A Conserved G XXX G Motif in APH-1 Is Critical for Assembly and Activity of the (cid:1) -Secretase Complex*

The multipass membrane protein APH-1, found in the (cid:1) -secretase complex together with presenilin, nicastrin, and PEN-2, is essential for Notch signaling in Caenorhabditis elegans embryos and is required for intramembrane proteolysis of Notch and (cid:2) -amyloid precursor protein in mammalian and Drosophila cells. In C. elegans , a mutation of the conserved transmembrane Gly 123 in APH-1 (mutant or28 ) leads to a notch / glp-1 loss-of-func-tion phenotype. In this study, we show that the corre-sponding mutation in mammalian APH-1a L (G122D) disrupts the physical interaction of APH-1a L with hypoglycosylated immature nicastrin and the presenilin holoprotein as well as with mature nicastrin, presenilin, and PEN-2. The G122D mutation also reduced (cid:1) -secre-tase activity in intramembrane proteolysis of mem-brane-tethered Biotechnology). Inhibitors -secretase activity ( -(3,5-difluoro- -phenylglycine ester and -(2-naph-thoyl)-Val-phenylalaninal (inhibitor (24)) Calbiochem. Coprecipitation Studies— A-agarose min each lysis buffer. Immunoprecipitated proteins were eluted M glycine 2.5) and 0.25% detergent, neutralized with 1.0 M Tris, subjected to Western blot analysis. Ni-NTA-agarose affinity pull- down experiments were performed as in 1% digitonin or 1% CHAPSO lysis buffer. Each binding experiment contained this simplified model, APH-1, PEN-2, presenilin, and nicastrin ( NCT ) exist in at least three major states during the initial assembly and subsequent maturation of the active (cid:2) -secretase complex: 1) unassembled immature components, 2) immature complex (probably labile in most cells), and 3) active/mature complex. Our data and results from other groups (10–16) suggest that APH-1 functions as a molecular scaffold for the (cid:2) -secretase complex. G XXX G-dependent helix-helix association plays an essential role in the initial assembly of the immature complex ( Step I ) and is also required for stabilizing the active mature complex ( Step II ). Transition of the immature complex to the mature complex is probably associated with a presenilin ( PS ) endoproteolysis-triggered conformational change ( Step II ). Note that multiple substeps likely exist within the two major steps described here. B , mutations of the G XXX G motif prevent APH-1 from performing its scaffolding role in the initial assembly of the immature complex (and consequently in maturation of the active complex). G XXX G mutant APH-1 may recruit additional molecule(s) (marked X ), self-oligomerize, or adopt an alternative folding. In agreement with this view, we found that mutant APH-1 could still associate with overexpressed (presumably unassembled/immature) PEN-2 and that mutant APH-1 existed in a high molecular weight complex without associating with presenilin and nicastrin. It should be noted that the cognate site of the APH-1 G XXX G motif is unknown and that APH-1, PEN-2, presenilin, and nicastrin could have multiple contacts with one another. It should also be noted that the stoichiometry of the (cid:2) -secretase complex is unclear and that (cid:2) -secretase subunits could potentially form homodimers or homo-oligomers. For simplicity, only one copy of each subunit was depicted. Im , immature; m , mature; PS-holo , presenilin holoprotein.

