Intramembrane proteolysis has been recognized as a general cellular mechanism that underlies many important physiological processes such as membrane protein turnover and signal transduction to the nucleus (Wolfe and Kopan 2004). One of the key intramembrane-cleaving proteases identified to date is γ-secretase. This multisubunit protease complex cleaves the Alzheimer’s disease (AD)-associated β-amyloid precursor protein (APP) and a number of other type I membrane proteins, including Notch1, a cell surface receptor crucial for cell differentiation during embryonic development and adulthood (Wolfe and Kopan 2004). Prior to γ-secretase cleavage, which occurs in the transmembrane domain (TMD), substrates need to undergo an initial processing step by a sheddase, which removes the bulk of the substrate ectodomain (Haass 2004). In case of APP, ectodomain shedding by β-secretase leaves a C-terminal APP fragment (CTFβ) in the membrane. Subsequent cleavage of CTFβ by γ-secretase releases the intracellular domain (ICD) of APP (AICD) and the 37–43 amino acid amyloid-β peptide (Aβ) including the highly neurotoxic Aβ42 that is believed to cause the disease (Haass 2004). An alternative ectodomain cleavage by α-secretase in the Aβ domain generates the smaller CTFα. The subsequent γ-secretase cleavage releases the non-pathogenic peptide p3 thereby precluding the formation of Aβ (Haass 2004). As inhibition and/or modulation of γ-secretase activity are obvious major therapeutic concepts in AD, γ-secretase is a prime AD drug target. In addition, γ-secretase is considered as therapeutic target of cancer because of its implication in enhanced Notch signaling mediated by the Notch ICD that is observed in certain cancers (Shih Ie and Wang 2007).
γ-Secretase is composed of four integral membrane proteins, presenilin (PS), nicastrin (NCT), anterior pharynx defective (APH)-1 and presenilin enhancer (PEN)-2 that are all required for its activity (Haass 2004). Biochemical and pharmacological evidence has established PS as catalytic subunit of γ-secretase (Steiner 2008). Mutations in PS account for the majority of familial forms of AD and shift the precision of γ-secretase cleavage towards an increased Aβ42/Aβtotal ratio. PS is cleaved by autoproteolysis into an N- and a C-terminal fragment (NTF, CTF) during the assembly of the γ-secretase complex. Each of the two fragments carries one of the two opposing active site aspartate residues in TMD6 and TMD7. These TMDs, together with TMD9, constitute at least a part of a water-containing cavity to accommodate γ-secretase substrates for cleavage (Sato et al. 2006a, 2008a; Tolia et al. 2006, 2008).
Presenilin has an active site that differs in its sequence from that of classical aspartyl proteases. The typical D(T/S)G(T/S) sequence is missing in PS, instead, the C-terminal active site aspartate is part of a highly conserved GxGD motif (Steiner et al. 2000). This motif is conserved in the type 4 prepilin peptidase (LaPointe and Taylor 2000), the signal peptide peptidase (SPP) and the SPP-like (SPPL) protease families (Weihofen et al. 2002), other polytopic aspartyl proteases of the GxGD-type. In addition to the signature GxGD motif, a C-terminal PxL motif is conserved besides the invariant N-terminal active site aspartate (Moliaka et al. 2004). Apart from that, GxGD-type aspartyl proteases display only very limited sequence homology to each other (Moliaka et al. 2004). Apparently, a novel active site has evolved for aspartyl proteases specialized in peptide bond hydrolysis directly at or in the membrane (Fluhrer et al. 2009). Previous studies with PS1 showed that mutations of the glycine immediately preceding the catalytically active site aspartate (G384 of PS1) can, depending on the mutation, cause either a dramatic shift in γ-secretase cleavage specificity, as in the case of the FAD-associated PS1 G384A mutation (De Jonghe et al. 1999; Steiner et al. 2000), or a total loss of γ-secretase activity (Steiner et al. 2000; Tolia et al. 2006). In addition, residue x of the motif (L383 in PS1) contributes to substrate identification/selectivity of γ-secretase (Yamasaki et al. 2006) probably by serving as part of a substrate binding site located in immediate vicinity to the catalytic site (Esler et al. 2002; Kornilova et al. 2005). A deeper understanding of the requirements of the γ-secretase active site domain in PS for substrate cleavage additionally requires a detailed functional analysis of the N-terminal glycine of the motif whose putative importance has not been addressed yet. Moreover, the questions of a putative general functional relevance of the glycine residues of the GxGD motif, and of why the glycine residues are invariant elements of the motif have remained unanswered. To address these issues, we have therefore now subjected the N-terminal glycine of the GxGD motif (G382 of PS1) to an extensive mutagenesis analysis. Our data identify a requirement of small side chains as optimally provided by the naturally occurring glycines at the respective positions in the GxGD motif enabling the accommodation of the wide range of different γ-secretase substrates close to the active site.
