Mitochondrial Determinants of Cancer Health Disparities

Abstract

Mitochondria are involved in the generation of energy, cell growth and differentiation, cellular signaling, cell cycle control, and cell death. To date, the mitochondrial basis of cancer disparities is unknown. The goal of this review is to provide an understanding and a framework of mitochondrial determinants that may contribute to cancer disparities in racially different populations.

Mitochondria, which are multi-functional, have been implicated in the initiation and progression of cancers in relation to metabolic alterations in transformed cells. Due to ethnic-based diversity, the mitochondrial genome (mtDNA) could be a basis for inherited racial disparities and for acquired somatic mutations during tumorigenesis. In African Americans, several germline, population-specific haplotype variants in mtDNA as well as depletion of mtDNA have been linked to cancer predisposition and cancer disparities. Indeed, depletion of mtDNA and mutations in mtDNA or nuclear genome (nDNA)-encoded mitochondrial proteins lead to mitochondrial dysfunction and promote resistance to apoptosis, the epithelial-to-mesenchymal transition, and metastatic disease, which in turn can contribute to cancer disparity and tumor aggressiveness related to racial disparities. Ethnic differences at the level of expression or genetic variations in nDNA encoding the mitochondrial proteome, including mitochondria-localized mtDNA replication and repair proteins, miRNA, transcription factors, kinases and phosphatases, and tumor suppressors and oncogenes may underlie susceptibility to high-risk and aggressive cancers found in African Americans and other ethnicities.

The mitochondrial retrograde signaling that alters the expression profile of nuclear genes in response to dysfunctional mitochondria is a mechanism for tumorigenesis. In ethnic populations, differences in mitochondrial function may alter the cross talk between mitochondria and the nucleus at epigenetic and genetic levels, which can also contribute to cancer health disparities. Targeting mitochondrial determinants and mitochondrial retrograde signaling could provide a promising strategy for the development of selective anticancer therapy for dealing with cancer disparities. Further, agents that restore mitochondrial function to optimal levels should permit sensitivity to anticancer agents for the treatment of aggressive tumors that occur in racially diverse populations and hence help in reducing racial disparities.

1. INTRODUCTION

Mitochondria, often called the powerhouse of the cell, are dynamic organelles found in nearly all eukaryotic cells. Mitochondria are the site of the citric acid cycle (TCA) and oxidative phosphorylation (OXPHOS), metabolic processes that convert pyruvate molecules into ATP, the energy currency of the cell. Apart from producing ATP, mitochondria are also involved in cellular activities such as maintenance of redox potential, apoptosis, fatty acid and heme biosynthesis, and regulation of oxidative stress. The role of mitochondria in cancer was described in 1931 as the discovery of a metabolic rearrangement in cancer cells, a phenomenon known as the Warburg effect. Otto Warburg discovered that cancer cells, due to a defect in mitochondrial oxidative phosphorylation (OXPHOS), produced abnormally high levels of lactate from glucose, even under aerobic conditions. Consistently, in human cancers, the mitochondrial genes that encode the components of the electron transport chain (ETC) of OXPHOS are frequently mutated and/or repressed. Furthermore, mitochondria are also the primary site of generation of reactive oxygen species (ROS). The ROS, upon exceeding a threshold, lead to extensive damage to mitochondria and changes in mitochondrial membrane potential. The resulting mitochondrial dysfunction can cause pro-cancer changes in expression of nuclear genes. This phenomenon is known as mitochondria-to-nucleus retrograde signaling.

Recent clinical advances in cancer prevention, diagnosis, and treatment have helped reduce cancer mortality rates in the Caucasian population. For this disease, however, there is a distinct disparity between African Americans (AAs) and other ethnic populations. Past studies have attributed racial disparities to differences in socioeconomic, educational, cultural and envirnonmental factors. However, it is increasingly recognized that these disparities may also be due to biological differences including differences in mitochondrial biology, genetics and function. Unfortunately, to date a comprehensive mitochondrial basis of cancer disparity is lacking. The present review provides an understanding of mitochondrial determinants involved in tumorigenesis and highlights, where available, the underlying mitochondrial basis for cancer disparities in AAs and other ethnic populations. Restoring mitochondrial function and targeting mitochondrial determinants and mitochondria-to-nucleus retrograde signaling could provide an effective strategy for the development of the selective anticancer therapy that would reduce cancer disparities between diverse populations.

2. DISPARITIES AT THE MITOCHONDRIAL GENOME LEVEL AND THEIR ROLE IN CANCER

The mitochondrial genome is a 16.6 kb, closed-circular, double-helical molecule that is inherited only through the mother. The mtDNA encodes 13 proteins of the OXPHOS system, two rRNAs, and 22 tRNAs [344]. The mtDNA-encoded proteins include 13 subunits that make up four OXPHOS complexes (Complex I, III, IV, and V; see Figure 1b.) Complexes I to IV are electron transport proteins, and Complex V is an ATP synthase [344, 370]. Nuclear DNA encodes the remaining OXPHOS complexes (73 subunits). Nuclear DNA-encoded subunits (38 for Complex I, 10 each for protein Complexes III and IV, and 15 for Complex V) are translated in the cytosol and imported into the mitochondrial compartment [344]. All four subunits constituting complex II are encoded by nuclear DNA (Figure 1).

