Three-dimensional Structure of the Signal Peptide Peptidase*

Signal peptide peptidase (SPP) is an atypical aspartic protease that hydrolyzes peptide bonds within the transmembrane domain of substrates and is implicated in several biological and pathological functions. Here, we analyzed the structure of human SPP by electron microscopy and reconstructed the three-dimensional structure at a resolution of 22 Å. Enzymatically active SPP forms a slender, bullet-shaped homotetramer with dimensions of 85 × 85 × 130 Å. The SPP complex has four concaves on the rhombus-like sides, connected to a large chamber inside the molecule. Intriguingly, the N-terminal region of SPP is sufficient for the tetrameric assembly. Moreover, overexpression of the N-terminal region inhibited the formation of the endogenous SPP tetramer and the proteolytic activity within cells. These data suggest that the homotetramer is the functional unit of SPP and that its N-terminal region, which works as the structural scaffold, has a novel modulatory function for the intramembrane-cleaving activity of SPP.

The intramembrane-cleaving proteases (I-CLiPs) 5 that sever the transmembrane domains of their substrates have been identified in a range of organisms and play a variety of roles in biological and pathological conditions (1). I-CLiPs have been classified into three groups: serine-, aspartyl-, and metalloprotease-type, according to the structure of active sites. Presenilin (PS) and signal peptide peptidase (SPP) family proteins belong to the group of aspartyl I-CLiPs (2,3). These polytopic proteases have nine transmembrane domains with the two catalytic aspartates as YD and GXGD motifs. Several ␥-secretase inhibitors cross-inhibit the SPP activity, suggesting that PS, the catalytic subunit of ␥-secretase, and SPP share a similar structure and proteolytic mechanism (4 -8). However, ␥-secretase requires three cofactor proteins (i.e. nicastrin, aph-1, and pen-2) in addition to PS (9 -11), whereas SPP alone exhibits catalytic function not requiring other protein cofactors (4). SPP is implicated in the clearance of signal peptides as well as misfolded membrane proteins (12)(13)(14). Moreover, some endoproteolytic products generated by SPP cleavage directly mediate signal transduction (15,16). In fact, loss-of-function studies of SPP in model animals resulted in severe developmental defects, inferring a vital role of SPP in metazoan development (17)(18)(19). Furthermore, a growing body of evidence indicates that SPP activity plays an important role in the maturation of several pathogens including the hepatitis C virus and the malaria parasite (7,20). Thus, understanding the structure and function relationship of SPP as well as the rational development of its inhibitors should have a significant therapeutic potential for these infectious diseases. Here, we found that SPP proteins formed a tetramer in the enzymatically active condition. Single particle reconstruction from electron microscopic images revealed that the purified SPP forms a bullet-like shape with concaves on the surface and a large chamber in the center. Intriguingly, overexpression of the N-terminal region of SPP, which is sufficient for the tetrameric assembly, led to the inhibition of the proteolytic activity. Our first study on the structure of SPP reveals its submolecular configuration and highlights a novel modulatory mechanism of the N-terminal region on the proteolytic activity of SPP.