Regulated intramembrane proteolysis of the Notch receptor, the ␤-amyloid precursor protein, and select type I membrane polypeptides represents a novel mechanism of signal transduction (1). It is now generally believed that the enzyme responsible for the intramembrane proteolysis of these substrates is the high molecular mass, multiprotein ␥-secretase complex, which consists of a heterodimer of the presenilin amino-and carboxyl-terminal endoproteolytic fragments (NTF 1 and CTF, respectively) as the putative catalytic subunit (2,3). A single transmembrane glycoprotein (nicastrin) has also been identified as a critical component of the ␥-secretase complex (4). Recent genetic studies in Caenorhabditis elegans have uncovered two additional genes, aph-1 and pen-2, both of which encode multipass membrane proteins that are required for ␥-secretase activity and notch/glp-1 signal transduction (5,6). Subsequent experiments in mammalian cells have indicated that the functionally conserved APH-1 and PEN-2 proteins physically associate with nicastrin and the presenilin NTF/ CTF heterodimer and are essential for Notch and ␤-amyloid precursor protein intramembrane proteolysis (7,8). Overexpression of presenilin, nicastrin, APH-1, and PEN-2 together produces or enhances ␥-secretase activity in yeast, insect cells, and mammalian cells, further supporting the hypothesis that these four proteins are essential for ␥-secretase activity (9 -13). Aside from the putative role of presenilin as the catalytic subunit of the ␥-secretase complex, the specific functions of nicastrin, APH-1, and PEN-2 remain unclear. Recent studies indicate that PEN-2 may be important in initiating presenilin endoproteolysis, whereas APH-1 may play a role in stabilizing the ␥-secretase complex (12)(13)(14)(15). Recent studies also suggest that APH-1 and nicastrin may interact to form a subcomplex prior to the assembly of the ␥-secretase complex (10,16). Despite these advances in research, the mechanism of ␥-secretase assembly in the lipid bilayers to form and maintain an active membrane protein complex capable of performing intramembrane proteolysis within a hydrophobic environment remains an enigma.
It has been reported that mutation of Gly 123 to aspartic acid within the fourth putative transmembrane region (TMR) of C. elegans APH-1 (mutant or28) inhibits notch/glp-1 signal transduction and leads to the "anterior-pharynx-defective" phenotype associated with the loss of aph-1, aph-2 (nicastrin), or sel-12 plus hop-1(presenilins) (5). This glycine residue in C. elegans APH-1 has been conserved during evolution and corresponds to Gly 122 in mammalian APH-1a L and APH-1a S and Gly 121 in mammalian APH-1b. APH-1a L and APH-1a S are alternative spliced forms of APH-1a that differ at the carboxyl termini (7). Because it is known that the role of the APH-1 proteins in Notch signaling is conserved during evolution and is associated with ␥-secretase function, we set out to investigate the biochemical function of the conserved glycine residue in the assembly and activity of the ␥-secretase complex. In this study, we used human APH-1a L as an example to demonstrate that the equivalent mutation in human APH-1a L (G122D) af-fects the ability of APH-1 to associate with the immature as well as the mature ␥-secretase complex and inhibits the intramembrane proteolysis of Notch. Moreover, we found that Gly 122 , Gly 126 , and Gly 130 in APH-1a L belong to a conserved GXXXGXXXG motif (where X represents any amino acid) generally accepted as a major determinant in transmembrane helix-helix protein interactions (17)(18)(19)(20)(21)(22) and show that these glycine residues are critical for ␥-secretase assembly and activity.

EXPERIMENTAL PROCEDURES
cDNA Constructs-Human full-length APH-1a L and PEN-2 cDNAs were obtained as described (7,15). Subsequent site-directed mutagenesis studies were performed using the QuikChange kit (Stratagene), and the identities of all clones were confirmed by DNA sequence analyses.
Cell Lines-Cells were maintained in Dulbecco's modified Eagle's medium and 10% fetal bovine serum (Invitrogen) with 5% CO 2 at 37°C. For transient expression, plasmids were transfected into the appropriate cell lines in 10-cm dishes using LipofectAMINE 2000 (Invitrogen), and samples were collected 48 h after transfection. HEK293 cells stably or transiently transfected with Myc/His-tagged APH-1a L and its glycine mutants were used to evaluate the interaction of APH-1 with presenilin, nicastrin, and PEN-2.
