Abstract
Clinical trials have extensively failed to find effective treatments for Alzheimer’s disease (AD) so far. Even after decades of AD research, there are still limited options for treating dementia. Mounting evidence has indicated that AD patients develop central and peripheral metabolic dysfunction, and the underpinnings of such events have recently begun to emerge. Basic and preclinical studies have unveiled key pathophysiological mechanisms that include aberrant brain stress signaling, inflammation, and impaired insulin sensitivity. These findings are in accordance with clinical and neuropathological data suggesting that AD patients undergo central and peripheral metabolic deregulation. Here, we review recent basic and clinical findings indicating that metabolic defects are central to AD pathophysiology. We further propose a view for future therapeutics that incorporates metabolic defects as a core feature of AD pathogenesis. This approach could improve disease understanding and therapy development through drug repurposing and/or identification of novel metabolic targets.
INTRODUCTION
The segment of population comprising people aged 60 and older is the fastest growing worldwide, and it is expected to more than double in the next 35 years [1]. However, living long does not necessarily mean living well, as age-associated diseases emerge and, for instance, the number of deaths caused by Alzheimer’s disease (AD) has considerably risen [2], despite some evidence for stable or declining incidence of dementia [3]. In fact, it is estimated that 47 million people live with dementia worldwide, causing huge economic and social hurdles.
AD is the most common cause of dementia, affecting around 60–70% of demented patients [4–6]. Although it primarily affects cognition, other debilitating non-cognitive symptoms may emerge including psychosis, mood alterations, impaired wake-sleep pattern, and appetite changes [7, 8], making AD an expensive and painful disease not only for patients, but also for caregivers and society. Recent works further suggest that central and peripheral metabolic disturbances, including insulin resistance and impaired glucose uptake, play key roles on AD risk and onset [9–11].
During the past decades, extensive basic and clinical research has expanded our knowledge on the cellular and molecular aspects of AD, contributing to the development of novel therapeutic approaches and tools for early diagnosis. However, even therapies initially considered extremely promising have been found somewhat disappointing in clinical trials [12, 13].
Recent results from Sevigny and co-workers suggest that the anti-aggregated amyloid-β (Aβ) antibody aducanumab reduces Aβ burden in a dose-dependent manner and causes slight improvements in memory scores in patients with prodromal or mild AD [14]. Conversely, several anti-amyloid strategies have failed to offer benefits to patients, and some clinical trials have even been halted due to safety concerns. For instance, the anti-Aβ antibody solanezumab failed in a large clinical trial tracking more than 2,100 people diagnosed with mild dementia due to AD for 18 months [15]. This suggests that current AD drug discovery pipeline might not be precisely addressing disease mechanisms. Still, additional candidate drugs with a wide range of mechanisms, including BACE and γ-secretase inhibitors, blockers of tau aggregation, and active and passive immunization strategies are currently in various stages of development and clinical trials, meaning there is hope that novel approaches will emerge shortly.
Mounting epidemiological studies and experimental evidence have supported a link between metabolic disorders and AD [10, 16–23]. A number of pilot clinical trials have recently indicated that drugs that modulate metabolism, including insulin and glucagon-like peptide-1 (GLP-1) analogs, may confer improvements in AD symptoms [24–26]. Hence, drugs targeting metabolic defects may have important therapeutic implications in AD.
Obesity and type 2 diabetes mellitus (T2DM) have been regarded as core metabolic disorders related to AD, as they share demographic profiles, risk factors, and clinical/biochemical features. Notably, both obesity/T2DM and AD have been associated with chronic inflammation, oxidative, and endoplasmic reticulum (ER) stress, and reduced neuronal sensitivity to insulin. Such molecular alterations are accompanied by energy metabolism deregulation and impaired glucose uptake [10, 27–32].
Since the emergence of initial studies proposing abnormal brain glucose metabolism as an important player in AD [33, 34], metabolic impairments that are hallmarks in diabetes and obesity have been postulated as core events in dementia [30, 36]. Indeed, PET-based measures have described abnormal glucose utilization in AD brains [26, 37] and evidence from ex vivo studies using AD tissues established that demented brains are less responsive to insulin/IGF-1 stimulation than controls [38]. Additional findings have demonstrated early Aβ deposition, decreased glucose metabolism, structural changes, and functional disruption at the same cortical midline brain regions vulnerable to AD changes [39–42]. This notion has been substantially reinforced after other molecular hints indicated that inflammation and defective insulin signaling are present in AD brains [10, 29–31].
