Structural Biology of Presenilins and Signal Peptide Peptidases

Abstract

Presenilin and signal peptide peptidase are multispanning intramembrane-cleaving proteases with a conserved catalytic GxGD motif. Presenilin comprises the catalytic subunit of γ-secretase, a protease responsible for the generation of amyloid-β peptides causative of Alzheimer disease. Signal peptide peptidase proteins are implicated in the regulation of the immune system. Both protease family proteins have been recognized as druggable targets for several human diseases, but their detailed structure still remains unknown. Recently, the x-ray structures of some archaeal GxGD proteases have been determined. We review the recent progress in biochemical and biophysical probing of the structure of these atypical proteases.

Introduction

Intramembrane proteolysis is an atypical hydrolysis of peptide bonds within the lipid bilayer. Several lines of evidence suggest that this unusual cleavage is involved in numerous biological processes encompassing all branches of life. So far, three families of intramembrane-cleaving proteases have been discovered: rhomboid, site-2 protease, and GxGD proteases, including γ-secretase and signal peptide peptidase (SPP).² γ-Secretase is a key enzyme in the production of amyloid-β peptide (Aβ), a major component of senile plaques in the brains of patients with Alzheimer disease, from amyloid-β precursor protein (APP) (Fig. 1) (1). In particular, γ-secretase cleavage determines the level of Aβ42, the most aggregation-prone species of Aβ predominantly deposited in the brains of Alzheimer disease patients (2, 3). Moreover, γ-secretase mediates the proteolysis-dependent signaling of several type I membrane proteins, including the Notch receptor, which is involved in the maintenance of stem cells and in the development of cancer (4). Thus, rational design of γ-secretase inhibitors (GSIs) and modulators (GSMs) based on the molecular mechanism of γ-secretase would pave the way toward development of novel drugs (5, 6). The identity of the γ-secretase had been a long-time mystery. PSEN genes were identified as familial Alzheimer disease (FAD)-linked genes encoding novel proteins, the presenilins (PSEN), lacking the traditional conserved aspartic protease motif (i.e. D(S/T)G) (7–9). Functional analyses in vitro and in vivo showed that presenilin (PS) is required for and modulates γ-secretase-mediated cleavage of APP (10–15). However, simple overexpression of PS does not significantly alter the γ-secretase activity in cells. Soon after the biochemical analysis of PS, it was recognized that PS forms a large membrane protein complex, and only overexpressed PS protein that is incorporated into the complex affects the APP cleavage. Crucial evidence of the enzymatic function of PS was obtained by chemical biology; transition state analog-type GSIs directly target PS (16, 17). In addition, the identification of membrane-embedded aspartate residues critical for enzymatic activity (18), as well as inhibitor binding experiments, revealed a novel conserved GxGD motif among species (19). In fact, the GxGD motif is conserved in type 4 prepilin peptidase proteins (20), and this motif is required for the secretion of type 4 prepilins or prepilin-like proteins in a wide range of bacterial species. Finally, another chemical biology approach unveiled a novel enzyme (SPP) that harbors exactly the same GxGD motif with a different primary sequence and topology compared with PS (Fig. 1) (21). Both the recombinant γ-secretase complex (22, 23) and SPP proteins reconstitute proteolytic activities, indicating that these proteins are truly proteolytic enzymes. Furthermore, recent x-ray crystallographic analyses of two archaeal membrane-spanning proteases carrying a GxGD motif revealed that the catalytic aspartates are facing a hydrophilic environment in the membrane (24, 25). Thus, now the identity of the γ-secretase is no longer in question. However, structural analyses of mammalian γ-secretase and SPP, which are required for the development of novel drugs, have not been fully achieved yet. In this minireview, we summarize the recent findings on the structure and function of γ-secretase and SPP proteins.

FIGURE 1.

Schematic depiction of γ-secretase and SPP. γ-Secretase executes the intramembrane proteolysis of several single membrane-spanning proteins, including APP. The core complex of γ-secretase is composed of PS, Nct, Aph-1, and Pen-2. SPP forms a homotetramer in its enzymatically active state. The topologies of PS and SPP are completely different. However, both PS and SPP harbor conserved catalytic motifs (aspartates are shown by stars) and the PAL motif (bold lines).

