While Aβ depositions and tau accumulation are the major hallmarks of AD, abnormalities in cerebral glucose metabolism in specific brain areas forecast the diagnosis of AD based on histopathological changes and can even be correlated with the disease progression [44, 45]. Cerebral glucose metabolism can be detected by [18F]fluorodeoxyglucose (FDG)-PET, an imaging technique based on the uptake of a radiolabeled glucose analogue. Thereby, reduced glucose metabolism (hypometabolism) was detected in AD brains, especially in the temporoparietal cortex including the precuneus and posterior cingulate cortex (Fig. 2) [46,47,48]. A negative correlation was observed between amyloid depositions and glucose metabolism bilaterally in the temporal and parietal lobes, brain areas that showed glucose hypometabolism in AD subjects [49]. Adding FDG-PET as a clinical tool improves the accuracy and confidence of diagnosing AD, indicating that glucose hypometabolism plays an essential role in AD pathogenesis [46, 50]. Additionally, FDG-PET is useful for staging the disease and was found to be more reliable in predicting the conversion of MCI to AD than MRI or CSF biomarkers [51,52,53]. FDG-PET even provides a higher predictive value compared to the amyloid PET, however, the combination of both methods provides the most accurate prediction of the conversion from MCI to AD [54, 55].
Type 2 diabetes mellitus (T2DM) is another risk factor for cognitive decline and late-onset AD. It is a metabolic disorder characterized by insulin resistance, impaired insulin signaling and glucose intolerance. The link between dementia including AD and diabetes has been shown [56,57,58,59,60].
Insulin is a peptide hormone which is secreted from beta cells located in the pancreas as a response to increased glucose levels in the blood. The insulin receptor (IR) can be found in various regions and most cell types of the developing and adult brain. It is still controversial whether insulin can also be produced in the brain [61]. By passing the BBB, peripheral-produced insulin can enter the central nervous system (CNS) where it regulates food intake but also affects cognitive functions such as memory. On the cellular level, insulin modulates multiple essential mechanisms [62]. In neurons, insulin regulates synaptic plasticity by enhancing neurite outgrowth, regulating the expression of NMDA, AMPA and GABA receptors and inhibiting apoptosis to promote neuronal survival. Even though IRs are more concentrated in neurons, they are also expressed in 20-40% of glia cells which provide support for neuronal cells in the brain. In astrocytes, IRs are involved in brain glucose sensing and in the entry of glucose and insulin into the brain [63].
Activation of the insulin signaling pathway starts with insulin binding to the alpha subunit of the IR located at the plasma membrane (Fig. 3) [64] Binding of insulin causes a conformational change and autophosphorylation of the intracellular β subunit of the IR leading to the endocytosis of the insulin-IR complex. At the same time, autophosphorylation of the IR enables growth factor binding protein 2 (GRB2) to bind to the Src homology 2 (SH2) domain of the phosphorylated tyrosine residues of the IR. This activates the downstream cascade of RAS, RAF and MEK1 kinase and finally ERK1 and ERK2, two MAP kinase isoforms. Next to the activation of the MAP kinase pathway, insulin also induces the AKT/PI3 kinase pathway. Phosphorylated IR substrate proteins interact with phosphatidylinositol 3-kinase (PI3K), which then catalyzes the formation of phosphatidylinositol (3,4)-bisphosphate (PIP2) to phosphatidylinositol (3,4,5)-trisphosphate (PIP3) [64]. As a result, 3-phosphoinositide-dependent protein kinase 1 (PDK1) is recruited which phosphorylates and activates AKT. AKT is a key regulator in the insulin signaling pathway and regulates several downstream proteins, such as the transcription factor FOXO1, glycogen synthase kinase 3 (GSK3) as well as mTOR and Forkhead transcription factors [65, 66]. While the MAP kinase pathway promotes mitogenic effects, the AKT/PI3 kinase pathway is involved in metabolic effects such as glucose uptake, glycogen synthesis and gluconeogenesis [67].