- Top of page
- Experimental procedures
To investigate the functional role of the conserved glycine residues of the GxGD motif further, we mutagenized the first glycine residue (G382) of the motif in PS1. To distinguish the constructs from endogenous PS, we used N-terminally hexahistidine-Xpress (H6X)-epitope tagged PS1 to introduce the mutants. Constructs encoding substitutions of G382 to alanine, isoleucine, proline, tryptophane, lysine and aspartate were generated. These constructs were stably transfected in HEK293 cells stably expressing Swedish mutant APP (HEK293/sw). HEK293/sw cell lines stably expressing H6X-tagged wt PS1 (Steiner et al. 2002) and the H6X-tagged PS1 D385A active-site aspartate mutant constructs (Yamasaki et al. 2006) were generated for comparison. The G382 mutants were first investigated for their capability to undergo γ-secretase complex formation. Like wt PS1 and the PS1 D385A mutant, all G382 mutants replaced endogenous PS1 suggesting that the capability to undergo γ-secretase complex formation was preserved in the mutants (Fig. 1a). Correct complex assembly is a prerequisite for the subsequent trafficking of a γ-secretase complex through the secretory pathway. No alteration in the maturation of NCT was observed, further suggesting that the mutants assembled normally with the other subunits to form γ-secretase complexes. Interestingly however, only the G382A mutant allowed PS endoproteolysis, whereas all other mutants investigated remained unprocessed. This observation suggests that stronger amino acid alterations than alanine at residue 382 of PS1 interfere with PS endoproteolysis.
Figure 1. PS1 G382 is crucial for PS endoproteolysis and APP CTF γ-secretase substrate processing. (a) HEK293/sw cells stably expressing H6X-tagged PS1 wt, PS1 D385A and the indicated PS1 G382 mutant constructs were analyzed for expression, endoproteolysis and replacement of endogenous PS by immunoblotting of cell lysates with antibody PS1N. Maturation of NCT was analyzed by immunoblotting with antibody N1660. The parental HEK293/sw cell line (endog. PS) was analyzed in parallel to assess the level of endogenous PS1 replacement in the H6X-tagged PS1 transfected cell lines. (b)–(e) PS1 transfected cell lines described above (a) were analyzed for APP processing. Cell lysates were analyzed for full length APP and APP CTFs by immunoblotting with antibody 6687 (b). Secreted APPs and Aβ were analyzed from conditioned medium by immunoblotting with antibody 5313 and by combined immunoprecipitation/immunoblotting with antibodies 3552/6E10, respectively (c). Levels of NCTm, APP CTFs, secreted APPs and Aβ analyzed as in (a–c), were quantified by measuring their chemiluminescence signal intensities. APP CTFs/NCTm ratios (d) and Aβ/APPs ratios (e) were expressed relative to those generated in cells expressing wt PS1 that was set 100%. Bars represent the mean of three independent experiments ± SE. Asterisks indicate the significance (one way anova with Dunnett’s post-test) relative to PS1 wt (*p < 0.05; **p < 0.01).
