Enzymatic Characteristics of I213T Mutant Presenilin-1/γ-Secretase in Cell Models and Knock-in Mouse Brains

Presenilin (PS)/γ-secretase-mediated intramembranous proteolysis of amyloid precursor protein produces amyloid β (Aβ) peptides in which Aβ species of different lengths are generated through multiple cleavages at the γ-, ζ-, and ϵ-sites. An increased Aβ42/Aβ40 ratio is a common characteristic of most cases of familial Alzheimer disease (FAD)-linked PS mutations. However, the molecular mechanisms underlying amyloid precursor protein proteolysis leading to increased Aβ42/Aβ40 ratios still remain unclear. Here, we report our findings on the enzymatic analysis of γ-secretase derived from I213T mutant PS1-expressing PS1/PS2-deficient (PS–/–) cells and from the brains of I213T mutant PS1 knock-in mice. Kinetics analyses revealed that the FAD mutation reduced de novo Aβ generation, suggesting that mutation impairs the total catalytic rate of γ-secretase. Analysis of each Aβ species revealed that the FAD mutation specifically reduced Aβ40 levels more drastically than Aβ42 levels, leading to an increased Aβ42/Aβ40 ratio. By contrast, the FAD mutation increased the generation of longer Aβ species such as Aβ43, Aβ45, and >Aβ46. These results were confirmed by analyses of γ-secretase derived from I213T knock-in mouse brains, in which the reduction of de novo Aβ generation was mutant allele dose-dependent. Our findings clearly indicate that the mechanism underlying the increased Aβ42/Aβ40 ratio observed in cases of FAD mutations is related to the differential inhibition of γ-site cleavage reactions, in which the reaction producing Aβ40 is subject to more inhibition than that producing Aβ42. Our results also provide novel insight into how enhancing the generation of longer Aβs may contribute to Alzheimer disease onset.


Presenilin (PS)/␥-secretase-mediated intramembranous proteolysis of amyloid precursor protein produces amyloid ␤ (A␤)
peptides in which A␤ species of different lengths are generated through multiple cleavages at the ␥-, -, and ⑀-sites. An increased A␤42/A␤40 ratio is a common characteristic of most cases of familial Alzheimer disease (FAD)-linked PS mutations. However, the molecular mechanisms underlying amyloid precursor protein proteolysis leading to increased A␤42/A␤40 ratios still remain unclear. Here, we report our findings on the enzymatic analysis of ␥-secretase derived from I213T mutant PS1-expressing PS1/PS2-deficient (PS ؊/؊ ) cells and from the brains of I213T mutant PS1 knock-in mice. Kinetics analyses revealed that the FAD mutation reduced de novo A␤ generation, suggesting that mutation impairs the total catalytic rate of ␥-secretase. Analysis of each A␤ species revealed that the FAD mutation specifically reduced A␤40 levels more drastically than A␤42 levels, leading to an increased A␤42/A␤40 ratio. By contrast, the FAD mutation increased the generation of longer A␤ species such as A␤43, A␤45, and >A␤46. These results were confirmed by analyses of ␥-secretase derived from I213T knock-in mouse brains, in which the reduction of de novo A␤ generation was mutant allele dose-dependent. Our findings clearly indicate that the mechanism underlying the increased A␤42/A␤40 ratio observed in cases of FAD mutations is related to the differential inhibition of ␥-site cleavage reactions, in which the reaction producing A␤40 is subject to more inhibition than that producing A␤42. Our results also provide novel insight into how enhancing the generation of longer A␤s may contribute to Alzheimer disease onset.
