Palmitoylated APP Forms Dimers, Cleaved by BACE1

A major rate-limiting step for Aβ generation and deposition in Alzheimer’s disease brains is BACE1-mediated cleavage (β-cleavage) of the amyloid precursor protein (APP). We previously reported that APP undergoes palmitoylation at two cysteine residues (Cys186 and Cys187) in the E1-ectodomain. 8–10% of total APP is palmitoylated in vitro and in vivo. Palmitoylated APP (palAPP) shows greater preference for β-cleavage than total APP in detergent resistant lipid rafts. Protein palmitoylation is known to promote protein dimerization. Since dimerization of APP at its E1-ectodomain results in elevated BACE1-mediated cleavage of APP, we have now investigated whether palmitoylation of APP affects its dimerization and whether this leads to elevated β-cleavage of the protein. Here we report that over 90% of palAPP is dimerized while only ~20% of total APP forms dimers. PalAPP-dimers are predominantly cis-oriented while total APP dimerizes in both cis- and trans-orientation. PalAPP forms dimers 4.5-times more efficiently than total APP. Overexpression of the palmitoylating enzymes DHHC7 and DHHC21 that increase palAPP levels and Aβ release, also increased APP dimerization in cells. Conversely, inhibition of APP palmitoylation by pharmacological inhibitors reduced APP-dimerization in coimmunoprecipitation and FLIM/FRET assays. Finally, in vitro BACE1-activity assays demonstrate that palmitoylation-dependent dimerization of APP promotes β-cleavage of APP in lipid-rich detergent resistant cell membranes (DRMs), when compared to total APP. Most importantly, generation of sAPPβ-sAPPβ dimers is dependent on APP-palmitoylation while total sAPPβ generation is not. Since BACE1 shows preference for palAPP dimers over total APP, palAPP dimers may serve as novel targets for effective β-cleavage inhibitors of APP as opposed to BACE1 inhibitors.


Introduction
Amyloid precursor protein APP undergoes sequential proteolysis by β-and γ-secretases to generate amyloid β (Aβ). Deposition of the amyloid (Aβ) peptide in senile plaques is a hallmark of Alzheimer's disease (AD) (reviewed in [1][2][3]). Shortly after synthesis in the ER, APP undergoes a number of post-translational modifications namely N-and O-glycosylation, acetylation and phosphorylation prior to trafficking to the Golgi and eventually to the plasma membrane. APP also undergoes novel lumenal palmitoylation in the ER where two cysteine

APP palmitoylation promotes the formation of APP dimers
Several studies have shown that ectodomain-mediated APP dimerization requires hyodrophobic interactions, apart from the dimerization domains 18-350 and 448-465 in the N-terminus of the protein [19,25]. Ectodomain-dependent dimerization of APP was shown to increase Aβ generation [23]. Since APP contains hydrophobic palmitic acid residues at its Cys 186 and Cys 187 in the ectodomain [4], we asked whether APP palmitoylation promotes APP dimerization. Here we performed an assay combining co-immunoprecipitation (co-IP) to assess dimerization and a modified acyl biotinylation assay (mABE) to assess palmitoylation of APP. For this assay, we used cells co-transfected with two expression plasmids. One expressed C-terminal V5-epitope tagged APP (APP V5 ), the other C-terminal YFP-and N-terminal HA-epitope tagged APP (HA-APP Y ). As expected, HA-APP Y efficiently co-immunoprecipitated with APP V5 (Fig 1A and 1B), confirming dimerization of APP. To test the presence of palmitoylated APP-dimers (palAPP V5 -palHA-APP Y ), the precipitate was subjected to an mABE assay that not only detected~100 kDa palAPP V5 , but also identified the~150 kDa palHA-APP Y (Fig 1A). This showed dimerization of palAPP V5 and palHAAPP Y . Next we determined the stoichiometry of APP-APP and palAPP-palAPP interaction. For this we measured band intensities of pulled-down palHA-APP Y and that of palAPP V5 . Quantitation revealed~1:1 (0.91 ± 0.07) stoichiometry for palHA-APP Y /palAPP V5 interaction (Fig 1C), suggesting that~91% of palAPP formed dimers. We then compared the band intensities of immunoprecipitated totHA-APP Y and totAPP V5 (Fig 1B). Quantitation showed that totHA-APP Y coimmunoprecipitates with totAPP V5 with a modest 0.19 ± 0.02 stoichiometry ( Fig 1C). These data demonstrate that palAPP undergoes near total (~91%) dimerization while only~20% of totAPP forms dimers. Similar results were obtained when HA-APP Y was immunoprecipitated prior to mABE assay. HA-APP Y pulled down~20% APP V5 , while pal HA-APP Y pulled down equal amount of palAPP V5 (data not shown). Our data provide evidence for the first time that palmitoylation of APP strongly promotes APP dimerization.
APP forms cis-and trans-dimers in vitro and in vivo [26,27]. The cellular localization and function of APP may determine whether it dimerizes in cis or trans orientation [28]. Here we tested the orientation of palAPP dimers. We used co-IP assays on co-culture systems to ask whether palAPP is primarily dimerized in cis or trans. For this, cells expressing N-terminally myc-tagged APP (mycAPP) were co-cultured with cells expressing HA-APP Y . HA-APP Y co-IPed mycAPP only in presence, but not in absence, of a cell impermeable cross-linker DTSSP at 4˚C (Fig 1D, panel b, compare lanes 1 and 2). Given that the two proteins were expressed in different cell lines, this result indicates that mycAPP and HA-APP Y dimerized in trans-orientation in presence of the cross-linker. Surprisingly, when we subjected the immunoprecipitates to mABE analysis, the same APP dimers in trans were found not to be palmitoylated ( Fig 1D, panel a, lane 2). In contrast, HA-APP Y not only pulled down mycAPP ( Fig 1D, panel a, lane  3), but both HA-APP Y and mycAPP were also palmitoylated ( Fig 1D, panel b, lane 3), in experiments where HA-APP Y and mycAPP were coexpressed in the same cell. A dimerizationdefective mycAPP mutant containing the H108/110A mutation in the Growth Factor Like Domain (GFLD) of APP (mycAPP(mut)) showed little or no co-immunoprecipitation with HA-APP Y (Fig 1D, panel a, lane 4) as expected from an earlier report [26]. Taken together, our data showed that palAPP did not form trans-dimers, thus suggesting palAPP-dimers were predominantly cis-oriented. Interestingly, cis-dimerization in particular is known to affect APP processing [22,23,29], increasing Aβ and sAPPβ generation [30].
To further confirm the direct correlation between APP palmitoylation and its dimerization, we have performed co-immunoprecipitation (co-IP) assays to test dimerization of APP in cells A. Cells expressing APP V5 or APP V5 plus HA-APP Y were subjected to co-immunoprecipitation assays to detect APP V5 /HA-APP Y interaction or APP-dimerization. APP V5 was immunoprecipitated with an anti-V5 antibody.
co-expressing HA-and V5-epitope-tagged palmitoylation-efficient APP mutants APP(C 133 S) and APP(C 158 S). We previously reported that substitution of Cys 133 or Cys 158 with serine (Ser) results in a~2 fold increase in APP palmitoylation compared to APP wt , by freeing the palmitoylatable cysteins (Cys 186 or Cys 187 ) from forming disulfide (S-S) bridges with Cys 158 and Cys 133 , respectively (Fig 2A, and [4]). Co-IP assays revealed that APP(C 133 S) and APP(C 158 S) form~2 fold increased dimerization compared to that of APP wt (Fig 2B), further confirming that increased palAPP levels increased APP-dimerization. In contrast, palmitoylation-deficient APP(C 186 S) and APP(C 187 S) showed little or no dimerization. Taken together, these results indicate that APP-palmitoylation promotes its dimerization.
Palmitoyl acyl transferases DHHC7 and DHHC21, but not DHHC1, increase APP palmitoylation and dimerization Because palmitoyl acyltransferases DHHC7 and DHHC21 consistently increased palAPP level without altering the amount of totAPP [4], we tested the effect of DHHC7 and DHHC21 on APP-dimerization. For this purpose, we used co-immunoprecipitation of HA-APP Y with APP V5 in presence or absence of DHHC7 or DHHC21 to assess the effect of the DHHCs on APP dimerization (Fig 3). DHHC1 was selected as a negative control, as expression of DHHC1 does not promote APP palmitoylation [4]. DHHC7 and DHHC21 consistently increased co-IP of HA-APP Y and APP V5 , while DHHC1-overexpression had no effect on HA-APP Y /APP V5 coimmunoprecipitation ( Fig 3A). Overexpression of DHHC7, in particular, not only increased APP dimerization (APP V5 /HA-APP Y interaction) (Fig 3B), but also consistently increased palAPP V5 and palHA-APP Y levels in our ABE analysis ( Fig 3B). Quantitation of APP dimerization revealed that DHHC7 increased APP dimerization by 2.3 ± 0.17 fold ( Fig  3C). DHHC7 also increased palAPP V5 and palHA-APP Y levels by~2 fold (Fig 3B), as expected. Thus, our data show a direct correlation between APP palmitoylation and APP dimerization.

