Mitochondrial Molecular Diversity in the Brain

Abstract

Mitochondria are not uniform organelles. Across the brain, they exhibit profound molecular, biochemical, and functional diversity shaped by cell type, anatomical region, subcellular compartment, and lived experience. Recent advances in cell-type- and subcellular domain-targeted proteomics, transcriptomics, advanced live imaging, and functional biochemistry have begun to map this landscape with unprecedented resolution. Together, these findings challenge the conventional view of mitochondria as generic metabolic engines and position mitochondrial molecular diversity as a fundamental feature of brain organization, with direct relevance to behavior, aging, and neurological disease. This mini review synthesizes key recent studies in this field, highlighting their findings, methodological novelty, and significance, and formulates theories and hypotheses for future investigations.

Introduction

The brain is the most energetically demanding organ in the body, consuming roughly 15–20% of the body’s total energy budget despite representing only 2% of its mass [1]. This energy is overwhelmingly supplied by mitochondrial oxidative phosphorylation (OXPHOS) [1,2], which produces ATP through the coordinated activity of the electron transport chain (ETC) complexes I-V. In addition to providing ATP, mitochondria are crucial signalling hubs [3], and play central roles in ion and metabolite storage and transport, cell death, and the synthesis of biomolecules [4–6]. The brain has extraordinary cellular heterogeneity, encompassing dozens of neuronal subtypes, astrocytes, oligodendrocytes, microglia, and vascular cells, each with distinct metabolic demands and functional identities. Even within a single brain cell, particularly in highly polarized neurons, the spatial and temporal metabolic requirements differ across functional subdomains. These variations create unique cellular microenvironments that give rise to distinct mitochondrial identities, characterized by differences in function and molecular makeup, which enable specialized activities such as vesicle release at the synapse. As such, disruptions in mitochondria’s ability to adopt cell-type- and compartment-specific identities could result in a spectrum of acute and chronic brain disorders [7].

Yet for much of the history of neuroscience, mitochondria were treated as generic, interchangeable batteries. The prevailing assumption was that mitochondrial function varied primarily across tissues (e.g., liver versus heart) rather than among specialized cell populations within a single organ or across microdomains of a single neuron. A wave of studies over the past decade has expanded our definition of mitochondrial diversity in the brain, illuminating not only the differences in mitochondrial functional and molecular identity between neurons and glia but also how subcellular compartments shape mitochondrial function, how mitochondria support compartment-specific activity, and the functional consequences of losing mitochondrial identity at the cellular, systemic, and circuit levels (Fig. 1). Here, we highlight landmark foundational work and emerging studies to synthesize up-to-date models and theories, and underscore the significance of mitochondrial diversity in the brain. To capture this rapidly evolving field, we discuss recent preprints, while noting that their peer review has not yet been completed. These findings, including cell-type- and compartment-specific molecular landscapes, have suggested intriguing hypotheses that warrant mechanistic testing.

Figure 1. Dimensions of Mitochondrial Diversity.

Mitochondria exhibit multidimensional diversity across (A) brain regions and (B) brain cell types. Variations in (C) transport dynamics, (D) molecular engine activity, and (E) unique subcellular microenvironments drive mitochondrial divergence. These axes collectively define mitochondrial identity, which underlies neuronal function, adaptability, and selective vulnerability in disease.

I. Mitochondrial molecular diversity is tailored with cellular functional heterogeneity

Ia. Diversity across cell types in the brain.

We now know that the brain is composed of extremely complex cell types. Mitochondria exhibit a wide range of molecular diversity to meet cell-type-specific metabolic needs. By using the MitoTag mouse system, Fecher and colleagues isolated intact mitochondria from three principal cerebellar cell types (Purkinje cells, granule cells, and astrocytes) in vivo and subjected them to deep proteomic profiling [8]. Although a common core mitochondrial proteome was identified across these three cell types, approximately 175 proteins with mitochondrial annotations, representing nearly all mitochondrial functions, showed differential expression. Strikingly, pathway analysis revealed that astrocytic mitochondria metabolize long-chain fatty acids more efficiently than neuronal mitochondria, suggesting potential cell-type-specific mitochondrial specialization. The study also identifies regulator of microtubule dynamics protein 3 (Rmdn3) as a cell-type-specific regulator of endoplasmic reticulum-mitochondria contact in Purkinje cells [8].

