Metabolic sensing and control in mitochondria

SUMMARY:

Mitochondria are membrane-enclosed organelles with endosymbiotic origins, harboring independent genomes and a unique biochemical reaction network. To perform their critical functions, mitochondria must maintain a distinct biochemical environment and coordinate with cytosolic metabolic networks of the host cell. This coordination requires them to sense and control metabolites and respond to metabolic stresses. Indeed, mitochondria adopt feedback or feedforward control strategies to restrain metabolic toxicity, enable metabolic conservation, ensure stable levels of key metabolites, allow metabolic plasticity and prevent futile cycles. A diverse panel of metabolic sensors mediates these regulatory circuits, whose malfunctioning leads to inborn errors of metabolism with mild to severe clinical manifestations. In this review, we discuss the logic and molecular basis of metabolic sensing and control in mitochondria. The past research outlined recurring patterns in mitochondrial metabolic sensing and control and highlighted key knowledge gaps in this organelle that are potentially addressable with emerging technological breakthroughs.

Summary

Mitochondria engage both feedback and feedforward control circuits to achieve the robust regulation of metabolism. A diverse array of molecular sensors monitors metabolite levels and fine-tunes enzymatic activities, ensuring swift adaptation to metabolic demand and effective protection against stress.

Mitochondria are essential organelles in all higher eukaryotes, supporting growth and viability through ATP production, biosynthesis and signaling functions. It is widely believed that mitochondria emerged during the early evolutionary history of eukaryotes from endosymbiotic alphaproteobacteria¹. The transition from an autonomous, facultatively aerobic pre-mitochondrial alphaproteobacterium to a permanently integrated organelle involves extensive integration of genomes, the evolution of protein and metabolite transport machinery, and coordination of metabolism between the endosymbiont and the host cell². These processes enable the formation of a distinct metabolic environment inside the mitochondria, which is critical for their proper functioning.

Central to metabolic compartmentalization is a highly regulated biochemical interface consisting of a selectively permeable mitochondrial inner membrane, dedicated metabolite transporters and asymmetrically distributed enzymes for the generation of chemical gradients. While it provides the cell improved metabolic versatility, rapid metabolic control and protection against toxic intermediates³, maintaining metabolic compartmentalization in mitochondria comes with its unique challenges. From a host cell-centric point of view, the host-mitochondria interface resembles a black-box computer program. For input, it accepts signals from the host reflecting the cellular nutrient and energetics status and metabolic demands; for output, it generates a two-pronged effect on cellular physiology: a metabolic output, including the consumption and production of metabolites; and a signaling output, transmitting metabolic stress, and malfunction sensed by the mitochondria that warrant additional responses by the host cell. Thus, functioning mitochondria must a) swiftly respond to the metabolic demands of the cell, b) generate proper metabolic output by producing an appropriate amount of energy and biosynthetic precursors and c) initiate correct signaling response that adequately reflects metabolic abnormalities encountered by the mitochondria. Any breach of these functional guardrails often triggers the adaptive response of the host to clear defective mitochondria via selective mitophagy or programmed cell death^4,5.

Sensing mechanisms for cellular metabolites have been documented for both prokaryotes and eukaryotes^6,7. This includes members of the alphaproteobacteria and Asgardarchaeota taxa^8,9, lineages that were predicted to give rise to modern eukaryotes and mitochondria. These mechanisms couple nutrient availability to not only the metabolic network but also translation and transcriptional machinery. While the mitochondria resemble their ancestors by maintaining an independent genome and a dedicated system for transcription and translation, there remains a knowledge gap on how mitochondria employ metabolic sensing and control mechanisms to fulfill their physiological functions. Previous research has highlighted some recurring themes in mitochondrial metabolic control. As discussed below, these principles underlie the wiring of the mitochondrial metabolic network and may inform future studies on unrecognized mechanisms for mitochondrial metabolic sensing and control.

Two strategies of mitochondrial metabolic control: homeostasis and adaptation

Like all biological systems, mitochondria must respond to changes in resources and adjust their metabolism accordingly. A unique feature of mitochondria is that their biochemical environment is already highly regulated by the host cell. This nested structure implies that an intracellular compartment could maintain a relatively stable internal environment by relying solely on the equilibration of metabolite transport. Therefore, mechanisms dedicated to mitochondrial metabolic sensing and control must serve a broader function beyond the homeostasis of intramitochondrial metabolites. As discussed below, requirements for metabolic sensing and control are derived from many aspects of mitochondrial function. Mitochondria are biosynthetic organelles whose output (such as the redox-active heme cofactor) may require tight regulation. They are also metabolically plastic and may adjust their metabolic pattern according to cellular demand. Additionally, some of the mitochondrial processes may have a stringent requirements for the internal environment distinct from the cytosol; for example, iron concentration in mitochondria is significantly higher¹⁰, likely in support of the assembly of iron-sulfur clusters. Finally, mitochondria are specifically the source or target of unique metabolic stresses. These features raise the possibility that metabolic sensing and control mechanisms that closely interact with mitochondria must be present to instruct the organellar function.

