Amyloid precursor protein mutation E682K at the alternative β-secretase cleavage β′-site increases Aβ generation†

BACE1 cleaves the amyloid precursor protein (APP) at the β-cleavage site (Met671–Asp672) to initiate the generation of amyloid peptide Aβ. BACE1 is also known to cleave APP at a much less well-characterized β′-cleavage site (Tyr681–Glu682). We describe here the identification of a novel APP mutation E682K located at this β′-site in an early onset Alzheimer's disease (AD) case. Functional analysis revealed that this E682K mutation blocked the β′-site and shifted cleavage of APP to the β-site, causing increased Aβ production. This work demonstrates the functional importance of APP processing at the β′-site and shows how disruption of the balance between β- and β′-site cleavage may enhance the amyloidogenic processing and consequentially risk for AD. Increasing exon- and exome-based sequencing efforts will identify many more putative pathogenic mutations without conclusive segregation-based evidence in a single family. Our study shows how functional analysis of such mutations allows to determine the potential pathogenic nature of these mutations. We propose to classify the E682K mutation as probable pathogenic awaiting further independent confirmation of its association with AD in other patients.


INTRODUCTION
Mutations in the amyloid precursor protein (APP) gene cause early onset autosomal-dominant Alzheimer's disease (AD) (Brouwers et al, 2008;De Strooper & Annaert, 2010;Selkoe, 2001;Wilquet & De Strooper, 2004). Pathogenic mutations (http://www.molgen.ua.ac.be/ADMutations/) located close to major APP cleavage sites (b-and g/e-sites) can increase total Ab production, but most affect the ratio of different Ab peptides without increasing total Ab, favouring a relative increase in Ab 42 versus other species (Bentahir et al, 2006;Citron et al, 1992;De Jonghe et al, 2001;Di Fede et al, 2009;Kwok et al, 2000;Scheuner et al, 1996;Suzuki et al, 1994). Ab 42 and Ab 40 are the main components of the senile plaques in AD brain parenchyma and the amyloid angiopathy in the cerebral blood vessels. Some other mutations have been identified in the Ab sequence itself. Those affect either the generation of Ab by the secretases, the aggregation properties of the Ab peptide or its proteolytic degradation (Betts et al, 2008;Nilsberth et al, 2001;Ono et al, in press;Tomiyama et al, 2008;Tsubuki et al, 2003). BACE1, a membrane-bound aspartic protease, is the bsecretase, which cleaves APP at the b-site Met 671 -Asp 672 of APP (Asp 1 of the Ab sequence; Hussain et al, 1999;Lin et al, 2000;Sinha et al, 1999;Vassar et al, 1999;Yan et al, 1999). This cleavage generates the APP carboxyterminal fragment C99, which is a substrate for the g-secretase complex, an intramembrane cleaving protease (De Strooper, 2003). g-Secretase processing of C99 yields a mixture of Ab peptides including Ab 1-38 , Ab 1-40 , and Ab 1-42 as the most abundantly detected species in cell culture and biological fluids (Takami et al, 2009). BACE1 cleaves APP in addition at a b 0 -site, a secondary cleavage site between Tyr 681 and Glu 682 (Glu 11 of Ab) to generate C89, which is further processed by g-secretase to produce truncated Ab 11-40/42 species. BACE1 cleavage at the Research Article APP E682K increases Ab Figure 1. Clinical and genetic studies of a single case of an AD patient carrying the APP E682K mutation. A. Amyloid imaging using [ 11 C] Pittsburgh compound-B (PIB). The PIB SUVR (standardized uptake value ratio) image of index patient E682K compared to a mean image of 13 AD patients with increased cortical PIB uptake and 13 elderly controls with normal aspecific uptake, all set to the same scale. B. Plot of mean SUVR values in the composite cortical and striatal VOI. C. Magnetic resonance imaging (MRI) of the index patient showing hippocampal atrophy. D. DNA sequencing revealed the heterozygous APP E682K mutation. E. Schematic representation of APP E682K mutation and secretase cleavage sites, numbers were indicated according to Ab sequence. b 0 -site was originally discovered in cell cultures overexpressing this protease (Vassar et al, 1999). An in vitro study showed that purified BACE1 cleaves synthetic peptides mimicking the sequence around the human APP b 0 -site in an enzymatic assay, but the enzymatic efficiency was lower than towards peptides containing the 'canonic' b-site sequence (Yang et al, 2004). Additional evidence suggested that the relative abundance of BACE1 cleavage at these two adjacent sites is governed by the expression levels of the protease: when BACE1 levels are low, bsite cleavage products are the major species, when BACE1 levels are high, b 0 -site cleavage products become predominant (Creemers et al, 2001;Qahwash et al, 2004). These findings were also taken to suggest that b 0 -site processing was only a minor event in APP processing (Creemers et al, 2001). The abundance of the b 0 -site cleavage for human APP processing as well as the functional significance of this alternative b-secretase cleavage site remains, therefore, elusive.
In this work, we identify a novel and unusual APP mutation in a Belgian patient showing early onset AD and seen in the University Hospital in Leuven. This mutation -E682K-is located at the b 0 -site within the Ab sequence ( Fig 1F). We examined the effect of the E682K mutation on the proteolytic processing of APP and found that this mutation caused significant increases in total Ab and in Ab 1-42/40 levels. We further analysed APP processing in neuronal cultures by short metabolic labelling experiments demonstrating that b 0 -site cleavage is a major processing event of wild-type (WT) human APP in neuronal cultures. The E682K mutation blocked this processing step and consequentially shifted BACE1 cleavage towards the b-site. The data demonstrate the functional significance of b 0 -site cleavage in preventing overproduction of Ab, which may potentially cause AD.