Coprecipitation Studies-Co-immunoprecipitations were performed as described previously (4,25,26). Briefly, proteins extracted using lysis buffer (1% digitonin or 1% CHAPSO in 50 mM sodium phosphate (pH 8.0), protease inhibitors (Roche Applied Sciences) and 300 mM NaCl) were pre-absorbed with preimmune serum and combined with anti-Myc antibody A14 overnight and then with protein A-agarose beads for 2 h at 4°C. The beads were washed four times for 15 min each with lysis buffer. Immunoprecipitated proteins were eluted with 0.1 M glycine HCl (pH 2.5) and 0.25% detergent, neutralized with 1.0 M Tris, and subjected to Western blot analysis. Ni-NTA-agarose affinity pulldown experiments were performed as described previously (7) in 1% digitonin or 1% CHAPSO lysis buffer. Each binding experiment con-tained equal amounts of proteins with similar expression of Myc/Histagged proteins.
␥-Secretase Activity Assay-The appropriate cell lines were assayed for ␥-cleavage of Notch using the N⌬E-GV luciferase reporter as described (7,27) in an adapted in vitro ␥-secretase activity assay (28). The in vitro activity assay involved lysis of membrane proteins with 1% CHAPSO, 50 mM PIPES (pH 7), 5 mM MgCl 2 , 5 mM CaCl 2 , and protease inhibitors (Roche Applied Science) and immunoprecipitation of Myctagged proteins with immobilized anti-Myc antibody 9E10. The beads were then washed with CHAPSO lysis buffer and subjected to incubation at 37°C in the presence of 0.25% CHAPSO, 50 mM PIPES (pH 7), 5 mM MgCl 2 , 5 mM CaCl 2 , 0.0125% phosphatidylethanolamine, 0.1% phosphatidylcholine, and N100 substrate. This N100 substrate, which was overexpressed and purified from Sf9 cells, harbors Val 1711 -Glu 1809 of the mouse Notch-1 receptor and a carboxyl-terminal FLAG/His tag. The ␥-secretase-cleaved N100-FLAG/His fragment was detected with anti-FLAG antibody.
Miscellaneous-All experiments are repeated at least four times, and representative data are presented. Proteins were separated by SDS-PAGE, immunoblotted with the appropriate antibodies, and processed using ECL reagents (Amersham Biosciences). Digitonin-solubilized membranes were fractionated using a 10 -30% linear glycerol velocity gradient as described (25), and 1-ml fractions collected from the top of the gradient were subjected to Western blot analysis. Immunofluorescence was analyzed using anti-Myc antibody A14 and Alexa Fluor® 568-conjugated secondary antibody (Molecular Probes, Inc.) on a Leica TCS SP2 laser scanning spectral confocal microscope.

RESULTS
Gly 122 in APH-1a L Is Critical for ␥-Secretase Assembly-To understand the effect of the G122D mutation on the ␥-secretase complex, we analyzed the ability of human APH-1a L (either wild-type (WT) or G122D mutant) to interact with the predominant in vivo active presenilin species, the endogenous presenilin endoproteolytic fragments. To assist in this objective, we generated HEK293 cells stably expressing APH-1a L WT-Myc/ His or APH-1a L G122D-Myc/His and examined cell lines that expressed a comparable level of either wild-type or mutant Myc/His-tagged proteins. No obvious difference was observed in the level of endogenous presenilin-1 NTF between the APH-1a L WT-Myc/His and APH-1a L G122D-Myc/His cells (Fig. 1A,  lanes 1-4). In Ni-NTA-agarose pull-down experiments, we observed that endogenous presenilin-1 NTF was able to coprecipitate with APH-1a L WT-Myc/His, an observation consistent with a previous report (7). However, the amount of presenilin-1  8) were subjected to Ni-NTA pull-down experiments. The Ni-NTA pull-down products (lanes 5-8) were resolved by SDS-PAGE and probed with antibodies against presenilin-1 NTF (PS1-NTF) and the Myc epitope as indicated. Similar results were also obtained using 1% CHAPSO extracts. Note that the Myc/His-tagged proteins and presenilin-1 NTF were expressed at similar levels in each set of cells (Input) and that the sizes of the Myc/His-tagged proteins are different, but were aligned to save space. B, HEK293 cells stably overexpressing APH-1a L WT-Myc/His (lane 1) or APH-1a L G122D-Myc/His (lane 2) were solubilized in 1% CHAPSO and immunoprecipitated (IP) using anti-Myc antibody A14. The immunoprecipitated products were resolved by SDS-PAGE, and the immunoblots were investigated with antibodies to presenilin-1 NTF, presenilin-2 NTF (PS2-NTF), and the Myc epitope as indicated.