Amyloid ␤ (A␤) 2 is a hydrophobic peptide that pathologically deposits in the brains of Alzheimer disease (AD) patients. The significant neurotoxicity of oligomeric and/or fibrillar A␤ aggregates indicates that accumulation of A␤ is a central pathogenic event in AD (1). Sequential cleavage of ␤-amyloid precursor protein (APP) by ␤and ␥-secretases releases A␤ into the luminal/extracellular space. ␤-secretase is a membranebound aspartic protease (identified as BACE) that cleaves the extracellular region of APP to produce membrane-spanning APP C-terminal fragment ␤ (termed CTF␤ or C99) and a N-terminal secreted form of APP␤. The intramembranous region of CTF␤ is cleaved next by ␥-secretase to produce A␤ and APP intracellular domain, and because of the loose site specificity of ␥-secretase-mediated proteolysis, various C-terminal truncated A␤ species, including two major species, A␤40 and A␤42, are also generated (1). A␤40 is the most predominant species of secreted A␤. A␤42 is hypothesized to be the trigger species for AD-related amyloid pathophysiology, because it has much faster aggregation potential than A␤40. Indeed, histochemical and biochemical studies have revealed that A␤42 primarily deposits within the brains of AD patients and several animal models (2)(3)(4). Most importantly, mutations of APP, presenilin-1 (PS1), and PS2 have been identified in familial Alzheimer disease (FAD) and have been found to specifically increase the A␤42/A␤40 ratio in cell medium or animal tissues and to accelerate the parenchymal accumulation of A␤ (5)(6)(7)(8)(9). In this way, ␥-secretase-mediated A␤ metabolism may be critically involved in the onset of AD.
Accumulating evidence strongly indicates that ␥-secretase is a high molecular weight membrane protein complex in which PS serves as a catalytic subunit (10). PS proteins are hydrophobic multiple membrane-spanning proteins that become activated through endoproteolysis of its large cytosolic loop domain, producing N-and C-terminal fragments (11)(12)(13)(14). The * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Fig. 1 N-and C-terminal fragments incorporate into the stable high molecular weight complex, which consists of at least three minimum cofactors: nicastrin, APH-1, and PEN-2 (11)(12)(13). Two conserved aspartate (Asp 257 and Asp 385 in PS1) residues, which are located within transmembrane domains 6 and 7, form the predicted active center of the PS/␥-secretase complex (10). Various type I membrane protein substrates, including APP and Notch, undergo regulated intramembranous proteolysis within the hydrophobic lipid bilayer environment (14,15).
On the basis of this model, we hypothesize that the mechanisms underlying the increased A␤42/A␤40 ratio observed in cases of FAD-linked PS mutations might directly reflect an alteration of the enzymatic characteristics of the PS/␥-secretase complex. More than 150 FAD-linked mutations have been identified in PS genes (1). These widely distribute to every sequence of PS proteins, indicating that each mutation can potentially modify the catalytic reactions of ␥-secretase enzyme in a distinct manner. Numerous cell culture studies have revealed various aspects of PS mutations and their effects; nevertheless, experimental differences, such as the variations in cell models, detection systems, and normalization methodology, among individual studies make the study of PS mutations through living cell-based methods a controversial issue (16,17). Traditionally, ␥-secretase was believed to mainly hydrolyze the covalent bonds at Val 40 -Ile 41 and Ala 42 -Thr 43 of CTF␤ (termed ␥-sites), generating A␤40 and A␤42, respectively (1). The ␥-sites were thus regarded as the major cleavage sites targeted by ␥-secretase. However, recent studies have revealed novel cleavage sites: ␥-secretase also mediates the cleavage ofand ⑀-sites, which are closer to the cytoplasmic membrane boundary (18 -20). Cleavage at theand ⑀-sites produces various longer A␤ (ϾA␤43) species, and evidence indicates that these species may be processed to shorter A␤ species in stages (21,22). Taken together, these findings suggest that ␥-secretase-mediated proteolysis consists of multiple complicated cleavage reactions that relate to each other along the transmembrane domain of APP-CTF␤. Nonetheless, it is still unclear how FAD mutations modify overall ␥-secretase-mediated cleavage of APP, leading to increased A␤42/A␤40 ratios. The molecular mechanisms underlying this process also remain elusive. This crucial issue requires more careful assessment from the aspect of the enzymatic characteristics of the PS/␥-secretase complex.
These findings prompted us to evaluate the activity kinetics of the wild-type (WT) and FAD mutant PS1/␥-secretase enzyme by using a CHAPSO solubilization ␥-secretase assay system (23,24). This strategy has great advantages because it enables us not only to directly assess the effect of mutations on the enzyme in cell models but also in brain tissues, which should better reflect the physiological status of the enzyme in vivo. We previously generated PS1/PS2-deficient cell lines that stably express either the WT or FAD mutant forms of human PS1 (25). Consistent with previous reports (8,9), we demonstrated that FAD mutations reduced the secretion of A␤ peptides (25). Thus, one possible mechanism underlying increased A␤42/A␤40 ratios is that FAD mutations might directly impair the enzymatic activity of ␥-secretase, thereby modifying the individual cleavage reactions that produce A␤40 and A␤42. In this study, we specifically focused on the effect of the I213T FAD-linked PS1 mutation on ␥-secretase and reported the enzymatic characteristics of ␥-secretase derived both from cell models and from the brains of knock-in mice.
