Inverse correlation between Alzheimer’s disease and cancer from the perspective of hypoxia

Sporadic Alzheimer's disease and cancer remain epidemiologically inversely related, and exploring the reverse pathogenesis is important for our understanding of both. Cognitive dysfunctions in Alzheimer's disease (AD) might result from the depletion of adaptive reserves in the brain. Energy storage in the brain is limited and is dynamically regulated by neurovascular and neurometabolic coupling. The research on neurodegenerative diseases has been dominated by the neurocentric view that neuronal defects cause the diseases. However, the proposal of the 2-hit vascular hypothesis in AD led us to focus on alterations in the vasculature, especially hypoperfusion. Chronic hypoxia is a feature shared by AD and cancer. It is interesting how contradicting chronic hypoxia's effects on both cancer and AD are. In this article, we discuss the potential links between the 2 diseases' etiology, from comparable upstream circumstances to diametrically opposed downstream effects. We suggest opposing potential mechanisms, including upregulation and downregulation of hypoxia-inducible factor-1α, the Warburg and reverse-Warburg effects, lactate-mediated intracellular acidic and alkaline conditions, and VDAC1-mediated apoptosis and antiapoptosis, and search for regulators that may be identified as the crossroads between cancer and AD.


Introduction
In the context of global population aging, the pathogenesis of aging-related diseases is a key topic in biomedical research, while Alzheimer's disease (AD) and cancer are among the leading causes of death and disability worldwide.
There are 2 forms of AD, which share the same neuropathological features (amyloid plaques and neurofibrillary tangles [NFTs]), but their pathogenesis is different. Familial AD accounts for less than 1% of the total number of AD cases (onset age < 65 years). It is an autosomal dominant disorder caused by mutations in amyloid precursor protein (APP), PS1, and PS2 and is characterized by more rapid disease progression, with an age distribution often described by a bell-shaped curve (Ashby et al., 2015;Swerdlow, 2007). Sporadic AD (SAD) (onset age > 65 years) is characterized by the progressive accumulation of age-related dysfunction. Some common genetic polymorphisms, such as the ApoE (apolipoprotein E) gene allele ε4, can affect susceptibility to SAD, and the incidence increases exponentially with age (Youmans et al., 2012). Since SAD is more prevalent, and metabolic and vascular factors may be more important in its pathogenesis, we will focus on SAD in this article.
Although research on AD and cancer is limited by selection bias, complex comorbidities, competing risks of death associated with aging, statistical methods, and confounding factors, cancer has been inversely associated with AD in a large number of epidemiological studies (Nolen et al., 2017;Rocca et al., 2011;Roderburg et al., 2021;Valentine et al., 2022). The first evidence came from Corsellis' autopsy series in 1962, with histological evidence confirming that subjects with AD had significantly lower rates of cancer than controls with other functional psychoses (Cumings, 1963). An inverse relationship between cancer and AD has also been found in the Framingham Heart Study, in which participants enrolled in a longterm follow-up cohort of up to 22 years were associated with the risk of developing AD from the time they entered the Framingham Heart Study. In this study, both AD and cancer were carefully defined, with cancer survivors having a lower risk of AD than those without cancer and individuals with AD having a lower risk of cancer. Neither observation can be explained by survival bias (Driver et al., 2012). Other studies, including Mendelian randomization studies and meta-analyses, have also supported this conclusion (Catalá-López et al., 2014;Seddighi et al., 2019).
There are different types of cancers, among which the negative correlation between smoking-related cancers (of the lung, pharynx, larynx, esophagus, stomach, pancreas, etc.) and AD was higher than that of nonsmoking-related cancers (prostate cancer, melanoma, lymphoma, and ovarian cancer; however, glioblastoma showed a positive correlation with AD (Sánchez-Valle et al., 2017;Seddighi et al., 2019).
Since both cancer and AD are diseases with complex and heterogeneous pathogenesis, it is challenging to explain the opposite relationship between their epidemiology. Although not every cellular or molecular pathway should have an inverse relationship, many existing oncology drugs have been validated to have favorable effects on AD, which supports the biological inverse correlation (Ancidoni et al., 2021;Araki, 2013;Monacelli et al., 2017).
AD and cancer may be thought of as opposite ends of the spectrum: cancer is the uncontrolled malignant proliferation of tumor cells, while AD is a process of premature death of nerve cells (Lanni et al., 2021;Schwartz et al., 2020;Zabłocka et al., 2021). Bhardwaj summarized the opposite relationship between the expression of genes involved in cell proliferation and apoptosis in these 2 diseases. In addition, they have multiple parallel genetic and biological pathways, but these parallel pathways often point in opposite directions (Bhardwaj et al., 2023;Nudelman et al., 2019;Plun-Favreau et al., 2010). It is important to explore the crosstalk between the 2 mechanisms (Ibáñez et al., 2014). Since both diseases have an important metabolic component, we focused on energy-related metabolic mechanisms, especially hypoxia. An interesting phenomenon is that chronic hypoxia is present in both AD and cancer. It is advantageous for cancer, leading to drug and radiation resistance, invasion, angiogenesis, increased tumor heterogeneity, and resistance to cell apoptosis. In contrast hypoxia, it is a disadvantage in AD, leading to hypoperfusion, altered brain metabolism, and apoptosis ( Fig. 1).
We note that while hypoxia is part of the pathogenesis of both AD and cancer, a cascade of downstream reactions ultimately leads to the opposite outcome ( Fig. 1: dashed box). What matters is how does this happen, as we discuss in more detail in the following.
A healthy brain requires a steady blood supply to maintain a steady energy state that meets all of its cellular and molecular needs. The brain is a specialized organ that is highly sensitive to both oxygen and energy: neurons have a high metabolic rate, and the brain's energy stores are limited (Iadecola, 2017). Therefore, the brain has evolved a unique autoregulation mode, that is, neurovascular and neurometabolic coupling, so as to regulate cerebral blood flow (CBF) quickly and accurately (Kannurpatti, 2017;Leybaert, 2005). This regulation is achieved through a variety of neurogenic, myogenic, and metabolic pathways under the coordination of the neurovascular unit (NVU), which can regulate local vascular blood supply and energy metabolism in response to local changes in neuronal activity levels. In the case of glial cells, at the synaptic level, astrocytes express glutamate transporter in the perisynaptic area to sense changes in neuronal activity, while, at the vascular level, astrocytes express glucose transporter-1 (GLUT1) at the luminal surface in contact with the vascular endothelium to facilitate glucose transport. Thus, neurovascular and neurometabolic coupling is achieved.
The relationship between cancer and AD appears to be inverse, but it is not clear how hypoxia contributes to this. In this context, the current paper will examine and analyze hypoxia and energy metabolism in an effort to identify key proteins and nodes underlying the regulation and effects of chronic hypoxia in cancer and AD.
We searched the PubMed, Web of Science, and Google Scholar databases using keyword combinations including "Alzheimer's disease," "ischemia," "cancer," "hypoxia," "metabolism," and "pathogenesis," with no date restrictions. However, inclusion was limited to materials published in English. In addition, we reviewed references in the selected publications to provide further information.

