|Home | About | Journals | Submit | Contact Us | Français|
Author contributions: R.B. and D.M.K. designed research; R.B. and C.B. performed research; R.B., C.B., and D.M.K. analyzed data; R.B. and D.M.K. wrote the paper.
Brains of patients affected by Alzheimer's disease (AD) contain large deposits of aggregated amyloid β-protein (Aβ). Only a small fraction of the amyloid precursor protein (APP) gives rise to Aβ. Here, we report that ~10% of APP undergoes a post-translational lipid modification called palmitoylation. We identified the palmitoylation sites in APP at Cys186 and Cys187. Surprisingly, point mutations introduced into these cysteines caused nearly complete ER retention of APP. Thus, either APP palmitoylation or disulfide bridges involving these Cys residues appear to be required for ER exit of APP. In later compartments, palmitoylated APP (palAPP) was specifically enriched in lipid rafts. In vitro BACE1 cleavage assays using cell or mouse brain lipid rafts showed that APP palmitoylation enhanced BACE1-mediated processing of APP. Interestingly, we detected an age-dependent increase in endogenous mouse brain palAPP levels. Overexpression of selected DHHC palmitoyl acyltransferases increased palmitoylation of APP and doubled Aβ production, while two palmitoylation inhibitors reduced palAPP levels and APP processing. We have found previously that acyl-coenzyme A:cholesterol acyltransferase (ACAT) inhibition led to impaired APP processing. Here we demonstrate that pharmacological inhibition or genetic inactivation of ACAT decrease lipid raft palAPP levels by up to 76%, likely resulting in impaired APP processing. Together, our results indicate that APP palmitoylation enhances amyloidogenic processing by targeting APP to lipid rafts and enhancing its BACE1-mediated cleavage. Thus, inhibition of palAPP formation by ACAT or specific palmitoylation inhibitors would appear to be a valid strategy for prevention and/or treatment of AD.
Deposition of the amyloid (Aβ) peptide in senile plaques is a hallmark of Alzheimer's disease (AD) pathology (for review, see Tanzi et al., 2004; Walsh and Selkoe, 2004; Haass and Selkoe, 2007). Aβ derives from the amyloid precursor protein APP, which is a type 1 transmembrane protein. APP undergoes sequential proteolysis by three types of proteases, α-, β- and γ-secretases. Proteolysis of APP by α- and γ-secretases releases soluble nonamyloidogenic p3 peptides, while β- and γ-secretase cleavages generate amyloidogenic Aβ peptides of various length ranging from 30 to 42 aa (Kojro and Fahrenholz, 2005; Steiner et al., 2008). Aβ42 peptides in particular form the most pathogenic oligomers and aggregates (Iwatsubo et al., 1994).
Although the proteolytic events resulting in Aβ production are well characterized, it is not clear why only a small portion of APP undergoes amyloidogenic processing (Haass et al., 2012). Shortly after synthesis in the ER, APP traffics to the Golgi and eventually to the plasma membrane. Approximately 10% of total APP is cleaved by α-secretase at the plasma membrane to generate a soluble ectodomain APP (sAPP), sAPPα, and a truncated C-terminal fragment (α-CTF), while the majority of the plasma membrane bound APP is rapidly endocytosed (Kojro and Fahrenholz, 2005; Thinakaran and Koo, 2008). Alternatively, APP can be processed by BACE1 to generate sAPPβ and β-CTF in the trans-Golgi, plasma membrane, and early endosomes (Selkoe, 1998; De Strooper and Annaert, 2000). The β-CTF is further cleaved in the endocytic recycling compartment or late endosomes by the γ-secretase complex to release Aβ (Haass et al., 1992). Interestingly, APP, BACE1 and γ-secretase components are all found in membrane microdomains known as detergent-resistant lipid rafts (Vetrivel et al., 2004, 2005; Hattori et al., 2006).
Protein palmitoylation consists of a thioester linkage between a 16 carbon palmitic acid and a cysteine residue, often resulting in lipid raft localization of the protein (Resh, 2004; Brown, 2006). The primary function of protein palmitoylation is to enhance hydrophobicity of proteins and thereby facilitate association with membrane components of the cell and hydrophobic domains of other proteins (Charollais and Van Der Goot, 2009). The proper activity of oncoproteins such as H-ras is dependent on their state of palmitoylation resulting in lipid raft localization (Rocks et al., 2005). In the brain, protein palmitoylation is the most abundant lipid modification among neuronal proteins (Fukata and Fukata, 2010). BACE1 and two components of γ-secretase, Aph-1 and nicastrin (Nct), are targeted to lipid rafts by post-translational palmitoylation of C-terminal cysteine residue(s) near their transmembrane domains (Cheng et al., 2009; Vetrivel et al., 2009). Although palmitoylation-deficient mutants of BACE1, Aph-1, and Nct do not exhibit altered APP processing in vitro (Cheng et al., 2009; Vetrivel et al., 2009), lack of Aph-1 or Nct palmitoylation may affect γ-secretase activity in vivo (Meckler et al., 2010).
Here, we report for the first time that APP undergoes palmitoylation in its N-terminal E1 lumenal domain. Mutations introduced into APP's palmitoylated cysteines result in ER retention of the protein. Palmitoylated APP is preferentially targeted to lipid raft domains where it serves as a good BACE1 substrate in cells and in vivo. Palmitoylation inhibitors severely impair the processing of APP by α- and β-secretases. Acyl-coenzyme A:cholesterol acyltransferase (ACAT) inhibitors, which redistribute cellular cholesterol, also inhibit APP palmitoylation and reduce Aβ generation. Thus, APP palmitoylation appears to promote lipid raft-bound Aβ production.
CHOAPP, H4, and naive CHO cells were maintained and transfected with expression plasmids as described previously (Huttunen et al., 2007b, 2009). CHO cells stably expressing APP and myc-epitope-tagged BACE1 (CHOAPP+BACE1) were maintained in DMEM containing 10% serum supplemented with G418 and Zeocine. AC29APP cells that lack ACAT activity were grown as described previously (Puglielli et al., 2001). Typically, 1.2 × 106 cells were used for transfection and palmitoylation assays.
Primary neuronal cultures were prepared as described previously (Puglielli et al., 2001). Briefly, hippocampi and frontal cortices of CD-1 mice (Charles River Laboratories International) were dissected from embryonic day 16 (E16)–E18. The tissues were triturated and plated for overnight on poly-(l-lysine)-coated (0.2 mg/ml) culture dishes at the concentration of 32,000 cells/cm2. Neurons were then maintained in Neurobasal medium with 2% B27 supplement (Invitrogen) at 37°C with 5% CO2 in air for 14 d. Postculture primary neurons were used for BACE1-inhibition and palmitoylation assays.