RESULTS
Clinical description of the index patient carrying the APP E682K mutation The index patient presented at the Memory Clinic, University Hospitals Leuven, at the age of 50 years with a prior diagnosis of early onset clinically probable AD. Around the age of 47, she had developed symptoms of depression followed by gradually progressive cognitive decline with significant impact on her instrumental activities of daily living. At presentation, the index patient was on citalopram 10 mg twice per day and donepezil 10 mg once per day. Neuropsychological evaluation of the patient revealed a significant episodic memory deficit (Auditory Verbal Learning Test total learning 22/75, delayed recall 0/15; Logical Memory Wechsler Memory Scale: 4/23 literal elements, 2/5 meaningful elements), working memory deficit (Benton Visual Retention test 8/15) and executive dysfunction on the Stroop and Trail Making Test, together with word finding problems (Boston Naming test score 38/60). Magnetic resonance imaging T2-weighted sequence revealed bilateral hippocampal volume loss ( Fig 1C). The cortical pattern of 11C-PIB uptake (Fig 1A and B) was closely similar to what is typically seen in clinically probable AD (Klunk et al, 2004;Nelissen et al, 2007). In contrast to what has been previously reported in presenilin 1 mutation carriers (Klunk et al, 2007), striatal uptake was within the normal range, similarly to what we observed for the K724N APP mutation (Theuns et al, 2006). Analysis of cerebrospinal fluid revealed decreased Ab 42 and increased total and 181T-phosphorylated tau (Table 1), a pattern corresponding to AD. Sequencing of the APP gene revealed a G to A substitution at g.278228G, which is predicted to result in an amino acid substitution at codon 682 (E682K). Such mutation was absent in 940 control samples. Her Apolipoprotein E genotype was e3/e3.
The patient is from a small family. Her father received a diagnosis of dementia at the age of 75 and died at the age of 83 years. His genetic status is unknown as no sample was available for genomic analysis. The patient's brother, however, is clinically, at the age of 53, an asymptomatic carrier. We considered the possibility that the APPE682K mutation could have a variable penetrance in this family. A strong variation in age of onset (range 45-88 years) has previously been observed with AD causing mutations in PSEN2 as well (Sherrington et al, 1996). We thus turned to functional assays to evaluate whether the identified mutation could be a genetic risk factor for AD or not.

E682K mutation increased Ab generation
We introduced the E682K mutation by site directed mutagenesis into human WT APP695 and expressed WT or mutant APP in mouse primary neuronal cultures and in transiently transfected CHO cells. The expression levels of WT and mutant full length APP were checked by Western blotting, and neuronal or CHO cell cultures expressing similar amounts of APP were subjected to further analysis. Conditioned media were analysed using Ab enzyme-linked immunosorbent assay (ELISAs; Table 2). In both

