Glucose Hypometabolism in AD

ApoE4 homozygote individuals exhibit reduced cerebral glucose metabolism in areas classicaly associated with AD [50, 104]. Although the extent of glucose hypometabolism was significantly less in ApoE4 individuals than AD patients, it is noteworthy because these individuals were cognitively normal at the time and were a decade younger than typical age when AD-like symptoms appear. T2D patients also exhibit abberant [18F]-fluoro-2-deoxy-2-D-glucose positron emission tomography (FDG-PET) uptake patterns at prediabetic and cognitively normal stages [4, 6, 11]. Alterations of glucose uptake at presymptomatice stages in both AD and T2D subjects suggest that metabolic disruption in the brain occurs early in the development of AD and T2D, and likely to contribute to disease progression or pathology. If metabolic disruption plays a disease-modifying role, it can potentially render room for altering diet as a way to influence the disease progression. In Drosophila, increasing energy flux by regulating mitochondrial activity in mushroom bodies that are involved in learning and memory led to improvement in long-term memory [96], suggesting energy metabolism can positively influence memory. Similarly, although the number of patients enrolled was small, individuals diagnosed with MCI and early stages of AD showed improvement when given special diet that included an increase in ketogenesis [14]. Albeit the fact that the science behind the influcence on brain function of a healthy lifestyle that includes regular exercise and healthier diet is yet to come, these data may be interpreted as suggesting that peripheral energy metabolism can potentially influence higher brain function.

FDG-PET Imaging to Detect Glucose Hypometabolism in AD

For several decades, FDG-PET imaging has been employed for examining the brains of various dementia and AD patients to study glucose metabolism in the brain. FDG is a short half-life radioactive analog of glucose that is transported into cells but not metabolized beyond the first step of the glycolysis generating FDG-6-phosphate inside the cell. The accumulation of radioactive signals thus reveals the areas in the tissue of interests that are active in glucose metabolism. AD patients present widespread FDG hypometabolism in brain region including parieto-temporal, posterior cingulate, and frontal cortices [22, 85, 86, 88] (Fig. 1A). By contrast, other regions, such as the visual cortex, thalamus, and cerebellum, seem to be preserved [83, 84, 111, 114, 115]. Indeed, FDG-PET hypometabolic signature is highly conserved in typical AD cases [114] as well as in early-onset AD patients [37]. In atypical presentations of AD, this “metabolic signature” is diluted in varied hypometabolic topographical patterns [100].

Fig. 1
figure 1

FDG metabolism in Alzheimer’s disease continuum and transgenic rats. (A) Axial FDG images of cognitive unimpaired (CU), mild cognitive impaired (MCI), and Alzheimer’s disease (AD) individuals. Standardized uptake value ratio (SUVR) was calculated using the pons as the reference area. (B) Sagittal and coronal FDG images in 11- and 19-month-old McGill-R-Thy1-APP transgenic rats. SUVR was calculated using the pons as the reference area

Full size image

Since an abnormal reduction in FDG uptake in the brain is detectable decades before symptoms appear [50], efforts have been made to develop it as a predictable “preclinical phase” biomarker [64]. Indeed, APOE4 homozygote individuals exhibited glucose hypometabolism in the same areas that showed a FDG reduction long before they showed symptoms, albeit to a less extent [104]. APOE is a lipid-binding protein that transports cholesterol across different tissues and cells [65], and 1 of its alleles, APOE4, has been identified as a major genetic risk factor for late-onset AD [75]. Having 1 or 2 copies of APOE4 increases the risk of developing AD 3- to 12-fold [69]. Although the use of “preclinical phase” to define this period is gaining acceptance, routine use of FDG-PET appears limited in diagnosing asymptomatic individuals as being at risk to develop AD. For instance, representative areas of FDG hypometabolism remains to be determined in amyloid-positive asymptomatic subjects, autosomal-dominant Alzheimer’s disease individuals, and in subjects presenting subjective cognitive decline [25]. In addition, cortical FDG hypermetabolism in the superior temporal gyrus was reported in cognitively unimpaired (CU) subjects with significant amyloid deposition in superior temporal gyrus [38]. The same pattern of hypermetabolism was also identified in APOE4 carriers in an amyloid status-independent manner [140]. This could be a compensatory mechanism associated with early toxic species of Aβ, such as Aβ oligomers that are produced from amyloid precursor protein (APP) following 2-step proteolytic processing at its extracellular and interamembrane domains [112].

