Mitochondrial dysfunction - the beginning of the end in Alzheimer's disease? Separate and synergistic modes of tau and amyloid-β toxicity

The pathology of Alzheimer's disease (AD) is characterized by amyloid plaques (aggregates of amyloid-β (Aβ)) and neurofibrillary tangles (aggregates of tau) and is accompanied by mitochondrial dysfunction, but the mechanisms underlying this dysfunction are poorly understood. In this review, we discuss the critical role of mitochondria and the close inter-relationship of this organelle with the two main pathological features in the pathogenic process underlying AD. Moreover, we summarize evidence from AD post-mortem brain as well as cellular and animal AD models showing that Aβ and tau protein trigger mitochondrial dysfunction through a number of pathways, such as impairment of oxidative phosphorylation, elevation of reactive oxygen species production, alteration of mitochondrial dynamics, and interaction with mitochondrial proteins. A vicious cycle as well as several vicious circles within the cycle, each accelerating the other, can be drawn, emphasizing the synergistic deterioration of mitochondria by tau and Aβ.

mito chondrial DNA, lipids of the mitochondrial membrane, and mitochondrial proteins. Dysregulation of mito chon drial function because of these insults leads to synaptic stress, disruption of synaptic transmission, apoptosis and, ultimately, neurodegeneration [14,15]. Th us, it is important to understand the mechanisms of mitochondrial stress related to the pathogenesis of AD and to exploit this insight for developing therapeutic strategies for AD.

Evidence of mitochondrial dysfunction in post-mortem AD brain and peripheral cells
Mitochondrial dysfunction has been proposed as an underlying mechanism in the early stages of AD since Figure 1. Amyloid-β-related mitochondrial impairment. Mitochondria were found to be the target for amyloid-β (Aβ), which interacts with several proteins, leading to mitochondrial dysfunction. Indeed, Aβ was found in the outer mitochondrial membrane (OMM) and inner mitochondrial membrane (IMM) as well as in the matrix. The interaction of Aβ with the OMM aff ects the transport of nuclear-encoded mitochondrial proteins, such as subunits of the electron transport chain complex IV, into the organelle via the translocase of the outer membrane (TOM) import machinery. Moreover, Aβ disturbs the activity of several enzymes, such as pyruvate dehydrogenase (PDH) and α-ketoglutarate dehydrogenase (αKGDH), decreasing NADH reduction, and the electron transport chain enzyme complex IV, reducing the amount of hydrogen that is translocated from the matrix to the intermembrane space (IMS), thus impairing the mitochondrial membrane potential (MMP). Taken together, these events cause abnormal mitochondrial electron activities, leading to decreased complex V activity and so to a drop in ATP levels, in addition to increasing reactive oxygen species (ROS) generation. Moreover, ROS induce peroxidation of several mitochondrial macromolecules, such as mitochondrial DNA (mtDNA) and mitochondrial lipids, contributing to mitochondrial impairment in the mitochondrial matrix. The complex of Aβ bound to binding alcohol dehydrogenase (ABAD) impairs the binding of NAD+ to ABAD, changes mitochondrial membrane permeability and reduces activities of respiratory enzymes, inducing further ROS production and leading to mitochondrial failure. Aβ binding also activates Fis1 (fi ssion protein) and promotes increased mitochondrial fragmentation; this increased mitochondrial fragmentation produces defective mitochondria that ultimately damage neurons. Furthermore, Aβ binding to cyclophilin D (CypD) enhances the protein translocation to the inner membrane, favouring the opening of the mitochondrial permeability transition pore, formed by the adenine nucleotide translocator (ANT) and voltage-dependent anion channels (VDACs). Cyt C, cytochrome C; DLP1, dynamin-like protein 1; PDH, pyruvate dehydrogenase; ProAp, proapoptotic factors; SOD, superoxide dismutase; TCA, tricarboxylic acid; TIM, translocase of the inner membrane. energy defi ciency is a fundamental characteristic of the AD brain [16,17] as well as of peripheral cells derived from AD patients [18]. Th e similar fi ndings in peripheral tissue highlight the potential of using surrogate tissue in the diagnosis of AD ( Figure 2).
Brain imaging studies have demonstrated defects in glucose use in living AD patients, an abnormality that may occur well before the onset of clinical symptoms [17]. Furthermore, several lines of evidence suggest that mitochondria intervene in the mechanism by which Aβ, which is derived from amyloid precursor protein (APP), triggers synaptic failure and neuro degeneration. Crucial to this fi nding was the notion that both Aβ and the full length protein, APP, accumulate within mitochondria in the brains of AD patients [6,[19][20][21].