C-terminally V5-epitope-tagged APP751 (APPwt-V5) were used as described previously (Huttunen et al., 2007b). N-terminal truncation mutants APP(Δ281) and APP(Δ343) were subcloned into pCDNA3.1-V5/His vector. cDNA of APPwt-V5 was used to prepare the cysteine mutants APP(C186,187S/A), APP(C186S), APP(C187S), C186/187E/Q/R/F/P), and APP(C133/158S/A) by using appropriate mutagenesis primer pairs and the QuikChange site-directed mutagenesis kit (Invitrogen). cDNAs encoding HA-epitope-tagged DHHC (Asp-His-His-Cys)-1, DHHC-7, and DHHC-21 (HA-DHHC-1, HA-DHHC-7, and HA-DHHC-21, respectively) were kind gifts from Dr. M. Fukata (National Institute for Physiological Sciences, Okazaki, Japan; Fukata et al., 2004). The following APP antibodies were used: C66 (APP C-terminus), 22C11 (APP N-terminus; Millipore Bioscience Research Reagents), anti-sAPPβ (IBL International), and 6E10 (Signet). Polyclonal antibody against BACE was obtained from Affinity BioReagents. Antibodies against epitope tags were anti-V5 (Invitrogen), anti-myc and anti-HA (Cell Signaling Technology). Antibodies against calreticulin (ER marker), GM130 (Golgi marker), and flotillin (lipid rafts marker) were obtained from Santa Cruz Biotechnology, BD Biosciences, and Abcam, respectively.
Cell lysates were prepared by directly extracting cells in a buffer containing 10 mm Tris-HCl at pH 6.8, 1 mm EDTA, 150 mm NaCl, 0.25% NP-40, 1% Triton X-100, and a protease inhibitor cocktail (Roche), followed by centrifugation at 16,000 × g. For modified acyl–biotinylation exchange assay (mABE assay), the lysis buffer was complemented with 10 mm tris(2-carboxyethyl)phosphine (TCEP; Sigma) and 10 mm N-ethylmaleimide (NEM; Thermo Scientific). Proteins (20–100 μg) were subjected to immunoprecipitation or ABE assay, or simply resolved on 4–12% gradient Bis-Tris gels, 12% Bis-Tris gels, or 16% Tricine gels (Invitrogen), depending on the individual experiment, as described. The blots were visualized by enhanced chemiluminescence. The images were captured by using BioMax film (Kodak) and quantified using QuantityOne software (Bio-Rad).
All experimental procedures were performed in accordance with the U.S. Public Health Service Guide for Care and Use of Laboratory Animals and were approved by the Massachusetts General Hospital Subcommittee on Research Animal Care. Mice used in the study were of either sex. BACE1-null and wild-type (WT) control mice (C57BL/6J background) were purchased from The Jackson Laboratory. BACE1-heterozygous knock-out mice (BACE1−/−) were generated by crossing BACE1-null and wild-type control mice as described previously (Kim et al., 2011). Three-month-old mice were killed, and brains were immediately removed and stored at −80°C until use. Cortices were obtained immediately after the brains were removed and immediately stored at −80°C until use. Membrane extracts from total brains or cortices were prepared as described previously (Kim et al., 2011). For ABE assays, extracts were prepared according to the published method (Wan et al., 2007). To assess changes in palmitoylated APP (palAPP) levels in cortices from young versus older mice, three 3-month-old and three 18-month-old nontransgenic (non-Tg) mice (C57BL/6J background; The Jackson Laboratory) were used. Mouse brain lysates were prepared as above.
Lipid rafts were isolated from 0.5% Lubrol WX (Lubrol 17A17; Serva) lysates of cultured cells by discontinuous flotation density gradients as described previously (Vetrivel et al., 2009). For the isolation of lipid rafts from mouse brain, we essentially followed published methods (Vetrivel et al., 2005). Briefly, mouse brains were homogenized in five volumes of HBS (50 mm HEPES pH 7.4, 0.15 M NaCl, 5 mm EDTA) before addition of 0.5% Lubrol. The extracts were sonicated three times for 30 s using a microsonicator and centrifuged at 2500 × g for 10 min to remove cell debris and nuclei. The supernatant were adjusted to 40% sucrose and separated on a discontinuous sucrose density gradient. Subcellular organelles were separated by subjecting cell lysates to OptiPrep fractionation analysis as done previously (Huttunen et al., 2009).
CHO-APP cells were metabolically labeled with [3H] palmitic acid ([3H]-C16; PerkinElmer) as described previously (Vetrivel et al., 2009). Briefly, cells were incubated with 10 mCi [3H]-C16 for 6 h before lysis in 10 mm Tris-HCl, pH 7.6, 2 mm EDTA, and 150 mm NaCl containing protease inhibitors. APP were immunoprecipitated from the labeled cells by anti-APP (C-terminus) antibody C66. The immunoprecipitates were subjected to fluorography as done previously (Bhattacharyya and Wedegaertner, 2000).
Briefly, cells were metabolically labeled with a chemical palmitic acid probe, alkylene palmitic acid (Alkyl-C16; Invitrogen) as described previously (Charron et al., 2009). Six hours after labeling, cells were subjected to immunofluorescence analysis as described previously (Charron et al., 2009). For Western blots, cells were lysed, and APP was immunoprecipitated from the lysates on agarose beads before reacting with a bioorthologal alkyne-labeled fluorescent chromofore, tetramethylrhodamine (alkyne-TAMRA; Invitrogen) via “click chemistry ” (Charron et al., 2009). The samples were then probed with an anti-TAMRA antibody (Invitrogen).
This assay is based on the substitution of biotin for palmitoyl modifications through a sequence of three chemical steps: unmodified cysteine thiols are blocked with NEM; palmitoylation thioesters are cleaved by hydroxylamine (+NH2OH); and finally, the newly exposed cyateinyl thiols are marked with thiol-specific biotinylating reagent (HPDP-biotin in our experiments). Biotinylated proteins are then affinity purified with streptavidin–agarose beads and probed for the protein of interest (Komekado et al., 2007; Kang et al., 2008). Briefly, cells were lysed with lysis buffer (50 mm Tris-HCl, pH 7.5, 150 mm NaCl, 5 mm EDTA) containing 1% SDS, 2% Triton X-100, 0.5% NP-40, protease inhibitors, 10 mm TCEP (Sigma), and 10 mm NEM (Thermo Scientific). Equal amounts of proteins were precipitated by chloroform-methanol before 1 m NH2OH treatment (untreated samples served as controls), HPDP-biotin addition, and affinity purification with StreptAvidin agarose. The precipitates were either probed with an appropriate antibody or Streptavidin-HRP to detect palmitoylated proteins. In some cases, cells were treated with palmitoylation inhibitors 2-bromopalmitate (2-BP) and cerulenin (Sigma) before ABE assay. In addition, while testing the effect of ACAT inhibition on APP palmitoylation, cells were treated with ACAT inhibitors CP-113,818 (CP) or CI-1011 [avasimibe (AV)], as described previously (Huttunen et al., 2007a, 2010) before ABE assay. CP and AV were generous gifts from Dr. J. Harwood (Pfiser, Groton, CT) and L.-F. Lau (GlaxoSmithKline, Shanghai, China), respectively. For lipid raft and nonraft palmitoylation studies, protein extracts from lipid rafts and nonrafts were subjected to ABE assay according to a method reported previously (Yang et al., 2010). Briefly, lipid raft fractions were extracted with 60 mm β-octyl glucoside (Sigma), and nonraft fractions were extracted with 1% Triton X-100. The protein extracts were precipitated using chloroform/methanol before ABE assay.
