Ca2+ Influx through Store-operated Ca2+ Channels Reduces Alzheimer Disease β-Amyloid Peptide Secretion*

Background: Dysregulation of Ca2+ homeostasis has been implicated in Alzheimer disease pathogenesis, but the effects of Ca2+ on amyloid precursor protein processing are not well understood. Results: Constitutive activation of the store-operated calcium entry pathway reduces β-amyloid generation. Conclusion: Elevation of Ca2+ influx affects amyloid precursor protein processing. Significance: Alteration of Ca2+ homeostasis in Alzheimer disease may influence pathogenesis directly through modulation of β-amyloid production. Alzheimer disease (AD), the leading cause of dementia, is characterized by the accumulation of β-amyloid peptides (Aβ) in senile plaques in the brains of affected patients. Many cellular mechanisms are thought to play important roles in the development and progression of AD. Several lines of evidence point to the dysregulation of Ca2+ homeostasis as underlying aspects of AD pathogenesis. Moreover, direct roles in the regulation of Ca2+ homeostasis have been demonstrated for proteins encoded by familial AD-linked genes such as PSEN1, PSEN2, and APP, as well as Aβ peptides. Whereas these studies support the hypothesis that disruption of Ca2+ homeostasis contributes to AD, it is difficult to disentangle the effects of familial AD-linked genes on Aβ production from their effects on Ca2+ homeostasis. Here, we developed a system in which cellular Ca2+ homeostasis could be directly manipulated to study the effects on amyloid precursor protein metabolism and Aβ production. We overexpressed stromal interaction molecule 1 (STIM1) and Orai1, the components of the store-operated Ca2+ entry pathway, to generate cells with constitutive and store depletion-induced Ca2+ entry. We found striking effects of Ca2+ entry induced by overexpression of the constitutively active STIM1D76A mutant on amyloid precursor protein metabolism. Specifically, constitutive activation of Ca2+ entry by expression of STIM1D76A significantly reduced Aβ secretion. Our results suggest that disruptions in Ca2+ homeostasis may influence AD pathogenesis directly through the modulation of Aβ production.

Alzheimer disease (AD) 3 is a progressive neurodegenerative disorder, the number one cause of dementia in the elderly and the sixth leading cause of death in the United States (1). Pathologically, AD is characterized by the accumulation of 38 -43amino acid-long amyloid ␤ peptides (A␤) in senile plaques and the presence of tangles composed of hyperphosphorylated tau in the brains of affected individuals (2). Clinically, Ͼ90% of cases of AD are classified as non-familial or sporadic disease, with aging as the main risk factor. However, causative mutations leading to early-onset familial AD have been identified in three genes: APP, PSEN1, and PSEN2 (3). These mutations all appear to lead to AD by increasing overall levels of A␤ or by promoting production of A␤ peptides (A␤ 42 ) that are more prone to oligomerization and deposition. As such, the "amyloid cascade" hypothesis of AD was developed, and A␤ is still considered to be one of, if not the most, important factors in the pathogenesis of AD (4).
A␤ peptides are generated through sequential proteolytic cleavage of amyloid precursor protein (APP), which is a type I transmembrane protein (5). In the A␤-producing amyloidogenic pathway, APP is first cleaved by the aspartyl protease ␤-secretase (␤-site APP-cleaving enzyme (BACE1)) within its extracellular domain, liberating the soluble ectodomain sAPP␤ and generating a membrane-tethered ␤-C-terminal fragment (␤-CTF) (6). The ␤-CTF is then cleaved within its transmembrane domain by ␥-secretase, releasing A␤ and the cytoplasmic APP intracellular domain (AICD). ␥-Secretase is an unusual aspartyl protease made up of four transmembrane subunits: nicastrin, APH-1, PEN-2, and presenilin (PS) 1 or PS2 (7). The ␤-CTF is cleaved serially by ␥-secretase at multiple sites producing A␤ fragments of varying size, with A␤ 40 and A␤ 42 being the most abundant (5). Alternatively, APP can be processed in a non-amyloidogenic manner. Cleavage by ␣-secretase generates sAPP␣ and an ␣-CTF, which is further cleaved by ␥-secretase to produce the small peptide p3 and AICD (5). Because ␣-secretase processing precludes formation of A␤, the non-amyloidogenic processing of APP is thought to be potentially beneficial.
