CALHM1 ion channel elicits amyloid-β clearance by insulin-degrading enzyme in cell lines and in vivo in the mouse brain

Alzheimer's disease is a neurodegenerative brain disorder and the first cause of dementia in elderly individuals. The cerebral extraneuronal lesions called senile plaques represent a defining histopathological characteristic of Alzheimer's disease (Duyckaerts et al., 2009; Serrano-Pozo et al., 2011). Senile plaques are formed by the deposition of amyloid-β (Aβ), a series of peptides produced by sequential endoproteolysis of the amyloid precursor protein (APP) by the action of two proteases, β- and γ-secretases (De Strooper et al., 2010; Haass and Selkoe, 2007; Marambaud and Robakis, 2005). This histological feature, together with concordant genetic and biochemical evidence, support the hypothesis that Aβ is a causative factor in Alzheimer's disease pathogenesis (Hardy and Selkoe, 2002). Notably, most of the familial forms of Alzheimer's disease have been found to be caused by mutations in the genes coding for APP or presenilins (the γ-secretase catalytic components), and some of these pathogenic mutations directly increase Aβ production, favor an increase in the relative abundance of the more neurotoxic Aβ isoform Aβ42 or change Aβ aggregation propensity (Lambert and Amouyel, 2011). Furthermore, recent work has demonstrated the existence of a protective mutation in APP in a large Icelandic Alzheimer's disease cohort (Jonsson et al., 2012) that lowers both production and aggregation of Aβ (Benilova et al., 2014; Maloney et al., 2014). In this context, there is strong interest in identifying selective interventions aimed at preventing or lowering Aβ accumulation in the Alzheimer's disease brain (Citron, 2010).

The calcium homeostasis modulator protein 1 (CALHM1) gene (Dreses-Werringloer et al., 2008) was identified by using an expression profiling method (Aguilar et al., 2008), which screened for genes that: (1) were preferentially expressed in the hippocampus – a brain region affected at the early stages of Alzheimer's disease (de Leon et al., 2007) – and (2) were located on susceptibility loci for Alzheimer's disease. CALHM1 does not appear to be a risk gene for Alzheimer's disease, but independent genetic studies have shown that it influences the onset of Alzheimer's disease (Dreses-Werringloer et al., 2008; Lambert et al., 2010). CALHM1 has also been proposed to control Aβ levels in cell cultures (Dreses-Werringloer et al., 2008; Vingtdeux et al., 2014) and in human cerebrospinal fluid (CSF) (Kauwe et al., 2010; Koppel et al., 2011, but see also Giedraitis et al., 2010). Taken together, these data warrant further investigation aimed at determining whether CALHM1 is involved in molecular mechanisms relevant to Alzheimer's disease pathogenesis (Berridge, 2010).

CALHM1 is expressed in the human and mouse brains, and in hippocampal and cortical neurons (Dreses-Werringloer et al., 2008, 2013; Ma et al., 2012). CALHM1 function in brain neurons remains incompletely understood. Several studies have determined, however, that CALHM1 is a voltage-gated pore-forming protein that forms a plasma membrane Ca²⁺-permeable channel regulated by extracellular Ca²⁺ concentration (Dreses-Werringloer et al., 2008; Ma et al., 2012; Tanis et al., 2013). Its expression and activation in different cell systems induce Ca²⁺ influx and elevate cytoplasmic Ca²⁺ concentration in response to a drop in extracellular Ca²⁺ concentration (Dreses-Werringloer et al., 2008, 2013). In mouse cortical neurons, CALHM1 responds to extracellular Ca²⁺ not only by elevating intraneuronal Ca²⁺ levels, but also by controlling cell excitability (Ma et al., 2012). Work performed in CALHM1-transfected hippocampal HT-22 cells, as well as wild type (WT) and Calhm1-knockout (KO) primary neurons, have further demonstrated that CALHM1-mediated Ca²⁺ influx controls intracellular Ca²⁺ signal transduction (Dreses-Werringloer et al., 2013). Thus, CALHM1 plays a key role in cerebral neuronal Ca²⁺ homeostasis and regulates neuronal function.