These functional necessities demand different logic for metabolite sensing. Mitochondria require a wide range of metabolic parameters to be sensed and processed to instruct the regulation of metabolic responses. While cellular parameters such as pH, osmotic pressure and cholesterol levels, could be sensed and controlled within a relatively narrow range, others (i.e. availability of oxygen and a subset of nutrients¹¹) are extrinsic in nature and cannot be regulated within an optimal range. Thus, metabolic processes reliant on these parameters need to be adjusted to prioritize the most essential physiological functions. Accordingly, we propose that mitochondria adopt two distinct circuits of metabolic control—feedforward and feedback control—to achieve either metabolic homeostasis or metabolic adaptation in response to these perturbations. These two strategies differ mainly in the coupling pattern between metabolic sensing and downstream effects. As elegantly reviewed by Ye and Medzhitov¹², both circuits require “sensors” that respond to metabolic perturbations, while only in feedback control circuits, the system exerts a direct effect on the sensed metabolic parameter retroactively. Oftentimes, changes in the “input” metabolic parameter stimulate a response in the system that counteracts such perturbation via a negative feedback mechanism. This circuit is instrumental for controlling the levels of metabolites within a narrow range for optimal physiological function. On the other hand, the feedforward control mechanism lacks such a homeostatic effector that attenuates perturbation; rather, it rewires the metabolic network to ensure adequate metabolic output in the presence of such perturbation. In some scenarios, these two circuits could act simultaneously, engaging downstream effector that both maintains homeostatic control and promotes adaptive response. For example, mitochondria-derived heme is sensed by the transcriptional repressor BTB Domain And CNC Homolog 1 (BACH1), which upon binding to heme derepresses the expression of the heme-degrading enzyme heme oxygenase-1¹³ in a classic feedback regulation mechanism. At the same time, Bach1 target genes also include redox regulators and iron storage proteins¹⁴ that mediate an adaptive response to increased iron released from heme degradation, thus forming an adaptive feedforward circuit. In mitochondria, metabolite sensing mechanisms engage both feedback and feedforward regulations, ensuring a proper mitochondrial internal environment and swift adaptation to cellular demand.

Mitochondria achieve homeostatic control of metabolism through feedback regulation

Feedback regulation is critical for effective metabolic regulation since it allows the metabolic flux to dynamically respond to demand while maintaining metabolite levels far from thermodynamic equilibrium¹⁵. This typically involves a sensing mechanism that responds to a perturbation in metabolic parameters and an effector to dampen the response. Without feedback inhibition, most enzymatic reactions would not swiftly adjust the metabolic flux in response to fluctuating demand. For mitochondrial metabolism, this feature is particularly desirable for restraining the toxicity of metabolite intermediates or avoiding resource-wasting through the synthesis of low-demand metabolites.

A. Feedback regulation restrains toxicity of mitochondria-produced metabolites

The metabolic versatility of mitochondria has made it an ideal compartment for the biosynthesis of not only building blocks for cell growth but also diverse cofactors and signaling molecules. These metabolites, while serving critical roles in biochemical catalysis or stress responses, often display intrinsic toxicity when produced at elevated levels. Therefore, mechanisms have evolved to limit their production and prevent excessive damage to mitochondrial functions. Given that many of these metabolites are produced via enzymatic actions within the mitochondrial matrix, the compartmentalization of their synthesis allows multiple layers of regulation, via tuning substrate import and export as well as enzyme expression, localization and activity. For example, itaconate is an important regulator of innate immunity produced primarily by myeloid cells during activation^16–18. The enzyme responsible for itaconate synthesis, ACOD1, resides in mitochondrial matrix¹⁹; once produced, itaconate exerts its anti-inflammatory effects partially by acting as a competitive inhibitor of succinate dehydrogenase. Additionally, as an electrophilic metabolite, itaconate alkylates thiol groups on target proteins such as Nuclear factor erythroid 2–related factor 2 (Nrf2) and triggers electrophile stress response²⁰. As an unrestrained electrophilic attack on thiolcontaining proteins is likely to disrupt cellular physiology, multiple downstream targets of itaconate function to restrain its overproduction or increase cellular buffering capacity against electrophilic stress. Nrf2 activation by itaconate leads to increased production of glutathione, a small molecule thiol that serves as the primary buffer against electrophiles and reduces itaconate content by forming adducts through its thiol group. It has also been proposed that itaconate modifies and inhibits the activity of mitochondrial isocitrate dehydrogenase (IDH2), therefore restraining the production of itaconate itself by reducing the supply of its precursor, isocitrate, via reductive carboxylation²¹.

Another interesting example of such regulation is the synthesis of heme, an essential cofactor produced via a multistep pathway that spans across the mitochondria and the cytosol. Despite the critical function of heme, free heme can participate in Fenton reactions that generate highly toxic hydroxyl radicals, which damage proteins and lipids and must therefore be tightly regulated²². In eukaryotes, control of heme synthesis occurs primarily at its rate-limiting enzyme, delta-aminolevulinate synthase (encoded by ubiquitously expressed ALAS1 and erythroid-restricted ALAS2). Multiple mechanisms act in concert to repress ALAS activity in response to heme accumulation. Heme suppresses ALAS1 levels by inhibiting the translation and enhancing the degradation of its mRNA^23,24. Moreover, heme binds ALAS1 via Cys-Pro motifs²⁵, inhibiting its mitochondrial import²⁶ and inducing the degradation of the protein via the mitochondrial protease CLPP/CLPX. These mechanisms ensure that free heme, as well as phototoxic porphyrin intermediates, are kept at a level that does not interfere with normal cellular physiology. Failure in the metabolic control for heme synthesis leads to porphyria, a family of conditions caused by the accumulation of toxic heme synthesis intermediates that damage the liver, kidney and nervous system²⁷. Loss-of-function mutations in multiple enzymes of the heme synthesis pathway have been found to cause porphyria, due to ineffective feedback inhibition on the rate-limiting enzyme ALAS. Much of the clinical symptoms could be attributed to intermediate toxicity rather than insufficient heme production, as demonstrated by several cases of liver transplantation from donors with porphyria²⁸. Such clinical observations highlight the essential functions of these feedback mechanisms in limiting the metabolite toxicity by maintaining the intermediates within the range required for normal and mitochondrial functions.

