Membrane Localization of β-Amyloid 1–42 in Lysosomes

β-Amyloid peptide (Aβ42) is the core protein of amyloid plaque in Alzheimer disease. The intracellular accumulation of Aβ42 in the endosomal/lysosomal system has been under investigation for many years, but the direct link between Aβ42 accumulation and dysfunction of the endosomal/lysosomal system is still largely unknown. Here, we found that both in vitro and in vivo, a major portion of Aβ42 was tightly inserted into and a small portion peripherally associated with the lysosomal membrane, whereas its soluble portion was minimal. We also found that the Aβ42 molecules inserted into the membrane tended to form multiple oligomeric aggregates, whereas Aβ40 peptides formed only dimers. Neutralizing lysosomal pH in differentiated PC12 cells decreased the lysosomal membrane insertion of Aβ42 and moderated Aβ42-induced lysosomal labilization and cytotoxicity. Our findings, thus, suggest that the membrane-inserted portion of Aβ42 accumulated in lysosomes may destabilize the lysosomal membrane and induce neurotoxicity.

␤-Amyloid peptide (A␤42) is the core protein of amyloid plaque in Alzheimer disease. The intracellular accumulation of A␤42 in the endosomal/lysosomal system has been under investigation for many years, but the direct link between A␤42 accumulation and dysfunction of the endosomal/lysosomal system is still largely unknown. Here, we found that both in vitro and in vivo, a major portion of A␤42 was tightly inserted into and a small portion peripherally associated with the lysosomal membrane, whereas its soluble portion was minimal. We also found that the A␤42 molecules inserted into the membrane tended to form multiple oligomeric aggregates, whereas A␤40 peptides formed only dimers. Neutralizing lysosomal pH in differentiated PC12 cells decreased the lysosomal membrane insertion of A␤42 and moderated A␤42-induced lysosomal labilization and cytotoxicity. Our findings, thus, suggest that the membrane-inserted portion of A␤42 accumulated in lysosomes may destabilize the lysosomal membrane and induce neurotoxicity.
Alzheimer disease (AD) 3 is the most common age-related neurodegenerative disorder, and the production and cerebral deposition of ␤-amyloid peptide (A␤) is widely believed to be central to the development of AD (1,2). The main isoforms of A␤ in AD are A␤40 and A␤42 (containing 40 and 42 amino acids, respectively) generated from the proteolytic processing of amyloid precursor protein (APP) (3). A␤42 is more neurotoxic than A␤40 and is the principle species associated with amyloid plaque (4 -7), but the exact molecular mechanisms of how A␤42 damages the neurons and deposits in brain still remain unclear.
Classically, extracellular deposition of A␤ was thought to be important in AD pathogenesis. More recent evidences have demonstrated that intraneuronal A␤ may play a crucial role in the early progression of the disease and pointed toward the importance of endosomal/lysosomal compartments in this pathogenic process (8 -12). The endosome/ lysosome pathway participates in A␤ production (13)(14)(15)(16)(17)(18), and A␤-released outside neurons in soluble or aggregated form can also be re-internalized and act inside endosomal/ lysosomal compartments (19). Nixon and co-workers (20) suggested that autophagosomes and other prelysosomal vacuoles are involved in AD. Recently Ling et al. (21) reported that A␤42 expression, but not that of A␤40, in Drosophila induces an age-dependent impairment of neuronal autophagy at a post-lysosomal stage, leading to extensive neuronal damage and death. Yang et al. (22,23) reported that loss of lysosomal membrane integrity occurs in response to A␤42 accumulation and is an early event in neuron death. Previous studies demonstrated that internalized A␤42 is largely resistant to degradation and accumulates as insoluble aggregates in late endosomes or secondary lysosomes in a variety of cells (24 -27); in contrast, shorter peptides such as A␤40 are rapidly degraded and do not accumulate (24,25,28). Notably, careful studies of human brain and brains from Alzheimer transgenic mice using C-terminal-specific antibodies against A␤40 and A␤42 established that most of the intraneuronal A␤ end at residue 42, not at residue 40, and are frequently co-localized with cathepsin D, a lysosomal marker (8,29). The oligomeric A␤ has been found to be most pathogenic (30 -32). In tissue derived from the human brain, A␤ oligomerization initiates within cells rather than in the extracellular space (33). Others have reported that A␤ oligomerization could occur in the endosomal compartments (34,35). The low pH of endosomes and lysosomes and their ability to concentrate solutes may provide an ideal environment in which to promote amyloid fibril assembly (24,25). Overall, the interaction between A␤42 and the lysosomal system seems to be pivotal for the preferential accumulation of A␤42 in neurons and association with AD pathogenesis.
Here, we show that a major portion of A␤42 accumulated in lysosomes was inserted into the lysosomal membrane, where they remained undegraded. We also present evidence that the multiorder oligomer of A␤42 formed in association with the lysosome membrane at low pH. The pH-dependent membrane insertion of A␤42 could cause membrane instability and lysosomal leakage. Our findings provide a possible mechanism for the lysosomal accumulation of A␤42 and its association with lysosome disruption, which have been hypothesized to be involved in AD pathology.

