The Chaperone BiP/GRP78 Binds to Amyloid Precursor Protein and Decreases Aβ40 and Aβ42 Secretion*

Recent studies of cellular amyloid precursor protein (APP) metabolism demonstrate a β-/γ-secretase pathway resident to the endoplasmic reticulum (ER)/Golgi resulting in intracellular generation of soluble APP (APPsβ) and Aβ42 peptide. Thus, these intracellular compartments may be key sites of amyloidogenic APP metabolism and Alzheimer’s disease pathogenesis. We hypothesized that the ER chaperone immunoglobulin binding protein (BiP/GRP78) binds to and facilitates correct folding of nascent APP. Metabolic labeling and immunoprecipitation of transiently transfected human embryonic kidney 293 cells demonstrated co-precipitation of APP with GRP78, revealing their transient interaction in the ER. Maturation of cellular APP was impaired by this interaction. Furthermore, the levels of APPs, Aβ40, and Aβ42 recovered in conditioned medium were lower compared with cells transfected with APP alone. Co-expression with APP of GRP78 T37G, an ATPase mutant, almost completely blocked cellular APP maturation as well as recovery of APPs, Aβ40, and Aβ42 in conditioned medium. The inhibitory effects of GRP78 and GRP78 T37G on Aβ40 and Aβ42 secretion were magnified by co-expression with the Swedish mutation of APP (K670N/M671L). Collectively, these data suggest a transient and direct interaction of GRP78 with APP in the ER that modulates intracellular APP maturation and processing and may facilitate its correct folding.

The major components of amyloid plaque in Alzheimer's disease (AD) 1 brain are A␤ peptides, including A␤40 and A␤42, that are derived from amyloid precursor protein (APP). APP is metabolized at or near the cell surface by an ␣-secretase that results in soluble APP (APPs␣) secretion and precludes A␤ formation. APP is also metabolized by an endosomal/lysosomal (endocytic) pathway that results in A␤ secretion (1)(2)(3). Recent data with human NT2 neurons demonstrates that A␤ is found intracellularly (4) with kinetics identical to APP synthesis, suggesting that a fraction of nascent APP is immediately metabolized to A␤ (5). Subsequently, a ␤-/␥-secretase pathway of APP metabolism resident to the endoplasmic reticulum (ER)/ Golgi of neurons was identified, resulting in intracellular APPs␤ and A␤42 generation (6 -11). This exocytic pathway may be specifically promoted by presenilin-1 and presenilin-2 mutations found in some pedigrees of familial AD, since these proteins localize to the ER/Golgi (12)(13)(14) and result in greater A␤42 secretion (15)(16)(17). In fact, common to all mutations of APP and presenilins linked to early-onset familial AD is their ability to promote A␤42 generation (1)(2)(3). Because A␤42 is intrinsically more amyloidogenic than A␤40 and deposits preferentially in brain, this relatively minor pathway of APP metabolism within the ER/Golgi may have major implications for AD pathogenesis.
All proteins destined for the cell membrane or secretion must first translocate into the ER. Newly translocated proteins are folded and assembled by a group of proteins, which include immunoglobulin-binding protein (BiP)/glucose-regulated protein, 78 kDa (GRP78), glucose-regulated protein, 94 kDa (GRP94), peptidyl prolyl isomerase, calnexin, and protein disulfide isomerase. Paramount among these ER residents is the highly conserved ATP-binding protein GRP78, which associates transiently with many polypeptides and more stably with misfolded or incompletely assembled proteins (18). Association with GRP78 is hypothesized to prevent misfolding and aggregation of nascent polypeptides during synthesis and assembly in the ER. Misfolded proteins in the ER remain bound to GRP78 and are destined for degradation. In mammalian cells, GRP78 is detected noncovalently bound to a variety of proteins, including immunoglobulin heavy chain, nicotinic receptor, HIV-1 envelope protein, T cell receptor ␣ chain variants, influenza hemagglutinin, and type I procollagen pro␣ chain (for reviews, see Refs. 19 and 20). This wide variety of substrates of GRP78 suggests that the correct folding of APP may also require transient binding to GRP78 in the ER.