Multiple lines of evidence suggest that Ca 2ϩ homeostasis is deregulated in AD (8,9). For example, alterations in the levels of Ca 2ϩ channels, exchangers, and Ca 2ϩ -dependent enzymes have been demonstrated in the brains of affected patients (10 -12). Several studies have also found altered Ca 2ϩ homeostasis in fibroblasts isolated from patients with AD compared with controls (13)(14)(15). In fact, both PS1 and APP have been shown to mediate changes in Ca 2ϩ homeostasis. Recent studies have proposed a variety of functions for PS1 in Ca 2ϩ homeostasis, including modulation of store-operated Ca 2ϩ entry (SOCE), formation of ER Ca 2ϩ leak channels, and regulation of sarcoendoplasmic reticulum calcium transport ATPase, inositol trisphosphate receptors, and ryanodine receptors (16 -21). APP appears to have numerous effects on Ca 2ϩ homeostasis as well. Expression of full length APP, for instance, affects spontaneous Ca 2ϩ oscillations in cultured neurons (22,23). Effects on intracellular Ca 2ϩ stores, on the other hand, have been attributed to the APP cleavage product AICD (24,25). Perhaps most intriguingly of all, however, are the effects mediated directly by the interaction of A␤ with Ca 2ϩ -permeable channels. These include functional alterations of plasma membrane ion channels such as voltage-gated Ca 2ϩ channels, nicotinic acetylcholine channels, and glutamate, serotonin, and dopamine receptors, alterations of intracellular Ca 2ϩ channels such as ryanodine receptors and inositol trisphosphate receptors, and even the direct formation of Ca 2ϩ -permeable ion channels (26).
Although a role for disruptions in Ca 2ϩ homeostasis in the pathogenesis of AD has been studied in the past using pharmacological manipulations, the effects of these changes on the processing of APP to generate A␤ are not well understood due to conflicting results. Therefore, we devised a genetic approach to alter [Ca 2ϩ ] i levels that would allow us to more precisely investigate the effects of Ca 2ϩ influx on A␤ generation. Recently, stromal interaction molecule 1 (STIM1) and Orai1 have been identified as the molecular components of the SOCE machinery. STIM1 is a type I transmembrane protein that resides within the ER membrane as dimers under basal conditions. Upon ER Ca 2ϩ store depletion, STIM1 rapidly oligomerizes and translocates within the ER membrane to plasma-membrane adjacent regions where it binds, clusters, and activates the store-operated Ca 2ϩ channel Orai1 (27,28). Coexpression of these components is sufficient to reconstitute and potentiate SOCE (29 -31). Additionally, expression of a well characterized mutant of the luminal EF-hand domain of STIM1, STIM1 D76A , leads to constitutive activation of Ca 2ϩ influx even under storereplete conditions (32). Therefore, we utilized these components of the SOCE pathway to specifically modulate Ca 2ϩ influx and isolate the effects of these manipulations on A␤ generation in the absence of confounding mutations in PSEN genes or the use of non-physiologic pharmacologic agents. In particular, we found that increased Ca 2ϩ influx resulting from overexpression of the constitutively active STIM1 D76A mutant led to dramatic reductions in the secretion of A␤. Our results indicate that Ca 2ϩ influx pathways have multiple effects on APP maturation and processing and provide insights into the importance of Ca 2ϩ homeostasis to neuronal pathophysiology and AD.
Analysis of APP processing by metabolic labeling with [ 35 S]Met/Cys was performed as previously described (40). CTM1 antiserum was used to immunoprecipitate full-length APP and APP CTFs from cell lysates, and mAb 4G8 was used to immunoprecipitate A␤ and p3 from conditioned media (40).
For total internal reflection microscopy (TIRF), HEK-APP cells plated on poly-L-lysine-coated 35-mm glass-bottom dishes were transfected with YFP-STIM1 or YFP-STIM1 D76A . After ϳ24 h cells were placed in warm HBSS before mounting on the microscope stage maintained at 37°C using a customdesigned environment chamber. TIRF images were acquired every 15 s using a 100ϫ TIRF objective (1.45 NA) and an EMCCD camera (Photometrics Cascade II). After acquiring base-line TIRF images, cells were briefly washed in Ca 2ϩ -free (0 Ca 2ϩ ) HBSS, and ER Ca 2ϩ stores were depleted by the addition of 1 M thapsigargin (Tg). Images were analyzed using Metamorph imaging software.
