|Home | About | Journals | Submit | Contact Us | Français|
It has been suggested that cellular cholesterol levels can modulate the metabolism of the amyloid precursor protein (APP) but the underlying mechanism remains controversial. In the current study, we investigate in detail the relationship between cholesterol reduction, APP processing and γ-secretase function in cell culture studies. We found that mild membrane cholesterol reduction led to a decrease in Aβ40 and Aβ42 in different cell types. We did not detect changes in APP intracellular domain or Notch intracellular domain generation. Western blot analyses showed a cholesterol-dependent decrease in the APP C-terminal fragments and cell surface APP. Finally, we applied a fluorescence resonance energy transfer (FRET)-based technique to study APP-Presenilin 1 (PS1) interactions and lipid rafts in intact cells. Our data indicate that cholesterol depletion reduces association of APP into lipid rafts and disrupts APP-PS1 interaction. Taken together, our results suggest that mild membrane cholesterol reduction impacts the cleavage of APP upstream of γ-secretase and appears to be mediated by changes in APP trafficking and partitioning into lipid rafts.
Amyloid-β protein (Aβ) is a 4-kDa peptide believed to play a central role in the pathogenesis of Alzheimer’s disease (Hardy and Higgins 1992). Aβ arises from cleavage of the β-amyloid precursor protein (APP), encoded by a gene on human chromosome 21. APP undergoes a proteolytic event by α-secretase within the Aβ region creating a large secreted ectodomain (α-APPs) and a shorter α-cleaved membrane-retained carboxyl-terminal fragment (α-CTF). The resultant 10-kDa α-CTF is cleaved by a presenilin (PS)-dependent γ-secretase to generate a small fragment called p3 (non-amyloidogenic pathway). In an analogous fashion, other APP holoproteins can be cleaved by another protease, β-site APP cleaving enzyme (BACE), generating a 12-kDa CTF (β-CTF) that is cleaved by the same γ-secretase to produce Aβ. In addition to releasing Aβ or p3, the same γ-secretase activity generates an APP intracellular fragment (AICD) that may be transcriptionally active (Cao and Sudhof 2004). Although the generation of the different isoforms of Aβ and p3 has been intensively studied, the normal biological function of APP proteolysis remains unclear.
The γ-secretase complex is a multiprotein complex composed of at least four proteins, presenilin (presenilin 1 or presenilin 2, PS1 and PS2, respectively), which is believed to contain the catalytic site, nicastrin, Pen-2, and Aph1 (Wolfe and Kopan 2004). PS1 is a 467 aminoacid nine-transmembrane domain (TM) protein that requires endoproteolysis to generate a functional heterodimer in which the C- and N-terminal fragments remain non-covalently associated (Selkoe and Wolfe 2007). This multiprotein complex is responsible for the cleavage of at least 30 different substrates, mostly type-I membrane proteins including APP and the Notch receptor among others (Lleo 2008). Notch receptor is a type I transmembrane receptor that is critically required for a variety of signaling events and cell fate decisions during embryogenesis and in adulthood (Artavanis-Tsakonas and Simpson 1991; Levitan and Greenwald 1995). Full-length Notch is cleaved in the presence of its biological ligand Delta, and the generated Notch intracellular domain (NICD) rapidly translocates to the nucleus where it acts as a transcriptional coactivator (Schroeter et al. 1998; Jack et al. 2001). Similar to APP proteolysis, Notch proteolysis is dependent on PS1 (Schroeter et al. 1998).
Because γ-secretase is responsible for the last step in Aβ generation, understanding how to modulate its activity is of considerable therapeutic interest. Several compounds (γ-secretase inhibitors) have been developed to decrease Aβ production by inhibiting the activity of this multiprotein complex. However, general γ-secretase inhibitors impair the cleavage of other substrates in addition to APP, giving rise to concerns about tolerability. Another therapeutic approach that has raised interest is γ-secretase modulation. For example, certain nonsteroidal anti-inflammatory agents have been observed to alter the site of γ-secretase cleavage in APP, rather than inhibiting the enzymatic activity (Weggen et al. 2001; Lleo et al. 2004). We have suggested that the mechanism of action of modulatory agents is allosteric modulation of γ-secretase (Lleo et al. 2004).
