Intracellular accumulation of beta-amyloid in cells expressing the Swedish mutant amyloid precursor protein.

β-Amyloid (βA) is a normal metabolic product of the amyloid precursor protein (APP) that accumulates in senile plaques in Alzheimer's disease. Cells that express the Swedish mutant APP (Sw-APP) associated with early onset Alzheimer's disease overproduce βA. In this report, we show that expression of Sw-APP gives rise to cell-associated βA, which is not detected in cells that express wild-type APP. Cell-associated βA is rapidly generated, is trypsin-resistant, and is not derived from βA uptake, indicating that it is generated from intracellular processing of Sw-APP. Intracellular and secreted βA are produced with different kinetics. The generation of intracellular βA is partially resistant to monensin and a 20°C temperature block but is completely inhibited by brefeldin A, suggesting that it occurs in the Golgi complex. Monensin, brefeldin A, and a 20°C temperature block almost completely inhibit βA secretion without causing increased cellular retention of βA, suggesting that secreted βA is generated in a post-Golgi compartment. These results suggest that the metabolism of Sw-APP gives rise to intracellular and secreted forms of βA through distinct processing pathways. Pathological conditions may therefore alter both the level and sites of accumulation of βA. It remains to be determined whether the intracellular form of βA plays a role in the formation of amyloid plaques.

␤-Amyloid (␤A) is a normal metabolic product of the amyloid precursor protein (APP) that accumulates in senile plaques in Alzheimer's disease. Cells that express the Swedish mutant APP (Sw-APP) associated with early onset Alzheimer's disease overproduce ␤A. In this report, we show that expression of Sw-APP gives rise to cell-associated ␤A, which is not detected in cells that express wild-type APP. Cell-associated ␤A is rapidly generated, is trypsin-resistant, and is not derived from ␤A uptake, indicating that it is generated from intracellular processing of Sw-APP. Intracellular and secreted ␤A are produced with different kinetics. The generation of intracellular ␤A is partially resistant to monensin and a 20°C temperature block but is completely inhibited by brefeldin A, suggesting that it occurs in the Golgi complex. Monensin, brefeldin A, and a 20°C temperature block almost completely inhibit ␤A secretion without causing increased cellular retention of ␤A, suggesting that secreted ␤A is generated in a post-Golgi compartment. These results suggest that the metabolism of Sw-APP gives rise to intracellular and secreted forms of ␤A through distinct processing pathways. Pathological conditions may therefore alter both the level and sites of accumulation of ␤A. It remains to be determined whether the intracellular form of ␤A plays a role in the formation of amyloid plaques.
The deposition of ␤-amyloid (␤A) 1 in senile plaques is a pathological hallmark of Alzheimer's disease (Glenner and Wong, 1984;Masters et al., 1985). ␤A is a normal product of the metabolism of amyloid precursor protein (APP) in the secretory pathway Seubert et al., 1992;Shoji et al., 1992;Busciglio et al., 1993). Mutations in APP that are linked to autosomal dominant inheritance of Alzheimer's disease have been found to alter the production of ␤A. Mutation of the two amino acids proximal to the N terminus of ␤A has been described in individuals from two Swedish families that develop early onset Alzheimer's disease (Mullan et al., 1992). Expression of APP containing the Swedish mutation (Sw-APP) in transfected cells increases the production of ␤A by about 5-fold, suggesting a causal link between altered APP processing and the development of Alzheimer's disease (Citron et al., 1992;Cai et al., 1993). In this report, we show that the Swedish APP mutation not only increases ␤A production but also results in abnormal intracellular accumulation of ␤A. The processing pathways that give rise to intracellular and secreted ␤A can be distinguished by their differential kinetics and sensitivities to metabolic inhibitors and temperature block.

