DZNE, German Center for Neurodegenerative Diseases, Munich, Germany
Address correspondence and reprint requests to Harald Steiner, Adolf-Butenandt-Institute, Biochemistry, Ludwig-Maximilians-University, and DZNE, German Center for Neurodegenerative Diseases, Schillerstr. 44, 80336 Munich, Germany. E-mail: firstname.lastname@example.org.
γ-Secretase plays a central role in the generation of the Alzheimer disease-causing amyloid β-peptide (Aβ) from the β-amyloid precursor protein (APP) and is thus a major Alzheimer′s disease drug target. As several other γ-secretase substrates including Notch1 and CD44 have crucial signaling functions, an understanding of the mechanism of substrate recognition and cleavage is key for the development of APP selective γ-secretase-targeting drugs. The γ-secretase active site domain in its catalytic subunit presenilin (PS) 1 has been implicated in substrate recognition/docking and cleavage. Highly critical in this process is its GxGD active site motif, whose invariant glycine residues cannot be replaced without causing severe functional losses in substrate selection and/or cleavage efficiency. Here, we have investigated the contribution of the less well characterized residue x of the motif (L383 in PS1) to this function. Extensive mutational analysis showed that processing of APP was overall well-tolerated over a wide range of hydrophobic and hydrophilic mutations. Interestingly, however, most L383 mutants gave rise to reduced levels of Aβ37–39 species, and several increased the pathogenic Aβ42/43 species. Several of the Aβ42/43-increasing mutants severely impaired the cleavages of Notch1 and CD44 substrates, which were not affected by any other L383 mutation. Our data thus establish an important, but compared with the glycine residues of the motif, overall less critical functional role for L383. We suggest that L383 and the flanking glycine residues form a spatial arrangement in PS1 that is critical for docking and/or cleavage of different γ-secretase substrates.
Alzheimer disease (AD) is the most common neurodegenerative disorder affecting the elderly population. AD is characterized by the massive cerebral deposition of the amyloid β-peptide (Aβ), which is widely believed to be disease-causative (Haass and Selkoe 2007). The small (~ 4 kDa) Aβ is generated by sequential processing of the β-amyloid precursor protein (APP), a type I membrane protein, by two aspartyl proteases, β- and γ-secretase (Lichtenthaler et al. 2011). Both secretases are key AD drug targets (Tomita 2009). In contrast to the type I membrane protease β-secretase, γ-secretase is a more complicated enzyme, comprised of four essential integral membrane protein subunits, which are necessary and sufficient for activity (Steiner et al. 2008). The catalytic subunit of the complex is presenilin (PS), an intramembrane-cleaving protease of the GxGD-protease family, which is activated by autoproteolysis into an N- and C-terminal fragment (NTF, CTF) (Steiner 2008). The other subunits nicastrin (NCT), APH-1, and PEN-2 are required for assembly, activation, stabilization, and maturation of the γ-secretase complex (Dries and Yu 2008).
Following ectodomain shedding of APP by β-secretase, γ-secretase cleaves the resulting CTF of APP in the transmembrane domain (TMD) thereby releasing Aβ species of various length (Aβ37–Aβ43), which are subsequently secreted (Lichtenthaler et al. 2011). Cleavage of the TMD starts at the ε-site close to the cytosolic border releasing the APP intracellular domain and is followed by stepwise carboxy-terminal trimming (Qi-Takahara et al. 2005; Takami et al. 2009). The resultant Aβ49 is further processed to the secreted Aβ43, Aβ40, and Aβ37 species. Imprecise cleavage at the ε-site leads to the generation of small amounts of the Aβ48 species from which Aβ42 and Aβ38 are produced. Subtle changes in the ratio of the principal Aβ40 species and its longer pathogenic byproduct Aβ42 towards a relative increase in Aβ42 are sufficient to cause neurotoxicity and ultimately trigger the onset of AD (Kuperstein et al. 2010). This disease-underlying principle is reflected by the majority of early onset familial forms of AD that are caused by changes of the Aβ42/Aβ40 ratio because of mutations in PS1, its homolog PS2, and APP (Scheuner et al. 1996). Although Aβ43 is a minor secreted species, which is generated at lower level than Aβ42, it apparently has a stronger pathogenicity than the latter (Saito et al. 2011).