B. Feedback regulation enables the conservation of metabolites consumed in mitochondria

Feedback metabolic control in mitochondria also contributes to metabolic efficiency by restricting the production of metabolites required at a stable, but relatively low levels. Without feedback regulation, producing such metabolites at a near-equilibrium rate would consume an unnecessarily large amount of energy or biosynthetic precursors. For example, coenzyme A is an essential cofactor for delivering acetyl and acyl groups in mitochondria. Synthesis of coenzyme A starts with Pantothenate kinase isozymes, among which PANK2 is predicted to localize in mitochondrial intermembrane space^29,30. This highly conserved pathway consumes four ATP molecules and cysteine per molecule of coenzyme A. Intracellular concentration of CoA species is estimated to be at low micromolar levels in the cytosol and millimolar levels in mitochondria³¹, and because CoA is used catalytically in most reactions, the demand to replenish CoA pool is likely modest, as suggested by the relatively long turnover time of CoA in mouse tissues compared to other cofactors^32,33. Thus, it may be desirable to limit the production of coenzyme A once its biological demand is satisfied. This is achieved through a common feature of all PANK enzymes: from E.coli to mammalian cells, these enzymes are all allosterically inhibited by CoASH, acyl and acetyl CoA^34,35. The high conservation of this regulatory mechanism likely suggests a significant fitness benefit for keeping CoA synthesis in check. Importantly, although PANK isozymes are also present in the cytosol, mutations in the mitochondrial PANK2 isoform led to Pantothenate kinase-associated neurodegeneration (PKAN), indicating that mitochondrial localized, regulated production of Coenzyme A serves a unique physiological function.

C. Feedback regulation ensures robust homeostatic control of metabolite levels in mitochondria

In principle, mitochondria could utilize feedback control mechanisms to achieve a stable chemical environment, particularly for metabolites that are synthesized or utilized only in mitochondria, or for metabolites that require stable mitochondrial concentrations. Similar mechanisms for regulating redox potential or pH have been reported for other organelles^36,37, but relevant studies for mitochondrial metabolites are scarce. One example is the synthesis of cardiolipin, the lipid species with exclusively mitochondrial distribution. The enzymatic pathway for cardiolipin synthesis resides primarily on the inner membrane of mitochondria, while its precursor phosphatidic acid (PA) needs to be transferred from the endoplasmic reticulum and the outer membrane by lipid transport proteins³⁸. Intriguingly, in S. cerevisiae, transport of PA between the inner and outer membrane by Ups1 is subjected to feedback regulation by cardiolipin itself. Accumulation of cardiolipin in the mitochondrial inner membrane traps Ups1 and limits its shuttling between membranes, effectively shutting down the inflow of PA to the cardiolipin synthesis pathway³⁹. This elegant mechanism likely functions to control the levels of cardiolipin in the mitochondrial inner membrane near its optimal range. Given the unique chemical structure of cardiolipin, such a regulatory mechanism is essential for the structural and functional integrity of mitochondria.

A comparable example is glutathione homeostasis in mitochondria. It has been long observed that following systemic inhibition of glutathione synthesis, the decline in mitochondrial glutathione levels displays much slower kinetics compared to cytosolic glutathione⁴⁰. It appears that the mitochondrial transport for glutathione is wired in a way that prioritizes the maintenance of the mitochondrial glutathione pool over the cytosolic pool. Recent studies have identified SLC25A39/SLC25A40 as candidates for mitochondrial glutathione importers^41,42. Remarkably, SLC25A39 displays a classic feedback regulation behavior as it accumulates dramatically upon pharmacological inhibition of glutathione synthesis⁴¹. Although the molecular mechanism remains elusive, these observations hint at the intriguing possibility that SLC25A39 may be part of the machinery that maintains mitochondrial glutathione levels via a feedback homeostatic mechanism. The SLC25 mitochondrial transporter family plays a pivotal role in shaping intramitochondrial metabolite pools⁴³; since it contains many members with unannotated functions, it is likely that many of them participate in homeostatic regulation via unappreciated regulatory circuits.

Mitochondria adapt to metabolic perturbations through feedforward regulation

Although a recurring theme in metabolic regulation, homeostatic control via feedback regulation is not sufficient for the optimal functioning of mitochondrial metabolism. This is particularly relevant for perturbations and stresses that are exogenous in nature, such as physical and chemical stresses, or stochastically occurring damage of organelles. Regulatory mechanisms in these scenarios typically function as feedforward circuits that drive an adaptive response from the cell. Similar to feedback regulation, these mechanisms involve sensing of metabolic cues but do not maintain homeostatic levels of a given metabolite under perturbations. Instead, mitochondria rely on these mechanisms to ensure efficient utilization of resources while these perturbations persist.

