Interactions between the amyloid precursor protein C-terminal domain and G proteins mediate calcium dysregulation and amyloid β toxicity in Alzheimer’s disease

Authors


E. Masliah, Department of Neurosciences, University of California, San Diego, La Jolla, CA 92093-0624, USA
Fax: +1 858 534 6232
Tel: +1 858 534 8992
E-mail: emasliah@ucsd.edu

Abstract

Alzheimer’s disease is characterized by neuropathological accumulations of amyloid β(1–42) [Aβ(1–42)], a cleavage product of the amyloid precursor protein (APP). Recent studies have highlighted the role of APP in Aβ-mediated toxicity and have implicated the G-protein system; however, the exact mechanisms underlying this pathway are as yet undetermined. In this context, we sought to investigate the role of calcium upregulation following APP-dependent, Aβ-mediated G-protein activation. Initial studies on the interaction between APP, Aβ and Go proteins demonstrated that the interaction between APP, specifically its C-terminal -YENPTY- region, and Go was reduced in the presence of Aβ. Cell death and calcium influx in Aβ-treated cells were shown to be APP dependent and to involve G-protein activation because these effects were blocked by use of the G-protein inhibitor, pertussis toxin. Collectively, these results highlight a role for the G-protein system in APP-dependent, Aβ-induced toxicity and calcium dysregulation. Analysis of the APP:Go interaction in human brain samples from Alzheimer’s disease patients at different stages of the disease revealed a decrease in the interaction, correlating with disease progression. Moreover, the reduced interaction between APP and Go was shown to correlate with an increase in membrane Aβ levels and G-protein activity, showing for first time that the APP:Go interaction is present in humans and is responsive to Aβ load. The results presented support a role for APP in Aβ-induced G-protein activation and suggest a mechanism by which basal APP binding to Go is reduced under pathological loads of Aβ, liberating Go and activating the G-protein system, which may in turn result in downstream effects including calcium dysregulation. These results also suggest that specific antagonists of G-protein activity may have a therapeutic relevance in Alzheimer’s disease.

Abbreviations
AD

Alzheimer’s disease

APP

amyloid precursor protein

amyloid β

ER

endoplasmic reticulum

G protein

guanine nucleotide-binding protein

GPCR

G-protein-coupled receptor

LDH

lactate dehydrogenase

PTX

pertussis toxin

Alzheimer’s disease (AD) is the most common neurodegenerative disorder in the elderly; it is characterized clinically by progressive cognitive decline and dementia, and neuropathologically by abnormal intracellular protein aggregates called neurofibrillary tangles and extracellular protein deposits known as amyloid plaques. Amyloid plaques are composed of amyloid β(1–42) [Aβ(1–42)], a cleavage product of the amyloid precursor protein (APP), and many studies have suggested that the abnormal deposition of Aβ may be causally linked to the pathogenesis of AD [1,2]. APP itself is reported to modulate Aβ-mediated toxicity [3–5], however, the precise mechanisms underlying APP-dependent Aβ toxicity remain a topic of intense research.

Recent work has highlighted the guanine nucleotide-binding protein (G protein) system as playing a key role in APP-dependent, Aβ-induced cell death [6]. The intracellular domain of APP (APP627-65) is reported to complex with and activate Go proteins [7,8], whereas G-protein inhibitors have been shown to block Aβ toxicity [6].

G proteins are a family of proteins involved in second messenger cascades and are important cellular signal-transducing molecules [9]. Heterotrimeric G proteins, composed of alpha, beta and gamma subunits, reside on the inner cell membrane surface, bound to G-protein-coupled receptors (GPCRs). Upon ligand binding, the GCPR undergoes a conformation change, causing it to release the alpha subunit of the G protein. Once activated, the free G-protein α subunit moves along the membrane and causes signal transduction throughout the cell [10]. A key role of the G-protein system is the regulation of intracellular calcium levels via receptors on the endoplasmic reticulum (ER) and the plasma membrane [11–14].

It has been reported that G-protein-associated signaling pathways are disrupted in AD post-mortem brains [15]. This disruption has been linked to the altered coupling of G proteins to GPCRs [16], or to altered levels of G proteins in different regions of AD brains, such as the frontal cortex and hippocampus [17]. Intracellular calcium levels, themselves regulated by G proteins, are also altered in AD [18–20].

Given the evidence implicating the G-protein system in APP-dependent, Aβ-mediated toxicity [6,7,21,22], and its central role in cellular calcium regulation [10,23,24], this study sought to investigate the role of calcium upregulation in APP-dependent Aβ toxicity in AD.

We demonstrate that, in neuronal cultures, Aβ is able to reduce the interaction between APP and Go, which in turn results in G-protein activation-dependent calcium dysregulation and subsequent cell death. These results were shown to be clinically relevant because immunoprecipitation analysis of the frontal cortex of patients at differing Braak stages of AD revealed a progressive decrease in the interaction between APP and Go, which was associated with an increase in membrane Aβ levels and G-protein activity.

Results

Aβ modulation of the interaction between APP and Go in neuronal cultures

In order to investigate the mechanisms underlying APP-dependent Aβ toxicity and the involvement of the G-protein system, APP-deficient B103 cells were transiently transfected with full-length APP (APP695) and then incubated with Aβ (10 μm) for 24 h. Cells were lysed, and the membrane fraction of the cell homogenate was isolated and immunoprecipitated with the APP-specific C-terminal G369 antibody. Immunoblot analysis of immunoprecipitated protein with an antibody for Go revealed that an interaction between APP695 and Go could be demonstrated in cells transfected with APP695 (Fig. 1A). There was no interaction in untransfected APP-deficient B103 cells, which served as a control in this and subsequent experiments. Aβ treatment caused a decrease in the interaction between APP and Go (Fig. 1A). Aβ treatment did not affect Go levels or the APP total expression levels (Fig. 1B), which suggests that the reduced interaction of APP and Go seen in the presence of Aβ is not simply caused by a reduction in their respective protein levels.

Figure 1.