Sample Preparation, Immunoprecipitation, Photoaffinity Labeling, Electrophoresis, and Immunoblotting-The cells were homogenized in 10% w/v glycerol-containing HEPES buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, and Complete protease inhibitor mixture (Roche Applied Science)) and subsequently centrifuged at 1,000 ϫ g for 10 min. The supernatants were centrifuged again at 100,000 ϫ g for 60 min to isolate the microsome fraction. The microsomes or cells were resuspended in 2% n-dodecyl-␤-D-maltopyranoside (DDM)-containing HEPES buffer to designate them as the solubilized microsomes or cell lysates. For large scale preparation, solubilized microsome fractions from infected Sf9 cells were loaded on anti-FLAG M2-agarose and eluted with 500 g/ml 3ϫFLAG peptide-containing 2% DDM-containing HEPES buffer after being washed three times. Eluate was further separated by a Superose 6 HR 10/30 column (GE Healthcare) on an Ä KTA Explorer chromatography system (GE Healthcare). For immunoprecipitation, DDM-solubilized fractions were coincubated with primary antibody and protein G-agarose (Invitrogen). Photoaffinity labeling experiments were performed as described previously, with some modifications (24,25,29). Briefly, microsomal fractions or 0.25% DDM-solubilized fractions were incubated with 50 nM photoprobes (i.e. L-852,505, L-852,646) and then irradiated for 90 min. Irradiated samples were adjusted to 1% SDS and rocked with immobilized streptavidin (GE Healthcare) overnight. Biotinylated proteins were eluted with SDS sample buffer by heating for 1 min and subjected to immunoblotting. SDS-PAGE and immunoblotting were performed as described previously (30). Blue-Native PAGE (BN-PAGE) was performed according to the manufacturer's protocol (Invitrogen). Briefly, membrane fractions were suspended in NativePAGE TM sample buffer containing 1% DDM. The mixture was centrifuged for 10 min at 15,000 ϫ g, and Coomassie Brilliant Blue was added to the supernatant to give a final concentration of 0.25% w/v. NativeMark TM unstained protein standard (Invitrogen) was used as a molecular weight standard. After electrophoresis, the gel was transferred to PVDF membranes. The membranes were destained briefly in methanol before being incubated with specific antibodies.
SPP Activity Assay-To measure SPP activity in vitro, solubilized samples were incubated with Myc-Prl-PP-FLAG peptide (BEX Co., Ltd.) (31) at 37°C. For SPP reporter assay in vivo (22), constructs encoding dSPP, ERSE-firefly luciferase, SPP sub , and Renilla luciferase were transfected to S2 cells. Luciferase activities were measured by the PicaGene Dual luciferase system (TOYO B-Net. Co., LTD. Tokyo, Japan) according to the manufacturer's instructions.
Transmission Electron Microscopy-Fractions were adsorbed by thin carbon films rendered hydrophilic by glow discharge in low air pressure and supported by copper mesh grids. Samples were washed with five drops of double-distilled water, negatively stained with 2% uranyl acetate solution for 30 s twice, blotted, and air-dried. For immuno-EM, purified 3FSPP and anti-FLAG M2 monoclonal antibody were mixed for 30 min at 4°C, and excess antibodies were removed by gel filtration chromatography (SMART system (GE Healthcare)). Samples were then negatively stained as described. Micrographs of negatively stained particles were recorded in a JEOL 100CX transmission electron microscope (JEOL, Tokyo, Japan) at ϫ53,100 magnification with 100-kV acceleration voltages. Images were recorded on SO-163 films (Eastman Kodak Co.), developed with a D19 developer (Kodak) and digitized with a Scitex Leafscan 45 scanner (Leaf systems, Westborough, MA) at a pixel size of 1.92 Å at the specimen level.
Automated Particle Selection and Image Analysis-Single particle image analysis (32) was performed using our SPINNS program and the IMAGIC V program (33-37). The projections were picked up by a combination of two automatic pickup programs: the auto-accumulation method using simulated annealing (34) and the three-layered neural network method (35). Initially, 531 particles, in 160 ϫ 160-pixel subframes, were selected from five EM images and used to train a pyramid-type neural network. Using the trained neural network, 4,692 particles were selected. The images were band pass-filtered with a low frequency cutoff of 384 Å and a high frequency cutoff of 4 Å, using IMAGIC V. The following image analysis was performed in three steps. First, the 4,692 images were rotationally and translationally aligned using the reference-free method (35). The aligned images were then classified into 150 clusters by the modified growing neural gas network method using a circular mask (37). Images in each cluster were averaged, and the averages with circular mask were used as new references. This cycle, from alignment to averaging, was repeated 25 times. The Euler angles of the class averages were automatically determined by the echo-correlated three-dimensional reconstruction method with simulated annealing (36) assuming C4 symmetry because of the tetrameric subunit stoichiometry of 3FSPP. These angles were used to calculate a preliminary three-dimensional density map by the simultaneous iterative reconstruction technique (38). The reprojections from the volume were employed as references for multi-reference alignment, and each image in the library was aligned and classified, providing improved cluster averages. From these averages, a new three-dimensional map was generated by the reconstruction method using simulated annealing without a three-dimensional reference. This cycle was repeated for three cycles. The density map was further refined by projection matching (39) followed by echo-correlated reconstruction. This cycle was repeated until convergence. Resolution was assessed without masking by dividing the data into two subsets and then calculated using the independent three-dimensional reconstructions of each, which were compared by Fourier shell correlation at the threshold of 0.5, using IMAGIC V.

