Two Endoplasmic Reticulum (ER)/ER Golgi Intermediate Compartment-based Lysine Acetyltransferases Post-translationally Regulate BACE1 Levels*

We have recently identified a novel form of post-translational regulation of BACE1 (β-site amyloid precursor protein-cleaving enzyme 1), a membrane protein that acts as the rate-limiting enzyme in the generation of the Alzheimer disease amyloid β-peptide. Specifically, nascent BACE1 is transiently acetylated in seven lysine residues clustered in a highly disordered region of the protein that faces the lumen of the endoplasmic reticulum (ER)/ER Golgi intermediate compartment (ER/ERGIC). The acetylation protects the nascent protein from degradation by PCSK9/NARC-1 in the ERGIC and allows it to reach the Golgi apparatus. Here we report the identification of two ER/ERGIC-based acetyltransferases, ATase1 and ATase2. Both proteins display acetyl-CoA:lysine acetyltransferase activity, can interact with and acetylate BACE1, and display an ER/ERGIC localization with the catalytic site facing the lumen of the organelle. Both ATase1 and ATase2 regulate the steady-state levels of BACE1 and the rate of amyloid β-peptide generation. Finally, their transcripts are up-regulated by ceramide treatment. In conclusion, our studies have identified two new enzymes that may be involved in the pathogenesis of late-onset Alzheimer disease. The biochemical characterization of the above events could lead to the identification of novel pharmacological strategies for the prevention of this form of dementia.

The efficiency of folding, conformational maturation, and molecular stability of nascent membrane and secretory proteins is greatly affected in the endoplasmic reticulum (ER) 2 by post-translational events that modify, either temporally or definitively, the protein (1). One of the best characterized forms of transient modification is the attachment of a glucose residue to improperly folded nascent glycoproteins by the UDP-glucose:glycoprotein glucosyltransferase (2,3). This transient event regulates the interaction of the nascent protein with the chaperone calnexin and its ability to leave the early secretory pathway. The mechanism works in such a way that successfully folded proteins dissociate from the calnexin/calreticulin cycle and advance toward the Golgi apparatus, whereas misfolded intermediates are directed toward the ER-associated degradation system (1,4).
We have recently reported the identification of a novel form of post-translational regulation of BACE1 (␤-site APP-cleaving enzyme 1) (5), a membrane protein that acts as the rate-limiting enzyme in the generation of the Alzheimer disease amyloid ␤-peptide (A␤) from the amyloid precursor protein (APP). Specifically, nascent BACE1 is transiently acetylated in seven lysine residues clustered in a highly disordered region of the protein that faces the lumen of the ER and ER Golgi intermediate compartment (ERGIC). The lysine acetylation of nascent BACE1 regulates its ability to advance toward the Golgi apparatus, where a Golgi-based deacetylase removes the acetyl groups. Nonacetylated intermediates of the nascent protein are retained in the early secretory pathway and degraded by a mechanism that involves the serine protease PCSK9/NARC-1 (proprotein convertase subtilisin kexin-type 9/neural apoptosis-regulated convertase-1) (6). The acetylation of nascent BACE1 in the lumen of the ER and/or ERGIC requires a membrane transporter that translocates acetyl-CoA, the donor of the acetyl group, from the cytoplasm to the lumen of the ER, and one or more ER/ERGIC-based acetyl-CoA:lysine acetyltransferases (5).
The acetylation/deacetylation process is tightly regulated by the lipid second messenger ceramide (5,7,8) and is under the control of the general aging program mediated by the insulinlike growth factor 1 receptor (IGF-1R) (9,10). Ceramide, the last output of the above pathway, regulates both efficiency of acetylation in the lumen of the ER/ERGIC and rate of deacetylation in the Golgi apparatus (5). As a result, the hyperactivation of ceramide signaling that occurs during normal aging (8) or in progeroid-like animal models (9) leads to increased acetylation and steady-state levels of BACE1 and to increased production of A␤ (8,9). The relevance of the above events for AD neuropathology is stressed by the fact that aging is the single most important risk factor for late-onset AD and that AD patients have very high levels of ceramide in the brain, when compared with age-matched controls (reviewed in Ref. 10).
Here we report the identification of two ER/ERGIC-based acetyltransferases, which we named ATase1 and ATase2. Both proteins display acetyl-CoA:lysine acetyltransferase activity in vitro and under native conditions can interact with and acetylate BACE1 in vitro and in vivo, and show an ER/ERGIC localization. They both regulate the steady-state levels of BACE1 and the generation of A␤, and are up-regulated by ceramide treatment of cultured cells.

EXPERIMENTAL PROCEDURES
The following experimental procedures, which have been extensively described in our previous work, are in the supplemental Methods: cell cultures and cell treatment; cell extraction, immunoprecipitation, and affinity purification; subcellular fractionation; A␤ determination; and siRNA treatment.