Chronic hypoxia has an important role in the pathogenesis of cancer
Cancer is not just an uncontrolled mass of proliferating cells but a complex disease (Futreal et al., 2004;Irigaray et al., 2007;Wu et al., 2016). Due to the interplay of many intrinsic and extrinsic etiologic aspects, tumors lose homogeneity early in their development and exhibit heterogeneity at multiple levels: between patients, between primary tumors and their metastases, and within individual tumors (Tabassum and Polyak, 2015). Tumors consist of heterogeneous populations of tumor cells that differ morphologically and phenotypically. They have unique levels of differentiation, proliferation, vascular distribution, inflammation, immunosuppression, and invasiveness in different regions. These subsets act as a major source of tumor metastasis and treatment resistance by interacting and competing with each other (Marusyk et al., 2014).
Hypoxia is a universal feature of solid cancers, caused by a mismatch between cellular oxygen supply and consumption (Schito and Rey, 2020). The causes of tumor tissue ischemia are complex. The microvessel density of tumor tissue is usually high. However, the interstitial pressure of tumor tissue increases, leading to the collapse of capillaries in the tumor, which will lead to hypoxia of tumor tissue (Heldin et al., 2004;Siemann, 2011). In addition, normal capillaries are relatively straight and regularly spaced according to the widely used Krogh's oxygen distribution model, while the abnormal newly developed intratumoral capillaries are tortuous and irregular, with extensive avascular space, which could not be reproduced (Baish et al., 1996;Siemann, 2011). Finally, systemic hypoxia of the host also contributes.
The relationship between tumor heterogeneity and hypoxia is bidirectional. On the one hand, hypoxia hinders DNA repair systems, including homologous recombination and mismatch repair, which may lead to genetic instability and mutation (Chan et al., 2008;Kondo et al., 2001;Luoto et al., 2013). Moreover, since spatially distinct cells in a tumor may be subject to different oxygen concentrations in the microenvironment, such differential selective pressures lead cells to develop adaptive metabolic responses to meet their energy and biosynthetic requirements, which is conducive to the emergence and maintenance of clonal heterogeneity (Hubert et al., 2016;Serganova et al., 2004). On the other hand, due to the heterogeneity of the tumor, it is difficult to form a regular and effective vascular system in the tumor area. This inefficient tumor vascular network of leaky and compressed vessels is characterized by telangiectasia, leakage, and slow blood flow and is unable to accomplish efficient blood flow and oxygen delivery.
There is also heterogeneity in hypoxia in tumors, and this heterogeneity shows both transient and long-term changes, creating a dynamic pattern of hypoxia levels that induce cellular responses and control interactions between tumor cells, stroma, and immune cells in the microenvironment (Bader et al., 2020). The heterogeneity can be considered from 4 aspects. First, there is heterogeneity in the range of hypoxia. The partial pressure of oxygen in most normal tissues is 40-60 mmHg (Brown and Wilson, 2004). However, due to the different initial oxygenation of solid tumor tissues originating from different tissues, different tumor sizes and stages, and different measurement sites of heterogeneous tumor tissues, the oxygen partial pressure in tumors varies greatly, ranging from 2 to 32 mmHg (Höckel and Vaupel, 2001;Horsman and Vaupel, 2016;McKeown, 2014). Therefore, tumor oxygenation is usually expressed as a median level. Polarographic electrodes can be used to measure the distribution of oxygen partial pressure (pO 2 ) in tumors and to evaluate the treatment and prognosis of tumors (Horsman and Vaupel, 2016).
Second, there is heterogeneity in the pattern of hypoxia. The different hypoxia patterns in tumor tissue can be divided into acute, chronic, toxic, and systemic hypoxia. They often exist at the same time and synergistically promote the occurrence, development, and drug resistance of tumors. Acute hypoxia is a sudden exposure to short-term hypoxia that is reversible and preserves cell viability through the activation of autophagy and metabolic adaptations (Chaplin et al., 1987). Chronic hypoxia, also known as diffuse hypoxia, is associated with diffusion. In 1971, Professor Folkman proposed the theory that "tumor growth and metastasis depend on angiogenesis" (Folkman, 1971). The oxygen diffusion limit of the vasculature ranges from 70 to 150 µm. Folkman suggested that new blood vessels are required for tissues, including cancer, which grow beyond 2-3 mm 3 . Both irregular vessels and unstable blood flow within the tumor result in the existence of a large amount of oxygen that cannot diffuse over a maximum distance and cannot meet the oxygen demand associated with tumor cell proliferation, especially in larger tumors ( Fig. 2) (Vaupel, 2004). In experimental settings, cells are considered to be chronically hypoxic when they are cultured under hypoxic conditions for hours to weeks. The duration of exposure affects the response of tumor cells to hypoxia, among other parameters; therefore, situations that can result in persistent and potentially permanent hypoxia need to be taken into consideration (Bayer and Vaupel, 2012). Hypoxic tumor cells may undergo necrosis; necrotic tumor tissue is not harmful to the host organism. Hypoxic cells may also be re-oxygenated, but the hypoxia-reoxygenation cycle may allow the tumor to acquire a more aggressive phenotype (Challapalli et al., 2017). Systemic hypoxia results from anemia or respiratory insufficiency and usually occurs after chemotherapy, radiotherapy, blood loss, and chronic obstructive pulmonary disease. Systemic hypoxia enhances the aggressiveness of cancer and is associated with a higher incidence of cachexia (Dubsky et al., 2008;Krzystek-Korpacka et al., 2007).
Third, there is heterogeneity in the extent of hypoxia. The distribution of hypoxia levels was highly heterogeneous, ranging from mild to severe (Hompland et al., 2021). The relative proximity of a tumor cell to the nearest blood vessel determines the oxygen gradient of the tumor cell, irrespective of whether the cell is located centrally or peripherally in the tumor. Because tumor cells grow faster than endothelial cells, tumor neovascularization often cannot fully match the perfusion requirements brought by continuous tumor growth, and the spacing between tumor cells and capillaries increases, which leads to varying degrees of hypoxia in the cells.
During mild and moderate hypoxia, the oxygen concentration is still sufficient to maintain cell activity. Chronic mild hypoxia is the most typical form of hypoxia in tumor tissue (Lee et al., 2020). At this point, hypoxia, far from being a disadvantage, may confer numerous functional benefits, including epigenetic modifications, tumor vascularization, metabolic alterations, metastatic signaling, invasion, and extravasation, all of which are regulated by hypoxia-inducible factor (HIF) (Brahimi-Horn et al., 2005;Semenza, 2012).
Because tumor tissue does not have an effective vascular network, the oxygen gradient within the tumor depends on the relative proximity of tumor cells to their nearest blood vessels and is not restricted to a specific distribution pattern. In this case, most of the available molecular oxygen is consumed by cells within 70-150 µm of the vessel, while cells in the deeper part are severely hypoxic and nutritionally deprived. Different types of hypoxia patterns overlap in space and time and work together to promote selective pressure on malignant cells and the development of adaptive metabolism.

