APP-Mediated Signaling Prevents Memory Decline in Alzheimer’s Disease Mouse Model

SUMMARY

Amyloid precursor protein (APP) and its metabolites play key roles in Alzheimer’s disease (AD) pathophysiology. Whereas short amyloid-β (Aβ) peptides derived from APP are pathogenic, the APP holoprotein serves multiple purposes in the nervous system through its cell adhesion and receptor-like properties. Our studies focused on the signaling mediated by the APP cytoplasmic tail. We investigated whether sustained APP signaling during brain development might favor neuronal plasticity and memory process through a direct interaction with the heterotrimeric G-protein subunit Gα_S (stimulatory G-protein alpha subunit). Our results reveal that APP possesses autonomous regulatory capacity within its intracellular domain that promotes APP cell surface residence, precludes Ab production, facilitates axodendritic development, and preserves cellular substrates of memory. Altogether, these events contribute to strengthening cognitive functions and are sufficient to modify the course of AD pathology.

In Brief

Deyts et al. find that APP-mediated signaling, which occurs in lipid-raft microdomains through an interaction between APP C-tail and the heterotrimeric G-protein subunit Gα_S, provides a positive feedback regulatory loop that promotes non-amyloidogenic APP processing. Continuous Gα_S/cAMP- dependent signaling through APP C-tail preserves spatial memory in an Alzheimer’s disease mouse model.

Graphical Abstract

INTRODUCTION

The genetic association of amyloid precursor protein (APP) with Alzheimer’s disease (AD) has spurred investigations on the pathophysiological roles of APP in the past few decades. The generation of multiple APP metabolites, including β-amyloid peptides (Aβ), translates to the intricate complexity of APP’s cellular functions. Whereas there is a wealth of information on the neurotoxicity associated with Ab peptides, our study explores the APP molecule as a whole cellular component that may influence neuronal function and impact the course of AD. Proteolytic processing of APP by α- and β-secretases releases the ectodomain and generates membrane-bound C-terminal fragments (APP-CTF) comprising the transmembrane and cytoplasmic domains (Andrew et al., 2016; Deyts et al., 2016b; Haass et al., 2012; Jiang et al., 2014; Műller et al., 2017). Subsequently, γ-secretase cleavage of APP-CTF releases Ab and the APP intracellular domain (AICD) from the membrane. Numerous type I transmem- brane proteins, including signaling receptors, undergo ectodomain shedding, and subsequent γ-secretase e cleavage terminates the cell surface signaling by releasing their C-terminal stubs (Fortini, 2002; Kopan and Ilagan, 2004; McCarthy et al., 2009). APP could also function as a receptor-like protein where the membrane-tethered APP-CTF mediates intracellular signaling until γ-secretase e cleavage and the release of the AICD from the membrane terminates it (reviewed in Deyts et al. 2016b). APP-mediated signaling is enhanced when γ-secretase e cleavage of the APP-CTF is inhibited or when the AICD is tethered to the membrane via a lipid anchor (mAICD), such that it is competent for signaling but remains unperturbed by g-secretase (Deyts et al., 2012, 2016a). We were able to capture and better characterize the intricate function of APP-CTF using an experimental mAICD construct fashioned after a membranetethered Caenorhabditis elegans homolog of DCC (UNC-40)/deleted in colorectal cancer (DCC) construct, used previously to characterize DCC signaling (Gitai et al., 2003; Zacharias et al., 2002). The likelihood that APP-mediated signaling could be mediated through membrane-bound APP-CTFs is supported by our current results, our recent in vitro and in vivo studies (Deyts et al., 2016a), and others (reviewed in Deyts et al., 2016b and Schettini et al., 2010). Experimentally targeting AICD at the membrane by expressing mAICD (devoid of the Ab sequence) allowed us to achieve constitutive activation of signaling through APP C-terminal tail (C-tail).

Our study examined the in vivo significance of APP C-tail and its associated signaling partners in contributing to cognitive function and amyloidogenic cascade. We used recombinant adeno-associated virus (rAAV) brain delivery to achieve sustained APP signaling through brain expression of mAICD in a transgenic mouse model that expresses familial AD (FAD)-linked mutant proteins. Our results indicate that mAICD expression in the brains of neonatal mice is sufficient to preserve cognitive function in an amyloidogenic 5XFAD (transgenic mice expressive five familial AD mutations) mouse model. We demonstrate that APP interaction with the G-protein subunit GαS (stimulatory G-protein alpha subunit) attenuates amyloidogenic APP processing and Ab production through the increase of cell surface residence of APP.

RESULTS

APP-Mediated Signaling Rescues Cognitive Deficit in AD Mouse Model of Amyloidosis

We previously observed that in vitro APP-mediated signaling through overexpression of mAICD causes axodendritic arborization as a result of direct coupling with GαS and subsequent activation of adenylate cyclase and CREB signaling (Figures S1C–S1E) (Deyts et al., 2012, 2016a). CREB associated signaling cascade strongly correlate with synaptic enhancement and memory consolidation (Abel and Nguyen, 2008; Alberini and Kandel, 2014; Pittenger et al., 2002; Rogerson et al., 2014). As a proof of concept, we delivered rAAV-mAICD in neonatal brain of 5XFAD mice to investigate if APP C-tail interaction with GαS could alleviate memory deficits in a model of amyloidosis. We employed the novel object recognition, the Y-maze, and the fear conditioning tasks to evaluate cognitive impairment in 5XFAD and non-transgenic (NTg) mouse littermates at 6 months of age (Andrew et al., 2017; Oakley et al., 2006) (Figures 1A–F). 5XFAD mice showed impairment in preferential recognition of the novel object (about 1.3%, compared with 38% of discrimination index in NTg littermates; p < 0.001; Figures 1A, 1B, S2A, and S2E), which was mitigated by the expression of mAICD as reflected by an increase of discrimination index to the levels seen in NTg mice (32% versus 38%, respectively p = 1.000). IRES-mediated lower level expression of mAICD also produced a trend of improvement in memory behavior in 5XFAD mice (18% discrimination; p = 0.213 as compared to 5XFAD-membranetethered control (mCtl); z-score analysis in Figure S3A), suggesting a dose-dependent effect. In contrast to mAICD, brain expression of mAICDmutAAA (mAICD variant of residues RHLSK mutated to AALSA), which lacks the GαS interacting site, significantly reduced the 24 h memory recognition in NTg mice (p = 0.009; Figures S2A and S3A) and failed to rescue memory function in 5XFAD mice (p = 1.000; Figures 1B, S2E, and S3A). Exploration activity of the objects was not significantly affected by the expression of mAICD constructs (Figures S2B and S2F), suggesting that the mouse cohorts under study exhibited similar explorative interest.

Figure 1. Sustained APP-Mediated Signaling Rescues Cognitive Deficit in an Amyloidogenic Mouse Model.

Cognitive mouse behaviors were assessed in NTg and 5XFAD mice that expressed mCtl, mAICD, and mAICD variants.

(A–J) At 5–6 months after neonatal intracerebroventricular (ICV) rAAV injection, mice were tested for novel object recognition (A and B), spontaneous alternation in Y-maze (C and D), contextual fear conditioning (E and F), open-field (G and H), and time spent in the dark (I and J).

(B) Novel object discrimination index indicates that 5XFAD mice spent less time exploring the novel object, a condition that was rescued in mice expressing mAICD.

(D) Percentage of spontaneous alternation is rescued in 5XFAD mice expressing mAICD. # p < 0.05 and ## p < 0.01 compared with 5XFAD-mCtl, using one-way ANOVA Kruskal-Wallis test followed by Dunn’s post hoc analysis.