Ca 2ϩ Imaging-Intracellular Ca 2ϩ concentration ([Ca 2ϩ ] i ) was measured in cells loaded with 5 M Fura-2 AM using a Nikon Diaphot inverted epifluorescence microscope and InCyt IM2 TM fluorescence imaging system as previously described (38,41). A three-part protocol was utilized to measure Ca 2ϩ entry under basal conditions (Ca 2ϩ stores full) and Ca 2ϩ entry after Ca 2ϩ store depletion. Specifically, after incubation for ϳ2-3 min in 0 Ca 2ϩ HBSS, cells were perfused in HBSS ([Ca 2ϩ ] ϭ ϳ1.3 mM) to measure basal Ca 2ϩ entry. Then cells were perfused in 0 Ca 2ϩ HBSS for 2 min before the addition of 1 M thapsigargin to deplete ER Ca 2ϩ stores. Finally, cell perfusion was switched back to HBSS to trigger SOCE. Individual responses from ϳ50 cells per coverslip were monitored and averaged. Each experiment was repeated on 2-4 independent coverslips. In some experiments cells were pretreated with 50 M 2-aminoethyldiphenyl borate (2-APB) during the Fura-2 AM loading and unloading periods (ϳ2 h), and the cells were perfused in HBSS ϩ 50 M 2-APB for initial Ca 2ϩ measurements (7 min). Then cells were perfused with HBSS for 7 min before switching back to HBSS ϩ 50 M 2-APB.
A␤ ELISA-Conditioned media from transfected HEK cells were collected after overnight incubation, and sAPP␣, A␤ 40 , and A␤ 42 levels were quantified by ELISA using specific mono-clonal antibodies for capture (B113 for A␤ 40 , A387 for A␤ 42 , and 5228 for sAPP␣) followed by alkaline phosphatase detection with monoclonal antibody B436 and CSPD Sapphire II Luminescence Substrate (Applied Biosystems) (42).
Statistics-All statistical analyses were calculated using Prism software (GraphPad Software, Inc.). Statistical tests used are indicated in the corresponding figure legends. All data are represented as the mean Ϯ S.E.

Generation of a Genetic Model to Modulate Ca 2ϩ Influx-To
establish a cell culture model for studying the effects of Ca 2ϩ influx on APP processing and A␤ production, we generated cell lines overexpressing the components of the SOCE pathway. Specifically, we transduced HEK cells stably overexpressing human wild-type APP with retroviruses carrying empty vector DNA (HEK-APP vector) or a cDNA encoding the store-operated Ca 2ϩ channel Orai1 (HEK-APP-Orai1). For individual experiments, we then transiently transfected these cells with empty vector DNA (Control), the store-operated channel activator YFP-STIM1, or the constitutively active YFP-STIM1 D76A mutant. Western blot analysis confirmed expression of Orai1 and YFP-STIM1 or YFP-STIM1 D76A in these cells (Fig. 1A). Importantly, steady state levels of transgene-derived full-length APP were similar between vector transduced cells and cells overexpressing Orai1 and STIM1 (Fig. 1A). Additionally, levels of nicastrin and PS1 N-terminal fragment were similar among all groups of cells, indicating that endogenous ␥-secretase subunit expression was unaffected by overexpression of the SOCE machinery (Fig. 1A). Similar results were obtained for HEK cells stably expressing APPswe (data not shown). Because expression of full-length APP and endogenous ␥-secretase components are unaffected by manipulation of the SOCE machinery, we reasoned that these cells would provide a good model for investigating the effects of Ca 2ϩ influx on APP processing.
Having established expression of the components of the SOCE pathway, we utilized immunofluorescence staining to characterize their subcellular localization. In accordance with previously reported studies, Orai1 appeared to localize on the surface with little intracellular accumulation (Fig. 1B). On the other hand, YFP-STIM1-transfected cells exhibited a diffuse reticular pattern of fluorescence that overlapped with the ER marker, protein disulfide isomerase ( Fig. 2A). As expected, YFP-STIM1 fluorescence appeared as multiple discrete puncta after ER Ca 2ϩ store depletion with Tg that was distinct from the ER ( Fig. 2A). In accordance with previous findings (32), STIM1 D76A formed puncta even under store-replete conditions ( Fig. 2A). YFP-STIM1 fluorescence showed some overlap with surface localization of Orai under basal conditions (Fig. 2B). After store depletion, obvious overlap between YFP-STIM1 and Orai1 could be observed in discrete puncta. Consistent with previous studies, puncta containing STIM1 D76A were also positive for Orai 1 under basal conditions (Fig. 2B).