 Aβ modulates the interaction between APP and Go in neuronal cultures. B103 rat neuroblastoma cells were transiently transfected with APP695 (A) or the C-terminal fragment comprised of the terminal 99 amino acids of APP (C99) (D), and subsequently treated with Aβ (10 μm, 24 h). Cell homogenates were purified to obtain membrane fractions which then were immunoprecipitated with the APP C-terminal-specific G369 antibody, run on a SDS/PAGE gel and Go protein binding was assessed by immunoblotting with a Go mAb. (A) Immunoprecipitation studies show a 40 kDa band representing the interaction between APP695 and Go in cells transfected with APP695. (B) Western blot analysis with the Go antibody (upper) demonstrates that transfection with APP695 and/or treatment with Aβ does not affect expression levels of Go or APP. (C) Similar immunoprecipitation studies in cells transfected with C99 show an interaction between the C99 region of APP and Go. As with full-length APP, this interaction was reduced in cells treated with Aβ. (D) Western blot analysis with the G369 antibody demonstrates that the expression level of APP695 is not affected by 24 h treatment with Aβ (10 μm), PTX (100 ng·mL−1) or a combination of the two.

Aβ has previously been reported to interact directly with APP and a number of interaction sites have been described; one reported site maps to the N-terminus of APP [3], another to the C-terminal end of APP, specifically to a peptide composed of the C-terminal 99 amino acids of APP (C99) [5]. It has been proposed that C99 may maintain Aβ-induced toxicity similar to full-length APP.

In order to examine this issue further, we investigated whether the C99 C-terminal fragment was capable of interacting with Go in a manner similar to full-length APP. To this end, B103 cells were transiently transfected with C99 and similarly treated with 10 μm Aβ for 24 h, and then underwent the immunoprecipitation procedure described above. In this case, similar to cells transfected with full-length APP, cells transfected with C99 also demonstrated an interaction with Go, which was absent in the untransfected controls (Fig. 1C). Aβ treatment was also able to modulate the interaction between C99 and Go, with cells treated with Aβ showing a decreased interaction between C99 and Go in comparison with Aβ-untreated cells (Fig. 1C).

Immunocytochemistry was conducted to verify the distribution of Go protein immunoreactivity in the presence and absence of Aβ (Fig. 2). In the absence of Aβ, Go immunoreactivity colocalizes with APP immunoreactivity in the cytoplasm of APP-transfected B103 cells (Fig. 2G–I). In the presence of Aβ, however, this colocalization is disturbed such that the addition of Aβ appears to reduce APP:Go colocalization/interaction (Fig. 2J–L, compare Fig. 2I,L).

Figure 2.

 Colocalization of APP:Go immunoreactivity in APP-transfected neuronal cultures. Immunoreactivity for Go and APP in untreated control B103 cells (A, B), Aβ-treated B103 cells (D, E), APP-transfected B103 cells (G, H) and Aβ-treated APP-transfected B103 cells (J, K), and colocalization of signals in cells under the different conditions (in C, F, I and L respectively). Scale bar = 20 μm.

G-protein activation contributes to Aβ-induced cell death in neuronal cultures

In order to investigate whether Aβ-mediated G-protein activation may impact Aβ-induced cell death, B103 cells were transiently transfected with APP695 and treated with Aβ (10 μm), the G-protein inhibitor pertussis toxin (PTX) (100 ng·mL−1) or a combination of the two, for 24 h. PTX prevents the two active components of the G protein, the α subunit and the βγ fragment, from separating and, in so doing, prevents the G-protein mechanism from reaching an active state. PTX administration had no effect on levels of APP in transfected or untransfected B103 cells (Fig. 1D). Following treatment, cell death was assessed using a lactate dehydrogenase (LDH) assay. Results from the LDH assay demonstrated that the effect of Aβ toxicity on untransfected B103 cells (which lack APP) was minimal, however, when Aβ treatment followed transfection with APP695 a significant increase in cell death was observed (Fig. 3A). PTX administration significantly reduced this APP-dependent, Aβ-induced cell death (Fig. 3A). These results support the involvement of the G-protein system in APP-dependent, Aβ-induced toxicity. The effects of APP and Aβ on neurite outgrowth were also assessed as an additional measure of cell viability (Fig. 3B). The presence of APP had a minimal effect on neurite outgrowth in B103 cells, however, Aβ treatment of APP-transfected cells resulted in a significant decrease in neurite outgrowth (Fig. 3B) in comparison with Aβ-treated untransfected B103 cells. PTX administration abrogated this reduction in neurite length when given in conjunction with Aβ treatment (Fig. 3B).

Figure 3.

 APP695-dependent, Aβ-induced cell death requires G-protein activation in neuronal cultures. B103 rat neuroblastoma cells were transiently transfected with APP695 and subsequently treated for 24 h with vehicle, Aβ (10 μm), PTX (100 ng·mL−1) or a combination of Aβ and PTX. Following treatment, cell death was analyzed via the LDH cell death assay. (A) Minimal cell death was observed in B103 cells with or without APP695 transfection, treatment with Aβ significantly increased cell death in APP695-transfected cells (*). Treatment with PTX alone had no significant effect upon cell death in cells with or without APP695 transfection, however, when administered in combination with Aβ, PTX significantly reduced the cell death associated with Aβ treatment of APP695-transfected cells (**). *Significant difference between vehicle- and Aβ-treated APP695-transfected cells (P < 0.05, by one-way ANOVA and post-hoc Fisher). **Significant difference between Aβ/PTX- and Aβ-treated APP695-transfected cells (P < 0.05, by one-way ANOVA and post-hoc Fisher). (B) Neurite outgrowth was assessed in cells transfected with APP695 and treated with Aβ. There was a significant reduction in neurite length in Aβ-treated, APP695-transfected cells which was abrogated upon the addition on PTX. *Significant difference (P < 0.05, by one-way ANOVA and post-hoc Fisher). (C) B103 rat neuroblastoma cells were transiently transfected with APP695, APP695ΔY or APP695ΔC31 and subsequently treated for 24 h with Aβ (10 μm). Following treatment, cell death was analyzed via the LDH cell death assay. APP-dependent, Aβ-induced cell death was clearly observed in cells transfected with APP695, whereas cells transfected with the APP695ΔY or APP695ΔC31 constructs failed to display significant cell death upon Aβ treatment in comparison with untransfected cells. *Significant difference between vehicle- and Aβ-treated APP695-transfected cells (P < 0.05, by one-way ANOVA and post-hoc Fisher). **Significant difference between Aβ-treated APP695-transfected cells and Aβ-treated APP695ΔY- and APP695ΔC31-transfected cells (P < 0.05, by one-way ANOVA and post-hoc Fisher). (D) Neurite outgrowth was assessed in cells transfected with APP695, APP695ΔY or APP695ΔC31, and subsequently treated for 24 h with Aβ (10μm). There was a significant reduction in neurite length in Aβ-treated APP695-transfected cells which was not evident in cells transfected with the APP695ΔY or APP695ΔC31 constructs. *Significant difference (P < 0.05, by one-way ANOVA and post-hoc Fisher).