Chronic hypoperfusion has an important role in the pathogenesis of AD
AD is an age-associated neurodegenerative disease. It is characterized by progressive and irreversible cognitive decline, brain atrophy, and eventual death (Hascup and Hascup, 2020). Traditionally, studies have identified 2 distinct neuropathological alterations in the disease, namely, amyloid beta peptides and NFTs, but the pathways that are dysregulated early have remained elusive (Krstic and Knuesel, 2013). Different hypotheses have been proposed for the pathogenesis of AD: (1) the β-amyloid (Aβ) deposition and hyperphosphorylated tau hypothesis, (2) the cholinergic hypothesis, and (3) the 2-hit vascular hypothesis. We will focus on the 2-hit vascular hypothesis and CBF alterations in AD, as it better corresponds to the characteristics of prodromal and progressing AD. In the 2-hit hypothesis, vascular injury and Aβ accumulation exert positive feedback on each other, so it is difficult to determine whether vascular injury is a cause or a consequence of AD. However, some studies have determined that changes in CBF precede cognitive decline. At the same time, increasing evidence supports the role of cerebral hypoperfusion in the transition from mild cognitive impairment (MCI) to AD, and cerebral hypoperfusion in the parietaltemporal junction area of individuals with MCI is a predictor of rapid conversion to AD (Chao et al., 2009;Lacalle-Aurioles et al., 2014;Matsuda, 2007). It should be noted that, in a recent study, an inverse association between MCI and AD was not found, which may be due to the fact that the pathology of MCI and AD does not completely overlap (van der Willik et al., 2018). There are different subtypes of MCI; not all MCI patients will progress to AD, and patients with longterm stable MCI may have microangiopathy (Sharma and Callahan, 2021;Solfrizzi et al., 2004;Srikanth et al., 2004). However, other vascular events (including transient ischemic attack and cerebrovascular accident) can increase the progression from MCI to AD and the rapid cognitive decline in AD patients (Jia et al., 2017;Li et al., 2011;Pąchalska et al., 2015).
According to the 2-hit vascular hypothesis, AD is a disease with mixed vascular pathology, including small vessel disease (Marchesi, 2011). Genetic predisposition and vascular risk factors (such as diabetes, hypertension, dyslipidemia, and obesity) (hit1) lead to blood-brain barrier (BBB) dysfunction and reduced CBF early in the disease or before neurodegeneration, triggering the cascade of events that precedes dementia (Biessels and Despa, 2018;Luchsinger et al., 2007;Nelson et al., 2016;Verghese et al., 2022;Whitmer et al., 2008). Increased Aβ (hit2) amplifies neuronal dysfunction, leading to the accumulation of Aβ and tau in the brain and the accumulation of Aβ around cerebral vessels, accelerating neurodegeneration and dementia, and contributing to the increase in the extent of pathological changes (Fig. 3).
The brain is a complex and functionally heterogeneous organ. The viability of the brain and its myriad functions are critically dependent on the continuous supply of energy substrates and oxygen through the bloodstream. The brain receives up to 20% of cardiac output at a mass of about 2% of total body weight and is responsible for consuming about 20% of oxygen and 25% of glucose. However, the brain's energy reserve is very limited, so it has evolved unique vascular characteristics to meet the physiological demands caused by neural activation (Attwell et al., 2010). This is mainly reflected in the NVU, an anatomical feature of CBF regulation, and neurovascular coupling, a functional feature of CBF regulation. When CBF stops, brain function ends within seconds, and neurons are damaged within minutes. Normal aging is known to reduce cerebral circulatory function, including detectable CBF attenuation in the limbic and commissural cortex, which is thought to underlie age-associated cognitive changes (Martin et al., 1991).
The NVU, which is composed of astrocytes, parietal cells such as vascular smooth muscle cells, pericytes, and endothelial cells, is the smallest functional unit of the brain. The NVU is essential for stabilizing the brain environment. First, the NVU maintains the integrity of the BBB. The BBB is a biological barrier structure composed of continuous endothelial cells, tight junctions, basement membranes, and astroglial endfeet that functions to control which molecules can enter the brain and which harmful proteins are removed from the brain parenchyma. Second, the NVU regulates CBF to meet the needs of neuronal activity through neurovascular coupling, thereby ensuring that adequate oxygen and nutrients are delivered to the desired brain tissue. In addition, the NVU is also involved in the formation of glymphatic systems. The glymphatic system is a component of the lymphatic system. Its function depends on aquaporin 4 channels in the terminal feet of astrocytes and cerebral arterial pulsations to regulate cerebrospinal fluid-interstitial fluid exchange and promote the clearance of interstitial solutes, including Aβ and tau (Iliff et al., 2013;Verghese et al., 2022). The disruption of the NVU, which can also lead to BBB dysfunction and decreased CBF, may be the cause of an early pathophysiological change in AD (hit1), which occurs when not only is the supply of oxygen and nutrients to the brain reduced but also the clearance of neurotoxic substances, such as Aβ and α-synuclein, from the brain parenchyma is reduced (Sweeney et al., 2018).
Neurovascular coupling is a unique mechanism of CBF control that involves communication between neurons, astrocytes, and cerebral vessels. During functional activation, regional CBF is expanded to meet the oxygen and energy demands through rapid regulation of microvascular diameter and accompanying upstream arteries and arterioles (supplying blood to the capillary bed). This mechanism ensures that the brain rapidly upregulates CBF and increases the rate of oxygen delivery to the activated site when the nerve is activated. Dysregulation of neurovascular coupling, particularly the functional effects associated with decreased CBF, is associated with cognitive dysfunction, the transition from a normal or mild cognitive impairment state to a dementia diagnosis, and agerelated functional changes.
The association between AD pathogenesis and reduced CBF in many brain regions has been supported by evidence from multiple neuroimaging studies (Zheng et al., 2019). A recent large study using arterial spin labeling magnetic resonance imaging demonstrated that CBF changes and vascular dysregulation are the initial events associated with cognitive decline prior to changes in the classical AD biomarkers Aβ and tau (Iturria-Medina et al., 2016). Several investigators have identified distinct changes in CBF as potential preclinical markers of AD (Hays et al., 2016). Decreases in CBF have been observed in individuals at high risk for AD before hippocampal atrophy, Aβ accumulation, and cognitive decline, and they continue to decline throughout the progression of AD (Badji and Westman, 2020;Bookheimer et al., 2000;Iadecola, 2004;Knopman and Roberts, 2010). The early stages of AD are marked by a decrease in CBF of approximately 20%. In the late stages of AD, metabolic and physiological abnormalities are further aggravated, leading to a decrease in CBF of 55%-65% (Hoyer and Nitsch, 1989).
Brain imaging studies in transgenic AD mice revealed that the brain exhibited marked hypoperfusion and cortical atrophy at 15 months of age (Hébert et al., 2013). In a rat model, bilateral carotid artery occlusion led to the accumulation of neurotoxic Aβ oligomers, memory impairment, neuronal dysfunction, and synaptic changes . Aβ accumulation and CBF reduction interact and promote each other. First, Aβ deposition will severely damage Fig. 3. The 2-hit hypothesis in AD. Abbreviations: AD, Alzheimer's disease; BBB, blood-brain barrier; NVU, neurovascular unit. cerebrovascular function, with increased arterial vasoconstriction and decreased CBF, while decreased CBF significantly increases Aβ cleavage from APP and leads to insufficient Aβ clearance and accelerated deposition by upregulating β and γ secretase (Thomas et al., 1996;Zhang et al., 2007). In addition, the presence of the apolipoprotein E4 allele, the strongest genetic risk for SAD, was also associated with reduced cerebral perfusion. Human apolipoprotein E (ApoE) has 3 major subtypes encoded by the E2, E3, and E4 alleles. The Ε4 allele of APOE increases the risk of SAD. ApoE is a glycoprotein of 299 amino acids. ApoE in the brain is mainly produced by astrocytes, acts as a lipid receptor to form lipoprotein particles, and participates in the pathological process of AD through a series of cascade reactions (ApoE cascade hypothesis) (Martens et al., 2022). Scarmeas found an association between the presence of apolipoprotein E4 allele and reduced cerebral perfusion with apolipoprotein E4 allele present in individuals as early as in their 20s (Brandon et al., 2018).
All these seem to indicate that chronic cerebral hypoperfusion seems to enhance different elements of AD pathology, such as increased accumulation of Aβ peptide and phosphorylated tau, and leads to disruption of the BBB, neuronal apoptosis, and white matter damage, ultimately causing cognitive impairment (Daulatzai, 2017).
There is a substantial overlap between risk factors for cerebrovascular disease and AD (e.g., diabetes, hypertension, dyslipidemia, and obesity). These risk factors act via the neurovascular system, leading to NVU destruction, BBB leakage, and neurovascular coupling dissociation, leading to cerebral hypoperfusion and a chronic hypoxic environment in the brain (hit1). Hypoperfusion can trigger the accelerated deposition of Aβ, and the accumulation of Aβ will worsen cerebrovascular function and further reduce CBF. Increased beta-amyloid (hit2) amplifies neuronal dysfunction, leading to synaptic damage and cognitive impairment.

