression of ABCA1 and the addition of apoA-I. Similarly, there was an increase in efflux of cellular phospholipids (1.9% of total cellular choline-labeled lipids). However, expression of ABCA1 alone did not increase cholesterol or phospholipid efflux compared with nontransfected controls. We also measured cellular cholesterol mass. Under our experimental conditions, there was no detectable cholesteryl ester in cells, either with or without expression of ABCA1.
As a control for possible nonspecific effects of ABCA1 expression unrelated to functional activity, we transfected Neuro2a cells with a mutant form of ABCA1 (Walker motif mutant); this mutant is well expressed on the cell surface but inactive both in lipid efflux and binding of apoA-I (18). When expressed at levels similar to those of wild type ABCA1, the mutant failed to alter A␤ secretion by Neuro2a cells (Fig. 3C), indicating that the effects of ABCA1 expression are related to the activity of the transporter, even though they do not require lipid efflux.
Because apoE is a major apolipoprotein in the central nervous system and the apoE4 isoform is associated with increased risk of AD (19,20), we also examined the effects of apoE on A␤ formation. Expression of ABCA1 with the addition of apoE also resulted in a profound decrease in A␤ formation (Fig. 4). Again, the major effect was attributable to ABCA1 expression alone. The addition of apoE2 resulted in a small but significant further decrease in A␤ formation, whereas apoE3 and apoE4 did not produce significant further reductions in A␤ secretion. The addition of apoE isoforms without ABCA1 expression did not affect A␤ secretion (not shown). Both ␤and ␥-cleavage of APP are required for the generation of A␤. We next carried out a series of experiments to determine which step is affected by ABCA1 overexpression. To measure ␤-cleavage, we immunoprecipitated the ␤-secretase cleavage product (C99) from cell lysates. There was an 85% decrease of cellular C99 in Neuro2a-APP Sw cells transiently transfected with ABCA1 compared with mock transfected control (Fig. 5A). ␥-Secretase mediates the final step in A␤ generation (1). To assess the direct effects of ABCA1 expression on ␥-secretase cleavage of APP, Neuro2a cells were transfected with constructs encoding C-terminal APP fragments (C99) containing a Myc epitope tag. ABCA1 expression was found to decrease the cleavage of C99 peptide, indicating a decrease in ␥-secretase processing (Fig. 5B). The addition of apoA-I did not provide a further significant decrease in A␤ generation (not shown).
In addition to targeting genes involved in cellular cholesterol efflux and transport, LXRs also activate synthesis of monounsaturated fatty acids (21). SCD is a key LXR target gene that catalyzes the conversion of stearoylCoA to oleoylCoA and increases the content of mono-unsaturated fatty acids in cell membrane phospholipids (22). We have recently shown that SCD activity decreases the amount of liquid ordered domains in the plasma membrane (13). There are two forms of SCD, both LXR targets, with similar catalytic activity and cellular effects (22). Using an antibody that recognizes both forms of SCD, we showed that treatment of Neuro2a cells with LXR activators resulted in a modest 1.6-fold increase in SCD protein (Fig. 6A). Transient transfection of SCD in Neuro2a cells resulted in a decrease in A␤ secretion into medium (Fig. 6B). Because Neuro2a cells have high basal levels of SCD activity, we also carried out similar experiments in 293 cells that have much lower basal SCD expression (13). Transient overexpression of SCD resulted in an increase in APPs␣ formation (not shown) and a profound decrease in A␤ generation (Fig. 7A) that was associated with a marked decrease in ␥-secretase cleavage of the C99 peptide (Fig. 7B). DISCUSSION The idea that changes in cellular lipid metabolism might favorably influence APP processing has attracted considerable attention (23). Cellular cholesterol depletion, based on treatment with cyclodextrins or statins, reduces membrane rafts, increases ␣-cleavage of APP, and reduces ␤-cleavage, leading to less A␤ secretion (24). One explanation for these observations is decreased partitioning or trafficking of APP into cholesterolenriched membrane domains where the ␤-cleavage enzyme resides (23,25). In contrast, in our study LXR activators and ABCA1 expression were observed to decrease A␤ secretion in FIG. 4. The effect of apoE isoforms on A␤ secretion. Neuro2a-APP Sw cells were transfected with empty vector (control) or ABCA1 vector, and apoE isoforms were added during the last 6 h of the experiment; data are the means Ϯ S.E. for five separate experiments conducted in duplicate or triplicate. *, p Ͻ 0.01 compared with mock transfected; #, p Ͻ 0.05 compared with ABCA1 transfected without apolipoprotein.

