Principles and functions of metabolic compartmentalization

Abstract

Metabolism has historically been studied at the levels of whole cells, whole tissues and whole organisms. As a result, our understanding of how compartmentalization—the spatial and temporal separation of pathways and components—shapes organismal metabolism remains limited. At its essence, metabolic compartmentalization fulfils three important functions or ‘pillars’: establishing unique chemical environments, providing protection from reactive metabolites and enabling the regulation of metabolic pathways. However, how these pillars are established, regulated and maintained at both the cellular and systemic levels remains unclear. Here we discuss how the three pillars are established, maintained and regulated within the cell and discuss the consequences of dysregulation of metabolic compartmentalization in human disease. Organelles are increasingly emerging as ‘command-and-control centres’ and the increased understanding of metabolic compartmentalization is revealing new aspects of metabolic homeostasis, with this knowledge being translated into therapies for the treatment of cancer and certain neurodegenerative diseases.

Humans ingest, metabolize or encounter more than 200,000 metabolites¹. Metabolite classes include peptides, lipids, amino acids, nucleic acids, carbohydrates and minerals found in the diet, as well as food additives, drugs, cosmetics, contaminants and pollutants incorporated from our modern life. A complex network of metabolic reactions removes unwanted or toxic substances and ensures adequate levels of energy and building blocks in a dynamic environment. Underscoring the importance of metabolic reactions in cellular and organismal fidelity, it is estimated that more than 30% of human genes are involved in metabolism (Human Metabolome Database 5.0), accounting for ~3,000 possible chemical reactions². Defects in these pathways or their regulation can result in human disease, with inborn errors of metabolism thought to underlie over 1,300 disorders^3–5. The vast complexity of human metabolism necessitates a high degree of organization to ensure efficiency through proper allocation of resources. A key organizational principle found throughout biology and that is tightly linked to the emergence of life is compartmentalization. In the context of metabolism, compartmentalization is a prerequisite for subdividing complex metabolic tasks into pathways amenable to regulation⁶. Compartmentalization promotes metabolic efficiency by enhancing the physical proximity of components in functionally related pathways and separating them from potentially competing processes.

For multicellular organisms, compartmentalization is found throughout physiology. Tasks such as metabolite storage, synthesis and consumption are distributed between organs and tissues at the systemic level and between different cell types within a tissue or organ (Fig. 1). Within cells, membrane-bound organelles subdivide the cytoplasm into chemically and physically unique reaction compartments. These compartments are in turn further parsed into unique environments at the level of multi-enzyme assemblies and condensates, which organize biochemical reactions at the molecular level^7–9.

Fig. 1 |. Metabolic compartmentalization across scales.

For multicellular organisms, compartmentalization is found throughout physiology. a, Tasks such as metabolite storage, synthesis and consumption are distributed between organs and tissues at the systemic level, and between different cell types within a tissue or organ. Within cells, membrane-bound organelles subdivide the cytoplasm into chemically and physically unique reaction compartments. These compartments are further parsed into unique environments at the level of multi-enzyme assemblies and condensates. b, As an example, glucose is distinctively stored and used between organs, tissues and cells. Storage occurs in the liver and skeletal muscle as glycogen, and glucose is the predominant fuel for the brain and substrate for lipid synthesis in adipocytes. Within liver lobules, zonation between periportal and periventricular areas compartmentalizes glycogen and glucose synthesis¹¹⁷. At the cellular level, an early step in glucose breakdown in the cytosol occurs by the action of 6-phosphofructokinase, which has been described to form large filaments²³. Glycogen granules in the cytosol serve as glucose storage sites¹¹⁸.

At its essence, compartmentalization fulfils three functions in metabolism, which we denote as pillars of metabolic compartmentalization (Fig. 2). Compartmentalization establishes the prerequisites for:

Fig. 2 |. The three pillars of metabolic compartmentalization.

At its essence, compartmentalization fulfils three functions in metabolism, by establishing the prerequisites for: (1) unique chemical and biophysical environments that facilitate biochemical reactions and metabolite storage; (2) detoxification and protection from toxic intermediates and by-products; and (3) providing metabolic control to regulate metabolite storage and release, prevent futile cycles and provide a mechanism by which metabolites function directly as signalling molecules.

1. (1) Unique chemical and biophysical environments for reactions and metabolite storage. Metabolic reactions often require specific reaction conditions (for example, pH and redox potentials) that are not compatible with other biological processes. Reaction compartments establish these unique chemical environments, enabling metabolic reactions to proceed under specific physiological conditions^10,11. Importantly, active concentration of metabolites and proteins is a means to drive chemical reactions, and compartments allow metabolites to be stored in conditions that promote their stability and regulated mobilization.

2. (2) Protection from toxic intermediates and by-products. Many metabolic reactions produce reactive end products or intermediates that result in damage or prevent other reactions from occurring. By restricting metabolic reactions to dedicated sites, compartmentalization protects the surrounding environment from reactive metabolites by concentrating detoxifying enzymes and metabolites at the site of reactive metabolite generation^12,13.

3. (3) Providing metabolic control. Metabolite levels must be regulated in response to changes in the nutrient state of a given system or environment to prevent unwanted futile metabolic cycles. Spatial separation of metabolic pathways enables rapid control of metabolite levels and coordination between pathways and the environment¹⁴. Moreover, compartmentalization provides a mechanism by which metabolites function directly as signalling molecules to relay organelle homeostasis from one compartment to another.

Studies investigating metabolism and its regulation at the levels of whole cells, whole tissues and whole organisms have revealed the core metabolic circuits within cells and organisms and established the importance of certain organelles in facilitating these reactions. However, bulk-level analysis misses the underlying heterogeneity, architecture and directionality of metabolic flux and thus occludes insights into the mechanisms and regulation of these reactions. With the emergence of new tools to systematically study metabolism at a cellular and subcellular resolution, a remarkable level of metabolite crosstalk continues to be uncovered between compartments. In this review, we use findings from the recent literature and well-established examples to illustrate how the three pillars of metabolic compartmentalization are established, maintained and regulated at the cellular level and discuss the consequences of dysregulation of metabolic compartmentalization in human disease.