This assay is based on a modification of the ABE assay as described previously (Cheng et al., 2009). Briefly, cells were lysed in lysis buffer (150 mm NaCl, 5 mm EDTA, 50 mm Tris-HCl, pH 7.4, 1% Triton X-100, protease inhibitors, 10 mm TCEP, and 10 mm NEM). Aliquots of lysates were incubated with appropriate antibodies to immunoprecipitate APP or sAPP. Immunoprecipitated proteins bound to agarose beads were treated with 1 m NH2OH, pH 7.4, followed by incubation with biotin-HPDP at 4°C for 2 h to label the reactive cysteine(s). A sample prepared in absence of NH2OH served as negative control. The beads were washed and immunoblotted with steptavidin-HRP (Cell Signaling Technology) to detect palmitoylation.
CHO cells expressing APP-V5 were grown on coverslips and metabolically labeled Alkyl-C16 (100 μm in DMSO) or DMSO. Six hours after labeling, cells were subjected to “click-it” assay using alkyne-TAMRA as described previously (Charron et al., 2009). After TAMRA labeling, cells were stained with anti-V5 antibody (Santa Cruz Biotechnology) to detect distribution of APP-V5 (Huttunen et al., 2007b). Confocal images were obtained using an Olympus DSU/IX70 spinning disc confocal microscope equipped with IPLab software (Scanalytics/BD Biosciences) for image processing.
For Aβ determination, cells stably transfected with wild-type or mutant APP751 were grown in six-well plates (Becton Dickinson Labware). When ~80–90% confluent, cells were washed in PBS and incubated in 1 ml of fresh medium for 24 h. Secreted Aβ40 and Aβ42 were quantified by standard sandwich ELISA using the commercially available Aβ ELISA kit (Wako Pure Chemical). Values were normalized by the total cellular protein amount. For the experiments with HA-DHHC-1, HA-DHHC-7 and HA-DHHC-21, CHOAPP cells were transiently transfected with the expression plasmids in six-well plates. Twenty-four hours post-transfected cells were washed with PBS and incubated in 1 ml fresh media for 24 h before Aβ ELISA assay.
CHOAPP cells were pretreated with increasing amounts of the BACE inhibitor IV (BACEi IV; 0–5 μm) for 12 h in presence of 10 μg/ml cyclohexamide (CHX; Calbiochem) for the last 6 h to block protein synthesis. Cells were incubated for another 4 h in absence of CHX but in presence of the BACE inhibitor before ABE assay to detect newly synthesized palAPP.
BACE-cleavage assays in lipid raft fractions were performed by combining methods reported previously by Wada et al. (2003) and Yamakawa et al. (2010). Briefly, lipid rafts were isolated from the NEM-supplemented Lubrol extract of cells or mouse brains. The lipid raft fractions were pooled and used as a source for palAPP for the β-cleavage assay. The substrate was mixed with a 50 mm Na-acetate buffer at pH 4 containing a complete protease inhibitor mixture (Roche Applied Science), the aspartic protease inhibitor pepstatin A (10 μM; Roche Applied Science), and the γ-secretase inhibitor N-[N-(3,5-difluorophenacetyl-l-alanyl)]-(S)-phenylglycine t-butyl ester (0.5 μm; Calbiochem). BACE activity was measured by incubating the mixture at 37°C in absence or presence of increasing concentration of the β-secretase inhibitors BACEi IV (Calbiochem) or dr9 (a kind gift from Dr. J. Tang, Oklahoma Medical Research Foundation, Oklahoma City, OK; Kim et al., 2007). After 1 h incubation, the reaction was terminated by increasing the pH to 7.6. The mixtures were reacted with 10 mm NH2OH to detach palmitate from the released sAPPβ. The samples were centrifuged at 100,000 × g for 1 h to remove membranes, and the supernatant were subjected to immunoprecipitation with anti-sAPP antibody 22C11. The mABE assays were performed on the precipitated samples using biotin-HPDP to label the exposed reactive cysteine(s). The beads were washed after biotinylation and probed with streptavidin-HRP or anti-sAPPβ antibody to detect palmitoylated sAPPβ (pal-sAPPβ) or sAPPβ, respectively.
All statistical analyses used a two-tailed Student's t test or one-way ANOVA, followed by a post hoc Tukey's test. Error bars represented in graphs denote the SEM. Significance was assessed at p < 0.05 and p < 0.01.
To test whether APP undergoes palmitoylation, we used CHO cells stably expressing wild-type APP751 (CHOAPP) in three different palmitoylation assays: metabolic labeling with the fluorescent reactive palmitic acid analog alkylene-palmitic acid (Alkyl-C16), metabolic labeling with [3H]palmitic acid ([3H]-C16), and ABE assay (Fig. 1). To visualize Alkyl-C16 incorporation, we metabolically labeled CHO cells expressing V5-epitope tagged APP (CHOAPP) with Alkyl-C16 and performed fluorescent microscopy analysis of cells after incorporation of fluorescent labeled alkyne-TAMRA by “click chemistry” (Charron et al., 2009). We observed colocalization of APP with TAMRA in Golgi-like puncta, indicating that proteins labeled with the palmitic acid probe Alkyl-C16 colocalized with APP at this site (Fig. 1A). To show incorporation of Alkyl-C16 into APP, CHOAPP cells were first metabolically labeled with (Alkyl-C16) followed by immunoprecipitation of APP using the anti-APP C-terminal antibody C66. The immunoprecipitates were then subjected to click chemistry using alkyne-TAMRA that interacted with Alkyl-C16. An anti-TAMRA antibody detected Alkyl-C16 labeling of APP (Fig. 1B, left). The palmitoylation inhibitor 2-BP reduced APP palmitoylation to a large extent (Fig. 1B, left). Staining for APP confirmed the presence of equal amounts of APP among the three immunoprecipitated samples (Fig. 1B, right). Direct incubation of CHOAPP cells with [3H]palmitic acid ([3H]-C16) followed by immunoprecipitation with the APP antibody C66 showed a single labeled band at ~100 kDa that was absent in samples precipitated with the preimmune serum (Fig. 1C). This band, visible after 2 months of exposure, corresponds to immature APP. Mature ([3H]-C16-labeled) palAPP, less abundant than immature palAPP (Fig. 1B,D–F), was not visible after 2 months of exposure of the film. These experiments indicate that APP is palmitoylated.