As evident from the processing of numerous other substrates apart from APP, γ-secretase activity is involved in a number of physiological cellular processes besides its central role in AD pathogenesis (Wakabayashi and De Strooper 2008). A primary function of γ-secretase is the degradation of membrane protein stubs, which remain after ectodomain shedding. In several cases, this can be coupled with signal transduction, where the ICD, which is released from the substrate by γ-secretase cleavage, translocates to the nucleus and regulates gene transcription. The most important substrate of γ-secretase of this type is Notch1, a crucial signaling molecule controlling cell differentiation during development as well as in adulthood via its γ-secretase-released Notch1 ICD (NICD) (Bray 2006).
Understanding how γ-secretase recognizes and selects its many substrates is a key issue for the development of drugs that inhibit or modulate Aβ generation of the enzyme while sparing cleavage of its other substrates (Tomita 2009). In addition, this knowledge is of general interest for understanding the mechanism of intramembrane-cleaving proteases (Wolfe 2009). Despite some advances, little is still known, however, about γ-secretase substrate recognition. The current model suggests a stepwise process in that the complex first recognizes the ectodomain-shedded substrate before it is presented to the active site where substrate cleavage occurs. NCT has been implicated as substrate receptor in this step, although this role is not firmly established and controversially discussed (Shah et al. 2005; Chavéz-Gutiérrez et al. 2008; Dries et al. 2009; Pamrén et al. 2011; Zhang et al. 2012). Whether substrate recognition is mediated by NCT or not, the substrate associates with the docking site, a substrate binding site spatially separated from the active site (Esler et al. 2002; Tian et al. 2002). This site has been identified in PS very close to and partially overlapping with the active site (Kornilova et al. 2005), and may involve TMDs 2, 6, 9 (Sato et al. 2008; Gong et al. 2010; Watanabe et al. 2010). Following binding at the docking site, the substrate comes in contact with the active site aspartates in TMDs 6 and 7, which enable catalysis of peptide bond hydrolysis, by a mechanism involving translocation or swinging of bound substrate into the catalytic site (Tian et al. 2002).
Consistent with a spatial proximity of docking and catalytic site (Kornilova et al. 2005), the highly conserved non-classical GxGD aspartyl protease family motif in the PS active site domain bearing the catalytic aspartate in TMD7 (Steiner et al. 2000) has been implicated in substrate docking/recognition and/or cleavage (Yamasaki et al. 2006; Pérez-Revuelta et al. 2010; Wu et al. 2010). Mutational analysis has revealed that the invariant glycine residues of the GxGD motif are critical for substrate recognition and/or cleavage, probably by ensuring efficient substrate accommodation at the active site. Maximally the very subtle amino acid change from glycine to alanine is functionally tolerated at this position (Steiner et al. 2000; Pérez-Revuelta et al. 2010). However, depending on the substrate, even this minimal amino acid change can block cleavage (Pérez-Revuelta et al. 2010). Similar findings have been made by mutational analyses for the conserved PAL motif (Nakaya et al. 2005; Wang et al. 2006), which although distant in primary sequence is close to the active site (Sato et al. 2008; Tolia et al. 2008).
To further our understanding of how the active site domain of γ-secretase contributes to substrate recognition and cleavage, we investigated the functional role of residue x of the GxGD motif in PS1, which is a leucine in almost all PSs. The leucine is replaced, however, by a phenylalanine in the C. elegans PS homolog SPE-4 and we recently identified this sequence variation as being critical for APP/Notch substrate selectivity of PS1 γ-secretase (Yamasaki et al. 2006). Apart from this initial information, however, the functional role of the leucine residue within the GLGD motif of PS1 has not been investigated yet and it is not known if it is of similar functional importance as the surrounding invariant glycine residues. In light of limited structural information for the γ-secretase active site, we decided to perform an extensive mutational analysis of the x residue. Our results demonstrate that aliphatic residues are preferred over aromatic, proline and positively charged residues, which strongly impair the cleavage of crucial signaling substrates such as Notch1 and CD44 (Miletti-González et al. 2012) without much effect on the efficiency of APP processing. Thus, we show that while L383 contributes to substrate selectivity it is overall the least critical residue of the GxGD motif in PS1. Finally, together with our previous data, we conclude that the spatial side chain arrangement of this motif is highly critical for γ-secretase substrate processing and selectivity/specificity.