A. Feedforward regulation allows metabolic plasticity of mitochondria

Mitochondria display remarkable metabolic plasticity that funnels biochemical intermediates into divergent fates under different metabolic states. This is achieved via tight coupling between metabolic sensing mechanisms and enzymatic activities, best exemplified by the regulation of the urea cycle. As urea production is essential for the clearance of nitrogen-containing metabolic waste, organisms must dynamically respond to the demand of protein catabolism^44,45, which fluctuates with food uptake and cannot be physiologically maintained within a narrow range. Therefore, a set of feedforward regulatory mechanisms has evolved to adjust the flux through the urea cycle in response to the demand for nitrogen waste disposal. Mitochondria harbors the rate-limiting enzyme of the urea cycle Carbamoyl phosphate synthetase I (CPS1), where nitrogen enters the urea cycle in the form of ammonia. First, this enzyme is regulated by an obligate allosteric activator, N-acetylglutamate⁴⁶, a metabolite produced in the mitochondria by N-acetylglutamate synthase (NAGS)^47,48. Arginine, an intermediate of the urea cycle, stimulates the activity of NAGS and leads to the higher activity of the urea cycle⁴⁷. Together, these mechanisms allow the urea cycle activity to respond to fluctuations in ammonia at enhanced sensitivity for its efficient clearance. Notably, NAGS mutations that abolish the stimulatory effect of arginine⁴⁹ cause late-onset hyperammonemia⁴⁹. Much of this regulatory mechanism is spatially constrained in the mitochondria, which creates a dedicated compartment for the accumulation of N-acetylglutamate and its allosteric regulation of CPS1 activity. Since N-acetylglutamate is produced in the mitochondria and degraded in the cytosol⁵⁰, compartmentalization allows an additional layer of regulation. Indeed, it has been suggested that glucagon increases mitochondrial N-acetylglutamate levels potentially by inhibiting its mitochondrial efflux^51,52, a poorly understood mechanism. These findings suggest that the confinement of CPS1 and its allosteric activator N-acetylglutamate in the mitochondrial matrix is critical for effective control of urea cycle activity.

A parallel example of the adaptive activation of a metabolic pathway is the unique molecular feature of PANK2, a mitochondrially-localized enzyme catalyzing the rate-limiting step of coenzyme A synthesis. As described above, all PANK isozymes display feedback inhibition by CoASH and acetyl-CoA, yet PANK2 seems to be highly sensitive to acetyl-CoA. Indeed, under physiological conditions, it is predicted to have minimal catalytic activity⁵³. This paradoxical behavior is resolved by the observation that this inhibition is released in the presence of acylcarnitine, the substrate of beta-oxidation. Since CoA is required for beta-oxidation reactions, the accumulation of acylcarnitine serves as a proxy for CoA limitation, and PANK2 responds to this signal by increasing the supply of CoA precursors, ensuring efficient catabolism of fatty acid chains.

B. Feedforward regulation prevents futile cycles

Substrate cycles, or futile cycles, consist of antiparallel chemical reactions that collectively do not result in the net conversion of substrates to products. These loop reactions could be desirable for processes such as thermogenesis; additionally, they form important nodes of metabolic regulation because a relatively minor change in forward or backward reaction could lead to a significant shift in the directional metabolic flux⁵⁴. In mitochondria, it was proposed that the forward and reverse reactions catalyzed by NAD⁺- and NADP⁺-dependent isocitrate dehydrogenases form a substrate cycle. Electron from reduced NADH is then transferred to NADP⁺ by nicotinamide nucleotide transhydrogenase (NNT), coupled to proton translocation across the mitochondrial inner membrane⁵⁵. This mechanism thus allows the finetuning of TCA cycle activity and particularly in response to proton motive force generated by the ETC complexes.

While these substrate cycles afford cells metabolic flexibility, excessive futile cycling needs to be carefully controlled to avoid unnecessary loss of energy. Metabolic compartmentalization by mitochondria provides a structural basis for separating forward and reverse biochemical reactions. Additionally, mechanisms that sense metabolic demands and switch between forward and reverse reactions must be integrated into mitochondrial metabolic networks, ensuring their efficiency and plasticity. These regulatory mechanisms often involve key enzymes at the mitochondria-cytosol interface that restrain the access of substrates to a certain compartment. A classic example is the regulation of β-oxidation and fatty acid biosynthesis via carnitine palmitoyltransferase (CPT1) by malonyl-CoA⁵⁶. As the entry site of most long-chain fatty acid to β-oxidation, CPT1 activity is allosterically inhibited by the substrate of fatty acid synthesis, malonyl-CoA. When the demand for energy is low, access of the mitochondrial beta-oxidation machinery to fatty acid is limited, as acetyl-CoA is shunted towards malonyl-CoA and blocks fatty acid entry into mitochondria through allosteric inhibition of CPT1^57–59. In contrast, when energy demand is high and AMPK is activated, fatty acid synthesis is shut down by the phosphorylation of acetyl-CoA carboxylase, reducing malonyl-CoA levels and releasing the inhibition on CPT1^60,61, thereby switching the direction of lipid metabolism to β-oxidation. Futile cycling from simultaneous fatty acid synthesis and catabolism is thus prevented jointly by a) segregating the forward and reverse reactions by the mitochondrial inner membrane and b) excluding the substrate access to compartmentalized enzymes when the opposite reaction is proceeding.