Accumulating reports have further established that life habits, feeding behavior, and environmental factors throughout life could contribute to increase susceptibility to sporadic AD [10, 44]. Obesity, for example, is associated with poorer cognition in non-demented subjects [45], and comprises a risk factor contributing to AD development [46]. A large-scale study that followed 10,136 participants for 36 years reported that participants that were overweight in midlife had three-fold increased risk to develop AD than those with normal weight [47]. Interestingly, this study found an association between obesity and AD even after corrected covariation for hyperlipidemia, hypertension, and diabetes, suggesting that body weight is an independent risk factor for AD. Another study conducted by the Cardiovascular Health Consortium also found positive associations between body mass index and AD, reporting that obesity in midlife was associated with a 40% increase in the risk for developing this form of dementia [48].
High adiposity has been associated with alterations in brain structure in late-life, such as brain atrophy and white matter lesions especially in brain regions involved in memory processing, such as the amygdala, hippocampus and frontal cortex [49]. Elevated fat consumption was also shown to increase levels of soluble and insoluble Aβ in the parietal-temporal cortex of aged transgenic AD mice as compared to wild-type animals, indicating that high adiposity could accelerate AD pathology [50]. Thus, controlling obesity and T2DM throughout midlife could represent a modifiable risk factor not only for cardiovascular and metabolic disease but also for AD, and future studies are warranted to explore such interventional approaches.
Therefore, attempts to halt metabolic defects in early stages of AD, as well as strategies aimed at preventing metabolic deregulation, could be key to slow AD progression or even reduce its risk. Here, we review clinical and preclinical evidence supporting metabolic deregulation as a core derangement in AD, and discuss potential therapeutic implications of such findings. Finally, we offer a perspective for future therapeutic approaches that takes metabolic dysfunction into account in AD.
METABOLIC DEFECTS IN AD
Mitochondrial dysfunction
Mitochondria play pivotal roles in cell survival by regulating energy metabolism, reduction-oxidation potential, and apoptotic pathways [51]. They have recently been demonstrated to actively participate in neurotransmission by locally controlling ATP and metabolite levels at synapses [52–55]. Thus, it is conceivable to speculate that alteration of mitochondrial structure, localization, and function could affect neurotransmission and neuronal function, ultimately impinging on cognition.
Current evidence suggests that mitochondrial abnormalities and oxidative damage are early events in AD, and may precede pathological hallmarks [56–58]. AD brains present reduced expression and/or activity of key enzymes of mitochondrial oxidative metabolism, including α-ketoglutarate dehydrogenase, pyruvate dehydrogenase, and cytochrome oxidase [59–61]. Neurons from AD patients exhibit overall decrease in mitochondrial mass, aberrant mitochondrial DNA release to cytosol, and increased mitophagy [62–69]. Aβ-induced mitochondrial dysfunction further potentiates the opening of the mitochondrial permeability transition pore (PTP) induced by Ca2 + [70, 71], which contributes to the release of pro-apoptotic proteins, such as cytochrome c and apoptosis-inducing factor.
Neuronal oxidative stress, a consequence on mitochondrial dysfunction, has been extensively demonstrated in the brains of patients and in experimental models of AD [56–58, 73]. Although physiological levels of reactive oxygen species are essential for brain function [74], neurons are especially sensitive to them, and prolonged oxidative stress may thus result in neurodegeneration [75, 76]. In line with this, blocking oxidative stress prevents AD-related neurotoxicity in AD models [62].
Mitochondria are highly dynamic organelles that undergo continual fusion and fission events, with impacts on mitochondrial biogenesis, morphology, trafficking, and degradation [77]. Mitochondrial fusion and fission events are imbalanced in AD [78–81], similar to obesity-related alterations [82], and experimental models have further revealed defective mitochondrial transport [31, 83] and increased fragmentation [62, 78] in neurons undergoing AD-related neurotoxicity. These events likely contribute to the metabolic failure germane to AD.
Brain glucose metabolism
For more than three decades now, 18F-fluorodeoxyglucose (FDG)-based PET has been used to demonstrate impaired glucose metabolism in AD, as compared to healthy subjects [84]. Such an approach has revealed that disease progression positively correlates with reduction of cerebral glucose metabolism with marked effects in areas notably affected by AD, including the posterior parietal lobe and portions of the temporal and occipital lobes [85]. Interestingly, APOE4 carriers, who are at higher risk of developing AD, present weaker FDG-PET signals decades before any clinical manifestation, suggesting that defects in brain metabolism may precede dementia onset [86]. Consistently, AD transgenic animal models also develop hypometabolic profiles in FDG– PET [87–89].