Materials and methods
Monoclonal and polyclonal antibodies used were to the PS1 N-terminus (PS1N) (Capell et al. 1997), the NCT C-terminus (N1660) (Sigma, St Louis, MO, USA), the APP ectodomain (22C11) (Weidemann et al. 1989), and C-terminus (6687) (Steiner et al. 2000), Aβ 1-16 (2D8) (Shirotani et al. 2007), Aβ 17-24 (4G8) (Signet Laboratories, Dedham, MA, USA), Aβ 1-40 (3552) (Yamasaki et al. 2006), the c-myc (9E10) (Santa Cruz Biotechnology, Santa Cruz, CA, USA) and the FLAG-epitope (M2) (Sigma). Anti-FLAG M2 antibody agarose conjugate was obtained from Sigma.
N-terminally-His6Xpress (H6X) epitope-tagged PS1 wt, PS1 D385A, and PS1 L383F cDNA constructs were described before (Steiner et al. 2002; Yamasaki et al. 2006). PS1 L383 mutants were generated by PCR-mediated mutagenesis of pcDNA4/HisC:PS1, which encodes H6X-PS1 wt (Steiner et al. 2002), using oligonucleotide primers encoding the respective mutations. All cDNA constructs were confirmed by DNA sequencing. For stable expression of wt PS1 and the corresponding mutants in mouse embryonic fibroblast (MEF) cells deficient for PS1 and PS2 (PS−/−), the PS cDNA constructs were subcloned without the H6X-epitope tag as BamHI/XhoI fragments into FU-∆Zeo vector (Kuhn et al. 2010). APPsw-6myc, F-NEXT and CD44∆E-Flag constructs have been described previously (Lammich et al. 2002; Okochi et al. 2002; Wang et al. 2004).
Cell culture, cell lines, cDNA transfections, and transductions
Human embryonic kidney (HEK) 293 cells stably expressing Swedish mutant APP (HEK293/sw) were cultured and stably transfected with H6X-tagged PS1 cDNA constructs as described (Steiner et al. 2000; Page et al. 2008). Pools of stably transfected cells were investigated to avoid clonal variations. Treatment of HEK293 cells with γ-secretase modulator (GSM)s (kind gift of Dr. Karlheinz Baumann) was carried out as described (Page et al. 2008). PS−/− MEF cells stably expressing wt and mutant PS1 cDNA constructs were generated by lentiviral gene transduction. Generation of lentiviral particles and administration to PS−/− MEF cells was carried out as described previously (Kuhn et al. 2010). The stably transduced PS−/− MEF cells were cultured and transiently transfected as described (Yamasaki et al. 2006).
PS1, NCT, APP, Notch1, CD44, and/or derivatives, respectively, were analyzed from cell lysates (Page et al. 2008) by direct immunoblotting as described (Lammich et al. 2002; Yamasaki et al. 2006). Total secreted APPs was analyzed by direct immunoblotting of conditioned media with antibody 22C11. Analysis of total secreted Aβ and CD44β was performed as described (Lammich et al. 2002; Yamasaki et al. 2006; Pérez-Revuelta et al. 2010). Mass spectrometry analysis of individual secreted Aβ species and their quantitation by an Aβ sandwich immunoassay was carried out as outlined previously (Page et al. 2008, 2010). Where indicated, immunoblotted proteins were quantified as described before (Pérez-Revuelta et al. 2010).
Quantitative data were subjected to statistical analysis by student's t test or one way anova with Dunnett′s post test using GraphPad Prism 5.0 (GraphPad Software Inc., San Diego, CA, USA) software where indicated. Data with p values lower than 0.05 were considered to be statistically significant.
