TRPC6 specifically interacts with APP to inhibit its cleavage by γ-secretase and reduce Aβ production

Generation of β-amyloid (Aβ) peptide in Alzheimer's disease involves cleavage of amyloid precursor protein (APP) by γ-secretase, a protease known to cleave several substrates, including Notch. Finding specific modulators for γ-secretase could be a potential avenue to treat the disease. Here, we report that transient receptor potential canonical (TRPC) 6 specifically interacts with APP leading to inhibition of its cleavage by γ-secretase and reduction in Aβ production. TRPC6 interacts with APP (C99), but not with Notch, and prevents C99 interaction with presenilin 1 (PS1). A fusion peptide derived from TRPC6 also reduces Aβ levels without effect on Notch cleavage. Crossing APP/PS1 mice with TRPC6 transgenic mice leads to a marked reduction in both plaque load and Aβ levels, and improvement in structural and behavioural impairment. Thus, TRPC6 specifically modulates γ-secretase cleavage of APP and preventing APP (C99) interaction with PS1 via TRPC6 could be a novel strategy to reduce Aβ formation.

A lzheimer's disease (AD) is characterized by extracellular senile plaques and intracellular neurofibrillary tangles in autopsied brain tissues. Senile plaques are mainly composed of b-amyloid (Ab) peptide, which is proposed to be responsible for AD pathogenesis 1 . The Ab is generated through a sequential cleavage of amyloid precursor protein (APP) by band g-secretases, while a-secretase cleavage precludes Ab formation and produces neurotrophic sAPPa (ref. 2). To regulate APP cleavage by secretases and reduce Ab production is a potential strategy for AD treatment.
The g-secretase cleavage is the final step in Ab production and attracts much attention in AD studies. However, g-secretase has diverse substrates besides APP, such as Notch, E-/N-cadherin and ErbB-4 and g-secretase cleavage of these proteins is essential for their physiological functions 3,4 . Drugs designed to inhibit g-secretase activity may thus suppress the cleavage of a wide range of substrates concurrently, leading to many side effects. For example, administration of semagacestat, a potent g-secretase inhibitor, resulted in reduced plasma Ab levels, but worsened cognitive performance as well as enhanced skin cancer risk, immune system abnormalities and gastrointestinal symptoms, all of which were attributed to the inhibition of g-secretase cleavage of Notch 5 . Indeed, semagacestat was found to be more potent to inhibit g-secretase cleavage of Notch than that of APP 6 . Thus, specific modulation, instead of complete inhibition, of g-secretase cleavage of APP might be an alternative avenue to reduce Ab levels and treat the disease 7 .
The transient receptor potential canonical (TRPC) is a family of Ca 2 þ -permeable nonselective cation channels, consisting of four subgroups, TRPC1, TRPC2, TRPC3/6/7 and TRPC4/5 (ref. 8). After activation by G-protein-coupled receptors or receptor tyrosine kinases, TRPC channels mediate Ca 2 þ influx and initiate cellular responses 9 . These channels have been reported to play important roles in development 10 and diseases 11,12 . Recently, presenilin 2, a g-secretase component, was reported to influence TRPC6 channel activity 13 , indicating that TRPC6 may be involved in Ab production. Further, AD patients usually have severe synapse and neuron loss, leading to memory decline [14][15][16] , whereas TRPC6 promotes neuronal survival 12,17 , synapse formation 18,19 and enhances spatial learning and memory 19 . We thus investigated whether TRPC6 affects Ab production.
Here, we report that TRPC6 reduces Ab levels both in cultures and in mice. TRPC6 interacts with APP (C99) to prevent the interaction between C99 and presenilin 1 (PS1) and thus suppresses g-secretase cleavage of APP (C99) without affecting Notch cleavage. A fusion peptide derived from TRPC6 also reduces Ab levels without effect on Notch cleavage. Therefore, targeting APP-PS1 interaction via TRPC6 may represent a novel intervention opportunity to reduce Ab levels without side effects induced by inhibiting g-secretase activity.