In the present study, we show that CALHM1 controls Aβ metabolism, in cell cultures and in vivo in the mouse brain, by facilitating a neuronal and Ca²⁺-dependent mechanism of extracellular clearance of Aβ. Aβ steady-state levels in the brain are known to be the result of a balance between its production from APP and its degradation by efficient clearance mechanisms. Aβ clearance is controlled by various pathways, which include endocytosis by several brain cell types (including glia and neurons), protein transport and/or binding mechanisms across the blood–brain barrier, and extracellular enzymatic degradation (Pflanzner et al., 2010; Zolezzi et al., 2014). In the enzymatic pathway, Aβ can be degraded in vivo by at least four peptidases from the metallopeptidase family: neprilysin (NEP, also known as MME), endothelin-converting enzyme-1 and -2 (ECE-1 and ECE-2) and insulin-degrading enzyme (IDE) (Turner and Nalivaeva, 2007). Here, we also report that CALHM1 activation lowered extracellular Aβ levels by promoting Aβ clearance by IDE through a mechanism facilitating IDE secretion. This work reveals a previously unrecognized function of CALHM1 in neurons, and identifies the CALHM1 ion channel as an endogenous regulator of Aβ degradation in the brain.

CALHM1 permeability is regulated by the extracellular Ca²⁺ concentration ([Ca²⁺]_o) (Dreses-Werringloer et al., 2008; Ma et al., 2012). We initially reported that CALHM1 can be activated in cell cultures by a procedure that consists of transiently removing extracellular Ca²⁺ and subsequently adding it back at physiological concentration to create a driving force for Ca²⁺ entry into the cells (Dreses-Werringloer et al., 2008). This procedure, termed the ‘Ca²⁺ add-back’ condition, induces CALHM1-mediated Ca²⁺ influx and elevates intracellular Ca²⁺ concentration ([Ca²⁺]_i). An analysis of the relationship between the CALHM1-mediated increase of [Ca²⁺]_i and CALHM1 channel activity revealed that CALHM1 is strongly regulated by [Ca²⁺]_o and is activated by a drop in [Ca²⁺]_o, with a half-maximal activation concentration of ∼0.2 mM Ca²⁺ (Ma et al., 2012). Concordant studies have demonstrated that normal electrical activity in neurons can result in a robust drop of [Ca²⁺]_o (up to concentrations <0.1 mM) in the interstitium and synaptic clefts of the brain (Rusakov and Fine, 2003). Thus, neuronal activity reduces [Ca²⁺]_o within the range of CALHM1 activation (Ma et al., 2012). In this context, and in order to use a more physiologically relevant procedure of CALHM1 activation than the Ca²⁺ add-back condition, we tested the effect of directly incubating the cells in artificial cerebrospinal fluid (aCSF) buffer containing 0.2 mM Ca²⁺.

Using Fluo-4 [Ca²⁺]_i measurements, we found that lowering [Ca²⁺]_o from 1.4 mM (physiological concentration) to 0.2 mM (half-maximal CALHM1 activation concentration) led to a significant increase in [Ca²⁺]_i in CALHM1-transfected cells, compared to control cells (Fig. 1A,B), confirming that Ca²⁺ influx through CALHM1 was observed in these conditions. The effect of CALHM1 activation on [Ca²⁺]_i was observed ∼5 min after switching to 0.2 mM Ca²⁺ buffer, but only once [Ca²⁺]_i stabilized after the drop generated by the decrease in [Ca²⁺]_o (Fig. 1A). Importantly, application of 0.2 mM Ca²⁺ aCSF also resulted in a strong reduction of extracellular Aβ accumulation in cells transfected with either human or mouse CALHM1, compared to vector-transfected control cells (Fig. 1C). These results are in line with our previous findings showing that the Ca²⁺ add-back condition lowers Aβ accumulation in CALHM1-expressing cells (Dreses-Werringloer et al., 2008; Vingtdeux et al., 2014), and further demonstrate that the effect of CALHM1 on [Ca²⁺]_i parallels its inhibitory effect on Aβ accumulation.