APOE influences insulin signaling at different levels. Sex as well as APOE genotype-based differences were observed in insulin BBB pharmacokinetics with APOE4 mice displaying higher insulin binding to brain vasculature than APOE3 mice [68]. Sex and APOE4 genotype were also found to influence CSF insulin levels. Women and APOE4 non-carriers had higher levels of CSF insulin which was associated with worse cognitive performance and higher CSF levels of (p-)tau indicating cerebral insulin resistance and impaired cerebral insulin signaling [69]. However, research have been mainly focusing on APOE3 and E4 carriers, leaving open questions about the effect of APOE2 on insulin signaling. High fat diet (HFD)-induced insulin resistance in female homozygous human APOE3 and APOE4 targeted replacement mice led to alterations in purine metabolism, glutamate metabolism and the pentose phosphate pathway (PPP). Further, cognitive deficits in hippocampal-dependent spatial learning and memory after HFD were enhanced in APOE4 mice [70]. APOE4, but not APOE3, also impaired insulin-stimulated mitochondrial respiration and glycolysis [71]. Protein expression of insulin signaling pathways and mitochondrial electron transport chain was changed in APOE4 mice [72]. LRP1, a major APOE receptor, interacts with the IR β-subunit in the brain. Deficiency of LRP1 resulted in impaired insulin signaling via reduced levels of IRβ in the presence but also absence of insulin and consequently also reduced levels of phosphorylated AKT (p-AKT). Additionally, in the absence of insulin the levels of LRP1 were downregulated. Treatment with insulin resulted in higher presentation of LRP1 at the cell surface [73]. LRP1 also regulates hepatic insulin response and somatic inactivation of LRP1 has been linked to insulin resistance [74].
Further, APOE can modulate insulin signaling directly. Recombinant APOE binds to IRs, with APOE4 showing a higher affinity than APOE3 resulting in less insulin-IR interaction with 50% reduction in the presence of APOE4 and 25% less insulin binding in the presence of APOE3. Interestingly, binding between APOE4 and IR increased with the age of the mice [71]. APOE4 suppresses activation of insulin signaling by blocking the interaction between insulin and IR. Thereby, IR internalization and signaling activation is downregulated. Further, APOE4 impairs IR trafficking by trapping it in the endosomes which leads to downregulation of IRs at the cell surface [71]. Bioinformatic analysis revealed insulin signaling pathway key molecules being affected in AD patients of which the six kinases, leucine-rich repeat kinase 2 (LRRK2), glycogen synthase kinase 3 beta (GSK3B), AKT serine/threonine kinase 1 (AKT1), epidermal growth factor receptor (EGFR), mitogen-activated protein kinase 1 (MAPK1), and the non-receptor tyrosine kinase FYN, played central roles [56]. In addition to the interaction with IRs, APOE can also activate the AKT pathway by direct binding to AKT. Interestingly, AKT is phosphorylated in an APOE isoform dependent manner [75, 76]. APOE can induce AKT phosphorylation via two pathways. In the first, APOE binds to heparan sulfate proteoglycan/ LDL receptor-related protein complex. This pathway is only activated by APOE3 and APOE4 but not by APOE2. The second and later activated pathway is mediated by calcium and also APOE genotype dependent [76]. Insulin treatment leads to activation of AKT pathway and higher phosphorylation of tau (at Ser202/ Thr205) in APOE4 mice compared to APOE3 mice, indicating that APOE4 potentiates insulin-dependent activation of AKT pathway and tau hyperphosphorylation in the brain as observed in APOE4 carriers [77]. APOE4 reduces brain insulin signaling by lowering insulin receptor substrate 1 (IRS1) at S636/639 and AKT phosphorylation at T308 which was linked to lower brain glucose in fasting human APOE4 targeted replacement mice and with increasing age reduction in IRS2 expression, IRS1 phosphorylation at Y608, AKT phosphorylation at S473 and MAPK activation [78].
