Signal peptide peptidase forms a homodimer that is labeled by an active site-directed gamma-secretase inhibitor.

Presenilin (PS) is the presumptive catalytic component of the intramembrane aspartyl protease gamma-secretase complex. Recently a family of presenilin homologs was identified. One member of this family, signal peptide peptidase (SPP), has been shown to be a protease, which supports the hypothesis that PS and presenilin homologs are related intramembrane-cleaving aspartyl proteases. SPP has been reported as a glycoprotein of approximately 45 kDa. Our initial characterization of SPP isolated from human brain and cell lines demonstrated that SPP is primarily present as an SDS-stable approximately 95-kDa protein on Western blots. Upon heating or treatment of this approximately 95-kDa SPP band with acid, a approximately 45-kDa band could be resolved. Co-purification of two different epitope-tagged forms of SPP from a stably transfected cell line expressing both tagged versions demonstrated that the approximately 95-kDa band is a homodimer of SPP. Pulse-chase metabolic labeling studies demonstrated that the SPP homodimer assembles rapidly and is metabolically stable. In a glycerol velocity gradient, SPP sedimented from approximately 100-200 kDa. Significantly the SPP homodimer was specifically labeled by an active site-directed photoaffinity probe (III-63) for PS, indicating that the active sites of SPP and PS/gamma-secretase are similar and providing strong evidence that the homodimer is functionally active. Collectively these data suggest that SPP exists in vivo as a functional dimer.

Presenilin (PS) is the presumptive catalytic component of the intramembrane aspartyl protease ␥-secretase complex. Recently a family of presenilin homologs was identified. One member of this family, signal peptide peptidase (SPP), has been shown to be a protease, which supports the hypothesis that PS and presenilin homologs are related intramembrane-cleaving aspartyl proteases. SPP has been reported as a glycoprotein of ϳ45 kDa. Our initial characterization of SPP isolated from human brain and cell lines demonstrated that SPP is primarily present as an SDS-stable ϳ95-kDa protein on Western blots. Upon heating or treatment of this ϳ95-kDa SPP band with acid, a ϳ45-kDa band could be resolved. Co-purification of two different epitopetagged forms of SPP from a stably transfected cell line expressing both tagged versions demonstrated that the ϳ95-kDa band is a homodimer of SPP. Pulse-chase metabolic labeling studies demonstrated that the SPP homodimer assembles rapidly and is metabolically stable. In a glycerol velocity gradient, SPP sedimented from ϳ100 -200 kDa. Significantly the SPP homodimer was specifically labeled by an active site-directed photoaffinity probe (III-63) for PS, indicating that the active sites of SPP and PS/␥-secretase are similar and providing strong evidence that the homodimer is functionally active. Collectively these data suggest that SPP exists in vivo as a functional dimer.
Presenilins (PSs) 1 were first identified through genetic studies demonstrating that mutations in them caused Alzheimer's disease (1,2). PSs are integral membrane proteins with multiple membrane-spanning domains that are thought to be the catalytic components of a high molecular mass complex referred to as ␥-secretase that carries out intramembrane cleav-age of multiple type I integral membrane proteins (3,4). Because ␥-secretase catalyzes the final cleavage that releases amyloid ␤ from its precursor protein, it is a target of emerging Alzheimer's disease therapeutic compounds (3). Recent studies have provided a more complete picture of the ␥-secretase complex. In addition to PS1 or PS2, three other proteins are required for activity: Nicastrin, APH-1, and PEN-2 (5)(6)(7)(8)(9). Both inhibitor studies and mutational analysis suggest that PSs are intramembrane cleaving aspartyl proteases (aspartyl I-CLiPs) (4). Active site-directed aspartyl protease inhibitors as well as other ␥-secretase inhibitors have been shown to bind PS (10 -13). Moreover mutation of either of two conserved aspartates present in adjacent transmembrane domain results in dominant negative PSs that inhibit ␥-secretase activity (14).
Despite these recent advances in the understanding of ␥-secretase, the multiprotein complex mediating this activity poses significant problems for studies that would inevitably lead to a more definitive structural and mechanistic understanding of this protease. Although the identification of additional members of the ␥-secretase complex has enabled reconstitution of ␥-secretase activity in a heterologous system (15), to date the active complex has not been purified to homogeneity.
Shortly after the identification of human PSs, close homologs were recognized in plants, invertebrates, and vertebrates. Some of these homologs such as those found in Caenorhabditis elegans and Drosophila have been extensively studied and shown to function in high molecular mass complexes like human PSs (for a review, see Ref. 16). More recently, other proteins with less obvious homology to PSs have been recognized by data base searching (17)(18)(19). These proteins have been referred to by various names. Herein they will be referred to as presenilin homologs/signal peptide peptidase (SPP). All of these proteins are predicted to be integral membrane proteins with multiple membrane-spanning regions and contain both the conserved transmembrane aspartates and the PAL motif near the COOH terminus (16,20).