Next, we used TIRF microscopy to observe the dynamic behavior of YFP-STIM1 at sites adjacent to the plasma membrane. Using this approach, we confirmed the dynamic trans-location of YFP-STIM1 to plasma-membrane adjacent sites in HEK-APP cells after store depletion with Tg ( Fig. 2C; supplemental movie, left panel). In contrast, after transfection with YFP-STIM1 D76A , we observed numerous puncta near the plasma membrane under store-replete conditions, and the sizes of these puncta were not significantly affected by store depletion ( Fig. 2C; supplemental movie, right panel). These results confirm that STIM1 and STIM1 D76A are not only expressed but also dynamically localize as expected in our cell culture model. STIM1 D76A Alters Ca 2ϩ Homeostasis in HEK-APP Cells-Next, we loaded cells with Fura-2 AM and directly measured [Ca 2ϩ ] i levels. Because the EF-hand mutation of STIM1, STIM1 D76A , activates SOCE independently of ER store depletion, we used a three-part protocol to measure both store-dependent (SOCE) and store-independent Ca 2ϩ influx (Fig. 3, A  and D). First, cells were switched from perfusion in 0 Ca 2ϩ FIGURE 2. Subcellular localization of YFP-STIM1 and Orai1. HEK-Orai1 cells transfected with YFP-STIM1 or YFP-STIM1 D76A (D76A) were fixed and imaged under basal, ER Ca 2ϩ store-replete conditions or after a 10-min treatment with 1 M Tg in 0 Ca 2ϩ HBSS. A, cells were immunostained with an antibody against the ER marker, protein disulfide isomerase (PDI). Scale bars, 5 m. B, cells were immunostained with mAb 9E10 to detect Orai1. Insets show enlarged regions indicated by boxes. Scale bars, 5 m. C, HEK-APP cells transfected with YFP-STIM1 or YFP-STIM1 D76A were imaged using TIRF microscopy. Cells were maintained in HBSS in a humidified environment at 37°C, and images were acquired every 15 s. After 2 min, 1 M Tg was added in 0 Ca 2ϩ HBSS to deplete Ca 2ϩ stores. Representative frames from the TIRF image sequence taken before (0 min) and 10 min after the addition of Tg (12 min) are shown.
HBSS to HBSS to measure store-independent, basal Ca 2ϩ entry. After [Ca 2ϩ ] i levels plateaued, cells were perfused in 0 Ca 2ϩ HBSS, and 1 M Tg was added to deplete ER Ca 2ϩ stores followed by Ca 2ϩ add-back in HBSS to trigger SOCE. The plateau for basal Ca 2ϩ entry (Fig. 3, B and E) and the peak for SOCE (Fig. 3, C and F) were then quantified for each stable cell pool. In vector-transduced HEK-APP cells, overexpression of YFP-STIM1 produced a small, but non-significant potentiation of basal Ca 2ϩ entry, whereas overexpression of YFP-STIM1 D76A dramatically increased basal Ca 2ϩ entry (Fig. 3B). Upon Ca 2ϩ add-back, robust SOCE was induced in all cases without any discernible difference between the three groups (Fig. 3C).
Consistent with previous reports showing a dominant negative effect of expression of Orai1 alone (29,31), HEK-APP-Orai1 cells, which stably overexpress Orai1, had reduced SOCE (Fig. 3F) as compared with parental HEK-APP cells ( Fig. 3C; Table 1). The transient expression of YFP-STIM1 in HEK-APP-Orai1 cells produced a modest increase in basal Ca 2ϩ entry and a large increase in SOCE compared with the transient transfection control (HEK-APP-Orai1 cells transfected with an empty vector), as expected (Fig. 3, E and F). The transient expression of YFP-STIM1 D76A in HEK-APP-Orai1 cells, in contrast, led to dramatic increases in basal Ca 2ϩ entry along with significantly elevated basal [Ca 2ϩ ] i levels even in 0 Ca 2ϩ HBSS (Fig. 3, D-F). Similar alterations in Ca 2ϩ homeostasis were observed in HEK-APPswe cells (Fig. 4). Thus, our data demonstrate significant modulation of Ca 2ϩ homeostasis in these cells (summarized in Table 1), with the most dramatic changes observed in cells expressing STIM1 D76A .
To confirm that the elevated basal Ca 2ϩ levels in the HEK-APP-Orai1 cells transiently expressing YFP-STIM1 D76A were indeed the result of elevated SOCE, the effect of the SOCE   (Fig. 3G). Furthermore, the Ca 2ϩ levels in these cells dynamically changed when 2-APB was removed for a brief period and then added back (Fig. 3H).
