Dysregulation of neuronal calcium homeostasis in Alzheimer’s disease – a therapeutic opportunity?

Abstract

Alzheimer’s disease (AD) is the disease of lost memories. Synaptic loss is a major reason for memory defects in AD. Signaling pathways involved in memory loss in AD are under intense investigation. The role of deranged neuronal calcium (Ca²⁺) signaling in synaptic loss in AD is described in this review. Familial AD (FAD) mutations in presenilins are linked directly with synaptic Ca²⁺ signaling abnormalities, most likely by affecting endoplasmic reticulum (ER) Ca²⁺ leak function of presenilins. Excessive ER Ca²⁺ release via type 2 ryanodine receptors (RyanR2) is observed in AD spines due to increase in expression and function of RyanR2. Store-operated Ca²⁺ entry (nSOC) pathway is disrupted in AD spines due to downregulation of STIM2 protein. Because of these Ca²⁺ signaling abnormalities, a balance in activities of Ca²⁺-calmodulin-dependent kinase II (CaMKII) and Ca²⁺-dependent phosphatase calcineurin (CaN) is shifted at the synapse, tilting a balance between long-term potentiation (LTP) and long-term depression (LTD) synaptic mechanisms. As a result, synapses are weakened and eliminated in AD brains by LTD mechanism, causing memory loss. Targeting synaptic calcium signaling pathways offers opportunity for development of AD therapeutic agents.

Introduction

Alzheimer’s disease (AD) is neurodegenerative disorder, which is characterized by alterations in memory formation and storage. Most cases of AD are sporadic and occur in the aging population (>60 years of age), but approximately 1%–2% of cases refer to early onset familial form of AD (40–50 years of age) [1]. Familial form of AD (FAD) is caused by mutations in genes encoding Presenilin 1 (PS1), Presenilin 2 (PS2) proteins and amyloid-precursor protein (APP) [1–4]. Several hypothesis about AD development have been proposed, but so-called «amyloid cascade hypothesis» is still considered to be the dominant. The amyloid cascade hypothesis states that overproduction and deposition of amyloid β-peptide (Aβ) in brain tissue driving AD pathogenesis [5]. Therefore, much effort was spent to develop agents which will reduce Aβ production or eliminate Aβ from the brain. Most drugs developed based on this approach failed to show any potential benefit for patients in clinical trials [6], although some promising trends have been recently reported in the trial of anti-Aβ antibody aducanumab [7]. These failures suggest that we urgently needed new approaches to AD treatment in addition to targeting Aβ. We and others previously proposed that neuronal Ca²⁺ dyshomeostasis has an important role in AD [8–12]. In this review, we will discuss Ca²⁺ signaling alterations observed in AD and point new ways of disease modifying drug development based on it.

Loss of synapses in Alzheimer disease

Several pathological hallmarks such as shrinkage of hippocampal and cortex brain regions, amyloid plaque deposition, intracellular neurofibrillary tangles and activated microglia are observed in AD patients. Another AD marker is a loss of synapses and changes in the shape and size of dendritic spines [13, 14]. Synaptic alterations observed at early stages of disease [15] and correlate strongly with cognitive decline in AD patients [16]. Synapse is formed by presynaptic axon ending and by postsynaptic dendritic spine. According to morphological structure, dendritic spines are classified into three groups: mushroom, thin and stubby spines [17, 18]. Mushroom spines, which are large and form strong synaptic contacts, were proposed to represent stable “memory spine”, while thin spines are “learning spines” which are responsible for formation of new memories [19]. We and others previously proposed that mushroom spines are strongly eliminated in AD and that loss of mushroom spines may underlie cognitive decline during the progression of the disease [20–22]. Consistent with this hypothesis, reduction in hippocampal mushroom spines was observed in different experimental mice models of AD, including PS1-M146V-KI mice [23], newly generated APP-KI mice [24, 25], in conditions of amyloid synaptotoxicity in vivo and in vitro [26], and ex vivo in hippocampal slice cultures from APP_SDL transgenic mice [27, 28]. We suggest that mushroom spines loss is an early event, which precedes more severe neurodegenerative changes. Thus, we reasoned that stabilization of mushroom spines may help to prevent eventual loss of memories of AD patients. What cell biological mechanism is responsible for dendritic spine alterations in AD? In this review we discuss the hypothesis that dysregulation of neuronal Ca²⁺ signaling plays an important role in destabilization of synaptic spines in aging and AD brains.