Shortly after its initial in silico identification, one of the presenilin homologs/SPP was identified as an aspartyl I-CliP (17). This protein has been termed signal peptide peptidase as it has been shown to carry out the intramembrane cleavage of signal peptides of major histocompatibility complex class I molecules and hepatitis C virus polyprotein following the initial cleavage of these type II membrane proteins by signal peptidase. SPP-mediated cleavage of major histocompatibility complex class I appears to play an important role in normal immune surveillance as HLA E epitopes are produced from the signal peptide of major histocompatibility complex class I by SPP cleavage (21). These epitopes are presented to natural killer cells at the cell surface. Such presentation is thought to indicate that the probed cell is healthy (22). SPP cleavage of an * This work was supported by NINDS, National Institutes of Health Grants NS39072 (to T. E. G.), NS44734 (to A. C. N.), and NS41355 (to M. S. W.) and by a Lefler postdoctoral fellowship (to A. Y. K.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
SPP was originally identified using an inhibitor labeling approach with a compound termed TBL 4 K, which is a photoaffinity probe, based on an inhibitor of SPP activity, (Z-LL) 2ketone (17). In contrast to human PS1, human SPP activity can be reconstituted in yeast without co-expression of other protein cofactors. Moreover mutation of the second of the conserved transmembrane aspartates (Asp-265) does not alter labeling with TBL 4 K but does block catalytic activity. More recent studies demonstrate that some ␥-secretase inhibitors can inhibit SPP activity and that relatively high concentrations of (Z-LL) 2ketone SPP inhibitors can inhibit ␥-secretase activity (23,24). Although the topography of SPP has not been definitively established, some evidence suggests that the transmembrane regions containing the proposed catalytic aspartates have an inverted topology relative to PS. If true, this topology would be consistent with the cleavage of type II membrane proteins by SPP and type I membrane proteins by PS (17).
Collectively these bioinformatic and experimental studies of SPP and PS1 support the notion that these proteins are members of a family of biomedically important aspartyl I-CliPs. Because both enzymes are potential targets for therapeutic intervention, PS for the treatment of Alzheimer's disease and SPP for anti-hepatitis C virus therapy, it will be important to identify compounds and conditions for the selective inhibition of each protease.
In the present study, we set out to characterize a ϳ95-kDa species of SPP. Intact SPP is theoretically a ϳ42-kDa protein that, when N-glycosylated, has been reported to migrate at ϳ45 kDa (17). Here we demonstrated that the ϳ95-kDa species is a homodimer and that the vast majority of SPP in mammalian cells and brain exists in this dimeric form. This dimer was SDS-stable but was partially heat-and acid-labile. Significantly a photoactivable active site-directed ␥-secretase inhibitor specifically labeled the dimeric form of SPP. These results suggest that SPP functions as a homodimer and provide additional evidence that the active sites of SPP and PS are structurally similar.

MATERIALS AND METHODS
SPP Constructs-Full-length SPP was cloned by amplifying a 1134-bp sequence from a human brain cDNA library. The wild type SPP cDNA contains the entire coding sequence of SPP and 17 bases of 5Ј untranslated sequence and 10 bases of 3Ј untranslated sequence and was cloned into HindIII (5Ј) and XbaI (3Ј) sites of the pAG3 expression vector (25). Full-length SPP was also cloned into the pcDNA6-V5-his vector (Invitrogen) to generate a carboxyl-terminally V5 ϩ His 6 epitopetagged SPP (SPP-CT V5his) and into pFLAG-CMV-2 (Sigma) to generate an amino-terminally FLAG epitope-tagged SPP (SPP-NT FLAG). In SPP-CT V5his, SPP was cloned into HindIII (5Ј) and XbaI (3Ј) of cDNA6-V5-his vector such that a vector-derived six-amino acid spacer (SRGPFE) exists between the end of the SPP coding sequence and the V5 tag. In SPP-NT FLAG, full-length SPP was cloned into the HindIII (5Ј) and XhoI (3Ј) sites of pFLAG-CMV-2 such that a vector-derived three-amino acid spacer (LLA) separates the FLAG tag from the amino terminus of SPP. All constructs were verified by sequencing.
Generation of Pooled Stable Lines-Chinese hamster ovary, human embryonic kidney 293 (HEK), or human neuroglioma (H4) cells were transfected with 2 g of plasmid DNA preincubated with 8 l of Fu-GENE 6 transfection reagent (Roche Applied Science) in serum-free OptiMEM (Invitrogen) overnight. Media were then replaced with either Ham's F-12 (Chinese hamster ovary cells) or Dulbecco's modified Eagle's medium (HEK cells and H4 cells) supplemented with 10% fetal bovine serum (Hyclone) and 2.5 g/ml blasticidin S (Calbiochem). Following selection, greater than 500 colonies were observed in each pooled stable line. Expression levels were monitored periodically by immunoblotting throughout the course of the experiments and remained stable through multiple passages. The HEK cell line that co-expresses the SPP-CT V5his and SPP-NT FLAG constructs was made by mixing 2 g of SPP-NT FLAG and 0.2 g of the SPP-CT V5his DNA in the transfection mixture. Stable cell lines were selected by selecting for blasticidin S encoded by pCDNA-6-SPP CT V5his. Cell lines were maintained at 37°C under 5% CO 2 .