Pixel densities (for signal area) of scanned images were calculated with Adobe Photoshop; densitometry (for signal-density) was analyzed with the EpiChemi3 Darkroom TM (UVP Bioimaging Systems) using Labworks Image Acquisition and Analysis Software 4.5. Additionally, samples were also imaged and quantified using the Odyssey Infrared Imaging System (LI-COR Biosciences, Lincoln, NE). For infrared imaging, IRDye 680 conjugated goat anti-rabbit IgG (catalog number 926-32221, LI-COR Biosciences) and IRDye 800CW conjugated goat anti-mouse IgG (catalog number 926-32210, LI-COR Biosciences) were used instead of horseradish peroxidase-conjugated secondary antibodies. For quantification, values were normalized to appropriate loading controls.
Acetyl-CoA:Lysine Acetyltransferase Activity-For the analysis of the acetyltransferase activity, we employed affinity-purified BACE1 as acceptor of the acetyl group and [ 3 H]acetyl-CoA as donor of the acetyl group. The acetyl-CoA:lysine acetyltransferases (ATase1 and/or ATase2) were provided either as pure enzymes following affinity purification with the ProFound kit (see supplemental Methods) or as membrane extracts of stably transfected CHO cells. The reaction was performed in 150 -200 l of acetylation buffer (50 mM Tris-HCl (pH 8.0), 0.1 mM EDTA, 1 mM dithiothreitol, 10% glycerol, 20 M acetyl-CoA) and let run for 1 h at 30°C. The reaction was stopped by adding an equal volume of ice-cold buffer and immediate immersion in ice; BACE1 was immunoprecipitated with an anti BACE1 N-terminal monoclonal antibody and then counted on a liquid scintillation counter. The experimental controls are described in the appropriate figure legends.
Additionally, the acetyltransferase activity of affinity-purified ATase1 and ATase2 was assayed with commercially available colorimetric (catalog number K332-100, BioVision Inc.) and fluorescent kits (catalog number 10006515, Cayman Chemicals). The assays were performed as recommended by the manufacturer.
Real Time PCR-Total RNA from H4 (human neuroglioma) and SH-SY5Y (human neuroblastoma) cells was extracted and purified using the RNeasy plus mini kit (Qiagen). cDNA was synthesized using the SuperScript double-stranded cDNA synthesis system (Invitrogen) and then PCR-amplified. Primers used for RT-PCR were as follows: GAPDH, forward 5Ј-GAAG-GTGAAGGTCGGAGTC-3Ј, reverse 5Ј-GAAGATGGTGAT-GGGATTTC-3Ј; ATase1, forward 5Ј-CGATTACTGAAGCT-GCCTCGA-3Ј, reverse 5Ј-GGTTTTTTGGCAAGGAACCA-C-3Ј; and ATase2, forward 5Ј-TCCTTGCCAAAAAACCCTGG-3Ј, reverse 5Ј-ATGCCCACCACCTTCTCTTCA-3Ј. RNA quantification was performed by real time PCR in a ABI PRISM 7000 sequence detection system using SYBR Green PCR master mix (Applied Biosystems). Amplifications were generated at 2 min at 52°C and 3 min at 95°C, followed by 40 cycles of denaturations at 95°C for 15 s, annealing, and synthesis (30 s at 61°C and 30 s at 72°C). ⌬⌬Ct values (14) were normalized with those obtained from the amplification of GAPDH and were expressed as foldchange over control. The assay was repeated three times, with each assay containing triplicate reactions and with each assay including an independent amplification of the probes.
Statistical Analysis-Results are always expressed as mean Ϯ S.D. of the indicated number of determinations. The data were analyzed by analysis of variance and Student's t test comparison, using GraphPad InStat3 software. Statistical significance was reached at p Ͻ 0.05.

RESULTS
Identification of ATase1 and ATase2 as Two Putative ER/ER-GIC Acetyl-CoA:Lysine Acetyltransferases-Our studies previously demonstrated the presence of acetyl-CoA:lysine acetyltransferase activity in the ER/ERGIC. They also demonstrated that the catalytic site of the putative ER/ERGIC-based acetyltransferase(s) faced the lumen of the organelle (5). To identify the corresponding gene(s), we searched in the human genome for proteins displaying structural similarities to the histone acetyltransferase catalytic domain. The search yielded 46 possible candidates, 2 of which were predicted to function in the ER/ERGIC. The candidates had unknown biochemical function and were initially named Camello-like 2 (accession number NM_016347; also called putative N-acetyltransferase 8B) and Camello-like 1 (accession number NM_003960; also called putative N-acetyltransferase 8). They appear to have orthologs in Xenopus laevis (15) but not in Caenorhabditis elegans or Drosophila melanogaster.