It is interesting to note that although the transient expression of STIM1 significantly potentiated SOCE in HEK-APP-Orai1 cells, the overall magnitude of SOCE was not increased compared with HEK-APP cells transiently expressing STIM1 (Fig.  3, F versus C; Table 1). Previously published data (29 -31) in HEK cells overexpressing both Orai1 and STIM1 show much higher levels of SOCE than seen in HEK-APP cells overexpressing Orai1 and STIM1. Our own experiments overexpressing Orai1 and STIM1 in HEK cells (without overexpression of APP) also showed a much higher level of SOCE (Fig. 4D), suggesting that the apparent difference in the magnitude of SOCE potentiation by overexpressing Orai1 and STIM1 may result from the overexpression of APP. In support of this notion, we observed that the magnitude of SOCE in HEK-APPswe cells overexpressing Orai1 and STIM1 was also different from that seen in HEK cells overexpressing Orai1 and STIM1 (Fig. 4). STIM1 D76A Expression Leads to Accumulation of APP CTFs and Reduced A␤ Secretion-Having confirmed significant alterations in Ca 2ϩ homeostasis in our cell culture system, we began to investigate the effects of these changes on APP processing. First, we performed Western blot analyses to assess the steady-state levels of full-length APP and APP CTFs derived from ␣-secretase and BACE1 processing (␣-and ␤-CTFs, respectively) (Fig. 5A). We found that overexpression of YFP-STIM1 D76A significantly increased ␤-CTF levels in both HEK-APP vector and HEK-APP-Orai1 cells compared with control transfection (Fig. 5B). Similarly, significant accumulation of ␤-CTFs also occurred in HEK-APPswe-Orai cells after the expression of YFP-STIM1 D76A (Fig. 5, C and D). Although we also observed a trend in the accumulation of ␣-CTFs after the overexpression of STIM1 D76A , the difference did not reach statistical significance (Fig. 5). These results raised the possibility that STIM1-mediated alterations in Ca 2ϩ homeostasis may have effects on APP processing and/or the fate of APP CTFs.
To further characterize the effects of Orai1 and STIM1 expression on APP processing, we performed metabolic labeling in the HEK-APP vector and HEK-APP-Orai1 cells transfected with YFP-STIM1 D76A (Fig. 6, left panels). In parallel, we also transfected HEK-APPswe vector and HEK-APPswe-Orai1 cells with YFP-STIM1 D76A (Fig. 6, right panels). As described above, expression of YFP-STIM1 D76A alone or coexpression with Orai1 results in significantly elevated levels of Ca 2ϩ entry, even in the absence of store depletion (Figs. 3 and 4, Table 1). We first measured APP synthesis by pulse-labeling cells for 15 min with [ 35 S]Met/Cys and found similar levels of immature full-length APP in control and STIM1 D76A -transfected cells (Fig. 6). After 3 h of continuous labeling [ 35 S]Met/Cys, we observed similar overall levels of full-length APP among groups but found that compared with control cells, STIM1 D76A -transfected cells had a shift in the ratio of mature to immature APP, favoring immature APP (Fig. 6). In contrast, steady-state levels of mature and immature APP were similar across cells lines (Figs. 1A and 4A), suggesting a delay rather than an absolute blockade in APP maturation induced by elevation of [Ca 2ϩ ] i levels.
Next, we immunoprecipitated APP CTFs from lysates of cells after 3 h of continuous labeling using the APP C-terminal-specific antibody CTM1. We observed a small but reproducible increase in both ␣and ␤-CTFs in STIM1 D76A -transfected cells (Fig. 6), consistent with the analysis of steady-state APP CTFs in these cells (Fig. 5). To examine the levels of secreted A␤-related peptides, we then subjected conditioned media from these experiments to immunoprecipitation using monoclonal antibody 4G8. Notably, in cells transfected with STIM1 D76A , we observed decreased levels of secreted A␤ (Fig. 6) both from cells stably expressing wild-type APP and from cells expressing APPswe. Furthermore, in STIM1 D76A -transfected cells expressing APPswe we were also able to observe decreases in secretion of the alternate ␤-site cleavage-derived product A␤  and a corresponding increase in the levels of ␣-secretase cleavage-derived p3 peptide (Fig. 6, long exposure (long exp.)). Unfortunately, the levels of A␤  and p3 were too low to detect in cells expressing wildtype APP. Importantly, overall protein secretion was not reduced in STIM1 D76A -transfected cells (data not shown), suggesting that the observed effects on A␤ secretion are not due to a generalized impairment in secretory protein trafficking or secretion of lumenal cargo.