Next we used an ABE assay to explore whether APP undergoes S-palmitoylation as opposed to N-palmitoylation (Roth et al., 2006). The assay consists of replacing thioester-linked palmitates attached to cysteine residues with a thiol-specific biotinylating reagent in the presence of NH2OH (Komekado et al., 2007; Kang et al., 2008). Palmitoylated proteins are isolated by affinity purification using NeutrAvidin beads. Both immature and mature APP showed NH2OH-dependent palmitoylation in CHOAPP cells, indicating S-palmitoylation of APP (Fig. 1D). Endogenous flotillin, a known palmitoylated protein, also showed NH2OH-dependent palmitoylation in our cell line (Fig. 1D). Despite weak expression of endogenous BACE1 in CHO cells, we were able to detect endogenous palmitoylated BACE1 in the NH2OH-treated sample (Fig. 1D) (Vetrivel et al., 2009). Quantitation of the ABE assay results revealed that 10.6 ± 1.5% of total APP undergoes palmitoylation in CHOAPP cells.
Finally, we asked whether endogenous APP undergoes palmitoylation in naive cell lines and mouse brain extracts. Our ABE assay detected both immature and mature palAPP in naive CHO cells, H4 human neuroglioma cells, and B104 rat neuroblastoma cells (Fig. 1E). Quantitative analysis revealed palmitoylation of ~7% endogenous APP in naive CHO cells. Strikingly, we could also identify both endogenous immature and mature palAPP by mABE using non transgenic mouse brain extracts (Fig. 1F). Together, all three different methods used showed for the first time that APP is palmitoylated.
Next, we asked which protein domain of APP harbors the palmitoylated cysteine residues. Given that APP lacks cysteine residues in its cytosolic domain, we used N-terminal deletion mutants of the protein. We performed ABE assays on the neuronal APP isoform APP695, which lacks the KPI domain containing 6 cysteine residues, on APP(Δ281) that lacks 12 N-terminal cysteine residues, and on APP(Δ343) lacking both the KPI and the cysteine-rich domains (Fig. 2A). Our results showed that APP695 was palmitoylated similarly to APP751 (~98%), indicating that one or more of the 12 cysteine residues in the N-terminal region between Cys38 and Cys187 are palmitoylated (Fig. 2B,C). Indeed, APP(Δ281) showed little or no palmitoylation (~9%), confirming that the six cysteine residues in the KPI domain play a minimal role in APP751 palmitoylation (Fig. 2B,C). APP(Δ343) resulted in complete loss of palmitoylation, as expected given the absence of cysteines in this protein (Fig. 2B,C). Therefore, effective palmitoylation of APP occurs in its N-terminal domain between Cys38 and Cys187. These data confirm that APP undergoes lumenal palmitoylation, similarly to Sonic Hedgehog, Spitz, or Wnt (Resh, 2006; Buglino and Resh, 2008).
A precise consensus sequence for protein palmitoylation is not known. However, a previously developed palmitoylation site prediction algorithm with a clustering and sorting strategy allows for identification of potential palmitoylation sites with >80% sensitivity and specificity (Zhou et al., 2006). Applying this strategy, we identified two cysteine residues at positions 186 and 187 as potential candidates for S-palmitoylation. This prediction was confirmed in our systematic mutation analysis, where we independently mutated all 12 cysteine residues between Cys38 and Cys187 and performed palmitoylation assays (data not shown). The top panel of Figure 2D shows that site-directed mutagenesis of cysteine residues 186 and/or 187 to serines or alanines prevented palmitoylation of APP.
Next, we characterized APP Cys186 and/or Cys187 for changes in trafficking and processing. Stably transfected CHO cells expressing APP(C186S), APP(C187S), APP(C186,187S), or APP(C186,187A) generated little or no detectable APP CTFs (Fig. 2D, bottom). APP maturation was also reduced by ~95%, indicating potential ER retention of the mutant proteins (Fig. 2D, bottom). To test for ER retention, we performed a subcellular distribution assay using OptiPrep fractionation. CHO cells stably expressing wild-type APP or the cysteine mutant APP(C186,187S) were subjected to fractionation in 7.5–30% OptiPrep gradients (Fig. 2E). As expected, the localization of wild-type APP showed a gentle shift from the immature protein in the ER (calreticulin) to the mature protein appearing in the Golgi fractions (GM130). In contrast, predominantly immature APP(C186,187S) partitioned exclusively in the calreticulin-positive ER fractions, confirming that APP(C186,187S) is indeed retained in the ER (Fig. 2E). To complement the fractionation analysis, cells expressing wild-type or mutant APP were subjected to indirect immunofluorescence analysis. Confocal microscopy showed typical Golgi-like punctate staining of wild-type APP (Fig. 2F, a–c). In contrast, APP(C186,187S) showed predominant localization to ER-like reticular structures stained with the ER marker calreticulin (Fig. 2F, d–f). Additionally, wild-type APP partially colocalized with the Golgi marker GM130, while APP(C186,187S) did not (data not shown). Not surprisingly, conditioned media from the mutant CHO cells showed >95% decrease in Aβ40 and Aβ42 levels as compared to WT APP-expressing cells (Fig. 2G). Similarly, PC-12 cells transiently expressing APP(C186S), APP(C187S), or APP(C186,187S) presented identical results, as all three proteins generated barely detectable APP CTFs, and their Aβ40 and Aβ42 levels were reduced by 85–95% as compared to transfected WT APP (data not shown). Collectively, these data show that APP is S-palmitoylated at Cys186 and/or Cys187 in the ER or early secretory compartments, and that these cysteine residues are required for ER exit of APP.
We initially reasoned that the nearly complete ER retention of our APP(C186,187S), APP(C186,187A), APP(C186S), or APP(C187S) mutants was due to the serine or alanine residues causing improper folding of the APP ectodomain. To overcome this issue, we created 15 additional APP mutant proteins, each containing a different amino acid replacement of Cys186 or Cys187. All 15 mutants generated undetectable or barely detectable levels of mature APP, presumably causing ER retention of each mutant protein (Fig. 2H; data not shown). Since only up to 11% of APP is palmitoylated in CHOAPP cells, lack of palmitoylation may not account for the dramatic ER retention of Cys186 or Cys187 APP mutants. Indeed, Cys186 and Cys187 reside within the copper binding domain of APP and are predicted to stabilize domain structure by forming disulfide bonds with Cys158 and Cys133, respectively (Fig. 2I) (Barnham et al., 2003). To confirm that disrupted disulfide bridges contribute to the observed ER retention, we mutated the partner Cys158 and Cys133 to serines to produce APP(C133S), APP(C158S), and APP(C133,158S) mutants. All three mutants showed enhanced palmitoylation in our ABE assays compared to wild-type APP, apparently due to increased availability of Cys186 and Cys187 to incorporate palmitate (Fig. 2J, top). This experiment suggests the interesting possibility that disulfide bond formation between Cys158 and Cys186 or Cys133 and Cys187 may regulate APP palmitoylation. Despite a slight increase in palmitoylation, we observed a modest decrease in the maturation of all three mutant APP proteins, confirming that disulfide bridges between Cys158 and Cys186 or Cys133 and Cys187 are also essential for ER exit of the protein (Fig. 2J, bottom). Interestingly, we noticed a trend for slightly increased β-CTF levels as compared to wild-type APP in all three mutants (Fig. 2J, bottom). Together, our data show that palmitoylation at Cys186 and Cys187 or disulfide bonds between Cys158 and Cys186 or Cys133 and Cys187 are important for ER exit of APP, and suggest for the first time that increased APP palmitoylation may enhance β-cleavage of APP.