Results
TRPC6 regulated Ab levels independent of its channel activity. We initially examined whether Ca 2 þ channels play a role in Ab production because Ca 2 þ entry can affect a-secretase cleavage of APP [20][21][22] . We downregulated several Ca 2 þ channel proteins ( Fig. 1a; Supplementary Fig. 10), including Ca v 1.2, Ca v 3.1 or Ca v 3.3, L-or T-type voltage-dependent Ca 2 þ channel proteins; TRPC5 or TRPC6, nonselective cation channel proteins, in primary cultured rat cortical neurons and found that downregulating TRPC6, but not others, greatly enhanced both Ab40 and Ab42 levels determined by enzyme-linked immunosorbent assay (ELISA; Fig. 1a), suggesting that TRPC6 specifically regulates Ab accumulation in cortical neurons.
However, treatment of the neurons with OAG or SKF96365, agents known to activate or block TRPC channels 23,24 , respectively, did not influence the Ab levels (Fig. 1b). These results suggest that TRPC6 regulates Ab accumulation likely independent of its channel activity.
We then tested the effects of TRPC6 on Ab levels in HEK293APP stable cells, a cell model commonly used for studying Ab accumulation. All the Ab40, Ab42, Ab38 and Ab1-x levels in the culture medium of the cells transfected with TRPC6 were markedly reduced ( Fig. 1c; Supplementary Figs 1a and 10). In contrast, expressing TRPC5 did not change the Ab levels ( Fig. 1c; Supplementary Fig. 10). These results show that TRPC6 specifically regulates Ab accumulation in HEK293APP stable cells. Similar to its effect on Ab accumulation in the neurons, SKF96365 did not suppress the inhibitory effect of TRPC6 on Ab levels in HEK293APP cells ( Fig. 1d; Supplementary Fig. 10). Meanwhile, expressing the dominant-negative form of TRPC6 (DN-TRPC6), a mutant known to suppress TRPC6 channel activity 25 , did not affect TRPC6 inhibition of Ab accumulation ( Fig. 1e; Supplementary Fig. 10). Moreover, expressing DN-TRPC6 also reduced Ab levels ( Fig. 1e; Supplementary Fig. 10). All together, these results provide evidences that TRPC6 inhibition of Ab accumulation is independent of its channel activity.
TRPC6 prevented APP (C99) cleavage by c-secretase. We next observed that expressing TRPC6 did not alter the APP levels, sAPPa production and expression of ADAM10 or BACE1 ( Supplementary Fig. 2a,b; Supplementary Fig. 15). Further, the enzymatic activity of aor b-secretase, determined by in vitro assay using their respective fluorogenic substrates was not changed by TRPC6 ( Supplementary Fig. 2c). The Ab can be cleared by enzymatic degradation or pumping into plasma 26,27 . However, the expression of insulin-degrading enzyme, neprilysin and clusterin, proteins known to affect Ab clearance, was not changed by TRPC6 ( Supplementary Fig. 2d,e; Supplementary  Fig. 15). Furthermore, the Ab levels in the medium and cell lysate of HEK293 cells assayed by the Ab clearance test were not altered by TRPC6 (Supplementary Fig. 2f). Together, these results suggest that aor b-secretase cleavage of APP and Ab clearance are not affected by TRPC6.
We then examined whether g-secretase cleavage of APP was affected by TRPC6. Cleavage of C99 by g-secretase releases Ab peptides and APP intracellular domain (AICD) 28 . TRPC6 greatly reduced both Ab40 and Ab42 levels in the culture medium of COS7 cells stably expressing C99 (Fig. 2a). TRPC6 also downregulated AICD generation in HEK293C99 stable cells  Supplementary Fig. 15). Further, expression of PS1, PS2, nicastrin or presenilin enhancer 2, proteins known as the gsecretase components, was not changed by TRPC6 ( Fig. 2d; Supplementary Figs 2l and 11). Consistently, the g-secretase activity determined by in vitro assay using the fluorogenic substrate was unchanged by TRPC6 (Fig. 2e). We also examined whether TRPC6 affected g-secretase cleavage of Notch by detecting Notch intracellular domain level generated from NotchDE, an immediate substrate of g-secretase 29 . As shown in Fig. 2f,g and Supplementary Fig. 11, TRPC6 did not alter the Notch intracellular domain level, suggesting that TRPC6 did not inhibit g-secretase cleavage of Notch. We further examined the CTF1 and CTF2 levels of E-/N-Cadherin, known as the endogenous substrates of g-secretase. As shown in Fig. 2h,i and Supplementary Fig. 11, TRPC6 did not alter the CTF2/CTF1 ratio of E-/N-Cadherin, suggesting that TRPC6 did not inhibit gsecretase cleavage of E-/N-Cadherin. Together, these results suggest that TRPC6 specifically reduces g-secretase cleavage of C99, but does not act as a general g-secretase inhibitor.