As is true for all Hsp70 proteins, the amino-terminal twothirds of GRP78 comprise an ATP binding domain, and the carboxyl-terminal third contains a peptide binding domain. GRP78 binds ATP and has weak ATPase activity. Elucidation of the three-dimensional structure of this domain led to the development of mutants such as GRP78 T37G with impaired ATP-induced release of bound protein (21,22). This ATPase mutant also blocks assembly and folding of immunoglobulin heavy chains (23). Thus the GRP78 T37G mutant functions as a molecular trap by stabilizing the normally transient interaction of newly synthesized polypeptide with GRP78. ATPasedefective GRP78 does not bind to and retard all secreted proteins. For example, Factor VIII secretion is reduced by ATPasedefective GRP78 co-expression, but monocyte/macrophage colony-stimulating factor is not (24). We hypothesized that: 1) GRP78 binds to APP in the ER, 2) this interaction has functional consequences on APP metabolism, and 3) the transient interaction of GRP78 with APP may be captured by co-expression with GRP78 T37G.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-Human embryonic kidney (HEK) 293 cells were cultured in Dulbecco's modified Eagle's medium containing 100 units of penicillin/ml and 100 g/ml of streptomycin sulfate supplemented with 10% heat-inactivated fetal bovine serum and 2 mM glutamine (Life Technologies, Inc.). Human APP or APP Swe was cloned into pRK5 (25). The APP751 isoform was used exclusively in this study. Hamster GRP78 and GRP78 T37G were each cloned into pcDNA3 (Invitrogen). Hamster GRP78 is greater than 99% identical to human GRP78. HEK 293 cells were split 1 day prior to transfection (1 ϫ 10 6 cells/6-cm dish) and transfected with 10 g of DNA by the calcium phosphate procedure or LipofectAMINE (Life Technologies, Inc.) as described by the manufacturer.
Metabolic Labeling and Immunoprecipitation-Forty-four hours after transfection, cells were labeled with [ 35 S]methionine and [ 35 S]cysteine for 4 h. Conditioned medium was recovered and cell lysates were prepared with 1 ml of lysis buffer (1% Nonidet P-40 in 50 mM Tris, 150 mM NaCl, and 5 mM EDTA, pH 8.0). Cell lysates were centrifuged to precipitate insoluble material. The cleared supernatants were equally divided into a pair of tubes containing either Karen, a polyclonal antiserum raised to the secreted amino terminus of APP, or anti-rodent GRP78 antibody (anti-GRP78), which does not cross-react with endogenous human GRP78. Protein-antibody complexes were incubated with protein A-Sepharose (Sigma, 25 l/sample) at 4°C for 30 min. After washing, the beads were boiled in Laemmli solution (50 mM Tris, pH 6.8, 2% SDS, 10% glycerol, 1% ␤-mercaptoethanol, and 0.01% bromphenol blue) and proteins separated by 8% SDS-PAGE. Radiolabeled proteins were detected by fluorography after incubation with Amplify (Amersham Pharmacia Biotech).
Immunoblot-Six hours after transfection, conditioned medium was collected for 40 h, and APPs was immunoprecipitated with Karen polyclonal anti-APP antibody. APP from cell lysates was similarly immunoabsorbed. GRP78 was detected by resolving equal sample aliquots of cell lysates by SDS-PAGE. All proteins were separated by 10% SDS-PAGE and transferred to PROTRAN (Schleicher and Schuell). The membranes were blocked in 5% non-fat dry milk in TBS (50 mM Tris, 150 mM NaCl, pH 7.6) and incubated with 22C11 (Boehringer Mannheim), a monoclonal antibody directed against an epitope in the extracellular domain of APP, or anti-GRP78 at 4°C overnight. Excess antibody was removed by washing in TBS-T (0.1% Tween) and then horseradish peroxidase-conjugated secondary antibodies were added for 1 h at room temperature. Membranes were washed and signals detected by chemiluminescence using the ECL system (Amersham Pharmacia Biotech).
ELISA-Six hours after transfection, conditioned medium was collected for 40 h and analyzed by a sandwich ELISA using BAN50 as the capture antibody and either horseradish peroxidase-coupled BA-27 or BC-05 as the detection antibody for A␤40 or A␤42, respectively (26). BAN-50 is a monoclonal antibody specific for A␤1-10.