Compartmentalization facilitates biochemical reactions and metabolite storage

Cells have developed mechanisms to create environments that support biochemical reactions under physiological conditions. For a metabolic reaction to proceed, exact concentrations of enzymes and substrates are needed in a permissive environment that controls for the required reaction conditions such as pH and osmolarity^15,16. To establish these metabolic reaction chambers, eukaryotic cells have evolved membrane-bound organelles, which divide the cytoplasm into chemically distinct compartments. For example, lysosomes concentrate protons within their lumens, which is required to unleash the activity of acid hydrolases, ensuring a centralized location for metabolite degradation and protection of other cellular compartments from destruction¹⁷. Similarly, the mitochondrial matrix, the labyrinth of compartments within the mitochondrial inner membrane, provides an electrochemical gradient required for ATP generation. When organelle metabolism is disrupted, it can lead to disastrous results, as discussed in Box 1. In addition to classical membrane-bound organelles, recent studies have revealed higher order enzymatic structures and condensates that achieve many of the same reaction conditions, albeit in a membraneless manner within the cytosol. Below we discuss the different strategies used by cells to create unique reaction chambers.

BOX 1. Genetic diseases caused by defects in metabolic compartments.

Mitochondria

Mitochondrial diseases (mitochondriopathies) primarily present during childhood with an incidence of 5–15 cases per 100,000 individuals¹³⁹. The most common presentation is Leigh syndrome, which encompasses more than 75 monogenic disorders¹⁴⁰. These diseases are highly heterogenous, often affecting multiple organ systems¹⁴¹. Although the majority of mitochondrial diseases stem from defects relating to oxidative phosphorylation, some are driven by dysfunctional phospholipid metabolism or nucleotide metabolism and detoxification¹³⁹. Management of mitochondrial diseases has been limited outside supportive care; however, new approaches including enzyme replacement therapy, which provides a recombinant version of the mutant enzyme, are being deployed. Enzyme replacement therapy has been successful in treating mitochondrial neurogastrointestinal encephalopathy syndrome, which is caused by loss-of-function mutations in thymidine phosphorylase. Erythrocyte-encapsulated thymidine phosphorylase has shown clinical success in one patient¹⁴².

Lysosome

Lysosomal diseases, also known as lysosomal storage diseases (LSD), encompass ~80 monogenic diseases. As a group, LSDs are fairly common, with an incidence of ~1:5,000. The most common LSDs are Fabry, Gaucher and Pompe diseases¹⁴³. LSDs affect multiple organs, and LSDs driven by enzyme deficiency typically present with central nervous system dysfunction. At the molecular level, the majority of LSDs are caused by defects in the degradation of macromolecules within lysosomes and concomitant accumulation of substrates¹⁴⁴. This can occur through mutation of specific enzymatic activities within the lysosomal lumen, transporters, lysosome biogenesis or trafficking. Some LSDs are treatable using enzyme replacement therapy, with Gaucher, Fabry and Pompe diseases being prime examples¹⁴³. For Niemann-Pick and Gaucher disease type C, substrate reduction therapy has proven successful¹⁴⁵.

Peroxisome

Peroxisomal diseases are mainly characterized as peroxisome biogenesis diseases (PBDs)¹⁴⁶. The most prominent group of PBDs is Zellweger spectrum disorder, which encompasses 12 different disorders¹⁴⁷. PBDs affect multiple organs, but those that affect the central nervous system are the most prevalent. PBDs are characterized by defects in importing peroxisomal proteins from the cytosol. Peroxisomal protein import is mediated by PEX proteins, which form a translocon-like import channel for peroxisomal proteins. Mutations in PEX genes lead to a decrease in catalase activity, defects in long-chain fatty acid oxidation and in ether lipid synthesis, resulting in metabolic imbalances¹⁴⁶. Most PBDs are managed with supportive care. However, cholic acid supplementation has been approved as the first treatment for patients with Zellweger spectrum disorders who have bile acid synthesis disorders¹⁴⁸.

Endoplasmic reticulum and lipid droplets

Diseases associated with defects in endoplasmic reticulum (ER) integrity and lipid droplet function have a range of symptoms, from lipodystrophy to neurological phenotypes. Mutations in genes encoding ER-shaping proteins, such as spastin and atlastin, cause hereditary spastic paraplegia characterized by axonal degeneration of upper motor neurons in the corticospinal tract¹⁴⁹. Physical therapy, electrical stimulation and muscle relaxants have proved useful to treat this condition¹⁵⁰. Mutations in genes encoding proteins involved in lipid droplet function and biogenesis, such as seipin, most prominently result in lipodystrophy, a disease that leads to an altered distribution of fat in the body and progressive metabolic complications such as diabetes¹⁵¹. The most successful treatment options for lipodystrophy include leptin replacement therapy and a low-fat diet¹⁵¹.

There is hope that genetic editing approaches may translate into treatments for many organelle-based diseases.

The mitochondrial pyruvate carrier

The identification of the long-sought mitochondrial pyruvate carrier (MPC) clarified an outstanding question in the field of how mitochondria import pyruvate for oxidation in the TCA cycle^152,153. Biochemical characterization of the yeast MPC and uptake studies with isolated mitochondria revealed that the MPC is a heterodimeric mitochondrial carrier that symports negatively charged pyruvate and a proton into the mitochondrial matrix^154,155. Thus, mitochondrial pyruvate import requires an intact mitochondrial membrane potential and metabolic activity. Alterations in mitochondrial fuel selection (for example, the inability to properly switch between carbohydrate and fat catabolism) are associated with mitochondrial dysfunction and are one of the hallmarks of metabolic disease¹⁵⁶. Elucidating the mitochondrial transporters underlying mitochondrial fuel choices and metabolite flux, such as the MPC, provides tools to study and perhaps modulate metabolite flux in metabolic diseases¹⁵⁷. Studies in animal models lacking the MPC have allowed researchers to reveal roles of mitochondrial pyruvate import in tumorigenesis, stem cell maintenance, neuronal excitability and control of systemic glycaemia^158–161. In addition, the use of colorectal cancer mouse models demonstrated that a glycolytic phenotype achieved by downregulation of the MPC was required during tumour initiation and that blocking mitochondrial pyruvate import increased both tumour frequency and grade¹⁵⁹. This study exemplifies how understanding mitochondrial transporters helps to dissect the metabolic contributions of cellular compartments to cell fate decisions and their effects on disease states, such as cancer.