Furthermore, morphometric analyses revealed that mito chondria display structural abnormalities, as illustrated by the accumulation of mitochondrial DNA and proteins in the cytoplasm and vacuoles; this accelerates mitochondrial degradation and thereby leads to a signifi cant decrease in mitochondrial numbers in AD [22]. Th e mitochondrial accumulation of APP and Aβ in brain tissues causes both altered activities in key mitochon drial enzymes, including cytochrome c oxidase (COX, complex IV), pyruvate dehydrogenase (PDH) and α-ketoglutarate dehydrogenase complex (αKGDH), and import of nuclear-encoded proteins [3,18,21,23]. Th e activity of αKGDH, but not of PDH, has been found to be reduced in cultured skin fi broblasts from sporadic AD patients and in a subset of patients with presenilin-1 mutations. Sorbi and colleagues [24] observed a reduction in PHD, ATP-citrate lyase and acetoacetyl-CoA thiolase in post-mortem brain tissues from AD-aff ected subjects and correlated the decrease of these enzymes to the decreased production of acetyl-coenzyme A and cholinergic defects that are observed in AD.
Th e impairment of mitochondrial oxidative phos phorylation in the AD brain was found to be proportional to clinical disability [3]. Th e most consistent defect in mitochondrial electron transport enzymes in AD is a defi ciency in COX, a fi nding reported in AD post-mortem brain as well as in platelets and fi broblasts from AD patients [25,26]. Dysfunction of COX increases ROS produc tion, reduces energy stores, and disturbs energy metabolism [27]. Altered levels of antioxidant enzymes, such as catalase, copper/zinc superoxide dismutase (Cu/ Zn-SOD) and manganese superoxide dismutase (Mn-SOD) have been found in AD brain, supporting increased oxidative stress as a pathomechanism in AD [28].
Laser-capture microdissection has been applied to investigate distinct brain areas as some undergo neurodegeneration whereas others do not: the analysis of the expression of 80 metabolically relevant nuclear-encoded electron transport chain genes from non-NFT-bearing AD neurons revealed that 60 to 70% of these were signifi cantly lower in metabolically aff ected areas such as the posterior cingulate cortex, the middle temporal gyrus, and the hippocampal CA1 region [29]. Post-mortem tissue also provides strong evidence for increased levels of cellular oxidative stress in vulnerable regions of AD brains compared to aged controls [30]. Increased protein oxidation, protein nitration, and lipid peroxidation were detected in brain areas with overt NFTs and amyloid plaques [31].
More recently, abnormal mitochondrial dynamics has been identifi ed in fi broblasts from sporadic and familial AD cases [32], a distortion probably mediated by altered expression of dynamin-like protein 1 (DLP1), a regulator of mitochondrial fi ssion and distribution, due to elevated oxidative and/or Aβ-induced stress. Moreover, deregulated mitochondrial dynamics might be due to enhanced nitrosative stress generated by Aβ, such as S-nitrosylation of DLP1 in AD brain [33]. Th is modifi cation can disturb the balance between fi ssion and fusion of mitochondria in favour of mitochondrial fi ssion followed by mito chondrial depletion from axons and dendrites and, subsequently, synaptic loss.
Altogether, these data indicate that mitochondrial dysfunction is a highly relevant event in AD pathophysiology. Post-mortem studies in AD brain allow the identifi cation of long-term synergistic in vivo eff ects of tau and Aβ in human because, in the majority of AD cases, both pathologies persist in parallel for many years.

Mitochondrial dysfunction in cellular and transgenic mouse AD models
Understanding the molecular pathways by which the various pathological alterations including Aβ and tau compromise neuronal integrity and lead to clinical symptoms has been a long-standing goal of AD research. Th e successful development of cell and mouse models that mimic diverse aspects of the disease process have facilitated this eff ort and assisted in understanding of the interplay of Aβ and tau on bioenergetics processes in vivo [10] (Figure 3).

Mitochondria and Aβ
Aβ may exert its toxicity via a plethora of pathways and thereby induce synaptic and neuronal degeneration [2].
