Presenilin mutations linked to familial Alzheimer's disease cause an imbalance in phosphatidylinositol 4,5-bisphosphate metabolism

Abstract

Phosphatidylinositol 4,5-bisphosphate (PIP₂) is an important cellular effector whose functions include the regulation of ion channels and membrane trafficking. Aberrant PIP₂ metabolism has also been implicated in a variety of human disease states, e.g., cancer and diabetes. Here we report that familial Alzheimer's disease (FAD)-associated presenilin mutations cause an imbalance in PIP₂ metabolism. We find that the transient receptor potential melastatin 7 (TRPM7)-associated Mg²⁺-inhibited cation (MIC) channel underlies ion channel dysfunction in presenilin FAD mutant cells, and the observed channel deficits are restored by the addition of PIP₂, a known regulator of the MIC/TRPM7 channel. Lipid analyses show that PIP₂ turnover is selectively affected in FAD mutant presenilin cells. We also find that modulation of cellular PIP₂ closely correlates with 42-residue amyloid β-peptide (Aβ42) levels. Our data suggest that PIP₂ imbalance may contribute to Alzheimer's disease pathogenesis by affecting multiple cellular pathways, such as the generation of toxic Aβ42 as well as the activity of the MIC/TRPM7 channel, which has been linked to other neurodegenerative conditions. Thus, our study suggests that brain-specific modulation of PIP₂ may offer a therapeutic approach in Alzheimer's disease.

Cerebral elevation and accumulation of the amyloid β-peptide (Aβ) are early and necessary steps in the pathogenesis of Alzheimer's disease (AD) (1–3). Aβ is produced by sequential proteolytic cleavages of the amyloid precursor protein (APP) by a set of membrane-bound proteases termed β- and γ-secretases (4). Heterogeneous γ-secretase cleavage at the C-terminal end of Aβ produces two major isoforms of Aβ, Aβ40 and Aβ42. Although Aβ40 is the predominant cleavage product, the less abundant, Aβ42, a longer and more amyloidogenic form of Aβ, is considered a key pathogenic agent in AD (5, 6). Approximately 5% of AD cases are familial (FAD), with most attributable to autosomal dominant mutations in the presenilin (PS1 and PS2) genes (1). It is now well accepted that PS1 or PS2 serves as a catalytic component of the γ-secretase complex that is essential for regulated intramembrane proteolysis (RIP) of select transmembrane proteins (4, 7), including APP to yield Aβ (8). PS1 and PS2 deficiency leads to impaired RIP and reduced Aβ generation (7–9), whereas FAD-linked mutations in the presenilins lead to misregulation of γ-secretase activity, evidenced by an increase in the ratio of Aβ42 to Aβ40 as well as the reduced generation of the APP intracellular domain (AICD) (10, 11).

Increasing evidence indicates that the presenilins are involved in several additional, γ-secretase-independent cellular functions including Wnt and Akt signaling (12–14), membrane protein trafficking (15), ion channel regulation (16, 17), and cell junction organization (18). Two of the most consistent cellular changes associated with both PS1 and PS2 FAD-associated mutations include ion channel dysfunction, such as defects in capacitative Ca²⁺ entry (CCE) (16, 17) and membrane trafficking defects, including diminished cell surface delivery of APP (15, 19–21). These Aβ42-independent cellular changes may also contribute to the onset and/or neuropathological characteristics of presenilin-dependent FAD (22). Interestingly, certain pathogenic mutations in PS1 cause frontotemporal dementia with profound neurofibrillary tangle pathology in the absence of amyloid plaques (23). However, it is currently unknown how the presenilins mediate multiple cellular processes in addition to their role as a protease subunit.

In the present work, we attempted to identify an ion channel(s) that may underlie the CCE defects associated with presenilin-dependent FAD. CCE, in which the emptying of intracellular Ca²⁺ stores activates Ca²⁺ influx, is the major Ca²⁺ entry pathway in nonexcitable cells, and also plays a role in excitable cells (24). Given the key role that Ca²⁺ influx plays in regulation of cell function, including exocytosis, enzyme activity, gene expression, and apoptosis, it is perhaps not surprising that CCE channels are increasingly associated with various disease conditions (24). However, the regulation and identity of channels that underlie CCE remain obscure.

