Co-expressed presenilin 1 NTF and CTF form functional γ-secretase complexes in cells devoid of full-length protein


Address correspondence and reprint requests to Jan Näslund, Karolinska Institutet, Novum Kaspac pl. 5, SE-141 57 Huddinge, Sweden. E-mail:


The enzyme γ-secretase catalyzes the intramembrane proteolytic cleavage that generates the amyloid β-peptide from the β-amyloid precursor protein. The presenilin (PS) protein is one of the four integral membrane protein components of the mature γ-secretase complex. The PS protein is itself subjected to endoproteolytic processing, generating stable N- and C-terminal fragment (NTF and CTF, respectively) heterodimers. Here we demonstrate that coexpression of PS1 NTF and CTF functionally mimics expression of the full-length PS1 protein and restores γ-secretase activity in PS-deficient mammalian cells. The coexpressed fragments re-associate with each other inside the cell, where they also interact with nicastrin, another γ-secretase complex component. Analysis of γ-secretase activity following the expression of mutant forms of NTF and CTF, under conditions bypassing endoproteolysis, indicated that the putatively catalytic Asp257 and Asp385 residues have a direct effect on γ-secretase activity. Moreover, we demonstrate that expression of the wild-type CTF rescues endoproteolytic cleavage of C-terminally truncated PS1 molecules that are otherwise uncleaved and inactive. Recovery of cleavage is critically dependent on the integrity of Asp385. Taken together, our findings indicate that ectopically expressed NTF and CTF restore functional γ-secretase complexes and that the presence of full-length PS1 is not a requirement for proper complex assembly.

Abbreviations used

amyloid β-peptide


Alzheimer's disease


APP intracellular domain


β-amyloid precursor protein


C-terminal fragment


Gal4 DNA-binding domain and VP16 transactivating domain


Notch intracellular domain


N-terminal fragment


presenilin 1


presenilin 2

γ-Secretase cleavage is an intramembrane proteolysis event, i.e. the enzyme cleaves the peptide bond within the transmembrane domain of its substrate (Wolfe and Selkoe 2002). Regulated intramembrane processing has recently been recognized as a general phenomenon for transducing a signal from the exterior to the interior of the cell through proteolysis (Brown et al. 2000), and the number of type I transmembrane proteins processed by this pathway is increasing rapidly (De Strooper 2003). γ-Secretase catalyzes the final step in the generation of the amyloid β-peptide (Aβ) from C99, an intermediate β-amyloid precursor protein (APP) fragment. Cerebral accumulation of the amyloid β-peptide (Aβ) is a central event in the pathogenic cascade leading to Alzheimer's disease (AD) (Selkoe 2001). γ-Secretase cleavage also determines the relative proportion of the Aβ40 and Aβ42 forms (Esler and Wolfe 2001). The longer and less abundant Aβ42 form is more prone to form fibrillar aggregates (Jarrett et al. 1993), and has been shown to be deposited early in AD brain (Iwatsubo et al. 1994). Concomitant with γ-secretase-mediated Aβ formation, a soluble C-terminal APP intracellular fragment is generated. This APP intracellular domain (AICD) associates with the cytosolic adaptor protein Fe65, translocates to the nucleus, and functions as a regulator of transcription (Cao and Südhof 2001). Surprisingly, the N-terminus of AICD is located 8–10 residues distal to the site predicted by the γ-secretase cleavages generating Aβ40 and Aβ42 (Sastre et al. 2001; Yu et al. 2001; Chen et al. 2002; Weidemann et al. 2002). This novel so-called ε-cleavage, occurring between Leu49 and Val50, resembles the cleavage generating the Notch intracellular domain (NICD) from the transmembrane and ligand-activated Notch precursor (Ebino and Yankner 2002). Competition studies in vitro suggest that cleavage at the ε site is intimately related to γ-secretase cleavage (Kimberly et al. 2003).