We next asked whether the ion channel properties of CALHM1 are required for its inhibitory effect on Aβ levels. CALHM1 permeability can be modulated pharmacologically or genetically. No specific CALHM1 ion channel blocker has been identified yet. However, the non-specific inorganic dye Ruthenium Red or the ion Zn²⁺ can fully inhibit CALHM1 currents and CALHM1-mediated [Ca²⁺]_i increase (Dreses-Werringloer et al., 2013; Ma et al., 2012). Ruthenium Red and Zn²⁺, at concentrations inhibiting CALHM1 permeability (Dreses-Werringloer et al., 2013; Ma et al., 2012), blocked the effect of CALHM1 on Aβ accumulation (Fig. 1D,E).

CALHM1 is N-glycosylated at residue Asp140 to yield a second immunoreactive band on western blots, and mutation of Asp140 to an alanine residue (N140A) prevents both CALHM1 N-glycosylation and CALHM1-mediated Ca²⁺ influx (Dreses-Werringloer et al., 2008, 2013). Two conserved residues in the third hydrophobic domain of CALHM1, Trp114 (Dreses-Werringloer et al., 2013) and Asp121 (Ma et al., 2012), have also been identified as being potentially important for pore formation and/or ion permeation of the channel. Indeed, mutation of Trp114 to an alanine residue (W114A) and Asp121 to an arginine residue (D121R) have been found to fully block CALHM1-mediated Ca²⁺ influx (Dreses-Werringloer et al., 2013) and CALHM1 currents (Ma et al., 2012), respectively. Fig. 1F shows that, compared to WT CALHM1, the three functionally defective CALHM1 mutants, N140A, W114A and D121R, all failed to reduce Aβ levels. The lack of effect of these three mutants on Aβ accumulation could not be explained by defects in CALHM1 protein expression or stability (Fig. 1F).

We then determined whether extracellular Ca²⁺ is required for the effect of CALHM1 on Aβ levels. As expected, CALHM1-mediated Ca²⁺ influx in 0.2 mM Ca²⁺ aCSF could be inhibited by the chelating agent EGTA (2 mM, see Fig. 1G). In these conditions, the effect of CALHM1 activation on Aβ levels was also noticeably blocked (Fig. 1I). Addition of a saturating concentration of Ca²⁺ (5 mM) in EGTA-containing 0.2 mM Ca²⁺ aCSF restored the effect of CALHM1 on both [Ca²⁺]_i (Fig. 1G,H) and Aβ levels (Fig. 1I). Taken together, these results show that functional CALHM1 ion channels and Ca²⁺ influx through CALHM1 control Aβ levels.

The effect of CALHM1 on Aβ levels could be due to a decrease in Aβ production from APP or to an increase in intracellular or extracellular Aβ degradation (Leissring and Turner, 2013). We observed no reduction in intracellular Aβ levels upon conditions of CALHM1 expression and activation in 0.2 mM Ca²⁺ aCSF (Fig. 2A). Moreover, treatment with the lysosomotropic drugs NH₄Cl and chloroquine, which neutralize Aβ lysosomal degradation (Nixon et al., 2008; Vingtdeux et al., 2010), did not prevent the decrease of extracellular Aβ by CALHM1 (Fig. 2A).