Palmitoylation inhibitors reduce APP dimerization
Next we asked whether inhibition of APP-palmitoylation affects APP-dimerization. Here, we tested the effect of two palmitoylation inhibitors, 2-bromopalmitate (2-BP) and cerulenin, on APP-dimerization because they had induced robust decrease of APP palmitoylation in our earlier report. Cells co-expressing HA-APP Y and APP V5 were subjected to cerulenin treatment prior to co-IP assay. Co-immunoprecipitation of HA-APP Y and APP V5 was decreased by cerulenin-treatment in a dose dependent manner (Fig 4A). In a separate experiment, cerulenin also decreased generation of Aβ 40 and Aβ 42 in APP-expressing cells (CHO APP ) in a dosedependent manner. Specifically, conditioned media from CHO APP cells generated 331.8±14.5 Immunoprecipitates were probed with an anti-HA antibody to detect pull-down of HA-APP Y . Subsequently the immunoprecipitates were subjected to mABE assay to detect palAPP V5 /HA-APP Y interaction (or palAPPdimerization). PalAPP V5 pulled down both palAPP V5 (M wt~1 02 kD) and palHA-APP Y (M wt~1 50 kD) from cells expressing APP V5 plus HA-APP Y but not from cells expressing only APP V5 . B. TotAPP-dimers (APP V5 / HA-APP Y ) only form in cells expressing both APP V5 and HA-APP Y . C. Quantitation of palAPP-dimers (palAPP V5 /palHA-APP Y ) versus totAPP-dimers (APP V5 /HA-APP Y ). Error bars show the s.e.m. (**p<0.01). D. palAPP dimerizes is cis-orientiation. Cells expressing HA-APP Y and cells expressing mycAPP were cocultured in absence or presence of 1mM cell-impermeable cross-linker DTSSP. Cell extracts were subjected to a pull-down assay, using an anti-HA antibody to immunoprecipitate HA-APP Y . To test for APP-dimerization, the precipitates were probed with an anti-myc antibody (panel b, co-culture). Cells co-expressing HA-APP Y and mycAPP were also subjected to a co-IP assay using the anti-HA antibody to pull-down mycAPP with HA-APP Y .(panel b, co-expression). To detect palAPP-dimerization, the immunoprecipitates were also subjected to mABE assay to detect co-IP of palHA-APP Y with pal-mycAPP (panel a). The experiment is a representative of three independent experiments. doi:10.1371/journal.pone.0166400.g001