EXPERIMENTAL PROCEDURES
Metabolic Labeling and Immunoprecipitation-COS-1 cells maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS) were transiently transfected with expressor plasmids encoding APP 695 , APP 751 , and APP 751 containing the Swedish mutation (KM changed to NL immediately N-terminal to the ␤A sequence, designated Sw-APP) (Mullan et al., 1992). HEK 293 cells stably transfected with the APP 695 cDNA containing the Swedish mutation were maintained in DMEM with 10% FCS and 0.25 mg/ml G418. Transfected COS cells were metabolically labeled with 125 Ci/ml [ 35 S]methionine, and immunoprecipitations were performed as described previously (Busciglio et al., 1993). For immunoprecipitation of cell-associated ␤A, three semi-confluent 10-cm plates of transfected COS cells were lysed and combined, and the supernatant of the 100,000 ϫ g centrifugation was immunoprecipitated. APP was immunoprecipitated with a polyclonal antibody to the 20 C-terminal residues of APP (RG117, 1:200) and resolved by 10% Tris-glycine SDS-polyacrylamide gel electrophoresis. ␤A was immunoprecipitated with a polyclonal antibody to synthetic ␤1-40 (1:200 in culture medium and 1:30 in cell lysates) and resolved by 10 -20% Tris-Tricine SDS-polyacrylamide gel electrophoresis. Gel loading was normalized to the protein content of the cell lysate (Bio-Rad protein assay), and quantitation was performed by PhosphorImager scanning. The specificity of these antibodies and the protocol for peptide preabsorption have been described previously (Busciglio et al., 1993). For pulse-chase analysis, stably transfected HEK 293 cells were labeled with 100 Ci/ml [ 35 S]cysteine/methionine in cysteine/methionine-free DMEM with 2% dialyzed FCS for 20 min and then chased in DMEM with 2% FCS and 1 mM cysteine/methionine. ␤A was immunoprecipitated with 5 g of purified mouse monoclonal antibody ␤1 to residues 1-40 of ␤A (Paganetti and Scheller, 1994). Quantitation of pulse-chase results was performed by direct counting of the gels using an InstantImager (Canberra Packard). Values were normalized for the number of cysteine and methionine residues in the different APP metabolites.
Treatments-The 4°C temperature block was performed by pulse labeling of transfected cells for 10 min at 37°C followed by a 90-min chase in unlabeled DMEM at either 4 or 25°C. Trypsinization of cell surface ␤A was performed by labeling transfected cells for 16 h followed by addition of 0.25% trypsin, 0.02% EDTA for 10 -20 min at 37°C. The reaction was stopped by 10-fold dilution in cold phosphate-buffered saline containing 5 g/ml phenylmethylsulfonyl fluoride. The cells were centrifuged at 500 ϫ g and washed with phosphate-buffered saline, and the cell pellet was lysed and immunoprecipitated. Brefeldin A and monensin were preincubated with transfected COS cells in methioninefree medium for 60 min followed by a 2-h labeling period in the presence of the drugs. The 20°C temperature block was performed by preincubation in methionine-free DMEM for 1 h at 20°C followed by labeling with 150 Ci/ml [ 35 S]methionine for 3 h at 20°C using a 20°C water bath in a 4°C cold room.

RESULTS
COS cells were transiently transfected with wild-type APP 695 and APP 751 and Sw-APP 751 cDNAs and metabolically labeled followed by immunoprecipitation of ␤A and APP from * This work was supported by National Institutes of Health Grants AG09229 and NS30352, the Alzheimer's Association, and Sandoz Pharma LTD. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed: Dept. of Neurology, Enders 260, The Children's Hospital, 300 Longwood Ave., Boston, MA 02115. Tel.: 617-355-7220; Fax: 617-738-1542. 1 The abbreviations used are: ␤A, ␤-amyloid; APP, amyloid precursor protein; Sw-APP, Swedish mutant amyloid precursor protein; DMEM, Dulbecco's modified Eagle's medium; FCS, fetal calf serum. THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 270, No. 45, Issue of November 10, pp. 26727-26730, 1995  the culture medium and cell lysate. Cells expressing Sw-APP secreted about 10-fold higher levels of the 4-kDa ␤A peptide than cells expressing wild-type APP 695 when normalized for APP (Fig. 1A), consistent with previous reports (Citron et al., 1992;Cai et al., 1993). Immunoprecipitation of the detergentsoluble lysate of cells expressing Sw-APP revealed the presence of a labeled 4-kDa peptide, which was not detected in cells expressing wild-type APP 695 and APP 751 or in mock-transfected cells (Fig. 1B). The 4-kDa peptide immunoreactivity was abolished by preabsorption of the ␤A antibody with synthetic ␤1-40 peptide (Fig. 1D) and did not appear after immunoprecipitation with an antibody to the APP C-terminal domain. When cells were incubated at 4°C, the generation of cellassociated ␤A was completely inhibited (Fig. 1C), suggesting that cell-associated ␤A is produced by active APP processing and not as an artifact of proteolysis during the immunoprecipitation protocol. Cell-associated ␤A was recovered in the supernatant after high speed centrifugation of the detergent-extracted lysate suggesting that it is soluble (data not shown). These results suggest that the cell-associated 4-kDa peptide is antigenically and physically similar to the secreted 4-kDa ␤A peptide. The low level of cell-associated ␤A precluded sequence analysis of the N terminus.