(F) Fear conditioning behavior is documented as a change in cumulative freezing time 24 h after exposure to foot shocks.

p values are compared with mCtl or mAICD respective conditions using two-way ANOVA followed by Sidak’s multiple time comparison test. All mice under study exhibited no change in anxiolytic behaviors (H and J). IRES-mAICD (purple), mAICD (orange), and mAICDmutAAA (blue) color-coded significances compared with mCtl, and mAICDmutAAA (green) color-coded significance compared with mAICD. The total number of animals is shown in parentheses. Error bars indicate SEM.

In the Y-maze, we tested short-term working memory through spontaneous alternation (Figure 1C). 5XFAD mice showed impairment in spontaneous alternation behavior compared with the NTg littermates (Figure 1D; p = 0.004). The reduced percentage of success in spontaneous alternation was rescued in 5XFAD mice expressing mAICD compared with NTg littermates (Figures 1D and S2G; p = 0.945). The expression of mAICDmutAAA did not alter the exploratory behavior in 5XFAD mice whereas it diminished spontaneous alternation of NTg mice (Figure S2C; p = 0.013). None of the mice displayed changes in the locomotor activity (Figures S2D and S2H). These findings suggest that the GαS protein–interacting motif in APP cytoplasmic tail could contribute to short-term working memory function.

We performed the contextual fear-conditioning task to evaluate stress-induced associative learning memory (Maren et al., 2013) (Figure 1E). We observed a significant increase of freezing time in 5XFAD mice expressing mAICD on the first day of the experimentation following the foot shock (Figures S2I and S2K; F(1,25) = 12.48, p = 0.002). These results are consistent with an improvement of short-term working memory, seen as well in the mAICD cohort performing the spontaneous alternation task (Figure 1D). Sustained memory retention was further observed on the second day of the experimentation (Figures 1F, S2J, and S2L; F(1,25) = 5.001, p = 0.034). 5XFAD mice expressing mAICDmutAAA also responded better to the fear environment (Figure S2K; day 1 F(1,29) = 16.01, p = 0.0004; and Figure 1F, day 2 F(1,29) = 13.17, p = 0.001). NTg mice were unaffected by the expression of mAICD constructs in this task (Figures S2I–Q). The anxiolytic behaviors recorded in open-field (Figures 1G and 1H) and dark-arm (Figures 1I and 1J) settings were not altered (see also Figures S3B and S3C).

APP-Mediated Signaling Reduces Ab Burden in AD Mouse Model

We investigated the consequence of sustained APP signaling on Aβ deposition by quantifying Aβ burden at early (3-month-old) and later (6-month-old) stages of amyloidosis in 5XFAD mice (Figure 2). Aβ deposition was very prominent in the subiculum area and greater in the deeper layers of the cortex (Figures S4A and S4B). mAICD expression reduced Ab burden in select brain areas compared with that seen in the mCtl cohort (Figures 2 and S4). Aβ burden, which consists of a combined contribution of the deposit number and deposit size, almost tripled between 3 and 6 months of age in the cortical and hippocampal areas (Figure 2B). However, a larger reduction of burden was detected only in the hippocampal areas in mAICD cohorts at 3 and 6 months of age (Figure 2B). The deposit number was the most significant contributor to a reduced burden by mAICD expression, but not by mAICDmutAAA, at 3 months of age in the hippocampus (Figure S4D). The size of the deposit was not affected by either mAICD or mAICDmutAAA at that age (Figure S4G; Table S1). On the other hand, we found that the size of the clusters was reduced in the subiculum area (p < 0.005) in 6-month-old mice expressing mAICD, while the deposit number was not changed in that area (p = 1.000; Table S1).

Figure 2. Sustained APP-Mediated Signaling Reduces Ab Burden in 5XFAD Mice.

Brain sections from 5XFAD mice, injected ICV with rAAV expressing mCtl, mAICD, or mAICDmutAAA, were immunostained for Aβ.

(A) Representative confocal images from 6-month-old mice depict a selective reduction in Ab burden in hippocampal areas of mice expressing mAICD (Figure S4B shows un-cropped images). Scale bar, 100 μm.

(B) Aβ burden was quantified in cortex and hippocampus 3 and 6 months after ICV injection.

(C) Percentage of changes in Aβ immunostaining between both time periods is shown.

(D) Relative changes of Aβ burden are quantified in the parietotemporal cortex (Cx), hippocampal CA1, CA3, and DG (shadow area).

(E) Average histogram representation between deposit number and size are shown for the hippocampus in 6-month-old mice. p values are compared with mCtl or mAICD respective conditions using two-way ANOVA followed by Sidak’s multiple size comparison test.

(F) Reduced number of large deposits is noticed in mice expressing mAICD.

(G) Aβ42/Aβ40 ratio analysis is represented using total brain lysates from 6-month-old 5XFAD mice extracted with formic acid. Data point distribution and columns of SD of the mean are shown. Error bars indicate the minimum and maximum point distributions.

* p < 0.05, ^** p < 0.01, and ^*** p < 0.001 compared with mCtl for each condition, using ANOVA Kruskal-Wallis test followed by Dunn’s post hoc analysis. mAICD (orange) and mAICDmutAAA (blue) color-coded significances are compared with mCtl, and mAICDmutAAA (green) color-coded significance are compared with mAICD. The total number of animals is shown in parentheses. Error bars indicate SEM unless otherwise specified.

Intriguingly, mAICDmutAAA expression promoted earlier Ab burden in the cortical area, seen mainly as an increase of deposit number in the younger cohort (Figures 2B and S4D; Table S1). Aβ burden and Aβ deposit number did not increase further with age in the cortex (or in the hippocampus) in mice expressing mAICDmutAAA, as supported by a reduction in the percentage of change between 3- and 6-month-old mice (Figures 2C, S4E, and S4K). The accelerated rate of Aβ burden is disrupted by the expression of mAICDmutAAA in the cortex and the hippocampus, whereas expression of mAICD slowed down the rate of Aβ deposition only in the hippocampus (Figures 2B and S4J). The lack of GαS interacting site in APP C-tail appeared to speed up Aβ deposition in the cortex while it did not increase further with age in either brain area (Figures S4J–L). This observation suggests that the cyclic adenosine monophosphate (cAMP)-dependent protein kinase A (PKA)-dependent pathway might prevent additional accumulation of Aβ deposits in the cortex.

When Aβ burden was analyzed in detail in 6-month-old mice, we observed that mAICD expression resulted in a significant decrease of Aβ staining in CA1, CA3, and dentate gyrus areas of the hippocampus (Figures 2D, S4F, and S4I; Table S1). Mice expressing mAICDmutAAA did not exhibit a reduction in Ab burden in the dentate gyrus (Figure 2D, highlighted image), suggesting that the GαS interaction site within mAICD is required for reducing Aβ burden only in select brain areas. We further observed that mAICD and mAICDmutAAA expression diminished the number of smaller deposits in the hippocampus of 6-month-old 5XFAD mice (Figure 2E; p = 0.0006 for mAICD, and p = 0.002 for mAICDmutAAA), whereas mAICDmutAAA did not affect significantly larger size plaque deposition, especially in the dentate gyrus (Figures 2A, 2F, S4B, S4H, and S4I; Table S1). As a point of interest, the average size of the deposit is considerably larger in the dentate gyrus than any in other brain areas, therefore suggesting that the PKA-dependent pathway might also facilitate Aβ clearance in this area.

To complement this analysis, we measured Aβ levels in the forebrains of 6-month-old 5XFAD mice. We observed no changes in soluble Aβ levels (Figures S4M and S4N). Insoluble Aβ42 and the Ab42/Ab40 ratio from formic acid homogenates were slightly reduced (but not significantly) in mAICD- and mAICDmutAAA- expressing mice (Figures 2G, S4O, and S4P), which is consistent with observed select decreases of Ab immunostaining in brain regions.

