Mitochondrial Tumor Suppressors—The Energetic Enemies of Tumor Progression

Cancer has been affecting people since the beginning of our history. Despite considerable effort in the last decades to find a cure, cancer-related death remains one of the leading causes of death worldwide. Around 18 million people are diagnosed with cancer worldwide and 9.6 million people succumb to this disease annually. The most common cancer types affecting humans are breast, colorectal, lung, and prostate cancer (1). Apart from the enormous toll cancer takes on lives of people, it also has a major impact on the world's economy; the total cost of cancer in Europe alone (in 2018) is estimated at 199 billion EUR (2). All these crucial factors create a great incentive to devise novel cancer treatments and diagnostic methods. Carcinogenesis, simplistically, requires two fundamental processes to occur: activation of proto-oncogenes to drive the cells' excessive proliferation and deactivation of tumor suppressor genes. Tumor suppressors are early defense molecules, which function as cellular brakes, stopping cells from uncontrollably dividing, therefore halting their progression toward carcinogenesis. Tumor suppressors were not specifically evolved to fulfil the tumor suppressor role; instead, each of them has a distinct and important role in cellular biology, which helps them to protect the cell from cellular transformation. Knudson's pioneering studies of retinoblastomas led to the discovery of the first tumor suppressor, the retinoblastoma gene and protein (3–6), which functions as a regulator of cell-cycle progression. The discovery of another famous tumor suppressor—p53, often referred to as the “guardian of the genome”—followed shortly thereafter (7–11). P53 was evolved to protect the genome of the cell from DNA damage, thus preventing genomic instability that precedes cellular transformation. Since then, many tumor suppressor genes and proteins have been discovered, including VHL (Von Hippel–Lindau), APC (adenomatous polyposis coli), and PTEN (phosphatase and tensin homolog). Although the great majority of the known tumor suppressor proteins reside in cytosol or nucleus, there is an expanding list of new, mitochondria-localized proteins that exert tumor-suppressive effects. Even though mitochondria are not considered the ultimate drivers of carcinogenesis, it is widely accepted that changes in mitochondrial processes, such as metabolic reprogramming and reactive oxygen species (ROS) dysregulation, can greatly facilitate this process.

The roles of the mitochondrial tumor suppressors are varied and include alterations in metabolic pathways, cellular differentiation, immune response, redox status, lipid biosynthesis, and mitochondrial dynamics. As such, they could provide an important link to cancer-associated processes such as metabolic reprogramming, the Warburg effect, inflammation, and stemness. Here, we review several members of the mitochondrial tumor suppressor family: lactamase-like B protein (LACTB), proline oxidase (POX), fumarate hydratase (FH), sirtuins (SIRT), fusion protein 1/tumor suppressor candidate 2 (FUS1/TUSC2), succinate dehydrogenase (SDH), and microtubule-associated scaffold protein 1 (MTUS1; Fig. 1; Supplementary Table S1). It is well documented that the function of many tumor suppressors is often context dependent; while these proteins can suppress tumorigenicity in certain contexts, such as specific tissues, tumor types, and (epi) genetic backgrounds, they can promote tumorigenicity in others. Therefore, to provide a complete review of the aforementioned tumor suppressors, we also briefly discuss the specific physiologic contexts where some of these proteins display a tumor promoter function, thus illustrating the complexity of tumor biology.

This review only describes those mitochondrial proteins whose tumor-suppressive function has been confirmed by four or more independent studies. Because the available information about these proteins is still sparse and fragmentary, it was our ambition to lay a foundation and create a resource upon which future research could be built.

The eukaryotic serine β-lactamase–like protein (LACTB) is a mitochondrial protein, evolutionary related to the bacterial penicillin-binding/β-lactamase proteins family (12). It is most likely derived from bacteria by horizontal or endosymbiotic gene transfer. Its conserved catalytic serine residue suggests that it still possesses peptidase or esterase activity. In bacteria, these classes of proteins are involved in bacterial cell wall biogenesis and peptidoglycan synthesis. However, because eukaryotic cells lack cell wall structures, the role of LACTB there remains unknown and is of great interest (12).

Mapped to the 15q22.1 chromosome, the human LACTB is a 547 amino acid long ubiquitously expressed protein, the highest expression being in skeletal muscles, heart, and kidneys (13, 14). LACTB is localized in the mitochondrial intermembrane space, and its localization depends on the N-terminal 62 AA-long mitochondrial targeting sequence (15).

In 2009, Polianskyte and colleagues demonstrated that LACTB can give rise to filaments formed by tetrameric subunits, suggesting its role in intramitochondrial membrane organization and micro-compartmentalization (15). LACTB belongs among the essential factors for proper activity of electron transport chain's (ETC) complex I, pointing to a regulatory role of LACTB in the respiratory chain (16). LACTB was also shown to be associated with mitochondrial ribosome (17). Transgenic mice with upregulated LACTB levels varied notably from the control mice, their fat-mass-to-lean-mass ratios being 20% higher than the wild type control. This suggests LACTB's role in obesity and fatty acid metabolism (18, 19). A study in 2016 showed that LACTB influences the progression of atherosclerotic plaques through modulating the levels of inflammatory chemokines, implying a role in regulation of immunity (20).

In 2017, while researching the expression profiles of “cancer-resistant” tissues, Keckesova and colleagues identified LACTB as a potential tumor suppressor (Fig. 2). Subsequent in vivo, in vitro, and clinical studies demonstrated LACTB's tumor-suppressive function in breast cancer (14). LACTB is downregulated in various breast cancer cell lines as well as in human breast cancer samples, where it showed 34% to 42% protein level decrease compared with normal breast tissue. LACTB overexpression, in a panel of tumorigenic and nontumorigenic cell lines, led to a decrease of proliferation rate in cancer cells, while nontumorigenic cells were minimally affected. This effect was also observed in vivo; the induction of LACTB in already formed tumors had a negative effect on their growth (14). LACTB knockdown in normal human epithelial cells in combination with overexpression of an oncogene (such as HRAS G12V or MYCT58A) led to malignant transformation of these cells, resulting in tumor formation, further confirming the tumor suppressive role of LACTB (14). Dysregulation of LACTB in breast cancer is associated with high expression of miR-374a, which promotes cancer cell survival and growth in vitro and in vivo. MiR-374a knockdown suppresses the cells' proliferative and colony formation activity, as well as migration and invasion capacity. LACTB silencing, however, reverses these changes (21). Further in vitro and in vivo studies demonstrated that LACTB has a negative impact on cancer growth, migration, and/or invasion also in gliomas, hepatocellular carcinomas, and colorectal cancer (22–25). In glioma A172 cells, the overexpression of LACTB significantly inhibited their proliferation. Its expression was shown to be involved in the invasion and angiogenesis process through the downregulation of proliferating cell nuclear antigen (PCNA), matrix metalloproteinase 2 and 9 (MMP2, MMP9), VEGF, proteins important in glioma cell proliferation, invasion, and angiogenesis (22). LACTB is also strongly downregulated in colorectal cancer, which is considered one of the most aggressive malignancies, and low levels of LACTB predict a poor prognosis in patients (23, 25). Interestingly, the downregulation of LACTB in colorectal cancer is not due to genetic mutations, but depends on epigenetic changes in its promoter: mainly promoter methylation and histone H3 hypoacetylation (25).