Antibodies, Western Blotting, and Immunostaining-Anti-peptide antisera were raised in rabbits to the amino-(anti-SPPnt, residues 1-20) and carboxyl (anti-SPPct, residues 358 -377)-terminal domains of human SPP (Covance). Peptides corresponding to these domains of SPP were synthesized and coupled to keyhole limpet hemocyanin prior to immunization. Lysates were prepared by lysis in either 2% CHAPSO in 50 mM HEPES (CHAPSO) and 1ϫ Complete PI or 1% Triton X-100 in Tris-buffered saline with 1ϫ Complete PI (TX-100). Except where indicated in the figure legends, sample loading buffer was added to each sample such that the sample contained a final concentration of 10% glycerol, 2% SDS, and 2.5% ␤-mercaptoethanol. Samples were run on 10 -20% Tris-HCl polyacrylamide gels (Bio-Rad) and then transblotted to Immobilon-P membranes (Millipore). Blots were probed with rabbit anti-SPPct at a 1:1000 dilution to detect the carboxyl-terminal 20 amino acids of SPP or rabbit anti-SPPnt at a 1:1000 dilution to detect the amino-terminal 20 amino acids of SPP. PS1 was detected with a 1:1000 mixture of antibodies that detect PS1 CTF (PS490, gifts of E. Koo) and PS1 amino terminus (852B3). Anti-V5 antibody (Invitrogen) and anti-FLAG (Sigma) were used at a 1:1000 dilution. Peptide preabsorption was performed by adding excess peptide dissolved in water and incubating it for 1 h at 37°C. The preabsorbed anti-serum was then centrifuged at 20,800 ϫ g for 2 min.
Trichloroacetic Acid Precipitation-Trichloroacetic acid precipitation was performed as described previously (17). Briefly 10% trichloroacetic acid was added to sucrose gradient fractionation samples. Samples were then centrifuged at 20,800 ϫ g for 2 min, and the supernatant was discarded. Pellets were washed twice in acetone and then allowed to dry. Pellets were resuspended in 0.5% CHAPSO, 50 mM HEPES, 150 mM NaCl, and 1ϫ PI.
Immunocytochemistry-Immunocytochemistry was carried out as described previously (26). Briefly HEK cells coexpressing SPP-CT V5his and SPP-NT FLAG were plated on 4-well chamber slides. Cells were fixed in 2% paraformaldehyde in phosphate-buffered saline and permeabilized in 0.05% saponin. Anti-V5 rabbit polyclonal antibody (Sigma) was used at a 1:500 dilution, and anti-FLAG mouse monoclonal (Sigma) was used at a 1:1000 dilution. Secondary goat anti-rabbit IgG (568) and goat anti-mouse IgG (488) were used sequentially at a 1:2000 dilution. Imaging was performed on a confocal Olympus microscope with Fluoview software system.
Ni 2ϩ Affinity Purification-HEK cells (one to two confluent 150-mm plates) stably expressing the SPP-CT V5his construct were lysed in CHAPSO lysis buffer with 1ϫ EDTA-free PI. Lysates were centrifuged at 3220 ϫ g for 5 min and the supernatant was diluted to 1% CHAPSO, 50 mM HEPES, and 20 mM imidazole and rocked with 100 l of nickel NTA slurry overnight at 4°C. Nickel NTA beads (Qiagen) were washed two times in 50 bed volumes of cold 50 mM HEPES, 1% CHAPSO, 150 mM NaCl, 20 mM imidazole, and 1ϫ EDTA-free PI. Two consecutive elutions were collected in 50 mM HEPES, 1% CHAPSO, 150 mM NaCl, and 200 mM imidazole. All buffers were at pH 7 and had 1ϫ EDTA-free PI.
Gradient Fractionation-Sucrose gradients were run as described previously (27). Glycerol velocity gradients were run as described previously (28). Briefly HEK wild type (wt) or HEK cells stably expressing the SPP-CT V5his construct (five confluent 150-mm plates) were lysed in either CHAPSO or TX-100. Lysates were spun at 3220 ϫ g to remove nuclei, insoluble material, and cellular debris, and 1 ml of supernatant was loaded on to the top of an 11-ml 10 -40% linear glycerol gradient with either 0.5% CHAPSO, 150 mM NaCl, and 25 mM HEPES or 0.1% TX-100 in Tris-buffered saline. The samples were then centrifuged in an SW 41 rotor for 15 h at 110,000 ϫ g at 4°C. One-milliliter fractions were collected from the top and analyzed by SDS-PAGE for SPP or PS1 proteins. Control gradients were prepared identically but included a combination of commercially available, characterized, molecular mass standards (Serva).
Metabolic Labeling and Immunoprecipitation of SPP-HEK wt or SPP-CT V5his stably expressing cells were labeled for 2 h with 1 mCi/ml [ 35 S]methionine. SPP was immunoprecipitated with either anti-SPPnt or anti-V5 antibodies from TX-100 lysates. Each analysis was performed in duplicate. Immunoprecipitated proteins were separated on 10 -20% Tris-HCl gels (Bio-Rad). Phosphorimaging analysis of the dried gels was performed. ImageQuant was used to quantitate the SPP in each sample and determine the half-life of SPP.
Covalent Labeling of SPP by Photoactivable and Biotinylated Active Site-directed ␥-Secretase Inhibitor-Photoprobe III-63 and analog III-31-C were synthesized as described previously (29). The photolabeling was performed as described previously (24). Briefly 500 nM compound III-63 was incubated with CHAPSO lysates for 1 h in the absence or presence of a parent compound, III-31-C (20 M), or an SPP inhibitor, (Z-LL) 2 -ketone (1 M). Samples were irradiated for 45 min at 350 nm. Biotinylated proteins were precipitated with immobilized streptavidin and detected by SDS-PAGE using anti-V5 antibody for SPP-CT V5hisexpressing cells or anti-SPPct antibody for the endogenous SPP in wt cells.

