AD (Alzheimer's disease) is the most prevalent form of dementia in the aged population. Definitive diagnosis of AD is based on the presence of senile plaques and neurofibrillary tangles that are identified in post-mortem brain specimens. A third pathological component is inflammation. AD results from multiple genetic and environmental risk factors. Among other factors, epidemiological studies report beneficial effects of caffeine, a non-selective antagonist of adenosine receptors. In the present review, we discuss the impact of caffeine and the adenosinergic system in AD pathology as well as consequences in terms of pathology and therapeutics.
Alzheimer's disease pathology
AD (Alzheimer's disease) is the most frequently encountered form of dementia. In most cases, AD appears as a sporadic multifactorial disease resulting from the interaction of different environmental, epigenetic and genetic factors . Various epidemiological studies have allowed the identification of ‘risk factors’ and ‘protective factors’. High blood pressure, diabetes and obesity are detrimental factors [2–4], whereas physical and intellectual activities, as well as fish consumption have protective effects (, and references therein, ).
The definitive diagnosis of AD is based on the observation of characteristic brain lesions: senile plaques and neurofibrillary tangles. Senile plaques are characterized by the extracellular accumulation of the Aβ (amyloid β-peptide), whereas neurofibrillary degeneration is due to the pathological accumulation in the neurons of the naturally present tau protein. Aβ derives from a precursor called APP (amyloid precursor protein) through the combined action of two distinct proteolytic enzymatic activities, β- and γ-secretase releasing N- and C-terminal fragments of Aβ respectively [7,8]. Notably, soluble Aβ oligomers are thought to be the most neurotoxic species driving the detrimental impact of amyloid pathology ( and references therein). Neurofibrillary degeneration arises from the intraneuronal accumulation of proteinaceous fibrils in PHFs (paired helical filaments) made of hyperphosphorylated tau proteins, forming flame-shaped neurons (for reviews, see [10,11]). Tau is a neuronal protein located within the axonal compartment, essential for the organization, stabilization and dynamics of microtubules [10,11], but tau has additional important neuronal functions at the dendritic and nuclear levels [12–14]. The physiological and pathological functions of tau are also regulated by post-translational modifications, such as phosphorylation. Changes in tau phosphorylation may affect multiple tau functions and facilitate tau aggregation [10,11]. Importantly, spreading of the neurofibrillary lesions in the cortex (first entorhinal cortex, then hippocampus and lastly neocortex), corresponds to the progression of the symptoms . This supports the hypothesis that tau pathology is instrumental in cognitive alterations as stressed by observations that it impairs various forms of synaptic plasticity and cognitive tasks in mouse models [16–21]. Finally, neuroinflammatory processes are considered as a third pathological component in AD as recently pointed out by several genetic studies [13,22]. Although their respective contribution to the amyloid and tau sides of AD remain unclear so far, astrogliosis and neuroinflammatory events, especially those mediated by microglial cells, have an instrumental role in AD. An extensive description of their role in AD is beyond the scope of the present paper, but has been reviewed elsewhere [23,24].
Caffeine and Alzheimer's disease: from epidemiology to pathophysiology
The methylxanthine caffeine (1,3,7-trimethylxanthine) is the world's most popular psychoactive drug. The reason for this popularity lies in its psychostimulant properties combined with the absence of substantial negative side effects. Caffeine is contained in coffee, tea, soft drinks and chocolate. Overall, the psychostimulant properties of caffeine are due to its ability to interact with neurotransmission in different regions of the brain, thereby promoting behavioural functions, such as vigilance, attention, mood and arousal as well as improvement of cognition .
In humans, cognitive benefits of caffeine have been reported . Brief exposure improves memory and cognitive function in paradigms of scopolamine-induced impairments . Caffeine also improves attention and information processing . In rodents, evidence from the last few years support the cognition-enhancing properties of caffeine in a variety of behavioural tasks that evaluated learning and memory . Different longitudinal studies have investigated the relationship between coffee consumption, cognitive decline or dementia/AD. In the FINE (Finland, Italy and The Netherlands Elderly) Study, among elderly men, drinking three cups of coffee per day was associated with the least 10-year cognitive decline . Furthermore, results from the Three City Study among 65-year-old persons indicated that a consumption over three cups per day was associated with less decline in verbal memory and, to a lesser extent, in visuospatial memory among women .