We find that the transient receptor potential melastatin 7 (TRPM7)-associated Mg²⁺-inhibited cation (MIC) channel (25) is responsible for the observed ion channel dysfunction in PS1 FAD mutant cells. We also provide evidence that an imbalance in phosphatidylinositol 4,5-bisphosphate (PIP₂) metabolism underlies the suppression of MIC channel activity in PS FAD mutant cells. Inhibition of γ-secretase activity does not reverse the deficits in either MIC channel activity or PIP₂ metabolism, suggesting that these changes are either independent or are upstream of γ-secretase function. Interestingly, the levels of cellular PIP₂ also closely correlate with levels of Aβ42. Given the role of PIP₂ in a variety of cellular processes, including ion channel gating and protein/membrane trafficking (26), PIP₂ imbalance may play a role in multiple cellular changes associated with presenilin FAD.

Results

Presenilins Negatively Modulate MIC/TRPM7 Channels.

We previously reported that store depletion-induced (CCE) currents were compromised in cells expressing FAD mutant forms of PS1 or PS2 (16). Several studies suggest that CCE channels are related to transient receptor potential homologs (24). TRPM7, the pore-forming subunit of MIC channels, is widely expressed, including in the brain (25), and has been associated with anoxic neuronal death and Guamanian neurodegenerative disorders (27, 28). We therefore revisited the ion channel mechanism that is affected by FAD-linked presenilin mutations to investigate whether MIC channels are subjected to presenilin-dependent regulation.

MIC currents were identified in heterologous cells based on reported electrophysiological characteristics, e.g., large-channel conductances (≈40 pS) and I–V curves with a reversal potential of 0 mV, with weak inward rectification in divalent-free solution [see supporting information (SI) Fig. 6 A–C]. Detailed characterization of MIC/TRPM7 channels is described in SI Supporting Materials and Methods and SI Fig. 6.

To determine whether the presenilins are required for regulation of MIC channel activity, we examined the effects of presenilin deficiency on MIC currents. To that end, we performed whole-cell recordings of MIC currents in mouse embryonic fibroblasts (MEFs) that lack both PS1 and PS2 genes (29). Time-dependent MIC currents were observed in both wild-type (WT) and PS1 and PS2 double-knockout (dKO) MEFs, and the steady-state maximum currents were similar between WT and dKO cells (SI Fig. 7A). Interestingly, MIC channels in PS1/2 dKO MEFs appear to be “preactivated” as evidenced by elevated levels of current at the 0 time point. Once fully activated, ramp I–V curves were very similar between WT and dKO MEF cells (SI Fig. 7B), indicating that preactivation of MIC currents may reflect an increase in the number of open channels rather than a change in channel properties.

To evaluate the number and activity of MIC channels in PS1/2 dKO MEFs, we carried out patch clamp recordings in the cell-attached configuration. Patch clamp recordings revealed a constitutive basal MIC channel activity in PS1/2 dKO MEFs even in the absence of current induction (Fig. 7C). Treatment with 2-aminoethoxydiphenyl borate, a reversible inhibitor of MIC channels, reduced the constitutively active MIC currents in PS1/2 dKO MEFs to levels comparable with those of WT cells (Fig. 7 D and E). Thus, our data suggest that presenilin deficiency leads to aberrant MIC channel gating, resulting in higher levels of basal MIC channel activity.

FAD-Associated PS1 Mutations Attenuate MIC Current Density.

We next examined whether FAD-associated PS1 mutations affect MIC currents. MIC currents were measured in HEK293 cells stably transfected with either WT or FAD mutant forms (L286V and ΔE9) of PS1. MIC current densities were significantly reduced in cells expressing the two FAD-linked mutants of PS1, L286V and ΔE9, compared with WT PS1-expressing cells (Fig. 1A and B). The FAD-associated suppression of MIC currents was highly reproducible, and was also confirmed in other cell types, including Neuro2a and CHO cells (data not shown). Although current densities were suppressed in FAD mutant cells, the rate of MIC current activation (I/I_MAX) in FAD mutant cells was not significantly different from WT PS1-expressing cells (data not shown). These data indicate that there might be fewer MIC channels present in the FAD PS1-expressing cells, or alternatively, the activation of MIC channels might be chronically suppressed by the presence of FAD mutant PS1.