RESULTS
Antibodies to SPP Recognize an ϳ95-kDa Band-Anti-peptide antisera were raised in rabbits to the amino (anti-SPPnt, residues 1-20) and carboxyl (anti-SPPct, residues 358 -377) terminal domains of human SPP. Both of these antibodies recognized a band of ϳ95 kDa in human brain and in human cell lines indicating that this band is likely to contain fulllength SPP (Fig. 1a). In HEK and H4 cells, a second band of ϳ45 kDa was also recognized by both anti-SPP antibodies (Fig.  1, a and d). In human brain, this band was not detected by either antibody even upon longer exposure (not shown). Preabsorption of the antibody with peptide completely blocked detec-tion of both the ϳ95and ϳ45-kDa bands (Fig. 1a), and neither band was present in preimmune sera (Fig. 1d). Variability in the relative amounts of the two bands could be seen depending on how the sample was handled post-lysis. For example, samples that had been manipulated for long periods of time (Ni 2ϩ affinity purification or overnight fractionation) or heated excessively had more of the ϳ45-kDa band relative to the ϳ95-kDa band (e.g. Fig. 1d). Although in most lysates the ϳ95-kDa band was always present, lower and more variable levels of the ϳ45-kDa band were seen. The ϳ95-kDa species was observed from lysates of cells that stably overexpress SPP with a V5 and His 6 epitope tag at its COOH terminus (SPP-CT V5his) or a FLAG epitope tag at the NH 2 terminus (SPP-NT FLAG) (Fig.  1b). The COOH-terminal V5his tag on SPP was placed at the very COOH terminus after the putative KKXX endoplasmic reticulum retrieval signal. In some cases, placement of an epitope tag beyond the endoplasmic reticulum retrieval signal can alter the fate of the protein. With SPP, we found no evidence for any effect of the epitope tags on the characteristics of the protein that we analyzed in this study. Except for expected FIG. 1. Antibodies to the NH 2 and COOH termini of SPP specifically detect ϳ95and ϳ45-kDa bands. a, human brain tissue, H4 cell, and HEK cell lysates were lysed in CHAPSO, and 20 g was loaded in each gel lane. The Western blot was probed with antisera to the COOH terminus (anti-SPPct) and NH 2 terminus (anti-SPPnt) of SPP as well as peptide-preabsorbed antisera. Panels labeled Human Brain, H4, and the first HEK are all 15-s exposures, and the last set of HEK panels are 60-s exposures. A ϳ95-kDa band is seen in all samples, and upon longer exposure a minor, but specific, ϳ45-kDa band is seen in the HEK cells. b, a ϳ95-kDa band is increased by stable overexpression of epitope-tagged SPP constructs. HEK wt cells, SPP-CT V5his-overexpressing HEK cells, and SPP-NT FLAG-overexpressing HEK cells were lysed in CHAPSO and then separated by SDS-PAGE. Blots were probed with anti-SPPct, anti-V5, or anti-FLAG antibodies. A small upward shift in the ϳ95-kDa band is observed in the overexpressing cell lines probed with anti-V5 and anti-FLAG due to the epitope tag. A minor ϳ45-kDa species detected by anti-V5 or anti-FLAG antibodies could be seen upon longer exposure (not shown). c, SPP-CT V5his can be quantitatively recovered from cell lysates by Ni 2ϩ affinity purification following lysis in CHAPSO (lane 1, original lysate; lane 2, flow-through). Nickel NTA was then washed twice (not shown). Elution was performed twice (lanes 3 and 4). Both the ϳ95and the ϳ45 kDa-forms of SPP are detected in the Ni 2ϩ affinity purification. d, H4 cell lysates were fractionated on a sucrose gradient, and fractions 4 and 5 were collected and probed with either anti-SPPct or anti-SPPnt antibodies. Preimmune serum (used at the same concentration, 1:1000) demonstrates that both the ϳ45and the ϳ95-kDa bands are specifically recognized by both antibodies. e, SPP-CT V5his was Ni 2ϩ affinity-purified following lysis in CHAPSO (lane 1, original lysate; lane 2, flow-through). Nickel NTA was then washed three times (not shown), and two elutions were collected (shown in lanes 3 and 4). The ϳ95-kDa form of SPP that is Ni 2ϩ affinity-purified is recognized by anti-V5 antibody as well as both the anti-SPPct and anti-SPPnt antibodies. There is a significant portion of non-epitope-tagged SPP that does not Ni 2ϩ affinity purify (lane 2) as detected by anti-SPPnt and anti-SPPct. The non-purified SPP runs at a slightly lower molecular mass relative to the epitope-tagged purified SPP (lane 2 versus 4).
shifts in molecular mass, identical results were obtained with COOH-terminally tagged SPP, NH 2 -terminally tagged SPP, and endogenous SPP. Upon longer exposure of the blots, a minor ϳ45-kDa band was also detected by anti-V5 and anti-FLAG antibodies in immunoblots of SPP-CT V5his-transfected HEK or SPP-NT FLAG-transfected HEK cells (not shown). Ni 2ϩ affinity purification of SPP-CT V5his from HEK cells stably expressing this protein resulted in purification of a ϳ45and a ϳ95-kDa band (Fig. 1c). Both ϳ45and ϳ95-kDa bands were detected by the anti-SPPnt and anti-SPPct antibodies in SPP lysates that had a higher relative proportion of the ϳ45-kDa species than typically seen; nevertheless these results again demonstrated that each band contained full-length SPP (Fig.  1d). The Ni 2ϩ affinity-purified ϳ95-kDa band was recognized by both anti-SPPnt and anti-SPPct antibodies demonstrating that the antibodies were recognizing both tagged and untagged versions of SPP (Fig. 1e). Notably the flow-through of the Ni 2ϩ affinity purification had a significant amount of untagged SPP, as detected by anti-SPPnt and anti-SPPct antibodies, that was not purified and ran at a slightly lower molecular mass relative to the eluted SPP-CT V5his (Fig. 1e, lane 2 versus lanes 3 and 4).