FIG. 5. ABCA1 overexpression decreases ␤-cleavage of APP Sw and ␥-cleavage of the 99-amino acid C-terminal fragment of APP. A,
Neuro2a-APP Sw cells were transfected with either empty vector or vector containing ABCA1 cDNA. The ␤-cleavage product C99 was detected by immunoprecipitation of cell lysates with antibody 4G8 followed by immunoblotting with 6E10. Cellular APP (cAPP) level was measured by immunoblotting cell lysates. The data are the means Ϯ S.E. for five separate experiments; *, p Ͻ 0.01 compared with mock (mk) transfected. B, Neuro2a cells were transfected with vector containing either C99 Myc alone or C99 Myc with ABCA1. A␤ was detected as in Fig. 3. Cellular C99 was detected by immunoprecipitation with anti-Myc antibody followed by Western blot with 6E10. The data are shown for three separate experiments conducted in duplicate. #, p Ͻ 0.05 compared with C99 transfected alone. the absence of cellular lipid efflux, and this was not accompanied by changes in ␣-cleavage and was mediated in part by a decrease in ␥-secretase cleavage of the C-terminal 99 amino acid peptide of APP. Our results suggest that ABCA1-mediated translocation of membrane cholesterol leads to a decrease in ␥-secretase cleavage. This suggests either sensitivity of ␥-secretase activity to membrane environment or altered trafficking of C99 to the site of ␥-cleavage and represents a new mechanism to link alteration in membrane lipid organization and A␤ production.
Our studies were undertaken as a test of the hypothesis that LXR activation would increase lipid efflux and thereby decrease A␤ generation. However, LXR activators decreased A␤ formation in the absence of extracellular acceptors (Fig. 1), and the addition of apoA-I or any of the three isoforms of apoE resulted in only minor further changes in A␤ (Figs. 3 and 4). Serving as a control for nonspecific effects related to ABCA1 expression, the Walker motif mutant of ABCA1 failed to decrease A␤ formation (Fig. 3C). These findings indicate that the decrease in A␤ is related to an intrinsic cellular activity of ABCA1. ABCA1 probably acts as a lipid translocase at the plasma membrane (26) and causes changes in plasma membrane morphology (27). Recently, Vaughan and Oram (28) showed that ABCA1 expression increases cell surface cholesterol oxidase-accessible domains, indicating a redistribution of cholesterol toward the outer membrane, independent of extracellular lipid acceptors. Decreased A␤ formation might result from an ABCA1-induced redistribution of membrane cholesterol either at the plasma membrane or in the Golgi or endocytic compartments.
Although ABCA1-mediated lipid translocation could affect raft organization, it seems unlikely that an alteration in plasma membrane rafts fully accounts for our findings. ABCA1 appears to be localized in and to induce cholesterol efflux from nonraft membrane regions (29), and ABCA1 expression did not alter the distribution or amount of plasma membrane liquid ordered regions in 293 cells. 2 In contrast, SCD activity decreases membrane liquid ordered regions (13), and was associated with increased ␣-cleavage of APP, similar to the effects of cholesterol depletion, However, changes in SCD protein levels in Neuro2a cells were modest and unlikely to account for a major part of the effect of LXR activators.