MATERIALS AND METHODS
A␤ Peptides, Antibodies, and Reagents-Lyophilized A␤ (AnaSpec Co.) was dissolved in dimethyl sulfoxide to obtain a 2 mM stock solution that was centrifuged (15,000 ϫ g) for 10 min to remove insoluble particulates (36) and stored at Ϫ70°C until use. Monoclonal antibody 6E10, which recognizes an epitope within residues 1-17 of human A␤, was purchased from Chemicon International Inc. Rabbit polyclonal antibody to the human C terminus of A␤42, which does not cross-react with A␤40 or APP, was from Signet Laboratories Inc. Mouse monoclonal antibody to human C terminus of A␤40, which does not cross-react with A␤42, was from Upstate Biotechnology. Antibodies against Lamp-1, Rab7, Rab6, and calnexin were from Santa Cruz Biotechnology. Secondary antibodies were obtained from ZhongShan Biotechnology. All lipids, bafilomycin A1 (BafA1), chloroquine, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), Lucifer Yellow, p-nitrophenyl phosphate, 4-methylumbelliferyl-␤-D-galactosidase, and a protease inhibitor mixture were purchased Sigma. Nerve growth factors and G418 were products of Invitrogen. All the other chemicals used were of analytical grade and were manufactured in China.
Monolayer Surface Pressure Measurements-Monolayer surface pressure (), defined as the change of surface tension after spreading a monolayer on water surface, was measured with a NIMA 9000 microbalance. Experiments were conducted as described in our previous work (37). Lipids were dissolved in a solvent of chloroform/methanol (7:1 v/v, 1.0 mg/ml). The maximum increase of surface pressure (⌬) induced by self-penetration of A␤42 into the air-water interface without a lipid monolayer was 15.6 mN/m, and the minimum concentration of A␤42 to reach such a maximum ⌬ was 800 nM, determined by pre-experiments (data not shown). Thus, we kept initial surface pressure ( i ) of the lipid monolayers at or above 16.0 mN/m and the concentration of A␤42 at 800 nM throughout experiments.
Hydrolysis of A␤42 and MALDI-TOF MS-Preparation of small unilamellar vesicles and hydrolysis of A␤42 was performed in accordance with the procedure described in our previous work (37). Lipids of the desired composition were mixed in chloroform/methanol (7:1, v/v) and dried under a stream of nitrogen. Lipid films were resuspended and sonicated in 5 mM Na 2 HPO 4 -HAc to near optical clarity. The concentration of phospholipid was determined (38). In a typical experiment A␤42 was incubated in the presence or absence of vesicles at 37°C. Hydrolysis reaction was initiated by the addition of papain, the specific hydrolytic sites of which are Glu-X, Gly-X, Tyr-X, His-X, Lys-X, and Arg-X. After 30 min, iodoacetamide was injected to stop the reaction. Then the mixture was subjected for MALDI-TOF MS analysis with a BIFLEX III MALDI-TOF mass spectrometer (Bruker Daltonics).
Preparation and Characterization of Lysosome-The procedure was processed as described (39). Briefly, cells were harvested by trypsinization and washed with phosphate-buffered saline and homogenization buffer (HB; 0.25 M sucrose, 10 mM Hepes, 1 mM EDTA (pH 7.4)). Hereinafter all manipulations were carried out at 4°C using pre-cooled reagents. Washed cells were resuspended in HB supplemented with protease inhibitor mixture at 1 ϫ 10 8 /ml and then homogenized using a Dounce glass Teflon homogenizer. Homogenates were spun at 800 ϫ g for 10 min to pellet nuclei and unbroken cells, which were then rehomogenized in a half-volume of HB. Supernatant was combined and centrifuged at 3000 ϫ g for 10 min to remove large heavy mitochondria. The resultant supernatant subsequently was centrifuged for 10 min at 18,000 ϫ g, obtaining a pellet resuspended in HB. The resuspension (0.5 ml) was layered on 6.5 ml of iso-osmotic (0.25 M sucrose) Percoll (GE Healthcare) at a concentration of 30% (pH 7.4), with a 1-ml cushion of 2.5 M sucrose at the bottom. After centrifuging at 44,000 ϫ g for 40 min in a fixed-angle rotor (Hitachi P70AT2), fractions of ϳ0.5 ml were carefully collected from the top of tube. Percoll was removed as described (40). Proteins were measured with BCA protein assay kit (Pierce). Organelle markers assayed for lysosomes were acid phosphatase (39) and ␤-hexosaminidase (41).
For preparation of a mouse brain lysosomal fraction, equal amounts of brain tissue derived from analogous cortical regions or hippocampus from brains of APPϫPS1 transgenic or wild type mice sacrificed at different ages (2 and 10 months) were processed as described for cells with slightly modifications. After dissection, brain tissue was immersed immediately in icecold HB and homogenized. Homogenate was digested with DNase I (250 g/ml for 30 min) and centrifuged at 1000 ϫ g for 10 min to remove nuclei and intact cells. The supernatant was collected and centrifuged again to remove blood cells. Next purification of lysosomes was carried out as described above.
Latency Measurements-Intactness of lysosomes was assessed by measuring the activity of ␤-hexosaminidase under isotonic conditions with or without 0.1% Triton X-100. Latency (%) of lysosomes is expressed as (activity with detergent minus activity without detergent)/(activity with detergent) ϫ 100. When the effect of A␤ on the intactness of lysosome was tested, lysosomes were incubated in the absence or presence of A␤ for 30 min before conducting latency measurement. The buffer used in the experiment was 5 mM citrate/phosphate (pH 4.5 or 7.4, isotonic osmolarity was adjusted with sucrose).
