Adiposity and dyslipidaemia

In diet-induced obese animals, increased body adiposity, attributed mostly to mesenteric and epididymal adipose tissues, may result in poorer performance on cognitive tests. In addition to fat storage, adipose tissue has a multifunctional characteristic defined by its endocrine activity. It communicates with other organs, including the brain, through the expression of signalling molecules (adipokines) and, thus, exerts systemic immunometabolic functions. However, it is postulated that excessive energy intake enables changes in adipose tissue and promotes the transition to a dysfunctional phenotype and, consequently, systemic disorders^{(Reference Kang, Wang and Oteiza19,Reference Tan and Norhaizan20)} .

Also, as a consequence of excessive body adiposity, cardiovascular risk factors emerge, which simultaneously pose risk factors for cognitive impairment and dementia: dyslipidaemia and hypertension. Dyslipidaemia and hypertension may induce endothelial dysfunction and cause cerebral hypoperfusion. As a result of impaired cerebral blood flow, there is a reduced supply of oxygen and nutrients, in addition to diminished removal of the products of cerebral metabolism and neurotoxic agents, both essential for brain health⁽⁶⁾.

Other hypotheses support a link between peripheral lipid metabolism and cognitive impairment. Exaggerated production of lipid metabolites, such as diacylglycerol and ceramide, can occur when lipids are redirected to other tissues under the saturation of adipose tissue storage capacity^{(Reference Cusi21)}. In mice fed a high-fat diet, increased hepatic ceramides were correlated with higher hippocampal ceramides levels and worse spatial memory retention. These compounds are lipid soluble and can readily cross the BBB. In the brain, they may exert central cytotoxicity, mediate neuroinflammation and impair insulin signalling, thus mediating cognitive dysfunctions^{(Reference Paraiso, Revel and Choi22)}.

From a pharmacological approach, the use of lipid-lowering drugs (that is, statins) has been shown to optimise the control of vascular risk factors, reversing cognitive deficits caused by high-fat diet (HFD) in animals^{(Reference Grundy, Stone and Bailey23,Reference Liu, Yang and Wang24)} . Alternatively, the supply of flavonoid-rich sources in similar animal models has demonstrated promising results. Dyslipidaemia biomarkers, such as triacylglycerol (TG), total cholesterol (T-chol), low-density lipoprotein-cholesterol (LDL-c) and high-density lipoprotein-cholesterol (HDL-c), showed improvement in circulating levels when obese animals received treatment with flavonoids. It was also possible to observe an improvement in the levels of atherogenic index (AI) and cardiac risk index (CRI)^{(Reference Metwally, Rashad, Mahmoud and Morus alba25)} and systolic blood pressure in diet-induced obese animals after treatment with flavonoid sources^{(Reference Micioni Di Bonaventura, Martinelli and Moruzzi7)}. From a central perspective, the anti-hyperlipidaemic effects of flavonoid intake were associated with lipid changes in the brain. In particular, xanthohumol, a flavonoid from the class of chalcones, controlled ceramide effects on the hippocampus by the hepatic–brain axis, which was correlated with improved spatial memory in high-fat-fed mice^{(Reference Paraiso, Revel and Choi22)}.

Furthermore, the significant reductions in body weight, body adiposity, visceral adiposity (epididymal, retroperitoneal and mesenteric adipose tissues) and serum levels of adipokines (leptin, resistin and adiponectin) and the browning of adipose tissue in response to ingestion of flavonoid sources suggest an improvement in obesity-associated immunometabolic disorders, which include cognitive impairment^{(Reference Kang, Wang and Oteiza19,Reference Metwally, Rashad, Mahmoud and Morus alba25–Reference Zhuang, Lu and Wang31)}

Low-grade systemic inflammation and oxidative stress

Inflammation is implicated in the pathogenesis of chronic non-communicable diseases such as diabetes and neurodegenerative diseases. In diet-induced obesity rodent models, systemic inflammatory markers co-exist with diminished brain integrity and poorer performance on cognitive tests^{(Reference Chen, Deng and Liu32,Reference López, Sánchez and Perez-Cruz33)} . Evidence indicates that circulating inflammatory molecules may influence brain health directly, when overcoming the BBB, or indirectly, inducing insulin resistance (IR) and oxidative stress in the peripheral organs, for example^{(Reference Batista, Soares and Mendonça9,Reference Mulati, Ma and Zhang34)} .

By far the greatest challenge of preventing and treating chronic low-grade systemic inflammation related to obesity is that it can be a result of numerous events that occur in isolation or parallel. Some of them can be cited:

1) Adipose tissue naturally produces pro-inflammatory molecules, such as interleukins (IL), tumour necrosis factor-alpha (TNF-α), leptin, resistin, retinol-binding protein 4 (RBP4), lipocalin 2, CC-chemokine ligand 2 (CCL2) and CXC-chemokine ligand 5 (CXCL5). Due to excessive storage of lipids, adipocytes increase in number (hyperplasia) and volume (hypertrophy), thus intensifying cytokine production. Furthermore, locally hindered tissue blood irrigation occurs because of adipocyte compression, culminating in oxygen depletion and hypoxia. This attracts immune system cells, such as macrophages, enhancing the generation of pro-inflammatory cytokines, which reach the bloodstream and other tissues, including the brain^{(Reference Metwally, Rashad, Mahmoud and Morus alba25,Reference Ouchi, Parker and Lugus35)} .

2) Oxidative stress has a bi-directional relationship with inflammation. The activated M1 macrophages due to the pro-inflammatory environment secrete reactive oxygen species (ROS) which, in turn, provoke pro-inflammatory cytokine release when activating the transcription factor nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB). In addition, ROS are the radical forms of oxygen and can cause damage to cellular proteins, lipids and nucleic acids that may impair cell signalling or lead to cell death. In obesity, imbalance between oxidant and antioxidant compounds in favour of the former, that is, oxidative stress, is frequently reported. Beyond macrophages, in obese mice oxidative stress can arise as a by-product of mitochondrial respiration due to excessive nutrients accumulated, or as a by-product of enzymatic oxidases (nicotinamide adenine dinucleotide phosphate (NADPH) oxidase and xanthine oxidase), and during advanced glycation end-product (AGE) generation^{(Reference Velayoudom-Cephise, Cano-Sanchez and Bercion36,Reference Zhou, Li and Xia37)} .

3) The gut dysbiosis and increased intestinal permeability in response to an obesogenic diet are associated with an increase in toxins in the bloodstream, mainly LPS. LPS is a main component of the Gram-negative bacterial cell wall which directly stimulates the immune system, triggers inflammation and leads to metabolic endotoxemia. LPS bind to the toll-like receptor (TLR)4 receptor complex expressed on cells, which increases the activity of the NF-κB pathway and secretion of pro-inflammatory cytokines (for example, TNF-α, IL-1β and IL-6)^{(Reference Kang, Wang and Oteiza19,Reference López, Sánchez and Perez-Cruz33)} .

4) Long-chain fatty acids have been reported to conduct pro-inflammatory cytokine syntheses through activating several molecular pathways, including mitogen-activated protein kinase (MAPK) (such as p38 and c-Jun NH2 terminal kinase (JNK)), NF-kB, activator protein (AP)-1, TLR4 and, thereby, TLR2^{(Reference Hwang, Kim and Lee38,Reference Kalivarathan, Chandrasekaran and Kalaivanan39)} . Furthermore, it has been suggested that a high-caloric diet and diet-induced hyperglycaemia can lead to endogenous production and accumulation of AGE and activation of its receptor (RAGE)^{(Reference Sampath, Rashid and Sang40)}. AGE/RAGE axis activation is closely related to inflammation, oxidative stress and mitochondrial dysfunction. AGE are formed by reactions between reducing sugars, such as glucose and fructose, and amino groups, and so they can result in irreversible morphofunctional changes of diverse biological structures^{(Reference Velayoudom-Cephise, Cano-Sanchez and Bercion36,Reference Kehm, Rückriemen and Weber41)} .