Aβ modulation of the interaction of Go:APP requires the C31 domain

It has previously been reported that a crucial event in APP-dependent, Aβ-induced toxicity is the caspase cleavage of full-length APP inside its cytoplasmic region to release the toxic C31 fragment [4]. The C31 peptide has been shown to cause apoptosis in vitro and has been described in brain samples from AD patients, but not in control samples [25].

In order to investigate the identity of APP domains involved in the interaction between Aβ, APP and Go, we examined the importance of the C31 domain and of regions within it. The -YENPTY- region in the C31 fragment was of particular interest because this domain has been previously reported to bind cytoplasmic proteins, including Fe65, X11/Mint1 and Jip-1b [26–28]. Most importantly, it has already been demonstrated that the APP-dependent component of Aβ toxicity was abrogated when this domain was deleted from APP [5].

B103 cells were transiently transfected with full-length APP (APP695), a construct in which the -YENPTY- region had been deleted (APP695ΔY) or a construct lacking the complete C31 fragment (APP695ΔC31). As described previously, cells were treated with Aβ (10 μm, 24 h), cell homogenates were collected, subjected to immunoprecipitation using the APP G369 antibody and blotted for Go proteins using the Go mAb. As described previously (Fig. 1A), cells transfected with APP695 displayed binding to Go which was absent in untransfected cells; this binding was largely absent in cells that had been treated with Aβ (Fig. 4A). Cells transfected with APP695ΔY or APP695ΔC31 also displayed an interaction with Go, however, unlike the case with full-length APP, this binding was not attenuated upon Aβ treatment (Fig. 4A). These findings suggest that the C31 domain is essential for Aβ-responsive modulation of the APP:Go interaction.

Figure 4.

 Aβ modulation of the APP:Go interaction requires an intact APP C-terminal domain. B103 rat neuroblastoma cells were transiently transfected with full-length APP (APP695), APP695 lacking the YENPTY domain (APP695ΔY) or APP695 lacking the C-terminal C31 domain (APP695ΔC31), and subsequently treated with Aβ (10 μm, 24 h). (A) For immunoprecipitation studies, cell homogenates were immunoprecipitated with the APP C-terminal-specific G369 antibody, run on a SDS/PAGE gel, and Go protein binding was assessed by immunoblotting with a Go mAb. The interaction between APP695 and Go (40 kDa) is reduced in cells that have been treated with Aβ. This interaction is present in cells transfected with APP695ΔY and APP695ΔC31, however, in these cells, Aβ failed to reduce the level of interaction observed. (B) Western blot analysis with G369 antibody demonstrates that treatment with Aβ did not affect the expression levels of APP695, APP695ΔY or APP695ΔC31. (C) B103 rat neuroblastoma cells were transiently transfected with APP695 and subsequently treated for 24 h with vehicle, Aβ (10 μm), PTX (100 ng·mL−1) or a combination of Aβ and PTX. Following treatment, levels of intracellular calcium were analyzed using the FLIPR calcium assay. Aβ treatment of APP695-transfected cells resulted in a significant increase in intracellular calcium levels compared with vehicle-treated APP695-transfected cells (*). Treatment with PTX alone had no significant effect upon calcium influx in cells with or without APP695 transfection, however, when administered in combination with Aβ, PTX significantly reduced the calcium influx associated with Aβ treatment of APP695-transfected cells (**). *Significant difference between vehicle- and Aβ-treated APP695-transfected cells (P < 0.05, by one-way ANOVA and post-hoc Fisher). **Significant difference between Aβ/PTX- and Aβ-treated APP695-transfected cells (P < 0.05, by one-way ANOVA and post-hoc Fisher). (D) B103 rat neuroblastoma cells were transiently transfected with APP695, APP695ΔY or APP695ΔC31, and subsequently treated for 24 h with Aβ (10 μm). Following treatment, intracellular calcium levels were analyzed via a FLIRP calcium assay. APP-dependent, Aβ-induced calcium influx was clearly observed in cells transfected with APP695, whereas cells transfected with the APP695ΔY or APP695ΔC31 constructs failed to display significant calcium influx upon Aβ treatment in comparison with APP695-transfected cells. *Significant difference between vehicle- and Aβ-treated APP695-transfected cells (P < 0.05, by one-way ANOVA and post-hoc Fisher). **Significant difference between Aβ-treated APP695-transfected cells and Aβ-treated APP695ΔY- and APP695ΔC31-transfected cells (P < 0.05, by one-way ANOVA and post-hoc Fisher).

In order to investigate the importance of APP C-terminal domains in APP-dependent, Aβ-induced cell death, B103 cells were transfected with APP695ΔY or APP695ΔC31 and treated with Aβ as before. Following treatment, cell death was assessed using a LDH toxicity assay. Consistent with previous results, the presence of full-length APP appeared to promote Aβ toxicity and cell death (Fig. 3C), however, Aβ treatment failed to induce cell death in cells transfected with the APP695ΔC31 or APP695ΔY constructs. Assessment of neurite outgrowth revealed that Aβ treatment of APP695-transfected B103 cells resulted in a significant reduction in neurite length, in comparison with cells transfected with the APP695ΔC31 or APP695ΔY constructs (Fig. 3D). These results are consistent with previous findings identifying the C31 domain, specifically the -YENPTY- region, as critical for Aβ- mediated toxicity [4,5,29,30].