Figure 1. mtDNA and nDNA encode subunits comprising the OXPHOS complexes.

mtDNA encodes 13 protein subunits involved in OXPHOS complexes; the other OXPHOS subunits are encoded by nDNA.

The remaining approximately two thousand mitochondrial proteins, including those involved in the replication, transcription, and translation of mtDNA, are encoded by nuclear genes and are translocated into the mitochondria [324]. In various aspects, mtDNA differs from nuclear DNA: a) although constituting <1% of the total cellular DNA content, mtDNA is polyploid or present as hundreds to thousands of copies; b) mtDNA lacks introns, and only 10% of the mitochondrial genome consists of non-coding sequences; c) mtDNA also lacks the histone bound nucleo-protein structure; d) transcription of mtDNA produces a polycistronic precursor RNA that is processed further into mRNAs, rRNAs, and tRNAs; and e) mtDNA replication occurs independently of the cell cycle [369]. Owing to proximity to the OXPHOS machinery, mtDNA is constantly exposed to the harmful effects of the ROS generated by the ‘leakage’ of electrons from the respiration complexes. This, combined with the absence of any histone-mediated protection and the natural error-prone replication of mtDNA polymerase γ [344], results in a tenfold higher accumulation of mutations in mtDNA compared to nuclear DNA [121]. Furthermore, due to the high density of coding sequences in mtDNA, mutations in mtDNA have more functional and epidemiological consequences than those in nuclear DNA. mtDNA mutations affect mitochondrial respiratory enzyme complexes and can promote oncogenesis through a) increased production of ROS [98], b) further damage to the mtDNA [95, 341], and c) resistance to apoptosis [190, 342].

2.1 Germline mitochondrial genome variations

Persistence of some mutations in certain populations creates haplogroups; a particular mtDNA haplogroup within a population will carry a unique mutation. A haplogroup can be further divided into haplotypes, generally based on restriction fragment-length polymorphisms [233]. Among human populations, there are more than 25 mtDNA haplogroups, and many of these correlate with a predisposition to cancer.

For a Japanese group, Tanaka et al. [359] retrospectively classified 30 haplotypes based on 149 polymorphisms in the coding region. The individual haplotypes were as follows: F, B5, B4a, B4b, B4c, A, N9a, N9b, Y, M10+M11+M12, M7a, M7b2, M7c, M8+Z+C, G1, G2, M9, D5, D4a, D4b, D4d, D4e, D4g, D4h, D4j, D4k, D4k, D4l, D4m, and D4n. The haplogroup M7b2 was associated with a higher risk of leukemia [372]. Booker et al. classified nine main European haplotypes (H, I, J, K, T, U, V, W, X) in patients with prostate and renal cancers and reported a higher risk of renal cancer for haplogroup U [38]. In a cohort study, they demonstrated an association for four germline mutations in cytochrome oxidase subunit I (T6253C, C6340T, G6261A, and A6663G).

A variety of studies have associated breast cancer risk with certain mtDNA single nucleotide polymorphisms (SNPs) in various populations. In European-American, Polish, Malay, North Indian and AA populations, the G10398A substitution in the N haplogroup affecting the ND3 locus is linked to a higher susceptibility to breast cancer [12, 50, 79, 80, 83, 191, 361]. For AA women, there is a cumulative increase in risk when the T4216C substitution in the ND1 locus is present along with G10398A. Although the A10398G and T16519C substitutions increase breast cancer risk, the T3197C and G13708A SNPs decrease the risk [12] (Figure 2).

Figure 2. mtDNA alterations associated with cancer disparity.

Three non-synonymous and tRNA substitutions are identified in epidemiologic studies as being associated with an increased risk of cancer in specific populations. G10398A in ND3 is associated with increased risk of breast (OR=1.6) and prostate cancer (OR=19.8) in AAs. A12308G in tRNA^Leu2 is a marker of the mtDNA haplogroup U. In Caucasians, this haplogroup is associated with increased risk of prostate (OR=1.95) and renal cancer (OR=2.52).^* In isolation, the T4216C substitution in ND1 confers no increased risk. However, when the T4216C substitution is present with G10398A, the risk of breast cancer is increased (OR=3.1) in AAs. Further, mtDNA depletion is reported to be associated with cancer disparities for AAs.