Another pharmacological intervention investigated in AD has been the use of cholesterol-lowering agents. There is growing evidence that links AD and cholesterol metabolism. Epidemiological studies have shown a decreased incidence and prevalence of AD among individuals treated with statins, widely used drugs that reduce cholesterol by inhibiting 3-hydroxyl-3-methylglutaryl coenzyme A (HMG CoA) reductase (Jick et al. 2000; Wolozin et al. 2000; Rockwood et al. 2002; Yaffe et al. 2002; Zamrini et al. 2004). Clinical trials with atorvastatin and simvastatin in AD have shown some promising results (Simons et al. 2002; Sparks et al. 2005), and larger clinical trials are currently ongoing (www.clinicaltrials.gov). Another line of evidence comes from studies showing that changes in cholesterol homeostasis affect APP processing. Most studies have shown that strongly reducing cholesterol levels with statins and/or cyclodextran causes a marked reduction in Aβ levels in vitro and in vivo (Simons et al. 1998; Fassbender et al. 2001; Refolo et al. 2001; Ehehalt et al. 2003; Ostrowski et al. 2007). However, it remains controversial whether milder cholesterol depletion similar to that observed in humans taking therapeutic levels of statins (Sparks et al. 2005) lowers Aβ production and the mechanism by which this occurs. Some studies have suggested that mild cholesterol reduction may actually enhance Aβ generation by facilitating the interaction between APP and BACE 1 (Abad-Rodriguez et al. 2004). On the other hand, the observation that Aβ generation depends on lipid rafts and that β-cyclodextrins, which rapidly extract cholesterol directly from the plasma membrane, are able to inhibit Aβ production raises the possibility that cholesterol reduction may alter APP processing by disrupting lipid rafts structure (Simons et al. 1998; Wahrle et al. 2002).
Since γ-secretase consists of multiple transmembrane proteins, and is preferentially distributed in association with lipid rafts, we explored the mechanism by which manipulation of the membrane lipid environment might impact APP processing. In this study, we show that mild cholesterol depletion led to a reduction in secreted Aβ, APP CTFs and cell surface APP, but preserved AICD generation and the γ-secretase-dependent cleavage of Notch. We also found that membrane cholesterol depletion reduced the association of APP with lipid rafts at the cell membrane in intact cells by using a fluorescence resonance energy transfer (FRET)-based microscopy approach.
We used the following cell lines: naive Chinese hamster ovary (CHO) cells, CHO cells stably overexpressing wild-type human PS1 and wild-type APP (PS70, a generous gift from Dr. Selkoe, Brigham and Women’s Hospital, Boston), human neuroglioma cell line (H4) stably expressing the double Swedish APP mutation (a generous gift from Bruno Imbimbo, Chiesi Farmaceutici, Parma) or human embryonic kidney (HEK) cells stably expressing the double Swedish APP mutation. Biotin acceptor peptide (BAP)-APP construct was used for cell surface biotinylation. The construct contains a BAP on the N-terminus of APP695 and a hemagglutinin (HA) tag on the C-terminus. Cells were cultured in DMEM with 10% fetal bovine serum at 37°C with 5% CO2 in a tissue culture incubator. For FRET (fluorescence resonance energy transfer)/FLIM (fluorescence lifetime imaging microscopy) and confocal microscopy, we used PS70 cells, H4 cells stably expressing the double Swedish APP mutation or CHO cells transiently transfected with wild-type APP695, or wild-type PS1 (a generous gift from Carlos Saura, Autonomous University of Barcelona, Barcelona) using Lipofectamine 2000 (Invitrogen) according to manufacturer’s instructions.