To confirm the results observed in our metabolic labeling experiments, we collected media conditioned by HEK-APP and HEK-APP-Orai1 cells after transfection with YFP-STIM1 D76A and quantified levels of A␤ 40 , A␤ 42 , and sAPP␣ by ELISA. In both wild-type APP-and APPswe-expressing cells, elevation of [Ca 2ϩ ] i levels by transfection of STIM1 D76A produced significant decreases in the amount of secreted A␤ 40 and A␤ 42 (Fig. 7,  A and C). Transfection of STIM1 D76A also resulted in increased secretion of sAPP␣ in cells expressing APPswe, although no increase in sAPP␣ levels was detected in cells expressing wildtype APP (Fig. 7, B and D).
Based on the observations from metabolic labeling experiments, we reasoned that accumulation of APP CTFs and diminution of A␤ secretion by elevation of [Ca 2ϩ ] i levels in cells expressing STIM1 D76A might be due to reduced ␥-secretase processing of APP ␤-CTFs. However, it was also possible that elevated [Ca 2ϩ ] i levels independently influenced BACE1 and ␥-secretase processing. To directly test the effects of [Ca 2ϩ ] i levels on ␥-secretase processing, we co-transfected HEK cells stably expressing Orai1 with YFP-STIM1 D76A and a plasmid encoding C99 -6myc, the APP ␤-CTF. This construct is often used to examine ␥-secretase cleavage of the APP ␤-CTF in the absence of any confounding effects of ␣and ␤-secretase processing of full-length APP. In agreement with our prediction, we found significantly reduced cleavage of C99 to AICD in cells expressing STIM1 D76A compared with controls (Fig. 7E). Together, these findings suggest that elevation of [Ca 2ϩ ] i levels through the SOCE pathway leads to multiple effects on APP metabolism, including reduced amyloidogenic processing of APP ␤-CTF by ␥-secretase.

DISCUSSION
In recent years several alternative hypotheses to the amyloid cascade have been advanced to explain the synaptic dysfunc-  tion and neuronal death that occurs in AD. Dysregulation of Ca 2ϩ homeostasis is one such example, and numerous studies have lent support to this hypothesis, from data demonstrating alterations in Ca 2ϩ handling in cells from affected patients to the discoveries of molecular roles for presenilins, APP, and A␤ peptides in the regulation of [Ca 2ϩ ] i levels (26). Although Ca 2ϩ has well established roles in neurotransmission and synaptic plasticity, its effects on the processing of APP and A␤ generation are less well known.
Studies to date have demonstrated conflicting results. In non-excitable cells, pharmacological elevation of [Ca ϩ2 ] i has been shown to both increase and decrease A␤ levels (43)(44)(45). In neurons, the data on the relationship between Ca 2ϩ influx and A␤ production is equally unclear. For example, Tg-and depolarization-induced elevations of [Ca ϩ2 ] i have been reported to selectively increase intraneuronal A␤ 42 (46). Likewise, ionomycin treatment of primary cortical neurons overexpressing human APPswe resulted in an increase in A␤ production through an increase in BACE1 expression (47). In contradic-tion, familial AD-linked mutations in PS1 that increase A␤ 42 production were found to decrease SOCE (16). Additionally, Ca 2ϩ influx through NMDA receptors has been shown to stimulate non-amyloidogenic ␣-secretase processing and inhibit A␤ production (48). In general, these studies have been hampered by one of two limitations: modulation of [Ca ϩ2 ] i by the expression of PS1 or PS2 mutants, which lends itself to ambiguity as PS1 or PS2 functions as the catalytic subunit of ␥-secretase, and the use of non-physiologic pharmacologic agents (Ca 2ϩ ionophore, thapsigargin) that often have pleiotropic effects on cellular function.
To avoid these limitations we utilized the molecular components of the SOCE system, STIM1 and Orai1, to genetically alter Ca 2ϩ levels in HEK cells. In this way we hypothesized that The bands corresponding to A␤ and ϩ11 A␤, which are generated by BACE1 cleavage at alternate sites followed by ␥-secretase cleavage, are indicated. The peptide p3 is released by sequential cleavage of APP by ␣and ␥-secretases. Imm, immature APP; Mat, mature APP. we would be able to observe the effects of a primary alteration in Ca 2ϩ homeostasis on APP processing in the absence of confounding effects due to presenilin mutations or pharmacologic manipulations. Based on data from previous studies (29 -31), we initially expected to observe a potentiation of SOCE in HEK-APP-Orai cells overexpressing STIM1. In contrast to this well characterized effect, we found that the magnitude of SOCE was similar in HEK-APP and HEK-293-APP-Orai1 cells transfected with STIM1. Because we were able to observe STIM1-mediated potentiation of SOCE in HEK-Orai1 cells, but not HEK-APP-Orai1 cells (Figs. 3 and 4), we conclude that this discrepancy is likely attributable to the overexpression of APP. Although numerous papers have demonstrated effects of APP expression on Ca 2ϩ homeostasis, the data are often conflicting (24, 25, 49 -51). Additionally, no studies to date have examined the effects of APP on SOCE in cells overexpressing STIM1 and Orai1, and thus the exact mechanism by which APP overexpression is affecting this process remains unclear. It will be interesting to further investigate this effect in future studies.