Other cancers that exhibit a higher prevalence in the presence of germline mtDNA mutations include esophageal and pancreatic cancers [369]. The 10398A polymorphism, for instance, confers a higher risk for esophageal cancer for Indian populations [83]. In Chinese, there is an association between mtDNA haplogroups D4a and D5a and increased risk of esophageal cancer [212]. In European haplotypes, the heritable SNP rs2857285 in the ND4 gene is associated with a more invasive form of ovarian cancer [104]. For patients with pancreatic cancer, germline mutation 16519T in the displacement loop (D-loop) worsened the prognosis [274]. Also, there is an association between a G5460A SNP (in haplogroup H) encoding an A331T substitution in the ND2 gene with pancreatic cancer [196]. Wang et al. [380] and Halfdanarson et al. [132], however, did not find any significant correlation between mtDNA polymorphisms and pancreatic cancer in Caucasian populations. In a multi-ethnic cohort study, the missense 4917 SNP in the ND2 gene was associated with a risk of colorectal cancer in European-Americans but not Africans, Asians, or Latinos. The same study also found that, for the T haplotype, the risk of colorectal cancer was elevated [206].

Although there are many studies correlating mtDNA germline polymorphisms and haplogroups to various cancer types, correlating a specific mutation to a specific cancer type is difficult. Since the mutation and polymorphism rates in mtDNA are high, the same mutation may arise in various populations and lead to different risks in those populations.

2.2 Germline mitochondrial genome depletion

Our previous study found that normal prostate and prostate tumors from AA men contain lower mtDNA copy numbers [185]. In this context, it is noteworthy that lower mtDNA content leads to the acquisition of aggressive, androgen-independent development of prostate cancers and to the epithelial-to-mesenchymal transition involved in metastasis [143, 273]. Our results demonstrate that depletion of mtDNA promotes apoptotic resistance and tumor aggressiveness [190, 281, 282]. In a prostate cancer cell line derived from an AA tumor, mtDNA depletion results in apoptotic resistance [63]. mtDNA depletion promotes features of stem cell-like properties such as over-expression of Oct3/3, Nanog, and CD44 [125, 223]. This feature is reported for prostate [223], ovarian [149], and breast [125] cancers. A low mtDNA content can be caused by mutations in the p53 gene, the POLG gene, and the gene for the mitochondrial transcription factor, TFAM [189, 338–340]. In a yeast screen designed to identify nuclear genes involved in the maintenance of mtDNA, Zhang and Singh identified more than 50 human homologs whose inactivation caused depletion of mtDNA [414]. It is likely that germline variants or somatic mutations in these genes alter mtDNA content and may promote apoptotic resistance and risk of cancers. The reduced mtDNA invokes mitochondria-to-nucleus cross talk, thus providing a molecular basis for risk of cancer and an underlying mechanism contributing to tumor aggressiveness. In addition to mtDNA depletion, mtDNA variants associated with different races also lead to mitochondria-to-nucleus cross-talk [191]. Such constitutive mito-nuclear cross talk may contribute to high risk, tumor aggressiveness, and cancer disparities in AAs and other ethnic populations.

2.3 Somatic mitochondrial genome mutations

Mutations in mtDNA have been reported in many cancers [263]. An exhaustive list of somatic mtDNA mutations and their frequencies has been prepared by Lee [202]. Although these mutations occur throughout the mitochondrial genome, in human cancers, the displacement loop (or D-loop) region is a mutational “hot spot.” The D-loop is a non-coding control region (np 16024-516) that houses cis-regulatory elements required for replication and transcription of mtDNA. Mutations in this region may affect copy numbers and expression of mitochondrial genes, thus making D-loop instabilities a driver of oncogenesis. Maurya et al., who analyzed the D-loop regions in 14 urothelial cell carcinomas, found 28 somatic mutations, which included nine insertion/deletion changes and two single-base substitutions [251]. In oral squamous cell carcinomas, nine mutations, including one point mutation, two base deletions, three insertion mutations, and three heterozygous mutations, were detected in the D-loop region [410]. In another study, somatic mutations in the D-loop were identified in a cohort of Chinese squamous cell carcinomas, but they did not correlate with prognosis or survival [217]. There are also D-Loop mutations in acute lymphoblastic leukemia (ALL) cases, with 89 G insertions, 95 G insertions, 182 C/T substitutions, 308 C insertions, and 311 C insertions, making a total of 132 mutations at 25 locations [397]. In a comprehensive study involving 54 hepatocellular carcinomas, 31 gastric cancers, 31 lung cancers, and 25 colorectal cancers, the incidence of somatic D-loop mutations in cancers of later stages was higher than that of early-stage cancers [203].