Plasmid DNA Constructs and Cell Cultures-PS ϩ/ϩ and PS1/ PS2-deficient (PS Ϫ/Ϫ ) mouse embryonic fibroblast (MEF) cells (kindly provided by Dr. B. De Strooper, K. U. Leuven and Flanders Interuniversitary Institute for Biotechnology, Belgium) (26,27) were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 100 g/ml gentamicin at 37°C in a 5% CO 2 incubator. PS Ϫ/Ϫ MEFbased cell lines stably expressing either WT or I213T FAD mutant human PS1 were described previously (25). To establish cell lines that stably express dominant negative mutant D257A human PS1, we generated cDNA by using a conventional two-step PCR protocol and a primer pair for D257A (5Ј-ATT TCA GTA TAT GCT TTA GTG GCT GTT-3Ј and 5Ј-AAC AGC CAC TAA AGC ATA TAC TGA AAT-3Ј). The cDNA was then subcloned into the EcoRI/XhaI site of a pCI-Bls vector; the DNA sequence was confirmed with an ABI DNA sequencer. After transfection into PS Ϫ/Ϫ MEF cells, several clonal lines were selected in 7.5 g/ml blasticidin S. All stable cell lines were maintained in growth medium containing 5.0 g/ml blasticidin S.
I213T Knock-in Mice-We genotyped 11.0 -12.5-month-old homozygous (I213T/I213T, n ϭ 4), hemizygous (ϩ/I213T, n ϭ 9), and control (ϩ/ϩ, n ϭ 5) I213T mutant PS1 knock-in mice for WT and mutant alleles by PCR, as described previously (28,29). Fresh whole brains were immediately dissected and used for further biochemical preparations and enzyme assays. All procedures involving animals and their care were approved by the Animal Care Use Committee of the RIKEN Brain Science Institute.
Calculation of Kinetics Constants-Michaelis-Menten kinetics constants (V max and K m ) were calculated by fitting experimental data to Lineweaver-Burk plots according to the equa- is the reaction velocity of A␤ generation, and [S] is the concentration of the APP-C99-FLAG substrate. To calculate the values of V max and K m , experimentally measured values of [V] determined from experiments using a range of high C99-FLAG concentrations were used for linear fitting because [V] measured from experiments using low C99-FLAG concentrations contains a large experimental error, which interferes with appropriate linear fitting.
Statistical Analysis-Statistical significance of data were tested with a Student's t test or a Tukey-Kramer test. Data were analyzed with InStat version 3.0a (GraphPad).

RESULTS
Enzymatic Characteristics of Wild-type PS1/␥-Secretase in PS Ϫ/Ϫ MEF Cell Models-To characterize ␥-secretase activity using a CHAPSO-solubilized assay system, first we solubilized purified crude microsomal membranes from PS ϩ/ϩ MEF, PS Ϫ/Ϫ MEF, and WT line 38 (WT38) cells, which stably express endogenous levels of WT human PS1 on a PS Ϫ/Ϫ MEF background, and incubated these membranes with 500 nM C99-FLAG. Immunoblotting with anti-A␤ N-and C-terminal endspecific antibodies showed that PS ϩ/ϩ MEF lysates displayed robust amounts of de novo A␤ generation, including A␤40 and A␤42 generation after 4 h of incubation (Fig. 1, A and B). Relatively higher A␤42/A␤40 ratios (about 40 -50%) compared with secreted A␤ ratios were observed in this assay system (25). By contrast, immunoblotting of PS Ϫ/Ϫ MEF lysates failed to detect any A␤, even after 4 h of incubation, thus completely excluding possible PS-independent nonspecific de novo A␤ generation activity due to contaminating proteases. As expected, expression of WT human PS1 significantly restored de novo A␤ generation activity to levels similar to those detected in PS ϩ/ϩ MEF lysates (Fig. 1, A and B). We observed a slightly lower A␤42/A␤40 ratio in reaction mixtures containing WT38 lysates compared with those containing PS ϩ/ϩ MEF lysates. This difference might reflect the absence of PS2 or differences between mouse and human PS1 species.
