Dynamin 1 Regulates Amyloid Generation through Modulation of BACE-1

Background Several lines of investigation support the notion that endocytosis is crucial for Alzheimer’s disease (AD) pathogenesis. Substantial evidence have already been reported regarding the mechanisms underlying amyloid precursor protein (APP) traffic, but the regulation of beta-site APP-Cleaving Enzyme 1 (BACE-1) distribution among endosomes, TGN and plasma membrane remains unclear. Dynamin, an important adaptor protein that controls sorting of many molecules, has recently been associated with AD but its functions remain controversial. Here we studied possible roles for dynamin 1 (dyn1) in Aβ biogenesis. Principal Findings We found that genetic perturbation of dyn1 reduces both secreted and intracellular Aβ levels in cell culture. There is a dramatic reduction in BACE-1 cleavage products of APP (sAPPβ and βCTF). Moreover, dyn1 knockdown (KD) leads to BACE-1 redistribution from the Golgi-TGN/endosome to the cell surface. There is an increase in the amount of surface holoAPP upon dyn1 KD, with resultant elevation of α–secretase cleavage products sAPPα and αCTF. But no changes are seen in the amount of nicastrin (NCT) or PS1 N-terminal fragment (NTF) at cell surface with dyn1 KD. Furthermore, treatment with a selective dynamin inhibitor Dynasore leads to similar reduction in βCTF and Aβ levels, comparable to changes with BACE inhibitor treatment. But combined inhibition of BACE-1 and dyn1 does not lead to further reduction in Aβ, suggesting that the Aβ-lowering effects of dynamin inhibition are mainly mediated through regulation of BACE-1 internalization. Aβ levels in dyn1−/− primary neurons, as well as in 3-month old dyn1 haploinsufficient animals with AD transgenic background are consistently reduced when compared to their wildtype counterparts. Conclusions In summary, these data suggest a previously unknown mechanism by which dyn1 affects amyloid generation through regulation of BACE-1 subcellular localization and therefore its enzymatic activities.


Introduction
Late-onset Alzheimer's disease (LOAD) typically manifests after the sixth decade, accounting for over 95% of all AD cases. Genetic studies of LOAD point to a number of risk factor genes, such as apolipoprotein E epsilon4 (ApoEe4) allele [1], and several endocytic proteins. For example, single nucleotide polymorphism (SNP) studies from two research groups studying Japanese and Belgian populations have independently identified an association of dynamin binding protein gene (DNMBP) on chromosome 10 to LOAD, particularly in individuals lacking the APOE e4 allele [2,3]. Furthermore, a significant association of LOAD with the dynamin 2 (DNM2) gene was detected by SNP analysis, especially in non-carriers of the ApoEe4 allele [4,5].
Dynamin is a GTPase that plays a critical role in endocytic vesicle fission [6]. It is encoded by three different genes (DNM1, DNM2, and DNM3) in mammals [7]. Dynamin 1 (dyn1) is highly and selectively expressed in the nervous system and represents the major dynamin isoform expressed in this tissue [8]. Dyn1 has been linked to the biology of AD. For example, dominant-negative dyn1 (K44A mutant), which blocks endocytosis, reduces Ab levels in interstitial spinal fluid (ISF) and prevents activity-dependent increases in Ab [9]. Dyn1 K44A mutant also reduces oligomer Ab 42 -induced neuronal death [10] and increases APP ectodomain shedding [11]. Others showed an increase in BACE-1 cleavage of APP and Ab generation at the cell surface in dyn1 K44A expressing HeLa cells [12]. Taken together, a role for dyn1 in AD is implicated but precise molecular mechanism(s) remains elusive.
Herein, we report that using gene silencing techniques to knockdown dyn1 levels reduces both secreted and intracellular Ab levels in cell culture. There is a dramatic reduction in beta-site APP-Cleaving Enzyme 1 (BACE-1) cleavage products of APP (sAPPb and bCTF). Moreover, dyn1 knockdown (KD) leads to BACE-1 redistribution from the Golgi-TGN/endosome to the cell surface. There is an increase in the amount of surface holoAPP upon dyn1 KD, with resultant elevation of a-secretase cleavage products sAPPa and aCTF. But no changes are seen in the amount of nicastrin (NCT) or PS1 N-terminal fragment (NTF) at cell surface with dyn1 KD. Furthermore, treatment with a selective dynamin inhibitor Dynasore leads to similar reduction in bCTF and Ab levels, comparable to changes with BACE inhibitor treatment. But combined inhibition of BACE-1 and dyn1 does not lead to further reduction in Ab, suggesting that the Ab-lowering effects of dynamin inhibition are mainly mediated through regulation of BACE internalization. Ab Levels in dyn1 2/2 primary neurons, as well as in 3-month old dyn1 haploinsufficient animals with AD transgenic background are consistently reduced when compared to their wildtype counterparts.
In summary, these data suggest a previously unknown modulatory mechanism by which dyn1 affects amyloid generation through regulation of BACE-1 subcellular localization and therefore its enzymatic activities. Together, our findings provide mechanistic evidence that inhibition of dyn1 functions may prevent certain pathologic changes associated with AD.