APOE has been shown to activate the PI3K/ AKT pathway to modulate food intake [79]. APOE4, bound to VLDL, inhibited the apoptotic pathway normally activated by HDL by reducing the phosphorylation of AKT through inhibition of PIP3. APOE4-VLDL inhibited PIP3 via SHIP2, known as risk factor for T2DM [80]. PDK1 mutant mice with reduced ability of AKT activation displayed enhanced membrane localization and activity of α-secretase and reduced neuronal resistance against neurotoxicity indicating that the AKT pathway also plays an important role in Aβ pathology. At the same time, PDK1 knock-in mice showed protection of neurons against ER stressors [81]. Pre-treatments with APOE2 and APOE3 reduced stress-induced neuronal apoptosis by regulating the AKT/ FOXO3a/ BIM signaling pathway, while no neuroprotective effect was found with APOE4 with higher concentrations of APOE4 showing a rather toxic effect [75]. These studies indicate that insulin signaling plays a crucial role in AD pathology revealing key elements such as the AKT pathway. However, more research is needed to decipher the exact mechanisms and how APOE isoforms influence key elements in these signaling cascades.
The human brain is a highly metabolically active organ that requires a constant delivery of fuels and oxygen to maintain its function. While the adult brain can utilize lactate and ketone bodies during special situations like exercise and fasting, its primary substrate is glucose [82]. This is taken up into the cytosol by highly efficient glucose transporters (GLUTs). In the cytosol, glucose is either stored as glycogen or further metabolized via glycolysis, the stepwise conversion to pyruvate. The presence of oxygen stimulates pyruvate dehydrogenase (PDH) to convert pyruvate to acetyl coenzyme A (Acetyl-CoA) in the mitochondria. Acetyl-CoA subsequently enters the tricarboxylic acid (TCA) cycle, also known as Krebs cycle, which finally produces NADH that feeds the electron transport chain (ETC) for oxidative phosphorylation (OXPHOS) to produce energy in the form of ATP. Under low oxygen conditions, cells depend on anaerobic glycolysis, in which pyruvate is converted to lactate by lactate dehydrogenase (LDH). This reaction is coupled to the conversion of NADH+H+ into NAD+ and is reversible, allowing cells to either produce or consume lactate dependent on their metabolic needs [83].
Glucose hypometabolism has been demonstrated as a common phenotype in MCI and AD, especially in APOE4 carriers. The metabolism of glucose is one of the processes that becomes dysfunctional at an early preclinical stage and correlates to cognitive impairment [84]. Elevated glucose concentration in the brain and reduced glycolytic flux are related to the severity of AD pathology and the establishment of clinical symptoms [85]. In recent years, the role of glycolysis in AD pathology has become more and more prominent. Analysis of glycolytic metabolites in the CSF of AD subjects showed that 5 of 11 metabolites, including phosphoenolpyruvate (PEP), 2-phosphoglycerate (2PG), 3-phosphoglycerate (3PG), pyruvate and dihydroxyacetone phosphate (DHAP), were significantly different from cognitively healthy non-AD controls. Except for pyruvate, a positive correlation with Aβ1-42 and Aβ1-42/Aβ1-40 was found [86]. Aβ resistant nerve cells have been shown to upregulate PDK1 expression and LDHA activity and subsequently lactate production a phenomenon called Warburg effect by which cells favor aerobic glycolysis over oxidative phosphorylation for energy production, even in the presence of oxygen, impacting metabolic efficiency [87]. This indicates that the Warburg effect is a contributing factor to support neuronal survival in the AD brain [88].
Different brain cell types display distinct metabolic phenotypes. Glucose can be taken up directly by neurons and used for oxidative phosphorylation. While neurons generally show high oxidative metabolism, astrocytes rather display glycolytic activity. Thus, glucose can be taken up by astrocytes and converted to lactate. According to the astrocyte-to-neuron lactate shuttle (ANLS) hypothesis, astrocytes are the primary producers of lactate which is then shuttled to neurons via monocarboxylate transporters (MCTs) and metabolized [89].