Studies have shown that PS proteins (PS1 and PS2) are required in order for γ-secretase cleavage to occur (Herreman et al. 2000; Zhang et al. 2000) and that PS1 is present in the high molecular weight complex that constitutes γ-secretase (Seeger et al. 1997; Li et al. 2000a). PS proteins contain 10 hydrophobic domains (HD), 7 or 8 of which form transmembrane domains according to most current models (Doan et al. 1996; Li and Greenwald 1998). Endogenously expressed PS is endoproteolytically cleaved in the large cytosolic loop, thus generating a 25-kDa N-terminal fragment (NTF) and an 18-kDa C-terminal fragment (CTF) (Thinakaran et al. 1996). The NTF and CTF remain associated and are incorporated into the active γ-secretase complex (Li et al. 2000a). Three other transmembrane proteins have been shown to act as PS partners in the γ-secretase complex: nicastrin (Yu et al. 2000), Aph-1 (Francis et al. 2002; Goutte et al. 2002), and Pen-2 (Francis et al. 2002). Expression of all four γ-secretase components is required for γ-secretase activity in cells (Edbauer et al. 2003; Kimberley et al. 2003; Takasugi et al. 2003). Accumulating evidence strongly suggests that PS provides the catalytic core of the γ-secretase complex. First, mutations of two conserved aspartate residues, Asp257 and Asp385, in transmembrane domains 6 and 7 of PS1 (corresponding to HD6 and HD8, respectively), dramatically reduce Aβ formation and PS1 endoproteolysis (Wolfe et al. 1999). Second, transition-state analogue inhibitors of γ-secretase bind specifically to the PS1 NTF and CTF heterodimers (Esler et al. 2000; Li et al. 2000b), where each subunit contributes one of the two critical aspartates. Finally, PS contains an aspartic protease motif (Steiner et al. 2000). Together, these arguments strongly implicate PS as the catalytically active component of the γ-secretase complex and that γ-secretase is an intramembrane-cleaving aspartyl protease. It should be noted, however, that PS-independent Aβ generation has also been observed (Wilson et al. 2002).

As noted above, the biologically active form of PS is the NTF and CTF heterodimer. However, endoproteolysis of PS is not an absolute requirement for PS bioactivity as shown by the familial AD-linked PS1ΔE9 mutation (Perez-Tur et al. 1995). PS1ΔE9 is not endoproteolytically processed due to the lack of the exon encoding the cleavage site (Thinakaran et al. 1996), but appears to functionally replace endogenous PS if overexpressed in cells (Thinakaran et al. 1997) and rescue an egg-laying defect in Caenorhabditis elegans lacking SEL-12 (Levitan et al. 1996). In addition, introduction of an artificial mutation in the endoproteolytic cleavage site in PS1 (M292D) inhibits PS endoproteolysis without affecting Aβ generation (Steiner et al. 1999). Endoproteolysis of PS1 has been proposed to be an autocatalytic event involving the Asp257 and Asp385 residues (Wolfe et al. 1999). However, the difference in the pharmacological profile between the intramembrane cleavage generating Aβ and the cleavage generating NTF and CTF suggests the existence of a distinct presenilinase (Campbell et al. 2002; Campbell et al. 2003).

Ectopic expression of NTF in cells containing endogenous PS results in rapid degradation of the exogenous fragment, presumably due to its lack of incorporation into stable complexes (Steiner et al. 1998; Saura et al. 1999). However, a recent study demonstrated that the PS-dependent egg-laying defect in C. elegans sel-12 mutants (Levitan et al. 2001) could be rescued by coexpression of the PS1 NTF and CTF. In this report we wanted to confirm and extend these results in mammalian cells devoid of endogenous PS1 and PS2 in order to directly study the bioactivity of NTF and CTF. We present data demonstrating that coexpression of NTF and CTF in PS-deficient cells reconstitutes both Aβ and AICD formation from a C99 reporter molecule. The exogenous NTF and CTF form functional γ-secretase complexes in PS-deficient cells as evidenced by the fragment's ability to form heterodimers and associate with endogenous nicastrin. Hence, functional γ-secretase complexes can assemble in the absence of full-length, un-cleaved PS. Expression of mutant NTF and CTF indicated that Asp257 and Asp385, the two proposed catalytic residues in PS1, have a direct effect on processing of the reporter molecule. We also show that C-terminally truncated PS1 constructs are endoproteolysis-defective and are unable to mediate γ-secretase activity in PS-deficient cells. Both endoproteolysis and γ-secretase activity can be rescued if these truncated PS molecules are coexpressed together with CTF, and this effect is critically dependent on the integrity of Asp385.

Materials and methods

Expression constructs

The reporter constructs that were used in the luciferase-based reporter system to monitor AICD- and NICD-generation have previously been described (Karlström et al. 2002; Taniguchi et al. 2002; Bergman et al. 2003). Briefly, a combined domain encoding the Gal4 DNA-binding domain and the transactivating VP16 domain (together abbreviated GVP) was incorporated into constructs designed to mimic the β-secretase processed APP fragment (C99), and the ligand activated site two-cleaved Notch-1 receptor (ΔEN1), respectively. The GVP-domain was introduced close to the transmembrane region of C99 between the triple lysine anchor and the remainder of the cytoplasmic tail to generate C99-GVP (Karlström et al. 2002). Similarly ΔEN1-GVP was generated by integration of the GVP-domain into the intracellular portion of the ΔEN1 (Taniguchi et al. 2002).