SPP Is an SDS-stable Dimer-Based on these immunoblotting results, we postulated that the ϳ95-kDa form of SPP is either an SDS-stable complex with another protein or a homodimer. To investigate the relationship between the two forms we conducted several experiments. First we attempted to dissociate the ϳ95-kDa form using various detergents for lysis. Lysis of HEK cells that overexpress SPP-CT V5his in 1% Triton X-100, 2% CHAPSO, 0.1% SDS, 1ϫ radioimmune precipitation assay buffer, or 6 M urea did not significantly alter the relative levels of the ϳ95and ϳ45-kDa bands detected after immunoblotting with anti-V5 antibody (not shown). In all instances, the ϳ95-kDa band was the predominant species. To obtain partially purified SPP, we performed sucrose gradient fractionation and collected the buoyant fractions 4 and 5, which contained detergent-resistant membranes and associated proteins. We reported previously that these detergent-resistant membranes contain ␥-secretase activity (27). Like PS1, SPP was enriched in these detergent-resistant membranes as well (not shown). Using the detergent-resistant membranes, we performed a number of experiments to determine whether we could disassociate the ϳ95-kDa band into the smaller 45-kDa band. Heating the sample without loading buffer did very little to the ratio of the 45-to 95-kDa bands (Fig. 2a, lane 2). However, heating the sample at 65°C for 20 min after loading buffer (SDS and ␤-mercaptoethanol) was added made a large difference in the amount of the 45-kDa band relative to the 95-kDa band. Additionally trichloroacetic acid precipitation of the sample generated large quantities of the 45-kDa species (Fig. 2a, lane 4). The combination of trichloroacetic acid precipitation and heating at 65°C for 20 min after loading buffer was added caused all the ϳ95-kDa species to be converted to the ϳ45-kDa species (Fig. 2a, lane 5). These are the same conditions (trichloroacetic acid precipitation and heated Western samples) that Weihofen et al. (17) used when SPP was initially characterized as a ϳ45-kDa protein. Integral membrane proteins with multiple transmembrane domains are often difficult to recover if cell lysates are heated excessively prior to SDS-PAGE; therefore, in the experiments shown previously, the lysates were only heated at 37°C for 10 min prior to gel loading. To further determine whether heating alters the relative levels of the two bands observed for SPP, we altered the incubation temperature of TX-100 lysates prior to PAGE. These results showed that as SPP-CT V5his-overexpressing HEK cell lysates were heated from 37 to 65°C for 20 min, the amount of the ϳ95-kDa species decreased, and the ϳ45-kDa species increased (Fig. 2b). Above 65°C the amount of both species decreased, and high molecular mass aggregates became predominant (Fig. 2b). Lysates heated above 85°C contained primarily an aggregated species that barely entered the gel at all (not shown). Similar results were obtained for CHAPSOlysed SPP-CT V5his-overexpressing HEK cells, TX-100-lysed SPP-NT FLAG-overexpressing HEK cells, and endogenous SPP in HEK cells (not shown).
To confirm that the ϳ45-kDa species was a product of the heated ϳ95-kDa species, CHAPSO lysates from HEK cells stably overexpressing SPP-CT V5his were separated by SDS-PAGE, and a gel slice in the 90 -100-kDa mass range was FIG. 2. Heating and trichloroacetic acid precipitation dissociate the SDS-stable ϳ95-kDa SPP to a ϳ45-kDa form. a, semipurified (detergent-resistant membrane preparation) SPP (lane 1) can be converted from the ϳ95-kDa species to the ϳ45-kDa species under a variety of conditions. Samples heated at 65°C for 20 min with loading buffer (lane 3) appear to convert more of the ϳ95-kDa species to the ϳ45-kDa species than samples that are heated without loading buffer (lane 2). Trichloroacetic acid precipitation of the semipurified SPP (lane 4) and trichloroacetic acid precipitation followed by heating at 65°C for 20 min (lane 5) converts nearly all of the ϳ95-kDa species to the ϳ45-kDa species. b, heating SPP-CT V5his-overexpressing HEK cell lysates alters the ratio of ϳ95-(**) to ϳ45-kDa (*) bands. The lysate from a single 150-mm plate was divided into six samples, which were incubated at various temperatures for 20 min, separated by PAGE, and probed with an anti-V5 antibody. c, the ϳ95-kDa SPP band is a partially heat-labile, SDS-stable protein complex. SPP-CT V5his protein migrates as an SDS-stable ϳ95-kDa complex (**, lane 1) that partially disassociates upon heating at 65°C for 20 min to a ϳ45-kDa band. The ϳ45-kDa band contains full-length SPP as it is recognized by antibodies to both the NH 2 and COOH termini of SPP (*, lane 2, only anti-V5 labeling shown). The ϳ95-kDa band seen in lane 1 can be excised from the gel and dissociated by heating at 65°C into the ϳ45-kDa band (lanes 3 and 4). Temp, temperature. excised, crushed in gel loading buffer, and either heated at 65°C or 37°C for 20 min prior to additional SDS-PAGE. Heating to 65°C for 20 min shifted the majority of the ϳ95-kDa band to the ϳ45-kDa band (Fig. 2c, lanes 3 and 4). Together these data indicate that the ϳ95-kDa species is indeed a detergent-stable, partially heat-labile protein complex that contains SPP.