Proteolytically Active SPP Polypeptides Form a Multimeric
Complex-To characterize the SPP molecule in an enzymatically active state, we analyzed endogenous SPP by BN-PAGE in the DDM-solubilized condition, in which SPP activity was preserved (31). Endogenous human SPP, which existed as a 45-kDa monomer and a heat-sensitive dimer (90 kDa) on SDS-PAGE analysis (40) (Fig. 1A), was detected as a single band at 200 kDa ( Fig. 1B), suggesting that active SPP forms a high molecular mass complex in the DDM-solubilized condition. Next we analyzed Drosophila SPP, which is composed of 389 amino acids (18). Endogenous dSPP polypeptide in S2 cells was detected as a single band of 40 kDa on SDS-PAGE without extensive boiling, indicating that dSPP did not form the SDS-resistant dimer (Fig.  1C). However, DDM-solubilized dSPP was migrated as a 180-kDa high molecular mass complex (Fig. 1D). Notably, endogenous dSPP was specifically co-immunoprecipitated with exogenously overexpressed dSPP, but not with other multispanning proteins, similarly to mammalian SPP (Fig. 1E, supplemental Fig. S1A) (22,40), suggesting that the ability of SPP to form a homo-oligomer is conserved beyond species irrespective of the formation of SDS-resistant dimer. In addition, DDM-solubilized SPPL2b, of which the mature form migrated at 90 kDa on SDS-PAGE (19), was detected at 400 kDa on BN-PAGE (supplemental Fig. S1, B and C). These data suggest that SPP protease family proteins maintain the high molecular weight complex structure in the DDM-solubilized condition.
To examine whether high molecular weight SPP complex is composed solely of SPP polypeptides or incorporates other components, we overexpressed 3FSPP ( Fig. 2A) by the recombinant baculovirus/Sf9 cell system. 3FSPP polypeptide, which migrated at 50 kDa as a monomeric form, was detected as a heat-sensitive, SDS-resistant 100-kDa dimer on SDS-PAGE similarly to that in mammalian cells (Fig. 2B). DDM-solubilized Sf9 cell lysates exhibited a proteolytic activity to cleave the synthetic SPP substrate, myc-Prl-PP-FLAG (31), suggesting that recombinant 3FSPP retained the proteolytic activity (see below). In fact, in agreement with previous results, photoaffinity probes based on transition state analog (Fig. 2C) as well as dipeptidic (DBZ-BpB3) (24) type inhibitors, all of which specifically targeted the endogenous SPP in mammalian cells (5,24,25), specifically labeled 3FSPP observed as a monomer band on SDS-PAGE. These data suggest that 3FSPP polypeptides overexpressed in Sf9 cells reconstituted the intramembrane-cleaving activity in a natural conformation.