The impact of Aβ on glycolysis and the relation between enhanced neuronal glycolysis and Aβ protection has been shown [85, 88, 90,91,92]. However, APOE genotypes have major impacts on neuronal glycolysis as well (Fig. 4). Lactate production was shown to be higher in differentiated N2a-APOE2-expressing cells and lower in N2a-APOE4-expressing cells compared to N2a-APOE3-expressing cells [93]. Neuro 2A (N2a) is a mouse neuroblastoma cell line with neural precursor cell properties. An APOE4-induced glycolytic phenotype became more severe in aged N2a cells and older primary astrocytes, while N2a-APOE2-expressing cells displayed more robust glycolytic activity [94, 95].
When combining APOE2- and APOE4-expressing cells, an increase in APOE protein levels was observed together with enhanced glycolytic function based on elevated basal glycolytic rate and maximum glycolytic capacity. This effect was only found with APOE2 and not APOE3 indicating a neuroprotective function for APOE2 [95]. ATP can be released by astrocytes to mediate neuronal activity and ATP levels were shown to be higher in N2a-APOE2-expressing cells compared to N2a-APOE3-expressing cells supporting the hypothesis that APOE2 provides better neuronal support [93].
APOE4-associated deficits in glycolysis were compensated by enhancing the activity of peroxisome proliferator-activated receptor gamma coactivator-1 (PGC-1), a key regulator of energy metabolism, which might be due to an increase hexokinase 1 (Hk1) expression [93]. Hexokinases are the initial enzymes of glycolysis, converting glucose to glucose 6-phosphate (G-6-P). Hk1 is mainly associated with mitochondria exclusively promoting glycolysis, whereas Hk2 translocate between mitochondria and cytosol mediating glycolysis when bound to mitochondria and glycogen synthesis when located in the cytosol [96]. APOE4 leads to reduced Hk1 and Hk2 protein expression and impaired functional activity as well as the upregulation of two other key rate-limiting enzymes of the glycolysis, phosphofructokinase (PFK) and pyruvate kinase M1 (PKM1) and deficient glycolytic activity. Gene expression profiles of Hk1 and Hk2 were significantly downregulated in APOE4 brains, similar to Slc2a4 which encodes a protein that functions as an insulin-regulated facilitative glucose transporter. In contrast, an increased Hk protein expression and enzymatic activity as well as basal and maximum glycolytic flux and ATP production was observed in APOE2 transgenic mice [93]. Additionally, APOE2-expressing cells show downregulation of PFK and PKM1. Interestingly, when combining APOE2- and APOE4-expressing N2A cells, APOE2 mitigates the APOE4-induced hexokinase impairments [95].
Alterations in glycolysis by APOE4 were observed in an insulin-dependent manner [71]. Increased glycolytic activity was associated with elevated energy-related metabolite D-Fructose 6-phosphate and TCA cycle metabolites malate, citrate and isocitrate in APOE4 mice compared to APOE3 [97]. Insulin treatment leads to an increase in glycolytic rate and capacity in APOE -/- cells and treatment with APOE4 inhibited this increase in glycolysis, whereas treatment with APOE3 had no effect on insulin-induced glycolysis [71]. However, not only APOE but also LRP1 can interact directly with the insulin receptor and thereby regulate insulin signaling and glucose uptake. Higher glucose levels suppress LRP1 expression, and the absence of insulin downregulates levels of LRP1. LRP1 downregulation in neurons leads to reduced levels of GLUT3 and GLUT4 and consequently in reduced glucose uptake. In mice, LRP1 deficiency causes glucose intolerance in the brain and hyperglycemia was observed to suppress LRP1 expression, finally causing insulin resistance and glucose intolerance [73]. Interestingly, LRP1 is a major GLUT4 storage vesicle cargo protein necessary for functional GLUT4 storage vesicle trafficking. LRP1 interacts with the luminal domains to constituents of the vesicle including GLUT4 and with the cytoplasmic tail to AKT substrate of 160 kDa (AS160), an insulin-signaling pathway target. LRP1 deficiency reduces GLUT4 expression and decreases insulin-dependent glucose uptake [98]. Even though APOE isoforms did not cause changes in the expression levels of major glucose transporters of the brain, GLUT1, GLUT3 and GLUT4 [95], APOE can indirectly alter glucose uptake in an isoform-dependent manner.