Different doses of lovastatin (Calbiochem), or methyl-β-cyclodextrin (MβCDX, Sigma-Aldrich) were used for different time periods to induce membrane cholesterol depletion. Cell toxicity was analyzed for all drugs by measuring adenylate kinase levels using Toxi-light reagent (Cambrex). The γ-secretase inhibitor DAPT (1μM, Sigma) was used as a control. All conditions were supplemented with mevalonate (0.25 mM, Sigma-Aldrich) to support the isoprenoid pathway. Total cell cholesterol levels were measured using the Amplex Red Cholesterol Assay Kit (Invitrogen). We tested lovastatin at different concentrations and for different time periods. We found that treatment for 48 hours with either 20μM lovastatin with 0.25 mM mevalonate and delipidated fetal bovine serum (DLFBS) was able to induce mild but consistent (25-30%) cell membrane cholesterol reduction without signs of cytotoxicity (Suppl. Fig. 1). We also examined the effects of methyl-β-cyclodextrin (MβCDX), which selectively extracts cholesterol from the plasma membrane in preference to other lipids. Consistent cholesterol reductions were only obtained after treating cells with 5mM MβCDX for 10 or 60 min as treatment for 48 hours induced significant cytotoxicity (Suppl. Fig. 1). Therefore, we used for all experiments 20μM lovastatin or 5mM MβCDX for 10 or 60 min which induced a consistent cholesterol reduction.
For human Aβ1-40 ELISA we used 6E10 (against Aβ1-17, Chemicon) as a capture antibody and a rabbit polyclonal Aβ1-40 (Chemicon) as a detection antibody. After incubation for 3 hours, wells were washed and a horseradish peroxidase (HRP)-conjugated Donkey anti-rabbit (Jackson Laboratories) was added. Wells were washed with PBS, Quantablue reagent (Pierce) was added and samples were read at 320 nm using a Victor3 Wallac plate reader (Perkin Elmer, Massachusetts). To measure human Aβ1-42 and Aβ1-x we used sensitive ELISA kits (Wako, Japan and IBL, Germany, respectively).
For the Western Blot analysis of Notch cleavage, CHO cells were transfected with the truncated Notch1 construct NotchΔECmyc (a generous gift from R. Kopan, Washington University, St Louis) and treated with lovastatin, DAPT or MβCDX for 48, 24 hours or 10 minutes, respectively. The cellular lysate was electrophoresed in 10-20% Tris-glycine gels and transferred to a nitrocellulose membrane. The immunoblotting was performed with a mouse 9E10 anti-myc antibody (Chemicon).
For the Western blot analysis of the APP C-terminal fragments (CTFs) we isolated cellular membranes, as described (Steiner et al. 1998), from PS70 cells treated with statins or MβCDX. Membrane preparations were electrophoresed in 5-16% Tris-Tricine gels, transferred to 0.2 μm nitrocellulose membranes, and detected by immunoblotting with a rabbit anti-APP C-terminal (Sigma) antibody. Incubation with primary antibodies was followed by detection with IR-fluorescent conjugated antibody (LI-COR Biosciences). The bands were quantitated using Odyssey software (LI-COR Biosciences), and the values normalized to APP full-length expression.
HEK cells were transfected with the BAP-APP-HA construct using Fugene 6 reagent (Roche) and the cells were treated with lovastatin, 5 or 10 mM MβCDX in serum-free medium for 60 min (48 hours for lovastatin) at 37°C. Cell surface was biotinylated in PBS with 0.3 μM BirA and 10μM biotin-AMP for 40 min at 30 °C (Chen et al. 2005). The remaining biotin was removed by washing 3 times with PBS. The cells were lysed in 70 μl PBS containing 1% Triton X-100, 0.1% SDS and protease inhibitors. Insoluble debris were removed by centrifugation (20000g for 30 min at 4°C). The lysate was analyzed for APP expression by Western Blotting. The membrane was probed with streptavidin(SA)-HRP for cell surface APP and normalized to HA-HRP (total APP).
AICD was generated in vitro from membrane preparations of PS70 cells as described (Sastre et al. 2001). Cells lysates were treated with vehicle, different concentrations of MβCDX or DAPT and incubated at 37 °C for 2 hours. As a negative control we incubated cells on ice. APP C-terminal fragments were detected as described above.