A wide range of amino acids at PS1 residue 383 supports APP processing
To investigate the role of the leucine 383 residue of PS1 for γ-secretase activity and substrate selectivity, we generated a variety of mutant PS1 constructs covering almost all possible amino acid substitutions at this site. We first were interested to investigate how these would affect processing of APP and thus stably expressed the L383 mutant constructs in HEK293/sw cells. PS1 wt and catalytically inactive D385A mutant constructs were used as positive and negative controls, respectively. To facilitate analysis, N-terminally-H6X-epitope-tagged PS1 constructs were used. We first analyzed how aliphatic and aromatic amino acids substitutions at the L383 residue would affect APP processing. As shown in Fig. 1a, all PS L383 mutant constructs replaced endogenous PS and were endoproteolysed like wt PS1 demonstrating functional γ-secretase complex formation and suggesting that the L383 residue of the GxGD motif is not critical for PS endoproteolysis. This characteristic conversion of PS into an N- and C-terminal fragment is believed to represent an activated state of γ-secretase and – being an autocatalytic process (Edbauer et al. 2003; Fukumori et al. 2010) – is not observed for the catalytically inactive PS1 D385A mutant (Steiner et al. 1999; Wolfe et al. 1999).
As judged from the levels of APP CTFs, which are the direct substrates of γ-secretase, processing of APP was overall largely supported as only a few mutations caused substrate accumulation indicative for activity loss (Fig. 1b). The aliphatic mutations did not affect activity, while the aromatic mutations and the proline mutant displayed a partial loss of function. Although the latter mutations caused an increase in APP CTFs, this accumulation of substrate was less pronounced when compared with the PS1 D385A mutant, which displayed a dramatic increase in these substrates. Consistent with the analysis of APP CTFs, the generation of Aβ was not significantly affected (Fig. 1c). A moderate reduction was only seen for the aromatic residues and proline, while all other mutants generated Aβ at levels comparable to wt PS1. Quantitation confirmed these results (Fig. 1d and e).
As all above described mutations maintained the hydrophobic nature of the side chain, we next investigated whether charged amino acid substitutions would have a stronger impact on APP processing. As shown in Fig. 1f, these mutants also allowed functional complex formation as judged from their ability to undergo efficient PS endoproteolysis. The negatively charged amino acids as well as their isosteric uncharged analogs did not affect γ-secretase activity (Fig. 1g). Likewise, the positively charged histidine substitution was supporting γ-secretase activity. As judged from the weak accumulation of the APP CTFs, a mild decrease in substrate turnover could, however, be attested for the arginine and lysine substitutions (Fig. 1g). This partial loss of function of these two mutants was also reflected in lower Aβ levels (Fig. 1h). Apart from these exceptions, however, total Aβ levels of the mutants were all very similar to those generated by wt PS1, suggesting that overall, the hydrophilic amino acid substitutions had no substantial impact on γ-secretase activity (Fig. 1h). Quantitation confirmed theses results (Fig. 1i and j).