As an alternative energy source, ketone bodies can provide energy in times of starvation. Ketone bodies include β-hydroxybutyrate (BHB), acetoacetate (ACA) and acetone can cross the BBB and ketones are rapidly metabolized by neurons. They can bypass glucose metabolism by entering the TCA cycle in the mitochondria for NADH synthesis and subsequently oxidation of the ETC and ATP synthesis [99]. In contrast to non-APOE4 carriers, elevated ketone body levels do not show cognitive benefits in APOE4 carriers indicating that APOE4 might impact the transport of ketone bodies executed by MCTs or the metabolism of ketone bodies in the mitochondria [100, 101]. Interestingly, cells expressing APOE2 or APOE4 both demonstrated a robust profile in uptake and metabolism of ketone bodies compared to APOE3 which showed less efficient ketone body utilization [93].
Neurons and astrocytes are tightly metabolically coupled and therefore dysfunction in astrocyte physiology has a great impact on the vitality of neurons [102].
Earlier studies showed that the presence of the E4 allele is associated with altered glucose metabolism [103, 104]. Astrocytes, known to be the major source of APOE production in the brain, displayed the highest glycolysis gene expression [105]. APOE4 astrocytes showed deficits in glucose uptake capability but increased glycolytic activity and decreased oxygen consumption compared to APOE3 [105, 106]. An APOE-directed global metabolic shift at the cost of glucose metabolism towards lipid oxidation and enhanced thermogenesis was observed in humanized APOE4 mice [107].
APOE2 astrocytes showed increased glucose uptake, higher maximum glycolytic capacity and glycolytic reserve compared to APOE3 cells [93]. APOE2 astrocytes also had a higher glycolytic flux whereas APOE4 astrocytes cycled more glucose through the PPP, a major regulator of redox homeostasis and biosynthesis, compared to APOE3 [106]. Treatment with Aβ led to an increase in astrocytic glycolytic activity, independent of the APOE isoform, however APOE4 astrocytes had a stronger increase in glycolytic ATP production compared to APOE3 astrocytes with no differences in the glycolytic output [108].
The production of lactate during glycolysis also dependent on the APOE isoform. Lactate synthesis from glucose as well as the total amount of NADH was higher in APOE4 compared to APOE3 and APOE2 [105, 106]. APOE4 was shown to increase LDH enzyme activity, whereas pyruvate entry into the TCA cycle was decreased and PDH enzyme activity was not altered indicating a shift towards aerobic glycolysis [105]. Compared to APOE2 and APOE3, APOE4 lead to an increase in the two main enzymes that control the entry of pyruvate into the TCA cycle, pyruvate carboxylase (PC) and PDH [106]. Interestingly, this shift towards aerobic glycolysis was also observed in humans. Decreased energy expenditure driven by a lower rate of oxygen consumption and increased lactate plasma levels was found in female APOE4 carriers compared to E4 non-carriers [105].
Mitochondria are essential for energy metabolism but also crucial for producing reactive oxygen species (ROS) to activate transcription factors and the release of cytochrome c to induce cell death [109]. Mitochondrial dynamics are regulated by the balance of continuous fission and fusion events and maintain mitochondrial shape, distribution, size and function [110].
Mitochondria are highly dynamic organelles that produce most of the chemical energy to cover the bioenergetic needs of the cell. In the mitochondrial cristae, a series of metabolic reactions and processes convert energy that is stored in macronutrients to the universal energy donor ATP. Enzymatic reactions in the TCA cycle, also known as Krebs or citric acid cycle, produce FADH2 and NADH that are required for the transfer of electrons in the ETC. This is composed of four major protein complexes (I-IV) that generate a charge and proton gradient across the inner mitochondrial membrane (IMM) which drives the ATP synthesis in complex V, known as ATP synthase. This process requires oxygen as a final electron acceptor and is called OXPHOS [109, 111].
Mitochondrial dysfunction presents a risk factor for neurodegenerative diseases including AD. The close link between early mitochondrial dysfunction and AD pathology has been demonstrated (for details see [112]) and research to untangle the role of mitochondria in AD pathology is still ongoing. Recently, the discovery of new mitochondrial components such as SHMOOSE, a small reading frame encoded microprotein, or a mitochondrial protein called cell-cycle exit and neuronal differentiation 1 (CEND1) that have been linked to AD pathology has been reported [113, 114].