We further purified the DDM-solubilized Sf9 cell lysates by affinity column using anti-FLAG antibody (Fig. 2D). Silver staining and immunoblot analysis of each fraction revealed that the major polypeptides in eluate fraction migrated at 50 and 100 kDa, corresponding to the monomer and the dimer of 3FSPP, respectively. Next we separated the purified 3FSPP by size-exclusion gel chromatography (SEC) (Figs. 2E and 3A and supplemental Fig. S2A). 3FSPP polypeptides were mainly detected at the fractions corresponding to ϳ250 -450 kDa. Silver staining revealed that these fractions contained exclusively 3FSPP polypeptides. Moreover, all samples containing 3FSPP showed the SPP activity to cleave myc-Prl-PP-FLAG in vitro, suggesting that the high molecular weight complexes were composed only of proteolytically active 3FSPP (Fig. 2F and supplemental Fig.  S2, A and B). The purified 3FSPP molecules were also subjected to BN-PAGE (Fig. 3B). DDM-solubilized 3FSPP was predominantly migrated as a 240-kDa high molecular mass complex similarly to endogenous SPP. In addition, 440-and 700-kDa complexes containing 3FSPP were detected. No additional major bands appeared between the 240-, 440-, and 700-kDa complexes, but three additional minor bands emerged below 240 kDa with an increased loading amount. Notably, these bands migrated as a monomer (60 kDa), dimer (120 kDa), and trimer (180 kDa) of 3FSPP, suggesting that the 240-kDa complex corresponds to a tetrameric complex of 3FSPP. Mutant

Three-dimensional Structure of SPP
3FSPP carrying the protease-inactive D219A mutation (6) was also detected as a 240-kDa complex on BN-PAGE, suggesting that SPP forms the high molecular mass complex irrespective of the enzymatic activity (supplemental Fig. S2C). These data suggest that active SPP was purified as a tetramer in the DDMsolubilized condition. Moreover, appearance of monomer and trimer species on BN-PAGE indicates that the tetrameric assembly is not a dimeric form of the SDS-resistant dimer. 440and 700-kDa complexes may represent oligomeric forms of the tetramer (i.e. octamer and dodecamer, respectively). The sizes of DDM-solubilized native SPP (monomer, 45 kDa; complex, 200 kDa), dSPP (monomer, 40 kDa; complex, 180 kDa), and tagged SPPL2b (monomer, 90 kDa; complex, 400 kDa) are also consistent with tetramer formation, supporting the notion that the formation of a tetrameric complex is a common characteristic of the SPP family proteins.
Electron Microscopy and Three-dimensional Reconstruction of SPP-We then negatively stained the purified 3FSPP particles with uranyl acetate and viewed them by electron microscopy. In the fractions 26/27 (corresponding to 250 kDa from standard proteins) separated by SEC (Fig. 3, A and B), variously shaped particles of a uniform size were observed (Fig. 3C). These particles were labeled by colloidal gold-conjugated anti-FLAG antibody, indicating the 3FSPP tetramer (Fig. 3D). The dimeric form of the particles, each of which also has the same dimensions, was observed in the 450-kDa fractions 20/21 of SEC (corresponding to 450 kDa from standards) (Fig. 3E), supporting the idea that the 440-kDa complex in BN-PAGE corresponds to the dimeric form of the 3FSPP tetramer. Most particles were rhomboid-or square-shaped with round corners (Fig.  3C). The variation in shape was interpreted to reflect different orientations of the same molecule on the grid. The squareshaped particles seemed to imply top views of the tetrameric form; the rhombuses would be side views. For three-dimensional reconstruction of the SPP molecule, image analysis was performed using our SPINNS program and IMAGIC V (33-37). The final reconstruction included 4,232 particles, 90.2% of all the selected images. Representative raw images are presented (Fig. 3F, first row), with their corresponding class averages (Fig. 3F, second row) and with surface representations and reprojections (Fig. 3F, third and fourth rows). Reprojections from the final volume are consistent with raw images and class averages, reflecting successful reconstruction from the original particle images. A plot of the Euler angles of the 117 adopted class averages (supplemental Fig. S3A) showed that SPP is almost randomly oriented on the grid surface. According to the Fourier shell correlation (FSC) function (41), 22 Å, if FSC Ͼ 0.5, was used as the resolution criterion (supplemental Fig. S3B).
