Functional and Topological Analysis of Pen-2, the Fourth Subunit of the γ-Secretase Complex^,

Abstract

The γ-secretase complex is a member of the family of intramembrane cleaving proteases, involved in the generation of the Aβ peptides in Alzheimer disease. One of the four subunits of the complex, presenilin, harbors the catalytic site, although the role of the other three subunits is less well understood. Here, we studied the role of the smallest subunit, Pen-2, in vivo and in vitro. We found a profound Notch-deficiency phenotype in Pen-2^−/− embryos confirming the essential role of Pen-2 in the γ-secretase complex. We used Pen-2^−/− fibroblasts to investigate the structure-function relation of Pen-2 by the scanning cysteine accessibility method. We showed that glycine 22 and proline 27 in hydrophobic domain 1 of Pen-2 are essential for complex formation and stability of γ-secretase. We also demonstrated that hydrophobic domain 1 and the loop domain of Pen-2 are located in a water-containing cavity and are in short proximity to the presenilin C-terminal fragment. We finally demonstrated the essential role of Pen-2 for the proteolytic activity of the complex. Our study supports the hypothesis that Pen-2 is more than a structural component of the γ-secretase complex and may contribute to the catalytic mechanism of the enzyme.

Introduction

γ-Secretase is a membrane-embedded aspartic protease composed of four subunits as follows: presenilin1 or 2 (PS1 or PS2), nicastrin (NCT),³ Aph-1a or Aph-1b, and Pen-2 (1). γ-Secretase is an attractive drug target for Alzheimer disease because it is responsible for the final cleavage in the processing of the amyloid precursor protein (APP) to Aβ peptides (2). However, because of its broad substrate specificity, general inhibition of γ-secretase is not without risks. The main problem is the involvement of γ-secretase in the cleavage of Notch, which releases the Notch intracellular domain (NICD) that is implicated in crucial signaling processes throughout life (3–5). The essential role of γ-secretase in early development is illustrated by the embryonic lethality of Drosophila melanogaster, Caenorhabditis elegans, or Mus musculus in which NCT, Aph-1a, or the two PSs together are genetically inactivated. In all cases, a severe Notch-deficiency phenotype is observed (5–15). However, it should be noted that other single knock-outs of subunits of the γ-secretase complex, like PS2 alone (7, 11) or Aph-1b (8, 15), do not result in Notch phenotypes. The effect of a genetic knock-out of Pen-2 in mice has not been reported until now.

The catalytic site of γ-secretase consists of two conserved aspartates, in transmembrane domains 6 and 7 (TMD6 and TMD7) of PS, which are crucial for both PS endoproteolysis and γ-secretase activity (16, 17). The catalytic aspartates are present in a water-containing cavity to which other parts of PS contribute as well (18–23); TMD9 and the loop connecting TMD6 and TDM7 containing the endoproteolysis site (hydrophobic domain VII (HDVII)) are in close proximity to the catalytic aspartates and play a role in γ-secretase activity. The proximity of TMD9 to the catalytic aspartates, its highly flexible nature and the fact that TMD9 binds directly to the substrate APP C99 further suggest that TMD9 is possibly involved in the transport of the substrate from the initial docking site to the catalytic cavity where it gets processed (19, 21, 24).

The role of the three other subunits of γ-secretase and their contribution to the catalytic activity of the protease are much less well understood. In this study, we focused on the smallest subunit, Pen-2 (presenilin enhancer 2), originally identified in a genetic screen for modulators of PS activity in C. elegans (13). Pen-2 is a 101-amino acid-long protein with two hydrophobic domains. A hairpin topology, with the loop domain exposed to the intracellular side of the cell membrane, has been proposed for the protein (25). RNAi-mediated down-regulation of Pen-2 in cell culture leads to decreased endoproteolysis of PS, which is associated with an increase of full-length PS and a decrease of PS N- and C-terminal fragments (PS NTF and PS CTF) (26–29). Additionally, mutational analysis has shown that the N-terminal part of hydrophobic domain 1 of Pen-2 interacts with the TMD4 of PS1 (30, 31) and is important for PS endoproteolysis (32). These observations suggest that Pen-2 is involved in the endoproteolysis of PS and therefore in the activation of the γ-secretase complex (26, 27, 29). Actually, Ahn et al. (33) recently demonstrated that, in an in vitro system, the combination of PS1 and Pen-2 was necessary and sufficient to induce PS endoproteolysis and γ-secretase-like activity, confirming the involvement of Pen-2 in the activation of PS.

It has been demonstrated that the conserved amino acid sequence motif, DYSLF, in the C terminus of Pen-2 as well as the length of this part of the protein are not only crucial for the assembly of the γ-secretase complex but also for the stabilization of the PS fragments after endoproteolysis (27, 34–36). Furthermore, incorporation of a FLAG tag at the N terminus of Pen-2 changes the conformation of PS, resulting in an increased Aβ₄₂/Aβ₄₀ ratio (32), similar to what is observed for familial Alzheimer disease mutations in the PS subunit (37). Interestingly, a γ-secretase modulator that decreases Aβ₄₂ production binds mainly to Pen-2 (38), further arguing for the crucial role of Pen-2 in the regulation of the activity of the complex, although the mechanism of this regulation has so far remained elusive.

We report here the phenotype of Pen-2^−/− mouse embryos and demonstrate the essential role of Pen-2 in γ-secretase processing of Notch and APP. In the absence of Pen-2, no processing of APP is observed, in contrast to NCT^−/− cells, which retain 5–6% of γ-secretase activity (39). We also made use of the Pen-2^−/− fibroblasts to perform a cysteine-scanning analysis of the Pen-2 protein (18–21, 40) to investigate the topology of Pen-2 and its spatial and functional relation to the water-accessible catalytic site of PS. Our data identify crucial residues in the Pen-2 sequence for γ-secretase complex assembly and stabilization and show that the hydrophobic domain 1 and the loop of Pen-2 are located in a water-containing cavity, in close proximity to PS1 CTF.

EXPERIMENTAL PROCEDURES

Generation of Pen-2^−/− Embryos

Pen-2^+/− male and female mice were purchased from the Texas Institute for Genomic Medicine and coupled to obtain Pen-2^−/− embryos.