Abnormalities in glucose metabolism are, however, not restricted to AD [49]. As such, the National Institutes on Aging (NIA) and the Alzheimer’s Association (AA) incorporated FDG-PET as a biomarker of “neurodegeneration” in the recent research framework, the NIA-AA 2018. FDG-PET is in fact being used for differential diagnosis, distinguishing AD from frontotemporal dementias (FTDs) with high sensitivity and specificity [101]. In addition, a systematic review indicates that FDG-PET can discriminate AD from CU individuals with a pooled sensitivity of 90% and specificity of 89%. The inclusion of mild cognitive impairment (MCI) subjects, a more heterogeneous group, surprisingly produced only slight variation, modifying sensitivity and specificity values to 92% and 78%, respectively [10]. These findings indicate that in a large group of subjects, FDG-PET is indeed a good biomarker for discriminating CU, MCI, and AD dementia individuals. In smaller groups or at the individual level, the validity of FDG to discriminate MCI patients remains to be determined.

Many in the field are applying FDG-PET to AD mouse models. The miniaturized version of PET, which is termed micro-PET, allows high-resolution noninvasive imaging in small animals, such as rats and mice [34, 63]. Transgenic animal models with pathological mutations in human APP exhibit progressive deposition of Aβ in parallel with cognitive abnormalities and are highly suited for longitudinal assessment with micro-PET [35]. Currently, several studies have used FDG imaging to investigate glucose metabolism in these animal models (for review, see [146]). The FDG patterns in APP mutant mice, however, do not appear to follow exactly what is observed in human cases. For instance, Tg2576 mice that express human APPswe under the Prp promoter exhibit early hypermetabolism at 7 months, but normal FDG patterns at 9–15 months [58, 74, 78]. APPswe/PS1∆9 mice under the Prp promoter also show early hypometabolism at 6 months coinciding with the beginning of amyloid plaque deposition, but normal metabolism at later stages at 13–15 months [103, 143]. These differences may be due to technical challenges that are associated with micro-PET and the small size of the mouse brain [146], or the nature of APP overexpression associated with transgenic mice. If it is due to technical challenges, transgenic rats may be better animal models for micro-PET imaging, due to a larger brain size than mice, such as McGill-R-Thy1-APP that expresses human APP751swe/ind [92] (Fig. 1B).

FDG-PET studies in T2D patients are not numerous, althougth it is well recognized that type 2 diabetes (T2D) patients exhibit regional brain atrophy and cognitive impairment even if they are not necessarily diagnosed as having AD [1, 67, 80]. The reason for the cognitive deficits in T2D patients is unclear, but it is believed to be due to insulin resistance in the brain [70]. Whether glucose metabolism is also reduced in T2D patients is unclear since some studies reported a reduction in FDG uptake [4], while some reported an increase [11]. Also in rodent models, the results are inconsitent: brain glucose metabolism in db/db mice was reduced based on [3H]-2-deoxy-D-glucose or [14C]-2-deoxy-D-glucose uptake [31, 133], while FDG-PET imaging of T2D rat brains showed an increase [6].