Lysosomal Subfractionation-All the fractionation procedures were conducted at 0 -4°C. Soluble (luminal) lysosomal proteins were obtained by resuspending the lysosomes in phosphate-buffered saline, freeze/fracturing them in dry ice/ethanol, and removing membranes by ultracentrifugation for 30 min at 200,000 ϫ g. The membrane pellet was incubated with 0.1 M Na 2 CO 3 (pH 11.0) (42) for 30 min and spun as above to give peripheral membrane proteins. The pellet of carbonatewashed membrane was further stripped by incubation with 1 M NaCl (43) for 30 min and spun as above. The ultimate membrane pellet was solubilized in 2% SDS and designated as membrane-inserted proteins. All buffers were supplemented with protease inhibitor mixture.
In Vitro Incubation with Biological Membranes-Samples containing 100 g of protein of isolated PC12 lysosomes were sonicated on ice for 15 s and ultracentrifuged at 200,000 ϫ g for 30 min. The obtained membrane pellet was resuspended in 100 l of Na 2 HPO 4 -citric acid buffers (20 mM, pH 4.5 or 7.4). Freshly dissolved A␤ peptides (0.1 M) were added and incubated at 37°C for 60 or 120 min. After incubation, the samples were ultracentrifuged as above to yield a supernatant (free unlabeled A␤) and membrane pellet. The pellet was extracted with 0.1 M Na 2 CO 3 (pH 11.0) and 1 M NaCl sequentially for loosely attached A␤. Finally, the pellet containing membrane integrated A␤ was solubilized in 2% SDS. For enzyme hydrolysis, 1 mg/ml trypsin was added to the 60-min incubation mixture of A␤42 and membrane at 37°C for 5 or 15 min just before ultracentrifugation.
Immunoblotting-Western blots for A␤ detection were performed as described (44). Analysis was performed using the public domain NIH ImageJ program (available on the Internet).
Measurement of Lysosomal Membrane Instability-Membrane instability and leakage of PC12 lysosomes were assessed by observing the distribution change of lysosomal fluorescent dye Lucifer Yellow or the leakage of lysosomal enzyme ␤-hexosaminidase into cytosol. For measuring the distribution of Lucifer Yellow, cells were labeled with Lucifer Yellow (0.1 mg/ml) at 37°C for 16 h and washed 3 times. Labeled cells with or without BafA1 pretreatment were then incubated with A␤42 at 37°C for 6 h and visualized on a Nikon E800 microscope. To assess the leakage of ␤-hexosaminidase, cells that had been incubated with A␤42 in the presence or absence of BafA1 were collected and homogenized ten strokes. Cytosol fraction was obtained by ultracentrifugation, and the activity of ␤-hexosaminidase was measured as described (41). The leakage of lysosomal ␤-hexosaminidase was expressed as a percentage of total activity obtained from cell homogenates. Assays were repeated in three independent experiments performed in duplicate. The significance of the results was assessed by Student's t test.
For lysosomes in brain, samples from the cortex or hippocampus of 2-and 10-month-old transgenic or wild type mice were homogenized and centrifuged. Then the activity of ␤-hexosaminidase was assayed as for cells. Purified lysosomal fraction were also incubated in 0.15 or 0.25 M sucrose at 37°C for 5 min, then the suspension was used for the assay of lysosomal integrity by measuring the activity of ␤-hexosaminidase (45,46).
Measurement of Cell Viability-Cell viability assays were carried out by MTT dye conversion assay (47) in 96-well cell culture plates. Cell viability was expressed as a percent of absorption of vehicle-treated control. In addition, cell death was evaluated by measuring the amounts of cytoplasmic lactate dehydrogenase released into medium. Lactate dehydrogenase activities were measured using a CytoTox96 nonradioactive assay kit (Promega), according to the manufacturer's instructions. Cytotoxicity was expressed as percentage of released lactate dehydrogenase per total cellular lactate dehydrogenase content. Assays were repeated in three independent experiments, each performed in triplicate.

A␤42 Can Insert into the Lipid Membrane via Its C Terminus in a pH-dependent
Manner-Previous studies implicate that a likely primary target of A␤ is the membrane, as the peptide may alter many of its important physical and biological properties (48 -50). Therefore, we first examined the membrane insertion ability of A␤42 via monolayer experiments. The surface pressure is believed to increase when the protein penetrates into the monolayer, and as reported earlier, it does not change if the protein only interacts with phospholipid head groups (51,52). Thus, the increase in surface pressure (⌬) of the lipid monolayer post-injection of proteins into subphase can only be interpreted as the result of actual insertion of the proteins into the phospholipid monolayer. ⌬ can be obtained as a function of the initial surface pressures ( i ) for each sample. Plotting a series of ⌬ versus i would yield a straight line with a negative slope, which intersects the abscissa at a point defined as the critical insertion pressure ( c ). The value of c is used to evaluate the penetration ability of the protein into this lipid monolayer.  Fig. 1C gives the correlation between c and pH, clearly indicating that the membrane insertion ability of A␤42 increased with decreasing pH. Same experiments were performed with other phospholipid mixtures, such as PC/phosphatidylethanolamine, PC/sphingomyelin, and PC/cerebroside, with similar effects (data not shown). These data suggest that the membrane insertion ability of A␤42 is critically dependent on pH and not sensitive to lipid composition. The biological membrane pressure had been established to be 31-34 mN/m (53). The packing density of the lipid monolayer with a surface pressure in this range is assumed to be comparable to that of the lipid bilayer (54,55). Thus, the monolayer results indicated that A␤42 is able to insert into a physiological bilayer at a pH lower than 5.5.