Download figure to PowerPoint
We next determined whether the G382 mutants have an impact on γ-secretase activity towards processing of APP. As shown in Fig. 1(b), with the exception of the G382A mutant, all other mutants caused an accumulation of APP CTFs, which are the direct substrates of γ-secretase. These data suggested that most G382 mutants strongly compromised the proteolytic activity of γ-secretase. To further substantiate these observations we analyzed the generation of Aβ. As expected from the above results only the G382A mutant allowed formation of Aβ, whereas the other mutants showed strongly reduced Aβ levels consistent with the strong accumulation of APP CTF stubs (Fig. 1c). Quantitation of APP CTFs (Fig. 1d) and Aβ (Fig. 1e) confirmed the effects exerted on γ-secretase activity by the PS1 G382 mutants.
Because G382A was the only mutant identified that retained γ-secretase activity, we next examined the activity of PS1 G382A in amyloidogenic APP processing in more detail. To investigate whether the G382A mutant potentially caused alteration(s) in the cleavage precision of γ-secretase, we analyzed the profile of Aβ species generated from PS1 wt and PS1 G382A expressing cells by MALDI-TOF mass spectrometry (Fig. 2a). This revealed Aβ40 as major species for both PS1 wt and G382A. Careful comparison of the mass spectra obtained for PS1 wt and the PS1 G382A mutant consistently revealed a relative increase of Aβ38 as well as a decrease of Aβ37 and Aβ39 for PS1 G382A. Most strikingly however, in addition to Aβ42, which was detected for both PS1 wt and PS1 G382A, the mass spectra of PS1 G382A consistently showed a peak corresponding to Aβ43, which was not detected for PS1 wt. Additional qualitative analysis of the Aβ species generated by PS1 G382A using Tris-bicine urea sodium dodecyl sulfate–polyacrylamide gel electrophoresis, which allows electrophoretic separation of the individual Aβ species, gave a consistent picture with the mass spectrometry analysis. Compared to PS1 wt, an increase of Aβ38 and an increased amount of Aβ42/Aβ43 species migrating as unresolved double band was observed for PS1 G382A (Fig. 2b). Quantitation of secreted Aβ38, Aβ40 and Aβ42 species by a highly sensitive and specific Aβ sandwich immunoassay (Page et al. 2008) confirmed that the PS1 G382A mutant produced increased amounts of Aβ38 besides Aβ40 as the major species (Fig. 2c). In addition, an increased production of Aβ42, which was difficult to detect by mass spectrometry (Fig. 2a) was clearly revealed by the Aβ sandwich immunoassay analysis (Fig. 2c and d). As shown in Fig. 2(d), comparison of the normalized Aβ38/Aβtotal, Aβ40/Aβtotal and Aβ42/Aβtotal ratios revealed a robust increase of Aβ38 (2.62 ± 0.12), a milder increase of Aβ42 (1.55 ± 0.06), and a very subtle decrease of Aβ40 (0.86 ± 0.02) for PS1 G382A relative to PS1 wt (1.00). We thus conclude that the PS1 G382A mutant causes alterations in the cleavage specificity at the γ-cleavage sites.
Figure 2. PS1 G382A affects γ-secretase cleavage precision. (a) Secreted Aβ in conditioned media of HEK293/sw cells stably expressing H6X-tagged PS1 wt or PS1 G382A was immunoprecipitated with antibody 4G8 and subjected to MALDI-TOF MS analysis. Note the relative peak height changes of individual Aβ species. (b) Secreted Aβ in conditioned media of cells described above (a) was immunoprecipitated with antibody 3552, subjected to Tris-bicine-urea sodium dodecyl sulfate–polyacrylamide gel electrophoresis to separate the individual Aβ species and analyzed by immunoblotting using antibody 6E10. (c) Levels of secreted Aβ38, Aβ40 and Aβ42 species in conditioned media of cells described above (a) were quantified by a highly specific Aβ sandwich immunoassay and plotted as a percentage of the total Aβ (i.e. the sum of Aβ38, Aβ40 and Aβ42) measured. Bars represent the mean of five independent experiments ± SE. (d) Data of (c) were plotted such that Aβ38, Aβ40 and Aβ42 to Aβtotal ratios produced by PS1 G382A were expressed relative to those of PS1 wt that were set 1.00. Asterisks indicate the significance (paired two-tailed Student’s t-test) relative to PS1 wt (**p < 0.01; ***p < 0.001).