Fig. 1.

FAD-linked PS1 mutations affect MIC current in a γ-cleavage-independent manner. (A) Whole-cell MIC currents in HEK293 cells stably expressing WT or FAD mutant forms of PS1. MIC currents were evoked by step pulse of −120 mV every 10 s in whole-cell configuration. Each data point represents mean ± SEM. (B) Comparison of average peak MIC current densities (pA/pF) from HEK293 cells expressing WT PS1 (PS1wt), L286V PS1, or ΔE9 PS1. Each column represents mean ± SEM. pA/pF: PS1wt, −103.4 ± 5.9 (n = 15); L286V, −80.7 ± 4.3 (n = 19); ΔE9, −59.9 ± 4.7 (n = 21). (C) Pharmacological inhibition of γ-secretase activity fails to reverse MIC current deficits associated with PS1 FAD mutations. Native HEK293 cells and L286V and ΔE9 PS1 FAD mutant cells were preincubated with two potent γ-secretase inhibitors, compound E (Comp E; 300 nM) or L-685,458 (2 μM), for 6 h before recording. Fresh γ-secretase inhibitors were applied after the first 3-h incubation. Each column represents mean MIC current density ± SEM at steady-state (pA/pF); control HEK293 cells, −98.4 ± 6.6 (n = 14); compound E-treated control, −61.5 ± 5.7 (n = 5); ΔE9, −51.5 ± 5.6 (n = 4); compound E-treated ΔE9, −35.9 ± 5.9 (n = 11); L-685,458-treated ΔE9, −42.4 ± 7.8 (n = 8); L286V, −76.3 ± 4.7 (n = 9); compound E-treated L286V, −66.0 ± 27.3 (n = 8); L-685,458-treated L286V, −108.7 ± 18.5 (n = 4). (D) Effects of Aβ42-lowering NSAIDs, sulidac sulfide (SS) and indomethacin (IND), on MIC current density in ΔE9 PS1 FAD mutant cells. Compound E treatment and MIC current measurements were as described in C. ∗, P < 0.001; ∗∗, P < 0.01.

Although the MIC channel serves as a nonselective cation channel permeable not only to Ca²⁺ but also other divalent cations, including Mg²⁺, other channels have been previously implicated as putative CCE channels (24), including the calcium release-activated calcium (CRAC) channel that mediates a highly Ca²⁺-selective, small-conductance current (30, 31). To determine whether FAD presenilin-mediated suppression is selective to MIC channels, we next examined whether FAD mutations affect CRAC currents. By depleting Ca²⁺ ions in the presence of internal Mg²⁺, we were able to detect the very small Ca²⁺ current that underlies CRAC conductance in these cells (SI Fig. 8A) (30, 31). In contrast to MIC currents, CRAC currents were not affected by PS1 FAD mutations (SI Fig. 8A), indicating that PS1 FAD mutations selectively affect MIC but not CRAC channels. These results further substantiate the idea that the MIC channel is a target ion channel activity that is subjected to presenilin-dependent regulation.

γ-Secretase Inhibitors or Modulators Fail to Restore MIC Currents in FAD-Associated Mutant PS1 Cells.

Subtle but consistent elevation in the ratio of Aβ42 to Aβ40 is the most common pathogenic phenotype in presenilin FAD (1, 4, 6). To determine whether observed MIC current suppression in FAD mutant PS1 cells is the result of altered γ-cleavage of APP, e.g., relative increases in Aβ42 production, we tested the effects of two of the most potent, widely used γ-secretase inhibitors on MIC currents in native HEK293 cells as well as cells stably expressing the ΔE9 and L286V FAD forms of PS1. Treatment of native HEK293 cells with either compound E (32) or L-685,458 (33) did not produce any significant effects on MIC currents (Fig. 1C) or on inositol 1,4,5-trisphosphate (IP₃)-mediated Ca²⁺ release (data not shown). Moreover, treatment with these γ-secretase inhibitors failed to rescue MIC current deficits in stable HEK293 cells expressing either ΔE9 or L286V FAD mutant forms of PS1 (Fig. 1C). By comparison, Aβ production has been shown to be efficiently inhibited by treatment with these inhibitors at concentrations far below those used in our experiments (32, 33). These data indicate that presenilin-associated proteolytic activity is dispensable for altered MIC channel activity in FAD mutant PS1-expressing cells.

