A New Functional Screening System for Identification of Regulators for the Generation of Amyloid β-Protein*

Presenilin (PS) is essential for γ-cleavage, which is required for the generation of amyloid β-protein (Aβ) from the β-amyloid precursor protein. However, it remains to be clarified how γ-cleavage is regulated. To elucidate the regulation of PS-mediated γ-cleavage, we developed a new functional screening method for identifying cDNA that enhances γ-cleavage. This screening system utilizes our own developed cell line, where the expression of cDNA that enhances γ-cleavage confers puromycin resistance. The cDNA library is retrovirally delivered to the above-mentioned cell line, allowing the identification of our target cDNAs by a combination of puromycin resistance selection and Aβ assay screening. With this screening method, we isolated several cDNAs enhancing γ-cleavage, including the previously reported Herp. Here we also demonstrate that Rab1A, identified with this screening, can be a regulator of Aβ generation. Thus, our established screening method is a powerful tool for identifying multiple regulators involved in γ-cleavage in the Aβ generation pathway, including modulators of γ-secretase activity or the intracellular trafficking of factors necessary for γ-cleavage.

A␤, 1 which is the major component of senile plaques in the brain of patients with Alzheimer's disease (AD), is generated from APP through its sequential proteolytic cleavage catalyzed by ␤and ␥-secretase (reviewed in Ref. 1). Although ␤-secretase was identified as a membrane-tethered aspartyl protease (2), the molecule responsible for ␥-secretase activity remains to be clarified. Mutations in the presenilin (PS) genes, PS1 and PS2, cause early-onset familial AD (reviewed in Ref. 1). Accumulat-ing evidence showed that PS is required for the proteolytic cleavage catalyzed by ␥-secretase, which occurs in the transmembrane domain of APP (␥-cleavage) (Refs. 3-5; reviewed in Ref. 6). Interestingly, although the ␥-cleavage is a critical step toward 〈␤ production, the major intramembranous cleavage site of APP is distinct from the ␥-cleavage (named as ⑀-cleavage site) (7,8). In addition, recent studies revealed that PS mediates several intramembranous cleavages including those of APP, Notch (3)(4)(5), ErbB4 (9,10), and E-cadherin (11), indicating that the PS-mediated intramembranous cleavage plays a critical role in biological functions. The PS complex appears to be responsible for inducing ␥-secretase activity (12,13); however, it is still controversial whether PS itself is ␥-secretase (Refs. 14 -16, reviewed in Ref. 17). The discrepancy between the intracellular major distribution of the PS complex and the intracellular site of ␥-cleavage also remains to be clarified (reviewed in Ref. 17). Therefore, the understanding of the mechanism underlying ␥-cleavage will require clarification of multiple factors involved in ␥-cleavage, including the components of the ␥-secretase complex and modulators of the ␥-secretase activity or the trafficking of factors necessary for ␥-cleavage. To elucidate how PS-mediated ␥-cleavage is regulated, we developed a new functional screening method for identifying cDNA that enhances ␥-cleavage using a combination of puromycin resistance assay and A␤ quantitation. To date, a number of PS-interacting proteins have been identified by the yeast two-hybrid screening method that utilizes the binding affinity with PS; however, no natural interactors with PS have been found to modulate A␤ generation except for nicastrin (13,18). Therefore, we employed the functional screening assay to measure the A␤ level, instead of the binding assay of PS. In addition, prior to the screening step using the A␤ assay, we selected cDNAs that confer puromycin resistance on the cells resulting from an increase in the degree of intramembranous proteolytic cleavage. This selection step facilitates the identification of our target cDNAs. Thus, the screening system that we developed is a powerful tool for identifying multiple factors involved in ␥-cleavage, including modulators of ␥-secretase activity or the trafficking of factors necessary for ␥-cleavage. Previously, we reported that the ER stress-inducible protein, Herp (19), which was identified by this method, increased A␤ generation (20). Here we present the details of our newly developed functional screening method for the identification of ␥-cleavage-enhancing factors, and also demonstrate that Rab1A, which is implicated in protein trafficking from the ER to the Golgi apparatus (reviewed in Ref. 21), can be another regulator of ␥-cleavage in the A␤ generation pathway.