The biochemical activity of both proteins was explored, and based on the results reported here they were re-named acetyltransferase 1 (ATase1) and 2 (ATase2) to indicate their biochemical function rather than the domain organization. Both ATase1 and ATase2 have a high probability of being ER/ ERGIC-resident proteins (k-NN prediction using the PSORT II software of the University of Tokyo, Japan and available on line) (16,17). Alignment of the amino acid sequences of ATase1 and ATase2 reveals an 88% sequence identity between the two proteins ( Fig. 1A); the Kyte-Doolittle hydrophilicity/hydrophobicity plot (Fig. 1B) indicates the presence of one single hydrophobic segment, corresponding to the membranespanning domain. In both cases the putative catalytic site is situated in the C-terminal domain and is predicted to face the lumen of the ER/ERGIC.
ATase1 Is an ER/ERGIC Resident Acetyl-CoA:Lysine Acetyltransferase-We initially targeted ATase1, which was cloned and inserted into a mammalian expression vector containing a Myc tag at the C terminus, to allow subcellular localization and affinity purification for the in vitro studies. Stable transfection in CHO cells yielded several colonies overexpressing ATase1 ( Fig. 2A). Membrane extracts from nontransfected control and ATase1-expressing cells were analyzed in vitro for acetyl-CoA:lysine acetyltransferase activity. The assay was performed in the presence of affinity-purified BACE1, which served as acceptor of the acetyl group, and radiolabeled acetyl-CoA, which served as donor of the acetyl group. Fig. 2B shows that the expression of ATase1 increased the acetyltransferase activity recovered from membrane extracts by ϳ2-fold. No activity was observed when the assay was performed in the absence of extracts (source of the enzymatic activity) or in the presence of extracts that had been boiled prior to the incubation (Fig. 2B).
Subcellular fractionation studies indicated a predominant ERGIC localization that extended to ER, but not to Golgi, fractions (Fig. 2C), which is consistent with our previous localization of the ER/ERGIC-based acetyl-CoA:lysine acetyltransferase activity (5). The subcellular localization of ATase1 overlaps with the compartments where the acetylation of nascent BACE1 occurs and the nonacetylated mutants of BACE1 are retained (5). In fact, we have previously demonstrated that the Lys-to-Ala (BACE1 Ala ) and the Lys-to-Arg (BACE1 Arg ) mutant forms of BACE1 cannot be acetylated in vivo and in vitro, are retained in the ER/ERGIC system, and rapidly degraded by a proteasome-independent system (5, 6). Next, we analyzed the acetyl-CoA:lysine acetyltransferase activity of the individual fractions from the gradient shown in Fig. 2C and found that the distribution pattern of ATase1 completely overlapped with the acetyl-CoA:lysine acetyltransferase activity (Fig. 2D). Finally, when compared with control (nontransfected) cells, ERGIC fractions from ATase1-expressing cells displayed increased (ϳ2-fold) acetyltransferase activity (Fig. 2E).
The above results indicate that ATase1 is localized in the compartments where the acetylation of nascent BACE1 normally occurs. They also indicate that the overexpression of ATase1 results in increased acetyl-CoA:lysine acetyltransferase activity. However, they do not rule out the possibility that ATase1 is only a co-factor in the reaction of lysine acetylation. To address this issue, we used affinity-purified ATase1 (instead of membrane extracts) for the in vitro acetylation of BACE1. As control, we used BACE1 Arg , which does not act as acceptor of the acetyl groups in vitro or in vivo (5). Fig. 3A shows that purified ATase1 was able to acetylate BACE1 WT but not BACE1 Arg proving that it is an acetyl-CoA:lysine acetyltransferase. The results obtained with BACE1 Arg are particularly important because they indicate that ATase1 can only acetylate the lysine residues that are found acetylated under physiological conditions (5).
Next, we incubated ATase1 with BACE1 in the presence of nonradiolabeled acetyl-CoA. Both BACE1 and ATase1 were purified by affinity chromatography prior to the incubation. BACE1 was then immunoprecipitated at the end of the reaction and analyzed by classical immunoblotting. In the absence of ATase1 only the immature form of BACE1 could be detected with an anti-acetylated lysine antibody (  domain of BACE1 or the Myc tag (Fig. 3B, middle and right panels, indicated by *). The band was never observed immediately after purification of BACE1 and appeared only after the 6-h incubation required for the in vitro acetylation assay, suggesting that it corresponds to the cleaved N-terminal ectodomain of BACE1 and most likely results from in vitro autocatalytic cleavage, as described previously (18).