Still, the specific reasons that lead to reduced FDG signals in AD are unclear. An immediate explanation for altered FDG signals in AD brains could be decreased expression/function on glucose transporters. Indeed, impaired expression of GLUT1 and GLUT3 has been observed in AD brains [90, 91]. Reduced levels of glucose transporters are likely to contribute to synaptic dysfunction, tau phosphorylation [90, 92], and vascular pathology [93] in AD models. Conversely, increasing GLUT1 expression was shown to rescue Aβ-induced neurotoxicity [94, 95], further supporting its potential role in AD.
In line with the impaired glucose uptake in AD, glucose phosphorylation by hexokinase appears to be reduced during the course of the disease [96]. It is noteworthy that aerobic glycolysis was recently shown to be reduced during normal aging [97], and this could be exacerbated in AD. Mechanistic hints for such events may have found place in that AD-associated soluble Aβ oligomers (AβOs) dampen hexokinase activity and reduce ATP levels in primary neurons in culture [65, 66]. Additionally, AβOs lead to transient inhibition of the metabolic sensor AMP-activated kinase (AMPK), which causes GLUT3 and GLUT4 removal from the neuronal surface [65]. Such metabolic responses could lead to compensatory longer-term increases in AMPK activity, thereby resulting in the aberrant AMPK overactivation described in the brains of AD patients and transgenic mouse models [98–101].
Thus, defective glucose uptake in AD brains could result from compromised metabolic routes, including perturbed exposure of GLUTs and impaired metabolic sensing by AMPK. Impaired neuron-astrocyte-vascular interactions and signaling could further exacerbate metabolic dysfunction, ultimately leading to the observed declines in FDG signals and brain function in AD patients.
Insulin resistance
An important player accounting for impaired glucose metabolism in AD could arise from defects in insulin signaling. Historically, the skeletal muscle, adipose tissue, and liver have been considered the main insulin-responsive tissues in control of peripheral metabolism. On the other hand, the brain was classically considered an insulin-insensitive organ until the initial observation that intracerebroventricular infusion of insulin reduces food intake and body weight in baboons [102].
In fact, insulin and insulin-like growth factor receptors are widely distributed throughout the encephalon [103]. The hippocampus and cortical formations present significant expression of theses receptors and are regions centrally involved in memory formation [104–106]. In accord, insulin was shown to be neuroprotective [31, 107–111], and to promote synapse plasticity [112, 113] and cognitive function in healthy subjects [108, 114–116]. Conversely, downregulation of brain insulin receptors was shown to promote tau phosphorylation [117], synaptic impairments, and memory loss [105, 118].
The sequence of events leading to brain insulin signaling dysfunction in AD is not completely understood, but resembles, in many aspects, the molecular steps described for T2DM in peripheral tissues. Hints into the mechanisms of neuronal insulin signaling dysfunction in AD came from experiment using primary hippocampal neurons showing that AβOs induced the removal of insulin receptors from the surface of neurons, an effect that was prevented by insulin itself or by insulin-sensitizing drugs [110]. Recently, tau deletion was shown to promote brain insulin resistance through aberrant PTEN activity, arguing for a role of tau loss-of-function in the deleterious effects in AD [119]. Further, ex vivo insulin stimulation in slices derived from human AD brains revealed an impairment of insulin signaling compared to tissue from age-matched controls [38]. Also, AD patients exhibited increased levels of serine phosphorylation in the insulin receptor substrate 1 (IRS-1 pSer616 and IRS-1 pSer636/639) that negatively correlated to memory scores [38]. This is in full accordance with early studies that established that expression and activity of brain insulin signaling components are reduced in AD [21, 120–122]. Further, soluble Aβ-injected mice [109] and cynomolgus monkeys [31] present increased levels of IRS-1 pSer636/639, in line with AD patient data. Thus, impaired brain insulin signaling could compromise survival and synaptic plasticity mechanisms, likely cooperating to memory defects in AD. Therefore, boosting the insulin signaling pathway in the brain may represent an important alternative strategy for AD treatment (Fig. 1).
Cellular basis of brain metabolic dysfunction in AD and potential therapeutic strategies. Brain accumulation of soluble Aβ aggregates causes increased microglial reactivity, thereby resulting in release of pro-inflammatory cytokines (e.g., TNF-α), which in turn triggers neuronal dysfunction. Soluble Aβ aggregates may further act directly on neuronal synapses to impair neuronal homeostasis. Neuronal stress signaling is characterized by elevated inhibitory phosphorylation of both insulin receptor substrate-1 (IRS-1pSer636) and eukaryotic translation initiation factor 2α (eIF2α-P). Such orchestrated response underlies brain insulin resistance and synapse impairments in brain regions relevant for memory (hippocampus) and peripheral metabolic control (hypothalamus) in AD, and approaches aimed at targeting these noxious mechanisms have been under investigation. Immunofluorescence images depict cultured rat hippocampal neurons stained for Aβ aggregates (upper right; red), IRS-1pSer636 (bottom left; yellow) or eIF2α-P (bottom right; green) after exposure to soluble Aβ aggregate preparations for 3 hours. Scale bar: 20 μm.