The CBF1-luciferase assay was performed as described previously (Berezovska et al. 2000). Briefly, CHO cells were transfected with a CBF1 luciferase reporter plasmid (a generous gift from D. Hayward) and β-galactosidase as an internal control for transfection efficiency. This assay detects activation of CBF1, a Notch1 downstream transcription factor, as a measure of Notch signaling. The treatment with lovastatin or MβCDX was begun 6 hours after transfection. Luciferase activity was measured 48 hours after treatment using a Victor3 Wallac plate reader (Perkin Elmer), and results were normalized to β-galactosidase expression levels.
For the analysis of APP-PS1 interaction, PS70 cells were double immunostained with antibodies against PS1 loop (amino acids 275-367, Chemicon) and against the C-terminal fragment of APP (amino acids 643-695, Chemicon). Pairs of primary antibodies were labeled with Alexa 488 or cyanine 3 (Cy3)-conjugated secondary antibodies.
For the detection of lipid rafts, living H4 cells were treated with MβCDX or a vehicle control, then immunostained with red-fluorescent Alexa 555 conjugate of cholera toxin subunit B (CT-B, Invitrogen) or fixed and immunostained with a Alexa 555-labeled flotillin antibody (BD Biosciences). For the detection of cell surface APP, cells were then immunostained without permeabilization with an APP antibody (Sigma-Aldrich). CT-B binds to the pentasaccharide chain of plasma membrane ganglioside GM1, which selectively partitions into and is one of the most widely used markers for lipid rafts (Sandvig and van Deurs 2002).
Confocal microscopy was performed using a Leica inverted fluorescent confocal microscope (Institut Ciencies Cardiovasculars de Catalunya, Leica TCD SP2-AOBS, Wetzlar, Germany). This microscope is equipped with a 405 diode pulsed laser, a PMC-100 detector (Leica, Germany) and a time-correlated single photon counting module (SPC730) to perform FRET/FLIM. The hardware/software package allows the measurement of fluorescence lifetimes on a pixel-by-pixel basis. Values were fitted to two exponential decay curves to represent a “non-FRETing” population with a longer lifetime (t2) and a “FRETing” population with a shorter lifetime (t1). FLIM has been described as a novel technique for the analysis of protein proximity (Berezovska et al. 2003; Lleo et al. 2004). The technique is based on the observation that fluorescence lifetimes of a donor fluorophore shorten in the presence of a FRET acceptor in close proximity (<10 nm). The decrease in lifetime is proportional to the distance between the fluorophores at R6.
For APP-PS1 FRET/FLIM experiments cells were fixed and double-immunostained for APP and PS1 as described previously (Berezovska et al. 2003; Lleo et al. 2004). We also applied this technique to detect lipid rafts and partition of APP into rafts by double staining H4 cells with Alexa 555-CT-B conjugate or a flotillin antibody and an APP antibody (Sigma-Aldrich). As a positive control, cells were immunostained against flotillin and CT-B, labeled with Alexa 555 and Alexa 488 secondary antibodies respectively, or with equimolar concentrations of Alexa488-CT-B and Alexa555-CT-B. All samples were compared to a negative control in which the donor fluorophore (Alexa 488) fluorescence lifetime was measured in the absence of the acceptor (no FRET ~2500 ps). As positive controls, we included two additional conditions. First, Alexa 488 lifetime was measured in the presence of a FRET acceptor (Cy3) in close proximity (Cy3-labeled antibody against Alexa488). Under these conditions we observed that Alexa 488 lifetime was shortened to ~1000 ps. Second, we performed photobleaching of the acceptor fluorophore and observed that the FRET signal was completely abolished (Suppl. Fig. 2).
One-way ANOVA was performed to analyze differences in lifetime or Aβ levels followed by least significant difference post hoc analysis. Levene’s test was also performed to determine whether variances were equal.