NTF that coprecipitated with APH-1a L WT-Myc/His was much higher than the amount that co-isolated with APH-1a L G122D-Myc/His (Fig. 1A, lanes [5][6][7][8]. Using a co-immunoprecipitation approach, we similarly observed that a minute amount of presenilin-1 NTF or presenilin-2 NTF coprecipitated with APH-1a L G122D-Myc/His compared with the ample amount of presenilin-1 and -2 fragments that coprecipitated with APH-1a L WT-Myc/His (Fig. 1B). Based on these observations, we concluded that the G122D mutation abrogates the association of APH-1a L with the mature presenilin species. Next, we tested whether Gly 122 in human APH-1a L is required for interaction with the immature presenilin holoprotein. Given that the endogenous presenilin holoprotein is usually maintained at a low or undetectable steady-state level, we transiently overexpressed full-length presenilin-1 in cell lines stably expressing APH-1a L WT-Myc/His and APH-1a L G122D-Myc/His and performed Ni-NTA-agarose pull-down experiments. We found that significantly less full-length presenilin-1, as well as presenilin-1 endoproteolytic fragments, coprecipitated with APH-1a L G122D-Myc/His compared with APH-1a L WT-Myc/His (Fig.  2, lanes 1-8). These findings suggest that the G122D mutation disrupts the ability of APH-1a L to interact with both the mature presenilin endoproteolytic fragments and the immature presenilin holoprotein.
Our next investigation focused on the ability of immature and mature nicastrin to coprecipitate with APH-1a L . According to published reports, hypoglycosylated immature nicastrin is mainly associated with immature full-length presenilin, whereas fully glycosylated mature nicastrin is associated with presenilin NTF/CTF fragments (29,30). In our study, we observed that APH-1a L WT-Myc/His coprecipitated with more of the immature than the mature species of nicastrin in cells overexpressing APH-1a L and the presenilin-1 holoprotein (Fig. 2,lane 6), an observation consistent with earlier studies (10,14). Moreover, we found that much less immature and mature nicastrin co-isolated with APH-1a L G122D-Myc/His compared with APH-1a L WT-Myc/His (Fig. 2, lanes 5-8). We also observed that cells stably expressing APH-1a L WT-Myc/His contained slightly more immature nicastrin than cells stably expressing APH-1a L G122D-Myc/His (Fig. 2, lanes 1-4), suggesting that the G122D mutation may also affect the reported ability of wild-type APH-1 to stabilize immature nicastrin (10,16). Based on these findings, we conclude that Gly 122 mediates the association of APH-1a L with both immature and mature nicastrin either directly or indirectly during the assembly process of the ␥-secretase complex.