Since mutations inserted into Cys186 or Cys187 resulted in nearly full ER retention of APP, these mutants could not be used to assess the effect of APP palmitoylation on APP metabolism. To begin assessing the effect of APP palmitoylation on APP trafficking, we first asked whether palAPP is enriched in lipid rafts compared to nonrafts. Indeed, a prominent function of palmitoylation is to recruit proteins to cholesterol-rich lipid raft microdomains (Resh, 2004; Brown, 2006). We first isolated lipid raft and non-lipid raft fractions from CHOAPP cells using discontinuous sucrose gradient fractionation (Fig. 3A). As expected, total full-length APP was more abundant in nonraft versus raft fractions. To assess relative distribution of palAPP, we adjusted the protein concentration in our ABE palmitoylation assay to start with approximately equal levels of total APP (Fig. 3B, left). Thus, approximately equal amounts of APP from lipid raft and nonraft fractions were subjected to an ABE assay. Interestingly, the ABE assay demonstrated high palAPP levels in lipid raft fractions, while palAPP remained nearly undetectable in nonraft fractions (Fig. 3B, right). Quantitative analysis revealed that while 19.7 ± 2.3% of lipid raft-bound APP was palmitoylated, only 2 ± 0.7% of nonraft APP showed palmitoylation (Fig. 3C). Since 2% of nonraft APP is palmitoylated and full-length APP is mainly localized to nonraft fractions, palAPP is also abundantly found in nonraft fractions (for non-raft APP levels, see Fig. 3A; 2% of this APP is palmitoylated in CHOAPP cells).
To show that palmitoylation also targets APP to lipid rafts in vivo, we isolated lipid raft and non-lipid raft fractions from non-Tg mouse brains (Fig. 3D). Similarly to the experiments performed in cells, palAPP was detected in mouse brain lipid raft and nonraft fractions (Fig. 3E). Quantitation of these direct ABE assays revealed that 18.86 ± 1.4% of raft-bound APP was palmitoylated, compared to only 2.68 ± 1.6% of non-raft APP showing palmitoylation (Fig. 3F). Modified ABE assays on nontransgenic mouse brain extracts confirmed these results (data not shown). Together, these in vivo and in vitro data indicate that one function of APP palmitoylation is to target palAPP to lipid raft fractions, thus perhaps promoting Aβ generation.
We next asked whether enhanced APP palmitoylation results in elevated Aβ production in cells. Palmitoylation is a reversible process regulated by palmitoylating enzymes [palmitoyl acyltransferases (PATs)] and depalmitoylating enzymes. Biochemical and genetic studies have identified a number of PATs, which share an ~50 aa cysteine-rich domain with a conserved Asp-His-His-Cys (DHHC) motif (Mitchell et al., 2006). Several DHHC PATs have been reported to increase palmitoylation of BACE1, especially of its immature form (Vetrivel et al., 2009).
We tested the effect of 23 HA-epitope-tagged DHHC PATs (DHHC-1 to DHHC-23) on APP palmitoylation in transiently transfected CHOAPP cells. Similarly to BACE1, several DHHC PATs enhanced APP palmitoylation in our ABE assay (data not shown). DHHC-7 and DHHC-21 were among the most consistent enzymes to increase APP palmitoylation, while DHHC-1 consistently did not (Fig. 4A, ABE assay). DHHC-7 or DHHC-21, but not DHHC-1, expression increased generation of both APP α- and β-CTFs (Fig. 4A, IB: C66). Importantly, DHHC-7 or DHHC-21, but not the control DHHC-1, increased secreted Aβ40 and Aβ42 levels in the conditioned media of CHO cells (Fig. 4B). Similarly, in PC-12 cells, overexpression of DHHC-7 or DHHC-21 also increased palmitoylation of endogenous APP (Fig. 4A), APP CTF generation, and Aβ40 and Aβ42 levels by approximately twofold compared to mock transfected cells (Fig. 4B). These data further confirm that APP is S-palmitoylated and indicate that palmitoylation promotes Aβ generation.
To directly address whether palAPP undergoes β- or α-secretase cleavages, we first looked for their palmitoylated cleavage products pal-sAPPβ or pal-sAPPα in the extracellular milieu of CHOAPP cells. Indeed, it is impossible to differentiate between Aβ or p3 peptides generated from palAPP or non-palAPP, given that palmitoylation of APP occurs N-terminally from its BACE1 cleavage site. Figure 5A shows that both pal-sAPPβ and pal-sAPPα were detectable in the media of CHOAPP cells, indicating that both β- and α-secretases constitutively cleave palAPP. Interestingly, levels of pal-sAPPβ were similar to those of pal-sAPPα (Fig. 5A, first lane, top). We usually see predominantly high levels of sAPPα when the palmitoylated pool of APP is not specifically examined (Fig. 5A, first lane, bottom).
To further investigate BACE1-mediated processing of palAPP, we selectively overexpressed or inhibited BACE1 in CHO cells (Fig. 5A). Stable overexpression of myc-epitope tagged BACE1 (CHOAPP+BACE1) not only increased the amount of sAPPβ by 1.90 ± 0.16-fold (p < 0.01) in the media as expected, but also elevated pal-sAPPβ levels by 2.21 ± 0.31-fold (p < 0.01) in our ABE assay, confirming that palAPP is cleaved by BACE1 (Fig. 5B). Interestingly, pal-sAPPα levels decreased upon BACE overexpression, further indicating that BACE1-mediated cleavage of palAPP is favored over α-secretase processing (Fig. 5A). BACEi IV reduced pal-sAPPβ levels in both CHOAPP and CHOAPP+BACE1 cells, again confirming that the released pal-sAPPβ is indeed a BACE1-cleaved product.