APP-Mediated Signaling Regulates APP Residence at the Cell Surface

To understand the mechanisms involved in reducing Aβ accumulation, we assessed whether mAICD expression might affect residence of full-length APP (APP-FL) at the membrane. Cell surface distribution of APP-FL was examined using an ectodomain APP antibody (APP-NT452) in a stable HEK293 cell line expressing wild-type APP (Figures 3A and 3B). A γ-secretase e inhibitor or the presenilin expression of the loss-of-function mutant PS1-D385A were used as positive controls for the accumulation of cell surface APP-FL (Figure 3B) (Leem et al., 2002). We found that mAICD expression favored APP-FL accumulation at the cell surface, an effect that was abolished by co-treatment with the adenylate cyclase inhibitor MDL-12,330A (Figures 3A and 3B, p < 0.0001). In contrast, mAICDmutAAA expression did not lead to alterated levels of cell surface APP-FL (p = 0.981). Adenylate cyclase inhibition also eliminated the increase of surface APP seen in γ-secretase e inhibited cells (Figure 3B;p = 1.000). Using flow cytometry (Figures 3C and 3D) and bio-tinylation approaches (Figure 3E), we confirmed preferential redistribution of mAICD-induced APP-FL to the cell surface in HEK293 cells, an effect that was similar to that observed in cells treated with the γ-secretase e inhibitor. Again, the inhibition of adenylate cyclase or the overexpression of mAICDmutAAA abrogated this effect. These findings demonstrate that the GαS-mediated signaling is necessary and sufficient to initiate a change of APP-FL localization associated with membrane accumulation of APP C-tail.

Figure 3. APP-Mediated Signaling Promotes cAMP-Dependent APP Residence at the Cell Surface.

(A) Confocal images of APP cell surface accumulation are visualized by APP-NT452 Aβ staining in stable APP-FL HEK293 cells co-transfected with EGFP and EV or PS1-D385A, or EGFP and mCtl, mAICD, or mAICDmutAAA treated with γ-secretase e inhibitor CompE and/or adenylate cyclase inhibitor MDL-12,330A.Scale bar, 20 μm.

(B) Quantification of cell surface APP is shown as relative value to untreated- (left) or mCtl- (right) conditions. ^** p < 0.01, ^*** p < 0.001, and ^**** p < 0.0001 compared with control untreated, mCtl, or EV condition, and # p < 0.05, and #### p < 0.0001 compared with MDL-12,330A treated condition, using one-way ANOVA Kruskal-Wallis test followed by Dunn’s post hoc analysis. mAICD orange color-coded significances compared with control conditions, and mAICDmutAAA green color-coded significance compared with mAICD. The total number of cell is shown in parentheses. Error bars indicate SEM.

(C) Flow cytometry analysis of APP surface expression was performed using APP-NT452 Aβ. Frequency distribution histograms of APP fluorescence intensity of stable APP-FL HEK293 cells are shown.

(D) Populations from higher intensity (>10⁴ gray scale intensity, shadow histogram area) were extracted.

(E) Biotinylation of cell surface APP was performed in HEK293 cells using the streptavidin pull-down assay revealed by the APP-CTM1 Ab.

Enhanced APP-Dependent Signaling through a Self-Regulatory Process Promotes Non-amyloidogenic Processing of APP

We determined that γ-secretase e inhibition or mAICD expression favored the accumulation of APP-CTF, an effect that was attenuated by a treatment with the adenylate cyclase inhibitor (Figure 4A). To further define the significance of adenylate cyclase cascade, we confirmed that the accumulation of APP-CTF derived from processing of endogenous APP-FL in primary neuronal cultures generated from mice lacking γ-secretase e activity (PS1-KO) is markedly decreased following adenylate cyclase inhibition (Figure 4B). Using co-immunoprecipitation strategy, we observed an increase in GαS protein interaction with APP-FL and APP-CTF consequent to γ-secretase e inhibition and enhanced APP residence at the cell surface (Figure 4C), therefore supporting the idea that APP-CTF-mediated increased redistribution of APP at the cell surface fosters an interaction with G-protein signaling components.

Figure 4. Enhanced APP-Dependent Signaling through a Self-Regulatory Process Promotes Non-amyloidogenic Processing of APP.

(A) Using APP-CTM1 Ab, APP-CTF accumulation is observed in stable APP-FL HEK293 cells expressing mAICD or treated with CompE (darker exposure below). This effect was reduced by adenylate cyclase inhibitor MDL-12,330A.

(B) Western blot analysis of cortical culture lysates from PS1-KO mice confirmed that MDL-12,330A treatment reduced the accumulation of endogenous APP-CTF, as reflected by a diminution in APP-CTF/APP-FL ratio.* p < 0.05 compared with control untreated condition, using a t test.

(C) Enhanced APP interaction with GαS is shown through co-immunoprecipitation (coIP) analysis in stable N2a cells expressing APP-FL that were treated with γ-secretase e inhibitor CompE.

(D) Lysates of stable HEK293 cells co-expressing APP-FL and mAICD were analyzed by high-resolution Tris-Tricine western blot in three independent pools of cell lines using CTM1 antibody (Ab) that recognized all APP-CTFs, and CTM1 Ab that recognized only tagged-mAICD. The darker set of exposure below identified several APP-CTF species. Overlapping readout of western blot scans is shown for CTM1 in red and CT11 in green. Secreted sAPP level in the conditioned media was analyzed using APP-NT452 Ab as shown at the bottom.

(E) Relative protein band densities from the darker CTM1 revealed blot are shown as plot lines in cells expressing mAICD (black) as compared to empty vector (EV, white).

(F–H) Stable HeLa APPSwe cells were transfected with EV or mAICD. The levels of secreted sAPPα (F), Aβ40 (G), and Aβ42 (H) were measured in the conditioned media. ^**** p < 0.0001 compared with control EV condition, using a t test. The total number of lysates is shown in parentheses. Error bars indicate SEM.

To establish the consequence of the accumulation of mem- brane-bound APP-CTF on APP processing and Ab production, we stably overexpressed mAICD and APP-FL in HEK293 cells (Figures 4D and 4E). As compared with cells expressing an empty vector, mAICD expression increased the levels of soluble APP (sAPP) in the media (Figure 4D). Whereas APP-FL levels remained unchanged, the levels of APP-CTFa, and to a minor extent the levels of APP-CTFb, were higher in the cell lysates (difference of 13.4 ± 1.8 versus 2.3 ± 0.3 AU in the band intensity, respectively; Figures 4D and 4E). We confirmed that the expression of mAICD elevated the levels of sAPPa and decreased the levels of secreted Ab40 and Ab42 in the conditioned medium of HeLa cells expressing FAD-linked APPSwe variant (Figures 4F–H), which is consistent with increased residence of APP at the cell surface and subsequent α-secretase cleavage.

DISCUSSION

Our investigation provides compelling evidence that the overex- pression of mAICD rescues memory impairments in an AD mouse model of amyloidosis. Through assessments of cognitive function, we demonstrate that GαS interaction site within AICD is required for hippocampus-dependent spatial memory. CREB signaling, a downstream effector of adenylate cyclase activation, strongly correlates with synaptic enhancement and memory consolidation (Abel and Nguyen, 2008; Alberini and Kandel, 2014; Pittenger et al., 2002; Rogerson et al., 2014). An elevation of cAMP levels through phosphodiesterase inhibition has been shown to enhance memory in Tg4510 Tau mice (P301L mutation) (Myeku et al., 2016), reverse the weakening of long-term plasticity mediated by Aβ (Vitolo et al., 2002), and restore synaptic structural alterations (Smith et al., 2009). These results, along with our observations, suggest that stimulating cAMP cascade could have translational potential.

