Intracellular Accumulation of Insoluble, Newly Synthesized Aβn-42 in Amyloid Precursor Protein-transfected Cells That Have Been Treated with Aβ1–42*

Our early study indicates that intracellular Aβ1–42 aggregates are resistant to degradation and accumulate as an insoluble residue in lysosomes, where they alter the normal catabolism of amyloid precursor protein (APP) to cause the accumulation of insoluble APP and amyloidogenic fragments. In this study, we examined whether the addition of exogenous Aβ1–42 also leads to the accumulation of newly synthesized intracellular Aβ. Here we describe that newly synthesized Aβ, especially Aβn-42, is generated from metabolically labeled APP and accumulates in the insoluble fraction of cell lysates after Aβ1–42 treatment. These results suggest that intracellular Aβ may derive from a solid phase, intracellular pathway. In contrast to the pathway that primarily produces secreted Aβ1–40, the solid-phase intracellular pathway preferentially produces Aβn-42 with ragged amino termini. Biochemical studies and amino acid sequencing analyses indicate that these intracellular Aβ also share the same types of Aβ structures that accumulate in the brain of Alzheimer’s disease patients, suggesting that a significant fraction of the amyloid deposits in Alzheimer’s disease may arise by this solid-phase pathway.

The major protein component of amyloid deposits associated with Alzheimer's disease (AD) 1 is a 39 -42-amino acid, selfassembling peptide known as the amyloid A␤ peptide. Although significant progress has been made in our understanding of the proteolytic processing of amyloid precursor protein (APP) and the secretion of soluble amyloid A␤ peptide, the mechanisms for the accumulation of insoluble amyloid deposits and their role in AD pathogenesis remains a matter of speculation. It is clear that at least two pathways for APP processing give rise to fragments bearing A␤ sequences at their amino termini: processing by ␣-secretase, which cleaves within the A␤ sequence, thereby precluding amyloid accumulation, and ␤-secretase processing, which generates carboxyl-terminal APP fragments containing the A␤ sequence. Amyloidogenic, ␤-secretase processing events may take place within several intracellular organelles, including the rough endoplasmic reticulum, trans-Golgi network, and lysosomes (1)(2)(3)(4)(5)(6)(7). Further processing of APP within the transmembrane domain by ␥-secretase releases soluble 3-and 4-kDa fragments containing all or part of the A␤ sequence (8). Recent evidence indicates that the familial AD amino acid substitutions within the APP transmembrane domain and presenilin favor the production of A␤1-42 form of A␤, which is preferentially localized within diffuse plaques and senile plaques in AD brain. This suggests that A␤1-42 is more closely associated with AD pathogenesis than shorter A␤ isoforms (9,10).
Biochemical studies of synthetic amyloid peptides have elucidated several important properties regarding their ability to assemble into the amyloid fibrils that characteristically accumulate in AD. Peptides that end at residue 42 aggregate much more rapidly than those ending at residue 39 or 40 (11,12). The pH optimum for ␤-sheet formation and aggregation is between pH 4.0 and 5.5 (11,13,14). Perhaps because of the fact that it aggregates much more rapidly, A␤1-42 is resistant to degradation once it has been internalized by endocytosis. It accumulates as an insoluble residue in late endosomes or secondary lysosomes, whereas A␤1-40 and shorter peptides are degraded and eliminated (15,16). The amyloid that accumulates in AD is structurally heterogeneous with ragged amino termini, but most of the A␤ peptides end at residue 42 (17,18). The amyloid that accumulates in AD brain has the hallmarks of a long-lived protein, such as D-amino acids and isopeptide bonds (19).