Research Article
Lujia Zhou et al.  neuronal and CHO cell cultures, the E682K mutation caused a two-to three-fold increase in Ab 1-40 and Ab 1-42 levels, but also a slightly higher Ab 1-42 /Ab 1-40 ratio ( Table 2). The alterations are quantitatively and qualitatively comparable to those caused by the previously characterized disease-causing 'Flemish' A692G mutation, which also caused increases in the Ab levels and a slightly higher Ab 1-42 /Ab 1-40 ratio ( Table 2; De Strooper et al, 1995;Haass et al, 1994;Hendriks et al, 1992;Nilsberth et al, 2001;Suzuki et al, 1994;Tian et al, 2010).
E682K mutation enhanced b-site cleavage of APP We next analysed the effects of the E682K mutation on APP processing in further detail. In primary neuronal cultures, this mutation increased C99 and sAPPb levels two-to three-fold (Fig 2), which correlates well with the overall increases in Ab levels as measured by ELISAs. Similar effects were observed in transiently transfected CHO cells, in which the E682K mutation caused a two-to three-fold increase in C99 and sAPPb levels (Fig S1 of Supporting Information). These data show that the E682K mutation increased Ab generation by favouring the b-site cleavage of APP. In contrast to the E682K mutation, the 'Flemish' A692G mutation did not significantly affect the b-secretase processing (as measured by C99 and sAPPb generation), confirming that the increased Ab generation with this mutant is caused by a different mechanism. It has indeed been shown that the 'Flemish' mutation affects an inhibitory domain in the APP sequence that modulates g-secretase activity (Tian et al, 2010).
E682K mutation blocked the b(-site cleavage, which is a major processing event of human APP in neuronal cultures We further examined how the E682K mutation might enhance b-site cleavage. This mutation is located at the previously identified b 0 -site in APP and the WT residue Glu 11 is known to occupy the P1' subsite of BACE1 involved in b 0 -site cleavage. A previous kinetic study has shown that the P1' subsite of BACE1 favours several residues including Asp, Glu, Met, Ala, Ser, and Gln, but that positively charged residues like Lys or Arg decrease affinity (Turner et al, 2001). The E682K mutation was thus predicted to block the b 0 -site cleavage. However, until now, b 0 -cleavage of human APP is considered as a minor processing event in cells expressing endogenous levels of BACE1, and has been mainly documented in cells overexpressing BACE1 (Creemers et al, 2001;Huse et al, 2002;Liu et al, 2002;Qahwash et al, 2004).
Here, we reanalyse the metabolism of human APP in neuronal cultures. We expressed WT or mutant human APP in neuronal cultures using the SFV system and then metabolically labelled the cells for 4 h. The carboxyterminal fragments (CTFs) of APP were immunoprecipitated from cell lysates using an APP carboxyterminal-specific antibody and separated on 16% Tricine SDS-polyacrylamide gel electrophoresis (SDS-PAGE) gels for further analysis by a phosphorimager. We identified five species of CTFs (Fig 3A), as previously described (Hoey et al, 2009;Kimberly et al, 2005). We conclude that the two highest molecular weight bands are phosphorylated C99 and nonphosphorylated C99; the next faint band is phosphorylated C89; the fourth band is a mixture of non-phosphorylated C89 and phosphorylated C83; and the lowest molecular weight band is non-phosphorylated C83. In another group of experiments, we treated the immunoprecipitates with lambda protein phosphatase (LPP; Fig 3B) to remove the confounding phosphorylation, and this resulted, as predicted, in three species of CTFs, i.e. non-phosphorylated C99, C89, and C83. Quantification of phosphor images showed that C99, C89, and C83 accounted for 51.7, 24.9, and 23.4%, respectively, of the total of the three species ( Fig 3C). These data indicate that b 0 -site cleavage is a major processing event that occurs with human APP in neuronal cultures. Interestingly, the E682K mutation ( Fig 3A), when expressed in neuronal cultures, generated more C99 (both phosphorylated and non-phosphorylated), no detectable phosphorylated C89, and apparently less of the mixture of non-phosphorylated C89 and phosphorylated C83 compared to WT APP ( Fig 3A). The level of the latter band was Primary cultured neurons were transduced with SFV expressing WT or mutant APP. Cell lysates were analysed by Western blotting using APP C-terminal antibody B63. Conditioned medium was analysed by Western blotting to determine secreted total sAPP (using 22C11 antibody) and sAPPb (using anti-sAPPb antibody). B, C. Semi-quantification of sAPPb and C99 levels from Western blotting, data were normalized to APP levels ( ÃÃ p < 0.01; ns, statistically not significant; mean AE SEM; n ¼ 4).

APP E682K increases Ab
equal to that of the band isolated from neuronal cultures treated with a highly selective BACE1 inhibitor compound 3 (Fig 3A), which is in accordance with our hypothesis that the E682K mutation blocked C89 generation. After treatment with LPP, this hypothesis was confirmed as only C99 and C83 were detected ( Fig 3B). We also immunoprecipitated Ab related peptides from the conditioned medium using Ab 40/42 carboxyterminal specific monoclonal antibodies (mAbs). WT APP generated two major species (Fig 3D), and both peptides were eliminated after treatment with BACE1 inhibitor (Fig 3D), but remained after treatment with a-secretase inhibitor TAPI-1 (Fig 3D). We thus concluded that these two peptides were products from BACE1 cleavage, i.e. they are Ab 1-40/42 and Ab 11-40/42 species. In agreement with the results obtained from analyzing the CTF levels, the E682K mutation blocked the generation of the Ab 11-40/42 peptide while it enhanced generation of Ab 1-40/42 peptide ( Fig 3D).
E682K mutation had little effects on a-secretase cleavage The E682K mutation (Glu 11 to Lys) is located close to the asecretase cleavage site Lys 17 . Therefore, we decided to investigate whether this mutation affected a-secretase cleavage as well. The antibody 6E10, that we use to detect sAPPa, did not bind the E682K mutant APP in Western blotting (data not shown), however, the typical a-secretase generated C83 carboxyterminal fragment remained unchanged (Fig 3C), while total sAPP levels showed a slight increase probably due to the increase in sAPPb (Fig 2A), suggesting that this mutation had little effects on the asite cleavage of APP. In contrast, the 'Flemish' A692G mutation appeared to slightly affect the generation of CTFa (Fig 2A), as published before (De Strooper et al, 1995;Haass et al, 1994).