HIF-1 regulates adaptive cellular responses to hypoxic conditions
At the molecular level, the cellular response to hypoxia is largely controlled by hypoxia-inducible factor-1 (HIF-1). HIF is part of a universal sensing system that regulates oxygen homeostasis and has 3 known isoforms: hypoxia-inducible factor-1α (HIF-1α), HIF-2α, and HIF-3α (Semenza et al., 1997). HIF1-α (encoded by HIF1A) is the most prevalent and controls the expression of more than 700 target genes that are directly or indirectly involved in adaptive and pathological processes related to hypoxia (Semenza, 2004;Wu et al., 2019).
The HIF complex is a heterodimeric transcription factor composed of 2 subunits, α and β (Leu et al., 2019). The β subunit is stably expressed and permanently present in the nucleus, whereas the stability of the α-subunit is dependent on oxygen availability. Under normoxic conditions, the alpha subunit is synthesized and hydroxylated by oxygen-dependent HIF prolyl-4-hydroxylases on 2 proline residues and rapidly degraded by ubiquitin-dependent proteasomes. Under hypoxic conditions, prolyl-4-hydroxylase is inactivated, allowing HIF-1α to escape degradation and translocate to the nucleus, where it forms a heterodimeric complex with HIF-1β, which acts as a transcriptional regulator and activates the hypoxic transcription program (Fig. 4) (Taylor and Scholz, 2022).
The major targets of HIF-1 include the genes for erythropoietin and vascular endothelial growth factor, GLUT1, pyruvate dehydrogenase kinase 1 (PDK1), and lactate dehydrogenase A (Leu et al., 2019). The adaptive response to hypoxia mediated by these genes is mainly reflected in the increased availability of oxygen and decreased metabolic oxygen consumption. The availability of oxygen to cells is increased thanks to erythropoiesis and angiogenesis. The adaptive response reduces mitochondrial mass and metabolic oxygen consumption by shifting glucose metabolites from mitochondrial metabolism, such as oxidative phosphorylation (OX-PHOS), the tricarboxylic acid (TCA) cycle, and fatty acid β-oxidation to glycolysis (ATP production in an oxygen-independent manner) (Solaini et al., 2010). In addition, some pathways that are not obviously related to hypoxia are also affected by HIF, such as apoptosis, pH regulation, inflammation, tumorigenesis, and histone demethylation .

HIF-1 is activated in cancer
HIF-1α expression is increased in many cancers, both as an adaptive event to hypoxia in the tumor microenvironment (TME) and as a consequence of alterations in many key genes in cancer cells Fig. 4. HIF-1 in normoxia and hypoxia. HIF-1 is composed of HIF-1β and HIF-1α. Under normoxia, HIF-1α is hydroxylated by PHD, and then it forms complexes with VHL and other proteins, followed by ubiquitination and degradation by the proteasome. Under hypoxia, HIF-1α lacks oxygen-dependent hydroxylation. It translocates to the nucleus and complexes with HIF-1β on HRE to form a transcription factor complex, which activates the transcription of HIF-1 target genes to mediate the adaptive response to hypoxia. Abbreviations: HIF-1α, hypoxia-inducible factor-1α; HRE, hypoxia response element; PHD, prolyl-4-hydroxylase; VHL, von Hippel-Lindau protein.
(such as loss of von Hippel-Lindau protein [VHL] tumor suppressor function or increased PI3K/AKT/mTOR activity) (Kaelin, 2008). The Warburg effect is a central factor in cancer progression and is driven by HIF stabilization (Vaupel and Multhoff, 2021). HIF shapes the plasticity of tumor cells and their microenvironment by regulating the expression of genes related to a variety of metabolic pathways. HIF-1 induces an increase in GLUT1 and GLUT3 to transport glucose into tumor cells. Glucose entering cells can serve multiple purposes, including glycogen synthesis, protein modification, and the PPP. In addition, tumor cells also reduce extracellular pH by activating HIF-1 target gene complexes, such as monocarboxylate transporter (MCT)-4, sodium hydrogen antitransporter (NHE-1), and carbonic anhydrase 9, forming an acidic environment, increasing the risk of tumor metastasis, and inhibiting cancer cell apoptosis (Brizel et al., 2001;Semenza, 2009).

HIF-1 is suppressed in AD
Activation of HIF-1α in cells represents an endogenous adaptive protection of cells, which helps increase oxygen supply to tissues and induces protection against ischemic brain injury (Chen and Sang, 2016;Zhu et al., 2014). The levels of HIF and its induced signaling pathways are both known to be downregulated during aging. However, in AD brains, HIF-1 levels were further decreased compared with age-matched controls, making the cells less resistant to hypoxia (Liu et al., 2008). Why HIF-1 is downregulated in the AD brain is not clear; one possible explanation is that increased oxidative stress in the AD brain disrupts and downregulates HIF-1, which triggers downstream hypoxic damage, including decreased expression of GLUT1 and GLUT3 in the brain. Under physiological conditions, the adult brain uses glucose almost exclusively as an energy substrate. Transport of glucose from blood to neurons requires the GLUT transporter family. GLUT1 is located in endothelial cells and astrocytes, whereas GLUT3 and GLUT4 are located in neurons. GLUT1 encoded by SLC2A1 is highly expressed in endothelial cells of the BBB and is then further transported to neurons via neuronal GLUT3. Intracellular glucose is metabolized through processes such as glycolysis (including lactate production and OXPHOS), the PPP, and gluconeogenesis. Under aerobic conditions, glycolysis is tightly coupled to the TCA cycle and mitochondrial OXPHOS, through which pyruvate, the product of glycolysis, is converted to a range of organic acids, and carbon dioxide is released. In addition, the TCA cycle reduces oxidized nicotinamide adenine dinucleotide to reduced nicotinamide adenine dinucleotide (NADH), which is sent to mitochondria for OXPHOS to produce ATP, the ultimate biochemical energy.
Decreased levels of GLUT1 and GLUT3 were also inversely correlated with tau hyperphosphorylation and increased NFT density, providing evidence that glucose transporters are involved in the formation of abnormal tau hyperphosphorylation and NFT, both of which are strongly associated with neurodegeneration (Liu et al., 2008).

Cancer is a metabolic disease characterized by mitochondriaassociated metabolic reprogramming (Warburg effect)
Hypoxia has been identified as a common feature of solid tumors, leading to different aspects of tumor progression. Low oxygen concentrations (hypoxia) are detrimental to most species. However, it is not a disadvantage but an advantage for the tumor, being beneficial for the growth, survival, and metastasis of cancer cells. In this context, the Warburg effect in cancer cells may be considered a selective reversible adaptive response to hypoxia (Fig. 5).
In 1930, Otto Warburg discovered the Warburg effect in cancer cells (Warburg, 1956). Comparing normal liver tissue with corresponding cancer cells, he found that, although liver cancer tissue sections breathed 20% less than normal tissue, they metabolized 10 times more glucose than expected. In addition, the amount of the glycolytic product lactate is 2 orders of magnitude higher in cancer cells than in normal tissues. Based on these phenomena, he proposed the Warburg effect, whereby tumor cells exhibit a higher rate of glucose metabolism than normal cells and preferentially utilize it in glycolysis rather than OXPHOS, even in the presence of sufficient oxygen (aerobic glycolysis) (Otto, 2016;Urbano, 2021). The incidence of stable HIF-1α under normoxic conditions is approximately 50% in cancer, and HIF-1α directly upregulates genes encoding glucose transporters such as GLUT1 and enzymes of the glycolytic pathway (Welsh et al., 2004). It is thought that glycolysis in tumors may be stably driven by HIF-1α, independent of the hypoxic environment (Maxwell et al., 2001). The relationship between cancer and the Warburg effect has been used to image tumors, by using the metabolic marker 18 F-deoxyglucose in positron emission tomography. 18 F-fluorodeoxyglucose is a radioisotopically labeled glucose that selectively accumulates in tumors because of increased glucose uptake due to enhanced glycolytic pathway activity that occurs in cancer cells (Carvalho et al., 2011).
The Warburg effect has been documented for more than 90 years, but its function remains unclear. It is currently believed that the Warburg effect is not simply an adaptation to hypoxia (Gatenby and Gillies, 2004). Rather, it is responsible for the malignant phenotype of cancer, participates in the metabolic profile that constitutes 70%-80% of human cancers, and is a central feature of the "selfish" metabolic reprogramming of cancer cells (Hanahan and Weinberg, 2011;Vaupel et al., 2019).