Several nonsteroidal antiinflammatory drugs (NSAIDs) have been shown to reduce Aβ42 levels, presumably through a direct modulation of the γ-secretase complex (34). We next tested whether Aβ42-lowering NSAIDs can reverse MIC current deficits in FAD mutant PS1-expressing cells. Pretreatment of ΔE9 PS1 FAD cells with Aβ42-lowering NSAIDs, sulindac sulfide and indomethacin, did not confer any detectable effects on MIC currents (Fig. 1D). Thus, these data support the idea that FAD PS1 mutations give rise to MIC current deficits independent of their effects on γ-secretase-mediated cleavage of APP or other cellular substrates.

Elevated PIP₂ Levels Ameliorate MIC Deficits Associated with Presenilin FAD.

Our data indicate that PS1 FAD mutations suppress the activity of the TRPM7-associated MIC channel (Fig. 1 A and B). It has been shown that PIP₂ (and not its metabolites) serves as a direct modulator of selected ion channels, including TRPM7 (26, 35). Consequently, phospholipase C (PLC)-mediated hydrolysis and depletion of PIP₂ have been shown to inhibit cation entry through the TRPM7 channel (35). Consistent with the above observations, the presence of anti-PIP₂ antibody in the intracellular pipette solution but not control antibody resulted in the inhibition of MIC currents in our recordings (SI Fig. 8B). To explore the possibility that aberrant PIP₂ metabolism may underlie the observed MIC channel deficits, we determined whether the augmentation of PIP₂ levels can restore MIC currents in FAD mutant PS1-expressing cells. Direct application of PIP₂ to the cytoplasmic side of the membrane led to the recovery of MIC current densities to levels comparable with those of WT PS1-expressing cells (Fig. 2A and B). In contrast, application of either the stereoisomer phosphatidylinositol 3,4-bisphosphate or PA failed to “rescue” MIC channel deficits in PS1 FAD mutant cells (Fig. 2B).

Fig. 2.

PIP₂ restores MIC currents in ΔE9 PS1 FAD mutant cells. (A) Recovery of MIC current by PIP₂ in FAD-linked ΔE9 PS cells. MIC currents were recorded in WT and ΔE9 mutant HEK293 cells. In some recordings from ΔE9 cells, 50 μM PIP₂ [PI(4,5)P₂], phosphatidylinositol 3,4-bisphosphate [PI(3,4)P₂], or phosphatidic acid (PA) was included in pipette solutions. (B) Each data point represents the mean ± SEM at steady state. PA/pF: WT PS, −103.4 ± 6.0 (n = 15); ΔE9, −60.0 ± 4.7 (n = 22); ΔE9 PIP₂, −89.28 ± 12 (n = 8); ΔE9 PI(3,4)P₂, −47.94 ± 9.61 (n = 8); and ΔE9 PA, −42.43 ± 3.22 (n = 8). (C) Recovery of MIC currents by pretreatment of FAD-linked ΔE9 PS1 cells with edelfosine (EDEL). Time-dependent activation of MIC current from control and edelfosine-pretreated cells is shown. (D) Each column represents mean current density ± SEM at steady state. PA/pF: control cells, −47.8 ± 7.0 (n = 7); edelfosine-pretreated cells, −78.3 ± 17.4 (n = 8). ∗∗, P < 0.05.