To test the effect of the G122D mutation on the interaction between APH-1a L and PEN-2, we transiently expressed HAtagged PEN-2 in either the APH-1a L WT-Myc/His or APH-1a L G122D-Myc/His cells. Detergent extracts of these cells were then subjected to Ni-NTA pull-down experiments. We observed that the association between APH-1a L and overexpressed HA-PEN-2 was not significantly affected by the G122D mutation, even though the interactions between APH-1a L and both presenilin-1 and nicastrin were disrupted (Fig. 3A). To examine whether the G122D mutation affects the association of endogenous PEN-2 with APH-1a L , we next performed the same Ni-NTA pull-down experiments with detergent-solubilized HEK293 cells stably expressing APH-1a L WT-Myc/His or APH-1a L G122D-Myc/His. In contrast to overexpressed HA-PEN-2, endogenous PEN-2 did not strongly coprecipitate with mutant APH-1a L compared with wild-type APH-1a L (Fig. 3B). The reason that the G122D mutation affected binding of APH-1a L to endogenous (but not overexpressed) PEN-2 is unclear. One possibility is that overexpressed PEN-2 exists in a state parallel to immature nicastrin and the presenilin holoprotein, whereas endogenous PEN-2 exists mostly in a "mature" state similar to the presenilin endoproteolytic fragments and mature nicastrin. In this context, "immature" overexpressed PEN-2 and APH-1a L can directly interact with each other in the absence of both presenilin-1 and nicastrin, perhaps via a site or sites other than Gly 122 in APH-1a L . Because the active ␥-secretase complex is not stable when any one of the four components is missing (2), the failure of mature PEN-2 to interact with G122D mutant APH-1a L might be an indirect consequence of the disruption of the association of APH-1 with presenilin and nicastrin (see "Discussion"). Mutation G122D in APH-1a L Inhibits ␥-Secretase Activity-Having established that the G122D mutation in APH-1a L disrupted the structural assembly of the ␥-secretase complex, we proceeded to investigate whether this structural defect could lead to deleterious effects on ␥-secretase activity. We tested the effect of the G122D mutation in APH-1a L on the production of the Notch intracellular domain (a product of ␥-secretase activity) in a cell-based Gal4/VP16-dependent luciferase transactivation assay. This assay, which indirectly measures ␥-secretase activity, utilizes a ␥-secretase substrate consisting of a chimeric membrane-tethered Notch receptor trimmed in the extracellular domain (N⌬E) and Gal4 and VP16 domains (GV) inserted at the junction between the transmembrane and intracellular domains of N⌬E (7,27). In HEK293 cells stably expressing robust amounts of APH-1a L G122D-Myc/His, we observed a 50 -60% reduction in luciferase reporter activity compared with control cells expressing comparable amounts of APH-1a L WT-Myc/His (Fig. 4A). No analysis could be performed on cells expressing higher levels of APH-1a L G122D-Myc/His because these cells did not grow well (possibly because of toxic effects of the G122D mutation). However, the modest reduction in ␥-secretase activity in cells stably expressing APH-1a L G122D-Myc/His obtained in the transactivation assay contrasts with the dramatic effect of the G122D mutation imposed on the structural assembly of the ␥-secretase complex and the loss-of-function notch/glp-1 phenotype observed in the C. elegans or28 mutant. A likely explanation for this discrepancy could be that the effect of the G122D mutation on ␥-secretase activity is masked by the presence of endogenous APH-1 proteins in the cells.
To examine the effect of the G122D mutation on ␥-secretase activity in the absence of endogenous APH-1, we immunoprecipitated the ␥-secretase complex containing either APH-1a L WT-Myc/His or APH-1a L G122D-Myc/His from the respective CHAPSO-solubilized membranes and measured the ability of the immunoprecipitate to cleave a purified Notch substrate containing the transmembrane S3/␥-secretase-like cleavage site (N100) in an in vitro ␥-secretase activity assay. We observed a much more robust N100 cleavage in the immunoprecipitate containing wild-type APH-1a L compared with mutant APH-1a L . The presence of the N100 proteolytic intracellular fragment was also sensitive to two different ␥-secretase inhibitors (Fig. 4B). Taken together, these findings show that the G122D mutation in APH-1a L has an inhibitory effect on ␥-secretase activity.
Our studies suggest that the G122D mutation perturbs the ability of APH-1a L to participate in the initial assembly process of the ␥-secretase complex and to associate with and stabilize the active ␥-secretase complex to directly modulate the intramembrane proteolysis of Notch. However, we currently cannot rule out the possibility that the G122D mutation affects proper trafficking of APH-1a L within cells and thus the formation of the ␥-secretase complex. We attempted to address this issue by assessing the cellular distribution of wild-type and G122D mutant APH-1a L . Immunofluorescence data revealed similar intracellular punctate/vesicular staining patterns for both wild-type and G122D mutant APH-1a L (Fig. 5A), suggesting that G122D mutant APH-1a L does not have a gross trafficking defect that could account for its effect on ␥-secretase assembly and activity. We also used linear glycerol density centrifugation to examine the native state of wild-type and G122D mutant APH-1a L and found that the G122D mutation did not significantly alter the native state of APH-1a L (Fig. 5B). Interestingly, G122D mutant APH-1a L , like wild-type APH-1a L , existed in similar high molecular mass glycerol density fractions (Fig. 5B). It is not clear why G122D mutant APH-1a L was found in high molecular mass fractions since it is incapable of tight association with presenilin and nicastrin. One probable explanation is that the APH-1a L mutant recruits additional molecules in the absence of presenilin and nicastrin.