Early mitochondrial dysfunction can be observed in AD patients [115]. In AD patient brains, the activity of the mitochondrial enzyme α-ketoglutarate dehydrogenase complex (KGDHC) correlates better with the clinical dementia rating than amyloid plaques or NFTs. This was especially found for APOE4 carriers, indicating that mitochondrial dysfunction might play an even more important role in subjects carrying the APOE4 allele [116]. Before the onset of any cognitive changes and amyloid or NFT pathology, APOE4 carriers display mitochondrial dysfunction especially in brain areas affected in AD supporting the hypothesis that mitochondrial dysfunction plays an essential role in early development of AD pathology [117].
Further, APOE4 carriers show downregulation of proteins responsible for mitochondrial biogenesis compared to non-APOE4 carriers. PGC-1α and NAD-dependent deacetylase sirtuin-3 (SIRT3) as well as the dynamic proteins mitofusin 1 (MFN1), mitofusin 2 (MFN2), and dynamin like protein 1 (DLP1) were all decreased in APOE4 carriers. These proteomic changes could be observed in AD compared to healthy control post-mortem brain tissue but not in MCI. APOE4 carriers display decreased levels of antioxidative stress response proteins such as superoxide dismutase 2 (SOD2) and forkhead-box protein O3a (Foxo3a) compared to non-APOE4 carriers.
Mitochondrial dysfunction is closely connected with AD pathology which has been primarily studied in neurons. ETC gene- and protein expression were altered in AD patients [118] and AD patient-derived neuronal cells [119]. Mitochondrial fission protein dynamin related protein 1 (DRP1) was shown to interact with Aβ, a process that increases with disease progression. Partial deficiency of DRP1 reduced Aβ production, mitochondrial dysfunction, maintained mitochondrial dynamics and increased mitochondrial biogenesis as well as synaptic activity in APP mice indicating that inhibition of DRP1-Aβ interactions protects Aβ-induced mitochondrial dysfunctions [120].
Furthermore, PGC-1α and PPAR-γ, both crucial factors in mitochondrial biogenesis, are implicated in Aβ generation and aggregation [121,122,123]. For example, PGC-1α, a co-factor for transcription factors including PPAR-γ, regulates the transcription of the β-APP cleaving enzyme BACE1. Both PGC-1α and PPAR-γ regulate mitochondrial dynamics and their levels were decreased in AD patients. Agonist-induced activation of PPAR-γ resulted in mitochondrial protection against oxidative stress by modulating mitochondrial fusion and fission events in hippocampal neurons [124].
The impact of APOE on neuronal energy metabolism has already been shown. APOE4 expression reduced protein expression of all subunits of mitochondrial respiratory complexes in neurons cultured from APOE transgenic mouse brains and N2A cells [125]. APOE4 neurons showed reduced expression of mitochondrial biogenesis regulators, PGC1α and PGC1β, as well as reduced maximal mitochondrial respiration and lower OXPHOS capacity. APOE4 neurons had smaller and more dispersed mitochondria in the cytoplasm compared to APOE3 neurons [126]. Enzymatic activity of complex IV was also shown to be decreased in APOE4 N2A cells. Interestingly, these reductions were only observed in neurons and not in astrocytes, indicating that APOE4 might have a different effect on neuronal mitochondria than on astrocytic mitochondria [125]. Mitochondrial dysfunction was not affected in mutant N2a-APOE4-expressing cells that lack the intramolecular interaction between the amino-terminal domain (Arg-61) and the carboxyl-terminal domain (Glu-255). Levels of complex IV were even restored in N2a cells with the treatment of a small molecule that disrupts the domain interaction, indicating that this domain interaction within APOE4 plays a crucial role in mediating mitochondrial dysfunction [125]. ETC complexes can be targeted directly by C-terminal-truncated APOE4 fragments and overexpressing APOE4 (1-272) fragment affects their activities resulting in altered mitochondrial function in neuronal cells [127].