Since both AD and T2D exhibit alterations in glucose metabolism, a hypothesis was proposed that there might be a defect in glucose-sensing machinery of the brain in both AD and T2D [102]. Glucose-sensing neurons and astrocytes in the brain are found in many regions, but they are mostly enriched in the hypothalamus and brain stem [30, 122]. Regardless of the regions involved, glucose-sensing neurons are divided into glucose-excited (GE) ones that increase and glucose-inhibited (GI) ones that decrease their firing rate under hyperglycemic or hypoglycemic conditions, respectively. The mechanisms by which GE neurons sense glucose are diverse, which depends on the type of glucose transporters expressed as well as the presence or absence of ATP-sensitive K channels and glucokinase, the major glucose sensor [122]. For GI neurons, on the other hand, activation of AMP-activated protein kinase (AMPK) appears to play a critical role [18]. Cells of another type in the hypothalamus that are involved in glucose sensing are tanycytes in the medial eminence. Tanycytes are a unique type of ependymal radial glia that participate in the fenestrated capillary barrier, and whose processes traverse various hypothalamic nuclei and surround ventrally located capillaries [26]. Like pancreatic beta cells, tanycytes express Glut1 and 2 and ATP-sensitive K channels, and respond to glucose puffs [8]. They also express glucokinase that couples the extracellular glucose levels to insulin secretion [79, 127]. But, whether there are significant changes in glucose transporter expression or their function among glucose-sensing cells of the brain in AD and T2D mouse models are currently unknown.

What is the significance of the reduction in glucose uptake in the brain? Since glucose is taken up by neurons and astrocytes [24, 30, 73, 147], a general hypometabolism of glucose in the brain is believed to represent reduced activities at synapses in these cells. Indeed, when brain glucose levels were increased acutely for 4 h by approximately 2-fold via a microdialysis techique coupled to glucose clamps, interstial fluid (ISF) Aβ levels increased by 25 to 39% during the hyperglycemic challenge at presymptomatic and symptomatic ages, respectively, in Appswe/PS1∆E9 mice [76]. Under the same setting, ISF lactate levels increased during hyperglycemic challenge. Since lactate released by astrocytes is shuttled to neurons, the increase in ISF lactate suggests that higher glucose levels in the brain increased neuronal activity [136], although recent studies demonstrated that neurons themselves take up glucose as well [24, 73].

Connection Between AD and T2D

Insulin Resistance

In addition to the fact that T2D increases the risk of developing AD later in life at least 2-fold [116], epidemiological studies also indicate a strong connection between AD and T2D both of which exhibit vascular lesions, hyperglycemia, hyperinsulinemia, atherosclerosis, and hypertension [116]. The underlying mechanism linking the 2 chronic diseases, however, remains unknown.

Insulin resistance is 1 of the 2 major cellular and physiological phenotypes characterizing T2D. It was demonstrated that the greater the extent of peripheral insulin resistance was, the lower the glucose uptake was in the brain when a late middle-aged individual cohort that has AD parental history was examined [134]. Contrary to the periphery, however, where glucose uptake is largely insulin dependent, the brain uses nearly 20% of all glucose in the body in a process that is largely insulin independent: when rats were rendered insulin deficient with streptozotocin treatment that destroys insulin-producing pancreas, intracerebral delivery of leptin restored glucose sensitivity [33]. These results also suggest that leptin in the brain can control blood glucose levels in insulin-independent manner.

Insulin receptor (IR) and insulin-like growth factor (IGF) 1 and 2 receptors are widely expressed in the brain [129, 130] (Allen Brain Atlas). IR and IGF receptors can form heterodimers with each other, which gives them different affinity to different types of ligands, such as insulin, IGF-1, and IGF-2 [57]. When IR was deleted in neurons or astrocytes, insulin resistance ensued suggesting IR is involved in insulin resistance. A neuron-specific knockout of IR study using nestin-cre line demonstrates that mice developed obesity, mild insulin resistance, and demonstrated elevated levels of insulin and leptin in plasma [15]. On the other hand, when IR was deleted in astrocytes using hGFAP-creERT2 and GLAST-creERT2 mice, not only insulin signaling was attenuated, but also glucose uptake [30]. These results suggest that astrocytic IR may participate in developing insulin resistance, but more importantly, it plays a critical role in glucose uptake. It should be noted that hypothalamic tanycytes also express GFAP [106] and become labeled in GLAST-creERT2 mice [105], suggesting it is plausible that tanycyte-mediated glucose uptake contributed to the phenotype.