Brain inflammation
Preclinical, clinical, and epidemiological evidence has indicated that inflammation is an important contributor to AD pathogenesis. Several studies have shown that markers of inflammation are increased in the brains, cerebrospinal fluid, and plasma of AD patients [123, 124]. These include TNF-α, IL-1β, IL-6 and other cytokines, as well as indicators of glial reactivity and infiltration of peripheral immune cells [125–128].
Although predicted in the original amyloid cascade in the early 1990s, only recently detailed insights on how brain inflammation may take place in AD came out. Microglia has been placed at the center of AD-linked inflammation and, while normal microglial activation is fundamental for Aβ clearance, chronic inflammation generates detrimental effects that promote AD pathology [125, 129]. Current notions suggest that microglial function go awry, resulting in increased pro-inflammatory signaling, reduced Aβ clearance and aberrant synaptic pruning [125, 130]. Microglia from AD mouse models present impaired phagocytosis capacity, degrade less Aβ, produce toxic signals, and exacerbate neuronal damage in AD models [126, 132].
Microglial-derived cytokines enhance AβPP processing, induce tau phosphorylation, and contribute to synapse plasticity impairment in neurons [133, 134]. Pro-inflammatory cytokines have further been implicated in memory deficits and depressive-like symptoms in AD models [109, 136]. Microglial signals may prompt astrocytes to assume neurotoxic phenotypes, further contributing to neuronal damage in AD [137]. Finally, reactive microglia induce synapse loss in AD models by stripping off synapses through a complement-dependent recognition system [138, 139].
Genetic studies have pointed that loss-of-function mutations in the Triggering Receptor Expressed on Myeloid Cells 2 (TREM2), notably expressed in microglia, increase up to 4 times the risk of AD in humans [140, 141]. TREM2 is essential for microglia survival, activation, and phagocytosis [142–145]. TREM2-deficient AD mice had impaired microglial metabolism [146], and failed to activate microglia surrounding plaques and to respond to injury, resulting in increased amyloid burden [144, 145]. It is noteworthy that a very recent study uncovered that, in addition to TREM2, variants of the microglial-expressed genes PLCD2 and ABI3 are associated to either protection or increased risk of AD [142]. Their results implicate innate immunity function in AD, further offering a genetic basis for the link between brain inflammation and AD.
Although microglial actions can trigger detrimental processes in the brain, they also play fundamental roles to maintain brain homeostasis. They release neurotrophic factors, such as BDNF and IGF-1, to influence neuronal survival and synaptic plasticity [147, 148]. It is thus possible to speculate that loss of proper microglial function could itself be harmful in brains undergoing neurodegeneration. Thus, compounds aimed at keeping microglia in good shape are highly warranted for preclinical and clinical AD testing.
In T2DM, increased levels of pro-inflammatory mediators, especially TNF-α, act in the hypothalamus and in peripheral tissues causing activation of intracellular stress kinases such as c-Jun N-terminal kinase (JNK), IκBα kinase (IKK) and double-stranded RNA-dependent protein kinase (PKR). These kinases trigger serine phosphorylation of IRS-1, thereby blocking downstream actions initiated by insulin. As in insulin-resistant peripheral tissues, TNF-α-induced hippocampal activation of JNK has been described in brains of AD transgenic mouse models, Aβ-injected mice and cynomolgus monkeys, and in postmortem analyses of AD brains [31, 121]. Defective insulin signaling has been shown as a consequence of stress kinase activation in AD brains and in several experimental models of AD, in which it was shown to contribute to memory impairment [31, 99].
Furthermore, involvement of both IKK and PKR has been described in AD-linked insulin signaling dysfunction in hippocampal neurons [31, 109]. Notably, blockade of TNF-α with the neutralizing antibody infliximab or genetic deletion of TNF-α receptor 1 led to improved insulin sensitivity [31], normalization of memory performance, and rescued depressive-like behavior in AD mouse models [109, 149–152]. A role for peripheral TNF-α has further been corroborated by recent findings that systemic infusions of anti-TNF-α antibodies rescue memory and glial reactivity [153]. These results provide additional evidence for a close parallel between inflammation-associated defective brain insulin signaling in AD and chronic inflammation-induced insulin resistance in peripheral tissues.