Presenilins

PS is a serpentine integral membrane protein with nine transmembrane domains (TMDs) (1). Two aspartyl residues in TMD6 and TMD7 of PS are required for its endoproteolytic activity, but not its expression and complex formation (18). Studies using transition state analog-type GSIs strongly indicate that PS is the enzymatic subunit of this atypical enzyme (16, 17). Biochemical purification and protein chemical analyses revealed Nct (nicastrin; NCSTN) as a major binding partner of PS (26). Moreover, genetic studies using Caenorhabditis elegans identified two more components required for γ-secretase activity that are conserved from worms to mammals: Aph-1 (anterior pharynx-defective 1; APH1) (27) and Pen-2 (presenilin enhancer 2; PSENEN) (Fig. 1) (28). Although several proteomic analyses of the γ-secretase complex have been performed to date, no other component critical for the enzymatic activity has so far been identified. In contrast, we and others have successfully reconstituted γ-secretase activity by coexpression of PS with these cofactors (22, 23). After complex assembly, PS undergoes auto-endoproteolysis between TMD6 and TMD7 to generate N- and C-terminal fragments (29, 30), which might reflect the active state of the enzyme. In fact, transition state analog-type GSIs bind only to the processed forms of PS (17). γ-Secretase-mediated cleavage is observed in the trans-Golgi network, on the cell surface, and in endocytic organelles, in which PS fragments are detected (31–33). In contrast, the PS holoprotein that is not incorporated into the complex is rapidly degraded in the endoplasmic reticulum, and trafficking of PS requires full assembly of the complex (17, 30). Recently, the proteolytic activity of purified recombinant PS protein was reported (34). This suggests that PS itself is a protease, although the proteolytic activity is quite low. These data strongly suggest that minimal γ-secretase is a membrane protein complex composed of PS fragments, Nct, Aph-1, and Pen-2.

Topological analyses of PS have implied that the catalytic aspartates are located at the center of the TMDs. This prediction prompted us to investigate the water accessibility of specific amino acid residues in a membrane-embedded state by the substituted cysteine accessibility method (SCAM) (35–37). SCAM is a procedure that has been frequently used to obtain structural information on various multipass membrane proteins in a functional state by covalently modifying the introduced cysteine residues using sulfhydryl reagents (38, 39). Using SCAM, we and others have revealed that PS1 harbors a hydrophilic “catalytic site” formed by TMD1, TMD6, TMD7, and TMD9 within the membrane (Fig. 2) (35–37, 40, 41). Residues at the luminal side of TMD6 and at the cytosolic side of TMD1, TMD7, and TMD9 form a subsite for the substrates. Especially, residues around the GxGD motif within TMD7 are highly water-accessible, suggesting the presence of a large water-filled cavity within the membrane at the cytoplasmic side of the complex. We also investigated the changes in water accessibility of the residues around the catalytic site during complex assembly. Unexpectedly, the water accessibility of the catalytic site of PS is widely increased before cofactor binding (42). This also fits with the hypothesis that the nascent PS polypeptide functions as a calcium leak channel in the endoplasmic reticulum (43). These ideas have now been supported by the results of structural studies on archaeal GxGD proteases (24, 25), and the hydrophilic milieu around the active site located within the lipid bilayer is a common structure across intramembrane-cleaving proteases.

FIGURE 2.

Predicted architecture of the catalytic site of PS1. Using SCAM, we identified that PS1 harbors a hydrophilic “catalytic site” formed by TMD1, TMD6, TMD7, and TMD9 within the membrane. Catalytic aspartates are shown by stars. Amino acid residues in the GxGD and PAL motifs are represented by circles. Putative entry of the substrate mediated by TMD9 is represented by an arrow.

Mechanism of γ-Secretase-mediated Cleavage

The molecular mechanism of proteolysis of the intramembrane sequence by γ-secretase has been enigmatic, as proteolytic processes require ionized water in general. However, extensive biochemical and enzymatic analyses revealed that γ-secretase executes an endopeptidase-like cleavage, followed by carboxypeptidase-like processive/successive cleavage (Fig. 3) (44). The transmembrane substrate is first endoproteolyzed at the border between the cytosol and membrane, which is called the ϵ-site (45, 46). This ϵ-cleavage allows liberation of the intracellular domains (ICDs) of the substrates from the membrane. ICDs are direct signaling mediators in several pathways, including Notch signaling (1, 4). Solid-state NMR and molecular dynamics simulation analysis of APP revealed that the structure around the ϵ-site is highly flexible in the membrane, which might allow access to the catalytic site of γ-secretase (47, 48). γ-Secretase then trims the remaining hydrophobic sequence in the membrane from the cytosolic side in a processive manner by every 3–4 residues (i.e. γ-cleavage). In the case of APP, the initial cutting at the ϵ-site forms Aβ48 or Aβ49, both of which are trimmed to generate the various C termini of the other Aβ peptides, ranging from 46 to 38 residues long (49, 50). Notably, the levels of Aβ48 and Aβ49 show some correlation (51), but not definitively, with the production of Aβ42 and Aβ40, respectively, suggesting the possibility that the proteolysis by γ-secretase occurs together with helix breaking of the substrate.