Research Article
Lujia Zhou et al. A. Cell lysates were subjected to IP with antibody B63 specific to the C-terminus of APP. Eluates were separated on 16% Tricine gels and analysed by phosphorimager. Five species of CTFs were detected from neurons expressing WT APP (lane 1), including phosphorylated C99, C99, phosphorylated C89, C89/ phosphorylated C83, and C83 (from high to low MW). Neurons expressing E682K APP (lane 2) eliminated the generation of phosphorylated C89 and C89. C89 can be measured by subtracting the density of the phosphorylated C83 band (lane 4) from the C89/phosphorylated C83 band (lane 2). In lanes 3 and 4, neurons were treated with BACE1 inhibitor, therefore, only phosphorylated and non-phosphorylated C83 is observed. B. Immunoprecipitates from cell lysates were treated with LPP. Neurons expressing WT APP (lane 2) generated C99, C89, and C83, while neurons expressing E682K APP (lane 3) largely eliminated the generation of C89. C. Quantification of radioactive bands of CTFs. Statistics were analysed by comparing each CTF species generated from E682K APP with that generated from WT APP ( ÃÃ p < 0.01; ÃÃÃ p < 0.0001; ns, statistically not significant; mean AE SEM; n ¼ 4). D. Conditioned media were subjected to IP with JRFcAb40/42 antibodies specific to the C-terminus of Ab 40 or Ab 42 , eluates were resolved on 10% Bis-Tris gels and analysed by phosphorimager. Neurons expressing WT APP generated Ab 1-40/42 and Ab 11-40/42 species (lane 1), and production of both peptides was eliminated after treatment with BACE1 inhibitor (lane 3), but was unaffected by the treatment with a-secretase inhibitor TAPI-1 (lane 5). E682K mutation enhanced the generation of Ab 1-40/42 but eliminated the generation of Ab 11-40/42 (lanes 2 and 6).
www.embomolmed.org EMBO Mol Med 3, 291-302 ß 2011 EMBO Molecular Medicine E682K mutation modulated g-secretase activity One interesting observation is that the E682K mutation increased the Ab 1-42 /Ab 1-40 ratio in neuronal and CHO cell cultures. It is known that pathogenic mutations near the Cterminus of the Ab sequence in APP can lead to increased generation of Ab 42 by modulating the active sites of g-secretase.
A recent report has shown that the 'Flemish' A692G mutation, which is located in the middle of Ab, has unexpectedly, in addition to its slight inhibitory effect on a-secretase processing (De Strooper et al, 1995;Haass et al, 1994), a quite pronounced effect on g-secretase activity (Tian et al, 2010), suggesting that the interaction between g-secretase and C99 may also occur at sites that are remote from the actual cleavage site in C99. Therefore, we wondered whether the increase in the Ab 42 /Ab 40 ratio caused by the E682K mutation might be caused by an additional effect on the modulation of g-secretase. We transiently transfected CHO cells with WT or E682K mutant C99. This APP fragment is the direct substrate of g-secretase and its processing will, therefore, largely be determined by g-secretase. The expression levels of both C99WT and C99E682K were investigated by Western blotting, no difference was observed. Conditioned media were analysed by Ab ELISAs (Fig 4). The E682K mutation generated the same amounts of Ab 1-42 but significantly less Ab 1-40 , leading to an increased Ab 1-42 /Ab 1-40 ratio. These data suggested that the E682K mutation affects also to a certain extent the Ab 1-42 /Ab 1-40 ratio via modulation of g-secretase.
Limited effects of E682K mutation on the aggregation kinetics and cytotoxicity of Ab peptide We studied the in vitro aggregation properties of synthetic Ab peptide carrying the E682K mutation (E11K mutant Ab). The aggregation kinetics of WT and mutant Ab 42 peptide were monitored by a Thioflavin T (ThT) fluorescence assay. Compared with WT Ab 42 peptide, E11K mutant Ab 42 showed a slight increase in the initial aggregation rate during the first 3 h of the aggregation process ( Fig 5A); however, the overall change in aggregation kinetics was small. Transmission electron microscopy (TEM) images further underlined that the aggregation process of E11K mutant Ab 42 was not significantly different from that of the WT Ab 42 peptide (Fig 5B). In parallel, we also analysed the cytotoxicity of E11K mutant and WT Ab 42 , no significant difference was detected (Fig S2 of Supporting Information).

