The X11α Protein Slows Cellular Amyloid Precursor Protein Processing and Reduces Aβ40 and Aβ42 Secretion*

Constitutive amyloid precursor protein (APP) metabolism results in the generation of soluble APP (APPs) and Aβ peptides, including Aβ40 and Aβ42–the major component of amyloid plaques in Alzheimer’s disease brain. The phosphotyrosine binding (PTB) domain of X11 binds to a peptide containing a YENPTY motif found in the carboxyl terminus of APP. We have cloned the full-lengthX11 gene now referred to as X11α.Coexpression of X11α with APP results in comparatively greater levels of cellular APP and less APPs, Aβ40, and Aβ42 recovered in conditioned medium of transiently transfected HEK 293 cells. These effects are impaired by a single missense mutation of either APP (Y682G within the YENPTY motif) or X11α (F608V within the PTB domain), which diminishes their interaction, thus demonstrating specificity. The inhibitory effect of X11α on Aβ40 and Aβ42 secretion is amplified by coexpression with the Swedish mutation of APP (K595N/M596L), which promotes its amyloidogenic processing. Pulse-chase analysis demonstrates that X11α prolongs the half-life of APP from ∼2 h to ∼4 h. The effects of X11α on cellular APP and APPs recovery were confirmed in a 293 cell line stably transfected with APP. The specific binding of the PTB domain of X11α to the YENPTY motif-containing peptide of APP appears to slow cellular APP processing and thus reduces recovery of its soluble fragments APPs, Aβ40, and Aβ42 in conditioned medium of transfected HEK 293 cells. X11α may be involved in APP trafficking and metabolism in neurons and thus may be implicated in amyloidogenesis in normal aging and Alzheimer’s disease brain.

The finding of miliary amyloid plaques in brain parenchyma is classically recognized as a hallmark of Alzheimer's disease (AD) 1 pathology, although the role of amyloid deposition in AD is controversial. Recent data of the effects of gene mutations linked to familial AD suggests that the deposition of amyloid plaque in brain may play a causal role in the cascades leading to dementia and the pathologic abnormalities seen in AD brain: the amyloid hypothesis of AD (1)(2)(3). The major components of amyloid plaque are A␤ peptides, including A␤40 and A␤42, derived by constitutive proteolytic cleavage of amyloid precursor protein (APP) encoded on human chromosome 21. APP is a type I cell surface protein with an extracellular region, a transmembrane region, and short intracellular carboxyl-terminal cytoplasmic region. The A␤ sequence encompasses half of the transmembrane domain and a short part of the extracellular domain of APP. A␤40 and A␤42 are released by ␤and ␥-secretase activities that cleave APP at the amino and carboxyl termini of A␤, respectively. By this pathway, A␤ and soluble APP (APPs␤) are released into the extracellular space. Alternate cleavage of APP within the A␤ sequence by an ␣-secretase activity releases APPs␣ and precludes full-length A␤ formation. In nonneuronal cell lines such as HEK 293 and Chinese hamster ovary cells, secreted APP fragments are generated primarily via the ␣-secretase pathway, although some A␤ is generated and secreted by ␤-/␥-secretases, primarily in the endosomal/lysosomal pathway. In these cells, endocytosis of cell surface APP requires the Tyr-Glu-Asn-Pro-Thr-Tyr (YENPTY) motif found in its intracellular carboxyl terminus and is thus necessary for A␤ generation (4,5).