The UAS-luciferase reporter construct (MH100) and the β-galactosidase encoding CMV-βgal plasmid have been described elsewhere (Taniguchi et al. 2002). Full-length PS1 and PS1 NTF were in the pcDNA3 backbone, and the CTF was in pcDNA3.1. Mutations were introduced by site directed mutagenesis using the Quick Change™ Site Directed Mutagenesis Kit according to the manufacturer's protocol (Stratagene). The DNA sequence of each construct was verified using the DYEnamic ET Terminator Cycle Sequencing kit (Amersham Pharmacia Biotech) and the ABI Prism 377 sequencer (Perkin Elmer).

Cell culture and transfection

Blastocyst derived embryonic stem cells lacking PS1 and PS2, BD8 cells (Donoviel et al. 1999), were cultured in DMEM supplemented with 10% fetal bovine serum, 1 mm sodium pyruvate, 2.4 mm l-glutamine, 0.1 mmβ-mercaptoethanol, and nonessential amino acids. All transfections were carried out using Lipofectamine PLUS according to the manufacturer's instructions (Invitrogen, Carlsbad, CA, USA). Serum free transfection medium (DMEM) was exchanged for culture medium after 2.5 h and the cells were grown for 24 h. γ-Secretase inhibitor L-685 458 (Calbiochem, San Diego, CA, USA) was included in the medium during recovery phase when indicated.

Luciferase-based reporter assays and immunoblotting

Transfections for the luciferase-based reporter assays were carried out in 24-well tissue culture plates. Cells were transiently transfected with a mixture of constructs coding for the reporter molecule (MH100), one of the substrates (C99-GVP or ΔEN1-GVP), PS1 (full-length or fragments) and β-galactosidase. Per well, cells were transfected with 100 ng MH100, 50 ng CMV-βgal, 100 ng of either C99-GVP or ΔEN1-GVP, and vectors encoding full-length PS1 (100 ng) or PS1 fragments (100 ng + 100 ng). Empty pcDNA3.1 vector (100 ng) was added to adjust for differences in DNA amounts. Cells were lysed in 100 µL lysis buffer per well (10 mm Tris, pH 8, 1 mm EDTA, 150 mm NaCl and 0.65% NP40) and luciferase activity was monitored luminometrically after addition of luciferin and ATP (BioThema, Haninge, Sweden). The β-galactosidase activities of the cell lysates were determined to equalize for differences in transfection efficiencies. Arbitrary β-galactosidase normalized luminescence units are presented as percent activity compared to wild-type PS1 response.

For immunoblotting, cell lysates were briefly sonicated and resolved on 10–20% Tricine gels. Proteins were transferred to nitrocellulose membranes that were subsequently probed with appropriate antibodies. The immunoblots were developed using horseradish peroxidase-conjugated secondary antibodies and ECL substrate (Pierce, Rockford, IL, USA).


The antibodies used for coimmunoprecipitations were polyclonal antibody Ab14, raised against the 25 N-terminal residues of human PS1 (Seeger et al. 1997), and monoclonal antibody MAB5232 (Chemicon, Temecula, CA, USA), which recognizes PS1 CTF. For immunoblotting, monoclonal antibody NT1, raised against the N-terminus of PS1 (Mathews et al. 2000), and polyclonal antibody AB5308 (Chemicon) recognizing PS1 CTF, were used. C99-GVP and myc-tagged ΔEN1-GVP were detected using antibodies 369 (Buxbaum et al. 1990) and 9E10 (Pharmingen, BD Bioscience, Heidelberg, Germany), respectively. Nicastrin was immunoprecipitated using antibody N1660 (Sigma, St Louis, MO, USA).


Transient transfections were performed in six-well plates in triplicates. Per well, cells were transiently transfected with 500 ng C99-GVP encoding vector and 500 ng of either the full-length PS1 construct or 250 ng PS1 NTF and 250 ng PS1 CTF. Cells were cultured for 24 h in 0.75 mL Opti-MEM supplemented with 5% fetal bovine serum. Cell media were flash frozen in liquid nitrogen, and subjected to sandwich ELISA measurements, essentially as described (Mathews et al. 2002).


Co-immunoprecipitations were performed on lysates from transiently transfected cells. Transfections were carried out in a six-well format (1 µg DNA per well), and the cells were allowed to recover for 48 h. Per well, cells were lysed in 0.5 mL coimmunoprecipitation buffer [50 mm Hepes, pH 7.4, 150 mm NaCl, 2 mm EDTA, Protease Inhibitor Cocktail (Roche, Indianapolis, IN, USA) and 1% CHAPSO] and briefly sonicated. Lysates were subjected to ultracentrifugation (100 000 × g, 20 min). The CHAPSO-soluble supernatant was incubated with primary antibody (dilution 1 : 300) over night and proteins were pulled down with a mixture of protein A and G Sepharose (Amersham). Monoclonal α-His antibody or preimmune sera were used as negative controls. Immunoprecipitated proteins were resolved on 10–20% Tricine gels (Invitrogen), transferred to nitrocelluose membranes and detected by immunoblotting as described above.