Fortunately, we also utilized the constitutively active STIM1 D76A mutant in our study. It is known that expression of this well characterized mutant of the luminal EF-hand domain of STIM1 leads to constitutive activation of Ca 2ϩ influx even under store-replete conditions (29,32). As expected, in co-transfected cells, Orai1 with STIM1 D76A oligomerized and formed numerous puncta near the plasma membrane even under store-replete conditions (Fig. 2), in agreement with constitutive Ca 2ϩ entry. Indeed, this reconstituted SOCE channel function was inhibited by preincubation with the SOCE blocker, 2-APB (Fig. 3) (31). Although in early studies 2-APB was thought to inhibit inositol trisphosphate receptors in addition to store-operated Ca 2ϩ channels, as a result of intense research conducted since the discovery of Orai and STIM family proteins, the complex actions of 2-APB effects on SOCE have been attributed to a direct block of Orai subunits at the channel level as well as an additional uncoupling of STIM1 and Orai subunits (52)(53)(54)(55)(56)(57). Thus, the most dramatic derangements in Ca 2ϩ homeostasis in HEK-APP-Orai1 cells transfected with STIM1 D76A is due to activation of constitutive SOCE. Interestingly, HEK-APP-Orai1 cells transfected with STIM1 D76A had significant accumulation of ␤-CTFs, suggesting potential alterations in APP processing and/or metabolism. Therefore, we chose to focus further investigations of APP processing on these cells with constitutive Ca 2ϩ entry. Using metabolic labeling and ELISA we found striking reductions in the secretion of A␤ with concomitant increases in the levels of both ␣and ␤-CTFs. Moreover, we observed reduced APP maturation, a small increase in p3 secretion, and reduced ␥-secretase cleavage of APP C99. These results suggest that elevations in [Ca 2ϩ ] i levels resulting from constitutive activation of Ca 2ϩ entry affects APP metabolism at multiple levels.
Notably, the magnitude of the effect on APP processing we observed appears to be proportional to the level of derangement in cellular Ca 2ϩ homeostasis. HEK-APP-Orai1 cells transfected with STIM1 D76A exhibited the most prominent elevations in [Ca 2ϩ ] i levels to such a degree that resting cytosolic Ca 2ϩ levels were significantly elevated even in nominally Ca 2ϩfree buffer. This elevation of basal Ca 2ϩ levels suggests signifi-cant alterations in the homeostatic mechanisms controlling resting [Ca 2ϩ ] i levels. Although this effect has been observed previously, the mechanisms mediating it have not been well characterized (29). However, the alterations in Ca 2ϩ homeostatic mechanisms are clearly dependent on the elevated SOCE, as the elevated basal Ca 2ϩ levels in HEK-APP-Orai1-YFP-STIM1 D76A cells were returned to almost normal levels by preincubating cells for 2 h with 2-APB, a widely used pharmacological agent that (at the concentration used in our study) inhibits SOCE and calcium release-activated Ca 2ϩ currents (Fig. 3, G and H). For the purposes of our study, these dramatic alterations in Ca 2ϩ handling were correlated with both greater accumulation of APP CTFs and reduced A␤ generation compared with STIM1 D76A -transfected cells that do not coexpress Orai1. Therefore, there may be a dose-response relationship between the magnitude of elevation in [Ca 2ϩ ] i levels and the impairment in A␤ generation, strengthening the correlation between dysregulation of Ca 2ϩ homeostasis and reduced amyloidogenic APP processing.