The cytoplasmic region of APP containing the YENPTY motif interacts with the PTB/PI (phosphotyrosine binding-protein interaction) domain of X11␣ (6), Fe65 (7,8), and their homologous genes X11-like and Fe65-like (9,10). X11 and Fe65 are highly expressed in neurons and contain a PTB domain originally described in Shc (11,12). The Shc PTB domain interacts with ⌿XNPXpY motifs (where ⌿ is hydrophobic, X is any amino acid, N is Asn, P is Pro, and pY is phosphotyrosine) found in tyrosine kinase receptors and other tyrosine-phosphorylated proteins. The PTB domain of Shc is likely involved in tyrosine kinase signal transduction cascades. However, the PTB domain is a more general protein-protein interaction domain found in several otherwise unrelated proteins such as X11, Fe65, Numb, and Disabled. Although the PTB domains of these proteins are homologous to Shc, they differ by binding to nonphosphorylated partners (13,14). The function of the newly described PTB domains is now being examined. For example, the PTB domain of Numb is crucial for the differentiation of sensory organ precursors in Drosophila (15,16). Although the PTB domain of X11␣ binds specifically to the YENPTY-containing region of APP, the functional significance of this interaction is unknown. Deletion of the last 18 amino acids of APP encom-passing the YENPTY motif or mutation of the amino-terminal tyrosine of the YENPTY motif of APP to glycine (Y682G) impairs binding to APP. Likewise, mutation of X11 at position 608 (F608V), previously referred to as the F479V mutation in the nonfull-length protein, impairs X11␣-APP interaction (6). The importance of these specific amino acid residues was confirmed by analysis of the crystal structure of the X11␣ PTB domain complexed to a peptide encompassing the YENPTY motif of APP (17).
Recently, we and others have identified a second X11 gene in humans. We refer to the newly isolated gene as X11␤ (Fig. 1). The goal of this study was to functionally characterize the interaction between X11␣ or X11␤ with APP. Coexpression of X11␣ with APP in human embryonic kidney (HEK) 293 cells results in comparatively greater levels of cellular APP (APPc), and less APPs, A␤40, and A␤42 recovered in conditioned medium of transiently transfected cells. These effects are 1) correlated with a prolonged half-life of APPc, 2) impaired by a single missense mutation of either APP (Y682G) or X11␣ (F608V), and 3) amplified by coexpression with the Swedish mutation of APP (APPswe; K595N/M596L) found in a pedigree of early onset familial AD. Thus, structural and functional data implicate a normal role for X11␣ in APP trafficking and metabolism via a specific protein-protein interaction.

EXPERIMENTAL PROCEDURES
Cell Culture-Human embryonic kidney 293 were grown in Dulbecco's modified Eagle's medium containing 100 units of penicillin/ml and 100 g of streptomycin sulfate/ml supplemented with 10% fetal calf serum.
Cell Transfection and Protein Extraction-Cells were split one day before transfection (1 ϫ 10 6 cells/6-cm plate) and transfected with 10 g of DNA by the calcium phosphate procedure. After 48 h, cells were washed twice with phosphate-buffered saline and lysed in lysis buffer (50 mM HEPES, pH 7.5, 10% glycerol, 150 mM NaCl, 1% Triton X-100, 1.5 mM MgCl 2 , 1 mM EGTA) supplemented with protease inhibitors (aprotinin, leupeptin, and phenylmethylsulfonyl fluoride). All constructs were cloned in pRK5 vector as described previously (6). The APP695 isoform was used exclusively in this study.
For [ 35 S]methionine labeling, cells were incubated with methioninedeficient Dulbecco's modified Eagle's medium containing 100 Ci/ml for 15 min followed by a chase in complete Dulbecco's modified Eagle's medium. After washing the cells with phosphate-buffered saline, proteins were extracted with lysis buffer. Conditioned media of transfected cells were collected before lysis, and proteins were immunoprecipitated overnight with Karen or 6E10 antibodies at 4°C. Bound proteins were recovered on protein A-agarose beads. After extensive washing with lysis buffer, proteins were separated by SDS-PAGE and detected by immunoblot or by PhosphorImager and autoradiography. Radiolabeled proteins detected by PhosphorImager were quantitated with Image-Quant software (Molecular Dynamics).