APP-dependent Aβ toxicity is mediated by calcium influx following G-protein activation

Calcium influx has previously been shown to increase following Aβ treatment in vitro [31]. This effect was explained by many different mechanisms, including the formation of calcium channels in the cell membrane by Aβ itself [31]. In order to ascertain whether the activation of G proteins led to changes in intracellular calcium levels, we conducted a calcium assay to measure intracellular calcium and its modulation by G-protein activation following Aβ treatment.

Cells were transiently transfected with APP695 followed by treatment with Aβ (10 μm), PTX (100 ng·mL−1) or a combination of the two, for 24 h. Following treatment, cells were applied to a calcium influx assay to measure intracellular calcium levels.

APP695-transfected cells treated with Aβ showed a significantly increased calcium influx in comparison with vehicle-treated APP695-transfected cells or uninfected cells treated with Aβ (Fig. 4C). PTX administration together with Aβ treatment significantly reduced the Aβ-induced calcium influx in APP-transfected cells (Fig. 4C), highlighting the importance of G-protein activation for APP-dependent, Aβ-induced calcium influx.

In order to further explore the contribution of various domains of APP to calcium influx modulation by Aβ, cells were transfected with APP695, APP695ΔY or APP695ΔC31 and treated with Aβ as before. Following 24-h incubation, calcium influx into the cells was measured (Fig. 4D). Transfection with APP695ΔY or APP695ΔC31 mutants failed to promote calcium influx into the cell following Aβ treatment compared with cells transfected with the APP695 construct (Fig. 4D). These results highlight the importance of the C31 domain, particularly the -YENPTY- region, in the ability of APP to invoke G-protein-modulated calcium influx.

Caspase cleavage of APP is required for Aβ modulation of APP and Go interaction

It has previously been reported that Aβ-mediated toxicity initiates a cascade of events that includes caspase activation, and APP caspase cleavage at residue 664 (APP695 numbering) generates the potentially cytotoxic peptide C31 [4]. This cleavage event is reported to be crucial for APP-mediated Aβ toxicity [4,5,29,30]. In order to investigate the impact of APP caspase cleavage on the interaction between APP and G proteins, we generated a construct of APP695 containing a point mutation within the consensus caspase cleavage site – blocking caspase cleavage at this site (APP695D664A) and the subsequent production of the C31 fragment. This mutation has been shown to significantly reduce Aβ-induced neuronal degeneration and cognitive impairment in transgenic mouse models of AD [29,32].

B103 cells were transfected with APP695 or APP695D664A and treated with Aβ, as in previous experiments. Following Aβ treatment, cells were homogenized and underwent immunoprecipitation with G369 antibody against APP and were blotted with the antibody against Go. By contrast to the results seen with APP695 (Figs 1 and 5A), Aβ failed to reduce the interaction between the APP695D664A construct and Go (Fig. 5A). Transfection with APP695 or APP695D664A and Aβ treatment had no effect on total levels of Go or APP (Fig. 5B).

Figure 5.

 The interaction between APP695 and Go requires APP caspase cleavage. B103 rat neuroblastoma cells were transiently transfected with APP695 or with APP695 with a substitution of alanine for aspartic acid at position 664 (APP695D664A). Cells were subsequently treated with Aβ (10 μm, 24 h). (A) For immunoprecipitation studies, cell homogenates were immunoprecipitated with the APP C-terminal-specific G369 antibody, run on a SDS/PAGE gel and Go protein binding assessed by immunoblot with a Go mAb. The interaction between APP695 and Go (40 kDa) is reduced in cells that have been treated with Aβ. This interaction is also present in cells transfected with APP695D664, however, Aβ treatment of APP695D664-transfected cells failed to reduce the band signal intensity. (B) Western blot analysis with the Go antibody demonstrates that transfection with APP695 or APP695D664 and/or treatment with Aβ does not affect expression levels of Go. Similar analysis with the G369 antibody demonstrates that Aβ treatment does not affect expression levels of APP695 or APP695D664. (C) B103 rat neuroblastoma cells transiently transfected with APP695 or APP695D664 and subsequently treated for 24 h with Aβ (10 μm) were analyzed via the LDH cell death assay to obtain a measure of APP695-dependent, Aβ-induced cell death. Minimal cell death was observed in B103 cells with or without APP695 transfection, whereas treatment with Aβ significantly increased cell death in APP695-transfected cells (*). Aβ failed to induce significant levels of cell death in cells transfected with APP695D664 (**). *Significant difference between Aβ-treated APP695-transfected cells and Aβ-treated untransfected cells (P < 0.05, by one-way ANOVA and post-hoc Fisher). **Significant difference between Aβ-treated APP695D664A-transfected cells and Aβ-treated APP695-transfected cells (P < 0.05, by one-way ANOVA and post-hoc Fisher). (D) Measurements of intracellular calcium demonstrate a significant increase in calcium influx in cells transfected with APP695 upon Aβ treatment in comparison with Aβ-treated untransfected cells (*). Aβ treatment of APP695D664A-transfected cells failed to elicit a significant increase in intracellular calcium levels in comparison with Aβ-treated untransfected cells (**). *Significant difference between Aβ-treated APP695-transfected cells and Aβ-treated untransfected cells (P < 0.05, by one-way ANOVA and post-hoc Fisher). **Significant difference between Aβ-treated APP695D664A-transfected cells and Aβ-treated APP695-transfected cells (P < 0.05, by one-way ANOVA and post-hoc Fisher).

LDH and calcium influx assays were also conducted on APP695D664A-transfected cells. As with the other mutant APP constructs examined, mutating the caspase cleavage site of APP reduced cell death observed upon Aβ administration (Fig. 5C) and the G-protein activation mediated calcium influx by Aβ into cells (Fig. 5D).