In addition, our findings are consistent with the non-amyloidogenic processing of APP by APP-mediated signaling at the cell surface. APP resides more at the cell surface in conditions where membrane-tethered AICD accumulates through the overexpression of mAICD or reduction of γ-secretase e activity (APP-CTFa accumulation). mAICD overexpression also enhances sAPPa secretion, which is coherent with the increased residence of APP at the cell surface and subsequent α-secretase cleavage. The increased residence of APP at the cell surface is no longer observed if adenylate cyclase is inhibited, or if the interaction of GαS with mAICD is abolished. Adenylate cyclase inhibition, and ineffective coupling with GαS, also reduces the accumulation of APP-CTF. These observations strongly infer that APP possesses a self-regulatory component within its intracellular domain that influences APP retention at the cell surface and promotes processing through the α-secretase cleavage pathway. This auto-regulatory processing of APP is observed in multiple systems under conditions where exogenous or endogenous APP is expressed. Although the endogenous APP-CTFs are not as abundant because they are released from the membrane by γ-secretase e processing, this positive feedback would ensure that APP-dependent signaling is amplified rapidly and locally, thus creating a strong APP-dependent signaling signature.

It is well documented that GαS-coupling reduces Aβ production (reviewed in Fisher et al., 2016; Postina, 2012; Thathiah and De Strooper, 2011; and Zhao et al., 2016), and alleviates cognitive impairment in AD mouse models (Shen et al., 2016). As supported by our findings, APP could be an important molecular substrate for memory acquisition through its partnership with GαS and associated adenylate cyclase activation within the lipid-raft microdomains. We previously reported that GαS localization in lipid rafts is required for mAICD,GαS interaction (Deyts et al., 2012). Substantial portion of endogenous APP-CTF resides in lipid rafts, especially when γ-secretase e activity is reduced (Vetrivel et al., 2005). APP accumulation at the cell surface and/or APP-CTF accumulation coincide with increased interaction with GαS. This is consistent with the idea that an enhanced APP-CTF level is initiated through mechanisms taking place first in lipid raft compartments.

What could account for the increased residence of APP at the surface in our experimental system? Activation of the PKA-dependent pathway is known to stimulate APP trafficking to the surface, through the formation of APP-containing vesicles from the trans-Golgi network (Xu et al., 1996). Although we have not fully explored the mechanisms responsible for greater APP retention at the cell surface beyond mechanisms that involve the cAMP-dependent pathway, it is well recognized that downstream phosphorylation signaling events could interfere with protein trafficking at the surface. Like APP, type 1 transmembrane cell-adhesion proteins and growth-factor receptors have intricate intracellular trafficking patterns and signaling events that could influence their residence in subcellular compartments (reviewed in Schoenherr et al., 2018). Endosomal sequestration and lysosomal degradation of activated receptors are widely used mechanisms of downregulation of receptor signaling (reviewed in Dobrowolski and De Robertis, 2011). Receptor internalization at the cell surface is critical for transient GαS/cAMP signaling to occur (Pavlos and Friedman, 2017; Sposini et al., 2017; Tsvetanova and von Zastrow, 2014). Nevertheless, more sustained cAMP signaling could be achieved within the early endosomes, an organelle that also contains lipid rafts (Simons and Toomre, 2000). Interestingly, neither APP-C83 (CTFα) nor AICD sojourn in early endosomes, as depicted by the lack of colocalization with Rab5, whereas APP-C99 (CTFβ) does (Xu et al., 2016), which is also consistent with the preferential localization of APP-CTFα at the cell surface and AICD in the nucleus. An accumulation of APP at the cell surface, following inhibition of its endocytic transit, facilitates α-secretase dependent reduction of Aβ production, whereas the diminution of APP at the cell surface due to internalization will increase beta-site APP-cleaving enzyme 1 or betα-secretase 1 (BACE1)-dependent amyloidogenic processing and secretion of Aβ (Andrew et al., 2016; Haass et al., 2012; Jiang et al., 2014).

While our findings indicate that mAICD expression produced about 6 times higher accumulation of APP-CTFα than APP-CTFβ in cell lines, we cannot rule out the possibility of endosomal contributions of APP-CTFβ-mediated signaling. Our previous studies indicated that APP-C99 expression, but not expression of APP-C99mutAAA that lacks interaction with GαS, is associated with cAMP-response element binding protein (CREB) signaling (Deyts et al., 2016a), therefore suggesting that the remote accumulation of APP-CTFβ could contribute to a long-lasting effect on GαS/cAMP-mediated signaling. Indeed, APP processing has intricate connections with the endo/lysosomal network (Nixon, 2017; Wang et al., 2018). APP-CTFβ accumulation has been implicated in early endosome abnormalities (through the recruitment of APPL1 leading to activation of Rab5) and defects in the lysosomal and autophagic systems (Hung and Livesey, 2018; Kim et al., 2016b). The excessive accumulation of APP-CTFβ in endosomal compartments could lead to enhanced endocytic recycling and alterations in growth factors signaling (Kim et al., 2016b; Xu et al., 2016). Elevated levels of APP-CTFβ disrupt retrograde axonal trafficking of signaling endosomes that impact nerve growth factor and other signaling pathways (Chen et al., 2018). Moreover, the activation of Rab5 is a key mechanism involved in the internalization of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors at the synapse (Brown et al., 2005; Hausser and Schlett, 2017; Parkinson and Hanley, 2018). Cell surface delivery of AMPA receptors is also closely associated with PKA-mediated signaling (Buonarati et al., 2019). The critical relevance on AMPA receptor trafficking in AD-related synaptic plasticity and cognitive function has been proposed but not fully characterized (Guntupalli et al., 2016; Hausser and Schlett, 2017; Parkinson and Hanley, 2018). It remains to be determined if mAICD expression would influence endosome signaling associated with AD pathology or AMPA receptor trafficking in a manner that impacts cognitive function.

The APP residence at the cell surface could also promote APP’s cell adhesion property and foster synapse formation (reviewed in Sosa et al., 2017). APP-induced synaptogenesis is tightly regulated by α-secretase activity (Stahl et al., 2014; Wang et al., 2009). The mAICD-mediated increase in cell surface residence of APP results in preferential α-secretase cleavage and the release of sAPPa, which may serve as a putative ligand for multiple receptors including APP (reviewed in (Deyts et al., 2016b; Műller et al., 2017). If we assume that sAPPα primarily acts via APP as the target receptor, this could result in a positive feedback loop to activate intracellular coupling that would enhance the memory process. It has been reported that rAAV-mediated expression of sAPPα can rescue memory deficits and reduce Ab burden in a FAD mouse model (Fol et al., 2016). sAPPα could also interact with GABA_B receptor, reducing release probability from glutamatergic synapses, consequently weakening synaptic transmission, but enhancing high-frequency mediated short-term plasticity (Rice et al., 2019). These are few examples that pinpoint the intricate contribution of APP metabolites in governing synaptic physiology that might have consequences on cognitive function.