Certain proteins can interact with APP (C99) to regulate Ab production 30 . We next studied whether TRPC6 and APP (C99) were subcellularly co-localized. As the anti-TRPC6 antibody on hand cannot be used for immunostaining, we transfected HEK293 cell with TRPC6-HA and APP-Myc (C99-Myc) and stained with antibodies against HA-and Myc-tag. We found that TRPC6 and APP (C99) might co-localize in certain organelles ( Fig. 3a; Supplementary Fig. 3a). These results provide the anatomic evidence to support the possible interaction between TRPC6 and APP. Next, in the fractionation experiments with mouse brain lysates, endogenous TRPC6 and APP (CTF) were also found in the same fractions ( Fig. 3b; Supplementary Fig. 12). We then explored whether TRPC6 and APP were in the same complexes. Immunoprecipitation of mouse brain lysates, HEK293 Ca v 3.1i Ca v 3.3i Ca v 1.2i Ca v 3. Ca v 1.2i Ca v 3.1i Ca v 3.3i   Supplementary Fig. 13), and the precipitated TRPC6 level was increased when its expression was enhanced ( Supplementary Fig. 4a,b; Supplementary Fig. 17). In addition, TRPC6 was found in the complexes precipitated by APP antibody in HEK293TRPC6 cells overexpressing APP-I716F or APP-V717F, two familial AD-linked APP variants around g-secretase cleavage sites (Supplementary Figs 4c-f and 17). In contrast, TRPC5 was not precipitated by APP antibody in HEK293APP cells overexpressing TRPC5 ( Fig. 3d; Supplementary Fig. 13). Moreover, TRPC6, but not TRPC5, and C99 were found in the complexes by reciprocal precipitation experiments in HEK293C99 cells ( Supplementary Figs 4g   and 17). In addition, NotchDE was not found in the complexes precipitated by TRPC6 antibody (Fig. 3e; Supplementary Fig. 14). All together, these results point to a possibility that TRPC6 selectively interacts with APP (C99) to specifically suppress its cleavage by g-secretase.