In vitro and in vivo experiments showed that in colorectal cancer cells, LACTB suppresses tumor growth and metastasis and can directly interact with p53 protein thus preventing its degradation by mouse double minute 2 (MDM2). This axis can potentiate the tumor suppressive effects promoted by both, p53 and LACTB, proteins. Moreover, p53 depletion reduces the tumor suppressive activity of LACTB when it is overexpressed in colorectal cancer (25). While this study provided interesting mechanistic insight into LACTB function, it did not clarify whether the interaction between p53 and LACTB occurs inside the mitochondria or in the cytoplasm. Thus far, two studies attempted to elucidate the function of tumor-suppressive function of LACTB outside the mitochondria with opposing conclusion. In breast cancer, deletion of mitochondrial localization sequence within LACTB abrogated its ability of tumor suppression, pointing to the importance of LACTB's mitochondrial localization for its tumor suppressor function (14). However, another study, performed in melanomas, showed cytoplasmic interaction of LACTB with serine/threonine–protein phosphatase (PP1A). This interaction was shown to prevent the binding between yes-associated protein (YAP) and PP1A in melanocytes thus repressing their malignant transformation (26). Future studies in this area are needed to clarify the importance of nonmitochondrial LACTB in tumor suppression and to define specific contexts when such nonmitochondrial tumor suppression occurs.

Breast cancer cells that survived LACTB induction displayed a more differentiated, epithelial-like morphology (14). This change in morphology was accompanied by an increase in the levels of epithelial differentiation markers, such as CD24 and epithelial cell adhesion molecule (EPCAM), and a decrease of certain mesenchymal markers [CD44, zinc finger E-box-binding homeobox 1 (ZEB1)]. This resulted in decreased proliferation and impaired tumorigenesis. A similar observation was noted in tumor tissues (14). The ability of LACTB to promote a differentiated phenotype was also observed in colorectal cancer (23, 25). Zeng and colleagues in 2018 demonstrated that in HCT116 and HCT8 cancer cells, stable LACTB overexpression promotes a morphologic change from a “spindle-like” into the tight “cell-to-cell” adhesion phenotype, while LACTB knockout reverses this process (25). These results showed that LACTB overexpression increased the expression of epithelial cell markers such as E-cadherin and β-catenin. In contrast, N-cadherin, vimentin, c-Myc, cyclin D1, and Twist1 mesenchymal markers were strongly downregulated, revealing a contribution of LACTB to the inhibition of epithelial–mesenchymal transition (EMT) in colorectal cancer (23, 25).

The role of LACTB in differentiation was also confirmed in normal cells, namely in skeletal muscle cells (14, 27). During the myoblast differentiation process in C2C12 cells, LACTB levels gradually increase. This process is negatively regulated by miR-351-5p expression. Overexpression of miR-351-5p downregulated LACTB and its inhibition had the opposite effect. When LACTB was silenced, the proliferation rate of C2C12 myoblasts increased (through upregulation of cyclin-regulated factors), which led to inhibition of the differentiation process and impaired myogenesis (27).

LACTB in colorectal cancer promotes autophagy through increasing the LC3-II/LC3-I ratio and the Unc51-like autophagy activating kinase-1 (ULK1) expression levels. The promotion of autophagy by LACTB and the consequent inhibition of EMT (in colorectal cancer) was shown to be the result of the inactivation of the phosphoinositide 3-kinase/protein kinase B/mTOR (PI3K/AKT/mTOR) signaling pathway via regulation of the PIK3R3 level and downregulation of PI3K (23).

The ability of LACTB to promote loss of tumorigenicity and onset of differentiation in breast cancer cells is partly dependent on its ability to reprogram the lipid metabolism (14). This is achieved through LACTB-dependent downregulation of the lipid-synthesizing mitochondrial phosphatidylserine decarboxylase (PISD) enzyme, which leads, in turn, to subsequent changes in the levels of mitochondrial lyso-phosphatidylethanolamine (LPE) and phosphatidylethanolamine (PE; Fig. 2). In hepatocellular carcinoma, metabolic dysregulation is an essential part of cancer progression and hepatocellular carcinoma is typically characterized by metabolic disorders (28). It was demonstrated that LACTB mRNA and protein levels were both downregulated in hepatocellular carcinoma, and decreased LACTB expression was associated with poor prognosis. On the other hand, its overexpression inhibited cancer cell proliferation, migration, invasion, and tumor growth (24). The expression and activity of important enzymes involved in lipid metabolism pathways in hepatocellular carcinoma (such as carnitine palmitoyltransferase 1A, acyl-coenzyme A dehydrogenase, and others) were shown to be significantly related to LACTB expression, reinforcing LACTB's role in lipid metabolism (24).

LACTB is involved in cancer progression in nasopharyngeal carcinoma and pancreatic adenocarcinoma where elevated LACTB expression is associated with poor survival rate (29). In the nasopharyngeal carcinoma context, LACTB expression promoted metastasis while suppression of LACTB reduced cellular mobility and metastatic ability. LACTB suppression was shown to enhance histone H3 stability and its acetylation, thus inhibiting the activation of receptor tyrosine–protein kinase/epidermal growth factor receptor–extracellular signal-regulated kinase (ERBB3/EGFR-ERK) pathway (30).

Another tumor suppressor protein localized in mitochondria is POX. POX, also known as proline dehydrogenase (PRODH), is the first enzyme of the proline metabolic pathway. Localized in the inner mitochondrial membrane, POX catalyzes the oxidation of proline into Δ¹-pyroline-5-carboxylate (P5C). Electrons produced by this reaction are transferred onto ubiquinone, explaining the dependence of POX on complex III and IV for its proper activity (31). The connection with the electron transport chain, however, seems to be more profound. POX's expression in DLD-1 (human colorectal cancer) cells negatively regulated the expression of components of the ETC and was inhibited by complex II inhibitors and succinate (31).

The interest in POX as a potential tumor suppressor was sparked in 1997 by a study from Polyak and colleagues, which identified POX as one of the several genes whose expression was significantly increased by the apoptosis-causing p53 tumor suppressor (32). Indeed, p53 response elements were identified in both promotor and intronic regions of the POX gene (33, 34). Decreased levels of POX have been detected in multiple tumor tissue samples, most notably in kidney and digestive tract tissues (35). In renal cell carcinoma (RCC), protein levels of POX are negatively regulated by miR-23b* (36). A follow up in vitro study suggested that the c-myc oncogene can indirectly downregulate POX levels by the miR-23b* both in P493 and PC3 cancer cells (37).