These results are consistent with and complement previous studies reporting that the caspase cleavage of APP is critical for the Aβ-induced cell death pathway [5,29].

Decreased interaction between APP and Go in human AD brains

Many reports have described disturbances in G-protein signaling in AD post-mortem brains [15], and it has been proposed that this disruption is caused by an altered coupling of G proteins to their receptors or by altered levels of G proteins in regions such as the hippocampus or frontal cortex in AD brains [16,17].

In order to further investigate G-protein characteristics and APP:Go binding in human tissue, brain samples from AD cases at different pathological stages of AD and age-matched non-dementia controls were obtained.

As expected, initial immunoblot characterization of purified membrane fractions (the fraction of the cell homogenate enriched with Aβ, APP and G proteins) revealed that disease progression was accompanied by a decrease in immunoreactivity for the neurodegenerative marker synaptophysin (Fig. 6A). APP protein levels in the membrane remained constant throughout the course of the disease (Fig. 6B), as did total levels of Go (Fig. 6C). As expected, levels of Aβ increased significantly with increasing disease severity (Fig. 6D).

Figure 6.

 Characterization of AD-related marker levels in human AD brain homogenates. Membrane fractions were obtained from human brains at various Braak stages of AD (I–II, n = 10; III–IV, n = 10; V–VI, n = 10) and nine non-dementia age-matched control brains (designated Braak stage 0). We used 10 μg of each sample for immunoblot analysis of various AD-related proteins. In each case, signal intensity was measured (mean ± SEM). (A) Levels of synaptophysin immunoreactivity show a decrease upon disease progression. (B) Levels of APP immunoreactivity remain stable across the different Braak stages. (C) Levels of Go immunoreactivity remain stable across the different Braak stages. (D) Levels of Aβ immunoreactivity increase as the disease progresses. *Significant difference between expression levels at Braak stage 0 and Braak stage V–VI (P < 0.05, by one-way ANOVA and post-hoc Fisher).

In order to investigate interactions between APP and G proteins in human AD brain, purified brain membranes from each of the series were immunoprecipitated for APP and recovered for Go. Our results demonstrated a decreasing interaction between APP and Go as the disease progressed (Fig. 7A). Reciprocal immunoprecipitation with the Go antibody and detection with the CT15 antibody revealed identical results; each immunoprecipitation was repeated twice with consistent results each time (data not shown). These results imply that as AD progresses and the pathological load of Aβ increases, the binding between APP and Go becomes increasing impaired, and at later stages of the disease binding may be absent altogether.

Figure 7.

 The APP:Go interaction and G-protein activation is modulated during AD progression. (A) Brain samples were homogenized and immunoprecipitated with the APP C-terminal specific CT15 antibody, run on a SDS/PAGE gel and Go protein binding was assessed by immunoblot with a Go mAb. Signal intensity of the 40 kDa band (corresponding to the interaction between APP and Go) was analyzed for each sample at each Braak stage. A decrease in the signal intensity of the 40 kDa band was observed with increasing disease severity (mean ± SEM). *Significant difference between expression levels at Braak stage 0 and Braak stage V–VI (P < 0.05, by one-way ANOVA and post-hoc Fisher). (B) In order to assess the activation capability of G proteins in AD, G-protein activation at different Braak stages was assessed using the classic G-protein activator norepinephrine (10 μm). Results from this assay demonstrate that G proteins in purified membrane fractions from human frontal cortex from each Braak stage are equally capable of being activated by norepinephrine (NE, 10 μm). (C) G-protein activation was assessed at different Braak stages and showed that G-protein activation increases with increasing disease severity. *Significant difference between activity levels at Braak stages I–II, III–IV and V–VI, and Braak stage 0 (P < 0.05, by one-way ANOVA and post-hoc Fisher).

It is possible that the decrease in APP:Go may be a result of global neurodegeneration in the brain and a ubiquitous decrease in protein–protein interactions, however, this would not explain our observation of an increase in Aβ levels during the disease (Fig. 6D). Because this dot blot analysis was performed on membrane fractions, this is not free Aβ, but is bound to APP and extracted with the membrane fractions. Given the increase in APP:Aβ interactions, we believe that the decrease in APP:Go interaction during disease progression is not a mere epiphenomena, but rather a specific trigger for G-protein activation.

Increased G-protein activation in human AD brains

In order to determine whether the decreased binding of APP and Go at later stages of AD and the subsequent liberation of Go had any effect on basal G-protein activation levels, we induced G-protein activation in membranes purified from the brain samples using norepinephrine (10 μm), a classic G-protein activator [33]. All membrane fractions displayed similar levels of GTP binding (Fig. 7B), indicating that the internal characteristics of G-protein activation remained the same throughout the disease.

Next, we sought to investigate G-protein activation levels in AD brains during progression of the disease. To this end, G-protein activation assays were conducted on purified membrane fractions from AD brains at different Braak stages and from non-dementia, age-matched controls. Results from these assays demonstrated a significant increase in the level of G-protein activation correlating with advancing Braak stage (Fig. 7C), suggesting that G-protein activation levels increased as the disease progressed. Because the inherent ‘excitability’ of the G proteins remained the same in AD and normal samples (Fig. 7B), the only parameter that correlated with the increased activation levels seen in AD was the disease-related increase in Aβ, we therefore propose that the increase in G-protein activation is caused by increased levels of Aβ, rather than by any inherent change in activation mechanisms.

Discussion

Previous studies have shown that calcium dysregulation plays a role in AD and that APP-dependent Aβ toxicity leads to cell death; this study is the first to link these two processes by demonstrating that calcium dysregulation occurs following G-protein activation resulting from the interaction between Aβ and the C-terminus of APP.