We next explored whether TRPC6 affected the interaction between C99 and PS1. The C99-Myc and PS1-HA in HEK293C99 cells were found in the complexes precipitated with the antibody against HA-or Myc-tag reciprocally ( Fig. 3f; Supplementary  Fig. 14). However, the reciprocal precipitation of C99 and PS1 was markedly inhibited when TRPC6 was expressed ( Fig. 3f; Supplementary Fig. 14). It has been reported that PS2 can affect TRPC6-mediated Ca 2 þ entry in HEK293 cells 13 . We next explored whether APP could regulate the activity of TRPC6. Application of hyperforin, an agonist of TRPC6 (refs 18,31), induced Ca 2 þ influx in HEK293TRPC6 cells and this Ca 2 þ entry was not affected by APP ( Supplementary Fig. 4h,i). Together, these results suggest that TRPC6 prevents APP (C99) interaction with PS1 and APP does not affect the channel activity of TRPC6.  4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Bottom kDa kDa ARTICLE TM2 domain was essential for TRPC6 to reduce Ab. To provide direct evidence that TRPC6 interaction with APP indeed inhibited APP cleavage by g-secretase and Ab formation, we investigated which domain of TRPC6 mediated its inhibitory effect on Ab production. TRPC6 has six transmembrane domains and cytoplasmic N-and C-terminals ( Supplementary Fig. 5a). As the cytoplasmic tail of C99 was reported to interact with several cytosolic proteins and these interactions may regulate Ab production 30 , we first hypothesized that cytoplasmic terminals of TRPC6 might interact with the cytoplasmic domain of C99 to reduce Ab production. We transfected HEK293APP cells with the N or C terminus (1-405 or 725-930 aa (amino acids)) of TRPC6 and examined Ab levels. These constructs did not affect Ab generation ( Supplementary Fig. 5b). After a series of truncation experiments ( Supplementary Fig. 5c), we found that all the constructs containing 437-508 aa of TRPC6 reduced Ab levels, whereas the constructs without this domain did not ( Supplementary Fig. 5d), suggesting that 437-508 aa of TRPC6 was important for its inhibition of Ab production. This domain contains the first and second transmembrane regions and the first extracellular loop. When the second transmembrane (TM2) region was mutated, by point mutation (C6-mut1 and C6-mut2, Supplementary Fig. 5e), replacement (C6-mut3) or reversal (C6-mut4), TRPC6 was not able to reduce Ab levels ( Fig. 4a), indicating that the TM2 domain was essential for TRPC6 to regulate Ab generation. Immunoprecipitation experiment revealed that C6-mut1 was not in the complexes precipitated by APP antibody (Fig. 4b; Supplementary Fig. 14), suggesting that TRPC6 may interact with APP through TM2 domain to inhibit Ab formation. To rule out the possibility that mutation of TRPC6 might affect its subcellular localization, we transfected HEK293 cell with C6-mut1-HA ( Supplementary Fig. 5e) and APP-Myc, and stained with HA and Myc antibodies. The C6-mut1 and APP might co-localize in certain organelles ( Supplementary Fig. 5f), similar to wild-type (WT) TRPC6 and APP (Fig. 3a).
TAT-TM2 inhibited Ab accumulation in vitro and in vivo. To mimic TRPC6 inhibition of Ab formation, a cell-permeable peptide TAT-TM2 containing the 487-508 aa of TRPC6 was synthesized and TAT-TM2, but not its mutated form (TAT-TM2-mut1, 2, Supplementary Fig. 6a) or TAT-TM1, greatly suppressed Ab production in HEK293APP cells (Fig. 4c). Further, TAT-TM2 dose-dependently reduced Ab levels in HEK293APP cells and in the primary cultured cortical neurons ( Supplementary  Fig. 6b,c). To determine the amino acids important for the inhibitory effect, TAT-TM2-N (N-terminal 10 aa of TM2 fused to TAT) and TAT-TM2-C (C-terminal 11 aa of TM2 fused to TAT) were synthesized. The TAT-TM2-C inhibited Ab generation, whereas TAT-TM2-N did not ( Supplementary Fig. 6d), suggesting that the C-terminal 11 aa of TM2 was critical for reducing Ab formation. Indeed, mutations within this domain blocked the effect of TAT-TM2-C on reducing Ab levels ( Supplementary  Fig. 6e). In the experiment to test whether TAT-TM2 also binds to APP, we found that TAT-TM2-biotin, but not TAT-biotin, interacted with APP similar to the full-length TRPC6 ( Fig. 4d and Supplementary Fig. 14), whereas it did not bind to TRPC6 ( Supplementary Figs 7a and 18). Addition of TAT-TM2 prevented the interaction between APP and TAT-TM2-biotin or TRPC6 (Supplementary Figs 7b-e and 18). Further, assayed by a double-blinded manner, injection of TAT-TM2 into APP/PS1 mice led to a marked reduction in Ab levels in the mouse brain (Fig. 4e). All these results show that TAT-TM2 reduces Ab accumulation in cultures and in mice.
