Linking mitochondrial dysfunction, neurotransmitter, neural network abnormalities and mania: Elucidating neurobiological mechanisms of the therapeutic effect of the ketogenic diet in Bipolar Disorder

Abstract

There is growing interest in the ketogenic diet as a treatment for Bipolar Disorder (BD), with promising anecdotal and small case study reports of efficacy. Yet, the neurobiological mechanisms by which diet-induced ketosis might ameliorate BD symptoms remain to be determined, particularly in manic and hypomanic states – defining features of BD. Identifying these mechanisms will therefore provide new markers to guide personalized interventions and provide targets for novel treatment developments for individuals with BD. In this critical review, we describe recent findings highlighting two types of neurobiological abnormalities in BD: 1) mitochondrial dysfunction; and 2) neurotransmitter and neural network functional abnormalities. We will consequently link these abnormalities lead to mania/hypomania and depression in BD and then describe the biological underpinnings by which the ketogenic diet might have a beneficial effect in individuals with BD. We end the review by describing future approaches that can be employed to elucidate the neurobiology underlying the therapeutic effect of the ketogenic diet in BD. In so doing, this may provide marker predictors to identify individuals who will respond well to the ketogenic diet, as well as offer neural targets for novel treatment developments for BD.

Bipolar Disorder (BD) is extremely disabling, yet current treatments often have severe side effects and high relapse rates (1, 2). There is interest in the ketogenic diet as a BD treatment, with promising anecdotal and small case study reports of efficacy (3). Indeed, a recent small, single-arm pilot trial revealed improved metabolic health in BD and schizophrenia patients (4). However, larger, more controlled clinical trials are needed. Likewise, the neurobiological mechanisms by which ketosis might ameliorate defining BD symptoms including mania in BD type I and hypomania in BD type II remain to be determined.

BD is characterized by cycling between mania/hypomania and depression (5); mixed mood episodes and rapid cycling are common (6); and there are high annual relapse rates into mood episodes of either polarity following manic/mixed episodes (1). Neural mechanisms of mania/hypomania, mood cycling, and depression are thus likely intrinsically linked. Specifically, depression might be the behavioral manifestation of neural mechanisms that compensate for pathophysiological processes underlying predisposition and onset of mania/hypomania, resulting in greater frequency of depressive rather than manic/hypomanic episodes (7). Thus, elucidating neural mechanisms underlying predisposition to mania/hypomania will provide targets for mania/hypomania as well as depression and mood cycling in BD. Here, we describe findings highlighting two types of neurobiological abnormalities in BD: 1) mitochondrial dysfunction; 2) neurotransmitter and neural network functional abnormalities, and how these abnormalities lead to mania/hypomania and depression in BD. We will also describe neurobiological mechanisms by which the ketogenic diet might have beneficial effects in BD. Finally, we propose future research for further elucidating these neurobiological abnormalities, and mechanisms underlying the ketogenic diet’s therapeutic effects in BD.

Neurobiological abnormalities in BD: mitochondrial dysfunction and energy dysregulation

BD is increasingly conceptualized as a disorder of energy dysregulation (8) and mitochondrial dysfunction (9). This is supported by studies showing altered brain energy metabolism in BD, including downregulation of NDUFS7, an essential mitochondrial complex I subunit; higher frequency of mitochondrial DNA mutations (10-13); polymorphism in autosomal mitochondrial complex I genes (14); high levels of mitochondrial metabolite lactate (15); decreased N-acetyl-aspartate (NAA):creatinine ratio (16); increased reactive oxygen species (ROS) production (17); and increased oxidative stress in dopamine (DA)-rich prefrontal cortical regions in postmortem brains of individuals with BD (11). We also recently showed that elevated lactate was significantly associated with metabolic syndrome in BD, a pattern that also extended to schizophrenia. Mitochondrial-targeted metabolomics further distinguished individuals with high lactate levels and metabolic syndrome from those without lactate changes but with metabolic syndrome (18).

Additional evidence supporting the role of mitochondrial dysfunction in BD comes from induced pluripotent stem cells (iPSCs) from individuals with BD, a useful in vitro model to examine neurobiological mechanisms of BD. iPSC-derived hippocampal cells showed upregulation of mitochondrial genes supporting protein kinases A and C and action potential firing, greater mitochondrial membrane potential and smaller mitochondrial size (19).