C. Feedforward regulation enables adaptive responses to extrinsic metabolic stress

The structural complexity of mitochondrial metabolic machinery dictates that numerous forms of biological stress could significantly limit its functioning. Cells must initiate appropriate responses to adapt to these environmental perturbations ranging from nutrient and oxygen deprivation to noxious chemical stimuli and infection. Mechanisms for metabolic regulation under these scenarios are akin to feedforward regulation illustrated above and typically trigger downstream effectors that systematically shift the metabolic status of the cell in response to mitochondria-derived stress.

Mitochondria dysfunction triggers the integrated stress response (ISR) and suppresses the mechanistic Target Of Rapamycin Complex 1 (mTORC1) activity, turning down the synthetic pathways for proteins and other building blocks as an adaptation to resource limitation. These pathways are wired in a fashion that senses functional defects triggered by diverse metabolic stresses. Short-term blockade of the mitochondrial electron transport chain activates AMP-activated protein kinase (AMPK) and acutely suppresses mTOR1 signaling, likely via sensing of decreased ATP/AMP ratio⁶². Extended inhibition of the mitochondrial electron transport chain, which disrupts membrane potential-coupled protein import, engages the integrated stress response via a mitochondrial matrix-targeted protein DAP3 binding cell death enhancer 1 (DELE1)^63,64, which is normally processed in the mitochondrial matrix but gets cleaved by the metalloendopeptidase OMA1 in the intermembrane space when its import is inhibited⁶⁵. Cleaved DELE1 triggers the integrated stress response by activating the eukaryotic translation initiation factor 2A (eIF2a) kinase heme-regulated inhibitor (HRI). Through a parallel mechanism, stresses that perturb the integrity of the mitochondrial outer membrane engage HRI to activate the integrated stress response via the release of cytochrome C from the intermembrane space⁶⁶. Expression of stress response genes downstream of ISR effector activating transcription factor 4 (ATF4) consequently enable adaptation to metabolic stress. In all these cases, the metabolic compartmentalization of mitochondria is instrumental for stress sensing and metabolic control. The general principle is that these compartments create segregated spaces for sequestering signaling components that are susceptible to perturbations. The compromised integrity of these compartments leads to the mislocalization of signaling components, which are interpreted by the cell as a proxy for metabolic stress. Interestingly, this paradigm is adopted by cells to clear defective mitochondria via selective mitophagy⁵, respond to mitochondrial damage by sensing mitochondrial DNA via the cyclic GMP–AMP synthase (cGAS)- stimulator of interferon genes (STING) pathway, and adapt to protein misfolding stress in the mitochondrial by activating transcription factor associated with stress (ATFS1)-mediated mitochondrial unfolded protein response (mtUPR) pathway⁶⁷. Regulation and functions of these pathways have been extensively discussed before^5,67,68.

Molecular basis of mitochondrial metabolic sensing

In metabolic control, the term “sensor” could be loosely interpreted to encompass any protein that a) physically associates with a metabolite or its derivative and b) changes its behavior according to the varying levels of that metabolite within its physiological range. This definition does not preclude the existence of additional sensing mechanisms that do not directly involve a protein component or protein-metabolite interaction (RNA-based riboswitches are classic examples)⁶⁹; but even this narrower category of metabolic sensors contains remarkable mechanistic diversity. The segregation of the metabolite pool of the mitochondrial matrix by the mitochondrial inner membrane plays a central role in metabolic sensing and control³. As specified below, this barrier serves as a structural basis for maintaining a distinct biochemical environment and restrains futile cycling by physically segregating antiparallel reactions. It restricts the access of reactive or toxic metabolites to the cytosol, enabling robust regulation of their production while limiting potential damage of cellular components. It also creates gradients of proton and electromotive force and allows the sequestration of signaling molecules such as cytochrome C and DELE1; upon metabolic stress that disrupts the integrity of this barrier, the release of these factors could be sensed as proxies for metabolic stress. Many of the sensing mechanisms described here either rely upon the intact mitochondria-cytosol barrier for robust activation or indirectly sense the defects of this barrier itself. We hereby describe four molecular mechanisms recurrently observed in mitochondrial metabolic control.

A. Sensing metabolites via allosteric regulation of mitochondrial enzymes

One of the most common mechanisms for metabolite sensing is the allosteric regulation of mitochondrial enzymes and transporters, allowing rapid tuning of biochemical fluxes in response to diverse metabolic cues⁷⁰. Factors sensed via allosteric regulation range from amino acids, lipids, carbohydrates and metabolic intermediates to metals and cofactors; in many cases, it allows the integration of multiple metabolic signals via the activity of a single enzyme. A well-studied example is the regulation of glutamate dehydrogenase (GDH) by GTP, NADH, leucine, Mg2+ and other metabolites. An important route of glutaminolysis, glutamate dehydrogenase has been shown by early studies to bind both ADP and GTP on different sites⁷¹. GTP inhibits the activity of the enzyme⁷² while ADP exerts an activating effect; an increased ADP/GTP ratio thus signals a low-energy status in mitochondria that demands the replenishment of tricarboxylic acid (TCA) cycle intermediates via the activity of glutamate dehydrogenase. Additionally, leucine potently activates the enzyme⁷³ via binding at the subunit interface of the hexameric enzyme complex⁷⁴; the physiological importance of this regulation is underlined by the finding that leucine and its analogs stimulate insulin secretion through allosteric activation of glutamate dehydrogenase⁷⁵. Perhaps most strikingly, pathogenic mutations have been identified in human glutamate dehydrogenase enzyme that specifically abolishes the allosteric inhibition by GTP⁷⁶. These mutations lead to a gain-of-function effect on the GDH enzyme and hyperactive insulin secretion in beta cells⁷⁷. Additionally, they increase hepatic ammonia production while restricting urea cycle capacity by reducing available glutamate for the synthesis of N-acetylglutamate, an allosteric activator of the urea cycle rate-limiting enzyme CPS1. The combination of these effects leads to infant and childhood hyperinsulinism and hyperammonemia syndrome, with the elevation of plasma ammonium levels and recurrent episodes of hypoglycemia, highlighting the essentiality of allosteric regulation of GDH for maintaining organismal metabolic homeostasis.

