Deregulation of mitochondrial gene expression in cancer: mechanisms and therapeutic opportunities

Abstract

“Reprogramming of energy metabolism” was first considered an emerging hallmark of cancer in 2011 by Hanahan & Weinberg and is now considered a core hallmark of cancer. Mitochondria are the hubs of metabolism, crucial for energetic functions and cellular homeostasis. The mitochondrion’s bacterial origin and preservation of their own genome, which encodes proteins and RNAs essential to their function, make them unique organelles. Successful generation of mitochondrial gene products requires coordinated functioning of the mitochondrial ‘central dogma,’ encompassing all steps necessary for mtDNA to yield mitochondrial proteins. Each of these processes has several levels of regulation, including mtDNA accessibility and protection through mtDNA packaging and epigenetic modifications, mtDNA copy number through mitochondrial replication, mitochondrial transcription through mitochondrial transcription factors, and mitochondrial translation through mitoribosome formation. Deregulation of these mitochondrial processes in the context of cancers has only recently been appreciated, with most studies being correlative in nature. Nonetheless, numerous significant associations of the mitochondrial central dogma with pro-tumor phenotypes have been documented. Several studies have even provided mechanistic insights and further demonstrated successful pharmacologic targeting strategies. Based on the emergent importance of mitochondria for cancer biology and therapeutics, it is becoming increasingly important that we gain an understanding of the underpinning mechanisms so they can be successfully therapeutically targeted. It is expected that this mechanistic understanding will result in mitochondria-targeting approaches that balance anticancer potency with normal cell toxicity. This review will focus on current evidence for the dysregulation of mitochondrial gene expression in cancers, as well as therapeutic opportunities on the horizon.

The mitochondrial ‘central dogma’

Mitochondria evolved from bacteria and became a major organelle for eukaryotes through endosymbiosis [1]. As originally independent organisms, mitochondria have a maternally inherited genome separate from the eukaryotic nuclear genome [2]. The mitochondrial genome is 16,569 bp, encoding 37 genes [3]. These 37 genes include 2 rRNAs, 22 tRNAs, and 13 mRNAs. The 13 mRNAs encode proteins for the electron transport chain (ETC) complexes I, III, IV, and V. Mitochondrial DNA (mtDNA) is circular and double-stranded with heavy and light strands. The light strand includes instructions for one mRNA encoding for the protein ND6, part of the core catalytic subunit of complex I, and eight tRNAs. The heavy strand encodes the other 12 mRNAs, 2rRNAs, and 14 tRNAs [4]. Unlike eukaryotic nuclei that are diploid, mitochondria can contain as many as 100 – 10,000 copies of mtDNA. Mitochondrial replication and transcription occur independently of the cell cycle and initiate in the displacement loop (D-loop) noncoding region [2]. Mitochondrial replication, transcription, and translation require a coordinated effort between nuclear-encoded proteins and mitochondrial-encoded rRNAs and tRNAs. Approximately 1000-1500 proteins localize within the mitochondria, most of which are encoded in the nuclear genome (nDNA) [5]. These nuclear-encoded mitochondrial proteins have a 20-30 amino acid mitochondrial localization sequence (MLS) that directs them to the mitochondrial matrix, which contains mitochondrial nucleoids that encase mtDNA [4, 6].

The Human Mitochondrial Genome: from basic biology to disease Chapters 1 and 2 comprehensively review mitochondrial replication, transcription, and translation [1, 7]. Briefly, mtDNA is replicated by nuclear-encoded proteins: DNA polymerase, POLG, mtDNA helicase, TWINKLE, and mitochondrial single-stranded DNA-binding protein, mtSSB [7]. Replication is based on the strand displacement model, where there is continuous replication in opposite directions from both the heavy- and light-strand until the circle is complete. Replication begins at the origin of replication on the heavy-strand by POLG and TWINKLE. The free parental heavy-strand is stabilized and protected by mtSSB to prevent mitochondrial RNA polymerase, POLRMT, from early RNA synthesis. When POLG and TWINKLE pass the origin of replication on the light strand, the stem-loop structure is formed by the parental strand and mtSSB can no longer bind. While POLRMT initially binds and initiates RNA synthesis, it is then replaced by POLG for DNA replication to occur on the light strand. Termination of mtDNA replication occurs when both the heavy- and light-strand have been replicated. First, the primers at the origins of replication are removed by the mitochondrial genome maintenance exonuclease 1, MGME1. Then, strands are ligated by DNA ligase 3 (LIG3). Finally, the new mtDNA is separated from the parental strand by topoisomerase TOP3A [7, 8].

mtDNA is packaged in nucleoids by mitochondrial transcription factor A (TFAM). TFAM is required for transcription initiation. TFAM binds to mtDNA upstream of promoters and recruits transcription factor B2 of the mitochondria (TFB2M) and POLRMT. TFB2M binds to the DNA and forms the initiation complex. Following transcription initiation, the human mitochondria elongation factor, TEFM, facilitates elongation by binding to the non-template strand of the mtDNA, which then leads to the release of TFB2M. Transcription occurs bidirectionally from the three promoters and generates polycistronic RNA transcripts. For transcription termination, mitochondrial termination factor 1 (MTERF1) binds to the mtDNA and blocks TWINKLE and the mitochondrial mtDNA replisome [4, 6]. Following transcription termination, the mRNA transcripts are transferred to mitochondrial RNA granules (MRGs) for storage or go directly to the mitochondrial translation machinery to be translated into protein [5].