The structures of mitochondrial respiratory complexes are key determinants of their respective functions and disruptions of respiratory complex structures alter mitochondrial bioenergetics (20), causing cell stress that impairs neuron function (21, 22). Indeed, mitochondrial structural and morphological disturbances are evident in BD, consistent with reports of disturbed neuronal metabolism (23-25). For example, postmortem brain studies indicate that mitochondria are smaller in BD samples versus controls (25). Similarly, primary skin fibroblasts from individuals with BD show altered mitochondrial morphology and differences in mitochondrial density and distribution (25). Mitochondria in iPSC-derived glutamatergic neurons from individuals with BD are also smaller compared to controls (26). These changes are associated with altered excitability and increased activity in BD individual-derived neurons (25, 26), suggesting that hyperexcitability/increased activity in glutamatergic neurons is a BD endophenotype (26-28). These BD-associated increases in activity are also associated with disease-related changes in mitochondrial dynamics (e.g., fission/fusion) responsible for altered mitochondrial size, morphology and function (23, 24, 26, 29).

Mitochondrial dysfunction impacts glucose utilization and glutamate synthesis which can lead to decreased capacity for glucose oxidation via the tricarboxylic acid (TCA) cycle. Alternate anaplerotic pathways can provide TCA precursors to compensate for these deficits in oxidative glucose metabolism. For example, TCA enzyme pyruvate carboxylase leads to the increased synthesis of glutamate which functions as an anaplerotic intermediate to maintain the TCA cycle. Specifically, pyruvate carboxylase plays a crucial role in this process by catalyzing the conversion of pyruvate to oxaloacetate, a key TCA cycle intermediate, helping to maintain brain glutamate homeostasis (30). Furthermore, in BD, elevated glutamate levels have been linked to increased pyruvate carboxylase activity (31). It is important to consider the possibility that mitochondrial dysfunction and glycolytic dysfunction may interact in a bidirectional manner. Thus, in different individuals with BD, one pathway may precede and influence the other, and vice versa. This underscores the heterogeneous nature of BD and the importance of individualized approaches in understanding its pathophysiology. Recent studies support this bidirectional relationship. For instance, mitochondrial dysfunction can lead to a decreased glucose oxidation capacity via the TCA cycle, necessitating a compensatory increase in glycolysis to meet cellular energy demands. Conversely, primary glycolytic dysfunction can impose additional stress on mitochondrial function, exacerbating mitochondrial dysfunction. There is also evidence linking mitochondrial abnormalities to symptom severity and functional outcomes in BD. Decreased levels of N-acetylaspartate (NAA) and elevated lactate levels, both indicators of mitochondrial dysfunction, have been associated with greater severity of depressive and manic symptoms, as well as poorer overall functional outcomes (32). This suggests that mitochondrial health is not only a disease marker but a determinant of clinical severity and prognosis.

Though evidence for mitochondrial dysfunction in BD is clear, mitochondrial genetic association with BD remains weak. This suggests that mitochondrial dysfunction serves as an intermediate phenotype influenced by other causal risk factors such as environment, infections, and drugs rather than being a direct genetic consequence. Mitochondria are a hub for many biological pathways, with genetic mutations in other pathways leading to mitochondrial dysfunction. For instance, DISC1 gene mutations disrupt mitochondrial function, resulting in impaired mitochondrial transport and morphology, which may contribute to neurobiological abnormalities in patients carrying these mutations (33). Furthermore, increased DA neurotransmission can induce oxidative stress, increasing ROS production that disrupts mitochondrial metabolism (34). Thus, understanding mitochondrial dysfunction as an intermediate phenotype can guide more precise interventions by targeting these secondary processes. Finally, we emphasize that mitochondrial dysfunction is not BD-specific. Postmortem brain analyses showed altered mitochondrial morphology in patients with schizophrenia and depression. In schizophrenia, mitochondrial dysfunction is linked to changes in mitochondrial-related gene expression, impacting energy metabolism and contributing to pathophysiology (35).