We developed a paradigm to induce mild cholesterol depletion by treating cells with lovastatin in the presence of low doses of mevalonate and DLFBS. Supplementation of statin-treated cells with 0.25 mM mevalonate is required to rescue the normal isoprenoid levels while blocking cholesterol biosynthesis (Brown and Goldstein 1980; Goldstein and Brown 1990; Keller and Simons 1998; Simons et al. 1998; Fassbender et al. 2001; Kojro et al. 2001; Meske et al. 2003; Cole et al. 2005). We found that treatment of CHO or HEK cells with 20 μM lovastatin in the presence of 3% DLFBS for 48 hours induced a consistent total cholesterol depletion (~30%) without signs of cytotoxicity (Suppl. Fig. 1). Treatment with statins did not induce cholesterol reduction in the presence of non-delipidated FBS. We also measured the total cholesterol levels after treatment with MβCDX, which selectively extracts cholesterol from the plasma membrane in preference to other lipids. Consistent cholesterol reductions were only obtained after treating cells with 5 mM MβCDX for 10 or 60 min. Treatment with MβCDX for longer periods of time affected cell viability (Suppl. Fig. 1).
We next investigated the effects of cholesterol depletion on APP processing. We treated APP/PS1 overexpressing CHO (PS70) or SweAPP overexpressing HEK cells for 48 hours in the presence or absence of lovastatin and DLFBS and levels of Aβ40 and Aβ42 were measured in the conditioned media by sandwich ELISA. After 48 hours of treatment both Aβ40 and Aβ42 levels in the conditioned media were significantly reduced compared to that in the vehicle control treated cells (~50 %, p<0.05, Fig. 1A). The reduction was observed in both PS70 and HEK cells confirming that the effect was not cell-type specific. To ensure that this effect was due to a decrease in cellular cholesterol levels and not by any other pleiotropic action of statins, we treated cells with lovastatin in the presence of 3 % FBS. The addition of cholesterol-containing serum was sufficient to reverse the reduction in Aβ levels (Fig. 1A) indicating that the observed effects on APP processing were specifically due to cholesterol depletion. As a positive control, we treated cells for 24 hours with 1μM DAPT, a commonly used γ-secretase inhibitor, and found that it reduced both Aβ40 and Aβ42 levels by ~95% consistent with a complete inhibition of γ-secretase (p<0.001).
We next examined whether cholesterol reduction would lead to changes in the levels of full-length APP or APP CTFs, the direct substrates of γ-secretase. We isolated cell membranes from PS70 cells treated with lovastatin/DLFBS or MβCDX, and the lysate was subjected to Western blot analysis. We found that cholesterol depletion did not alter total APP levels, but reduced the levels of both α- and β-CTF by ~30% compared to a vehicle control (Fig. 1B). Addition of serum to cells was able to restore the CTF levels confirming that the effect was cholesterol-dependent. As expected, DAPT led to a strong accumulation of APP CTFs. These data indicate that cholesterol reduction lowers Aβ, affects APP processing and APP CTF generation.
The reduction in APP CTFs observed after cholesterol reduction could reflect a possible effect on APP trafficking or an inhibition of α- and β-cleavages. To distinguish between these two possibilities we performed experiments to analyze APP trafficking by using a BAP-APP construct in HEK cells. We found that treatment with lovastatin or MβCDX in cells transfected with BAP-APP led to a reduced cell surface APP without changes in total APP levels (Fig. 1C). Combination of lovastatin and MβCDX showed an additive effect. Overall, our results suggest that cholesterol depletion reduces APP CTFs by altering APP trafficking and reducing substrate availability.
This assay measures in vitro generated AICD with results from a preexisting β-CTF (Sastre et al. 2001). We incubated cell membrane preparations with different concentrations of MβCDX for 2 hours at 37 °C and analyzed the levels of AICD by Western blot. As a control, membranes were incubated at 4 °C or with DAPT and minimal amounts of AICD are detected consistent with γ-secretase inhibition (Fig. 1D). Treatment of cell membranes with different concentrations of MβCDX, which was able to reduce total cholesterol in a dose-dependent manner up to 70%, had no effect on AICD generation. Because this assay measures de novo AICD from preexisting CTFs (Sastre et al. 2001), we did not observe any effects on the levels of CTFs. We also measured in parallel the levels of total Aβ in this assay and results were normalized to levels of APP CTFs. We found that cholesterol reductions up to 70% did not reduce total Aβ (Fig. 1E).