Although POX was initially discovered as one of the genes induced during p53-dependent apoptosis, it is also able to mediate apoptosis independently of p53, and its proapoptotic, tumor-suppressive effects have been confirmed both in vitro and in vivo (35, 38–40). As it turns out, ROS play a pivotal role in POX's mechanism of action (Fig. 1). Thus, increased POX expression is accompanied by a surge in ROS generation (41, 42), the presence of which has been found essential for the apoptotic effects of POX in cancer cells (42–45). Mechanistically, at least in some instances, ROS generated by POX is able to mobilize Ca²⁺, which in turn activates calcineurin and nuclear factor of activated T cells (NFAT), both of which were crucial for POX-mediated apoptosis in lung, renal, colon, and ovarian carcinoma cells (44). Interestingly, POX has been shown to activate the extrinsic apoptotic pathways in human colon cancer cells as well. Its expression was accompanied by increased cleavage of caspase-8 and higher levels of death receptor 5 (DR5) and TNF-related apoptosis-inducing ligand (TRAIL) mRNA which seems to be, at least partially, mediated by NFAT (45). Another study showed that troglitazone, a PPARγ ligand, also induced the expression of POX and increased its catalytic activity (46). The induction of PPARγ led to an increased formation of ROS by POX and onset of apoptosis in multiple human cancer cell lines (47, 48).

In colorectal cancer, the phosphorylation status and hence the activity of mitogen-activated protein kinase kinase (MEK), ERK (45), EGFR and glycogen synthase kinase-3β (thus the stability of β-catenin; ref. 42) was negatively affected by POX expression. Cyclooxygenase-2, which is frequently upregulated in colorectal cancer and often contributes to cancer development (49), was also downregulated as was the production of prostaglandin E2 (PGE₂; ref. 42). Constitutively active MEK partially blocked apoptosis induced by POX (45) as did the addition of PGE₂ (42). These effects were, to some degree, reversed by the introduction of manganese superoxide dismutase (MnSOD; refs. 42, 45), further reinforcing the importance of ROS as the main mediator of POX's tumor suppressive function. Hypoxia-inducible factor 1α (HIF1α) levels were shown to be greatly reduced by POX in both normoxic and hypoxic conditions in colon cancer cells. This is achieved through POX-dependent upregulation of α-ketoglutarate, a critical substrate for prolyl hydroxylation and degradation of HIF1α (35). This effect has been reversed by the aforementioned miR-23b*, where miR-23b* did induce the expression of HIF1α by inhibiting POX (36).

In addition to apoptosis, POX was able to induce G₂ arrest in human colon cancer cells. The arrest was accompanied by an upregulation of genes from the growth arrest and DNA damage–inducible gene family (GADD) such as GADD45α, GADDβ, GADDγ, and GADD34 (35). POX activity has been shown to be vital in inducing senescence in both transformed (human osteosarcoma, U2OS) and untransformed (human fibroblast, Hs68) cell lines in response to DNA damage and POX's ability to generate ROS was essential for this effect. Furthermore, ROS produced by the overexpression of POX could cause DNA damage itself (50).

In certain contexts, namely during hypoxia and/or nutrient stress, POX was shown to act as a promoter of tumor survival. Under hypoxia, the expression of POX rose considerably in various cancer cell lines in vitro and in the hypoxic tumor microenvironment of the breast cancer MDA-MB-231 cells in vivo. The increased levels of POX contributed to the survival of cancer cells under hypoxia and glucose deprivation. In the hypoxic conditions, this was explained mechanistically by the ability of POX to induce ROS-dependent protective autophagy instead of apoptosis in cancer cells (51). Under nutrient stress, glucose-starved or rapamycin treated colorectal cancer cells displayed significant increase in the catalytic activity of POX. This led to the degradation of proline, which can be obtained by cells from breakdown of the extracellular matrix. Subsequent increase in ATP levels allowed the survival of cancer cells. These results indicated that the induction of proline cycle initiated by POX under conditions of nutrient stress may be a mechanism by which cells switch to a catabolic mode for maintaining cellular energy levels (52). Similar results were also observed in pancreatic ductal adenocarcinoma cells, where under conditions of nutrient stress POX-dependent oxidation of collagen-derived proline promoted cancer cell survival and proliferation (53). Cancer progression was also shown to be mediated by POX-dependent proline catabolism in non–small cancer lung cells (NSCLC). In this cancer type, POX is upregulated by lymphoid-specific helicase (LSH), an epigenetic driver of NSCLC, in a p53-dependent manner. The activity of POX then promotes EMT as well as expression of inflammatory cytokines [most likely by producing ROS and affecting inhibitor of nuclear factor kappa-B subunit alpha (IKKα) phosphorylation], whereas its knockdown reduced cell growth, impaired colony formation, downregulated markers of mesenchymal state, and upregulated epithelial marker (E-cadherin; ref. 54).

FH, also known as fumarase, is a member of the tricarboxylic acid (TCA) cycle that catalyzes reversible hydration of fumarate to malate. Although the TCA cycle takes place in mitochondrial matrix, a so called “echoform” of FH can also be found in cytosol and can localize to nucleus upon DNA damage, where it participates in DNA damage response (55, 56).

Mutations in FH predispose to malignant paragangliomas and pheochromocytomas (57, 58) as well as to hereditary leiomyomatosis and renal cell carcinoma (HLRCC; ref. 59, 60). FH loss has also been reported in glioma, ependymoma, adrenocortical carcinoma, neuroblastomas, osteosarcoma, and Ewing sarcoma (61).

The underlying mechanisms of FH's tumor suppressive role are yet to be completely understood; accumulation of fumarate (through inactivation of FH), however, seems to be beneficial to cancer cells and could be considered the common denominator of most negative (i.e., cancer-promoting) effects in FH-defective cancer cells (Fig. 1). These effects can be roughly divided into two categories: fumarate itself directly acting as an inhibitor of various enzymes and fumarate as a source of cysteine modifications that can hinder protein activity. The accumulated fumarate can spontaneously modify proteins via Michael addition reaction with the sulfhydryl group of cysteine, thus creating S-(2-succinyl) cysteine (2SC; ref. 62). This modification inactivates proteins such as glyceraldehyde-3-phosphate dehydrogenase (with a possible link to diabetes and mitochondrial stress; refs. 62, 63), mitochondrial aconitase 2 (64), iron regulatory protein 2 (65), and Kelch-like ECH-associated protein 1 [KEAP1; which is a negative regulator of nuclear factor erythroid 2-related factor 2 (NRF2); ref. 66]. 2SC modification can be considered a reliable biomarker for detection of FH-defective tissues (67).

HIF1α expression in hypoxic regions triggers cancer plasticity and heterogeneity, promoting an aggressive and metastatic phenotype. Together with NF-B, they regulate over 1,000 genes, enhancing cell survival through the expression of growth factors, such as vascular VEGF, which contribute to a new vascularization within the tumor (68). In renal cancer, higher levels of stabilized HIF1α were detected in HLRCC-derived cell lines with inactive FH (69) as well as in FH-defective fibroblasts (70). HLRCC exhibits higher angiogenesis in association with increased transcript expression from the VEGF and BNIP3 hypoxia-responsive genes (71). In addition, HIF stabilization, in FH knockdown mouse embryonic fibroblasts (MEF), led to increased expression of MET oncogene, which in turn, helped further stabilize HIF, creating a feed-forward loop. Expression of MET positively affected motility and could contribute to tumor progression (72). Thus far, there are two proposed mechanisms of HIF stabilization in FH-defective cells: it has been shown that fumarate acts as an inhibitor of HIF prolyl hydroxylase thus blocking the path to HIF degradation (69, 73). A different study suggests that increased ROS (formed, at least partially, by NADPH oxidase and protein kinase C-δ activity in FH-deficient cells) is responsible for HIF prolyl hydroxylase inhibition, possibly though oxidizing cellular pool of Fe²⁺ to Fe³⁺, thus depriving HIF prolyl hydroxylase of its cofactor (74). FH-deficient cancer cells were, in fact, found to be iron deficient and had notably higher levels of HIF1α (75), pointing to a role of FH in iron metabolism.