Neurobiological abnormalities in BD: neurotransmitter and neural network dysfunction

Predisposition to mania/hypomania is associated with elevated phasic DA neurotransmission (36, 37), given that amphetamine can induce mania/hypomania in individuals with BD, and induces greater euphoria in those with higher levels of behavioral traits like reward sensitivity and low harm avoidance that predispose to mania/hypomania (38, 39). Conversely, antipsychotic medications which diminish DA neurotransmission treat or prevent recurrence of mania/hypomania (37, 40). Similarly, the therapeutic properties of lithium, one of the most prescribed mood stabilizers, are linked to the drug’s ability to modify DA biosynthesis and signaling, and to potentially diminish DA neuron hyperactivity (40-42). Though the precise mechanisms for lithium’s actions have yet to be fully elucidated, lithium likely acts on multiple intracellular targets to modify DA signaling including AKT and glycogen synthase kinase 3, protein kinase C, brain-derived neurotrophic factor (BDNF), adenyl cyclase, and phosphoinositide pathways (40, 41, 43, 44). Midbrain ventral tegmental area (VTA) DA neurotransmission is also regulated via gamma-aminobutyric acid (GABA) afferents from the rostromedial tegmental nucleus (45), GABAergic projection neurons from the nucleus accumbens (46), and VTA GABAergic neurons (45, 47). Elevated GABA levels are evident in prefrontal regions by Magnetic Resonance Spectroscopy (MRS) studies in individuals with BD (48), which might result in inhibition of prefrontal cortical function and, in turn, disinhibit midbrain DA neurotransmission. However, no studies have examined VTA GABA in individuals with BD. Lower levels of GABA marker glutamic acid decarboxylase (GAD67), which catalyzes GABA synthesis from glutamate, are observed in postmortem studies of individuals with BD (49). Moreover, inhibition of VTA GABA neurons in animals causes mania-like behaviors (50). Together, this suggests that a combination of elevated glutamate and lowered GAD67 leads to elevated glutamate to GABA ratio in BD, and that this altered GABA/glutamate balance results in elevated DA transmission, especially in the reward neural network.

Elevated activity in the DA-modulated reward neural network is especially evident in the ventrolateral prefrontal cortex (vlPFC) during reward anticipation in individuals with BD and in young adults at future risk of mania/hypomania due to their having heightened levels of behavioral traits that predispose to mania/hypomania (51-54). Furthermore, elevated euphoric response to amphetamine, which occurs in individuals at heightened risk of mania/hypomania, is associated with elevated vlPFC and wider reward neural network activity to reward anticipation in healthy individuals (55). The vlFPC receives dopaminergic input from the midbrain, but, relative to other cortical regions, has a thicker cortical layer V (56-58), where DA D₂-like receptors are expressed (59, 60). These findings suggest that the vlFPC might be especially responsive to midbrain DA release in contexts like reward expectancy (RE) where there is possibility of future reward. Greater RE-related left vlFPC activity in such contexts might reflect elevated phasic mesocortical DA transmission and is a potential neurobiological mechanism underlying mania/hypomania risk. Overall, these findings indicate that mania/hypomania (and related euphoria) are likely triggered in individuals with predisposition to mania/hypomania, due to underlying functional abnormalities in the DA-modulated reward neural network (associated with altered GABAergic and glutamatergic function). Nevertheless, despite these important basic and clinical neuroscience findings, more translational research is needed to elucidate the causal relationships among specific regional abnormalities in DA, GABA, glutamate transmission and reward (and other) neural network dysfunction that predispose to mania/hypomania, mood cycling, and, ultimately, chronic BD.

Linking mitochondrial, neurotransmitter and neural network abnormalities in BD

Elevated DA neurotransmission can lead to oxidative stress and aberrant mitochondrial metabolism (34). Studies in mice and humans showed that sustained elevations in DA synthesis raise cytosolic DA, boosting DA metabolism by monoamine oxidase (MAO) (61). MAO, anchored to the outer mitochondrial membrane, transfers electrons generated by DA deamination into the mitochondrial intermembrane space to enhance mitochondrial electron transport chain activity. Resulting increased ATP production supports elevated DA synthesis and release. Increased metabolic demand can, however, ultimately impair mitochondrial function (62) which elevates lactate and decreases N-acetylaspartate (NAA) due to increased dependence upon glycolysis rather than oxidative phosphorylation (32). These processes may culminate in elevated cytosolic DA due to reduced metabolism by MAO and DA oxidation. Such findings are consistent with work in Parkinson’s disease indicating a toxic cascade where elevated oxidative stress-induced mitochondrial dysfunction results in cytosolic DA oxidation and lysosomal dysfunction, further contributing to mitochondrial oxidative stress (63, 64). Consistent with this, we found increased protein oxidation of DA transporter-immunoreactive regions in postmortem prefrontal cortex from individuals with BD, supporting the link between oxidative stress and DA (65) (Figure 1; red and gray boxes).