Phosphorylation cascade downstream of G-protein coupled receptor signaling is known to modulate α-secretase or BACE1 cleavages of APP in a manner that significantly affects cerebral Aβ burden (Fisher et al., 2016; Haass et al., 2012; Ly et al., 2013; Saftig and Lichtenthaler, 2015; Small and Gandy, 2006; Suh et al., 2013; Thathiah and De Strooper, 2011; Zhang et al., 2009). Intriguingly, the effect of APP-mediated signaling on Aβ burden in the cortex varies from that seen in the dentate gyrus, therefore highlighting the complexity of APP metabolism in various brain areas. Aβ production is tightly regulated by differences in α-secretase activity (Endres and Deller, 2017; Postina, 2012; Saftig and Lichtenthaler, 2015). APP and α-secretase brain distributions overlap differently at distinct developmental stages, with increasing colocalization during aging (Marcinkiewicz and Seidah, 2000). Of note, a variety of receptor-like γ-secretase e substrates could also influence APP-mediated signaling and, consequently, APP cell surface residence, through significant crosstalk with cAMP-dependent signaling. Indeed, transcriptional Notch activity represses the cAMP signaling cascade (Hallaq et al., 2015), while we have previously reported that accumulation of membrane-tethered DCC through γ-secretase e inhibition enhanced PKA-dependent signaling (Parent et al., 2005). Likewise, these conjectures might infer on the complexity of the cognitive behavior performance. A reduction of Ab burden in the hippocampus CA1/CA3 areas might be sufficient to prevent cognitive impairment under contextual fear conditioning, whereas Aβ reduction in the dentate gyrus through a PKA-dependent cascade might also be needed to preserve memory in tasks that require greater working memory acuity. It would be of interest to determine if behavior associated with specific brain regions where Aβ burden is not recovered would be rescued by mAICD expression.

Altogether our studies highlight the critical role of APP metabolites (other than Aβ) in influencing the hippocampal circuitry and memory proficiency in AD mouse models. Unlike the AICD-induced transcriptional functions, which are detrimental and exacerbate in AD pathology (reviewed in Bukhari et al., 2017; Pardossi-Piquard and Checler, 2012; and Pousinha et al., 2017), we demonstrate that sustained APP signaling favors brain plasticity and memory process. We found that retaining APP C-tail at the membrane preserves spatial working memory and intrinsic processing of APP in an amyloidogenic mouse model, an outcome that requires a direct interaction of APP C-tail with the heterotrimeric G-protein subunit GαS. Our results support the idea that APP possesses an autonomous regulation motif within its intracellular domain, which promotes non-amyloidogenic processing of APP at the cell surface.

STAR★METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the senior corresponding author, Angèle Parent (aparent@uchicago.edu).

EXPERIMENTAL MODEL DETAILS

Mouse models

Wild-type C57BL/6J (B6), C3H/HeJ (C3), SJL/J (SJL) and 5XFAD mice were purchased from The Jackson Laboratory (Bar Harbor, Maine). C57BL/6J x C3H/HeJ (B6C3) and C57BL/6J x SJL/J (B6SJL) F1 chimeras were obtained and maintained in our breeding colony. Transgenic 5XFAD mice expressing five familial AD mutations [3XAPP (Swedish K670N/M671L, Florida I716V, and London V717I), and 2XPS1 (M146L and L286V) mutations], with neuronal expression driven by the Thy-1 promoter, were obtained by intercrossing hemizygous 5XFAD with B6SJL mice (Oakley et al., 2006). PS1-KO mice were generated by intercrossing heterozygote PSEN^+/− mice, which are maintained in B6C3 background (Wong et al., 1997). Mouse handling procedures were performed in accordance with National Institutes of Health guidelines and reviewed by the University of Chicago Institutional Animal Care and Use Committee. The laboratories of Dr. Sangram Sisodia (University of Chicago) kindly provided PS1-KO mice for colony expansion. PCR-amplified regions using specific primers for each strain were performed to confirm the genetic identity of the mice. The levels of protein expression were periodically verified in the cohorts of mice generated by immunoblotting with the appropriate antibodies. As indicated in Supplement Data Figure S5, we validated that no-virus injected naive 5XFAD exhibited similar behavior performance and Ab accumulation as compared with mice injected with mCtl control virus. Accordingly, we decided to include both control groups in our analysis. Both sexes were used in all the experimentations.

Stable cell lines, transfections, and treatments

HEK293 cells stably expressing APP695 and mCtl, mAICD or mAICDmutAAA were generated by retroviral infection as described previously (Deyts et al., 2012). Briefly, retroviral supernatants collected 48 h after transfection of Phoenix cells were used to infect HEK293 APP695 cells in the presence of 4 μg/ml polybrene. Stably transduced cells were selected in the presence of 1 μg/ml puromycin and pooled for further analysis. HEK293 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Invitrogen) supplemented with 10% fetal bovine serum, 1% glutamine and 1% penicillin-streptomycin (GIBCO). Transient transfections of HEK293 cells were performed with LipoD293™ (SignaGen Laboratories, Rockville, MD) according to the manufacturer’s protocol. Plated cells were treated with Compound E (10 nM) and/or MDL-12,330A (10 nM) for 24 h before being lysed for western blotting or fixed for fluorescence imaging (Deyts et al., 2012; Seiffert et al., 2000).

Neuronal cultures

Primary mouse cortical neuron cultures were prepared from embryonic E16 Wt or PS1-KO mice as previously described (Parent et al., 2005) and maintained at 37°C in Minimal Essential Medium (Invitrogen) supplemented with 1% glutamine, 5% horse serum, 0.5% D-glucose, 0.15% HCO₃, and nutrients, in a humidified 10% CO₂ incubator. Neurons were cultured in 0.1% polyethyleneimine-coated 18 mm glass coverslip for immunostaining or in poly-L-Lysine coated 60 mm dishes for western blot. PS1-KO neurons (8DIV) were treated with MDL-12,330A (10 nM, 24h) before being lysed for western blotting experiments. For neurite outgrowth experiments (Figures S1C–S1E), neurons generated from Wt mice were transiently transfected at 7 DIV using Lipofectamine 2000 (Invitrogen) in Neurobasal medium (Invitrogen). After 3 h, the transfection medium was replaced by 50% original medium and 50% supplemented Minimal Essential Medium without serum. Neurite outgrowth was performed in 8DIV culture as previously described (Deyts et al., 2012, 2016a).

METHOD DETAILS

rAAV constructs and viral preparation

All cDNAs were generated by PCR and cloned into pTR-UF22 vector, a chimeric rAAV serotype 2/8 hybrid recombinant gene delivery system consisting of rAAV2 inverted terminal repeats (ITR) and rAAV8 capsid genes (rAAV2/8) (Chakrabarty et al., 2013; Kűgler et al.,2003; Passini et al., 2003). cDNAs are expressed bicistronically with enhanced green fluorescent protein (EGFP) through IRES sequence and under the control of the cytomegalovirus enhancer/chicken b-actin promoter (CAG), and a woodchuck hepatitis virus post-transcriptional regulatory element (WPRE) (Klein et al., 2002; Kűgler et al., 2003; Levites et al., 2006). cDNAs encoding mCtl, mAICD, and mAICDmutAAA were cloned between SpeI and EcorV sites of pTR-UF22 to generate mCtl-IRES-EGFP (mCtl), mAICD-IRES-EGFP (mAICD), and mAICDmutAAA-IRES-EGFP (mAICDmutAAA), respectively (Figure S1A). To reduce mAICD expression, mAICD was introduced after IRES sequence (Figure S1B). The brighter copepod GFP (copGFP) fluorescent reporter (Löw et al., 2010; Veeraraghavalu and Sisodia, 2013) and mAICD-tagged flag were separated by a ribosome skip Thosea asigna 2A short peptide site element (Kim et al., 2011; Tang et al., 2009). copGFP-2A-mAICD was cloned between Not I sites to generate IRES - copGFP-2A-mAICD (IRES-mAICD). The short 2A peptide sequence [GAG GGC AGA GGA AGT CTG CTA ACA TGC GGT GAC GTC GAG GAG AAT CCT GGA CCT] has been added by overlap PCR amplification. All PCR-amplified regions were verified by sequencing. All bacterial transformations were performed in SURE-2 competent cells (Stratagene). Large-scale plasmid preparations of the viral constructs were performed (Maxiprep kit, QIAGEN) and packaged into rAAV8 capsid as described (Chakrabarty et al., 2013). The genomic titer of each virus (between 1–10×10¹³ genome copy/ml) was determined by quantitative PCR.