Whole Mount in Situ Hybridization

Embryos were isolated from the mother at embryonic day 8.5 (E8.5) and fixed in 4% paraformaldehyde in PBS. After dehydration to 100% methanol and bleaching in 6% hydrogen peroxide for 1 h at room temperature, embryos were rehydrated in 100% PBS + 0.1% Tween. The embryos were treated with 10 μg/ml proteinase K for 5 min and post-fixated in 0.2% glutaraldehyde and 4% paraformaldehyde. Prehybridization was done in 50% formamide and 5× SSC, pH 4.5, complemented with 50 μg/ml yeast RNA and 50 μg/ml porcine heparin during 1 h at 60 °C, after which the prehybridization mix was replaced with the hybridization mix containing 1 ng/μl digitonin-labeled riboprobes against Notch signaling pathway components as indicated. After overnight incubation in the hybridization mix, the embryos were washed extensively in PBS with 0.1% Tween and treated with RNase A (100 mg/ml) for 15 min.

Blocking was performed in 2% Boehringer Blocking Reagent and 20% fetal calf serum in MABT buffer (100 mm maleic acid, 150 mm NaCl, pH 7.5, 0.1% Tween). After blocking, the embryos were treated with an alkaline phosphatase-coupled anti-digitonin antibody (final concentration 1:2000) overnight at 4 °C. The next day, the embryos were washed in MABT and treated with levamisole (2 mm). BM purple, a chromogenic substrate for alkaline phosphatase, was added, and the staining was visualized by light microscopy.

Generation of Pen-2^−/− Fibroblasts and Cell Culture

Pen-2^−/− fibroblasts were generated from E9.5 embryos and immortalized by transfection with the plasmid pMSSVLT, driving expression of the large T-antigen. Immortalized mouse embryonic fibroblasts were cultured in Dulbecco's modified Eagle's medium/F-12 containing 10% fetal bovine serum (Sigma).

Generation of Pen-2 Mutants and Corresponding Stable Cell Lines

All mutations in Pen-2 were generated with the XL site-directed mutagenesis kit (Stratagene) and confirmed by DNA sequence analysis. Retroviruses were generated by co-transfecting pMSCVpuro vector containing Pen-2 and the PIK helper plasmid into HEK293 cells. Viral particles harvested at 48 h post-transfection were used to infect Pen-2^−/− fibroblasts at 30–40% confluency. Transduced cells were selected with 5 μg/ml puromycin.

Antibodies

Polyclonal antibodies against mouse Pen-2 (B126.2), PS1 NTF (B19.3), Aph-1a (B80.2), and APP C terminus (B63.3) and monoclonal 9C3 against the C terminus of nicastrin have been described previously (41, 42). The following antibodies were purchased: anti-N-cadherin from BD Biosciences; anti-NICD (cleaved Notch1 Val-1744) from Cell Signaling Technologies; mAb 9E10 (Sanver Tech); and MAB5232 against PS1 CTF (Chemicon). The anti-FLAG M2 antibody was purchased from Sigma.

SDS-PAGE

Total cell lysates were prepared in lysis buffer (5 mm Tris-HCl, pH 7.4, 250 mm sucrose, 1 mm EGTA, 1% Triton X-100, and complete protease inhibitors (Roche Applied Science)). Post-nuclear fractions were taken, and protein concentrations were determined using standard BCA assay (Pierce). 20 μg of protein was separated on 4–12% BisTris gels and transferred to nitrocellulose membranes to perform Western blot analysis. Signals were detected using the chemiluminescence detection with Renaissance (PerkinElmer Life Sciences). Quantifications were performed by means of densitometry.

Blue Native PAGE

Microsomal membrane fractions were prepared in lysis buffer containing 0.5% dodecyl maltoside, 20% glycerol, and 25 mm BisTris/HCl, pH 7. Upon ultracentrifugation (55,000 rpm), supernatant was taken; protein concentrations were measured, and 5 μg of protein was supplemented with 5× concentrated sample buffer (2.5% Coomassie Blue G-250, 100 mm BisTris/HCl, 500 mm 6-aminocaproic acid, pH 7, 15% sucrose). Samples were loaded onto a 5–16% polyacrylamide gel and run for 4 h at 4 °C. Subsequently, the gel was incubated with 0.1% SDS in transfer buffer for 10 min at room temperature and transferred to a polyvinylidene difluoride membrane. Membranes were destained in water/methanol/acetic acid (60:30:10, v/v) and incubated with antibodies to detect the γ-secretase complex.

Water Accessibility Assay

Cells were plated in 10-cm dishes, washed with PBS, and treated with the biotinylated sulfhydryl-specific reagents indicated below for 30 min at 4 °C. In case of pretreatment with sodium (2-sulfonatoethyl)methanethiosulfonate (MTSES), cells were incubated with MTSES or DMSO for 30 min at 4 °C before treatment with the biotinylated reagent. After extensive washing in PBS to remove unbound reagent, cells were collected and lysed in 25 mm HEPES, pH 8, 150 mm NaCl, 2 mm EDTA, 1% Triton X-100. Biotinylated proteins in the total cell lysates were precipitated with immobilized NeutrAvidin protein beads (Pierce). Bound proteins were eluted by boiling in Nu-PAGE sample buffer, and SDS-PAGE was performed as mentioned above. Pen-2 was detected in input and bound fractions.

Sulfhydryl-specific reagents used were as follows: EZ-Link biotin-HPDP (N-(6-(biotinamido)hexyl)-3′-(2′-pyridyldithio)propionamide, Pierce) (200 μm), MTSEA-biotin (N-biotinylaminoethyl methanethiosulfonate, Biotium) (500 μm) and TS-XX-biotin ethylenediamine (biotin-XX ethylenediamine thiosulfate, sodium salt, Biotium) (500 μm), and MTSES (sodium (2-sulfonatoethyl)methanethiosulfonate, Biotium) (200 μm).

Cross-linking

Cross-linking studies were performed by incubating microsomal membrane fractions with the heterobifunctional amine sulfhydryl-reactive cross-linker SPDP (N-succinimidyl-3-(2-pyridyldithio)propionate) (1 mm). As negative controls, membranes preincubated with 10 mm N-ethylmaleimide, 10 mm EDTA, or membranes incubated with DMSO were used.