To determine whether this complex is a homodimer or a complex of SPP and another protein, we generated a pooled stable HEK cell line expressing both SPP-CT V5his and SPP-NT FLAG. Immunocytochemistry demonstrated that the majority of cells expressed both tagged SPP proteins (Fig. 3a) and that the proteins co-localized. Stably transfected HEK lines expressing only the SPP-NT FLAG or the SPP-CT V5his constructs were generated and used as controls. For these studies, cells were lysed in a zwitterionic detergent, CHAPSO, which preserves ␥-secretase activity and is almost identical to CHAPS, which has been shown to maintain SPP activity (23). Ni 2ϩ affinity purification of the SPP-CT V5his protein from the SPP-CT V5his⅐SPP-NT FLAG-co-expressing HEK cells resulted in enrichment of the ϳ95-kDa SPP-CT V5his and detection of some ϳ95-kDa SPP-NT FLAG, providing strong evidence for the existence of a one to one complex of SPP-CT V5his and SPP-NT FLAG (Fig. 3b, anti-FLAG, lane 3). In these studies, the co-purifying ϳ95-kDa SPP-NT FLAG migrated at a slightly higher molecular mass than the majority of the SPP-NT FLAG present in the original lysate and flow-through (Fig. 3b, lanes 1   and 2, respectively). The disparity in dimer molecular masses observed is attributed to the presence of the epitope tags. Dimers containing a single SPP-CT V5his or SPP-NT FLAG protein and a wild type SPP protein are ϳ1.5-3 kDa larger than dimers of wild type SPP alone, and the dimers with two epitope-tagged constructs are ϳ3-6 kDa larger. Thus, a Ni 2ϩ affinity-purified SPP dimer that is anti-FLAG-positive from the SPP-CT V5his⅐SPP-NT FLAG-co-expressing cell line will contain two epitope tags and runs at a slightly higher molecular mass relative to a SPP-NT FLAG complexed with wt SPP that remains in the flow-through. The relatively low amount of SPP-NT FLAG complexed with SPP-CT V5his is attributable to the multiple dimeric forms of SPP that potentially form in these cells. In all, six different dimers are possible: (SPP) 2 , (SPP-CT V5his) 2 , (SPP-NT FLAG) 2 , SPP⅐SPP-NT FLAG, SPP-CT V5his⅐SPP, and SPP-CT V5his⅐SPP-NT FLAG). Although we do not have precise data on the level of overexpression of each species, the amount of SPP-CT V5his was approximately equivalent to endogenous SPP, and the level of SPP-NT FLAG was lower than endogenous SPP. Thus, the amount of SPP-CT V5his⅐SPP-NT FLAG is predicted to represent less than 10% of all SPP dimers. To confirm that the association of SPP-CT V5his and SPP-NT FLAG was not occurring in vitro, cells lines expressing the individual constructs were lysed and mixed together, and Ni 2ϩ affinity purification was performed (Fig. 3b, postlysis control). No anti-FLAG-positive bands were observed in the elution of the mixed lysates. Moreover purification of lysates from cells expressing either tagged version alone did not result in the purification of any FLAG-immunoreactive proteins.
Glycerol Velocity Gradient Sedimentation of SPP and PS1-Glycerol velocity gradient fractionation can be used to provide estimates of the molecular masses of proteins and protein complexes. This technique provided initial evidence that PS1 existed in a high molecular mass complex (28). To explore whether SPP exists in a high molecular mass complex, we analyzed the distribution of SPP in a glycerol velocity gradient and compared it to both exogenous molecular mass marker proteins and the distribution of endogenous PS1. Cells overexpressing SPP-CT V5his and lysed in either CHAPSO (Fig. 4) or TX-100 (not shown) showed similar distributions of SPP upon glycerol velocity sedimentation. The SPP dimer distributed primarily in the 100 -160-kDa range with a small portion of the dimer SPP sedimenting at higher molecular mass. Comparable results were observed in gradients probed for the wt protein with anti-SPPnt (Fig. 4) and anti-SPPct antibodies (not shown). Only a small amount of monomer detected by anti-SPPnt was observed in fractions 2 and 3. These results contrast with the distribution of PS1 under varying detergent lysis conditions. Under conditions that maintain ␥-secretase activity (CHAPSO lysis), the NTF and CTF of PS1 sedimented to a high molecular mass fraction (Ͼ200 kDa, Fig. 4). Using detergents that disrupt and inactivate the ␥-secretase complex (TX-100), PS1 was found in lower molecular mass fractions (Fig. 4).