B. Sensing metabolites via protein post-translational modifications

Post-translational modifications of mitochondrial proteins may occur enzymatically or non-enzymatically. These modifications create molecular platforms for sensing the availability of diverse metabolites⁷⁸. Of note, the distinctive chemical environment in the mitochondrial matrix is essential for shaping the target spectrum of such modifications. For example, high pH in the mitochondrial matrix favors the deprotonation of cysteine and lysine residues, which are critical for their reactivity. Indeed, non-enzymatic modifications of both cysteine and lysine are involved in sensing metabolic cues ranging from ROS to TCA cycle intermediates^79,80.

The widespread distribution of reactive lysine residues renders a significant portion of mitochondrial proteome susceptible to acetylation by acetyl-CoA⁸¹, mostly via non-enzymatic reactions⁸². These modifications modulate the activities of diverse enzymes involved in the TCA cycle, electron transport chain, fatty acid and carbohydrate metabolism while providing interfaces for metabolic regulations. For example, acetylation of long-chain acyl-coenzyme A dehydrogenase (LCAD) on K42 reduces its enzymatic activity by approximately 40%⁸³, thus potentially functioning as a coupling mechanism between cellular energetic state and fatty acid catabolism. In mitochondria, the non-enzymatic acetylation is reversed by NAD⁺-dependent, enzymatic deacetylation catalyzed by SIRT3, a member of the Sirtuin deacylation enzymes. Modulation of metabolic flux controlled by SIRT3 activity might be particularly important in adaptation to stress conditions such as starvation and hypoxia^84,85. Similar to SIRT3, SIRT4 and SIRT5 also reside in mitochondria where they catalyze the removal of negatively charged lysine acyl modifications^86,87. Together these post-translational modification mechanisms allow responses to different environmental perturbations and coordinate cellular metabolism via diverse downstream targets^88,89.

Another intriguing example is the S-glutathionylation of protein cysteine residues, which have been shown to modulate the activity of diverse targets ranging from respiratory complexes to uncoupling proteins^90,91. In mitochondria, a dedicated enzyme, glutaredoxin 2, catalyzes the reversible glutathionylation of thiol proteins. The thiol groups on both protein cysteine residues and glutathione could form disulfide bonds, and glutaredoxins catalyze protein S-glutathionylation by facilitating the disulfide bond exchange from oxidized glutathione (GSSG) to the protein-glutathione mixed disulfide bond. Remarkably, for proteins with two adjacent thiol groups capable of forming intra-molecular disulfide bonds, thermodynamic equilibrium between reduced and oxidized protein is predicted to be sensitive to glutathione levels⁹². Thus, in the case of S-glutathionylation, reversible modification by a single enzyme at equilibrium is sufficient to translate the availability of a given metabolite into altered protein behavior.

C. Sensing metabolites via the availability of cofactors, co-substrates and coenzymes

The mitochondria harbor multiple pathways for the synthesis, trafficking and functioning of coenzymes and cofactors, ranging from lipoic acid, iron-sulfur clusters to Coenzyme Q and heme. These processes offer effective leverages for metabolic control as their synthesis is often sensitive to the availability of metabolites. In many cases, reactions for the synthesis of these cofactors reside exclusively in mitochondria and would therefore be uniquely sensitive to fluctuations in mitochondrial metabolite levels.

The mitochondria are the primary location for the synthesis of iron-sulfur clusters⁹³, which are essential cofactors involved in redox catalysis and the structural integrity of numerous protein complexes. Unsurprisingly, mitochondria also harbor perhaps the largest pool of intracellular iron, accounting for up to more than 50% of total cellular iron content⁹⁴. While iron is indispensable for cellular metabolism, unbuffered free iron participates in Fenton reactions that produce reactive hydroxyl radicals. These toxic byproducts damage proteins, DNA and lipid bilayers and trigger cell death through ferroptosis. Thus, iron levels must be sensed and kept within a narrow range. In mammalian cells, two iron sensor proteins, iron regulatory protein 1 (IRP1) and regulatory protein 2 (IRP2), respond to iron levels and regulate the translation of proteins involved in iron metabolism via binding to the untranslated regions of their mRNAs⁹⁵. The synthesis of iron-sulfur cluster cofactor plays a central role in both IRP1 and IRP2: for IRP1, an iron-sulfur cluster blocks its association with cognate mRNAs; iron deficiency prevents the incorporation of this iron-sulfur cluster to the apoprotein and releases it for the binding to the iron-responsive elements (IREs) on mRNAs. For IRP2, iron regulates its degradation through the E3 ligase F-box and leucine rich repeat protein 5 (FBXL5); an iron-sulfur cluster at the FBXL5-IRP2 interface mediates their association and also confers this system sensitivity to oxygen⁹⁶. The reliance of iron sensing machinery on mitochondrially produced iron-sulfur cluster prosthetic groups orients whole-cell iron metabolism towards the demand for iron in the mitochondria, underscoring the critical role of this subcellular compartment in iron sensing and utilization. Additionally, iron-sulfur clusters are involved in control of mitochondrial-localized heme synthesis: two enzymes in the heme synthesis pathway, aminolevulinic acid dehydratase (ALAD) and ferrochelatase (FECH), have been shown to harbor iron-sulfur clusters as an essential prosthetic group for their action^97,98. Of note, the iron-sulfur cluster is absent in the homologs of FECH in lower eukaryotes, indicating that this mechanism might be selected for its ability to coordinate heme backbone synthesis with iron availability.