FIGURE 3.

Schematic model of the γ-secretase-mediated intramembrane cleavage. Nicastrin is a putative substrate receptor on the membrane. The substrate is captured and then incorporated into the catalytic site within PS. First, endopeptidase-like ϵ-cleavage occurs to liberate the ICD of the substrate. Subsequently, carboxypeptidase-like processive/successive γ-cleavage within the TMD results in release of the Aβ peptides.

SCAM analyses revealed the GSI/GSM-binding sites and conformational alterations upon compound binding. Transition state analog-type GSIs decrease the water accessibility of the residues around the GxGD and PAL (Pro-Ala-Leu) motifs (35, 36), the latter being highly conserved in PS and SPP proteins among species. Point mutations in the PAL motif also diminish the γ-secretase and SPP activities (52, 53). A cross-linking experiment also showed the proximity of the PAL motif to the catalytic center. Using this and the photoaffinity labeling technique, we identified that N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester, a PS-selective GSI, directly targets the C-terminal region of PS1 (54). A peptide with a stabilized helical conformation also inhibits γ-secretase activity, presumably by competing with the substrate (55–58). Intriguingly, the effect of a helical peptide in SCAM and photoaffinity labeling is totally distinct from that of a transition state analog-type GSI (35, 36, 57, 59), supporting the notion that PS harbors two enzymatically important sites: the catalytic pocket and the initial substrate-binding site, which might be involved in substrate selectivity (56). In addition, TMD2, TMD6, and TMD9 are implicated in the formation of the initial substrate-binding site (Fig. 2) (36, 41, 59). Nevertheless, detailed structural analyses of the initial substrate-binding site are critical for the development of substrate-specific GSIs. We also identified the movement of TMD1 of PS1 downward to the cytosolic side by SCAM (37). Intriguingly, inhibition of the motion of TMD1 by a specific antibody results in inhibition of γ-secretase activity (60). Moreover, we have identified TMD1 as the direct molecular target domain for the potent Aβ42-lowering compound GSM-1 (61). Interaction of GSM-1 with the luminal side of TMD1 induces a conformational change in the structure of the cytoplasmic region of TMD1, which reduces Aβ42 production. These data suggest the critical function of TMD1 in the intramembrane-cleaving mechanism. Notably, Aβ42 is capable of binding to γ-secretase to be processed to Aβ38 (62), although the binding site still remains unclear. Nevertheless, structural information on inhibitor/modulator-bound PS at the atomic level is mandatory to understand the molecular mechanism of the γ-secretase-mediated cleavage. Thus, a combination of SCAM with other structural analyses such as x-ray crystallography, NMR, and subsequent computer simulations would be ideal to understand the structure and function of PS.

Cofactor Proteins of the γ-Secretase Complex

Other subunits have also been implicated in the enzymatic process of γ-secretase. Nct is a single membrane-spanning protein with a heavily glycosylated extracellular domain (26). This extracellular domain has been implicated in the stability of the enzyme as well as substrate binding (63), although the latter has been opposed by results showing that recombinant PS (34) and SPP (64) are able to cleave peptide bonds without any cofactor proteins. However, we and others have shown that antibodies against the Nct extracellular domain are capable of inhibiting γ-secretase activity by competition with substrates (33, 65, 66), suggesting the possibility that Nct assists in capturing γ-substrates on the membrane (Fig. 3). Aph-1 is a highly hydrophobic protein with unknown function. There exist two or three mammalian aph-1 genes (Aph1a and Aph1b; Aph1c is only in mice), and two aph-1a isoforms are transcribed by alternative splicing. Recently, interaction of Aph-1 with arrestin, which regulates the trafficking and activity of the γ-secretase complex, was reported (67, 68). The fact that different Aph-1 proteins never exist in the same complex (69) suggests the possibility that Aph-1 functions as a binding scaffold for regulatory proteins to determine the specific activity of the γ-secretase complex at different subcellular localizations, as well as substrate preference. In addition, Aph-1 and Nct form a subcomplex in the early secretory pathway (70) and stabilize the γ-secretase complex (71) by binding to the very C-terminal end of PS (72, 73). Pen-2 is a small polypeptide with a hairpin like conformation and is required for activation of the γ-secretase subcomplex composed of PS, Nct, and Aph-1 (30). Pen-2 interacts directly with TMD4 of PS (74, 75), whereas its exact function still remains unclear. However, systematic mutagenesis analyses suggest that Pen-2 is involved in the stability of the complex (76) and the production of Aβ42 (77). Supporting this idea, an immobilized GSM with a phenylimidazole moiety pulled down Pen-2 (78). Nevertheless, the three cofactor proteins are required for full activity of γ-secretase on the cell membrane.