Neuronal stress signaling and defective proteostasis
Defects in protein homeostasis, or proteostasis, have been recently associated with neuronal malfunction and cognitive impairment in AD. Phosphorylation of the eukaryotic translation initiation factor 2α (eIF2α) at serine 51 (eIF2α-P) by stress kinases, including PKR, attenuates general protein synthesis, and its sustained elevation has already been associated to memory impairment in rodents [154–156]. Importantly, increased levels of eIF2α-P were found in the brains of AD patients [99, 157–160], as well as in animal models, including APP/PS1 mice, and Aβ-injected mice and cynomolgus monkeys [109, 157]. Moreover, PKR-dependent eIF2α-P appears to be initiated by TNF-α signaling, ultimately leading to hippocampal synapse loss and memory failure [109].
Suppression of two additional eIF2α kinases, PERK and GCN2, was shown to alleviate AD-linked inhibition of long-term potentiation, and to restore spatial memory impairment in AD transgenic mice by replenishing hippocampal protein synthesis [157]. These results were confirmed by additional studies showing neuroprotective actions of PERK inhibition/ablation in AD models [161, 162]. A recent study demonstrated that metabotropic glutamate receptor-dependent long-term depression, which is also impaired in AD mutant mice, is recovered when PERK activation is suppressed, further indicating that normalization of eIF2α-P levels improves synaptic plasticity [163]. In accordance, activation of PKR and PERK has been reported in postmortem AD brains [159, 164–167], likely suggesting clinical relevance to these experimental findings.
In addition to attenuation of general protein synthesis, sustained eIF2α-P paradoxically leads to the enhanced translation of selective mRNAs, including that of activating transcription factor 4 (ATF4), a repressor of long-term synaptic plasticity and memory that counteracts pro-memory signaling [168, 169]. Increased levels of ATF4 have been found in the brains of AD patients and transgenic mouse models [157, 170], and aberrant translation of ATF4 in axons mediates neurodegeneration in the AD brain [158]. Finally, elevated eIF2α-P has been shown to increase BACE activity and amyloidogenesis in mouse models [171, 172], likely contributing to amyloid build-up in human AD.
Taken together, these results point to novel molecular mechanisms of cognitive decline in AD initiated by metabolic impairments and resulting in defective proteostasis and synaptic function. Findings further suggest that targeting stress kinases and eIF2α-P levels might be interesting future approaches to restore neuronal homeostasis and synaptic function in AD (Fig. 1). Pharmacological modulators of eIF2α-P actions have now begun to emerge [173–177], and future studies may assess their preclinical potential in AD and other forms of neurodegeneration.
Peripheral metabolic dysfunction and the hypothalamus
The hypothalamus is a brain region with prominent endocrine actions that regulate, among other physiological functions, sleep/wake cycle, body temperature and, importantly, food intake and lipid/carbohydrate metabolism [178]. Neuroendocrine studies have revealed that aberrant pro-inflammatory and stress signaling pathways in the hypothalamus are sufficient to deregulate peripheral metabolism in diabetes/obesity pathophysiology [179]. Hypothalamic nuclei are highly responsive to peripheral signals, such as those mediated by insulin and leptin. However, sustained hyperinsulinemia, typical of metabolic derangements such as obesity and diabetes, were shown to cause signal-resistance in the hypothalamus [179].
Extensive evidence indicates that low-grade inflammation takes core place in the hypothalamus to impair body metabolism. Hypothalamic disturbance is driven at a molecular level by several of the inflammatory pathways mentioned above to mediate peripheral effects of altered metabolism [180].
Overfeeding and obesity cause a nutrient overload that includes an elevation in circulating levels of free fatty acids [181], which, in turn, stimulate ER stress in hypothalamic neurons [182, 183]. In parallel, high free fatty acid levels directly activate toll-like receptors triggering immediate transduction of pro-inflammatory intracellular cascades in hypothalamic neurons [184]. In addition, central and peripheral cytokines appear to contribute to hypothalamic dysfunction [185, 186].
Such orchestrated response, mediated by stress kinases and transcription factors, will result in defective proteostasis, neuronal insulin/leptin resistance and in a transcriptional shift toward a neurotoxic profile [178, 188]. The main outcome is aberrant hypothalamic function and impaired control of body metabolism in obesity [189]. Therefore, mechanisms that actively operate to damage hippocampal/cortical neurons in AD resemble those that mediate central deregulation of body metabolism in obesity. This notion has led to the hypothesis that the hypothalamus might be affected in AD, thereby offering an explanation on why AD patients develop peripheral metabolic impairments, such as insulin resistance and hyperglycemia.