In addition to D-loop disruptions, deletions, point mutations, insertions, and duplications in other parts of the mitochondrial genome are known. Somatic mutations in mtDNA genes have been noted in human ovarian, thyroid, salivary, kidney, liver lung, colon, gastric, brain, bladder, head and neck, prostate, and breast cancers, and in leukemia [369, 398]. For example, a 40-bp insertion localized in the COX I gene appears to be specific for renal cell oncocytoma [386], and a deletion mutation is resulting in the loss of mtDNA within NADH dehydrogenase subunit III is associated with renal carcinoma [147, 186]. In cancers of AA women, two variants of the cytochrome b gene, 13G and I2-992T, are more frequent in breast cancer patients compared to healthy controls. As determined in a multi-ethnic study, there is also a positive correlation between the SNP T4216C in the ND1 gene and colorectal cancer [4]. In a population-based study involving prostate cancer patients of European and AA descent, the frequency of COX I missense mutations was higher in cancer patients compared to healthy controls, with some of these sequence variants possibly representing germ line mutations [300]. Specifically, the authors associated two SNPs, T6221C and T7389C, with a higher incidence of prostate cancer. In renal cell carcinomas, the co-occurrence of somatic and germ-line mtDNA mutations has been reported [321]. In contrast, somatic variants of the ATP6 and ND3 genes in the Mexican Mestizo populations could not be linked to a higher risk of prostate cancer [54].

Brandon et al. [42] suggested that mtDNA mutations in tumors can be divided into two main groups: (1) severe mutations that inhibit OXPHOS, increase ROS production, and promote tumor cell proliferation and (2) milder mutations that permit tumors to adapt to new environments. The severely deleterious tumorigenic mutations that inhibit mitochondrial respiration could be advantageous in the initial phases of tumor growth when the tumor requires mitochondrial ROS to drive cell proliferation. However, the adaptive mtDNA mutations may permit the tumor cells to flourish in new environments as they metastasize. Since the migratory tumor cells could be exposed to similar environmental challenges as the humans who migrated out of Africa, the same mtDNA mutations might be adaptive in both tumors and people [42].

2.4 Somatic mtDNA copy number variations

Both an increase and a decrease in mtDNA copy number are associated with an increased risk for tumorigenesis [68, 220, 244, 244,164, 235, 377, 381, 382]. Depletion in mtDNA has been reported in breast [235, 367], colorectal [202], gastric [393, 202] hepatocellular [202], kidney [423, 424], lung [202], ovarian [381], and prostate [185] cancers. An increase in mtDNA content has been reported in Burkitt’s lymphoma, chronic lymphocytic leukemia, head & neck cancer, non-Hodgkin’s lymphoma, small lymphocytic lymphoma [425], and papillary thyroid carcinoma [235]. Depletion of mtDNA may result in disruption of mitochondrial respiration, and the ensuing dysfunction in mito-nuclear signaling can drive oncogenesis. An increase in mtDNA copy number might occur as a compensatory response to mitochondrial dysfunction, resulting in increased ROS production, altered mito-nuclear cross talk, and tumorigenesis.

Mutations in or depletion of mtDNA can have a causative effect in carcinogenesis through disruption of the OXPHOS enzyme complexes and the ensuing oxidative stress and retrograde signaling. Conversely, mtDNA dysfunction may not directly generate the cancer phenotype but influence tumor progression and maintenance by causing a metabolic shift from respiration to aerobic glycolysis [369]. To a great extent, the physiological role of mtDNA in initiating and maintaining tumorigenesis has been elucidated through the use of trans-mitochondrial cybrids –hybrid cells that combine nuclear genes from one cell and mitochondrial genes from another. This technique is useful in dissociating the function of mtDNA genes from those of the nuclear genes. Prostate cell (PC3) cybrids harboring the T8993G mtDNA mutation generate tumors that are seven times larger than wild-type cybrids, which barely grew in mice [38]. Additionally, after transplantation into nude mice, cybrids constructed using a common HeLa nucleus and mitochondria containing a point mutation in ATP synthase subunit 6 conferred a growth advantage in early tumor stages, possibly through the prevention of apoptosis [137].

2.5 Contribution of numtogenesis

Intact mitochondria containing mtDNA, mitochondrial RNA (mtRNA), and mitochondrial proteins localize into the nucleus [13, 33, 41]. Indeed, the nuclear copies of mtDNA described as NUMTs (nuclear mtDNA) are present in at least 85 sequenced eukaryotic genomes [139]. We have named this phenomenon leading to the presence of nuclear mitochondria as numtogenesis [422]. In humans, numtogenesis is estimated to occur at a rate of ~5 × 10⁻⁶ per germ cell per generation [89].

Evolutionary studies suggest that, in humans and other mammals, the insertion sites of germline NUMTs are distributed non-randomly [379]. Germline NUMTs tend not to originate from the mtDNA displacement loop (D-loop); they tend to be located in damage-prone regions of the nuclear genome, such as open chromatin and fragile sites [379]. These results implicate NUMTs in the repair of double-strand breaks [138]. The mechanism for NUMT accumulation is not well understood. The most parsimonious mechanism explaining NUMT accumulation involves de novo transposition from the mitochondria to the nucleus; however, NUMTs also accumulate via segmental duplication (sometimes within repetitive elements), and possibly by RNA retro-transposition [64, 130, 270, 271]. The human genome contains between 755 and 1,105 germline NUMTs, with mtDNA identities ranging from 64–100% [270]. Our preliminary data suggest the existence of NUMT diversity between AAs and Caucasian Americans and among other populations.