The selective resistance of aggregated A␤1-42 to degradation provides a simple and direct mechanism for why A␤1-42 preferentially accumulates in the brain (17,18). The accumulation of insoluble A␤1-42 in lysosomes also alters the catabolism of APP and causes the accumulation of APP and a series of potentially APP amyloidogenic fragments (20,21). Like the internalized A␤1-42, these fragments accumulate in the insoluble fraction of the cell and display very long half-lives (20). Several features of this accumulation are analogous to prion replication (22). The prion model postulates a conformation change in the precursor protein preceding their proteolytic conversion to more prions. The APP amyloidogenic fragments appear to undergo such a conformation change, since they display an epitope that is specifically associated with A␤ aggregates (20). This suggests that they may have the same shape as aggregated A␤ and are therefore capable of adding on to the fibril lattice established by the internalized exogenous A␤1-42. The fact that they co-purify in the insoluble fraction of the cell is consistent with the suggestion that they co-aggregate (20). The accumulation of amyloid is autocatalytic as predicted for prion replication, and once amyloid core is seeded, the continued presence of exogenous A␤1-42 is not required for further accumulation of APP and amyloidogenic fragments (20). To complete the prion-like cycle, the APP amyloidogenic fragments would need to be further proteolytically processed to A␤.
In this report, we examined whether the addition of exogenous A␤1-42 causes the accumulation of newly synthesized intracellular A␤. Here we describe that 4-kDa A␤, especially A␤n-42, is produced from metabolically labeled APP molecules and accumulates in the detergent-insoluble fraction of cells that have been incubated with synthetic A␤1-42. Most of the newly synthesized A␤ peptides that accumulate have ragged amino termini and end at residue 42. The structure of these peptides is remarkably similar to the structure described for amyloid A␤ isolated from Alzheimer's brain tissue (17,18), suggesting that much of the brain amyloid may be derived from this solid-phase, intracellular pathway.

MATERIALS AND METHODS
Metabolic Labeling and Immunoprecipitation-Transfected cell cultures (1 ϫ 10 7 cells in a 10-cm plate) were incubated with methioninedeficient Dulbecco's modified Eagle's medium for 2 h before labeling. The cells were then incubated in 2 ml of methionine-deficient Dulbecco's modified Eagle's medium containing 25 M amyloid peptide and 1% bovine serum albumin and labeled with 100 Ci/ml [ 35 S]methionine/ cysteine (1000 Ci/mmol, Tran 35 S-label, ICN) for 16 h. At the end of incubation time, the cells were washed twice with cold phosphatebuffered saline and lysed with Nonidet P-40 lysis buffer (50 mM Tris (pH 8.0), 150 mM NaCl, 1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 5 mM EDTA, 2 mg/ml leupeptin, 0.2 unit/ml soybean trypsin inhibitor, 1 mg/ml aprotinin). The insoluble cell lysate was collected by centrifugation at 10,000 ϫ g for 10 min, solubilized in 88% formic acid (v/w), and lyophilized. After lyophilization, the dried sample was resolubilized with 2ϫ radioimmune precipitation buffer, sonicated until clarified, diluted to 1ϫ radioimmune precipitation buffer, and centrifuged at 10,000 ϫ g for 10 min. The supernatant of sample was then subjected to immunoprecipitation analysis with various antibodies and analyzed by SDS-polyacrylamide gel electrophoresis.

Isolation of [ 35 S]Met-labeled A␤ from the Insoluble Fractions of Cell
Lysates-APP751-overexpressing cells were preincubated with methionine-deficient Dulbecco's modified Eagle's medium for 2 h before labeling. The cells were then incubated in 2 ml of methionine-deficient Dulbecco's modified Eagle's medium containing 25 M A␤1-42 and 1% bovine serum albumin and labeled with 100 Ci/ml [ 35 S]methionine/ cysteine (1000 Ci/mmol; Tran 35 S-label, ICN) for 16 h. At the end of labeling period, the cells were then washed twice with cold phosphatebuffered saline and lysed in Nonidet P-40 lysis buffer. The insoluble cell pellet was then collected by centrifugation at 10,000 ϫ g for 10 min and solublized in 88% formic acid. The formic acid-solublized cell fraction was then injected over a Superdex 75 gel filtration column that had been previously equilibrated in 60% formic acid. The sample was eluted from the gel filtration column in 60% formic acid at 1 ml/min. The fractions containing A␤ were then pooled and lyophilized. The lyophilized material was then redissolved in 60% formic acid and subjected to a second round of reverse-phase HPLC chromatography in a Vydac C-4 column. The sample was eluted in a 5-95% ACN gradient, and fractions coeluted with synthetic A␤1-42 were collected for further analysis.