Throughout our investigations we utilized cells overexpressing wild-type APP and the familial AD-linked Swedish APP mutation. In almost all experiments we found a similar effect of alterations in Ca 2ϩ homeostasis on APP metabolism. These include a delay in maturation of APP, accumulation of APP CTFs, reduced secretion of A␤, and a small increase in p3 levels. The one exception was the effect of STIM1 D76A transfection on secretion of sAPP␣. In cells expressing APPswe we found that elevation of [Ca 2ϩ ] i levels resulted in increased sAPP␣ secretion, and again greater alterations in Ca 2ϩ homeostasis were correlated with larger increases in sAPP␣ secretion (Fig. 7D). However, in cells expressing wild-type APP, STIM1 D76A transfection produced no change in secretion of sAPP␣ (Fig. 7B). We suggest that the reason for this difference is likely due to the extent to which full-length APP molecules are subject to amyloidogenic versus non-amyloidogenic processing in these cells. Whereas most wild-type APP undergoes non-amyloidogenic processing, the presence of Swedish mutations in APP leads to preferential BACE1 cleavage and consequently a greater proportion of APP undergoing amyloidogenic processing (58). Notably, BACE1 processing of APPswe can occur as early as during transit of nascent APPswe polypeptides through the cis-Golgi (58). Thus, in HEK-APPswe cells even a small reduction in amyloidogenic processing would allow more APP to reach the cell surface and be subject to non-amyloidogenic processing, resulting in a readily observable increase in sAPP␣. On the other hand, in HEK-APP-overexpressing cells where most APP is already undergoing non-amyloidogenic processing, further increases in APP available for non-amyloidogenic processing result in a proportionally smaller effect on the total levels of sAPP␣ produced.
Taken together, our results demonstrate that elevation of [Ca 2ϩ ] i levels by Ca 2ϩ influx through store-operated channels leads to reduced amyloidogenic processing of APP and a dramatic decrease in the generation of A␤ 40 and A␤ 42 . Although A␤ has been implicated in the disruption of intracellular Ca 2ϩ homeostasis through a variety of mechanisms, including membrane disruption, Ca 2ϩ pore formation, and ion channel modulation, our data suggest that the relationship between Ca 2ϩ and A␤ may be reciprocal. Specifically, it appears that A␤ species (peptides, oligomers, and/or fibrils) may lead to elevations in [Ca 2ϩ ] i levels that then negatively regulate amyloidogenic APP processing, reducing further production of A␤. This reciprocity could serve as a protective cellular mechanism, which limits production of A␤ when extracellular concentrations are high, preventing pathologic accumulation of potentially toxic A␤ peptides. Alterations in the mechanisms regulating Ca 2ϩ accumulated during aging or through the acquisition of a mutation in PSEN1 or other AD-associated genes could then potentially disrupt this homeostatic balance, favoring AD pathogenesis. Alternatively, Ca 2ϩ induced accumulation of APP CTFs may result in alterations in intracellular signaling, as has been recently demonstrated (59).
In neurons, the principal cell type affected in AD, the relationship between [Ca 2ϩ ] i levels and A␤ generation and secretion is likely to be more complex than observed in our experiments in HEK cells. Calcium-regulating systems are markedly more complex in neurons, and Ca 2ϩ signals are involved in diverse processes such as protein and secretory vesicle trafficking for neurotransmission, endocytosis, gene transcription, and synaptic plasticity. Neurons are also polarized cells, and many Ca 2ϩ signaling events are restricted to specific microdomains. Overall, this results in a system in which the effects of Ca 2ϩ signaling on APP processing will depend on the localization, magnitude, and mode of Ca 2ϩ entry. For example, in presynaptic nerve terminals A␤ secretion and intraneuronal accumulation of A␤ have been linked to Ca 2ϩ -dependent neuronal activity (46,60). On the other hand, at post-synaptic sites Ca 2ϩ influx through NMDA receptors has been reported to reduce A␤ release (48).
We chose to utilize simplified non-neuronal cells for this work precisely because we wanted to avoid the complexity in neuronal Ca 2ϩ -regulating systems. Thus, we believe our work presents strong evidence of a direct role for elevated [Ca 2ϩ ] i levels in the negative regulation of amyloidogenic APP processing. However, because we utilized the components of the SOCE pathway (STIM1 and Orai1) to manipulate [Ca 2ϩ ] i levels, we cannot rule out the possibility that the effects we have observed are specific to STIM1-mediated Ca 2ϩ influx through store-operated Ca 2ϩ channels. The implications of this possibility on disease pathogenesis are difficult to predict because, although some studies in neurons have demonstrated functional SOCE, the precise roles of STIM1 and Orai1 in the central nervous system remain unknown (61,62). In fact, STIM1 likely has functions independent of SOCE in neurons as it has been shown to be a negative regulator of voltage-gated Ca 2ϩ channels (63,64). Further studies of the specific role of STIM1 on neuronal Ca 2ϩ regulation and APP processing are warranted in the future.