Warburg effect is a faster way to produce ATP
Due to the heterogeneity of tumor cells, not all cancers rely on glycolysis. Glycolysis contributes from 1% to 64% of the total ATP in cancer cells (Zu and Guppy, 2004). However, glycolysis is still an advantage of tumor cells, and the glycolytic activity is related to the degree of tumor malignancy. In general, rapidly proliferating tissues rely more on glycolysis to produce ATP, while differentiating tissues mainly rely on OXPHOS to produce energy (Berridge et al., 2010). Pfeiffer et al. hypothesized that a high-rate but low-yield ATP production pathway, glycolysis, would have a selective advantage in the competition for shared energy (Pfeiffer et al., 2001).
Glycolysis is known to be much less efficient in providing energy than OXPHOS because only 2 ATP molecules are produced per glucose molecule through glycolysis (Martínez-Reyes and Chandel, 2021). On the other hand, the TCA cycle in mitochondria typically produces about 36 ATP molecules for each glucose molecule. In cancer cells, this inefficiency of ATP production by aerobic glycolysis appears to be offset by the high rate of glycolysis, which is 100-fold faster than OXPHOS due to the much shorter reaction pathway. However, these data are not entirely accurate. Since not all glucose metabolized by the glycolytic pathway will actually produce lactate molecules, a fraction of glucose may be drained into the branching reaction. Thus, these figures still need to be revised further.

Tumor cells generate lipids, proteins, and nucleic acids to promote proliferation through the branching pathway of aerobic glycolysis
To explain the downregulation of OXPHOS in tumor cells, Warburg hypothesized that there are mitochondrial defects in tumor cells that result in impaired aerobic respiration and subsequent dependence on glycolytic metabolism. However, it was subsequently found that the mitochondrial OXPHOS function is intact in the majority of cancers. Moreover, ATP produced by glycolysis is sufficient for cancer growth. For cancer cells with relatively long doubling times, ATP may only be needed to maintain the cell, rather than to support proliferation. This suggests that the Warburg effect has other contributions to cancer cell metabolism besides rapid ATP production (Graziano et al., 2017).
For example, glycolytic products are used through branching reactions as a carbon source for anabolic processes and for de novo construction of nucleotides, lipids, and proteins, and can be redirected to several pathways generated by glycolysis. To undergo rapid replicative division, cancer cells must replicate their genome, synthesize proteins and lipids, and assemble these components to form new cells. Rate-limiting enzymes in the glycolytic branching pathway are frequently upregulated in tumors. Many glycolytic intermediates can be transferred to branched pathways to produce different biosynthetic precursors, such as N-acetylglucosamine, serine, and glycine, to support the biosynthesis of macromolecules that determine cancer cell proliferation (Lunt and Vander Heiden, 2011;Vander Heiden et al., 2009).
The most important of the branching pathways is the PPP. The pathway's key enzymes, transketolase-like 1 and transaldolase, are overexpressed in cancer cells in order to produce NADPH and ribose-5-phosphate (Pavlova and Thompson, 2016). NADPH and 5-phosphate ribose are synthetic precursors of amino acids and nucleotides, which provide substrates for tumor growth and proliferation. In addition, NADPH plays an important role in protecting cancer cells against antineoplastic drugs by maintaining sufficient levels of reduced glutathione, which maintains the redox state and protects cells from reactive oxygen species (ROS) (Pavlova and Thompson, 2016;Traverso et al., 2013). When the PPP intermediates, such as fructose 6-phosphate and glyceraldehyde 3-phosphate, are not required for metabolism, they can be recycled back into glycolysis to produce pyruvate and lactate.

Lactate, the product of aerobic glycolysis, causes extracellular acidosis and makes tumor cells more aggressive
Another benefit of the Warburg effect on cancer progression is the reverse gradient of intracellular and extracellular pH and acidification of the TME. Alkalosis within cancer cells leads to cell proliferation, represents an antiapoptotic defense mechanism in cancer cells, and is important in driving the increased dependence on glycolysis and the decreased dependence on OXPHOS (Porporato et al., 2011). In contrast, extracellular acidosis is strongly associated with invasion, angiogenesis, metastasis, and resistance to therapy.
Among the TME metabolites, lactate is one of the most important players in almost all types of cancer. The Warburg effect, which causes tumor cells to preferentially convert pyruvate to lactate rather than enter the TCA cycle, may be due to the fact that NADH and ATP produced by the TCA cycle are major negative regulators of glucose metabolism. By converting excess pyruvate to lactate, negative feedback can be avoided, and glucose metabolism in proliferating cells can be enhanced. Excessive secretion of lactate by cancer cells in the extracellular environment results in lactate concentrations as high as 40 mM measured in the serum of patients with different cancers, whereas lactate concentrations in healthy tissues and serum range from 1.5 to 3 mM (Walenta et al., 2000). High lactate levels are generally associated with a worse prognosis. Hirschhaeuser found that intratumoral lactate levels were inversely correlated with overall and disease-free survival, and tumor irradiation triggered a rapid and transient decrease in lactate levels (Brizel et al., 2001;Hirschhaeuser et al., 2011). In addition, lactate has many other effects. Lactate can upregulate vascular endothelial growth factor expression, thereby driving angiogenesis. Lactate can also strongly inhibit T cell activation and, thus, play a role in immune escape (Fischer et al., 2007).

Glycolytic enzymes enhance the resistance of tumor cells to apoptosis
Hexokinase (HK) is a phosphorylase of 6-carbon sugars and is considered to be the "starting point" of glycolysis and the PPP (Ge et al., 2020). This enzyme is widely distributed, and 4 isozymes exist in mammals (Sun et al., 2008). It has been shown that tumor cells have elevated levels of HK-I and HK-II, which have an inhibitory effect on apoptosis in tumors (Pastorino and Hoek, 2003). Isoforms HK-I and HK-II bind to the outer mitochondrial membrane through the N-terminal part of voltage-dependent anion channel 1 (VDAC1), which enhances glycolysis and regulates cellular energy metabolism through phosphorylation of glucose, and the substrate ATP produced by OXPHOS in mitochondria. An additional function of HK binding to VDAC1 is to prevent VDAC1 from forming the mitochondrial permeability transition pore (mPTP), thereby inhibiting cell apoptosis (Abu-Hamad et al., 2008;Shoshan-Barmatz et al., 2017). mPTP is a multiprotein complex composed of cyclophilin D in the mitochondrial matrix, adenine nucleotide transporter in the inner membrane, and VDAC1 in the outer membrane. mPTP formation causes matrix swelling and MOM rupture, releasing intermembrane space proteins, such as cytochrome c. Cytochrome c, by binding to the central regulator of apoptosis Apaf-1, leads to the formation of apoptotic bodies with subsequent recruitment and activation of caspase-9. Caspase-9 cleaves and activates caspase-3 and caspase-7, which regulate apoptosis by cleaving key substrates (Taylor et al., 2008). Various cellular stress conditions, including Ca 2+ overload, phosphate concentration, oxidative stress, and reduced ATP availability, can lead to mPTP formation (Baumgartner et al., 2009;Seidlmayer et al., 2015).
The 3D structure of VDAC1 reveals that it consists of a β-barrel consisting of 19 transmembrane β-chains connected by flexible loops and a 25-residue-long N-terminal domain containing an α helix within the pore (Hiller et al., 2010). This region could regulate the transport of substances through VDAC1 channels (Ujwal et al., 2008), as it is voltage gated. The diameter of the VDAC1 channel pore ranges from 2.6 to 3.0 nm under normal conditions but can be reduced to approximately 1.5 nm when the VDAC1 N-terminus is located within the pore (Bayrhuber et al., 2008). Passage of cytochrome c requires channels formed by a minimum number of 6 VDAC1 monomers (Hiller et al., 2008). When they oligomerize, they form large channels that allow the release of apoptotic proteins. It was estimated that a minimum number of 6 VDAC1 monomers arranged in a circular shape would create a central pore diameter of 4.0 nm and allow the transport of cytochrome c, which has 3.4 nm outer diameter.
Recent models suggest that the structure of VDAC1 is in a dynamic equilibrium between monomeric and oligomeric states; therefore, normally expressed VDAC1 can directly induce apoptosis (Shoshan-Barmatz et al., 2010). However, compared with normal cell lines, many human cancer cell lines have increased expression of VDAC1 and do not undergo apoptosis, reflecting the antiapoptotic contribution of HK-I and HK-II by binding to VDAC1. Notably, the HK reaction product glucose 6-phosphate (G6P) can mediate the dissociation of HK-I from VDAC, leading to the activation of apoptosis. However, this inhibition was almost completely ineffective against HK-II (Azoulay- Zohar et al., 2004). In addition, tumor cells can produce large amounts of ATP through glycolysis under the Warburg effect, thereby avoiding ADP accumulation and the conversion of ADP to adenine monophosphate (AMP) and phosphate (the promoter of mPTP) (Bonora and Pinton, 2014). The binding of HK-1 and HK-2 to VDAC proteins was also related to the alkalosis of tumor cells. Alkaline pH promoted the coimmunoprecipitation of HK and VDAC proteins, which further emphasized the role of acid-base balances in cancer (Björkholm and Monteggia, 2016).
In hypoxic tumor cells, VDAC1 is cleaved to form (VDAC1-ΔC), a novel form produced by C-terminal truncation of the protein that is dependent on TP53 or p73 (Brahimi-Horn et al., 2012). Cells expressing VDAC1-ΔC have increased mitochondrial size and are reprogrammed to utilize more metabolites. It is also associated with the upregulation of glycolysis and mitochondrial respiration, which enhances cell resistance to apoptosis and facilitates cell growth in the hypoxic microenvironment (Mazure, 2016).