DISCUSSION
This work describes the identification and characterization of a novel APP mutation E682K from a single case of early onset AD. The index patient is a Belgian female; she was diagnosed as probable AD at the age of 49 years. Neuropsychological evaluation, CSF biomarkers (Ab 42 , total-tau and phosphotau), amyloid imaging (using 11 C-PIB PET) as well as magnetic resonance imaging of brain all confirmed the diagnosis of AD. The patient's father developed late onset AD, while the brother is an asymptomatic mutation carrier currently at the age of 53, and there is no extended familial information. This raised the interesting question whether the identified mutation is a benign polymorphism or should be considered a genuine genetic risk factor for disease. The issue of rare genetic variants contributing moderate risk to disease is hotly debated in genetic research, as it is very difficult to detect such mutations while they are likely to contribute significantly to the total genetic risk for disease (Singleton et al, 2010). We hypothesized that the APP E682K mutation could have variable penetrance, which has previously also been proposed for disease causing mutations in PSEN2, carriers of which display a strong variation in age of onset (Sherrington et al, 1996) and decided to use functional assays to test whether the identified mutation could be considered as pathogenic or not.

Research Article
APP E682K increases Ab Figure 4. Effects of E682K mutant C99 on g-secretase activity in cell-based assay. A-C. CHO cells were transfected with WT or E682K mutant C99, conditioned media were analysed by Ab ELISAs ( ÃÃ p < 0.01; Ã p < 0.05; ns, statistically not significant; mean AE SEM; n ¼ 3). D. Western blotting analysis of C99 expression levels. Cell lysates containing either C99WT or C99E682K were loaded in a series of two times dilution.
Our analysis showed that this novel mutation increased full length Ab release and also the Ab 1-42 /Ab 1-40 ratio to similar extents as found with a previously well characterized disease associated APP A692G or 'Flemish' mutation. While the observed effects on Ab processing suggest the possibility that this novel mutation is pathogenic, the genetic support is inconclusive. According to the criteria discussed by Guerreiro and coworkers (Guerreiro et al, 2010), the APP E682K mutation was found in a single case of early onset AD. This mutation was absent from 940 healthy controls and affects Ab generation. Although strictly spoken a second case with this mutation needs to be identified as independent confirmation (Guerreiro et al, 2010), we propose that the accumulated evidence is clearly in favour to classify this mutation as probable pathogenic. Further follow up of the brother of the index case could yield final confirmation, but because of the study design, we are not allowed to contact the brother for further investigation. Therefore, we will have to wait for additional cases from other investigations in the future to classify this mutation as definitively pathogenic. Nevertheless, the functional analysis of this mutation provided an important and interesting novel insight in the biological significance of the b 0 -site processing of APP, which appeared quantitatively much more important for normal APP metabolism than previously thought.
Our data indicate indeed that processing at this site should be considered anti-amyloidogenic, counteracting partially the amyloidogenic b-secretase cleavage of APP. In agreement with this assumption, rodents, which tend to cleave their endogenous APP at the b 0 -site (De Strooper et al, 1995), do not spontaneously develop amyloid plaques. The functional significance of b 0 -site