To further characterize membrane-associated A␤42, papain was used to hydrolyze A␤42 after it reacted with lipid vesicles based on the fact that A␤42 inserted into a lipid membrane would be resistant to digestion. The hydrolysis products were analyzed by MALDI-TOF MS. The MS peaks from a typical enzyme hydrolysis of the A␤42 solution were identified and summarized in supplemental Table S1. Fig. 2, A and B, show the MS spectra of the hydrolysis products of A␤42 in the absence and presence of vesicles at pH 4.0. The two MS spectra exhibit some obvious differences; the peaks of m/z 2932, 3672, and 4072 disappeared in the presence of vesicles. In comparison, we found that the inaccessible cleavage sites were Gly 33 -Leu 34 and Gly 37 -Gly 38 , both of which are located in the C-terminal transmembrane domain of A␤42. In contrast, the mass spectra in the presence and absence of vesicles at pH 7.0 were almost identical (supplemental Fig. S1), indicating that papain could access nearly all the A␤42 cleavage sites after reacting with vesicles at neutral pH. Therefore, we concluded that A␤42 inserts into lipid vesicles via its C-terminal at low pH, thus protecting its C-terminal sites from cleavage.
Lysosome-accumulated A␤42 Mainly Inserts into the Lysosomal Membrane in Cells-To investigate whether A␤42 could integrate into biological membranes of intracellular acidic organelles, we incubated differentiated PC12 cells with 1 M freshly dissolved A␤42 for 6 h at 37°C, then the cells were washed with phosphate-buffered saline and transferred into fresh medium for 6 h. After incubation, the cells were harvested, and the lysosomes were isolated on a self-generated Percoll gradient as described under "Materials and Methods." The two lysosomal enzyme markers, acid phosphatase and ␤-hexosaminidase, are shown to be concentrated in fractions 11 and 12, indicating that the bulk of lysosomes was contained in the two fractions (Fig. 3A). Intactness of the organelles was above 90% based on a latency measurement. We also characterized our lysosomal preparations by Western blot (supplemental Fig. S2) and electron microscopy (supplemental Fig. S3). The lysosomal proteins Lamp-1 and Rab7 were markedly enriched in the lysosomal fraction, whereas the endoplasmic reticulum FIGURE 1. A␤42 can insert into lipid membranes in a pH-dependent manner. The interaction between A␤42 and total brain lipid monolayers (A) or eggPC (B) was detected via monolayer surface pressure measurements at various pH values. The surface pressure change (⌬) caused by peptide insertion into the lipid monolayer can be obtained as a function of the initial surface pressure I . A plot of ⌬ versus i yields a straight line with a negative slope that intersects the abscissa at a limiting surface pressure, the critical insertion pressure, c . The value of c is a quantitative measure to evaluate the insertion ability of the peptide into the lipid monolayer. The concentration of A␤42 was 800 nM, and Na 2 HPO 4 -citric acid was the buffer used to supply a sufficient pH range. Panel C gives the correlation between the critical insertion pressure c and pH.  Table S1. a.i., arbitrary intensity.
(ER) and Golgi markers calnexin and Rab6 were undetectable, suggesting that the ER and Golgi were not present in this lysosomal fraction. Immunoelectron microscopy analysis further confirmed that the lysosomal fraction was mostly composed of lamp-1-positive lysosomes.
We also examined the subcellular location of endocytosed A␤42 in PC12 cells. Western blot analysis of gradient fractions with a specific A␤42 antibody showed that A␤42 applied extracellularly was enriched in the lysosomal fraction (Fig. 3B). Also, immunofluorescence analysis displayed a co-localization of A␤42 with Lyso Tracker (data not shown). These results indicated that exogenous A␤42 could be taken up by PC12 cells and located to lysosomes.
To determine the intralysosomal distribution of A␤42, isolated lysosomes were fractionated into a soluble and a membrane fraction. The membrane fraction was then sequentially extracted with carbonate and sodium chloride, which are widely used to strip peripherally membrane-attached proteins (42,43). After these treatments, the proteins remaining in the membrane fraction were considered to be tightly inserted into the lysosomal membrane. Thus, the proteins of intact lysosomes were fractionated into "soluble," "membrane-associated" (Na 2 CO 3 -extractable and NaCl-extractable), and "membrane-inserted" (detergent-extractable) fractions. As controls, ␤-hexosaminidase, a lysosomal matrix marker, was detected entirely within the soluble fraction (Fig. 4A), whereas Lamp-1, a lysosomal membrane marker, was enriched in the membraneinserted fraction (Fig. 4B, top panel), indicating that the lysosome was substantially fractionated.
The lysosomal distribution of internalized A␤42 was shown in Fig. 4B, bottom panel. The internalized A␤42 was markedly enriched in the membrane-inserted fraction, whereas only a small portion was extracted by 0.1 M Na 2 CO 3 , and its soluble portion was nearly undetectable.