L383 mutants preferentially lower the generation of short Aβ species
We next investigated whether the mutants would change γ-secretase cleavage specificity and if they would potentially cause pathogenic production of longer Aβ species, such as Aβ42. Using an Aβ immunoassay specific for Aβ38, Aβ40, and Aβ42 (Page et al. 2008), differences in the profiles of the Aβ species generated by the hydrophobic mutants compared with wt PS1 were noted (Fig. 2a). In particular, we found that aliphatic amino acid substitutions did not affect the Aβ42/Aβtotal ratios, while in contrast, the aromatic amino acids, as well as the L383P mutant showed a mild – moderate increase of Aβ42 generation (Fig. 2b). Interestingly, all mutants substantially decreased the generation of the shorter Aβ species Aβ38 (Fig. 2c). Because we could not detect all Aβ species with our assay, we asked how other Aβ species would be affected by the mutations of L383 and thus additionally analyzed these by mass spectrometry. In addition to confirming the increase in Aβ42 for the aromatic amino acids, this analysis revealed that the L383P mutant caused an exceptional, very strong increase in the generation of the long Aβ43 species, which exceeded that of Aβ42. An increase in Aβ43 was also noted for the L383Y mutant. Mass spectrometry analysis also showed that nearly all hydrophobic mutants of L383 reduced the levels of the shorter Aβ species Aβ37, Aβ38, Aβ39. Only the L383G mutant still produced Aβ37. Fig. 2d illustrates these findings by showing representative examples for each observed mass spectrometry profile of the Aβ species obtained for the mutants. As shown in Fig. 3a, also the hydrophilic amino acid substitutions showed effects on the cleavage specificity of γ-secretase. The negatively charged amino acids aspartate and glutamate as well as the positively charged amino acid substitutions did mildly increase the generation of Aβ42 (Fig. 3b). From the isosteric analogs of aspartate and glutamate, asparagine and glutamine differed in their behavior as only glutamine caused a mild increase in Aβ42 generation (Fig. 3b). A decrease in Aβ38 was noted as well, but not so pronouncedly as for the hydrophobic mutants, and for the L383N mutant even an increase in Aβ38 was observed (Fig. 3c). Mass spectrometry analysis of the mutants was in good agreement with the above measurements of the individual Aβ species. For the L383N mutant, the generation of short Aβ species was increased, in particular that of Aβ37, which besides Aβ40 became the major species generated. In addition to an increase in Aβ42, the positively charged L383K and L383R mutants robustly increased the generation of Aβ43. Like above, these findings are illustrated by showing representative examples for each observed mass spectrometry profile of the Aβ species obtained for the mutants (Fig. 3d).
Taken together, we conclude that, unlike the neighboring glycine residues G382 and G384, which are highly critical residues on γ-secretase processing, mutation of L383 of the GxGD motif is considerably tolerable for APP substrate processing. However, most mutants reduced the generation of the short Aβ37–39 species indicating a reduced efficiency in the intrinsic carboxy-terminal trimming activity of γ-secretase. Only a subset of the mutants showed a partial loss of function and these changed γ-secretase cleavage specificity toward an elevated generation of the pathogenic Aβ42/43 species.
Most Aβ42/43-increasing L383 mutants are resistant to GSMs
The above results indicate conformational changes of γ-secretase induced by the majority of the L383 mutants that may alter APP substrate positioning thus giving rise to the observed changes in the Aβ profiles including increased Aβ42/43 productions. Changed PS conformations as a consequence of e.g. Aβ42/43-increasing FAD-associated mutations (Berezovska et al. 2005) might also cause an unwanted resistance to GSMs, Aβ42-lowering drugs that from a therapeutic point of view, are expected to be superior to γ-secretase inhibitors, as they do not block γ-secretase activity and thus leave the cleavage of crucial substrates such as Notch1 unaffected (Imbimbo and Giardina 2011). A few PS FAD mutants are locked in a conformation, which render γ-secretase even resistant to the Aβ42-lowering capacity of highly potent second generation GSMs (Page et al. 2008; Hahn et al. 2011; Kretner et al. 2011). Having this in mind, we next investigated whether L383 mutants would be responsive to such GSMs using a representative selection for each amino acid class of the above mutants and including those which showed the strongest pathogenic, i.e. Aβ42/43-increasing activities. As shown in Fig. 4a, only a subset of the L383 mutants showed reduced Aβ42 production in response to GSM-1, a frequently used acidic GSM (Page et al. 2008; Hahn et al. 2011; Kretner et al. 2011). The positively charged L383K and R mutants, the aromatic L383F, Y and W mutants, and the L383P mutant were highly resistant to GSM-1. For the latter mutant, GSM-1 also failed to lower the generation of Aβ43, the preferentially generated long Aβ species by L383P (data not shown). These mutants showed also a striking resistance to a structurally different non-acidic bridged aromatic GSM with increased potency developed by Eisai (Kretner et al. 2011), with the exception of the L383Y mutant, which showed a partial response (Fig. 4c). However, all mutants clearly increased the generation of Aβ38 showing that the resistant mutants not simply failed to be targeted by the GSMs (Fig. 4b and d). We conclude that the generation of Aβ42(43) species cannot efficiently be lowered by advanced GSMs of two different structural classes for most of the pathogenic L383 mutants.