The mitochondrial NAD transporter

Essential to mitochondrial function is the import of enzymatic cofactors and prosthetic groups into the mitochondrial matrix and their proper incorporation into functional holoenzymes. Mitochondrial respiration and many mitochondrial dehydrogenases depend on the redox cofactor NAD, whose redox balance also determines directionality of cellular metabolic flux. Depending on the cell type, mitochondria contain up to approximately eight times higher concentrations of NAD than the cytosol¹⁶². A long-standing debate in the field has been the contribution of mitochondrial NAD uptake relative to intramitochondrial synthesis. A significant advance in answering how mitochondria acquire NAD came from the identification of a mitochondrial NAD transporter in metazoans by gene co-essentiality analysis^163–165. Three independent studies showed that SLC25A51 (also known as MCART1) and its close homologue SLC25A52 are required for mitochondrial NAD uptake and respiration. Mitochondrial NAD levels appear to correlate with oxidative capacity across tissues, suggesting NAD levels might be rate-limiting for mitochondrial respiration, which has raised the possibility of modulating mitochondrial NAD content in the treatment of mitochondrial dysfunction¹⁶². The separation of metabolite pools is particularly important during times of acute starvation, when cytosolic metabolite levels such as NAD drop, whereas mitochondrial levels are maintained, thus preventing cell death during this time of stress^166,167. Separation of the mitochondrial and cytosolic chemical environments is thus a mechanism to preserve mitochondrial integrity and ensure the fidelity of cells under stress. Similarly to what has been observed for the MPC, SLC25A51/MCART1 expression appears to be regulated and in turn affects metabolic processes, possibly through mitochondrial sirtuins, which use NAD as a cosubstrate. Circadian rhythm, feeding state and microRNA-mediated regulation have been proposed as regulatory mechanisms^168,169.

The mitochondrial glutathione transporter

The recent identification of SLC25A39 and SLC25A40 as mitochondrial transporters for reduced glutathione (GSH), the most abundant antioxidant molecule in cells, confirmed the importance of GSH in maintaining mitochondrial redox potential^170,111. These studies also demonstrated, unexpectedly, that GSH is required for mitochondrial iron–sulfur–cluster synthesis. Iron–sulfur clusters are found in various cellular proteins including multiple ETC proteins and are required for electron transport within these complexes, as well as the TCA enzymes aconitase and succinate dehydrogenase¹⁷¹. The observation that mitochondrial GSH is important for iron–sulfur cluster biosynthesis was unexpected given that, for cytosolic and nuclear iron–sulfur proteins, only the initial steps of synthesis occur in the mitochondria, and suggests that mitochondrial GSH is important in maintaining the stability and activity of enzymes containing these clusters. A decline in GSH levels or alterations in GSH synthesis or redox balance has previously been associated with increased oxidative stress and the onset of neurodegenerative disease and anaemia, and is thought to underlie ferroptosis (discussed below)^172,173. Loss of SLC25A39 has been associated with neurodegeneration and anaemia in flies and mice, respectively^170,174. With the identification of the mitochondrial GSH importer, deciphering the role of this metabolite in mitochondrial function and the role of the mitochondrial GSH pool in the onset of neurodegeneration are now within reach, as are the mechanisms underlying mitochondrial GSH sensing.

Mechanisms to promote biochemical reactions without membranes

At the fundamental level of metabolic compartmentalization, metabolic reactions are facilitated by the spatial proximity between enzymes, often provided by higher order structures. These structures include stable enzyme complexes, dynamic assemblies of sequential enzymes, also known as metabolons¹⁸ (for example, tricarboxylic acid (TCA) cycle enzymes, or purinosomes), as well as larger enzyme assemblies such as the electron transport chain (ETC)¹⁹. Collectively, these enzymatic complexes facilitate directed and effective metabolic flux by preventing diversion of intermediates into alternative pathway branches as well as limiting diffusion of labile or toxic intermediates away from reaction centres (reviewed in ref. ⁹). Substrate channelling between reaction centres in stable multi-enzyme proteins is well illustrated by the fatty acid synthase (FASN) complex responsible for fatty acid synthesis in the cytosol. Conversion of acetyl-coenzyme A (acetyl-CoA) and malonyl-CoA to palmitate requires seven catalytic activities that are elegantly organized into an assembly line through the formation of a FASN homodimer. This 272 kDa behemoth encompasses three catalytic N-terminal domains fused to four C-terminal domains^20,21. Structural studies have revealed that the arrangement of FASN catalytic domains prevents incorrect crossreactions between intermediates. Encompassing all reactions of the fatty acid elongation cycle within one complex while handing substrates from one functional domain to another allows for efficiency. To what extent substrate channelling occurs in other enzyme complexes and concomitant effects of complex formation on catalytic efficiency remains debated⁹.

Some enzymes, when clustered at high enough concentrations, can undergo phase separation into a condensate or membraneless organelle^7,8,22. For example, 6-phosphofructokinase forms high-molecular-mass polymeric structures that can undergo a liquid-to-solid transition to assemble into filaments in the presence of its substrate²³. Similarly, cytidine triphosphate synthase forms large cytoplasmic structures in the cytosol of multiple organisms that increase enzymatic activity^24,25. A more recent example has been the discovery of the purinosome, which involves liquid–liquid phase separation, the condensation of macromolecules into a dense liquid phase distinct from the cytosol. Researchers have found that six enzymes involved in de novo purine biosynthesis appear to form large bodies close to the mitochondria, the major producers of one-carbon units in the cell²⁶. Purinosome formation is dynamic. When the demand for purines outstrips the amount supplied by the purine salvage pathway, purinosomes assemble, and this has been hypothesized to enable substrate channelling and higher catalytic efficiencies. The assembly and disassembly of purinosomes is probably controlled by feedback regulation by adenosine or guanosine levels via HIF-1α signalling and STAND–Nwd1 complex signalling^27,28. The metabolic logic for placing multiple copies of each enzyme in a cluster is the increased catalytic rate due to increased local concentration of enzymes and metabolites^9,19. The regulation of enzyme phase separation by metabolites (for example, assembly and disassembly of polymers) reveals how metabolite levels can be influenced by compartmentalization and can in turn regulate compartmentalization.