Download figure to PowerPoint
We next investigated whether and how the processing of other γ-secretases substrates would be affected by PS1 G382 mutations. To this end, we assessed γ-secretase substrate cleavage by a widely used experimental paradigm using PS1/2 double knockout (PS−/−) MEF cells transiently co-transfected with wt or mutant PS1 and the respective substrate. In all experiments using this experimental system and as exemplified in Fig. 3(a), we confirmed that γ-secretase complex formation was restored upon the transfection of PS, which was monitored by the maturation of NCT (Fig. 3a and data not shown). We first examined APP processing by monitoring the generation of AICD from APPsw-6myc (Wang et al. 2004), a C-terminally 6myc-tagged APP substrate. Consistent with the results shown above, the investigation of APP processing in this experimental system showed again that only the G382A mutant supported γ-secretase activity in APP processing, while all other G382 mutants were not capable to process this substrate (Fig. 3a). We then examined the processing of the well-characterized γ-secretase substrate Notch1 (De Strooper et al. 1999) using F-NEXT, a Flag-tagged Notch1ΔE derivative (Okochi et al. 2002) as substrate. As assessed by the F-NEXT cleavage product Notch1 intracellular domain (NICD), only the PS1 G382A mutant allowed processing, while all other PS1 G382 mutants were impaired, very similar to the result obtained with APPsw-6myc (Fig. 3b). Quantitation confirmed that the PS1 G382A mutant retained a partial, but substantial γ-secretase activity towards the APP (Fig. 3c) and Notch1 substrate (Fig. 3d).
Figure 3. PS1 G382 mutants display similar effects on APP and Notch1 γ-secretase substrate processing. (a) PS−/− MEF transiently co-transfected with APPsw-6myc and the indicated PS1 constructs. Cell lysates were analyzed for PS1 expression and endoproteolysis by immunoblotting with antibody PS1N. Maturation of NCT was analyzed by immunoblotting with antibody N1660. APP CTFs and AICD were analyzed by immunoblotting with anti-myc antibody 9E10. (b) PS−/− MEF were transiently co-transfected with F-NEXT and the indicated PS1 constructs. Expression and processing of F-NEXT was analyzed by immunoblotting with antibody 9E10. (c, d) Generation of AICD (c) and NICD (d) was analyzed as in (a) and (b), respectively, and the intensities of their immunoblot signals were quantified densitometrically. Levels of AICD and NICD were normalized to the levels of NCTm [analyzed as in (a) and quantified densitometrically] and expressed relative to those generated by PS1 wt that was set 100%. Bars represent the mean of three independent experiments ± SE. Asterisks indicate the significance (paired two-tailed Student’s t-test) relative to PS1 wt (*p < 0.05; ***p < 0.001).
Download figure to PowerPoint
The above data showed that the G382 mutants affected γ-secretase processing of Notch1 very similar to that of APP, i.e. that only a very small substitution of glycine to alanine was tolerated at this residue position to support substrate cleavage. They therefore indicated that the distinct profile on γ-secretase substrate processing might possibly be a general feature of the G382 mutants. To further explore this, we next tested the processing of CD44, another well-studied γ-secretase substrate (Lammich et al. 2002; Murakami et al. 2003), using the previously described CD44ΔE-Flag construct (Lammich et al. 2002) as substrate. As shown in Fig. 4, none of the G382 mutants, and surprisingly not even the G382A mutant, allowed processing of CD44ΔE-Flag as judged from the lack of the CD44ICD and the CD44β cleavage products (Fig. 4a). Quantitation confirmed this observation and revealed that the block of CD44 processing by the PS1 G382A mutant was comparable to the catalytically inactive PS1 D385A mutant (Fig. 4b). We conclude that PS1 G382A has a differential impact on processing of the APP and Notch1 substrates versus the CD44 substrate.