Research Article
Lujia Zhou et al. The aggregation kinetics is not dramatically affected by the E11K mutation, apart from a small increase in the initial rate during the first 3 h of aggregation process. B. TEM images further underline that the aggregation process of Ab 42 upon introduction of the E11K mutation is not significantly modified. After 1.5 h aggregates are observed which become larger upon further incubation.
www.embomolmed.org EMBO Mol Med 3, 291-302 ß 2011 EMBO Molecular Medicine cleavage of APP by BACE1 has indeed been debated since a long time. On the one hand, it was previously suggested that this cleavage is a minor part of physiological APP processing (Creemers et al, 2001). Our new study now clarifies that this is not the case and that about one-fourth of the major APP CTF species is generated by this pathway under our experimental conditions. Second, it was reported that N-terminal truncation of Ab may enhance the aggregation properties as well as the cytotoxic effects of Ab (Demeester et al, 2001) and that in sporadic AD cases, significant amounts of Ab 11-40/42 species were found in the plaques in the brain (Huse et al, 2002;Liu et al, 2006). Therefore, b 0 -site cleavage was proposed to contribute to AD pathogenesis (Huse et al, 2002;Liu et al, 2006). However, our data argues for a protective role of b 0 -site cleavage, as blocking this cleavage disrupted the balance of APP processing by BACE1 leading to increased full length Ab. The fact that the b 0 -site mutation E682K is associated with early onset AD speaks clearly in favour of this argument. Of note, our data suggest additional effects of the E682K mutation on g-secretase activity as well. APP carrying the E682K mutation increased significantly the Ab 1-42 /Ab 1-40 ratio in cellbased assays. This effect occurs at the g-secretase level because expressing a C99 construct that bypasses the ectodomain shedding step needed for g-secretase (Lichtenthaler et al, 1999) led to a similar increase in this ratio. It has been recently reported that a substrate inhibitory domain (ASID) located at the middle of Ab (17-23) is modulating g-secretase activity by binding to an allosteric site within the g-secretase complex and that the 'Flemish' mutant C99 increased g-secretase activity via disruption of this ASID (Tian et al, 2010). The E682K mutation is located at the Ab 11 site, which is not part of the previously delineated ASID, but had nevertheless a significant effect on Ab 1-40 levels and little effects on Ab 1-42 levels. We suggest the possibility that the interaction domain between g-secretase and C99 is more extended than ASID and mutations such as the E682K mutation, as well as other similar substitutions, may be considered as new tools to investigate the possible allosteric modulation of g-secretase by APP substrate.
The current data indicate that BACE1 cleaves APP at two distinct sites and that both are physiologically significant events in APP processing. So far, BACE1 is the most attractive drug target for AD treatments, and the evaluation of inhibitors under development for treatment of the disease focuses on the effect of blocking the bsite cleavage. This could be problematic because BACE1 inhibitors will also affect b 0 -site cleavage. Kinetic data showed that the BACE1 has a much higher enzymatic efficiency towards peptide containing the b-site than peptide containing the b 0 -site sequence (Yang et al, 2004), which raises a theoretical possibility that BACE1 inhibitors may have differential inhibition effects on the two distinct sites depending on the concentrations used. This could cause relative shifts from b 0to b-site cleavage, as observed with the current mutation. We thus propose that inhibitor drugs should be carefully monitored for their dose-effect on both BACE1 cleavage sites and to measure, therefore, the effects on Ab 11-40/42 of any such drug when used in patients.
Finally it is likely that in the near future, increasing exon-and also exome-based sequencing efforts will increase the number of putative pathogenic mutations identified without further conclusive segregation-based evidence in a single family. Functional analysis of such mutations as demonstrated here will allow to determine the putative pathogenic nature of some of these mutations as more extensively discussed in (Guerreiro et al, 2010). Such functional study may also be crucial to fill part of the existing gap in genetic research with regard to the postulated low frequency, moderate risk genetic loci that are now very difficult to identify and were called the 'dark matter of disease risk' (Manolio et al, 2009;Singleton et al, 2010).

Genetic analysis
Genomic DNA was extracted from whole blood by automated DNA extraction. Mutation analysis was performed by direct sequencing of exon 16 and 17 of APP. Twenty nanogram of genomic DNA was amplified in a polymerase chain reaction (PCR), and PCR products were purified using ExoSAP-IT 1 (USB Corporation, Cleveland, OH, USA) according to the manufacturer's protocol. Purified PCR products were sequenced in both directions (primer sequences available upon request), by fluorescent cycle sequencing using the 'Big Dye 1 Terminator v3.1 Cycle Sequencing' kit (Applied Biosystems, Foster City, CA, USA) according to manufacturer's recommendations. Labelled products were separated on an Applied Biosystems 3730 DNA analyser (Applied Biosystems). Sequence traces were analysed using NovoSNP (Weckx et al, 2005). The presence of g.278228G>A in exon 16 of APP in the index patient, and its absence from 940 control chromosomes was assessed using a pyrosequencing assay (primer sequences available upon request) on a PSQ TM HS96A System (Biotage, Uppsala, Sweden) according to the manufacturer's protocol. Healthy control individuals were unrelated, age and gender matched and from the same geographical region.

Antibodies and compounds
Rabbit polyclonal antibody B63 raised against the C-terminus of APP has been described previously . mAbs JRFcAb40/ 28, JRFcAb42/26 recognizing the C-terminus of Ab species terminating at 40 or 42, respectively, and JRFAbN/25 recognizing the Nterminal 1-7 amino acids of human Ab were provided by Janssens Pharmaceutica. The following antibodies were purchased: mAb 22C11 recognizing the N-terminus of APP (Chemicon/Biognost); sAPPb polyclonal antibody recognizing the N-terminus to the b-secretase cleavage site of APP (Covance). Compound TAPI-1 was purchased from Calbiochem; BACE1 inhibitor compound 3 was kindly provided by Merck Research Laboratories.

Amyloid imaging
Using a Siemens Biograph PET-CT scanner, 11 C-Pittsburgh Compound B images were acquired dynamically from 40 to 70 min following injection of 335.1 MBq. After realignment, PET frames were divided by the mean uptake in cerebellar cortex to obtain standard uptake ratio (SUVR) images. The index patient's SUVR image was nonlinearly warped to a PIB template in MNI space consisting of 13 AD patients previously rated as having increased uptake (Nelissen et al, 2007). Volume-of-interest (VOI) masks were derived from the Automatic Anatomical Labelling (AAL) atlas, whereby, a composite cortical VOI was defined as the union of parietal, lateral temporal and frontal areas excluding primary sensory motor cortex.