Fumarate was found to inhibit the activity of ETC's complex II directly, via product inhibition, and complex I indirectly, thus reducing cellular respiration. As complex I contains iron–sulphur clusters, and some targets of fumarate-caused succinylation are proteins important for iron–sulphur clusters biogenesis, this could provide an explanation of complex I inactivity in FH-defective cancer cells (76). In HLRCC, accumulated fumarate dysregulates iron metabolism in two major ways, by covalently modifying iron regulatory protein 2, which is consequently unable to repress ferritin translation (65), and by enhancing ferritin translation via the activation of NRF2; ferritin then promotes the expression of forehead box protein M1 (FOXM1), which in turn promotes the proliferation (65). Moreover, the upregulated ferritin can lower cytosolic iron levels; this seems to correlate with the observation that FH-deficient cancer cells had lower cytosolic iron levels (75). This was further explained by reduced AMP-activated protein kinase (AMPK) levels leading to lower expression of iron divalent metal transporter 1 (DMT1; ref. 75), which seems to suggest that the dysregulation of iron metabolism in FH-defective cells happen via multiple ways.

FH deficiency was shown to induce the process of EMT, which promotes migration, invasion, and cancer progression. Fumarate acts as an inhibitor of α-ketoglutarate–dependent dioxygenases [such as ten-eleven translocation (Tet) enzymes responsible for DNA demethylation], thus affecting genome-wide histone methylation (77). Vimentin and other EMT-related genes were upregulated in Fh1^−/− normal and cancer renal cells. The reintroduction of Fh1 into these cells was sufficient to diminish the expression of EMT-related genes and restore the expression of E-cadherin, leading to more epithelial-like morphology of the cells. Further experiments showed that the fumarate-induced inhibition of Tet enzymes and subsequent DNA hypermethylation, which led to downregulation of antimetastatic miRNAs from miR-200 family, was responsible for the expression of EMT-related transcription factors (78). The importance of FH inhibition in EMT is further reinforced in the example of nasopharyngeal carcinoma. LSH, which acts as a chromatin modifier, is upregulated in nasopharyngeal carcinoma and is important for its progression. LSH mediates silencing of FH transcription, which then promotes EMT and cancer progression through modulation of IKKα activity (79).

FH was shown to be able to promote DNA repair. Upon ionization-induced DNA damage, DNA-dependent protein kinase (DNA-PK) that participates in the nonhomologous end joining (NHEJ) DNA repair phosphorylates nuclear FH, which then associates with chromatin at the double-strand break (DSB) region. Locally produced fumarate enhances the recruitment of DNA–PK complex at the DSB region which promotes NHEJ-mediated DNA repair and survival (80). The involvement of FH in DNA damage response seems to be highly conserved as it has been primarily discovered in yeast, where it was shown to participate in homologous recombination (HR; refs. 81, 82). Subsequent studies confirmed and extended the importance of FH and fumarate in DNA repair and genomic stability (83, 84).

Accumulated fumarate can upregulate antioxidant response element-controlled genes (85). This might be accomplished by covalently modifying cysteine residues of KEAP1 (succinylation), which is then unable to repress NRF2-mediated antioxidant response pathway (66). An antioxidant phenotype could prevent ROS-induced DNA damage, thus providing cells with growth advantage. On the other hand, fumarate was also found to succinylate gluthathione (GSH), leading to persistent state of oxidative stress and causing senescence in vitro and in vivo (86). Fumarate can also indirectly activate Abelson tyrosine-protein kinase 1 (ABL1) proto-oncogene possibly via oxidative stress–dependent mechanism (87).

FH is an important metabolic enzyme and its dysregulation in cancer cells is accompanied by considerable metabolic reprogramming of cancer cell metabolism. FH-defective HLRCC cells (UOK262) had significant changes in protein phosphorylation, leading to inhibition of pyruvate dehydrogenase's enzymatic activity, thus preventing carbon from entering the TCA cycle (88). This led to enhanced aerobic glycolysis and higher lactate production to compensate for the defects in energy production (75, 89). It is worth mentioning, however, that the rate of oxidative phosphorylation was not insignificant. Pentose phosphate pathway, specifically its oxidative branch, was also enhanced, leading to increased amounts of ribose and NADPH, the latter of which can be used in fatty acid synthesis (89). Cells with defective FH also used other compensatory mechanisms, such as glutamine-dependent reductive carboxylation (to produce Ac-CoA and citrate) and oxidative metabolism of glutamine to fumarate to restore the energy homeostasis (90).

AMPK is generally regarded as a master sensor of metabolism (91). The accumulated fumarate protected both FH-defective fibroblasts and renal cells from apoptosis by activating AMPK (70). In support of these findings comes a screen of synthetic lethality, where it was revealed that the inhibition of adenylate cyclase in FH-deficient cell lines (human embryonic kidney cells and HLRCC-derived UOK262) negatively affects the cells' survival (92). In contrast, a different study found that AMPK levels in FH-deficient kidney cancer cells were reduced and the cells exhibited dysregulation in iron metabolism (deficiency in cytosolic iron) as well as the activation of anabolic proteins (acetyl-CoA carboxylase and ribosomal protein S6) and iron regulatory proteins, all of which was accompanied by increased expression of HIF1α. The activation of AMPK and silencing of HIF1α decreased the tumorigenicity of these cells (75).

Paradoxically, while cancer arises from errors in the DNA repair mechanisms, to divide correctly, cancer cells also need to have functional DNA repair. As such, the abovementioned DNA repair function of FH is sometimes beneficial to cancer cells. For example, in gastric cancer tissue, FH mRNA and protein levels are significantly increased. Its knockdown impaired DNA damage repair and resulted in better response to cisplatin-mediated chemotherapy both in vitro and in vivo (93). FH was also able to promote cell survival upon exposure to ionizing radiation in other cancer cell lines, such as U2OS from human osteosarcoma and GSC11 (human primary glioblastoma cells; ref. 80).

Thus, we have a paradoxical behavior of FH, where in cancer cells which do express FH, mitochondrially localized FH acts as a tumor suppressor (through HIF1 stabilization and other mechanisms) and nuclear-localized FH acts as a tumor promoter (through DNA repair mechanism). As to which of these two branches of FH action manifests more in cancer cells is currently thought to be context dependent and influenced by localization of FH, concentration of fumarate and malate and the stage of tumor development (61, 94, 95).