Interestingly, other studies support the hypothesis that caffeine consumption might prevent AD. A retrospective study has shown an inverse correlation between coffee intake and the occurrence of AD later on in life since patients with AD had an average daily caffeine intake of 73.9±97.9 mg during the 20 years that preceded diagnosis of AD, whereas the controls had an average daily caffeine intake of 198.7±135.7 mg during the corresponding 20 years of their lifetimes . In the prospective CSHA (Canadian Study of Health and Aging), daily coffee drinking decreased the risk of AD by 31% during a 5-year follow-up . In line with those findings, another study from the CAIDE (Cardiovascular Risk Factors, Aging and Dementia) population reports that coffee drinkers at midlife had lower risk of dementia and AD later in life compared with those drinking no or only little coffee. The lowest risk (65% decreased) was found in people who drank three to five cups per day . Finally, recent prospective data indicated that high plasma caffeine levels were associated with a reduced risk of dementia or a delayed onset in patients with MCI (mild cognitive impairment) . It is noteworthy that in a recent study from the Honolulu–Asian Aging Study, authors did not find a significant association between caffeine intake and dementia risk . However, interestingly, they reported that, at autopsy, patients in the highest quartile of caffeine intake (>277.5 mg/day) were less likely to have any of the neuropatholgical lesions, such as AD-related lesions, microvascular ischaemic lesions, cortical Lewy bodies, hippocampal sclerosis or generalized atrophy. Therefore the available epidemiological data support the hypothesis that caffeine consumption is able to slow down cognitive decline in the elderly and reduces the risk of developing AD. Of note, although this also looks to be the case in Parkinson's disease , caffeine has been recently suggested to exhibit detrimental effects in Huntington's disease , suggesting that caffeine is not protective in all neurodegenerative conditions and that its effects depend on underlying instrumental mechanisms.
Several scientific studies support the hypothesis that caffeine is also beneficial in animal models which mimic the amyloid or tau sides of AD. Caffeine mitigates cognitive decline induced by Aβ and reduced the amyloid burden in transgenic mice overexpressing mutant APP (APPSw) in preventative, but also in therapeutic, paradigms. Indeed, APPSw mice chronically treated from 4 to 9.5 months of age with caffeine (300 mg/l by drinking water) were cognitively improved in several behavioural tasks that evaluated working and spatial memory and exhibited reduction of hippocampal Aβ1–40 and Aβ1–42 . Importantly, a similar treatment of APPSw mice at late pathological stages (18–19 months) for 4–5 weeks reversed memory deficits and reduced amyloid deposition as well as soluble Aβ levels in both entorhinal cortex and hippocampus . Such beneficial effects of caffeine against Aβ production were recently confirmed by a different group using an experimental model of sporadic AD based on feeding rabbits with a cholesterol-enriched diet that elevates both Aβ levels and tau phosphorylation in the brain . In this study, rabbits fed on the cholesterol-enriched diet were treated with low doses of caffeine (0.5–30 mg/day) through drinking water, corresponding to a maximal 60 mg/day consumption in humans (i.e. approximately one cup of coffee). In this paradigm, caffeine significantly decreased Aβ production in accordance with the results of Arendash et al. [39,40]. Interestingly, reduced production of Aβ1–40 and Aβ1–42 was also observed in a neuroblastoma model overexpressing mutant APP following treatment with concentrations of caffeine below 10 μM , supporting further the notion that caffeine affects mechanisms underlying Aβ production. In accordance, chronic caffeine treatment of APPSw mice has been associated with decreased PS1 (presenilin 1) and BACE1 (β-site APP-cleaving enzyme 1) protein expression as well as increased IDE (insulin-degrading enzyme) levels, the latter presumably contributing to enhanced Aβ degradation [39,41]. The effect of caffeine on BACE1 expression could relate to its ability to reduce c-Raf1 activity, possibly through PKA (protein kinase A) activation . In addition, caffeine would reduce GSK3 (glycogen synthase kinase 3) expression and/or activity and thereby influencing Aβ production . However, a direct effect of caffeine on γ-secretase activity remains elusive, and mechanisms linking caffeine and Aβ production/clearance deserve further evaluation. It is finally noteworthy that, although the beneficial effects of coffee on cognitive decline and decreased AD risk in humans has been mostly ascribed to caffeine, other coffee constituents may also play an important role towards amyloid pathology. Indeed, two previous studies have shown that non-caffeine components of coffee display neuroprotective effects in Drosophila melanogaster and Caenorhabditis elegans amyloid models through activation of the Nrf2 (nuclear factor-erythroid 2-related factor 2) detoxification pathway [42,43]. Interestingly, recent data indicate that caffeine may also have an impact on tau phosphorylation. Indeed, cultured neuronal cells exhibited reduced tau phosphorylation under caffeine treatment . Although caffeine doses used are far above the levels normally obtained following habitual consumption, this indicates a possible relationship to tau. In accordance, we observed recently that chronic caffeine consumption by tau transgenic animals (THY-Tau22 strain) developing neurofibrillary lesions, prevented memory alterations as well as decreased tau phosphorylation in the hippocampus .