The assembled γ-secretase complex contains 19 TMDs, which may cause difficulties in crystallization of the fully active enzyme. Thus, we used an indirect approach: single particle analysis of the purified γ-secretase complex. We analyzed the complex overexpressed in Sf9 cells and found that the active γ-secretase is a very large complex with a volume of 560 × 320 × 240 Å at 48 Å resolution (79). Osenkowski et al. (80) extensively analyzed the purified complex by cryo-EM and revealed the minimal γ-secretase structure with dimensions of 8 × 9 nm in the top view and 8.5 nm in height at 12 Å resolution. They also identified the internal chamber and cavity that might be water-accessible in the putative TMD. Renzi et al. (81) also reported the structure of the γ-secretase complex analyzed by single particle analysis at 18 Å resolution. They compared this with the structure of the intermediate complex lacking Pen-2 and found a widening of the internal chamber of the complex. These data are consistent with our SCAM data demonstrating that the binding of the cofactors narrows the water accessibility of the catalytic site. Differences in the predicted shape and size may reflect a monomeric or oligomeric state and/or may be caused by distinct expression and purification protocols. Nevertheless, further structural studies of the assembled γ-secretase complex are required to understand the functional role(s) of these subunits.

Signal Peptide Peptidase

SPP cleaves remnant signal peptides in the membrane after their production by signal peptidase during the biogenesis of membrane proteins at the endoplasmic reticulum (82). This process is also critical to the immune surveillance system, in which signal peptides from major histocompatibility complex type I are cleaved by SPP (83). Moreover, SPP is implicated in the maturation of the core protein of hepatitis C virus and GB virus B, suggesting SPP as a potential target for antiviral therapy (84–86). A bioinformatics method identified four SPP-like (SPPL) proteins in fruit flies and mammalians: SPPL2a–c and SPPL3. SPPL2a and SPPL2b are highly expressed in various immune cells and are implicated in TNFα cleavage and interleukin-12 production (87, 88). Moreover, recent phenotypic analyses of knock-out mice revealed the critical function of SPPL2a in the development of B and dendritic cells via cleavage of the invariant chain (89–91).

Human PS and SPP share identical active site GxGD and highly conserved PAL motifs, pointing to a common catalytic mechanism. Supporting this notion, a subset of GSIs and GSMs directly target the SPP protein to inhibit/modulate its proteolytic activity (61, 92–95). SPP cleaves substrates at multiple sites in a processive manner similar to γ-secretase (94). Thus, the structure of the catalytic site of SPP should resemble that of PS, although the water accessibility of residues around the active site in SPP remains unknown. Moreover, expression of human SPP in yeast and in bacteria reconstitutes the proteolytic activity (21, 64), suggesting that the SPP protein has activity on its own and does not require other cofactors. Intriguingly, the recombinant C-terminal half of SPP is sufficient for proteolytic activity in vitro, indicating that this region is the minimal catalytic domain of SPP. We have purified the human SPP protein using the baculovirus/Sf9 cell system and analyzed its structure by single particle analysis (96). Enzymatically active SPP forms a bullet-shaped homotetramer with dimensions of 85 × 85 × 130 Å at 22 Å resolution. The SPP complex has a larger chamber within the molecule, and the N-terminal region of SPP is sufficient for its tetrameric assembly. This tetramer formation is a common feature of SPP family proteins. Thus, the ability to express active SPP/SPPL as a single protein indicates that a presenilin-like protease may be amenable to crystallization and high-resolution structural analysis.