Reactions were quenched with N-ethylmaleimide/EDTA. After centrifugation at 55,000 rpm, membrane pellets were solubilized in 5 mm Tris-HCl, pH 7.4, 250 mm sucrose, 1 mm EGTA, 1% Triton X-100; equal amounts of proteins were separated in SDS-PAGE in nonreducing conditions and detected by Western blotting using antibodies against γ-secretase components.

Activity Assays

Cell-free APP Processing Assay

Cell-free assays were performed as described by Kakuda et al. (43) with some minor modifications. Briefly, microsomal membrane fractions solubilized in 1% CHAPSO were mixed with recombinant APPC99–3×FLAG substrate (0.5 μm final concentration), 0.0125% phosphatidylethanolamine, 0.1% phosphatidylcholine, and 2.5% DMSO. Reactions were incubated at 37 °C for 3 h. AICD was detected by Western blot analysis with the anti-FLAG M2 antibody and Aβ species by AlphaLISA. Aβ and AICD levels were normalized to the amounts of γ-secretase complex in the in vitro assay, which were estimated from the PS1 NTF levels.

Cell-based APP Processing Assay

Fibroblasts were infected with Ad5/cytomegalovirus bearing human APP-695 containing the Swedish mutation. The cells were then cultured in Dulbecco's modified Eagle's medium supplemented with 0.2% fetal bovine serum for 16 h, and the conditioned medium was collected and used to analyze APP processing. Aβ₄₀ and Aβ₄₂ levels were quantified by AlphaLISA and soluble APP levels by SDS-PAGE and Western blot analysis. Cell lysates were prepared, and APP CTF fragments were analyzed by SDS-PAGE followed by Western blot. Aβ levels were normalized to γ-secretase complex levels and, to infection efficiency, quantified from the sAPP expression levels.

Cell-based Notch Processing Assay

Fibroblasts were infected with Ad5/dE1dE2a/cytomegalovirus containing Myc-tagged NotchΔE. At 24 h post-infection, 10 μm lactacystin was added to the cultures, and after 4 h cell lysates were prepared. NICD and NotchΔE levels were estimated by Western blot analysis using a neo-epitope (cleaved Notch1 Val-1744) and an anti-Myc antibody, respectively. NICD levels were normalized to the levels of γ-secretase complex and to infection efficiency (levels of NotchΔE).

AlphaLISA (PerkinElmer Life Sciences)

Conditioned medium was mixed with PBS supplemented with 0.1% casein, streptavidin-coated AlphaLISA donor beads, and biotinylated antibody against the neo-epitope of Aβ₄₀ or Aβ₄₂. After overnight incubation at 4 °C, acceptor-beads coupled to an antibody against the N terminus of Aβ were added. After 1 h incubation, light emission (615 nm) was detected upon laser excitation at 680 nm.

Statistical Analysis

Data from at least three independent experiments were used for calculation of p values using the Student's t test.

RESULTS AND DISCUSSION

Pen-2^−/− Embryos Display a Notch-deficiency Phenotype

We generated a Pen-2^−/− mouse model by crossing Pen-2^+/− mice obtained from the Texas Institute for Genomic Medicine. The Pen-2 gene was inactivated by exchanging three of the four Pen-2 exons (exons 2–4) for a lacZ/neomycin cassette. Pen-2 gene disruption resulted in embryonic lethality, confirming the essential role of Pen-2 in the γ-secretase complex. Up to E9.5, living Pen-2^−/− embryos were recovered in a nearly Mendelian ratio (21:93), and they were completely absorbed by E11.

At E9.5, the yolk sac surrounding the Pen-2^−/− embryos had only a primitive immature blood vessel plexus and a blistered appearance compared with the yolk sac of the Pen-2^+/+ or Pen-2^+/− littermates, which displayed a normal complex vasculature (Fig. 1A). The Pen-2^−/− embryos were smaller than the Pen-2^+/+ littermates. They had an abnormally large pericardial sac, a kinked neural tube, and a truncated posterior region. Somitogenesis was initiated in the Pen-2^−/− embryos but appeared severely delayed compared with the Pen-2^+/+ littermates. The optic and otic vesicles, the first branchial arch, and the forelimb buds were visible, but the fusion of the head folds was delayed (Fig. 1A).

FIGURE 1.

Notch deficiency phenotype of the Pen-2^−/− embryo. A, E8.5 Pen-2^−/− embryos display a severe Notch deficiency phenotype. Upper panels, although an initial vascular plexus and primitive red blood cells have formed (arrow), organization into a discrete network of vitelline vessels is lacking. The yolk sac of Pen-2^+/+ littermates has a complex vascular network and is stretched. Lower panels, Pen-2^−/− embryos are smaller than Pen-2^+/+ embryos and are characterized by a large pericardial sac (arrow), a distorted neural tube, a truncated posterior, and a delayed development. B, Hes-5 and Dll-1 mRNA expression detected by whole mount in situ hybridization with digitonin-labeled riboprobes confirms a disturbed Notch signaling pathway. The Hes-5 probe (upper panels) reveals a lack of Hes-5 expression in Pen-2^−/− embryos (no blue-brownish color) compared with Pen-2^+/+ embryos (blue-brownish color going from the head to the tail along the spinal cord). Dll1–1 (lower panels) is ectopically expressed in the neural tube and the head in Pen-2^−/− embryos (blue-brownish color, indicated by arrows).

The phenotype of the Pen-2^−/− embryos was similar to the previously characterized phenotype of the PS1^−/− PS2^−/− embryos and reflects a lack of γ-secretase activity and the consequent effects on the Notch signaling pathways (7, 11). To confirm the disturbance in Notch signaling, we performed whole mount in situ hybridization on embryos at E8.5 using Hes-5 and Delta-like1 probes to detect expression of the respective genes. If Notch cleavage by γ-secretase is blocked, Hes-5 signals are expected to be down-regulated and Delta-like1 signals to be up-regulated. We observed indeed the absence of Hes-5 mRNA and ectopic expression of Delta-like 1 mRNA in Pen-2^−/− compared with the Pen-2^+/+ embryos (Fig. 1B).