Using neuronal PC12 cells, we found that expression of one known APP gene mutation, the 'Swedish APP (APP Sw )' double mutation (KM670/671NL), leads to an enhanced vulnerability of these cells to oxidative stress and mitochondrial dysfunction, mediated by diff erent caspases and the stress-activated protein kinase pathway [34][35][36] (Figure 3). Of note, we are able to study the eff ects of Aβ in a dose-dependent manner as cells transfected with APP Sw had fi vefold higher Aβ levels than wild-type APP (APP wt ) transfected cells, similar to the increase found in human carriers with the APP Sw mutation compared to sporadic AD cases. Th e data suggest that the mitochondrial dysfunction induced by Aβ is probably mediated via enhanced nitric oxide production that reduces complex IV activity. Th is reduction leads to a depletion of intracellular ATP levels, fi nally initiating cell death [37]. When intracellular Aβ production is prevented by a γ-secretase inhibitor, this restores nitric oxide and ATP levels, indicating a direct involvement of Aβ in these mechanisms.
Mitochondria were found to be a target for APP toxicity as both the full-length protein and Aβ accumulate in the mitochondrial import channels, and both lead to mitochondrial dysfunction. Recently, APP has been demonstrated as a substrate of the mitochondrial γsecretase in cultured SH-SY5Y human neuroblastoma cells [38]. Pavlov and colleagues [38] proposed that APP processing in mitochondria may lead to local Aβ production, thereby contributing to mitochondrial dysfunction. Recently, our group investigated specifi c eff ects of Aβ on mitochondrial function under physiological conditions by analysing mitochondrial respiratory function and energy homeostasis in both control and APP wtexpressing SH-SY5Y cells [8]. One possibility as to how chronic exposure to soluble Aβ may result in an impairment of energy homeostasis is through decreased respiratory capacity of the mitochondrial respiratory chain, mainly of complex IV, which, in turn, may accelerate neuronal demise.
Eventually, APP, through Aβ production, causes an imbalance of mitochondrial fi ssion/fusion that results in mitochondrial fragmentation and abnormal distribution, which contributes to mitochondrial and neuronal dysfunc tion [39].
An interesting observation was made when the toxic eff ects of Aβ were compared with human amylin (HA), a protein aggregating in the pancreas of type 2 diabetic patients [40]. Both agents share amyloidogenic properties and are toxic to primary neuronal cultures, whereas the non-amyloidogenic rat amylin is not [41]. In a next step, HA and Aβ toxicity to SH-SY5Y neuroblastoma cells was assessed by iTRAQ quantitative proteomics [42]. Th is revealed the surprising observation that identical proteins are deregulated by HA and Aβ, but not rat amylin, and a quarter of these were mitochondrial, supporting the notion that mitochondrial dysfunction is a common target in these two amyloidoses. Functional validation revealed that Aβ and HA both exert a shared toxicity, at least in part, via deregulation of identical mitochondrial proteins and impairment of complex IV activity [42].
AD is believed to be a disease of the synapses, characterized by massive synaptic and, eventually, neuronal loss, and these features have been detected in AD transgenic mouse models (Figure 3). A decrease of the mitochondrially localized chaperone hsp70, which stabilizes proteins during their import into mitochondria, has been observed in synaptosomal fractions derived from the APP Sw transgenic strain Tg2576, pointing to a mito chondrial stress response [43]. Signifi cant diff erences were revealed in the protein subunit composition of respiratory chain complexes I and III. Moreover, endo genous Aβ was found to be associated with brain mitochondria in these mice. Cu 2+ ions have been shown to render Aβ more toxic, possibly in its dimeric conformation, as a potent inhibitor of COX, thereby contribut ing to neurodegeneration in AD [44]. In APP Sw transgenic mice, reduced glucose metabolism was observed after [14C]-2-deoxyglucose infusion as well as impaired Cu/Zn SOD activity, both of which contribute to oxidative damage. In another transgenic mouse model, APPS/L, which combines the Swedish (K670M/N671L) with the London (V717I) mutation in the human APP gene, an early energy dysfunction was found as shown by a decreased mitochondrial membrane potential as well as decreased ATP levels at 3 months of age, when Aβ levels are elevated but plaques are not yet present [5,14,37]. A stronger reduction in mitochondrial membrane potential and in ATP levels was found in double transgenic APP/ PS1 (APPS/L/PS1 M141L) mice, which generate higher levels of Aβ compared to their single transgenic littermates (APPS/L), and exhibit Aβ plaques already at an age of 3 months [4]. Together, this implies that an Aβdependent mitochondrial dysfunction starts at a very young age and accelerates substantially with increasing age, as does the Aβ plaque load [4]. Consistent with this observation, APP/PS2 (APP Sw /PS2 N141I) double transgenic mice display age-related cognitive defi cits associated with discrete brain Aβ deposition and infl ammation [45]. Furthermore, an Aβ insult in APPS/L mice caused early defi cits in synaptic mitochondria, as shown by increased mitochondrial permeability transition, a decline in both respiratory function and activity of COX, and increased mitochondrial oxidative stress [46].