APH-1a L Contains a Conserved GXXXG Motif Important in Transmembrane Helix-Helix Association-We have noticed that the conserved transmembrane Gly 122 and Gly 126 in APH-1a L are part of a sequence arrangement that resembles a sequence motif important in transmembrane protein interactions (Fig. 6). This highly conserved sequence arrangement known as the GXXXG motif was originally identified as a requirement for the dimerization of a single TMR of glycophorin A (17,18,22,31) and is recognized as a high affinity transmembrane helixhelix binding motif for many other membrane proteins, including aquaporin-1, ErbB-4, ATP synthase, and the anion-selective membrane channel VacA (19 -21, 32-37).
To determine whether the 122 GXXXG 126 sequence is indeed part of a bona fide transmembrane interaction motif, we evaluated the possibility that the disruptive effect of G122D APH-1a L on the ␥-secretase complex is not dependent exclusively on changing the glycine to a residue with a charged side chain such as aspartic acid. We replaced Gly 122 with either the nonpolar residue alanine or proline and tested their ability to interact with endogenous presenilin, nicastrin, and PEN-2. In our Ni-NTA-agarose pull-down experiments, transiently transfected APH-1a L G122A-Myc/His coprecipitated with similar amounts of presenilin-1 NTF, nicastrin, and PEN-2 as APH-1a L WT-Myc/His (Fig. 7A, lanes 12 and 14), whereas APH-1a L G122P-Myc/His failed to interact with presenilin-1 NTF, nicastrin, or PEN-2 (lane 15), suggesting that the effect of the Gly 122 mutation is not solely dependent on the aspartate. The subtle/no difference in the binding of G122A mutant APH-1a L to presenilin-1 NTF, nicastrin, or PEN-2 compared with wildtype APH-1a L is likely due to the conservative substitution of alanine for glycine. In accordance with previous observations (18,22), in some (but not all) cases, less steric clashes may occur upon replacing the glycine with alanine compared with other residues and thus could be less disruptive for the ␥-secretase complex.
To further establish that the APH-1a L 122 GXXXG 126 sequence is part of an authentic transmembrane interaction motif, we replaced the conserved transmembrane glycine residue of APH-1a L at position 126 with either alanine or leucine. Compared with APH-1a L WT-Myc/His, the G126A mutation diminished the interaction of APH-1a L -Myc/His with endogenous presenilin-1 NTF, nicastrin, and PEN-2 (Fig. 7A, compare  lanes 12 and 16). Moreover, the G126L mutation in APH-1a L ablated the interaction to the same extent as the G122P or G122D mutation (Fig. 7A, compare lanes 13, 15, and 17). Mutations of other transmembrane or cytosolic glycine residues to alanine in APH-1a L such as at position 15 (in TMR-1), position 130 (in TMR-4), or position 145 (in the cytosolic region after TMR-4) did not significantly affect the interaction of presenilin-1 NTF, nicastrin, or PEN-2 with APH-1a L (Fig. 7A, lanes  18 -20).