We next examined whether cholesterol depletion altered the processing of other γ-secretase substrates. The γ-secretase cleavage of the Notch receptor was monitored by using a luciferase assay that reflects Notch signaling as previously described (Hsieh et al. 1996; Lleo et al. 2003). CHO cells were transfected with a CBF1 luciferase reporter construct or an empty vector and β-galactosidase as an internal control for the transfection efficiency. Cells were treated for 48 hours with lovastatin, MβCDX or DAPT, and CBF1 luciferase activity was detected in the lysates. The results were normalized to β-galactosidase expression levels. There was no difference in the CBF1 luciferase activity after cholesterol depletion compared to that in vehicle treated control cells (Fig. 2A). As expected, a marked reduction in CBF1 activity was observed after treatment with DAPT (p<0.05).
To confirm these results, we analyzed the levels of NICD, the γ-cleaved product of Notch, by Western blot. CHO cells were transfected with a constitutively active form of Notch (NΔEC) which undergoes cleavage and generation of the NICD domain. NΔEC-transfected cells were treated with lovastatin, DAPT, MβCDX or a vehicle control. After 48 hours the cells were harvested and lysates were subjected to Western blot analysis (Fig. 2B upper panel). After quantification, we did not observe any differences in the generation of NICD after cholesterol depletion (Fig. 2B lower panel). As expected, treatment with DAPT led to a marked reduction in the generation of NICD with accumulation of the NΔEC fragment, consistent with inhibition of γ-secretase. Taken together, these results indicate that cholesterol depletion does not impair - either Notch signaling or γ-secretase dependent Notch S3 cleavage.
Since lipid rafts are one of the main sites where APP amyloidogenic processing takes place (Lee et al. 1998), we next explored whether cholesterol removal had any impact on raft-associated APP. We developed a novel FRET-based assay (FLIM) to measure association of APP into lipid rafts by staining cells flotillin or CT-B-Alexa555. This FRET assay is based on the principle that when the two fluorophores are in close proximity (<10 nm), the measured lifetime of the donor fluorophore (Alexa 488-APP) is shortened in proportion to the distance between the fluorophores. For these experiments we tested CHO, HEK or H4 cells and found that only the latter ones had sufficient lipid raft staining to perform FRET experiments. We observed that only a small percentage of APP colocalized with rafts on the cell surface (Fig. 3). However, this small amount was enough to shorten Alexa488 lifetime by ~25 % in the presence of the acceptor probe (flotillin or Alexa555-CTB) indicating the presence of FRET (Table 1). As a positive control, we stained cells with CT-B and flotillin or equimolar concentrations of Alexa488-CT-B and Alexa555-CT-B and observed wide colocalization as well as a further 25% reduction in fluorescence lifetime (Table 1). Therefore, this assay reflected the association of APP into lipid rafts and could be applied to measure changes under cholesterol reducing conditions. We observed that treatment with MβCDX increased Alexa488 lifetime (15%) compared to vehicle-treated cells (Fig. 3, Table 1). We interpret these data as a reduced partition of APP into lipid rafts.
We postulated that the observed effects on APP processing and trafficking under mild cholesterol-lowering conditions might have altered γ-secretase-APP interactions. To confirm this possibility we used a FRET-based (FLIM) assay to detect APP CTFs-PS1 interactions in intact cells (Berezovska et al. 2003; Berezovska et al. 2005). We measured the proximity of the loop region of PS1, which is adjacent to the putative catalytic site of γ-secretase, to the C-terminus of APP (Berezovska et al. 2003; Lleo et al. 2004; Berezovska et al. 2005). We checked that treatment with lovastatin or MβCDX did not alter APP or PS1 cellular distribution in PS70 cells as assessed by confocal microscopy. Interestingly, PS1 partially colocalized with lipid rafts at the cell surface but colocalization was not affected by cholesterol depletion (Suppl. Fig. 3). PS70 cells were treated with 20 μM lovastatin for 48 hours or 5 mM MβCDX for 10 min. The donor fluorophore (Alexa 488-PS1) had a lifetime of ~2500 ps in the absence of a FRET acceptor. When the acceptor (Cy3-labeled APP epitope) is in close proximity to the donor, the lifetime is shortened by 40 % (~1600 ps). By contrast, treatment with lovastatin or MβCDX diminished this effect and increased Alexa 488 lifetime by 11.5% and 38.5% respectively compared to the baseline conditions (Table 2, Fig. 4). We interpret these data as reflecting a reduced interaction between APP CTFs and PS1 in cells with depleted cholesterol.