Fig 2. Palmitoylation-prone APP mutants exhibit increased APP dimerization compared to wtAPP. A.
Schematic representation of the Cys to Ser mutants of APP used for the following co-immunoprecipitation assays. B. Co-immunoprecipitation assay in cells co-expressing APP V5 and HA-APP Y and its mutants containing indicated Cys to Ser substitution. HA-APP Y pulls down APP V5 , indicating APP-APP dimerization. APP(C 133 S) and APP(C 158 S) show 2 fold increase in dimerization, while APP(C 186 S) and APP(C 187 S) fail to dimerize. APP(C 186 S) and APP(C 187 S) generated trace amounts of palmitoylation-independent dimers (* and **). C. ABE assay of cells overexpressing indicated APP mutants show 2 fold increased palmitoylation of APP(C 133 S) and APP(C 158 S), where as APP(C 186 S) and APP(C 187 S) were defective in palmitoylation.
doi:10.1371/journal.pone.0166400.g002 Thus, inhibition of APP palmitoylation not only leads to disruption of APP-dimerization, but also reduces Aβ generation. Similar to cerulenin, 2-BP also dramatically reduced HA-APP Y /APP V5 co-immunoprecipitation ( Fig 4B). Quantitation showed~56% reduction of HA-APP Y /APP V5 coimmunoprecipitation by 50 μM 2-BP, while 25 μg/ml cerulenin -treatment reduced the interaction by nearly 58% (Fig 4C). As expected, cerulenin and 2-BP also caused a 50% reduction in palmitoylated APP (palHA-APP Y and palAPP V5 ) levels without affecting total HA-APP Y and APP V5 levels. Our data showed that inhibition of APP palmitoylation reduced APP dimerization, further confirming a correlation between these two modifications.
To validate the effect of 2-BP and cerulenin on APP dimerization we determined the 2pFLIM efficiency of two interacting APP molecules in absence or presence of the inhibitors. For this purpose we transfected cells with APP C-terminally tagged with mGFP (APP mGFP ) C. Quantitation showed 54 and 58% decrease of APP-APP dimerization by cerulenin (25 μg/ml) and 2-BP (50 μM), respectively. D. Naïve CHO cells were either transiently transfected with an expression plasmid encoding APP mGFP or co-transfected with expression plasmids encoding APP mGFP and APP mCherry . After 24 h, cells were either treated with DMSO or with indicated palmitoylation inhibitors cerulenin or 2-BP for 6 h prior to formalin treatment. FRET/FLIM analysis was employed to measure decay time constant T m of APP mEGFP in cells expressing APP mEGFP (n = 54), and APP mGFP and APP mCherry (n = 46). T m of life-time decay of APP mEGFP in APP mEGFP /APP mCherry co-expressing cells was monitored in absence and in presence of palmitoylation inhibitors (25 μg/ml cerulenin or 50 μM 2-BP). Quantitation revealed that inhibitors increased T m values by~1.5 fold, indicating disruption of dimerization between APP mGFP and APP mCherry . and mCherry (APP mCherry ). FRET measurements were taken by using APP mGFP as donor and APP mCherry as acceptor, as described by Fogel, H. et al (Fig 4D). Briefly, the 2pFLIM method is based on the fact that that shortening of donor lifetime indicates FRET. APP mEGFP alone showed lifetime decay, displaying a time constant T m of 2.65 ± 0.06 ns (Fig 4E). FRET between APP mEGFP and APP mCherry decreased the T m to 1.3 ± 0.02 ns (Fig 4E), indicating a strong APP mEGFP -APP mCherry interaction. 2-BP (50 μM) and cerulenin (25 μg/ml) treatment brought up the time constant to 1.76 ± 0.06 and 1.72 ± 0.09 (Fig 4E), respectively, as these compounds reduced APP mEGFP -APP mCherry interaction. FRET analysis revealed a~32 and a~35% reduction in APP dimerization by 2-BP and cerulenin, respectively. Here, we further demonstrated that reduction in palAPP levels by palmitoylation-inhibitors (cerulenin and 2-BP) reduced APP dimerization.
So far, we confirmed that palmitoylation inhibitors decreased APP dimerization. While co-IP assays and FRET/FLIM analyses yielded the same result, the decrease in APP dimerization was more pronounced in our co-IP assays whereas FRET assays did not completely correlate with decrease in APP palmitoylation and dimerization. Specifically, co-IP assays showed that 25 μg/ml cerulenin and 50 μM 2-BP reduced both APP-palmitoylation and APP-dimerization by 50%. FRET/FLIM analyses only yielded a consistent~33% in APP-dimerization by the same inhibitors that reduced palAPP level by~50%.
To further confirm the effect of palmitoylation inhibitors on APP-dimerization we attempted another approach. We employed bimolecular fluorescence complementation (BiFC) assays on cells co-expressing two APP constructs, APP-GFP(1-10) and APP-GFP(11), containing split-GFP. BiFC assays were performed as described by Isbert et al. [17]. APP-GFP (1-10) and APP-GFP(11) expression plasmids contained two non-fluorescence portions of GFP fused separately to the N-terminus of APP. GFP's full fluorescence property, called BiFC, is restored when APP-GFP(1-10) and APP-GFP(11) are brought together due to APP-APP association. Here, we co-expressed expression plasmids encoding APP-GFP(1-10) and APP-GFP(11) in naïve CHO cells. Cells expressing APP-GFP(1-10) or APP-GFP(11) did not generate any fluorescence (Fig 5A, a and b, respectively), as expected, while cells co-expressing the split-GFP plasmids, APP-GFP(1-10) and APP-GFP(11), produced robust fluorescence (Fig 5A, c) as shown before [17], suggesting APP-APP dimerization. The co-expressing cells were then sorted by a fluorescence-activated cell sorter (FACS) to obtain homogenous cultures of cells expressing APP-GFP(1-10) and APP-GFP (11). FACS sorted cells co-expressing APP-GFP(1-10) and APP-GFP(11) were grown on coverslips over night before treating with increasing amounts of cerulenin (0-100μg/ml) for 3 h ( Fig 5A, d, e and f, respectively), although 25, 50 and 100μg/ml cerulenin-treatment decreased palAPP(1-10) levels by~19,~57 and~99%, respectively (Fig 5B and 5C). It was surprising that the effect of APP-dimerization by palmitoylation-inhibitors in our BiFC analysis did not yield even 33% decrease that was observed in our FLIM/FRET analysis. It is possible that BiFC APP and untagged APP differ in their stability, thus masking the effect of the palmitoylation inhibitors. Constitutively expressed BiFC APP (APP(1-10)) and untagged APP (APP) showed half-lives of 2-3 h (S2A Fig) similar to earlier reports describing the half-life of APP as~4 h [31]. Thus, the little or no effect of cerulenin on BiFC APP dimerization is not due to stronger stability of the BiFC APP mutants. We also tested the half-life of palAPP by pulse chase analysis where CHO APP cells were first labeled with chemically reactive palmitic acid, Alkyl-C16, as before [4] followed by chasing with unlabeled palmitic acid for 0.5, 1, 3 and 6 h. palAPP was detected after labeling Alkyl-C16 incorporated APP with fluorescent TAMRA using Click-iT technique as before [4]. Half-life of palAPP appeared to be 3-6 h (S2B Fig), suggesting no change in the stability of palAPP compared to totalAPP. Although the data does not indicate whether palAPP dimers are more stable than non-palAPP, it will be interesting to determine the half-life of dimerized palAPP in future when and if an antibody specific for palAPP becomes available.