Furthermore, we examined the lysosomal localization of the endogenous A␤42 accumulated in N2a APPsweϫPS1-transfected cells. Immunoblots with a specific A␤42 antibody, A␤40 antibody, and 6E10 (recognizing residues 1-17 of A␤) showed that A␤42 was markedly localized in the membrane-inserted fraction with only a small portion loosely attached to the membrane (Na 2 CO 3 extraction), and its soluble portion was also undetectable (Fig. 4C, top panel). A␤40, however, was undetected in any fraction of the lysosomes (Fig. 4C, middle panel), indicating that A␤42 preferably accumulated in the lysosomes of cells. Combining the results from PC12 and N2a cells, one may suppose that A␤42 mostly tightly inserted into the lysosomal membrane may play an important role in neuronal toxicity. Additionally, we investigated the subcellular localization of A␤42 in brains from APPϫPS1 transgenic mice using immunoelectron microscopy. The results in supplemental Fig. S4, in agreement with the recent findings (29,56), clearly showed that the immunoreactivity of intraneuronal A␤42 increased with aging (there were more gold particles in the lysosomes of a 10-month-old mouse than a young mouse) and localized predominantly to lysosomes of neurons in transgenic mice (comparatively less gold particles in the endoplasmic reticulum and mitochondria).
Effect of Alkalizing Drugs on the Insertion of A␤42 into the Lysosomal Membrane-To assess the effect of pH on the lysosomal distribution of A␤42, the lysosomal pH was altered by BafA1 or chloroquine. BafA1 specifically inhibited the transmembrane component of ATPases, the H ϩ pump responsible for the acidification of late endosomes and lysosomes (57). Chloroquine, a weak base, accumulated in lysosomes where it raises the pH (58). Differentiated PC12 cells were incubated with 1 M A␤42 for 6 h in the presence or absence of 10 nM BafA1 or 10 M chloroquine, then washed with phosphatebuffered saline and transferred into fresh medium for 6, 12, or 24 h (Fig. 4, D and E). An equal quantity of isolated lysosomes was fractionated as described above and analyzed by Western blotting. BafA1 induced an obvious effect on the lysosomal distribution of A␤42, decreasing the level of membrane-inserted A␤42 from ϳ90% to nearly 50% (Fig. 4D). A similar effect was also observed with chloroquine (Fig. 4E), indicating that A␤42 can insert into the lysosomal membrane in a pH-dependent manner.
Extending the incubation time from 6 to 12 and 24 h, only A␤42, which tightly integrated into the membrane, remained detectable in the lysosomes, whereas the fraction peripherally attached to the membrane became undetectable (Fig. 4, E and F). The lysosomal system is known to process or digest cargoes from phagocytosis, endocytosis, and autophagy (59,60). Therefore, the fact that the membrane-inserted A␤42 remained undegraded suggests that this form may resist hydrolysis by lysosomal proteases.
A␤42 and A␤40 Differ in their Interaction with the Lysosomal Membrane-To explore whether another pathologically important peptide, A␤40, and a shorter peptide, A␤28, also inserted into and accumulated at the lysosomal membrane after internalization by differentiated PC12 cells, we incubated syn- thetic A␤40 and A␤28 with cells as A␤42. Fig. 5 shows the intralysosomal distribution of A␤42, A␤40, and A␤28 after 6 h of incubation. Interestingly, the total amount of the lysosomal accumulation of A␤42 (that is, the sum of soluble, membraneattached, and membrane-inserted A␤42) was remarkably higher than that of A␤40 and A␤28. Compared with A␤42, A␤40 had only a very low remnant level located at the membrane fraction. A␤28 appeared to be eliminated to the same background level as the control. Alkalizing drug treatment (i.e. cells were preincubated with 10 M chloroquine for 30 min before A␤ was added) also caused the same effect with A␤40 as it did with A␤42; however, the level of A␤40 in each fraction was very low.
To further clarify the molecular basis of pH affecting the interaction of A␤ peptides with biological membranes, we incubated A␤ monomers of different lengths with purified lysosomal membrane for 60 min in acidic (pH 4.5) or neutral (pH 7.4) buffer in vitro. Fig. 6A shows the membrane interaction of A␤ monomers, indicating that pH has a significant effect on the distribution of A␤42 and A␤40. The levels of membrane insertion and membrane association decreased at pH 7.4, whereas the water-soluble fractions increased compared with those at pH 4.5. In addition, the amount of membrane insertion of A␤42 was greatly higher than that of A␤40 at both pH 4.5 and 7.4. Compared with A␤42, A␤40 existed mainly in the water-soluble (pH 7.4) or membrane-associated state (pH 4.5). Moreover, regardless of pH 4.5 or 7.4, A␤28 only existed in the free form without any membrane association (data not shown). Extending the incubation time from 60 to 120 min (Fig. 6B) made it clearer that pH plays an important role in A␤-membrane interactions. Note that at lower pH, A␤42 preferred to form many multiple-molecular weight oligomeric aggregates in both the membrane-associated and membrane-inserted fractions, whereas A␤40 occurred only in dimers.
To demonstrate that membrane insertion may protect A␤42 from hydrolysis by proteases, we treated an incubation mixture of A␤42 and the lysosomal membrane with trypsin for different times. As expected, trypsin degraded most free and membrane-attached A␤42 within 15 min, whereas membrane-inserted A␤42 mostly persisted (Fig.  6C). As a control, mixture of A␤42 with only buffer showed complete degradation within 15 min (data not shown). These data provided further evidence that A␤42 molecules resist degradation in the membrane-inserted state. Here, we also examined the membrane-insertion characteristics of A␤42 with an electron microscope. An incubation mixture of A␤42 and the lysosomal membrane was washed first with carbonate and NaCl to remove loosely attached A␤42 and then immunogold-labeled with 6E10 (the N terminus antibody of A␤42) or the C terminus antibody of A␤42. As expected, the sample stained positively with 6E10 but negatively with the C terminus A␤42 antibody, further indicating that A␤42 inserted into the lysosomal membrane through its C terminus (supplemental Fig. S5).