Radiochemical Sequencing of Amyloid Peptide-APP751-transfected cells were metabolically labeled with either [ 35 S]Met or [ 3 H]Phe, and the secreted or intracellular A␤ were then purified as described above. The amyloid peptides were then subjected to automated Edman degradation amino acid sequencing analysis, and the amount of [ 3 H]Phe radioactivity eluted from each sequencing cycle was then determined by liquid scintillation counting. In the case of [ 35 S]Met-labeled amyloid peptide sequencing reaction, 4,000 cpm of purified labeled A␤ was mixed with 10 g of synthetic A␤1-42 in 0.2 M ammonium bicarbonate buffer (pH 8.0) and digested with 0.1 g of L-1-tosylamido-2-phenylethyl chloromethyl ketone-trypsin (Sigma) at 37°C for 16 h. The [ 35 S]Metcontaining tryptic fragments (A␤29 -42) were then collected by centrifugation on a tabletop centrifuge at 14,000 rpm for 30 min. The pellet was then washed two times with water and dissolved in 40% ACN, 60% formic acid right before the radiochemical sequence analysis.

RESULTS
Immunoprecipitation of Newly Synthesized A␤ from the Detergent-insoluble Fraction of A␤1-42-treated Cells-Our previ-ous work indicates that A␤1-42 preferentially accumulates in late endosomes and lysosomes of both cultured human fibroblast and PC12 cells and is resistant to degradation. We examined whether the presence of intracellular A␤ affects the catabolism of APP and A␤ in APP-overexpressing human embryonic kidney 293 cells since both nonamyloidogenic and amyloidogenic APP-processing pathways have been demonstrated in this cell line. Previous studies on the uptake of 125 I-labeled A␤1-42 demonstrated that most of the internalized 125 I-labeled A␤ is sedimentable at 10,000 ϫ g (15,16). Because the amyloidogenic fragments of APP also accumulate in the nonionic detergent-insoluble fraction of cells treated with A␤1-42 (20), we investigated whether some of these fragments are ultimately converted to 4-kDa A␤.
APP-overexpressing human embryonic kidney 293 cells were treated with A␤1-42, metabolically labeled with [ 35 S]Met for 16 h, and then lysed with Nonidet P-40 lysis buffer as described under "Materials and Methods." The Nonidet P-40-insoluble fraction was then dissolved in 88% formic acid, lyophilized, and resuspended in radioimmune precipitation buffer, and the 100,000 ϫ g-soluble supernatant was immunoprecipitated with antibodies raised against A␤1-42, A␤1-28, or the carboxyl terminus of APP (13G8). All three antibodies immunoprecipitate a broad size range of labeled products from A␤1-42treated cells that are absent in untreated cells (Fig. 1A). Because of the low efficiency of immunoprecipitation with anti-A␤ antibodies due to the presence of the exogenously supplied, unlabeled A␤1-42, we devised a more specific method of immunoprecipitating A␤ using cells that were treated with synthetic A␤4 -42 instead of A␤1-42. We verified that A␤4 -42 is not recognized by an anti-A␤ monoclonal antibody, 3D6, that recognizes the first 5 residues of A␤ (Fig. 1B). When the detergent-insoluble fraction of cells treated with A␤4 -42 is immunoprecipitated with this antibody, a small amount of 4-kDa A␤ is detected (lane 5, Fig. 1A indicated by arrow). To further demonstrate that the accumulation of intracellular newly synthesized A␤ is because of the A␤ treatment, a time course analysis was performed. APP-overexpressing cells were treated with 25 M A␤, and the amount of 35 S-labeled A␤ was then detected by immunoprecipitation at indicated times. Our previously published data (20) and the results from this study ( Fig.  1A) indicate that there is a large amount of APP carboxylterminal fragments in this insoluble fraction after A␤ treatment. To improve the sensitivity of A␤ detection by eliminating the cross-reactivity between the APP carboxyl-terminal fragments and the anti-A␤ antibody, cell extracts were preabsorbed with anti-APP carboxyl-terminal antibodies (13G8) before immunoprecipitation with the anti-A␤ antibody, 3D6. As shown in Fig. 1C, there is no detectable newly synthesized [ 35 S]Metlabeled A␤ in the first 4 h of A␤ incubation. Newly synthesized A␤ is first observed after 6 h of A␤ treatment, suggesting that the 35 S-labeled A␤ is derived from the amyloidogenic APP fragments that accumulate in the insoluble fraction of cells that have internalized A␤1-42. This observation is consistent with the recent findings that A␤1-42 can be produced intracellularly in the detergent-insoluble fraction of both cultured human NT2 neurons and APP-overexpressing human embryonic kidney cells (1-3), suggesting that a common amyloidogenic pathway exists in these cells.