Organelles are unique membrane-bound environments

Although enzyme assemblies and condensates are found throughout the kingdoms of life, it is the segregation of metabolites and pathways between the cytosol and membrane-bound organelles that defines eukaryotic cells. To generate chemically unique environments, cells rely on the localization of organelle-specific enzymes through cellular trafficking pathways (reviewed in ref. ²⁹) and the activity of transporters and channels³⁰.

Transporters generate chemically unique environments.

Transporters and channels span the hydrophobic lipid bilayers of organelles and function as gates for metabolites and ions to pass selectively into and out of organelles. Transporters switch between conformational states opened to either side of the membrane to transport a substrate along or against its concentration gradient driven by hydrolysis of ATP or by coupling to another substrate that is transported along its concentration gradient³¹. Channels function by a different mechanism, forming pores to facilitate membrane transport. They allow the free flow of substrates along their concentration gradients once activated by an electrochemical gradient or ligand. More than 700 transporters and channels are thought to be encoded in the human genome, accounting for ~3.5% of all coding genes^32,33. Whereas polar metabolites require transporters to move between compartments, some lipophilic compounds, such as drugs, diffuse through membranes, bypassing boundaries between compartments³⁴. In addition, some metabolites transverse membranes as distinct metabolic intermediates, as in the case of fatty acids: rather than being directly imported, fatty acids enter mitochondria after being conjugated with carnitine, which is critical for directing them towards fuelling energy production³⁵.

The relatively large number of transporters belies the fact that biological characterization is limited for many of them. Challenges in the study of transporters arise from their requirement for a membrane environment to be active, functional redundancies and overlapping substrate specificities^33,36,37. Clever combinations of computational, genetic, biochemical and metabolite profiling approaches have brought new vigour to defining the roles of transporters in metabolism and compartmentalization (discussed in Box 2). Transporters carry out three essential functions: (1) facilitate import and export of metabolic substrates and products to and from membrane-bound compartments; (2) facilitate the import of enzymatic cofactors and prosthetic groups; and (3) create the chemical environment necessary for biochemical transformations. Below we provide three examples of recently discovered transporters that accomplish these tasks for mitochondria, a critical hub of metabolic compartmentalization within the cell.

Box 2. New approaches for studying organelle metabolism, transport, signalling and deducing transporter function.

OrganelleIP and resolving organellar metabolite pools

Because organelles comprise only a small portion of the cell, organellar isolation is needed to understand the identities of their metabolic constituents. Isolation of organelles and subsequent metabolite profiling has been conducted for several decades; however, the development of rapid and specific immunopurification methods (OrganelleIP methods) has enabled a new level of resolution and parallel sample analysis not previously matched by traditional approaches¹¹⁹. Such methods have been developed for the isolation of mitochondria (MitoIP¹²⁰), lysosomes (LysoIP⁸⁹), melanosomes (MelanoIP¹²¹) and peroxisomes (PeroxoIP¹²²).

To reduce artefacts from organelle isolation, stable isotope labelling of essential nutrients in cell culture–subcellular fractionation technology has been developed recently, and was used to demonstrate compartment-specific acyl-CoA profiles⁵¹.

Genetically encoded biosensors

The development of genetically encoded biosensors for metabolites such as lactate, ATP, glutathione, H₂O₂ and calcium when targeted to specific compartments enables researchers to detect changes in metabolite levels under physiologically relevant conditions without a need to isolate organelles¹²³. The SONAR and iNAP sensors for NAD/NADH and NADPH, respectively, have revealed both compartment-specific levels of NAD/NADH and NADPH and the differences in NAD/NADH abundance across cell types^124,125 Moreover, these tools have been used in a small molecule screen to identify compounds that modulate NADH levels within cells and tumours^124,125.

Genetic tools to modulate organelle redox state

An important toolkit for locally regulating NAD(P)H levels has been the localization of heterologous enzymes such as bacterial LbNOX and TPNOX or fungal NDI1 (refs.^126–129). These proteins oxidize the nucleotides NADH (LbNox, NDI1) and NADPH (TPNOX), respectively. A chemogenomic approach to generate ROS is the use of d-amino acid oxidase, which converts d-amino acids to their corresponding alpha keto acids producing H₂O₂ (refs.^130,131). By localizing them to the mitochondria and cytosol, they provide a tool to help decipher the contribution of compartmentalized NAD(P)H/NA(P)D levels to cell physiology and have been used to reveal insights into the link between reductive stress and metabolic disease, as well as the Warburg effect^{127,128,132,133}.

Other tools to study metabolism with subcellular resolution

Additional tools are exquisitely geared towards elucidating metabolic regulation with subcellular resolution. These include metabolic imaging such as correlative anti-Stokes Raman spectroscopy¹³⁴ and fluorescence lifetime imaging¹³⁵ that can visualize native lipids and redox cofactors, respectively, in live cells, proximity labelling approaches to interrogate metabolic regulation at specific cellular sites^136,137, and imaging mass spectrometry¹³⁸.