Communication
To determine whether cell-associated ␤A in cells expressing Sw-APP is intracellular or associated with the cell surface, cells were treated with trypsin prior to lysis and immunoprecipitation. The level of cell-associated ␤A was only slightly decreased by trypsinization (Fig. 1D), suggesting that it is mostly intracellular. In some labeling experiments, the 3-kDa peptide was detected in a cell-associated form. In contrast to the 4-kDa peptide, the cell-associated 3-kDa peptide was trypsin-sensitive (Fig. 1D), suggesting that it is mostly associated with the plasma membrane. We then determined whether intracellular ␤A could be due to uptake of ␤A from the medium. Conditioned medium from Sw-APP-expressing cells containing high levels of radiolabeled ␤A was incubated for 6 and 24 h with nontransfected COS cells. The level of labeled ␤A in the medium was only slightly decreased after this incubation period (Fig.  1E). A low level of radiolabeled ␤A was detected in association with nontransfected cells after incubation with radiolabeled ␤A but did not exceed 20% of the level of cell-associated ␤A in Sw-APP-transfected cells (Fig. 1E). Thus, most of the intracellular ␤A produced from Sw-APP is not derived by uptake of ␤A from the medium. These results suggest that cell-associated ␤A is intracellular and is predominantly generated from intracellular processing of Sw-APP.
The kinetics of production of intracellular and secreted ␤A were examined by pulse-chase labeling of HEK 293 cells stably transfected with Sw-APP. Intracellular and secreted ␤A both appeared by 30 min of chase time (Fig. 2, A and B). Intracellular ␤A continued to accumulate after 45-and 60-min chase times and then declined by 120 min. In contrast, ␤A in the medium continued to accumulate up to 120-min chase time (Fig. 2, A and B). Intracellular ␤A reached a maximum of about 15% of secreted ␤A at 60-min chase time. We also examined the time course of appearance of the 11.5-and 9-kDa C-terminal fragments of APP, which are metabolic intermediates in the processing of APP to ␤A and the 3-kDa peptide, respectively (Busciglio et al., 1993;Haass et al., 1993;Higaki et al., 1995). The 11.5-kDa fragment appeared by 15-min chase time, which slightly preceded the appearance of intracellular and secreted ␤A (Fig. 2, A and B). In contrast, the ␣-secretase-generated 9-kDa C-terminal fragment accumulated more slowly, reaching a maximal level after 90-min chase time (data not shown). The differential kinetics of production of intracellular and secreted ␤A suggest that they are processed differently and provide further evidence that intracellular ␤A is not a contaminant of extracellular ␤A or a product of ␤A endocytosis.
To determine the sites of generation of intracellular ␤A from Sw-APP, we examined the effects of inhibitors of processing Transfected COS cells expressing APP 695 , APP 751 , or Sw-APP 751 and mock-transfected cells were metabolically labeled with [ 35 S]methionine for 16 h. The culture medium and cell lysates were immunoprecipitated for ␤A as described under "Experimental Procedures." C, temperature dependence of intracellular ␤A generation. Sw-APP-expressing cells were pulse-labeled at 37°C for 10 min and then chased for 90 min at either 4 or 25°C. Incubation at 4°C completely prevents generation of intracellular (Cell) and secreted (Sec) ␤A. D, trypsin resistance of cell-associated ␤A. Sw-APP-expressing cells were untreated (Cntrl) or incubated with trypsin for 10 or 20 min at 37°C, and the cell lysates were immunoprecipitated with the ␤1-40 antibody. Preabsorption of the antibody with synthetic ␤1-40 peptide (Preab) abolishes immunoreactivity. E, cell-associated ␤A in Sw-APP-expressing cells is not due to ␤A uptake from the medium. The first three panels show cellassociated ␤A in Sw-APP transfectants (Cntrl) and in non-transfected COS cells incubated for 6 or 24 h with medium containing radiolabeled ␤A. The last two panels show ␤A remaining in the conditioned medium after 6-and 24-h incubations with non-transfected COS cells. Radiolabeled ␤A is derived from the medium of metabolically labeled Sw-APP transfectants. and transport in the secretory pathway. Pretreatment of cells with brefeldin A causes resorption of the proximal Golgi into the endoplasmic reticulum and inhibits anterograde transport and maturation of APP in the secretory pathway (Caporaso et al., 1992;Gabuzda et al., 1994). Brefeldin A completely inhib-ited the production of both intracellular and secreted ␤A from Sw-APP, suggesting that these processing steps are unlikely to occur in the endoplasmic reticulum or proximal Golgi (Fig. 3A). We also examined the effects of the ionophore monensin, which inhibits the maturation of newly synthesized proteins at the trans-Golgi and their transport past the trans-Golgi network (Tartakoff, 1983). Incubation with 7.5 M monensin almost completely inhibited ␤A secretion in both Sw-APP and wildtype APP-expressing cells but only partially inhibited the generation of intracellular ␤A in cells expressing Sw-APP (Fig. 3A). PhosphorImager analysis showed that monensin inhibited ␤A secretion by 99 Ϯ 0.2% but inhibited the generation of intracellular ␤A by only 49 Ϯ 5%. A monensin doseresponse analysis showed that ␤A secretion was almost completely inhibited by 2.5 M monensin, whereas intracellular ␤A production was only partially inhibited by 2.5 M monensin, and no further inhibition was evident up to 10 M monensin (Fig. 3B). Cell-associated APP increased in monensintreated cells, as previously reported (Gabuzda et al., 1994). The differential sensitivity of intracellular ␤A to monensin and brefeldin A is consistent with a site of generation in the trans-Golgi. In contrast, the inhibition of ␤A secretion by monensin, without increased cellular retention of ␤A, is consistent with the generation of secreted ␤A in a post-Golgi compartment.