In another recent study, Mosharov and colleagues developed a physical voxelization strategy that partitioned a frozen human coronal hemisphere section into 703 voxels at neuroimaging resolution (3 × 3 × 3 mm) [9]. In each voxel, they measured multiple mitochondrial phenotypes: OXPHOS enzyme activities, mitochondrial DNA (mtDNA) copy number and volume density, and mitochondrial-specific respiratory capacity. The resulting atlas reveals that the human brain harbors highly diverse mitochondrial phenotypes driven by both anatomical topology and local cell-type composition. Grey matter consistently displays higher mitochondrial content and respiratory capacity than white matter, and subcortical areas differ systematically from cortical ones. Single-nucleus RNA-sequencing of selected voxels enabled integration of mitochondrial biochemistry with cell-type composition, revealing that cell-type differences are a primary driver of the observed regional variation [9]. Furthermore, Sanchez-Contreras and colleagues profiled somatic mtDNA mutations across multiple tissues in young and aged mice using high-accuracy next-generation duplex sequencing [10]. They found that mutation burden increases with age in a highly tissue-specific manner. The brain, in particular, exhibits its own characteristic pattern of mutation accumulation. The spectrum of clonally expanding mutations differs from that of non-clonal variants, and reactive oxygen species (ROS)-linked mutation classes are not proportionally expanded with age, suggesting active removal rather than unrestricted accumulation. These data highlight the complexity of heteroplasmy in tissues with mixed cell-type composition, such as the brain.

Less is known about the cell-type-specific mitochondrial heterogeneity in nonneuronal cells in the brain. Mitochondria in astrocytes and oligodendrocytes are thought to support neuronal metabolism under certain circumstances, such as intense synaptic firing or glucose deprivation [7]. There are clear metabolic distinctions across brain cells: neurons are overall heavily OXPHOS-dependent; oligodendrocytes shift from OXPHOS in progenitor states to glycolysis upon maturation; astrocytes are metabolically flexible; and microglia toggle their metabolism between oxidative and glycolytic states during immune activation [11]. Mitochondrial diversity may underlie the neuronal cell-type-specific vulnerability observed in mitochondrial diseases such as Leigh syndrome [11]. These studies, along with others, reinforce the principle that mitochondrial molecular identity is a cell-type-specific property in the brain, and that brain mitochondria exhibit distinct molecular evolution across aging trajectories, cognitive functions, and disease. It confirms that the spatial distribution of energetic capacity reflects the brain’s functional architecture. Brain tissue heterogeneity, which complicates bulk analysis, underscores the need for cell-type-resolved approaches.

Ib. Diversity within the same brain cell.

The highly polarized nature of neurons creates distinct microenvironments that display differences in mitochondrial behavior and molecular identity. Mitochondrial abundance and positioning exhibit substantial cell-type- and compartment-specific variation in and this heterogeneity is correlated with regional activity and postsynaptic targets [12,13]. Mitochondria form an extensively elaborate and dynamic network in the somatodendritic region, yet are mostly discrete and highly mobile in the axon. At synapses, mitochondria are distinguishable by their unique morphological [14,15], proteomic [14,16], enzymatic [14,17], and Ca²⁺ handling features [18]. Mitochondria are larger and closer at presynapses than at postsynapses, and presynapses typically have a mitochondrion within one micron. Notably, Kenyon cells, which are important for learning and memory, have approximately one mitochondrion per two presynapses [13].