Antibodies and ELISA-The anti-myc antibody 9E10 (Oncogene Science) at 1 g/ml was used for immunoblotting. The 22C11 monoclonal antibody (Boeringer Mannheim) was directed against an epitope of the extracellular region of APP. The polyclonal antisera 369 was directed to the cytoplasmic carboxyl terminus of APP. Karen is a goat polyclonal antisera directed to the secreted amino-terminal domain of APP (18). The monoclonal antibody 6E10 (Senetek) was raised to A␤1-17. The A␤ sandwich ELISA was performed as described previously (19) using BAN50 as the capture antibody and either horseradish peroxidasecoupled BA-27 or BC-05 as the detection antibody for A␤40 or A␤42, respectively. BAN-50 is a monoclonal antibody specific for A␤1-10.

APP Mutations Affected Recovery of APPs in Conditioned
Medium-The YENPTY motif in the intracellular carboxyl terminus of APP is involved in its cellular processing. For example, deletion of this motif results in greater recovery of APPs␣ in conditioned medium (4,5). Either deletion of the YENPTY sequence or mutation of the amino-terminal tyrosine of the motif (APP Y682G) abrogates binding to the X11␣ PTB domain (6). Thus, we hypothesized that the APP Y682G mutation would recapitulate the effects of YENPTY deletion on APPs␣ recovery. HEK 293 cells were transiently transfected with APP, APP Y682G, or APPswe constructs. The APPswe double mutation resulted in comparatively greater A␤ and less APPs␣ recovery in conditioned medium. Comparable levels of APP expression were verified by immunoblot of cell lysates with the anti-APP antibody 369 ( Fig. 2A, upper panel). APPs␣ in conditioned media was immunoprecipitated and detected by immunoblot with 6E10 ( Fig. 2A, lower panel). The APP Y682G mutation resulted in greater release of APPs␣ in conditioned medium, similar to deletion of the YENPTY motif ( Fig. 2A, lower panel). As expected, APPswe resulted in a decrease in APPs␣ in conditioned medium. Transfected HEK 293 cells were also labeled with [ 35 S]methionine, and APPs␣ was recovered by immunoprecipitation with 6E10, separated by SDS-PAGE, and revealed by autoradiography (Fig. 2B). Similar to the results of Fig. 2A, APPs␣ release was markedly increased by the Y682G mutation, although comparable expression of APP was found in cell lysates (data not shown). Thus, the X11␣ PTB domain binding site in APP is functionally important for APPs␣ release. Interference with this interaction either by deletion of the carboxyl terminus or mutation of the YENPTY domain of APP resulted in greater APPs␣ release into conditioned medium.
X11␣ Expression Impaired ␣-Secretase Processing of APP-In HEK 293 cells APP is metabolized primarily by an ␣-secretase activity at the cell surface, resulting in APPs␣ secretion. Because a fraction of X11␣ protein is localized at the cell membrane, we hypothesized that X11␣ would influence APPs␣ release. HEK 293 cells were transiently cotransfected with APP and myc-tagged X11␣ constructs or control vector (Fig. 3A). APP and X11␣ in cell lysates were detected by immunoblot. APPs␣ in conditioned medium was immunoprecipitated with Karen and detected by immunoblot with 6E10.
Coexpression of APP with X11␣ resulted in decreased recovery of APPs␣ in an X11␣ dose-dependent manner. X11␣ also resulted in a large increase of APP in cell lysates. Although APPs␣ in medium was barely detectable when transfected with 5 g of the X11␣ construct, no further increase of APP in the cell lysate was detected compared with 1 g of X11␣. This might suggest that APP is being processed by a ␤-secretase pathway. Conditioned media were immunoprecipitated with Karen, and bound proteins were detected by immunoblot with 22C11, an anti-APP antibody directed against all APPs species. The same decrease in APPs was documented with this antibody, ruling out this possibility (Fig. 3A). We speculate that the generation of APP may be decreased by high expression of X11␣ or that the APP level reaches a plateau in the cell. Cells were also cotransfected with APP mutations (Fig. 3B). Coexpression of 5 g of X11␣ construct with either APP or APPswe resulted in far less APPs␣ in conditioned medium. This result was expected because APPswe does not influence X11␣ binding (data not shown). Coexpression of APP Y682G with X11␣ led only to a small decrease of APPs␣ compared with APP Y682G transfection only (Fig. 3B). Accordingly, APP Y682G retained only 5-10% of binding activity with X11␣ in vivo and in vitro (6). Collectively, this data suggested that the interaction of the PTB domain of X11␣ with the intracellular region of APP containing the YENPTY motif impaired release of APPs␣.