rAAV neonatal injection

Injection of rAAV2/8 into the cerebral ventricles of newborn mouse pups has been shown to induce widespread neuronal transduction and life-long expression of transgenes (Chakrabarty et al., 2013; Kim et al., 2013; Levites et al., 2006). rAAV injections were performed blind to genotype, and transgenic status determined by genotyping at the time of weaning. The following procedure was adapted as previously described (Chakrabarty et al., 2013; Kim et al., 2013; Kim et al., 2016a; Levites et al., 2006). Briefly, cryoanesthetized neonates (8–12 hours after birth) were injected using 100μl Hamilton syringe (Hamilton, 7653–01) with a 32 gauge needle (Hamilton, 7803–04, RN 6PK PT4) at an angle of 45°to a depth of 1.5 mm. The injection site was located two-fifths of the distance along a line defined between each eye and the lambda intersection of the skull. 2 μl of virus (2×10¹⁰ genome copy) was slowly injected into each lateral ventricle, followed by slow retraction of the needle. After injection pups were allowed to completely recover on a warming pad and then returned to the home cage with the mother.

Behavior tasks

Behavioral experiments were performed blind to genotype and conducted by three observers between 8 am and 5 pm, in a softly illuminated room. The behavioral test battery consisted of six memory tests performed in the following order: elevated-plus maze, Y-maze spontaneous alternation, Y-maze light/dark exploration test, open-field, novel object recognition, and lastly contextual fear conditioning. The mice were placed in the testing room at least 24h beforehand. Except for the elevated-plus and the fear conditioning tasks, the tests were performed in a Biological Safety cabinet, under dimmed light and black noise ventilation conditions to reduce anxiety, noise and smell cue distraction behaviors. At the end of the trial, the mice were returned to their home cage. Mazes, chambers, and objects were wiped with Labsan C-dox to prevent odor cues after each handling test.

Y-Maze Spontaneous Alternation

This maze tested the willingness of mice to explore new environments (Hodges, 1996). It implies that a mouse can remember which arm they entered and will alternate to the opposite arm. Spontaneous alternation tests were conducted using a symmetrical Y-maze composed of a white bottom and wall Plexiglas (35L x 7W x 15H cm). At the start of each trial, the mouse was introduced in the center of the Y-maze and the arm entries were recorded. The trial ended when either 36 arms were visited or the trial had lasted no longer than 10min. Re-entries into the same arm were rated as separate entries (Holcomb et al., 1999). Mice were excluded from the analysis if they performed less than 8 arm entries during the 10min trial. Percentage of spontaneous alternation was defined as the number of consecutive entries into all three arms divided by the number of possible alternations (total arm entries minus 2) (Hsiao et al., 1995). Locomotor activity was defined as the number of transitions between arms by the time of the experimentation.

Light/Dark Transition

Anxiolytic behavior is known to interfere to some extent with cognitive function (Curzon et al., 2009b; Walf and Frye, 2007). In order to address this possibility, Light/Dark exploration test was conducted to assess anxiolytic behavior associated with the propensity of the mouse to explore dark compartment (Bourin and Hascoët, 2003). Using the Y-maze, the test was conducted 24 h after the spontaneous alternation task. The mice were exposed to a novel environment with one protected (dark compartment) and two unprotected (light compartment) arms. At the start of the trial, the mouse was placed in the center of the Y-maze. Mice were allowed to freely explore the Y-maze for 5 min while the number of entries and time spent in the dark compartment are recorded. Percentage of time spent in the dark was quantified. Mice that explored the arms with less than 4 entries were excluded.

Open-Field Activity

The open-field test was used to assess anxiety-like behavior associated with locomotor activity and exploration habits in a freely moving environment. Assessment took place in a 44 l griddled plastic storage box with light-obstructed white walls (50L x 40W x 28H cm). A direct light was placed 90 cm above the center. The mouse was positioned in the center of the open-field chamber and allowed to freely explore the chamber for a 5 min period. For analysis, the chamber was divided into a central brighter zone (26×16 cm²) and a peripheral darker zone (12 cm wide). The number of inside exploration was reported as a percentage of the total number of exploration.

Novel Object Recognition

The novel object recognition trials were conducted to determine memory acquisition based on the tendency for mice to explore unfamiliar objects (Leger et al., 2013). The task was performed in the open-field plastic chamber described above. A familiar blue-painted wooden triangle and an unfamiliar yellow-painted wooden cube were used as the old and new objects, respectively (Figure 2A). On the first day of the test, a habituation trial was conducted for each animal, which consisted of a 5 min exploration period with no objects within the open field arena. Further exploration time was allowed following the introduction of two identical objects (blue triangles) into the box, which are placed equidistant from both corners, 12 cm apart from the wall. The mouse was allowed to freely explore the objects for 5 min. On the second day of the test (24 h later), the familiar object (blue triangle) and the unfamiliar object (yellow cube) were introduced and the object interaction was recorded for 5 or 10 min. Positive exploration was defined as touching the object with the nose or directing the nose to the object within a 12 cm grid across the object. The time of exploration was manually recorded. Sitting or leaning on the object or grooming within the grid were not considered as exploration time (Clark and Martin, 2005; Ennaceur and Delacour, 1988). The total exploration time for both objects was recorded. Results were reported as the time spent (T) per exploration number (N) of the object. Mice presenting less than five total object interactions (or less than two interactions per object) were excluded from the analysis. The discrimination index was calculated using the percentage of the ratio between the time spent per exploration number of the novel object by the time spent per exploration number of both objects (Tnew/Nnew – Told/Nold / Tnew/Nnew + Told/Nold) (Bevins and Besheer, 2006). Exploratory activity was defined as the total number of object exploration by the time of the experimentation. Locomotor activity was calculated by dividing the distance recorded as one arbitrary unit for each transition trajectory by the time of the experimentation.

Contextual Fear Conditioning

To assess learning and memory behaviors under a stress-induced fear situation, we used a fear-conditioning chamber that could deliver a foot shock from a bottom grid floor (Coulbourn Instruments, Whitehall, PA) (Andrew et al., 2017; Curzon et al., 2009a). On the first day, the mouse was allowed to habituate to the chamber for a period of 2 min prior to receiving a first shock (1 mA, 2 s). A second shock of the same intensity and duration was delivered 2 min later. Three minutes later the training session was terminated and the mouse was returned to the home cage. On the next day (24 h later), the mouse was placed in the same conditioning chamber for 5 min without any shock to test contextual memory retrieval. Immobilization associated with fear behavior was defined as the complete lack of motion for a minimum of 1 s for the B6SJL and 0.5 s for the B6C3 background mice. Freezing time was recorded using FreezeFrame 4 Software (ActiMetrics, Wilmette, IL). Cumulative freezing time was quantified using a 30 s epoch (Andrew et al., 2017; Curzon et al., 2009a). Quantitative analysis on the first-day recording allowed us to assess the learning acquisition rate associated with the conditioning session. A similar analysis was applied on the second day to assess the contextual memory acquisition. In order to confirm the acquisition of freezing behavior associated with contextual fear, differences in freezing time between the first and the second day were also calculated.

Brain collection

Mice were anesthetized with isoflurane and then transcardially perfused with PBS. Brains were extracted and dissected at the midline. The right hemi-brains were immediately snap frozen with dry ice, and stored at 80°C until use for ELISA detection as described below. The left hemi-brains were post-fixed 24–48 h in 4% paraformaldehyde/4% sucrose at pH 7.4 dissolved in PBS, equilibrated in PBS/20% sucrose for 24 h at 4°C, embedded into Tissue-Tek®O.C.T. compound (Sakura, USA), snap frozen, and stored at −80°C until further processing. Coronal sections (40 μm) were collected and stored in a cryoprotective solution containing 30% ethylene glycol, 30% glycerol, and 0.1 M phosphate buffer at −20°C until processing for immunohistochemistry.