It is suggested that anti-inflammatory and antioxidant abilities in the peripheral perspective are mechanisms by which flavonoids offer indirect neuroprotection. As can be noted in Table 1, a diversity of flavonoid sources reduced pro-inflammatory biomarkers (TNF-α, IL-1β, IL-6 and monocyte chemoattractant protein-1 (MCP-1)) in blood, adipose tissue or liver^{(Reference Liu, Fu and Lan27,Reference Chen, Deng and Liu32,Reference López, Sánchez and Perez-Cruz33)} and improved endogenous antioxidant system activity (superoxide dismutase (SOD), catalase (CAT), glutathione peroxidases (GPx) and glutathione reductase (GR)) with a positive impact on spatial learning and memory ability^{(Reference Batista, Soares and Mendonça9,Reference Santos, Batista and Padilha Mendonça18,Reference Batista, Mendonça and Soares42,Reference Xia, Xie and Qiao43)} . The modulation of the immune system probably is through the down-regulation of the NF-κB pathway^{(Reference Kalivarathan, Chandrasekaran and Kalaivanan39,Reference Fu, Qin and Yu44)} , while improvement in the antioxidant system could be associated with increased up-regulation of nuclear factor E2-related factor (Nrf2) and, therefore, antioxidant enzyme expression^{(Reference Fu, Qin and Yu44)}. Redox homeostasis can also be a result of ROS scavenger activity of flavonoids. Moreover, modulating gut microbiota and controlling endotoxemia (LPS levels and its receptor TLR4 expression), as observed with the supplementation of genistein and (−)-epicatechin, are also proposed to indirectly mitigate neuronal disorders^{(Reference Kang, Wang and Oteiza19,Reference López, Sánchez and Perez-Cruz33)} .

Peripheral insulin response and diabetes

A growing body of population studies and meta-analyses demonstrate the adverse effects of diabetes mellitus on brain and cognition^{(Reference Biessels and Despa5,Reference Cheng, Huang and Deng45,Reference Xue, Xu and Ou46)} . Gudala et al. (2013) estimate an increased risk of 73% of all types of dementia, 56% of Alzheimer’s disease and 127% of vascular dementia for individuals with type 2 diabetes mellitus (T2DM)^{(Reference Gudala, Bansal and Schifano47)}.

One of the most well-known effects of impaired peripheral IR on brain health is its impact on the integrity and permeability of the BBB. This is related to sustained hyperglycaemia, AGE and oxidative stress exposure^{(Reference Rom, Zuluaga-Ramirez and Gajghate48)}. Moreover, an increased concentration of systemic saturated fatty acid caused by a HFD or obesity can enhance the NF-kB-mediated inflammation at the BBB via the TLR4 receptor^{(Reference Van Dyken and Lacoste49)}. Therefore, BBB dysfunction favours the influx of neurotoxic and pro-inflammatory agents from the periphery, eliciting neuroinflammation-mediated processes in the brain which result in neuronal damage and neurodegeneration. At the same time, it disfavours physiological processes through impaired BBB transport of molecules from the periphery to the brain (for example, insulin and leptin) and efflux of molecules from the brain to the periphery (for example, toxic Aβ peptides)^{(Reference Mainardi, Spinelli and Scala50,Reference Takata, Nakagawa and Matsumoto51)} . Other mechanisms that contribute to cognitive impairment and dementia in T2DM include cerebrovascular abnormalities once insulin acts as a vasoconstrictor at high concentrations^{(Reference Kellar and Craft52)}.

Obesity is a consistent risk factor for peripheral IR and, thereby, T2DM. In rodents, 8 or more weeks of obesogenic diet promote altered systemic markers of glucose metabolism and IR concurrently with worse performance in different cognitive tasks (Table 1). Thus, many pathophysiological fronts of obesity-induced peripheral IR are quite relevant as a strategy to collaborate with brain health and cognitive function.

The antidiabetogenic potential of flavonoids was demonstrated in diet-induced obese rodents based on the usual analysis of glucose homeostasis. Improvements in the glucose tolerance test (GTT), insulin tolerance test (ITT), glycaemic and insulin levels, and HOMA-IR index could explain, at least partially, the better cognitive performance or brain physiology of obese rodents (Table 1). However, it is important to note that the studies covered in this review focused mainly on the central IR (see ‘Brain insulin response’ section) to the detriment of peripheral IR, and therefore, limited mechanisms of systemic glucose homeostasis control were explored. Some authors highlighted the improvement of insulin and inflammatory sensitising adipokine levels in the blood of obese rodents after flavonoid-rich-diet intake: leptin, resistin and adiponectin by epigallocatechin gallate (EGCG)^{(Reference Mi, Qi and Fan28)}; resistin and adiponectin by luteolin^{(Reference Liu, Fu and Lan27)}; and adiponectin by jaboticaba peel^{(Reference Batista, Soares and Mendonça9)}. The ability to optimise the peripheral antioxidant status is also considered an important mechanism to preserve insulin cell signalling and glucose metabolism, as presented by jambo fruit, apigenin, quercetin and leaves of Morus alba L.^{(Reference Metwally, Rashad, Mahmoud and Morus alba25,Reference Kalivarathan, Chandrasekaran and Kalaivanan39,Reference Batista, Mendonça and Soares42,Reference Xia, Xie and Qiao53)} . In HFD-induced NAFLD model rats, quercetin prevented hepatocyte steatosis and metabolic dysfunction^{(Reference Gao, Chen and Fang54)}. Finally, circulating free fatty acids (FFA) have been decreased and glucagon-like peptide-1 (GLP-1) increased after naringin and apigenin consumption, respectively, in two different diet-induced obesity models^{(Reference Kalivarathan, Kalaivanan and Chandrasekaran55,Reference Wang and Chen56)} . Interestingly, in patients with Alzheimer’s disease, 6-month treatment with the GLP-1 analogue liraglutide could prevent the decline of brain glucose metabolism, proving that it also directly affects brain function^{(Reference Gejl, Gjedde and Egefjord57)}. Dysregulation of secretion and actions of the incretin has been observed in obesity and related metabolic disorders; therefore, it represents a new non-insulin therapeutic target for these pathological conditions. For more detailed flavonoids antidiabetogenic insights, see the work of Shahwan et al. (2022)^{(Reference Shahwan, Alhumaydhi and Ashraf58)}.

Neuroinflammation and oxidative stress

Neuroinflammation occurs through the activation of glial cells (astrocytes and microglia) in response to different insults and aims to protect the CNS from pathogens and to repair/restore nerve cells. However, under excessive or chronic activation, the glia inflammatory response can lead to neural damage and neurodegeneration by different mechanisms. In these cases, microglia and astrocytes, which represent most of the brain’s cellular identities, became non-negligible sources of neurotoxic compounds, such as pro-inflammatory cytokines, nitric oxide (NO•), Reactive Oxygen and Nitrogen Species (RONS) and prostaglandins^{(Reference Micioni Di Bonaventura, Martinelli and Moruzzi7,Reference Williams and Spencer12,Reference Calis, Mogulkoc and Baltaci13)} . In addition to the neuronal damage caused by the production of RONS by activated glial cells, it is imperative to consider that the brain is an organ that is very sensitive to oxidative stress owing to its high lipid content in cell membranes and its high oxygen consumption, associated with a low concentration of antioxidant enzymes^{(Reference Jha, Jha and Kumar59,Reference Maciejczyk, Żebrowska and Chabowski60)}