We show that interactions between APP and Go are attenuated upon treatment with Aβ, resulting in APP-dependent, Aβ-induced cell death, which was significantly abrogated upon treatment with PTX, a G-protein inhibitor. These results highlight the role of the G-protein system in the mechanism underlying APP-dependent, Aβ-induced toxicity, and are consistent with number of studies suggesting that the activation of G proteins may mediate APP-dependent Aβ toxicity in AD [34–36]. In order to pinpoint specific domains in APP involved in the binding to Go and the responsiveness to Aβ, we transfected B103 cells, which themselves lack endogenous APP, with constructs containing full-length APP (APP695) or APP lacking either the entire C31 domain (APP695ΔC31) or the C-terminal -YENPTY- region (amino acids 681–687, APP695ΔY). Although all constructs are able to bind Go, the presence of the C31 domain, more specifically the -YENPTY- region, is necessary for the responsiveness to Aβ. These results are consistent with a recent study investigating the regions of APP important in Go binding, which also found that APP binding to Go could be demonstrated in rat hippocampal cells transfected with an APP construct lacking the YENPTY-region [6]. These results are also consistent with studies that have linked the C-terminal fragment of APP to calcium dysregulation [37]. However, unlike the study by Sola Vigo et al. [6], which demonstrated Aβ toxicity in hippocampal cells transfected with the APP construct lacking the -YENPTY- region, we did not observe significant cell death upon Aβ treatment of B103 cells transfected with APP695ΔY (or APP695ΔC31). It is possible that in the aforementioned study, in which transfections were conducted into primary cells, enough endogenous APP containing the C31/YENPTY region was present to cause toxicity, whereas in B103 cells, the lack of endogenous APP meant that the effect of C31 or -YENPTY- deletion was not masked by the presence of endogenous APP. The lack of cell toxicity upon Aβ treatment in these cells demonstrated the need for an intact C31/ YENPTY region for APP-dependent Aβ-induced cell toxicity. The study by Sola Vigo et al. [6] also demonstrated a lack of toxicity in rat hippocampal cells transfected with constructs containing the -YENPTY- region (but lacking other regions of the intracellular domain of APP); although this discrepancy is puzzling, it may suggest that multiple sites on APP are involved in G-protein binding, and Aβ-induced cell death.

Our results highlight the importance of the C-terminal region of APP for the interaction between APP and Go, because without it there is no domain for the G protein to bind to. We indicate that the C31 region, in particular the YENPTY domain, is crucial for release of the Go subunit from intracellular APP. The site of direct interaction of APP with G protein has previously been mapped to the sequence 657–674 (APP695 numbering) [7,21]. We further extended these findings by adding the importance of the YENPTY domain. Although the YENPTY domain is outside the binding region 657–674, it can still regulate the binding of Go to this region. In this way, we combine the data on the importance of G proteins in Aβ-induced cell death with other data such as the importance of the C31 region, in particular the YENPTY area, in Aβ-induced toxicity [5,29,32]. The mechanism we propose links APP:Aβ extracellular binding leading to APP homodimerization, followed by caspase recruitment and activation, to cleave C31 [4]. Once C31 is removed, the YENPTY region, a domain within this C31, is actually cleaved off, thereby regulating the binding of Go to APP, and providing a possible illustration of how C31 cleavage activates G-protein signaling-dependent calcium influx and subsequent cell death.

Aβ-induced G-protein activation may be an early step in Aβ toxicity, and activation of the G-protein system may produce many of the observed downstream cellular responses to Aβ. A major cellular function of the G-protein system is the regulation of intracellular calcium levels [10,23,24] and a number of neuronal functions are highly dependent on calcium signaling, the intensity of which is tightly regulated both temporally and spatially throughout the cell [38].

Calcium has been linked to AD, with the calcium hypothesis of AD, first proposed over 10 years ago, suggesting that sustained intracellular calcium disturbances may underlie many neurodegenerative disorders, including AD [39]. This theory was supported by studies demonstrating that alterations in calcium signaling were a feature of both sporadic and familial AD [40,41]. A subsequent study [42] examined the effect of APP and its cleavage product Aβ on calcium regulation. Leissring et al. [42] found that APP−/− cells displayed a reduced release of ER calcium, which could be rescued only upon overexpression of APP fragments containing the intracellular domain of APP and not upon expression of a fragment lacking the intracellular domain of APP. These results highlighted the importance of this region in the regulation of calcium signaling.

Aβ oligomers have been shown to rapidly elevate intracellular calcium levels in human neuroblastoma cells and to increase membrane permeability and conductance [43,44]. Aβ aggregates have also been reported to cause calcium release from ER stores, by both inositol triphosphate and ryanodine receptors [45–47]. Increased intracellular calcium levels have, in turn, been proposed to act in a ‘feedback’ manner; several studies suggest that increased influx of calcium through channels on the plasma membrane or ER may modulate APP processing, resulting in an increase in Aβ production [48–50]. Consistent with this model, a recent study by Resendeet al. [51] examined the effects of oligomeric Aβ and found that these oligomers deplete ER calcium levels via phospholipase C activation, leading to intracellular calcium dyshomeostasis. Resendeet al. further showed that administration of dantrolene, an inhibitor of ER ryanodine receptors, prevented oligomer-induced apoptosis, demonstrating the involvement of ER in Aβ toxicity induced by calcium dysregulation [51].

Although the increases in intracellular calcium observed in AD are frequently ascribed to calcium entry through calcium channels, a number of alternative theories have been proposed to explain the changes in calcium regulation. One such theory is the formation of cation-selective ion channels by Aβ, which has been demonstrated to form calcium pores in artificial membrane systems, as well as in living cells [52,53]. Another channel-independent manner by which Aβ may interfere with intracellular calcium homeostasis is the formation of reactive oxygen species, which may induce membrane lipid peroxidation [54]. The resultant alterations in membrane properties may affect membrane transporter and ion channel function and could lead to elevated intracellular calcium levels [55].