Figure 1. Links among mitochondrial dysfunction, elevated dopamine neurotransmission and lowered GABA neurotransmission in BD.

Elevated dopamine (DA) synthesis, a potential neurobiological mechanism underlying predisposition to mania/hypomania, results in elevated cytosolic DA, which in turn leads to increased increase ATP production to sustain elevated DA synthesis and elevated metabolism of DA by monoamine oxidase (MAO) anchored to the outer mitochondrial membrane. Increased metabolic demand can, however, ultimately result in impaired mitochondrial function. This can lead to elevated cytosolic DA, due to reduced metabolism by MAO, DA oxidation and mitochondrial oxidative stress. In parallel, the mitochondrial transporter Aralar, responsible for uptake of GABA into mitochondria, is activated by mitochondrial membrane polarization resulting from high levels of mitochondrial metabolism. Sequestration of GABA by mitochondria reduces GABA availability for vesicular uptake. Additionally, inhibition of class I or II histone deacetylases (HDACs), which can protect neurons from oxidative stress-induced neuronal damage, might be reduced in individuals with BD.

There are also abnormalities in GABA neurotransmission potentially caused by dysregulation of mitochondrial homeostasis (66). The mitochondrial transporter Aralar, responsible for mitochondrial GABA uptake, is activated by mitochondrial membrane polarization resulting from elevated mitochondrial metabolism. GABA sequestration by mitochondria reduces GABA availability for vesicular uptake (Figure 1; blue box). Thus, sustained elevated DA synthesis and release can lead to: 1) inability of mitochondria to maintain sufficient ATP production, leading to increased dependence on glycolysis; 2) mitochondrial oxidative stress and DA oxidation and, in turn, 3) reduced GABA availability.

Inhibition of class I or II histone deacetylases (HDACs) can protect neurons from oxidative stress-induced neuronal damage frequently observed in BD (67, 68). Altered expression of class I HDACs was found in hippocampi of postmortem brains from individuals with BD (69). Using in vivo imaging of HDAC-specific radiotracers in individuals with BD, altered levels and activity of HDACs were related to attention and emotional regulation, further suggesting a role for HDACs in BD pathophysiology (70) (Figure 1; gray box). HDAC inhibitors also reversed mania-like behaviors in rat models (71, 72), potentially via changes in growth factors and other factors related to neuronal and synaptic plasticity (68). This could be mediated through increased BDNF signaling via TrkB receptors on VTA GABAergic neurons, leading to inhibition of dopaminergic activity (73). Furthermore, inhibition of class I HDAC protein, HDAC2, in the VTA is necessary and sufficient for the therapeutic actions of the mood stabilizer valproate (74). This highlights additional links between mitochondrial dysfunction, elevated DA neurotransmission, and lowered GABA neurotransmission in BD.

Together, while the specific links among mitochondrial, neurotransmitter, and neural network abnormalities in BD remain to be elucidated, the findings above suggest that sustained elevated DA synthesis and release in BD offer a neurobiological mechanism underlying predisposition to mania/hypomania. We posit that this can lead to increased mitochondrial metabolism/ATP production, and to reduced GABA availability resulting from mitochondrial polarization and GABA uptake. Ultimately, insufficient mitochondrial ATP production to meet demand for elevated DA synthesis may impair mitochondrial function, increasing cytosolic DA and mitochondrial oxidative stress. In parallel, reduced inhibition of class I and II HDACs in BD can compound this process by increasing mitochondrial oxidative stress (Figure 1).