The data confirm that Pen-2 is an essential component of γ-secretase. Because of the severe Notch phenotype, it is difficult to detect potential non-Notch signaling-dependent phenotypes (as in animals with the other γ-secretase components inactivated (7, 8, 10, 11, 44)). This makes it difficult to assess the contribution of other γ-secretase substrates in the Pen-2^−/− phenotype and to rule out the possibility that Pen-2 has additional non-γ-secretase-related functions. However, taking into account that Pen-2 protein expression is strongly dependent on the presence of the three other γ-secretase components (8, 44, 45), it seems logical to assume that Pen-2 only functions as part of the γ-secretase complex.

Effect of Pen-2 Knock-out on γ-Secretase Activity and Assembly

We derived Pen-2^+/+, Pen-2^+/−, and Pen-2^−/− fibroblasts from mouse embryos at E9.5 and immortalized the cells. Western blot analysis of Pen-2^−/− cell membrane extracts demonstrated the complete absence of Pen-2 protein. In contrast, the three other components of the γ-secretase complex were present at normal levels (i.e. equal to the expression in Pen-2^+/+ fibroblasts). Interestingly, PS1 was mainly present as full-length protein, and NCT failed to become fully glycosylated. We observed an accumulation of APP CTF, indicating a γ-secretase activity deficiency in the Pen-2^−/− fibroblasts (Fig. 2A). γ-Secretase maturation and activity correlated with the expression levels of Pen-2 in these cells as assessed in Pen-2^+/− cells, suggesting a limiting role of Pen-2 in the assembly and activity of the complex in this cell type (Fig. 2, A and B).

FIGURE 2.

Characterization of fibroblasts heterozygous (Pen-2^+/−) or homozygous (Pen-2^−/−) for the Pen-2 deletion gene and of Pen-2^−/− fibroblasts expressing the cysteine-less Pen-2 mutant. A, 4–12% BisTris SDS-PAGE/Western blot analysis of Pen-2^+/+, Pen-2^+/−, and Pen-2^−/− fibroblasts derived from E9.5 mice. 20 μg of solubilized membrane protein was applied per lane. Antibody 9C3 was used to stain immature (immat) and mature (mat) NCT, and B19 was used to stain full-length presenilin1 (FL PS1) and presenilin1 N-terminal fragments (PS1 NTF) and Mab5232 for the C-terminal fragments (PS1 CTF). Antibody B63 was used to stain full-length APP (FL APP) and APP C-terminal fragments (APP CTF) and antibody B80 for Aph-1a. Actin, loading control. B, Blue Native PAGE/Western blot analysis of Pen-2^+/+, Pen-2^+/−, and Pen-2^−/− DDM extracted membranes (5 μg/lane) stained with antibodies against PS1 CTF, NCT, and Aph-1a as in A. Full γ-secretase complex is absent in Pen-2^−/− fibroblasts. The trimeric complex in Pen-2^−/− fibroblasts is composed of FL PS1, NCT, and Aph-1a. *, artificial complex because of detergent-dependent dissociation, composed of PS1 CTF, Aph-1a, and NCT (59). C, Blue Native PAGE/Western blot analysis (5 μg/lane) of membrane fractions of Pen2^−/− fibroblasts reconstituted with cysteine-less Pen-2 (CL Pen-2) or wild type Pen-2 (WT Pen-2) and Pen-2^+/+ or Pen-2^−/− fibroblasts. The cysteine-less Pen-2 mutant is incorporated in the mature γ-secretase complex to the same extent as the wild type Pen-2. *, artificial complex (A). D, SDS-PAGE/Western blot analysis of membranes of the same fibroblasts as in C and using the same antibodies as in A. The cysteine-less Pen-2 mutant rescues NCT maturation, PS1 endoproteolysis, and APP CTF cleavage to the same extent as wild type Pen-2.

In Blue Native-PAGE analysis, a trimeric complex composed of full-length PS1, NCT, and Aph-1a was detected in Pen-2^−/− fibroblasts (Fig. 2B). Because this trimeric complex contains the catalytic subunit of γ-secretase (PS), we wondered whether this subcomplex had some remaining enzymatic activity. We measured Aβ levels in the conditioned medium of Pen-2^+/+ and Pen-2^−/− fibroblasts using the AlphaLISA technology. We also tested the activity of the complex in a cell-free assay using microsomal membrane fractions solubilized in 1% CHAPSO and recombinant APP-3×FLAG substrate. AICD levels in the in vitro reactions were quantified by SDS-PAGE and Western blotting.

Cell-based activity assays using the Pen-2^−/− fibroblasts showed that APP CTF fragments accumulated, and Aβ₄₀, Aβ₄₂, and NICD production were abolished (Fig. 3, B and C). A dilution series of conditioned medium of Pen-2^+/+ fibroblasts demonstrated that Aβ production in Pen-2^−/− fibroblasts was less than 0.3% of the Aβ production in Pen-2^+/+ fibroblasts (Fig. 3D). This contrasts with at least 10-fold higher Aβ production from NCT^−/− fibroblasts (∼3% of the Pen-2^+/+ Aβ production), confirming that NCT is dispensable for γ-secretase activity (39). Pen-2, however, appears to be absolutely necessary to generate an active complex in cell culture. In a recent in vitro reconstitution paradigm, it was also shown that Pen-2 is needed to activate (wild type) PS (33).

FIGURE 3.

Lack of γ-secretase activity in Pen-2^−/− fibroblasts is not simply caused by a lack of PS endoproteolysis. A, Blue Native PAGE/Western blot analysis of γ-secretase complexes from Pen-2^+/+ PS1/2^+/+ fibroblasts (WT), Pen-2^−/− fibroblasts, Pen-2^−/− fibroblasts expressing the human PS1 ΔE9, and PS1^−/− PS2^−/− fibroblasts (PS dKO) rescued with PS1 ΔE9. PS1 ΔE9 is incorporated in the γ-secretase complexes. +, two wild type alleles; −, two knock-out alleles; T, transfected; /, not applicable. WB, Western blot. B, cell-based activity assay overexpressing the APPswedisch mutant indicated the lack of Aβ₄₀, Aβ₄₂, and NICD production and an accumulation of APP CTFs in Pen-2^−/− fibroblasts to a similar extent as observed in PS double KO fibroblasts. Processing of APP and Notch was partially rescued upon PS1 ΔE9 expression in PS double KO cells but not in Pen-2^−/− cells. C, AICD was absent in cell-free reactions using Pen-2^−/− microsomal fractions. Expression of human PS1 ΔE9 in Pen-2^−/− fibroblasts did not restore γ-secretase activity. D, dilution series of conditioned medium of WT fibroblasts showed that Aβ levels in Pen-2^−/− fibroblasts are lower than 0.3% of the Aβ levels in WT fibroblasts. Aβ levels were not detectable in Pen-2^−/− fibroblasts rescued with the PS1 ΔE9 mutant either. The minor Aβ levels in NCT^−/− fibroblasts could be detected as demonstrated before (39), confirming the sensitivity of the assay.