Purification of Intracellular Amyloid Peptide by Gel Filtration and HPLC Column Chromatography-To further characterize the newly synthesized 4-kDa A␤ generated by cells treated with A␤, we size-selected the 4-kDa fraction using gel filtration methods that were developed to isolate and characterize 4-kDa A␤ from insoluble brain amyloid deposits (19). APP-overexpressing cells were treated with A␤1-42 and met-abolically labeled with [ 35 S]Met for 6 -12 h and then lysed with Nonidet P-40 lysis buffer as described under "Materials and Methods." The Nonidet P-40-insoluble fraction was then dissolved in 60% formic acid and then loaded onto a Superdex 75HR size exclusion column and eluted in 60% formic acid as described (19,24). A peak elutes between fractions 27 and 30 that co-migrates with authentic A␤1-42 and is absent from control cells that were not treated with A␤1-42. (Fig. 2A). The A␤42-containing fractions (fractions 27-30, as indicated on Fig.  2A) were pooled and further purified by a semi-preparative reverse-phase HPLC. As shown in Fig. 2B, the majority of the 35 S-labeled material in the 4-kDa peak from the gel filtration column elutes as a broad peak from fractions 31-51, and the peak is coincident with the elution profile of synthetic A␤1-42. Therefore, most of the newly synthesized, [ 35 S]Met-labeled A␤ elutes with a profile that is identical to authentic A␤1-42. As we reported earlier and shown in Fig. 2B (inset), A␤1-40 elutes as a sharp peak on an analytical reverse-phase HPLC that precedes the broad elution profile of A␤1-42 (11). The sharp peak of [ 35 18 -22). This result indicates that the [ 3 H]Phe-labeled A␤ is "ragged" at its amino terminus, containing both longer and shorter peptides than those beginning at residue 1. To confirm that the radiochemical sequence is derived from amyloid A␤ peptides, we radiochemically sequenced [ 35 S]Met-labeled peptides after cleaving them with trypsin to create homogeneous ends. The A␤ sequence contains one methionine residue at position 35 that would occur at cycle 7 after trypsin cleavage at lysine 28. The purified intracellular [ 35 S]Met-labeled A␤ was subjected to tryptic fragmentation, and the hydrophobic carboxyl-terminal fragment from residues 28 -42 was purified as described previously (19) and subjected to radiochemical Edman sequence analysis. Sixty percent of the total starting 35 S label was recovered in the purified carboxyl-terminal fragment. As indicated in Fig. 3B, the [ 35 S]Met radioactivity begins in cycle 7, which is consistent with the predicted A␤ sequence. The lag in [ 35 S]Met radioactivity eluting at cycles 8 and 9 was also observed in the Met yield obtained chemically in the same cycles. The [ 35 S]Met is quantitatively released during sequencing, and no labeled material remained associated with the filter. These results provide independent confirmation that the radiochemical sequence obtained is derived from A␤ rather than from contaminating peptides. The presence of ragged amino termini suggests that the ends of the A␤ peptides accumulating in the insoluble fraction of the cell may be generated by relatively nonspecific proteolysis. The proteolytic trimming of exogenously supplied synthetic A␤1-42 has been previously observed for exogenously added peptide after internalization in human fibroblasts (15). In contrast, the radiochemical se- quence of A␤ isolated from cultured media is much more homogeneous, and [ 3 H]Phe radioactivity can only be detected in fractions 4, 19, and 20, suggesting the amino terminus of soluble A␤ is generated by a more specific mechanism (25). Interestingly, A␤ isolated from the plaques of the brains of Alzheimer's patients also contained such amino-terminal truncation as we describe here (18,24,26,27).