TRPC6 reduced Ab to improve spatial learning of AD mice. To test whether TRPC6 could regulate Ab levels in animals, we crossed TRPC6 transgenic mice, in which TRPC6 was overexpressed in the forebrain excitatory neurons 19 , with APP/PS1 mice, a mouse model commonly used to study Ab formation 32 . Amyloid plaques were examined by Ab immunostaining with 6E10 antibody on the brain sections of APP/PS1 and APP/PS1/TRPC6 mice ( Fig. 5a; Supplementary Fig. 8a). The plaque load was markedly reduced in APP/PS1/TRPC6 male ( Supplementary Fig. 8b) and female (Fig. 5b) mice compared with that in APP/PS1 mice. Plaque staining with Ab specific antibody MOAB-2 (ref. 33) confirmed that the plaque load was greatly reduced in APP/PS1/TRPC6 mice compared with that in APP/PS1 mice ( Supplementary Fig. 8f-i). Total Ab levels in guanidine-HCl (GuHCl) extraction of APP/PS1/TRPC6 male ( Supplementary  Fig. 8c) and female (Fig. 5c) mouse forebrains were greatly reduced than those of APP/PS1 mouse forebrains. The TBST soluble and insoluble Ab levels in APP/PS1/TRPC6 male ( Supplementary Fig. 8d,e) and female (Fig. 5d,e) mice were also substantially reduced compared with those in APP/PS1 mice. To further confirm the effect of TRPC6, we examined whether TRPC6 could reduce amyloid plaque in APP/PS1 mice with another genetic background. The plaque load in the hippocampus of APP/PS1/TRPC6 mice in B6C3 background stained with thioflavin S (Fig. 5f) was also greatly reduced than that in APP/PS1 mice (Fig. 5g). Taken together, TRPC6 reduces Ab accumulation in APP/PS1 mice.
As synapse loss was found to be best correlated with dementia degree 14,15 and Ab was proposed to be responsible for the synapse loss and dementia, we next asked whether reduced Ab levels by TRPC6 could improve AD pathology. At 6 months, the spine density, an indicator to evaluate synapse numbers revealed by Golgi staining, of TRPC6 mouse hippocampus was higher than that of WT mice (Supplementary Fig. 9a-c). But at 11 month the spine density between TRPC6 mice and WT mice was likely similar (Fig. 6a-c). The spine density of APP/PS1 mice either at 6 or 11 months was greatly reduced than that of WT mice, whereas it was markedly increased in APP/PS1/TRPC6 mice than that in APP/PS1 mice, suggesting that TRPC6 prevented the spine loss in APP/PS1 mice. In the behavioural test, we found that at 6 months, TRPC6 mice performed better than WT mice both in the training and probe session of Morris Water Maze ( Supplementary  Fig. 9d,e), while at 11 months, there was no difference in the performance between TRPC6 and WT mice (Fig. 6d,e). However, compared with APP/PS1 mice, APP/PS1/TRPC6 mice spent shorter time to find the hidden platform in training session at 11 months (Fig. 6d), and crossed the platform region more times in the probe test at 6 months ( Supplementary Fig. 9e). Therefore, overexpression of TRPC6 in neurons improves learning and memory in APP/PS1 mice. All these results are consistent with an explanation that TRPC6 downregulates Ab levels to maintain the structural and behavioural plasticity of AD mouse brains.

Discussion
In this study we reported that TRPC6 could function as a novel endogenous g-secretase modulator and specifically regulate Ab production. Several lines of evidence support this conclusion. First, TRPC6 inhibited g-secretase cleavage of APP without effect on aor b-secretase cleavage of APP, Ab clearance or g-secretase cleavage of other substrates tested ( Fig. 2; Supplementary Figs 2,  11 and 15). Second, TRPC6 selectively interacted with APP (C99), but not with Notch (Fig. 3c,e; Supplementary Figs 3b,c,13,14 and  16). Third, the fusion peptide TAT-TM2 prevented g-secretase cleavage of APP, but not that of Notch (Fig. 4h,I; Supplementary  Fig. 14). Finally, elevating TRPC6 in the forebrain neurons reduced Ab levels and improved structural and behavioural plasticity of APP/PS1 mice (Figs 5,6; Supplementary Figs 8 and 9). Thus, affecting TRPC6 levels may be a new strategy to specifically inhibit g-secretase cleavage of APP and reduce Ab production.