Neurobiological mechanisms underlying therapeutic effects of the ketogenic diet in BD

There are at least two plausible mechanisms by which the ketogenic diet might impact mitochondrial dysfunction, neuroinflammation, DA, and GABA in BD. First, the ketogenic diet is thought to increase synthesis of GABA and reduce glutamate, the major excitatory neurotransmitter (75). This occurs via elevated production of acetyl-CoA from ketone bodies such as β-hyroxybutyrate (βOHB), leading to increased oxaloacetate consumption in the TCA cycle. The resulting higher glutamate:aspartate ratio drives increased GABA synthesis from glutamate (76). Second, βOHB is an inhibitor of class I HDAC proteins (77, 78). This leads to an increase in histone acetylation and increased expression of multiple genes that result in pleiotropic effects including: 1) an antioxidant response, 2) reduction of oxidative stress and inflammation, 3) a balanced redox state, and 4) improved mitochondrial function. In the kidney, these effects are mediated by upregulation of Forkhead Box O3 (FOXO3A) and Metallothionein 2 (MT2), master regulators of many of these processes (77). FOXO3A is a renal antioxidant regulator, reducing oxidative stress and inflammation, maintaining balanced redox state, and enhancing mitochondrial function; MT2 similarly acts a kidney antioxidant defense regulator. Together, these factors improve kidney mitochondrial function by reducing stress (77).

We propose a model where similar impairments in brain mitochondrial function and redox state imbalances contribute to altered mood states in BD, consistent with earlier studies (79) (Figure 2A). Alterations to the TCA cycle may elevate excitatory glutamate levels, and likely lower levels of VTA GABA (despite evidence of elevated prefrontal cortical GABA), resulting in an imbalance in glutamate/GABA signaling. Concomitant mitochondrial stress may also raise mitochondrial ROS, further negatively impacting brain function in BD. Increased neuroinflammation and its connections to mitochondrial dysfunction and oxidative stress may also contribute to BD pathology. Individuals with BD have elevated pro-inflammatory cytokine release, likely mediated by the NLRP3 inflammasome, a multiprotein complex (consisting of NLRP3, ASC, and pro-caspase-1) that detects pathogenic microorganisms and stress signals (80). Mitochondrial dysfunction activates the NLRP3 inflammasome through mechanisms like excessive ROS production, release of mitochondrial DNA into the cytosol as a damage signal, altered mitochondrial antiviral signaling protein (MAVS), and changes in ionic fluxes. This activation leads to cleavage of pro-caspase-1 into caspase-1, which then processes pro-inflammatory cytokines IL-1β and IL-18, triggering neuroinflammation (81). Conversely, the ketogenic diet’s therapeutic mechanisms in BD may derive from the ability of ketone bodies like βOHB to enhance mitochondrial metabolism and reduce neuroinflammation. βOHB increases oxaloacetate consumption in the TCA cycle, leading to increased GABA generation and a higher GABA/glutamate ratio which may normalize excitatory/inhibitory imbalances (82-84). In parallel, HDAC inhibitors such as βOHB lead to increased gene expression and inhibit the NLRP3 inflammasome, thereby decreasing release of pro-inflammatory cytokines to reduce pathogenic neuroinflammation (85) (Figure 2B).

Figure 2. Impact of ketogenic diet on mitochondrial metabolism in BD.

(A) Dopamine (DA) neurons in individuals with BD exhibit impaired mitochondrial metabolism. This is reflected in an altered tricarboxylic acid (TCA) cycle which yields elevated levels of glutamate, and likely lower levels of VTA GABA, although there is evidence of elevated prefrontal cortical GABA from MRS studies. This is compounded by elevated levels of cytotoxic reactive oxygen species (ROS) that further negatively impact neuronal function. In parallel, elevated levels of the class I histone deacetylase HDAC2 diminish gene expression. (B) The ketone body β-hyroxybutyrate (βOHB), a key component of the ketogenic diet, exerts its therapeutic effects in BD through several mechanisms, including improving mitochondrial metabolism. βOHB increases consumption of oxaloacetate in the TCA cycle, leading increased GABA production and a higher GABA:glutamate ratio. βOHB also diminishes HDAC2 levels, which results in increased gene expression as well as diminished oxidative stress and neuroinflammation.