Mitochondrial translation is carried out by the 55S mitoribosome, which consists of two subunits: the 28S small subunit (mtSSU), made up of 29 proteins, and the 39S large subunit (mtLSU), made up of 50 proteins. The two mitochondrial rRNAs are the small 12S and large 16S [9]. Mitochondrial translation begins with the mitochondrial initiation factors mtIF1 and mtIF3, which signal the formation of the initiation complex to allow the mRNA to bind to the mtSSU. The mitochondrial elongation factor Tu (mtEFTu) directs elongation by binding the tRNA to the mitoribosome and mitochondrial elongation factor G1 (mt EFG1). Translation termination is facilitated by mitochondrial release factor 1a (mtRF1a) when a stop codon (UAA or UAG) is reached, leading to hydrolysis of the tRNA and nascent peptide [2].The mitoribosome, with the assistance of oxidase (Cytochrome C) assembly 1-like (OXA1L) and transmembrane protein 126A (TMEM126A), translates the 13-mitochondrial encoded proteins then inserts them into the inner mitochondrial membrane (IMM) where the ETC resides [2, 10, 11]. The ETC itself is made up of 92 proteins; 79 of them are nuclear-encoded and localized to the translocase outer mitochondrial membrane (TOMM) complex by mitochondrial chaperone proteins [2]. Interestingly, recent findings support the notion that most mitochondrial-localized nDNA-encoded proteins are locally translated at or near the surface of the outer mitochondrial membrane (OMM) [12, 13]. The proteins are then transferred through TOMM to the translocase inner mitochondrial membrane (TIMM) 23 complex to be inserted directly into the IMM. Alternatively, if proteins have the presequence translocase-associated import motor (PAM), they are imported into the matrix. Once these nuclear-encoded proteins reach the matrix, the mitochondrial processing peptidase cleaves the PAM sequence, and the protein associates with OXA1L to be inserted in the IMM to form the five ETC complexes [5].

Mitochondrial oxidative phosphorylation (OXPHOS) is a major source of energy and redox balance, and plays a crucial role in many cancer phenotypes. It is important to note that the role of OXPHOS in cancer is highly context-dependent, with many studies demonstrating pro- or anti-tumor roles of mitochondrial metabolism, as recently reviewed [14]. This context dependence is likely the result of the tissue of origin, local oxygenation, access to nutrients in the tumor microenvironment, the tumor’s genetic background, and exposure to exogenous stressors such as systemic or radiation therapies. Each of the ‘central dogma’ steps in mitochondrial gene expression is adaptable and responsive to external stimuli and stressors, thus providing cancer cells with opportunities to deregulate them for survival benefit. A clear understanding of these mechanisms in the context of cancer holds promise to support the development of anticancer mitochondria-targeting therapies. Below, we review studies from the cancer literature yielding insights into how tumor cells deregulate mitochondrial gene expression. Further, we explore the status of the efforts to therapeutically target these mechanisms in cancer (Fig. 1 and Table 1).

Fig. 1. Targeting the mitochondrial ‘central dogma’ in cancer.

Mitochondria contain their own genome, which enables them to have their own machinery for replication, transcription, and translation. Deregulation of these processes has been observed in various cancer contexts, providing a rationale for therapeutic targeting of the mitochondrial central dogma. Red text with blunt red arrows: the current and developing inhibitors to target the ‘central dogma’ processes. Blue text with an arrow: this drug stabilizes and promotes the ‘central dogma’ process. Small blue circles on the ETC represent the number of mt-encoded proteins for the given ETC subunit (e.g., Complex I has seven mt-encoded proteins, and Complex II does not have any mt-encoded proteins). Abbreviations; POLG (DNA polymerase gamma), TWINKLE (mtDNA helicase), mtSSB (mitochondrial single-stranded DNA-binding protein), NRTI (nucleotide reverse transcriptase inhibitor), ddC (nucleoside analog 2’3’-dideoxycytidine), TFAM (mitochondria transcription factor A), TFB2M (transcription factor B2 of the mitochondria), POLRMT (mitochondrial RNA polymerase), MTERF1 (mitochondrial termination factor 1), IMT1/IMT1B (inhibitor of mitochondrial transcription), D26 (POLRMT inhibitor compound), TMP (tetramethylpyrazine), mitoribosome (mitochondrial translation machinery), MRP (mitochondria ribosomal proteins), rRNA (ribosomal RNA), mtIF1/mtIF3 (mitochondrial initiation factors), mtEFTu/mtEFG1 (mitochondrial elongation factors), mtRF1 (mitochondrial release factor 1a), COL-3 (chemically modified tetracycline, 6-deoxy, 6-demethyl, 4-de-dimethylamino tetracycline), JG-98 (allosteric inhibitor of HSP70), SAMMSON ASO (survival associated mitochondrial melanoma specific oncogenic non-coding RNA antisense oligonucleotide).