L383 is critical for γ-secretase substrate selectivity
The above results showed that mutation of L383 of the GLGD active site motif of PS1 have overall little impact on total γ-secretase activity toward APP substrate processing. In the next set of experiments, we compared the processing of APP, Notch1, and CD44 in mouse embryonic fibroblast (MEF) cells derived from PS1/2 (PS−/−) double knockout mice to further investigate the previously identified substrate selectivity provided by this residue (Yamasaki et al. 2006). We first concentrated on the analysis of the aliphatic and aromatic amino acid substitutions, as these keep the hydrophobic nature of the L383 residue and thus represent conservative exchanges at this position. APPsw-6myc (Wang et al. 2004), F-NEXT (Okochi et al. 2002), and CD44∆E-Flag (Lammich et al. 2004) substrates were transiently transfected into PS−/− MEF cells stably transduced with the various wt and mutant PS1 constructs. To assess γ-secretase activity on APP and Notch1 substrate cleavage, formation of APP intracellular domain and NICD was analyzed. Because of the difficulty to detect the ICD of CD44, the Aβ-like CD44β peptide was analyzed for this substrate, given the ease of its analysis (Pérez-Revuelta et al. 2010). Processing of APP was in line with the results obtained in the stably transfected HEK293 cells above and confirmed that the aromatic amino acid substitutions maintained γ-secretase activity substantially, although they were lower in their activity than their aliphatic counterparts (Fig. 5a). Consistent with our previous findings (Yamasaki et al. 2006), processing of Notch1 was strongly impaired by the PS1 L383F mutant (Fig. 5b). The other aromatic amino acid substitutions PS1 L383Y and L383W as well as the L383P mutant behaved like the L383F mutant and displayed a similar severe impairment of Notch1 substrate processing. In contrast, the aliphatic amino acid substitutions supported normal γ-secretase activity on this substrate (Fig. 5b). Additional analysis using a NICD-specific antibody confirmed these results (Figure S1a, b). As shown in Fig. 5c, the same profile was observed for processing of CD44. For this substrate the aromatic acids and proline caused a nearly complete block of substrate cleavage. Quantitation confirmed the differential effects imposed by the aromatic amino acid substitutions of L383 on APP, Notch1, and CD44 and substrate processing (Fig. 5d). We conclude that the L383 residue of the active site of PS1 is critical for γ-secretase substrate selectivity.
We next analyzed the hydrophilic substitutions in the same experimental paradigm. In agreement with the results obtained above in HEK293 cells, all mutants except the positively charged L383K and L383R mutants, which displayed a partial loss of function, supported the processing of APP (Fig. 6a). All hydrophilic mutants supported the processing of Notch1 (Fig. 6b), and CD44 (Fig. 6c) except the L383K and L383R mutants, which almost completely inhibited the cleavage of these substrates. Quantitation, as well as validation of the Notch1 processing data using a NICD-specific antibody (Figure S1c, d) confirmed these results (Fig. 6d).
Taken together, these data show that L383 residue tolerates a number of amino acid substitutions without loss of γ-secretase activity. Only a subset of the mutants, i.e. the aromatic mutants, the proline mutant and the positively charged lysine and arginine mutants strongly inhibit substrate cleavage of at least two substrates, Notch1 and CD44. In contrast, the same mutations affect APP processing to a much lesser extent. We conclude that while providing γ-secretase substrate selectivity, residue x is also the least critical residue of the GxGD motif in PS1.
Several studies have implicated the active site domain of γ-secretase in PS1 in both substrate recognition and cleavage, with the residues N-terminal to the TMD7 active site probably providing subsites, i.e. substrate-binding pockets (Schechter and Berger 1967), of the protease (Yamasaki et al. 2006; Pérez-Revuelta et al. 2010). The invariant glycine residues of the GxGD motif have been shown to be highly critical for substrate processing (Steiner et al. 2000; Pérez-Revuelta et al. 2010). In this study, we have in detail investigated the role of the x position of the GxGD motif of the active site domain of γ-secretase in PS1. By domain exchange experiments between PS1 and its C. elegans homolog SPE-4 we had previously mapped L383 as an APP/Notch substrate selectivity-providing residue (Yamasaki et al. 2006). To further define the functional role of this leucine residue, we were interested to investigate how mutations of it and thus of residue x of the GxGD motif, affect PS endoproteolysis and γ-secretase substrate processing, how they affect γ-secretase cleavage specificity of APP, how they influence substrate selectivity of γ-secretase and finally, whether the spatial side chain arrangement (i.e. small-big-small) of the GLGD motif is critical for γ-secretase activity.