In AD mouse background, the role of IR signaling appears mixed. When nestin-cre-derived IR null mice were crossed with Tg2576, Aβ peptide levels were reduced with an attenuation in p-AKT and p-GSK3 levels [124], suggesting IR signaling contributed positively to amyloid deposition. When IGF-1 receptor was deleted from APPswe/PS1∆9 mice, however, it resulted in modestly higher scores in cognitive tests, a reduction in astroglyosis and synaptic loss, but increased Aβ aggregation [21], suggesting unlike IR, IGF-1 receptor contributes negatively to the development of AD pathology. The authors argued that IGF-1 signaling is also involved in defibrillation of Aβ plaques, but the mechanisms remain to be investigated.

Like the receptor knockout studies, a clear consensus has not been reached as to whether brain insulin levels are indeed elevated and whether insulin resistance is being detected in the brain from AD mouse models as well as in human AD brain samples [120]. For instance, some reported a reduction in p-AKT and p-GSK3 levels [70, 121], downstream effectors of insulin signaling, while others reported an increase [36, 94, 139]. A complicating issue in these studies is that the kinases tested can also be modulated by signals other than insulin/IGFs, rendering it difficult to assign the observed changes to insulin exclusively. Perhaps most convincing report came from the study that measured direct phosphorylation of IR and its associated IRS-1/2, after stimulating human postmortem tissue explants from AD and control subjects with insulin [126]. Similarly, Bomfim et al. also reported an increase in p-IRS-1 in cynomolgus monkeys following direct infusion of oligomeric Aβ into their brains for 24 days [12]. The latter report suggests that Aβ peptides themselves signal to activate IR, perhaps initiating insulin resistance as Aβ peptides accumulate. Aβ peptides also appear to influence metabolic dysruption in a different AD mouse model of 3XTG mice that expresses APPswe, tauP301L under the Thy1.2 promoter in PS1M146V knockin background. Only the female and not male mice developed glucose intolerance and insulin resistance that worsened progressively in correlation with pronounced Aβ deposits [132], suggesting insulin resistance develops with amyloid pathology.

Leptin Resistance

Leptin resistance is another cellular and physiological phenotype of T2D that can potentially influence AD pathology [48]. Leptin is a hormone produced by white adipocytes [40] but signals in the hypothalamus after crossing the blood-brain barrier [40]. With obesity, circulating leptin levels increase resulting in hyperleptinemia, which can elicit leptin resistance [28]. When APP23 mice that express APPswe under the mouse Thy1 promoter were crossed with leptin mutant ob/ob mice, the mice exhibited glucose intolerance that is worse than ob/ob mice, and also developed insulin resistance at 3 months, well before Aβ begins to accumulate, while ob/ob mice did not [125]. Already at 2 months, mice exhibited severe cognitive deficits although there was very little Aβ detectable. The authors also crossed the AD mice with NSY mice that develop T2D-like symptoms [113], and a similar phenotype was observed albeit at much attenuated levels. Although the phenotype of premature cognitive deficit can be attributed to obesity-associated pathology, these data clearly highlight the role of leptin in cognition. It also implicates defective leptin metabolism in the development of AD pathology.