Cell-free activity assays confirmed the cell-based data as there was no in vitro Aβ production (data not shown). Furthermore, AICD production was completely abolished (Fig. 3C) indicating that not only the γ-cleavage of APP was disturbed but the ϵ-cleavage as well.

Because Pen-2 is generally considered to induce PS endoproteolysis by an unknown mechanism, we wondered whether the lack of activity of the trimeric complex in Pen-2^−/− fibroblasts was simply the consequence of the fact that PS fails to undergo endoproteolysis in the absence of Pen-2. To investigate this, we stably expressed the PS1 Δexon9 (PS1 ΔE9) mutant in the Pen-2^−/− fibroblasts. PS1 ΔE9 is active without the need for endoproteolytical processing (46–48). Even though PS1 ΔE9 was incorporated in the trimeric complex (Fig. 3A), we failed to detect any γ-secretase activity (Fig. 3, B and C). In contrast, the PS1 ΔE9 mutant rescued γ-secretase activity in fibroblasts lacking both PS1 and PS2, proving that it is indeed active in its uncleaved form. This result indicates that, at least in vivo, Pen-2 is essential for activity of the γ-secretase complex per se, independently of its role in the endoproteolysis of PS. Potential mechanisms underlying this regulation include the following: (i) the stabilization of the PS fragments after heterodimerization (34); (ii) the specificity of γ-secretase activity (32); and (iii) the accessibility or the affinity of γ-secretase to the substrate (49).

Ahn et al. (33) demonstrated that, in an in vitro system, PS1 ΔE9 exhibited activity on its own, without the need for other γ-secretase components. In contrast, in our experimental conditions, PS1 ΔE9 is in complex with NCT and Aph-1 (trimeric complex), in a native state in the membrane. Our activity assays clearly show that there is no γ-secretase activity associated with the trimeric complex. Therefore, we hypothesize that, without Pen-2, PS1 ΔE9 is either kept in an inactive conformation by the association with NCT/Aph-1 or that the substrate cannot enter the catalytic site, which would suggest a role for Pen-2 in the gating mechanism. In both scenarios Pen-2 is involved in the catalytic mechanism of γ-secretase, beyond endoproteolysis. Furthermore, we cannot exclude that interaction with other unknown membrane-associated factors may play a role as well.

Scanning Cysteine Accessibility Method Analysis of Pen-2

Little is known about the topology and exact position of Pen-2 in the context of the γ-secretase complex. Therefore, we decided to use the scanning cysteine accessibility method, which has proven to be a valuable approach in the analysis of the structure-function of the PS1 subunit of the complex (18–21).

We first generated a cysteine-free form of Pen-2 (CL Pen-2) by replacing the unique cysteine (Cys¹⁵) in wild type Pen-2 by an alanine. The CL Pen-2 variant was able to rescue γ-secretase complex formation, PS1 endoproteolysis, and NCT maturation in the Pen-2^−/− fibroblasts (Fig. 2, C and D). Importantly, as judged by APP CTF protein levels, γ-secretase proteolytic activity was restored to the same extent as in Pen-2^−/− fibroblasts rescued with wild type Pen-2 (Fig. 2D). Therefore, we concluded that CL Pen-2 is a suitable backbone for further analysis with scanning cysteine accessibility method.

We incorporated cysteine residues in the two hydrophobic and the hairpin-loop domains of Pen-2 by site-directed mutagenesis (overview Fig. 5A). The Pen-2 mutants were stably expressed in the Pen-2^−/− fibroblasts, and their ability to rescue γ-secretase complex formation and maturation was verified by SDS-PAGE and Blue Native PAGE analysis.

FIGURE 5.

Water accessibility pattern of unique cysteines in Pen-2 reveals an unexpected topology for Pen-2. A, amino acid (AA) sequence of murine Pen-2 is shown. The amino acids that were mutated to cysteine are indicated. Putative hydrophobic domain 1 and 2 of Pen-2 are displayed as α-helices according to the predictions of TOPPRED. An overview of the results of the water accessibility assays is shown in A. Dark and light gray circles represent cysteines with high or low reactivity towards the biotinylated reagents, respectively. Light ringed white circles represent cysteines that do not react with the biotinylated reagents. B, intact cells were exposed to the membrane-permeable sulfhydryl-specific reagent EZ-linked biotin-HPDP (upper panels) or the membrane-impermeable sulfhydryl-specific reagent MTSEA-biotin (lower panels). Biotinylated proteins were precipitated by neutravidin beads in the presence of Triton X-100. Input and bound fractions were separated by SDS-PAGE (4–12% BisTris) and Western blotting, and Pen-2 was detected using the B126 antibody recognizing the N terminus. Wild type Pen-2 (WT) with its lumenal cysteine at the N terminus was used as a positive control; cysteine-less Pen-2 (CL) was a negative control. C, as a control for the reliability of the MTSEA reagent, intact cells were treated with the membrane-impermeable TS-biotin-XX in the same way as in B. The Pen-2 cysteine mutant E49C shows reactivity to the impermeable TS-XX-biotin, confirming the results of the MTSEA-biotin reagent. D, pretreatment with the bulky and charged MTSES-reagent results in a decreased MTSEA-biotin labeling of the cysteine at the C terminus of Pen-2 (W85C). In contrast, the endogenous cysteine at the N terminus of Pen-2 (Cys¹⁵) does not display labeling differences with or without pretreatment, indicating that this cysteine is present in a restricted environment. Pretreatment with MTSES did not change the labeling with MTSEA-biotin for the cysteine in the N-terminal part of hydrophobic domain 1 (F25C) either. E, model of the mode of action of the different sulfhydryl-specific reagents used in this research. The different positions (numbers 1–5) of cysteines in a hypothetical membrane spanning protein and their labeling (+ or −) by the sulfhydryl-specific reagents used in this study are indicated.