EXPERIMENTAL PROCEDURES
Antibodies and Cell Lines-The monoclonal antibody 6E10 specific to human A␤1-17 was purchased from Senetek (St. Louis, MO). The other A␤ antibodies have all been characterized previously (22). The antibody against Rab1A was purchased from Santa Cruz Biotechnology, Inc. The anti-APP N-terminal antibody was purchased from Sigma. The affinitypurified rabbit antibody, B12/4, was raised against 20 C-terminal amino acid residues of APP695 (23). Ba/F3 cells of the murine pro-B cell line were maintained as previously described (24). PS-knockout and wild-type murine fibroblasts immortalized with large T antigen were maintained as previously described (20).
All resulting constructs were verified by DNA sequence. Constructs of Human Hippocampus-derived cDNA Library-Human hippocampus cDNAs were synthesized from human hippocampus mRNA (CLONTECH) using the SuperScript TM Choice system (Invitrogen) according to the instructions from the manufacturer. The cDNAs were inserted into the BstXI site of pMX, using the BstXI adaptor (Invitrogen). The resulting cDNA library contains 4 ϫ 10 6 independent colonies, and the average size of the insert DNA is 1.6 kb.
Retrovirus-mediated Gene Expression and Functional Screening Procedure-Retrovirus-mediated gene expression in cells was carried out as previously reported (24,34). Briefly, we transfected with cDNA library or pMX carrying the identified cDNA into Phoenix-Eco cells. The viral supernatant obtained after 24 h of culture at 48 h after transfection was used to infect cells. For the screening, we used BaF/3 cells (24,36) stably transfected with pCxN-C53NICD and pHES-1-pac (designated as A5-9) or A5-9 cells stably expressing human PS1 (designated as A5-9-PS1) as the target cells. The target cells (4 ϫ 10 6 ) were retrovirally expressed with human hippocampus cDNA library and grown on 96well plates (the initial cell number and density were 10 4 cells/well and 10 5 cells/ml, respectively). At 48 h after retroviral infection, the cells were subjected to puromycin resistance selection using the minimum lethal dose of puromycin (9 g/ml puromycin for A5-9-PS1 cells; 20 g/ml puromycin for A5-9 cells). The functional screening procedure was described in the text. The cDNAs transduced into the cells were extracted and amplified by PCR using vector primers (primer sequences: 5Ј-GGTGGACCATCCTCTAGACTG-3Ј and 5Ј-GTTACTTAAGCTAGCTT-GCC-3Ј) and sequenced.
Detection of A␤ and Other Immunoblotting Techniques-The secreted A␤ was immunoprecipitated with 6E10 and detected using a highly sensitive immunoblotting technique with BA27 (for A␤40) or BC05 (for A␤42) as previously described (30,35). ELISA for A␤ was performed as previously described (22). ELISA data were statistically analyzed by analysis of variance using StatView-J.4.11. Intracellular C99 was immunodetected as previously described (30).
Luciferase Assay-The cells cotransfected with PGV-B-HES-1 and pRL-tk were analyzed using a luciferase assay system (Promega). The Renilla luciferase expression plasmid, pRL-tk (Promega), was used as an internal control for transfection.

RESULTS AND DISCUSSION
A Chimeric Protein of C53NICD Generates A␤ and Activates HES-1 Promoter-mediated Transcription in Cells-Our functional screening system utilizes the transcriptional activity of the Notch-1 intracellular domain (NICD) released from the membrane following transmembranous cleavage of Notch-1 (25). We generated a chimeric gene encoding C53NICD by replacing the C-terminal intracellular domain of C99 (␤-secretase-cleaved APP fragment; reviewed in Ref. 6) with the murine NICD (Fig. 1A). We also generated a puromycin resistance gene (pac) driven by the HES-1 promoter (a DNA element responsible for Notch-dependent gene expression) (25, 32) (Fig.  1B). Using the two DNAs, we designed a functional screening system based on the following idea. (i) In the cells stably expressing C53NICD and pac driven by the HES-1 promoter, an increase in the degree of ␥-cleavage should confer an increase in puromycin resistance and A␤ generation (Fig. 1B). (ii) Therefore, when the above-mentioned stable cells were transfected with a cDNA library, the cells harboring a cDNA that enhances ␥-cleavage should be first selected as puromycin-resistant clones, and then identified by the screening for cells generating a higher A␤ level.
