Glucose is the brain's primary fuel source in meeting its intensive energy demand. However, temporal variations in the brain's energy demand and supply necessitate alternative additional fuel sources to meet its metabolic and energy challenges. Circulating ketone bodies (AcAc, BHB and acetone) provide metabolic fuel the supply of which can be elevated through carbohydrate fasting-induced hepatic catabolism of fatty acids or exogenous supplementation. Passing through the brain blood barrier and entering the mitochondria of cells in the brain through monocarboxylate transporters, BHB is metabolized into AcAc and then into acetyl-CoA, which feeds into the TCA cycle (Figure 1). Studies of D-BHB supplementation have shown benefits of providing ketones as an alternative to glucose as an energy source for the brain. Some of these benefits include a elevation in the nicotinamide adenine dinucleotide (NAD) redox state (NAD+/NADH) (49, 50), which is important for mitochondrial function, an increase in the free energy for ATP synthesis in neurons (49, 51, 52), and furnishing the cell with acetyl-CoA and citric cycle intermediates (53). These findings lend support to the therapeutic potential of nutritional ketosis in pathologies characterized by glucose insensitivity by providing an alternative energy substrate.

Substantial research has examined the metabolomic and bioenergetic effects of alcohol on the brain. Acute alcohol administration changes the brain's energetics, decreasing glucose metabolism while increasing the metabolism of acetate, a metabolite of alcohol (54). This alcohol-induced shift in brain energetics appears to be accentuated in AUD patients who, during sobriety, show higher brain acetate metabolism (55, 56) and lower brain glucose metabolism (54, 57) than non-alcohol dependent controls. These findings suggest that a shift from glucose to acetate metabolism persists beyond acute intoxication in individuals with AUD (Figure 2). During alcohol detoxification, when acetate supplies are low, this could lead to a central energy deficit that could contribute to the AWS and associated neurotoxicity (56) (Figure 2). The energy substrate deficit can be alleviated by increasing plasma ketone concentrations. Indeed, nutritional ketosis induced by a KD or oral D-BHB (ketone salts) decreased brain glucose metabolism, assessed with fludeoxyglucose ([¹⁸F]FDG-PET) and increased brain acetate metabolism, with [¹¹C]acetoacetate binding in healthy controls (12, 58). However, aging may influence this effect, as Roy et al. (59) showed both elevated brain [¹⁸F]FDG and [¹¹C]acetoacetate in aging rats after KD.

Brain studies in Alzheimer's disease can provide a useful parallel for AUD, as both diseases are associated with reductions in the global cerebral metabolic rate of glucose, which is estimated at 20–25% (60, 61). Reduced glycolytic flux and uptake (62) could help to explain this hypometabolism. Ketogenic diets and ketone supplementation have been shown to be protective in in vitro neuronal cell models (63) and benefits in clinical trials of Alzheimer's Disease (37, 64–66). Nutritional ketosis induced by the administration of MCT supplements has been shown to improve memory (67), and to double brain AcAc consumption in individuals with Alzheimer's disease, thereby increasing total brain energy metabolism without affecting brain glucose utilization (68). The relationship between plasma ketones and brain ketone uptake was the same in individuals with Alzheimer's Disease as in healthy young adults (58), indicating that there is intact AcAc utilization in Alzheimer's Disease. Thus, there is a potential for interventions that elevate circulating ketone bodies, primarily the administration of D-BHB, to be useful in treating pathologies characterized by impaired glucose metabolism and supply such as Alzheimer's Disease and AUD.

AWS is characterized by a general hyperexcitability of the central nervous system (69). The amino acids glutamate and γ-aminobutyric acid (GABA) are respectively the major excitatory and inhibitory neurotransmitters in the brain. Although alcohol initially inhibits excitatory effects by glutamate transmission and facilitates the inhibitory actions of GABA, chronic alcohol exposure results in compensatory changes in these amino acid transmitter systems that are opposite those seen with acute exposure and may contribute to alcohol withdrawal (70). There have been contradictory findings on brain glutamate concentrations in AUD from proton magnetic resonance spectroscopy (¹H-MRS) studies. Glutamate levels in the nucleus accumbens (71) and thalamus (72) have been shown to be elevated in individuals with AUD compared to non-dependent controls. However, glutamate levels in the anterior cingulate cortex have been reported to be higher (73), lower (74, 75) or unchanged (71, 72, 76) in AUD individuals during early withdrawal compared to non-dependent controls.