We now asked if dimerization of APP C-terminal fragments (CTFs), only detected in our FRET/FLIM or BiFC analyses, in addition of full-length APP, could explain the discrepancy between the co-IP analysis and the fluorescence-based methods. To distinguish between fulllength APP and APP-CTF dimerization, we tested the effect of the palmitoylation inhibitors on APP-APP and CTF-CTF interactions in cells expressing C-terminally V5-or HA-epitope tagged APP (APP V5 and APP HA , respectively) ( Fig 6A). Again, we detected CTF V5 /CTF HAdimerization in addition to full-lenghth APP V5 /APP HA -dimerization ( Fig 6A). Similar to cells  1-10)) and APP C-terminal split GFP 11 (APP(11)), FACS sorted for equal intensity. Cells showing green fluorescence represents the APP dimers. These cells were sorted in a fluorescence activated cell sorter (FACS) to obtain equal intensity cells. Next the sorted cells were grown on coverslips for 18 h before treating with 0 (DMSO), 25, 50 and 100μg/ml cerulenin for 6 h showed little or no change in fluorescence intensities. B. Cells expressing APP(1-10) or APP(1-10)+APP(11) were subjected to ABE assay. Probing the samples with anti-GFP detected palmitoylaed APP(1-10) (palAPP(1-10)). Anti-Flotillin antibody detected palmitoylated flotillin (palFlotillin) in ABE assay. C. BiFC intensities of the cells were quantitated using ImageJ software. Intensities of more than 50 cells were measured for each treatment. Average intensities are plotted in percent (%) change in dimerization, using no treatment as 100% describing changes in APP-dimerization percent by increasing amount of cerulenin (solid line). The discontinuous line represents decrease in palAPP levels obtained from ABE analysis upon cerulenin treatment. Error bars show the s.e.m.
Next we tested the effect of palmitoylation inhibitors on APP-dimerization in human neural stem cells (ReN cells, Millipore) differentiated into mature neurons. Differentiated Ren lines containing FAD mutants inside a 3-D matrix has been demonstrated as a potential cellular model for AD [32]. Here, we infected naïve ReN-VM cells with lentiviral particles containing expression vectors for APP mGFP or APP mCherry . Cells were then sorted by a fluorescenceactivated cell sorter (FACS) to obtain homogenous cultures of cells expressing APP mGFP (ReN-A G ) or APP mGFP +APP mCherry (ReN-A GC ) (S3A Fig). The sorted cells were allowed to differentiate into neurons as described by D'Avanzo et al. [33] (S3B Fig) prior to co-IP analysis in absence or presence of palmitoylation inhibitors. A pull-down assay using an antibody specific for the mGFP (anti-GFP) epitope co-immunoprecipitated APP mGFP and APP mCherry , suggesting APP mGFP -APP mCherry interaction ( Fig 6B). Interestingly, the C-terminal fragment of APP mGFP (CTF mGFP ) also co-precipitated with the CTF of APP mCherry (CTF mCherry ), suggesting CTF mGFP -CTF mCherry interaction. Cerulenin (25 μg/ml) and 2-BP (50 μM) reduced APP mGFP -APP mCherry interaction by~50%. In contrast, cerulenin or 2-BP showed little or no effect on CTF mGFP -CTF mCherry interaction (Fig 6A), further confirming that inhibition of APP palmitoylation prevents full length APP-APP dimerization without affecting CTF-CTF dimerization.
palAPP dimers serve as substrates for β-cleavage in cell membranes The APP ectodomain (E1) regulates a number of APP-functions, such as synaptogenesis [26] or glutamate release via the G i/o -signaling pathway [27]. Induced dimerization of APP via its E1 domain increased Aβ production and sAPPβ release [23,24]. Since palAPP is a better substrate for β-cleavage compared to total APP, we asked whether palAPP dimers undergo βcleavage. We hypothesized that dimerized palAPP would produce dimerized pal-sAPPβ after proteolysis by BACE1. First, we attempted to identify sAPP dimers in the conditioned media of cells co-expressing N-terminally myc-epitope tagged APP (mycAPP) and HA-APP Y . Both myc-sAPP and HA-sAPP were detected in the conditioned media, but we were unable to detect any dimerized sAPP in the CM via co-IP assays (data not shown). We then reasoned that palAPP dimers, upon proteolysis by β-secretase, may produce pal-sAPPβ dimers anchored to cell membranes. However, we again failed to detect pal-sAPPβ dimers in total or lipid raft membranes (data not shown). Although this observation did not entirely eliminate the possibility of pal-sAPP dimers in the conditioned media, it appears that their presence may be below our detection limit. Developing a pal-APP-specific antibody will be necessary to detect pal-sAPP dimers in the conditioned media.
To increase the sensitivity of our assay, we turned to in vitro experiments. We previously reported that palAPP is a better substrate than totAPP for BACE1-mediated β-cleavage in in vitro studies, using detergent resistant lipid raft microdomains. Thus, we next asked whether palAPP dimers are better substrates than totAPP for β-cleavage in an in vitro BACE-activity assay in detergent resistant membranes (DRM). DRMs were rich in lipid rafts as evident from enriched amounts of raft-resident protein flotillin in these membrane fractions compared to that in non-DRM fractions (data not shown). DRMs also showed the presence of high levels (~20%) of palAPP compared to that in non-DRM fractions. To detect the generation of sAPPβ dimers in these palAPP-rich DRMs, we performed co-IP experiments after in vitro BACE1-activity assays of DRMs isolated from HA-APP Y /mycAPP-expressing (Fig 7A). To stabilize released palHA-sAPP β /palmyc-sAPP β dimers, we pretreated HA-APP Y and mycAPP coexpressing cells with the mild crosslinker DSS (disuccinimidyl suberate) at a low concentration (50 μM) as described by Fogel, H et al. [27] prior to DRM preparation. DRMs isolated from these cells were incubated in acetate buffer, pH 4 at 37˚C for 1h to generate sAPP β . The membranes were then subjected to co-IP analysis where myc-sAPP β generated upon β-cleavage of the mycAPP/HA-APP Y -dimer were precipitated using an anti-myc antibody to pull-down HA-sAPP β (Fig 7A). Interestingly, the anti-myc antibody not only pulled down~100 kD HA-sAPP to confirm myc-sAPP/HA-sAPP interaction, but also co-precipitated~150 kD flHA-APP Y indicating the presence of residual myc-APP/HA-APP Y dimers in the assay ( Fig  7A, upper panel). Most importantly, an anti-sAPP β antibody stained a pulled-down~100kD sAPP (Fig 7A, IB:anti-sAPPβ), indicating that the myc antibody pulled-down HA-sAPP ( Fig  7A, upper panel) as sAPP β . This showed that myc-sAPP β /HA-sAPP β dimers were present after β-cleavage of APP-APP (myc-sAPP/HA-sAPP Y ) dimers.
Pretreatment of the cells with the palmitoylation inhibitor cerulenin prior to BACE1-activity assay revealed dramatic reduction of the HA-sAPP β pull-down with myc-sAPP β in a dose dependent manner (Fig 7A, upper panel), demonstrating palmitoylation-dependent release of sAPP β -dimers. Importantly, cerulenin showed virtually no effect on total sAPPβ release as we observed little or no reduction in tot-sAPPβ levels (Fig 7A, IB:sAPPβ). Notably, cerulenin decreased both palHA-APP Y and palmyc-APP levels in a dose-dependent manner (Fig 7B). Similar results were obtained when the cells were pretreated with the palmitoylation inhibitor 2-bromopalmitate (Fig 7A and 7B, 2-BP lanes). Together, our observations revealed that APP dimers (HA-APP Y /mycAPP-dimer) released sAPP β -dimers (HA-sAPP β /myc-sAPP β -dimer) in a palmitoylation-dependent manner, while tot-sAPPβ release was independent on protein palmitoylation. This observation indicates that palAPP dimers are a better substrate for BACE1 cleavage compared to totAPP in DRMs.