Ethics Statements
All animal studies were carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The animal studies were approved by the Institutional Animal Care and Use Committee of James J Peters VA Medical Center (Permit Number: CAI-10-044a). All efforts were made to minimize suffering.

Cell Lines
N2a cells stably transfected with cDNAs encoding human APP695 provided by Dr. Greengard at The Rockefeller University [17] were transfected with dyn1siRNA and maintained for 4-5 days to achieve about 50-80% knockdown of dyn1 protein levels. Fibroblast cell lines from dynamin 1, dynamin 2 conditional knockout mice expressing tamoxifen-inducible Cre recombinase (dyn KD) provided by Dr. Pietro De Camilli at Yale School of Medicine [18] lose dynamin expression upon treatment with 1mM 4-hydroxy-tamoxifen. Wild type fibroblast cells were treated by tamoxifen as a control. Alternatively, cells were treated with a dyn inhibitor Dynasore [19] at 10 mM (Sigma), or a BACE inhibitor at 15 mM (Inhibitor IV, EMD) overnight before subjected for further analysis.

Cell Transfection
For siRNA analysis, N2a 695 cells were seeded at 50% to 60% confluence and transfected with 200 pmol siRNA dyn1 versus control duplex using Lipofectamine RNAimax (Invitrogen). For transient transfection of cDNA constructs, dyn KD fibroblasts were seeded at 70% confluence and transfected with 0.5 mg of BACE1-EGFP cDNA provided by Dr. Bradley Hyman at Harvard Medical Center, [20] using Lipofectamine 2000 (Invitrogen).

Cell Lysate Analysis
After transfection, cells were harvested in lysis buffer [21]. Equal amounts of total protein were loaded for Western blot using 6E10 to detect holoAPP (full-length APP). Ab 40 and Ab 42 levels in media were determined by human ELISA kits (WAKO), according to the manufacturer's instructions. Alternatively, mouse Ab 40 levels in media of Dyn KD fibroblasts after various treatments were determined by mouse ELISA kits (WAKO).

Immunoprecipitation
Media or lysates were diluted with immunoprecipitation buffer [22] and immunoprecipitated using 4G8 followed by immunoblotted with 6E10 for Ab and bCTF detection. Media were immunoprecipitated using MAB348 (Millipore) followed by immunoblotted with sAPPb antibody (for sAPPb) or 6E10 antibody (for sAPPa).

Biotinylation/Cell Surface Assay
Dyn KD fibroblasts were incubated at 4uC with 0.5 mg/ml sulfo N-hydroxysuccinimide biotin (Pierce) to label surface proteins as described previously [17]. Samples of both biotinylated and non-biotinylated BACE-1 were analyzed by Western blot using MAB5308 or RU690. Purified lysates from BACE-1 expressing 293 cells were used as positive controls. Alternatively, samples of both biotinylated and non-biotinylated holoAPP were analyzed by Western blot using 6E10.