We next assessed the effect of CALHM1 on the rate of Aβ production. To this end, we used the approach of transient block of protein transport through the secretory pathway by the reversible Golgi disassembling agent brefeldin A (BFA) (Langhans et al., 2007). As well as having well-established reversible effects on cellular trafficking, BFA has the advantage of specifically interrupting the export of newly synthesized proteins from the ER (Lippincott-Schwartz et al., 1989) without affecting protein synthesis, including APP synthesis (Knops et al., 1993). As we previously observed, BFA induced the accumulation of APP in the ER (Chapuis et al., 2011) and completely inhibited intracellular and extracellular Aβ accumulation (Fig. 2B,C). BFA washout allowed the reactivation of APP trafficking to progressively restore APP processing and Aβ production (Fig. 2B). We found no effect of CALHM1 expression and its activation on the rate of restoration of Aβ production, whereas, as expected, extracellular Aβ accumulation was strongly inhibited by CALHM1 activation (Fig. 2C). Thus, CALHM1 had no effect on either Aβ production or intracellular Aβ degradation by the lysosomes.

We then investigated whether CALHM1 promotes extracellular Aβ degradation and whether a membrane-associated or soluble proteolytic activity is involved in this mechanism. In a cell-free assay, we incubated (37°C) the conditioned medium of CALHM1-transfected cells (which were activated in 0.2 mM Ca²⁺ aCSF) with exogenous Aβ from the conditioned medium of APP-transfected cells (see Fig. 2D). In these conditions, we observed a substantial degradation of exogenous Aβ by the conditioned medium of cells expressing activated CALHM1, whereas the conditioned medium of non-transfected cells or cells transfected with a control vector had no effect on Aβ levels (Fig. 2E). These data show that upon CALHM1 activation, Aβ is targeted for degradation in the extracellular space by a soluble proteolytic activity.

Extracellular Aβ proteolytic clearance is mediated in vivo by at least four peptidases from the metallopeptidase family: NEP, ECE-1, ECE-2 and IDE (Turner and Nalivaeva, 2007). Treatment with phosphoramidon, which inhibits NEP and ECEs, or with thiorphan, which is more specific for NEP inhibition, failed to prevent the effect of CALHM1 on Aβ (Fig. 3A,D). In contrast, 1,10-phenanthroline (PNT), which inhibits IDE, or insulin, which acts as a potent competitive inhibitor for Aβ degradation by IDE, fully blocked Aβ degradation by CALHM1-transfected cells (Fig. 3B–D). The inhibitory effect of PNT and insulin on CALHM1-mediated Aβ degradation was also observed in vitro in the cell-free assay using the conditioned medium of CALHM1-transfected cells (Fig. 2E). Because Aβ can be degraded by PNT- and insulin-sensitive proteolytic activities independently of IDE (Espuny-Camacho et al., 2010), we verified the implication of IDE in this mechanism by directly targeting IDE expression using small interfering RNAs (siRNAs). We found that IDE expression reduction by RNA interference (RNAi) lowered the effect of CALHM1 on Aβ degradation (Fig. 3E), confirming that IDE mediates Aβ degradation upon CALHM1 activation.

Because our data incriminate extracellular soluble IDE in CALHM1-mediated Aβ degradation, we analyzed IDE levels in the conditioned medium of CALHM1-transfected cells. We found that CALHM1 transfection and its activation in 0.2 mM Ca²⁺ aCSF triggered a substantial secretion of IDE from different cell lines, such as N2a (expressing or not human APP) and HT-22 cells (Fig. 3F, see also Fig. 3E). Importantly, IDE secretion was fully blocked by conditions that inhibit CALHM1 permeability, such as Ruthenium Red and Zn²⁺ treatments (Fig. 1D,E) or expression of the functionally defective CALHM1 mutant W114A-CALHM1 (Fig. 3G). Taken together, these results show that CALHM1 ion channel promotes extracellular Aβ degradation by IDE through an increase in IDE secretion.

To determine whether CALHM1 controls cerebral IDE activity, we first assessed the in vitro degradation of exogenous Aβ and insulin by Calhm1 KO and WT brain homogenates. Calhm1-KO mice were recently generated in our laboratory and are viable and fertile (Dreses-Werringloer et al., 2013; Ma et al., 2012; Taruno et al., 2013). Histological analyses have revealed that Calhm1 KO mice display no defects in brain development or brain integrity in adulthood (P.M. and V.V. unpublished observations), which make them a relevant model to study CALHM1 functions in the brain.