Initial discoveries suggested that the hypothalamus might indeed be a key brain region that presents amyloid plaque pathology, and that hypothalamic dysfunction occurs early in disease [7, 180]. These studies identified amyloid deposits in AD brains [190, 191], and brain imaging studies revealed reduced hypothalamic volume in early AD patients when compared to non-cognitively impaired subjects [192]. Another study described neurodegeneration in the hypothalamus, with shortened dendritic arborization and synapse pathology in early AD patients [193]. In Aβ-injected rats, accumulation of fibrillar aggregates in the hypothalamus was detected up to three weeks after the injection and was accompanied by hypothalamic astrocytosis [194].
Hypothalamic inflammation, ER stress, and insulin resistance were demonstrated in Aβ-injected mouse and cynomolgus monkeys [32]. Hypothalamic dysfunction was associated with development of persistent peripheral glucose intolerance, which was further observed in different AD transgenic mouse models [32, 196]. Blockade of TNF-α mediated signaling pathways or alleviation of ER stress normalized glucose tolerance [32], indicating that diabetes-linked mechanisms may operate in the hypothalamus to impair peripheral metabolism in AD.
In addition to peripheral metabolism deregulation, it is noteworthy that hypothalamic defects could also underlie other non-cognitive aspects of AD. Hypothalamic nuclei responsible for circadian rhythm maintenance are affected in AD patients and animal models [197–201], raising the possibility that impaired neuronal function in the hypothalamus accounts at least partially for sleeping pattern disruption and aggressive behavior. For instance, several studies have now investigated how sleep becomes deregulated in AD [198, 203], and further studies are warranted to investigate whether hypothalamic inflammation could, at least in part, mediate sleep disturbances in AD.
NOVEL GROUNDS FOR AD RESEARCH
Is it established that metabolic dysfunction comprises a risk factor for AD?
A substantial body of evidence supports that brain insulin signaling deregulation in AD could represent a clinical link between T2DM and dementia [30, 204]. Confirmation of this notion is central for the development of effective approaches in dementia. As not all reports have found significant effects of metabolic impairments in AD, more discussion on whether T2DM/obesity and other metabolic defects are causally linked to dementia is warranted.
A meta-analysis of 12 studies in large cohorts found a mild association between metabolic syndrome (MetS) onset and poorer cognitive performance [205]. When individuals were separated by age, however, a stronger correlation was observed among the younger (<70 years old) rather than elderly patients, who might present other age-related factors masking putative effects. Indeed, patients that suffer from T2DM and/or MetS comprise very heterogeneous populations. They may take different medications, have different lifestyle habits, and often present different comorbidities, and this should be taken into account in such epidemiological studies.
Furthermore, clinical assessment of MetS does not follow unified criteria. With its onset defined by any three out of five parameters, it is very likely that components of MetS have differential impact on cognition and risk of dementia. Accordingly, an association study demonstrated that obesity had a closer relationship to mild cognitive impairment than other MetS factors [206]. Moreover, a significant positive association between fasting plasma insulin levels and cognitive dysfunction has been reported in a Danish MetS study [207]. Therefore, though a connection between impaired body metabolism and dementia has become increasingly clear, clinical and epidemiological analyses should use careful methodologies to isolate principal components associated with AD onset and progression [208].
Additionally, most epidemiological approaches addressing the connection between impaired metabolism and dementia are rather descriptive, and have not been often confirmed by interventional studies [209]. For example, while higher blood glucose levels have been associated to declining memory [23, 210], data from ACCORD (Action to Control Cardiovascular Risk in Diabetes) trial has not shown beneficial effects of managing glycemia or lipidemia on cognition [211, 212]. Conversely, multidomain interventions aimed at reducing risk factors and controlling body metabolism have been proposed as effective means of reducing dementia incidence [43, 44]. A pioneer investigation has already shown that such approach could preserve cognition in at-risk subjects [213], and replication attempts are underway in the FINGER study (NCT01041989) [214]. Future clinical studies and meta-analyses should investigate possible effects that might explain the contrasting observations in the field and should test the propelled hypothesis that lifestyle interventions aimed at improving metabolism could reduce AD risk at later stages of life.
The future of translational research in AD
Eleven decades have passed since AD was initially described [215] and, despite several proposed etiogenic hypotheses, no drug or approach has been shown to effectively reverse or even slow down dementia progression. Moreover, AD remains largely idiopathic, with only a small fraction of the cases explained by familial mutations, and with yet unknown bona fide molecular predictors or diagnostic biomarkers. It is thus not surprising that clinical trials have failed or been halted. Given that this apparently unsuccessful story is not due to lack of interest or research, it tells us that shifts in current AD research pipeline might be required.
