11.1 Alzheimer’s Disease (AD): A Continuum Pathological Process

Increasing evidence support that the pathological process of Alzheimer’s disease (AD) initiates years or decades before the diagnosis of dementia, a syndromal state that encompasses memory deficits, problems in language, in planning, and execute familiar tasks and in other cognitive skills that affect the individual’s ability to perform daily actions. AD constitutes the most common cause of dementia in the elderly people, accounting for an estimated 70–80% of cases (Braak and Braak 1991; Selkoe et al. 2012). Currently, nearly 44 million people have AD or a related dementia worldwide, and this number is expected to triple by 2050 as the size of elderly population rise (Prince et al. 2016). AD frequently co-occurs with other pathologies that might also contribute to dementia, a situation designated as multiple etiology dementia , resulting in AD plus Lewy body dementia (LBD) or AD plus vascular cognitive impairment (VCI)/vascular dementia (VaD), although both LBD and VCI/VaD can lead to dementia in the absence of AD (Montine et al. 2014). This cognitive neurodegenerative landscape is one of the most devastating brain illnesses in terms of handicap, clinical evolution and treatment, thus representing a major socioeconomic concern for the twenty-first century (Prince et al. 2016). Moreover, it is also a challenge for the scientific community, because it remains to be defined the precise biological changes that cause AD, why it progresses more quickly in some individuals than in others, and how the disease can be prevented, slowed or halted.

AD is an age-related progressive disorder that usually begins with a preclinical asymptomatic stage , in which it is claimed to occur an evolving brain damage underlying very subtle deficits in cognition (Sperling et al. 2011; Musiek and Holtzman 2015). This stage may precede the mild cognitive impairment (MCI), a condition in which subjects are only mildly impaired in memory with relative preservation of other cognitive domains and functional activities; thus subjects are cognitively altered, but they do not yet meet the criteria for dementia, and this MCI condition is also designated as prodromal AD state (Petersen et al. 2009; Dubois et al. 2010). Although not all MCI cases progress to a dementia state, epidemiological studies showed that MCI patients, who display significant heterogeneity in their profile of cognitive impairment, progress to AD at rate of 10–15% per year (Weiner et al. 2015). Thus, forecasting who, among a group of MCI patients will be more likely to further decline in cognition and convert into an Alzheimer’s dementia, would be crucial to ensure an early intervention and to establish effective therapeutic strategies to treat, halt or prevent AD. Over the last years, neuroimaging techniques have provided information regarding the morphology and the physiology of the brain of patients with MCI or AD, and also the analyses of biomarkers in the cerebrospinal fluid (CSF) and in the plasma have been useful to notice and study the disease pathophysiological process and progression in vivo (Blennow et al. 2010; Sperling et al. 2011). Although the accumulation of amyloid-beta (Aβ) protein is a biomarker and putative trigger of cognitive decline in AD, anatomo-pathological observations of elderly people with normal cognitive function revealed the presence of extensive Aβ deposition, forming amyloid plaques, in the brain (Serrano-Pozo et al. 2011). These findings were confirmed in more recent in vivo studies using brain imaging techniques and compounds to label Aβ deposits (Elman et al. 2014). Moreover, functional magnetic resonance imaging (fMRI) studies, which were performed during an episodic memory encoding task, showed an increased activation in medial temporal lobe (MTL) regions in both aged cognitively normal individuals with brain Aβ deposition and individuals with MCI as compared with age-mates people without Aβ deposits. This MTL activation was postulated: (i) to be a compensatory mechanism to manage the underlying Aβ deposition and to delay the onset of cognitive decline or (ii) to promote the Aβ deposition in aged people (Sperling et al. 2010; Mormino et al. 2012). These findings certainly contribute to the controversy of whether Aβ protein overproduction and accumulation are beneficial or harmful (Sperling et al. 2010; Elman et al. 2014). Currently, there still exists a great need of better defining biomarkers, imaging markers, and cognitive profiles that predict with greater confidence the progression from preclinical to clinical stages of MCI and AD dementia and to define if Aβ is a causative agent of these pathological conditions.

11.2 Clinical and Neuropathological Features of AD

There are two forms of Alzheimer’s: (i) the early-onset (under 65 years of age) familial AD (fAD ), which comprises 1–5% of total AD population and has an aggressive and rapid progression with relative short survival time; and (ii) the late-onset (after 65 years of age) sporadic AD (sAD) that represents the majority (>95%) of the cases (Tanzi 2013; Prince et al. 2016).

In both AD forms, the most noticeable clinical symptom is the impairment in formation and retention of new episodic memories. The syndrome of memory decline in AD has been attributed to neuronal loss in the perforant pathway of the MTL (Hyman et al. 1984); however, further functional imaging studies have suggested that memory processes are subserved by a set of distributed, large-scale neural networks (Raichle et al. 2001). In addition to the hippocampus and surrounding MTL cortices, these memory networks are comprised of a set of cortical regions, collectively named as the default network, which typically deactivate during memory encoding and other cognitive demanding tasks focused on external stimuli processing (Sperling et al. 2010). Frontoparietal cortical networks, which support executive function and attentional processes, also likely interact with these memory systems, and thus multiple cognitive domains become dysfunctional as AD progresses (Dickerson and Sperling 2009; Sperling et al. 2010 and references therein).