3. DISPARITIES AT THE NUCLEAR GENOME LEVEL AND THEIR ROLE IN CANCER

3.1 Mitochondrial oxidative phosphorylation

Except for 13 proteins, all other mitochondrial proteins, as well as the regulatory apparatus of the mitochondria, are encoded by nuclear genes (Figure 1). In human cancers, there are mutations in genes encoding OXPHOS proteins, TCA cycle enzymes, mtDNA regulatory elements, and mitochondrial biogenesis.

Succinate dehydrogenase (SDH), or complex II of the ETC, serves as a link between the TCA cycle and the ETC, and the gene encoding it is frequently mutated in paraganglioma, breast cancer, gastric cancer, and renal carcinoma. The SDH complex consists of four subunits, A, B, C, and D, encoded respectively by the SDHA, SDHB, SDHC, and SDHD genes. Patients with hereditary paraganglioma carry germline mutations in the SDHD gene [22], whereas mutations in SDHC cause an autosomal dominant form of paraganglioma [275]. The main mechanism of SDH mutation-derived tumorigenesis is disruption of the ETC, leading to increases in the levels of cell-damaging ROS [153]. Also, murine fibroblasts deficient in SDHA cause succinate accumulation and translocation of HIF-1α into the nucleus. Mutations in SDHA are seen in pituitary adenoma [104], and SDHB mutations are associated with an early onset of familial renal carcinoma [305]. Patients suffering from gastrointestinal stromal tumors [159] and breast cancers [176] also show SDH inactivation in the tumor tissues. Germline mutations in the gene encoding fumarate dehydrogenase (FDH), the TCA cycle enzyme that reversibly hydrates fumarate to malate, is seen in skin leiomyomata, renal cell cancers, and uterine leiomyomas [204, 205]. Bennedbaek et al. identified, in Danish patients with pheochromocytoma or paraganglioma, eight germline variants in the SDHB, SDHC, and SDHD genes [24].

Heterozygous missense mutations in the isocitrate dehydrogenase (IDH) gene are frequently seen in acute myelocytic leukemia (AML), gliomas, and astrocytomas [401]. Both the cytosolic (IDH1) and mitochondrial (IDH2) isoforms display mutated versions and are implicated in tumorigenesis. In AML, IDH2-R140 is the most common mutation [241, 284], whereas the IDH2-R172 mutation is seen in gliomas [400]. Under physiological conditions, IDH de-carboxylates isocitrate to α-ketoglutarate and reduces NADP+ to NADPH. Mutated forms of IDH convert isocitrate to 2-hydroxyglutarate, an oncometabolite that alters global gene expression patterns through DNA methylation and chromatin remodeling [71, 229]. In some recent studies, the relevance of germline variants of IDH genes has been investigated. In a cohort of AML patients, the G-allele of the IDH1105 SNP was associated with shorter survival and a poorer prognosis compared to the T-allele [107]. In another study, the SNPs rs12478635 in the IDH1 gene and rs11632348 in the IDH2 gene exhibited associations with death risk for patients with hepatocellular carcinoma [413].

3.2 Mitochondrial antioxidant system

The reverse electron transport from complex II to complex I in the ETC leads to the formation of reactive oxygen species (ROS). Although low levels of ROS serve as signal transducers and promote normal cellular functions, excessive levels induce mtDNA damage and destabilize the mitochondrial membrane potential, ultimately initiating cancer [347]. Mitochondria, therefore, house antioxidant enzymes that quench these free radicals and prevent cellular damage. Deregulation/mutations in manganese superoxide dismutase (MnSOD/SOD2), glutathione peroxidase (Gpx), and thioredoxin-2 (Trx2) are seen in several human cancers [69]. SOD2 catalyzes the conversion of the superoxide anion (O₂₋) to hydrogen peroxide (H₂O₂). The role of SOD2 in cancer initiation is concentration-dependent and akin to that of a “two-edged sword.” Under physiological conditions, SOD2 scavenges ROS and acts primarily as a tumor suppressor. This activity of SOD2 is aided by Sirt2 deacetylase, and loss of the latter promotes oncogenesis. Under high ROS conditions, however, the scavenging properties of SOD2 prevent ROS-induced apoptosis and thereby allow malignant cells to survive [151]. Likewise, in breast cancers and skin carcinomas, SOD2 is associated with a protective effect [209], but it has a pro-tumorigenic function in ovarian clear cell carcinoma [141].