A recent study provides a systematic, in vivo, cell-type-specific comparison of somatodendritic and axonal mitochondrial proteomes across various neuron types, including functional validation [19]. The central finding is that axonal mitochondria are not defined by axon-specific proteins but are broadly depleted of proteins that support mtDNA expression, OXPHOS, and proteostasis relative to their somatodendritic counterparts. Integration with the RiboTag translatome and published protein half-life data reveals that local translation of short-lived proteins selectively counteracts proteome impoverishment for specific metabolic pathways [19]. Similarly, advances in synaptic molecular biology revealed that local translation at the synapse is critical for synaptic function and identity [20]. Thousands of mRNAs are actively trafficked to specific synapse types, where local translation shapes synaptic protein composition independently of the soma [20]. Supported by additional recent publications, axonal mitochondria are being redefined as primarily lipid-metabolizing rather than glucose-oriented organelles. Parallel work shows DDHD2-mediated fatty acid release as an axonal energy source [21], and evidence suggests that fatty acids and triglycerides are also important bioenergetic fuel reserves for synaptic function in the brain [22,23]. These findings echo other studies demonstrating differences in OXPHOS and mitochondrial vulnerability to oxidative damage at synapses compared to the soma [16,24,25].

In addition, via direct high-resolution imaging of neurons in culture and in vivo, mitochondria in the main axonal shaft were found to be largely depleted of mtDNA and defective in ATP synthesis [26,27]. By contrast, those stationed at branch points preferentially engage in mtDNA replication and transcription, accumulate nuclear-encoded mitochondria-targeted mRNAs, and are spatially linked to sites of nascent cytosolic peptide synthesis. This branch-point pool undergoes asymmetric division: one daughter retains mtDNA and remains at the branch point (self-renewal of the biogenesis-competent pool), while the other daughter is launched as a highly motile, mtDNA-negative organelle that disperses into neurites. This asymmetric genome partitioning rejuvenates the membrane potential of the mtDNA-rich parent mitochondrion [27].

These findings pose an interesting question: What is the function of axonal mitochondria lacking mtDNA? One possibility is presynaptic Ca²⁺ buffering, yet proteomic analysis reveals a paradox [19,26]. Axonal mitochondria have less MCU protein, the major component of the Ca²⁺-import channel in the inner mitochondrial membrane, but buffer Ca²⁺ better during overload [28,29]. These findings demonstrate that organelle function cannot be inferred from protein or RNA levels alone, and that stoichiometry of regulatory subunits (for example, EMRE in the MCU complex) as well as local compartment-specific cues may override abundance. Thus, further work to elucidate not only mitochondrial composition but also function across distinct brain cell types and neuronal subcompartments is warranted to understand mitochondria’s contributions to compartment- and cell-type-specific activities under both homeostatic and pathological conditions. Certain discrepancies in the reported metabolic and proteomic profiles across neuronal compartments likely stem from differences in experimental design, such as compartment definitions and proteomic workflows. These inconsistencies could be resolved through future studies employing directly comparable methodologies and matched neuronal models.

Ic. Diversity in time.

The classical model suggests that throughout neuronal development and maturation, precise positioning of mitochondria is required to maintain local ATP homeostasis. During neurite outgrowth, on-site mitochondria supply ATP for the extension and branching of neuronal axons and dendrites [30]. A recent study employing the Drosophila horizontal system (HS) visual interneurons forces a reconsideration of this model, showing that mature HS dendrites tolerate substantial mitochondrial depletion without functional impairment [31]. The authors propose that activity-dependent glycolytic regulation and phosphagen buffering (via the arginine kinase/phosphoarginine system) can substitute for OXPHOS in meeting acute dendritic ATP demands. This mirrors observations in mice, where postnatal MIRO1 knockout depletes cortical pyramidal dendrites of mitochondria for months before morphological defects emerge [32]. The results suggest that neuronal vulnerability to mitochondrial dysfunction is heavily dependent on both developmental timing and cell type, a principle with direct implications for understanding why only specific neuron populations degenerate in mitochondria-associated neurological diseases.

II. Mitochondrial transport and beyond: molecular diversity and adaptor moonlighting

IIa. Molecular engines powering mitochondrial movement.