Prior to carrying out the above functional screening, we first investigated whether the expression of C53NICD in cells leads to A␤ generation and activates the HES-1 promoter. As shown in Fig. 2A, Neuro 2a cells transiently expressing C53NICD generated A␤40 and A␤42 as well as the cells expressing C99, indicating that C53NICD undergoes ␥-cleavage in the cells as APP does. We next addressed whether C53NICD activates the expression of the gene under the control of the HES-1 promoter. For this purpose, we constructed a plasmid carrying the luciferase-coding gene driven by the HES-1 promoter, and performed luciferase assay. As shown in Fig. 2B, the expression of the luciferase gene was induced by the HES-1 promoter when it was cotransfected with a plasmid encoding C53NICD, but not with a plasmid encoding C99. This result indicates that the expression of C53NICD causes NICD-dependent HES-1 promoter-mediated gene activation in the cells. Furthermore, to exclude the possibility of the activation of the HES-1-driven gene by nonspecific generation of NICD-like fragments from C53NICD, we next determined using PS-deficient cells whether the activation of the luciferase gene depends on PS. As shown in Fig. 2C, the activation of the luciferase gene completely depended on PS, indicating that the release of NICD from C53NICD, as well as that from Notch-1, depends on PS. This result strongly supports the reliability of the puromycin selection step in the above-described screening system. Thus, we decided to develop our designed functional screening method, because we ascertained that ␥-cleavage of C53NICD generates A␤, accompanied by PS-dependent activation of HES-1 promoter-mediated transcription. For a functional screening assay, a stable expression of cDNA in cells is preferable. Therefore, we employed the retrovirus-mediated gene expression method ( Puromycin-resistant clones were obtained following the retroviral expression of cDNA library in A5-9-PS1. A␤ in the conditioned medium (8 ml) of a puromycin-resistant clone was immunodetected with a highly sensitive immunoblotting method. Immunoprecipitates from one culture dish were each loaded in one lane. The initial cell density was adjusted to 10 5 cells/ml, and the cells were cultured for 4 days. After we confirmed that the final cell density was equal between the clones, A␤ secreted from each clone was detected. Upper panel, the level of A␤40; lower two panels, the levels of A␤40 and A␤42. Numbers denote the numerical designation of the clone that exhibited puromycin resistance phenotype. The data except those for clone 3 are representative of puromycin-resistant clones exhibiting an increase in A␤ level. The data of clone 3 are representative of puromycin-resistant clones exhibiting no increase in A␤ level.

FIG. 2. C53NICD generates A␤ and activates the HES-1 promoter in the cell.
A, A␤ secreted from N2a cells transiently expressing C53NICD after a 48-h culture was immunoprecipitated with 6E10 and detected using a highly sensitive immunoblotting technique. Cells were transiently transfected with pCxN-C53NICD or pcDNA-C100 (harboring the cDNA encoding the first methionine plus C99) or pCxN2 (mock transfection). B, N2a cells were co-transfected with PGV-B-HES-1 (see "Experimental Procedures") and a pRL-tk plus pCxN-C53NICD or pcDNA-C100 or an empty vector. At 1 day after transfection, luciferase assay was performed. Renilla luciferase expression plasmid, pRL-tk, was used as an internal control for transfection. The results are presented as -fold induction, which is the relative luciferase activity (ratio of HES-1 promoter-driven luciferase/Renilla luciferase) of the cells over that of control cells (the cells transfected with PGV-B-HES-1 and a pRL-TK plus an empty vector). C, PS1/PS2 double-deficient fibroblasts (PSϪ/Ϫ) or wild-type fibroblasts (wt) were transiently co-transfected with PGV-B-HES-1 and a pRL-tk plus pCxN-C53NICD or an empty vector (mock). At 2 days after transfection, luciferase assay was performed. Renilla luciferase expression plasmid, pRL-tk, was used as an internal control for transfection. The results are presented as luciferase activity (arbitrary unit), which is the ratio of HES-1 promoter-driven luciferase/Renilla luciferase. pac (see "Experimental Procedures") (designated as A5-9). We also established A5-9 cells stably expressing human PS1 (designated as A5-9-PS1), because the use of the cells overexpress-ing PS1 might help to identify the activator of ␥-secretase, as PS is essential for the ␥-secretase activity.