Characteristic imaging features of sAD and fAD include atrophy of the medial temporal lobe, the precuneus, the ventrolateral temporal, lateral parietal, and posterior cingulate cortices, the amygdala and the anterior hippocampus. Furthermore, hypometabolism and amyloid deposition in these regions can be detected using fluorodeoxyglucose and Pittsburgh compound B positron emission tomography (FDG-PET and PiB-PET), respectively (Tentolouris-Piperas et al. 2017 and references therein).

Neuropathologically, both forms of AD are characterized by the presence of extracellular amyloid plaques and of intracellular neurofibrillary tangles (NFT) , which are deposits of abnormal proteins that progressively grow and spread in the brain parenchyma (Blennow et al. 2006). The amyloid plaques are mainly composed by the amyloid-β (Aβ ) protein with 42 or 40 amino acids (Aβ40 and Aβ42), being the first form more abundant than Aβ40 within the plaques (Citron et al. 1996). The intracellular NFTs are primarily composed of paired helical filaments (PHF) consisting of hyperphosphorylated tau, a microtubule-binding protein of cytoskeleton (Bramblett et al. 1993), although more recent studies reported that acetylated tau was present throughout all stages of AD and seems to precede the tau hyperphosphorylation and, eventually, later truncation (Irwin et al. 2012). Postmortem studies were crucial to categorize the progression of both amyloid and tangles pathologies, which contributed to the development of AD diagnostic criteria that are currently used (Braak and Braak 1991; Gómez-Isla et al. 1997). The amyloid plaques occur initially and most severely in the precuneus and frontal lobes, whereas neuronal death begins and arises most readily in the entorhinal cortex and hippocampus, regions with relatively few Aβ deposits. In contrast, the NFT correlate more closely with neuronal loss, both spatially and temporally. That is, amyloid pathology appears to begin in the cortex and spreads inward, while tau pathology exhibits an opposite progression. Therefore, brain regions with neuritic plaques, which are amyloid plaques associated to NFT, display an extensive neuronal death (Musiek and Holtzman 2015).

Amyloid plaques are usually surrounded by dystrophic neurites and reactive glial cells (astrocytes and microglia), forming structures known as senile plaques. The activation of glial cells has been considered a double-edge event: it can be seen as an endogenous defensive mechanism against amyloid deposition and neuronal damage, while on the other hand, the persistent activation of glia might trigger a neuroinflammatory process, with increased production of pro-inflammatory cytokines and other neurotoxic factors, such as reactive oxygen species (Akiyama et al. 2000; Agostinho et al. 2010, and references therein). Neuroinflammation is largely driven by glial cells but also by mononuclear phagocyte (10% of cell population in CNS) and by dying neurons; however, the exact role of this process during AD onset and progression is still not well established (Agostinho et al. 2010; De Strooper and Karran 2016).

In the last decades, it was recognized that the reduction of synapse number is perhaps the strongest quantitative neuropathological correlate of dementia in AD (Terry et al. 1991; Selkoe 2002). Indeed, both APP and α- and β-secretases are more abundantly localized in cortical synapses (Pliássova et al. 2016a, b), and synaptic activity is tightly linked to APP and the formation of Aβ (Agostinho et al. 2015; Müller et al. 2017). Moreover, it was shown that soluble Aβ oligomers, but not amyloid plaques cores, collected directly from the cerebral cortex of subjects with AD, impair both synaptic function (e.g., long-term potentiation) and synaptic structure (e.g., dendritic spines) and, consequently, the memory of a learned behavior in healthy adult rats (Shankar et al. 2008; Selkoe and Hardy 2016). These data are consistent with observations made in transgenic AD mice (with human mutant APP), which showed that amyloid plaques have a penumbra of soluble Aβ oligomers in which the synaptic density is low (Koffie et al. 2009). There are also evidences in humans that Aβ oligomers predict the presence of dementia more accurately than amyloid plaques burden and that these Aβ species are more closely related with tau pathology (Lesné et al. 2013; Musiek and Holtzman 2015). Indeed, these findings have contributed to (i) shift the assumption that amyloid plaques are the likely causative agent of pathology into the concept that soluble Aβ oligomers are more potent in causing synaptic dysfunction and loss and (ii) explain why the amyloid plaques distribution does not match with neuronal damage nor with tau pathology. Hence the prevalent view is that amyloid plaques sequester Aβ oligomers and, thus, protect neurons from their synaptotoxicity. Noteworthy, increasing evidences in the last years also support that astrocytes, a major type of glia cells, reciprocally regulate synaptic connectivity and transmission, forming a “tripartite synapses” (Perea et al. 2009). Aβ peptides can trigger astrocyte activation or astrocytic dysfunction, leading to metabolic and synaptic alterations that might underlie several pathological states, such as AD (Matos et al. 2012; Rial et al. 2016; De Strooper and Karran 2016).