Future research

Multimodal neuroimaging

While neuroimaging approaches alone cannot elucidate intracellular, extracellular, or cell-specific abnormalities in neurotransmitter function, different neuroimaging techniques can examine neural networks implicated in mania and thus BD (86). We used functional Magnetic Resonance Imaging (fMRI) to examine the DA-modulated reward neural network, showing elevated activity in individuals with BD and in young adults at future risk of mania/hypomania (51-54). MRI-based brain iron quantification provides a valuable proxy for midbrain DA synthesis, storage, and release because iron co-localizes with DA vesicles, and is involved in DA synthesis as a co-factor for tyrosine hydroxylase, the rate-limiting enzyme for DA biosynthesis (87, 88). More specifically, midbrain neuromelanin (NM), a dark pigment synthesized via iron-dependent oxidation of cytosolic DA, accumulates with iron complexes inside lysosomes in substantia nigra pars compacta (SNc) and VTA DA neuron soma. NM contrast to noise ratio (CNR) can detect DA abnormalities in patients, supporting its use as a proxy for DA synthesis capacity (88-96), with decreases and increases evident in neurodegenerative disorders (89) and schizophrenia (90, 97), respectively. NM is thought to reflect cumulative DA synthesis capacity rather than acute, phasic DA release (98). Midbrain NM is positively correlated with striatal DA transporter availability (99) and amphetamine-induced striatal DA release (90), though, thus far, no BD studies have used NM imaging. In addition, MRS can examine glutamate and GABA concentrations in the reward neural network in BD. These approaches can also probe concentrations of lactate, a proxy marker of aberrant mitochondrial oxidase phosphorylation dysfunction. Future research can employ these neuroimaging modalities to: 1) examine reward neural network activity, 2) VTA DA synthesis capacity, 3) GABA, glutamate, and lactate concentrations in prefrontal and VTA regions, 4) relationships among these measures to mania/hypomania severity, and 5) the extent to which ketogenic diet reduces the magnitude of these abnormalities.

Induced pluripotent stem cell (iPSC)-derived 3D cerebral organoids

Most studies examining BD pathophysiology have used postmortem tissue or neuroimaging technologies. However, these techniques cannot completely capture the dynamic complexity of the human brain. iPSC-derived three-dimensional (3D) cerebral organoids (COs) offer a new opportunity to address this challenge by directly studying neural networks and cellular connections that closely model human brain. Unlike two-dimensional cell models that are limited to only a few cell types, 3D COs generate a large diversity of cells such as oligodendrocytes, radial cells and more mature neurons, that can self-organize into spherical shapes (100) – features crucial for understanding the impact of mitochondrial dysfunction on neurotransmission. For example, glutamatergic neurotransmission is mediated by tripartite synapses involving pre- and post- synaptic neurons wrapped by glial cells; these organizations cannot be recapitulated in other human-derived models and are essential for proper redox balance.

Furthermore, despite the success of recent GWAS studies in identifying hundreds of risk variants and genetic risk characterization of BD (101), questions remain regarding determination of the substantial deflation of heritability compared with in-kin studies. Many interpretations have been suggested such as disparate contributions of environmental factors, risk posed from rare variants, and unaccounted non-additive genetic effects (102). These uncharted complex effects and the immense intricacies of the human genome have led to challenges in translatable biological interpretation for precision psychiatry (103, 104). An advantage of iPSC-derived 3D COs for BD research is that they can be generated from individual patients, enabling tailored treatments based on the patient's unique genetic makeup.

Development of 3D COs from iPSCs derived from human peripheral blood mononuclear cells (PBMCs) has enabled monitoring of mitochondrial health from primary, reprogrammed, and differentiated stages. Importantly, this approach preserves mitochondrial genetics, function, and treatment responses across PBMCs to iPSCs to COs, and measurable neuronal activity in the COs derived from PBMCs (105). Other recent studies showed that iPSC-derived COs from individuals with BD exhibit abnormal neural connectivity (106), consistent with neuroimaging findings (107, 108). iPSC-derived human cortical spheroids from individuals with BD demonstrated that lithium lowers neuronal excitability, regulates secretion of pro-inflammatory cytokines, and increases mitochondrial oxygen reserve capacity (109). Furthermore, another study showed that young neurons derived from iPSCs of individuals with BD exhibit hyperactive action potential firing, which is selectively reversed by lithium (42). Using COs with iPSC-derived neuronal cells will therefore serve as a powerful approach for future BD studies examining mitochondrial function across different stages of neuronal development and cellular specificity.