Local production of metabolites

By providing privileged chemical and metabolic environments, compartmentalization enables the on-site production of metabolites required in specific cellular locations. The purpose of such locally produced metabolites can be synthesis of macromolecules, to provide structural components and to mediate signalling cascades. An example is the synthesis and remodelling of the lipid cardiolipin in the inner mitochondrial membrane. Cardiolipin has a dimeric structure, consisting of two phosphatidic acid moieties joined through a glycerol backbone. The remodelling of its fatty acyl chains has been proposed to be directly linked to the high protein density of the inner mitochondrial membrane³⁸. Cardiolipin is unique to mitochondrial membranes (and to a lesser degree, to peroxisomes) and is required for respiratory complex function as well as for the function of mitochondrial carriers, which have specific binding sites for cardiolipin molecules³⁹. Cardiolipin itself also has been proposed to act as a precursor to signalling lipids⁴⁰. Oxidized cardiolipin is a substrate for hydrolysis by mitochondrial calcium-independent phospholipase A₂γ and is required for the release of oxidized fatty acids with various signalling functions, including in inflammatory processes and the perception of pain^41–45. Interestingly, when cardiolipin is externalized to the surface of mitochondria, as occurs upon stress, it acts as a signal for mitophagy and apoptosis^46,47.

Epigenetic modifications, such as the methylation and acetylation of histones, have been appreciated as a sink of metabolites in the nucleus and highlight the important and unexpected production of TCA cycle metabolites in the nucleus⁴⁸. Histone modifiers require acetyl- and S-adenosyl methionine units derived from the TCA cycle and one-carbon metabolism, and the TCA cycle intermediate α-ketoglutarate serves as cosubstrates of DNA- and histone-modifying enzymes. Levels of metabolites structurally related to α-ketoglutarate, such as succinate, fumarate and hydroxyglutarate, can inhibit chromatin modifiers at the concentrations present inside cells. Although the precise source (mitochondrial or nuclear/cytosolic) of these metabolites has not been conclusively demonstrated, the presence of isoenzymes of TCA cycle enzymes in the nucleus has been proposed to provide metabolites required as substrates to mediate epigenetic changes, in addition to their established role of energy production in mitochondria^49,50. The measurement of acyl-CoA levels specifically in the nucleus has revealed distinct acyl-CoA profiles between the nucleus and cytosol and identified isoleucine as a source for histone propionylation⁵¹. The presence of another subset of metabolic enzymes, serine hydroxymethyltransferase 1, dihydrofolate reductase and thymidylate synthase involved in one-carbon metabolism/thymidylate synthesis, has been attributed to demands of nucleotide synthesis for DNA replication. The enzymes have been reported to translocate and assemble in the nucleus in a cell-cycle dependent manner and ascribed to contribute directly to DNA synthesis during S-phase^52–54. Together, these examples illustrate that subcellular location of metabolic pathways is important for specifying functions of metabolites.

Compartmentalization protects from reactive metabolites

Many metabolic pathways generate reactive intermediates or end products that not only disrupt forward chemical reactions, but also damage cellular components. Reactive molecules include both reactive oxygen and nitrogen species (reactive oxygen species (ROS) and reactive nitrogen species), side-products following reactive nitrogen species/ROS reaction with macromolecules (for example, lipid peroxides), metabolic intermediates (for example, succinate, reactive acyl-CoAs⁵⁵, glycolytic products such as methylglyoxal⁵⁶, and formaldehyde produced by demethylase enzymes in the nucleus⁵⁷), signalling molecules with reactive groups (arachidonic acid) and xenobiotics (for example, aspirin⁵⁸)⁵⁹. By restricting the movement of these reactive metabolites throughout the cell and concentrating detoxifying enzymes and small molecules used in their neutralization at their sites of generation, compartmentalization offers one mechanism that cells use to combat damage from reactive metabolites. To illustrate the principles of compartmentalized protection, we focus on reactive metabolites generated at the mitochondria and plasma membrane/lysosome and how these organelles contend with them. The mitochondria play an outsized role in the production of reactive metabolites, and lipid peroxidation-mediated cell death, which originates in the plasma membrane and lysosome (for example, ferroptosis), has gained traction in recent years as a new therapeutic modality for cancer treatment⁶⁰. Throughout this discussion, we provide examples of how new tools are revealing the molecular mechanism underlying compartment-specific reactive metabolite generation and corresponding detoxification pathways.

Compartmentalized ROS generation

Within the metabolism field, ROS are the quintessential reactive metabolites. They encompass at least five different molecules, whose reactivity is largely believed to dictate their target scope. While in vitro these molecules can react with all macromolecule classes, compartmentalization within cells makes the interpretation and identification of bona fide ROS targets more complex. A notable example is superoxide, which is produced within the mitochondrial matrix as a product of complexes I, II and III activity within the ETC as well as other mitochondrial enzymes⁶¹. Whereas in vitro superoxide readily reacts with multiple protein residues, lipids and nucleotides, in the mitochondria ROS is rapidly converted by superoxide dismutase to less-reactive hydrogen peroxide (H₂O₂)⁶². Thus, cytosolic macromolecules rarely come into contact with superoxide, but rather with H₂O₂, which is membrane permeable. An important question in the field is how far from its source of synthesis mitochondrial-derived H₂O₂ can act within the cell, given the highly reductive nature of the cytosol. The fact that mitochondrial ROS production has been proposed to underlie ROS-induced cell death and damage in thousands of studies adds urgency to resolving these questions. One increasingly popular approach has been the development and application of ultrasensitive genetically encoded H₂O₂ reporters such as ro-GFP (reduction–oxidation sensitive GFP)⁶³ and HyPer7 (ref. ⁶⁴) (Box 2). When localized to different organelles, they provide a stunning portrait of compartmentalized H₂O₂. Using nucleus-localized HyPer7, recent studies revealed that mitochondria-derived H₂O₂ can indeed make it to the nucleus by disabling cytosolic antioxidant systems such as peroxiredoxins⁶⁵. Once in the nucleus, mitochondrial-derived H₂O₂ has been found to damage nucleic acids, resulting in double-stranded breaks and replication fork arrest leading to a block in cell proliferation⁶⁶. How the nucleus contends with an increase in ROS to safeguard the genome through the direct regulation of ROS-generating organelles such as the mitochondria, in addition to transcription-based anterograde signalling, is poorly understood.