Discussion
We previously reported that palAPP serves as a better BACE1-substrate than totAPP. Here we show that over 90% of palAPP is found in dimers and that palAPP dimers are 4.5-times more enriched than totAPP dimers. BACE1 cleaves palAPP dimers more efficiently than totAPP. This finding may prove to be important for the design of effective β-cleavage inhibitors of APP, as opposed to BACE1 inhibitors.
APP forms homodimers and higher-order oligomers in heterologous expression systems and in brain homogenates [23,34,35]. Dimerization via the GXXXG motif alters γ-secretase activity [24] and plays an important role in the processing of Aβ40/Aβ42 into shorter Aβ species [16]. APP dimerization through the GXXXC-motif had no effect on its BACE-cleavage, but loss of dimerization via GXXXC has been shown to inhibit the production of Aβ 42 [24]. APP dimerization via the ectodomain (E1 and E2), in contrast, appears to play significant role in APP processing [22]. E1-mediated APP dimerization has been reported to mediate APP's synaptogenic functions [26]. E1-mediated dimerization of APP has also been shown to induce APP-APP conformational changes and presynaptic enhancement, leading to Aβ40mediated APP/Gi/o-induced glutamate release [27]. Forced dimerization of APP increased Aβ production, while inducing dimerization of APP C-terminal domain via mutation of the increasing amounts of cerulenin (Cer) prior to membrane preparation. The in vitro BACE-activity assay was followed by co-IP analysis as described in Materials and Methods. Myc-sAPP β pulled down HA-sAPP β in absence of cerulenin. In presence of cerulenin, pull-down of HA-sAPP β with myc-sAPP β decreased in a dose dependent manner. 2-bromopalmitate (100μM) dramatically reduced co-IP of HA-sAPP β and myc-sAPP β . B. ABE assay of cells co-expressing HA-APP Y and myc-APP. pal HA-APP Y and palmyc-APP decreased upon cerulenin-treatment in a dose-dependent manner. C. Representation of dose-dependent decrease in HA-sAPP β pull-down with myc-sAPP β in presence of cerulenin. doi:10.1371/journal.pone.0166400.g007 transmembrane GxxxG motif reduced Aβ generation [23,24]. The ectodomain-mediated APP-dimerization appeared to play a more significant role in AD pathophysiology than its Cterminal mediated dimerization. Yet, APP-dimerization via its ectodomain is controversial because purified or overexpressed APP-ectodomain(s) are primarily monomeric, and can only form dimers at very high concentrations or in presence of heparin [22,36,37]. Although recent reports show that the GFLD (Growth factor like domain) and CuBD (Copper binding domain) of APP are essential for APP-dimerization [26], GFLD generated as a stable proteaseresistant degradation product of the APP ectodomain (18-350) is monomeric [25]. We were also unable to detect dimeric sAPP in the conditioned media of APP-overexpressing cells, as expected from earlier reports [22,36,37]. It has been predicted earlier that ectodomain-mediated APP-dimerization requires unknown hydrophobic interactions [35]. Since post-translational palmitoylation provides hydrophobicity for protein-lipid and protein-protein interactions, our discovery that palAPP forms dimers 4.5-times more efficiently than totAPP is consistent with the earlier report.
An interesting question is whether APP-palmitoylation directly mediates its dimerization. APP-dimerization initiates in the ER [19]. Palmitoylation-deficient APP mutants (APP(C 186 S) and APP(C 187 S)) show little or no dimerization and are retained in the ER [4]. It is worth noting that we often detect a~200 kD band (Fig 2B, Ã ) appearing from APP(C 186 S) mutant. APP (C 187 S) also generates similar band to much lesser extent (Fig 2B, ÃÃ ). Since both APP(C 186 S) and APP(C 187 S) mutants are predominantly ER-bound [4], we speculate that the bands are APP-dimers because ER-targeted APP (ER-APP) has been reported to generate a strong~200 kD band indicating APP-dimerization initiating in the ER [17]. In addition to strong ER-binding these mutants also lack palmitoylation. Thus the trace amount of APP-dimers from these mutants may be palmitoylation-independent APP-dimers. Increased palmitoylation of APP, either by APP(C 133 S) and APP(C 158 S) mutations or by overexpression of the palmitoylating enzyme DHHC7 that can increase palAPP levels also show concomitant increase in APPdimer levels (Figs 2 and 3). In contrast, co-IP, FRET/FLIM and BiFC analyses of cells treated with pharmacological inhibitors of palmitoylation showed robust (in co-IP assay) or moderate (FRET/FLIM and BiFC assays) decrease in APP dimerization (Figs 4 and 5). The palmitoylation inhibitors cerulenin not only reduced palAPP levels, but also inhibited APP-APP dimerization in a dose-dependent manner (Fig 4A). Moreover, palmitoylation-inhibitors specifically reduced flAPP-dimerization without affecting CTF-dimerization (Fig 6). Since palmitoyl moieties are incorporated in the ectodomain of APP, reduction of APP-palmitoylation and APPdimerization by these inhibitors strongly suggests a direct effect of APP-palmitoylation on its dimerization. However, the significance of other dimerization sites such as the growth-factorlike domain (GFLD), the copper binding domain (CuBD) or the E1 (91-111) region in APPdimerization cannot be ruled out. We may have uncovered a series of sequential events initiating with APP-palmitoylation, that promotes its ectodomain-mediated dimerization. Further studies will be required to verify this hypothesis. Since APP-dimers show differential susceptibility towards external stimuli based on the subcellular localization of the dimers [27], we predict that APP requires palmitoylation domains and/or additional domain(s) for dimerization in a spatial and temporal manner.