The ability of neurons to distribute and move mitochondria is essential to maintain their exceptionally polarized axonal and dendritic processes and synaptic activity [33–37]. Mitochondria are delivered from the cell body to the terminus via microtubule-based, long-range transport [38–40]. In mammalian neuron axons, microtubules are uniformly aligned, with all (+)-ends pointing to the axonal terminus, while in dendrites, their polarities are mixed [41]. Microtubule motor-adaptor complexes serve as both engines and carriers, loading mitochondria onto microtubule tracks and driving their movement. To date, a suite of motors and mitochondrial adaptors has been characterized. However, the cell-type-specific activity of these proteins in the brain is less understood. Kinesin motors regulate movement toward the (+)-ends of microtubules. At least 14 families and 45 genes encoding kinesins have been identified in mammals so far [42–48]. Cytoplasmic dynein moves mitochondria toward the (−)-ends of microtubules [49]. The best-known mitochondrial motor adaptor complex is the KHC-Milton-MIRO complex. MIRO attaches to the outer mitochondrial membrane (OMM) via its C-terminal transmembrane (TM) domain [50]. MIRO’s N-terminal cytosolic domains bind to Milton (TRAK1/Milton-1/OIP106 and TRAK2/Milton-2/GRIF1) and kinesin heavy chain (KHC) [51–54]. Together, these complexes anchor mitochondria to microtubules. Importantly, kinesin light chain (KLC) appears dispensable for this complex [54]. The dynein complex can coordinate with kinesin’s activities on the same mitochondrion [49,55,56].

The ability of a neuron to halt mitochondria once they reach their final destinations in distal neurites is as important as moving them. Neuronal activity can unload mitochondria from microtubules by increased concentrations of intracellular Ca²⁺ ions, which bind the EF-hands of MIRO, leading to conformational changes of the KHC-Milton-MIRO complex and the dissociation of mitochondria from microtubules [52,57–60]. This mechanism is reversible and instantaneous, likely to rapidly meet the high energy demands required to maintain electric firing and buffer the heightened Ca²⁺ concentrations. In contrast to Ca²⁺-dependent mitochondrial arrest observed in cultured neurons, spontaneous and sensory-evoked Ca²⁺ transients did not alter mitochondrial motility in mature retinal ganglion-cell dendrites, suggesting that activity-dependent regulation of transport varies with neuronal cell type, subcellular compartment, developmental state, and experimental context [61]. Glucose also regulates mitochondrial positioning in neurons. O-linked N-acetylglucosaminyltransferase (OGT) O-GlcNAcylates Milton by adding uridine diphosphate N-acetylglucosamine (UDP-GlcNAc), a derivative of glucose [62,63]. O-GlcNAcylated-Milton next complexes with four and a half LIM domains protein 2 (FHL2), which traps mitochondria onto actin filaments [64]. This mechanism may represent how mitochondria seek and utilize the fuel. In mammalian axons, syntaphilin can dock mitochondria onto microtubules close to presynaptic boutons, thereby restricting their motility and supporting local energetic and calcium-buffering demands of synaptic activity [65,66].

IIb. Heterogeneity in mitochondrial transport across brain cells.

Although the molecular mechanisms underlying mitochondrial transport are likely shared across distinct cell types, a key question in the field is what accounts for the heterogeneity in transport among brain cell types. One challenge in answering this question is the technical complexity of live-imaging mitochondria, both in culture and in vivo, as well as the analytical methodologies that follow [67]. Recently, Niedermeier, Feyen and colleagues used MitoTag mice crossed to three Cre driver lines, labeling parvalbumin-expressing local inhibitory neurons (PV+), excitatory CamKIIα+ projection neurons, and noradrenergic locus coeruleus (LC) neurons, to perform in vivo two-photon imaging of axonal mitochondrial transport in awake adult mice via high-frequency 3D acousto-optic scanning [68]. The study challenges the previous view that over 90–99% of cortical mitochondria in adult animals are stationary (at P45-P120) [69,70]. Instead, the authors report a mobile fraction of 14–32% across different cell types, which is up to ten times higher than earlier findings [68]. They suggest that the previous discrepancy was due to undersampling caused by low acquisition rates. Strikingly, LC neuron axons exhibit substantially faster mitochondrial transport velocities than PV or CamKIIα neurons. Aggregated tau, an early pathological hallmark of Alzheimer’s disease (AD), impairs mitochondrial transport preferentially in LC neurons in vivo, providing mitochondrial motility-based evidence of selective neuronal vulnerability in AD [68]. This compelling demonstration of neuronal heterogeneity in mitochondrial movement sets an exciting stage for future investigations into the brain cell-type-specific molecular engines required for these sophisticated regulations.