Several compounds have been developed to target g-secretase, but they were not successful in clinical trials to treat AD patients 5,34 . Although different reasons may have contributed to the failures, the unwanted side effects due to the strong inhibition   40 , and overexpressing TMP21 in mice causes post-natal growth retardation, severe neurological problems and premature death 41 . Here, we report that TRPC6 interaction with APP leads to the specific reduction in Ab production and improvement in structural and behavioural plasticity in APP/ PS1/TRPC6 mice. It is thus likely that the substrate selectivity of g-secretase could be finely regulated by additional proteins, such as TRPC6, to confer the selective modulation of Ab production.
While a few proteins have been found to regulate the substrate selectivity of g-secretase, the mechanism remains largely unknown. Recently, g-secretase-activating protein (GSAP) was reported to increase g-secretase cleavage of APP (C99) and Ab generation, whereas Notch cleavage was not influenced 42 . It has been shown that GSAP could specifically interact with C99, but not with Notch, an observation similar to TRPC6 reported here (Fig. 3c,e; Supplementary Figs 3b,c, 13, 14 and 16). Thus, selective substrate targeting by interaction may confer a specific modulation of g-secretase cleavage. Actually, it has been reported that small-molecule g-secretase modulators, which regulate g-secretase cleavage of APP, but not that of Notch, bind to C99 with a high affinity, but to Notch with a much lower affinity 43 . However, different from GSAP, which also binds to PS1, TRPC6 does not interact with PS1. Several other proteins have been reported to interact with PS1 to modulate g-secretase cleavage of APP 36,39,40 . In this context, the APP-specific interaction of TRPC6 to regulate Ab generation may basically minimize the risk of inhibiting g-secretase to avoid unexpected side effects, a critical problem encountered in clinical trials. Representative images (f) and quantification (g) of thioflavin S staining of amyloid plaque in the hippocampal region of 6 and 11 months APP/PS1 and APP/PS1/TRPC6 mice with B6C3 background (n ¼ 4-5 mice). Scale bar, 200 mm. Data were presented as means ± s.e.m. Two-tailed Student's t-test was performed. *Po0.05, **Po0.01 versus APP/PS1.
The TRPC channels are nonselective cation channels which play important roles in development 10 and diseases 11,12 . To date all the reported functions of TRPC6 channels require its channel activities. Here, we reported that TRPC6 inhibited g-secretase cleavage of APP likely through protein-protein interaction, but not requiring its channel activity. Actually, Ca v 1.2, an L-type Ca 2 þ channel protein, has been reported to regulate transcription of certain genes independent of its channel activity, but through the C-terminal fragment 44 . TRPM6 and TRPM7, two channel proteins with kinase activity, have been found to phosphorylate myosin IIA and annexin I through their kinase domains 45 . In addition, TRPC1 could promote neurite outgrowth independent of its channel activity 46 . Thus, TRPC6 is a new channel protein to exert cellular functions independent of its channel activity. It is not clear for now whether TRPC6 interaction with APP is regulated. Since downregulation of TRPC6 and DN-TRPC6, all affect Ab production, the level of TRPC6 is critical for its modulation of Ab production. Therefore, it will be informative to examine the expression levels of TRPC6 in AD patients.
In conclusion, our results showed that TRPC6 regulated Ab levels independent of its channel activity. TRPC6 inhibited g-secretase cleavage of APP, but not that of Notch, through its selective interaction with APP. These results suggest that preventing APP interaction with PS1 via TRPC6 could be a novel strategy to reduce Ab formation and reveal a novel cellular function of TRPC6. Finding molecule(s) functioning as TAT-TM2, which is able to reduce Ab production without affecting Notch cleavage, may provide novel intervention candidates to reduce Ab levels and treat AD patients while bypassing the side effects induced by the direct inhibition of g-secretase activity.