The role of ectodomain-mediated dimerization of APP in APP processing is still under investigation. Palmitoylation inhibitors not only abrogate APP-APP interaction, but also reduce APP-CTF α/β generation and Aβ production in cells [4]. However, our co-IP experiments failed to directly detect HA-sAPP β /myc-sAPP β -dimers in the conditioned media of cells co-expressing HA-APP Y and mycAPP (data not shown). We were also unable to detect HA-palsAPP β /myc-palsAPP β -dimers in the conditioned media. Although high levels of both myc-and HA-sAPP were released in the conditioned media (data not shown), our failure to detect both total and palmitoylated sAPP-dimers is not surprising because purified ectodomain of APP is primarily monomeric in solution [22,37]. We have reported earlier that palAPP is targeted to the detergent resistant cholesterol rich microdomains called lipid rafts. We have also shown that raft-associated palAPP serve as a better BACE1-substrate compared to totAPP [4]. Thus, we reasoned that palsAPP β dimers could be embedded in lipid-rich membranes because of the hydrophobic nature of palmitoyl-moiety. Consistently, in presence of a mild cross-linker we detected sAPP β -dimers in detergent resistant membranes (DRMs) rich in lipid rafts. Importantly, release of sAPP β -dimers, but not that of tot-sAPP β , was susceptible to the palmitoylation inhibitors cerulenin and 2-BP (Fig 7). Generation of sAPP β -dimers was decreased by cerulenin in a dose-dependent manner corresponding to the decrease in palAPP levels (Fig 7A and 7B). Decrease of sAPP β -dimer formation in presence of palmitoylation inhibitors suggests that sAPP β -dimers were formed from pal-APP dimers in DRMs. A direct detection of palsAPP β dimers in vivo is necessary for further studies on the role of palAPP dimerization in APP processing.
BACE1-mediated β-cleavage of APP is the rate-limiting step for Aβ generation. Unfortunately, development of BACE1 inhibitors for AD treatment is difficult due to the promiscuity of BACE1 for its substrates. Moreover, APP is primarily a substrate for α-cleavage producing non-amyloidogenic peptides. Relocalization of APP into cholesterol-rich lipid rafts shifts APP towards amyloidogenic cleavage by BACE1 (reviewed in [38]). We have reported that palAPP is enriched in lipid rafts favoring β-over α-cleavage [4]. We also reported that lipid raft-associated palAPP is a better substrate in vitro and in vivo. Now we show that palAPP forms stronger dimers than totAPP primarily in cis-orientation (Fig 1D), which is considered more favorable dimer orientation for β-cleavage. Although, a direct proof that BACE1 cleaves palAPP-dimers more efficiently than non-palAPP requires further investigation, our data strongly indicate that palAPP-dimers are better substrates for β-cleavage in DRMs than totAPP. It is encouraging that a recent High Throughput Screen of nearly 77,000 compounds identified two small molecule modulators of totAPP dimerization that may lower sAPP β levels without affecting αor γ-cleavage [39]. Our data indicate that small molecules designed to specifically reduce palAPP-dimer formation would be potent inhibitors of APP's β-cleavage in the brains of patients affected by AD.
Our finding that APP-palmitoylation is a novel contributor to APP-dimerization adds the palmitoylated cysteines (Cys 186 and Cys 187 ) to the previously identified multiple dimerization interfaces in APP. The multi-fasceted dimerization of APP provides a potential for different conformations of the protein, leading to different effects on β-cleavage and Aβ generation. Thus, the dimerization domain plays an essential role in predicting whether APP-dimers increase or decrease Aβ production. Palmitoylation targets palAPP to the detergent resistant lipid raft membranes (DRMs), and palAPP-dimers exhibit β-cleavage in the DRMs.
A complete loss of APP palmitoylation by 100μg/ml cerulenin resulted in pronounced loss of APP-APP dimerization, but not that of APP-CTFs (Fig 6B). This points to the fact that palAPP is predominantly dimerized. Accordingly, palAPP was found to form~4.5 fold increased dimers compared to totAPP. In addition, palAPP exclusively formed cis-oriented dimers, which is a preferred orientation of APP-dimers for β-cleavage. We also found that palAPP dimers undergo β-cleavage in lipid-rich DRMs, which are one of the critical microdomains for amyloidogenesis. In conclusion, majority of palAPP appears to form cis-dimers undergoing β-cleavage. Although we cannot entirely exclude the possibility of a pool of monomeric palAPP, our data overwhelmingly supports the conclusion that palAPP in its cis-dimerized form is a potential drug target for AD treatment. Identification of specific small molecule modulators for palAPP-dimers may become an effective targeted therapeutic strategy to lower Aβ in AD brains.