Defective leptin metabolism has actually been shown in Tg2576, which had lower serum leptin levels at 3 months compared to the wild-type controls [48]. This is in part due to a significant reduction in total body fat, detectable at ages younger than 2 months old in Tg2576 mice. Since Tg2576 mice do not show plaques until they are 11–13 months old, these data illustrate the presence of an early metabolic disruption in the mice. Correlative to low serum leptin levels, Tg2576 mice appear to have a functional defect in the hypothalamus; their fasting glucose levels are higher than in the wild-type mice, and fasting-induced increase in NPY and AgRP RNAs was completely lost in Tg2576 mice. NPY-positive arcuate neurons from Tg2576 mice also failed to respond to leptin electrophysiologically, and the authors demonstrated this was likely due to Aβ42 interfering with leptin action. In line with these data, we found that 5XFAD mice exhibit leptin resistance at 6 months even on normal chow but not insulin resistance, unless fed with 60% high-fat diet (HFD, Fig. 2, insulin resistance not shown). Whether leptin levels are altered in human AD cases, on the other hand, is not conclusive: higher cerebrospinal fluid (CSF) leptin levels were reported in AD subjects without any correlation with BMI [13], but serum leptin levels were lower in AD cases compared to control individuals [97].

Fig. 2
figure 2

5XFAD mice develop leptin resistance at 6 months on normal chow. 5XFAD mice were subjected to leptin sensitivity assay by measuring 24-h food consumption after 5 mg/kg of human leptin injection

Full size image

If Aβ42 indeed interferes with leptin signaling, the mechanism by which Aβ42 inhibiting leptin signaling needs to be elucidated. If the phenotype is due to an inhibitory crosstalk between the Aβ42 and leptin signaling pathways, potential targets should be those that are modulated by both Aβ42 and leptin. A candidate is AMP-activated protein kinase (AMPK), since Aβ42 activates AMPK [77, 128, 142], while leptin inhibits it [82]. Indeed, AMPK phosphorylation was increased in APP/PS2 mice that express hAPPswe and hPS2-N141I under the Thy1 and Prp promoters, respectively [71], as well as in APPsw/PS1∆E9 mice under the Prp promoter [118]. We have also found that AMPK activity is significantly increased in 5XFAD mice (data not shown), which increases further with HFD. Although the fact that an increase in AMPK activity is observed in multiple AD models is encouraging in placing the focus on AMPK as the putative target of leptin resistance, the role of AMPK activation and leptin resistance itself in overall AD pathology needs to be fully examined.

The transport of leptin into the brain constitutes an important part of leptin metabolism since leptin acts mostly in the brain. Leptin is produced from adipocytes and transported into the brain through microvessles and the fenestrated barrier in the medial eminence. Studies that utilized targeted leptin receptor knockout mice suggest that the leptin receptor itself is involved in leptin uptake. When leptin receptor was deleted in endothelial cells using Slco1c1-creERT2 line, 125I-leptin uptake was reduced by 60% in the cortex and ventral tegmental area, and ~ 40% in the hypothalamus, reflecting 50% reduction in leptin receptor RNA detected in the brain [23]. In the global leptin receptor knockout mice, however, leptin uptake was reduced only by 40% [43]. It is possible that the discrepancy is due to some leptin molecules crossing the blood-brain barrier that does not rely on endothelial cells, such as the fenestrated barrier of the arcuate and medial eminence. Tanycytes in the medial eminence were shown to take up leptin both in vitro and in vivo [5], which were not targeted in endothelial deletion of the leptin receptor.

Whether leptin levels are altered in human AD subjects is not completely settled. In a 9-year follow-up study of elderly patients, it was found that the higher the serum leptin levels were, the lower the risk of developing dementia and AD [68]. Similar results were also reported by others [9, 44, 51, 56]. In contrast, other groups reported that lower serum leptin levels were associated with cognitive impairment in T2D patients, and that leptin levels in the CSF and plasma were higher in AD and dementia patients [141] and a significant increase in CSF leptin levels among AD cases [13]. The reasons for the opposite results are not clear at the present time.