To examine the reaction kinetics of in vitro ␥-secretase, next we monitored the time course of A␤ generation. The levels of each A␤ species increased in a linear fashion for up to 5 h, indicating that the initial velocity of the enzyme-substrate reaction was stable during this period (data not shown). The rate of A␤ generation fit to a Michaelis-Menten-like reaction curve (Fig. 1C). V max values for PS ϩ/ϩ MEF and WT38 lysates were 790.60 Ϯ 37.01 pM/min and 686.80 Ϯ 198.73 pM/min, respectively, and K m values were 2.08 Ϯ 0.13 M and 2.26 Ϯ 1.40 M, respectively (Fig. 1C). These results approximately corresponded to those of previous studies (23,24,31) and suggest that the enzymatic characteristics of ␥-secretase in both types of cell lysates are not much different. The V max and K m values of A␤40 for PS ϩ/ϩ MEF lysates were 150.99 Ϯ 33.77 pM/min and 0.90 Ϯ 0.30 M, respectively, and the V max and K m values of A␤40 for WT38 were 124.27 Ϯ 12.68 pM/min and 0.70 Ϯ 0.07 nM, respectively. However, we could not estimate the V max and K m of A␤42 because de novo A␤42 levels continuously increased with increasing concentrations of C99-FLAG, resulting in a linear fit to the Lineweaver-Burk plot, which did not provide us with appropriate intercept values of the 1/[V] axis and 1/[S] axis. This difference in kinetics suggests the possibility that A␤40 and A␤42 are generated from distinct reactions. As recent studies have mentioned, the A␤42/ A␤40 ratio gradually increases with increasing C99-FLAG concentrations (24,32).
Accumulating evidence strongly indicates that PS itself works as the catalytic subunit of the ␥-secretase complex. We confirmed this premise by demonstrating that expression of dominant negative human D257A mutant PS1 failed to restore de novo A␤ synthesis in PS Ϫ/Ϫ MEF cells and that 1% Triton X-100 solubilization, which disrupts the interaction of PS complex components, completely abolished A␤ generation (data not shown). The ␥-secretase-specific inhibitors DAPT (data not shown) and WPE-III-31C (Fig. 1D) significantly suppressed the generation of each A␤ species in a dose-dependent manner at IC 50 values of ϳ100 nM DAPT and ϳ10 nM WPE-III-31C. To evaluate expression levels of PS1 and de novo generation level of A␤, we analyzed six independent stable cell lines, including the WT38 line (supplemental Fig. 1) (25). After incubation of the membrane fractions with 500 nM C99-FLAG, a linear correlation between PS1 levels and A␤ levels was observed (Fig. 1E), even though the A␤42/A␤40 ratio remained unchanged among these cell lines (supplemental Fig. 1C). Taken together, these results indicate that this assay system reflects exactly genuine PS-dependent ␥-secretase activity. Thus, in all of the following experiments, we regarded PS1 protein levels as a relative unit of active enzyme.  Fig. 1). After incubation with 500 nM C99-FLAG at 37°C for 4 h, de novo A␤ species were measured. The relative level of PS1 in WT (line 38) lysate was set as 100% in the value of the x axis. Dots represent the mean values of triplicate analysis using each lysate. JUNE 13, 2008 • VOLUME 283 • NUMBER 24 I213T FAD-linked Mutations Partially Impair ␥-Secretasemediated ␥-Site Cleavage-In our previous study, we reported that expression of FAD mutant PS1 in PS Ϫ/Ϫ MEF cells failed to restore the ability of these cells to secrete A␤ peptides, leading us to propose that FAD mutations might attenuate ␥-site cleavage reactions by directly modifying the enzyme characteristics of ␥-secretase (25). To test this hypothesis, we assessed the rate that PS1/␥-secretase generated A␤s in cells expressing either WT PS1 or I213T FAD mutant PS1. WT and I213T FAD mutant cell lysates were coincubated with C99-FLAG at 0 -2.5 M concentrations, and the rate of A␤ generation was measured. Cell lines expressing similar levels of each PS1 variant were chosen in order to minimize experimental error ( Fig. 2A), and the relative amounts of A␤ per molecule of WT and I213T FAD mutant PS1/␥-secretase were evaluated after normalization with PS1 protein levels. As shown in Fig. 2B, the I213T mutation moderately attenuated the rate of A␤ generation compared with that of WT. Suppression was specific and significant at high concentrations (2.0 -2.5 M) of C99-FLAG substrate. Moreover, V max of A␤ generation also tended to decrease by 30% (Fig. 2, B and C). This suggests that the I213T mutation directly and specifically impaired ␥-secretase enzymatic activity responsible for overall A␤ generation.