Of note, age-dependent impairment of oxygen consump tion, such as a decrease of state 3 and of uncoupled respiration, was observed in several APP transgenic mouse models compared to aged-matched controls [5,14,43,47]. Th is indicates that mitochondrial deregulation is a common feature in Aβ-generating mice and independent of the mouse model used.
Th ere is broad experimental proof that Aβ is indeed present in mitochondria. Aβ binds specifi cally to the mitochondrial Aβ-binding alcohol dehydrogenase (ABAD) [6], a mitochondrial matrix protein that is up-regulated in the temporal lobe of AD patients as well as in APP  transgenic mice [48]. Th e Aβ-ABAD interaction causes elevated ROS production, cell death, as well as spatial learning and memory defi cits in 5-month-old APP/ ABAD double transgenic mice [49]. Th e investigation of the crystal structure of ABAD-Aβ demonstrated that the formation of the complex prevents the binding of NAD+ to ABAD, thereby changing the mitochondrial membrane permeability and reducing the activities of respiratory enzymes, which then may lead to mitochondrial failure [6]. Aβ in mitochondria further interacts with cyclophilin D (CypD), an integral part of the mitochondrial permeability transition pore that potentiates free radical production, causes synaptic failure, and promotes opening of the pore leading to apoptosis [50]. It has been demonstrated previously that CypD is capable of forming complexes with Aβ within mitochondria of cortical neurons from APP mutant mice, increasing the translocation of CypD from the matrix to the inner membrane. Furthermore, in APP transgenic mice, the abrogation of CypD was capable of attenuating Aβ-mediated abnormal mitochondrial dysfunction, such as calcium-induced mitochondrial swelling, and it lowered mitochondrial calcium uptake capacity and impaired mitochondrial respiratory function. Aβ impaired calcium storage in mitochondria, altering neuronal function, as it is exported to the cytosol, together with other apoptotic factors (ProAp), such as cytochrome c (Figure 1). Finally, Anandatheerthavarada and Devi [51] showed that APP contains a mitochondrial targeting sequence and that an accumulation of APP in mitochondrial membranes leads to mitochondrial dysfunc tion in Tg2576 neurons. Taken together, these fi ndings are in line with the recently proposed hypothesis of an intracellular Aβ toxicity cascade, which suggests that the toxic Aβ species that cause molecular and biochemical abnormalities are in fact intracellular oligomeric aggregates rather than the extracellular, insoluble plaques [6,20].

Mitochondria and tau protein
How does tau interfere with mitochondrial function? In its hyperphosphorylated form, tau, which forms the second hallmark lesion in AD, the NFTs, has been shown to block mitochondrial transport, which results in energy deprivation and oxidative stress at the synapse and, hence, neurodegeneration [52][53][54][55]. In transgenic pR5 mice that overexpress human P301L mutant tau [56], the biochemical consequences of tau pathology have been intensively investigated using proteomics followed by functional validation [57,58]. A mass-spectrometric analy sis of the brain proteins from these mice revealed mainly a deregulation of mitochondrial respiratory chain complex components (including complex V), antioxidant enzymes, and synaptic proteins (Figure 4). Th e functional analysis demonstrated a mitochondrial dysfunction in the mice, together with reduced NADH-ubiquinone oxido reductase (complex I) activity and, with age, impaired mitochondrial respiration and ATP synthesis. Mitochondrial dysfunction was associated with higher levels of ROS in aged transgenic mice. Increased tau pathology as in aged homozygous pR5 mice revealed modifi ed lipid peroxidation levels and the upregulation of antioxidant enzymes in response to oxidative stress [57].