Surprisingly, the anti-acetylated lysine antibody recognized a band of ϳ28 kDa, which corresponds to ATase1 (Fig. 3B, left panel,  ϩATase1). In fact, the same band could be observed with anti-Myc antibodies (Fig. 3B, right panel, ϩATase1) but not with antibodies against the N-terminal (data not shown) or C-terminal domain (Fig. 3B, middle panel) of BACE1. Therefore, these results indicate that ATase1 can interact with BACE1 in vitro and can undergo acetylation in one or more lysine residues, most likely through an autocatalytic mechanism (discussed below).
ATase2 Is an ER/ERGIC Resident Acetyl-CoA:Lysine Acetyltransferase-We next cloned ATase2 into a mammalian expression vector containing a Myc tag at the C terminus. This construct was used to generate cells stably expressing ATase2 and ATase1 ϩ ATase2 (Fig. 4A). Similarly to ATase1, the expression of ATase2 alone increased (by ϳ3-fold) the acetyl-CoA:lysine acetyltransferase activity recovered from total membrane extracts (Fig. 4B). In addition, the subcellular distribution of ATase2 (Fig. 4C), even though slightly different, appeared to overlap with ATase1 (Fig. 2C) and with the normal distribution of the acetyl-CoA:lysine acetyltransferase activity (Fig. 2D). Finally, ER/ERGIC fractions generated from ATase2-expressing cells displayed increased acetyltransferase activity, when compared with corresponding fractions from control nontransfected cells (Fig. 4D). As with ATase1-expressing cells (Fig. 2D), no activity was found in fractions corresponding to the Golgi apparatus (data not shown, see also Ref. 5).
When affinity-purified ATase2 was incubated with BACE1 and nonradiolabeled acetyl-CoA, we found a marked increase in the acetylation of both biosynthetic intermediates of BACE1 (Fig. 5A, left  panel). Interestingly, the acetylation pattern produced by ATase2 seems somehow different from the one observed with ATase1 (compare Fig. 5A to Fig. 3B) suggesting potential differences in substrate recognition activities of the two enzymes. Similarly to ATase1, the immunoprecipitation of BACE1 resulted in the pulldown of ATase2, suggesting physical interaction, at least in vitro (Fig. 5A, right panel). However, in contrast to ATase1, we could not clearly resolve a band corresponding to ATase2 with an anti-acetylated lysine antibody (Fig. 5A, left  panel). This occurred even though ATase2 could clearly be identified with an anti-Myc antibody (Fig. 5A, right panel). These results might indicate potential differences in the biochemical properties and functional regulation of ATase1 and ATase2 (discussed below). Finally, incubation of ATase2 with BACE1 and radiolabeled acetyl-CoA resulted in acetylation of BACE1 WT but not BACE1 Arg (Fig. 5B) proving that ATase2 can only acetylate the lysine residues that are found acetylated under physiological conditions (5).
We next used a commercially available colorimetric assay that employs a recombinant peptide corresponding to the N-terminal tail of the histone protein, as acceptor of the acetyl group, and a highly concentrated nuclear histone acetyltransferase (HAT) preparation, as enzyme. The assay was partially modified to employ our affinity-purified ATases, instead of HAT, as source of the enzymatic activity. Fig. 5C shows that both ATase1 and ATase2 displayed robust acetyl-CoA:lysine acetyltransferase activity, which was in the same range of that observed with HAT. Finally, when ATase1 and ATase2 were included together in the assay, we detected a clear additive effect indicating that the two enzymes can act independently of each other, at least in vitro. It remains to be determined whether the additive effect is caused by their ability to target different lysine residues of the substrate-peptide or by the presence of excess substrate in the assay. The ability of ATase1 and ATase2 to acetylate an appropriate substrate was also confirmed by using a fluorescent assay that employs the histone H3-(5-23) peptide as acceptor and acetyl-CoA as donor (supplemental Fig. 2).
Catalytic C-terminal Domain of ATase1 and ATase2 Faces the Lumen of the ER/ERGIC System-We have previously shown that the transient acetylation of the N-terminal and globular domain of nascent BACE1 occurs in the lumen of the ER/ERGIC system (5). We also demonstrated that the acetyltransferase activity responsible for the lysine acetylation of the nascent protein was positioned in the luminal face of the secre-tory compartment (5). The hydrophilic/hydrophobic plot (Fig.  1B) together with the apparent topology (using the MTOP prediction model (19)) of ATase1 and ATase2 predicts the C-terminal domain, which contains the acetyl-CoA:lysine acetyltransferase homology domain, to face the lumen of the ER/ERGIC system.