IDE is secreted but can also be found as a membrane-bound enzyme outside the cell (Vekrellis et al., 2000). Soluble and membrane-associated brain homogenates were prepared from adult WT (Calhm1^+/+) and Calhm1-KO (Calhm1^−/−) mice, and were analyzed for synthetic Aβ and recombinant insulin degradation in an in vitro assay (see Materials and Methods). Strikingly, both soluble and membrane-associated brain homogenates contained significantly lower levels of Aβ degradation activity in CALHM1-deficient mice, compared to WT mice (Fig. 4A,B). Aβ degradation activity measured in this in vitro assay in the membrane-associated brain homogenates was almost entirely sensitive to insulin and PNT (Fig. 4C), indicating that the proteolytic activity directed against Aβ measured in this assay is mostly due to IDE. Insulin degradation activity in these homogenates was significantly inhibited by PNT, confirming that IDE is also the main proteolytic activity directed against insulin in our assay (Fig. 4D). In line with the results obtained for synthetic Aβ degradation, Calhm1-KO brain homogenates showed a clear trend of reduction in insulin degradation activity, compared to WT brain homogenates (Fig. 4E).

The APP intracellular domain (AICD) is the C-terminal fragment of APP produced by γ-secretase (Marambaud and Robakis, 2005). Previous studies have found that IDE targets AICD for degradation in cell cultures (Edbauer et al., 2002) and in vivo in mice (Farris et al., 2003; Miller et al., 2003). IDE subcellular localization is unconventional and the exact trafficking pathways leading to its presence both outside the cell and in the cytosol remain to be clearly defined. Using subcellular fractionation, we confirmed that intracellular IDE was mainly found as a soluble pool in the cytosol (Fig. 5A). This localization of IDE in the cytosol is consistent with its role in AICD degradation (Edbauer et al., 2002; Vekrellis et al., 2000).

AICD metabolism (production and degradation), like the metabolism of other protein fragments produced by γ-secretase, can be followed in vitro by incubating membranes isolated from cell or brain tissue homogenates at 37°C (Marambaud et al., 2003). Using this membrane assay, we preincubated brain homogenates in vitro for 2 h to produce the γ-secretase inhibitor-sensitive AICD fragment (Fig. 5B). AICD degradation was then assessed by incubating the homogenates for 5 h (Fig. 5B, 0, 2, 5 h). In these conditions, a decrease of AICD levels was readily detectable and the rate of AICD degradation could be quantified (Fig. 5B,C). Importantly, the decrease over time of AICD levels was fully protected by PNT treatment (Fig. 5B), confirming that IDE is the enzyme responsible for AICD degradation in this assay. Because no increase in AICD levels were observed over time in the presence of PNT, we concluded that AICD production after the 2 h preincubation reached saturation levels (Fig. 5B). Under these conditions, we found a significant decrease in AICD degradation activity in Calhm1-KO brain membranes, compared to membranes isolated from WT brains (Fig. 5B,C). Taken together, these data show that CALHM1 deficiency decreases brain IDE activity.

