Random Mutagenesis of Presenilin-1 Identifies Novel Mutants Exclusively Generating Long Amyloid β-Peptides

Familial Alzheimer disease-causing mutations in the presenilins increase production of longer pathogenic amyloid β-peptides (Aβ42/43) by altering γ-secretase activity. The mechanism underlying this effect remains unknown, although it has been proposed that heteromeric macromolecular complexes containing presenilins mediate γ-secretase cleavage of the amyloid β-precursor protein. Using a random mutagenesis screen of presenilin-1 (PS1) for PS1 endoproteolysis-impairing mutations, we identified five unique mutants, including R278I-PS1 and L435H-PS1, that exclusively generated a high level of Aβ43, but did not support physiological PS1 endoproteolysis or Aβ40 generation. These mutants did not measurably alter the molecular size or subcellular localization of PS1 complexes. Pharmacological studies indicated that the up-regulation of activity for Aβ43 generation by these mutations was not further enhanced by the difluoroketone inhibitor DFK167 and was refractory to inhibition by sulindac sulfide. These results suggest that PS1 mutations can lead to a wide spectrum of changes in the activity and specificity of γ-secretase and that the effects of PS1 mutations and γ-secretase inhibitors on the specificity are mediated through a common mechanism.

heterogeneous series of N-and C-terminal products, named amyloid β peptides (Aβ) and APP intracellular domain (AICD), respectively.
The process regulating these heterogeneous intramembranous cleavages is poorly understood but is a central issue in Alzheimer's disease (AD) research for three reasons. First, the longer Aβ species ending at residue 42 or 43 (Aβ 42/43 ) are thought to play a critical role in the pathogenesis of AD, although they are minor products of normal γ-secretase activity (6). Second, clinical mutations in the presenilins (PSs), which are responsible to the majority of familial AD (FAD) pedigrees, cause increased ratio of Aβ 42/43 isoforms to total Aβ (7). Third, medication that modulates the Aβ 42/43 production by altering γ-secretase activity could provide an effective therapy for AD.  (8,9). This endoproteolysis yields N-and C-terminal fragments (NTF and CTF, respectively) that are stabilized and assembled into functional macromolecular complexes together with at least three other membrane proteins, nicastrin (Nct), APH-1 and PEN-2 (10,11). The abundance and stoichiometry of PS complexes are tightly regulated, and exogenously over-expressed PSs can replace endogenous PSs (8). Recent studies have suggested that these multiprotein complexes mediate γ-secretase cleavage and probably PS endoproteolysis and that PSs may be the catalytic components of the complexes (2,(12)(13)(14)(15).
Furthermore, it has been presumed that FAD-associated mutations of the PSs significantly alter γ-secretase activity and enhance production of Aβ 42/43 through a gain-of-function effect.
However, it is still unclear how more than a hundred of FAD-linked missense mutations scattered throughout the sequence of PSs can selectively upregulate the enzymatic production of Aβ 42/43 .
While trying to address this conundrum, we realized that the FAD-linked mutations and the few site-directed mutagenesis studies described to date (12,(16)(17)(18)(19) did not provide an unbiased view of the effects of PS mutations on γ-secretase activity. Thus, the clinical mutations were ascertained by linkage to FAD, whereas the site-directed mutations in conserved residues were created to test specific hypotheses. We therefore undertook a screen of randomly mutagenized PS1 to identify single-residue missense mutations that affected the physiological endoproteolysis of PS1 and/or selectively changed Aβ isoform production.
We report here the discovery and characterization of novel mutant PS1 that exclusively generate longer Aβ species, but which do not support the activities for Aβ 40 generation or PS1 endoproteolysis.

EXPERIMENTAL PROCEDURES
Antibodies and reagents -The rabbit polyclonal antibody against the N-terminus of PS1 was previously described (10   which were S141R, T245I, R278I, P433L, A434T