11.3 Putative Initiators and Risk Factors for AD

The “amyloid cascade hypothesis” was proposed in the early 1990s and posits that Aβ deposition initiates a sequence of events leading to progressive tau pathology, synaptic dysfunction, neuroinflammation, vascular damage, neuronal loss, and ultimately impairment of higher cortical activity, such as memory and cognition (Hardy and Higgins 1992). Although, this hypothesis has been the source of considerable controversy, mainly because the clinical trials targeting Aβ failed and there are evidences supporting that tau might by an initiator of neurodegeneration (Selkoe et al. 2012; Musiek and Holtzman 2015), it still is the main theoretical construct for AD. Moreover, this premise has contributed undoubtedly to the replacement of the earlier descriptive studies by more mechanistic and functional studies, and this has contributed to the development of diagnostic and therapeutic strategies for a disease believed before to be either incurable or an inevitable consequence of aging (Hardy and Selkoe 2002).

Human genetics studies strongly support the role of Aβ as disease initiator (see Sect. 11.2). The autosomal dominant familial AD (fAD) is due to mutations in three genes, which are all integrally involved in Aβ production, the (i) amyloid precursor protein (APP) gene, (ii) presenilin 1 (PSEN1) gene, and (iii) presenilin 2 (PSEN2) gene. The APP is a ubiquitous transmembrane protein that can be proteolytically cleaved via amyloidogenic pathways, involving β-and γ-secretases that give rise to Aβ40 or Aβ42; PSEN1 and PSEN2 are catalytic subunits of the γ-secretase complex that cleaves APP to generate Aβ (Hardy and Selkoe 2002; Agostinho et al. 2015). Notably, some APP mutations in the middle of the Aβ coding region are also associated with fAD by promoting not Aβ production but fibrillation and aggregation or by inhibiting Aβ degradation or clearance (Tomiyama et al. 2008). These forms of AD provide an opportunity to examine a pure Aβ-driven disease in relatively young, often healthy, individuals (Musiek and Holtzman 2015).

The majority of AD cases is considered sporadic (sAD) and can be instigated by aging along with a complex interaction of genetic, metabolic, and environmental risk factors still not understood. The genetics of sAD is also multifactorial, being the strongest genetic risk factor the ε4 allele of apolipoprotein E gene (APOE). This polymorphic gene has three common alleles, ε2, ε3, and ε4, resulting in single amino changes in the apolipoprotein E that is involved in the transport of lipids in the blood. The APOE ε4 allele increases the risk for AD, and the ε2 allele decreases the risk for AD relative to the most frequent ε3 allele. About 20–25% of the population carries at least one copy of ApoE4, which quadruplicates the risk of AD, whereas 2% of the population carries two E4 alleles, having an increased risk of around 12-fold, compared to individuals with the more common ApoE3/E3 genotype (Verghese et al. 2011). Numerous studies have shown accelerated amyloid plaque accumulation in human ApoE4 carriers, and it was also reported that ApoE influences Aβ metabolism, and, in particular, ApoE4 protein promotes amyloid aggregation and deposition (Liu et al. 2013). Noteworthy, there are other genes that might contribute to sAD, such as genes related with proteins involved in the metabolism of cholesterol and lipids, in the immune response, and in synaptic structure and function (Bertram and Tanzi 2004; Tanzi 2013). A recent study had also reported that epigenetic factors (differences in DNA methylation) are associated with sAD in monozygotic twins (Mastroeni et al. 2009).

Two other risk factors chiefly contributing for sAD are age (the probability of developing AD increases from 10% under the age of 65 to 50% over 85 years old) and gender (sex); the prevalence of AD has been greater in women. In fact, the incidence for AD is comparable in women and men at younger ages, but at a later age the incidence becomes greater in women probably because women have a greater longevity of 4.5 years on average (Solomon et al. 2013). Epidemiological studies have also identified several risk factors for cognitive decline and AD more associated with metabolic alterations or lifestyle habits, such as diabetes mellitus, obesity, atherosclerosis, cardiovascular disease, and hypercholesterolemia at midlife, as well as physical and mental inactivity, low educational attainment, poor diet, sleep deprivation, and smoking (Crous-Bou et al. 2017; Tariq and Barber 2017). Thus, sAD can be considered a multifactorial, genetically complex, and heterogeneous disorder formatted by several non-genetic factors (Tariq and Barber 2017).

11.4 Therapeutic and Preventive Strategies for AD

Currently, drugs approved by the FDA (Food and Drug Administration) for the treatment of cognitive decline and AD, such acetylcholinesterase inhibitors (that avert the catabolism of the neurotransmitter acetylcholine to improve cognition) and NMDA receptor antagonists (that reduce glutamate-induced neuronal excitotoxicity), have limited effects and none of them prevent or reverse the disease pathology (Huang and Mucke 2012). Thus, the establishment of strategies to delay the onset of AD or slow its progression would have a significant impact on public health (Sperling et al. 2011; Prince et al. 2016). In the last years, several intervention studies have moved their focus toward cognitively healthy people at risk of developing AD, which are likely to have not yet substantial irreversible neuronal network dysfunction and loss, as the best strategy to reduce the incidence and prevalence of AD. Moreover, the estimate that a third of AD cases are potentially attributable to modifiable risk factors related with metabolic problems or lifestyle habits has prompted the development of several multi-domain intervention programs to prevent cognitive decline among elderly people. The data from large, long-term, randomized controlled trial have demonstrated that a multi-domain intervention, including equilibrated and healthy diet, exercise, cognitive training and monitoring vascular risk can ameliorate or preserve cognitive functioning in aged individuals (60–77 years old) compared to the general population at risk of dementia (Ngandu et al. 2015; Crous-Bou et al. 2017). Recent prospective studies reported that the adherence to a Mediterranean-style diet, which includes proportionally larger consumption of olive oil, legumes, unrefined cereals, fruits and vegetables, moderate consumption of fish, dairy products and wine, and low consumption meat products, might be related with slower cognitive decline and a reduced risk of progression from MCI to AD (Solfrizzi et al. 2011; Panza et al. 2015). Another lifestyle measure shown to attenuate the incidence of cognitive deterioration with aging is a diet rich in nuts (Rajaram et al. 2017), with a regular consumption of ω-3 unsaturated fatty acids (Burckhardt et al. 2016), which has also been confirmed to afford benefits in controlled studies using animal models of AD (Hooijmans et al. 2012; Muthaiyah et al. 2014).