To test whether the recovery of PS1 FAD-associated MIC current deficits may be achieved by manipulation of endogenous PIP₂ levels, we used a pharmacological approach. A major catabolic pathway for PIP₂ in eukaryotic cells involves PLC-mediated hydrolysis of PIP₂ to generate diacylglycerol and IP₃ (36). We therefore tested whether an inhibitor of the PLC pathway (and therefore PIP₂ breakdown), edelfosine (36), could restore MIC currents in FAD mutant PS1 ΔE9 cells (Fig. 2 C and D). The use of another widely used PLC inhibitor, U73122, was excluded in the present study because of reported off-target effects and toxicity (36). Similar to direct PIP₂ administration, edelfosine pretreatment led to the recovery of MIC currents in FAD mutant PS1 ΔE9 cells (Fig. 2 C and D). Thus, pharmacological augmentation of cellular PIP₂ reverses MIC current deficits in PS1 FAD cells, raising a possibility that aberrant PIP₂ regulation may underlie MIC channel dysfunction associated with presenilin FAD.

FAD-Associated Mutant Presenilins Cause an Imbalance in PIP₂ Metabolism.

To monitor cellular PIP₂ levels, we first carried out HPLC analyses to measure the steady-state levels of PIP₂. Total (steady-state) PIP₂ lipid mass in HEK293 cells stably expressing either WT or FAD PS1 (ΔE9, L286V) showed a very consistent, but small (≈5–10%) reduction in PIP₂ levels in PS1 FAD cells compared with control cells (SI Fig. 8C). This result was also confirmed in CHO cells overexpressing the M146V FAD PS1 mutation (data not shown). Because PIP₂ is subject to a high turnover, we suspected that steady-state levels of PIP₂ may not provide a very sensitive measure of cellular PIP₂ metabolism. Thus, to determine more directly whether the presence of presenilin FAD mutations results in aberrant PIP₂ metabolism, we carried out analysis of phosphoinositide turnover by using a cell-free radiolabeling assay in the presence of [γ-³²P]ATP (Fig. 3A) (37). Incorporation of radioactive phosphate into phospholipids occurs through phosphorylation reactions mediated by lipid kinases but also reflects the contribution of phosphatases and phospholipases present in cell extracts.

Fig. 3.

Alteration of PIP₂ metabolism in presenilin FAD mutant cells. (A) Lipid kinase and TLC analysis of extracts prepared from HEK293 cells stably transduced with either WT or FAD mutant (ΔE9, L286V) PS1. PIP, phosphatidylinositol 4-monophosphate. (B) Quantification of TLC data on PIP₂. Data are from five independent experiments, and they are expressed as mean ± SEM. (C) Quantification of TLC analyses on extracts prepared from HEK293 cells stably expressing WT or FAD mutant (N141I) PS2. Data are from at least three independent experiments, and they are expressed as mean ± SEM. ∗, P < 0.01; ∗∗, P < 0.05.

We began by analyzing extracts prepared from HEK293 cells stably expressing human WT or FAD mutant PS1 (ΔE9, L286V) (Fig. 3A), and we found a ≈40% reduction in γ-³²P incorporation into PIP₂ in PS1 FAD-expressing cells relative to control cells (Fig. 3B). To determine whether diminished PIP₂ radiolabeling is a common phenotype for both PS1 and PS2 FAD, we next examined PIP₂ turnover in extracts prepared from WT and FAD mutant PS2 (N141I)-expressing cells. In PS2 FAD-expressing cells, PIP₂ labeling was reduced by ≈26% cells compared with WT controls (Fig. 3C). In both PS1- and PS2-expressing cells, no consistent changes in the labeling of PIP (SI Fig. 9A) and PA (SI Fig. 9B) were observed. Similar defects in PIP₂ metabolism were also observed in Neuro2a cells stably expressing either the E280A or the H163R FAD-associated PS1 variant (SI Fig. 9C). Thus, consistent with electrophysiology data (Fig. 1), our lipid analyses indicate that the presence of PS1 and PS2 FAD mutations leads to an imbalance in PIP₂ metabolism that may be attributed to either diminished synthesis or enhanced breakdown of PIP₂.

Inhibition of PIP₂ Breakdown but Not γ-Secretase Reverses Aberrant PIP₂ Turnover in FAD Mutant PS1 Cells.