A␤42 Induces Lysosomal Labilization of PC12 Cells in a pHdependent Manner-To assess whether pH has an effect on A␤42-induced lysosomal labilization, we performed a combination of a fluorescence assay and an enzyme activity analysis. For the fluorescence assay, we used Lucifer Yellow, a membrane-impermeable marker of fluid-phase pinocytosis that accumulates in secondary lysosomes, to determine lysosomal leakage as previously described (22,25). Differentiated PC12 cells were incubated first with Lucifer Yellow and then with A␤42 for 6 h at 37°C and examined under a fluorescence microscope. As shown in Fig. 7A, untreated control cells dis- FIGURE 4. Lysosome accumulated A␤42 is enriched in the membrane fraction. Intralysosomal distribution of ␤-hexosaminidase activity (lysosomal matrix marker (A)), Lamp-1 (lysosomal integral membrane protein marker) and internalized A␤42 in differentiated PC12 cells (B) is shown. Purified lysosomes were fractionated into a soluble fraction (soluble), 0.1 M carbonate extracted fraction (Na 2 CO 3 ), 1 M NaCl extracted fraction (NaCl), and membrane-inserted protein fraction (insertion) as described under "Materials and Methods." C, the distributions of endogenous A␤ accumulated in lysosomes of N2a APPsweϫPS1 cells were examined in each lysosomal fraction with specific anti-A␤42, anti-A␤40, and 6E10 antibodies. BafA1 (D) or chloroquine (CHQ, E) reduced the level of the membrane-inserted A␤42. Differentiated PC12 cells were pretreated with or without alkalizing drugs, BafA1 (10 nM), or chloroquine (10 M) for 30 min, incubated with 1 M A␤42 at 37°C for 6 h, and then transferred into fresh medium for the indicated times. Lysosomes were purified, and equal volumes were subfractionated and analyzed by Western blotting with an antibody specific to the C terminus of A␤42, which does not recognize A␤40 or APP. F, a graphic presentation of E shows the membrane-inserted A␤42 level in the presence or absence of chloroquine. The A␤42 level in the absence of chloroquine was set at 1, and the data represent those from two individual experiments. played a punctate pattern of fluorescence, revealing small, circumscribed, vesicular structures resembling intact lysosomes. After treatment with A␤42, the cells displayed a diffuse intracellular pattern of fluorescence, indicating lysosomal leakage into the cytosol. However, this redistribution of Lucifer Yellow could be counteracted partly if the cells were pretreated with 5 nM BafA1 for 20 min before the addition of A␤42.
The effect of a neutralizing pH on A␤42-induced lysosomal labilization was examined by measuring the release of ␤-hexosaminidase into the cytosol in differentiated PC12 cells in the presence and absence of 5 nM BafA1. As shown in Fig. 7B, the activity of the released ␤-hexosaminidase in the cytosolic fraction of A␤42-treated cells was 40% higher than that in the untreated control. BafA1 treatment resulted in a pronounced decrease in A␤42-induced lysosomal leakage, whereas BafA1 alone did not affect lysosomal leakage of ␤-hexosaminidase.
These results indicate that BafA1 moderates the A␤42-induced instability of lysosomes.
We also separately examined the effects of A␤42, A␤40, and A␤28 on the latency of isolated lysosome under neutral or acidic pH. As shown in Fig. 7C, we observed deleterious effects of A␤42 on lysosomal membranes that were especially prominent under acidic pH, whereas A␤40 and A␤28 did not obviously affect lysosomal membrane intactness. These results consistently demonstrated that A␤42 caused membrane instability under lower pH.
Alkalizing Drugs Provide Dose-dependent Protection against A␤42-mediated Cytotoxicity-The effects of BafA1 and chloroquine on A␤42-induced cytotoxicity were determined using an MTT assay and lactate dehydrogenase release. Fig. 8A shows that cell toxicity measured by the MTT assay increased with the concentration of A␤42, whereas the addition of BafA1 partly prevented cell death induced by A␤42 at all concentration levels. A similar phenomenon was observed in the lactate dehy- FIGURE 5. Preferential accumulation of A␤42 versus A␤40 and A␤28 in differentiated PC12 cells. Cells were pretreated with or without 10 M chloroquine (CHQ) for 30 min and then incubated with or without 2 M A␤ isoforms at 37°C for 6 h. Lysosomes were purified, and equal amounts of lysosomes were subfractionated into soluble, 0.1 M carbonate-extractable (Na 2 CO 3 ), and 1 M NaCl-extractable (NaCl) membrane-inserted (Insertion) fractions. All subfractions were separated in a 16% Tris-Tricine gel and analyzed by Western blotting with 6E10. Ctrl, control. were then added and incubated at 37°C for 60 min. After incubation the samples were fractionated into free A␤ (free), 0.1 M carbonate extract (Na 2 CO 3 ), 1 M NaCl extract (NaCl), and membrane-inserted fraction (insertion) as described under "Materials and Methods." B, furthermore, samples were prepared after incubation with fresh A␤42 or A␤40 for 120 min. C, after incubation for 60 min at pH 7.4, a mixture of A␤42 and membrane was digested with 1 mg/ml trypsin (Try) at 37°C for the indicated times. After digestion, reaction mixtures were ultracentrifuged as above and analyzed by immunoblotting. drogenase release assay (Fig. 8B). The effect of chloroquine on A␤42-induced cytotoxicity was identical to that of BafA1 (Fig.  8, C and D). These results suggest that lower pH plays a critical role in A␤42-induced cell toxicity.