Except for G22C and P27C, both in hydrophobic domain 1, all Pen-2 mutants were able to rescue γ-secretase complex formation, NCT maturation, and PS endoproteolysis (supplemental figure), suggesting that the introduction of cysteines in Pen-2 in general does not interfere with complex formation.

Glycine 22 and Proline 27 in Pen-2 Hydrophobic Domain 1 Are Involved in Complex Formation and/or Stabilization

Interestingly, although the Pen-2 G22C mutant was expressed in Pen-2^−/− fibroblasts to a sufficient extent, complex formation, maturation, and activity were severely compromised by the mutation (Fig. 4, A and B). However, the Pen-2 P27C mutant showed low expression levels that directly correlated with γ-secretase complex levels (Fig. 4, A and B). This effect was observed in several independent cell lines (data not shown), suggesting that the P27C mutant is unstable, but once incorporated in the complex, it can replace wild type Pen-2.

FIGURE 4.

Amino acids Gly²² and Pro²⁷ of Pen-2 are involved in complex formation and/or stabilization. A, Blue Native PAGE/Western blot analysis of different mutants. Expression of CL Pen-2 in Pen-2^−/− fibroblasts rescues full γ-secretase complex formation (1st lane). In contrast, expression of the G22C mutant compromises complex formation to a large extent (2nd lane). Replacement of the Gly²² with an alanine (G22A) restores complex assembly (3rd lane). Expression of the P27C mutant in Pen-2^−/− fibroblasts results in a partial rescue of γ-secretase complex formation (4th lane). Replacement of the Pro²⁷ with an alanine (P27A) has similar effects as the P27C mutation (5th lane). * indicates the artificial complex composed of PS1 CTF, Aph-1a, and NCT (59). B, SDS-PAGE/Western blot analysis of the Gly²² and Pro²⁷ mutants. The Pen-2 G22C mutant hardly rescues NCT maturation or PS1 endoproteolysis (2nd lane), which is in agreement with the low levels of full complex seen in Blue Native PAGE analysis (A). Replacement of Gly²² by alanine (3rd lane) restores NCT maturation and PS1 endoproteolysis to comparable levels as CL Pen-2 (1st lane). Mutation of the Pro²⁷ to cysteine results in very low Pen-2 levels, probably because of a problem in stability of Pen-2. The Pen-2 P27C and P27A molecules that get assembled in the γ-secretase complex give rise to mature NCT and PS endoproteolysis (4th and 5th lanes). C, cell-free activity assays indicate that the G22C mutant does not rescue AICD production, whereas the G22A mutation rescues AICD production to a large extent. In contrast, the low levels of Pen-2 P27C and Pen-2 P27A that are stably assembled in γ-secretase complexes allow production of equal levels of AICD as the CL Pen-2. AICD levels were corrected for PS1 NTF levels to obtain specific γ-secretase activity. *, p < 0,005.

According to secondary structure predictions for Pen-2 (TOPPRED (50)), the Gly²² and the Pro²⁷ are located in the first helical hydrophobic region of Pen-2 (Fig. 5A). Several studies have shown that glycine and proline residues are special in terms of protein backbone conformation: whereas glycine provides flexibility, proline imposes rigidity due to its side-chain ring structure. Thus, one possible explanation for our results is that the conserved residues Gly²² and Pro²⁷ are part of a critical hinge region in hydrophobic domain 1 important for γ-secretase complex formation and/or stabilization. To test this hypothesis, we replaced Gly²² and Pro²⁷ by an alanine residue, which has a high propensity for helix formation, and therefore, if the available model is correct, if should stabilize the predicted α-helical transmembrane region.

The G22A mutation displayed normal γ-secretase complex levels (Fig. 4A). Furthermore, NCT maturation and PS1 endoproteolysis were rescued to a large extent (Fig. 4B), and cell-free activity assays demonstrated substantial AICD production (Fig. 4C). These results show that the backbone flexibility provided by Gly²² is not essential for the function of the γ-secretase complex and indicate that the mutation of a smaller (glycine) to a larger (cysteine) side chain may sterically disturb a protein-protein interaction involved in the assembly or stability of the complex, for instance at the interface between Pen-2 and PS1 NTF (30, 31).

However, the P27A mutation had the same effect on Pen-2 levels and γ-secretase maturation and activity as the P27C mutation. The low levels of Pen-2 P27A that assemble into the enzyme complex give rise to γ-secretase complexes that display similar specific activity as CL Pen-2, indicating that the mutant assembles into functional γ-secretase complexes (Fig. 4, A–C). Prolines impose important structural restraints on the overall structure of proteins and can be involved in conformational changes (51–54). Moreover, because of its restricted backbone, proline is usually found in irregular structures such as β-turns and α-helical capping motifs (55). Therefore, it seems likely that Pro²⁷ kinks or breaks the helical hydrophobic domain 1 in Pen-2, an effect that is necessary for the stability of both Pen-2 and the γ-secretase complex.

Hydrophobic Domain 1 and the Intracellular Loop of Pen-2 Are Water-accessible from the Lumenal/Extracellular Side

To delineate the solvent accessibility of Pen-2 amino acid residues, we tested the reactivity of the introduced cysteines to EZ-linked biotin-HPDP (HPDP-biotin) (Fig. 5, A and B). HPDP-biotin is a membrane-permeable sulfhydryl-reactive reagent that can only react with free sulfhydryl groups exposed to a hydrophilic environment (Fig. 5E). As expected from the predicted hairpin-like topology of Pen-2 (25), cysteines at the N terminus, cytosolic loop, and C terminus were labeled by HPDP-biotin. Surprisingly, we found that many of the cysteines introduced in hydrophobic domain 1 were water-accessible; cysteines at positions 18–29 were clearly labeled, and cysteines at positions 30–32, 34, and 36 were weakly biotinylated by the HPDP-biotin. In contrast, cysteines at positions 33, 35, and 37 and 38 did not react with the probe. These results indicate that hydrophobic domain 1 is not a classic membrane-embedded α-helix and support a model in which the N-terminal part of this hydrophobic region is unstructured and exposed to a hydrophilic environment, whereas its C-terminal part (Pro²⁷–Arg³⁹) is behaving as a partially accessible transmembrane helix that may be terminated N-terminally by Pro²⁷ as discussed above. In contrast, Ser⁶⁰ to Gln⁷⁹ Pen-2 residues were not labeled by the HPDP-biotin, indicating that the hydrophobic domain 2 is an authentic, 21-amino acid-long transmembrane helix comprising amino acids Ser⁶⁰ to Gln⁷⁹ (Fig. 5, A and B). All biotinylation experiments were repeated at least three times in independent experiments.