We next tested the effects of some of these glycine mutations  1 and 11), APH-1a L WT-Myc/His (lanes 2 and 12), or APH-1a L -Myc/His harboring a single mutation of the glycine residue indicated (lanes 3-10 and 13-20) were lysed in 1% digitonin. After determining that similar amounts of the Myc/His-tagged proteins, endogenous presenilin-1, nicastrin, and PEN-2 were expressed in each set of cells, we subjected equal amounts of proteins (lanes 1-10) to Ni-NTA pull-down experiments, and the resultant products (lanes [11][12][13][14][15][16][17][18][19][20] were resolved by SDS-PAGE and probed with antibodies to presenilin-1 NTF (PS1-NTF), nicastrin (NCT), PEN-2, and the Myc epitope as indicated. B, cell lines stably expressing LacZ-Myc/His (lane 1), APH-1a L WT-Myc/His (lanes 2 and 6), or APH-1a L -Myc/His harboring a single mutation of the glycine residue indicated (lanes 3-5 and 7-8) were processed as described in the legend to Fig. 4B and subjected to incubation with N100-FLAG/His at 37°C for 0 and 6 h. Samples were electrophoresed on SDS-polyacrylamide gel and probed with anti-FLAG antibody (upper and middle panels) or with anti-Myc antibody (lower panel). IP, immunoprecipitation. on ␥-secretase activity. Because the in vitro assay reliably measured ␥-secretase activity when cells stably overexpressing wild-type or mutant APH-1a L -Myc/His were used, we generated stable HEK293 cell lines for G15A, G122A, G122P, G126L, in addition to the cells stably expressing APH-1a L WT-Myc/His and APH-1a L G122D-Myc/His described above. We immunoprecipitated the ␥-secretase complex containing either wild-type or mutant APH-1a L -Myc/His from the respective CHAPSO-solubilized membranes and measured in vitro ␥-secretase activity as described in the legend to Fig. 4B. Like G122D, G122P and G126L abrogated N100 cleavage, whereas G122A and G15A did not have noticeable effects on N100 cleavage (Fig. 7B). The observation that the G122A mutation did not inhibit ␥-secretase activity is consistent with our earlier finding that G122A mutant APH-1a L did assemble with the other subunits of the ␥-secretase complex. Based on our findings, we conclude that the conserved Gly 122 and Gly 126 in APH-1a L are critical residues that make up a highly specific GXXXG motif required for the assembly and activity of the multimeric ␥-secretase complex.
Sequence comparison and ␣-helical wheel analysis suggested that residues surrounding APH-1a L Gly 122 and Gly 126 , such as Ala 118 , Ser 129 , Gly 130 , and Ser 133 , may also be important for transmembrane helix-helix association. In particular, Gly 130 is highly conserved, with its equivalent residue in C. elegans or Arabidopsis thaliana APH-1 being alanine (Fig. 6). Our earlier experiment showing that the G130A mutation did not affect the ability of APH-1a L to associate with the other ␥-secretase subunits (Fig. 7A) suggests that the conserved alanine-for-glycine substitution at position 130 may be tolerated, as at position 122. To test whether Gly 130 (like Gly 122 and Gly 126 ) is also a critical residue for transmembrane interaction, we replaced Gly 130 with aspartate and tested the ability of G130D mutant APH-1a L to interact with the other ␥-secretase components. In our Ni-NTA-agarose pull-down experiments, much less endogenous presenilin-1 NTF, mature and immature nicastrin, and PEN-2 coprecipitated with transiently expressed APH-1a L G130D-Myc/His compared with APH-1a L WT-Myc/His (Fig.  8). These findings suggest that Gly 130 (like Gly 122 and Gly 126 ) is critical for mediating interactions in the multimeric ␥-secretase complex, providing evidence that the glycophorin A-like helix-helix binding motif in putative TMR-4 of APH-1a L is longer ( 122 GXXXGXXXG 130 ). Further studies will be needed to elucidate the exact sequence and structural requirements for this critical motif in the assembly and activity of the ␥-secretase complex. DISCUSSION As the ␥-secretase complex is largely buried in the cellular membrane, and ␥-secretase activity is carried out within the hydrophobic lipid bilayer, it is conceivable that molecular interactions among the four components of the ␥-secretase complex occur mainly through their TMRs and that helix-helix interactions are essential for the assembly and molecular actions of ␥-secretase. To date, little is known about the transmembrane interactions among the four components of the ␥-secretase complex. In this study, we have identified the conserved transmembrane Gly 122 , Gly 126 , and Gly 130 in TMR-4 of mammalian APH-1a L as part of the GXXXG motif, which is widely accepted as an important determinant in transmembrane helix-helix interactions. Indeed, mutations of Gly 122 , Gly 126 , and Gly 130 in the GXXXG motif prevent the stable and specific association of APH-1a L with the ␥-secretase complex. The disruption in the assembly of the ␥-secretase complex is the most probable explanation for the loss of ␥-secretase activity associated with G122D mutant APH-1a L shown in this study and for the recessive loss-of-function notch/glp-1 phenotype associated with the C. elegans or28 mutant, which harbors an equivalent mutation (5). We propose that the GXXXG motif of APH-1 is a major and highly specific docking or packing site for the assembly of the ␥-secretase complex.