Disruption in Circadian Rhythm

It has been reported since the early 1980s that AD patients experience disturbed sleep, which is now being called as “sundowning” behavior [98, 99]. Disruption in circadian rhythm (CR) can have significant influence in overall metabolic balance [27]. There is a strong correlation between circadian disruption and cardiovascular disease, obesity, and hyperglycemia, as observed among individuals who work on the night shift [54, 55, 93, 131]. In addition, expression levels of many metabolic enzymes as well as proteins involved in detoxification and stress responses cycle according to CR. Some of these include glucose transporter, glucokinase, glucagon, pyruvate kinase, acyl CoA dehydrogenase, aldehyde dehydrogenase, fatty acid synthase, glycogen phosphorylase, nicotinamide phosphoribosyltransferase, and AMPK [19, 29, 59, 60, 66, 91]. Levels of Aβ peptides and tau in the brain also fluctuate according to CR, suggesting CR influences APP and tau metabolism.

Recent research demonstrates a clear link between disturbed sleep and accumulation of Aβ peptides and tau. For instance, disrupting normal sleep in mice led to a 2-fold increase in ISF tau [46]. An increase in tau in CSF was also observed in humans by sleep intervention. Similarly to tau release, Aβ peptide levels increased upon sleep deprivation both in mouse ISF and human CSF upon sleep perturbation [53, 72, 107].

ISF and CSF Aβ peptides are known to cycle according to diurnal patterns both in rodents and humans [53]. Since administration of tetrodotoxin (TTX) that blocks sodium channel reverses the increase in ISF Aβ, it is hypothesized that neuronal activity is responsible for the increased release of Aβ peptide at nighttime. The link between neuronal activity and an increase in ISF Aβ and APP processing has been demonstrated [20, 52]. What would be the basis for increased neuronal activity in synaptopathic diseases such as AD? It was reported that although there were general depression in excitatory Ca2+ transients in APP23:PS45 mice compared to the wild-type controls, a group of neurons that were located within 60 μm from amyloid plaques exhibited a surprising hyperactivity [16]. The hyperactivity was attributed to defective GABAnergic synaptic inhibition based on pharmacological studies [16]. These results may lend support for the observation that epileptic seizures are more common among AD patients [2, 41, 42, 81, 90, 108, 110]. If abberrant seizure-like activity is prevalent in AD brains, one can envision that such activity can deliver stimuli that ultimately leads to accumulation of Aβ peptides. Whether ISF Aβ peptides are released preferentially from plaque-associated ones is unknown, and so is the question whether Aβ released upon neuronal activity is in some way related to hyperactive neurons: increasing neuronal activity facilitates cleavage of APP [52]. Could sleep deprivation increase perhaps aberrant or excessive neuronal activity like epileptic seizures at night? It is noteworthy that when mice were optogenetically stimulated in the perforant pathway to provide chronic neuronal activation, they exhibited seizure soon after each optogenetic stimulation [138]. These data suggest that the intensity of neuronal activity evoked by experimental stimulation paradigm is higher than normal neuronal activity. Whether seizure occurrences are higher at night in AD patients is currently not known.

An alternative explanation for the link between sleep deprivation and increased Aβ and tau release in ISF and CSF may be that sleep deprivation inhibits efficient disposal of Aβ peptides, since glymphatic system becomes more efficient during sleep by creating convective fluxes in periarterial space [137]. Decreased disposal of Aβ peptides upon sleep disturbances also provide explanation to the observed correlation between lower CSF Aβ 42 and rapid progression of dementia [117]. It is unlikely, however, that reduced ISF tau disposal underlies the increased CSF tau levels, since the higher CSF tau correlates with more severe dementia in humans [117].

When sleep is disturbed, it also disrupts our normal CR. It is well known that there is a strong connection between the CR and metabolism. The central clock in suprachiasmatic nucleus (SCN) is influenced by internal metabolite levels including blood glucose and lipids [7, 89]. The identity of metabolic nutrient sensors that are coupled to the CR is currently unknown, but the focus is on AMPK and the pathway involving NAD+/Sirt1 metabolism, since both AMPK and Sirt1 activities are regulated by the clock as well as nutritional status [7]. AMPK phosphorylates crytochorme, Cry, directly [62]. Accordingly, the CR in the liver of LKB1 knockout mice is disrupted. LKB1 is not the upstream kinase for AMPK in the brain [135], but the identity of a kinase that phosphorylates and activates AMPK in the brain in a CR-dependent manner is yet to be deterimined. Sirtuins are part of a feedback loop with the core clock in NAD-dependent manner [3, 87]. Clock-BMal1 in turn activates the genes involved in NAD+ synthesis, whose cellular concentration cycles in alignment with the CR.