Co-expression of PS1 NTF and CTF rescues γ-secretase activity and AICD/NICD formation from reporter constructs in PS-deficient cells

Co-expression of NTF and CTF has previously been reported to rescue a PS-dependent egg-laying defect in C. elegans sel-12 mutants (Levitan et al. 2001). We first wanted to corroborate these findings in mammalian cells lacking both PS1 and PS2 (Donoviel et al. 1999). To assay γ-secretase activity and AICD formation in PS-deficient cells, we made use of a C99 construct containing a Gal4 DNA-binding domain and VP16 transactivating domain (together abbreviated GVP) (Karlström et al. 2002). After intramembrane processing of C99, the AICD fragment containing the GVP moiety translocates to the nucleus and initiates expression of the luciferase reporter gene. The use of this chimeric C99 substrate has two distinct advantages, one being that AICD formation can be faithfully quantified and the other that the presence of the GVP domain reduces the cytosolic degradation of the AICD fragment, allowing for detection of intracellular AICD-GVP by immunoblotting. The luciferase assay does not distinguish between γ- and ε-cleavage-mediated AICD formation. However, in order for AICD-GVP to elicit a luciferase response it must be soluble, and all soluble AICD fragments analyzed to date start at a site corresponding to the ε site (Sastre et al. 2001; Yu et al. 2001; Chen et al. 2002; Weidemann et al. 2002).

Blastocyst-derived PS-deficient cells (Donoviel et al. 1999) were transfected with vectors encoding full-length PS1 or the endoproteolytic NTF and CTF polypeptides. AICD generation from the C99-GVP reporter molecule was quantified by bioluminescence spectroscopy 24 h after transfection. Transfection of the PS-deficient cells with full-length PS1 resulted in robust intramembrane proteolysis of C99-GVP and subsequent AICD formation (Fig. 1a). Transfection of NTF or CTF alone did not rescue intramembrane proteolysis of C99-GVP. However, coexpression of NTF and CTF restored γ-secretase activity and AICD generation in the PS-deficient cells (Fig. 1a). These findings were confirmed by immunoblotting of lysates, where an AICD-GVP fragment was detected from cells transfected with full-length PS1 or the coexpressed NTF and CTF (Fig. 1b). Intramembrane processing of C99-GVP was dramatically reduced if NTF and CTF were coexpressed in the presence of the transition state mimic L-685 458 γ-secretase inhibitor (Fig. 1a) (Shearman et al. 2000).

Figure 1.

Co-expression of the PS1 endoproteolytic fragments in PS-deficient cells rescues intramembrane proteolytic processing of C99-GVP and ΔEN1-GVP. (a) A luciferase-based reporter assay was used to monitor the intramembrane processing of C99-GVP and formation of AICD-GVP in PS-deficient cells. Transfection of full-length (FL) PS1 or cotransfection of NTF and CTF rescued intramembrane processing of C99-GVP. The cleavage activity recorded from cells cotransfected with NTF and CTF was inhibited by treating the cells with the γ-secretase inhibitor L-685 458 (5 µm). Expression of NTF or CTF alone did not induce intramembrane cleavage above background (C99-GVP alone). The luciferase-induced bioluminescence corresponding to γ-secretase activity is presented in the bar graphs as percent activity compared to full-length PS1. Error bars represent standard error of the mean (SEM) for triplicate samples. (b) Expression of FL PS1 and the PS1 endoproteolytic fragments was confirmed by immunoblotting. C99-GVP expression and the generation of the derived fragments (C83-GVP and AICD-GVP) were also visualized by immunoblotting. Note that AICD-GVP generation (indicated by open circles; lanes 2 and 5) corresponds well to luciferase response in the reporter assay (see Fig. 1a), whereas C99- and C83-GVP accumulates in the absence of functional PS or in the presence of L-685 458 (lanes 1,3,4 and 6). (c) Intramembrane processing of the Notch reporter molecule, ΔEN1-GVP, was assayed as in (a). Bar graphs indicate percent activity compared to full-length PS1. Error bars represent standard error of the mean (SEM) for triplicate samples.

PS proteins also control intramembrane processing of Notch receptors. In the absence of PS, no NICD signaling is recorded from a ΔEN1-GVP reporter molecule (Karlström et al. 2002), that mimicks a ligand-activated and ectodomain-shedded Notch-1 protein (Schroeter et al. 1998; Mumm et al. 2000). Ectopic expression of full-length PS1 as well as coexpression of PS1 NTF and CTF in PS-deficient cells restores NICD signaling from ΔEN1-GVP (Fig. 1c). Thus, intramembrane processing of both APP and Notch can be faithfully reproduced by concurrent expression of the PS1 endoproteolytic fragments.