AD has been classically viewed as a proteinopathy in which accumulation of Aβ and tau plays significantly roles on synapse and memory dysfunction [216–218]. From the early concept of insoluble plaques as drivers of memory impairment to a more recent and refined concept of neurotoxicity of soluble Aβ species, the amyloid cascade hypothesis has reigned as the most influential paradigm for AD pathogenesis in the past 30 years [219, 220]. Not less important, however, are the notions that non-canonical forms of tau and ApoE4 trigger neuronal dysfunction and memory loss in animal models [170, 221–228], likely accounting for human AD pathophysiology. There is growing indications that soluble Aβ and tau can be secreted, diffuse trans-synaptically, adopt prion-like behaviors in the brain, and impair synaptic plasticity [229–237]. Notably, synapse plasticity and memory deficits triggered by both soluble Aβ or tau appear to depend on interactions with AβPP [225].
Although Aβ levels start to rise decades before initial clinical symptoms appear in AD-linked mutation carriers [40, 239], the underlying causes for increased brain Aβ and tau in sporadic AD remain poorly understood. An attractive hypothesis postulates that accumulation of injuries throughout life may sum up with poor habits and lifestyles to favor AD onset at later stages [10, 241] (Fig. 2). Thus, midlife metabolic diseases, including obesity and diabetes, as well as traumatic injuries, could create the neurotoxic conditions for gradual increases in amyloid and tau pathologies to take place in sporadic AD. Aβ and tau, in turn, could exacerbate brain dysfunction by acting on neurons, astrocytes and microglia, and promote the neurodegeneration observed in late AD.
Risk factors and metabolic defects in AD. Several life conditions have been associated with an increased risk of developing AD. Although some of these disorders, including depression and traumatic brain injury, have been primarily linked to changes in the brain, emerging evidence indicates they might also impact peripheral tissues, such as liver, skeletal muscle, and the adipose tissue. These diseases could also harm additional organs, including pancreas and the gut. In addition, midlife obesity, type 2 diabetes, sedentarism, and poor sleeping habits appear to negatively affect both the brain and periphery. The deleterious impact of such conditions may result in the metabolic defects that favor the onset of AD, including inflammation and insulin resistance (in brain and periphery), and defective mitochondrial function and cellular proteostasis in the brain. Furthermore, improper bidirectional communication between the brain and peripheral tissues through neurotransmitters, hormones, and cytokines might contribute to originate the pathophysiological features of AD. Reducing the metabolic impact of AD risk factors might be key to reducing the number of new cases of dementia in the future.
On the other hand, individual genetic features, such as single-nucleotide polymorphisms, mutations or alleles, could represent additional predisposition traits to set the pace for AD onset in association with environmental conditions. Genetic studies have recently taken major steps forward with the identification of TREM2 variants as a risk factor for AD [140, 141]. Follow-up discoveries that have implicated TREM2 loss-of-function in the impairment of microglial function [142–144], with consequences to neuroinflammation and Aβ clearance, as the pathophysiological underpinnings of increased AD risk.
Resolution offered by molecular biology and genetic studies has substantially increased with advancing technologies and will shed light on genomic variations that affect the risk for AD and other forms of dementia. Exciting news are that US government has supported a large, controlled, clinical study to map genome variations in AD patients [242], and understanding genetic variations in AD has been set as one of the priorities of the newborn UK Dementia Research Institutes.
As most drugs and therapies tested in rodent models that advanced into clinical trials had disappointing outcomes, history tells us that strategies that were developed based on single assumptions for AD pathogenesis have been misleading, and that early intervention are key to success. Thus, the future of AD research may benefit from 1) improving AD modeling, taking sporadic variables and human-specific traits into consideration; 2) developing efficient diagnostic tools to detect at-risk cohorts as early as possible; and 3) testing combination therapies that target more than a single aspect of disease.
In this context, animal models of AD have recently taken major leaps forward by ongoing attempts to develop non-human primate models [243, 244], which likely better resemble human pathology, and by next-generation AβPP knock-in models [245, 246] that are less prone to neurotoxicity by non-Aβ fragments of AβPP. Still, the field demands future models that are less based on familial AD mutations and that comprise more features of human AD, including neurofibrillary tangles. Second, it is imperative that the complex nature of AD be discriminated by better diagnostic and prognostic biomarkers. Recent discoveries have raised interest and excitement, and a combination of imaging, neuropsychology, and fluid biomarkers might result in more accurate tracking of AD onset and progression. [247–249]. Lastly, although reigning Aβ-targeting approaches should not be completely put off the game, there is an exciting trend to move forward with preclinical and clinical testing of combination therapies, which might likely yield more favorable results in large trials.