To further assess the site of intracellular ␤A generation, we examined the effects of a 20°C temperature block, which results in protein retention in the trans-Golgi (Matlin and Simons, 1983). APP synthesis occurred at 20°C but was significantly reduced (data not shown). Hence, determinations of ␤A were normalized for the level of APP. The 20°C temperature block almost completely inhibited ␤A secretion but only partially inhibited the generation of intracellular ␤A (Fig. 3C). The level of intracellular ␤A generated at 20°C was similar to that observed in the presence of monensin (Fig. 3, B and C). These results provide additional evidence that intracellular and secreted ␤A are generated at distinct sites in Golgi and post-Golgi compartments, respectively.  (MON). B, monensin dose response of intracellular and secreted ␤A. Note that monensin almost completely inhibits secretion of ␤A but only partially inhibits generation of intracellular ␤A. Values are expressed as percent of IC ␤A in the absence of monensin (control) and represent the mean Ϯ S.E. of three independent experiments. C, incubation at 20°C almost completely inhibits generation of secreted ␤A but only partially inhibits generation of intracellular ␤A. Values are molar ratios of ␤A:APP normalized to the 37°C value and represent the mean Ϯ S.E., n ϭ 3. *, p Ͻ 0.01 relative to control by ANOVA. DISCUSSION These experiments suggest that APP harboring the Swedish mutation is processed to ␤A at an early step in the secretory pathway giving rise to a stable intracellular pool of ␤A. Hence, the Swedish mutation results in both increased secretion and intracellular accumulation of ␤A. Although intracellular ␤A is a small fraction of the total ␤A produced, it may nevertheless play a potentially important pathogenic role in plaque formation. The appearance of ␤A in a cell-associated form has also been observed in a neuronal cell line (Wertkin et al., 1993). Although we have not observed intracellular ␤A in transfected cells that overexpress wild-type APP, we cannot exclude the possibility that there is a small pool that is below the limits of resolution. Nevertheless, our results demonstrate that the Swedish mutant APP gives rise to significantly increased accumulation of intracellular ␤A.
Several lines of evidence suggest that the intracellular and secreted forms of ␤A arise through distinct processing pathways. First, the kinetics of generation of intracellular and secreted ␤A are different. Second, the generation of intracellular ␤A is inhibited by brefeldin A but is partially resistant to monensin and a 20°C temperature block, suggesting that it occurs in the Golgi complex, most likely the trans-Golgi. In contrast, both monensin and a 20°C temperature almost completely inhibit the secretion of ␤A without increasing cellular retention of ␤A. Thus, secreted ␤A is generated in a post-Golgi compartment, which is distinct from the Golgi site of generation of intracellular ␤A. These findings are consistent with previous reports, which suggest that secreted ␤A is produced in a post-Golgi compartment (Busciglio et al., 1993;Haass et al., 1993;Higaki et al., 1995;Yamazaki et al., 1995), and are also consistent with the recent observation that the cellular site of ␤-secretase cleavage is altered in MDCK cells expressing Sw-APP (Lo et al., 1994).
The effects of the Swedish mutation on APP metabolism suggest that inherited mutations alter not only the level but also the sites of accumulation of ␤A. Increased ␤A production has also been demonstrated to result from aberrant APPs produced by deletion mutations, incorrect APP isoform expression, and APP overexpression (Zhong et al., 1994). Furthermore, we have previously demonstrated that inhibition of energy metabolism markedly increases the generation and intracellular accumulation of potentially amyloidogenic C-terminal fragments of APP (Gabuzda et al., 1994). Thus, a variety of genetic and non-genetic pathological situations can alter the processing of APP and affect both the level and cellular sites of accumulation of ␤A. It remains to be determined whether intracellular accumulation of ␤A predisposes to ␤A aggregation and plays a role in the formation of amyloid plaques.