Functional γ-secretase complex formation by coexpressed NTF and CTF and authentic secretion of Aβ40 and Aβ42

In cells, the NTF and CTF remain associated as heterodimers together with other proteins, among them nicastrin, in high molecular weight complexes (Yu et al. 2000). PS has been shown to be obligatory for nicastrin maturation (Edbauer et al. 2002; Leem et al. 2002), and ablation of nicastrin reduces PS stability (Edbauer et al. 2002; Leem et al. 2002). To ascertain that coexpression of NTF and CTF also results in functional complex formation, coimmunoprecipitations were performed on lysates from PS-deficient cells transfected with full-length PS1 or the endoproteolytic fragments. Immunoprecipitations using an antibody against CTF coprecipitated the NTF in cells expressing full-length PS1 or cells coexpressing NTF and CTF (Fig. 2a, lanes 2, 5, upper panel). Reciprocal immunoprecipitations against the NTF pulled down the CTF (Fig. 2a, lanes 2, 5, lower panel). These results were extended by immunoprecipitations using an antibody against the C-terminal part of nicastrin. PS1 NTF and CTF were only associated with nicastrin in cells transfected with full-length PS1 or the NTF and CTF combined (Fig. 2b, lanes 2 and 5). Collectively, these results suggest that coexpression of PS1 NTF and CTF functionally mimics expression of full-length PS1 in PS-deficient cells in terms of γ-secretase complex assembly and activity.

Figure 2.

Co-expressed PS1 NTF and CTF are associated with each other and to nicastrin and induce authentic Aβ-generation in PS-deficient cells. (a) NTF is coimmunoprecipitated with CTF in PS-deficient cells transfected with either FL-PS1 or both of the endoproteolytic fragments (lanes 2 and 5). In reciprocal experiments, CTF was coimmunoprecipitated with NTF. (b) The NTF and CTF are coimmunoprecipitated with endogenous nicastrin (NCT) in cells transfected with either FL PS1 or the endoproteolytic fragments (lanes 2 and 5). Note that neither the NTF nor the CTF coprecipitates with NCT, when expressed in the absence of the CTF. (c) Aβ40- and Aβ42-specific ELISAs confirmed authentic Aβ production in cells transfected with the endoproteolytic fragments. Aβ was not detected in cells transfected with vector alone (not shown). Graphs represent Aβ40 levels (white bars) and Aβ42 levels (black bars) in conditioned media collected 48 h after transfection. Error bars indicate SEM for triplicate transfections. (d) The ratios between the levels of Aβ42 and Aβ40 do not differ between media collected from PS1 FL transfected cells, or cells transfected with the NTF and the CTF.

Assessment of ε-cleavage-mediated AICD formation determines only one of the two cleavage events involved in γ-secretase-mediated intramembrane processing of APP. Therefore, γ-cleavage activity was directly assayed by measuring Aβ production from the C99-GVP substrate. Levels of the two main secreted Aβ40 and Aβ42 variants in conditioned media were determined by sandwich ELISAs. Expectedly, expression of either fragment alone did not result in Aβ secretion (not shown). However, coexpression of NTF and CTF re-established secretion of Aβ, albeit at lower levels compared to cells transfected with full-length PS1 (Fig. 2c). No significant change in the Aβ42/Aβ40 ratio was observed between cells expressing full-length PS1 and the endoproteolytic fragments (Fig. 2d). Hence, expression of PS1 in the endoproteolytic fragment form in PS-deficient cells restores γ-secretase-mediated secretion of both Aβ40 and Aβ42 from the C99-GVP reporter molecule.

Asp257 and Asp385 in PS1 are critical for γ-secretase activity

Two highly conserved aspartate residues in PS1, Asp257 in transmembrane domain 6 and Asp385 in transmembrane domain 7, have been shown to be critical for endoproteolysis and γ-secretase activity (Wolfe et al. 1999). Because the cointroduction of NTF and CTF described here bypasses the endoproteolysis step, we wanted to study the direct effect of these mutations on ε-cleavage and AICD formation. The D257A and D385A mutations were introduced in the NTF and CTF, respectively, and expressed in PS-deficient cells. As seen, AICD generation was abolished whenever an aspartate-mutant fragment was present (Fig. 3a), consistent with similar studies performed in C. elegans sel-12 mutants by Levitan et al. (2001). The inability of the aspartate-mutant fragments to rescue γ-secretase activity was not due to lack of expression, as assessed by immunoblotting (Fig. 3b). The aspartate mutations did not appear to affect heterodimer formation, since immunoprecipitations using an antibody against CTF coprecipitated NTF regardless of whether the mutation was present in the NTF or the CTF (Fig. 3c). Taken together, these results suggest that the D257A and D385A mutations, when PS1 is expressed under conditions bypassing endoproteolysis, affect PS1-mediated γ-secretase activity directly and not by interfering with NTF and CTF association. This was also recently confirmed in PS null mouse embryonic fibroblasts stably expressing D275A and D385A mutants (Nyabi et al. 2003).