Characterization of the Carboxyl Terminus of the Newly Synthesized, Intracellular A␤-Our early study from the biochemical properties of synthetic A␤ indicated that A␤1-42 rather than A␤1-40 is able to form SDS-resistant aggregates and migrates as a 16-kDa band on SDS-polyacrylamide gel electrophoresis at high concentrations (11,28). The specificity of SDSresistant A␤ aggregates is demonstrated in Fig. 4A, where trace amounts of 14  buffer. The insoluble cell lysate was then extracted with 60% formic acid. After centrifugation, the formic acid-soluble fraction was injected over a Superdex 75 gel filtration column. The sample was then eluted from the gel filtration column in 60% formic acid at 1 ml/min, and the absorbance was monitored at 254 nm. B, the intracellular [ 35 S]Metlabeled A␤ was mixed with 20 g of synthetic A␤ in 1 ml of 88% formic acid and subjected to reverse-phase HPLC in a semi-preparative Vydac C-4 column as described (15). The sample was then developed in a 5-95% ACN gradient, fractions were collected, and the amount of radioactivity was then determined by scintillation counting. As indicated, 35 S-labeled peptide (q) was coeluted with unlabeled synthetic A␤1-42 and displays a broad elution profile that is a characteristic feature of A␤1-42 peptide and is not a reflection of heterogeneity (see inset, 2 g each of synthetic A␤1-40 and A␤1-42 were injected onto a Vydac C-4 analytical reverse-phase column. The sample was then eluted by a 5-95% ACN gradient).

FIG. 3. Radiochemical sequencing of intracellular A␤ peptide.
A, APP-overexpressing cells treated with A␤ 1-42 were metabolically labeled with [ 3 H]Phe, and A␤ purified from the reverse-phase HPLC were pooled and radiochemically sequenced by automated Edman degradation. The results indicates that the amino terminus of the intracellular A␤ was ragged and may have been subjected to limited proteolysis as previously observed for the exogenously added peptides. B, in the case of [ 35 S]Met-labeled amyloid peptide-sequencing reaction, 4,000 cpm of purified, labeled A␤ was mixed with 10 g of A␤1-42 in 0.2 M ammonium bicarbonate buffer (pH 8.0) and digested with 0.1 g L-1tosylamido-2-phenylethyl chloromethyl ketone-trypsin (Sigma) at 37°C for 16 h. The [ 35 S]Met-containing tryptic fragments (A␤29 -42) were then collected by centrifugation on a tabletop centrifuge and subjected to Edman radiochemical sequence analysis. The majority of 35 S-Met was only detected in the cycle 7, which is consistent with the predicted A␤ sequence. metabolically labeled A␤1-42 and A␤1-40 and shorter A␤ peptides. The HPLC-purified, [ 35 S]Met-labeled intracellular A␤ was mixed with synthetic A␤1-28, A␤1-42, and A␤1-40 at a concentration of 500 g/ml at pH 5.0 for 24 h. The labeled peptide forms 16-kDa SDS-resistant aggregates only with synthetic A␤1-42 but not A␤1-40 (indicated by the arrowhead, Fig. 4B), suggesting that a fraction of the newly synthesized [ 35 S]Met A␤ from the insoluble fraction of cell lysate is A␤1-42 or closely related products that are capable of co-assembling with synthetic A␤1-42 into SDS-resistant aggregates.