Overall, we found that L383 is the least critical residue of the GxGD protease active site motif of PS1. Residue x is less sensitive to mutation and more permissive than the glycine residues with respect to PS endoproteolysis and γ-secretase substrate processing. We found that aromatic residues are interfering with Notch1 and CD44 processing, whereas they are relatively well-tolerated for APP. The presence of aliphatic acid residues at position x such as leucine in the wt case represents a critical determinant for processing of substrates such as Notch1 and CD44. Thus, we confirm and extend our previous findings that γ-secretase substrate selectivity depends on the side chain at residue x of the GxGD active site motif (Yamasaki et al. 2006). Similar to the situation with the G382 and G384 mutants, Notch1 and CD44 are substrates, which are critical to alteration of these key residues of the active site domain (Pérez-Revuelta et al. 2010). In contrast with mutation of the glycine residues, however, a number of L383 alterations for which Notch1 and CD44 are already severely affected are still tolerable for APP. Aromatic residues cause a weak loss of function on APP processing while other substrates such as Notch1 and CD44 are much more sensitive. This suggests that in contrast with the latter substrates, APP might be conformationally more flexible, such that its residues coming in contact with the γ-secretase active site domain may more stably interact with the subsites of the protease. Despite the only mild impact of the L383 mutations on the generation of total Aβ, robustly increased Aβ42(43)/Aβtotal ratios were observed for the substitution of L383 with aromatic amino acids, proline, and positively charged amino acids. Thus, some of the mutants, which displayed a weak loss of function, also affected the cleavage specificity of γ-secretase and increased the generation of Aβ42 or Aβ43, respectively. Almost all mutations showed reduced amounts of the short Aβ species Aβ37–Aβ39 compared with wt PS1. Thus, compared with wt, L383 mutants affect APP substrate positioning such that the γ-37, γ-38, and γ-39 sites are less exposed to the active site. One of the mutations (L383N) was, however, exceptional in that regard and displayed an increase in the generation of short Aβ species, in particular of Aβ37. Collectively, these observations suggest that most of the L383 mutants reduce carboxy-terminal trimming of the APP TMD, i.e. the processivity of γ-secretase. The impaired generation of short Aβ species is, however, not a strict consequence of reduced γ-secretase processivity and premature release of longer Aβ species as recently suggested (Chavéz-Gutiérrez et al. 2012), as an concomitant accumulation of longer Aβ42/43 species was observed only for some of the L383 mutants. Interestingly, although total γ-secretase activity and processivity are not necessarily coupled (Moehlmann et al. 2002; Quintero-Monzon et al. 2011; Chavéz-Gutiérrez et al. 2012), it is striking that most of the mutants which showed a weak-moderate loss of function in APP processing and increased the generation of Aβ42/43 were also resistant to GSMs, and were those mutations, which displayed strong loss of function effects on Notch1 and CD44 processing (see Table 1 for a summary).
Table 1. Summary of the activities of PS1 L383 mutants and their responses to GSMs
γ-Secretase activities of PS1 L383 mutants for APP, Notch1 and CD44 processing relative to that of PS1 WT (set to 100%) as determined by generation of AICD, NICD, and CD44β respectively, are indicated (+, robust activity, > 75%; ↓, reduced activity, > 50–75%; ↓↓, only weak activity, 25–50%; ↓↓↓, almost no activity, < 25%).
Changes in the cleavage specificity of γ-secretase were determined with regard to the different secreted Aβ species.