Effect of FAD Mutation on Stepwise APP Proteolysis
We further examined the generation rate of both A␤40 and A␤42 species. As expected, the I213T mutation drastically attenuated A␤40 levels compared with WT levels (Fig. 2, B and D) by about 80% at the maximum concentration of C99-FLAG. Although the I213T mutation attenuated A␤42 levels at C99-FLAG concentrations of 1.5-2.5 M (Fig. 2, B and D), levels observed at 0.1-1.0 M C99-FLAG were comparable with those measured from cells expressing the WT enzyme. The reduction of A␤40 levels was more drastic than that of A␤42, indicating that the I213T mutation certainly increases the A␤42/A␤40 ratio (Fig. 2E). Taken together, these results indicate that the I213T mutation impairs ␥-site cleavage reactions affecting both A␤40 and A␤42 generation and that differential efficiency in reducing A␤40 and A␤42 ultimately results in a higher A␤42/ A␤40 ratio.
I213T Mutant PS1 Enzyme Facilitates the Generation of Longer A␤ Species-Recent studies have identified various longer A␤ species (ϾA␤43) generated from ␥-secretase-mediatedand ⑀-site cleavage (21,33). The results shown in Fig. 2  strate that the I213T mutation clearly impairs the production of A␤40 and A␤42, affecting the rate at which these A␤s are generated. Nevertheless, the overall levels of A␤ produced were not greatly affected by the I213T mutation. This suggests that the mutant enzyme may facilitate the generation of longer A␤ species. To address this possibility, we measured the levels of each A␤ species in the reaction mixtures by SDS-PAGE of 8 M ureacontaining modified Tris-Tricine gels, which enabled us to separate various long A␤ species by the lengths of their C-terminal amino acid residues (21). We successfully observed PS-dependent de novo generation of not only A␤40 and A␤42 but also of A␤43, A␤45, and ϾA␤46 species in samples from WT PS1/␥secretase-expressing cells (Fig. 3, A and B). The levels of each A␤ species increased in a linear, time-dependent fashion, and the ␥-secretase inhibitors (DAPT and WPE-III-31C) clearly suppressed this increase in a dose-dependent manner (data not shown). Next, we compared the levels of each A␤ species in cells expressing I213T FAD mutant PS1/␥-secretase to those in cells expressing WT PS1/␥-secretase. Interestingly, the I213T mutation increased the generation rate of longer forms of A␤s, including A␤43, A␤45, and ϾA␤46 species, compared with WT (Fig. 3, A and C). A␤43 levels were significantly enhanced at C99-FLAG substrate concentrations of 0.1-1.5 M, plateauing at higher concentrations. In contrast, ϾA␤46 levels were strongly enhanced at C99-FLAG concentrations of 1.5-2.5 M. Unlike A␤43 and ϾA␤46 levels, A␤45 levels remained enhanced regardless of the C99-FLAG concentration. These results indicate that the I213T mutation facilitates the production of longer A␤ species as opposed to inhibiting the production of shorter A␤ species. These results also suggest that a highly related mechanism may exist that links the generation of A␤ species of different lengths.

demon-
␥-Secretase Activity in I213T Mutant PS1 Knock-in Mouse Brain-To extend the findings obtained from cell models and to examine the physiological status of ␥-secretase activity in vivo, we assessed ␥-secretase activity in samples purified from I213T PS1 knock-in mouse brain. The protein levels of PS1 and of other ␥-secretase complex components in CHAPSO-solubilized microsomal fractions were not significantly different among mice of each genotype (Fig. 4A). Thus, with this model, we can strictly evaluate the characteristics of native ␥-secretase activity without having to consider possible artifacts stemming from overexpression and/or integration of the transgene. To stimulate A␤ generation, we incubated CHAPSOsolubilized fractions containing ␥-secretase with 500 nM C99-FLAG for 4 h. We detected de novo A␤ synthesis in all samples, regardless of genotype; 10 M WPE-III-31C inhibitor almost completely blocked ␥-secretase activity (Fig.  4B). As we observed with the cell models, in I213T PS1 knock-in mice, the I213T mutation significantly reduced A␤40 levels without obviously affecting A␤42 levels, leading to higher A␤42/A␤40 ratios (Fig. 4, C and D). The mutant enzyme also increased A␤43 and A␤45 levels (Fig. 4, E and F). Interestingly, the effects of the I213T mutation on ␥-secretase activity were much more pronounced in brain samples from homozygous (I213T/I213T) mice than in samples from hemizygous (ϩ/I213T) mice, suggesting that the effect of the I213T mutation is dose-dependent. These findings strongly support the results obtained from the cell models and suggest that FAD mutations directly lead to abnormal alterations in ␥-secretase activity, even in the brain under physiological conditions.