Metabolic Labeling of SPP-Having established that SPP is a dimer, we next evaluated the synthesis and degradation of SPP in both wt and SPP-CT V5his-overexpressing HEK cells with pulse-chase studies. Cells were labeled with [ 35 S]methionine and lysed in TX-100, and SPP was then immunoprecipitated using anti-SPPct, anti-SPPnt, or anti-V5 antibodies. Immunoprecipitation of SPP from either wt HEK cells or stably transfected HEK cells expressing SPP-CT V5his showed the SPP ϳ95-kDa dimer (Fig. 5a). As seen previously (Fig. 1b), the molecular mass of the dimer was shifted to a slightly higher molecular mass in the SPP-CT V5his-overexpressing cell line. Although a ϳ45-kDa nonspecific band was present in all of the The co-purification of SPP-NT FLAG and SPP-CT V5his together causes a small increase in the molecular mass of the dimer that is purified due to the presence of two epitope tags.
immunoprecipitates, including the protein A and G control, this band was augmented by the addition of antibody to the protein A beads and in the SPP-CT V5his-overexpressing cell line. Additionally the band decreased with increasing chase time. Although some ϳ45-kDa SPP monomer was detected, it was only a small fraction of the total SPP observed in these 2-h labeling studies. Specific detection of the ϳ45-kDa monomer was not altered by shorter (15-min) or longer (4-h) labeling periods. These studies demonstrate that the dimer form of SPP assembles rapidly. The half-life of SPP monomer and dimer was only about 2 h, but the remaining 50% of the total SPP dimer was stable for a very long period and had a half-life of 12 h. Although dimer and monomer SPPs had identical halflives in these experiments, the dimer was still detectable after 24 h, whereas no monomer was detected after 24 h. It is possible that SPP rapidly dimerizes coincident with translation, and little or no monomer is present in vivo.
〈n Active Site-directed ␥-Secretase Inhibitor Photoaffinity Labels SPP Dimer-To determine whether ␥-secretase and SPP contained similar active sites, compound III-63, a previ-ously characterized, photoactivable, biotinylated ␥-secretase inhibitor, was used to covalently label SPP (Fig. 6). This compound is a potent, active site-directed inhibitor of ␥-secretase that covalently labels PS1 in CHAPSO lysates and microsomes upon photoactivation (24). For these studies, CHAPSO-solubilized lysates were incubated with the photoactivable III-63 in the absence or presence of a parent compound, III-31-C, or an SPP inhibitor, (Z-LL) 2 -ketone (both competitors are unbiotinylated and not photoactivable). After pulling down labeled proteins with streptavidin beads, we analyzed the precipitates by Western blotting. We observed labeling of the SPP dimer in HEK, Chinese hamster ovary, and H4 cells overexpressing the SPP-CT V5his construct and wt lysates from untransfected H4 (not shown) and HeLa cells as well (Fig. 6). III-63 labeling of the dimer was completely blocked by the parent compound III-31-C and the SPP inhibitor (Z-LL) 2 -ketone (Fig. 6). In addition, III-63 labeling appears to be specific for the dimer form of SPP as no monomer labeling was observed. Heating the inhibitor-bound samples only generated a small amount of monomer relative to the heated control samples (no inhibitor). FIG. 4. SPP sediments at a size consistent with a homodimer. A 10-40% linear glycerol velocity gradient of HEK wt cells lysed in CHAPSO was probed with the anti-SPPnt antibody and shows that a dimer sediments in fractions consistent with a ϳ160-kDa complex and a monomer that sediments in even smaller fractions. SPP-CT V5his-overexpressing HEK cells were lysed in either CHAPSO or TX-100 (not shown) and following sedimentation and SDS-PAGE were probed with anti-V5 antibody, anti-PS1 CTF, and anti-PS1 NTF. Significant differences in the PS1 fragment sedimentation are observed in the two lysis conditions. However, no significant differences are observed in SPP-CT V5his sedimentation profiles with either of the lysis conditions. Gradient MW kDa at the bottom refers to control proteins loaded on separate but identically poured gradients. Lysate and control gradients were performed in triplicate. Comparable results were observed for endogenous SPP from H4 cells (not shown).
FIG. 5. SPP rapidly associates into a homodimer and is stable. a, HEK wt and SPP-CT V5his-overexpressing cells were [S 35 ]methioninelabeled for 2 h. Cells were lysed in TX-100 and immunoprecipitated with either anti-V5 or anti-SPPnt antibody. The SPP dimer at ϳ95 kDa (**) is detected above the background control (protein A ϩ G only). The dimer is augmented by the overexpression of SPP-CT V5his (ctV5his) and shifted slightly upward due to the presence of the epitope tag. Protein A (only)-specific bands are observed at 220, 105, and 14 kDa (NS) and do not appear to increase or decrease in any of our pulse-chase experiments performed. The protein A (only)-specific 45-kDa band is augmented with antibody-treated protein A and does decrease with increasing chase time. b, 50% of the SPP dimer is gone at the 2-h time point, but the remaining 50% has a half-life of ϳ12 h. This suggests that the III-63 binding may stabilize the dimer form of SPP, making it resistant to heat dissociation DISCUSSION We found that SPP exists primarily as a dimer and that this dimer is detergent-stable but heat-and acid-labile. The dimeric form of SPP could be specifically labeled with an active sitedirected ␥-secretase inhibitor, compound III-63, and importantly, this labeling could be blocked by the addition of the SPP inhibitor (Z-LL) 2 -ketone. Because transition state analog inhibitors of ␥-secretase are known to only interact with active protease (36), specific labeling of dimeric SPP by such analogs strongly implies that the dimer is active (17,19,30).
Glycerol velocity sedimentation was most consistent with the majority of SPP existing either as a dimer alone or as a dimer plus a small molecular mass component. However, the technique was not precise enough to distinguish between the two possibilities. Following CHAPSO detergent lysis, the majority of SPP sedimented with a peak at ϳ160 kDa. Because the dimer is ϳ95 kDa, it is possible that an additional component of ϳ50 kDa was present. Nevertheless to date we have no other data that suggest that SPP has other binding partners. Based on previous studies showing that SPP is active when expressed in a heterologous system (17,23), if there is indeed an additional SPP binding partner, then either it is conserved in yeast or it is not necessary for function. Pulse-chase experiments indicated that the SPP dimer assembles rapidly and is quite stable.