Since BACE1 overexpression increases pal-sAPPβ levels, levels of its precursor full-length palAPP would be expected to concomitantly decrease. To show this, we quantitated total full-length palAPP levels in CHOAPP as compared to CHOAPP+BACE1 cells. Indeed, overexpression of BACE1 in CHOAPP cells reduced total palAPP levels by ~ 41% (Fig. 5 C,D). This was somewhat surprising, as elevated BACE1 activity is known to leave total APP largely unchanged, despite a significant increase in β-CTF/C99 levels (Fig. 5C, bottom). BACE1 expression did not significantly change palmitoylated flotillin levels (Fig. 5C). These data not only show that palAPP undergoes BACE1-mediated cleavage, but also suggest that BACE1 cleaves a larger percentage of palAPP than total APP proteins.
After showing that BACE1 overexpression decreases full-length palAPP levels, we next asked whether full-length palAPP levels increase with BACE or α-secretase inhibition. For these experiments, we used two BACE inhibitors (BACEi IV and dr9) and one α-secretase inhibitor (TAPI) in CHOAPP cells. All three inhibitors worked in our experiments, as evidenced by changes in the positive controls sAPPβ and sAPPα levels (Fig. 5E, bottom). To detect full-length palAPP levels, we used our ABE assay. BACEi IV and dr9, but not the α-secretase inhibitor TAPI, increased the levels of mature and immature palAPP (Fig. 5E). BACEi IV also induced a dose-dependent increase of newly synthesized palAPP in cyclohexamide-treated CHOAPP cells (Fig. 5F, ABE assay, G). In the same experiment, we detected an inverse correlation between palAPP and β-CTF levels (Fig. 5G), while total APP levels remained largely unaffected (Fig. 5F, compare palAPPm/palAPPim, βCTF). α-CTF levels also decreased to a lesser extent. Subcellular fractionation on an OptiPrep gradient showed that both mature and immature APP colocalize with BACE1 in Golgi compartments, a potential site for BACE1-mediated cleavage of palAPP (Fig. 5H). These experiments further confirm that palAPP is more efficiently processed by β- than α-secretase in cultured cells.
To test for BACE1-mediated palAPP processing under physiological conditions, we asked whether BACE inhibitors would also increase palAPP levels in primary neuronal cells. Primary neurons were treated with BACEi IV or dr9 for 16 h. ABE analysis showed elevated palAPP levels in BACE inhibitor-treated neurons (Fig. 6A). Treatment with 0.5 μm BACEi IV increased palAPP levels by 2.17 ± 0.2-fold compared to vehicle-treated cells (Fig. 6B). As a positive control for effective BACE inhibition in primary neurons, sAPPβ generation was strongly reduced (Fig. 6A, CM). Alkyl-C16 labeling and TAMRA incorporation confirmed these results in a separate experiment (Fig. 6C).
Finally, we asked whether lack of BACE1 expression affects in vivo palAPP processing in cortices isolated from age-matched wild-type controls or BACE1-knock out (BACE−/−) mice. APP was immunoprecipitated from membrane extracts, and equal amounts of APP were subjected to mABE analysis to detect palAPP (Fig. 6D, b). These experiments showed increased levels of mature palAPP in BACE−/− compared to wild-type cortices (Fig. 6D, a), further indicating that palAPP is a substrate for BACE1 in vivo. Interestingly, this ~20% increase was specific for mature palAPP (Fig. 6E), while immature palAPP and total APP remained largely unchanged in vivo (Fig. 6D, a and b, respectively). Importantly, all our in vivo and cell-based data show that palAPP is a better substrate for BACE1 than α-secretases. Additionally, the finding that BACE1 cleaves a larger percentage of palAPP than total APP proteins indicates that BACE1 may cleave palAPP more readily than non-palAPP.
Since palAPP is efficiently cleaved by BACE1 and is enriched in lipid rafts, we next asked whether BACE1 could cleave palAPP in lipid raft fractions. To this end, we performed cell-free in vitro BACE-activity assays on lipid raft-associated APP and assayed each reaction for pal-sAPPβ levels (Fig. 7). We first separated lipid raft and non-lipid raft fractions from CHOAPP cells using a discontinuous sucrose gradient fractionation. Lipid raft fractions were kept at 0°C in a low-pH buffer (pH 4) and warmed to 37°C for BACE1-mediated cleavage of APP and palAPP. N-terminal APP fragments were then immunoprecipitated with the anti-APP N-terminal antibody 22C11 and subjected to mABE palmitoylation assay to detect pal-sAPPβ. Interestingly, pal-sAPPβ levels dramatically increased upon incubation at 37°C (7.3 ± 1.1, p < 0.1 times over 0°C), while the increase in total sAPPβ did not reach statistical significance (Fig. 7A,B). To further confirm that pal-sAPPβ was indeed generated by BACE1 cleavage, we performed the assay in presence of increasing concentrations of the BACE inhibitor BACEi IV. The inhibitor prevented the increase in pal-sAPPβ levels in a dose-dependent manner (Fig. 7A,B), confirming BACE1-dependent processing of palAPP. Although the experiments were performed at a low pH (pH 4) to specifically test BACE1 activity, we probed for the release of sAPPα in the assay. We detected residual sAPPα in the reaction mixture, but the levels of sAPPα remained unchanged upon incubation at 37°C, confirming that our in vitro assay at pH 4 indeed released pal-sAPPβ and sAPPβ, but not sAPPα. BACEi IV also had no effect on sAPPα levels, as expected. Similarly to BACEi IV, 1 μm BACE inhibitor dr9 also decreased pal-sAPPβ levels in a cell-free in vitro BACE-activity assay performed at 37°C (data not shown). The large (~7.3 times over control) increase of pal-sAPPβ release at 37°C, not mirrored by total sAPPβ, indicates that palAPP is cleaved more efficiently by BACE than non-palAPP in our in vitro BACE-activity assay in lipid rafts.
Finally, we performed BACE-activity assays on lipid rafts isolated from mouse brains (Fig. 7C). Similarly to cells, we observed increased pal-sAPPβ upon BACE activation, which was reduced by BACEi IV in a dose-dependent manner (Fig. 7C). Total sAPPβ levels followed the same trend, although the increase from 0 to 37°C was less pronounced. Together, our data show that lipid raft-bound palAPP undergoes BACE1-mediated processing in cells and in vivo. Together with the DHHC PAT and BACE overexpression/inhibition data above, our results strongly advocate that palmitoylation promotes amyloidogenic processing of APP.
Aging is closely associated with the impaired cognitive performance of AD patients (Buckner, 2004). Here, we asked whether APP palmitoylation is age dependent. To assess the effect of aging on brain palAPP levels, we performed ABE assay on cortical extracts from 3-month-old (young) and 18-month-old (older) wild-type nontransgenic mice (Fig. 8A). Eighteen-month-old mice showed a 1.87 ± 0.7-fold (p < 0.05) increase in palAPP levels compared to younger mice (Fig. 8B), suggesting a direct correlation between age and increased endogenous APP palmitoylation in mouse brain.