Lactate shuttle by stromal cells constitutes the reverse-Warburg effect in tumors
The Warburg effect is most likely due to the adaptive metabolic reprogramming of cells to utilize all available glucose as a result of decreased oxygen levels. That is, cells in the hypoxic zone take up glucose in large amounts for glycolysis, while more actively proliferating cells in the surrounding area may use and oxidize the lactate secreted by the hypoxic zone (Sonveaux et al., 2008). Because of the heterogeneity in the origin and differentiation of tumor cells, the heterogeneity in the degree of hypoxia within the tumor, and the constant changes in the TME, the metabolic phenotypes of cancer cells can and do vary even within the same tumor mass (Lee and Yoon, 2015). Metabolic abnormalities and heterogeneity also occur in the TME, especially cancer-associated fibroblasts (CAFs).
Tumors are "fibrotic wounds that never heal," and the CAFs are a major source of collagen (Piersma et al., 2020). Recent studies have suggested that there is a wide range of molecular crosstalk between CAFs and tumor cells, which can enhance cell migration and alter tumor cell metabolism. In addition, the TME also establishes a cycle of metabolism-related functions between tumor cells and CAFs. The unexpectedly slow proliferation rate of CAFs compared to the proliferation rate of cancer cells and matched NFs suggests that the increased glycolysis in CAFs is not used for the proliferation of CAFs themselves but rather provides metabolic support for tumor cells. Tumor cells hijack CAFs and reprogram their metabolism, inducting a dramatic shift from OXPHOS to aerobic glycolysis, which is known as the reverse-Warburg effect (Martinez-Outschoorn et al., 2011).
CAFs secrete and shuttle metabolites such as pyruvate and lactate to tumor cells during aerobic glycolysis, and this lactate shuttle also occurs in brain tissue (Li et al., 2021;Najafi et al., 2019). Lactate is transported in and out of cells through a family of MCTs with different isoforms. MCT1 (for lactate uptake) and MCT4 (for lactate excretion) have been shown to be upregulated in a variety of cancers and are associated with poor prognosis and high mortality. When lactate is secreted by MCT4 in CAF, it is subsequently taken up by MCT1 in adjacent tumor cells and used as its alternative carbon source. In addition, CAFs also produce glutamine and other amino acids, which are transported to tumor cells to provide nitrogen nutrients (Fig. 6).

Sporadic AD is a metabolic disease characterized by an agerelated energy deficit in neuronal mitochondrial activity (reverse-Warburg effect)
The complex brain bioenergetic deficits in SAD and their association with age suggest that AD may be a bioenergetic disease (Ding et al., 2013;Gil-Iturbe et al., 2020). This may be due to a variety of causes, including the decrease in CBF, the significant reduction of GLUTs, the decrease in insulin and insulin receptor levels in the brain, mitochondrial damage, and changes in the expression of enzymes involved in OXPHOS and glycolysis (Talbot et al., 2012). Here, we focus on degenerative changes in the metabolic machinery of neurons and astrocytes that link excitatory neural energy coupling, aerobic glycolysis and OXPHOS, lactate uptake and release, and cellular resistance to oxidative stress.
With aging, mitochondria respond through 2 types of mitochondrial metabolic reprogramming: (1) the reverse-Warburg effect, where an additional energy source is derived from lactate produced by astrocytes, leading to upregulation of some mitochondrial OXPHOS activity in cortical neuronal cells and (2) the Warburg effect, where upregulation of glycolysis in astrocytes, as indicated by astrocytes expressing high levels of the glycolysis-promoting enzyme PFKFB3, utilizes most of the glucose in the brain. Lactate, the product of glycolysis, is released into the extracellular environment and used to replenish the energy substrates required by neurons, in a pattern similar to the intracellular utilization of anaerobic glycolysis by peripheral tissues (Pellerin et al., 2007).
This mitochondrial metabolic reprogramming is an attempt to maintain the integrity of neurons under stress and an explanation for the selective vulnerability of neurons in AD (Kapogiannis and Mattson, 2011;Swerdlow et al., 2010). The entropic selection principle states that the outcome of substrate competition between neuronal populations depends on the neuronal environment and the metabolic rate of the population. When neuronal substrates are scarce, neurons with higher metabolic rates have a selective advantage, but when neuronal substrates are abundant, neurons with a lower metabolic rate have the advantage. Under resource-limited conditions, cells that do not undergo upregulated OXPHOS will eventually die due to the inability of neurons to compete for energy substrates and produce enough energy, while damaged cells acquire a dominant competitive phenotype through metabolic reprogramming, which, in turn, leads to the spread of disease. SAD targets specific neural networks with high metabolic demands, including the hippocampus, posterior cingulate cortex, medial frontal cortex, and lateral frontal, temporal, and parietal regions. Bioenergetic failure of neurons in these regions may be the main driver of SAD (Watanabe et al., 2021).