Generation of Dyn1 Haploinsufficient Mice with AD TRANSGENIC Mouse Background
Human Swedish bAPP and FAD-linked PS1 DE9 mutant transgenic mice from Jackson Laboratory [25,26] were mated with heterozygous dyn1 null mice (dyn1 +/2 ) provided by Dr. Pietro De Camilli from Yale School of Medicine [8]. Double heterozygous F1s were then bred with heterozygous dyn1 null mice only to generate offsprings that express human Swedish APP and FADlinked PS1 DE9 in the dyn1 +/+ or dyn1 +/2 background. Genotypes were determined by PCR amplification as described previously.

Brain Lysate Preparation and Analysis
Mouse brains from APP/PS1 +/2 dyn1 +/+ or APP/PS1 +/2 dyn1 +/2 at 2-3 months of age, were rapidly dissected and snap frozen for further analysis. Each frozen hemi-brain was then processed via step-wise solubilization [21,27]. Lysates were analyzed by Western blot using 6E10 to determine levels of holoAPP and bCTF. Levels of Ab 40 were determined by human Ab 40 ELISA kits (WAKO), according to the manufacturer's instructions. Results were normalized to wet brain weight and expressed as picomoles per gram.

Statistical Analysis
Densitometric analysis of western blot bands (integrated density) was performed using Multigauge v3.1 software (Fujifilm). Levels of holoAPP, bCTF, sAPPb/a, Ab and BACE-1 were normalized to actin and expressed as percentage of control. Absolute Ab 40 concentrations were quantitatively determined by sandwich ELISA (Wako). Independent-samples t tests (parametric design) were used to determine significant mean differences between groups. Significance for t tests are reported with a p,0.05 using two-tailed tests with an a level of 0.05. All statistical analysis was performed using SPSS v18.0.

Dynamin 1 (Dyn1) Knockdown in N2a 695 Cells Reduces Ab and bCTF Levels
To determine whether dyn1 modulates APP metabolism, we first analyzed the effects of dyn1 knockdown on levels of APP and its metabolites in N2a cells stably expressing human APP695 (N2a 695 cells). Using siRNA transfection methods, 50-80% reduction in dyn1 protein levels was achieved after 4-5 days. Both secreted and intracellular Ab levels were significantly decreased by 64.5% and 77.8% respectively, compared to control siRNA transfection ( Fig.1A, p,0.001). There was 62.5% reduction in Ab 40 and 44.7% reduction in Ab 42 levels measured by sandwich ELISA (Fig.1B, p,0.001). There was a dramatic reduction in other APP metabolites sAPPb and bCTF levels (57.4%, p = 0.021 and 45.7%, p,0.001 respectively), without significant changes in holoAPP levels (119.16% of controls, p = 0.101). In addition, the levels of BACE-1 as well as c-secretase components such as PS1 NTF and Pen-2 were unchanged ( Figure S1A).
To confirm the specificity of effects of dyn1 siRNA on APP metabolism, three different dynamin 1 siRNA duplex constructs were synthesized (IDT Inc.) and transfected into N2a 695 cells to determine APP processing. Secreted Ab levels were similarly decreased by all three siRNA constructs compared to control siRNA transfection ( Figure S1B). The specificity of Ab detection was determined by comparing to transfection of APP siRNA. Moreover, to confirm the specificity of bCTF detection, lysates from N2a 695 cells with dyn1 or control siRNA transfection were immunoprecipitated with 4G8 followed by immunoblotting with 6E10 ( Figure S1B, left panels) to determine levels of bCTF and Ab. Alternatively, lysates were directly blotted with 6E10 to detect holoAPP, bCTF and Ab, followed by reprobing with 369 for detection of a/bCTF ( Figure S1C, right panels).
Based on our findings, the consistent changes in levels of bCTF and sAPPb support the notion that dyn1 knockdown inhibits APP cleavage by BACE-1 rather than accelerating degradation of bCTF.