To investigate whether palAPP levels can be reduced by pharmacological treatments, we used two different palmitoylation inhibitors: 2-bromopalmitate and cerulenin (Figs. 1B, ,99A). Both inhibitors decreased palmitoylation of APP in a concentration-dependent manner while also preventing maturation of APP and APP CTF generation (Fig. 9A). Unfortunately, palmitoylation inhibitors are known to be toxic as they severely disturb cellular lipid metabolism (Draper and Smith, 2009).
We and others have found that ACAT1 inhibition reduces Aβ generation in vitro, in cells, and in vivo and is a potential therapeutic target to lower Aβ (Puglielli et al., 2001; Hutter-Paier et al., 2004; Huttunen et al., 2007a, 2010; Bhattacharyya and Kovacs, 2010; Bryleva et al., 2010). Interestingly, the preferred long-chain acyl substrate for ACAT1 is palmitoyl-CoA. Since ACAT inhibition does not affect total APP levels or trafficking of total APP to lipid rafts (see below), we reasoned that it might affect generation of palAPP instead. First, we examined palmitoylation of APP in AC29APP cells, a genetically mutated CHO cell line that overproduces cholesterol while lacks ACAT1 activity (Puglielli et al., 2001). We have shown previously that this cell line is almost fully unable to produce APP CTFs and Aβ (Puglielli et al., 2001) (Fig. 9B, left). Glycosylation and trafficking of APP in these cells were largely unaffected by lack of ACAT activity (data not shown). To our surprise, AC29APP cells produced little or no palAPP (Fig. 9B, left). Most interestingly, reduction in palAPP was also observed in cells treated with the well-known ACAT1 inhibitor CP-113,818 (10 μm; Fig. 9B, right). We have shown previously that CP-113,818 effectively reduces Aβ generation in vivo and in vitro (Hutter-Paier et al., 2004).
Next, we treated CHOAPP cells with another well-established ACAT1 inhibitor, CI-1011 or avasimibe. CI-1011 was originally produced by Pfizer and previously reached phase III clinical trials for cardiovascular disease (Tardif et al., 2004). We reported previously that CI-1011 reduces APP CTF levels and Aβ generation in cells and in vivo (Huttunen et al., 2010). Four days of treatment with 5 or 10 μm CI-1011 severely reduced β- and α-CTF levels, as expected (Fig. 9C). Similarly to CP-113,818, 10 μm CI-1011 also reduced total palAPP level by ~57% (Fig. 9C,D).
To test the effect of ACAT1 inhibition on lipid raft-bound palAPP, we first performed lipid raft fractionation on CHOAPP cells in absence or presence of 10 μm CI-1011. Distribution of total APP to lipid raft and nonraft fractions remained largely unaltered upon treatment with CI-1011 (Fig. 9E). Equal amounts of raft and nonraft fractions were isolated from the cells and subjected to ABE assay. As expected, palAPP was highly enriched in lipid raft fractions of the untreated cells (Fig. 9F). Importantly, lipid raft-bound palAPP levels were reduced by 76.15 ± 8.9% in cells pretreated with CI-1011 (Fig. 9F). These experiments show that ACAT1 inhibition reduces palAPP levels in lipid rafts, likely resulting in the observed decrease in amyloidogenic processing of APP.
Our study shows for the first time that APP undergoes palmitoylation. We have identified the palmitoylation sites in the N-terminal E1 lumenal domain of APP, at Cys186 and/or Cys187. Cys186 and/or Cys187 mutants are retained in the ER. Palmitoylation targets APP to lipid rafts, where palAPP serves as a good BACE1 substrate. Palmitoylation or ACAT inhibitors severely impair the processing of palAPP by α- and β-secretases, presumably due to ER retention and degradation of APP. Together, our data indicate that palmitoylation preferentially targets APP to lipid rafts where palAPP serves as a good substrate for BACE1 cleavage and Aβ generation (Fig. 10).
Protein palmitoylation is a reversible process requiring palmitoylating enzymes (PATs), depalmitoylating enzymes (protein palmitoyl thioesterases), and access to palmitoyl-CoA. At least 24 PATs containing a common ~50 aa cysteine-rich domain with a conserved DHHC motif have been shown to catalyze cytoplasmic S-palmitoylation in mammals (Ohno et al., 2006). However, palmitoylation of APP necessarily occurs in the lumen of the secretory pathway since APP's cytoplasmic tail lacks cysteines. Examples of lumenal palmitoylation by membrane-bound O-acyltransferases (MBOAT) are becoming increasingly available (Buglino and Resh, 2008). In mammals, secreted proteins such as the morphogens Hedgehog and Wnt, the EGF-receptor ligand Spitz, as well as the peptide hormone ghrelin are post-translationally modified with fatty acids by MBOAT PATs such as Hhat and Porc (Miura and Treisman, 2006; Miura et al., 2006; Buglino and Resh, 2008; Yang et al., 2008). Modification of these secreted proteins occurs in the lumen of the secretory pathway. In addition to MBOAT PATs, the multiple transmembrane domain DHHC PATs could also catalyze lumenal palmitoylation if their active sites occasionally faced the lumen. Indeed, several transmembrane proteins such as aquaporin-1 reorient one or more of their transmembrane domains post-translationally (Lu et al., 2000). APP-interacting DHHC protein (AID)/DHHC-12 was previously reported to interact with APP and suppress APP processing (Mizumaru et al., 2009). However, DHHC-12 does not appear to palmitoylate APP in our experiments (data not shown). To our knowledge, APP is the first transmembrane protein that undergoes S-palmitoylation in the lumen. Identification of APP's palmitoylating enzyme(s) will further clarify the mechanism of lumenal palmitoylation of transmembrane proteins and may open up a novel pathway to reduce Aβ generation.
Lumenal palmitoylation of APP also requires transport of highly hydrophobic long-chain palmitoyl-CoA into the lumen (Gooding et al., 2004). Indeed, a brain-specific carnitine palmitoyltransferase 1c has been reported previously to localize in the ER membrane to ensure the entry of palmitoyl-CoA into the ER lumen (Sierra et al., 2008). Alternatively, free palmitic acid may be transported into the ER lumen before being converted to palmitoyl-CoA by lumenal acyltransferases (Rys-Sikora and Gill, 1998). Palmitoyl-CoA hydrolase activity was detected in the ER lumen, further confirming that protein palmitoylation and depalmitoylation are not confined to the cytoplasm (Mentlein et al., 1988).