Notably, several epidemiological studies in humans have demonstrated beneficial effects on cognition upon caffeine consumption, which has been associated with a decreased risk of developing AD and other neurodegenerative disorders, as well as age-related cognitive decline (Maia and de Mendonça 2002; van Gelder et al. 2007; Simonin et al. 2013; Flaten et al. 2014; Panza et al. 2015). Caffeine is a methylxanthine with psychostimulant properties, which exists in coffee, tea and chocolate, which are regularly consumed by millions of people around the world (Fredholm et al. 1999). The psychostimulant properties of caffeine are due to its capacity to interact with adenosine receptors in different brain regions (Fredholm et al. 1999), thereby enhancing vigilance and attention, stabilizing mood and improving cognition, mainly memory performance when it is perturbed by pathological conditions, either in human or animal studies (Nehlig et al. 1992; Yu et al. 2017). Apart from these acute effects subjective of role of caffeine as a cognitive enhancer, the regular consumption of caffeine is instead considered as a cognitive normalizer (Cunha and Agostinho 2010). Indeed, some case-control and cross-sectional and longitudinal population-based studies evaluated the long-term effects of caffeine on brain function and provided evidence that its consumption or higher plasma caffeine levels might be protective against cognitive decline and dementia (Santos et al. 2010; Panza et al. 2015). Interestingly, although epidemiological studies support that caffeine consumption can slow down cognitive decline in the elderly and reduces the risk to develop AD or Parkinson’s disease, it has been reported that this methylxanthine is harmful for Huntington’s disease, suggesting that caffeine is not beneficial for all neurodegenerative conditions and its effects depend on pathogenic mechanisms (Simonin et al. 2013; Flaten et al. 2014).

11.5 Caffeine and Adenosine Receptors: Impact in AD

Caffeine , at a concentration of 1–30 μM in the body (equivalent to ingestion of 1–5 cups of coffee), exerts its primary effect in the central nervous system through the inhibition of adenosine receptors, mainly A1 and A2A receptors, and subsequent modulation of neurotransmitter release (Fredholm et al. 1999; Kerkhofs et al. 2018). However, higher concentration of caffeine (millimolar, not corresponding to normal coffee consumption) can affect the release of calcium from intracellular stores, interfere with GABAA receptors, and inhibit 5’-nucleotidases and alkaline phosphatase (Fredholm et al. 1999, 2005; Cunha and Agostinho 2010). The rapid metabolization (within 1 h) of caffeine in the liver (Nehlig 2018) introduces an additional difficulty in the understanding of caffeine molecular mechanisms in the brain. The main metabolites of caffeine are paraxanthine, theophylline, and theobromine, which are also adenosine receptor antagonists; however, these metabolites can cause distinct, although sometimes also, overlapping responses (Yu et al. 2017). A case-control study provided initial evidence that the plasma levels of caffeine upon coffee intake were associated with a reduced risk of dementia, particularly for those who already have MCI (Cao et al. 2012). However, a subsequent study reported that it was the CSF levels of theobromine rather than levels of caffeine that correlated with a more favorable profile of AD biomarkers in the CSF, mainly Aβ1–42 and tau protein (total and phosphorylated), in patients with MCI or AD (Travassos et al. 2015), in agreement with the protection afforded by the consumption of chocolate (rich in theobromine) in AD (Moreira et al. 2016).

The receptors for adenosine (an endogenous neuromodulator) regulate both synaptic transmission and plasticity either by directly modulating synaptic responses or by interfering with other receptors (Cunha 2016). Adenosine receptors are coupled to G proteins and can be antagonized by methylxanthines, such as theophylline and caffeine, which have greater affinities for human than rodent brain adenosine receptors (Kerkhofs et al. 2018). These receptors are classified into A1 (A1R), A2A (A2AR), A2B (A2BR) and A3 (A3R) receptors. The A1R and A3R inhibit adenylyl cyclase through Gi/o (inhibitory) proteins, while A2AR and A2BR stimulate adenylyl cyclase through Gs/olf (stimulatory) proteins. A1R inhibits calcium channels and stimulates potassium channels, decreasing stimulus-evoked release of neurotransmitters, in particular of glutamate, as well as postsynaptic responsiveness to glutamate; in contrast, A2AR facilitates the evoked release of glutamate and promotes synaptic plasticity phenomena (Fredholm et al. 2005; Burnstock et al. 2011). A1R and A2AR are widely located throughout the brain, being present in both neuronal and glial cells, which is in agreement with the predominant effects of caffeine on brain-related functions (Fredholm et al. 2005; Cunha and Agostinho 2010; Cunha 2016; and references therein). In pathological conditions, A1R mostly acts as a hurdle that needs to be overtaken to begin neurodegeneration; thus, A1R only controls neurodegeneration if activated in the temporal vicinity of brain insults (de Mendonça et al. 2000). In contrast, the blockade of A2AR alleviates the long-term burden of brain injuries in different neurodegenerative conditions, such as AD and Parkinson’s disease (Gomes et al. 2011).