Table 1.

Evidentiary summary on current and developing mitochondrial-targeted therapies in cancer.

Targeted aspect of mitochondrial gene regulation	Molecular Target	Therapy	Reference #	Cancer Type	Type of Evidence
mtDNA Replication	POLG	ddC	68	AML	Pre-clinical
			72	PDAC	Pre-clinical
			69	ALL	Pre-clinical
			70	Cervical	Pre-clinical
			71	Pancreatic	Pre-clinical
mt-transcription	POLRMT	IMT1/IMT1B	75	Cervical and Colon	Pre-clinical
			77	Endometrial	Pre-clinical
			78	Colon, Cervical, PDAC, Lung	Pre-clinical
			79	Osteosarcoma	Pre-clinical
		D26	76	Ovarian	Pre-clinical
	TFAM	Melatonin	80	Glioblastoma	Pre-clinical
		TMP	81	Cervical and Colon	Pre-clinical
mt-translation	Mitoribosome	Antibiotics	84	Breast	Pre-clinical
			85	Leukemia	Pre-clinical
			86	Lung, Colon, PDAC	Pre-clinical
			88	Breast, Ovarian, Lung, Melanoma, Prostate, Pancreatic, Glioblastoma	Pre-clinical
			90	AML	Clinical
			91	AML	Pre-clinical
			96	Breast	Clinical
			93	Gastric	Pre-clinical
			92	Lymphoma	Pre-clinical
			94	Ovarian	Pre-clinical
			95	Breast	Pre-clinical
		Antibiotics, COL-3	87	Lung, colon, PDAC	Pre-clinical
		COL-3	83	Prostate	Pre-clinical
	HSP70	JG-98	98	Prostate	Pre-clinical
	lncRNA SAMMSON	SAMMSON ASO	59	Uveal Melanoma	Pre-clinical

A summary of therapeutic agents and their molecular targets discussed in the review.

Deregulation of mtDNA packaging and epigenetic modification in cancer

The integrity of mtDNA can be impacted by mtDNA mutation, mtDNA damage and repair, mtDNA packaging into nucleoids, and epigenetic modifications. As the roles of mtDNA mutations [15, 16] and mtDNA damage and repair [17, 18] in cancer have been highlighted in recent reviews, they will not be covered here. mtDNA is packaged into nucleoids by TFAM and localized in the mitochondrial matrix [7]. Only recently have the mechanisms of mtDNA packaging and accessibility for replication and transcription been investigated. TFAM was found to bind specific mtDNA sites, which drove the binding of the rest of the mtDNA through cooperative binding. Furthermore, it was observed that TFAM levels modulate mtDNA accessibility, where cells with a high TFAM:mtDNA ratio had increased compaction of mtDNA into nucleoids inaccessible to the transcription and replication machinery. It was speculated that these inaccessible nucleoids could serve as a reservoir of healthy mtDNA that could be utilized when increased transcription and replication output was necessitated [6]. Supporting the idea of nucleoid formation as protection of mtDNA, increased packing of newly replicated mtDNA into nucleoids with higher levels of TFAM was observed upon treatment with doxorubicin, a commonly used DNA-damaging chemotherapy, or ethidium bromide (EtBr), suggesting the protective effect of TFAM and nucleoid packaging under stress [19]. Thus, given its role in nucleoid packaging and mtDNA accessibility, TFAM is a major regulator of mtDNA. Several studies outlined below support the importance of TFAM in cancers, although its dual role in both mtDNA replication and transcription is tightly intertwined and thus not easily distinguished without further mechanistic investigation.

Epigenetic modifications of mtDNA affect mitochondrial gene regulation [20]. There are limited reports of mtDNA being epigenetically modified, but of those in  the literature, it is of mtDNA methylation. In breast cancer, eight mtDNA methylation sites were observed to be significantly methylated in the mtDNA extracted from the peripheral blood of breast cancer patients compared to matched samples from family members without breast cancer [21]. In contrast to breast cancer, mtDNA extracted from colorectal tumors had greater demethylation within the mtDNA D-loop region compared to corresponding non-cancer tissues. This demethylation also correlated with greater complex I ND2 protein expression [22]. It was speculated that high ND2 expression would support Complex I and thus OXPHOS, but this was not tested functionally [22]. In liver cancer, N6-methyldeoxyadenosine (6mA) was found to be enriched in purified mtDNA compared to the total gDNA from HepG2 cells. This reduced the binding affinity of TFAM, leading to decreased mtDNA copy number and mtDNA transcription [23]. Furthermore, the investigators showed knock down of the 6 mA methyltransferase, METTL4, increased mtDNA copy number, increased mtDNA transcription, and increased both OXPHOS and glycolysis. This study provides proof of concept that mtDNA methylation can play a functional role in cancer by modulating mtDNA expression and downstream metabolic pathways, suggesting cancer cells may hijack the mtDNA machinery to enable pro-survival mitochondrial functioning. Clearly, studies providing these functional insights in cancer are extremely limited and should be undertaken more broadly.