Although ROS are canonically viewed as damaging molecules within the cell, they have important signalling functions as well⁵⁹. This is perhaps best illustrated by recent discoveries demonstrating that mitochondrially derived ROS (superoxide and hydrogen peroxide) are critical regulators of adipocyte thermogenesis⁶⁷. Using chemical proteomic platforms, which allow for the global identification of ROS protein targets⁶⁸, Chouchani and colleagues demonstrated that cysteine 253 (C253) in uncoupling protein 1 (UCP1) is modified by thermogenesis-derived mitochondrial ROS⁶⁹. Structural modelling of the oxidation of C253 to a sulfenic acid and detailed molecular analysis revealed disruption of a critical UCP1 interface resulting in UCP1 activity and increased mitochondrial respiration. Mutation of C253 to an alanine (C253A) resulted in a substantial decrease in thermogenic response in brown adipose tissue, higher glucose levels and increased immune populations in brown adipose tissue, which was attributed to redox stress in the C253A mice⁶⁹. Given the overlap between ROS sensitive cysteines and those which are amenable to covalent modification with small molecules⁶⁹, development of a small molecule targeting C253•UCP1, resulting in its activation, would be an interesting approach to pharmacologically regulate thermogenesis and its ensuing health benefits.

Ferroptosis is a compartmentalized form of metabolic cell death

Outside the nucleus and mitochondria, the plasma membrane, peroxisomes and lysosomes have emerged within the past decade as critical sites of compartmentalized ROS responses. These organelles have been brought to the forefront of compartmentalized ROS biology with the discovery of ferroptosis, a new form of cell death. The plasma membrane appears to be the main compartment that is disrupted upon ferroptosis induction, with compromise of this lipid bilayer a characteristic of ferroptosis⁷⁰. At the plasma membrane phospholipids containing polyunsaturated fatty acids (PUFAs) become oxidatively damaged and generate highly reactive lipid peroxides, which disrupts plasma membrane integrity. Glutathione peroxidase 4 (GPX4), a GSH-dependent hydroperoxidase, converts lipid peroxides to harmless alcohols using GSH. Inhibition of GPX4, or alteration of cellular redox status by blocking cysteine import or GSH synthesis are the main mechanisms for initiating ferroptosis⁷¹. An open question in the field is the source of lipid peroxides. Recently, peroxisome-based lipid ether synthesis has been linked to ferroptosis induction. Using a genome-wide CRISPR suppressor screen in renal cell and high-grade serous ovarian carcinomas, Schreiber and colleagues found that loss of peroxisome biogenesis genes, and in particular enzymes critical for ether lipid biosynthesis (for example, alkylglycerone phosphate synthase and fatty acyl-CoA reductase 1), mediated strong resistance to ferroptosis⁷². Importantly, peroxisomal enzymes traditionally associated with anti-oxidative responses including catalase and superoxide dismutase did not alter ferroptosis sensitivity⁷³. Moreover, AGPAT3, an ER-resident enzyme, was identified as a transferase required for the synthesis of ferroptosis-relevant ether lipids that contain a PUFA tail, providing an important clue as to the origin of lipids that are required for ferroptosis⁷⁴. In addition to lipid peroxidation, an important regulator of ferroptosis is iron availability. In its ferrous state (Fe²⁺), iron can react with H₂O₂ in the Fenton reaction to generate a hydroxyl radical that readily oxidizes lipids. Iron is transported into the cell through the transferrin receptor, which is engulfed and releases its cargo in the lysosomal lumen⁷⁵. A genome-wide CRISPRi screen in induced pluripotent stem cell-derived neurons revealed that the lysosomal lumen protein saposin is an important regulator of lysosomal iron levels. Deletion of saposin in post-mitotic neurons results in accumulation of lipofuscin (a mixture of lipids, oxidized proteins, sugars and metals) and leads to a substantial increase in lysosomal iron levels, which result in lipid peroxidation and ferroptosis⁷⁶. The last component required for ferroptosis is a source of H₂O₂. Multiple studies have suggested that the mitochondria could be the source of this ROS. However, how this ROS navigates from the mitochondria to the site of PUFA synthesis remains unknown⁷⁷. Although the physiological roles of ferroptosis in development are just beginning to emerge, there is substantial interest in leveraging ferroptosis to target drug-resistant persister cells in a variety of cancers⁷⁸. Whereas the bulk of a tumour is removed following chemotherapy, surgery, radiation or precision therapy treatments, persister cells, as their name suggests, remain and are thought to underlie tumour progression⁷⁹. Although the mechanisms underlying persister cells’ escape from therapy are under active investigations, it has been shown that persister cells specifically rely on GPX4 for survival in vitro. Moreover, loss of this enzyme prevented tumour relapse in mice, suggesting a viable strategy to target persister cells⁷⁸. A key question surrounding the use of ferroptosis as a therapeutic modality is the therapeutic index that can be achieved by inducing lipid peroxidation in tumours versus normal tissue.

Compartmentalization provides a means for metabolic control

The metabolic state within compartments is integrated through multiple levels in an organism by the activity of signal transduction pathways. Eukaryotes have evolved mechanisms to locally sense metabolite concentrations and appropriately control synthesis and degradation pathways to regulate respective metabolite abundance and quality. The paradigm for organellar metabolic signalling was first established by Brown and Goldstein’s discovery of sterol sensing at the ER and canonically relies on a protein-based sensing mechanism for a metabolite and a corresponding transcriptional response. A drop in sterol levels results in the translocation of the transcription factor SREBP from the ER to the Golgi, where intermembrane proteolysis releases SREBP, enabling nuclear translocation and induction of genes involved in sterol biosynthesis (reviewed in ref. ⁸⁰). Elegant sensing mechanisms have now been described for multiple metabolic pathways in their own organelle-specific iterations. Below, we highlight three forms of metabolic sensing based on recent discoveries that illustrate how these pathways are both dependent on a given compartment to orchestrate signalling, and in turn regulate its metabolic content and activity.