Αβ in the brain is mainly produced by neurons. IDE, however, is ubiquitously expressed and could potentially control extracellular Αβ levels by being released from cell types other than neurons. Because CALHM1 is expressed in cerebral neurons (Dreses-Werringloer et al., 2013; Ma et al., 2012), we asked whether CALHM1 deficiency affects the steady-state levels of endogenously produced Aβ in isolated whole forebrain primary neuron cultures. We found that Calhm1-KO neurons accumulated significantly more Aβ42 over time than neurons isolated from WT mice (Fig. 6A). The effect of CALHM1 deficiency on Aβ42 accumulation was abolished by the IDE competitive inhibitor insulin (Fig. 6B), indicating that the increase in Aβ42 in Calhm1-KO neurons was due to a deficit in IDE-mediated degradation. These data not only confirm that endogenous CALHM1 promotes Aβ degradation by IDE, but also reveal that this mechanism is controlled by neurons. To go further, we measured and quantified the steady-state levels of endogenous Aβ40 and Aβ42 in Calhm1-KO brains. CALHM1 deficiency resulted in a significant increase of Aβ40 and Aβ42 levels [51% increase of Aβ40 – WT, 0.6±0.04 pmol/g versus KO, 0.9±0.02 pmol/g, P=0.0005; 40% increase of Aβ42 – WT, 0.12±0.007 pmol/g versus KO, 0.17±0.009 pmol/g, P=0.0067 (mean±s.e.m., Student's t-test); Fig. 6C,D]. Furthermore, the total levels of AICD were increased in Calhm1-KO brain homogenates, compared to homogenates isolated from WT brains (Fig. 6E,F). Thus, CALHM1 deficiency increases endogenous Aβ and AICD levels in the mouse brain.

In the current study, we show that CALHM1 is a repressor of Aβ accumulation in cell lines and in vivo in the mouse brain. We determined that CALHM1 acted by promoting an extracellular mechanism of Aβ degradation. Using pharmacological and RNAi approaches, we identified IDE as the protease solely responsible for CALHM1-dependent Aβ degradation. These results were corroborated by an analysis of Calhm1-KO mice, which displayed a consistent decrease in cerebral IDE activity accompanied by a significant elevation of endogenous Aβ levels in both the whole brain and primary neurons. Additional investigation in cell cultures revealed that CALHM1 promoted IDE-mediated Aβ degradation by triggering IDE secretion.

The importance of these results is twofold. A main therapeutic strategy for drug development in Alzheimer's disease is focusing on approaches aimed at interfering with Aβ accumulation in the brain (Citron, 2010). Furthermore, ion channels, and, more specifically, cell surface ion channels that are more easily accessible, are targets of choice for drug development (Clare, 2010). In this context, our data not only reveal new insights into CALHM1 biology and its relevance to Alzheimer's disease pathogenesis, but also warrant further investigation aimed at determining whether anti-amyloidogenic CALHM1 activating interventions can be identified.

The localization and substrate accessibility of IDE is complex and its exact trafficking is only partially understood (Leissring and Turner, 2013). IDE can be found in the cytosol, where it degrades AICD, or outside the cell, where it regulates the catabolism of several soluble peptides (Turner and Nalivaeva, 2007). At the mechanistic level, our study identified a role for CALHM1 in IDE cellular secretion. By using a cell membrane fractionation method, we found a reduction in the activity of not only soluble IDE but also membrane-associated IDE in the Calhm1-KO brain. Our data therefore suggest that CALHM1 could also be involved in other mechanisms of IDE activation or substrate accessibility, not only extracellularly but also in the cytosol, as suggested by the observed effect of CALHM1 on AICD levels. Further studies will be needed to address this possibility and determine which mechanisms of IDE maturation or trafficking might also be modulated by CALHM1.

We further determined that CALHM1 ion channel and extracellular Ca²⁺ are required for the CALHM1 anti-amyloidogenic effect. Indeed, pharmacological and genetic approaches (using Ruthenium Red, Zn²⁺ or inactive CALHM1 mutants) revealed that CALHM1 promoted IDE-mediated Aβ clearance through its ion channel properties. Moreover, manipulation of [Ca²⁺]_o revealed that extracellular Ca²⁺ was required for the CALHM1 anti-amyloidogenic effect. These data strongly support the notion that Ca²⁺ influx through CALHM1 is the trigger for the observed effect on Aβ clearance by IDE. It is important to note, however, that CALHM1 has been reported to form a large pore that can allow significant permeabilities for other ions, such as Na⁺, K⁺ and Cl⁻, or for larger molecules, such as ATP (Ma et al., 2012; Siebert et al., 2013; Taruno et al., 2013). CALHM1 permeability activation is therefore expected to result in the control of multiple signaling pathways that could potentially also contribute to different mechanisms of IDE activation and to the observed effect on Aβ clearance. In this context, additional studies will have to determine whether Ca²⁺ influx through CALHM1 is the only trigger for Aβ clearance or whether other pathways controlled by CALHM1 permeability for different ions or molecules are also required.