Figure 4. CD44 processing is sensitive to PS1 G382 mutations including the smallest possible G382A substitution. (a) PS−/− MEF cells were transiently co-transfected with CD44ΔE-Flag and the indicated PS1 constructs. Cell lysates were analyzed for PS1 expression and endoproteolysis and for maturation of NCT by immunoblotting as described in Fig. 3(a). Expression and processing of CD44ΔE-Flag was analyzed by immunoprecipitation with anti-c-myc agarose and immunoblotting with anti-myc antibody 9E10. The circle denotes the CD44ICD band detectable only for PS1 wt. CD44β levels were analyzed from conditioned media by combined immunoprecipitation/immunoblotting with anti-FLAG-M2 antibody. (b) Generation of CD44ICD and CD44β was analyzed as in (a) and the intensities of their immunoblot signals were quantified densitometrically. Levels of CD44ICD and CD44β determined were normalized to the levels of NCTm [analyzed as in (a) and quantified densitometrically] and expressed relative to those generated by PS1 wt that was set 100%. Bars represent the mean of three independent experiments ± SE. Asterisks indicate the significance (paired two-tailed Student’s t-test) relative to PS1 wt (**p < 0.01; ***p < 0.001).
Download figure to PowerPoint
To investigate whether side-chain size was responsible for the striking substrate discrimination of CD44 by PS1 G382 mutant γ-secretase, we next exchanged G382 for other small residues and generated the G382S, G382T and G382V mutants thereby steadily increasing the side chain length from glycine over alanine to serine, threonine and valine. In this and the subsequent experiments we assessed the processing of CD44 by analyzing CD44β because of its ease in analysis compared to the more difficult to detect ICD. As shown in Fig. 5(a), a residual activity on PS endoproteolysis was observed for the G382S mutant, which was considerably lower than that of G382A, whereas the G382T and G382V mutants did not support PS endoproteolysis. Again, none of the mutants were capable of CD44 processing (Fig. 5a). With respect to processing of APP, the G382S mutant displayed a strongly reduced, extremely weak activity, much lower than that of G382A, while the G382T and G382V mutants were inactive (Fig. 5b). The same profile was observed for the processing of Notch1 (Fig. 5c). Quantitation of CD44β, AICD and NICD levels confirmed the strong block (APP and Notch1) and total inactivation (CD44) of processing, respectively, by the G382S mutant (Fig. 5d–f). Taken together, these data show that glycine G382 of the GxGD motif of PS1 is highly critical for γ-secretase substrate processing. They demonstrate a strict requirement for small side chains for this residue of the active site domain and show that γ-secretase substrates exist, such as CD44 that do not tolerate the smallest possible alteration (i.e. to an alanine side chain) of this residue.
Figure 5. Identification of a small side chain requirement at G382 for γ-secretase substrate processing. (a) PS−/− MEF cells were transiently co-transfected with CD44ΔE-Flag and the indicated PS1 constructs. Cell lysates were analyzed for PS1 expression and endoproteolysis, maturation of NCT by immunoblotting as described in Fig. 3(a), and for CD44 expression by immunoblotting with anti-myc antibody 9E10. Note that the CD44ICD cannot be detected by direct immunoblotting. Generation of CD44β was analyzed from conditioned media as in Fig. 4(a). (b) PS−/− MEF were transiently co-transfected with APPsw-6myc and the indicated PS1 constructs and analyzed for APP CTFs and AICD by immunoblotting as described in Fig. 3(a). The asterisk denotes an unspecific band and the circle the faint AICD band detectable for the PS1 G382S mutant. (c) PS−/− MEF were transiently co-transfected with F-NEXT and the indicated PS1 constructs. Expression and processing of F-NEXT was analyzed by immunoblotting as in Fig. 3(b). The circle denotes the weak NICD band detectable for the PS1 G382S mutant. (d)–(f) Levels of CD44β (d), AICD (e), NICD (f) were analyzed as in Figs 4(a) and 3(a),(b), respectively and quantified by measuring their immunoblot signal intensities densitometrically. These were normalized to the levels of NCTm [analyzed as in Fig. 4(a) and quantified densitometrically] and expressed relative to those generated in cells expressing wt PS1 that was set 100%. Bars represent the mean of three independent experiments ± SE. Asterisks indicate the significance (one way anova with Dunnett’s post-test) relative to PS1 wt (**p < 0.01).