Dyn1 Down-regulation Changes BACE-1 Intracellular Distribution
Next, we evaluated the effects of dyn1 knockdown on BACE1 intracellular traffic using dyn1 knockdown (KD) fibroblasts derived from dyn conditional knockout mice expressing tamoxifen-inducible Cre recombinase [18]. These experimental conditions allowed us to study dynamic changes in BACE-1 with inducible dyn1 down-regulation. After tamoxifen treatment at 1mM for 5-7 days, dyn1 protein levels were reduced by 56.6% ( Figure S2A, p = 0.03) in the KD fibroblasts. Alternatively, wildtype fibroblast cells were treated with tamoxifen as a control and there was no change in dyn1 protein levels with tamoxifen treatment ( Figure  S2B, right panel). Similar to the results using dyn1 siRNA in the N2a 695 cell line, reduced bCTF levels ( Figure S2A; 66.7% of reduction, p = 0.008) were observed in dyn1 KD fibroblasts with tamoxifen treatment. Tamoxifen treatment in wildtype cells did not cause any changes in a/bCTF levels, ruling out an effect of tamoxifen itself ( Figure S2B, left panel).
BACE-1 intracellular distribution in dyn1 KD cells was studied after transfection with EGFP-BACE1 construct [20]. As shown in Fig. 2, amounts of EGFP-BACE-1 at the cell surface were significantly increased with dyn1 knockdown. Under control conditions, BACE-1 (green fluorescence) was mainly localized in the endosomes and TGN/Golgi, determined by an endosomal marker EEA1 (red fluorescence, Fig. 2A,) as well as a TGN marker c-adaptin (red fluorescence, Figure 2B). However, after dyn1 knockdown, amounts of green fluorescence (EGFP-BACE1) were dramatically increased at cell surface with little remaining within the TGN/Golgi and endosomes. Transfected EGFP-BACE-1 expression levels were comparable between control cells and dyn1 KD cells ( Figure S3A).

Dyn1 Knockdown Increases Cell Surface BACE-1 and holoAPP Levels
Using surface biotinylation methods, we next analyzed the intracellular and surface distribution of endogenous BACE-1 in dyn1 KD fibroblasts without over-expression of EGFP-BACE-1. As shown in Fig.3A (top panel), surface BACE-1 levels were elevated by 251.8% with dyn1 knockdown when compared to controls (p = 0.047). There was a modest elevation in total BACE-1 protein levels without achieving statistical significance (increase by 78.5%, p = 0.08). The specificity of BACE-1 detection was confirmed by using cell lysates extracted from BACE-1 expressing 293 cells as positive controls.
Interestingly, cell surface holoAPP levels in dyn KD cells were also elevated by 131.1% (p = 0.02) upon tamoxifen induction, without any significant changes in total APP protein levels (111.6% of controls, p = 0.615) as shown in Figure 3B. In addition, the levels of sAPPa in N2a 695 cells determined by immunoprecipitation with MAB348 followed by immunoblotting with 6E10 antibody, was increased by 286.3% (p = 0.042) upon dyn1 knockdown by siRNA transfection. Consistently, the levels of aCTF were also elevated with dyn1 KD in both N2a 695 cells ( Figure S1C) and in dyn1 KD fibroblasts ( Figure S2A; increase by 84.9%, p = 0.011). The increased holoAPP at cell surface likely promotes a-secretase cleavage to generate sAPPa and aCTF.
However, it should be noted that dyn1 knockdown does not have a general effect on internalization of all proteins because no significant changes were seen in the amounts of nicastrin (NCT) or PS1 NTF at cell surface upon dyn KD conditions ( Figure S3B). In addition, the total protein levels of NCT and NCT were unchanged with dyn1 KD ( Figure S3B, input panels). The changes in both BACE-1 and holoAPP at cell surface with dyn1 KD suggest that dyn1 regulates a common trafficking pathway for those two proteins between endosomes and plasma membrane.
Taken together, these data suggest that dyn1 knockdown inhibits endocytosis of BACE-1, thereby preventing delivery of BACE-1 to its active sites in endosomes [28]. The BACE-1 on the cell surface is likely to be inactive and unable to process holoAPP to generate bCTF and Ab. Therefore, increased holoAPP availability at surface for a-secretase cleavage leads to elevation in sAPPa ( Figure 3B) and aCTF ( Figure S1C and S2A).