In animal models, activation of glia cells and neuroinflammation as a result of diet-induced obesity are considered key factors for cognitive function impairment (Figure 2)^{(Reference Micioni Di Bonaventura, Martinelli and Moruzzi7,Reference Qin, Sun and Mu30,Reference López, Sánchez and Perez-Cruz33,Reference Zhou and Wang61)} . Besides direct activation of glial cells, other mechanisms can lead to neuroinflammation in obesity: (1) activation of NF-κB and its translocation to the cell nucleus, where it mediates transcription of target genes, including pro-inflammatory cytokines and iNOS^{(Reference Li and Mi26,Reference Mi, Qi and Fan28)} ; (2) activation of NADPH oxidase (NOX), resulting in the production of superoxide radical (O2•⁻), which can react with NO• to generate peroxynitrite (ONOO⁻) that inhibits mitochondrial respiration and induces apoptosis^{(Reference Calis, Mogulkoc and Baltaci13,Reference Spencer, Vafeiadou and Williams62)} ; (3) reduced activity of antioxidant endogenous defences elicited by FFA in the brain, potentiating central oxidative damage^{(Reference Sripetchwandee, Chattipakorn and Chattipakorn63)}; (4) increased brain resistance to agents that regulate food intake, brain glucose homeostasis and neuroinflammation, including adipokines and hormones such as GLP-1, IGF-1, leptin and adiponectin^{(Reference Spielman, Little and Klegeris64)}; (5) cytokines and RONS produced peripherally, mainly in adipose tissue, that can cross the BBB together with other potential pro-neuroinflammatory agents, such as ceramides and FFA^{(Reference Spielman, Little and Klegeris64,Reference Calabrò, Rinaldi and Santoro65)} ; (6) other neurotoxic compounds not of peripheral origin, including Aβ peptides and phosphorylated Tau protein aggregates which are generated locally in mice obese brains^{(Reference Träger and Tabrizi66)}, along with those already mentioned, can activate neuronal death signalling cascades^{(Reference Spencer, Vafeiadou and Williams62)}.

Fig. 2.

Molecular modulation in the central nervous system by the ingestion of flavonoid sources in rodent models of diet-induced obesity to improve cognitive functions. AICD, amyloid-intracellular domain; AKT, protein kinase B; APP, amyloid precursor protein; Aβ, amyloid beta peptide; BACE1, β-site APP cleaving enzyme 1; BDNF, brain-derived neurotrophic factor; CAMK, calmodulin-dependent protein kinase; CAT, catalase; CREB, cAMP-response element binding protein; DDP4, dipeptidyl peptidase-4; ERK, extracellular signal-regulated kinase; GFAP, glial fibrillary acidic protein; GLP1, glucagon-like peptide-1; GLP1R, glucagon-like peptide-1 receptor; GLUT4, glucose transporter 4; GPX, glutathione peroxidases; GR, glutathione reductase; GSH, reduced glutathione; GSK3β, glycogen synthase kinase 3 beta; GTP, guanosine triphosphate; IBA1, ionised calcium-binding adapter molecule 1; IL, interleukin iNOS, nitric oxide synthase; IR, insulin receptor; IRS-1, insulin receptor substrate 1; JNK, c-Jun NH2 terminal kinase; MAPK, AMP-activated protein kinase; mBDNF, mature brain-derived neurotrophic factor; mTOR, mechanistic target of rapamycin; NeuN, neuronal nuclei; NO•, nitric oxide; NOX, NADPH oxidase; Nsf, N-ethylmaleimide–sensitive factor; NT, neurotrophin; O2•⁻, superoxide radical; ONOO⁻, peroxynitrite; PI3K, phosphoinositide 3-kinase; PKC, protein kinase C; PLCγ1, phospholipase Cγ-1; PSD-95, postsynaptic density protein 95; SOD, superoxide dismutase; Syn-1, synapsin-1; Syt-1, synaptatogmin-1; TAU, tubulin-associated unit; TNFR, tumour necrosis factor receptor; TNF-α, tumour necrosis factor-alpha; TRKB, tropomyosin receptor kinase B; Vamp1, vesicle-associated membrane protein 1.

Emerging evidence supports the use of immunotherapeutic agents in the management of neurodegenerative diseases. It aims to control the production of secreted and cell surface inflammatory mediators driving neuronal dysfunction and death^{(Reference Mortada, Farah and Nabha67)}. In diet-induced obesity models, flavonoids’ modulatory effects of neuroinflammation and cerebral oxidative stress are related to controlling the activation of glial cells, reducing the release of cytokines, inhibiting the induction of iNOS and the activation of NOX, and reducing the activity of pro-inflammatory transcription factors such as NF-κB and its downstream targets. Glial fibrillary acidic protein (GFAP) and ionised calcium-binding adapter molecule 1 (IBA-1) were used as markers of astrocyte and microglia activity, respectively, and although high-fat diets increased their expression, consumption of tart cherry, purple sweet potato colour, naringenin, epicatechin and genistein significantly reduced its expression in the hippocampus of animals receiving these treatments^{(Reference Micioni Di Bonaventura, Martinelli and Moruzzi7,Reference Kang, Wang and Oteiza19,Reference Qin, Sun and Mu30,Reference López, Sánchez and Perez-Cruz33)} Moreover, endothelial inflammatory markers such as intercellular adhesion molecule 1 (ICAM-1) and vascular cell adhesion molecule 1 (VCAM-1) were also reduced by the consumption of tart cherry juice^{(Reference Micioni Di Bonaventura, Martinelli and Moruzzi7)}

Flavonoids seem to be effective in modulating the expression of cytokines at the brain level, once IL-1β, IL-6 and TNF-α were down-regulated after intervention with different food sources and/or isolated forms of flavonoids^{(Reference Batista, Soares and Mendonça9,Reference Li and Mi26,Reference Liu, Fu and Lan27,Reference Mulati, Ma and Zhang34,Reference Fu, Qin and Yu44,Reference Zhou and Wang61)} . NF-κB activation in animal-fed obesogenic diets was inhibited by concomitant administration of flavonoids such as purple sweet potato colour, EGCG, luteolin, formononetin and sea-buckthorn flavonoids^{(Reference Li and Mi26–Reference Mi, Qi and Fan28,Reference Mulati, Ma and Zhang34,Reference Fu, Qin and Yu44)} . Expression of iNOS at the brain level was reduced in animals given purple sweet potato colour, EGCG, sea-buckthorn flavonoids and Cinnamomum cassia Presl flavonoids^{(Reference Li and Mi26,Reference Mi, Qi and Fan28,Reference Mulati, Ma and Zhang34,Reference Li, Zhou and Liu68)} ; while consumption of epicatechin inhibited the expression of NOX^{(Reference Kang, Wang and Oteiza19)}. Finally, the brain’s antioxidant endogenous defence has been consistently improved by treatment with a diversity of flavonoids sources, as demonstrated in several studies which found higher levels of SOD, CAT, GPx and reduced glutathione (GSH)^{(Reference Batista, Soares and Mendonça9,Reference Santos, Batista and Padilha Mendonça18,Reference Liu, Fu and Lan27,Reference Zhuang, Lu and Wang31,Reference Batista, Mendonça and Soares42,Reference Xia, Xie and Qiao53,Reference Zhou and Wang61,Reference Li, Zhou and Liu68)} , in agreement with the lower values of malondialdehyde (MDA), an end product of polyunsaturated fatty acid peroxidation^{(Reference Liu, Fu and Lan27,Reference Xia, Xie and Qiao53,Reference Zhou and Wang61,Reference Li, Zhou and Liu68)} .

Altogether, these findings suggest that dietary interventions with sources of flavonoids may represent a neuroprotective tool adjuvant for preventing or mitigating the damage caused by neuroinflammation associated with obesity.