Our data show that the defects in MIC activity associated with FAD mutations are independent of γ-secretase activity of presenilins (Fig. 1C). Because these defects can be rescued by PIP₂ (Fig. 2 A–D), we reasoned that the observed PIP₂ alteration could also be independent of the alteration in γ-secretase activity present in presenilin FAD. Pretreatment with compound E, a potent γ-secretase inhibitor, failed to rescue PIP₂ defects in PS1 FAD cells (Fig. 4A and B). This result suggests that aberrant PIP₂ metabolism is either independent or is upstream of FAD-associated misregulation in γ-secretase activity. In contrast, inhibition of PIP₂ breakdown by pretreatment with edelfosine led to substantial elevation in PIP₂ labeling in PS1 ΔE9 FAD mutant cells to levels comparable with or higher than WT controls (Fig. 4 C and D), as shown for MIC current deficits (Fig. 2 C and D).

Fig. 4.

Inhibition of PLC but not γ-secretase reverses FAD-associated reduction in PIP₂ turnover. (A and B) HEK293 cells stably expressing either WT or FAD mutant (ΔE9) PS1 were pretreated with either DMSO or γ-secretase inhibitor (compound E; CpdE) for 6 h and subjected to lipid kinase/TLC analysis. (C and D) TLC analysis of extracts prepared from HEK293 cells stably expressing either WT or FAD mutant (ΔE9) PS1 that were pretreated with either DMSO or edelfosine (EDEL) for 6 h.

PIP₂ Levels Inversely Correlate with Levels of Aβ42.

Several recent studies indicate that brain lipids are essential for reconstitution of γ-secretase activity using isolated γ-secretase components (38, 39). Thus, it is conceivable that changes in PIP₂ levels (and subsequent alterations in surrounding membrane composition) may influence γ-secretase activity. We therefore tested the hypothesis that modulation of cellular PIP₂ levels correlates with levels of Aβ. HeLa cells stably expressing the Swedish mutant form of APP were treated with DMSO (control), the PLC inhibitor edelfosine, or PLC activator m-3M3FBS (40). Edelfosine treatment (10 μM) resulted in a ≈10% increase in the steady-state levels of PIP₂ (Fig. 5A). Meanwhile, activation of PLC by treatment with 15 μM m-3M3FBS yielded a ≈10% decrease in the steady-state levels of PIP₂ (Fig. 5B). We found that PLC inhibition also results in a significant down-regulation of Aβ42 (Fig. 5C), whereas activation of PLC gave rise to an up-regulation of Aβ42 (Fig. 5D). Thus, the changes in PIP₂ levels showed a strong correlation with changes in secreted Aβ42 levels (37.3% decrease for edelfosine and 37.2% increase for m-3M3FBS) (Fig. 5 C and D). The effect of PLC inhibition on Aβ42 was also confirmed with miltefosine, an active analog of edelfosine (Fig. 5C). We did not observe any significant effects of treatment on steady-state levels of full-length APP, as determined by Western blotting of cell lysates with 6E10 (data not shown). Modulation of other lipid metabolites by MAFP, a phospholipase A₂ inhibitor, did not confer any detectable effects on Aβ42 levels (data not shown). The dose-dependent effect of PLC activity modulators on Aβ42 levels was also confirmed in HEK293, CHO, and Neuro2a cells stably expressing human APP (data not shown). Our data indicate that inhibition (activation) of PLC activity can rescue (mimic) the Aβ42 FAD phenotype (e.g., Aβ42 overproduction).

Fig. 5.

Changes in PIP₂ levels correlate with Aβ42 levels. (A and B) Treatment with PLC activity modulators leads to changes in steady-state levels of PIP₂. HeLa-APPsw were treated with either PLC inhibitor edelfosine (EDEL, 10 μM) or PLC activator m-3mFBS (M3M, 15 μM) for 6 h and cell membranes analyzed for steady-state levels of PIP₂. The data were from 3 independent experiments and are expressed as mean ± SEM. (C and D) Corresponding Aβ42 levels are shown. The culture media from the HeLa-APPsw cells were assayed for secreted Aβ42 by sandwich ELISA. (E and F) PIP₂-mediated regulation of γ-secretase cleavage of APP. HEK293 cells stably expressing WT human PS1 were transiently transfected with the myc-tagged C-terminal stub of APP (C99-myc). Twenty-four hours posttransfection, cells were treated with either PLC inhibitor [edelfosine (EDEL, 10 μM) (E)] or PLC activator [M3M, 15 μM (F)] for 6 h. Total secreted Aβ was detected by immunoprecipitation with 7N22 and NuPAGE Western blotting with 6E10 (Bottom). Secreted Aβ42 was analyzed by using sandwich ELISA. Data are from at least three independent experiments, and they are expressed as mean ± SEM. ∗, P < 0.01; ∗∗, P < 0.05.