X11␣ Coexpression with APP Decreased A␤40 and A␤42 Recovery in Conditioned Medium-A␤ peptides, particularly A␤40 and A␤42, are also produced by constitutive APP metabolism. In contrast to ␣-secretase cleavage of APP, which precludes generation of full-length A␤ peptides, A␤40 and A␤42 are generated by ␤and ␥-secretase activities. In HEK 293 cells, A␤ peptides are generated almost exclusively by an endosomal pathway, which required a functional YENPTY motif (4). We assessed the effect of X11␣ coexpression with APP on A␤40 and A␤42 recovery in conditioned medium (Fig. 4), as measured by a sensitive and specific ELISA (19). Compared with cells transfected with X11␣, transfection with APP resulted in measurable A␤ concentrations in medium. As expected, transfection with APPswe resulted in much greater levels of A␤ in conditioned medium (Fig. 4), in parallel with diminished APPs levels (Fig. 2). Conversely, the APP Y682G mutation resulted in a slight decrease in A␤40 and A␤42 release (Fig. 4A), in parallel with a slight increase in APPs␣ (Fig.  2). Coexpression of X11␣ with APP reduced the levels of A␤40 and A␤42 in medium. This inhibitory effect of X11␣ was amplified by coexpression with the APPswe mutation. As predicted by the binding data, APP Y682G metabolism was resistant to X11␣ effects.
Thus, mutation of APP within the YENPTY motif impaired the inhibitory effects of X11␣ on A␤ secretion. We also determined the effect of coexpression of APP or APPswe with the mutation X11␣ F608V. This mutation lies within the carboxylterminal ␣-helix of X11␣ PTB domain and is a critical residue for PTB domain interaction (6). HEK 293 cells were cotransfected with APPswe and X11␣ constructs. As predicted from the binding data, the inhibitory effect of X11␣ on A␤ secretion was attenuated by X11␣ F608V coexpression (Fig. 4B). Thus the specific interaction of X11␣ with APP appears to inhibit release of A␤40 and A␤42 into conditioned medium. The ELISA data was not corrected for level of APPc, because apparent effects on A␤ secretion would be artifactually magnified.
X11␣ Coexpression Stabilized Cellular APP-Our data suggests that X11␣ coexpression blocks the production of soluble APP metabolites and results in apparent greater levels of cellular APP, thus appearing to stabilize APP in the cell (Fig. 3). To test this hypothesis, we performed a pulse-chase analysis of . In panel B, 5 g of pRK5-myc-X11␣ DNA was transfected. Proteins in cell extracts and conditioned medium were treated as in Fig. 2A. A c-myc epitope tag fused to the amino terminus of X11␣ allowed detection of this protein by anti-myc antibody. Proteins in cell lysates were detected by antibody 369 or anti-myc. Soluble APP in conditioned media was immunoprecipitated with Karen and detected by immunoblot with 6E10 for APPs␣ or 22C11 for total APPs. HEK 293 cells transiently transfected with APP alone or with X11␣. Cells were labeled for 15 min with [ 35 S]methionine before chase for up to 8 h. APP in cell lysates was immunoprecipitated by Karen antibody, separated by SDS-PAGE, and detected by PhosphorImager analysis and autoradiography. Radiolabeled APP was quantitated by ImageQuant software. Similar to previous reports, APP had a half-life of ϳ2 h in HEK 293 cells (Fig. 5). Coexpression of X11␣ prolonged APP half-life to ϳ4 h. To further evaluate the stabilization of cellular APP by X11␣, we transiently expressed X11␣ in HEK 293 cells stably expressing APP and measured APP levels in cell lysates. In these experiments, X11␣ expression resulted in an apparent increase in the amount of APP in cell lysates. As predicted, the X11␣ F608V mutation had less effect (Fig. 6A, lower panel). Comparable amounts of X11␣ and X11␣ F608V were expressed in the cells (Fig. 6A, upper panel).