We propose a mechanism in which, under normal conditions (Fig. 8A), G proteins are bound to APP, perhaps at multiple sites, and remain in an inactive condition. Under pathological conditions such as AD (Fig. 8B), when Aβ binds to APP, the G proteins are released, activated and go on to affect various downstream effectors, resulting in an increase in intracellular calcium levels, perhaps via regulation receptors on the ER and plasma membrane [11–14]; this may eventually result in cell death. The dimerization of APP and its caspase cleavage may also be involved. Although these results do not preclude multiple G-protein binding sites on APP, they highlight the importance of the -YENPTY- region in APP-dependent, Aβ-mediated toxicity. Collectively, our results bolster previous work that identified an interaction between G-protein activity and Aβ-induced toxicity, and provide the first demonstration that this toxicity is clinically relevant and related to G-protein activation-dependent calcium dysregulation. Our data suggest a direct interaction between APP and Go, as previously demonstrated [7], however, they do not exclude the possibility that APP and Go may be members of a complex; the identity of these members (if any) remains to be determined. Our model is also consistent with studies of G-protein signaling that report a dissociation of G-protein heterotrimers following activation [56–58].

Figure 8.

 Aβ-mediated G-protein activation dependent calcium influx – proposed mechanism. APP binding to Go under basal conditions (A) allows the G protein to remain in an inactive state. Increasing levels of Aβ, such as those found in AD (B), disrupt APP:Go binding, this Aβ responsiveness requires an intact C31/-YENPTY- region. The Go is liberated and goes on to achieve an active conformation and activate downstream pathways, including the influx of calcium from ER stores or from the extracellular space. Elevated levels of intracellular calcium lead to eventual cell death.

In addition to providing the first evidence linking APP-dependent, Aβ-induced cell toxicity and the well-documented calcium dysregulation observed in AD, the results presented here suggest a mechanism underlying APP-dependent Aβ toxicity. In this mechanism, basal APP binding to Go is reduced under pathological Aβ loads, thus liberating Go and activating the G-protein system, which may in turn result in calcium dysregulation. Our results also suggest that specific antagonists of G-protein activity may have a therapeutic relevance in AD.

Materials and methods

Human brain samples

For this study a total of 39 aged human brains (frontal cortex) from the Alzheimer’s Disease Research Centre (University of California, San Diego, CA, USA) were used. Samples were obtained from patients at various stages of AD (Braak stages: I–II, n = 10; III–IV, n = 10; V–VI, n = 10) and from non-dementia age-matched controls (these were designated as Braak stage 0, n = 9). Table 1 describes the demographic characteristics and AD categorization of the patient samples used in this study. All cases had been extensively characterized clinically during life and histopathologically post-mortem. Cases with a history of trauma, hemorrhage or infarction were not studied. Post-mortem examination was conducted within 8 h and patients included in this study died of acute bronchopneumonia, myocardial infarction or sepsis.

Table 1.   Demographic information of human samples. MMSE, Mini Mental State Examination.
DiagnosisBraak stage rangeNAgeSexMMSEPlaques frontal cortex (per 0.1 mm2)Tangles frontal cortex (per 0.1 mm2)Brain weight (g)
Mean ± SEMF/MMean ± SEMMean ± SEMMean ± SEMMean ± SEM
Non-dementia control0973.0 ± 7.104/527.9 ± 1.231.1 ± 0.560 ± 0.0993.6 ± 153.60
Mild ADI–II1088.7 ± 1.667/325.3 ± 2.1122.1 ± 4.410 ± 0.01016.9 ± 132.23
Moderate ADIII–IV1089.6 ± 1.865/515.6 ± 2.3838.9 ± 4.950.2 ± 1.33941.6 ± 118.90
Advanced ADV–VI1079.3 ± 2.545/59.11 ± 2.3344.2 ± 3.633.3 ± 1.241087.2 ± 53.90

Preparation of human membrane samples

Membranes were prepared as previously described [59]. Briefly, human frontal cortex samples were homogenized in 10 vol of homogenizing buffer (5 mm Hepes pH 8.0, 0.32 m sucrose, 5 mm benzamidine, 2 mmβ-mercaptoethanol, 3 mm EGTA, 0.5 mm MgSO4, 0.05% NaN3) containing protease inhibitors (10 μg·mL−1 leupeptin, 5 μg·mL−1 pepstatin A, 5 μg·mL−1 aprotinin, 10 mm phenylmethanesulfonyl fluoride) and phosphatase inhibitors (10 mm sodium orthovanadate, 2 mm KF, 1 pm okadaic acid), using a Teflon/glass homogenizer at 4 °C. Homogenized samples were centrifuged at 100 000 g for 1 h at 4 °C. The supernatant was used as the cytosolic fraction, and the pellet, rehomogenized in the original volume of homogenizing buffer, was used as the membrane (particulate) fraction. The membrane fraction was subsequently used for G-protein activation assays and western blot analysis.

GTP activation assays

The DELFIA GTP-binding kit (Perkin-Elmer, Wellesley, MA, USA) was used according to the manufacturer’s instructions. This assay is based on time-resolved fluorescence and measures GPCR by the use of Europium-tagged GTP. Stimulation of the GPCR by agonists leads to the exchange of GDP for GTP on the α subunit of the G protein. Europium-tagged GTP cannot be hydrolyzed and so remains attached to the Gα subunit, allowing the monitoring of the agonist-dependent activation of G proteins.

Antibodies

The following antibodies were used in this study: G369 (C-terminal APP specific; kindly donated by S. Gandy, Thomas Jefferson University, Philadelphia, PA, USA), CT15 (C-terminal APP specific; kindly donated by E. Koo, University of California, San Diego, CA, USA), anti-Go (rabbit polyclonal; Upstate Biotechnology, Lake Placid, NY, USA), anti-synaptophysin (mouse mAb; Chemicon, Temecula, CA, USA), anti-Aβ (6E10, mouse mAb; Covance, Princeton, NJ, USA) and anti-(β-actin) (rabbit polyclonal; Sigma, St Louis, MO, USA).

Western blot analysis

Neuronal cell homogenates, obtained as previously described [60], and membrane fractions of human brain samples were analyzed by immunoblot. Twenty micrograms of total protein per sample were loaded onto 10% Bis/Tris (Invitrogen, Carlsbad, CA, USA) SDS/PAGE gels, transferred onto Immobilon membranes, incubated overnight with primary antibodies. After incubation with primary antibodies, membranes were incubated in appropriate secondary antibodies, reacted with enhanced chemiluminescence, and developed on a VersaDoc gel-imaging machine (Bio-Rad, Hercules, CA, USA). Anti-(β-actin) (1 : 1000) was used to confirm equal loading.