Aβ immunostaining

To quantify Aβ deposition, whole brain sections were stained with the 3D6 antibody using a free-floating procedure as previously described (Andrew et al., 2017; Deyts et al., 2016a). Briefly, coronal brain sections were washed in Tris-buffered saline, then blocked for 2 h in TBS containing 5% horse serum and 0.25% Triton X-100 followed by incubation with the 3D6 primary antibody (1:10,000) for 72 h at 4°C. After multiple washes in TBS, sections were incubated for 2 h at room temperature with Alexa555 secondary antibody. Sections were then washed and mounted using Vectashield mounting medium (Vector Laboratories, Burlingame, CA). Images were acquired with a 20X LWD Zeiss objective (0.95 NA) using a CRi Pannoramic Scan Whole Slide Scanner (Perkin Elmer/3DHistech) coupled to a Zeiss AxioCam MRm high sensitivity camera. We used extended depth of field to project 16 mm of brain section volume into a single slice, which was then converted into 16-bit files and analyzed in MetaMorph software (Molecular Devices). Same threshold intensity values were applied to each series to eliminate background fluorescence. The number of deposits and the deposit size were measured using Integrated Morphology Analysis plug-in feature of MetaMorph. Aβ burden was analyzed by reporting the threshold stained area divided by the total area under study. The values were averaged from 5–7 sections per mouse.

Aβ ELISA measurement

Aβ levels were measured in 6-month-old 5XFAD mouse brains as previously described (Chakrabarty et al., 2015). Mouse hemi-brains were homogenized in RIPA extraction buffer containing 50 mM Tris-HCl, 150 mM NaCl, 1% Triton X-100, 0.5% deoxycholate, 0.1% SDS, and 1X EDTA-free protease inhibitor cocktail (cOmplete EDTA-free, Roche) and a phosphatase inhibitor mixture (PhosSTOP EasyPack, Roche). After centrifugation at 100,000 g for 1 h at 4°C, the supernatant was retained as the Triton-soluble fraction (soluble Aβ). The pellet was then sonicated in the presence of 70% formic acid and centrifuged at 100,000 g for 1 h at 4°C. The resulting formic acid-extracted supernatant representing the insoluble fraction was neutralized with 1 M Tris-base, 0.5 M NaH₂PO₄ buffer. The levels of Ab40 and Aβ42 were quantified using end-specific Aβ antibodies by sandwich ELISA as described (Chakrabarty et al., 2015). Aβ40 and Aβ42 concentration values were normalized to total brain protein concentrations (pmol per gram of total protein). Aβ analysis was also performed from stable APPSwe HeLa cells. Conditioned media were collected 48 h after plating the cells as described previously (Vetrivel et al., 2011). The levels of secreted Aβ40 and Aβ42 were quantified by ELISA using mAb B608 for capture and mAb B436 for detection (Vetrivel et al., 2009). All samples were processed in duplicate.

Protein analysis

Western blotting

Cells were lysed in buffer containing 150 mM NaCl, 50 mM Tris-HCl, pH 7.4, 0.5% NP-40, 0.5% sodium deoxycholate, 5 mM EDTA, 0.25% SDS, 0.25 mM phenylmethylsulfonyl fluoride and protease inhibitor mixture (1:200, Sigma-Aldrich), and briefly sonicated on ice. Equal amounts of proteins were resolved on SDS-PAGE gels and transferred to a PVDF Immobilon-FL membrane (Millipore/Fisher Scientific). Endogenous or overexpressed APP-FL and APP-CTF were detected by immunoblotting with the CTM1 antibody. APP-CTFα and APP-CTFβ were separated on 16.5% Tris-Tricine gels. The C terminus of mAICD was detected either with CT11 or FLAG antibodies. Soluble APP fragments (sAPP) were fractionated on 4%–20% SDS gels and revealed using APP-NT452 polyclonal antibody on western blots made from conditioned media lysates collected 12–14 h before lysates (Vetrivel et al., 2009). Detection of GAPDH protein was used as a loading control. To increase accuracy and sensitivity of protein detection, western blots were quantified by fluorescence using Odyssey infrared imaging system (LI-COR Biosciences, Lincoln, NE). A fixed size area was selected surrounding the band of interest and quantified within the same gel (Deyts et al., 2016a).

Co-immunoprecipitation

Co-immunoprecipitation (coIP) analysis of APP-CTF interaction with GαS was evaluated in stable N2a cells expressing APP-FL following transient transfection of HA-tagged GαS-wt with or without Compound E treatment (10 nM, 24h) as previously described (Deyts et al., 2012). Briefly, cells were lysed in ice-cold lysis buffer (25 mM HEPES, pH 7.4; 5 mM MgCl₂, 1% NP-40, 125 mM potassium acetate, 10% glycerol, and protease inhibitor mixture). Lysates were clarified by centrifugation at 13,000 rpm for 20 min at 4°C, and aliquots of the supernatants were incubated overnight with mAb HA at 4°C. The immune complex was captured on protein G-agarose beads (Thermo Fisher Scientific). Bound proteins were eluted in Laemmli buffer and analyzed by immunoblotting with CTM1 to detect APP-CTF.

Cell surface protein biotinylation

Cell surface proteins from transfected HEK293 cells were biotinylated with 0.5mg/ml NHS-SS-biotin (Thermo Fisher Scientific) in PBS, for 45 min on ice as previously described (Govind et al., 2012; Thinakaran et al., 1996). Cells were washed with Tris-buffered saline pH 7.4 to quench unreacted biotin. Cells were then lysed for 1h on ice in buffer containing 150 mM Tris-HCl, pH 7.4, 0.5% NP-40, 0.5% sodium deoxycholate, 5 mM EDTA, 0.25% SDS, 0.25 mM phenylmethylsulfonyl fluoride and protease inhibitor mixture (1:200, Sigma-Aldrich). The resulting cell lysates were subsequently incubated overnight at 4°C with streptavidin-agarose beads to capture biotinylated APP. Immunoprecipitated proteins were dissociated by boiling in Laemmli buffer containing 1% SDS and resolved on SDS-PAGE gels along with aliquots of the total input lysates. Endogenous or overexpressed APP-FL and APP-CTF were detected by immunoblotting with the CTM1 antibody.

Cell surface protein immunofluorescence

To label cell surface APP, HEK293 cells were fixed with 4% paraformaldehyde/4% sucrose in phosphate buffer saline (PBS) for 30 min at 4°C. Cells were incubated first for 2 h at room temperature with APP-NT452 primary followed by Alexa555 secondary antibodies to label cell surface-bound APP. Next, cells were permeabilized on ice with 0.2% Triton X-100 solution for 8 min before incubation with CTM1 for 2 h at room temperature followed by incubation with Alexa647 secondary antibodies to detect total APP.

Flow cytometry analysis of cell surface APP

HEK293 cells were transfected with rAAV-mCtl or rAAV-mAICD constructs using LipoD293™. 24h after transfection, some dishes were treated with Compound E (10 nM) and/or MDL-12,330A (10 nM). Cells were collected 24 h later using magnesium and calcium-free cold PBS. Cells (about 0.5×10⁶) were suspended in 1 mL cold PBS with polyclonal APP-NT452 (1:1000) for 30 min at 4°C and subsequently labeled with Alexa555 antibody for an additional 30 min at 4°C. Between each antibody applications, cells were washed twice with cold PBS containing 2% fetal bovine serum, centrifuged for 5 min at 800 rpm, and re-suspended in PBS. At the end of the experimentation, cells were fixed with 4% paraformaldehyde/4% sucrose in PBS for 30 min, washed, and incubated at room temperature for 30 min with 0.5 mg/ml Hoechst diluted in PBS to label nuclei. The final cell product was re-suspended in 25–50ml PBS for flow cytometry analysis as previously described (Wang et al., 2015). Surface-bound APP-Alexa555 and Hoechst fluorescence intensities were acquired from 5,000–10,000 cells using a FACSCanto system and analyzed using FlowJo software (BD Biosciences). Frequency distribution histograms were generated using the normalized mode condition (normalized to higher cell frequency for a given intensity).