We decided to assess whether this was indeed the case by assaying in vitro the acetyl-CoA:lysine acetyltransferase activity recovered from ERGIC vesicles of ATase1-or ATase2-expressing cells under two different experimental conditions as follows: in the presence or absence of mild concentrations (0.2-0.5%) of Triton X-100 (Fig. 6A). When prepared, vesicles are sealed and of the same membrane topographical orientation as in vivo (5, 20 -22). Therefore, under normal conditions (Fig. 6A, Sealed) the acetylation of affinity-purified BACE1 can occur only if the catalytic site of the enzyme (ATase1 or ATase2) is facing the outside/cytosolic face of the vesicles. Conversely, if the catalytic site of the enzyme resides in the luminal face of the vesicles, the reaction will only occur in the presence of mild concentrations of detergent, which will allow access of both the acceptor (purified BACE1) and donor (radiolabeled acetyl-CoA) to ATase1 or ATase2 (Fig. 6A, Opened). Fig. 6B clearly shows that the acetyltransferase activity of native ATase1 and ATase2 could only be observed in the presence of Triton X-100, indicating that the catalytic site is facing the lumen of the organelle in its native conditions. The activity recovered in the absence of Triton X-100 was in the normal range of latency (3-5%) observed with the glucose-6-phosphatase and sialyltransferase methods (see supplemental Methods).
Next, we took advantage of the fact that both ATase1 and ATase2 have a Myc tag at the C terminus, which is predicted to face the lumen of the ER/ERGIC system. Therefore, the above vesicles were incubated with an anti-Myc antibody covalently attached to aldehyde-activated agarose beads (ProFound system) for immunoprecipitation. The experiment was performed under both the "opened" and "sealed" conditions ( Fig. 6A) to determine the topology of the enzymes. Fig. 6C shows that ATase1 and ATase2 could be immunoprecipitated only when the vesicles were used under the opened condition, which allowed the anti-Myc antibody to interact with the C-terminal Myc tag. Therefore, when taken together, the above results indicate that both ATase1 and ATase2 have the predicted topology, with the catalytic C-terminal domain facing the lumen of the ER/ERGIC system, where the lysine acetylation of nascent BACE1 normally occurs.
ATase1 and ATase2 Influence the Steady-state Levels of BACE1 and the Generation of A␤-The results shown in Figs. 3B and 5A indicate that both ATase1 and ATase2 can be coimmunoprecipitated with BACE1 suggesting physical interaction. However, this experiment was performed after co-incubation of affinity-purified ATases and BACE1 in vitro. Therefore, we decided to assess whether the physical interaction between the enzyme (ATases) and the substrate (BACE1) also occurred in vivo. For this experiment we used cell lines that had been transfected with ATase1-myc or ATase2-myc, but not with BACE1-myc, so that we could clearly analyze the possible interaction with endogenous BACE1. ATase1 and ATase2 were purified (Fig. 7A, upper panel) with the anti-Myc-aldehyde-ac- Reaction was stopped by lowering the temperature; BACE1 was then purified again and analyzed on a scintillation liquid counter. The same assay was repeated with BACE1 Arg , a mutant form of BACE1 where the lysine residues that serve as acceptors of the acetyl groups have been mutated to arginine (5). Results are the average (n ϭ 6) Ϯ S.D. Asterisk indicates statistical significance (p Ͻ 0.005). B, affinity-purified ATase1 and BACE1 were co-incubated in the presence of acetyl-CoA (50 M) for 6 h at 30°C. Reaction was stopped by lowering the temperature. BACE1 was then immunoprecipitated with a biobead cross-linked antibody specific for the N-terminal domain of BACE1 and analyzed by Western blotting with the indicated antibodies (Ac. Lys., antiacetylated lysine; BACE1 C-term., anti-BACE1 C-terminal domain; Myc, anti-Myc). Numbers indicate the different species of BACE1 (see supplemental Fig.  1 for detailed explanation). Asterisk indicates a band that could be recognized with an antibody against the N-terminal domain of BACE1 and migrated with the expected molecular mass of the cleaved ectodomain of BACE1. Ͻ indicates ATase1.

ATase1 and ATase2 Acetylate Nascent BACE1
tivated agarose beads (ProFound system) employed in Fig. 6C and then analyzed by immunoblotting with an anti-BACE1 antibody. Fig. 7A (middle panel, anti-BACE1) clearly shows that BACE1 was present in the immunoprecipitate indicating that both acetyltransferases interact with endogenous BACE1 in vivo. Importantly, only the two lower bands corresponding to biosynthetic intermediates of BACE1 (Fig. 7A, compare the middle panel with the normal migration pattern of affinitypurified BACE1 on a similar SDS-PAGE system, shown in Figs. 3B and 5A) could be co-immunopurified with ATase1 and ATase2 (see also supplemental Fig. 1). These results are perfectly consistent with the fact that ATase1 and ATase2 are localized in the ER and ERGIC (Figs. 2C and 4C), where both biosynthetic intermediates of BACE1 can be found (supplemental Fig. 1A; see also Ref. 5). However, they are also surprising because we initially thought that only the lower species, migrating at ϳ52-55 kDa, was acetylated. In fact, the ϳ52-55-kDa band is the only one that can be detected with anti-acetylated lysine antibodies after immunoprecipitation of BACE1 from cell lysates (5) (see also Fig. 7D). This apparent discrepancy could be caused by the low levels and rapid metabolism of the upper biosynthetic intermediate that migrates at ϳ65 kDa (see also Fig. 7E). In fact the detection of this species is enhanced by subcellular fractionation (see supplemental Fig.  1A) or affinity purification (see Figs. 3B and 5A). To address this issue, we analyzed the acetylation pattern of BACE1 after immunopurification of ATase1 and ATase2. Fig. 7A (lower panel, anti-Ac. Lys.) clearly shows that both BACE1 biosynthetic intermediates are acetylated. Therefore, when taken together, these results indicate that ATase1 and ATase2 interact with nascent BACE1 in the ER/ERGIC system (see also supplemental Fig. 1B for a  schematic overview).