Brain insulin response

Insulin originates almost exclusively from the pancreas and arrives at the brain parenchyma after crossing the BBB via transcytosis mediated by insulin receptors. Insulin receptors are also widely distributed in brain structures, most prominently in the hypothalamus, cerebellum, cortical regions, midbrain and hippocampus, where insulin may elicit specific responses^{(Reference Kullmann, Kleinridders and Small69)}. Distinct from peripheral tissues, insulin in the brain exerts minimal effects on glucose uptake, presenting many more insulin-independent glucose transporters (GLUT), mainly GLUT1 and GLUT3, than GLUT4^{(Reference Kyrtata, Emsley and Sparasci70)}. Now it is known that insulin in the brain plays a key role in the regulation of both metabolic and non-metabolic functions with systemic and central outcomes. It influences eating behaviours (food intake and satiety) and peripheral metabolism (energy balance) as well as neuronal survival, neuroplasticity, mood and cognition. Importantly, insulin also mediates many neuroprotective pathways, for example, reducing Aβ peptide and phosphorylated-Tau protein aggregation^{(Reference Reagan, Cowan and Woodruff71–Reference Chen, Cai and Hoover73)}.

Some authors suggest that the impaired insulin sensibility in the brain is the master phenomenon that links obesity to cognitive dysfunction^{(Reference Biessels and Despa5,Reference Batista, Soares and Mendonça9,Reference Qin, Sun and Mu30)} . Despite the complexity of attributing cause and effect in multifaceted phenomena like these, two causal elements are pointed out here as central: peripheral IR and neuroinflammation. In obesity and T2DM, there is a relative contribution of the peripheral IR to the brain IR via disrupted BBB insulin transport and vascular dysfunctions, as discussed in the previous section. Besides that, there is evidence that IR in the brain and periphery are independent events sharing pathogenic mechanisms that simultaneously affect brain health. Therefore, insulin treatment has been discussed as a therapy for Alzheimer’s disease even in the absence of diabetes^{(Reference Kellar and Craft52,Reference Guillemot-Legris and Muccioli74)} .

Neuroinflammation can result from many pathophysiological processes associated with obesity or from direct insults to the brain from dietary components and has a significant impact on the degree of IR. On a molecular level, insulin receptor substrate (IRS) phosphorylation on tyrosine is mandatory for insulin-mediated responses in brain structures, similar to peripheral tissues. Concurrently, serine phosphorylation of IRS by the kinases JNK and IKK, inflammatory mediators, FFA, ceramides, palmitate and RONS blocks insulin signalling^{(Reference Qin, Sun and Mu30,Reference Milstein and Ferris72)} .

Decreased expression levels of IRS-1 and other key insulin signalling cascade proteins were also reported in the hippocampus of HFD-induced obese mice^{(Reference Wang and Chen56)}. Other events previously associated with impaired insulin signalling in the brain promoting cognitive decline include IGF-1 resistance, insulin-degrading enzyme (IDE) hypofunction and Aβ accumulation^{(Reference Nguyen, Ta and Nguyen75)}

Little is known of how brain insulin sensitivity can be restored in humans, and non-invasive biomarkers of central IR are scarce. In addition to transgenic (insulin receptor knockout) and local IR induction (intracerebroventricular streptozotocin injection) animal models, the diet-induced obesity animal is emerging as a model to incite and revert central IR^{(Reference Qin, Sun and Mu30,Reference Milstein and Ferris72)} . In this sense, recent studies showed that flavonoid sources directly restored insulin signalling in the hippocampus of diet-induced obese animals by means of increasing the IRS-1 expression and phosphorylation in tyrosine. Subsequently to IRS, flavonoids directly up-regulate insulin downstream targets by activating the PI3K/AKT pathway (Figure 2). The IRS(1/2)/PI3K/AKT pathway coordinates much of the insulin functions in the hippocampus, such as promoting maturation of dendritic spines, potentiation of synaptic strength and increase of adult neurogenesis^{(Reference Mainardi, Spinelli and Scala50)}.

In addition, the IRS(1/2)/PI3K/AKT signalling pathway regulates the hippocampal neuroprotective property of insulin, among others, that is, its ability to keep the protein glycogen synthase kinase-3 β (GSK3β) inactive (Figure 2)^{(Reference Fu, Qin and Yu44)}. GSK3β is a serine/threonine kinase, and when is active, it contributes to Tau hyperphosphorylation, increases production of Aβ peptide, apoptosis and mitochondrial dysfunction and, therefore, may favour neuronal loss. Another important point is that activated GSK3β negatively regulates the cAMP-responsive binding protein (CREB)/brain-derived neurotrophic factor (BDNF) pathway involved in neuronal survival and synaptic plasticity^{(Reference Mi, Qi and Fan28,Reference Khan, Tantray and Alam76)} . In diet-induced obese mice brains, the inhibitory activity of GSK3β was observed after intervention with jambo fruit and jaboticaba peel, naringin, formononetin and EGCG, both by up-regulating of GSK3β Ser9 phosphorylation. Moreover, naringenin down-regulated the GSK3β active form (Tyr216 phosphorylated) in the hippocampus of old obese mice (Table 1). In these studies, interest in GSK3β inhibition was mainly directed to prevent Tau phosphorylation, which will be discussed below. Beyond that, therapeutic interest in targeting GSK3 for various brain disorders has been extensively reviewed^{(Reference Rippin and Eldar-Finkelman77)}.

Tau protein abnormality

Tau is a microtubule-associated protein that provides important cytoskeletal stabilisation of neurons. It influences neuronal shape and polarity, growth, transport and migration, especially in axons. Misregulation of Tau, mostly by hyperphosphorylation events, decreases its affinity for microtubules and may cause cytoskeletal dysfunctions compromising synapses and neuronal survival^{(Reference Papanikolopoulou, Skoulakis, Barrio, Sutherland and Rodriguez78)}. Moreover, hyperphosphorylated Tau may form intracellular aggregates that are neurotoxic and cause brain oxidative stress, mitochondrial damage, deregulation of intracellular signalling pathways, synapse failure and memory deficits^{(Reference Chen, Deng and Liu32,Reference Papanikolopoulou, Skoulakis, Barrio, Sutherland and Rodriguez78)} . Of great importance, Tau filaments amount (that is, neurofibrillary tangles) have been linked to the pathogenesis of several dementing conditions grouped as tauopathies, for example, Alzheimer’s disease^{(Reference Kimura, Sharma and Ishiguro79)}.

Increased levels of hyperphosphorylated Tau in the cortex and hippocampus have been documented in obesity diet-induced rodent models^{(Reference Fu, Qin and Yu44,Reference Zhou and Wang61)} . Furthermore, in ageing models, HFD feeding accelerated cognitive impairment accompanied by Tau hyperphosphorylation^{(Reference Liang, Gong and Ye80)}. In these preclinical models, it is often assumed that Tau hyperphosphorylation is a consequence of the dysregulation of the central insulin signalling pathway, given that insulin inactivates GSK3β, one of the main kinases responsible for Tau phosphorylation (Figure 2). In agreement, similar results occur in the diabetic brain of intracerebroventricular streptozotocin-injected rats. It is exciting that 6 weeks of treatment with intranasal insulin in these animals greatly attenuated the hyperphosphorylation of Tau at Thr205, Ser262 and Ser396 sites and improved cognitive performance in the Morris water maze (MWM) test^{(Reference Guo, Chen and Mao81)}. The presence of the Aβ peptide also results in the activation of GSK3β and Tau hyperphosphorylation in obese brain rodents, although addressed less frequently^{(Reference Kim and Amin82)}.

In addition, there is evidence that defective brain clearance and degradative systems are implicated in Tau accumulation and tauopathies. Tau is degraded by both the proteasome and autophagy systems. Phosphorylated forms and certain cleavage products of Tau, as well as Tau aggregates and monoubiquitylated Tau are selectively degraded by autophagy^{(Reference Papanikolopoulou, Skoulakis, Barrio, Sutherland and Rodriguez78,Reference Chesser, Pritchard and Johnson83)} . Autophagy deficiency in the hippocampus was reported in diet-obesity feed mice^{(Reference Zhuang, Lu and Wang31)}. Therefore, this could be another mechanism by which obesity exacerbates Tau accumulation and aggregation. Thus, understanding how these pathological Tau species can be formed enables us to develop effective therapeutic approaches to prevent their formation, although more studies are needed to optimise the degradation and cleaning processes of these proteins.