It should be added here that AD transgenic mice are known to exhibit a disrupted CR. 5XFAD mice exhibited dramatic alterations in the CR at 8 months of age, but even at 2 months, BMal1 and Per2 RNA levels in the SCN began to lose daily oscillations [119]. The difference was attributed to Aβ inducing degradation of BMal1 in the nucleus, thereby inhibiting Per2 induction. Also in 3xTg-AD mice, CR was disrupted [123]. Whether Aβ is responsible in disrupting the CR is unclear, and neither the mechanism by which Aβ signals to induce degradation of clock proteins, especially because APOE null mice also exhibited a defect in the CR [145]. Nonetheless, these studies suggest that restoration of normal CR can certainly be a way to improve metabolic imbalances, which would have positive impact on certain aspects of AD pathology.

Metabolic Disruption in AD Brains

The literature clearly suggests a metabolic dysfunction exists in AD brains, and it begins long before syptoms appear. But, there are many outstanding questions that need to be addressed. What cellular processes or molecules are responsible for initiating metabolic disruption? What are the consequences of metabolic disruption in AD brains? Does peripheral metabolic dysfunction influence the brain metabolism? If so, what is the underlying mechanism? Can we reverse the metabolic defect with diet?

Of the mutations that confer increased susceptability for developing AD [61], ApoE and perhaps clusterin, may be more relevant to understanding metabolic dysfunction in AD, since both are lipid-binding proteins. ApoE4 homozygosity in particular presents the greatest genetic risk for late-onset AD, and 50–60% of AD cases harbor either 1 or 2 copies of ApoE4 alleles [47]. Individuals who are homozygous for E4 alleles are 12 times more likely to develop AD, and those who are heterozygotes, 3.7 times [120]. Glucose hypometabolism in their brain is detectable very early long before cognitive impairment surfaces [104]. The risk of developing cerebral amyloid angiopathy (CAA) is also much higher among ApoE4 carriers with T2D than those with diabetes or E4 alone [95]. ApoE4 is also hypothesized to be the critical molecule that could relay the peripheral insulin metabolism to the CNS: when ApoE3 and ApoE4 knockin mice were rendered insulin resistant in the periphery by HFD, there was a reduction in p-AKT and p-GSK3 in the hippocampus and cortex from ApoE4- but not in ApoE3-KI mice, suggesting that ApoE4 may be involved in conveying peripheral insulin resistance to the brain [144]. Insulin-mediated AKT activation via infusion with reverse microdialysis was also attenuated in greater extent in ApoE4 compared to ApoE3 knockin mice, mainly due to ApoE4 interfering with recycling of the IR to the neuronal surface. Whether there exists a crosstalk between ApoE4 and leptin signaling has not been tested. Of note, clusterin was shown to augment leptin signaling in the hypothalamus [17].

Concluding Remarks

Recent analyses of human AD brain samples revealed that AD is a complex disease with many comorbities associated with it [39]. To better combat AD, we need a greater understanding of the disease progression in human cases not only in AD but also other diseases that exhibit similar comorbidities that are linked to AD. For instance, although some in the field begin to name AD a type 3 diabetes due to metabolic disruption observed in the AD brain [121], only limited data are currently available on FDG-PET from T2D patients at presymptomatic stages. Progress in therapy will be much greater in close communication between clinical scientists who examine the pathology and basic scientists who study the underlying mechanisms or the effect of a particular etiology in animal models.

Required Author Forms

Disclosure forms provided by the authors are available with the online version of this article.