We previously published that the lysine acetylation of nascent BACE1 results in molecular stabilization and increased steady-state levels of the ␤-secretase (5). Therefore, we analyzed whether the overexpression of ATase1 and ATase2 affected the levels of BACE1 and/or the production of A␤. The results shown in Fig. 7B indicate that both ATase1 and ATase2 caused an increase in the steady-state levels of endogenous BACE1, with ATase2 displaying an apparent stronger effect, which could be related to the more powerful acetyl-CoA:lysine acetyltransferase activity observed in vitro (compare Fig. 4B with Fig. 2B for total membrane extracts and Fig. 4D with Fig.  2E for the individual fractions). These results were supported by the increased levels of A␤ secreted in the conditioned media (Fig. 7C), by the increased levels of endogenous C99 (supplemental Fig. 3), and by the apparent increase in the lysine acetylation of endogenous BACE1 (Fig. 7D) in ATase1-and ATase2expressing cells. C99 is the C-terminal fragment of APP that results from BACE1-mediated cleavage of APP and represents the immediate substrate for the generation of A␤.
The detection of the acetylated intermediates of nascent BACE1 from total cell lysates is normally hampered by the rapid deacetylation that occurs in the Golgi apparatus (5). To circumvent this obstacle, we also decided to immunoprecipitate BACE1 from ER/ERGIC vesicles prior to immunoblotting with FIGURE 4. ATase2 is localized in the early secretory pathway and has acetyl-CoA:lysine acetyltransferase activity. A, Western blot analysis of ATase1, ATase2, and ATase1 ϩ 2 expressing CHO cells. B, total membrane extract from control (nontransfected) and ATase2-expressing cells was assayed for acetyltransferase activity as in Fig. 2B. Results are expressed as percent of control and are the average (n ϭ 6) Ϯ S.D. Asterisk indicates statistical significance (p Ͻ 0.005). C, subcellular distribution of ATase2 was analyzed by SDS-PAGE and immunoblotting after separation of intracellular membranes on a 10 -24% discontinuous Nycodenz gradient. The appropriate subcellular markers are indicated. D, fractions 16, 18, and 20 of a subcellular fractionation gradient (see C) from control and ATase2-expressing cells were compared for acetyl-CoA:lysine acetyltransferase activity. The assay was performed as in Fig. 2E. Results are the average (n ϭ 6) Ϯ S.D. Asterisk indicates statistical significance (p Ͻ 0.005). JANUARY 23, 2009 • VOLUME 284 • NUMBER 4 an anti-acetylated lysine antibody. Fig. 7F clearly shows that under the above conditions the detection of BACE1-acetylated intermediates is enhanced, and the effect caused by the overexpression of ATase1 and ATase2 is much more evident. In fact, both acetyltransferases caused a marked increase in the acetylation of the immature form of endogenous BACE1.

ATase1 and ATase2 Acetylate Nascent BACE1
Next, we decided to down-regulate ATase1 and ATase2 and assess the effects on the steady-state levels of BACE1. Fig. 8, A-C, shows that two different siRNAs targeting ATase1 successfully reduced the steady-state levels of endogenous BACE1. Interestingly, the effects observed on the levels of BACE1 were proportional to the degree of ATase1 down-regulation induced by the different oligonucleotides. The fact that the down-regulation of ATase1 resulted in a marked reduction of endogenous BACE1 also indicates that ATase2 cannot replace the activity of ATase1 in vivo. It is worth stressing that the siRNA against ATase1 did not affect the expression of ATase2 (or vice versa; data not shown). This conclusion is consistent with the results shown in Fig. 5C, which indicate that ATase1 and ATase2 act as independent enzymes in vitro. The down-regulation of ATase2 also resulted in a modest decrease of BACE1 levels (Fig. 8, D-F). However, it is important to point out that the down-regulation of ATase2 was often accompanied by macroscopic cell death, which interfered with the analysis of BACE1 metabolism. Such an effect was independent of the transfection procedures and was never observed following down-regulation of ATase1 or other unrelated genes. Therefore, we interpret this as evidence of an important biological function of ATase2 in the cell. The down-regulation of ATase1 or ATase2 was also accompanied by decreased secretion of A␤ into the conditioned media (Fig.  8G) and by decreased levels of endogenous C99 (Fig. 8, B and E).