Download figure to PowerPoint
Based on the latter conclusion, we finally asked whether CD44 processing would also be sensitive to the corresponding alanine substitution of the C-terminal glycine in the GxGD motif of PS1, i.e. G384A, a severe FAD mutation and the only known glycine substitution functionally tolerated at this site (Steiner et al. 2000; Sato et al. 2006a; Tolia et al. 2006). Very similar to the analogous PS1 G382A mutant, CD44 cleavage was strongly impaired and nearly completely blocked by the PS1 G384A mutant (Fig. 6a). In contrast, the PS1 C92S or PS1 Δexon9 FAD mutations investigated for comparison were capable of maintaining processing of CD44 at substantial levels (Fig. 6a), PS1 C92S being more active than PS1 Δexon9. In contrast to their effect on CD44 substrate cleavage, both PS1 G382A and PS1 G384 substantially supported processing of APP (Fig. 6b), as did the other two FAD mutants investigated for comparison. Quantitation of the CD44β levels generated from the CD44 substrate by wt PS1 and the mutants investigated further confirmed these observations (Fig. 6c). Thus, processing of CD44 does also not tolerate a glycine to alanine substitution of the C-terminal glycine residue of the GxGD active site motif. To further prove that processing of CD44 is not supported by mutation of G384, we additionally analyzed the previously described inactive PS1 G384P and G384K mutants (Steiner et al. 2000) on the processing of this substrate. As expected from our previous results, which demonstrated that these drastic amino acid substitutions inactivate γ-secretase toward APP and Notch1 substrate processing (Steiner et al. 2000), the PS1 G384P and G384K mutants were not capable to support CD44 cleavage (Fig. 6d). Taken together, we conclude that processing of CD44 is sensitive to the smallest possible side chain alteration of both glycine residues of the GxGD active site motif.
Figure 6. Processing of CD44 is sensitive to side chain alterations of G384. (a) PS−/− MEF cells were transiently co-transfected with CD44ΔE-Flag and the indicated PS1 constructs. Cell lysates were analyzed for PS1 expression and endoproteolysis and for maturation of NCT by immunoblotting as described in Fig. 3(a), and for CD44 expression by immunoblotting as in Fig. 5(a). Generation of CD44β was analyzed from conditioned media as in Fig. 4(a). (b) PS−/− MEF were transiently co-transfected with APPsw-6myc and the indicated PS1 constructs. APP CTFs and AICD were analyzed as described in Fig. 3(a). (c) Levels of CD44β and NCTm, analyzed as in Figs 4(a) and 3(a), respectively, were quantified by measuring their immunoblot signal intensities densitometrically. CD44β/NCTm ratios were expressed relative to those generated in cells expressing wt PS1 that was set 100%. Bars represent the mean of three independent experiments ± SE. Asterisks indicate the significance (one way anova with Dunnett’s post-test) relative to PS1 wt (**p < 0.01). (d) PS−/− MEF cells were transiently co-transfected with CD44ΔE-Flag and the indicated PS1 constructs and analyzed for PS1 expression and endoproteolysis and for maturation of NCT by immunoblotting as described in Fig. 3(a), and for generation of CD44β as in Fig. 4(a).