Metabolic regulation at the lysosome

Over the past decade, the lysosome has emerged as a cellular metabolic signalling hub with the discovery that the mTORC1 pathway localizes to its surface under amino acid-replete conditions⁸¹. Localizing a master growth regulator to the lysosome allows mTORC1 to survey both the extracellular environment (by providing access to endocytosed components which have reached the lysosome) and the intracellular environment (by facing outwards to the cytosol). mTORC1 promotes an anabolic programme and inhibits catabolism in nutrient-replete conditions, upregulating nucleotide, lipid and carbohydrate synthesis through phosphorylation of a number of downstream targets to promote cell growth. The lysosome in mammals possibly serves as a storage site for amino acids, similar to the vacuole in yeast, and provides the evolutionary logic for placing mTORC1 at this organelle. Recent research has demonstrated that leucine, arginine and lysine binding to their cognate amino acid sensors lead to the inactivation of the GATOR1 tumour suppressor, which in turn controls the activity of the heterodimeric RAG GTPases that recruit mTORC1 to the lysosomal surface, where mTORC1 is activated by the small GTPase Rheb^81–86. Rheb is in turn controlled by the activity of the tuberous sclerosis complex 1/2 (TSC1/2), whose lysosomal localization is also dynamically regulated by growth factors and functions as a further hub for energy and stress signalling⁸⁷.

When cells are starved of nutrients or growth factors, mTORC1 is released from the lysosomal surface and is inactive. Subsequently, catabolic processes including autophagy are activated, allowing the cell to recycle components providing building blocks for anabolic processes. Counter-balancing the anabolic activity of mTORC1 is its substrate: the transcription factor EB, which is a member of the TFE3 family of transcription factors⁸⁸. Upon a decrease in nutrient levels, mTORC1 activity and dephosphorylation by calcineurin, transcription factor EB translocates to the nucleus. There, it binds the promoters of genes involved in lysosome biogenesis, lysosomal function and autophagy, resulting in a concomitant increase in lysosome-based catabolic processes⁸⁸. Furthermore, mTORC1 has been shown to regulate the compartmentalization of essential amino acids, by controlling their efflux from lysosomes to the cytosol, where they are used in anabolic process⁸⁹. This interplay between anabolic and catabolic regulation is at the centre of metabolic signalling, where countervailing cellular processes must be balanced to maintain metabolic homeostasis. The transcription factor ATF4 ties metabolism to the control of proteostasis and redox balance by regulating protein and GSH synthesis, while connecting nutrient sensing at the lysosome with mitochondrial metabolism through its activation of mTORC1 (refs. ^90–92).

Mitochondrial retrograde signalling

Given the substantial role of the mitochondria in controlling cellular energy levels and metabolism, multiple signalling pathways have evolved to communicate the levels of mitochondrial metabolites or stress. These pathways are collectively referred to as mitochondrial retrograde signalling⁹³. Retrograde signalling is largely conserved and has three main mitochondrial signalling outputs: (1) energy levels, (2) calcium and (3) ROS. As with sterol and nutrient sensing, changes in mitochondrial outputs are relayed from protein sensors to elicit a transcriptional response. A drop in ATP synthesis upon mitochondrial dysfunction activates AMP-activated protein kinase, which in turn activates the nuclear hormone receptor PGC-1α and promotes the expression of numerous genes involved in mitochondrial energy metabolism and biogenesis^94,95. Disruption in the mitochondrial membrane potential triggers the release of Ca²⁺ into the cytosol, thus activating the transcription factors NFκB and NFATC. NFκB and NFATC then translocate to the nucleus and promote the expression of genes involved in glycolysis, gluconeogenesis as well as calcium metabolism^96,97. Multiple Ca²⁺-regulated kinases are also activated in response to mitochondrial calcium export (JNK, CAMKIV and p38 MAPK), resulting in a varied transcriptional response that increases mitochondrial adaptation to different stressors⁹⁸.

The integrated stress response plays an important role within retrograde signalling and is regulated by many external stressors (for example, hypoxia, nutrient deprivation) and internal stressors, most notably unfolded proteins. Recent discoveries have demonstrated that inhibition of ETC complex V leads to the activation of ATF4. Using genome-wide CRISPRi screens, Kampmann and Jae and colleagues identified DELE1 as a new component required to communicate ETC dysfunction to the cytosol^99,100. Disruption of complex V activity is sensed by the protease OMA1, which cleaves DELE1, promoting its translocation from the inner mitochondrial membrane to the cytosol. There, it binds to and increases the kinase activity of EIF2AK1, which phosphorylates ATF4, resulting in its nuclear translocation to regulate gene expression.

Compartmentalized reactive metabolite sensing

Because each cellular compartment faces separate challenges from reactive metabolites and compartments are metabolically connected, sensors for reactive metabolites in different cellular compartments are able to elicit activation of a coordinated stress response. These pathways often simultaneously undergo activation in disease states characterized by high levels of cellular damage, including Alzheimer’s disease, metabolic disorders and cancer⁵⁹. When triggered, these pathways launch a coordinated transcriptional response that results in both an upregulation of enzymes required for detoxification and a remodelling of cellular metabolism across compartments to support the function of detoxification enzymes, ensuring that detoxification mechanisms are not rate-limited by redox cofactor availability¹⁰¹. The basic leucine zipper domain-containing transcription factor NRF2 functions as the master regulator of the cellular antioxidant response. Under basal reductive conditions, NRF2 binds to KEAP1, a ROS-sensing protein that functions as an adaptor for the E3 ubiquitin ligase CUL3, which ubiquitylates NRF2, leading to its proteasomal degradation¹⁰¹. KEAP1 contains key ROS and reactive metabolite sensitive cysteines that, upon modification, block its interaction with NRF2, allowing the transcription factor to translocate to the nucleus where it promotes the expression of cellular programmes involved in antioxidant activities and corresponding metabolic pathways to support their activity¹⁰². For example, NRF2 induces the expression of rate-limiting enzymes in the cytosolic pentose phosphate pathway to increase NADPH generation. This is done in coordination with a battery of NADPH-dependent detoxification enzymes (for example, thioredoxin reductase) in the mitochondria and cytosol that are also upregulated upon NRF2 activity. The placement of NRF2–KEAP1 in the cytosol is thought to provide a convenient location for ROS/reactive metabolite sensing, because ROS levels in the ER and mitochondria have been purported to be sensed by this pathway^103,104. As discussed above, nuclear ROS levels are an important determinant in cell proliferation and recent findings have linked genotoxic stress pathways to its sensing. ATM is a PIKK kinase that is canonically activated by double-stranded DNA breaks and phosphorylates a multitude of proteins to promote DNA repair, cell-cycle arrest and apoptosis. Work by Paull and colleagues identified that cysteine 2991 in ATM functions as an H₂O₂ sensor and upon oxidation, results in its dimerization and activation distinct from its DNA-sensing functions¹⁰⁵. Through an undefined mechanism, ROS-based activation of ATM results in higher pentose phosphate pathway activity through upregulation of the rate-limiting enzyme glucose-6-phosophate dehydrogenase¹⁰⁶, which generates higher NADPH levels resulting in lower overall nuclear ROS levels. Whether additional members of genotoxic stress sense nuclear ROS, or whether there are other direct mechanisms by which these pathways impact ROS generation or detoxification has not been elucidated.