Random mutagenesis screen
and L435H ( Fig. 1A and B). The S141 and T245 residues are located within the second and sixth although the complexes containing S141R and T245I mutants were also distributed broadly into the lower MW ranges (Fig. 1C). The mutant holoproteins also displaced the proteolytic fragments of endogenous presenilins (Fig. 1C). To assess the γ-secretase activity of these mutants in the absence of endogenous PSs, we transfected each of these PS1 mutants together with human APP into PS-null MEFs using retroviral vectors and then measured Aβ levels in the conditioned media from these cells (Fig. 3A).  (Fig. 3C). These results confirm that γ-secretase activity was modified by the specific amino acid substitution rather than due to the lack of endoproteolysis of the mutant holoprotein.  (Fig. 4A and B). Both PS1 species appeared to be similarly transported from the ER to the Golgi apparatus. Second, the pattern of Nct glycosylation in R278I-PS1-expressing cells was indistinguishable from that in WT-PS1-expressing cells (Fig. 4C).
Finally, the selective over-production of Aβ 42/43 by R278I-or L435H-PS1 cells was still evident even when Aβ peptides were generated in a solubilized membrane fraction prepared from these cells (Fig. 4D).
Different amino acid substitutions at the same residue can result in distinct modifications of γ-secretase activity -The PS1 R278I mutation conferred exclusive generation of Aβ 43 , whereas the P433L mutation resulted in loss-of-function.
To address whether these substitutions have a specific role in PS1 endoproteolysis and γ-secretase activities, we generated and tested additional mutations at these residues. We introduced five other mutations into the R278 codon, including FAD-linked R278K and R278T as well as substitutions to S, E, and P, which have polar, negatively charged, and imino group side chains, respectively. Unlike the R278I mutation, all of these substitutions increased the levels of secreted Aβ 42/43 , but did not affect either PS1 endoproteolysis or Aβ 40 secretion (Fig. 5A).
Next we generated seven mutations at the P433 residue and tested γ-secretase activity (  Additionally, it appears that PS1 LA mutations cause similar but exaggerated modification on Aβ secretion as compared to FAD-linked mutations. It has been reported that DFK167 inhibits PS1 endoproteolysis in an in vitro assay system (39). However, when DFK167 was applied to cells expressing WT-PS1 or FAD-linked I143F-PS1, the endoproteolysis was unaffected (Fig. 6C). Unexpectedly, when DFK167 was applied to cells expressing the R278I-or L435H-PS1 mutant, the endoproteolysis was activated in a dose-dependent manner (Fig. 6C).
This paradoxical activation was highly specific to these mutants; DFK167 had no effect on the failed endoproteolysis of loss-of-function mutants such as D385A-or P433L-PS1.

DISCUSSION
In this study, we found six missense mutations of PS1 by a random mutagenesis screen for PS1-endoproteolysis-impaired mutants.
These mutations fell into two categories: PS1 LA and the loss-of-function mutant P433L-PS1.
These mutations did not significantly affect the complex assembly and stabilization of uncleaved mutant holoproteins. PSase cleavage is tightly related to γ-secretase activity, although PS endoproteolysis is not absolutely required for the activation of γ-secretase (12,40). Both activities required two intramembranous aspartyl residues at the putative catalytic sites of PS (D257 and D385 of PS1) (12), and certain γ-secretase inhibitors blocked PSase cleavage as well (41,42).
However, the profiles of their response to inhibitors were different from one another (39,41).
Additionally, a recent report proved that D257Aor D385A-PS1 holoprotein was cleaved, although with low efficiency, without any concomitant γ-secretase activity, when stably expressed in PS-null cells (43). Nevertheless, in most cases, except mutants harboring mutations at the cleavage site (e.g. M292D-PS1) (40,44,45), the previously reported cleavage-defective mutations simultaneously conferred definitive modification of γ-secretase activity and/or specificity (12,16,17,25,46). This is the case for PS1 LA  Meanwhile, during preparation of this manuscript, R278I PS1 LA mutation has been reported to co-segregate with family members affected with early-onset, atypical FAD (48). These suggest that PS1 LA mutants have similar but exaggerated effects on γ-secretase activity as compared to typical FAD mutants. PS1 mutations appear to result in a wide spectrum of γ-secretase activity that is associated with a shift in cleavage specificity.
Two of the five PS1 LA mutations and the P433L loss-of-function mutation were clustered at the highly conserved PAL sequence in the C-terminal tail. The R278 residue is remote on the sequence from, but can be located close to the PAL motif near the cytoplasmic face of the membrane on the tertiary conformation based on the eight-TM model of PS1 proteins ( Fig. 1B; 49).
These residues may compose of a conformational domain that is important not only for the activity but also for the cleavage specificity of γ-secretase.
However, the two other mutations S141R and T245I, which had a weaker effect, were not binding (38,47,50,51). Several of our observations suggest that the effects of PS1 mutations and the allosteric effects of γ-secretase inhibitors are likely to be tightly related, and we speculate that these effects might even be mediated by a shared mechanism. First, our results and previously published data (32)(33)(34) show that PS1 mutations clearly modify the response of γ-secretase to these inhibitors.