Presenilins and neuronal calcium homeostasis

Genetic mutations in presenilins is one of the major causes of familial AD (FAD) [1–4]. Presenilins together with nicastrin, APH-1 and PEN-2 form the γ-secretase complex that cleaves several substrates including amyloid precursor protein (APP) [29]. Sequential cleavage of APP by β- and γ-secretases constitutes amyloidogenic pathway that leads to production of toxic Aβ peptides. FAD-associated mutations in presenilins disrupt the function of γ-secretase. There is a debate on whether FAD mutations in presenilins upregulate or downregulate γ-secretase activity [30–32]. Presenilins also exert a number of γ-secretase- independent functions [33]. One of these functions is related to Ca²⁺ signaling. The connection between presenilins and Ca²⁺ signaling was initially uncovered when it was reported that fibroblasts from FAD patients release supranormal amounts of Ca²⁺ in response to InsP₃ [34]. Similar results were obtained in experiments with cells expressing FAD mutant presenilins [35] and in cortical neurons from FAD presenilin mutant knock-in mice [36, 37]. To explain these results it has been suggested that mutant presenilins affect store-operated Ca²⁺ influx [38, 39], increase activity and/or expression of intracellular Ca²⁺ release channels such as RyanR [37, 40–42] and InsP₃R [43–45] or modulate function of SERCA ER Ca²⁺ pump [46]. We proposed that presenilins form passive ER Ca²⁺ leak channel, the function that is disrupted by many but not all FAD mutations [47–50] (Fig 1). This idea created some controversy and it was challenged [51–53]. However, independent experimental support for leak function of presenilin began to accumulate [54, 55]. Importantly, recent unbiased screen for modulators of intracellular Ca2+ homeostasis revealed key role of presenilins in mediating passive Ca²⁺ influx from ER, in agreement with the “leak channel” hypothesis [56, 57]. A large hole that traverses through the entire protein was observed in the recent high resolution crystal structure of archaeal presenilin homologue PSH1 [58]. The authors noted that this hole is large enough to allow passage of small ions [58]. Interestingly, PSH1 monomer in this crystal structure adopts a fold similar to the seven-helix fold of the ClC chloride channel family [59]. These latest evidence provide additional support to the “ER Ca²⁺ leak channel hypothesis”.

Figure 1. Ca²⁺ dysregulation in AD and synaptic loss.

Amyloid β-peptide (Aβ) is generated by sequential cleavages of amyloid-precursor protein (APP) by β-secretase (β) and γ-secretase (γ). Aβ is able to form Ca²⁺-permeable pore in cell membrane. Aβ affects activity of synaptic NMDAR and mGluR5. Glutamate stimulates activation of mGluR1/5 receptors, production of IP3 and IP3R1-mediated Ca²⁺ release from the ER. Presenilins (PSEN) acts as Ca²⁺-leak pore. Many familial AD mutations disrupt this function of presenilins, which leads to ER Ca²⁺ overload and subsequent downregulation of neuronal store-operated calcium entry (nSOC) gated by STIM2. Increased ER Ca²⁺ levels result in enhanced Ca²⁺ release through IP3R1 and RyanR2. Dysregulated spine Ca²⁺ signals lead to reduction in CaMKII activity and enhanced CaN activity, subsequent facilitation of LTD and inhibition of LTP and loss of synapses. Abbreviations used in figure: AD - Alzheimer’s disease, NMDAR - N-methyl-D-aspartate receptor, mGluR1/5 - metabotropic glutamate receptor type 1 or 5, IP3 - inositol trisphosphate, IP3R1- inositol trisphosphate receptor, ER - endoplasmic reticulum, RyanR2 - ryanodine receptor type 2, CaMKII - Ca²⁺/calmodulin-dependent protein kinase II, CaN – calcineurin, LTD - long-term depression, LTP - long-term potentiation, Glu - glutamate.