Identification of cDNA Encoding ␥-Cleavage-enhancing Factor-The scheme of the screening procedure used is shown in Fig. 3. To initiate screening, a human hippocampus-derived cDNA library was retrovirally transduced into A5-9 or A5-9-PS1 cells. At 48 h after retroviral infection, the cells were subjected to puromycin resistance selection. The cells were treated with a lethal dose of puromycin to select the cells expressing cDNA encoding a ␥-cleavage-enhancing factor. After 2 weeks of growth, surviving clones were picked up and again treated with the same dosage of puromycin to eliminate clones that were transiently resistant to puromycin. To identify the clones exhibiting an increased A␤ level, A␤ secreted from puromycin-resistant clones was detected using a highly sensitive immunoblotting method. We found that some of the puromycin-resistant clones clearly exhibited an increased A␤ level (Fig. 4). Table I shows the number of puromycin-resistant clones and the number of the clones with an increased A␤ (A␤40) level. Out of 4 ϫ 10 6 cells transduced with cDNA library, approximately 30 (target cells, A5-9-PS1) or 10 (target cells, A5-9) clones exhibited the puromycin-resistant phenotype (Table I). Among them, approximately 10 clones exhibited an increase in the A␤ level (Table I). Next, the cDNAs from the puromycin-resistant clones with an increased A␤40 level were amplified by PCR using vector primers. One of these DNAs encoded Herp (19), which was previously reported to enhance A␤ generation from APP (20). From another puromycin-resistant clone with an increased A␤ level (clone 28 in Fig. 5A), we obtained 1.4-kb PCR product containing the open reading frame encoding the entire Rab1A amino acid sequence. Rab1A is implicated in protein trafficking from the ER to the Golgi apparatus (reviewed in Ref. 21). We further determined whether Rab1A is responsible for conferring puromycin resist- FIG. 5. Expression of Rab1A in the parental cells resulted in an increase in A␤ generation from C53NICD. Clone 28 was isolated as a puromycin-resistant clone with increased A␤ level by our screening method using A5-9-PS1 cells. PCR product from clone 28 encoded Rab1A. A, A␤ secreted from clone 28 was detected as described in Fig.  3 legend. B, puromycin-resistant cells were obtained following the retroviral expression of Rab1A in A5-9-PS1 cells as described in Table II. The levels of A␤ secreted from the puromycin-resistant cells and A5-9-PS1 cells were detected. The cell lysates (25 g) were immunoblotted with the anti-Rab1A antibody. The radioimmune precipitation assay buffer-solubilized lysates (25 g) were immunoblotted with the anti-APP N-terminal antibody. D, the intracellular C99 level in radioimmune precipitation assay buffer-solubilized lysates (1.5 mg) was immunoprecipitated and immunoblotted as previously described (20). The lysate prepared from one culture dish were each loaded in one lane. The lysate prepared from fibroblasts retrovirally expressing C99 was also immunodetected with B12/4 (left lane).

TABLE I Number of puromycin-resistant clones exhibiting an increased A␤ level
obtained by the screening method After the target cells (4 ϫ 10 6 ) retrovirally expressing a cDNA library were treated with puromycin as described in the text, the surviving clones were counted. The level of A␤40 secreted from puromycin-resistant cells was determined as described under "Experimental Procedures." Figures in parentheses indicate the numbers of spontaneously surviving clones, which were counted using the target cells without retroviral expression of the cDNA library. ance and an increased ability of A␤ generation on A5-9-PS1 cells. As shown in Table II, the expression of Rab1A conferred puromycin resistance on A5-9-PS1 cells. We also confirmed that the puromycin-resistant cells obtained after the retroviral transduction of Rab1A cDNA secreted higher levels of both A␤40 and A␤42 than A5-9-PS1 cells (Fig. 5B). These results clearly indicated that the Rab1A cDNA is a gene responsible for exhibiting puromycin resistance and enhanced A␤ generation.
Characterization of other such cDNAs are under way. Expression of Rab1A Increases A␤ Generation from Fulllength APP-We next investigated whether Rab1A increases A␤ generation from full-length APP using murine fibroblasts.
As shown in Fig. 6A, the expression of Rab1A clearly increased secreted A␤40 and A␤42 levels. Quantification by ELISA revealed that the expression of Rab1A increased the levels of A␤40 and A␤42 3-and 2-fold, respectively (Fig. 6B). The difference in the degrees of increase in A␤40 and A␤42 levels suggested distinct molecular mechanisms underlying the generation between the two A␤ variants. The level of soluble APP (sAPP; reviewed in Ref. 1), which is generated through ␣or ␤-cleavage, was also slightly increased by Rab1A expression, whereas no significant change in the intracellular APP level was observed (Fig. 6C). The degree of an increase in sAPP level was less than 2-fold, because the band intensity of one-half the amount of sAPP secreted from the fibroblasts expressing Rab1A was lower than that of total sAPP secreted from the mock fibroblasts (Fig. 6C). These results indicated that Rab1A expression increased the degree of ␥-cleavage greater than that of ␣or ␤-cleavage. In addition, the intracellular level of C99, which is a substrate of ␥-secretase, was significantly reduced by a high expression level of Rab1A (Fig. 6D), providing strong evidence that Rab1A expression enhances ␥-cleavage, resulting in an increase in A␤ generation.