A Selective Dynamin Inhibitor Dynasore Reduces bCTF and Ab Levels
Next, to determine whether pharmacological inhibition of dyn1 could affect APP processing, N2a 695 cells were treated with a selective dynamin inhibitor dynasore at 10mM [19,29,30]. As shown in Figure 4A, levels of bCTF and Ab secreted in media were decreased by 47.4% (p = 0.008) and 70.4% (p,0.001), comparable to that with a BACE inhibitor treatment [31] (reduction of bCTF by 38% p = 0.006, and Ab by 50.4% p = 0.011).
Furthermore, the effects of dyn1 knockdown on BACE-1 cleavage of endogenous APP were confirmed in dyn1 KD fibroblast cell lines. Similar to the results using siRNA and chemical inhibitors in N2a 695 cells, reduction in bCTF levels ( Figure S2A; 66.7% of reduction, p = 0.008) was observed in these cells upon dyn1 knockdown. The total Ab 40 levels measured by sandwich ELISA were reduced by 39.4% (p,0.001), comparable to the reduction induced by a BACE inhibitor treatment (41.3% reduction, p,0.001; Fig. 4B). However, the combination of BACE inhibitor IV with tamoxifen-induced dyn1 KD did not lead to further reduction in Ab 40 levels than either treatment alone (41.3% reduction). Consistently, Ab 40 levels were reduced by 40% in dyn1 KD fibroblasts with dynasore treatment alone. The combination of dynasore and BACE inhibitor treatment did not lead to further reduction in Ab 40 levels as well (43.5% reduction). The lack of additive effects with combined treatments suggests that the Ab-lowering effects of dyn1 KD, dynasore and BACE inhibitor IV all act through the same step: BACE1 inhibition.