Two previous reports of SPP show bands at both ϳ45 and ϳ95 kDa but do not identify the larger molecular mass species as a dimer (19,30). Although each of the reports show both monomer and dimer, we found that SPP is predominantly a dimer. It is unclear whether the ratio of monomer to dimer differences are cell line-specific, but more likely they are due to different postlysis conditions. Many laboratories commonly heat or boil Western samples prior to SDS Western analysis, which would, at least, partially dissociate the dimer. SPP has been detected as an N-glycosylated, monomeric ϳ45-kDa species (17). In that study SPP was isolated and detected using the photoaffinity labeling reagent TBL 4 K followed by trichloroacetic acid precipitation, and there was no evidence for TBL 4 K labeling a ϳ95-kDa species. Significantly no anti-SPP antibody was ever used to detect SPP present in the labeled or unlabeled cell lysates (17). As our results demonstrated the vast majority of SPP is dimeric; either TBL 4 K labeling apparently does not enable detection of the SPP dimer or the conditions used in the analysis (trichloroacetic acid precipitation) converted all the dimer to monomer. We found that the parent compound of TBL 4 K, (Z-LL) 2 -ketone, could block binding of a ␥-secretase inhibitor to the dimeric form of SPP. These data suggest that (Z-LL) 2 -ketone binds dimeric SPP, and in all likelihood TBL 4 K does as well. Dimeric SPP could be labeled by TBL 4 K, but access to the biotin tag may be sterically hindered in the dimer, compromising detection of the labeled form by streptavidin. Alternatively TBL 4 K may bind specifically to monomeric SPP or disrupt the dimer. More likely, however, TBL 4 K covalently binds dimer, and as our data show (Fig. 2a), during trichloroacetic acid precipitation the dimer is disrupted.
Because of the difficulty in purifying an active ␥-secretase complex to homogeneity and, until very recently, reconstituting the ␥-secretase activity in a heterologous system, much of what has been learned about the nature of the ␥-secretase has come from mutational analysis and inhibitor studies. Mutation of the two transmembrane aspartates demonstrated a key role for these residues in ␥-secretase cleavage and provided key data that supported the hypothesis that PSs are aspartyl proteases (14). Inhibitor studies are also consistent with the hypothesis that ␥-secretase is an aspartyl protease, and the labeling of PS by different ␥-secretase inhibitors, including several designed to be active site-directed inhibitors of aspartyl proteases, further supports this hypothesis (10 -13, 31). Based on its multiple transmembrane regions and homology with a few key motifs, SPP has also been postulated to be an aspartic I-CLiP. Nevertheless the type of definitive structural data that is available for other families of proteases (e.g. classic aspartyl proteases such as pepsin and renin) has not been generated for ␥-secretase/PS activity. The finding that an active site-directed ␥-secretase inhibitor could label the dimeric form of SPP and that this labeling could be blocked by an SPP-specific inhibitor indicates that the active sites of these proteins are structurally related. Together with studies showing that several ␥-secretase inhibitors can inhibit SPP activity and an SPP inhibitor can inhibit ␥-secretase, these data provide further experimental support that the proteolytic activities of PS and SPP are related (23,24).
The finding that SPP exists primarily as a dimer may have implications for how the entire family of aspartyl I-CLiPs functions. There is some evidence that PS may also exist as a dimer. Yeast two-hybrid studies show various fragments of PS1 or intact PS1 can self-associate (32,33). Minor high molecular mass forms of PS have been detected by Western blotting after denaturing SDS-PAGE (32,33). More compelling are recent data based on cross-linking studies with inhibitors demonstrating that PS may indeed exist as a dimer or oligomer within the ␥-secretase complex (34). In that study, PS1 NTF⅐NTF dimers were detected following labeling with a transition state analog ␥-secretase inhibitor with two photolabile moieties.
The finding that SPP exists as a dimer has possible implications for how an aspartyl I-CLiP carries out intramembrane cleavage. Further study will be needed to determine precisely how dimerization is related to function and to definitively establish whether other members of the family form dimers. SPP is apparently a simpler aspartyl I-CLiP than ␥-secretase: SPP can be expressed in active forms without additional cofactors in heterologous systems and, unlike PS, does not appear to undergo endoproteolysis (35). Thus, SPP may be expected to be more easily characterized at a molecular and structural level than ␥-secretase, and future study of SPP dimers may provide FIG. 6. The SPP dimer is specifically labeled by a ␥-secretase inhibitor. ␥-Secretase inhibitor III-63 covalently labels the dimer (**) form of SPP in both SPP-CT V5his-overexpressing H4 cell lysates and wt HeLa cell lysates (ctV5his and Endogenous, respectively). No monomer (*) labeling was observed in any of the experiments performed. Heating the inhibitor-bound samples only generated a small amount of monomer relative to the control samples suggesting that the inhibitor has further stabilized that dimer form. Labeling of the SPP dimer is blocked by parent compound III-31-C (20 M) or an SPP inhibitor, (Z-LL) 2ketone (1 M). NS refers to a nonspecifically labeled band detected in HeLa cells by the anti-SPPct antibody. This band is present in human wt lysates probed with either the preimmune serum or peptideabsorbed antiserum of anti-SPPct antibody.