The setup of adenosine receptors in the brain is altered in AD and other dementia. This was first observed in the early 1990s, in studies comparing brain samples from AD patients and age-matched non-dementia individuals (control) that reported a reduction in A1R levels in the dentate gyrus and CA3 regions of the hippocampus of AD patients, which are regions of NFT spread and of neuronal loss (Kalaria et al. 1990; Ułas et al. 1993). Deckert et al. (1998) proposed that the reduction of A1R in CA1 region of hippocampus might not be specific for AD cases, since it also occurred in other types of dementia. Positron emission tomography (PET) in vivo studies using a radioligand for A1R (11C-MPDX, 8-dicyclopropylmethyl-1-[11C]methyl-3-propylxanthine) revealed a decreased binding of [11C]MPDX in the medial temporal cortex of AD patients when compared with normal elderly individuals, which is consistent with postmortem autoradiographic studies showing a reduction of A1R levels in patients with AD (Fukumitsu et al. 2008). In contrast, Angulo et al. (2003) showed that the protein levels of A1R are slightly augmented in the hippocampus of AD patients, mainly in degenerating neurons with NFT and in dystrophic neurites of amyloid plaques, although no significant changes were observed in A1R mRNA expression. This study also described that A2AR, which are located mainly in the striatal neurons in control individuals, appeared in microglia cells in the hippocampus and cerebral cortex of AD patients (Angulo et al. 2003). In the frontal cortex of AD patients, it was also described as an upregulation of A1R and A2AR , as compared with age-matched non-dementia (control) individuals (Albasanz et al. 2008). The levels of A1R and A2AR were considerably increased in AD patients in early and advanced stages, without differences with disease progression, being the upregulation of these receptors associated with sensitization of the corresponding transduction pathways (Albasanz et al. 2008). Interestingly, it also reported increased levels of A2AR in astrocytes, but not in microglia cells, in the hippocampus of AD patients, and it was postulated that increases in astrocytic A2AR levels contribute to memory loss in AD conditions (Orr et al. 2015). Interestingly, a study performed in peripheral blood mononuclear cells (PBMCs) of MCI and AD patients and age-matched healthy people reported increased A2AR levels in MCI patients, which might indicate the involvement of A2AR in early AD stages (Gussago et al. 2014). Since this increase in A2AR levels was not observed in peripheral cells of individuals with vascular dementia, it was postulated that A2AR could be a biomarker to distinguish these two types of dementia (Gussago et al. 2014).

Since most of the available studies in human tissue samples are from deceased patients with advanced stages of AD, it still remains to be investigated how the adenosine receptors change in terms of density, localization, and function and the onset and progression of cognitive decline and AD; however, this is difficult to accomplish in human patients. Thus, the use of experimental models to mimic the disease is useful to better understand the initiating mechanisms and to test and validate novel therapeutic strategies for dementia-associated diseases.

11.6 Evidences for the Involvement of Adenosine Receptors in Experimental Model of AD

11.6.1 Caffeine Studies

Numerous epidemiological studies showed that caffeine can decrease cognitive deficits and AD (Chen 2014). Caffeine and its metabolites, at low to moderate doses, act mainly on adenosine receptors (Fredholm et al. 1999); however, the neuroprotective mechanisms of these methylxanthines are not completely defined.

The first study showing a protection by caffeine in AD-like conditions was performed in cultured neurons exposed to the toxic synthetic fragment – Aβ25–35 – to mimic AD-like conditions. This study showed that caffeine prevented neuronal cell death, and this neuroprotective effect was mimicked by a selective antagonist of A2AR (ZM 241385), but not when an A1R antagonist was used, thus indicating that the neuroprotection afforded by caffeine against Aβ -toxicity was mainly mediated by the blockade of A2AR (Dall’Igna et al. 2003). This pioneering in vitro study prompted follow-up studies in animal models of AD to test the effect of caffeine on memory impairment and neurochemical alterations that occur in AD. Arendash and collaborators (2006) treated a transgenic AD mice model (TgAD-APPsw, with a Swedish mutation in APP gene) with caffeine orally for 5 months (1.5 mg consumption daily per mouse, which is equivalent to human intake of five cups a day) and reported an amelioration of reference, working, and recognition memory. This same study indicated that this caffeine protection on memory performance involved a reduction of the hippocampal levels of β- and γ-secretases and consequently in Aβ production (Arendash et al. 2006). The same group further showed that caffeine (oral administration) prevents memory impairment and plaque deposition in aged TgAD-APPsw mice (18 months old) relatively to age-matched Tg mice that were not consuming caffeine (Arendash et al. 2009). This ability of caffeine to prevent Aβ-induced memory deterioration was also observed in an animal model of sAD, consisting in the intracerebroventricular Aβ25–35 injection (icv-Aβ25–35 that cause memory deficits 1-week later). In fact, the acute or sub-chronical intraperitoneal (ip) caffeine administration prevented memory deficits induced by icv-Aβ25–35 and effect mimicked by the A2AR antagonist (SCH58261), whereas selective A1R antagonists were devoid of effects (Dall’Igna et al. 2007). These studies strongly support a neuroprotective role of caffeine, mainly mediated by A2AR the antagonism, in neurodegeneration triggered by Aβ overload. In a similar way, it was reported that caffeine or the selective blockade of A2AR by SCH58261, both administered by oral gavage, was able to ameliorate age-related memory impairment (Prediger et al. 2005; Leite et al. 2011).