When taken together, the above results indicate that ATase1 and ATase2 influence the levels of BACE1 and the rate of A␤ generation. The results obtained with ATase1 and ATase2 siRNAs are particularly important because they indicate that biochemical inhibition of ATase1 and ATase2 might be useful for pharmacological approaches aimed at the prevention of late-onset AD.
Lipid Second Messenger Ceramide Increases the Levels of Endogenous ATase1 and ATase2-We previously published that ceramide treatment of different cell lines results in increased acetylation and stabilization of BACE1 (5,7,8). We also published that the increased activation of ceramide in the brain caused by either normal aging in wild-type mice or by the hyperactivation of IGF-1R signaling in p44 ϩ/ϩ transgenic mice, a model of accelerated aging (23), results in increased acetylation and stabilization of BACE1 (8,9). Therefore, we decided to analyze the effects induced by ceramide on the mRNA levels of ATase1 and ATase2 in H4 (human neuroglioma) and SH-SY5Y  Fig. 3A. Both the enzyme (ATase2) and the acceptor substrate (BACE1) were purified by affinity chromatography prior to the incubation. BACE1 Arg was used as control because the lysine residues that serve as acceptors of the acetyl groups have been mutated to arginine (5). Results are the average (n ϭ 6) Ϯ S.D. Asterisk indicates statistical significance (p Ͻ 0.005). C, acetyl-CoA:lysine acetyltransferase activity of ATase1 and ATase2 was assayed with a commercially available colorimetric assay employing a pool of histone-tail-peptides as acceptor of the acetyl group. Both ATase1 and ATase2 were purified by affinity chromatography prior to the reaction. The acetyl-CoA:lysine acetyltransferase activity of a highly purified pool of nuclear HAT is also shown. Results are the average (n ϭ 3) Ϯ S.D. (human neuroblastoma) cells. Under basal conditions the mRNA levels of both ATases were higher in SH-SY5Y than in H4 cells, suggesting possible cell-type differences (data not shown). However, when treated with ceramide, both cell lines displayed a marked activation of transcription of ATase1 and ATase2, as detected by quantitative real time PCR (Fig. 9). The effect could be observed as early as 4 h after treatment and was still evident after 4 days, which is consistent with the time course that we initially described (7,8). In conclusion, the above results clearly demonstrate that ATase1 and ATase2 are upregulated by ceramide treatment and suggest a possible implication in AD pathogenesis. In fact, AD patients have increased levels of ceramide (more than 3-fold, when compared with agematched controls) in the brain (24,25).

DISCUSSION
We have recently described a novel form of post-translational regulation of nascent BACE1 that involves transient acetylation of seven lysine residues in the lumen of the ER/ERGIC system followed by deacetylation of the mature protein in the lumen of the Golgi apparatus (5). The acetylated protein is able to proceed along the secretory pathway and achieve conformational maturation/stability, whereas the nonacetylated intermediates are retained and disposed in the ERGIC (5). Being a novel form of posttranslational regulation, the biochemical machinery responsible for these events is unknown. Here we report the identification of two ER/ERGIC-based acetyl-CoA:lysine acetyltransferases that are directly involved in the lysine acetylation of nascent BACE1. The proteins, initially named Camello-like 1 and Camello-like 2 for their structural organization, have no previously identified biochemical activity and were named here ATase1 and ATase2 to indicate their function.
Both ATase1 and ATase2 were shown to have acetyl-CoA:lysine acetyltransferase activity in vitro and under native conditions, and to be able to acetylate BACE1 in the lysine residues that we previously identified (5). Both enzymes were found to localize in the compartments where the lysine acetylation of BACE1 normally occurs, displaying a perfect co-distribution with the endogenous ER/ERGIC acetyl-CoA:lysine acetyltransferase activity, and to have the required topographical arrangement with the catalytic site facing the lumen of the ER/ERGIC, as predicted by our previous results (5). Finally, they were shown to interact with BACE1 in vitro and in vivo and to affect the steady-state levels of their immediate substrate.