The highly conserved sirtuin protein family has several roles spanning from metabolism to oxidative stress (96–98). Seven human sirtuins have been described, three of which, SIRT3, SIRT4, and SIRT5, were identified also in mitochondria (99–102). The biological function of these proteins is NAD⁺-dependent deacetylase and mono-ADP-ribosyltransferase (103–105). There is growing evidence, however, that they may possess other enzymatic activities such as desuccinylation and demalonylation (106). Despite a relatively large amount of information known about these proteins, there is still a lively debate about their complex role in cancer. As our review is primarily concerned with mitochondrial tumor suppressor, we will only focus on the function and properties of sirtuins 3 and 4 (Fig. 1).

Studies have shown SIRT3 to be downregulated in various cancers, such as mantle cell lymphoma, leukemia, hepatocellular carcinoma, pancreatic cancer, lung cancer, or gastric cancer (107–111). Sirt3^−/− MEFs presented gene instability, and a single oncogene expression by viral transduction induced cell transformation in vitro. In addition, Sirt3^−/− mice generated estrogen receptor/progesterone receptor (ER/PR)–positive tumors in the mammary gland over 24 months (112). A dataset study performed with genomic RNA from 38 normal and 36 human breast cancer samples showed that SIRT3 expression is decreased in cancer compared with the normal tissue (112).

In mantle cell lymphoma, SIRT3 protein levels are reduced and relate to a poor clinical outcome (107). Similar results were found in chronic lymphocytic leukemia samples and malignant B cells (107). The absence of SIRT3 was accompanied by an increase in superoxide dismutase 2 (SOD2) and isocitrate dehydrogenase 2 (IDH2) acetylation as well as an increase in ROS level. Conversely, when SIRT3 is overexpressed, these cells exhibit a decrease of the Warburg effect (decreased lactate production and downregulation of several glycolytic genes), which is further diminished by a drop in glucose uptake (107), supporting the idea that SIRT3 negatively regulates glycolysis. SIRT3 mediates the metabolic reprogramming through the destabilization of HIF1α (113, 114). Expression of SIRT3 is downregulated in human breast cancer and its loss correlates with upregulation of HIF1α target genes, such as GLUT1, HK2, PGK1, PDK1, LDHA, and VEGFA. On the other hand, overexpression of SIRT3 represses glycolysis and proliferation in breast cancer cells, confirming the metabolic mechanism for tumor suppression (113). Expression of SIRT3 can be controlled by the ZEB1 (109), a key regulator in metastasis and EMT (115). It has been demonstrated that ZEB1 can interact with the methyl-CpG binding domain protein 1 (MBD1) and suppress, through an interaction with the promoter in the E-box and Z-box, SIRT3 expression in a pancreatic cancer cell line, enhancing the aerobic glycolysis (109).

When compared with normal adjacent tissue, SIRT3 expression is downregulated in human lung adenocarcinoma and its reexpression significantly inhibits the cancer cells growth through Annexin V and caspase 3–mediated apoptosis. In this cell model, SIRT3 overexpression not only increased the BAX/BCL-2 and BAD/BCL-X/L ratios but also increased the levels of p53 and p21 and decreased ROS levels (110). In chronic myelogenous leukemia and promyelocytic human leukemia, SIRT3 activates apoptosis through caspase-3, Bcl-2, and cytochrome c pathway (116).

In gastric cancer, SIRT3 protein and mRNA levels are significantly reduced both in cancer tissues and cell lines and overexpression of SIRT3 decreases cell proliferation and colony numbers (111). Conversely, its knockdown promotes cell growth and colony formation ability. Mechanistically, it was shown that SIRT3 inhibits the expression of Notch-1 both at mRNA and protein levels. Moreover, Notch-1 overexpression drastically reduces the inhibitory effects of SIRT3 on tumor cell proliferation (111). In addition, in cells from hepatocellular carcinoma, overexpressing SIRT3 leads to higher levels of p53 most likely through MDM2 downregulation (108). SIRT3 seems to play an important role in a plethora of pathways as its overexpression reduces the NAD⁺ level, activates the AKT and c-Jun N-terminal kinase (JNK) signaling pathways, and represses the ERK 1/2 pathway (108). JNK signaling pathway enhances apoptosis (117), whereas AKT and ERK stimulate cell growth (118). Imbalance in these pathways would affect cell proliferation and cell growth (in hepatocellular carcinoma), indicating the tumor-suppressive role of SIRT3.

The oncogenic role of SIRT3 was demonstrated in oral squamous cell carcinoma (OSCC), in which SIRT3 protein levels are elevated; its downregulation enhanced apoptosis and induced colony formation. Furthermore, cells with low levels of SIRT3 are more sensitive to cisplatin and ionizing radiation, indicating that presence of SIRT3 can induce resistance to chemotherapy and apoptosis in these cancer cells (119). The tumor promoting properties of SIRT3 were also reported in melanomas and RCC, where mRNA and protein levels of SIRT3 were increased compared with their nontumorigenic counterparts (119–121). Knockdown of SIRT3 in multiple melanoma cell lines resulted in decreased proliferation, inhibited colony formation, and reduced migration capacity and these cells exhibited senescence-like phenotype and dysregulation of the cell cycle. In a xenograft mouse model, SIRT3 knockdown caused a decrease in tumor growth (122). In RCC cell lines, SIRT3 was shown to have a role in glutamine oxidation, as SIRT3 knockdown had impaired proliferation, glutamine oxidation, and significantly reduced activity of glutamate dehydrogenase (GDH; ref. 123). SIRT3 has also been identified as rescuing EJ bladder carcinoma cells from p53-induced senescence and growth arrest by interacting with p53′s MASD (mitochondria-associated senescence domain) region between amino acid 64 and 209. It was shown to exert deacetylation activity in vitro on p53 peptide sequences (124).

Sirtuin 4 is widely expressed in multiple tissues such as pancreas or human muscle (125). As is the case with SIRT3, its main function is also related to metabolism; it decreases the activity of GDH (126) and inhibits insulin secretion in pancreatic β cells (127). Under nutrient-depleted conditions, SIRT4 is activated and deacetylates malonyl-CoA decarboxylase, thus inhibiting the conversion of malonyl-CoA to acetyl-CoA, a crucial reaction in lipid metabolism. Indeed, mice lacking SIRT4 show an increase in malonyl-CoA decarboxylase activity and display an imbalance in lipid metabolism, protecting them against diet-induced obesity (128). It has been found that SIRT4 also has a lysine deacylase enzymatic activity; it removes acyl moieties from lysine residues (methylglutaryl-, hydroxymethylglutaryl-, and 3-methylglutaconyl-lysine), thus affecting leucine metabolism and insulin secretion (129).