Materials and Methods
Cell culture and transfection CHO APP , and naïve CHO cells were maintained and transfected with expression plasmids as described before [40,41]. CHO cells stably expressing APP (CHO APP ) were maintained in DMEM containing 10% serum supplemented with G418. Typically, 1.2 X 10 6 cells were used for transfection and palmitoylation assays.

Maintenance of immortalized hNPC cell line ReNcell VM (ReN cells)
ReN cells were maintained as described by Kim, Y.H. et all [32]. Briefly, ReN cells (Millipore) were maintained in Proliferation medium (484.5 ml DMEM/F12 (Gibco/Life Technologies) with 0.5 ml of heparin (2 mg/ml stock, STEMCELL Technologies), 10 ml of B27 (Life Technologies) 5 ml of 100X penicillin/streptomycin/amphotericin B (Lonza), 80 μl of bFGF stock and 100 μl of EGF stock) on Matrigel (Sigma-Aldrich) coated flasks at 37˚C CO 2 incubator. For differentiation the media were changed to Differentiation media, which is Proliferation media containing no growth factors, bFGF or EGF. The cells were maintained in Differentiation media for~6 days to obtain neuronal structure prior to co-IP assays.

Lentiviral infection of ReN cells
To transfect the ReN cells with the lentiviral constructs containing APP mGFP and APP mCherry expression plasmids (very generous gifts from Dr. Inna Slutsky, Sackler Faculty of Medicine, Tel Aviv University, Israel), we obtained the lentiviral vectors packaged by MGH viral core fascility. 1 X 10 6 viral particle was used to infect 85% confluent proliferating ReN cells in 6-well dishes. After 24 h the cells were washed three times to stop the infection. The expression of the infected genes was confirmed by mGFP or mCherry GFP expression by fluorescence microscopy and western blot analysis. To probe APP mGFP expression, anti-mGFP (Abcam, USA) antibody was used in immunostaining that specifically detects mGFP epitope. For APP mCherry detection we used anti-mCherry antibody from Abcam.

FACS enrichment of the transfected ReNcells
The infected ReNcells were washed with PBS and then incubated with Accutase (Millipore) for 5 min. The cell pellets were resuspended in PBS supplemented with 2% serum replacement solution (Life Technologies) and 2% B27, and then passed through a cell strainer filter (70 mm Nylon, BD Biosciences). The cell concentrations were adjusted to~200,000 cells per ml and then enriched by using FACSAria cell sorter (MGH core facility, Charlestown, MA). GFP and/ or mCherry channels were used to detect the expression of the transfected genes in the individual cells. The sorted/enriched cells were maintained in normal proliferation media. To sort CHO cells co-expressing APP mGFP +APP mCherry similar procedure was used after 24h transfection of these cells with APP mGFP and APP mCherry expression plasmids using Effectene reagents following the manufacturer's protocol.

Modified ABE assay (mABE assay)
This assay is based on a modification of the ABE assay as described earlier [4,13]. 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 NH 2 OH (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 NH 2 OH served as negative control. The beads were washed and immunoblotted with Steptavidin-HRP (Cell Signaling) to detect palmitoylation.

APP and palAPP dimerization assays (Co-IP assay)
Cells were co-transfected with expression plasmids encoding HA-APP Y and APP V5 were lysed and subjected to pulled-down assay where HA-APP Y was precipitated with anti-HA antibody followed by immunoblotting with anti-V5 antibody to detect co-immunoprecipitation of APP V5 . Reverse pull-down was also performed where APP V5 was first pulled-down using anti-V5 antibody followed by probing with anti-HA antibody to detect co-IP of HA-APP Y . Co-IP assays were performed either in absence or presence of palmitoylation inhibitors 2-bromopalmitate or cerulenin to test APP-APP dimerization Co-immunoprecipitation of HA-APP Y with APP V5 suggested APP-APP dimerization. Similar co-IP assay was performed on cells coexpressing APP mGFP and APP mCherry . Antibody specific for mGFP (anti-GFP) was used to pulldown APP mGFP prior to probing with anti-mCherry antibody to detect APP mGFP -APP mCherry dimerization.
To detectermine palAPP-dimerization, co-IPd samples from cells co-expressing HA-APP Y and APP V5 were subjected to mABE assay as described above. Streptavidin-HRP detected pal-HA-APP Y and palAPP V5 at molecular weights~150 and~100, respectively. Identification of both bands indicated palAPP-dimerization.

Assays for trans-dimerization of APP and of palAPP
To test for APP trans-dimerization, co-IP assay was performed in mixed cell culture where CHO cells expressing HA-APP Y were co-cultured with Neuro-2A cells expressing APP V5 . Briefly, the co-cultured cells were grown to confluency prior to incubation without or with 1mM cell-impermeable crosslinker 3.3'-dithiobis[sulfosuccinimidyl propionate] (DTSSP, Sigma) at 4˚C to crosslink proteins at the cell surface as described by Soba et al. [20]. After lysis, APP V5 was pulled-down using an anti-V5 antibody prior to probing the precipitate with an anti-HA antibody to detect co-IP with APP V5 indicating APP-dimerization. We next subjected the co-IPed APP forms to mABE assay to detect palHA-APP Y and/or palAPP V5 .

FLIM Imaging and analysis
FLIM imaging was carried out as described by Fogel, H. [27]. For imaging in CHO cells, cells were transfected with expression plasmids encoding mEGFP-tagged APP alone or together with expression plasmid encoding mCherry-tagged APP (APP mGFP or APP mGFP + APP mCherry , respectively). For imaging in neuronal progenitor RenVM, the cells were incubated with lentiviral particles containing expressionng vectors for APP mGFP or APP mGFP + APP mCherry (the lentiviral vectors were gifts from Dr. Inna Slutsky, Tel Aviv University, Tel Aviv, Israel). Cells were sorted by a cell sorter to obtain homogenous cultures of APP mGFP -expressing or APP mGFP + APP mCherry -expressing cells. The cells were fixed and the FLIM analysis was performed as described previously [45]. Briefly, pulsing Chameleon Ti:Sapphire laser (Coherent Inc., Santa Clara, CA) was used to excite GFP donor fluorophore (two-photon excitation at 780 nm wavelength). The baseline lifetime (t 1) of the mGFP fluorophore (APP mGFP ) was measured in the absence of the mCherry acceptor fluorophore (APP mCherry ) (negative control, FRETabsent). Donor fluorophore lifetimes were recorded using a high-speed photomultiplier tube (MCP R3809; Hamamatsu, Bridgewater, NJ) and a fast time-correlated single-photon counting acquisition board (SPC-830; Becker & Hickl, Berlin, Germany). In the presence of the acceptor fluorophore, if the two fluorophores are <5-10 nm apart, FRET occurs and the donor fluorophore lifetime (t 2 ) shortens. The acquired FLIM data were analysed using SPC image software (Becker & Hickel, Berlin, Germany) to fit the raw data from each pixel to multi-exponential fluorescence decay curves to calculate mGFP fluorescence lifetimes. The degree of donor life time T m, was calculated (T m = (t1-t2)/t1), where t is the fluorescence lifetime of the donor fluorophore (mGFP) measured in nanoseconds after pulse. To represent a "non-FRETing" population with a longer lifetime and a "FRETing" population with a shorter lifetime, the fluorescence lifetimes are plotted in a bar-graph as described by Fogel, H. et al [27].