Although they are 88% identical at the amino acid level and share several biochemical features, ATase1 and ATase2 seem to display two important differences. First, ATase1 could recognize and acetylate the fully mature form of BACE1, at least in vitro, whereas ATase2 could not; second, ATase1 could be detected with an anti-acetylated lysine antibody, whereas ATase2 could not. The former might indicate differences in substrate recognition properties of the two enzymes, whereas the latter might suggest different catalytic mechanisms. These differences might become important when attempting to design biochemical inhibitors for pharmacological purposes (26). In this regard, it is important to point out that p300 HAT contains an "activation/regulatory loop" that undergoes autoacetylation and functions as an allosteric switch that regulates its own acetyltransferase activity; when hypoacetylated it inhibits, and when autoacetylated it activates the activity of the enzyme (27, 28). Therefore, it is possible that ATase1 can  1A). The migration of the different forms of affinity-purified BACE1 on a similar SDS-PAGE system is shown in Figs. 3B and 5A. B, steady-state levels of endogenous BACE1 in ATase1-, ATase2-, and ATase1 ϩ 2-expressing cells were analyzed by immunoblotting of total cell lysates. C, A␤ levels in the conditioned media were determined by standard sandwich enzyme-linked immunosorbent assay. Results are the average (n ϭ 3) Ϯ S.D. Symbols indicate statistical significance (*, p Ͻ 0.05; catalog number, p Ͻ 0.005). D, endogenous BACE1 was immunoprecipitated (I.P.) from total cell lysates of ATase1-, ATase2-, and ATase1 ϩ 2-expressing cells prior to Western blot analysis with an anti-acetylated lysine antibody. As expected (5), only the ϳ52-55-kDa band corresponding to immature BACE1 could be visualized with an anti-acetylated lysine antibody. The migration of the different forms of BACE1 on a similar SDS-PAGE system is also shown for comparison (E). F, experiment described in D was repeated after immunoprecipitation of endogenous BACE1 from a pool of ER/ERGIC fractions (representative gradients are shown in Fig. 2C, Fig. 4C. and in supplemental Fig. 1).
which are found in the ER/ERGIC system. Interestingly, ceramide regulates the steady-state levels of the transporter in vivo (data not shown) and, consequently, the efficiency of acetyl-CoA transport into the lumen of the ER (5). Finally, we previously reported that the deacetylase, which is responsible for removing the acetyl groups from nascent BACE1, is solely found in the Golgi apparatus (5). Therefore, when taken together, these results indicate that the different components of the acetylation/deacetylation machinery have been strategically inserted along the secretory pathway to ensure availability of the required substrate and enzymes and the correct sequence of the biochemical events.
The subcellular distribution profile of ATase1 and ATase2 shows a predominant ERGIC localization that extends to the ER but not to the Golgi apparatus (Fig. 2C, Fig. 4C, and supplemental Fig. 1). This is consistent with the fact that the immunopurification of ATase1 or ATase2 pulls down the two biosynthetic (acetylated) intermediates of BACE1 that are found in the ER/ERGIC system ( Fig. 7A and supplemental Fig. 1). These results might suggest that the ATases act both as modifying enzymes (acetyltransferases) and chaperones of nascent BACE1. The fact that the lysine residues that are acetylated in BACE1 appear to be on the same face of the protein and to surround the catalytic pocket of the enzyme (5) suggests that the ATases "shield" nascent BACE1. Consistently with the above model, the "loss-of-acetylation" mutant versions of BACE1 are retained and degraded in the early secretory pathway by PCSK9/NARC-1 even though efficiently synthesized (5,6). In contrast, the "gain-of-acetylation" mutant is protected from degradation in the early secretory pathway. Therefore, we could envision a mechanism in which a stoichiometric competition between the acetyltransferases and the ERGIC-based degrading machinery regulates the steady-state levels of BACE1; ATase-bound (acetylated) BACE1 would be protected from degradation whereas ATase-unbound (nonacetylated) BACE1 would not.
The Camello ortholog Xcml in X. laevis has been shown to regulate cell migration and adhesion very early in embryogenesis, during the process of gastrulation (15). The molecular mechanism, even though unexplored and unknown, appeared to involve some sort of protein modification (presumably acetylation) in the early secretory pathway (presumably in the lumen of the ER). Even though we now offer a clear biochemical function for the human Camello-like proteins ATase1 and ATase2, the findings reported by Popsueva et al. (15) point to a crucial role of these enzymes in basic cellular events such as cell migration, cell-cell interaction, and development. The cell toxicity observed following down-regulation of ATase2 seems to support this conclusion. Because post-translational modifications of membrane proteins in the ER and Golgi apparatus are crucial for the correct assembly and function of the mature proteins, it is not surprising that the enzymes involved in such modifications might be crucial themselves (1,2,34).
The fact that the events described here act downstream of the aging program mediated by IGF-1R, the common regulator of life span in all organisms (10,(35)(36)(37), in regulating the rate of A␤ generation (5, 7-9) suggests a possible implication in late-onset AD neuropathology. This scenario would be supported by the fact that ceramide levels are higher in late-onset AD brains, when compared with age-matched controls (24,25). Therefore, it is possible that the biochemical characterization of the acetylation/deacetylation machinery may lead to the identification of novel pharmacological strategies for the prevention of late-onset AD.