Mammalian APH-1 Interacts with Presenilin and Nicastrin and Is Required for Intramembrane Proteolysis of Amyloid- (cid:1) Precursor Protein and Notch*

Presenilin and nicastrin are essential components of the (cid:2) -secretase complex that is required for the intramembrane proteolysis of an increasing number of membrane proteins including the amyloid- (cid:1) precursor protein (APP) and Notch. By using co-immunoprecipita-tion and nickel affinity pull-down approaches, we now show that mammalian APH-1 (mAPH-1), a conserved multipass membrane protein, physically associates with nicastrin and the heterodimers of the presenilin amino-and carboxyl-terminal fragments in human cell lines and in rat brain. Similar to the loss of presenilin or nicastrin , the inactivation of endogenous mAPH-1 using small interfering RNAs results in the decrease of presenilin levels, accumulation of (cid:2) -secretase substrates (APP carboxyl-terminal fragments), and reduction of (cid:2) -secretase products (amyloid- (cid:1) peptides and the intracellular domains of APP and Notch). These data indicate that mAPH-1 is probably a functional component of the (cid:2) -secretase complex required for the intramembrane proteolysis of APP and Notch. Regulated intramembrane proteolysis of the amyloid- (cid:1) precursor protein (APP), 1 Notch, and other selected membrane proteins is a conserved control mechanism essential for cell signaling (1). The cleavage of Notch within its transmembrane Cell Lines and Transfection— HEK293 cells stably or HeLa cells transiently transfected with the Myc-His-tagged mAPH-1

Biochemical Methods-Co-immunoprecipitation and Western blotting were performed as described previously (3-5, 10 -11) with the exception that the SuperSignal West Femto Reagents (Pierce) were used. Nickel-nitrilotriacetic acid (Ni-NTA)-agarose affinity pull-down experiments were performed according to the manufacturer's instruction (Qiagen).
APP and Notch Cleavage Assays-Native HeLa cells were transfected with mAPH-1, nicastrin, or presenilin-1 siRNAs or control RNA oligonucleotides for 56 -96 h to inactive target genes. The resultant cells were subsequently transfected with proper reporter constructs and analyzed as previously described for APP and Notch cleavage in either the luciferase reporter gene assays (2,28) or the N⌬E immunoblotting analysis (15,17,29). Luciferase reporter systems were provided by T. Sudhof and J. Lundkvist. N⌬E and NICD constructs were from R. Kopan. For A␤ analysis, HeLa/APP Swedish stable cell lines were treated with the appropriate siRNAs for 56 -96 h and then changed to fresh media for 16 h. A␤ was immunoprecipitated from the conditioned media by polyclonal antibody, separated on Bicine/Tris SDS gel (30), and immunoblotted with A␤ monoclonal antibody. All experiments in this paper were performed at least four times.

RESULTS AND DISCUSSION
From a human glioblastoma cDNA library, we isolated two mAPH-1 (mAPH-1a and mAPH-1b) genes that are homologous to worm aph-1. mAPH-1a has at least two splice variants: 1) mAPH-1a L has seven exons and encodes a longer (L) open reading frame of 265 residues; and 2) mAPH-1a S has six exons and encodes a shorter (S) open reading frame of 247 residues (Fig. 1, A and B). mAPH-1b has six exons and encodes 257 amino acids (Fig. 1, A and B). We made cDNA constructs that express Myc and His epitope-tagged mAPH-1a L , mAPH-1a S , and mAPH-1b. Western blots of lysates from HEK293 cells stably transfected with these cDNAs revealed Myc-immunoreactive bands with relative molecular masses of approximately 30 kDa, which are the predicted sizes of mAPH-1 amino acid sequences ( Fig. 2A, lane 1) (data not shown). Additional Mycimmunoreactive bands at 13-15 and 50 kDa can sometimes be detected (data not shown).