In obesity diet-induced models, isolated flavonoids and flavonoid-rich fruits elicit neuroprotective effects on the hippocampus by decreasing Tau phosphorylation in GSK3β-mediated pathways, positively impacting performance in cognitive behavioural tests (Table 1). Jaboticaba peel and jambo fruit minimise Tau phosphorylation at Thr205^{(Reference Batista, Soares and Mendonça9,Reference Batista, Mendonça and Soares42)} , and naringenin, quercetin and formononetin at Ser396^{(Reference Chen, Deng and Liu32,Reference Fu, Qin and Yu44,Reference Zhou and Wang61)} . Tau has several potential phosphorylation sites; however, these studies represent a crucial initial step towards modulating Tau misregulation in the context of obesity^{(Reference Yamauchi, Kamon and Minokoshi84)}. Moreover, purple sweet potato solution, an extract rich in anthocyanin, and naringenin improved the deficit in autophagy and reduced the generation of Aβ in obese mice hippocampus, respectively, both phenomena that potentially prevent the unwanted forms of Tau^{(Reference Zhuang, Lu and Wang31,Reference Zhou and Wang61)} .

Amyloid-β peptides and amyloid plaques

Along with neurofibrillary tangles, amyloid plaques comprise the other pathological hallmark of Alzheimer’s disease. Amyloid plaques consist of forty or forty-two amino acids called Aβ peptides that are aggregated in fibrils containing a high β-sheet insoluble structure. The amyloid hypothesis proposes that neurodegeneration can occur because of too much Aβ production, a change in ratio with more Aβ42 than Aβ40, or impaired clearance of Aβ in the brain^{(Reference Blennow and Zetterberg85)}. The Aβ peptide is derived by sequential cleavage of the amyloid precursor protein (APP), a transmembrane glycoprotein, by the β-secretase (β-site APP cleaving enzyme, BACE1) and δ-secretase at N- and C-termini, respectively (Figure 2). As a result, three protein fragments are formed: soluble APP (sAPP) and Aβ in the extracellular space and the amyloid intracellular domain (AICD) in the cytoplasm. While in the preferable state, APP is initially cleaved by α-secretase and thereby by δ-secretase to generate sAPP and a P3 non-toxic peptide into the extracellular space and the AICD into the cytoplasm. The presence of sAPP is related to normal synaptic plasticity, learning and memory, emotional behaviours and neuronal survival. Extracellular amyloid plaques, in turn, are typically surrounded by reactive astrocytes and microglia and by dystrophic neurites (axonal swellings) which together comprise a neuritic plaque, reinforcing a vicious cycle of neuroinflammation and neurodegeneration^{(Reference Jankowsky and Zheng86)}. The neurotoxicity of Aβ is closely connected with its potential to generate free radicals and oxidative stress^{(Reference Kim and Kim87)}, Tau hyperphosphorylation and neuroapoptosis related to GSK3β activation^{(Reference Zhou and Wang61,Reference Kim and Amin82)} . Moreover, Aβ may disrupt the CREB/BDNF pathway and the insulin signalling and function in the brain (Figure 2). Together, these deleterious effects may lead to neurodegeneration, cognitive dysfunctions and, ultimately, neurodegenerative disease^{(Reference Zhou and Wang61,Reference Nguyen, Ta and Nguyen75,Reference Amidfar, de Oliveira and Kucharska88)} .

Importantly, Aβ42 in the cerebrospinal fluid (CSF) declines over one decade before clinical symptoms of dementia, and the presence of amyloid plaques precedes the formation of neurofibrillary tangles, suggesting that the first Alzheimer’s disease marker may exacerbate the formation of the second (Figure 2)^{(Reference Buchhave89)}. In an attempt to delay or prevent dementia, efforts to identify risk factors for amyloid plaque deposits have been made.

Previous clinical studies have shown that obesity in midlife may exacerbate subtle cognitive decline associated with Aβ deposition^{(Reference Clark, Koscik and Allison90,Reference Sullivan, Blackshear and Simino91)} . In animal models, a chronic obesogenic diet intensifies the amyloid parameters in transgenic models of Alzheimer’s disease^{(Reference Bracko, Vinarcsik and Cruz Hernández92)} and senescence-accelerated prone mouse (SAMP8)^{(Reference Zhou and Wang61)} and even in healthy animals^{(Reference Nakandakari, Muñoz and Kuga93)}. The relationship between obesity-induced metabolic dysfunction (that is, hypercholesterolaemia, non-alcoholic fatty liver disease and IR) and Aβ is currently considered:

The cholesterol metabolism is affected in obesity, resulting in hypercholesterolaemia and high TG and LDL serum values^{(Reference Metwally, Rashad, Mahmoud and Morus alba25,Reference Zhuang, Lu and Wang31)} . Evidence supports that dyslipidaemia control could have a therapeutic role in Alzheimer’s disease by amyloid cascade inhibition because enhancing the cholesterol cellular membrane content may result in increased lipid rafts (distinct membrane domains characterised by high concentrations of cholesterol and glycosphingolipids where β-secretases are more active), β-secretase activity, APP processing and, finally, Aβ production^{(Reference Fonseca, Resende and Oliveira94,Reference Pac-Soo, Lloyd and Vizcaychipi95)} .

The hepatic functionality should be considered when the Aβ production/clearance balance between peripheral organs and brain is addressed: brain-derived Aβ in the blood can be cleared physiologically when it goes through the peripheral organs, mainly the liver. At the same time, the liver appears to be the major site of peripheral Aβ synthesis which can cross the BBB and contribute to brain neurotoxicity^{(Reference Sutcliffe, Hedlund and Thomas96)}.

In a previous clinical study, midlife HOMA-IR predicted late-life brain Aβ accumulation after 15 years^{(Reference Ekblad, Johansson and Helin97)}, reinforcing a strong link between IR (as an eventual obesity dysfunction) and amyloid pathology^{(Reference Mi, Qi and Fan28,Reference Kellar and Craft52)} . One hypothesis points to obesity leading to reduced cerebral blood flow and BBB disturbance, which may reduce brain Aβ clearance and increase Aβ peripheral influx, amyloid plaque production and, thus, neuronal dysfunction and even Alzheimer’s disease. In addition, insulin and Aβ are both degraded by the same enzyme in the brain, IDE. IDE is down-regulated in IR and high levels of insulin can compete with Aβ for IDE and block Aβ degradation^{(Reference Qiu and Folstein98)}.

Concerning the natural therapeutic approach, only Ref.^{(Reference Zhou and Wang61)} have explored flavonoid efficiency to stave off Aβ pathology in diet-induced obesity animal models and the related cognitive decline. They noted in HFD-fed SAMP8 mice that naringenin consumption could regulate soluble and insoluble Aβ40, Aβ42, BACE1 and APP levels while improving the mice’s performance in the Barnes maze and the MWM tests^{(Reference Zhou and Wang61)}. However, the underlying biomechanisms were not discussed. In previous non-obesity animal model studies, flavonoids demonstrated neuroprotectant abilities by inhibiting and destabilising amyloid fibrils through enhancing α-secretase activity^{(Reference Rezai-Zadeh, Arendash and Hou99)}; binding directly to the β-sheet-rich aggregates and mediating a non-toxic conformational change^{(Reference Bieschke, Russ and Friedrich100,Reference Xie, Wang and Yau101)} ; preserving the endothelial function and cerebral blood flow in a dyslipidaemic environment^{(Reference Bolduc, Baraghis and Duquette102)}; ameliorating the brain autophagic process which is essential for Aβ clearance^{(Reference Singh, Singh and Garg103)}; and improving brain insulin sensitivity^{(Reference Mi, Qi and Fan28)}. Other studies demonstrated flavonoids acting as brain antioxidant agents suppressing oxidative stress, another stimulator of Aβ peptide aggregation^{(Reference Kim and Kim87)}. These relevant results support that further studies are needed to explore flavonoids’ potential to prevent the onset of the obesity-induced amyloidogenic pathway, from both central and peripheral body perspectives.