Methods
Animals. All animal experiments were performed in accordance with the Institutional Animal Care and Use Committee of the Institute of Neuroscience, Shanghai, China. APP/PS1 mice were from Jackson laboratory (#004462) in B6C3 background, and TRPC6 transgenic mice were constructed in C57 background 19 . Animal age and sex is provided in the related figure legends. To obtain mice with the same genetic background, APP/PS1 mice or TRPC6 mice were first crossed with C57 mice for several generations or with C3H mice, respectively. Then, APP/PS1 mice and TRPC6 mice in the same background were crossed and off springs were used. All the mice were randomized according to their genotype.
Thioflavin S staining, immunohistochemistry and immunocytochemistry. Briefly, mouse brain sections were stained with Thioflavin S (Sigma, 20 mg ml À 1 ) and differentiated in 70% ethanol. Immunostaining of amyloid plaque with Ab antibody was performed according to the instruction of VectaStain Universal ABC kit (VECTOR, PK-6200). Briefly, mouse brain sections were first quenched with H 2 O 2 and antigen retrieved in formic acid. After incubation with 6E10 or MOAB-2 antibody overnight, sections were further incubated with secondary antibody (VECTOR, PK-6200, 1:50), ABC complex and visualized with DAB (VECTOR, SK-4100) as a substrate. These sections were examined under Neurolucidar (Nikon) after mounting. For immunocytochemistry, HEK293 cells from ATCC were co-transfected with TRPC6-HA/APP-Myc, TRPC6-HA/C99-Myc or C6-mut1-HA/ APP-Myc. The cells were trypsinized 1 day after transfection and cultured on 12 mm coverslips for 2 days. The cells were then fixed with paraformaldehyde (PFA), permeabilized with Triton, blocked with normal donkey serum and then stained with primary antibodies overnight, and subsequently incubated with Alexa Flour 488/546/647-conjugated secondary antibodies (Invitrogen and Jackson, 1:1,000). Images were obtained using FV10I confocal microscope (Olympus).
Ab ELISA. Mouse forebrains were dissociated and homogenized with automatic homogenizer (Shanghai Jingxin, JXFSTPRP-48) in TBST. The homogenate was mixed with GuHCl and rotated at room temperature for total Ab quantification. For TBST soluble and insoluble Ab quantification, the homogenate was centrifuged and the supernatant was collected for TBST soluble Ab, and the pellet was further homogenized in GuHCl for TBST insoluble Ab. Ab levels of the mouse forebrain and in culture medium of the cells were quantified by the ELISA kit (Invitrogen, KHB3482 and KHB3544, or IBL, 27717 and 27729).
Ab clearance assay. Ab clearance assay was carried out as previously described 47 .
HEK293 cell was transfected with YFP or TRPC6 for 1 day, and switched to serumfree DMEM with 1 mg ml À 1 dissolved Ab42 (American Peptide) and 10 mg ml À 1 BSA for 12 h. After incubation, the culture medium was collected and Ab42 level was measured by ELISA. The cells were washed with PBS and lysed in 100 ml of 5 M GuHCl, and Ab42 level was measured by ELISA.
Golgi staining. Golgi staining was carried out using Rapid GolgiStain Kit (FD Neurotechnologies, PK401) according to the instruction of the manufacturer. Briefly, after immersion in the impregnation solution for 1 week, mouse brains were cut into 100 mM sections. Sections were stained and dehydrated in graded ethanol, cleared in xylene and mounted. The sections were examined with Zeiss LSM 510 confocal microscope.
Morris water maze. Morris water maze experiment was conducted as previously described 19 . The water maze was a circular pool of 1 m in diameter and divided into four equal quadrants with a platform submerged 1 cm underneath the water surface and kept in the centre of a certain quadrant. During training session, animals were placed into water in randomized order of quadrant each day, with every quadrant once per day. Training session consisted of 8 days, four trials per day, and the time spent to find the hidden platform was recorded as escape latency for each mice. After training, a probe test was carried out with removed platform. Mice were allowed to swim for 1 min and the number of platform area crossing was recorded. Animals showed obvious behavioural abnormality, like never stand on the platform, always rotate or swim around the edge of the maze, were excluded for analysis. Behavioural data were statistically analysed using repeated measures and multivariate analysis in general linear model with LSD post hoc test (SPSS 16.0).