While iPSC-derived COs represent a significant advancement in modeling human brain development and disease, it is essential to acknowledge their limitations. One challenge is the heterogeneity between organoids even when derived from the same source or patient cell line. This variability complicates consistency and reproducibility of experimental results. Additionally, these models predominantly represent early developmental stages (mostly late-stage neural progenitor stem cells), and lack the complexity, maturity, and decades of cumulative stress of the adult brain. Consequently, while organoids provide valuable insights into human neurodevelopment and disease mechanisms, they do not fully recapitulate the intricate environment of the mature brain. These caveats highlight the need for cautious data interpretation and continued development of more advanced models to better mimic the complexities of human neurobiology (103, 104). Scientists continue work to identify better methods, as demonstrated by Sivitili and colleagues (110), who developed a pipeline capable of successfully and reliably generating morphologically consistent human organoid models.

Examining mitochondrial supercomplexes

The mitochondrial electron transport chain is composed of five respiratory complexes (CIV) responsible for ATP synthesis through oxidative phosphorylation and are integral for DA neuron function (111, 112). Conversely, respiratory complex dysfunction leads to mitochondrial bioenergetic abnormalities associated with an array of neurodegenerative and psychiatric illnesses including BD (113-116). The structural organization of the respiratory complexes is integral to the effective function of the mitochondrial electron transport chain and its ATP biosynthetic capacity (117). Indeed, respiratory complexes do not function in isolation but are organized into higher-order assemblies called ‘supercomplexes’ in defined ratios (e.g., CI/CIII/CIV)(117-119). Supercomplexes are critical for 1) regulating efficiency of electron transport to make ATP (20, 119, 120); 2) stabilizing individual respiratory complexes (121); and 3) decreasing ROS production (120, 121). Supercomplex assembly is also dynamic, depending on mitochondrial membrane composition/shape (i.e., crista morphology) as well as on expression levels and local distribution of the individual complexes. Thus, disease-associated changes in respiratory complex expression, structure, local mitochondrial membrane composition, or some combination of these factors, can impair supercomplex organization to disturb cell energetics (118, 119, 122). We therefore posit that disturbances in supercomplex organization and function may contribute to BD pathology and deserve further study. Finally, mitochondria, can organize into higher-order structures with glutamate transporters (e.g., GLT-1, GLAST) and Na⁺/K⁺ transporters, particularly in astrocytes (123). Such co-compartmentalization is disrupted in schizophrenia leading to impaired bioenergetic coupling that may explain some of the metabolic deficits associated with the illness (35). This raises the question of whether similar disruptions in mitochondrial/transporter associations also contribute to BD pathophysiology.

Prior work examining mitochondria in BD employed either conventional light imaging and/or transmission electron microscopy (TEM) approaches (25, 124, 125). While these methods can detect grossly altered mitochondrial morphology, they cannot directly resolve structural defects specific to individual respiratory complexes or higher-order supercomplex organization (21). Sample preparation (e.g., fixation, staining, dehydration, sectioning) may also alter mitochondria, making it still more difficult to disentangle disease changes from those caused by sample preparation (126, 127). Thus, mitochondrial structural changes in BD may be significantly underestimated. Recent work by our group and others used in situ cryo-electron tomography (cryo-ET) to image mitochondria in human cells in the native state, free of extrinsic manipulations that might alter intracellular architecture (21, 127). Critically, with in situ cryo-ET’s subnanometer resolution, future studies will be able to directly identify 3D respiratory complex structures and their higher-order supercomplex organization to ascertain if these structures are altered directly in cells from individuals with BD, and if these changes are improved in response to ketogenic diet therapy.