Repurposing drugs to accelerate disease targeting
Drug development targeting the central nervous system has traditionally high failure rates. For instance, the approval likelihood of new AD drugs between 2002 and 2012 reached only 0.4%, whereas cancer and cardiovascular drugs hit the approval rate of 6.7%, and 7.1%, respectively [12, 250]. Given the urgent need to combat the global burden of AD, policymakers and science leaders have gathered efforts to find ways to effectively treat or prevent AD by 2025 [251]. Nonetheless, if one considers the traditional pipeline from basic research to ultimate clinical testing, it is inevitable to realize the long road until a novel treatment can be labeled as safe and effective for any human disease. Therefore, strategies aimed at repurposing already marketed drugs become an interesting option to accelerate drug discovery for AD and other diseases [252].
The abundant body of data indicating that anti-diabetic compounds could be neuroprotective in preclinical AD studies and in pilot clinical trials has fostered clinical trials in larger cohorts [24, 29]. Insulin, the most well-known anti-diabetic compound, has advanced to clinical trial aimed at determining whether mild-to-moderate AD patients may present memory benefits with continued intranasal delivery (SNIFF; clinical trial ID NCT01767909). This investigation has received significant support after demonstration that intranasal insulin enhances memory in non-cognitively impaired and early AD subjects, and that insulin is neuroprotective against AD-related synapse loss [31, 242]. Exenatide and liraglutide, two compounds already labeled for T2DM management, have advanced into initial clinical trials (NCT01255163 and NCT01843075, respectively) after substantial preclinical investigation [26, 253–257].
Given the support for a role of neuroinflammation in AD pathogenesis, it is tempting to hypothesize that anti-inflammatory approaches could also represent effective therapeutics in AD. Results from clinical trials, nonetheless, have been contradictory. Although lifelong use of non-steroidal anti-inflammatory drugs (NSAIDs) was associated with reduced risk of developing AD [258], clinical trials, unfortunately, did not reveal beneficial outcomes for AD patients [259, 260].
Studies with aspirin, nimesulide, ibuprofen, rosiglitazone, and pioglitazone, for instance, have not shown positive effects in randomized clinical trials so far [259, 261–264]. A more detailed investigation in one clinical trial, though, has revealed that naproxen effect vary depending on the stage of the disease. It accelerated AD pathology on later stages of the disease, while it reduced AD risk on preclinical stages [265]. This dual effect of NSAID on AD depending on the disease stage possibly mirrors pleiotropic roles of microglia on the disease. Such apparently disappointing results coming from NSAIDs could be due to the fact that anti-inflammatory agents target generic rather than specific neuroinflammatory components in AD. Thus, future studies may reveal effects of labeled drugs on more refined targets, including microglial modulators, hopefully resulting in more effective strategies for AD therapy. Additionally, it is likely that chronic low-grade inflammation takes place in the brain to cause abnormal elevation of Aβ, tau and neuronal dysfunction and long before initial symptoms emerge in AD. Thus, preventing pro-inflammatory conditions and treating inflammation as early as possible need to be tested as potential ways of slowing AD progression.
Fostering tests on repurposed drugs could be key to accelerate disease targeting pipeline in AD. Table 1 summarizes the current depth of pre-clinical and clinical evidence for drugs that could be repurposed for AD treatment, thus representing new hopes for treating dementia. Advancing on the therapeutic pipeline, it will be now needed to determine whether promising modulators of metabolism could indeed represent disease-modifying approaches in AD.
Repurposing drugs for AD therapy: current preclinical and clinical evidence from metabolism-targeting approaches
CONCLUDING REMARKS
Knowledge on the complex nature of AD pathophysiology has considerably evolved over the past decades, even though this gain of information has not translated into effective therapies yet. Accumulating observations have implicated metabolic defects, including dyshomeostasis of glucose metabolism, insulin resistance, and disturbed proteostasis, in the course of AD pathogenesis. Impaired metabolism may arise from a combination of genetic and environmental components to increase the risk of AD development, and could further drive cognitive and non-cognitive symptoms, and neurodegeneration.
Considering AD as a metabolic disease and understanding the mechanistic links among AD, obesity, and T2DM could be helpful steps toward developing effective strategies for AD prevention and treatment. Repurposing agents already approved for the treatment of metabolic disorders may have clinical relevance for AD, as they have already been through preclinical toxicology assessments, human safety, tolerability, and pharmacokinetic assessments. Some clinical trials are now underway and conclusive results might be available in the upcoming years. There is growing agreement that combination therapies might yield more effective results in AD. In this scenario, targeting metabolic impairments might open new avenues to develop alternative therapeutic strategies with higher chances of success.
Footnotes
ACKNOWLEDGMENTS
Work in the authors’ laboratories has been supported by grants from Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq/Brazil), Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ), International Society for Neurochemistry (ISN), and the National Institute for Translational Neuroscience (INNT/Brazil).