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analyses were performed using GraphPad Prism software (La Jolla, CA). Most of the graphs originated from data are presented as mean ± SEM using two-tailed analysis or otherwise as specified. Numbers of animals used for each condition are indicated in parentheses within each bar graph or within the legend. The significance levels are indicated by asterisks * p < 0.05, ^** p < 0.01, ^*** p < 0.001, and ^**** p < 0.001 if compared to control groups: NTg, Wt, EV, or mCtl; # p < 0.05, ## p < 0.01, ### p < 0.001, and #### p < 0.0001 if compared to other conditions within the same study as specified. Applied color-coded was used to identify significance within groups. Mann-Whitney t test, one-way ANOVA Kruskal-Wallis test followed by Dunn’s post hoc analysis, or two-way ANOVA followed by Sidak’s multiple time comparison test were used as indicated in the figure legends. π p < 0.05, ππ p < 0.01 and πππ p < 0.001 to compare values within same animals using Wilcoxon paired t test. Z-scores were calculated using the following equation: (X-μ) / (σ) where X is the observed value, μ is the mean of the control, and σ is the standard deviation of the control. Where relevant, we provided exact p values for both significant and non-significant conditions. We provided F values and degrees of freedom for two-way ANOVA analysis, as indicated in the text.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit polyclonal anti-APP-CTM1	Dr. Gopal Thinakaran (Vetrivel et al., 2009)	N/A
Rabbit polyclonal anti-APP-CT11	Dr. Gopal Thinakaran (Vetrivel et al., 2009)	N/A
Rabbit polyclonal anti-APP-NT452	Dr. Gopal Thinakaran (Vetrivel et al., 2009)	N/A
Mouse monoclonal anti-N-terminal Aβ antibody (clone 3D6)	Dr. Dale Schenk (Elan Corporation PLC2) (Andrew et al. 2017)	N/A
Mouse monoclonal anti-flag (clone M2)	Sigma Aldrich	Cat#F3165; RRID:AB_259529
Rabbit polyclonal MAP2	Sigma-Aldrich	Cat#M3696; RRID:AB_1840999
Mouse monoclonal GAPDH	Abeam	Cat#ab8245; RRID:AB_2107448
Mouse monoclonal anti-HA (clone 6E2)	Cell Signaling	Cat#2367; RRID:AB_10691311
Goat anti-mouse Alexa 488	Invitrogen	Cat#A-11029; RRID:AB_2534088
Goat anti-mouse Alexa 555	Invitrogen	Cat#A-21424; RRID:AB_141780
Goat anti-rabbit Alexa 555	Invitrogen	Cat#A-21429; RRID:AB_141761
IRDye 680LT donkey anti-mouse	LI-COR Biosciences	Cat# 926–68022; RRID:AB_10715072
IRDye 680 donkey anti-rabbit	LI-COR Biosciences	Cat# 926–32223; RRID:AB_621845
IRDye 800CW donkey anti-mouse	LI-COR Biosciences	Cat# 926–32212; RRID:AB_621847
IRDye 800CW goat anti-rabbit	LI-COR Biosciences	Cat#827–08365; RRID:AB_01796098
Aβ end-specific sandwich ELISA using end-specific antibodies (2.1.3: Aβ 42; 13.1.1: Aβ 40 for capture) and Ab5-HRP (pan Aβ) for detection	Dr. Todd Golde (Chakrabarty et al., 2015)	N/A
Mouse monoclonal Aβ anti-B608 for capture and anti-B436 for detection of sandwich ELISA	Dr. Steven Wagner (Vetrivel et al., 2009)	N/A
Bacterial and Virus Strains
rAAV2/8 caps id packaging	Dr. Todd Golde (Chakrabarty et al., 2013)	N/A
Chemicals, Peptides, and Recombinant Proteins
γ-secretase inhibitor Compound E	Dr. Todd Golde Mayo Clinic core facility (Seiffert et al., 2000)	N/A
Cis-N-(2-phenylcyclopentyl) azacyclotridec-1-en-2-amine (MDL-12,330A)	Enzo Life Science	Cat# BML-CN 145–0005
cOmplete EDTA-free - EDTA-free protease inhibitor cocktail	Roche/Sigma-Aldrich	Cat# 04693132001
PhosSTOP EasyPack - phosphatase inhibitor mixture	Roche/Sigma-Aldrich	Cat# 4906845001
Vectashield mounting medium	Vector Laboratories	Cat#H-1000
Lipofectamine 2000	Invitrogen/ThermoFisher	Cat#11668019
LipoD293™	SignaGen Laboratories	Cat# SL100668
Streptavidin-agarose beads	Pierce/ThermoFisher	Cat#20347
G-agarose beads	Pierce/ThermoFisher	Cat#20398
Premium sulfo NHS-SS-biotin	Pierce/ThermoFisher	Cat#PG82077
Experimental Models: Cell Lines
HeLa cells expressing APP-FL containing the Swedish mutation (APPSwe)	Dr. Gang Yu (UT Southwestern) (Leem et al., 2002)	N/A
Human embryonic kidney (HEK293)	ATCC	Cat# CRL-1573
Stable HEK293 cells overexpressing APP695 and empty vector	This paper	N/A
Stable HEK293 cells overexpressing APP695 and mAICD	This paper	N/A
SURE-2 competent cells	Stratagene/Agilent	Cat#200152
Experimental Models: Organisms/Strains
Mouse: C57BL/6J (B6)	The Jackson Laboratory	JAX: 000664
Mouse: C3H/HeJ (C3)	The Jackson Laboratory	JAX: 000659
Mouse: SJL/J (SJL)	The Jackson Laboratory	JAX: 000686
Mouse: transgenic 5XFAD	The Jackson Laboratory (Oakley et al., 2006)	MMRRC: 34840-JAX
Mouse: knockout PSEN1 (PS1-KO)	Dr. Sangram Sisodia (University of Chicago) (Wong et al., 1997)	N/A
Recombinant DNA
Plasmids: EV, APP-FL, mCtl, mAICD, mAICDmutAAA, GaS-HA, PS1-D385A	Deyts et al., 2012	N/A
pTR-UF22-CAG-EV-IRES-EGFP	Dr. Todd Golde (Chakrabarty et al., 2013; Küugler et al., 2003; Passini et al., 2003)	N/A
pTR-UF22-CAG-mCtl-IRES-EGFP	This paper	N/A
pTR-UF22-CAG-mAICD-IRES-EGFP	This paper	N/A
pTR-UF22-CAG-mAICDmutAAA-IRES-EGFP	This paper	N/A
pTR-UF22-CAG-IRES-copGFP-2A-mAICD-flag	This paper	N/A
Software and Algorithms
MetaMorph software	Molecular Devices	N/A
FreezeFrame 4 software	ActiMetrics	N/A
FlowJo™ software	BD Biosciences	N/A
Prism software	GraphPad	N/A

Highlights.

• Membrane retention of APP C-tail preserves spatial memory in Alzheimer’s disease model

• APP interaction with Gα_S attenuates amyloidogenic APP processing

• A self-regulatory component of APP promotes APP processing at the cell surface

• Sustained APP signaling modifies the course of Alzheimer pathology