The main focus of this study was to determine whether mild membrane cholesterol depletion affects APP processing and the mechanism by which this occurs. This is a relevant question because cholesterol-lowering agents are being investigated as a possible treatment for AD. While strong reduction of the cholesterol levels has been shown to reduce Aβ production (Simons et al. 1998; Fassbender et al. 2001; Kojro et al. 2001; Refolo et al. 2001; Ehehalt et al. 2003; Xiong et al. 2008), the effect of mild cholesterol depletion on Aβ and the mechanism by which this occurs is controversial. Here we find that treatment of different cell lines with lovastatin reduced Aβ40 and Aβ42 in a cholesterol-dependent manner. The reduction of Aβ was observed in cells supplemented with mevalonate, and was reversed by addition of cholesterol-containing serum (Cole et al. 2005; Cordle and Landreth 2005; Ostrowski et al. 2007). This indicates that the observed effects are directly related to cholesterol depletion and not to any other pleiotriopic effect of statins. Although statins have also been shown to lower Aβ by inhibiting protein isoprenylation (Ostrowski et al. 2007), our data suggest that statins can also display the same effect by reducing membrane cholesterol.
We next investigated whether γ-secretase is sensitive to membrane cholesterol reduction. Experiments to address this issue have yielded conflicting results so far. Some authors have shown that γ-secretase depends on rafts but is not cholesterol-dependent (Wada et al. 2003), while others have found that indeed the enzyme can be modulated by cholesterol (Wahrle et al. 2002; Xiong et al. 2008). Our data support the notion that the neither the ε-cleavage (that generates AICD) not the overall γ-secretase cleavage are sensitive to mild cholesterol depletion. Mild cholesterol reduction did not impair the generation of AICD or the cleavage of the truncated form of Notch, NΔEC, which is the direct substrate of γ-secretase. We cannot exclude that mild cholesterol depletion had an effect of specific γ-secretase cleavages, since we did not measure all Aβ species. High cholesterol reductions are difficult to assess in cell culture models because cholesterol depletion may affect cell viability. However, in a cell-free assay of γ-secretase (Sastre et al. 2001) moderate reductions (<70%) in cholesterol content were not able to reduce total Aβ production. A recent report (Ostrowski et al. 2007) describes a direct and potent influence of cholesterol on γ-secretase activity. However, the authors used a purified mammalian γ-secretase system that allows drastic manipulations of the lipid environment that are not achievable in living cells and therefore their results are not directly comparable to ours. We also found that cholesterol depletion reduced both APP α- and β-CTF generation, indicating a possible effect on APP trafficking or an inhibition of α- and β-cleavages. To distinguish between these 2 possibilities we performed experiments to study APP trafficking by using a BAP-APP construct. Our results indicate that cholesterol depletion alters APP trafficking and reduces cell surface APP. Although MβCDX is known to interfere with endocytosis, the same effects were also observed in cells treated with lovastatin alone suggesting that the effects are due to cholesterol depletion. Furthermore, the fact that mild cholesterol depletion lowers Aβ in cell culture models but not in cell-free assays supports the notion that the effect takes place upstream of γ-secretase cleavage. Overall, our results suggest that the reduction in APP CTFs could be due to reduced substrate availability.