A␤42 Inserts into Lysosomal Membrane and Induces Lysosome Instability in Aged Transgenic Mice-Finally we investigated whether the insertion of A␤42 into the lysosomal membrane and the resulting labilization occurred in vivo. For this, we isolated lysosomes from cortical regions or hippocampus from brains of transgenic mice at different ages (2 and 10 months). Analogous wild type mice were used as control. In 10-month-old transgenic mice, the immunoblotting of subfractions of cortex lysosomes with an antibody specific to A␤42 showed that most of the A␤42 accumulated in lysosomes inserts into the membrane, with a little loosely attached (Fig.  9A). As for 2-month-old transgenic mice and wild type mice, the accumulation of A␤42 in brains is not obvious, and we did not observe similar phenomenon (data not shown).
Next we evaluated the stability of lysosomes in transgenic mice brain by studying the subcellular distribution of the lysosomal matrix enzyme ␤-hexosaminidase after homogenization and the leakage of this enzyme from the lysosomal fraction after incubation in hypotonic condition. First, we found that the percentage of ␤-hexosaminidase in the cytosolic fraction showed an apparent increase in aged transgenic mice, with about an 80% increase in the 10-month-old transgenic mice compared with the same old wild type mice (Fig. 9B). The increase of free ␤-hexosaminidase reflects the increase of lysosomal instability in the older transgenic mice brains. The lysosome labilization in aged transgenic mice brain was further established by the evidence that the integrity of the lysosomes from aged transgenic mice decreased greatly after incubation at 37°C for 5 min in hypotonic sucrose than that of the control lysosomes ( Fig. 9C and D). It indicates that the lysosomes of aged transgenic mice were more susceptible to the osmotic imbalance across their membranes and, therefore, lost their enzyme latency markedly.

DISCUSSION
Although the pathogenic pathway of A␤-induced neuronal death in AD has been investigated for years, the definitive pathogenesis is still unclear. Recently, a lysosomal branch of the cell death cascade may be important in the disease process. The prevailing model describes that A␤ accumulation in lysosomes resists degradation and eventually causes the release of degradation-resistant insoluble aggregates and other lysosome contents into the cytosol, consequently leading to neuron toxicity (21)(22)(23)61). However, the molecular mechanism involved in the lysosomal accumulation of A␤42 remains to be clarified. Additionally, lysosomes are organelles that contain a variety of digestive enzymes. Thus, knowing how a mass of A␤42 escapes degradation and accumulates in lysosomes is important in understanding A␤42-specific neurotoxicity.
We first employed two model membrane systems, monolayers and vesicles (a closed bilayer), to explore the effect of pH on A␤42-membrane interactions. We found that A␤42 could insert into the lipid vesicles via its hydrophobic C terminus under acidic conditions and that the membrane insertion ability of A␤42 increased with decreasing pH. This implies that A␤42 can insert into the membranes of endosomes (pH 5-6.5) and lysosomes (pH 4 -5). Therefore, we isolated and purified lysosomes from differentiated PC12 cells with extracellularly applied A␤42 and N2a APPsweϫPS1-transfected cells and examined the intralysosomal distribution of exogenous and endogenous A␤42.
For the first time we determined that lysosomal-accumulated A␤42 could be divided into three parts: a soluble state in the lysosome lumen; a membrane-attached state, loosely adsorbed onto the lysosomal membrane surface that can be were then incubated at 37°C for 6 h and visualized under a Nikon fluorescence microscope. A␤42 treatment caused lysosomal leakage as revealed by fluorescence that diffused into the cytoplasm; pretreatment with BafA1 moderated the A␤42-induced release of Lucifer Yellow, and the cells displayed the same discrete punctate pattern of fluorescence as the control (Ctrl). The inset is the corresponding phasecontrast image. B, cytosolic ␤-hexosaminidase activity was determined as described under "Materials and Methods" to evaluate the effect of BafA1 on A␤42-induced lysosome leakage. Values are the means Ϯ S.D. of three independent experiments performed in duplicate. Statistical analysis was conducted using Student's two-tailed t test (*, p Ͻ 0.05; **, p Ͻ 0.05). C, shown is the effect of A␤ isoforms on the latency of isolated lysosomes. Purified lysosomes were resuspended in isotonic citrate/phosphate buffer (5 mM, adjusted with sucrose) and incubated with 1 M A␤ at 37°C for 30 min before measuring ␤-hexosaminidase activity in the presence or absence of 0.1% Triton X-100. detached by cations; and a membrane-inserted state, tightly associated with the membrane, that cannot be extracted by cations. In differentiated PC12 cells, with the addition of 1 M A␤42, we found that a great majority of the lysosomal-accumulated A␤42 was in a membrane-associated state, whereas the A␤42 in the lysosomal lumen was minimal and nearly undetectable under the experimental conditions. Moreover, we found that the membrane-related A␤42 was mostly tightly inserted into the lysosomal membrane, whereas only a small portion was peripherally associated with the membrane. Identical results were obtained from the N2a cells stably expressing human ␤APPswe and PS1.