Outlook

The intersection of new genomic, proteomic and metabolic tools with a renewed interest in biological compartmentalization has resulted in a greater understanding of how segregation impacts metabolic pathways. The three central pillars that encapsulate the functions of compartmentalization in metabolism provide a framework to understand metabolic processes throughout physiology and direct the development of new tools to appreciate how compartmentalization shapes metabolism. The past century has seen a remarkable characterization of the enzymes, substrates and products of metabolic reactions, yet important questions remain for metabolic compartmentalization.

How are unique chemical environments established?

How metabolites and redox cofactors, such as pyridoxal phosphate and their derivates, are imported into organelles, and how their uptake is balanced with local synthesis are critical areas that have not been addressed. Moreover, how metabolite-specific transport is coupled to metabolic demand remains largely unknown for the vast majority of metabolites. Another pertinent question is how organelles exchange metabolites between each other and what the roles of the ubiquitous, recently identified organelle–organelle contact sites are in this process. How different concentrations of metabolites (for example, NAD or acyl-CoAs) are established between cytoplasm and nucleoplasm is poorly understood^51,107. Finally, heterogeneity is increasingly appreciated as a determining factor for many biological processes¹⁰⁸. The advent of high-resolution techniques has begun to reveal a high degree of heterogeneity between organelles. How organelle heterogeneity shapes compartmentalized metabolism and in turn metabolic flux and reactions is an important direction for future research.

How are compartments protected from toxic intermediates and by-products?

The type of reactive metabolites found in most organelles, and their corresponding targets still remains a mystery. The field has done an extraordinary job of developing specific reporters for some reactive metabolites including many species of ROS. Accurate sensors for other ROS as well as the plethora of reactive metabolites, including lipids, will shed much needed light in the field. Accordingly, the identification of compartmentalized ROS sensors is also an area that will continue to grow in the next decade.

What are the mechanisms of metabolic regulation and how do organelles communicate?

We still have an unsophisticated understanding of the sensors for many metabolites in the cytosol and in different compartments. Moreover, our knowledge of how organelles communicate with each other through metabolic signalling pathways remains in its infancy. Recent insights have revealed that not only do organelles control metabolite levels, but in turn metabolites influence the position of organelles within the cell. Prominent examples include amino acid regulation of mTORC1 influencing perinuclear lysosomal positioning, and active transport of mitochondria in neurons to firing synapses in response to a rise in intracellular calcium, with the small mitochondrial GTPase Miro functioning as a calcium sensor required for mitochondria motility^109,110. In the past decade, the identification of metabolite sensors has exploded, and we see this as just the beginning of a major expansion in research in compartmentalized metabolism. With the advent of new functional genomic approaches, such as paralogue-targeting CRISPR libraries^111,112, we can now systematically identify and characterize metabolite sensors in different organelles and deconvolute the function of paralogues and perceived functional redundancy at different compartments. Using these approaches, we can begin to address open questions, including whether nuclear NAD is sensed by the same pathways as cytosolic and mitochondrial NAD and whether there are unique non-overlapping roles for cytosolic versus mitochondrial isoenzymes and the mechanisms governing their cellular distribution. Separation of metabolic pathways across compartments creates prerequisites for metabolites functioning directly as signalling molecules relaying organelle homeostasis from one compartment to another and requires mechanisms of intercompartment metabolite exchange. Although organelle–organelle contact sites have been observed for decades, structural and functional information on the molecular tethers mediating such contacts has only recently become available¹¹³. Established functions of organelle–organelle contact sites include lipid synthesis and exchange, as well as calcium signalling^114,115. In addition to the discovery of tethers, detailed characterization of the protein composition of organelle contact sites will allow insights into how cellular compartments share nutrients and communicate, which in turn requires metabolites to be sensed in a compartment-specific manner¹¹⁶.

Although new studies have transformed our understanding of compartmentalized metabolism at the cellular level, much work remains to elucidate the pathways that control organismal metabolic compartmentalization. Metabolite levels are net outcomes of reversible reactions, thus, steady-state measurements, and even measurements for short periods, mask the underlying complexity of metabolism. Quantitative metabolic flux analysis is an important step towards resolving the dynamic regulation of metabolism. The characterization of metabolite transporters holds much promise not only as a mechanism to understand metabolic wiring and regulation, but also as a tool for manipulating metabolite levels within specific compartments. Such studies can address how changes in metabolic compartmentalization and flux are driven at the molecular level and how these determine fuel choices for cell types, tissues and whole organisms in both normal and disease states. Implicit in the study of metabolic compartmentalization is the realization that many human diseases originate from defects in the proper control of metabolite levels at different cellular locations. By focusing on the mechanisms underlying these compartmentalization disorders, the expectation is that new medicines will be developed to help patients suffering from these diseases.