For surface representation, the three-dimensional map was contoured at an isosurface containing a volume corresponding to 327 kDa: 182% of the tetrameric SPP mass (179 kDa) calculated from the amino acid composition. The additional volume seems attributable to glycosylation (4,12,22) and the attached lipids and detergents. The surface representation depicts a bullet-shaped molecule with a pointed bottom tip and a boat-tail top (Fig. 3G). Viewed from the top, SPP is a square with round corners: 100 ϫ 85 Å. The height estimated from the side views is 130 Å. The structure has four low density interior regions connected to exterior openings, which are capped by a plug-like structure at the top ( Fig. 3H  and supplemental Fig. S3C). Moreover, a cleft-like concave on the top of each rhomboid side connected to the large chamber was identified in the structure. Biochemical analyses revealed that SPP exposes N and C termini to the luminal and the cytosolic sides, respectively (42). To confirm the topology of the 3FSPP particle, purified 3FSPP was coincubated with anti-FLAG antibody and observed by electron microscopy. The particles were labeled with antibodies near the corners of the large domain (Fig. 3, D and I), suggesting that the boat-tail domain faces the luminal side and the tip locates at the cytosolic side. Taken together, single particle analysis revealed that the purified 3FSPP tetramer forms a bullet-shaped structure with an internal chamber.
N-terminal Region of SPP Is Responsible for Tetrameric Assembly-It was reported that the recombinant C-terminal region of SPP containing five transmembrane domains with the catalytic aspartates was able to reconstitute the substrate binding as well as proteolytic activity in vitro and was present as a monomer (43). We thus hypothesized that the remaining N-terminal half of SPP, which is the counterpart domain of the C terminus, is responsible for tetramer formation. To examine this, we first purified the N-terminal region of 3FSPP (3FSPP/ NT; truncated human SPP (1-191 amino acids) fused with 3ϫFLAG tag at its N terminus) overexpressed in Sf9 cells ( Fig.  4A and supplemental Fig. S4). In addition to a monomer (30 kDa), an SDS-resistant homodimer of 3FSPP/NT was detected similarly to the SPP holoprotein (Fig. 4B). Recombinant 3FSPP/NT polypeptides were also detected as 130-and 260-kDa complexes on BN-PAGE, corresponding to the tetramer and the dimeric tetramer, respectively (Fig. 4C). Purified 3FSPP/NT was also observed as a half-size particle of the 3FSPP holoprotein in negatively stained images (Fig. 4D). These data strongly suggested that the N-terminal region is a scaffold domain for the tetramer formation of SPP.
To examine the physiological relevance of the tetramer formation of SPP within cells, we overexpressed dSPP/NT, which encodes the truncated dSPP (1-200 amino acids) equivalent to human 3FSPP/NT (Fig. 5, A and B, and supplemental Fig. S4). Consistent with the putative scaffolding function observed in 3FSPP, overexpressed dSPP/NT was co-immunoprecipitated with endogenous dSPP (Fig. 5C). Next we analyzed the possible function of dSPP/NT using a chemical biology approach. SPP utilizes two functional sites for the intramembrane cleavage, i.e. the initial substrate binding site and the catalytic site, which are directly targeted by the helical peptide-and the transition state analog-type inhibitors, respectively (8,22,31). A previous report indicated that both sites were present in the recombinant C-terminal fragment of dSPP (43). To test whether the dSPP/NT polypeptide harbors enzymatically functional sites, we performed a photoaffinity labeling experiment using the helical peptide-and the transition state analog-based photoprobes (pep.11-Bt and 31C-Bpa, respectively) (25,44,45). Consistent with the previous results, both wild-type dSPP and dSPP/ D228A were labeled by pep.11-Bt, whereas 31C-Bpa was bound only to wild-type dSPP, suggesting that the initial substrate binding site was formed irrespective of the endoproteolytic activity of dSPP (Fig. 5D). However, no labeling of dSPP/NT was observed by either of the photoprobes, indicating that dSPP/NT does not harbor a functional site for the enzymatic activity. We then examined the effect of dSPP/NT overexpression on the proteolytic activity in cells. dSPP activity was detected by the transcriptional activation of the ERSE-driven firefly luciferase (26), which was mediated by specific intramembrane cleavage of the recombinant SPP substrate (SPP sub ) (22). We found that the coexpression of dSPP with ERSE reporter and SPP sub in S2 cells increased luciferase activity, which was decreased by the coexpression of an aspartate mutant dSPP, suggesting that the reporter assay is applicable to dSPP in S2 cells similarly to mammalian cells. Intriguingly, the overexpression of dSPP/NT inhibited FIGURE 4. The N-terminal region of SPP is sufficient for tetrameric assembly. A, schematic depiction of 3FSPP/NT. B, immunoblot analysis of affinitypurified 3FSPP/NT by anti-FLAG antibody. The dimer and monomer forms of 3FSPP/NT are indicated by the arrowhead and arrow, respectively. C, affinitypurified 3FSPP/NT was analyzed using BN-PAGE in the DDM-solubilized condition. 3FSPP/NT proteins were mainly detected as tetramers (arrow) and dimeric tetramers (arrowhead). D, electron microscopic observation of negatively stained affinity-purified 3FSPP/NT as compared with 3FSPP. The scale bar represents 100 Å.