Next, raising the intracellular pH with alkalizing drugs reduced the level of the membrane-inserted A␤42 and increased that of the membrane-attached portion. Moreover, after prolonging the incubation time of A␤42 with PC12 cells from 6 to 12 and 24 h, quite a large amount of A␤42 was found to be retained in the membrane-inserted fraction, whereas the membrane-attached A␤42 disappeared. Trypsin digestion of the incubation mixture of A␤42 and the membrane also indicated that the membrane-inserted A␤42 cannot be easily hydrolyzed. Based on these results, one can reasonably presume that the massive accumulation of A␤42 in lysosomes may result from nondegradation by lysosomal proteases because of its membrane insertion, thereby providing the physical prerequisite for A␤42 neurotoxicity.
Our previous study showed that the insertion ability of A␤40 is strongly dependent upon the ratio of cholesterol to phos-pholipids under neutral pH (37). However, in a separate examination we found that cholesterol has little influence on the membrane insertion ability of A␤42; the corresponding c for eggPC and the cholesterol-containing eggPC monolayer are almost the same (data not shown). Here we show that there are significant differences in the membrane insertion between A␤42 and A␤40, and the level of membrane-inserted A␤40 was very low or negligible compared with A␤42 at either pH 4.5 or 7.4 (Fig.  6A). Our investigation on the hydrophobic exposure of A␤42 by an 1-anilino-8-naphthalenesulfonate fluorescence probe showed that the hydrophobic exposure of A␤42 increases with decreasing pH (data not shown), corresponding to its increasing ability of membrane insertion. These data suggested that the magnitude of hydrophobicity may be the main factor involved in the membrane insertion, and the difference in membrane insertion ability between A␤42 and A␤40 may be due to the difference of the two hydrophobic amino acid residues, as A␤28, the isoform without C-terminal transmembrane domain, cannot insert into the membrane at all.
Unlike A␤42, in cell culture models A␤40 was degraded much more rapidly after internalization (23,25), and in AD patients and animal models, it is A␤42, not A␤40, that is detected in intracellular accumulation (8,29,56,62). Our comparison of the plasma membrane binding of A␤ in differentiated PC12 cells (supplemental Fig. S6) shows that the binding of A␤42 is only about 2-fold that of A␤40. Thus, the great difference in the cumulative amount of internalized A␤42 and A␤40 may be caused mainly by the different rate of degradation due to their difference in membrane insertion ability. Additionally, incubation of A␤ with the lysosomal membrane showed that membrane-inserted A␤42 tends to form multi-oligomeric aggregates that are poorly degraded, whereas A␤40 forms only dimers. Thus, this differential interaction of A␤42 and A␤40 with the membrane could be the underlying cause for the preferential accumulation of A␤42 in lysosomes and neuronal death. This finding may explain the paradoxical observations that A␤40 is the primary proteolytic product of APP in neurons (63), whereas it is A␤42 that predominantly exhibits intraneuronal accumulation and neurotoxicity (64,65).
However, our present study does not account for the formation of A␤ oligomers and the particular mechanism of membrane insertion of the protein. From the in vitro interaction results of A␤ with membrane in this current manuscript (Fig.  6A), we can make sure that monomeric A␤ can insert into the membrane. There may be many aggregates in solution for the aggregating property of A␤42. A␤ oligomerization has also been shown to occur during interaction with lipid bilayers, in particular lipid rafts (66,67). Therefore, we cannot distinguish the origination of the oligomeric A␤42 on the membrane. It is unclear whether the membrane insertion of A␤42 takes place before, after, or during its oligomerization. The mechanism and dynamics by which A␤42 oligomerizes when it interacts with membrane are very complicated and remain to be established.
Previous reports have suggested that the lysosomal accumulation of A␤42 may play an important role in neuronal cell death and that the loss of lysosomal membrane impermeability is involved in A␤42 pathogenesis (22,23,68,69), but what is essential for the A␤42-induced lysosomal damage remains to be shown. We confirmed the view that A␤42 applied extracellularly leads to fluorescence diffusion of endocytosed punctate Lucifer Yellow and lysosomal enzyme leakage. Treatment with alkalizing drugs, such as chloroquine and BafA1, obviously ameliorated lysosome damage and cell death. A recent publication by Ji et al. (68,69) reported that apoE4 and A␤42 might work in concert at lower pH to increase the susceptibility of lysosomes to disruption. They also pointed out that the question of whether the apoE4/A␤42 complex or the two molecules, independently, destabilize the membranes still remains to be resolved. Our study demonstrates that A␤42 itself in the membraneinserted form can destabilize the lysosomal membrane. Consistent with the observations that A␤42 inserts to the lysosomal membrane at acidic pH and destabilizes them in vitro, aged transgenic mice expressing human APP and PS1 also showed tight binding of A␤42 to the lysosomal membrane in the cortical region and fragile lysosomes in cortex and hippocampus. These data in vivo emphasized the importance of lysosomal pathway induced by A␤42.
Taken together, if A␤42 resulting from extracellular uptake or intracellular generation accumulates within the endocytic compartments of neurons, the acidic characteristics of endosomal/lysosomal organelles could provide a favorable environment for the membrane insertion of A␤42 that may protect it from degradation and result in lysosome instability. Additionally, the concentration effect of lysosomes and the pathological accumulation of membrane-bound A␤42 may favor the oligomerization of A␤42, which probably then acts as a seed for further aggregation. Our findings suggest that lysosomal A␤ insertion may be pathogenic, and regulating intracellular pH pharmacologically may have a therapeutic effect on AD.