Th us, this evidence demonstrated for the fi rst time that not only Aβ but also tau pathology can lead to metabolic impairment and oxidative stress as in AD. Consistent with observations of a cytosolic accumu lation of the α-chain of ATP synthase observed at early stages of neurofi brillary degeneration in AD, one mechanism proposed is that tau accumulation could have direct consequences on mitochondrial activity through the cytosolic accumulation of the α-chain of ATP synthase. Reciprocally, hyperphosphorylation of tau may be directly attributable to mitochondrial oxidative stress in a mouse model lacking the mitochondrial detoxifying enzyme superoxide dismutase 2 (Sod2 −/− ), consistent with a synergistic interaction of APP and mitochondrial oxidative stress in contributing to an AD-like neocortical pathology [59]. Furthermore, chronic respiratory chain dysfunction through inhibition of complex I led, besides a concentration-dependent decrease in ATP levels, to a redistribution of tau from the axon to the cell body, the retrograde transport of mitochondria and, fi nally, cell death [60]. Together, these fi ndings support the notion that tau pathology involves a mitochondrial and oxidative stress disorder possibly distinct from that caused by Aβ.

Pathogenic convergence of Aβ and tau on mitochondria
Although Aβ and tau pathologies are both known hallmarks in AD, the underlying mechanisms of the interplay between plaques and NFTs (or Aβ and tau) have remained unresolved. However, a close relationship between mitochondrial impairment and Aβ on the one hand and tau on the other has been established. How do both features in AD relate to each other and might the two molecules synergistically aff ect mitochondrial integrity? Several studies suggest that Aβ aggregates and hyper phosphorylated tau may block mitochondrial carriage to the synapse, leading to energy defects and neuro degeneration [61]. Moreover, transport of APP into axons and dendrites may be inhibited by enhanced tau levels, causing impaired axonal transport, which suggests a link between tau and APP [52,55]. Remarkably, Aβ injections amplify a pre-existing tau pathology in several transgenic mouse models [56,62,63], whereas lack of tau abrogates Aβ toxicity [54,64]. Our fi ndings indicate that mitochondria in tau transgenic pR5 mice display enhanced vulnerability towards Aβ insult in vitro [57,65], suggesting a synergistic action of tau and Aβ pathologies on this organelle (Figure 3). Th e Aβ insult caused a greater reduction of mitochondrial membrane potential in cerebral cells from pR5 mice [57]. Furthermore, incubation of isolated mitochondria from pR5 mice with either oligomeric or fi brillar Aβ preparations resulted in impairment of the mitochondrial membrane potential and of respiration. Interestingly, aging particularly increased the sensitivity of mitochondria to oligomeric Aβ insult compared to that of fi brillar Aβ. Th is suggests that while both oligomeric and fi brillar Aβ are toxic, they exert diff erent degrees of toxicity [65]. In a related study, Crouch and colleagues [66] demonstrated that increasing the bioavailability of intracellular copper can restore cognitive function by inhibiting the accumulation of neurotoxic Aβ trimers and phosphorylated tau in APP Sw / PS1 transgenic mice. In yet another study, it was shown that exposure to Aβ induces tau hyperphos phory lation by promoting glycogen synthase kinase (GSK)3β activity. Th is demonstrates an intimate relationship between Aβ accumulation and abnormal tau phosphory lation in causing the cognitive defi cits that characterize AD, and highlights GSK3β activity as an important intermediate [67]. In contrast, overexpression of the longest form of human wild-type tau (tau441) in mouse neuroblastoma (N2A) showed an anti-apoptotic protec tive function of tau phosphorylation, which likely inhibited competitively phosphorylation of β-catenin by GSK-3β, facilitating the function of β-catenin. Th us, overexpression of tau seems to attenuate Aβ-mediated cell death via suppression of the mitochondria-caspase-3 pathway [68,69]. Taken together, these studies illustrate a complex interplay between the two key proteins in AD.
In a step further, triple transgenic mouse models have recently been established that combine Aβ and tau pathologies, and the contributions of both AD-related proteins to the mitochondrial respiratory machinery and energy homeostasis have been investigated in vivo ( Figure 3). Indeed, our group demonstrated a mitochondrial dysfunction in a novel triple transgenic mouse model (pR5/APP Sw /PS2 N141I) -tripleAD mice -using proteomics followed by functional validation [9]. Notably, deregulation of complex I was found to be tau-dependent, while deregulation of complex IV was Aβ-dependent, at both the protein and activity level ( Figure 5). Whereas down-regulation of several subunits of complex I was observed in tripleAD mice compared to both pR5 and APP Sw /PS2 mice, deregulation of several subunits of complex V was seen in tripleAD mice compared to pR5 mice but not compared to APP Sw /PS2 mice (Figure 4). Th e convergent eff ects of Aβ and tau led already at the age of 8 months to depolarized mitochondrial membrane poten tial in tripleAD mice. Additionally, we found that age-related oxidative stress at 12 months of age may exaggerate the disturbances in the respiratory system and synthesis of ATP and, in turn, take part in the vicious cycle that fi nally leads to cell death.