Genetic Perturbation of Dyn1 in Animals Reduces Ab Generation
Dyn1 has been implicated in regulating activity-dependent Ab secretion through endocytosis [9]. To determine whether dyn1 has a direct effect on endogenous Ab generation, we next analyzed levels of newly synthesized Ab secreted from dyn1 2/2 neurons [8] using S 35 -methionine labeling approach. When compared to dyn1 +/+ neurons, levels of S 35 -labeled Ab (both Ab 1-40/42 and Ab 11-40/42 species as described previously [23,24]) secreted into the media were decreased by 55.9% in dyn1 2/2 neurons after 5 hours of incubation (Fig. 5A, p = 0.004). The dyn1 genotypes were confirmed by PCR results (Fig. 5A, middle panel) and the levels of Figure 2. Dyn1 down-regulation changes BACE-1 intracellular distribution. A) Dyn1 fibroblast cells treated without (control; ctrl) or with tamoxifen (dyn1 KD) were transiently transfected with EGFP-BACE-1 followed by fixation. Localization of BACE-1 is shown as green fluorescence (labeled as EGFP-BACE). In addition, cells were double stained with an endosomal marker, EEA1 (red; labeled as EEA1). Overlay of both signals is shown as yellow (labeled as Overlay). Nuclear staining by DAPI is shown as blue (labeled as DAPI). B) Localization of BACE-1 is shown as green fluorescence (labeled as EGFP-BACE). In addition, cells were double stained with a TGN marker, c-adaptin (red; labeled as c-adaptin). Overlay of both signals is shown as yellow (labeled as Overlay). Nuclear staining by DAPI is shown as blue (labeled as DAPI). BACE-1 (green fluorescence) is mainly localized in the endosomes and TGN/Golgi in control cells. However, after dyn1 knockdown, the amounts of green fluorescence (EGFP-BACE1) are dramatically increased at cell surface with a little remaining within the TGN/Golgi and endosomes. doi:10.1371/journal.pone.0045033.g002 b-actin were comparable between dyn1 +/+ and dyn1 2/2 neurons (Fig.5A, bottom panel).
Dyn1 knockout animals can only survive up to 2 weeks [8]. To determine whether genetic perturbation of dyn1 affects Ab generation in adult animals in vivo, we next analyzed the levels of Ab, holoAPP and bCTF in the brains of dyn1 haploinsufficient (dyn1 +/2 ) mice with AD transgenic background expressing human Swedish APP and PS1DE9 mutations [25,26]. At 2-3 months of age, levels of Ab 40 determined by sandwich ELISA in total brain lysates from AD transgenic mice carrying dyn1 haploinsufficiency (n = 8), were decreased by 35.5% when compared to controls (n = 9; p = 0.013; Fig. 5B). However, levels of Ab 42 at this age group were too low for detection of any statistically significant changes. The levels of bCTF in the brains of dyn1 haploinsufficient animals (dyn1 +/2 ) were slightly decreased (13.7%) when compared to dyn1 wild-type (dyn1 +/+ ) animals ( Figure S4). There was a trend of reduction in bCTF in dyn1 +/2 mouse brains, but due to large variations in these transgenic animals (even within the same group), this effect did not achieve statistical significance (p = 0.613). The levels of holoAPP were comparable between the two groups ( Figure S4:107.4% of controls in dyn1 +/2 mice; p = 0.679). . Dyn1 knockdown increases cell surface BACE-1 and holoAPP levels. A) Dyn KD fibroblast cells were treated with tamoxifen to induce dyn1 knockdown and then labeled with biotin followed by immunoprecipitation with streptavidin beads to pull down surface BACE-1. The amount of endogenous BACE-1 at cell surface was compared between control and dyn1 KD conditions. Total protein levels of endogenous BACE-1 were also determined from lysates as input. Protein levels were normalized to b-actin content and expressed as percentage of control. Data were collected in duplicate or triplicate from three independent experiments. Elevations in BACE-1 at cell surface (*p,0.05) without significant changes in total protein levels, were observed upon dyn1 KD. Cell lysates of BACE-1 expressing 293 cells were used as positive controls in these experiments. B) Alternatively, the amount of surface holoAPP was compared between control and DKO conditions. Elevations in holoAPP at cell surface (*p,0.05) without significant changes in total protein levels, were observed upon dyn1 KD. The sAPPa levels were determined in N2a 695 cells after dyn1 siRNA transfection, by immunoprecipitation with MAB348 followed by immunoblotting with 6E10 antibody. The levels of sAPPa was increased by 286.3% (p = 0.042) upon dyn1 KD. doi:10.1371/journal.pone.0045033.g003