Recently, an analytical, chromatographic method for separating A␤1-42 from A␤1-40 was described that employed reverse-phase separation at high temperature (29). As indicated in Fig. 5A, A␤1-42 can be resolved as a single peak on a Zorbax C-18 reverse-phase column at 65°C and is well separated from A␤1-40 by almost 2 min. The purified 4-kDa radiolabeled peptides were dissolved in 1 ml of 70% formic acid and injected onto a Zorbax C-18 reverse-phase column. The radioactive sam-ple was then eluted by a 20 -45% ACN gradient at 65°C, and the fractions that coeluted with synthetic A␤1-42 peptides were collected for further analysis (Fig. 5B). The radiolabeled peptides elute as two major peaks at the positions of both A␤1-40 and A␤1-42, even though the amino termini of the peptides were determined to be ragged by radiochemical sequencing. Our experience is that the carboxyl terminus of A␤ has a dominant influence on their chromatographic behavior (11), and synthetic amino-terminal deletions of A␤1-42 and A␤1-40 tend to co-migrate with the peptides beginning at position 1 (data not shown). Taken together, these results indicate that a mixture of A␤n-42 and A␤n-40 are produced from APP and accumulate in cells treated with A␤1-42. DISCUSSION Our results suggest that a solid-phase pathway for the simultaneous production and accumulation of amyloid A␤ exists within late endosomes or secondary lysosomes of cells that contain degradation-resistant A␤1-42 aggregates. We have previously shown that the presence of intracellular A␤1-42 aggregates alters the normal catabolism of APP to cause the accumulation of APP and potentially amyloidogenic APP fragments in lysosomes of cultured cells. In this report, we have demonstrated that newly synthesized 4-kDa A␤ also accumulates in the same insoluble fraction as the amyloidogenic APP fragments. Because of the requirement for exogenously added synthetic A␤1-42 to initiate the accumulation of insoluble frag- ments of APP, we must rely on radiochemical methods to unambiguously distinguish the newly synthesized A␤ that is derived from the transfected APP gene from the synthetic A␤1-42 added to the cell cultures. A [ 35 S]Met-labeled 4-kDa band is immunoprecipitated from the formic acid-soluble fraction with a variety of anti-A␤ antibodies. The efficiency of immunoprecipitation of 4-kDa A␤ is enhanced by using A␤4 -42 instead of A␤1-42 to prime the cells and immunoprecipitating with 3D6, a monoclonal antibody specific for the first 5 residues of A␤. Similarly, the efficiency of immunoprecipitation of 4-kDa A␤ is improved by immunodepletion of crossreacting carboxyl-terminal fragments of APP.
Like the amyloid deposits from AD brain, the detergentinsoluble fraction of A␤1-42-treated cells is largely solubilized in formic acid; therefore, we have used the purification methods employed for purifying A␤ from AD brain to purify and characterize the 4-kDa peptides in the insoluble fraction of A␤1-42-treated cells. The fractions that elute in formic acid from a Superdex HR75 column at the position of denatured 4-kDa A␤1-42 were pooled and further purified by reversephase HPLC. Most of the radiolabeled material elutes from the reverse-phase HPLC at the same position as A␤1-42. Radiochemical sequencing of this purified material indicates that the amino termini of these peptides is heterogeneous, with peptide ends both longer and shorter than those beginning at residue 1. The heterogeneity may also be because of non-A␤ contaminants, but after trypsin cleavage to create homogeneous cleavage sites, most of the [ 35-S]Met label is observed at cycle 7 as expected for A␤. The carboxyl terminus of the newly synthesized 4-kDa product was characterized by analytical reversephase HPLC under conditions that resolve A␤1-42 and A␤1-40. Although labeled material is recovered in both peaks, most of the radioactivity is associated with A␤1-42. Perhaps the most convincing evidence that the newly synthesized 4-kDa peptides are A␤1-42 and closely related structures is the ability to co-assemble into SDS-resistant aggregates with authentic A␤1-42 at concentrations above the critical concentration for aggregation. This property is specific to A␤1-42 and A␤1-43, since no SDS-resistant aggregates are observed for A␤1-40 even in the presence of a vast excess of A␤1-42. Thus, the intracellular A␤ displays a number of similarities to the amyloid that accumulates in AD brain tissue. The intracellular amyloid is insoluble in nonionic detergents, but like amyloid isolated from brain, a substantial fraction of the amyloid is soluble in formic acid (17,18). Both the intracellular amyloid and brain amyloid display considerable amino-terminal heterogeneity, and the majority of the amyloid peptides end at residue 42 (17,18).