Cell culture and transfection
CHO cells were maintained at 378C in a humidified, 5% CO 2 controlled atmosphere in DMEM (Gibco) supplemented with 10% foetal bovine serum (FBS). Transfections were performed using Fugene (Invitrogen) according to the manufacturer's instructions. When 80% confluence was reached, cells were transiently transfected with APP or C99 cDNA constructs. After 24 h, media were replaced with DMEM supplemented with 5% FBS and conditioned for 16-24 h and then collected for Ab ELISA assays in the presence of complete protease inhibitor cocktail (Roche). Conditioned media were also analysed by Western blot for the detection of sAPP, while cell lysates were analysed by Western blot for the detection of APP CTFs.

Neuronal cultures and Semiliki Forest virus transfection
Primary cortical neurons were derived from E14 embryos from C57BL/6 mice, the preparation procedures have been described previously (Annaert et al, 1999). Briefly, the brains of E14 embryos were dissected in HBSS medium (Gibco), trypsinized and plated on 6-cm dishes (Nunc) precoated with poly-L-lysine (Sigma-Aldrich). Cultures were maintained in neurobasal medium (Gibco) supplemented with B27 (GibcoBRL). A 5 mM cytosine arabinoside was added to prevent glial cell prolification. Semliki Forest viruses (SFV) were produced as described (Annaert et al, 1999). Three days cultured neurons were transduced with SFV expressing WT or mutant human APP695. After 1 h, media were replaced with normal culture media and left for 2 h. Media were refreshed and conditioned for $6 h. Conditioned media were analysed by Ab ELISA or Western blot for the detection of sAPP, while cell lysates were analysed by Western blotting for the detection of APP CTFs.

Metabolic labelling and Immunoprecipitation
Neuronal cultures were transduced with SFV for 1 h and left in normal culture medium for 2 h as described above. Media were then replaced with methionine-free MEM (Gibco BRL) containing 100 mCi/ml [ 35 S] methionine (ICN Biomedicals). After 4 h metabolic labelling, supernatants were collected, cells were washed and lysed in DIP buffer (20 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS) supplemented with complete protease inhibitors. Supernatants were immunoprecipitated using a combination of JRFcAb40 and JRFcAb42 antibodies and incubated with protein G-Sepharose at 48C overnight. The immunoprecipitants were resolved on 10% NuPAGE gels (Novex; Invitrogen) running with MES buffer. Cell extracts were immunoprecipitated using APP C-terminal antibody B63. The immunoprecipitants were resolved on 16% Tricine gels (Novex; Invitrogen) running with Tricine-SDS buffer (Novex; Invitrogen). The intensity of the radioactive bands was quantified using a phosphorimager (Melecular Dynamics, Suunyvale, CA) and Aida Image Analyzer v.4.22 software.

Ab ELISA
The concentration of Ab in conditioned medium was measured by end-specific ELISAs. JRFcAb40/28 or JRFcAb42/26 was used as capturing antibodies for Ab species terminating at 40 or 42, respectively. HRP-conjugated JRFAbN/25 recognizing the N-terminal 1-7 amino acids of human Ab was used as detecting antibody for full length Ab. The affinity of JRFAbN/25 for E11K mutant Ab peptide was not different from its affinity for WT Ab peptide (Fig S3 of Supporting Information). Plates were coated with 60 ml capturing antibodies at 2 mg/ml at 48C overnight in coating buffer (10 mM Tris-HCl, 10 mM NaCl, 10 mM NaN 3 , pH 8.5). The next day, plates were rinsed once in phosphate buffer saline (PBS) and blocked with blocking buffer (0.1% casein in PBS, pH7.4) either at 48C overnight or at room temperature for 4 h. Plates were then rinsed twice with PBS. Thirty microlitre of detecting antibody diluted in blocking buffer was first added to each well, followed by adding 30 ml conditioned medium diluted in blocking buffer. Plates were incubated at 48C overnight, and rinsed and developed the next day using 60 ml mixture of 3,3 0 ,5,5 0tetramethylbenzidine (TMB) solution. The reaction was stopped by addition of 60 ml of 2 N H 2 SO 4 . Plates were read immediately at 450nm.

Preparation of Ab peptide
Ab peptide (JPT Peptide Technologies GmbH) was dissolved at a concentration 1 mg/ml in 1,1,1,3,3,3-hexafluor-2-propanol (HFIP 99%, Aldrich Cat.# 10,522-8). HFIP was evaporated using a gentle stream of argon gas and the peptide film was resolved using dimethyl sulphoxide, (DMSO Sigma Cat. #D4540) at a final concentration of 1 mg/ml. The peptide was separated from DMSO with a 5 ml HiTrapTM Desalting column (GE Healthcare, Sweden). Complete removal of DMSO was confirmed by Fourier transform infrared spectroscopy (FTIR): DMSO provides spectral maxima at 1011 and 951 cm À1 . The peptide was eluted into a 50 mM Tris, 1 mM EDTA buffer, pH 7.5 and the peptide concentration was measured using Bradford assay. The samples were kept on ice until experiments started, with a maximum lag time of 20 min.