To determine whether mAPH-1 physically associates with the ␥-secretase complex, HEK293 cells stably expressing mAPH-1a L -Myc-His were used in reciprocal co-immunoprecipitation experiments. We found that in cell lysates extracted with CHAPSO, a detergent compatible with ␥-secretase activity, mAPH-1a L co-precipitated with endogenous nicastrin and endogenous presenilin-1 NTF and CTF heterodimers ( Fig. 2A,  lanes 2-4). Conversely, endogenous presenilin-1 NTF and CTF and nicastrin (data not shown) co-precipitated with mAPH-1a L -Myc-His ( Fig. 2A, lane 5). Presenilin-2 NTF also co-precipitated with mAPH-1a L -Myc-His and nicastrin, but not presenilin-1 NTF or CTF ( Fig. 2A, lanes 2-5), suggesting that the two presenilin proteins are in separate complexes (3) that share other common components including nicastrin and mAPH-1a L and that the interactions revealed in our co-immunoprecipita- mAPH-1 Is a Component of the ␥-Secretase Complex tion experiments are highly specific. The specificity of the mAPH-1a L , presenilin, and nicastrin interaction was corroborated by the absence of co-immunoprecipitation of these proteins with preimmune serum (Fig. 2A, lane 6) or with antibodies to irrelevant proteins such as calnexin and Rab5 ( Fig. 2A, top panel) (data not shown).
To study the role of endogenous mAPH-1 in presenilin-and nicastrin-associated ␥-secretase activity, we used 21-nucleotide duplexes of siRNAs and performed RNA interference studies in HeLa cells (27) to inactivate endogenous mAPH-1. Reverse transcriptase-PCR experiments indicated that mAPH-1b mRNA is hardly detectable, whereas mAPH-1a mRNA can be easily detected in HeLa cells (Fig. 4A, lanes 2 and 3). mAPH-1a but not mAPH-1b siRNAs significantly reduced mAPH-1a mRNA levels in HeLa cells (Fig. 4A, lanes 4 and 5). The fact that siRNAs against a highly homologous gene did not affect mAPH-1a illustrates the specificity of the siRNA technique. We next examined whether the presenilin NTF:CTF heterodimers are stable after mAPH-1 is inactivated in mammalian cells. We found a dramatic reduction of presenilin-1 and presenilin-2 but not other proteins such as ␣-tubulin in mAPH-1a and mAPH-1a plus mAPH-1b siRNA-treated HeLa cells (Fig. 4, B and C). Consistent with the low expression level of mAPH-1b in HeLa cells, a minor reduction of presenilin proteins was observed in cells treated with mAPH-1b siRNAs (Fig. 4B). The down-regulation of the steady-state presenilins by mAPH-1

FIG. 3. Endogenous mAPH-1 interacts with nicastrin and the presenilin heterodimers in vivo.
A, 1% digitonin or 1% CHAPSO (data not shown) extracted proteins from HeLa or HEK293 (data not shown) cells expressing mAPH-1a L -Myc-His were subjected to Ni-NTA-agarose pull-down experiments in the lysis/washing buffer plus 20 mM EDTA (lane 1), 100 mM imidazole (lane 2), or no EDTA/imidazole (lane 3). The pull-down products were resolved on SDS-polyacrylamide gel, and the immunoblots were investigated with antibodies to mAPH-1a L (affinitypurified H2D-2, top panel) and presenilin-1 NTF (bottom panel). B, rat brain membranes were solubilized in 1% digitonin and immunoprecipitated using the mouse monoclonal anti-presenilin-1 CTF antibody (MAB5232, Chemicon) (lane 1) or normal mouse IgG (lane 2). The immunoprecipitation products were resolved on SDS-polyacrylamide gel, and the immunoblots were investigated with antibodies to mAPH-1a and presenilin-1 NTF as in panel A. C and D, rat brain membranes were solubilized in 1% digitonin and immunoprecipitated using the antibodies as indicated at the top. The immunoprecipitation products were resolved on SDS-polyacrylamide gel, and the immunoblots were investigated with antibodies to nicastrin (C) or presenilin-1 CTF (D). The asterisk indicates IgG. The lane labeled input represents ϳ20% of the starting digitonin lysate utilized for the immunoprecipitation. siRNAs is most probably at the protein levels in that mAPH-1 siRNAs did not cause major changes in presenilin-1 mRNA levels (Fig. 4A). As expected (7-8), we found that nicastrin siRNAs reduced the steady-state levels of presenilin-1 and presenilin-2 and that presenilin-1 siRNAs reduced mature nicastrin (Fig. 4, B and C). However, we did not observe significant changes in the overall nicastrin levels in cells treated with mAPH-1a siRNAs (Fig. 4C). This is probably because of the incomplete inhibition of mAPH-1 by the mAPH-1a siRNA oligonucleotides we used. Indeed, further investigation into the mAPH-1a plus mAPH-1b siRNA-treated cells did reveal a reduction in the level of mature nicastrin (Fig. 4D). Together with the observation that mAPH-1 co-immunoprecipitates with both presenilin and nicastrin, these data suggest that mAPH-1 functionally and physically associates with the ␥-secretase complex under physiological conditions.