The finding that A␤n-42 accumulates inside cells treated with amyloid A␤1-42 provides further support for the hypothesis that amyloid accumulation is an autocatalytic process mechanistically related to prion replication. Like PrP sc , the core structure of aggregated A␤n-42 is resistant to proteolysis, both inside cells and in vitro (15,16,30). The fact that metabolically labeled, newly synthesized A␤ accumulates indicates that at least a fraction of the APP molecules are converted to more A␤ as the prion replication model predicts. Because the amyloid core is resistant to degradation, this conversion may be carried out by nonspecific proteolysis and exo-peptidase activities. This may explain why the A␤ that accumulates in the insoluble fraction displays ragged amino and carboxyl termini. Misfolded APP molecules and amyloidogenic fragments appear to be intermediates in amyloid accumulation because they also accumulate in the nonionic detergent-insoluble fraction of the cell where they turn over very slowly (20). At least a subset of these amyloidogenic fragments appear to have adopted the same conformation as aggregated A␤ because they display an unique conformation-dependent epitope that is only detected in aggregated A␤ (20). This suggests that the misfolded amyloidogenic APP fragments may be capable of binding to aggregated A␤n-42 and extending the amyloid lattice. Because both amyloidogenic carboxyl-terminal APP fragment precursors and the 4-kDa A␤ product accumulate in the insoluble fraction, the simplest hypothesis is that this conversion takes place in the solid phase.
Co-aggregation of poorly metabolized A␤ and amyloidogenic fragments of A␤ may explain why the intracellular A␤ immunoreactivity is not morphologically recognizable as amyloid fibrils. Because the insoluble aggregates are a heterogeneous collection of fragments, perhaps their underlying fibrillar lattice is not revealed until the fragments have been digested to their protease-resistant amyloid core. The fact that the intracellular amyloidogenic fragments themselves are poorly degraded suggests that this conversion process may be quite slow and may even occur extracellularly, after the insoluble residue has been externalized, either by exocytosis or the death of the cell. This accumulation appears to mimic the "granulovacuolar" pathophysiology of degenerating neurons and dystrophic neurites, where A␤ and APP immunoreactivity have been localized to granular or globular deposits (31)(32)(33) that are also positive for ubiquitin immunoreactivity (34,35). This may represent the site at which the intracellular A␤ and APP-insoluble residue may be externalized, perhaps by breaking off of vesicles from neurite termini. How this insoluble residue is ultimately converted to A␤ is unknown, but the microglia surrounding the plaque may phagocytose the residue and digest it to the protease-resistant amyloid core. It has been proposed that microglia may play a role in depositing amyloid fibrils in senile plaques (36). Because the center of neuritic plaques and mature or cored plaques are not positive for the non-A␤ epitopes (31)(32)(33), these may represent regions of the amyloid deposits where the conversion process is complete.
Although the significance of intracellular A␤ immunoreactivity and the origin of extracellular A␤ deposits is still controversial, a recent report by Cataldo et al. (37) indicates that lysosomal hydrolases can be detected in extracellular amyloid deposits of AD and Down syndrome brains. This is consistent with the lysosomal origin of these deposits under pathological conditions (37)(38)(39). A␤ immunoreactivity has been detected in the vacuole of chloroquine-induced rat soleus muscle cells, and immunohistochemical studies indicate that most of A␤ in the vacuoles reacts with anti-A␤1-42, and only a few react with A␤1-40-specific antibodies (40). Our results are also consistent with several recent findings that there is an intraneuronal amyloid pool that accumulates with time in culture (1,23). How the solid-phase pathway for A␤1-42 accumulation relates to the intracellular A␤ immunoreactivity in human brain and AD pathogenesis remains a challenge for further experimentation.