SIRT4 was found to be downregulated in hepatocellular carcinoma, colorectal, gastric, and lung cancer and esophageal squamous cell carcinoma (130–133). In gastric cancer, its downregulation was related with depth invasion and lymph node–positive cases (134). SIRT4 was discovered as a tumor suppressor involved in cellular metabolic response to DNA damage in lung cancer. The mechanism of action involves regulation of glutamine metabolism. This occurs via inhibition of GDH activity, which converts α-ketoglutarate (α-KG) into l-glutamine (135). Moreover, SIRT4 reconstitution decreased glutamine uptake, but not glucose, and repressed proliferation (135). Since this initial study, the involvement of SIRT4 in glutamine metabolism was observed and confirmed in other cellular models. A microarray study of 89 colorectal cancer cases showed that SIRT4 protein levels were decreased compared with the normal tissue and this decrease correlated with a worse prognosis. Its overexpression inhibited the proliferation of colorectal cancer cells in vitro and in vivo (131). Modulation of glutamine metabolism was once again found to be important for these effects as SIRT4 weakened the ability of colorectal cancer cells to utilize glutamine and enhanced cell death caused by glucose metabolism inhibitor 2-deoxyglucose. SIRT4′s role in metabolic regulation was further confirmed by its ability to inhibit pyruvate dehydrogenase (PDH) complex through the hydrolysis of the lipoamide cofactor (136), which binds to the E2 catalytic subunit and it is essential for PDH function (137). In colorectal cancer, SIRT4 influences EMT, leading to upregulation of E-cadherin expression and inhibition of migration, proliferation, and invasion by negatively affecting glutamine metabolism (138). This occurs through SIRT4′s repression of the activity of GDH. Supplementation of cells with α-KG abrogates the expression of E-cadherin by SIRT4 (138). The EMT-suppressive role of SIRT4 was further confirmed in gastric cancer, where overexpression of SIRT4 increased the expression of E-cadherin, and downregulated N-cadherin and vimentin (132). SIRT4 is also downregulated in lung cancer (133) and lung cancer cells transfected with SIRT4 exhibited inhibition of cell proliferation, cell cycle, cell invasion, and migration through the inhibition of the mitochondrial fission protein dynamin-1-like protein (Drp1) phosphorylation and downregulation of MEK/ERK signaling pathway (133).

Recent studies in esophageal squamous cell carcinoma (ESCC) demonstrated, that SIRT4 is upregulated in tumor tissues when compared with the adjacent normal tissue and SIRT4 levels inversely correlated with the mean survival time of ESCC patients (139). A similar phenotype was found in breast cancer, where SIRT4 proteins levels were increased compared with the nonneoplastic tissue counterparts and promoted migration, invasion, and proliferation of breast cancer cells (140). In a cell line derived from liver carcinoma, HepG2, SIRT4 overexpression under cellular stress conditions increased the cell survival and protected cancer cells against stress-induced cell death. Furthermore, SIRT4 promoted growth after DNA-damaging conditions such as radiation, UV irradiation, and cisplatin exposure (141).

FUS1 (fusion 1 protein), also known as TUSC2 (tumor suppressor candidate 2) is a small, 110 AA-long, ubiquitously and highly conserved mitochondrial protein. It has been described as a regulator of Ca²⁺ ion uptake in the mitochondria, having a protective role in autoimmunity, cancer, and inflammation (142).

FUS1, localized on the chromosome 3p21.3 (143), is frequently involved in cancer. Its expression is absent or reduced in majority of lung cancers (144). In nasopharyngeal carcinoma, its downregulation is controlled by miR-663b (145). miR-93, miR98, and miR-197 were also shown to bind to the 3′UTR region of the FUS1 transcript and decrease its expression (146). The expression of FUS1 is negatively regulated by sequence elements in both, 5′UTR and 3′UTR, regions of FUS1 mRNA (147). The tumor-suppressive properties of FUS1 were demonstrated through overexpression studies in lung cancer cells, which resulted in induction of apoptosis and inhibition of cell growth due to a cell-cycle arrest (Fig. 1; refs. 148, 149).

Expression of both FUS1 and p53 synergistically decreased tumor growth and induced apoptosis in NSCLC. Levels of MDM2 were decreased when cells were cotransfected with p53 and FUS1, suggesting that the tumor suppressive role may be linked to the accumulation of p53 through FUS1-mediated downregulation of MDM2 (150).

FUS1 was characterized as an inhibitor of several tyrosine kinases such as c-Abl, EGFR, platelet-derived growth factor receptor (PDGFR), and c-Kit (149, 151). Moreover, it was found that FUS1 is linked to the AMPK/AKT/mTOR signaling pathway in liver kinase B1 (LKB1)–defective NSCLC cells (152). LKB1 is a tumor suppressor, which is mutated or inactivated in 50% of NSCLC and was shown to activate the apoptotic regulator AMPK (153, 154) what leads to inhibition of AKT. FUS1 increases the effects of the AKT inhibitor, MK2206, by enhancing cell's sensitivity to this inhibitor. Combination of both FUS1 reexpression and MK2206 decreased tumor growth in a human LKB1-defective H322 xenograft mouse model. In vitro experiments showed that increased AMPK phosphorylation and the resulting increased kinase activity are responsible for this effect. AMPK knockout disrupted the FUS1–MK2206 cooperation. Moreover, combination of MK2206 treatment and FUS1 reexpression also decreased the phosphorylation and the enzymatic activity of mTOR and AKT. These results establish a connection between FUS1 expression and the AMPK/AKT/mTOR signaling pathway axis in the LKB1-deficient cells (152).

In mitochondria, FUS1 regulates calcium homeostasis. Epithelial cells, splenocytes, and activated CD4⁺ T cells with FUS1 knockout exhibited reduced calcium uptake into mitochondria what led to increased mitochondrial membrane potential and ROS generation. Increased ROS generation and NF-κB activation can lead to downregulation of programmed cell death protein 1 (PD-1) and programmed-death ligand 1 (PD-L1) levels in T lymphocytes (142), thus enhancing the autoimmune response and suppressing tumor immune evasion. FUS1 protein levels are increased during immune activation and are repressed by asbestos, tobacco exposure, or other environmental assaults in vivo (155). Fus1^−/− mice showed an increase in chronic inflammation, suppression of antitumor defense, dysregulation of mitochondrial membrane potential, increased oxidative stress and genotoxicity, and dysregulation of the mitochondrial uncoupling protein 2 (UCP2) expression (155). These experiments support the role of FUS1 as a link between the mitochondrial homeostasis and the inflammatory response.

SDH is an enzymatic complex located within the inner mitochondrial membrane, responsible for the oxidation of succinate to fumarate accompanied by the reduction of ubiquinone to ubiquinol in electron transport chain. SDH is a heteromeric complex formed by four nuclear-encoded subunits: SDHA, SDHB, SDHC, and SDHD (156, 157). All subunits are essential for its proper function; for this reason, deleterious mutations in any of the SDH genes invariably result in a decreased SDH activity. Defects of SDH are comparatively rare in humans (156). Mutations or its inhibition leads to accumulation of succinate in mitochondria, causing disruption in ATP generation and mitochondrial impairment (156, 157). SDH is known to be involved in neuroprotection and its mutations can correlate with neurodegenerative disorders such as Parkinson disease (158). Moreover, inherited defects of SDH in humans are associated with encephalomyelopathy namely Leigh syndrome (157).