The brain supports neuronal development through aerobic glycolysis in the young age and protects against Aβ damage in the old age
OXPHOS and glycolysis are the 2 main pathways by which cells rely on glucose for energy production. When the rates of glycolysis and OXPHOS are fully coupled, glucose is almost completely oxidized to pyruvate and CO 2 . Glycolytic and OXPHOS rates are variable and imbalanced throughout the stages of the human life cycle. The childhood brain is predominantly glycolytic, consuming large amounts of glucose to meet developmental needs. The overall metabolic decline in old age is accompanied by a loss of aerobic glycolysis (Erecinska et al., 2004;Goyal et al., 2017). Glucose metabolism generates more than 15 times as much energy as aerobic glycolysis through OXPHOS. However, the purpose of glycolysis is not just to produce energy. Various metabolic intermediates produced in the glycolytic pathway will enter a range of biosynthetic processes, including the TCA cycle, gluconeogenesis, PPP, lipid metabolism, and other anabolic pathways that contribute to biomass production. Glycolytic enzymes are highly expressed in the postsynaptic density and regulate postsynaptic transmission (Bas-Orth et al., 2017). In addition, the PPP can also coordinate with glycolysis to play an important role in maintaining cellular redox state and preventing reactive oxygen species generation. In the adult brain, glucose consumption exceeds oxygen consumption by about 10%-12%. Aerobic glycolysis, responsible for this excess glucose consumption, is mainly present in areas such as the medial frontal gyrus, precuneus, and posterior cingulate cortex, and reflects the support for synapse and axon formation and turnover in the brain throughout life (Goyal et al., 2014). However, local aerobic glycolysis decreases significantly with aging in adults and may be associated with the development of later AD (Goyal et al., 2017;Vlassenko et al., 2010). The brain is thought to be predominantly glycolytic in early life. The newborn brain accounts for about 13% of body weight but consumes up to 60% of total body energy. Aerobic glycolysis accounts for about 35% of neonatal brain glucose consumption (Settergren et al., 1976). Elevated levels of glycolysis during childhood were associated with the highest synaptic growth rates in children. This energy utilization continues throughout childhood to support maturation changes in neurons, including axon elongation, synaptogenesis, and myelination. In the adult brain, the area of aerobic glycolysis was narrowed by about 10% and was highest in the default mode network (Settergren et al., 1976;Vlassenko and Raichle, 2015). Regional high aerobic glycolysis may reflect ongoing developmental processes in order to generate synaptic changes associated with learning and memory (Bauernfeind et al., 2014). However, the default mode network is also the first functional brain network to be identified as affected by AD (Greicius et al., 2004). The default mode network is responsible for the processes associated with episodic memory and self-reference, which coincides with the selective vulnerability of neurons, which is a hallmark of AD. One of the characteristics of AD is the massive loss of neurons in major neuronal circuits, which distinguishes it from other pathological processes that can lead to dementia (Silverman et al., 2001).
Interestingly, regions of the adult brain, which retain high levels of glycolysis, are also the most sensitive to AD, proving that these regions are more susceptible to energy depletion (Goyal et al., 2014). A correlation between lower aerobic glycolytic regions and higher tau deposition was also observed in preclinical and very mild symptomatic AD individuals, independent of age (Vlassenko et al., 2018). This may represent a toxic effect of tau on glycolysis in the aging brain or deprivation of synaptic maintenance associated with aerobic glycolysis that may somehow contribute to the acceleration of neurodegeneration, including tau pathology. Recently, Zhiqiang Qiu et al. developed a glycolysis-brain cell marker connectivity network in an attempt to delineate the extensive interactions of glycolysis-related genes with the brain microenvironment in AD and verified the predictive performance of the glycolysis index in normal populations (Qiu et al., 2022).
Some individuals who exhibit high plaque accumulation at autopsy have not had cognitive impairment, and one explanation for this is that these individuals may acquire and exhibit mechanisms of resistance to the toxic effects of Aβ. That is, in the initial stage of AD, Aβ-resistant cells increase the expression of lactate dehydrogenase A or pyruvate dehydrogenase kinase 1 by increasing HIF-1 activity, increase glucose uptake and glycolysis, and become resistant to the toxicity of the Aβ peptide, thereby alleviating the progression of AD (Soucek et al., 2003). This seems to mimic the Warburg effect, but the imitation is botched; as in late AD, PDK1 decreases, eliminating the protection from the Warburg effect.

Astroglia-neuron lactate shuttle hypothesis forms the structural basis of the reverse-Warburg effect
In the human cerebral cortex, astrocytes are significantly more abundant and have significantly higher glucose uptake than neurons, yet neurons account for 70%-80% of the total energy expenditure of the brain, while glial cells consume the remaining fraction (20%-30%) (Bass et al., 1971;Nedergaard et al., 2003). This suggests that brain energy metabolism is a compartmental process, with different anatomical structures and metabolic characteristics in different parts, and different cells respond to energy demands through complementary metabolic coupling, which can be described by the currently proposed astroglia-neuron lactate shuttle hypothesis: astrocyte responses to neural excitation are achieved by uptake of glutamate at synapses to trigger aerobic glycolysis and subsequent lactate secretion. Neurons then take up lactate and preferentially use it in neural activity to generate energy (Fig. 7).
Astrocytes have special processes called perivascular endfeet, which contact vascular capillaries and almost completely cover their surface, and both endothelial cells and astrocytes express the glucose transporter GLUT1 on their surface, making the endfeet a major site for glucose entry into the brain (Kacem et al., 1998). Another part of the astrocyte process wraps around the synapse, forming the communication interface between neurons and astrocytes. These 2 specialized processes of astrocytes allow them to obtain large amounts of glucose, which are disproportionate to their energy requirements and serve as key regulators of connective synaptic activity, local blood flow regulation, and brain energy metabolism (Gordon et al., 2008).
Glutamate is an essential excitatory neurotransmitter that is released into the synaptic cleft during synaptic activity. Astrocytes trigger aerobic glycolysis of astrocytes through glutamate uptake in order to produce lactate. Subsequently, glutamate is taken up by astrocytes and enters the glutamate-glutamine cycle, and lactate is released into the extracellular space by MCT (Bak et al., 2006;Dienel, 2019). Lactic acid then enters neurons as an energy substrate and is converted into pyruvate and acetyl-CoA; the latter is used in the TCA cycle to generate NADH, which promotes OXPHOS through the mitochondrial electron transport chain. It has been observed that there is a gradient of lactate concentration between astrocytes and neurons that favor lactate efflux from astrocytes and inflow into neurons (Mächler et al., 2016). This phenomenon has also been computationally validated by the model (Coggan et al., 2018;Mächler et al., 2016).
Theoretically, both astrocytes and neurons are able to completely oxidize glucose and lactate (Zielke et al., 2009). The underlying mechanism for their differential preferential utilization of energy substrates lies in the differential expression of glycolysis-promoting enzymes. PFKFB3 is a glycolysis-promoting enzyme expressed at a low level in neurons. PFKFB3 in neurons is continuously degraded by the proteasome under the action of the E3 ubiquitin ligase APC/C-Cdh1. This makes neurons extremely sensitive to energy depletion and degeneration (Almeida et al., 2004). Astrocytes express high levels of PFKFB3 and exhibit low APC/C-Cdh1 activity, which results in high astrocyte glycolytic rates. Furthermore, the expression of LDH isoforms in neurons and astrocytes leads to different conversion tendencies; Fig. 7. Warburg and reverse-Warburg effects in AD. The neurovascular unit in the brain determines the existence of energy coupling between nerve cells. Astrocytes mainly produce lactate through glycolysis and deliver it to neurons (Warburg effect), and neurons mainly use lactate as an energy substrate for oxidative phosphorylation (reverse-Warburg effect). Abbreviations: AD, Alzheimer's disease; GLUT1, glucose transporter-1; MCT, monocarboxylate transporter; TCA, tricarboxylic acid. LDH1 is preferentially expressed in neurons (favoring lactate to pyruvate conversion), and LDH5 is preferentially expressed in astrocytes (favoring pyruvate to lactate conversion) (Bittar et al., 1996).
Hypoxia can also induce metabolic regulation and reprogramming in some neurons, thus directly upregulating glucose uptake in response to energy crisis. However, the effect of such reprogramming is limited because neurons lack direct contact with blood vessels. Moreover, in neurons, glucose metabolism mainly targets the oxidative branch of the PPP as an antioxidant strategy to produce reducing equivalents necessary to maintain the protective antioxidant state of neurons. If the inhibitory effect of PFKFB3 is removed, the enhanced glycolysis in neurons may aggravate cell damage through oxidative stress and neuronal apoptosis since glucose utilization in the PPP is significantly reduced (Herrero-Mendez et al., 2009).