RESULTS
To determine whether GRP78 was capable of binding to APP within the ER, HEK 293 cells were co-transfected with APP and either empty vector (pcDNA3), GRP78, or GRP78 T37G. Processing of the Swedish mutant APP (APP Swe; K670N/ M671L) in HEK 293 cells results in elevated levels of A␤40 and A␤42 compared with wild type APP (27,28). Therefore, similar co-transfections were also performed with APP Swe and GRP78. Following the co-transfections, cells were metabolically labeled with [ 35 S]methionine and [ 35 S]cysteine, and subsequently, proteins from cell lysates or conditioned medium were immunoprecipitated with the anti-APP antiserum Karen (Fig.  1, A and C) or anti-GRP78 (Fig. 1B) as described under "Experimental Procedures." Both mature (top band) and immature (indicated by arrow) forms of APP and APP Swe were detected in the lysates of transfected 293 cells (Fig. 1A). When APP or APP Swe were immunoprecipitated with anti-APP antibody, an additional 78-kDa protein was faintly detected in samples co-transfected with GRP78 (Fig. 1A, lanes 2 and 5, respectively). This 78-kDa protein was not detected in co-transfections of APP or APP Swe with empty vector (Fig. 1A, lanes 1 and 4,  respectively). As predicted, when the ATPase-defective mutant, GRP78 T37G, was co-transfected with either APP or APP Swe, a much greater amount of the 78-kDa protein co-precipitated from cell lysates with anti-APP antibody (Fig. 1A, lanes 3 and  6, respectively). Additionally, a decrease in the amount of mature APP and APP Swe was consistently detected in lysates of cells co-transfected with GRP78 T37G.
We next sought to determine which forms (mature and/or immature) of APP and APP Swe were bound by BiP. Therefore, the same experiment as above was performed, but lysates were immunoprecipitated with anti-GRP78. A small but reproducible amount of 95-kDa APP or APP Swe co-transfected with GRP78 was precipitated by anti-GRP78 (Fig. 1B, lanes 2 and  5). Again, this transient interaction was stabilized by co-transfection of APP or APP Swe (lanes 3 and 6) with GRP78 T37G. A single protein corresponding in molecular weight to immature APP or APP Swe was detected. No larger molecular weight protein equivalent to mature APP was detected in these lanes, suggesting that GRP78 and GRP78 T37G bound only to immature APP.
We hypothesized that the association of APP with GRP78 in the ER may impede the secretion of APPs into conditioned medium. To test this hypothesis, radiolabeled APPs in condi- tioned medium was immunoprecipitated with Karen antibody (Fig. 1C). A moderate but reproducible decrease in APPs was detected in the media of cells co-transfected with GRP78 (lanes 2 and 5) compared with empty vector (lanes 1 and 4). A more significant reduction of APPs recovery was observed for cells co-transfected with GRP78 T37G and either APP or APP Swe (lanes 3 and 6).
The radiolabeled protein migrating above immature APP (Fig. 1A) may be either mature, fully glycosylated APP, or an unrelated co-precipitating protein. To distinguish between these possibilities, similar experiments were conducted and proteins in conditioned medium or cell lysates detected by immunoprecipitation and immunoblot (Fig. 2). The larger molecular weight protein above immature APP was recognized by the mouse monoclonal anti-APP antibody (22C11), verifying that it is mature, glycosylated APP (labeled M, Fig. 2A). Cotransfection of APP or APP Swe with GRP78 or GRP78 T37G impaired APP maturation ( Fig. 2A, lanes 5, 6, 8, and 9). In addition, transfection with GRP78 T37G consistently resulted in accumulation of immature endogenously expressed APP in cell lysates (labeled I, Fig. 2A, lane 3). Recovery of APPs in conditioned medium was reduced by co-transfection with GRP78, and this inhibitory effect was more pronounced with the GRP78 T37G mutation (Fig. 2C), despite equivalent levels of GRP78 or GRP78 T37G expression (Fig. 2B).
We next hypothesized that the binding and retention of APP in the ER by GRP78 would modulate A␤40 and A␤42 secretion. HEK 293 cells were transfected with vector (pRK5), APP, or APP Swe in the absence or presence of vector (pcDNA3), GRP78, or GRP78 T37G. A␤40 and A␤42 were measured by ELISA (Fig. 3). Transfection with APP resulted in measurable and reproducible levels of A␤40 and A␤42 in conditioned medium. Analogous to results with APPs (Figs. 1C and 2C), the secreted A␤40 and A␤42 levels were reduced by co-transfection of APP with GRP78 and more strongly diminished with GRP78 T37G. The levels of total APP expression from all these cell lysates were equivalent (not shown). Thus, the decreases in secreted A␤ peptides appear due to an effect of GRP78 on the catabolism of APP and not on APP synthesis. This conclusion is further supported by transfections with APP Swe, which results in greater concentrations of both A␤40 and A␤42 compared with normal APP (27,28). When co-expressed with APP Swe, the inhibitory effects of GRP78 and GRP78 T37G on A␤40 and A␤42 secretion were magnified.