Ryanodine receptors in AD

There are several lines of evidence supporting dysregulation on RyanR-mediated Ca²⁺ signaling in AD. Post-mortem hippocampal brain specimens from early-stage AD patients display increased [H]³ryanodine binding, indicative of increased RyanR protein levels hippocampal regions (subiculum, CA2 and CA1) compared to non-demented controls [60]. These results were recently supported in a study where post-mortem analysis of brain from individuals with mild cognitive impairment (MCI) at high risk for developing AD revealed the up-regulation of RyanR2 [61]. Recent data suggest that RyanRs are increased in expression and function in FAD models, particularly in the hippocampus and cortex of PS1-M146V knock-in (KI) mice [37, 40, 50] and in transgenic CRND8 (APP695(KM670/671NL + V717F) mice [62]. The up-regulation of RyanRs may be a part of AD pathology, but it may also be a protective and/or compensatory response to neuronal Ca²⁺ dysregulation (reviewed in [63]).

One approach used to dissect this issue is to obstruct the effects of RyanRs using pharmacological inhibitors or functional blockers. In our previous studies, we observed that long-term feeding of the RyanR inhibitor dantrolene exacerbated amyloid plaque formation and resulted in the loss of hippocampal synaptic markers and neuronal deterioration in 8 month old APPPS1 mice [50]. In contrast, studies from other groups showed that short term treatment with dantrolene was able to stabilize Ca²⁺ signals, ameliorate cognitive decline and reduce neuropathology, amyloid load and memory impairments in various AD mouse models [64–66], suggesting that blocking RyanR activity may actually be beneficial in the context of AD. One potential problem with interpreting these results is that specific RyanR inhibitors do not exist and the drug dantrolene, used in most studies, has additional targets such as store-operated Ca²⁺ channels [67]. Moreover, dantrolene is specific for skeletal muscle isoform RyanR1 [68], and does not block neuronal RyanR2 and RyanR3 effectively.

To resolve this issue, in the recent studies we took a genetic approach and generated APPPS1x RyanR3^−/− mice [69]. We compared the phenotype of APPPS1x RyanR3^−/− to the phenotype of WT, RyanR3^−/− and APPPS1 mice. In this analysis we discovered that RyanR3 appears to play a dual role in the context of AD pathology, rather than an invariable positive or negative effect. In the young APPPS1 mice (≤3 month old) deletion of RyanR3 was detrimental and enhanced AD pathology [69]. We concluded that RyanR3 plays an important protective role in early stages of AD by helping to reduce neuronal excitability and activity-dependent Aβ production. These data support the hypothesis that blockade of RyanRs in the early stages of AD progression would produce a more aggressive AD phenotype compared to the placebo group, as we previously suggested [50, 63, 70]. In contrast, in older APPPS1 mice (≥6 month old) deletion of RyanR3 resulted in beneficial effects and reduced AD pathology [69]. These results consistent with reports that dantrolene exerted beneficial effects in several AD mouse models [64–66] and suggested that blocking RyanR is a viable therapeutic strategy. This idea is reviewed in depth by Dr. Grace Stutzmann and her colleagues in the accompanying review article [71].

Disruption of nSOC signaling in mouse models of AD

Another Ca²⁺ signaling defect in AD neurons is related to dysfunction of neuronal store-operated Ca²⁺ entry (SOC) pathway. The neuronal function of SOC pathway is poorly understood, however key molecular components of SOC are expressed in the brain and enriched in hippocampus. SOC activation is mediated by STIM proteins, which act as ER Ca²⁺ sensors. STIM2 protein is highly enriched in hippocampus and STIM1 is enriched in cerebellum. Recent report demonstrated impaired spatial memory in forebrain-specific double knockout of STIM1 and STIM2 proteins [72]. Differential role of STIM1 and STIM2 proteins in control of neuronal SOC was demonstrated [73]. In our studies we discovered that expression of STIM2 is downregulated in hippocampal neurons PS1-M146V-KI and APP-KI neurons and in post-mortem samples from AD patients [23, 24]. We reasoned that reduction in STIM2 expression level is a compensatory response to ER Ca²⁺ overload in these models. We further proposed that downregulation of STIM2 and synaptic SOC is responsible for loss of mushroom spines in PS1-M146V-KI and APP-KI in hippocampal neurons [23, 24] (Fig 1). Indeed, expression of STIM2 resulted in rescue of mushroom spine loss in PS1-M146V-KI and APP-KI hippocampal neurons [23, 24] and in neurons exposed to Aβ oligomers [26].