Lactate, the product of glycolysis, causes intracellular acidosis and promotes apoptosis in AD
In the context of the astroglia-neuron lactate shuttle, it has been found that astrocytes are more likely than neurons to respond to hypoxic stimuli that promote glycolysis, with the production of lactate, but not pyruvate, which continues to be transported to neurons and acts as an alternative energy substrate (Marrif and Juurlink, 1999;Pellerin et al., 1997).
Recently, the understanding of lactate has been transformed from it being a metabolic waste product in the brain to being a dominant energy source in the context of neuronal damage (Pellerin et al., 1997). Boumezbeur et al. demonstrated by 13 C magnetic resonance spectroscopy that the brain can support up to 10% lactate metabolism at physiological levels and up to 60% lactate metabolism at supraphysiological levels (Boumezbeur et al., 2010). Ketone bodies produced in the liver and lactate produced by exercising skeletal muscle can also be important energy substrates for the brain when glucose levels are limited, and lactate produced by exercise has been shown to be taken up by the brain and improves memory function (Roig et al., 2012). In addition to being an energy substrate, lactate appears to be involved in many other neuroprotective mechanisms (Cunnane et al., 2020). Brain-derived neurotrophic factor (BDNF) is an important neurotrophic factor in the brain, which promotes neuronal differentiation, nerve regeneration, and synaptic plasticity, and enhances memory (Björkholm and Monteggia, 2016). A decrease in BDNF in the brain, especially the hippocampus, and downregulation of BDNF mRNA are the hallmarks of AD. Lactate infusion can alleviate ischemic injury and increase BDNF expression in a mouse model (Berthet et al., 2012;Coco et al., 2013).
Lactate levels increase with age, and elevated CSF lactate levels have also been observed in AD patients (Redjems-Bennani et al., 1998). Notably, while lactate production is beneficial for the healthy aging brain and can improve memory, it may be detrimental to AD. A recent study by Harris et al. suggested that elevated aerobic glycolysis may be detrimental to AD-related cognitive decline in later stages of the disease due to inadequate lactate handling. This inadequate handling may arise from increased neuronal uptake resulting from increased lactate secretion by glial cells, which leads to pronounced cytoplasmic acidification (Harris et al., 2016;Schwartz et al., 2020). pH sensitivity is a characteristic of many pairs of nerve cell membrane proteins, which regulate a series of physiological functions, such as nerve excitability, synaptic transmission, and neurotransmitter uptake. Cells must maintain intracellular pH (pHi) within the physiological range. The pHi of normal cells fluctuates between 6.8 and 7.3. Alkaline pHi is a characteristic of tumor cells, triggering DNA breakdown, replication, and cell division (Aerts et al., 1985;da Veiga Moreira et al., 2015). pHi is acidic in AD neurons, caused by excessive secretion of lactate by glial cells, and this plays a crucial role in determining the hallmarks of AD (Fang et al., 2010), such as decreased glucose uptake, apoptosis, amyloid plaque deposition, and tau phosphorylation, ultimately leading to cognitive decline (Basurto-Islas et al., 2013).

Glycolytic enzyme HK is involved in the regulation of cell apoptosis
Mitochondria play a key role in cellular functions, including energy production, cell proliferation, and apoptosis (Birch-Machin, 2006). In view of the metabolic dysfunction in AD, mitochondrial defects are closely related to both the pathogenesis and progression of AD (Swerdlow et al., 2014). Mitochondria contain approximately 1000 different proteins with tissue-specific features. One of the key proteins regulating mitochondrial function is VDAC1, which is the most abundant protein in the mitochondrial outer membrane. It connects the energy, redox, and signaling pathways between mitochondria and other cellular compartments.
VDAC1 has the property of voltage gating, and being in an open or closed conformation has a major impact on mitochondrial metabolism and cellular bioenergetics. VDAC1 has a symmetric bell-shaped conductance that is elevated at low potentials from −20 to +20 mV, with open channels that are permeable to organic anions, including respiratory substrates, ATP, ADP, phosphate, and metabolites. At higher positive and negative potentials (> 30-60 mV), the conductance decreases, and the channel closes, allowing the selective transfer of small cations, for example, K + , Na + , and Ca 2+ (Rostovtseva and Colombini, 1996). VDAC1 can interact with more than 150 other proteins, including phosphorylated tau, Aβ, and γ-secretase, and is involved in their toxicity. In addition to this, VDAC1 plays a key role in mitochondriamediated apoptosis by releasing apoptotic proteins located in the intermembrane space and binding to proapoptotic and antiapoptotic proteins. Thus, VDAC1 has been proposed as a promising target for the control of apoptosis.
It is known that overexpression of VDAC1 triggers apoptosis through its oligomerization, and this can be prevented by HKs. HKs competitively inhibited the binding of Bax to VDAC1, thereby reducing the formation of the VDAC1-Bax complex. In contrast, detachment of HKs from VDAC1 induces apoptosis, increasing VDAC1 binding to Bax, or the formation of heterologous or homologous oligomers.
VDAC1 is overexpressed in the postmortem brains of AD patients and in the brains of APP transgenic mice, and VDAC1 expression increases with age (Cuadrado-Tejedor et al., 2011;Manczak and Reddy, 2012;Pérez-Gracia et al., 2008). However, in AD, the expression level of VDAC1 was increased, but the interaction of VDAC1 with glycolytic enzymes, such as HKs, was reduced, which may be induced by Aβ (Magrì et al., 2018;Smilansky et al., 2015). When dissociated from VDAC1, HK loses access to mitochondrial ATP, thereby reducing the supply of ATP for glycolysis and glucose metabolism and making the cell susceptible to apoptosis. HK dissociation from VDAC also activates signaling by the NLRP3 inflammasome and may contribute to neuroinflammation in AD.
In addition, a set of metabolomics data analyses demonstrated the accumulation of G6P in humans and mice with AD (Demarest et al., 2020). The accumulation of G6P can reduce HK activity by competitively inhibiting ATP binding to the HK active site. This observation echoes another study that showed that glucose metabolism was enhanced during the early stages of apoptosis and VDAC1 bound to HKs to mediate cell resistance to injury. However, in the late stage of apoptosis, the accumulation of G6P disrupted the interaction between VDAC1 and HK, which led to apoptosis (Bobba et al., 2015).

Short concluding remarks
In this article, we discuss the mechanisms by which hypoxia regulates different cell fates in AD and cancer. The difference between the 2 is not only "too young cells" and "too old cells" but also a series of parallel changes, but, due to the characteristics of nerve cells and cancer cells, these parallel mechanisms proceed in opposite directions. Moreover, we have observed "clumsy mimicry" of cancer cell metabolism in AD, which may be attributed to the nonregenerative nature of neuronal cells, with the accumulation of G6P in neurons preventing attempts to upregulate glycolysis against apoptosis. Due to the demand for substrates for cell proliferation, cancer cells open a large number of glycolytic branch pathways so that the reaction can be sustained. In addition, the brain has elaborate NVUs that regulate metabolism in an orderly manner under physiological conditions through the cooperation of neurons and astrocytes. In contrast, in AD brains, we find that this impairs the ability of neurons to take up glucose and upregulate glycolysis. However, in the disordered blood vessels of cancers, cells can unimpededly upregulate glucose uptake and glycolysis. In addition, we wondered whether there might also be an inverse relationship between vascular dementia and cancer, as a recent longitudinal cohort study supported an inverse relationship between cancer and vascular dementia, whereas other studies have suggested no such relationship (Kinnunen, 2010;Zhang et al., 2022). Given that most patients with vascular dementia have mixed forms (AD + vascular) and vascular dementia is divided into 4 major subtypes (subcortical ischemic vascular dementia, cortical dementia, poststroke dementia, and mixed dementia), it is uncertain whether there is a relationship between cancer, and vascular dementia and its different subtypes. Arteriosclerosis, lacunar infarction, and small vessel disease can all be risk factors for vascular dementia, and the same changes may also occur in AD (Kalaria, 2016). We speculate that the same vascular factors may be present in different types of dementia and contribute to the opposite relationship with cancer. In this article, we considered hypoxia as a crossroads between AD and cancer although the crossroads are complicated, and they need to be better understood in order to develop new therapies against both diseases.

Disclosure statement
The authors declare no conflict of interest.

Author Contributions
Drafting the manuscript: ZS; Reviewing the literature and revising the manuscript: GZ and XL; Designing and supervising the study: HZ. All authors read and approved the final manuscript.