Together, our studies highlight the key role of mitochon dria in AD pathogenesis, and the close interrelationship of this organelle and the two main pathological hallmarks of this disease. Our data complement those obtained in another triple transgenic mouse model, 3xTg-AD (P301Ltau/APP Sw /PS1 M146L) [70]. Mitochondrial dysfunction was evidenced by age-related decreased activity of regulatory enzymes of the oxidative phosphory lation system, and of PDH and COX in 3xTg-AD mice aged from 3 to 12 months [71]. Th ese mice also exhibited increased oxidative stress and lipid peroxidation. Most of the eff ects on mitochondria were seen at the age of 9 months, whereas mitochondrial respiration was signifi cantly decreased at 12 months of age. Importantly, mitochondrial bioenergetics defi cits were found to precede the development of AD pathology in the 3xTg-AD mice. More recently, Hyun and colleagues [72] demonstrated that the plasma membrane redox system (PMRS) is impaired in the 3xTg-AD mice and that the activities of PMRS enzymes may protect neurons against Aβ toxicity, suggesting enhancement of PMRS function as a novel approach for protecting neurons against oxidative damage in AD and related disorders. Collectively, these recent data consolidate the idea that a synergistic eff ect of tau and Aβ augments the pathological deterioration of mitochondria at an early stage of AD.

Conclusion
We discuss here the critical role of mitochondria and the close inter-relationship of this organelle with the two main pathological features in the pathogenic process underlying AD.
Mitochondrial impairment has been recognized as a prominent and early event in AD in terms of metabolic energy production associated with respiratory chain dysfunction and the regulation of levels of second messengers such as ROS or Ca 2+ . Th is is partially due to either accumulated damage to mitochondrial DNA or direct harmful eff ects of oxidative stress and/or Aβ on mitochondrial components. Also, mitochondrial traffi cking as well as mitochondrial fusion/fi ssion are known to be altered in AD, compromising normal neurophysiology Figure 5. Impairment of the electron transport chain in Alzheimer's disease. Complexes I (NADH:ubiquinone oxidoreductase) and II (succinate dehydrogenase, belongs to the tricarboxylic acid (TCA) cycle) receive electrons from NADH and FADH2, respectively. Electrons are then driven from complexes by the mobile carrier molecules coenzyme Q/ubiquinone (UQ) and cytochrome c (Cyt c) to the fi nal acceptor, molecular oxygen (O 2 ). Electron fl ow is coupled to proton movement across the inner mitochondrial membrane (IMM) in complexes I, III and IV. The resulting proton gradient is harvested by complex V to generate ATP. In Alzheimer's disease, abnormal mitochondrial electron activities have been observed, predominantly in complexes I and IV, leading to impaired mitochondrial membrane potential, decreased production of ATP (complex V) and enhanced reactive oxygen species (ROS) levels. Interestingly, deregulation of complex I is mainly tau-dependent, while deregulation of complex IV is amyloid-β (Aβ)-dependent, at both the protein and activity level. IMS, intermembrane space. and a healthy cellular mitochondrial pool, which eventually leads to apoptosis. Specifi cally, the mitochondrial impairment integrates the close interplay of the two common hallmarks of AD, plaques and NFTs, or Aβ and tau, which act independently as well as synergistically on this vital organelle. With regards to the involvement of AD-related proteins in pathogenesis, a vicious cycle as well as several vicious circles within the cycle, each accelerating the other, can be drawn, emphasizing the potency of the synergistic destruction in mitochondria. Finally, the key role of mitochondrial dysfunction in the early pathogenic pathways by which Aβ leads to neuronal dysfunction in AD is particularly challenging with respect to establishing therapeutic treatments. Besides the modulation and/or removal of both Aβ and tau pathology, strategies involving eff orts to protect cells at the mitochondrial level by stabilizing or restoring mitochondrial function or by interfering with energy metabolism appear to be promising; these eff orts also emphasize the importance of mitochondria in the pathogenesis of AD. Cellular models for AD as well as transgenic mice, and particularly triple transgenic models that combine both pathologies, are valuable tools in monitoring therapeutic interventions at the mitochondrial level. Eventually, this may lead to therapies that prevent the progression of Aβ deposits and tau hyperphosphorylation at an early stage of disease.