Mitochondrial dysfunction and energy metabolism

The brain is the main energy consumer of the body utilising 25% of the total body energy supplies. Approximately 80% of this energy is used by neurons, primarily at synapses, where mitochondria play its major role providing ATP for basic brain functions. Cognition completely depends on neurogenesis, synaptic remodelling and neurotransmission, which requires an uninterrupted supply of mitochondrial ATP (Figure 3). This key role of mitochondria benefits from a very efficient machinery for energy usage; however, it also makes the brain highly vulnerability to perturbances in the energy supply^{(Reference Kaliszewska, Allison and Martini104)}. Naturally, the lifelong bioenergetic imbalance causes disturbances in brain reduction–oxidation (redox) homeostasis, that is, ageing, with well-known effects on cognitive function, mainly impaired learning and reduced memory. Diet-induced obesity acts in the brain like accelerated ageing at the mitochondrial level causing similar cognitive impairment^{(Reference Kaliszewska, Allison and Martini104,Reference Penna, Pizzella and Cimmino105)} .

Fig. 3.

Amyloid beta peptides associated with cognitive impairment in diet-induced obesity models and flavonoid intervention modulatory effect. AKT, protein kinase B; AMPK, AMP-activated protein kinase; BAK, Bcl-2-associated X protein; BCL, B-cell lymphoma; ERK, extracellular regulated protein kinase; FFA, free fatty acids; LC3, microtubule-associated protein light chain 3; LDL, low-density lipoprotein cholesterol; MAPK, mitogen-activated protein kinase; MMP, mitochondrial membrane potential; mTOR, mechanistic target of rapamycin; ROS, reactive oxygen species; ULK1, Unc-51-like autophagy-activating kinase 1.

The energy metabolism and mitochondrial function are deeply related, where mitochondria play an essential role in the glucose homeostasis. Mitochondrial processes lead to the production of ROS that generate oxidative damage, which is controlled by the cell’s detoxification system. The excessive intake of nutrients provokes the mitochondria to become overloaded with fatty acids and glucose, leading to an excess of ROS well above the capacity of the antioxidant system (Figure 3)^{(Reference Wang and Chen56)}.

It is well demonstrated in diet-induced obesity models that mitochondrial dysfunction is linked to impaired glucose metabolism, insulin secretion and reduced mitochondrial membrane potential (MMP). The MMP reflects the mitochondrial activity, defined by the proton gradient and transmembrane potential of hydrogen ions through the electron transport chain and oxidative phosphorylation. These are the key forces behind ATP production. Continuous high energy supplies can disrupt MMP, break down mitochondrial energetic functions impairing ATP synthesis and enhance ROS production, further causing additional mitochondrial damage. Moreover, excessive food intake causes alterations in mitochondrial fission and fusion processes, leading to altered mitochondrial morphology (Figure 3). The structural relation and functional contacts of mitochondria with other subcellular organelles are very important for the regulation of energy metabolism. Not surprisingly, overfeeding induces cognitive impairments through mitochondria dysfunction, accelerating neurological ageing and, thereby, favouring neurodegeneration^{(Reference Cunarro, Casado and Lugilde106)}. Accordingly, in hippocampal mitochondria of mice on 20 weeks of HFD, decreased MMP and ATP contents as well as an increase in ROS production were observed, with consequently impaired learning and memory function^{(Reference Wang and Chen56)}.

AMPK is considered the central protein regulating the body’s energy homeostasis. The energy status of the cell can be detected by AMPK, which is ultrasensitive and ubiquitously expressed, being a key energy sensor and regulator of cellular metabolic activity^{(Reference Ahmad, Serpell and Fong107,Reference Ahmad, Friar and Vohra108)} . AMPK signalling in the brain is functionally linked to behaviours, including cognition and affective symptoms^{(Reference Zhuang, Lu and Wang31)}; it acts as a modulator of the brain’s long-term potentiation (LTP), which is required for memory formation^{(Reference Wang and Chen56)}.

GSK-3β is an insulin-sensitive serine/threonine (Ser/Thr) protein kinase, highly expressed in almost all brain regions. GSK-3β phosphorylates and regulates more than 100 substrates, mainly recognised by its ability to inactivate glycogen synthase. Depending on the activation state of GSK-3β, it can be involved in various brain cognitive-related functions, such as neurogenesis, synaptic plasticity, memory formation and neurotransmitter signalling. Overexpression of GSK-3β in the brain can induce apoptosis and neuronal death, regulating the cell fate within the central nervous system. Moreover, GSK-3β plays a role in cholinergic neurons, since it modulates the synthesis of acetyl-CoA in mitochondria, then regulates acetylcholine production^{(Reference Fan, Zhao and Wang109)}. Martin et al. (2018) demonstrated that GSK3β regulated mitochondrial energy metabolism in neurons and glia and extends its action to cellular NAD metabolism^{(Reference Martin, Souder and Miller110)}. They showed that GSK-3β inhibition stimulated mitochondrial regulator PGC-1α and led to the activation of mitochondrial and redox pathways in the hippocampus. In mouse models of HFD and GSK-3β deletion or GSK-3β knockout, significant improvements in glucose tolerance and lower body weights were observed compared with controls. Therefore, GSK-3β has a fundamental role in metabolic disorders and is a potential therapeutic target for managing obesity disorders^{(Reference Gupte, Tousif and Lemon111)}.

Flavonoid consumption reversed brain metabolic changes and attenuated hippocampal mitochondria dysfunction improving learning and memory in obese mice (Table 1 and Figure 3). Naringenin increased MMP and ATP contents in hippocampal mitochondria. Naringenin^{(Reference Zhou and Wang61)} and other isolated flavonoids (Table 1), such as EGCG^{(Reference Mi, Qi and Fan28)}, apigenin^{(Reference Kalivarathan, Chandrasekaran and Kalaivanan39)} and luteolin^{(Reference Liu, Fu and Lan27)}, showed positive effects on regulating energy metabolism through improving insulin sensibility and glucose homeostasis locally in the brain, bringing the mentioned benefits to mitochondria function. Several studies have shown the ability of flavonoids to regulate glucose and fat metabolism by acting directly in many molecular targets, such as SIRT1, SIRT6, 67LR, ERα/β, ERp57 and SUMO1, controlling the expression/activity of central metabolic proteins (for example, AMPK, UCP and PO3K/Akt)^{(Reference Aoi, Iwasa and Marunaka112)}. Naringin and anthocyanins from purple sweet potato colour also were identified as activators of AMPK in HFD-fed mice presenting a decrease in their activity in the hippocampus and reflected in improved cognitive performance^{(Reference Zhuang, Lu and Wang31,Reference Wang and Chen56)} . Conversely, ingested and absorbed anthocyanins positively modulate the density and function of glucose transporters (GLUT4), through AMPK activation, restoring insulin sensitivity in humans^{(Reference Solverson113)}. Therefore, AMPK should be better explored as a target for phytochemical mechanisms of action, since it represents an essential link between systemic energy metabolism, mitochondrial dysfunction and the brain’s cognitive function. Anthocyanins food sources such as jaboticaba fruit^{(Reference Batista, Soares and Mendonça9)} and jambo fruit^{(Reference Batista, Mendonça and Soares42)}, as well as naringin^{(Reference Wang and Chen56)}, were able to inhibit GSK3β in the hippocampus of obese mice. GSK3β is implicated in the control of brain energy metabolism, specifically in the hippocampus, where it is involved in mechanisms of memory, behaviour and neuronal fate determination. When inhibited, GSK3β improves and regulates neuronal mitochondrial energy metabolism^{(Reference Martin, Souder and Miller110)}.