Glutamate in the cingulate of AUD patients was inversely correlated with the number of heavy drinking days in the 14 days preceding the MRS scan (76). Additionally, the number of drinking years but not drinks per day was associated with higher concentrations of glutamate and glutamine (Glx) in AUD. Mon et al. (75) concluded that sobriety may normalize glutamate levels over the course of abstinence. GABA levels in plasma (70) and cingulate cortex (77) have been shown to be low during acute alcohol withdrawal. Moreover, initially low cingulate GABA levels may normalize within 3 days of last alcohol intake but only in treatment-naïve individuals with more severe AUD (77). Thus, more research is needed to better understand the dynamics of brain glutamate and GABA in individuals with AUD.

Glucose metabolism through the TCA cycle is both the main source of energy to the brain and the main source of carbon for the synthesis of glutamate and GABA (78, 79). In a murine model with reduced brain-specific pyruvate dehydrogenase activity, reduced flux through the TCA cycle reduced the glutamate content of the brain and elicited epileptiform discharges, which were ameliorated by acetate administration (80). A recent study in a mouse model of Alzheimer's Disease (triple transgenic Alzheimer's 3xTgAD, which shows reduced brain glucose utilization) showed higher hippocampal glutamate and α-ketoglutarate (a precursor of glutamate) in animals who received a KE diet compared to regular chow and a positive correlation between glutamate and α-ketoglutarate levels in both groups (50). Thus, nutritional ketosis appears to furnish mitochondria with TCA cycle substrates (38), as evidenced by the finding that a 4-month KD elevated glutamate and glutamine in young adult rats (81). Although patients with epilepsy did not show differences from controls in posterior cingulate cortical glutamate measures, patients' elevated glutamate concentrations predicted short-term freedom from seizures, supporting the clinical relevance of glutamate concentrations in epilepsy (82). This underscores the need to elucidate the mechanism(s) underlying the association of KD-induced changes in glutamate with brain excitability.

Vesicular glutamate transporters (VGLUT) are required for packaging and exocytotic release of glutamate. VGLUT is inhibited by AcAc and BHB through a competitive interaction with the VGLUT allosteric activator Cl⁻ (83). A decrease in the concentration of glutamate per vesicle from VGLUT inhibition reduces glutamatergic activation, thereby dampening excitation. BHB and AcAc may also dampen neuronal excitability via their effect on K+/ATP channels, having been shown to reduce the spontaneous firing rate of substantia nigra pars reticulata neurons in vitro (84). This effect was abolished by the genetic or metabolic elimination of metabolically sensitive K⁺/ATP channels (84). In a recent study, acetone and BHB acted as inhibitors of glutamate at NMDA receptors (85). In addition, D-BHB and acetoacetate reduced neuronal death and changes of neuronal membrane properties in rat neocortical neurons subjected to glutamate excitotoxicity (86). Further, calorically restricted KD increases the expression of glutamic acid decarboxylase, the enzyme responsible for the conversion of glutamate to GABA (87), in the brain which increases the conversion of glutamate and thereby reduces excitation. Thus, there is conflicting evidence and potential mechanistic roles of glutamate in the context of AUD and further elucidation is needed with a particular emphasis on intracellular versus extracellular changes.

The efficacy of a KD in preventing or reducing seizures in epilepsy (88) may have direct relevance to alcohol withdrawal, which can be complicated by seizures. The following mechanism(s) have been proposed for the reduction of seizures by a KD: (1) restoring glutamatergic neurotransmission and enhancement of GABA synthesis, (2) circumventing glycolysis and providing Acetyl-CoA for the TCA cycle through fatty acid oxidation, (3) stimulating ATP-sensitive K⁺ channels, and (4) inhibiting voltage-dependent Ca²⁺ channels (78, 89). However, seizures are uncommon in AUD patients undergoing detoxification, partly because benzodiazepines, which are widely use to manage the AWS, have anticonvulsant activity (90). A KD may reduce overall neuronal excitability, mitigating the severity of alcohol withdrawal symptoms and reducing the need for benzodiazepine treatment during acute withdrawal. Thus, rodent studies are needed to investigate the effect of a KD on alcohol-induced seizures, as these would inform efforts to prevent alcohol withdrawal-induced seizures in patients (90). However, it is unclear how the hypothesized reductions of neuronal excitability with KD would associate with brain glutamate concentrations.

Ghrelin is a homeostatic hormone that stimulates human appetite, having effects opposite to those of leptin (91, 92). Endogenous peripheral ghrelin levels decrease during alcohol drinking and increase during alcohol abstinence (93–99). Studies have shown that genetic or metabolic reductions in ghrelin levels decrease alcohol intake (100, 101). In addition, higher ghrelin levels are associated with greater self-reported craving (97, 98, 101, 102), longer and more intense subjective responses to alcohol (103), and activation of the bilateral insulae (104) and ventral striatum (105) as measured with functional magnetic resonance imaging during alcohol cue exposure. Higher levels of ghrelin and activation of the ghrelin receptor stimulate the cholinergic-dopaminergic reward link, which has implications for the reinforcing effects of ghrelin in AUD (106). In healthy volunteers, a single administration of a KE was associated with decreased self-reported hunger and plasma ghrelin levels than the ingestion of isocaloric dextrose (15).