Heterozygous germline mutations in SDHA, SDHB, SDHC, and SDHD have been identified to cause hereditary paragangliomas, pheochromocytomas, gastrointestinal stromal tumor (GIST), RCC, and breast cancer (159–161). Because of SDH's role in the electron transport chain, it has been suggested that ROS production can be induced by SDH-inactivating mutations showing the potentially big impact of an enzymatic complex on the genomic instability, apoptosis, and neoplastic transformation due to ROS accumulation (Fig. 1; refs. 162, 163). In cells with inhibited SDH, succinate accumulated in mitochondria is transported to the cytosol. Elevated cytosolic levels of succinate inhibit the activity of HIF prolyl-hydroxylase (PHD). This process stabilizes HIF1α what leads to expression of genes involved in the promotion of angiogenesis, metastasis, and metabolism, contributing to more aggressive tumor phenotypes (164).

The subunit A and B of the SDH complex are affected by germline mutations such as missense or nonsense mutations resulting in a loss of expression of both subunits (159, 165). It was also demonstrated that the SDHB promoter could be subjected to epigenetic modifications in primary neuroblastomas and phaeochromocytomas (166). The mutations of SDHB subunit are mostly present in paraganglioma, phaeochromocytoma, and in neuroblastoma cancer development, but SDHB and SDHA are also involved in breast cancer albeit with low frequencies; SDHA is mutated in 3.19% and SDHB in 0.1% of patients with breast cancer (161). SDHB is known to be involved in GIST. Twenty-seven percent of patients with GIST lack SDHA expression, which can be caused by heterozygous and somatic loss of both alleles (165).

Subunit B is involved in the EMT process in various cancer types. A study published in 2016, demonstrated that SDHB downregulation facilitates the EMT process through hyperactivation of TGFβ signaling pathway and an increase in SNAIL1-SMAD3/4 mRNA and protein levels in colorectal cancer (167). SDHB-deficient metastatic pheochromocytomas/paragangliomas displayed a dramatic change in SNAI1/2 localization, which was translocated to the nucleus in all metastatic tumors. This translocation reflects the activation of SNAI1/2 as a transcription factor, which promotes the initial steps of the EMT (168). A connection between low SDH expression and EMT has also been described in serous ovarian cancer (162). In mice, Sdhb knockdown promotes histones hypermethylation (such as hypermethylation of H3K27), which, through promotion of EMT, contributes to the induction of a tumorigenic and metastatic phenotype. SDHB knockdown led to upregulation of various transcription factors involved in EMT, in particular to a dramatic increase of TWIST2 mRNA levels with a consequent downregulation of E-cadherin. These global epigenetic changes contributed to the reprogrammed carbon source utilization and tumor progression of ovarian cells in mice (162).

SDHC also seems to be involved in gastrointestinal and breast cancer (169, 170). In breast cancer, its low expression promotes the expression of EMT genes such as vimentin (VIM), SNAI2, TWIST, and AXL (170).

Mutations of subunit SDHD cause hereditary paraganglioma (171). Furthermore, SDHD mutations are associated with sporadic thyroid cancer (172). The mechanism, which promotes cancer progression in thyroid cells, is connected to the aberrant autophagy process in cells expressing mutated variant of SDHD. When functioning properly, SDHD prevents the degradation of forkhead box protein O3 (FOXO3a), a positive autophagy regulator (172). In cutaneous melanomas, SDHD expression is downregulated due to promoter mutations. The lower SDHD expression is correlated with a worse prognostic feature (173, 174).

MTUS1, also known as mitochondrial tumor suppressor 1, was discovered in 2003 and is located on chromosome 8p21.3–22 (175). The MTUS1 gene encodes six isoforms also known as Angiotensin-II type 2 receptor-interacting proteins (ATIP), signaling transducers that interact with AT₂-receptor: ATIP1, ATIP2, ATIP3a, ATIP3b, ATIP4, and MTUS1 transcript variant 7 (176, 177). MTUS1 was named “mitochondrial tumor suppressor gene 1” according to its function, but its role is not strictly of a tumor suppressor, as it is involved in vascular remodeling, adipocyte differentiation, lymphocytosis, and cardiac hypertrophy (178). In endothelial cells, MTUS1 knockdown inhibits endothelial tube formation and migration. Moreover, MTUS1 is important for mitochondrial dynamics, motility and ROS production (178). MTUS1 is ubiquitously expressed in all tissues (175). Its major transcripts ATIP1, ATIP3, and ATIP4 have different tissue distribution. ATIP1 and 4 are abundantly present in the brain (cerebellum and fetal brain). ATIP3 is expressed in prostate, bladder, ovary, colon, and breast (179). In 2016, Wang and colleagues, described MTUS1 as a mitochondrial protein able to regulate the cytokine production, thus attributing it with an anti-inflammatory role. Silencing MTUS1 in endothelial cells activates NF-κB and p38MAPK, which are the main signaling pathways responsible for the inflammatory response in cells (180).

MTUS1 deficiency is associated with multiple types of cancer. A significant (50%) downregulation of MTUS1 was found in patients suffering from colon cancer (181). This downregulation was not a consequence of mutations in the gene but rather occurred on a post-transcriptional level. The downregulation resulted in increased cellular proliferation, whereas MTUS1 overexpression reduced the proliferation (181). MTUS1 expression is also reduced in gastric cancer and is associated with lymph node metastasis and poor overall survival rate (182). Furthermore, MTUS1 mRNA expression is downregulated in patients with RCC and correlates with poor prognosis, whereas its expression resulted in significantly longer survival time (183).

In human prostate cancer, the downregulation of ATIP/MTUS1 is influenced by EGF (184). Treatment with EGF decreased the mRNA levels of ATIP1 with a subsequent stimulation of cell growth. This effect was reversed by the activation of the AT₂ receptor, which, via an unknown mechanism, promotes the formation of the ATIP/SHP-1 (Src homology 2 domain-containing protein–tyrosine phosphatase 1) complex. This complex, upon translocation to nucleus, promotes the inhibition of the proliferation-signaling cascade. The formation of this complex prevents the ATIP downregulation by EGF, leading to a decrease of the proliferation rate in prostate cancer cells (184). MTUS1 expression levels are also regulated by DNA methylation; in NSCLC, its promoter is methylated causing protein downregulation, whereas inhibition of DNA methylation restored MTUS1 expression levels (185). Another mode of MTUS1 regulation comes from miRNAs control, specifically in osteosarcoma, breast cancer, colorectal carcinoma, and lung cancer (186–189). In osteosarcoma, MTUS1 is a direct target of miR-765. The overexpression of the miR-765 promotes cell proliferation, migration, and invasion and is accompanied by activation of the ERK/EMT pathway (188). MiR-19a and miR-19b coregulate MTUS1 to promote cell proliferation and migration in lung cancer (186).

Failure to undergo cellular differentiation is one of the key events during cancer formation and in 2003, a study from Seibold and colleagues demonstrated a potential role of MTUS1 in this fundamental process (175). In human umbilical vein endothelial cells (HUVEC), under 3D culture conditions, MTUS1 is upregulated during the initial stages of the differentiation process. Similarly, the mRNA levels of MTUS1 in colon tumor tissue and different pancreatic tumor cell lines showed low expression in undifferentiated proliferating cancer cells and higher expression in differentiated and slowly proliferating cancer cells (175, 181). Further studies confirmed that downregulation of MTUS1/ATIP is a frequent event during the progression of oral tongue squamous cell carcinoma where this process is correlated with poor differentiation and enhanced proliferation (190).