Three-dimensional Structure of SPP
the proteolytic activity of endogenous dSPP in a similar manner to that observed by the expression of aspartate mutant dSPP, indicating that dSPP/NT functions as a dominant negative mutant (Fig. 5E). Moreover, the amount of the endogenous dSPP tetramer was reduced in cells expressing dSPP/NT (Fig. 5F). Considering that dSPP/NT lacked any mechanistically functional sites for enzymatic activity, interaction of dSPP/NT with the endogenous dSPP disrupted the tetrameric assembly, thereby causing the dominant negative effect. Collectively, these data suggest that the N-terminal region of SPP is responsible for the tetrameric assembly, which might be the prerequisite for the intramembrane cleaving activity of SPP.

DISCUSSION
In this study, we analyzed the structure and biochemical character of SPP using in vitro as well as cell-based assays. We found that SPP forms a bullet-shaped tetramer with a large interior chamber. The tetrameric assembly was conserved among the SPP and SPP-like protease (SPPL) family proteins and was mediated by its N-terminal region. Overexpressed SPP/NT was incorporated into the SPP complex and inhibited the enzymatic activity in living cells, implying that the tetramer formation is the prerequisite for the proteolytic activity of SPP.
It was reported that human SPP proteins formed a SDS-resistant dimer (40). In contrast to mammalian SPP proteins, dSPP was solely detected as a monomer in SDS-PAGE (Fig. 1B) (43). However, DDM-solubilized human SPP, SPPL2b, and dSPP polypeptides were mainly detected as a tetramer complex on BN-PAGE. Thus, the tetramer formation is mediated by a common molecular mechanism, and the binding mode of the tetramer should be distinct from the formation of the SDSresistant dimer observed in human SPP or SPPL2b. SDS-resistant dimer formation is not a critical mechanistic feature; rather, it appears to be an artificial phenomenon in human SPP caused by SDS. We also found that the N-terminal half is sufficient for the tetrameric assembly of SPP, although the precise mode of interaction still remains unknown. However, SPP/ SPPL proteins were never cofractionated with other member proteins of the SPP family (19,31), suggesting a specific mechanism of recognition and interaction behind homo-oligomerization. Very recently, 200-, 400-, and 600-kDa complexes containing SPP and its substrates were reported, in accordance with our results (46). EM showed a dimer form of the tetramer in SEC fractions 20 and 21, suggesting that the tetramer is an essential and minimal form of functional SPP. Further analyses will be needed to identify the critical domain(s) for the tetramerization.