Discussion
It is well established that sorting mechanisms which cause APP, BACE-1 and c-secretase to colocalize in the same subcellular compartment play important roles in the regulation of Ab production [32]. Further, endocytic pathways are important for regulation of BACE-1 activity (Huse et al., 2000). Substantial data regarding the nature and mechanisms underlying APP endocytosis have been reported, but the role of endocytosis in regulating BACE-1 distribution among endosomes, TGN and plasma membrane remain unclear.
As an important endocytic adaptor protein, the involvement of dynamin in Ab biogenesis has been described [9,11,12], raising the possibility that it may affect Ab biogenesis as a step in the pathogenesis of AD. However, there are discrepancies between prior reports regarding its effects on Ab. One group showed increased cell surface Ab production with dominant negative dyn1 K44A [12] but other groups reported reduced Ab production by dyn1 K44A [9,11]. It was also suggested that impaired APP endocytosis by dyn1 K44A contributes to reduction in activitydependent Ab secretion [9]. A similar situation has been proposed for deficiencies of other sorting proteins (e.g., SorL1 and Vps35), both of which, like dynamin, will undoubtedly alter sorting of multiple substrates. Yet, deficiencies of either SorL1 or Vps35 can promote Ab generation and accumulation as prominent phenotypes at the whole brain level [21,33].
Our data, using gene silencing in three different cell lines (including N2a cells overexpressing human APP695 cells, fibroblasts with endogenous APP expression, and primary neurons derived from dyn1 2/2 animals), as well as in animal models, demonstrates a previously unknown mechanism by which dyn1 regulates Ab generation; i.e., through modulation of BACE-1 traffic between the plasma membrane and endosomes. By reducing BACE-1 localization to the endosomal compartment, dyn1 inhibits Ab generation and secretion. These data further support the idea that endosomes are the most optimal compartments for BACE-1 enzymatic activities and that prevention of BACE-1 distribution into endosomes can reduce APP processing into Ab. The reduction of both Ab and bCTF/sAPPb through dyn1 knockdown may be a better therapeutic strategy compared to those targeting c-secretase, given that recent data suggest bbut not c-secretase-mediated proteolysis of APP causes synaptic and memory deficits in AD [34]. We would suggest that the previously reported effects of dyn1 knockdown on Ab generation [9,11,12] are more likely to be attributable to modulation of BACE-1 sorting and less likely to be attributable to the conventional wisdom that dyn1 is primarily acting via reduction of APP internalization. A combination of dynasore (or dyn1 knockdown) and BACE inhibitor treatment did not cause any further reduction in levels of Ab generation when compared with the maximal effect of either treatment on its own ( Figure 4B), consistent with the interpretation that they both act on the same pathway. If dynamin knockdown/blockade caused an inhibition of Ab generation by acting via inhibition of APP internalization, then there would be further reduction of Ab when the BACE inhibitor was added. Together, our data suggest that the Ab-lowering effects of dynamin inhibition are mainly mediated through regulation of BACE-1 internalization.
Interestingly, a recent report demonstrated that an immediate early gene, Arc/Arg3.1 which recruits dynamin and endophilin to early/recycling endosomes at post-synaptic terminals, can mediate PS1/c-secretase trafficking through the endosomal pathway, and thereby affect activity-dependent Ab generation [35]. These data suggest that dyn1-regulated endocytosis may also regulate c-secretase function. However, our data demonstrate that dyn1 knockdown affects only BACE-1 and holoAPP traffic to plasma membrane but not the c-secretase components nicastrin and PS1 NTF ( Figure S3B), suggesting a shared pathway for BACE-1/ holoAPP selectively regulated by dyn1.