To test whether mAPH-1 like presenilin or nicastrin modulates ␥-cleavage of APP within its transmembrane domain, we examined whether the ␣or ␤-cleaved carboxyl-terminal fragments (C83 and C99, respectively), the immediate substrates of the ␥-secretase complex, accumulated when endogenous mAPH-1 was inactivated in the mammalian cells. mAPH-1a siRNA treatment of HeLa cells stably expressing full-length APP Swedish protein (HeLa/APP Swedish ) resulted in marked increase of both C83 and C99 (Fig. 5A). Inactivation of mAPH-1a in native HeLa cells also caused the accumulation of C83 and C99 (data not shown). Similarly, and as expected when ␥-secretase activity is inhibited, C83 and C99 accumulated when the cells were treated with either nicastrin or presenilin-1 siRNAs. In contrast, we observed no significant change in the total level of full-length APP or ␣-tubulin (Figs. 4C and 5A). We next used A␤-specific antibodies to determine whether amyloid-␤ peptides, the products of ␥-secretase complex, are inhibited in mAPH-1a siRNA-treated HeLa/APP Swedish cells. Immunoprecipitation followed by immunoblotting experiments revealed a significant reduction of total A␤ in the conditioned media from mAPH-1a siRNA-treated cells as compared with the control cells (Fig. 5A).
We also tested whether mAPH-1a is required for the production of APP intracellular domain, another product of the  -1b (top left panel, lanes 1 and 2) or mAPH-1a (top right panel, lanes 3-5). Please note that mAPH-1b mRNA expression is low in HeLa cells (top left panel, lane 2). The human glioblastoma U118-MG cDNA library was used as a control sample for the mAPH-1b primers (top left panel, lane 1). All of the samples were also amplified using primer pairs specific for presenilin-1 (middle panel) and 18 S rRNA (bottom panel) to control for equal PCR amplifications. B, HeLa cells were transfected with either the indicated siRNA duplexes, sense oligonucleotides (Ϫ), or no oligonucleotide (No) harvested 72 h post-transfection and immunoblotted for presenilin-1 NTF, presenilin-2 NTF, and ␣-tubulin with the appropriate antibodies. C, HeLa/APP Swedish stable cells were subjected to transfection with either the appropriate sense oligonucleotide (Ϫ) or siRNA duplexes (ϩ) corresponding to mAPH-1a, nicastrin, or presenilin-1, respectively, and immunoblotted for presenilin-1 NTF, presenilin-2 NTF, nicastrin, and ␣-tubulin. D, Sense oligonucleotides (Ϫ) or mAPH-1a plus mAPH-1b siRNA-treated samples from panel B were probed with anti-nicastrin antibody. Please note that a lower percentage SDS-polyacrylamide gel was used for this panel as compared with the rest of the panels. Please also note that the presenilin levels are lower in the sample treated with mAPH-1a plus mAPH-1b siRNAs (panel B) than in the sample treated with mAPH-1a siRNAs alone (panels B and C). mAPH-1 Is a Component of the ␥-Secretase Complex ␥-secretase activity. To do so, we used a previously established cell-based assay (2,18,28). In this assay, a UAS-luciferase target gene was co-expressed with a chimeric protein (C99-GV), which was generated by inserting a transcription factor composed of the yeast Gal4 DNA-binding domain and the viral VP16 transcriptional-activating domain (GV) at the junction between the transmembrane and intracellular domains of the APP ␤-cleavage product (C99). C99-GV can only activate the UAS-luciferase target gene if the GV domain is released from the membrane and travels to the nucleus. ␥-Cleavage and nuclear trafficking as measured by luciferase activity was reduced in mAPH-1a (to 55%) or nicastrin (to 50%) siRNAtreated cells when compared with control cells (Fig. 5B). This finding suggests that mAPH-1a is essential for ␥-cleavage of APP to enable the intracellular domain to travel to the nucleus. Taken together, these data indicate that mAPH-1 like presenilin and nicastrin modulates intramembrane proteolysis of APP at the ␥-site in the mammalian cells.