To define more clearly the role of PIP₂ in modulating γ-secretase activity (e.g., Aβ42 levels), we next expressed an APP-C99 construct, an ectopic γ-secretase substrate that resembles the β-secretase-generated, membrane-associated APP stub in heterologous cells. In C99-transfected cells, the Aβ42-reducing activity of edelfosine and the Aβ42-promoting activity of m-3m3FBS were still observed (Fig. 5 E and F), indicating that PIP₂-mediated modulation of Aβ42 levels may occur at the level of presenilin/γ-secretase modulation. Under these conditions, total Aβ levels were unaltered.

We next examined the possibility that PIP₂ itself rather than its PLC-derived metabolites affects Aβ42 levels. In the brain, synaptojanin 1 (Synj1) is the major inositol 5-phosphatase that converts PIP₂ into PI (4)P (41), thereby reducing the levels of cellular PIP₂. Expression of constructs encoding the Synj1 inositol 5-phosphatase domain (Synj1-IPP) resulted in an enhanced levels of Aβ42 (SI Fig. 10), mimicking the effects of FAD mutant presenilins. This result implies that PIP₂ itself rather than its metabolites (e.g., PLC metabolites) is likely to mediate the observed regulatory role on Aβ42 levels.

Discussion

The major insight from the present study is that altered PIP₂ metabolism may play a key role in presenilin FAD-associated cellular changes. We show that the MIC/TRPM7 channel is a target ion channel activity that is affected by FAD-associated pathogenic presenilin mutations through a PIP₂-dependent mechanism. Furthermore, PIP₂ appears to modulate Aβ42 levels.

It has been well documented that all FAD-associated presenilin mutations result in misregulation of the presenilin-dependent γ-secretase (leading to an increased Aβ42:Aβ40 ratio and reduced AICD generation) (9, 11) and alterations in cellular cation homeostasis (16, 17). However, the identity of presenilin-modulated cation channels and the relationship between the proteolytic function of the presenilins and presenilin-dependent regulation of ion entry have not been elucidated. The observation that FAD-associated presenilin mutations affect the activity of Ca²⁺-permeable ion channels was described more than a decade ago. However, reports on the molecular mechanism that underlies presenilin FAD-dependent Ca²⁺ dyshomeostasis have been inconclusive and often contradictory. It has been suggested that changes in γ-secretase-mediated production of the AICD are responsible for aberrant Ca²⁺ signaling observed in cells expressing FAD mutant presenilin (42). However, Herms and colleagues (43) have demonstrated that CCE deficits associated with FAD mutant presenilins occur even in the absence of APP (and AICD). Our study identifies the MIC/TRPM7, a Ca²⁺-permeable nonselective cation channel, as a target ion channel that mediates the observed alterations in Ca²⁺ entry in either presenilin-deficient cells or cells expressing FAD mutant forms of the presenilins. Our studies also show that MIC channel dysfunction observed in presenilin FAD-expressing cells may be attributed to altered PIP₂ metabolism (Fig. 4). Interestingly, dysregulation of TRPM7 has been associated with at least two neurodegenerative conditions, including ischemia and Guamanian neurodegenerative disorders (27, 28). Thus, it is conceivable that altered MIC/TRPM7 activity may contribute to the neurodegenerative changes associated with FAD.