Although there is some indication that KD may suppress ghrelin levels (see review by Roekenes and Martins (107)), there are some inconsistencies in the literature (108–111). Leptin and peptide YY have effects opposite to ghrelin, in that they promote satiety (112, 113). A KD has been shown to increase serum peptide YY levels (114), though it has also been shown to decrease leptin levels (115, 116).

Fibroblast growth factor 21 (FGF21) is a hormone of hepatic origin whose targets include white and brown adipose tissue, the hypothalamus, and the hindbrain (117, 118). A KD has been shown to increase the concentration of FGF21 in murine models (119, 120), but this effect was not seen in humans (121–123). Nevertheless, in humans, FGF21-based pharmacotherapy decreased body weight (124) and variation in the FGF21 gene has been associated with macronutrient preference (carbohydrate, fat, and protein) (125). Moreover, FGF21 administration reduced a preference for alcohol in mice and for sweets in mice and monkeys (126). Therefore, FGF21 may be a key factor involved in the effects of ketosis on alcohol preference and warrants further investigation.

Glucagon-like peptide 1 (GLP1) is an intestinal hormone that enhances insulin secretion, inhibits glucagon secretion, and decreases gastric motility (127). There is some clinical evidence that the concentration of GLP-1 is increased in response to high fat KDs (109, 128), although experiments in cell culture have yielded contradictory evidence (129). GLP-1 receptor activation by GLP-1 agonists suppresses the effects of alcohol on the mesolimbic dopamine system and decreases alcohol consumption and operant self-administration (130–135). However, the GLP-1 receptor agonist Exendin-4 failed to attenuate morphine conditioned place preference or remifentanil self-administration (132). In addition, there is limited evidence that GLP-1 receptor agonists affect cocaine consumption (136, 137). Taken together, GLP-1 receptor activation induced by increased GLP-1 levels produced by a KD could serve as a suppressor of alcohol intake. Further research is needed to establish the role of the KD effect on circulating GLP-1 levels.

Evidence suggests that alcohol dependence is associated with dysregulation of the hypothalamic-pituitary-adrenal (HPA) axis and extrahypothalamic glucocorticoid signaling as well as other stress (e.g., corticotropin-releasing factor [CRF]) and anti-stress (e.g., neuropeptide Y) systems (138). However, the few available studies of the effects of ketosis on the HPA axis and other stress systems have yielded contradictory findings. For example, in one rat study, neither a KD nor a ketone supplementation diet affected plasma levels of adrenocorticotropic hormone or corticosterone (139). In another study, both KD and MCT increased HPA axis activity (140). Interestingly, in female but not male rats exposed to chronic mild stress, a KD prevented stress-related blood corticosterone and hypothalamic NPY expression; this effect was not accompanied by altered CRF mRNA expression (141). Furthermore, continuous microinjection of D-BHB into the prefrontal cortex attenuated the effects of a chronic unpredictable stress on depression-like behavior and HPA axis activity (142). More research on the effects of ketosis on stress systems is needed.

NAD⁺ is present in all living cells and plays a vital role in cellular metabolism as a coenzyme for redox reactions, including those required for mitochondrial energy production. NAD⁺ decreases with age (143, 144) and lower NAD⁺ levels are associated with neurodegenerative and neuropsychiatric disorders including Alzheimer's Disease and schizophrenia (145). Although individuals with AUD have low liver concentrations of NAD⁺ (146), it remains to be determined whether their brain NAD⁺ concentrations are affected by chronic heavy alcohol consumption. Because NAD⁺ and pyruvate are implicated in both the oxidation of alcohol (147, 148) and in the metabolic effects of fasting (149), it is possible that these compounds mediate the clinical efficacy of nutritional ketosis in AUD. An intravenous infusion of NAD+ during alcohol or opioid withdrawal attenuated both craving and withdrawal symptoms (150). A 7-Tesla magnetic resonance spectroscopy study in healthy volunteers showed that ketone supplementation elevates the concentration of NAD⁺ in the brain (151). Mice who received dietary supplementation with KE had higher cortical and hippocampal free cytosolic [NAD⁺]/[NADH] than mice fed a control diet (50). A KD also increased cellular concentrations of NAD⁺ (152), along with concentrations of Sirt1, Parp-1, and 8-hydroxy-2′-deoxyguanosine, which could improve brain health by increasing resilience to DNA damage and oxidative stress (153). However, the effects of ketone modulation of NAD⁺ in patients with AUD and its clinical and cognitive effects are unstudied. Because nucleotide coenzymes and their corresponding oxidizing forms are compartmentalized and bind at a subcellular level, their measurement and the interpretation of the results require great care to ensure accuracy. For example, from fed, freeze-clamped, rat liver the calculated free cytoplasmic [NAD⁺]/[NADH] from lactate dehydrogenase was approximately 200 times higher than the ratio calculated measured using total concentrations of the coenzymes. Conversely, the free cytoplasmic [NADP⁺]/[NADPH] from isocitrate dehydrogenase was ~20 times lower than the ratio calculated from measured total respective amounts (149).