A series of x-ray crystallographic studies revealed that the serine-and metalloprotease-types of I-CLiPs, e.g. rhomboid and S2P, respectively, harbor the active site residues within the hydrophilic cleft in the lipid bilayer (47). Moreover, we have biochemically shown that PS has the catalytic pore structure (48), suggesting that the hydrophilic interior structure of a catalytic site is a common feature essential to the intramembrane cleaving mechanism of I-CLiPs. Here, we analyzed the structure of SPP using single particle analysis and found that the SPP tetramer has a bullet-like structure with low density internal regions, to which the concaves near the boat-tail domain were connected. The boat-tail domain with its plug-like structure was predominantly labeled by an antibody targeting the N terminus, suggesting that SPP also has a hydrophilic chamber accessible from the luminal side (Fig. 6A). Intriguingly, the plug-like structure near the catalytic site was also predicted in the cytosolic side of PS (48). Considering the inverse topology of SPP as compared with PS (42), the four chambers within the SPP tetramer and the plug-like structures at the predicted luminal side represent the hydrophilic catalytic sites of SPP within the lipid bilayer, similarly to those of PS. The concaves connected to the chambers may represent the substrate entry sites. Furthermore, considering the previous study reporting that the C-terminal region of SPP restored proteolytic function in the monomeric state (43), we speculate that the N-terminal region functions as a scaffold for tetramer formation located at the center of the bullet-like structure. This may explain why a single anti-FLAG antibody bound to one tetramer despite high binding efficacy. Collectively, these data suggest that SPP forms a bullet-like tetramer with its N termini at the center, whereas the enzymatically active C terminus is located as an outer ring (Fig.  6B). Nevertheless, further fine structural analyses, e.g. single particle reconstruction using cryo-EM or x-ray crystallography, will be needed to clarify the precise structure and function relationships of SPP.
Overexpression of dSPP/NT inhibited the activity of endogenous dSPP similarly to that of the catalytic site mutant. In accordance with previous studies of the C-terminal region of SPP (43), our chemical biology approach revealed that dSPP/NT contains neither the initial substrate binding site nor the catalytic site, suggesting that the inhibitory effect of dSPP/NT is independent of the proteolytic machinery (i.e. defects in catalytic function or capturing of substrates). In addition, the possibility of a nonspecific hydrophobic interaction of the substrate with dSPP/NT was excluded because nonspecific labeling of dSPP/NT with photoprobes, including a highly hydrophobic helical peptide, was not detected. Rather, we found that the amount of dSPP tetramer was reduced in cells expressing dSPP/NT, which is capable of binding to the holoprotein, suggesting that the interaction of dSPP/NT affected the tetramer formation of the endogenous dSPP (Fig. 6B). These results are totally different from those observed with PS, which is a homologous protease; exogenous N-or C-terminal fragments (NTF or CTF, respectively) of PS failed to be incorporated into a functional complex, nor did it affect the levels or activity of endogenous ␥-secretase (49,50). Moreover, we failed to overexpress the C-terminal region of dSPP with or without dSPP/NT (data not shown), whereas the coexpression of PS NTF and CTF reconstituted the functional ␥-secretase complex (51). Thus, the tetrameric assembly of SPP with its N-ter-

Three-dimensional Structure of SPP
minal region is a unique feature of the SPP/SPPL family I-CLiPs. Oligomer formation is the prerequisite for the function of several channels, receptors, and transporters. In particular, the tetramer formation of aquaporin is critical to its water permeability and ion selectivity (52)(53)(54), whereas each aquaporin subunit has a water channel in the center. We speculate that the interaction of dSPP/NT resulted in a conformational change of each subunit that affects the formation and/or the stability of the dSPP tetramer, thereby causing the dominant negative effect (Fig. 6B). These results also imply that the N-terminal region is the scaffold domain for the formation of the enzymatically active SPP tetramer.
In sum, we have revealed that SPP forms a bullet-shaped tetramer with its N-terminal region in the proteolytically active state. However, we were not able to fully characterize the structure of SPP in the membrane-embedded state. Further fine structural analyses, e.g. reconstruction of SPP in the presence of lipids and x-ray analysis of its Type-I three-dimensional crystals composed of two-dimensional crystal layers in lipid, should eventually identify the mechanism by which the SPP complex recognizes the substrates, which in turn leads to the entrance of the substrates into the active site.