During the course of these studies with X11␣, we cloned a second human X11 gene we named X11␤. This gene is located on chromosome 15 (human genome project) and encodes a 120-kDa protein with a predicted topology as X11␣. Both X11␣ and X11␤ are highly expressed in brain. 2 The PTB and PDZ domains of X11␤ were 80 -90% identical to X11␣ domains and were thus predicted to bind similar targets. However, their amino termini were quite divergent. Accordingly, the X11␤ PTB domain bound efficiently to APP in vitro and in cells in culture (data not shown) and had similar results on APP proc-essing, i.e. diminished soluble metabolites and stabilization of APP in cells. Hence, expression of X11␤ in HEK cells stably transfected with APP notably increased the amount of APPc (Fig. 6B). This effect is almost completely abolished by a mutation of X11␤ analogous to X11␣ F608V. Taken together, these data demonstrate that both X11␣ and X11␤ inhibit APP metabolism when overexpressed in HEK 293 cells.

DISCUSSION
Nonneuronal cell lines are instructive model systems of APP trafficking and metabolism. HEK 293 cells produce A␤, primarily A␤40, via ␤-/␥-secretase activities in an endosomal pathway, although the primary metabolic products of APP in this cell line result from ␣-secretase activity (4). The intracellular carboxyl-terminal domain of APP, in particular the YENPTY consensus sequence required for endocytosis by clathrin-coated pits, plays an important role in APP processing and A␤ generation by the endosomal pathway (4,5). X11␣, a protein highly expressed in neurons, specifically interacts with a peptide encompassing the YENPTY motif of APP (6). We now demonstrate that the interaction of X11␣ or X11␤ with APP has significant effects on its metabolism in HEK 293 cells and has implied effects on cellular trafficking of APP.
Coexpression of X11␣ with APP decreased the recovery of its soluble fragments APPs␣, A␤40, and A␤42 in conditioned medium of HEK 293 cells. These effects are specific, as demonstrated by the use of a single point mutation within either the cytoplasmic YENPTY motif of APP or the PTB domain of X11␣, which impairs their interaction and thus, X11␣ effects. The decreased recovery of soluble APP fragments in conditioned medium was observed in parallel with an apparent increase in APP in cell lysates. This suggested prolongation of the half-life of APPc, which was confirmed by pulse-chase analysis. Expression of a second member of the X11 gene family, i.e. X11␤, in HEK 293 cells had similar effects on APP processing. X11␤ bound as efficiently as X11␣ to APP in vivo and in vitro (data not shown). Thus, the specific interaction of the X11 PTB domain with the YENPTY motif-containing region of APP appeared to retard its processing and prolong its half-life, resulting in decreased recovery of soluble proteolytic fragments. When coexpressed with APP, X11␣ slowed both the ␣-secretase pathway and the endosomal/lysosomal pathway leading to A␤ generation. The mechanisms and intracellular site(s) of X11␣ effects on APP metabolism are unknown, but one may hypothesize that X11␣ slows endosomal trafficking of APP. Alternatively, X11␣ may prevent the secretion of APPs, leading to an accumulation of APP in the cell. Interestingly, recent evidence suggests increased neuronal endocytosis and thus increased A␤ secretion in neurons of sporadic AD brain compared with agematched control brain (20).