Figure 3.

The conserved Asp257 and Asp385 residues are critical for intramembrane proteolysis of C99-GVP independently of their effect on endoproteolysis. (a) The luciferase response corresponding to AICD-generation was inhibited when the dominant negative mutations D257A and D385A were introduced into the endoproteolytic fragments. Co-expression of mutant NTF (D257A) and native CTF (wt) or native NTF (wt) and mutant CTF (D385A) failed to rescue activity, as did coexpression of both mutant fragments. The bars represent luciferase activity presented as percent activity compared to cells transfected with wild-type PS1 endoproteolytic fragments. Error bars indicate SEM for triplicate transfections. (b) Expressed PS1 fragments and the various GVP fragments were detected by immunoblotting of cell lysates. Note the accumulation of C99- and C83-GVP in lysates from cells expressing mutant fragments (lanes 3–5) and the appearance of a band corresponding to AICD-GVP (indicated by an open circle) in lysates from cells expressing the wild-type endoproteolytic fragments (lane 2). (c) Co-immunoprecipitations with an antibody towards PS1 CTF pulled down the NTF, regardless of whether mutations were present or not.

The CTF is functionally involved in PS1 endoproteolysis and γ-secretase activity

Encouraged by the finding that coexpressed NTF and CTF rescued γ-secretase activity in PS-deficient cells, we set out to extend these findings by studying the role of NTF and CTF in γ-secretase complex formation. We found that modifications of the C-terminus of PS1 affected γ-secretase activity. Truncated PS1 constructs terminating after HD8 (HD8stop) and HD9 (HD9stop) could not catalyze intramembrane proteolysis and AICD generation from the C99-GVP reporter molecule (Fig. 4a). In addition, neither HD8stop nor HD9stop underwent endoproteolytic cleavage (Fig. 4b). As we had shown that expression of CTF in the presence of native NTF functionally substitute for expression of full-length PS1, we next asked whether CTF, when coexpressed with C-terminally truncated and non-functional PS1 molecules, could rescue γ-secretase activity. Strikingly, coexpression of HD8stop and HD9stop together with the native CTF reconstituted AICD formation (Fig. 4a). Furthermore, expression of CTF enabled endoproteolytic processing of HD8stop and HD9stop (Fig. 4b). Thus, it appears that re association of the endoproteolysis-defective HD8stop and HD9stop molecules with the CTF triggers endoproteolytic processing of the former molecules and assembly of a functional γ-secretase complex.

Figure 4.

The C-terminal end of PS1 is required for both PS1 endoproteolysis and C99-GVP intramembrane processing. (a) Either FL PS1 or PS1 variants truncated either after HD9 (HD9stop) or HD8 (HD8stop) were expressed either with or without wild-type CTF and assayed with the luciferase-based reporter system. Intramembrane processing of C99-GVP is severely reduced in cells transfected with either of the C-terminally truncated constructs alone. However, upon cotransfection with the CTF, activity is restored. Bars represent percent activity as compared to the response for FL PS1, and error bars represent SEM for triplicate transfections. (b) Immunoblotting revealed that endoproteolysis is inhibited for the C-terminally truncated constructs when expressed in PS-deficient cells. Co-transfection with wild-type CTF triggers endoproteolysis.

The Asp385 residue is critical for PS1 endoproteolysis

The results shown above indicated that the endoproteolysis-defective and non-functional HD9stop molecule could be used as a tool in studying endoproteolysis in PS-deficient cells. In this regard, we wanted to directly address the role of the CTF Asp385 residue in the endoproteolytic cleavage of PS1. PS-deficient cells were cotransfected with HD9stop and a CTF containing the D385A mutation. Cells were lysed 24 h after transfection, and as indicated in Fig. 5(a), coexpression of HD9stop and mutant CTF did not restore intramembrane proteolysis and AICD formation from the C99-GVP reporter molecule. Importantly, the CTF containing the D385A mutation also failed to rescue endoproteolysis of the HD9stop molecule (Fig. 5b). These results indicate that Asp385, in addition to its role in intramembrane proteolysis, is required for endoproteolytic activity.

Figure 5.