In addition to its direct action in mitochondrial metabolism, D-BHB may exert therapeutic effects as a signaling molecule. D-BHB has been suggested to have direct involvement in epigenetic regulation due to its ability to act as an inhibitor of class 1 histone deacetylases (HDAC) that increases global acetylation levels in a dose-dependent manner (154). During withdrawal from chronic alcohol, anxiety-like behaviors were correlated with an increase in HDAC activity and a decrease in H3/H4 acetylation, but the behaviors could be reversed with the HDAC inhibitor trichostatin A (155). Furthermore, alcohol withdrawal-induced hyperalgesia was attenuated by the HDAC inhibitor suberoylanilide hydroxamic acid (156). In addition, the anxiolytic-like responses to acute alcohol administration were associated with increased histone acetylation and HDAC inhibition in the amygdala (155, 157). In vitro application of D-BHB also increased the expression of FOXO3, MnSOD, CAT, and MT2, genes that encode oxidative stress resistance factors (154). Bolstering this mechanism, various studies have demonstrated D-BHB's neuroprotective effect against oxidative stress (158–160). In humans, alcohol is processed by ADH enzymes into acetaldehyde, which produces unstable free radicals like hydrogen peroxide and superoxide (161, 162). Furthermore, chronic alcohol consumption depletes mitochondrial glutathione, a potent antioxidant (163). Thus, D-BHB may be unique in its capacity to respond to epigenetic and oxidative stress changes that occur during AUD.

Brain-derived neurotrophic factor (BDNF), a neurotrophin that helps control neurogenesis, has been implicated in the development of AUD (164–166). Although individuals with current AUD had lower overall serum BDNF levels than non-AUD controls (167), preclinical studies indicate that the directionality of the BDNF change is brain region-specific (168, 169). Further, BDNF levels raise during alcohol withdrawal in preclinical models (170), clinical populations (171–173), and raise during withdrawal from other addictive drugs (174, 175). Mechanistically, D-BHB enhances the expression of BDNF through downstream targeting of CREB and acetylation of BDNF promoters (176–178). While some clinical evidence points to serum BDNF being significantly increased following adherence to a KD (116, 179, 180), Vizuete et al. (181) found a KD decreased striatal BDNF levels and had no effect on hippocampal levels of BDNF in Wistar rats. The KD and D-BHB's effect on BDNF expression in the context of AUD warrants investigation.

D-BHB is a ligand of the hydroxyl carboxylic acid receptor type 2 (Hca2) (182), a GPCR encoded by the Hcar2 gene that mediates anti-inflammatory effects (183). In a rodent stroke model, a KD and D-BHB separately rescued stroke-induced neurological deficits but the effect was not seen in Hcar2 knockout mice (Hcar2^−/−; (184). These findings reinforce the critical role of Hca2 as an intermediate for D-BHB's neuroprotective effects. In addition to lower hepatic D-BHB levels in humans with alcohol-associated hepatitis, D-BHB attenuated abnormalities in plasma ALT levels, steatosis, and hepatic trigylceride levels induced by the β-oxidation inhibitor etomoxir and alcohol (185). The protective effect of D-BHB was not seen in Hcar2^−/− mice (185).

The NLR family pyrin domain containing 3 (NLRP3) inflammasome complex is a predominately macrophagic protein that mediates caspase-1 activation and the secretion of proinflammatory cytokines in response to mitochondrial dysfunction, ROS and more. Evidence suggests that the NLRP3 inflammasome complex is activated by alcohol consumption (186, 187) and inhibited by D-BHB supplementation (188). Although deficiencies of NLRP3 were shown to attenuate alcohol-associated steatosis (189), a study showed that this inhibition can increase the rate of hepatic damage (190), suggesting that the NLRP3 inflammasome complex may be protective during alcohol-induced hepatic damage. In vitro inhibition of the NLRP3 inflammasome by D-BHB was decreased by high insulin or high glucose, suggesting an influence of the metabolic state of the cell (191). Finally, a single dose of D-BHB clinically was shown to increase markers of NLRP3 inflammasome activation blood cells (192); however, this finding failed subsequent replication in patients with obesity (193). Further research is needed to elucidate the role of metabolic ketosis on the NLRP3 inflammasome in the context of AUD.