Caffeine targets and Alzheimer's disease: the role of adenosine receptors
Only at high millimolar concentrations, which are irrelevant for normal consumption, can caffeine act at the level of ryanodine receptors and cyclic nucleotide phosphodiesterases, but it is now well established that, under normal habitual caffeine consumption, the effects exerted in the brain by caffeine depend on its ability to block adenosine A1 and A2A receptors . Adenosine receptors have a crucial neuromodulatory role and regulate both synaptic transmission and plasticity either by directly modulating synaptic responses or by interfering with other receptors .
During aging, we and others have found compelling evidence of cortical and hippocampal upsurge of A2A receptor expression/function. Specifically, in the hippocampus of aged rats, A2A receptor expression is nearly 2-fold higher compared with young rats [47,48]. More importantly, the A2A receptor-dependent activation of glutamate release becomes more pronounced as aging progresses, and shifts from PKC (protein kinase C)-mediated signalling to cAMP-dependent effects [48,49]. This is accompanied by clear behavioural deficits in hippocampus-dependent tasks, such as spatial memory in rats . Accordingly, rats overexpressing hippocampal A2A receptors also exhibit behavioural deficits including spatial memory defects as well as LTP (long-term potentiation) impairments (, and V.L. Batalha, J. Valadas, J. Coelho, M. Bader and L.V. Lopes, unpublished work). Interestingly, other detrimental conditions associated with cognitive impairment, such as hypoxia, diabetes or epilepsy, share similar A2A receptor overactivation ([48,52], and for a review, see ). Recently, we demonstrated decreased adult hippocampal LTP and cognitive/memory impairment in a chronic-stress-induced aging-like model, generated by maternal separation during the early postnatal period, in association with increased hippocampal A2A receptor expression . Strikingly, we observed, in adults, a normalization of synaptic and cognitive dysfunctions following A2A receptor blockade with the selective antagonist KW6002 , indicating an instrumental role of A2A receptor dysregulation in the genesis of synaptic dysfunctions underlying cognitive impairment within such an aging context.
Importantly, additional convergent data indicate that caffeine protects against the synaptoneurotoxicity induced by Aβ through blockade of A2A receptors. In primary cultures of cerebellar granule cells, low doses of caffeine (1–25 μM), comparable with those achieved following caffeine treatment in animals or moderate coffee consumption in humans , are able to counteract Aβ25–35 toxicity, an effect mimicked by ZM241385, an A2A receptor antagonist, but not CPT (8-cyclopentyltheophylline), a selective A1 receptor antagonist . These protective effects were confirmed in vivo. Subchronic treatment with caffeine at 30 mg/kg was shown to be protective against aversive and working memory deficits induced by i.c.v. (intracerebroventricular) injection of Aβ25–35 in mice  and mimicked by administration of SCH58261, a selective A2A receptor antagonist. A2A receptor blockade, through intraperitoneal injection of SCH58261 and KW6002 or genetic knockout, was also shown to prevent working memory impairment as well as synaptic loss induced by i.c.v. injection of Aβ1–42 [56,57]. Working memory improvement observed following A2A receptor blockade was thought to be related to the prevention of synaptotoxicity promoted by Aβ through modulation of p38 MAPK (mitogen-activated protein kinase) and mitochondrial function . Interestingly, it has been demonstrated that memory improvement promoted by A2A receptor blockade following i.c.v. injection of Aβ was not observed in amnestic conditions induced by MK801 or scopolamine . A2A receptor blockade would then mitigate detrimental synaptic effects of Aβ. To date, no studies have been published on the impact of A2A receptor modulation upon tau pathology in AD. However, we demonstrated recently the beneficial impact of A2A receptor deletion in a transgenic mouse model of AD-like tau pathology .