Detailed Structure of GxGD Proteases

To date, no detailed structural information on mammalian PS and SPP is available. A C-terminal fragment of PS1 in SDS micelles was analyzed by NMR (97). However, no activity has been detected in the recombinant PS1 C-terminal fragment so far (98), suggesting that structural information on the molecule carrying two catalytic aspartates is important. Recently, x-ray crystallographic analyses of the archaeal GxGD protease have been reported (Fig. 4). In the first study, Hu et al. (24) reported the structure of FlaK from Methanococcus maripaludis at 3.6 Å resolution. The structure contains six TMDs, and the GxGD motif is located in TMD4. The second essential aspartyl residue is located away from the GxGD motif, in TMD1, suggesting that the protease must undergo conformational changes to bring the two aspartates into close proximity for catalysis. Although FlaK belongs to the type 4 prepilin/preflagellin peptidase family and shows no sequence homology to PS and SPP, several important key residues for PS activity were mapped around the catalytic site of FlaK (24, 42), indicating that the active site of the prokaryotic enzyme has a similar architecture. In the second study, Li et al. (25) reported the crystal structure of a PS homolog (PSH; MCMJR1) from the archaeon Methanoculleus marisnigri (99), which harbors nine TMDs, at 3.3 Å resolution. The structure of PSH contains a large hole traversing through the entire protein and a cavity that reaches the active site from the cytosolic side, in a manner similar to that in PS. However, the aspartates are separated by a distance of 6.7 Å, which is not enough for proteolysis. Thus, substrate binding must cause a conformational change in the catalytic site, as suggested in the case of FlaK. Notably, PSH shows considerably high sequence homology to PS/SPP and conserves several critical residues for endoproteolytic activity. In the structure of PSH, TMD9, which is implicated in the substrate binding and gating function of PS1 (36, 41), is facing the open space, raising the possibility that the substrate is likely to enter the active site laterally through the space between TMD6 and TMD9. Li et al. (25) generated a structural model of PS1 and annotated the FAD-linked mutations. They found that several artificial mutations in PSH corresponding to the residue in FAD-linked PS1 mutations decreased the proteolytic activity. This is highly reminiscent of the recent report that FAD-linked mutations cause malfunction in recombinant PS1 (34). These mutations are located at the interface of TMD helices or in proximity to the catalytic site in the PSH structure, suggesting that the local conformation of active enzyme is altered by amino acid substitution to decrease the proteolytic activity (25). However, it still remains difficult to annotate the effect of FAD mutations on the γ-secretase activity in this model, as FAD-linked mutations cause several malfunctions at multiple processes of the γ-secretase-mediated cleavage in a qualitative manner (100). For instance, mutation in TMD4, which is implicated in the Pen-2 binding to PS1, shows no effect on PSH activity. Thus, a detailed structural analysis of the whole γ-secretase complex is still required. Intriguingly, PSH forms a tetrameric assembly with its N-terminal half, which is reminiscent of the single particle analysis data on human SPP (96). The majority of the catalytic site architecture in PSH is formed by its C-terminal half, which is in accordance with the result that the C-terminal portion is the minimal proteolytic core domain of SPP (64). Thus, accurate modeling of the SPP structure based on PSH would provide further molecular information on the intramembrane-cleaving process by SPP/SPPL. Nevertheless, structural analyses of these archaeal GxGD proteases provide several hints and clues to the molecular mechanism of intramembrane-cleaving activity. However, much more effort is required to determine the detailed structure of γ-secretase, which is composed of PS and three cofactors.

FIGURE 4.

Structures of FlaK and PSH. Shown are ribbon representations of FlaK (left; Protein Data Bank code 3S0X, molecule A) and PSH (right; code 4HYG, molecule A). Putative catalytic aspartates are represented by spheres and indicated by dashed circles. These illustrations were generated using PyMOL.

Conclusion

Aspartyl intramembrane-cleaving enzymes are now found in all forms of life and play essential roles in biology and disease. This, together with advances in structural research, including rhomboid and site-2 proteases, has led to our understanding of the universal principle that intramembrane cleavage is achieved by creating an aqueous cavity inside the membrane-embedded protease. In addition, identification of small compounds and their utilization in chemical biology have provided several clues to understanding the mechanism of action of these atypical enzymes. However, there are still many unsolved issues in the structural biology of GxGD proteases. How do γ-secretase and SPP family proteins recognize and incorporate the substrate into the catalytic site? What molecular mechanism underlies the processive cleavage? What are the functional roles of the γ-secretase cofactors? How are the activities of these enzymes regulated? Can we design compounds to regulate the activities specifically to develop novel therapeutics against human diseases? Furthermore, the biggest issue is to determine the structure of the γ-secretase complex/SPP at atomic resolution in the enzymatically active state. The revolutionary advances in biochemical and structural biology with the development of novel techniques should give us answers to these questions in the near future.