DISCUSSION
This study demonstrates co-precipitation of APP with GRP78 and functional consequences of this interaction on APP metabolism. Specifically, the maturation of cellular APP was impaired, and APPs, A␤40, and A␤42 recovery in conditioned medium was reduced. The binding of GRP78 to proteins is ordinarily transient and may be difficult to detect. Thus, coprecipitation and functional effects obtained with GRP78 and APP were magnified by the GRP78 T37G ATPase mutation, which fails to release transiently bound protein, and by the APP Swe mutation that results in greater A␤40 and A␤42 FIG. 2. GRP78 and GRP78 T37G retard APP maturation and inhibit APPs secretion. Cells were transiently transfected with the constructs indicated. GRP78 in cell lysates was detected by immunoblot with anti-GRP78 (B). APP in cell lysates and APPs in conditioned medium were detected by immunoprecipitation with Karen followed by immunoblot with 22C11 (A and C). Immature APP is indicated by I and mature glycosylated APP by M. secretion. The APP detected bound to GRP78 was of a single molecular weight corresponding to the immature form. The mature or intermediate forms of APP did not co-precipitate even with the mutant GRP78 T37G. Thus, the transfected APP bound to BiP was not saturating the system retaining it in the ER. These results also suggest that the membrane-spanning APP is normally transiently bound to and retained in the ER as a nascent polypeptide by the chaperone GRP78. GRP78 may subsequently interact with the KDEL receptor causing a net retention of the GRP78⅐APP complex in the ER lumen. Alternatively, because GRP78 binding impedes APP maturation in the ER, it may be retained via interaction with other chaperones such as calnexin (for review, see Ref. 29). Future experiments will distinguish between these possibilities.
Levels of APPs, A␤40, and A␤42 secretion were reduced by interaction of APP with GRP78. This transient interaction may impair access of APP to ␤-/␥-secretases within the ER/Golgi or may influence APP metabolism by facilitating its correct folding. Because secreted A␤42 is generated primarily in the ER, these data suggest that GRP78 binding to APP may directly or indirectly confer protection from ␤-/␥-secretases within this cell compartment. One may speculate that immediate processing of nascent APP is exclusive to misfolded APP. However, the possibility remains that less A␤42 is secreted, because it is bound to GRP78 or other proteins in the ER, such as ERAB (endoplasmic reticulum-associated binding protein) (30,31). In contrast to A␤42, secreted A␤40 is derived primarily from a post-Golgi compartment. The decrease observed in APPs and A␤40 secretion may, therefore, result from APP retention in the ER and thus substrate depletion. HEK 293 cells stably expressing APP V717G produce intracellular A␤42 even in the presence of brefeldin A (32), suggesting that APP cleavage occurs in the ER or cis Golgi of these non-neuronal cells. The effects of GRP78 co-expression with APP V717G or other carboxyl-terminal APP mutations found in some pedigrees of familial AD on its metabolism are as yet unknown.
Because amyloid in the aged or AD brain is thought to be generated primarily by neurons, and metabolism of APP is more amyloidogenic in neuronal than in non-neuronal cells, it will be of interest to examine the effects of GRP78 overexpression on APP metabolism in a neuronal cell line. Furthermore, intracellular A␤ is more readily detectable in neuronal cells (4,5,9,10,33). As described above for HEK 293 cells, intracellular A␤42 is detected in NT2 neurons even in the presence of brefeldin A, suggesting that A␤42 is generated in the ER or intermediate compartment (6,7). Some intracellular A␤42 exists in a detergent-insoluble form and thus is detected by ELISA only after formic acid extraction, suggesting A␤42 aggregates intracellularly (11). Thus, in view of recent hypotheses concerning a seminal role for A␤42 generation and aggregation in the ER in AD pathogenesis, GRP78 binding to APP in neurons may have even greater functional consequences than in non-neuronal cells.