Variants of the SOD2 gene are linked to increased susceptibility to various cancers. Five SNPs are currently identified: rs7855, rs5746151, rs5746136, rs2758331, and rs4880. The rs4880 SNP in the mitochondrial targeting sequence involves a T47C substitution that results in an Ala to Val change in the mitochondrial targeting sequence. The Ala allele is present at a higher frequency in colorectal cancers of Hispanic patients compared to non-Hispanic white patients, even though this SNP is not particularly associated with higher rates of colorectal cancer [348]. In a recent study of pediatric medulloblastoma patients, the T47C SNP was linked to a higher incidence of the disease [43]. In a Chinese population, T47C is associated with a poorer prognosis for gastric cancer [395] and for squamous cell carcinoma [224]. In a Korean cohort, the allelic frequency of the T5482C SNP was higher in patients compared to healthy controls [134]. In the Hispanic population, this substitution is linked to higher incidence of colorectal cancer [348].

Gpx is involved in the removal of peroxides from the cytosol and mitochondria. The mitochondrial isoforms, Gpx1 and Gpx4, act as tumor suppressors by reducing ROS [165, 210]. The Gpx1 Pro198Leu SNP is associated with a higher risk of bladder cancer [55]. Gpx4 expression is downregulated in pancreatic and breast cancer cells [59, 405], and there is a Gpx4 deletion in large B-cell lymphomas and renal cancers [403].

The mitochondrial thioredoxin (Trx) superfamily of thiol-disulfide oxidoreductases comprises Trx2, Trx2 reductase (TrxR2), and Trx-dependent peroxidase (Prx3) [303]. Try2 transfers reducing equivalents from cysteine residues to disulphide bonds and maintains proteins in a reduced state. Trx2 modulates the activities of transcription factors and apoptosis signaling factors. On the other hand, Prx3 reduces H2O2 generated during mitochondrial respiration [145]. The Trx superfamily is overexpressed in human cancers; Trx2 and Prx3 are elevated in multiple myeloma and thymoma, respectively, and protect cancer cells from apoptosis.

3.3 Mitochondrial DNA replication

The replication of mtDNA is regulated by several nuclear-encoded proteins, including DNA polymerase γ (POLG1 and POLG2) [88], TFAM [298], and mitochondrial RNA polymerase (POLMRT) [40]. POLG and TFAM are essential for maintaining mtDNA integrity and copy number. Mutations in either can lead to depletion of mtDNA, which in turn initiates mitochondrial dysfunction. In POLG1, there are numerous somatic and germ-line mutations that are associated with various cancers. Singh et al. detected POLG1 mutations in 63% of breast tumors [340]. In malignant breast tissue, there were 17 mutations spanning over the exonuclease domain and linker region of POLG associated with a complete loss of mtDNA. Mutations in the POLG exonuclease domain correlated with a higher frequency of rare point mutations in the mtDNA control region [93]. A recent study revealed that POLG1 expression is upregulated in myelomas, melanomas, and prostate cancers and is downregulated in lung, head and neck, brain, bladder, and esophageal cancer and in leukemia [338]. In addition, the expression levels of POLG1 correlates with the copy numbers of POLG1. In primary tumors, there is a higher frequency of POLG1 mutations, mostly missense mutations and substitutions. At the 1143 amino acid position in the polymerase domain, there is a missense variation that changes glutamic acid to glycine. The European-American population shows a six-fold higher allele frequency of E1143G compared to the AA population. It remains to be determined if E1143G, along with the POLG1 variants, T251I and P587L, are involved in the predisposition to cancer. As determined with a German cohort of patients, an SNP in the promoter region of POLG (rs2856268, A>G) is associated with higher risk of breast cancer [290]. In the AA population, variants in the CAG repeat sequence length of the POLG gene are associated with increased risk of breast cancer [10] and testicular cancer [32]. As determined for an Indian population, the SNPs, rs41553913 at POLRMT and rs9905016 at POLG2, increase the risk of oral leukoplakia and cancer [85]. The authors speculate that, in cancer tissues, these polymorphisms are associated with increased mtDNA replication, although the mechanism remains elusive.

In colorectal cancers [126] and breast cancers [15], truncating mutations in TFAM are associated with mtDNA depletion and oxidative stress. In murine models, complete knockout of TFAM leads to mtDNA loss, OXPHOS breakdown, and embryonic lethality [373].

3.4 Mitochondrial DNA repair

There are primarily six kinds of lesions that affect mtDNA: 1) Alkylation of bases that is usually induced by chemotherapeutic agents [317]. Alkylation damage can also be caused by the endogenous pool of S-adenosylmethionine that exists in the mitochondria [146]. 2) Hydrolytic deamination of bases and formation of abasic sites [90, 219]. 3) Adduction of DNA bases with carcinogens such as acetaldehyde, cisplatin, and DMBA, resulting in crosslinking of bases [195]. 4) Mismatched bases that occur due to replication errors [170] or incorporation of damaged bases during replication [35]. 5) DNA single- and double-strand breaks that are either induced by carcinogens or occur as a result of inefficient repair of other lesions. 6) Oxidative damage to bases and the sugar-phosphate backbone from ROS.