APOE4 causes impairments in insulin signaling and a link between brain insulin signaling and mitochondrial function is evident. RNA sequencing analysis on endothelial monolayers showed that mitochondrial function is one of insulin-responsive pathways in the BBB endothelium [66]. Treatment with insulin increased basal and maximal mitochondrial respiration, glycolysis rate and glycolytic capacity in APOE-/- neurons and neurons treated with recombinant APOE3, while recombinant APOE4 inhibited insulin’s ability to increase mitochondrial respiration and glycolysis. This was also observed for lipidated APOE4 secreted from astrocytes, indicating that APOE4 but not APOE3 suppresses insulin-stimulated mitochondrial respiration and glycolysis in neurons [71]. Changes in proteins involved in insulin signaling and mitochondrial ETC were also observed in APOE4 mice. Most prominent differences were found in mitochondrial signaling including higher expression of complexes I, II and III which stands in contrast to other studies reporting reduced expression of all subunits of mitochondrial respiratory complexes in neurons cultured from APOE transgenic mouse brains as discussed above. Further, increased expression of pErk1/2 but decreased pGSK3β levels in APOE transgenic mice were observed while the pAKT/AKT ratio was not affected [72].
A study based on fMRI, electrophysiology and metabolomics showed that APOE4 disrupts mitochondrial function by causing a hyperactivity phenotype in the brains of aged mice, especially in the entorhinal cortex. This neuronal hypermetabolism was supported by the APOE4-induced increase of mitochondrial metabolites that are involved in cellular energy metabolism including the TCA cycle metabolites malate, citrate and isocitrate as well as carnitine, an important metabolite for the transport of FAs into the mitochondrial matrix for energy usage via β-oxidation. Additionally, APOE4 caused an upregulation of ATP and upregulation of several metabolites that are involved in mitochondria-related ammonia regulation such as leucine, inosine monophosphate and succinoadenosine [97]. This is in agreement with increased ATP production and OXPHOS observed in APOE4 iPSC-derived neurons, compared to APOE2 and APOE3 [128].
Mitochondrial impairments in neurons and astrocytes intricately converge within the complex pathogenesis of AD, highlighting the interconnected nature of these cellular processes in both cell types in AD. APOE4-driven astrocyte TCA cycle dysfunction in AD may impair neurotransmitter recycling, potentially contributing to synaptic dysfunction and neurodegeneration [129].
Dysfunctions in ETC have been shown to induce a 49-fold upregulation of APOE transcript and protein as well as APOE secretion in diverse cells including iPSC-derived human astrocytes. Further, a correlation between APOE expression and ETC subunits was shown in the 5xFAD mouse model, indicating a direct link between APOE expression and mitochondrial function [130].
Mitochondria in APOE4 astrocytes were more fragmented and formed less tubule structures. The fission protein DRP1 was upregulated, but no change in levels of the fusion protein OPA1 was observed. When adding APOE3 astrocytes to APOE3 neuronal cultures, mitochondrial function and aerobic glycolysis was increased. Adding APOE4 astrocytes however caused a shift towards glucose metabolism which was demonstrated by a higher basal respiration on glucose, a higher glycolytic rate, increased expression of multiple OXPHOS subunit complexes, higher levels of ATP and lower AMPK activation [126]. These studies indicate that APOE4 induces mitochondrial fission and a shift towards increased basal glucose metabolism.
Elevated function of mitochondria-associated ER membranes (MAM), a subdomain of the ER involved in lipid and cholesterol homeostasis, have been shown in AD patients. Fibroblasts treated with APOE4-conditioned astrocyte medium displayed an increase in ER-mitochondrial communication and MAM function compared fibroblasts treated with APOE3-containing medium, indicating that mitochondrial dysfunction affects multiple cellular communication pathways. Interestingly, this effect was only observed with lipoprotein-enriched but not unlipidated APOE protein [131].