Deregulation of mtDNA copy number and mtDNA replication in cancer

There is strong evidence supporting the significant association of mtDNA copy number (mtDNA-CN) with cancer incidence, progression, and treatment responses. mtDNA can be extracted from either the peripheral blood or from tumor tissue itself. The first set of studies reviewed here will be correlative studies evaluating mtDNA-CN from peripheral blood. In a large population-based prospective study, 3225 middle-aged women without cancer had peripheral blood drawn when they joined the study and then were monitored for an average of 15 years until cancer diagnosis, death, or the end of the study. It was found breast cancer, respiratory system cancer, and skin cancer incidences were all associated with higher mtDNA-CN, although only the breast cancer association reached statistical significance [24]. Interestingly, lower mtDNA-CN was associated with the incidence of genital organ, urinary system, hematological, nervous system, and endocrine gland cancers [24]. For genital organ cancers, lower mtDNA-CN was found to be associated with an increased risk of cancer-specific mortality [24]. A similar analysis of mtDNA extracted from the peripheral blood of participants from the prospective European Prospective Investigation into Cancer and Nutrition (EPIC) study revealed that elevated mtDNA-CN was associated with a reduced risk of developing pancreatic ductal adenocarcinoma [25]. Additionally, others have observed an increase in mtDNA-CN, extracted from blood, matched to control samples in breast cancer [26], renal cell carcinoma [27] and non-Hodgkin lymphoma [28]. Collectively, these studies demonstrate that blood mtDNA-CN associations are cancer-type specific. Furthermore, the timepoint the blood sample is taken in relationship to the cancer diagnosis has been shown to affect the interpretation of mtDNA-CN associations. In a prospective study, a direct association between breast cancer risk and increased mtDNA-CN was observed only in the participants who had been diagnosed three years or less before their peripheral blood draw [29]. In the first study mentioned above with the 3225 middle-aged women, they found participants who currently had breast cancer when their blood sample was taken had significantly increased mtDNA-CN compared to the control participants who later were diagnosed with breast cancer [24]. From these studies evaluating mtDNA-CN extracted from peripheral blood, we overall find cancer-type and timing-specific associations. These studies raise the intriguing possibility that mtDNA-CN, known to be impacted by aging, stress, and oxidative damage, in non-tumor cells may be a marker of, or even functionally contribute to, initiation of some types of tumors.

mtDNA can also be extracted from tumor tissue; the following correlative studies have evaluated mtDNA extracted from tumor tissue compared to mtDNA extracted from non-tumor tissue. One group observed higher mtDNA-CN from tumor tissue of breast cancer patients compared to the matched patient non-tumor tissue [30]. In the mining of 22 different tumor types profiled by The Cancer Genome Atlas Project (TCGA), a group of cancers were observed to have significantly decreased mtDNA-CN compared to non-cancerous tissue, including breast, bladder, kidney, esophageal, and head and neck cancers [31]. Conversely, lung adenocarcinoma tissue, compared to non-cancerous control tissue, had an increase in mtDNA-CN [31]. While these studies do not provide direct evidence that altered replication of mtDNA underpins the above associations of mtDNA-CN with cancer outcomes, they raise the intriguing possibility that cancer cells harbor unique mechanisms to maintain mtDNA levels in accordance with their survival needs.

A few studies have provided mechanistic insights into how cancers may modulate mtDNA replication for survival benefit. Depletion of mtDNA using EtBr is a common approach to test the role of mtDNA in cancer. Two murine tumor cell models, B16 (melanoma) and 4T1(breast carcinoma), were treated with EtBr to investigate tumor formation independent of mtDNA. After subcutaneous injection of these mtDNA-depleted B16 cells into either C57BL/6 or NOD/Scid mice, there was a 20–25 day delay in tumor growth compared to the parental B16 cells [32]. Additionally, the mtDNA-depleted 4T1 cells were either implanted subcutaneously or orthotopically in Balb/c mice, and a 20–25 delay in tumor growth was also observed [32]. Furthermore, i.v. injection of the mtDNA-depleted B16 and 4T1 cells ablated lung metastases [32]. mtDNA-CN was also investigated in glioblastoma multiforme progression compared with human neural stem cells (hNScs). mtDNA content was increased during differentiation of the hNScs but not in the progression of glioblastoma multiforme that instead expressed pluripotency markers such as OCT4, NANOG, and SOX2 [33]. Further work from these investigators showed depletion of mtDNA-CN disrupted multiple myeloma tumor growth [34]. These findings from direct mtDNA depletion suggest mtDNA-CN is important for both tumor growth and metastasis.