To determine whether siRNA-mediated inactivation of mAPH-1a in HeLa cells would inhibit Notch intramembrane cleavage, a chimeric receptor N⌬E-GV with the Gal4 and VP16 domain (GV) inserted at the junction between the transmembrane and intracellular domains of the membrane-tethered Notch fragment (N⌬E) was used. N⌬E lacks most of the extracellular domain and is an immediate substrate for presenilindependent intramembrane cleavage. N⌬E-GV, like C99-GV, can only activate the UAS-luciferase target gene if the GV domain travels to the nucleus. Treatment of HeLa cells with mAPH-1a or nicastrin siRNAs reduced the luciferase activities to approximately 40ϳ55% of that in the control cells (Fig. 5B), suggesting that nuclear trafficking of N⌬E and hence ␥-secretase activity is reduced when mAPH-1a is down-regulated. A requirement for mAPH-1a in Notch cleavage per se was further supported by direct examination of the NICD production by an immunoblotting method (15,17,29). Similar to the loss of presenilin-1, siRNA-mediated inactivation of mAPH-1a or nicastrin inhibited NICD production (Fig. 5C). Taken together, these data indicate that mAPH-1 like nicastrin and presenilin is specifically required for intramembrane processing of Notch to generate the intracellular domain in the mammalian systems.
Our data thus show that mAPH-1 satisfies three major criteria for a subunit of the ␥-secretase complex in mammalian cells. First, endogenous mAPH-1, nicastrin, and the presenilin NTF:CTF heterodimers physically interact with each other in vivo and under the conditions that are compatible with ␥-secretase activity. Second, like presenilin and nicastrin, mAPH-1 is required for the intramembrane proteolysis of both APP and Notch. Third, the down-regulation of mAPH-1 like presenilin or nicastrin leads to concurrent reduction of the mature nicastrin or the steady-state levels of the presenilin endoproteolytic fragments, the mature presenilin species within the ␥-secretase complex. This later observation indicates that mAPH-1 and nicastrin may to a certain degree limit the formation of the ␥-secretase complex. However, when overexpressed, neither mAPH-1, nicastrin, nor mAPH-1 plus nicastrin was able to boost the levels of the presenilin NTF:CTF heterodimers in the mammalian cells (data not shown). This suggests that mAPH-1 and/or nicastrin are not sufficient to account for the "limiting cellular factors" (31) and that at least one additional protein is required for the restricted assembly of the active ␥-secretase complex. One such protein is probably PEN-2 (25). Indeed, our preliminary data suggest that presenilin, nicastrin, and mAPH-1 can be co-isolated with Myc-His epitope-tagged mammalian PEN-2 in Ni-NTA-agarose pull-down experiments (data not shown). Future studies to understand how mAPH-1 interacts with the other components (presenilin, nicastrin, and possibly PEN-2) during the folding, assembly, and trafficking processes to form an active ␥-secretase complex should shed light on the molecular mechanism of this unusual enzyme in the intramembrane proteolysis and in the pathogenesis of Alzheimer's disease.
FIG. 5. mAPH-1 modulates ␥-cleavage of APP and Notch. A, the siRNA-treated samples from Fig. 4C were blotted for full-length APP, C99, and C83 using an APP-CTF polyclonal antibody, whereas the levels of A␤ in the 16-h conditioned media were analyzed as described under "Experimental Procedures." B, ␥-secretase-mediated cleavage of C99-GV and N⌬E-GV luciferase reporter assays. HeLa cells treated with either the appropriate sense oligonucleotides (Ϫ) or siRNA duplexes (ϩ) corresponding to either mAPH-1a or nicastrin were subjected to transfection with ␤-galactosidase and either C99-GV, N⌬E-GV, or a plasmid containing the Gal4 DNA-binding domain and assayed for luciferase reporter activity. Data collected were normalized to the ␤-galactosidase activity and then converted to fold activation over Gal4. The graph represents a typical experiment performed in quadruplets and is shown as percent activity (means Ϯ S.E.) where 100% activities for the C99-GV (gray bar) and the N⌬E-GV (open bar) assays are 150-and 100-fold activation over Gal4, respectively. C, HeLa cells treated with either sense oligonucleotides (Ϫ), mAPH-1a siRNA duplex, or nicastrin siRNA duplex were subjected to transfection with carboxyl-terminal Myc-tagged N⌬E. Samples were lysed and blotted with anti-Myc 9E10. The migration position of the cleaved NICD is shown on the left. mAPH-1 Is a Component of the ␥-Secretase Complex