IIc. Moonlighting motor-adaptor.

The OMM protein MIRO was first identified as a microtubule motor-adaptor that couples mitochondria to microtubules through interactions with Milton and KHC. We have found that MIRO can interact with additional proteins, but these interactions are highly context dependent. For example, upon mitochondrial depolarization, MIRO1 interacts with PINK1, PARKIN, and LRRK2, leading to its removal from the OMM and the initiation of mitophagy. This step is impaired in neurons from patients with Parkinson’s disease (PD) [71–74]. Furthermore, we have recently found that MIRO1 binds to ribosome subunits and organizes mitochondria-related protein biogenesis by spatially confining OMM-associated ribosomes, defining a new role for MIRO1 and opening a new area of investigation [75]. During oxidative stress, MIRO stabilizes its interaction with oxidized MIC60, an intramitochondrial redox sensor, exacerbating mitochondrial malfunctions in aging and neurodegenerative disease models [76]. MIRO1 can also bind to BAX, facilitating BAX macropore formation in the OMM and mtDNA leakage in models of glioma and PD-relevant neurons exposed to oxidative stress [77]. Intriguingly, the same MIRO1-BAX complex acts as a mitochondrial switch, dictating distinct cell death pathways. In glioma, MIRO1-dependent mtDNA leakage via BAX pores activates the cGAS-STING-pIRF3 signaling axis, promoting GPX4 expression and resistance to ferroptosis, whereas in PD neurons, the same pore triggers apoptosis [77]. We have developed a series of chemically diverse small molecules that bind the C-terminal GTPase of human MIRO1 and can allosterically disrupt aberrant MIRO1-BAX and MIRO1-MIC60 complexes in a non-competitive manner, leading to beneficial outcomes in models of neurodegeneration and glioma [71,77,78]. These findings highlight the value of targeting this domain for broad therapeutic interventions.

How are MIRO1’s multifaceted functions coordinated in different types of brain cells and within neuronal subcellular compartments? Does MIRO1 play a part in establishing neuronal mitochondrial diversity via its function in mitochondrial transport or via its newly discovered role in regulating mitochondria-related protein biogenesis? Can the same MIRO1 complex or signaling be regulated differently across neuronal compartments, giving rise to distinct mitochondrial behaviors? How are these functions involved in disease pathogenesis? The answers could help us better target cell-type-specific pathologies across neoplastic and neurodegenerative diseases.

Conclusions

Researchers have now collectively established that the neuron is not just an energetically demanding cell. Instead, it is an energetically differentiated cell, and this differentiation is encoded at every scale from the axon branch point to the brain-wide circuit (Fig. 1). Mitochondria across brain cell types, compartments, and neurite topologies differ in their proteomes, transport dynamics, mtDNA content, renewal mechanisms, and vulnerability to pathological protein aggregation. In parallel, synapses are diversified by active mRNA sorting that enables local translation to maintain synapse-type-specific protein stoichiometries independently of somatic gene expression. These findings converge on a new framework for thinking about neuronal identity and vulnerability: molecular diversity at the level of individual organelles and synaptic compartments is not incidental but is the principle upon which brain health and disease operate.

Despite rapid progress, fundamental questions remain. For example, how mitochondria at the remote neurite terminus maintain their quality or initiate stress responses, which typically require rapid trafficking of stress effectors between the nucleus and the mitochondrion. It is possible that mitochondrial stress signaling mechanisms not only diverge from non-polarized cells but also between the soma, dendrites, and axons to accommodate the unique spatial complexity of the neuron. Local translation of mediators of mtUPR, ISR, or mitophagy may be utilized to preclude the need to transport materials in and out of the distal cell body. Brain cell-type-dependent molecular mechanisms and signaling paradigms that link neurite topology to the positioning of biogenesis-competent mitochondria and the recruitment, movement, and retention of diverse mitochondrial populations (for example, mtDNA-rich or lipid-metabolizing) will be important to elucidate. Understanding and targeting mitochondrial diversity across cell types and subcompartments in the brain may be essential for developing therapies precisely matched to the neurons that matter most in a given disease.