Download figure to PowerPoint
- Top of page
- Experimental procedures
To broaden our understanding of the requirements of the γ-secretase active site domain in PS for substrate cleavage, we addressed the functional role of the glycine residue G382 in PS1 on γ-secretase activity. This residue is the N-terminal one of two conserved glycine residues, which comprise the non-classical GxGD signature active site motif of PS and related aspartyl proteases of the GxGD-type (Fluhrer et al. 2009). Our data demonstrate that glycine 382 is a functionally highly critical residue of the GxGD motif of PS1, which affects the processing of at least three different γ-secretase substrates, APP, Notch1 and CD44 in a rather similar manner. Only a very subtle alteration of this residue to alanine is tolerated functionally whereas all other mutations investigated consistently block the proteolytic activity of γ-secretase. The requirement for a small side chain at this residue is rather strict as residues with slightly larger side chains than glycine and alanine, such as serine had either strongly reduced activity, or were inactive such as threonine, aspartate or valine. In support of our findings, the G382C mutant, which was not included in our analysis, also rendered γ-secretase inactive (Sato et al. 2006a; Tolia et al. 2006). Strikingly, however, as found for CD44, the presence of glycine can even become an absolute requirement for γ-secretase. CD44 is a major cell adhesion molecule that is implicated in tumor invasion and metastasis and its ICD has been shown to activate gene transcription via 12-O-tetradecanoylphorbol-13-acetate (TPA)-responsive elements including its own gene (Okamoto et al. 2001). The presence of G382 at the GxGD motif in PS1 thus ensures the ICD-mediated signaling function of CD44. Unlike APP and Notch1, this highly critical substrate is already sensitive to the minimal possible side chain alteration of glycine to alanine at the position of residue 382. The same biochemical behaviour was observed for the PS1 G384A mutant, i.e. the C-terminal glycine of the GxGD motif, which also blocks the processing of CD44 but substantially maintains that of APP suggesting a similar sensitivity of this substrate to steric alteration at this residue position of the protease active site domain of PS1.
Comparing our results to the data we obtained previously in a mutagenesis study of G384 of PS1 (Steiner et al. 2000) we note that the very same mutations of G382, which inhibited γ-secretase activity were also inhibitory when introduced at the G384 residue. Apparently, a critical requirement for small side chains residues also exists for substrate cleavage at this residue (Steiner et al. 2000; Sato et al. 2006a; Tolia et al. 2006). Only the subtle G384A mutation was functionally tolerated and allowed γ-secretase activity, whereas the G384C mutant (Sato et al. 2006a; Tolia et al. 2006) or the G384D mutant (Steiner et al. 2000), with their slightly larger side chains were already inactive in γ-secretase substrate processing as were all other and more bulkier substitutions tested (Steiner et al. 2000). With respect to APP processing, the subtle G382A and G384A mutations, while allowing γ-secretase activity, differentially altered their cleavage specificity. Whereas the PS1 G384A mutant displayed a strongly increased Aβ42/Aβtotal ratio, the PS1 G382A mutant not only increased the generation of Aβ42 but also that of Aβ43 and the shorter species Aβ38.
Of further note in comparison with the G384 mutagenesis data is the observation that the very same G382 mutations that affected γ-secretase activity also inhibited PS endoproteolysis. These data suggest that mutations of G382 in PS1 have a two-fold impact on γ-secretase both on PS endoproteolysis and on γ-secretase activity. This behavior is different from our previous mutagenesis study of glycine 384 of PS1. Among the G384 mutants investigated only the G384K and G384D mutant inhibited PS endoproteolysis, while the other mutants underwent endoproteolysis (Steiner et al. 2000) (see also Fig. 5d). In agreement with our observations, differential effects on PS endoproteolysis were also observed for the corresponding glycine to cysteine substitutions G382C (not endoproteolysed) and G384C (endoproteolysed) (Sato et al. 2006a; Tolia et al. 2006). These data are consistent with several other observations indicating distinct biochemical and pharmacological properties of PS endoproteolysis and γ-secretase substrate cleavage (Xia 2008).