Due to its proximity to the ETC chain and sites of ROS generation, mtDNA is particularly susceptible to oxidative damage. In human fibroblasts, an H₂O₂-producing system of glucose and glucose oxidase leads to a greater accumulation of strand breaks, oxidized bases, and abasic sites in mtDNA compared to nDNA [31]. In fact, strand breaks induced by H₂O₂ occur at a tenfold higher frequency in mtDNA, with the sugar-phosphate backbone as the primary target [337]. Oxidative damage to DNA bases results in two main products: thymine glycol, the main modified pyrimidine [383], and 7,8-dihydro-8oxo-20-deoxyguanosine (8-oxodG), the main modified purine [90]. Although both oxidized bases block DNA polymerase, 8-oxodG is more mutagenic and is responsible for the characteristic G→T substitutions [135]. Other oxidative lesions include 8-hydroxyguanine, FAPy-adenine, 8-hydroxyadenine, 5,6-dihydroxyuracil, 5-hdroxyuracil, 5-hydroxycytosine, and 5-hydroxymethyluracil [5,8].

Previously, the consensus was that the DNA repair pathways in mitochondria were less developed than in the nucleus since it was thought that mitochondria were unable to repair UV-induced dimerization [74] and alkylation damage [259]. Recently, improvements in sub-cellular localization techniques using fluorescent tagging of proteins and immunogold labeling have provided evidence that the major DNA repair pathways also exist in mitochondria [35].

During DNA replication, base mismatches are often introduced by ‘slippage’ errors of DNA polymerase or randomly by alkylation, deamination, or oxidation of the bases. The DNA mismatch repair (MMR) pathway operates in the nucleus to correct these errors. Briefly, the MSH proteins (MutS homolog family) recognize and bind to the mismatched base sites [194], followed by recruitment of the MLH (MutL homolog) proteins [264], downstream excision, and re-synthesis of the DNA strand [252]. MMR was first identified in rat mitochondrial lysates wherein the G→T mismatch was corrected in a bi-directional, ATP-dependent manner [245]. In another study, the mechanism for mitochondrial MMR was somewhat elucidated with the discovery of the MLH1 protein in mammalian mitochondria [332]. However, mitochondrial MMR appears to be largely independent of the MLH proteins, as the Y-box binding protein (YB-1) is the major MMR initiator in mitochondria of HeLa cells [100]. Loss of YB-1 in the HeLa cells lead to higher rates of mtDNA mutagenesis.

The best-understood pathway for DNA repair in mitochondria is base excision repair (BER). Since BER is the preferred pathway for the repair of oxidative DNA base lesions [352], mitochondria use it because of the susceptibility mtDNA to ROS. Indeed, the most common oxidative base lesion, 8-oxodG, is more efficiently repaired in the mitochondria than in the nucleus. The first indication of mitochondrial BER was the discovery of mitochondrial uracil-DNA glycosylase (UDG) by Anderson and Friedberg in 1980 [7]. Later studies provided direct evidence of BER in the mitochondria [103, 201].

Mitochondrial BER starts with a) lesion recognition and strand scission, followed by b) gap tailoring, and converging at the c) DNA ligation step [352]. The initial step of recognizing the DNA lesion is mediated by DNA glycosylases. Mono-functional DNA glycosylases like UDG1 recognize the damaged base and hydrolyze the N-glycosidic bond, leaving the DNA in an abasic state. Following removal of the damaged base, the resulting apurinic/apyrimidinic (AP) site is incised by AP endonuclease. The mitochondrial isoform of APE1 is generated by truncation of the nuclear localization signal [62]. In contrast, bi-functional DNA glycolysis involves intrinsic AP-lyase activity. In mammalian mitochondria, there are at least four bi-functional DNA glycosylases: 8-oxoguanine DNA glycosylase (OGG1) [99], nth-like 1 (NTHL1) [172], Nei-like 1 (NEIL1), and Nei-like 2 (NEIL2) [237].

DNA glycosylases produce single-strand breaks that have a blocking 5′ or 3′ end that must be removed before the correct base can be inserted, and DNA ligation can occur. The 5′-deoxyribose phosphate (dRP) generated by the action of UNG1 and APE1 is removed by the dRP lyase activity of DNA polymerase γ, leaving a single strand gap with 5′-P and 3′-OH ends [226]. Similarly, the 3′-phospho-α,β-unsaturated aldehyde left by OGG1 is removed by the phosphodiesterase activity of APE1 [352], leaving a single-strand gap with 5′-P and 3′-OH ends. The final step is DNA synthesis and ligation by DNA polymerase γ and DNA ligase IIIα, respectively [352].