Based on these results we propose that positive modulators of nSOC activity may be considered as potential therapeutics agents for preventing synaptic and memory loss in AD [21, 23, 24, 26, 74]. However, in order to develop effective drug the knowledge of molecular identity of nSOC channels is necessary. The candidates for such role are members of Orai and/or TRPC families. Expression of Orai2 is enriched in hippocampus [75]. Association of neuronal STIM2 with Orai1 was demonstrated in biochemical experiments [76] and the functional role of Orai1 protein in control of synaptogenesis in hippocampal neurons has been recently described [77]. TRPC proteins have been demonstrated to play a role in neuronal Ca²⁺ signaling in multiple studies [78, 79]. Brain overexpression of TRPC6 channels resulted in spine proliferation and enhanced memory performance in transgenic mice [80], suggesting a potential role for TRPC6 in mediating nSOC. Interestingly, TRPC6 is activated by hyperforin [81] and it has been demonstrated in the previous studies that hyperforin and its derivatives were able to prevent beta-amyloid neurotoxicity and spatial memory impairments in AβPPSwe/PSEN1 E9 (AβPP/PS1) transgenic mice [82–84]. TRPC6 was also recently demonstarted to interact directly with APP and to affect APP cleavage by γ-secretase [85]. Future studies will determine the potential role of Orai and TRPC channels in supporting STIM2-gated nSOC and will help to establish if positive modulators of these channels exert beneficial effects in AD.

CaMKII and CaN “tug of war” and AD pathogenesis

Another possibility for the development of disease preventing therapies is to target signaling pathways downstream from synaptic Ca²⁺ dysregulation. Spine Ca²⁺ signaling sets up a balance between activities of Ca²⁺-calmodulin-dependent kinase II (CaMKII) and Ca²⁺-dependent phosphatase calcineurin (CaN). Both CaMKII and CaN are enriched in the brain and extremely abundant in synaptic locations [86, 87]. CaMKII is a holoenzyme of 12 subunits, each derived from one of four genes (α, β, γ and δ). In the brain αCaMKII and βCaMKII are the most abundant subunits, expressed at the ratio 3:1. The total concentration of CaMKII in the hippocampal spines was estimated to be on the order of 100 μM [88, 89]. CaMKII holoenzymes are activated by the binding of Ca²⁺/CaM. Following activation, CaMKII can undergo inter-subunit autophosphorylation at residue T286 (for αCaMKII), that results in “locking” CaMKII in an active state independently from Ca²⁺ levels. Presence of p(T286)-αCaMKII at synaptic locations is essential for LTP and it has been proposed to be critical for memory formation [90, 91], an idea supported by observation of LTP and memory defects in T286A αCaMKII mutant mice [92]. The role of αCaMKII T286 autophosphorylation in memory maintenance is less clear [91, 93]. CaN (PP2B) is a protein phosphatase composed of a large catalytic (CaNA) and a small regulatory subunit (CaNB). Increase in Ca²⁺ concentration leads to Ca²⁺ association with EF hand motifs in CaNB, partial activation of CaN and exposure of Ca²⁺-CaM binding site on CaNA. Association of Ca²⁺/CaM with CaNA then causes full activation of CaN [86, 94]. Activation of CaN in hippocampal neurons is essential for induction of LTD [95, 96]. Overexpression of CaN in the forebrain of mice impaired the transition from short-term to long-term memory as well as an intermediate form of LTP [97]. Conditional genetic knockout of CaN in mouse forebrain resulted in impairment of hippocampal-dependent tasks including working and episodic memory and blocked LTD [98].

As it is clear from this brief description, both CaMKII and CaN play key and opposing roles in synaptic plasticity and both enzymes are regulated by spine Ca²⁺ in a complex manner. Experimental evidence indicated that CaMKII functions as a high-frequency activity detector that stabilizes the spines (by LTP-like mechanism), whereas CaN is responsive to low frequency stimulation and destabilizes the spines (by LTD-like mechanism) [96, 99]. A proposed allosteric model suggests that CaM is more likely to activate either CaN or CaMKII depending on local Ca²⁺ concentration [100]. Intense and transient Ca²⁺ increase through NMDARs following tetanic stimulation result in the preferential activation of CaMKII within the dendritic spines. However, as Ca²⁺ decreases, but before it returns to baseline, CaM is more likely to bind and activate CaN [100]. CaN activates PP1, and PP1 in turn dephosphorylates CaMKII, consequently decreasing its kinase activity [95].