Even though mitochondrial tumor suppressors are a diverse group of proteins, they share some underlying characteristics (Supplementary Table S1). First, several of them either directly (POX, SDH, FH) or indirectly (LACTB, sirtuins) participate in metabolic pathways, showing the importance of metabolic dysregulation in cancer. Second, many of them are connected to ROS generation (POX, SIRT3) as well as HIF signaling (FH, SDH) confirming that the cellular reprogramming triggered by (pseudo)hypoxia is an important element of tumor development. Third, several of them are connected to the notorious p53 tumor suppressor, either being induced by it (POX) or modulating its activity (LACTB, SIRT3, FUS1). Because p53 is mutated in a large number of cancers, these interactions could prove to be crucial in improving our concept of tumor suppression in general. Because the discovery and characterization of mitochondrial tumor suppressors is a new and emerging field, its translation into a therapeutic setting is still in its beginnings. Concerning LACTB tumor suppressor pathway, one possible therapeutic approach involves the ability of LACTB to downregulate/inhibit the mitochondrial PISD enzyme, thus promoting cancer cell death and/or differentiation. Indeed, it was shown that doxorubicin, a drug widely used in chemotherapy, is able to modify the mitochondrial membrane composition via PISD pathway inducing cell death in HeLa cells (191). In colorectal cancer, LACTB inhibits EMT and proliferation through PI3K inactivation and the use of PI3K inhibitors was already suggested as a possible therapeutic intervention in colorectal cancer (23). Untangling the dual, physiologic, and cancer type–specific role of POX (and proline metabolism as a whole) could lead to novel ways of treatment. The inhibition of collagen biosynthesis together with upregulation of POX might represent one such example (192). Furthermore, several downstream effectors of POX, such as COX-2, β-catenin, and EGFR, are already being used and researched as promising therapeutic targets in various types of cancer (193–195). Alterations in DNA damage response that accompany FH deficiency (and subsequent fumarate accumulation) could be exploited in therapy since these cells were shown to be vulnerable to synthetic-lethal targeting with PARP inhibitors (83). Because FH is an enzyme of a fundamental metabolic pathway, looking for potential therapy targets in the “rewired” metabolic pathways has likewise proven successful. Synthetic lethality has been discovered between FH and adenylate cyclases as well as genes of heme metabolism (92). As reviewed in Kancherla and colleagues, several other therapies have been proposed to target the FH tumor-suppressive pathway (196). FH deficiency leads to HIF accumulation and increased hypermethylation. Bevacizumab and erlotinib are used in clinic to treat patients with increased HIF levels. Regarding the hypermethylation, a current phase II clinical trial aims to evaluate the efficacy of SGI-110 (guadecitabine), a DNA methyltransferase inhibitor (clinical trial identifier: NCT03165721), in patients with HLRCC-associated RCC. Increased MET oncogene expression has also been detected in FH-deficient cells. An ongoing phase II clinical trial is aiming to compare survival between RCC patients treated with c-Met inhibitors versus sunitinib, the standard of care for RCC (clinical trial identifier: NCT02761057). Given the physiologic importance of sirtuins in the mitochondria and their key role in cancer, they have become attractive targets for drug design. The first described sirtuin activator, called resveratrol, is a plant metabolite present in grapes. Although it was originally described to activate human sirtuin 1 (197), later it also reported to affect the activity of sirtuin 3 and sirtuin 5 (198). However, several serious adverse reactions were reported that prevented its use in clinic and new analogues are being developed that could activate sirtuins at much lower doses (199). There is a growing field of nature-derived sirtuin 3 activators, such as honokiol (200, 201) and oroxylin A (202) that could prove useful in therapy development. Downstream effectors of sirtuins might also be taken into consideration. For instance, SIRT3 induces apoptosis through BCL, caspase-3 and p53 stabilization might increase the SIRT3 downstream effects. SIRT4 inhibited cells invasiveness through ERK–Drp1 signaling pathway, impairing mitochondrial dynamics. Inhibition of Drp1 by Mdivi-1 blocked cell cycle progression through G₂–M arrest (203). Regarding FUS1 protein, a phase I clinical trial of FUS1-nanoparticles mediating functional gene transfer in humans was already completed (204), clinical trial identifier: NCT00059605. Combination of FUS1-nanoparticles with cisplatin or erlotinib showed an enhancement of the therapeutic effects of these drugs, pointing out the important role of FUS1 in modulating chemosensitivity of lung cancer cells (205, 206). Taking into account that c-Abl is suppressed by FUS1 in lung cancer, inhibitors targeting c-Abl, such as GNF5, might be used in future for treatment of lung cancer (207). Concerning SDH tumor-suppressive pathway, the most promising approaches are based on the prevention of the succinate accumulation, which is thought to be the major actor of SDH-related oncogenic processes. Some of the therapeutic approaches in SDH-mutated metastatic cancers involve targeting metabolic reprogramming, redox imbalance, pseudohypoxia, or epigenetic reprograming. Among these drugs are inhibitors of oxidative phosphorylation, HIF2a and PARP, or demethylating agents that currently are or will soon be tested in clinical trials (208).

Although the studies of mitochondrial tumor suppressors are still at a relatively early stage, they already provided many valuable insights into biology and dependencies of cancer cells and deepened our appreciation on the important role of mitochondria in tumor progression. Numerous in vitro and in vivo studies highlighted the importance of these tumor suppressors in various biological processes, ranging from regulation of metabolic pathways, ROS production, and hypoxia to promotion of differentiation, apoptosis, and tumor-suppressive metabolic states. These studies also led to emergence of many, as yet unanswered question regarding their role and mechanism of action in cancer cells. How are these tumor suppressors inactivated in cancer cells, and what are the genetic and epigenetic variations of these genes in human cancers? What are the identities of their physiologic upstream regulators and how do changes in mitochondrial processes lead to cancer cell differentiation? Which one of the oncogenic cascades activated in the absence of these tumor suppressors should be targeted to give us the best therapeutic benefits? Does the inactivation of these genes occur in the early stages of transformation or do they emerge at a later stage of tumor progression? Do these mitochondrial tumors suppressor interact with each other? Given the fact that many of them share similar mechanisms of action, it is probable that they might interact and modulate each other's activity, thus creating a general tumor-suppressive landscape within mitochondria. Answers to these questions are important for fully defining the role of these tumor suppressors in tumorigenesis and for applying this knowledge toward more efficient cancer treatments.

No disclosures were reported.

The authors thank Tom DiCesare for graphical assistance with presented images. This work was supported by the Czech Science Foundation (18–24473Y), IOCB Postdoctoral Fellowship (RVO 61388963), EMBO Installation Grant and Ministry of Education, Youth and Sports, and the European Social Fund [OP RDE; Project: “IOCB MSCA Mobility III” (No. CZ.02.2.69/0.0/0.0/19_074/0016322)].