Several studies attempted to further grasp the mechanism underlying the ability of caffeine to attenuate different features pertinent to AD , besides the control of the amyloidogenic processing of APP. In fact, caffeine intake through drinking water at an early pathologic stage in a THY-Tau22 transgenic mouse modelling progressive AD-like tau pathology prevented the development of spatial memory deficits, reduced hippocampal tau phosphorylation and proteolytic fragments, and dampened several upregulated proinflammatory and oxidative stress markers found in the hippocampus (Laurent et al. 2014). Accordingly, Cao and collaborators (2009) reported that high levels of caffeine and their metabolites correlate with the decrease of cytokines in the hippocampus of TgAD-APPsw mice. In this AD mice model, it also described alterations in the intracellular signalling pathways associated with adenosine receptors, such as in protein kinase A (PKA), phospho-AMP responsive element-binding protein (CREB), phospho-c-Jun N-terminal kinase (JNK) and in phospho-extracellular signal-regulated kinase (ERK), which are crucial for synaptic plasticity and oxidative stress (Zeitlin et al. 2011). Furthermore, treatment of TgAD-APPsw mice with caffeine prevented the decrease of mitochondrial membrane potential and respiratory rate and, consequently, the reduction of ATP levels and the reactive oxygen species (ROS) overproduction in the hippocampal and cortical mitochondria (Dragicevic et al. 2012). Also, in rabbits fed with 2% cholesterol-enriched diet, which was considered a model of AD-like conditions, it was reported that the oral administration of caffeine prevented the downregulation of A1R and the increase in Aβ levels and tau phosphorylation, as well as the endoplasmic reticulum (ER) stress (Prasanthi et al. 2010). Interestingly, it was reported that caffeine was able to decrease low-density lipoprotein (LDL) cholesterol internalization, which impairs the APP internalization into lysosomes and consequently reduces the production of Aβ, in neuronal cultures (Li et al. 2015b). A final group of studies has exploited a model of sporadic AD, based on an intracerebroventricular intoxication with streptozotocin, to show an ability of caffeine to prevent memory deterioration as well as the associated loss of synaptic markers (Espinosa et al. 2013). In keeping with the evidence that the loss of synaptic markers is the current best morphological correlate of AD-related memory impairment (Terry et al. 1991; Selkoe, 2002; Scheff et al. 2007), caffeine and A2AR blockade prevented hippocampal synaptotoxicity and memory deficits in a variety of animal models (Cognato et al. 2010; Duarte et al. 2012; Machado et al. 2017).

All these studies performed in experimental models confirmed the neuroprotective role of caffeine in AD pathology and strength the contention that caffeine reduced the incidence of AD in humans.

11.6.2 Inhibitory Adenosine Receptor: A1R and A3R

In aged animals , there is a decrease of A1R levels and a decreased synaptic transmission in the hippocampus that are mediated by A1R (Pagonopoulou and Angelatou 1992; Cunha et al. 1995; Sperlágh et al. 1997; Lopes et al. 1999a; Cheng et al. 2000; Sebastião et al. 2000; Meerlo et al. 2004; Canas et al. 2009a). Confirming previous results, in a mice model of accelerated senescence with short life span and memory deficits, there is also a decrease in the levels of A1R, an abnormal accumulation of Aβ, and later on, formation of amyloid plaques when compared with a resistant senescence mice strain (Castillo et al. 2009).

The activation of A1R (with R-PIA) increased the levels of soluble APP in a dose-dependent manner, through a protein kinase C (PKC) signalling pathway, in a neuronal cell line (SH-SY5Y, Angulo et al. 2003). Moreover, in this cell line the activation of A1R, via ERK-signalling pathway, leads to tau translocation, from cytosol to the cytoskeleton, increasing tau phosphorylation, which supports a possible role of A1R in key AD events (Angulo et al. 2003). On the contrary, in mixed (glia and neurons) cultures of rat cerebellum, the antagonism of A1R (with DPCPX) did not prevent Aβ25–35-induced neurotoxicity (Mitchell et al. 2009). This finding was later confirmed by the chronic (60 days) administration of DPCPX in a TgAPP-PS1 model, in which no protective effect on memory impairment was granted by this A1R antagonist; notoriously it was shown that DPCPX deteriorated memory performance in nontransgenic mice (Vollert et al. 2013). However, in an ex vivo model of cultured organotypic hippocampal slice of mice expressing pro-aggregant tau, the blockade of A1R (by rolofylline) was able to restore synaptic function and morphology (Dennissen et al. 2016). This study also demonstrated that the chronic treatment by rolofylline of pro-aggregant tau transgenic mice prevented memory impairment evaluated by different behavioral tests (y-maze, novel object recognition, and fear conditioning), suggesting that the A1R antagonism could be a potential therapeutic strategy for memory deficits and neurochemical alterations triggered by tau pathology (Dennissen et al. 2016). However, all the evaluation of the role of A1R in AD should take into account the known opposite impact of acute and chronic manipulation of A1R on brain function (see Jacobson et al. 1996) and the fact that there is a tight interaction between A1R and A2AR (Lopes et al. 1999a; Ciruela et al. 2006).