The cytoplasmic tail of APP contains important sorting determinants (37) and sequence elements required for interaction with proteins including X11, which influences APP processing and A␤ generation (38 -40). Therefore, to know whether the effect of Rab1A on A␤ generation depends on the cytoplasmic tail of APP, we next addressed whether the expression of Rab1A increases the A␤ generation from APP lacking the cytoplasmic tail (APP⌬C). As shown in Fig. 7A, the expression of Rab1A increased A␤ generation even from APP⌬C, although the degree of increase was slightly small. Therefore, the cytoplasmic domain of APP or the intracellular trafficking of APP mediated by the cytoplasmic domain is not essential for an increase in 〈␤ generation by the expression of Rab1A. The deletion of the cytoplasmic tail of APP drastically enhanced sAPP generation as previously reported (33,38) (Fig. 7B). It is also noted that the expression of Rab1A did not result in an enhanced sAPP generation from APP⌬C ( Fig. 7B; for the quantitative analysis of sAPP production, see Fig. 7C). An increase in A␤ generation from APP⌬C caused by Rab1A expression, which was accompanied by no increase in sAPP generation, clearly indicates that the expression of Rab1A increases ␥-cleavage, which is not the result of the simple stimulation of the trafficking of the APP⌬C from the ER to the Golgi apparatus. We also found that a high expression level of Rab1A did not alter the steady-state levels of full-length PS and its endocleaved product, N-terminal fragment (data not shown), strongly suggesting that a high expression level of Rab1A did not affect the trafficking of PS from the ER to the Golgi apparatus, because the endoproteolysis of PS is likely to occur in the ER-cis-Golgi intermediate compartment (reviewed in Ref. 17). Thus far, the involvement of several Rab proteins in the pathway for A␤ or soluble APP generation has been addressed (41)(42)(43). However, the role of Rab proteins in the ␥-cleavage pathway is not yet clarified. Our study provided strong evidence that Rab1A, which is known to be involved in protein trafficking from the ER to the Golgi apparatus (reviewed in Ref. 21), plays a role in the ␥-cleavage pathway. Previously, the ER-cis-Golgi intermediate compartment has been suggested to be a site for A␤ generation (44,45). In this regard, ER localization of Rab1A well agrees with the ER being the intracellular site of A␤ generation. In addition, in PS1-deficient cells, C99 is accumulated in the ER without undergoing ␥-cleavage (46,47). One possible explanation for this is that C99 in PS1deficient cells is retrogradely trafficked to the ER after it is generated through the ␤-secretase cleavage of APP in a late Golgi compartment or trans-Golgi network. Therefore, Rab1A could enhance ␥-cleavage of ER-localized C99, which is retrogradely trafficked back to the ER. However, accumulating evidence showed that the major intracellular site of ␥-cleavage is not the ER, but, most likely, the trans-Golgi network (30,31,48,49). In this regard, it is noteworthy that Rab1A not only exists in the ER, but is also associated with the transcytotic vesicle (50). This indicates that the function of Rab1A is not limited to protein trafficking from the ER to the Golgi apparatus. Rab1A could be involved in ␥-cleavage in its other destinations or during its travel to the transcytotic vesicle or an as yet identified compartment. At present, we do not know whether Rab1A activates ␥-secretase or modulates the trafficking of the factors involved in ␥-cleavage, including PS and nicastrin. Further study on the mechanism underlying the enhancement of ␥-cleavage caused by a high expression level of Rab1A is necessary.
In this study, we showed a new functional screening method for identifying factors involved in ␥-cleavage and also identified Rab1A by this method. Previously, we reported that the ER stress-inducible protein, Herp, which was identified by this method, enhanced ␥-cleavage and bound to PS (20). These results strongly suggest that Herp is an inducible regulator of ␥-secretase. In contrast, Rab1A appears to constitutively reg-ulate ␥-cleavage in the A␤ generation pathway. Interestingly, both of the two identified proteins are ER-localized proteins, suggesting that the ER or an ER-derived compartment plays a critical role in the ␥-cleavage pathway. The other factors identified by this method are now under investigation. We believe that our established functional screening method is a powerful tool for identifying multiple factors, including activators of ␥-secretase (or ␥-secretase itself) and modulators of the trafficking of the factors involved in ␥-cleavage in the A␤ generation pathway. Identification of the factors required for ␥-cleavage will enable the comprehensive study of the mechanism underlying the intermembranous proteolytic cleavage mediated by PS. In addition, such factors could be new candidate therapeutic targets for AD.