We have identified Cys186 and Cys187 as palmitoylated Cys residues in the E1 domain of APP. We have also found that these cysteines strongly regulate trafficking of APP out of the ER. These same cysteines were predicted previously to form disulfide bonds with Cys133 and Cys158 to stabilize the copper-binding domain of APP. Our data show that lack of disulfide bond formation in Cys133 and Cys158 mutants may promote palmitoylation at Cys186 and Cys187, suggesting that disulfide bond formation at Cys186 and Cys187 in APP may regulate its palmitoylation (Fig. 2J). Conversely, Wnt-3a palmitoylation is predicted to regulate disulfide bond formation (Coudreuse and Korswagen, 2007). Interestingly, reduced forms of Cys186 and Cys187 appear to promote zinc binding to APP (Bush et al., 1993). This form would also allow the same cysteines to incorporate palmitate via a thioester bond. The precise relationship between palmitoylation, copper/zinc binding, and disulfide bond formation remains to be established for APP's Cys186 and Cys187. Our results are consistent with the idea that proper conformation of the E1 domain in APP is important for ER exit and intracellular trafficking of the protein. Disulphide bridges or protein palmitoylation at Cys186 and/or Cys187 facilitate this conformation.
palAPP being enriched in lipid rafts is in line with the importance of lipid rafts in APP processing. Palmitoylation often targets proteins to lipid rafts (Cheng et al., 2009; Vetrivel et al., 2009; for review, see Levental et al., 2010). APP, BACE1, and γ-secretase components are all found in detergent-resistant lipid rafts (Vetrivel et al., 2004; 2005; Hattori et al., 2006). Amyloidogenic processing of APP partially depends on lipid rafts whereas nonamyloidogenic α-cleavage of APP occurs mainly in the phospholipid-rich domain of the plasma membrane (Ehehalt et al., 2003). Both monomeric and dimeric species of Aβ are concentrated in lipid rafts in Tg2576 mice and human AD brains (Lee et al., 1998; Kawarabayashi et al., 2004). Palmitoylation also targets BACE1 and two components of γ-secretase, Aph-1 and Nct, to lipid rafts (Cheng et al., 2009; Vetrivel et al., 2009). Double-transgenic mice expressing APP and palmitoylation-deficient Aph-1 or Nct show reduced Aβ deposition (Meckler et al., 2010). Interestingly, artificial targeting of BACE to the lipid rafts enhanced β-cleavage of APP (Cordy et al., 2003). Our data indicate, but do not confirm, enhanced β-secretase-mediated processing of palAPP in lipid rafts. Surprisingly, wild-type APP is considered a poor substrate for BACE1 (Stockley and O'Neill, 2008). Strict biochemical quantitation of BACE1-mediated cleavage of palAPP will be required to determine whether palmitoylation does enhance APP processing in lipid rafts. An additional question is whether palAPP outside lipid rafts serves as a good BACE1 substrate. The main pool of APP resides outside the lipid rafts, but the relative concentration of palAPP in these fractions was too low for our in vitro BACE cleavage assays. Nevertheless, palAPP is also found outside of lipid rafts, where it may regulate APP processing.
Cortical extracts from nontransgenic mice showed an age-dependent increase in palAPP levels (Fig. 8). It is well known that both AD and aging are characterized by elevated Aβ accumulation, at a much higher degree in the brains of patients affected by AD than during normal aging (Price et al., 1992; Haass and Selkoe, 2007; Reddy and Beal, 2008). An age-dependent increase in BACE1 activity and levels may lead to a rise in Aβ generation in some, but not all, AD patients (Fukumoto et al., 2002, 2004). It remains to be investigated whether palAPP levels are increased in human aging and AD brains. Elevated palAPP, perhaps leading to more Aβ generation, may contribute to widespread Aβ accumulation in these brains.
Treatment with the palmitoylation inhibitors cerulenin and 2-bromopalmitate not only prevented APP palmitoylation, but also resulted in severe loss of its processing and maturation, suggesting that palmitoylation plays an important role in APP metabolism. In recent years, the key role of palmitoylation has become increasing clear in multiple diseases such as Huntington's disease, various cardiovascular and T-cell mediated immune disorders, as well as cancer (Draper and Smith, 2009). To date, two types of palmitoylation inhibitors have been identified, lipid-based palmitoylation inhibitors such as 2-BP or cerulenin, and non-lipid palmitoylation inhibitors such as the chemotypes Compounds I, II, III, and IV (Ducker et al., 2006). Lipid-based inhibitors broadly inhibit the palmitoylation of proteins and affect fatty acid biosynthesis. In contrast, non-lipid palmitoylation inhibitors have been shown to selectively inhibit the palmitoylation of different PAT recognition motifs (Ducker et al., 2006; Draper and Smith, 2009). However, no small-molecule PAT inhibitor has yet been developed for therapeutic purposes.
ACAT inhibitors reduced palAPP levels in lipid rafts, and lack of ACAT activity in AC29 cells prevented palmitoylation of APP. One possible explanation for how ACAT inhibitors reduce palAPP is that ACAT inhibition slightly increases ER free cholesterol (Huttunen et al., 2009), which then may limit availability of palmitic acid and/or palmitoyl-CoA inside the ER lumen. Alternatively, slightly increased free cholesterol levels may indirectly impact lipid raft levels enriched with proteases for the amyloidogenic pathway. Cholesterol binds directly to the transmembrane domain of APP (Bodovitz and Klein, 1996; Kojro et al., 2001; Barrett et al., 2012). Cholesterol has also been reported previously to directly interact with the palmitoyl moiety of the μ-opioid receptor (OPRM1) and thus regulate receptor function (Zheng et al., 2012). Simvastatin treatment reduced OPRM1 signaling (Zheng et al., 2012). Thus, changes in cellular cholesterol induced by ACAT inhibition or other CNS-relevant lipids such as 24-hydroxycholesterol could regulate palAPP levels and/or function. More research will be needed to explore this event at a mechanistic level. Numerous ACAT inhibitors have already been developed against atherosclerosis and hypercholesterolemia and tested in clinical trials (Farese, 2006). However, ACAT inhibitors are currently not marketed. Our previous studies confirmed by others clearly support ACAT inhibition as a strategy for regulating amyloid pathology in the brain (Puglielli et al., 2001; Hutter-Paier et al., 2004; Huttunen et al., 2010). Collectively, our previous studies and the current mechanistic finding that ACAT inhibitors reduce palmitoylation of APP warrant further development of ACAT inhibitors as a therapeutic strategy for AD.
This wok was supported by Cure Alzheimer's Fund grants and NIH–NINDS Grant R01NS45860 (D.M.K). We thank Dr. Vivek Gautam (Massachusetts General Hospital, Charlestown, MA) for his generous help in preparing primary neurons; Dr. Masaki Fukata (National Institute for Physiological Sciences, Okazaki, Japan) for providing us with the DHHC expression plasmids; Dr. Kiyohito Mizutani (Kobe University Graduate School of Medicine, Kobe, Japan; previously at Massachusetts General Hospital, Charlestown, MA) for generating APP(Δ343)-expressing cells; Dr. Jordan Tang (Oklahoma Medical Research Foundation, Oklahoma City, OK) for dr9; Lit-Fui Lau (GlaxoSmithKline, Shanghai, China; previously at Pfizer, Groton, CT) for CI-1011; and Dr. Doo Yeon Kim (Massachusetts General Hospital, Charlestown, MA) for helpful advice and discussions.
The authors declare no competing financial interests.