Effect of Asp385 on endoproteolysis of the truncated PS1 construct HD9stop. (a) The inactive and endoproteolysis-defective HD9stop fragment was coexpressed with wild-type CTF, mutant CTF (D385A) or PS1ΔE9 and intramembrane proteolysis of C99-GVP was assayed using the luciferase reporter system. Co-expression of HD9stop and mutant CTF did not restore intramembrane proteolysis. The ability of PS1ΔE9 to mediate processing of C99-GVP was not affected by cotransfection of HD9stop. Bars represent percent activity as compared to the response for FL PS1, and error bars represent SEM for triplicate transfections. (b) Immunoblotting of cell lysates confirmed that the wild-type CTF enabled endoproteolytic cleavage of HD9stop, whereas no endoproteolysis was seen in the presence of the mutant CTF or PS1ΔE9. (c) Functional complex formation of HD9stop in the absence or presence of CTF was assessed by reciprocal coimmunoprecipitations. L = lysate (4%), IP = Immunoprecipitate, and C = control (immunprecipitation using preimmune sera). (d) Co-immunprecipitations of PS1ΔE9 in the absence or presence of HD9stop. The lower circle indicates PS1ΔE9 and the upper circle indicate HD9stop.

In the previous experiments we had studied the effects of expression of ‘free’ CTF, therefore we asked whether a CTF that was still attached to the N-terminal part of PS1 could rescue endoproteolysis of the HD9stop fragment. To allow detection of HD9stop endoproteolysis we made use the non-cleavable PS1ΔE9 molecule, instead of full-length PS1, as donor of a native CTF domain. As seen in Fig. 5(a), PS1ΔE9 mediates AICD generation in PS-deficient cells. However, no endoproteolysis of HD9stop was observed in the presence of PS1ΔE9 (Fig. 5b). Co-transfection of HD9stop and PS1ΔE9 did not affect the latter molecule's ability to restore γ-secretase activity (Fig. 5a). The HD9stop molecule did not assemble into functional complexes in the absence of the CTF, as determined by coimmunoprecipitations (Fig. 5c). Moreover, HD9stop assembled into the complexes in its uncleaved form (Fig. 5c, lanes 5 and 8). Only the PS1ΔE9 molecule associated with nicastrin (Fig. 5d, lane 4).


The assembly and maturation of a fully functional γ-secretase complex is an intricate process involving at least four different protein components and the endoproteolytic cleavage of PS to yield the NTF and CTF. These proteins influence each other's stability, maturation, and trafficking (as reviewed in Ref. (De Strooper 2003). The PS moiety is likely to provide the catalytic subunit of the complex (this study, and Wolfe and Selkoe 2002). Putative functions for the other members of the complex have recently been proposed. For instance, overexpression of Aph-1 stabilizes the full-length PS protein (Takasugi et al. 2003). Full-length PS also accumulates when Pen-2 is downregulated, whereas overexpression of Pen-2 promotes the formation of NTF and CTF (Luo et al. 2003; Takasugi et al. 2003). This suggests that Pen-2 plays a central role in PS endoproteolysis, although further work is required to show how Pen-2 modulates this reaction. Interestingly, Pen-2 also associates with the PS1ΔE9 molecule (Steiner et al. 2002), which is not endoproteolytically processed.

Here we show that coexpression of the PS1 NTF and CTF in blastocyst-derived embryonic stem cells lacking both PS1 and PS2 (Donoviel et al. 1999) functionally substituted for expression of full-length PS1. Co-expression of NTF and CTF restored γ-cleavage, generating Aβ40/42, and ε-cleavage, generating AICD, from the C99 reporter construct. NICD signaling from the Notch ΔEN1 fragment could also be rescued by simultaneous expression of NTF and CTF. Significantly, coimmunoprecipitations demonstrated that the fragments associated within the cell and assembled with endogenous nicastrin, another integral component of the γ-secretase complex. These results in mammalian cells confirm and extend the studies performed in C. elegans by Levitan et al. (2001), where the NTF and CTF rescued the sel-12 mutant phenotype. Collectively, these results indicate that the presence of full-length PS1 is not a requirement for functional γ-secretase complex assembly, since coexpressed NTF and CTF can replace the full-length molecule. However, PS endoproteolysis may serve as an important biochemical checkpoint for regulating the amount and quality of γ-secretase complexes generated.

The putatively catalytic Asp257 in HD6 and Asp385 in HD8 affect both γ-secretase processing and PS1 endoproteolysis (Wolfe et al. 1999). By using the system described here we could distinguish between these two proteolytic processes. Whenever an aspartate was substituted for alanine, either D257A in the NTF or D385A in the CTF, intramembrane proteolysis of C99-GVP was abolished. Co-immunoprecipitations indicated, however, that neither mutation disrupted the formation of the NTF and CTF heterodimer complex. Consequently, when studying these mutations under conditions that distinguish effects on endoproteolysis from effects on γ-secretase processing, it appears that the conserved aspartates have a direct role in the catalytic process, in keeping with the original hypothesis proposed by Wolfe et al. (1999). In addition, pharmacological data imply that PS endoproteolysis is an autocatalytic event (Beher et al. 2001).