Immunoprecipitation. For the immunoprecipitation, mouse brains or cells were lysed in the buffer containing 0.01 M PB (pH 7.4), 1% NP40, 1% DOC, 0.1% SDS, 0.15 M NaCl and 2 mM EDTA or 20 mM Tris-HCl (pH 8), 137 mM NaCl, 10% glycerol, 1% NP40 and 2 mM EDTA, respectively. The antibodies or control IgG and protein A/G beads were sequentially added into the precipitation system and rotated at 4°C overnight. The beads were washed several times and the immunoprecipitated proteins were resolved by SDS-polyacrylamide gel electrophoresis and analysed by immunoblots.
In vitro secretase activity assay. Briefly, cells were first homogenized and centrifuged, and secretases in the pellet were re-solublized. The a-secretase activity was assayed in 100 mM sodium acetate (pH 7.0) with 2 mg fluorogenic substrate (Calbiochem, 565767), while b-secretase activity assayed in 100 mM sodium acetate (pH 4.5) with 2 mg fluorogenic substrate (Calbiochem, 565758) and g-secretase activity assayed in 50 mM Tris-HCl (pH 6. In vitro purified C99 cleavage assay. The purified C99 recombinant protein was used as the substrate of g-secretase. HEK293 cells were homogenized in lysis buffer (10 mM Tris-HCl, 1 mM EGTA, 1 mM EDTA, pH7.4) and centrifuged at 1,000g for 15 min. The supernatant was further centrifuged at 13,200 r.p.m. at 4°C for 15 min. g-Secretase was reconstituted by resolving the pellets in lysis buffer with 0.2% (w/v) CHAPSO. The in vitro C99 cleavage assay was performed by incubating 20 mg of g-secretase fraction and 2 mg C99-His recombinant protein in 100 ml reaction buffer (50 mM Tris-HCl, 2 mM EDTA, 0.25% CHAPSO, pH 6.8) at 37°C for 4 h. Then, Ab levels were determined by ELISA.
Peptide injection. TAT fusion peptides were synthesized by ChinaPeptides. Male littermate APP/PS1 mice were injected intra-peritoneally with 200 ml 2.5 mM TAT or TAT-TM2 and sacrificed for Ab quantification 3 h after injection in a blinded manner.
Fractionation. Fractionation was carried out as reported previously 49 . Briefly, mouse brain was homogenized in 1 ml homogenization buffer (HB: 20 mM Tris-HCl, pH 7.4, 250 mM sucrose, 1 mM EGTA, 1 mM EDTA and protease inhibitors) by 20 strokes followed by 10 passages through a 21-gauge needle. Lysates were centrifuged at 3,000 r.p.m. for 15 min at 4°C to generate the post-nuclear supernatant. The post-nuclear supernatant was mixed 1:1 with OptiPrep (Sigma, D1556, 60% iodixanol in water). The resultant 2 ml 30% iodixonol mixture was placed at the bottom of an ultracentrifuge tube and was overlaid successively with 1 ml 25, 20, 17.5, 15, 12.5, 10, 7.5, 5 and 2.5% iodixonol in cold HB. The gradients were centrifuged for 20 h at 27,000 r.p.m. at 4°C using a Beckman SW41 rotor. The resulting gradient was collected in 500 ml fractions from the top of the ultracentrifuge tube and prepared for western blot analysis. For each blot, two gels with 11 lanes were run in parallel and the results were merged together.
FAD-linked APP variants were mutated from WT APP with following primer pairs and then cloned into AscI/ClaI sites of pCAGGS vector.
Statistical analysis. All data excluding behavioural experiment were analysed by two-tailed Student's t-test for two groups, and one way ANOVA with Newman-Keuls post hoc test for more than two groups using GraphPad Prism. For all experiments, the sample sizes were chosen according to previous literatures in the field. For all, *Po0.05, **Po0.01, ***Po0.001. All data were presented as mean ± s.e.m.