Immunoprecipitation studies

Human brains (frontal cortex) or harvested cells were homogenized as previously described [61]. Briefly, cells or brain homogenates were lysed in buffer A (50 mm Tris/HCl, pH 8.0, 1 mm EDTA, 100 mm NaCl, 1% Triton X-100, 0.5% sodium cholate, 0.2 mm phenylmethanesulfonylfluoride, 1 mg·mL−1 trypsin inhibitor). Lysates were centrifuged at 4 °C at 100 000 g for 20 min, supernatant fractions were used for immunoprecipitation.

Homogenates were immunoprecipitated by using an APP polyclonal antibody (G369 or CT15) and added to 20 μL slurry of protein A/G beads (Santa Cruz Biotechnology, Santa Cruz, CA, USA). Following an overnight incubation at 4 °C the beads were washed three times with lysis buffer containing 50 mm Tris/HCl pH 7.5, 1 mm EDTA, 100 mm NaCl, 0.2 mm phenylmethanesulfonylfluoride, and 1 mg·mL−1 trypsin inhibitor. Proteins were eluted by boiling in 25 μL of loading buffer and resolved by SDS/PAGE on a 4–12% Bis/Tris gel. Following electrotransfer to nitrocellulose membrane (Millipore Corp, Bedford, MA, USA), Go was recovered by incubation with the antibody against Go, followed by incubation with the appropriate secondary antibody. Proteins were visualized by enhanced chemilluminescence and analyzed with a Versadoc XL imaging apparatus (Bio-Rad). Densitometrical analysis of immunological signals was performed using quantity one software (Bio-Rad).

Neuronal cell culture and APP transfection

B103 rat neuroblastoma cells were routinely maintained in Dulbecco’s modified Eagle’s medium, supplemented with 10% fetal calf serum and 5% horse serum in a humidified atmosphere of 5% CO2/95% O2. B103 cells were chosen because they have previously been shown to lack endogenous APP and to become sensitive to Aβ toxicity upon APP transfection [62].

In order to investigate interactions between APP, Aβ and Go, B103 cells were transiently transfected using the Amaxa transfection system (Amaxa, GmbH) according to the manufacturer’s instructions. Cells were transfected with full-length wild-type human APP695 or mutant forms of APP. Mutant forms of APP were generated on human APP695 and were constructed in expression plasmid pcDNA3.1 (Invitrogen). These mutants included deletion of the entire C31 domain (amino acids 664–695, to give construct APP695ΔC31), deletion of the -YENPTY- sequence (amino acids 681–687, to give construct APP695ΔY), as well as a point mutation D664A in which the aspartic acid at position 664 was replaced by alanine in order to block APP caspase cleavage (APP695D664A) [29]. Transfection efficiency was determined by parallel transfection with a GFP-tagged construct and was consistently determined at ∼ 70% (data not shown). Control cultures were transfected with identical amounts of the corresponding empty vectors. Cells were transfected and left for 24 h before treatment for a further 24 h with Aβ.

Preparation of Aβ

For Aβ toxicity study, cells were treated with 10 μm Aβ(1–42) (American Peptide, USA). Aβ was prepared as described previously [5]. In short, the peptide was dissolved in HCl (10 mm) and subsequently lyophilized. The peptide was dissolved in dimethylsulfoxide and added immediately to the cell culture medium to give a final concentration of 10 μm (at a dimethylsulfoxide final concentration of 0.5%). This freshly solubilized Aβ preparation yielded primarily monomeric and oligomeric Aβ species (Fig. S1) [4].

Analysis of calcium levels FLIPR assay

In order to evaluate whether the toxic effect of Aβ was calcium dependent, B103 cells were treated for 24 h with Aβ. After 24 h, cells were stained for the fluorescent imaging plate reader (FLIPR4) calcium assay according to the manufacturer’s instructions (Molecular Devices, Palo Alto, CA, USA). Briefly, cells were grown in black, 96-well plates (5 × 104 cells per well) with a clear bottom (Costar 3904; Corning, NY). Before loading the fluo-4 dye (supplied in the FLIPR kit; Molecular Devices), the media was replaced with serum-free Hanks balanced salt solution. Cells were added with an equal volume of loading buffer and incubated for 1 h at 37 °C, 5% CO2, in the presence of probenecid (2 mm). The plate was loaded into a DTX 880 multimode detector (Beckman Coulter). Cells were excited at 485 nm and fluorescence emission determined at 535 nm (integration time 0.4 s). Experiments were performed in triplicate and values are means ± SD of three experiments for relative fluorescence units (RFU).

Assays of cell viability and neurite outgrowth

A LDH release assay (CytoTox 96 assay; Promega, Madison, WI, USA) was performed to measure levels of cell death. Levels of released LDH are inversely proportional to an intact mitochondrial membrane potential and proportional to the total protein released from cells, providing an accurate index of membrane integrity and cytotoxicity. Briefly, cells were grown in 96-well plates. Following treatment with Aβ, culture medium was collected and used to determine viability, according to a colorimetric methodology. Measurements were performed with a DTX 880 plate reader as absorbance at 490 nm. All assays were performed in triplicate.

In addition, neurite outgrowth was evaluated as previously described [63]. For this purpose, averages of 10 images per condition of neuronal cells in culture were captured for subsequent analysis with the imagepro plus program (Media Cybernetic, Silver Spring, MO, USA). A neurite was defined as a cellular process longer than the diameter of one cell body. Neurite length was analyzed in at least 100 cells for each experimental group.

Statistical methods

Differences between groups were tested using one- and two-factor ANOVA with Fisher PLSD post hoc tests. Additional preliminary analysis between control and treated groups was by unpaired, two-tailed, Student’s t-test. All the results are expressed as mean ± SEM.

Acknowledgements

This work was funded by NIH grants AG 5131, AG 18440 and AG 022040.

Ancillary