Although, it was described that activation of A3R is beneficial in ischemic brain injury (Chen et al. 2006) and retinal degeneration (Galvão et al. 2015); the role of this type of receptors in AD was not yet described. There is only a suggestion that A3R might control APP internalization into lysosomes and consequently reduce the production of Aβ in neuronal cultures (Li et al. 2015b).

11.6.3 Facilitatory Adenosine Receptor: A2AR and A2BR

Several studies reported an increase in A2AR levels and binding sites in the hippocampus and cortex of aged animals, evaluated by different techniques (Cunha et al. 1995; Lopes et al. 1999a, b; Cheng et al. 2000; Canas et al. 2009a). Moreover, cortical A2AR also undergoes a gain of function in aged animals, confirmed through different measures such as facilitation of synaptic transmission, increase of acetylcholine release, and increase in long-term potentiation in hippocampus of aged animals when compared with young adults (Lopes et al. 1999b; Rebola et al. 2003; Rodrigues et al. 2008; Costenla et al. 2011).

As previously mentioned, several studies reported that A2AR blockade mimicked the neuroprotective role of caffeine in different AD experimental models. In a sAD model (Aβ1–42 icv administration), the blockade (by SCH58261) or genetic inactivation of A2AR was able to prevent memory impairment (evaluated by Y-maze and novel object recognition tests) and synaptotoxicity (decrease of SNAP-25 and synaptophysin synaptic levels) (Canas et al. 2009b). This study also demonstrated that the blockade of A2AR was able to prevent synaptotoxicity and subsequent neuronal loss, through a p38 mitogen-activated protein kinase (MAPK) signalling rather than by controlling cAMP/protein kinase A in hippocampal cultured neurons (Canas et al. 2009b). Moreover, this neuroprotective effect of A2AR blockade may be explained by predominant localization of A2AR at the synapse (Rebola et al. 2005), since in a synaptosomal preparation, the blockade of A2AR was able to prevent mitochondrial dysfunction triggered by Aβ1–42 (Canas et al. 2009b). A2AR blockade also prevented or reverted memory deficits associated with other chronic brain diseases where synapse deterioration is present, such as diabetic neuropathy (Duarte et al. 2012), convulsions in early life (Cognato et al. 2010) or chronic stress (Kaster et al. 2015). In contrast, this neuroprotective effect afforded by A2AR blockade does not occur in other pharmacological models which comprise an acute deterioration of memory and that do not involve synaptotoxicity, such as in rats administrated with scopolamine (antagonist of muscarinic receptors) or MK-801 (antagonist of glutamate receptor) (Cunha et al. 2008).

The A2AR seems also to be involved in the production of Aβ, since the activation of these receptors by HENECA enhanced the activity of γ-secretase and, subsequent, Aβ1–42 formation in a human neuroblastoma cell line (Nagpure and Bian 2014). In addition, Lu et al. (2016) demonstrated that A2AR controls Aβ levels by a mechanism that involves a physical interaction of A2AR with γ-secretase, in particular with the catalytic subunit PS1. The A2AR activation, through Gs protein and subsequent cAMP/PKA signal pathway, could cause Aβ overproduction by increasing APP levels and γ-secretase activity; in contrast, the blockade of A2AR decreased the interaction of A2AR with PS1, which promotes in Aβ production (Lu et al. 2016). There are also evidences that A2AR could have a role in tau pathology, similar to that described in models based on Aβ-induced pathology. Laurent et al. (2016) developed a tau pathology animal model with a genetic deletion of A2AR, by crossing A2AR global KO mice with THY-Tau22 mice, where they showed that the genetic silencing of A2AR prevented spatial memory impairment, the decrease of hippocampal long-term depression (LTD), the imbalance of glutamate and GABA, and also attenuated neuroinflammation and tau hyperphosphorylation. This study also showed that an antagonist of A2AR (MSX-3, oral administration) was able to improve memory and reduced tau phosphorylation in THY-Tau22 mice (Laurent et al. 2016). Another important link between A2AR and memory deterioration was provided by the conclusion that A2AR controls the expression and function of glucocorticoid signalling in the hippocampus (Batalha et al. 2016), which has a strong impact on memory performance (Lupien et al. 1999) and is disrupted in AD (Popp et al. 2015).