Although controversial (Chavez-Gutierrez et al. 2008), γ-secretase substrate recognition and cleavage likely involves an initial substrate-binding step by NCT (Shah et al. 2005; Dries et al. 2009). Following or along with this putative first recognition step which might serve to monitor whether the substrate has undergone the necessary ectodomain truncation (Struhl and Adachi 2000), the substrate comes in contact with one or possibly more binding sites prior to its movement to the separate catalytic site (Annaert et al. 2001; Esler et al. 2002; Tian et al. 2002, 2003). Such a substrate-binding site, termed docking site (Esler et al. 2002; Kornilova et al. 2005) has been identified in PS very close to and probably partially overlapping with the catalytic site (Kornilova et al. 2005). The docking site has not been clearly defined yet and may actually represent several distinct substrate binding sites. Initially, the site was mapped at the NTF/CTF interface (Kornilova et al. 2005). However, recent data have also selectively identified the NTF (Sato et al. 2008b) as well as the CTF (Sato et al. 2008a). In the latter, a single docking site residue has been identified in the luminal oriented side of TMD9 (Sato et al. 2008a), which has been proposed to possibly function as lateral gate for substrate entry to the catalytic active site (Sato et al. 2008a; Tolia et al. 2008). Our data indicate that the GxGD motif may be an additional critical part of the γ-secretase active site domain coming in contact with the substrate prior to the scissile bond cleavages at the catalytic aspartate residues. This is suggested by the lack of processing of three different γ-secretase substrates, APP, Notch1 and CD44 with a very similar profile by the inactive G382 mutants. The apparent requirement for small side-chains, which suggests strict steric requirements for substrate-protease interactions close to the active site, is consistent with this view. Recent data indicate that the conserved PxL motif (PAL in PS) at the cytosolic membrane border of TMD9 could be part of the same functional region of the active site domain as the GxGD motif (Sato et al. 2008a; Tolia et al. 2008), both motifs possibly providing subsite residues. This would be consistent with the observation that γ-secretase activity is very sensitive to mutational alteration in the PAL region. Like for the invariant glycine residues of the GxGD motif mutations only very few amino acid residue exchanges are tolerated for γ-secretase activity in the PAL motif (Nakaya et al. 2005; Wang et al. 2006). Mutational analysis has further shown that the PAL motif is critical for the normal conformation of the γ-secretase catalytic site (Wang et al. 2006), which is in line with the motif being an essential part of the γ-secretase active site domain. These properties of the PAL motif have also been demonstrated for SPP (Wang et al. 2006). It is therefore likely that the glycines of the GxGD motif are also critical residues in the active site domains of the other GxGD-type protease families. Although systematic mutagenesis studies addressing this issue have not been reported yet, mutants in SPP and in SPPL2b analogous to the PS1 G384A mutant have, like the latter, a lower proteolytic activity than the wt enzyme (Sato et al. 2006b; Fluhrer et al. 2008). Interestingly, residue x of the GxGD motif is a longer hydrophobic side chain residue in PS, SPP and SPPL. The reason for this particular side chain arrangement preceding the catalytic aspartate (i.e. small-large-small) is not clear yet and will possibly only be resolved when structural information on the active site domain of γ-secretase or SPP becomes available at atomic resolution. The finding, however, that the intermediate residue x confers APP/Notch substrate selectivity reinforces that a certain spatial side chain arrangement in the active site domain of γ-secretase is critical for substrate binding and cleavage (Yamasaki et al. 2006).
We conclude that the presence of small side chains in the GxGD motif, which is optimally fulfilled by the naturally occurring two conserved glycines as part of this motif is an important structural requirement for γ-secretase substrate cleavage. We propose that the evolutionary conserved presence of these particular glycine residues serve to minimize steric hindrance effects in protease-substrate interactions close to the active site or to provide flexibility required for effective substrate binding and/or positioning. Providing space and/or flexibility are unique features of glycine, which has the smallest possible amino acid side chain, i.e. a hydrogen atom, allowing access of the main chain to a much larger conformational space. These features make glycine particularly suitable for conformationally constrained close contacts as possibly required for the interaction of γ-secretase with its substrates. The presence of glycine can thereby allow optimal binding/docking of the various γ-secretase substrates, all varying in their TMD sequences, close to the active site prior to their subsequent cleavage. Glycine residues would also facilitate conformational changes potentially required for efficient substrate binding or processing. The presence of glycines at the particular positions in vicinity to the catalytic site provides a plausible way of how the γ-secretase ensures cleavage of critical substrates such as the signaling protein CD44 shown in this study, and possibly of others as well. We propose that glycines 382 and 384 of the GxGD motif in PS1 are probably conserved during evolution to allow that γ-secretase can efficiently cleave all of its substrates including substrates very sensitive to steric and/or conformational alteration of the PS active site domain, such as CD44. These findings shed light on the requirements of an unusual aspartyl protease active site evolved for intramembrane cleavage and broaden our understanding of γ-secretase substrate recognition and cleavage, which may prove crucial for therapeutic targeting of the enzyme.