In summary, this work shows for the first time that CALHM1 is an ion channel that has anti-amyloidogenic properties in vivo in the mammalian brain. Specifically, we found that CALHM1 promoted a neuronal and Ca²⁺-dependent mechanism of extracellular Aβ degradation by IDE. These data warrant further investigation aimed at determining whether CALHM1 ion channel activity can be pharmacologically targeted for promoting amyloid clearance in the Alzheimer's disease brain.

Ruthenium Red, ZnCl₂, NH₄Cl, chloroquine, 1,10-phenanthroline and insulin were purchased from Sigma-Aldrich. Phosphoramidon was from Cayman Chemical and thiorphan from Enzo Life Sciences. BFA was from Epicentre Technologies. Plasmids containing Myc- or V5-tagged WT and mutant (N140A and W114A) human and mouse CALHM1 cDNAs were as described previously (Dreses-Werringloer et al., 2008, 2013). V5-tagged D121R-CALHM1 mutant was created by using the QuikChange II site-directed mutagenesis kit (Agilent Technologies) and confirmed by sequencing. Anti-Aβ1-17 (6E10) and anti-Aβ17-24 (4G8) antibodies were from Covance, and anti-APP1–200 antibody (LN27) was from Zymed Laboratories Inc. Anti-APP C-terminal domain antibody (R1) was as described previously (Vingtdeux et al., 2010). Anti-IDE and anti-TSG101 antibodies were from Abcam, anti-Myc and the antibody against both MEK1 and MEK2 (also known as MAP2K1 and MAP2K2, respectively) was from Cell Signaling Technology, anti-V5 antibody was from Invitrogen, anti-actin antibody was from BD Transduction Laboratories, and anti-transferrin receptor antibody was from Invitrogen.

All animal experiments were performed according to procedures approved by the Feinstein Institute for Medical Research Institutional Animal Care and Use Committee. The Calhm1-KO mice were as described previously (Dreses-Werringloer et al., 2013; Taruno et al., 2013).

Naïve N2a cells were obtained from the ATCC. N2a cells stably transfected with human APP695 (APP-N2a) and HT-22 cells were described and maintained as reported previously (Dreses-Werringloer et al., 2008). Cells were transiently transfected in complete medium without antibiotics with the different CALHM1 cDNAs using Lipofectamine 2000 reagent (Invitrogen), as per the manufacturer's instructions. At 16 h post-transfection, cells were incubated for 1 h in aCSF buffer (in mM: 120 NaCl, 26 NaHCO₃, 1 NaH₂PO₄, 2.5 KCl, 1.3 MgSO₄, 10 D-Glucose, and 0.2 CaCl₂) to activate CALHM1. IDE RNAi was performed by transfection of IDE-targeting siRNA (Silencer Select, Ambion, Life Technologies) or negative control siRNA (Silencer Select, Ambion) for 24 h with Lipofectamine RNAiMAX reagent (Invitrogen). Primary neurons from day 17 embryos were prepared as previously described (Dreses-Werringloer et al., 2013; Vingtdeux et al., 2010).

[Ca²⁺]_i was measured using Fluo-4 (Fluo-4 NW Ca²⁺ Assay Kit, Life Technologies) as described previously (Dreses-Werringloer et al., 2013). Transiently transfected HT-22 cells were incubated for the indicated times in aCSF buffer containing 0.2 mM CaCl₂ and supplemented or not with 2 mM EGTA. In one experimental setting, Ca²⁺ concentration was restored in EGTA-supplemented aCSF buffer by adding 5 mM CaCl₂.