APOE is also implicated in mitochondrial fatty acid (FA) metabolism. Active neurons produce toxic FAs but have a low capacity to store or consume them. Thus, they release these FAs, which can be taken up and metabolized by astrocytes. Lindner and colleagues showed that APOE4 astrocytes with ectopic accumulation of fat, a stress-associated condition, had a reduced capacity to clear potentially neurotoxic FAs because APOE4 reroutes them to secretion [132].
Increased accumulation of damaged mitochondria in the lysosome was linked with lysosomal dysfunction which we earlier showed to be affected in APOE4 astrocytes [24]. APOE4 causes cholesterol accumulation which hinders the lysosomal-dependent clearance of damaged mitochondria resulting in a shift in energy production with increased glycolysis but impaired mitochondrial function in human iPSC-derived APOE4 astrocytes [133]. The main type of mitochondrial damage in APOE4 transgenic mice was mitochondria-derived lysosomes associated with lipofuscin [134]. Mitochondrial function and cell survival could be improved by reducing DRP1/FIS1-mediated mitochondrial damage in human cardiomyocytes. Treatment with P110, a selective inhibitor for FIS1-mediated DRP1 recruitment to the mitochondria, enhanced mitochondrial structure [135].
Mitochondrial function is closely related with fatty acid metabolism, LD lipolysis, autophagy and glycolysis which could be especially observed in astrocytes. However, defects in mitochondrial metabolism also affect neuronal functions including synaptic function, oxidative stress, and finally cognitive function. APOE4 disrupts mitochondrial structure and function in the brain by targeting several mechanisms making it even more challenging for single target treatment approaches. A limitation of most of these studies, however, is the focus on APOE3 and E4 genotypes, leaving open the questions on how APOE2 affects mitochondrial function.
Increased brain metabolism (hypermetabolism) has been observed in the brain of individuals with MCI [136, 137]. Longitudinal FDG PET imaging in AD patients revealed hypermetabolism 25 years before symptoms onset. Remarkably, hypermetabolism was detected in brain regions that were hypometabolic at later stages of the disease [138]. However, the role of glucose metabolism in disease pathophysiology and whether it is cause or consequence, remains unclear.
Interestingly, APOE4 has been linked to increased brain metabolism and activity, especially in young individuals. Functional MRI analysis in 20-35-year-old healthy individuals showed greater task-related activation in APOE4 carriers relative to non-carriers in hippocampal regions [139]. Further, young APOE4 carriers exhibited greater hippocampal activity relative to non-carriers while accomplishing the same encoding task [140]. This indicates that APOE4 influences brain activity and metabolism decades before the onset of clinical symptoms or pathological hallmarks. This is supported by studies in APOE4 knock-in mice showing brain hypermetabolism in the absence of neurodegeneration [97, 141].
While this appears to contradict the reports of a hypometabolism in AD patients, it may be explained by an age- and pathology-dependent APOE effect on the brain. Increased metabolic activity in young brains of APOE4 carriers may drive the classical AD pathology and contribute to the accumulation of β-amyloid aggregates and phospho-tau [97] leading to neuroinflammation, synapse loss and neurodegeneration. This may subsequently result in a hypoactive state once brain pathology has reached a certain threshold (Fig. 5). Natural aging may further support this metabolic switch. This hypothesis is further supported by a recent study showing that APP knock-in mice display mitochondrial hypermetabolism before onset of pathology. Upon increasing pathologies, such as neuroinflammation, synapse loss and amyloid plaque accumulation, the brain shifted to a state of hypometabolism [142]. A time-dependent metabolic shift was also shown for cultured human astrocytes. Upon exposure to Aβ, astrocytes initially enhanced mitochondrial fusion to stabilize their energy production and increased glycolytic ATP production. This however, changed with prolonged Aβ-mediated stress, and cells showed impaired mitochondrial function and a shift towards fatty acid β oxidation to maintain constant ATP levels [90]. We hypothesize that this metabolic shift is further exacerbated by the presence of APOE4, leading to the typical brain hypometabolism observed in AD patients.
Nevertheless, an APOE4-dependent metabolic switch during aging, from an early hypermetabolism to a disease-associated hypometabolism, needs to be confirmed by further research and requires more attention in order to better understand the APOE4-induced transition from a healthy to a diseased brain.