Because mutations of the GXXXG motif disrupt the interaction of APH-1a L with both the immature and mature forms of presenilin and nicastrin and subsequently affect ␥-secretase activity, it is likely that the GXXXG-mediated docking or packing of APH-1 is an early event in the formation of the active complex and that the GXXXG motif in APH-1 also exerts its functional effect to maintain the stability of the active mature ␥-secretase complex (Fig. 9). This model is in agreement with the proposed stabilizing role for APH-1 in the assembly of the ␥-secretase complex (10 -15). However, the exact mechanism of how the APH-1 GXXXG motif regulates the assembly and maturation process of the ␥-secretase complex awaits further investigation. Among the key unresolved issues are the identities of the direct APH-1-binding partner and the sequence of the opposing transmembrane helix that associates with the APH-1 GXXXG motif. In several reported cases, the homo-and hetero-oligomerization of two transmembrane helices require the packing of the GXXXG sequences in conjunction with their surrounding residues contributed by the respective transmembrane helices (20,31). In this regard, it is possible that the opposing sequence that associates with the GXXXG sequence in APH-1a L is also a transmembrane GXXXG motif or a variation of the motif. It is thus interesting to speculate that two or more copies of APH-1 could associate via their GXXXG sequences to serve as a molecular scaffold for oligomerization of the other ␥-secretase subunits, such as the putative presenilin dimer (38). Another possibility is that the APH-1 GXXXG sequence may directly interact with presenilin. The presenilin family proteins harbor several conserved transmembrane glycine residues that could potentially constitute the GXXXG motif. On the other hand, recent biochemical reports suggest that APH-1 and nicastrin interact to form a stable intermediate subcomplex at the early stage of ␥-secretase assembly (10,16). It is possible that the inability of the presenilin proteins to be co-isolated with the GXXXG mutants of APH-1a L is caused in some way by a prior defect in the direct binding of the TMRs of APH-1a L and nicastrin. However, the single TMR of nicastrin does not possess a typical GXXXG motif. It remains to be determined whether APH-1 interacts with nicastrin via a GXXXG dependent or GXXXG-independent mechanism. Finally, it is possible that mutations of the GXXXG motif in putative TMR-4 of APH-1 could disrupt interaction with another TMR in APH-1 itself or with another APH-1 molecule.
In summary, our findings have revealed a specific role for APH-1 in the scaffolding of the ␥-secretase complex. We have demonstrated that the conserved membrane helix-helix interaction GXXXG motif in APH-1 is a critical determinant for the specific interactions among the constituents of the ␥-secretase complex. Although the GXXXG motif is unlikely to be the sole determinant in this process, this study, which focused on the packing/docking of APH-1a L to the other ␥-secretase components, should provide a framework for understanding the molecular mechanism underlying the assembly, folding, and other regulatory processes of the ␥-secretase complex in intramembrane proteolysis. Further investigation into the role of the GXXXG motif in the assembly and activity of the ␥-secretase complex should shed new light on transmembrane proteinprotein interactions within the lipid bilayers, a largely uncharted challenge that extends beyond the boundaries of the ␥-secretase complex. Finally, our findings suggest that targeting the GXXXG motif in APH-1 could be an alternative therapeutic strategy for Alzheimer's disease and related disorders because compounds that block the docking or packing sites in the ␥-secretase complex might control the assembly and activity of the active enzyme.