Besides its direct action on synapses, the effects of A2A receptor modulation upon AD pathophysiology could be non-neuronal. The role of A2A receptors expressed by astroglial and microglial cells is far from understood (for a review, see ). However, a few studies indicate that A2A receptors are up-regulated in both microglial and astroglial cells treated by Aβ [59,60]. A2A receptors may promote activation and proliferation of astroglial cells [61,62], and thereby regulate their ability to release glutamate  by controlling glutamate uptake , or affect the homoeostasis of the endogenous neuroprotectant adenosine via adenosine kinase . Astroglial A2A receptor up-regulation may thus contribute to the pathophysiology of AD. This idea is in line with the observation of the reinstatement of glutamate uptake in Aβ-treated glial cells following A2A receptor blockade . In addition, it has been shown that A2A receptor stimulation causes microglial activation  and potentiates the release of nitric oxide (NO) as well as prostaglandin E2 release from these cells [65,66]. Different experimental evidence supports the anti-inflammatory effect of A2A receptor blockade in different neuropathological situations [67–69]. Furthermore, A2A receptor blockade mitigates LTP defects as well as microglial activation and IL-1β (interleukin 1β) release following LPS (lipopolysaccharide) administration . Then, blocking microglial A2A receptor could be beneficial in AD. However, as microglia may play a Janus role in AD , this conclusion needs further evaluation. In particular, the role of A2A receptor activation towards the pathophysiology of AD will require further in vivo studies with appropriate and reliable cell-type-specific models.
Although recent data indicate that perinatal caffeine may impair interneuron migration and hippocampal network function , epidemiological and pre-clinical data, including ours, support the notion that caffeine might be not only a cognitive enhancer, but also a disease modifier in AD. The qualities of caffeine as a safe , inexpensive and brain-penetrating agent deserves the translation of these findings into a pilot clinical trial in AD patients. Despite epidemiological data on the effects of caffeine in aged and AD subjects, and data from animal studies, no clinical trials have been performed to date to evaluate the extent by which caffeine can slow down disease progression in patients that have already developed AD. In addition, encouraging preliminary experimental data have been obtained related to the role of A2A receptors in AD. Although more work is still needed to uncover specific A2A receptor functions in AD, all findings to date indicate that antagonists tested in human trials, even when not optimal in terms of efficacy, have reliably been shown to be safe and tolerable (, for reviews, see [73,74]). Therefore A2A-based trials will be feasible in the future, if we are able to better delineate the function of A2A receptors in the pathophysiology of AD.
Brain Disorders Across the Lifespan: Translational Neuroscience from Molecule to Man: An Independent Meeting held at University College Cork, Ireland, 12–13 September 2013. Organized and Edited by Eoin Fleming (University College Cork, Ireland).
D.Bl. and L.B. are supported by France Alzheimer, La Ligue Européenne Contre la Maladie d’Alzheimer (LECMA)/Alzheimer Forschung Initiative, LabEx (excellence laboratory) DISTALZ (Development of Innovative Strategies for a Transdisciplinary approach to ALZheimer's disease), Inserm, Centre National de la Recherche Scientifique (CNRS), Université Lille 2, Région Nord/Pas-de-Calais, Agence Nationale de la Recherche (ADORATAU), ERA-Net (ABeta-ID) and Fonds Unique Interministériel MEDIALZ. V.F. holds a grant from Inserm/Region Nord pas de Calais. C.L. holds a doctoral grant from Lille 2 University. D.Bo. is supported through grants from the National Institutes of Health [grant numbers NS065957, MH083973 and NS061844], the U.S. Department of the Army [grant number W81XWH-12-1-0283] and the Legacy Good Samaritan Foundation. L.V.L. is an FCT Investigator (iFCT). Her laboratory has support from Fritz-Thyssen Foundation, Bial and Fundação para a Ciência e a Tecnologia (FCT) (Portugal). D.B. is an Inserm investigator.
These authors contributed equally to this review.