Alterations of the mitochondrial replication machinery, including POLG alpha subunit (POLGA), TFAM, the D-loop polycystosine tract, and mtSSB, have also been observed in cancer. Extensive nDNA methylation of POLGA exon two was observed in several cancer cell lines (glioblastoma multiforme, multiple myeloma, hepatocarcinoma, ovarian, and breast) that were associated with less POLGA protein and decreased mtDNA-CN [35]. In the glioblastoma multiforme cancer cell line, this hypermethylation prevented the increase of mtDNA-CN and differentiation [35]. In the case of colorectal cancer cell lines, over-expression or knock-down of TFAM increased or decreased mtDNA levels, respectively. TFAM overexpression led to heightened proliferation, colony formation, apoptosis resistance, and increased micro-metastases in vivo with human cell line xenograft models delivered via tail vein injection [36]. A correlative study found that an increase in cytosine residues within the mtDNA D-loop polycytosine tract (C-tract), the site of replication and transcription initiation, was associated with oral cancer severity. Investigators speculated the additional cytosine residues impacted mtDNA replication machinery from binding properly [37]. In osteosarcoma, increased mtDNA-CN was found in osteosarcoma cell lines compared to non-transformed osteoblast cells which was then attributed to the overexpression of mtSSB, for when mtSSB was knocked-down (KD) with shRNA, it reduced mtDNA levels. Upon KD of mtSSB, slowed growth rate of cells in vitro and tumors in vivo with subcutaneous tumor xenografts was observed [38].

Overall, mtDNA is essential for tumor initiation and progression, as shown by the in vitro and in vivo studies demonstrating tumor delay when mtDNA is depleted. Association studies are only able to suggest a functional relationship, as they vary between cancer subtypes, timing of sample collection, and type of sample collected (either blood or tissue), which can all impact the prediction cancer risk and prognosis.

Deregulation of mtRNA production and processing in cancer

Steps of mtDNA transcription include promoter recognition, transcription initiation, elongation, poly-adenylation, and termination, producing a polycistronic RNA transcript [4]. Four core components, POLRMT, TFAM, transcription factor B1 of the mitochondria (TFB1M) and TFB2M, are necessary and sufficient for mammalian mitochondrial transcription [39]. Post-translational modifications of TFAM regulate its function and thus affect mtRNA production. For example, acetylation at lysine sites within the high-mobility-group (HMG) box domain 1 affects interactions with mtDNA that dictate compaction of the mtDNA in nucleoids [40], phosphorylation at serine sites within the HMG1 domain disrupts TFAM localization and binding to promoter sites, leading to TFAM degradation [41], and polyubiquitination hampers TFAM trafficking to the mitochondria [20, 42]. The potential de-regulation of TFAM modifications and functionally how this may contribute to pro-tumor phenotypes is not yet understood but merits investigation. mtRNA is post-transcriptionally modified through cleavage of the polycistronic mtRNA and by RNA chemical modifications at ten identified sites [20, 43, 44]. These mtRNA modifications can impact mitoribosome formation and subsequent mitochondrial translation and function.

Deregulation of mitochondrial transcription in cancer has been reviewed recently [45], with relevant studies limited to correlative gene expression analyses. Mechanistic understanding remains lacking but certainly merits investigation. In non-small cell lung cancer (NSCLC), levels of mitochondrial transcription termination factors (MTERF 1–4) were strongly associated with improved overall survival [46]. Similarly, MTERF3 was found to be highly expressed in brain, liver, pancreatic, lung, and breast tumors relative to normal tissues [46]. In gliomas, MTERF3 immunohistochemical staining positively correlated with tumor grade and with RNA levels of other nDNA-encoded mitochondrial transcription factors [47]. Similar associations of MTERF3 and mitochondrial transcription factors were noted in hormone receptor positive breast cancer [48]. Another study documented increased TEFM levels in glioma compared to normal brain tissue [49]. The retinoblastoma 1 (RB1) transcriptional cofactor binds E2F, preventing E2F binding to DNA thus blocking its transcriptional activation activity; phosphorylation of RB1 releases E2F to enable its transcriptional activity. Analysis of the ChIP-seq data from ENCyclopedia Of DNA Elements (ENCODE) of MCF7 cells revealed E2F1 was recruited to the promoters of mitochondrial protein translation genes within nDNA. Upon RB1 loss, mRNA and protein abundance for mtDNA-encoded genes increased [50], although the mechanistic underpinning of this is not yet clear. While mechanistic insights are extremely limited in present cancer models, these studies raise the intriguing possibility that mtRNA production by the mitochondrial transcription machinery may be up-regulated in some cancers to promote aggressiveness, perhaps by supporting mitochondrial metabolism and/or other functions.

Deregulation of mtDNA translation in cancer

The assembly and function of the mitoribosome requires coordination of nuclear and mitochondrial RNA production, as the two rRNAs and 22 tRNAs are mitochondrial-encoded, and the rest of the mitochondrial ribosomal proteins (MRPs) and accessory factors are encoded in nDNA. Modifications of the rRNA, assembly co-factors, and tRNAs all affect the function of the mitoribosome and resulting translation efficiency, although deregulation of these processes is only beginning to be appreciated in cancer.