We examined the functional role of the PS1 CTF further by generating the C-terminally truncated PS1 molecules HD8stop and HD9stop. These molecules were found to be defective both in mediating γ-secretase cleavage and in being endoproteolytically processed. Although the critical intramembrane aspartate residues are present, both constructs lack the so-called PALP motif, where the first proline residue is important for endoproteolysis and stabilization (Tomita et al. 2001). Our findings are in accordance with previous studies emphasizing the importance of the integrity of the PS1 C-terminus (Tomita et al. 1999). Whether the truncations render the PS1 molecule inactive due to misfolding of the C-terminal domain or whether an interaction site with the other γ-secretase components is disrupted remains to be shown. Importantly, coexpression of HD8stop and HD9stop with CTF resulted in the conversion of both truncated molecules into functional NTFs, and restored intramembrane processing of C99-GVP. One possibility is that the expressed CTF may interact and assemble with the N-terminal portion of the truncated constructs prior to γ-secretase complex assembly and endoproteolysis. Alternatively, the coexpressed CTF may displace the non-functional C-terminal domain of the truncated constructs in the nascent γ-secretase complex. Nevertheless, the wild-type PS1 CTF can functionally substitute for an inactive C-terminal domain, resulting in endoproteolysis of the truncated PS1 constructs and γ-secretase activity.

Finally, we made use of the HD9stop fragment as an ‘endoproteolysis probe’ in a set of experiments where we wanted to investigate whether a CTF containing the D385A mutation or an intact C-terminal domain that remains linked to the N-terminal domain (i.e. PS1ΔE9) could support endoproteolysis of HD9stop. Co-expression of HD9stop and the mutant CTF did not lead to endoproteolysis, and, as would be expected, the combination of the two failed to restore intramembrane processing of C99 (Fig. 5). Thus, in addition to its direct role in γ-secretase activity, as suggested by the data presented in Fig. 3 and in previous work by Levitan and coworkers (Levitan et al. 2001), these results also indicate that Asp385 is central for the endoproteolytic process. It has been hypothesized that PS endoproteolysis is an autocatalytic process (Wolfe et al. 1999) and the results presented in this study are consistent with this notion. However, the possibility remains that the integrity of the Asp385 residue may also be necessary for the interaction between PS1 and an unidentified presenilinase. Such a putative presenilinase must also be an aspartyl protease, but may have an activity pharmacologically distinct from γ-secretase (Campbell et al. 2002; Campbell et al. 2003). Alternatively, Asp385 may have a role in the interaction with one or more of the other γ-secretase complex components, thus impeding functional complex assembly. Co-expression of PS1ΔE9 failed to convert HD9stop into NTF, suggesting that the CTF has to be liberated from the rest of the PS molecule in order to elicit endoproteolysis complex assembly of HD9stop. It appears that the activity of functional PS1ΔE9 is not affected by the presence of HD9 (Fig. 5a).

Taken together, our findings demonstrate that: (i) coexpression of PS1 NTF and CTF can functionally substitute for expression of the full-length molecule in rescuing γ-secretase activity in PS-deficient mammalian cells; (ii) mutations in Asp257 or Asp385 affect intramembrane processing independently of their effect on PS1 endoproteolysis; (iii) the Asp385 residue in the CTF has a direct role in mediating endoproteolytic cleavage of PS1. Accordingly, it appears that assembly of a functional γ-secretase complex is not dependent on the presence of full-length, nonendoproteolyzed PS1. Given that the system described here allows one to dissociate PS endoproteolysis and γ-secretase complex formation, the temporal relationship between these events and the factors regulating them can now be further dissected. In this context, it will be critical to establish whether a distinct presenilinase exists or whether PS endoproteolysis is an autocatalytic event, possibly activated by the assembly of PS and the other components of the γ-secretase complex. Finally, the suggestion that NTF and CTF can be functionally coexpressed in mammalian cells, as described here, opens up novel opportunities to probe for sequences in each fragment that are critical for mediating endoproteolysis but not intramembrane processing, and vice versa.


We are grateful to Drs Donviel and St. George-Hyslop for the gift of the PS1/2-deficient BD8 cells. We thank Dr Marc Mercken of Janssen Pharmaceutica, Johnson and Johnson Pharmaceutical Research and Development, for the Aβ antibodies used in the ELISAs. The study was partly funded by grants from Loo och Hans Ostermans Stiftelse, Stiftelsen Gamla Tjänarinnor, Åke Wibergs Stiftelse, Gun och Bertil Stohnes Stiftelse, Stiftelsen Clas Groschinskys Minnesfond, Fonden för Åldersforskning vid Karolinska Institutet, Petrus och Augusta Hedlunds Stiftelse, Alzheimerfonden and from the National Institute on Aging of the National Institutes of Health.