Cells were washed with phosphate-buffered saline (PBS) and solubilized in RIPA buffer (Millipore) supplemented with 1× Complete protease inhibitor mixture (Roche Applied Science). Depending on the primary antibody used, 5–20 μg of cell extracts were analyzed by SDS-PAGE. Conditioned medium containing secreted total Aβ, secreted (s)APPα and IDE was analyzed by direct western blots, as described previously (Vingtdeux et al., 2010). Intracellular Aβ levels were analyzed in cells harvested by trypsinization and by immunoprecipitation using the 4G8 antibody, as described previously (Vingtdeux et al., 2010).

Brains from 6-month-old Calhm1^+/+ and Calhm1^−/− mice were homogenized in 8 volumes (w/v) of 250 mM sucrose in 50 mM Tris-HCl pH 7.4. Nuclei were pelleted and the supernatants were centrifuged at 100,000 g for 1 h. Resulting supernatants were saved as soluble fractions and membrane pellets were resuspended in 100 mM Na₂CO₃, pH 11.3 and centrifuged at 100,000 g for 1 h. Pellets were resuspended in 50 mM Tris-HCl pH 7.4 and sonicated. Protein concentration in soluble and membrane fractions were evaluated using BCA assay. To quantify proteolysis of Aβ by brain fractions, 100 pM of synthetic human Aβ1-42 (Invitrogen) was incubated at 37°C with 100 µg/ml proteins of soluble or membrane fractions. At each time point, an aliquot of the sample was removed and stored at −80°C until analysis. To measure proteolysis of insulin, 100 pM of recombinant human insulin (Sigma) was incubated at 37°C with 50 µg/ml of proteins from the membrane fraction. ELISA was used to determine Aβ (Wako) and insulin (ALPCO Diagnostics) concentrations.

N2a cells were homogenized in Buffer A (10 mM NaCl, 1.5 mM MgCl₂, 2 mM DTT, 10 mM Tris-HCl pH 7.4) using 50 strokes of the pestle of a tight-fitting Dounce. Buffer B (525 mM mannitol, 175 mM sucrose, 2.5 mM EDTA, 12.5 mM Tris-HCl, pH 7.4) was then added at a 4:10 ratio to the cell homogenate in Buffer A. The resulting homogenate was centrifuged at 1300 g for 10 min to remove the nuclei. Supernatant was centrifuged at 20,000 g for 45 min to isolate a first fraction (denoted 20 K, enriched in the cell surface marker transferrin receptor). Supernatant was then centrifuged at 100,000 g for 3 h to isolate a second fraction (denoted 100 K, enriched in both transferrin receptor and the ESCRT-I subunit TSG101). The final supernatant contained the soluble cytosolic proteins and was enriched in the cytosolic markers MEK1 and MEK2. Intact vesicles from the 100 K fraction were trypsinized (0.25% trypsin for 10 min at 37°C) to degrade proteins exposed to the cytosol.

Cell membranes were obtained from brains of 6-month-old Calhm1^+/+ and Calhm1^−/− mice using the in vitro γ-secretase assay described previously (Marambaud et al., 2003). AICD was generated by incubation of the membrane preparations at 37°C for 2 h in a volume of 25 µl/assay. Samples were further incubated for 5 h to monitor AICD degradation. AICD levels were then analyzed by western blotting using the R1 antibody.

Endogenous Aβ1-40 and Aβ1-42 levels were assessed in total brain homogenates of 15-month-old Calhm1^+/+ and Calhm1^−/− littermate mice. Aβ1-40 and Aβ1-42 concentrations were determined by ELISA (Wako).

We thank Dr Luca Giliberto (The Feinstein Institute for Medical Research, Manhasset, New York 11030, USA) for critical reading of the manuscript.