PIP₂ plays a major regulatory role at the plasma membrane as a precursor for several signaling molecules, e.g., diacylglycerol and IP₃. Moreover, localized membrane changes in PIP₂ itself are likely an important signal because PIP₂ is known to modulate a variety of channels and transporters (26). PIP₂ has been shown to play a role in a variety of cell functions, including ion channel modulation, rearrangement of cytoskeleton, and membrane protein trafficking (26). Intriguingly, cellular phenotypes in PIP₂-compromised cells are reminiscent of those observed in cells expressing FAD-associated presenilin mutations, e.g., ion channel and trafficking deficits (15). The potential role of presenilin in PIP₂ metabolism can be further evidenced by occurrence of PS1 in PIP₂-enriched subcellular compartments, including lipid rafts (44), phagocytic cups (19), lamellipodia (45), and adherent junctions (18). Thus, our study raises the possibility that PIP₂ may play a key role in the multiple cellular defects associated with presenilin FAD mutations. Cai et al. (46) recently reported that PS1 FAD is associated with aberrant phospholipase D (PLD) activity, and PLD supplementation reverses FAD phenotypes. Interestingly, PLD enzymatic activity is critically dependent on PIP₂, and the major product of PLD activity (PA) stimulates the production of PIP₂.

We also show that modulation of cellular PIP₂ levels correlates with changes in Aβ42 levels (Fig. 5). Modulation of APP processing by various membrane lipids has been extensively described (47): brain lipids are required to reconstitute fully the activity of the γ-secretase (38), and several single lipid components directly modulate γ-secretase activity (47). Thus, it is plausible that changes in PIP₂ levels give rise to a change in the composition of the lipid environment where the γ-secretase complex resides, thereby altering its proteolytic activity. Alternatively, PIP₂ may modulate the trafficking of the γ-secretase complex (or its individual components), thereby influencing the final activity of the complex. Recent reports demonstrate that the targeting of the γ-secretase complex into different subcellular sites has direct consequences on the functional properties of the γ-secretase enzyme (38, 48). Although the exact mechanism of the PIP₂ effect on γ-secretase activity remains to be elucidated, modulation of PIP₂ may represent an approach to control Aβ production. In addition, such modulation may allow inhibition of additional downstream signaling pathways that are involved in other secretase-independent cellular changes.

Materials and Methods

Electrophysiology.

Patch clamp experiments were conducted in the standard whole-cell recording configuration as described in ref. 49.

Lipid Kinase Assay and TLC.

In vitro lipid phosphorylation assays on postnuclear supernatant of HEK293, Neuro2a, or CHO cells expressing either WT or FAD mutant PS1 and PS2 were carried out as described in ref. 37.

Compound Treatment and APP/Aβ Analyses.

HeLa-APPsw or HEK293 cells transiently transfected with APP-C99myc were treated with either the PLC inhibitors edelfosine or miltefosine (10 μM), DMSO, or the PLC activator m-3m3FBS (15 μM) for 6 h. Conditioned media were collected, and cell debris were removed by centrifugation at 20,817 × g for 20 min. Aβ42 levels in conditioned media were determined by sandwich ELISA (Biosource International, Camarillo, CA) according to manufacturer's instructions. Total Aβ was detected by immunoprecipitating 500 μl of conditioned media with 6E10 (1:100) (Signet, Williston, VT) or 7N22 (1:500) (Biosource) overnight at 4°C. Samples were separated by using the NuPAGE gel system (Invitrogen, Carlsbad, CA), and Aβ peptides were detected by quantitative Western blotting with 6E10. Full-length APP was detected by Western blot analysis using 6E10 or APP-CTmax antibody. APP-CTmax was generated by immunizing rabbits with keyhole limpet hemocyanin-conjugated peptides corresponding to the C-terminal region of APP, (C)HLSKMQQNGYENPTYKFFEQMQN. An expression construct encoding Synj1-IPP was kindly provided by Pietro De Camilli (Yale University, New Haven, CT).

Abbreviations

Aβ

amyloid β-peptide

AD

Alzheimer's disease

AICD

amyloid precursor protein intracellular domain

APP

amyloid precursor protein

CRAC

calcium release-activated calcium

CCE

capacitative Ca²⁺ entry

dKO

double knockout

FAD

familial Alzheimer's disease

IP₃

inositol 1,4,5-trisphosphate

MEF

mouse embryonic fibroblast

MIC

Mg²⁺-inhibited cation

NSAIDs

nonsteroidal antiinflammatory drugs

PA

phosphatidic acid

PIP₂

phosphatidylinositol 4,5-bisphosphate

PLC

phospholipase C

PLD

phospholipase D

PS1

presenilin 1

PS2

presenilin 2

TRPM7

transient receptor potential melastatin 7.