Thioflavin T fluorescence
Ab protein concentrations were normalized to 25 mM by further dilution using 50 mM Tris, 1 mM EDTA containing buffer and a final concentration of 12 mM ThT was added in a Greiner 96-well plate. The fibrillation kinetics were followed in situ using a Fluostar OPTIMA fluorescence plate reader at an excitation wavelength of 440 nm and an emission wavelength of 480 nm. Readings were recorded in triplicate every 10 min for a period of 10 h.

Transmission electron microscopy
Aliquots (5 ml) of the Ab preparation were adsorbed to carbon-coated FormVar film on 400-mesh copper grids (Plano GmbH, Germany) for

Cell viability assay
Mouse primary hippocampal neurons were grown in Neurobasal medium (Invitrogen) for 1 week, and then were treated with different concentrations of Ab42 (WT or mutant) pre-aggregated for 1.5 h. After 72 h treatment, 10 ml Cell-Titer-Blue dye (Promega) was added to 200 ml of the growth culture medium on the cells. After 3 h, the fluorescence intensity of the samples was measured at an excitation wavelength of 560 nm and an emission wavelength of 590 nm.

Statistical analyses
Statistical significance was determined by the Student's t-test.

Author contributions
LZ analysed the effects of APP mutants in Ab generation and wrote the manuscript. NB and KS performed the genetic analysis. MM suggested experiments and contributed to Ab ELISA measurements. KVL, PVD, DD and FVL analysed patient's brain image and blood samples. AV, KB and IB characterized the aggregation property and cytotoxicity of mutant Ab peptide in vitro. CVB and RV identified the AD case and the APP mutation. BDS designed the study and wrote the manuscript. investigator of the Research Foundation -Flanders (FWO). BDS is the Arthur Bax and Anna Vanluffelen chair for AD.

APP E682K increases Ab
The paper explained PROBLEM: Direct genomic sequencing of APP exons 16 and 17 in a group of Belgian early onset Alzheimer's disease (AD) patients (n ¼ 543) identified a novel mutation in exon 16 (g.278228 G>A, c.2044G>A), predicting a glutamine-to-lysine substitution at codon 682 (p. E682K). This mutation was absent from healthy control group (n ¼ 940). The index patient was diagnosed as probable AD at the age of 49 years. Neuropsychological evaluation, CSF biomarkers (Ab42, total-tau and phospho-tau), amyloid imaging (using 11C-PIB PET) as well as magnetic resonance imaging of brain all confirmed the diagnosis of AD. The patient's father developed late onset AD, while the brother is an asymptomatic mutation carrier currently at the age of 53, and there is no extended familial information. Genetic information is nonconclusive in determination of the pathogenicity of the identified mutation. This work explored the possible pathogenic nature of the novel APP mutation by functional analysis.

RESULTS:
In both neuronal and non-neuronal cell cultures, APP E682K mutation caused a two-to three-fold increase in Ab 1-40 and Ab 1-42 levels. The E682K mutation is located at the alternative b 0 -secretase cleavage site on APP, the quantitative significance of which remained elusive. This work shows that the b 0 -processing of human APP is a major event in neuronal cultures, whereas, the E682K mutation blocked the b 0 -site cleavage and shifted APP processing to the b-site, thus enhanced the generation of full length Ab as well as other b-cleavage products from APP.
Interestingly, the E682K mutation also led to a significant increase in the ratio of Ab 1-42 to Ab 1-40 . Our data suggest that this is due to modulation of g-secretase, as introducing the mutation to C99, a direct substrate for g-secretase, significantly increased the ratio of Ab 1-42 to Ab 1-40 in cell cultures. The alterations on Ab generation caused by E682K mutation are quantitatively similar to those caused by the previously reported disease causing 'Flemish' A692G mutation. Functional analysis strongly supports the possible pathogenic nature of the novel mutation. IMPACT: This work first demonstrates that b 0 -site cleavage of human APP is a major event in neurons and plays a 'protective' role in APP metabolism, while disruption of this cleavage by mutation or other effects will enhance the amyloidogenic processing of APP and may consequentially lead to AD. Our work shows how a combination of clinical and cell biological methods works to determine the possible pathogenicity of novel mutations identified in single families when segregation-based evidence is not available. This work is important from a drug development perspective as well, adding a potentially important layer of complexity to the development of the b-secretase (BACE1) inhibitors in the clinic. We propose that b-secretase inhibitors should be monitored for their effects on both the band the b 0 -site. Inhibitors at certain concentration blocking the b 0 -site and shifting the cleavage of APP to b-site could be problematic.