The observed effects of X11␣ on inhibition of A␤ secretion with APP coexpression are qualitatively similar but amplified by coexpression with APPswe. In effect, coexpression of X11␣ with APPswe decreased A␤40 and A␤42 secretion to that seen with APP expression only. Similar to APP, when coexpressed with APPswe, X11␣ retarded both the ␣-secretase pathway and A␤ generation. The Swedish mutation of APP promoted its metabolism by ␤-/␥-protease activities, resulting in a 5-10-fold increase in A␤40 and A␤42 secretion, with a concomitant decrease in the ␣-secretase pathway (21,22). There are other important differences between APP and APPswe metabolism. For example, in contrast to APP, transfection of an APPswe construct lacking a cytoplasmic tail, which precludes reinternalization, did not reduce the secretion of A␤ peptides. Thus, an additional ␤-/␥-protease pathway in Golgi-derived vesicles or the Golgi itself is present in APPswe metabolism to A␤ in nonneuronal cells (23)(24)(25)(26). X11␣ may affect metabolism of APPswe by this cellular pathway as well as the endosomal pathway of A␤ generation.
X11␣ and X11␤ are neuronal proteins that contain two PDZ (PSD-95/Dlg/ZO-1) domains in addition to the PTB domain. The PDZ domains found in other neuronal membrane proteins such as the PSD-95 family and nitric oxide synthase are implicated in their membrane clustering and localization. Clustering and localization of proteins may serve to stabilize proteins and prolong their half-life (27). This is similar to the effects of X11 on APP that we observed in this study. Although the binding partners of the PDZ domain of X11␣ are unknown, a heterotrimeric complex of PDZ partner/X11␣/APP may be implicated in X11␣ effects and APP localization. The Fe65 gene family is also expressed primarily in neurons, and the encoded protein contains a PTB domain that binds to APP (8) and thus may influence its metabolism. In addition to two PTB domains, Fe65 contains in its sequence a WW protein interaction domain that binds to proline-rich sequences (28). Thus, Fe65 and X11␣ may have differential effects on APP trafficking and metabolism based on the formation of alternate and potentially competing heterotrimeric complexes of APP with either a PDZ binding partner of X11␣ or a WW binding partner of Fe65.
Neuronal processing of APP is in some ways distinct from its metabolism in nonneuronal cells and results in greater A␤ generation compared with nonneuronal cells (29 -32). In addition to having the more ubiquitous ␣-secretase pathway at or near the cell surface and endosomal/lysosomal processing of APP to A␤, neuronal cells have additional ␤/␥ processing of APP within the endoplasmic reticulum/early Golgi. This neuronal exocytic pathway favors the generation of a higher ratio of A␤42 to A␤40 compared with the ␤-/␥proteases of the endocytic pathway (31,18,33). Interestingly, presenilins are localized primarily to the endoplasmic reticulum and Golgi (34 -36), suggesting that presenilin-1 or presenilin-2 mutations linked to familial AD exert their effect on APP metabolism, specifically increased A␤42 secretion, by promotion of ␤-/␥cleavage within this exocytic pathway. The effects of X11 on the metabolism of APP717 mutations or on APP metabolism coexpressed with presenilin mutations, all of which result in a higher ratio of A␤42 to A␤40 secretion (19,37), are unknown.
It will be of interest to probe the effects of X11 coexpression in transgenic mice harboring mutations of human APP linked to familial AD. With aging, these transgenic animals develop a partial AD-like phenotype, in particular behavioral changes and amyloid deposition in brain (38 -40). Examination of X11 and Fe65 expression and their binding partners in normal aging and AD brain may shed light on two unanswered questions in AD research, namely, the selective anatomic localization of amyloid plaques in brain and the increased risk of AD with aging. Finally, if the amyloid hypothesis of AD proves tenable, knowledge of the effects of X11 and Fe65 on APP metabolism may serve as the basis for novel therapeutic strategies to delay the onset or slow the progression of amyloid formation and thus the clinical dementia of AD.