As with mitochondrial transcription, the cancer literature largely consists of correlative studies relating mitoribosome component levels to oncogenic phenotypes. A recent review highlighted the elevation of mitoribosome subunits and assembly factors (mL10m, mL38, mL52, uS2m, mS27, and NSUN4) in several tumor types, including breast, liver, lymphoma, melanoma, prostate, and thyroid, relative to normal tissue [51]. Conversely, the same study found other mitoribosome subunits and assembly factors (bl17m, bl33m, mL54, bS6m, and uS10m) were decreased in breast, colon, lung, lymphoma, pancreas, prostate, and renal tumors versus normal tissue [51]. Thus, the actual output and activity of the mitoribosome in those contexts remains unclear. Other studies comparing tumor to normal tissue have found increased MRPs and mitochondrial translation gene expression pathways in non-small cell lung cancer (namely, increased MRPL35) [52], breast cancer (increased mtEF4) [53], high-grade serous ovarian cancer (increased mt-COXII and accessory factors TUFM and DARS2) [54], cervical and lung cancers (increased mtEF4) [55], and hepatocellular carcinoma (increase in several MRPs) [56]. It is important to note that all these studies are correlative and do not actually measure translation or metabolic output. An interesting study in hormone receptor-positive breast cancer patient biopsies showed a direct correlation between mitochondrial translation evaluated by ³⁵S-Met pulse labeling, with both MRPs and mitochondrial translation factor proteins expression and Complex IV activity [57].

Several functional studies have shed light on mechanisms by which mitochondrial translation may be deregulated in cancer cells. The nDNA-encoded transcription factor core binding factor beta (CBFB) regulates transcription of many nDNA genes and was recently found to also localize to the mitochondrial matrix. CBFB frequently harbors loss-of-function mutations in cancers, significantly co-occurring with phosphoinositide 3-kinase alpha (PIK3CA) oncogenic gain-of-function mutations. In breast cancer cell line xenografts and genetically engineered mouse models, a novel role for CBFB in mitochondrial translation was established. CBFB bound to mt-mRNAs and elongation factor TUFM, enhancing binding between the two. CBFB loss of function mutation was found to promote tumor progression. Upon introduction of the loss of function mutation in CBFB, mitochondrial translation and OXPHOS were disrupted, and cells increased dependence on aerobic glycolysis [58]. Another group found survival associated mitochondrial melanoma specific oncogenic non-coding RNA (SAMMSON, encoded in nDNA) expression in more than 90% of skin melanoma tumor samples by mining the TCGA pan-cancer RNA sequencing data [59]. They validated expression through RT-qPCR in uveal melanoma cell lines and PDX-derived cell lines. They observed that SAMMSON directly promoted mitochondrial translation through binding with MRPs in the large ribosomal subunit using RNA immunoprecipitation-qPCR to first identify the binding partners with SAMMSON and then the SUnSET assay for mitochondrial translation. SAMMSON KD decreased mitochondrial translation, OXPHOS, and mitochondrial membrane potential. Additionally, pharmacologic inhibition of SAMMSON, discussed below, delayed tumor growth in melanoma patient-derived xenograft models [59]. Continued and expanded functional studies such as these, in which translation as well as downstream effects on metabolic pathways are measured, will be of crucial importance for moving translation-targeting approaches into the clinic.

Chemical modification of mt-tRNA is an established mechanism regulating mitochondrial translation of methylated RNAs in normal physiologic systems. One study in oral cancer exemplified the possibility that mt-tRNA methylation can be manipulated by tumor cells for survival benefit [60]. Methylation of mt-tRNA^Met by NOP2/Sun RNA methyltransferase 3 (NSUN3) at position 34 converts 5-formylcytosine (f⁵C) to 5-methylcytosine (m⁵C). This methylation mark promoted mitochondrial translation and subsequent OXPHOS in vitro and in vivo models of human oral cancer. This was observed to be critical for cells’ bioenergetics to invade the extracellular matrix and disseminate from the primary tumor [60].

While the pro- or anti-metastatic role of OXPHOS in general, regardless of mechanism, is controversial amongst tumor types, a couple of studies have addressed how modification and expression of mitoribosome core and accessory proteins can impact mitochondrial translation and metastasis. One study documenting the pro-invasive effect of low OXPHOS demonstrated coordination between methylation of MRPS23 at arginine 21 (R21) by arginine methyltransferase 7 (PRMT7) and at lysine 108 (K108) by SET-domain-containing protein 6 (SETD6). This stabilized the MRPS23 protein to persist at a low abundance, protecting the protein from ubiquitin-mediated degradation, thus maintaining low OXPHOS [61]. Another group found increased expression of mitochondrial elongation factor 4 (mtEF4) in breast tumors promoted OXPHOS and cell migration [53]. These studies exemplify the promise of a deep mechanistic understanding of pro-tumor mitochondrial translation for future therapeutic targeting development.

Therapeutically targeting the mitochondrial ‘central dogma,’ considering potency and toxicity

There are a myriad of different cancer therapies targeting mitochondrial functions, many of which have been reviewed elsewhere [62–64]. Here, we review the few current and developing cancer therapies aimed at mitochondrial replication, transcription, and translation (Fig. 1 and Table 1).