Conclusions
In summary, here we report a previously unknown regulatory mechanism by which dyn1 modulates APP proteolysis and Ab homeostasis through regulation of BACE-1 subcellular distribution. Without directly targeting BACE-1 enzymatic activities, modulation of BACE-1 subcellular localization by targeting dyn1 suggests new directions for developing therapeutic strategies for AD. Figure S1 Genetic silencing of dyn1 in N2a cells reduces bCTF and Ab levels. A) Western blot analysis of PS1 NTF by Ab14, Pen-2 by rabbit polyclonal anti Pen-2 antibody NE1008 (Calbiochem), and BACE-1 by clone 61 and RU690, were The levels of b-actin were also determined in dyn1 +/+ and dyn1 2/2 primary neuron lysates. Data were collected in duplicate or triplicate from three independent experiments. B) Levels of Ab 40 in hemi-brain lysates of 3-month old APP/PS1 +/2 dyn1 +/+ (n = 9) and APP/PS1 +/2 dyn1 +/2 (n = 8) mice were determined by sandwich ELISA analysis. Statistically significant reduction in Ab 40 was observed in APP/PS1 +/2 dyn1 +/2 mice in comparison to their control littermates (APP/PS1 +/2 dyn1 +/+ ; *p,0.05). Data were presented as pmol/gram of wet brain weight of each animal. doi:10.1371/journal.pone.0045033.g005 performed in N2a695 cells with transfection of dyn1 siRNA or control duplex. There were no significant differences in total protein levels of BACE-1 or c-secretase components with dyn1 KD conditions. B) Three different dyn1 siRNA duplex were synthesized (IDT Inc.) and transfected into N2a 695 cells to determine the specificity of observed effects of dyn1 KD on APP processing. The secreted Ab levels were similarly decreased in three different dyn1 siRNA treatment, compared to control siRNA transfection. The specificity of Ab detection was determined by comparing to transfection of APP siRNA. Sequences of the dyn1siRNAs are as follows: Duplex 1 forward 59-rGrGrC rUrUrA rCrArU rGrArA rUrArC rCrArA rCrCrA rCrGA A-39; reverse, 59-rUrUrC rGrUrG rGrUrU rGrGrU rArUrU rCrArU rGrUrA rArGrCrCrArG-39. Duplex 2 forward 59-rGrGrA rCrArU rArGrA rCrGrG rCrArA rGrArCrATC-39; reverse, 59-rGrArU rGrUrC rCrUrU rCrUrU rGrCrC rGrUrC rUrArU rGrUrC rCrUrU-39. Duplex 3 forward 59-rCrGrG rUrUrA rGrArC rArGrU rGrCrA rCrCrA rArGrA rArGrC T-39; reverse, 59-rArGrC rUrUrC rUrUrG rGrUrG rCrArC rUrGrU rCrUrA rArCrC rGrUrG-39. Stealth siRNA Control Medium GC Duplex was used as a negative control (IDT Inc.). Sequences are described as before. APP siRNA sequences are as follows: forward 59-rUrCrC rUrCrC rGrUrC rUrUrG rArUrA rUrUrU rGrUrC rArArC rCrCrA-39, reverse 59-rGrGrU rUrGrA rCrArA rArUrA rUrCrA rArGrA rCrGrg rArGGA-39. C) Media and lysates from N2a 695 cells with dyn1 or control siRNA transfection were immunoprecipitated with 4G8 followed by immunoblotted with 6E10 to determine levels of bCTF and Ab. Alternatively, lysates were directly blotted with 6E10 to detect holoAPP, bCTF and Ab, followed by reprobed with 369 for detection of a/bCTF. (TIF) Figure S2 Dyn1 knockdown in fibroblast cells changes endogenous APP processing. A) Dyn1 KD cells were treated with tamoxifen to induce dyn1 knockdown. After tamoxifen treatment at 1 mM for 5-7 days, dyn1 protein levels were reduced by 56.6% (p = 0.03) in the dyn1 KD cells. Similar to the results in N2a 695 cells, changes in metabolism of endogenous APP including reduction in bCTF levels (66.7% of reduction; p = 0.008) were observed in these cells upon dyn1 knockdown. Elevation of aCTF levels was also observed (84.9% of increase; p = 0.011). B) Wild type fibroblast cells were treated with tamoxifen at 1 mM for 5-7 days as a control. The levels of bCTF and aCTF, as well as dyn1 protein levels were unchanged with or without tamoxifen treatment (ctrl versus Tamoxifen). (TIF) Figure S3 Dyn1 knockdown increases cell surface BACE-1 levels. A) Dyn KD fibroblast cells were treated with tamoxifen to induce dyn1 knockdown followed by transiently transfected with EGFP-BACE-1. The levels of EGFP-BACE1 in dyn DKO cells with or without tamoxifen induction (three independent transfection experiments in both control and DKO cells) were comparable as shown in top panel (determined by Invitrogen anti-GFP antibody 3E6). The levels of dyn1 and actin were also determined (bottom panels). B) The amounts of nicastrin (NCT) or PS1 NTF at cell surface were also determined by biotinylation followed by streptavidin pull down. Western blot analysis of surface and total (input) NCT by anti-nicastrin antibody (BD Transduction) and RU717, as well as surface and total (input) PS1 NTF by Ab14 was performed. No significant changes were seen in the amounts of NCT and PS1NTF at cell surface or total cell lysates upon dyn KD conditions. (TIF) Figure S4 Genetic perturbation of dyn1 in animals slightly reduces bCTF levels. Levels of bCTF and holoAPP in total hemi-brain lysates of 3-month old APP/PS1 +/2 dyn1 +/+ (n = 9) and APP/PS1 +/2 dyn1 +/2 (n = 8) mice were determined by western blot with antibody 6E10. Levels of bCTF and holoAPP were normalized to actin and presented as ratio to wild type counterparts. The levels of bCTF in the brains of dyn1 haploinsufficient animals were slightly decreased (13.7%) when compared to dyn1 wild-type animals. But due to large variations in these transgenic animals, this effect did not achieve statistical significance (p = 0.613). The levels of holoAPP were comparable between two groups (107.4% of controls in dyn1 +/2 mice; p = 0.679). (TIF)