Both ABCA1 and SCD expression were associated with a decrease in ␥-secretase cleavage of the C-terminal 99 amino acids of APP (Figs. 5B and 7). In the case of APP Sw , ABCA1 decreased ␤-cleavage by 85% (Fig. 5A), but the overall decrease in A␤ secretion was only around 60% (Fig. 3B), which is comparable with the effect of ABCA1 on ␥-cleavage (Fig. 5B). This suggests that under these conditions ␥-cleavage is the ratelimiting step in A␤ secretion. Our findings suggest that this key enzyme of A␤ formation is highly sensitive to its membrane lipid environment, possibly reflecting the fact that ␥-secretase mediates an intramembranous cleavage of APP. Alterations in membrane fluidity caused by lipid translocation or increased content of mono-unsaturated phospholipids could lead to a decrease in ␥-secretase activity. Although the cellular compartmentalization of ␥-secretase activity is not completely understood (1), several reports suggest that A␤ formation mainly occurs in the endocytic compartments (30) where components of ␥-secretase, including presenilins, are shown to be present (31). An alternative interpretation of our results is that induction of ABCA1 activity results in altered cellular trafficking of APP or the C-terminal fragment of APP, providing less substrate for ␥-secretase cleavage.
The finding that cellular cholesterol depletion leads to less A␤ secretion has fostered the idea that statins could be useful in the treatment of AD. Epidemiological studies have suggested that statin therapy is associated with decreased prevalence of AD (32,33). However, brain cholesterol is derived by local synthesis (not from plasma low density lipoprotein), and statins would have to be present in the brain at high levels to alter neuronal lipid metabolism. Although human studies show an association between statin treatment and decreased prevalence of AD, such associations can reflect the influence of confounding factors, as appears to be the case for statins and bone disease (34). A recent placebo-controlled prospective trial of statin therapy in the elderly failed to show any improvements in cognitive function (35).
Inhibition of cholesterol esterification by acylCoA:cholesterol acyl transferase inhibitors has also been proposed to favorably affect processing of APP (36). Because cholesteryl esters have minor solubility in membranes and are thought to be present in cells as inert lipid droplets, it is unlikely that these effects are related to cholesteryl ester accumulation. One possibility suggested by our studies is that acylCoA:cholesterol acyl transferase inhibition leads to accumulation of cellular free sterol and conversion to LXR ligands via endogenous oxysterol synthesizing enzymes such as 24-cholesterol hydroxylase (37). Cholesteryl esters were not stored in cells in appreciable amounts under the conditions of our experiments, and changes in cellular acylCoA:cholesterol acyl transferase activity are thus unlikely to account for the findings.
Our findings, along with another recent report (38), suggest that LXR activators currently being developed for the treatment of atherosclerosis might have therapeutic efficacy in Alzheimer's disease. LXR/RXR activation markedly decreased A␤40 but had less effect on A␤42 secretion. Although A␤42 is thought to be more amyloidogenic than A␤40, much more A␤40 is secreted by neurons, and both forms are found in amyloid plaques (2). Therefore, controlling the predominant A␤40 secretion could be beneficial. Our results differ from another report where a 65% increase in A␤42 and no significant change in A␤40 was found in Neuro2a cells treated with TO and 9-cis-retinoic acid (10 M) (39). The reasons for this discrepancy are not clear. However, our findings substantially agree with another report (38), which appeared while our paper was under review. Our results extend these latter studies by the direct demonstration that ABCA1 and SCD expression decreases A␤ formation and that these effects are observed without lipid efflux from cells and involve a decrease in ␥-secretase cleavage.
In contrast to the more general approach of cholesterol bio-synthesis inhibition, LXR activation may directly regulate genes that favorably modulate plasma membrane composition and structure in the brain (8). Tangier disease patients have not been reported with premature dementia, suggesting that ABCA1 may not have an essential role in protecting against AD. However, this does not rule out the possibility that increased expression of ABCA1 has a protective role, just as it does in atherosclerosis (40).