Maintenance of homeostatic mtDNA levels is important for tumor cell growth and progression as discussed above. Therefore, targeting mitochondrial replication is a reasonable strategy for suppressing tumor growth. Nucleoside reverse transcriptase inhibitors (NRTIs) are a class of mitochondrial replication targeting drugs that were originally developed to treat human immunodeficiency virus (HIV) but have been shown to inhibit POLG [65, 66]. NRTIs, including 2’,3’-dideoxycytidine (ddC; zalcitabine), are nucleoside mimetics that can be incorporated into newly synthesized mtDNA by POLG, causing premature replication termination [66, 67]. In addition to inhibiting POLG, ddC has been shown to reduce mtDNA content in a variety of different cancers [65, 68–72]. ddC treatment of acute myeloid leukemia (AML) in vitro and in vivo models reduced mtDNA-CN, reduced tumor growth, and impeded tumor initiation [68]. Interestingly, ddC induced mtDNA replication stress which activated ferroptosis through the cyclic GMP–AMP synthase (cGAS)–stimulator of interferon genes (STING) pathway in pancreatic ductal adenocarcinoma (PDAC) models [72]. Furthermore, ddC treatment reduced tumor growth in pre-clinical PDAC models [72]. These findings are positive pre-clinical indications for using NRTIs to treat cancer. To date there are no published clinical trials exploring the effectiveness of NRTIs in treating cancer. Encouragingly, NRTIs have long been used for the treatment of HIV and are generally tolerated with neuropathy, stomatitis, and rash being the most common side effects of ddC treatment [73].

POLRMT and TFAM have been investigated as potential cancer therapeutic targets [45]. Inhibitors of mitochondrial transcription (IMTs) were identified through screening compounds that selectively bind and inhibit POLRMT [74]. The leading compound was LDC195943 (known as IMT1) and was further developed to improve function, leading to compound LDC203974 (known as IMT1B) [74, 75]. Recently, IMT1B was improved upon through structure-activity relationship studies, in which the novel compound D26 was conceived [76]. All these compounds are effective POLRMT inhibitors and have been shown to reduce mitochondrial transcription and subsequent mitochondrial protein levels and mitochondrial function in cancer cells [75–77]. These studies also demonstrated IMT drugs have minimal toxicity in small animals [75–77]. Importantly, IMT1 was tested in 89 cancer cell lines as well as primary cells and drastically reduced cell viability in approximately one third of the cancer cell lines, while primary cells were unaffected by treatment [75]. IMT sensitivity was not correlated with a particular cancer type, but studies investigating therapeutic effectiveness have been carried out in cell models of colon, ovarian, cervical, endometrial, and pancreatic cancers as well as osteosarcoma [75–79]. Given the desire to move this drug into the clinical setting, studies were conducted to understand the mechanisms of resistance to IMT1 as well as to identify potential biomarkers to predict positive outcomes after treatment. Using a CRISPR/Cas9 whole genome screen, it was discovered that loss of genes in the von Hippel-Lindau (VHL) and mammalian target of rapamycin complex 1 (mTORC1) pathway led to IMT1 resistance in colon cancer cells [78]. In models of acquired IMT1 resistance, IMT1 sensitivity could be re-established by targeting other mitochondrial proteins and processes such as TFAM using siRNA knock-down and mitochondrial translation using the antibiotic chloramphenicol [78]. Few studies have investigated TFAM modulation in a cancer therapeutic context. While there are currently no drugs specifically targeting TFAM, treatment of glioma cells with melatonin unexpectedly reduced TFAM levels [80]. Conversely, TFAM can be stabilized by tetramethylpyrazine (TMP) leading to increased mtDNA content in cancer cell lines [81]. POLRMT inhibitors and TFAM modulators show potential as cancer treatments, further work is needed to bring these promising strategies to the clinic.

Of the three mitochondrial ‘central dogma’ processes, mitochondrial translation is by far the most developed as a cancer therapeutic target. This is due to the high degree of homology between the mitoribosome and bacterial ribosome, for which inhibitors have been long utilized as antibiotic agents. Of course, there are numerous FDA-approved antibiotics routinely used in the clinic: >50 target the bacterial ribosome and, thus, the human mitoribosome. Tetracycline analogs, such as doxycycline (DOX) and tigecycline (TIG), are a class of antibiotics that target the 30S ribosomal subunit in bacteria [82]. DOX and COL-3 (also known as CMT-3) are two TET analogs that were originally found to inhibit the expression of matrix metalloproteases (MMPs), leading to a reduction in tumor growth and metastasis [83–85]. The antiproliferative effect of TET analogs was later shown to be due to direct targeting of the 28S small mitoribosome subunit, which reduced mitochondrial protein synthesis and OXPHOS activity [86–88]. TIG is a member of the glycylcycline class of antibiotics which were derived from TET and have similar structures and mechanisms [89]. A phase I clinical trial of TIG treatment in AML patients established a maximum tolerated dose of 300 mg daily, but sustained target inhibition was not optimal [90]. Nevertheless, TIG treatment may be a promising cancer therapy, especially in combination with other drugs. Other antibiotics also target the mitoribosome and have similar effects on mitochondrial function yet have not been further developed to understand their toxicity and efficacy in the clinical setting [88].