A similar principle applies to the regulation of ETC assembly by mitochondrial fatty acid biosynthesis. Mitochondria possess an independent pathway, the mitochondrial fatty acid synthesis (mtFAS) pathway, to produce short-chain fatty acid primarily for the synthesis of lipoic acid, an essential cofactor for numerous mitochondrial enzymes⁵⁰. This pathway produces long acyl chains attached to the acyl carrier proteins (ACP). The acylated ACP interacts with key ETC assembly factors and is predicted to be essential for their biogenesis⁹⁹. As the mitochondrial fatty acid synthesis is believed to be sensitive to the limitation of acetyl CoA, this mechanism enables the cells to detect acetyl CoA levels using cofactor production (lipoic acid and acyl-ACP) as a proxy, and control downstream metabolism via ETC assembly and lipoic acid-containing enzyme activities.

An intriguing mechanism by which TCA cycle intermediates are sensed involves alpha-ketoglutarate-dependent dioxygenases, a versatile group of iron-containing enzymes that includes key players in epigenetic regulation¹⁰⁰, oxygen sensing, lipid metabolism, and other critical processes. These enzymes couple the decarboxylation of alpha-ketoglutarate with the oxidation of the substrate¹⁰¹, and in many case the predicted K_M of those enzymes to alpha-ketoglutarate (αKG) overlaps with its physiological levels¹⁰², suggesting that their activity may dynamically respond to intracellular αKG levels. Indeed, it has been shown that these enzymes sense changes in cellular αKG¹⁰³ and mediate the effect of altered mitochondrial metabolism on epigenetic landscape and cell fate¹⁰⁴. Importantly, alpha-ketoglutarate-dependent dioxygenases are also competitively inhibited by TCA cycle-derived metabolites fumarate, succinate and oncometabolite 2-hydroxyglutarate produced by mutant isocitrate dehydrogenases (IDH1/2); these mitochondrial metabolites accumulate under environmental stress (such as hypoxia¹⁰⁵ and inflammation¹⁰⁶) or oncogenic mutations¹⁰⁷ and trigger profound changes in metabolic and epigenetic profile of the cell.

D. Sensing metabolites via the physical state of the mitochondrial membrane

A unique feature enabled by a compartmentalized organelle is the rich variation in membrane physical and chemical properties that serves as the basis for sensing diverse metabolic cues. Fluctuations in mitochondrial membrane lipid composition leads to changing thickness, curvature and asymmetry of membranes and alter their interaction with membrane proteins, leading to conformational changes that could be captured by downstream signaling pathways.

Carnitine Palmitoyltransferase 1A (CPT1A) is a mitochondrial outer membrane-anchored protein that serves as an important checkpoint for the entrance of fatty acid into beta-oxidation. Apart from being allosterically regulated by malonyl-CoA, CPT1A also harbors an N-terminal domain that adopts different conformation in response to changes in membrane curvature^108,109. This conformational change modifies the sensitivity of CPT1A to the regulation of malonyl-CoA, creating an additional layer of regulation that may reflect the long-term changes in the metabolic state. The chemical composition of the lipid bilayer could also directly impact its physical property and engage relevant sensors. Cardiolipin is a phospholipid species enriched specifically in the mitochondria and essential for its function^110,111. The last step of cardiolipin production is catalyzed by Tafazzin, an unspecific phospholipid-lysophospholipid transacylase that leads to the enrichment of linoleoyl hydrocarbon chains in cardiolipin. Sensing of substrate saturation, however, is achieved through the preferential catalysis of Tafazzin in the negatively curved membranes, which are enriched in unsaturated phospholipid and cardiolipin^112,113. Inferring metabolic state via membrane physical state appears to be a recurring theme coopted for metabolic signaling in organelles.

Outlook

Decades of research have unveiled the pivotal role of metabolic sensing and regulation in the mitochondria. These findings have far-reaching implications for our understanding of cellular metabolism and human disease, as numerous inborn errors of metabolism have been attributed to defects of metabolic sensing and control mechanisms involving mitochondria. Clinical manifestations of these diseases underscore the significance of intact metabolic control mechanisms that ensure efficient metabolic output while adapting swiftly to environmental perturbation.

In recent years, renewed interest in metabolic compartmentalization has led to the deorphanization of numerous components in organellar metabolite transport and regulation^{41,42,114–116}, aided by an array of novel technologies. Among them include improved workflows for the isolation of organelles^117–120, many of which shorten the time interval between cell lysis and extraction, allowing better preservation of metabolome and proteome for downstream analysis. Some workflows involve labeling the organelle of interest with genetically-encoded tags, which enables cell type-specific isolation of organelles when applied to animal models. In parallel, an expanding toolkit based on proximal labeling enables the compartment-specific isolation of proteins without isolation of organelles themselves¹²¹. Modification of this strategy also enables profiling of proteins with high temporal resolution¹²², integration with cysteine chemical proteomics¹²³ and characterization of compartments that cannot be isolated with conventional methods¹²⁴.