Next, we evaluated the reactivity of cysteines to MTSEA-biotin (Fig. 5B). MTSEA-biotin is a membrane-impermeable reagent that can only react with free, hydrophilic cysteines that are accessible from the extracellular site in intact cells (Fig. 5E). As expected, the N and the C termini of Pen-2 were labeled by MTSEA-biotin, confirming their lumenal/extracellular position. Moreover, all residues in the N-terminal part of hydrophobic domain 1 that were labeled by HPDP-biotin were labeled by MTSEA-biotin as well, indicating that the hydrophilic environment to which these amino acid residues are exposed is accessible from the extracellular/lumenal site. A remarkable finding was the labeling of several residues in the loop of Pen-2 by the MTSEA-biotin reagent (Fig. 5B). This result was confirmed using another membrane-impermeable reagent (TS-XX-biotin) (Fig. 5C), which contains, in contrast to the neutral methanethiosulfonate group of MTSEA-biotin, a negatively charged thiosulfate group and is therefore even less likely to cross the cell membrane.

Clearly, many of our results cannot be explained by the current hairpin structure model of Pen-2 (Fig. 6A) (25). Taking into account the predictions for the secondary structure and our accessibility results, we postulate that the N-terminal part of the hydrophobic domain 1 of Pen-2 is unstructured and water-accessible and positioned either as an extension of the lumenal/extracellular N terminus, and therefore not in the membrane at all, or in a hydrophilic cavity in the membrane (Fig. 6, B and C).

FIGURE 6.

Topology models for Pen-2. A, hairpin model for the topology of Pen-2 as proposed by Crystal et al. (25). B and C, two possible models taking into account the results of the water accessibility assays. The loop domain is present in a hydrophilic cavity that is accessible from the extracellular site. The N-terminal part of hydrophobic domain 1 is either located in the membrane in a hydrophilic cavity (B) or it is not part of the transmembrane domain but rather an extension of the N-terminal part of Pen-2 (C). In both cases, the N-terminal part of Pen-2 is present in a constricted environment (gray circles).

To discriminate between these two possibilities, we repeated the MTSEA-biotin experiment with or without pretreatment with MTSES, a large and charged sulfhydryl-specific reagent. In these experimental conditions, a cysteine present in a restricted environment cannot be labeled by the large, charged reagent, and therefore, no competition will be observed with the following MTSEA-biotin labeling. However, if the cysteine is present in a spatially unrestricted environment, MTSES pretreatment results in decreased subsequent MTSEA-biotin labeling.

We performed the competition assay with the Pen-2 F25C mutant of hydrophobic domain 1 and included as well wild type Pen-2, which contains only one endogenous cysteine (Cys¹⁵) and the Pen-2 W85C mutant, with a cysteine at the N and the C terminus, respectively. Both Cys¹⁵ and Trp⁸⁵ are supposed to reside in the unrestricted extracellular space.

As predicted, a clear decrease in labeling was observed in the case of the Pen-2 W85C mutant consistent with the presence of its cysteine in the extracellular space. Interestingly, there was no difference in the MTSEA-biotin labeling in untreated or pretreated Pen-2 F25C cells suggesting that this residue is present in a restricted environment. However and unexpectedly, the labeling of wild type Pen-2 (C15) was also unaffected by treatment with the bulky MTSES (Fig. 5D). This implies that the proximal N terminus of Pen-2 (Cys¹⁵ and Phe²⁵) is present in a constricted aqueous environment, either by chemical-spatial restrictions in a membrane cavity or by the presence of a protein-protein interface, for instance the close proximity of other γ-secretase subunits. This result is in agreement with previous observations that antibodies against the N terminus of Pen-2 fail to co-immunoprecipitate the complete γ-secretase complex (56).

Our data correspond to the observations of Crystal et al. (25) as regards the lumenal position of the N and C termini. However, for the putative loop domain, we demonstrate water accessibility from the extracellular environment. Therefore, we postulate that the loop of Pen-2 is re-entering in a hydrophilic cavity in the membrane. At the first sight, this seems contradictory to the results of Crystal et al. (25), who placed the loop in the cytosol based on observations using a protease protection assay. However, the selective permeabilization used in the proteolysis assay might have disrupted the membrane integrity or might have caused partial destabilization of the core structure of γ-secretase, exposing the loop of Pen-2 to the protease. Another possibility to explain the accessibility of the loop of Pen-2 to impermeable reagents is that the reagents can travel through a channel in γ-secretase, connecting the extracellular space with the cytosol. However, this scenario is not supported by the water accessibility pattern observed for cytosolic cysteines in PS1 (21).

In conclusion, we propose a new model for Pen-2 in which the N and C termini are at the extracellular site, in a constricted and a free environment, respectively. The hydrophobic domain 2 is an α-helical domain spanning the membrane, whereas the hydrophobic domain 1 is partially unstructured and in a restricted environment. The loop of Pen-2 is present in a hydrophilic cavity in the membrane that is accessible from the extracellular environment (Fig. 6, B and C).

Cross-linking Studies Locate the Loop of Pen-2 and PS1 CTF in Close Proximity

The catalytic site of γ-secretase lies at the interface between PS NTF and CTF (16, 18, 20, 57) in a water-containing cavity (18, 20). In agreement, a cryo-EM study reported the existence of low density regions in the membrane core of the γ-secretase that can be interpreted as solvent-accessible pores (58). However, there has not been any indication yet as to the localization of the other subunits in relation to these pores. Our accessibility data provide evidence that Pen-2 might spatially be very close to the catalytic cavity of PS and potentially directly affecting its activity. To investigate the relative position of the loop of Pen-2 and PS1 in the complex, we performed cross-linking assays involving E49C in the loop of Pen-2.