Targeting mitochondrial translation has been shown to be more efficacious in cancer types and cells that have high mitochondrial activity and dependency (e.g. leukemias). TIG was discovered as a cancer therapeutic through a drug screen to reduce the viability of leukemia cells and was shown to inhibit mitochondrial translation and OXPHOS [91]. Furthermore, TIG reduced the growth of lymphomas [92], gastric [93], ovarian [94], and breast cancers [95]. TIG treatment also was observed to synergize with chemotherapy [94] and the anthelmintic drug pyrvinium pamoate [95] to kill ovarian and breast cancer cells. In addition to TIG, other antibiotics (DOX, azithromycin, and chloramphenicol) prevented tumor sphere formation in a variety of cell lines, suggesting cancer stem cells (CSCs) are sensitive to mitochondrial translation inhibition [88]. It is important to note that neither mitochondrial translation nor function was assessed after treatment with those antibiotics, so the actual importance of mitochondrial translation for CSC biology remains unclear. Nevertheless, DOX treatment of breast cancer patients prior to surgical resection led to a reduction in CSCs at the time of surgery [96]. Lastly, mitoribosome targeting antibiotics reduced lung metastases in pre-clinical oral squamous cell carcinoma models, demonstrating that targeting mitochondrial translation may also be effective at preventing metastasis as well as reducing tumor burden [60].

In addition to direct inhibition of the mitoribosome, a few studies have discovered alternative ways to impede mitochondrial translation by reducing MRP abundance. JG-98 is an allosteric inhibitor of mitochondrial heat shock protein 70 (HSP70; HspA9) [97]. Inhibition of HSP70 by JG-98 led to reduced levels of MRPs, reduction in OXPHOS, and subsequent cell death in castration resistant prostate cancer [98]. Additionally, as mentioned above, SAMMSON promotes MRP stability and increases OXPHOS in uveal melanoma (UM) [59]. SAMMSON reduction using an antisense oligonucleotide (ASO) in UM PDX models reduced tumor growth and increased apoptosis [59]. The use of either of these drugs has not moved to the clinic. However, there are many ASO treatments that have gained FDA approval in recent years and are a very exciting therapeutic technology [99].

There is the important question of toxicity of targeting mitochondrial function, particularly when promising OXPHOS inhibitors have failed in the clinic. As recently reviewed, this area of drug development still has potential because as the mechanistic understanding of basic mitochondrial biology evolves, an appropriate therapeutic window that will be both efficacious and tolerated is more likely to be found [100]. Indeed, the recently disappointing failure of the highly potent Complex I inhibitor IACS-010759 [101], is likely due to the extremely potent blockade of Complex I activity, resulting in excessive glycolytic lactate production. Promisingly, other oxphos inhibitors, such as metformin and nitric oxide, have been used for many years in the clinic and better balance efficacy with safety, as highlighted in a recent commentary [100].

Conclusion

Mitochondria possess their own genome that relies on several highly regulated processes for mitochondrial RNA production, protein production, and subsequent mitochondrial functions. All branches of the mitochondrial ‘central dogma’ have been implicated in various cancers. Most investigations in this area currently are associative studies making correlative observations, while there are a handful of particularly informative functional studies providing powerful evidence for a functional role in cancer. It is also important to note that the majority of mitochondrial gene expression studies do not address the likely important role of the tumor microenvironment. Our current understanding is mostly about the role of the microenvironment on mtDNA integrity, as recently reviewed [15]. Further, mtDNA methyltransferase METTL4 was found to be upregulated by hypoxic conditions in HepG2 cancer cells [23]. Early-stage studies such as these raise the interesting possibility that the tumor microenvironment may play an important role in regulating mitochondrial gene expression of tumor cells, and this may have therapeutic implications.

Today, there is still a knowledge gap on how each change within these processes affect cancer globally, as it is often observed that mitochondrial functions and vulnerabilities are highly cancer-type and context-specific. We have highlighted an emerging opportunity to therapeutically target the mitochondrial gene expression apparatus in cancer using novel and, sometimes, well characterized FDA-approved agents. This is an area of developing interest and urgency in understanding the fundamental mitochondrial biology driving both cancer progression and treatment responses. Understanding these mechanisms should help us address the pressing question of if and how a therapeutic window for targeting cancer mitochondria exists.

Author contributions

MB: conceptualization and visualization, writing-original draft and editing, writing-revision and editing. SW: visualization, writing-original draft and editing, writing-revision and editing. GE: conceptualization and visualization, funding acquisition, writing-original draft and editing, writing-revision and editing, supervision.

Competing interests

GVE receives sponsored research funding from Chimerix, Inc and receives experimental mitochondria-targeting compounds from the Lead Discovery Center of Germany.