We found nine studies (Table 1) that evaluated the effect of flavonoids over oxidative stress of HFD rodent brains; most were very specific in analysing hippocampus, demonstrating that the antioxidant activity of flavonoids is an important mechanism of action for ameliorating obesity-related cognitive impairment. Most of these studies also evaluated the antioxidant defences, in the brain and/or periphery, showing that flavonoids’ overall action involves boosting the redox system through increasing the content of SOD, CAT, GPx and GRd antioxidant enzymes and GSH. There is extensive evidence that flavonoids can protect humans from diseases such as inflammatory, neurodegenerative and metabolic syndrome owing to their potent antioxidant activity. Ingested and absorbed flavonoids can easily access the brain through the BBB and act in the hippocampus, preventing/treating mitochondrial dysfunction by improving the antioxidant defences and reducing ROS^{(Reference Sarkar, Chaudhary and Jamaddar114,Reference Yang, Bai and Li115)} .

Autophagy and apoptosis

Autophagy and apoptosis are two distinct but interrelated cellular processes that regulate cell survival and death (Figure 3). Autophagy is a degradation process of damaged and defective cytoplasmic constituents which is fundamental for repairing the neuronal machinery and function. Apoptosis is the programmed neuronal death activated when inner repair is not sufficient to maintain function. Both processes are essential for maintaining proper cellular function and survival, including in the hippocampus. The hippocampus is a vital brain region associated with learning and memory, and disruptions in apoptosis and autophagy in this region have been linked to cognitive decline^{(Reference Fairlie, Tran and Lee116–Reference Melrose118)}. Increased apoptosis in the hippocampus has been shown to impair autophagy, leading to the accumulation of damaged organelles and proteins. Similarly, impaired autophagy has been shown to increase apoptosis in the hippocampus, leading to the loss of neurons and cognitive dysfunction^{(Reference Fairlie, Tran and Lee116,Reference Sharma119)} .

Deficient autophagy can reduce the expression of BNDF, which is found in high concentrations in the hippocampus and cerebral cortex and is considered a key molecule in the maintenance of survival of neuronal cells. Moreover, autophagy is necessary to maintain redox homeostasis. Therefore, the brain’s impaired autophagy can result in oxidative stress and BNDF down-regulation inducing neuronal cell death^{(Reference Penna, Pizzella and Cimmino105,Reference Fairlie, Tran and Lee116)} . The p-ULK1 signalling pathway plays a crucial role in autophagy initiation. while Bcl 1 is a pro-apoptotic protein that can inhibit autophagy by interacting with Beclin-1. LC3-II and p62 are markers of autophagosome formation and degradation, respectively^{(Reference Fairlie, Tran and Lee116)}. Also, the caspase family of proteases (twelve identified) is vital in the apoptosis process. Four caspases are initiators of an autoproteolytic process which will activate another three caspases, which are executioners of apoptosis, and cleave the Bcl-2 protein propelling the cellular death. Dysregulation of caspase-mediated apoptosis is linked to the pathogenesis of neurological disorders and metabolic diseases^{(Reference Gupta, Ambasta and Kumar117)}. Noteworthily, these events cited above are frequent in obesity.

In obesity disease, the expression of Bcl 1 is increased, leading to a decrease in Beclin-1 expression and subsequent inhibition of autophagy. This dysregulation of the Bcl 1/Beclin-1 complex can contribute to the development of insulin resistance and other metabolic disorders. In obese individuals, the expression of LC3-II is decreased, while the expression of p62 is increased, indicating a decrease in autophagy flux^{(Reference Sharma119–Reference Lim, Cho and Kim121)}. AMPK, PI3K and GSK3β are also directly involved in the modulation of autophagy and apoptosis while sharing the main molecular pathways that are altered in obesity-derived cognitive impairment. AMPK is a positive regulator of autophagy via directly phosphorylating and activating ULK1^{(Reference Gupta, Ambasta and Kumar117)}. HFD significantly decreased the phosphorylation of hippocampal ULK1(ser555), along with the protein levels of Beclin 1 and microtubule-associated protein (LC3)-II in the mouse hippocampus v. the control group. The accumulation of p62 was observed in the hippocampus of HFD-fed mice. Consistently, immunofluorescence staining showed a weaker signal of LC3 presented in the cytoplasm^{(Reference Zhuang, Lu and Wang31)}. Moreover, Hou et al. (Reference Hou, Jeon and Baek2022) showed that HFD induced damage in mice brains through disruption of autophagic flux, indicated by decreased levels of ATG12 and Beclin1, less LC3 conversion and significant accumulation of p62, which are important markers of autophagy machinery. HFD consistently shows a marked increase in caspase levels in mouse hippocampus and consequent cognitive impairment^{(Reference Hou, Jeon and Baek122)}. Previous evidence suggested that HFD led to decreased levels of hippocampal p-AMPK and cognitive dysfunction in mice^{(Reference Zhuang, Lu and Wang31)}.

The pathways/markers that were found to be affected by flavonoids from Table 1 include the UN51-like Ser/Thr kinases (ULK), the proteolytic cleavage of microtubule-associated protein 1 light chain 3 (LC3), the Bcl family of proteins, the apoptotic caspases and the p62 receptor of autophagy. This review shows that flavonoids can directly regulate autophagy and apoptosis through various mechanisms (Table 1 and Figure 3). The present collected evidence indicates that flavonoids have a direct positive effect on hippocampal cell autophagy and apoptosis through the regulation of the p-ULK1–Bcl 1, LC3-II and p62 signalling pathways^{(Reference Qin, Sun and Mu30,Reference Zhuang, Lu and Wang31)} . In addition, flavonoids can control apoptosis by reducing the expression of caspase-3 and BaK^{(Reference Qin, Sun and Mu30)}. Flavonoids can also indirectly influence autophagy and apoptosis by modulating PI3K/Akt and MAPK signalling pathways, which stimulate anti-apoptotic proteins and suppress pro-apoptotic proteins^{(Reference Zhuang, Lu and Wang31)}. The activation of these pathways can lead to an increase in autophagy flux, which is crucial for maintaining hippocampal cell homeostasis, thus leading to regulated apoptosis and the observed improvement in synaptic plasticity^{(Reference Metwally, Rashad, Mahmoud and Morus alba25,Reference Xia, Xie and Qiao53)} .

Specifically, glucosides of acylated anthocyanins from purple sweet potato were able to reduce caspase enzymes in the hippocampus of HFD-fed mice (Table 1). The studies have shown improved autophagy and reduced apoptosis in the hippocampus; consequently, better neuron viability was observed in treated animals. Moreover, obese mice when treated with purple sweet potato anthocyanins presented higher levels of p-ULK1, Bcl 1, Bcl 2 and LC3-II, and reduced expression of p62 in hippocampal cells, indicating an increase in autophagy flux and reduced apoptosis^{(Reference Qin, Sun and Mu30,Reference Zhuang, Lu and Wang31)} . The ingestion of purple sweet potato anthocyanins reduced neuroinflammation, brain oxidative stress and microglia activation while promoting an improvement of insulin sensibility in obese mice^{(Reference Li and Mi26)}. The purple sweet potato anthocyanins were able to protect against high-fat-diet-induced cognitive deficits through AMPK-mediated autophagy in mouse hippocampus^{(Reference Qin, Sun and Mu30)}. Zhuang et al. (Reference Zhuang, Lu and Wang2019b) showed that the protective effects of purple sweet potato anthocyanins against HFD-induced cognitive deficits were associated with improving AMPK-mediated autophagy, blocking oxidative stress and restoring BNDF protein levels, ultimately suppressing hippocampal apoptosis^{(Reference Zhuang, Lu and Wang31)}.