Our work also indicates that neither the γ-secretase-dependent cleavage of Notch nor Notch signaling is affected by cholesterol depletion. This is supported by other studies that show that Notch CTFs are predominantly found in non-raft membrane domains while APP CTFs reside in lipid rafts (Vetrivel et al. 2005). This has implications for cholesterol-lowering strategies in AD, since these compounds may target Aβ production without interfering with Notch function.
Several lines of evidence suggest lipid rafts as the principal sites in cellular membranes where Aβ is generated (Lee et al. 1998; Cordy et al. 2003; Ehehalt et al. 2003; Wada et al. 2003; Vetrivel et al. 2005; Hur et al. 2008). Lipid rafts are highly dynamic sphingolipid- and cholesterol-rich membrane microdomains with important roles in cellular signaling and trafficking (Allen et al. 2007). Biochemically, rafts are characterized by their insolubility in non-ionic detergents such as Triton X-100 (Brown and Rose 1992). By this method, some proteins relevant to Aβ production, such as APP, BACE and PS1, have been shown to be present in rafts prompting the hypothesis that amyloidogenic processing of APP take places in rafts (Parkin et al. 1999; Ehehalt et al. 2002; Cordy et al. 2003). However, the use of detergents to solubilize cell membranes in these studies has the potential to introduce significant artifacts (Munro 2003). To avoid this limitation, here we used FLIM, a FRET-based technique, to assess lipid rafts in intact cells. Compared to conventional FRET experiments, FLIM has the advantage that it does not depend on the fluorophore concentration, is not destructive, and is not sensitive to miss-excitation phenomenon. We used a FLIM assay to detect FRET between Alexa488-APP (donor) and Alexa555-CTB or Alexa555-flotillin (acceptor), which are known rafts markers (Sandvig and van Deurs 2002). We found that cholesterol depletion did result in increased fluorescence lifetime, suggesting that APP is less associated to rafts under these conditions. These results combined with our biochemical data showing reduced trafficking of APP to the cell surface, and reduced APP CTFs-PS1 interaction assessed by FLIM under cholesterol-lowering conditions, would suggest that less APP CTF is available for γ-secretase, therefore decreasing both Aβ40 and Aβ42.
In summary, we report that mild cholesterol depletion impairs APP processing but without affecting Notch cleavage or the APP ε- or γ-secretase cleavage. The effects of cholesterol reduction are observed upstream of γ-secretase by altering APP trafficking, reducing APP CTF generation and raft-associated APP.
Membrane cholesterol reduction. We treated cells with different concentrations of lovastatin or MβCDX for different time periods. We found that treatment for 48 hours with 20μM lovastatin, 0.25 mM mevalonate and delipidated fetal bovine serum (DLFBS) was able to induce mild but consistent (~30%) cell membrane cholesterol reduction without signs of cytotoxicity. In the case of MβCDX, consistent cholesterol reductions were only obtained after treating cells with 5 or 10 mM MβCDX for 10 or 60 min as treatment for 48 hours induced significant cytotoxicity. * One way ANOVA, p< 0.05.
FRET-based assay (FLIM) detects APP-PS1 interactions in intact cells. PS70 cells were immunostained with PS1-loop and APP-C terminal antibodies labeled with Alexa488- and Cy3-donor and acceptor fluorophores, respectively. As described, these two domains remain in close proximity mainly near the cell surface, represented as red pixels in the pseudocolored image (Berezovska et al. 2003; Lleo et al. 2004). Photobleaching of the FRET acceptor in part of the cell (square) leads to an almost complete loss of the FRET signal, reflected as blue pixels in the pseudocolored lifetime image.
Cholesterol depletion does not impair PS1 subcellular distribution. H4 cells were stained with a PS1 antibody and Alexa555-CT-B. As shown, endogenous PS1 partially colocalized with lipid rafts at the cell surface. Colocalization was not affected by mild cholesterol depletion.
This work was supported by a grant from the Fondo de Investigación Sanitaria (FIS04/1893) to A.L., NIH AG026593 to O.B. and NIH AG15379 to B.T.H. We would like to thank Esther Peña for technical assistance.