DISCUSSION
Although an increased A␤42/ A␤40 ratio is a common consequence of FAD-linked PS1 mutations, the molecular mechanism underlying this increase still remains controversial (16,17). Possibly, the increased A␤42/A␤40 ratio may result from direct alterations in the cleavage reactions catalyzed by mutant PS/␥-secretase such that A␤42 generation is facilitated, A␤40 generation is attenuated, or both A␤40 and A␤42 generation change with a different manner. To clarify this crucial issue, we examined the enzymatic characteristics of WT and I213T FAD mutant PS1/␥-secretase in this study. First, we observed that de novo A␤ generation depended on PS protein levels, indicating that our assay system reflects genuine PS-dependent ␥-secretase activity. Kinetics analyses revealed that the I213T mutation reduced de novo A␤ generation compared with that measured under WT conditions and specifically inhibited the generation of A␤40 more drastically than of A␤42. Importantly, these findings were clearly confirmed in experiments of ␥-secretase derived from I213T mutant knock-in mouse brains, in which reduction of de novo A␤ generation was mutant allele dose-dependent. We reported previously that expression of FAD mutant PS1 in PS Ϫ/Ϫ MEF cells failed to restore the ability of these cells to secrete A␤ (25). Here, our findings suggest that the reduced secretion levels of A␤ directly correspond to a reduction in the rate of mutant PS1/␥-secretase-mediated cleavage of ␥-sites. Taken together, these observations have led us to conclude that the mechanism underlying the increased A␤42/A␤40 ratio observed in cases of FAD mutations most likely has to do with the differential reduction of A␤40 and A␤42, in which A␤40 generation is reduced more than A␤42 generation. Accordingly, the I213T mutation would be expected to lead to a partial loss of ␥-site cleavage, which would affect both A␤40 and A␤42 generation.
Besides ␥-site cleavage at Val 40 and Ile 42 of the C terminus, ␥-secretase also cleaves APP-CTF␤ at two novel sites located closer to the cytoplasmic membrane boundary; these sites are termed -site (Val 46 of the C terminus) and ⑀-site (Thr 48 and Leu 49 of the C terminus) (18 -20). Funamoto et al. (34) demonstrated that cell models expressing A␤49 or A␤48 prefer to secrete A␤40 or A␤42, respectively. In addition, recent studies identified various longer A␤ species (A␤43 to A␤49) and demonstrated that DAPT-treated cells contained elevated A␤43 and A␤46 levels but reduced A␤40 levels (21,22). On the basis of this evidence, it is hypothesized that A␤40 and A␤42 are generated as end-products of the ␥-secretase-mediated stepwise cleavage of A␤49 and A␤48, which occurs at every third residue along the ␣-helix structure of A␤49 and A␤48 after the ⑀-cleavage site (21,22). Interestingly, we observed that in cells expressing I213T mutant ␥-secretase, de novo levels of A␤43 and ϾA␤46 species significantly increased in stages in a C99-FLAG concentration-dependent fashion. If all A␤ species are spontaneously generated from A␤49/A␤48 through a one-step reaction after ⑀-site cleavage, then one would expect the rate of A␤40, A␤43, and ϾA␤46 generation to demonstrate similar reaction curves because these reactions should have similar K m values. However, in cells expressing WT PS1 ␥-secretase, we clearly observed that only A␤40 levels reached a plateau at C99-FLAG concentrations over 1.0 M, whereas A␤43 and ϾA␤46 levels increased linearly in parallel with increasing C99-FLAG concentrations. This finding indicates that each A␤ species is generated from stepwise cleavage (A␤49 3 A␤46 3 A␤43 3 A␤40) rather than spontaneous cleavage. This would suggest that A␤43 3 A␤40 might be the rate-limiting step of this reaction pathway. Thus, a reasonable explanation for our results is that the I213T mutation specifically inhibits the ␥-site cleavage reaction A␤43 3 A␤40 and disrupts the kinetic balance of pre-␥-site cleavage reactions (A␤49 3 A␤46 3 A␤43), leading to higher concentrations of longer A␤ intermediates. Our finding that the A␤43-generating reaction curve of the I213T mutant enzyme clearly reached a plateau, as did the A␤40-generating reaction, supports this idea well.