Neurogenesis and synaptic plasticity

Neurogenesis throughout life is a fascinating ability of some mammalian species. In humans, this remains under debate. In brain rodents, neurogenesis can be observed in the subventricular zone of lateral ventricles and the subgranular zone of the hippocampal dentate gyrus^{(Reference Olude, Olopade and Ihunwo123)}. However, the contribution to behavioural output by adult-born neurons requires overcoming many steps, from stem cell activation and differentiation to the survival and integration of newborn neurons into pre-existing circuits^{(Reference Denoth-Lippuner and Jessberger124)}.

BDNF, nerve growth factor (NGF), neurotrophin (NT)-3 and NT-4/5 belong to the neurotrophin family of growth factors. Neurotrophic factors exert crucial functions in the nervous system including neuronal growth, survival and regeneration to reinforce the strength and efficacy of synaptic transmission, that is, synaptic plasticity^{(Reference Mi, Qi and Fan28,Reference Inoue, Antunes and Maideen125)} . Therefore, improving NT levels may be positively associated with cognitive performance as well. Moreover, emerging evidence indicates that NT, especially BDNF and NGF, are also regulators of many metabolic and immunologic processes in both neuronal and non-neuronal cells^{(Reference Inoue, Antunes and Maideen125,Reference Colardo, Martella and Pensabene126)} .

Some studies suggest that obesity can interfere negatively with NT levels and action, but there is no consensus^{(Reference Cefis, Chaney and Quirié127–Reference Sandrini, Di Minno and Amadio129)}. BDNF, one of the most abundant NT in the brain, can be determined in serum and plasma. However, these values seem to fluctuate in the face of factors such as sex, ethnic origin, exercise, hormonal and peripheral production conditions^{(Reference Inoue, Antunes and Maideen125,Reference Cefis, Chaney and Quirié127,Reference Sandrini, Di Minno and Amadio129)} . This may explain the failure of clinical studies to determine a consistent correlation between circulating BDNF levels and measures of weight and adiposity in humans^{(Reference Kaur, Gonzales and Tarumi128–Reference Roth, Elfers and Gebhardt130)}. In this sense, animal models could better elucidate this relationship at a brain location.

In obese rodent models, unrestricted access to a HFD or high-fat, high-fructose diet (HFFD) resulted in a down-regulation of cortical BDNF and hippocampal BDNF, NGF, NT-3 and NT-4 simultaneously, worsening hippocampus-dependent learning and memory^{(Reference Liu, Fu and Lan27,Reference Mi, Qi and Fan28,Reference Zhuang, Lu and Wang31,Reference Xia, Xie and Qiao53,Reference Kalivarathan, Kalaivanan and Chandrasekaran55,Reference Kim and Kim131)} . Thus, because of immunotrophic to metabotrophic functions of NT, a bidirectional relationship between NT and obesity can be stated, where obesity disfavours NT actions, which, in turn, favour the obesity pathophysiology^{(Reference Inoue, Antunes and Maideen125,Reference Colardo, Martella and Pensabene126)} .

Obesity appears to implicate NT functions at various degrees of molecular signalling. The suppression of the BDNF transcription by the extracellular regulated protein kinase (ERK)/CREB pathway is often cited, where oxidative stress and chronically elevated glucocorticoids, such as cortisol, are causal candidates^{(Reference Mi, Qi and Fan28,Reference Xia, Xie and Qiao53,Reference Inoue, Antunes and Maideen125)} . In sequence, the impairment of the conversion of pro-BDNF to mature BDNF (mBDNF) by interfering with the action of proteases involved in this process was also raised^{(Reference Inoue, Antunes and Maideen125)}. Importantly, the BDNF binding with plasma membrane receptor tropomyosin receptor kinase B (TrkB) may be blocked in obesity. mBDNF/TrkB activates four main intracellular signalling cascades via activation of Ras/MAPK, phospholipase Cγ-1 (PLC-γ)/protein kinase C (PKC), PI3K/AKT and GTP-ases, resulting in neurogenesis, dendritic growth and branching, neuron survival and synaptic functions^{(Reference Xia, Xie and Qiao53,Reference Inoue, Antunes and Maideen125,Reference Kowiański, Lietzau and Czuba132)} . In addition, TrkB mRNA levels and phosphorylation can be decreased, as demonstrated in obese HFFD-fed rats^{(Reference Kalivarathan, Kalaivanan and Chandrasekaran55)}. Recently, it was also discussed that deficient BDNF/TrkB activity underlies neurodegeneration in Alzheimer’s disease, for example, by increasing the JAK2/STAT3 inflammatory pathway and δ-secretase expression, leading to both APP fragmentation and Tau hyperphosphorylation (Figure 2)^{(Reference Amidfar, de Oliveira and Kucharska88)}.

The promising benefits of the administration of flavonoids in neurogenesis and synaptic plasticity in the obesity context are consistent and mostly surround BDNF gene expression (Table 1). Interestingly, improvements in hippocampal dentate gyrus levels of doublecortin (DCX) (an immature neuron biomarker), dendritic arborisation of the DCX-positive neurons and BDNF were associated with blueberry consumption by HFD-fed mice^{(Reference Carey, Gildawie and Rovnak10)}. Purple sweet potato colour solution, another anthocyanin-rich food, could also ameliorate BDNF loss in the hippocampus of HFD-fed mice^{(Reference Zhuang, Lu and Wang31)}. EGCG has presented the capacity to modulate the ERK/CREB/BDNF pathway, as well as the expression of other growth factors (NT-3, NT-4 and NGF) in mouse brains while reversing the HFFD-elicited deterioration of spatial learning abilities^{(Reference Mi, Qi and Fan28)}. Quercetin (isolated quercetin or quercetin-derivative-rich extract of Morus alba L.), in turn, could normalise the expression of PI3K and downstream AKT, CREB and BDNF in the obese rodent hippocampus, which was correlated with the modulation of brain antioxidant defences and oxidative damage^{(Reference Metwally, Rashad, Mahmoud and Morus alba25)}.

CREB phosphorylation is shown to be induced by GLP-1 analogues and inhibitor of dipeptidyl peptidase (DPP)-4 because DPP4 can rapidly degrade and inactivate GLP-1. In the hippocampus of HFFD-fed rats, increased expression of GLP-1 receptor and inhibition of DPP-4 previously associated with improvement of CREB-BDNF signalling by apigenin consumption added to the increased TrkB expression and phosphorylation abilities^{(Reference Kalivarathan, Kalaivanan and Chandrasekaran55)}.

Neuronal nuclei (NeuN or RBFOX3) contribute significantly to neural and synaptic maturation and function in distinct brain regions, referred to as a neuronal viability marker. Accordingly, RBFOX3 homozygous knockout mice exhibited impaired learning abilities as well as cold hyperalgesia, a sensory abnormality^{(Reference Lin, Wang and Huang133)}. In obese mice under HFD, consumption of anthocyanin-rich foods (jaboticaba peel and purple sweet potato colour), elicited enhanced hippocampal NeuN expression and immunofluorescence staining accompanied by better cognitive performance in hippocampal-dependent behavioural tests^{(Reference Batista, Soares and Mendonça9,Reference Zhuang, Lu and Wang31)} .

Finally, some flavonoids previously up-regulated synaptic protein expressions in the brain of diet-induced obese rodents: N-ethylmaleimide-sensitive factor and vesicle-associated membrane protein 1 by EGCG^{(Reference Mi, Qi and Fan28)}; synapsin I and postsynaptic density protein 95 by luteolin^{(Reference Liu, Fu and Lan27)}; and synaptatogmin-1 and synapsin-1 by quercetin^{(Reference Gao, Chen and Fang54)}. Synaptic proteins participate in synaptic vesicle fusion, neurotransmitter release and synaptic transmission, suggesting they are another mechanism by which dietary flavonoid intervention may enhance synaptic plasticity and cognitive function^{(Reference Liu, Fu and Lan27,Reference Lin, Wang and Huang133)} .