Microsomal membrane fractions from the Pen-2 E49C mutant were treated with SPDP, a heterobifunctional cross-linker with spacer arm of 6.8 Å, at 4 °C. SPDP conjugates primary amine (mainly lysines) and sulfhydryl (cysteines) groups of proteins. After cross-linking, conjugated products were separated by SDS-PAGE in nonreducing conditions and analyzed by Western blotting.

Western blotting against Pen-2 showed a higher mobility band for the Pen-2 E49C mutant (Fig. 7A, 6th lane, arrow). This band was also detected with an antibody against PS1 CTF (Fig. 7B, 6th lane, arrow), but not with antibodies against the other γ-secretase components (NCT, Aph-1, and PS1 NTF) (data not shown). Moreover, the band was not observed when the free sulfhydryls were blocked with the alkylating agent N-ethylmaleimide prior to the cross-linking reaction (Fig. 7, A and B, 5th lane) or when reducing conditions were applied (data not shown), and no cross-linked products were detected with the CL Pen-2 (Fig. 7, A and B, 1st to 3rd lanes). The molecular weight of the band (∼26 kDa) is in accordance with a cross-linking product between Pen-2 and PS1 CTF. Therefore, our results show that the loop of Pen-2 and the PS1 CTF are at a distance equal to or less than 6.8 Å from each other in the γ-secretase complex. Taking into account the previously reported interaction of PS1 NTF (TMD4) with hydrophobic domain 1 of Pen-2, we placed Pen-2 at the interface of PS1 NTF-TMD4 and PS1 CTF with parts of its structure in a hydrophilic cavity.

FIGURE 7.

Cross-linking studies reveal close proximity of the loop of Pen-2 to PS1 CTF. Membrane fractions of Pen-2^−/− fibroblasts rescued with cysteine-less (CL) Pen-2 or cysteine-less Pen-2 E49C were treated with the heterobifunctional cysteine-amine cross-linker SPDP (spacer arm, 6.8 Å). Preblocking of the free sulfhydryl groups with N-ethylmaleimide was performed as negative control. As additional negative control, cross-linker was omitted in the sample (DMSO). Protein extracts were separated in SDS-PAGE under nonreducing conditions on a 4–12% BisTris gel and visualized in Western blot with antibodies against γ-secretase components. A, band with a molecular mass of ∼26 kDa was observed with Pen-2 antibody (arrow). B, same band of 26 kDa was observed with PS1 CTF antibody (arrow).

Cryo-EM studies demonstrated two low density areas in the γ-secretase complex, which were suggested to be water-containing cavities: the catalytic cavity and the substrate-binding cavity (Fig. 8A) (58). Combined with the fact that modifications of Pen-2 result in changes in the Aβ₄₀ to Aβ₄₂ ratio (32), we postulate that Pen-2 participates in one of the two functional cavities of γ-secretase (Fig. 8, B and C).

FIGURE 8.

Model for the topology of Pen-2 and the interaction between Pen-2 and PS1. A, model for PS1 containing the catalytic site (18–21, 60). PS1 has 10 hydrophobic domains from which nine cross the membrane as transmembrane domains (TMDs, indicated by the bundle of numbered cylinders). The other one is present in the large loop between TMD6 and TMD7 and contains the endoproteolytic site (HDVII). In the figure, HDVII has undergone endoproteolysis and is divided in two parts. TMD6 and TMD7 delineate a water-containing cavity inside the membrane with the two catalytic aspartates facing each other. TMD9 and HDVII participate in the formation of this cavity as well. Water accessibility data of PS1 suggest that the catalytic cavity is open toward the cytosol (black line). Furthermore, an initial substrate-binding site has been proposed to allow lateral gating of the hydrophobic substrates into the hydrophilic catalytic cavity. TMD9 has been suggested to be part of this substrate binding site (pale line). Therefore, TMD9 is part of both the substrate-binding site and the catalytic site. Moreover, its flexible nature, induced by the PAL motif, is proposed to be important for the gating mechanism. B and C, model for the interaction of Pen-2 and PS1 as a drawn-open (B) and closed (C) picture. Hydrophobic domain 1 of Pen-2 (I) is supposed to interact with TMD4 of PS1 NTF (30, 31), whereas our cross-linking data position the loop of Pen-2 close to PS1 CTF. Furthermore, our water accessibility data of Pen-2 reveal that the loop of Pen-2 is accessible from the lumenal site. Similarly, the N-terminal part of TMD9 of PS1 containing the PAL motif, is accessible to impermeable sulfhydryl-reactive reagents as well and therefore exhibits a comparable accessibility pattern as the loop of Pen-2 (21). Therefore, we draw the loop of Pen-2 (red) in close proximity to PS1 TMD9. The N terminus of Pen-2 is present in a constricted environment indicated by the gray circles.

Conclusions

Our water accessibility and cross-linking data for hydrophobic domain 1 and the loop domain of Pen-2 reveal an unexpected topology for Pen-2 and a close proximity to PS1 CTF. Together with the previously shown interaction of PS1 NTF with the hydrophobic domain 1 of Pen-2 (30, 31), these results position Pen-2 at the interface of the PS NTF TMD4 and PS CTF fragments. Until now, only Pen-2 and PS1 modifications have been implicated in changes in γ-secretase activity (Aβ₄₂/Aβ₄₀ ratio changes) (32, 37), implying that Pen-2 plays a functional role in the γ-secretase activity. The complete lack of γ-secretase activity we observed, despite the clear formation of a trimeric complex (PS-NCT-Aph-1a), in the Pen-2^−/− cells containing either wild type PS1 or the endoproteolytically deficient PS1 ΔE9 mutant, further corroborates this notion. Taking into account that the cytosolic part of PS1 TMD9 is accessible to membrane-impermeable sulfhydryl-reactive reagents (21), and therefore displays a similar water accessibility pattern as the loop of Pen-2, and that PS1 TMD9 is proposed to be involved in the gating mechanism of γ-secretase, we suggest that Pen-2 is part of the dynamic gating mechanism that involves PS1 CTF (Fig. 8, B and C).