Importance of glial cells

In addition to changes in neuronal function, a ketogenic diet also impacts the function of astrocytes, oligodendrocytes, and microglia which help to support neuronal activity, regulate synaptic connections, and mediate inflammatory responses. Previous studies found that mice that have undergone chronic stress in the presence of a ketogenic diet have fewer cellular stress markers in hippocampal microglia compared to animals fed standard chow (128). Furthermore, treatment with the ketone βOHB in the presence of a lipopolysaccharide challenge in microglial cell cultures results in reduced expression of pro-inflammatory markers, suggesting that βOHB generally acts as on microglia as an anti-inflammatory and neuroprotective agent (129). In addition to microglia, astrocytes are logical key players in the response to ketosis as they are the main site of fatty acid oxidation in the brain. Studies suggest that ketone bodies produced by fatty acid oxidation are the dominant source of oxidative fuel for neurons in a glucose-restricted state (130). Furthermore, ketone bodies can suppress astrocytic glutamate release by inhibiting the actions of vesicular glutamate transporters, further impacting nearby neuronal function by changing the excitatory/inhibitory balance (130). Given these important effects in glial cells, it is likely that they play a prominent role in modulating neuronal activity and overall neuroinflammatory state in response to the ketogenic diet. It will be interesting in future studies to determine how a ketogenic diet impacts the role of astrocytes in DA dynamics and clearance.

Use of the ClockΔ19 mice as a model for the study of mania, mitochondrial function, and treatment development

One of the most consistent findings in BD is strong circadian rhythm disruption. Several human genetic studies have identified single nucleotide polymorphisms in the Clock gene, a core regulatory component of circadian rhythms, in the development, severity, and response to treatment in individuals with BD (131-136). We previously described a behavioral profile strikingly similar to the manic phase of BD in mice with a single-base mutation in the Clock gene (ClockΔ19)(137-139). This profile includes hyperactivity, increased exploratory drive, lower levels of behavioral despair, increased impulsivity, and a hyper-hedonic response to multiple rewarding stimuli, and these phenotypes have been consistently described across multiple laboratories (138-143). Chronic mood stabilizer treatment with lithium or valproate, reverses this manic-like phenotype (74, 139), making the ClockΔ19 mouse one of the few, well-validated, animal models of human mania which shows face, construct, and predictive validity. We found that most of their behavioral phenotypes are the result of abnormally high dopaminergic neurotransmission in midbrain regions (144). Moreover, levels of DA synthesis are directly regulated by an interaction between the CLOCK protein and mitochondrial factors that indicate the cell’s redox state (145). The CLOCK protein directly binds to the promoter of the tyrosine hydroxylase (TH) gene, leading to regulation of normal circadian rhythms in TH expression and DA levels. Interestingly, this regulation of TH by CLOCK is mediated by an interaction with the metabolic sensing protein, Sirtuin 1 (SIRT1), whose levels and activity vary based on cellular redox state controlled by mitochondria. Thus, levels of DA are tightly tuned to the metabolic state of the cell via SIRT1/CLOCK interactions. In ClockΔ19 mice, TH levels are abnormally high and rhythms are disrupted, leading to overall greater levels of DA, particularly at times of day when DA levels are normally low, and there is a disconnect between cellular metabolism and DA synthesis. CLOCK also directly regulates the expression of multiple genes related to mitochondrial fission, fusion, and autophagy (146). In the ClockΔ19 mouse heart, there is impaired mitochondrial turnover, leading to the accumulation of ROS-producing mitochondria with impaired mitophagy and ultrastructural defects (147). Furthermore, mitochondria of ClockΔ19 mice exhibit excessive fission due to changes in RNA stability of Dynamin-related protein 1 (148). ClockΔ19 mice are thus an ideal animal model to examine pathophysiological processes including mitochondrial disfunction and abnormalities in DA neurotransmission in BD, as well as the restoration of phenotypes by a variety of treatments.

Summary

Abnormalities in the DA-modulated reward neural network, mitochondrial metabolism, prefrontal cortical GABA, and lactate characterize BD. Promising studies using human iPSC-derived COs and rodent models of mania highlight the feasibility of cellular and rodent models alongside human neuroimaging to provide a comprehensive understanding of mitochondrial function, related gene expression, reward neural network dopaminergic and GABAergic neurotransmission abnormalities, and their relationships with mania/hypomania in BD. Moreover, the combination of these techniques may elucidate the cellular- and neural network-specific neurobiological mechanisms by which the ketogenic diet can ameliorate these abnormalities in BD. Future studies can use these approaches to determine new markers of BD pathophysiology, and then use this knowledge to identify those most likely to benefit most from the ketogenic diet. This approach can also examine mechanisms for other dietary interventions that may involve altered ketogenic metabolism, such as intermittent fasting, to ameliorate neural network and neurotransmitter abnormalities in BD. These new research directions will help guide personalized interventions and provide targets for novel therapies for BD.