So far, the question arises if the prevention afforded by A2AR in AD animal models is a general effect or is cell or brain region-specific. Viana da Silva et al. (2016) demonstrated that there is an increase of A2AR in synaptic membranes of CA3 hippocampal region of TgAPP-PS1 mice model (6 months old). Moreover, the blockade of A2AR (SCH58261 or ZM241385) or the neuronal genetic silencing of these receptors in these transgenic AD mice prevented the suppression of long-term potentiation in hippocampal CA3 pyramidal cells and ameliorated memory impairment (Viana da Silva et al. 2016). This study and other studies of our group strongly suggest that in early phases of AD, the synaptic loss and dysfunction could be prevented by A2AR blockade/silencing in neurons (Canas et al. 2009b; Viana da Silva et al. 2016). Furthermore, Li et al. (2015a) also validated the important role of neuronal A2AR in memory, since the optogenetic activation of neuronal A2AR intracellular signalling in hippocampus is sufficient to impair memory, through the control of CREB phosphorylation and long-term potentiation (LTP), which are crucial events involved in memory process (Li et al. 2015a). This role of A2AR in memory was also confirmed by the activation of A2AR, achieved by icv administration of CGS21680, which led to memory impairment (Pagnussat et al. 2015).

Remarkably, astrocytic A2AR may also have a role in AD pathology. Matos et al. (2012) observed that the activation of A2AR triggers astrogliosis in cultured astrocytes similar to Aβ1–42, an effect prevented by the blockade of A2AR. This study showed that Aβ1–42 increased the A2AR in astrocytes and caused astrocytic dysfunction, decreasing the glutamate uptake capacity and the levels of glutamate transporters, GLAST, and GLT-1, and these alterations were prevented by the genetic silencing or pharmacological blockade of A2AR (Matos et al. 2012). Moreover, TgAPP-PS1 mice also display increased levels of astrocytic A2AR, and the conditional genetic silencing of astrocytic A2AR increases memory performance in this transgenic model (Orr et al. 2015). In another study performed by the same group, the authors also showed an increase in astrocytic A2AR in a transgenic AD mice models (with several familial mutations in human APP) that exhibit amyloid plaques; however, this upregulation of A2AR was not observed in animals that overexpress human APP (wild-type) but never form amyloid plaques (Orr et al. 2017). In these animal models the antagonist of A2AR, KW-6002, at low doses (4–10 mg/kg per day) was able to prevent memory deficits (Orr et al. 2017). These studies point to a role of astrocytic A2AR in a late phase of AD, when there is already amyloid plaques.

So far, there is scarce information about the role of A2AR in microglia. In transgenic mice of AD, 5 × FAD mice, with increased number of microglia, it was observed that blockade of A2AR with preladenant decreased the hypermobility of microglia associated with amyloid plaques in hippocampal slices but did not reestablish microglia motility toward tissue damage (Gyoneva et al. 2016). In addition, in an animal model of neuroinflammation, triggered by lipopolysaccharide (LPS) administration, the blockade of A2AR with SCH58261 prevented the decrease of hippocampal LTP, as well the recruitment of activated microglia cells and the overproduction in interleukin-1β (Rebola et al. 2011). Altogether, these findings suggest an involvement of microglial A2AR in controlling memory impairment and pathology of neurodegenerative diseases. Nevertheless, more studies are necessary to understand the role of microglia A2AR in AD.

When designing a therapeutic strategy, it should be taken into account that adenosine receptors (A1R and A2AR) can control the permeability of the blood-brain barrier (BBB), which defines the accessibility of molecules to the brain (Carman et al. 2011). The activation of A1R or A2AR increased BBB permeability, and this information was later confirmed by the opposite effect displayed by A1R and A2AR KO mice (Carman et al. 2011). Moreover, administration of NECA (an agonist of adenosine receptors) in Tg-AD animal model (double mutation in APP/PSEN1) allowed the entry of anti-Aβ antibodies and posterior label of β-amyloid plaques (Carman et al. 2011).

In the literature there is limited information about the role of A2BR in AD experimental models, probably due to the lack of good selective drugs for A2BR. However, it is worth noting that A2BR controls glucose uptake and availability in astrocytes (Magistretti et al. 1986; Allaman et al. 2003; Lemos et al. 2015), an observation of particular relevance in view of the characteristic cortical hypometabolism used to diagnose AD (Chen and Zhong 2013).

11.7 Conclusions

AD pathology begins several years before clinical symptoms. The complex and multifactorial nature of this pathology implies the use of multimodal interventions to manipulate possible initiators and risk factors of disease. Lifestyle modifications, mainly the mental and physical exercise and the practice of a Mediterranean diet, have been shown to be beneficial for AD, in particular the consumption of coffee, tea, and chocolate that contain caffeine, which affords a general benefit to bolster the quality of life on aging (Freedman et al. 2012). Numerous studies showed that caffeine, the most widely consumed psychoactive drug worldwide, and its metabolites, by acting through adenosine receptors, are a cognitive normalizer and also restrain Aβ overproduction. In fact, alterations of the setup of adenosine receptors, mainly of A2AR, can be viewed as a potential biomarker and a target of therapeutic interventions for cognitive decline and AD (Fig. 11.1). Nevertheless, the role of adenosine receptors in early phases of AD and in brain regions affected in AD remains to be further clarified. Although the research in experimental models regarding the impact of adenosine receptors on AD have been an added value, it is mandatory to hustle the translation of basic scientific findings into potential treatments for this disease.

Fig. 11.1
figure 1

Adenosine receptors might be a key target to prevent, slow and treat AD. The multifactorial nature of AD implies the use of multiple interventions to manipulate possible initiators and risk factors of disease. The manipulation (genetic or pharmacological) of adenosine receptors (AdoR), in particular A2AR that has been reported to be upregulated in AD, can be can be viewed as a potential preventive and therapeutic strategy for cognitive decline and AD

Full size image