Because both CaMKII and CaN are Ca²⁺-dependent, subtle changes in synaptic Ca²⁺ signaling cause shift in the balance of CaMKII and CaN activities. The balance between activities of CaMKII and CaN appear to be shifted in AD synapses in favor of CaN (Fig 1) [11, 101]. Consistent with this hypothesis, a shift in the balance between LTP-like and LTD-like mechanisms has been recently reported based on the analysis of synaptic plasticity in mouse model of AD [102]. It has been proposed that deranged synaptic Ca²⁺ signaling causes aberrant metaplasticity in AD by shifting a balance in induction of LTP and LTD [103]. In support of these functional arguments, biochemical analysis revealed alterations in CaMKII localization and expression in AD brains. It was discovered that p(T286)-αCaMKII is reduced at synaptic locations in hippocampus of AD patients and in mouse hippocampal neurons treated with Aβ [104]. The degree of p(T286)-αCaMKII loss at synaptic locations correlated with severity of the disease [104]. These results suggested that reduction in synaptic CaMKII activity may play an important role in AD pathogenesis [105]. In our experiments we observed direct correlation between reduced nSOC, levels of synaptic p(T286)-αCaMKII and mushroom spine loss in mouse models of AD [23, 24, 26] (Figure 1). Importantly, STIM2 overexpression rescued p(T286)-αCaMKII levels [23, 24, 26]. The exact mechanism how nSOC influences synaptic CaMKII function is unclear. Dendritic spines have a poor intrinsic buffering capacity for Ca²⁺, and action potentials may increase Ca²⁺ only very briefly. It is possible that nSOC activity is necessary to support CaMKII autophosphorylation when high-frequency stimulation ceases. While activity of synaptic CaMKII is reduced in AD, activity of CaN appear to be enhanced. Superactivation of CaN in human AD samples was reported in biochemical experiments [106, 107]. Activation of CaN in AD human samples appears to occur due to calpain-mediated cleavage and hyperactivation. Importance of CaN in AD has been highlighted by previous studies in AD cellular and animal models. It has been shown that CaN mediates both the neurotoxic and cognitive effects of Aβ oligomers [108–114]. Beneficial effects of CaN inhibition have been demonstrated in several AD mouse models [108, 110, 111, 114–116]. It has been shown that CREB phosphorylation and LTP expression, which are disrupted by Aβ oligomers, are restored following FK506 treatment in hippocampal slice experiments [111]. These findings lead to proposal that CaN overactivation in one of the driving forces of AD pathology [117]. Very importantly, recent analysis revealed significantly reduced incidence of AD in transplant patients treated with calcineurin inhibitor FK506 [118]. These findings provide strong support to the hypothesis that excessive activation of CaN plays a key role in spine loss in AD.

Future directions

Synaptic Ca²⁺ dysregulation appears to play an important role in synaptic loss in AD. This knowledge provides a number of potential therapeutic targets for prevention of memory loss in AD (Fig 1). Potential approaches include modulation of RyanR2 activity and activity of nSOC channels. Molecular identity of nSOC channels needs to be established to facilitate these efforts. Pharmacological tools aimed at restoring the balance between CaMKII and CaN activities in synaptic spines may also provide a potential for therapeutic interference. It is necessary to establish if beneficial effects can be achieved following inhibition of CaN without immunosuppression side-effects. It is also important to identify downstream relevant targets of CaMKII and CaN at the synapse.

Highlights.

• Synaptic loss is a basis for memory loss in Alzheimer’s disease (AD)

• Dysregulation of synaptic calcium (Ca²⁺) signaling plays an important role in synaptic loss in AD

• Function of ryanodine receptors and store-operated calcium channels is abnormal in AD neurons

• There is a shift in the balance of Ca²⁺-calmodulin-dependent kinase II (CaMKII) and Ca²⁺-dependent phosphatase calcineurin (CaN) activities at the synapse

• Balance between long-term potentiation (LTP) and long-term depression (LTD) synaptic mechanisms is tilted in AD spines, causing elimination of synapses by LTD-like mechanism