BiFC assay
Bio-immunofluorescence of split GFP constructs (BiFC) assay was performed as described before [17] with modification. Briefly, naïve CHO cells were transiently transfected (Lipofectamine 2000, Invitrogen) with plasmids encoding split-GFP APP, APP(1-10) and APP (11). The split-GFP plasmids were generous gifts from Dr. Claus U. Pietrzik from Department of Pathobiochemistry, University Medical Center of the Johannes Gutenberg-University Mainz, Germany. 24hs after transfection cells were sorted by a fluorescence activated cell sorter (FACS). Cells were sorted based on the fluorescence intensity to obtain a homogenous population of cells with equal BiFC signals. Approximately 100,000 cells were plated on coverslips on 12-well plates, and grown for 18 h. Cells were then treated with increasing amounts of cerulenin (0-100 μM) for 3 h before fixing in 4% paraformaldehyde in 1x PBS at room temperature (RT) for 30 min. Cells were washed with 1 x PBS for three times before mounting the coverslips on DAPI containing mounting media (ProLong Gold antifade with DAPI, Life Technologies). Fluorescence microscopy was performed under Nikon confocal microscope using 40X objective. The fluorescence intensity was measured by ImageJ software.

Detergent Resistant Membrane (DRM) preparation
DRMs were purified as described in Navarro-Lerida et al [46] with modification. Briefly,~2.5 X 10 5 cells were resuspended in 5 volume (weight:volume) HEPES buffer (50 mM HEPES, pH 7.4, 0.15 M NaCl, 1 mM PMSF plus 0.5% Triton X-100) at 4˚C. Cells were homogenized by passing through a syringe (0.5x16 mm) on ice for 10 times. The homogenate was brought up to 4 ml by adding 2 ml 80% sucrose in HEPES, and placed at the bottom of a Beckman SW40 Ultraclear tube. The discontinuous sucrose gradient (40-30-5%) was formed by sequentially loading 4 ml 30% sucrose and 4 ml 5% sucrose in HEPES. Cells fractions were separated by centrifugation at 200,000 g for 18 h in a SW40 rotor (Beckman) at 4˚C. A light, scattered band confined to the 5-30% sucrose interface was observed that contained most flotillin, which is a subcellular marker for lipid raft-rich membranes. This fraction was collected as detergent resistant membranes or DRMs.

In vitro BACE-activity assay
To test sAPPβ-sAPPβ dimers from APP-APP dimers upon BACE1 activity, we collected detergent resistant membrane (DRMs) from cells co-expressing HA-APP Y and mycAPP. DRMs were mixed with 50mM Na-acetate buffer of pH 4 containing complete protease inhibitor mixture (Roche Applied Science), the aspartic protease inhibitor pepstatinA (10 M; Roche Applied Science), and the γ -secretase inhibitor N -[N -(3,5-difluorophenacetyl-L -alanyl)]-S phenylglycine t -butyl ester (10M; Calbiochem). BACE-activity was measured by incubating the mixture at 37˚C. After 1 h incubation, the reaction was terminated by bringing the pH to 7.6. The samples were centrifuged at 100,000 g for 1 h to remove membranes, and the supernatant were subjected to immunoprecipitation with anti-myc antibody to IP myc-sAPPβ. The precipitate was probed with anti-HA antibody to detect co-IP of HA-sAPPβ suggesting sAPPβ-sAPPβ dimerization. DRMs were also collected from HA-APP Y +mycAPP expressing cells after treatment with increasing amounts of cerulenin or with 50 μM 2-bromopalmitate (2-BP). These DRMs were also subjected to in vitro BACE1-activity assay followed by co-IP experiment as described above.
Pulse chase assay CHO APP cells were were metabolically labeled with 100 μM chemical palmitic acid probe, alkylene palmitic acid (Alkyl-C16; Invitrogen) as described previously [4]. After 6 h labeling cells were washed once with DMEM media, and incubated for 0.5-6 h at 37˚C in DMEM supplemented with penicillin/streptomycin, 3.6 mg/ml fatty acid free BSA and 100 μM unlabeled palmitic acid (Sigma), as described before [47](Chen, C. and Manning, D. 2000). Cells were collected at 0.5, 1, 2, 3 and 6 h prior to IP with C66 antibody and subsequent labeling with TAMRA via Click-iT chemistry as we did before [4]. IPed samples were probed with anti-TAMRA antibody to detect palAPP.
Half-life assay CHO cells were transfected with BiFC APP (APP(1-10) expression plasmid. 24 h after transfection, cells were treated with 100 μM cyclohexamide and MG132 for 0-6 h as described before [48]. After treatment cells were lysed and equal amount of lysates were subjected to SDS-PAGE and immunobloted with C66, anti-mGFP, or anti-actin antibodies to detect the levels of endogenous APP, APP(1-10) and actin.

Aβ 40 and Aβ 42 determinations
For Aβ determination, CHO APP cells were grown in six-well plates (Becton Dickinson Labware) till 80-90% confluency. After washing the cells once with PBS the cells were layered with 1 ml media for 6 h before adding increasing amounts of cerulenin (0-100 μg/ml). After 6 h of cerulenin treatment the conditioned media were collected and immediately subjected to Aβ ELISA assay. The levels of secreted Aβ 40 and Aβ 42 in the condition media were quantified by standard sandwich ELISA using the commercially available Aβ ELISA kit (Wako Pure Chemical) as before [4]. Aβ levels (in pmol/L) were plotted against cerulenin concentrations.

Statistical analysis
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 s.e.m. Significance was assessed at Ã p<0.05 and ÃÃ p<0.01. The lysates were also probed with anti-actin antibody. B. CHO APP cells were metabolically labeled with chemically-labeled palmitic acid (Alkyl-C16) for 6 h followed by chasing with unlabeled free palmitic acid for 0.5-6 h, as indicated. After immuoprecipitation of APP with C66 antibody from the labeled cells, the precipitates were subjected to Click-iT assay to incorporate TAMRA on Alkyl-C16. Immunobloting the precipitates with anti-TAMRA antibody detected Alkyl-C16 labeled APP (palAPP) and showed half-life of palAPP to be~3 h. The blot is the representation of duplicate experiments. (TIF)

S3 Fig. Fluorescence-activated cell sorting and differentiated ReN cells.
A. ReN cells expressing APP mGFP +APP mCherry via Lentiviral infection were subjected to FACS analysis at the MassGeneral Hospital core fascility (MGH. Charlestown). Only 8.1% cells expressed both APP mGFP +APP mCherry (P3 polulation) compared to 14% expressing APP mCherry (P4 population) and 12.8% expressing APP mGFP (P5 population) alone. B. After sorting the P3 population from the infected cells (panel a), the cells were differentiated into neuronal cells (panel b) prior to co-IP analysis. (TIF)