The A␤42-producing pathway remains controversial. In previous work, Qi-Takahara et al. (21) also pointed out that DAPT treatment did not lead to the accumulation of A␤45 and A␤48 species, even though it did have a clear inhibitory effect on secreted levels of A␤42. In the case of PS2/I141I FAD mutant cells, a small accumulation of A␤45 was observed after DAPT treatment (22). Based on our observations, cleavage reactions of A␤42 and A␤45 at C99-FLAG concentrations of Ͼ1.5 M should fit well to stepwise cleavage reactions similar to A␤48 3 A␤45 3 A␤42. Nevertheless, under lower concentrations of C99-FLAG, we noticed that the I213T mutation constantly facilitated the generation of A␤45 without affecting A␤42 levels, suggesting that an additional cleavage pathway that generates A␤42 may exist. Zhao et al. (33) reported recently that A␤46 may be converted not only to A␤43 but also to A␤42, although not through a major pathway. If this were the case, then one reasonable explanation for our observations is that in cells expressing I213T mutant ␥-secretase, the A␤45 3 A␤42 reaction, which is inhibited, may be partially compensated by the A␤46 3 A␤42 reaction. In this case, a portion of the A␤42 pool may be a by-product of a miscleavage reaction of A␤46 3 A␤43. Further detailed enzymological assessments will be required to clarify this important issue.
Increasing the A␤42/A␤40 ratio would be sufficient to cause parenchymal A␤ accumulation and formation of toxic A␤ oligomers. In our previous work, we observed that certain A␤42/ A␤40 ratios accelerate A␤ aggregation and cell toxicity (35). Recently, Kim et al. (28) also reported that luminal/extracellular overproduction of A␤40 clearly attenuates amyloid pathology in Tg2576 mouse brain (36). These observations indicate that A␤40 has protective effects against the aggregation of A␤42. As a previous study demonstrated, formic acid-extracted fractions derived from I213T mutant knock-in mouse brains actually contain enhanced levels of A␤42. Here, we demonstrated that, even though the I213T mutation impaired ␥-site cleavage, A␤42/A␤40 ratios in the brains of both hemizygous and homozygous knock-in mice remained elevated compared with that in control mice. Thus, in mutant mouse brains, imbalanced A␤ metabolism would ultimately shift the equilibrium of A␤ aggregation potential to a situation that facilitates the accumulation of A␤42. If this were indeed the case, one would expect the levels of non-aggregated soluble A␤ in the brains of FAD patients to be partially reduced. Recent studies indicate that A␤ may have physiological functions, such as regulation of synaptic plasticity, neuronal survival, or intracellular lipid metabolism (37)(38)(39)(40). Thus, the partial lack of functional A␤ species within brain terminals and/or synaptic terminals may also contribute to part of the neurodegenerative process in FAD patients.
Little evidence has been accumulated on longer A␤ species in AD patients and in other A␤-related abnormal events, and thus it is still unclear whether longer A␤ species contribute to AD pathophysiology. The majority of longer A␤ species are less efficiently secreted and are detected specifically in low density lipid fractions, as they are partially anchored to the membrane (41). Thus, if these membrane-anchored A␤s start to concen-trate and form abnormal A␤ clusters on low density lipid domains, it is conceivable that these clusters can act as scaffolds to further the formation of toxic A␤ oligomers and/or fibrillar aggregates, which can disrupt membrane fluidity and/or receptor-mediated signaling functions. In this case, the I213T mutation could specifically enhance the steady state levels of longer A␤ species and accelerate abnormal membrane-related A␤ metabolism. Our preliminary results indicate that, like the I213T mutation, some FAD mutations also facilitate the generation of longer A␤ species. 3 Different compositions of different A␤ species may explain the morphological variations in amyloid plaques reported by some immunohistochemical studies of FAD patient brains (42). To understand the overall A␤-related cascade leading to the onset of AD, continuous effort will be necessary in the near future to elucidate the detailed molecular mechanisms underlying abnormal, FAD-related A␤ metabolism and ␥-secretase function.