Metabolite-protein interaction is central to sensing of metabolic signals and remains challenging to reliably study. New platforms such as Mass spectrometry Integrated with equilibrium Dialysis for the discovery of Allostery Systematically (MIDAS) utilize purified proteins to achieve the enrichment of protein-binding metabolites and enables metabolome-wide profiling given any single protein of interest¹²⁵. Conversely, for a given metabolite, association with its protein binding partners induces altered protein stability, solubility and protease accessibility, which would be profiled at whole proteome-scale by quantitative mass-spectrometry^126–128. These new methods provide unprecedented opportunity for the systemic discovery of potential metabolic sensors and signaling components.

These advances reveal exciting new directions of research toward understanding the mechanistic logic of metabolic regulation in organelles. The mitochondria harbor distinct biochemical environments that do not automatically equilibrate with the cytosol, yet compared with our understanding of nutrient sensing and signaling in the cytosol ⁶, parallel mechanisms in the mitochondria are still understudied. Below, we listed some questions that need to be addressed in this area:

• Which mitochondrial metabolites require homeostatic regulation to maintain a relatively stable concentration distinct from the cytosol? What are the sensing and control mechanisms underlying such regulation?

• Metabolite transporters on the mitochondrial inner membrane could function as control sites for mitochondrial metabolites. Indeed, for substrates such as glutathione and iron, evidence suggests that they participate in post-translational regulation of their transporters^41,129. What is the mechanism coupling metabolite sensing to mitochondrial import?

• Mechanisms for retrograde signaling from the mitochondria to the cytosol have been described from yeast to humans that trigger a multitude of effects through nuclear gene expression. However, much less is known about how mitochondrial transcription and translation respond to the metabolic status of the cytosol. How are metabolic signals and stresses relayed to the mitochondria from the cytosol or other organelles?

Addressing these questions would require the concerted effort from researchers at the intersection of enzymology, mitochondrial biology, cell signaling and physiology to push the frontier of our understanding of this awe-inspiring symbiosis between mitochondria and host eukaryotic cells.

Figure 1. Mitochondria adopt feedback and feedforward control circuits to achieve metabolic homeostasis or adaptation.

(A) Schematic of a mitochondrial feedback circuit that ensures metabolic conservation by limiting the synthesis of metabolites demanded in mitochondria. PANK2, a mitochondrial enzyme in the Coenzyme A synthesis pathway, is allosterically inhibited by CoA and acetyl-CoA. (B) Schematic of a mitochondrial feedback circuit dedicated to maintaining the mitochondrial levels of a metabolite. Glutathione has been observed to downregulate its mitochondrial importer SLC25A39, functioning in a feedback mechanism likely to achieve homeostatic regulation of mitochondrial glutathione levels. (C) Schematic of a mitochondrial feedback circuit that restrains the production of toxic mitochondrial metabolite. Heme inhibits the import of the rate-limiting enzyme in its de novo synthesis, ALAS1/ALAS2, to avoid the accumulation of toxic porphyrin intermediates. (D) Schematic of a mitochondrial feedforward circuit that enables metabolic plasticity. Urea cycle intermediate arginine stimulates the synthesis of N-acetylglutamate, an allosteric activator of urea cycle enzyme CPS1, allowing robust activation of the urea cycle upon the influx of ammonium nitrogen. (E) Schematic of a mitochondrial feedforward circuit that prevents futile cycles. Fatty acid synthesis substrate malonyl-CoA inhibits the entrance of fatty acid into the reverse reaction, β-oxidation, by allosterically inhibiting CPS1. (F) Schematic of mitochondrial feedforward circuits that trigger adaptive responses to stress. The release of mitochondrial DNA or cytochrome C triggers stress response signaling via the cGAS-STING pathway or the integrated stress response (ISR).

Figure 2. Molecular basis of metabolite sensing in mitochondria.

Schematics for different mechanisms by which metabolites alter the properties of a sensor protein. (A) Metabolites could directly interact with metabolite sensors and modify their behavior through allosteric regulation or post-translational modification. For example, GTP allosterically inhibits glutamate dehydrogenase, which connects the mitochondrial energy state with the replenishment of TCA cycle intermediates. (B) Metabolites could affect enzymatic activity by serving as substrates of protein post-translational modifications. The long-chain acyl-CoA dehydrogenase responds to mitochondrial acetyl-CoA levels via acetylation of residue K42, a reversible modification that reduces its enzymatic activity. (C) Metabolites could modulate the activity of sensor proteins by affecting the incorporation of cofactors and prosthetic groups into apoproteins. Human aminolevulinic acid dehydratase requires an iron-sulfur [4Fe-4S] cluster for its activity, coordinate heme intermediate synthesis with iron availability. (D) The physical properties of mitochondrial membranes are sensitive to their chemical composition, providing a platform for membrane-bound sensor proteins to respond to metabolic stimuli. The cardiolipin-remodeling enzyme Tafazzin preferentially incorporates unsaturated acyl chains into cardiolipin through activation by negative membrane curvature.