The relative change in the Aβ38, Aβ40, or Aβ42 to Aβtotal ratio as measured by Aβ immunoassay for each mutant compared with PS1 WT is indicated by arrows (for the Aβ38/Aβtotal and Aβ40/Aβtotal ratio: ↓↓, < 0.5-fold; ↓, 0.5–0.75-fold decrease, ↑, > 1.25-fold increase; for the Aβ42/Aβtotal ratio: ↑, = 1.5–2.5-fold; ↑↑, > 2.5–3.5-fold; ↑↑↑, > 3.5-fold increase; –, no change).
Aβ37, Aβ39, and Aβ43 were determined by mass spectrometry, arrows reflect qualitative changes compared with PS1 WT (↓, reduced peak; ↑, increased peak; ↑, only a very small peak was observed; –, no change).
Aβ37 was the major short Aβ species for PS1 L383N.
Aβ43 was the major pathogenic species for PS1 L383P.
Additionally, the relative levels of the Aβ42/Aβtotal ratios in response to treatment with GSM-1 and the Eisai compound (each 1 μM) compared with vehicle control are indicated (++ = 0–25%; +, > 25–50%; +/−, > 50–75%; –, > 75%). N.D. not determined.
The evolutionary basis for the highly conserved presence of aliphatic amino acids at the x position of the GxGD motif in PS is likely that only these amino acids allow efficient processing of γ-secretase substrates other than the overall very robust substrate APP. From the hydrophobic amino acids analyzed, aromatic amino acids and proline at this site are not supported and/or preferred by all substrates. This feature provides some selectivity in γ-secretase substrate processing and may contribute to the fact that not all type I membrane proteins, which are potential substrates are processed by γ-secretase (Hemming et al. 2008). Apparently, a certain side chain arrangement that in approximation can be defined as G-π/ψ/ζ-G-D (π = small side chain amino acid, ψ = aliphatic amino acid, ζ = hydrophilic amino acid) is critical for γ-secretase substrate processing and selectivity. Exceptions are the positively charged amino acid substitutions lysine and arginine, as well as proline, which support APP processing only partially and besides Aβ42 also increase the generation of the longer species Aβ43. The side chain length per se of the amino acids at L383 residue does not play a similar important role compared with the strict requirement for short side chains at the neighboring G382 and G384 residues (Pérez-Revuelta et al. 2010).
Interestingly, consistent with the GxGD motif providing substrate selectivity, the GxG sequence has recently been suggested to represent a putative nucleotide-binding site in γ-secretase, which may be the target of novel Notch-sparing γ-secretase inhibitors (Wu et al. 2010). Structurally, the motif is part of an unstructured flexible loop (Sato et al. 2006; Sobhanifar et al. 2010), which has also been found for the PS-related archael GxGD protease FlaK, for which crystal structure information has recently been obtained (Hu et al. 2011). In PS1, L383 likely faces away from the both flanking glycine residues (Wu et al. 2010) forming a spatial arrangement that is critical for docking and/or cleavage of different γ-secretase substrates.
In conclusion, by determining the sequence requirements at residue L383 of PS1 for three critical γ-secretase substrates, our study complements previous mutational analyses of the GxGD active site motif in PS1 such that the side chains for substrate selectivity and cleavage efficiency provided by this critical motif are now fairly well defined for all three residues of the GxG sequence. Further major insights into the structure of the γ-secretase active site domain in PS and the mechanism of substrate selection and cleavage are awaited from a crystal structure, as it will come available.
We thank Bart De Strooper for PS−/− MEF cells; Ralph Nixon for monoclonal antibody PS1N; Alison Goate, Masayasu Okochi, and Sven Lammich for the APPsw-6myc; F-NEXT and CD44∆E constructs, respectively; and Karlheinz Baumann for GSMs. This work was supported by the Deutsche Forschungsgemeinschaft (SFB596) (C.H., H.S.), the BMBF (KNDD) (C.H., H.S., S.F.L.), the FöFoLe program of the Ludwig-Maximilians-University (LMU) Munich (B.K., H.S.), and the Center for Integrated Protein Science Munich (CIPSM). The LMUexcellent program supports C.H. with a research professorship. The authors declare that they do not have conflicts of interest to disclose.