Variations in OGG1 correlate with various cancers. The C1245G SNP has a Ser326Cys change in the catalytic subunit of OGG1, which has an impaired capacity to excise 8-oxoguanine [160]. In a Brazilian group, this variant of OGG1 was present in 42% of the non-small cell lung carcinoma (NSCLC) patients and in 34% of healthy controls [78]. In a meta-analysis of Asian patients, Ser326Cys was associated higher risk of lung cancer [207]. A higher susceptibility to head and neck cancer was linked to the C1245G SNP [345]. Breast cancer risk is increased in individuals with the OGGI C >G variant [178, 268]. In certain North Indian ethnic groups, the SNP C1245G at exon seven was detected in about a quarter of the population, although its relevance in cancer risk remains to be evaluated [236].

3.5 Transcription factors in the mitochondria

The regulation of mtDNA genes depends on various nuclear-encoded transcription factors. The mitochondria-localized transcription factors, called mitoTFs, mediate respiration, apoptosis, and mtDNA transcription. The mitoTFs are broadly classified into two groups: a) the hormone and steroid receptors, such as those for estrogen, progesterone, androgen, and glucocorticoids and b) downstream factors activated upon binding of hormones and cytokines to their respective receptors (e.g., p53 HER1, and HER2).

3.5.1 Estrogen receptor in the mitochondria

First described by Jenson [162], the estrogen receptor (ER) is present as two isoforms, ERα and ERβ, each encoded by separate genes, ESR1 and ESR2 [187]. ERα and ERβ bind to estrogens, resulting in receptor dimerization and recruitment to the promoter region of target genes at their estrogen response elements (EREs) [278]. ERβ is predominant in mitochondria, whereas ERα is concentrated in the cytosol [49, 66]. Chen et al. [66] identified, in ERβ, a mitochondrial-targeting peptide signal (mTP) that is essential for proteins to be translocated into the mitochondria. Furthermore, the presence of EREs in the mtDNA points to the mitochondrial localization of ERs. The binding of ERβ to mtDNA was confirmed in MCF-7 breast cancer cells [65]. Apart from directly regulating the genes of mtDNA, ER also influences mitochondrial functions indirectly through the transcriptional regulation of nuclear genes encoding mitochondrial proteins. For instance, in MCF-7 cells, ERα activates respiratory factor-1 (NRF-1), which in turn activates TFAM and promotes mitochondrial biogenesis [250].

In breast cancer cells, binding of the estrogen, 17β estradiol (E2), to ERα and ERβ results in an increase in proliferation, survival, and invasion [183]. Also, in MCF-7 cells, exposure to E2 increases the binding of ERα and ERβ to mtDNA and increases the expression of genes encoding cytochrome c oxidase subunits I and II [65]. In lung cancer cells, there is a pro-survival, anti-apoptotic role for ERβ [415]. Tamoxifen, an anti-cancer drug that induces oxidative stress and apoptosis via a mitochondria-dependent pathway also acts through binding with the ERs as both agonist and antagonist [301]. SNPs in the estrogen receptor gene are involved in breast cancer susceptibility. In a meta-analysis, three SNPs (rs2077647:T>C, rs2228480:G>A, and rs3798577:T>C) correlated with breast cancer risk in various ethnic populations [211]. Amongst Caucasians, the rs2228480 AA genotype was associated with a lower risk compared to the GG genotype, whereas the rs3798577TT genotype correlated with increased risk in an Asian population. In all the ethnic groups, rs2077647:T>C was associated with a higher risk of breast cancer, albeit non-significantly. In a recent study, the 13950T/C SNP of ERβ was associated with higher risk of uterine leiomyomas [374].

3.5.2 Progesterone receptor in the mitochondria

The response to progesterone is mediated by two nuclear isoforms of the progesterone receptor (PR), PRA and PRB [16]. The nuclear PRs (nPRs) contain an N-terminal transcriptional regulatory domain, a DNA-binding domain, a D-domain or hinge region, and a hormone binding domain [157]. Although, under normal conditions, the two isoforms are co-expressed in progesterone-targeted tissues (breast, endometrium, cervix, and ovaries), their ratios vary widely during malignant transformation. For instance, PRB is overexpressed in highly invasive forms of cervical, endometrial, and ovarian cancers [112]. In contrast, in breast cancers, an excess of PRA correlates with a poor prognosis, and a greater proportion of PRB predicts better survival [16].

MCF-10A breast epithelial cells lacking nPRs proliferate rapidly in the presence of progestin [23]. This phenomenon is attributed to stimulation of mitochondrial respiration and a concomitant inhibition of apoptosis by progestin. The authors hypothesize two scenarios: an indirect paracrine action of progesterone on mitochondrial function and/or the direct action of progesterone through ligand binding with a mitochondrial PR. In fact, a truncated PR (PR-M) has been cloned from human adipose and aortic cDNA libraries [320]; the encoded protein lacks the DNA-binding and regulatory domains but retains the hormone-binding domain. Subsequently, Dai et al. showed that the PR-M localizes to the mitochondria of T47D breast epithelial cells and HeLa cells [81]. The functionality of PR-M was established by an increase in mitochondrial membranes and increased cellular respiration upon exposure to progestin. Progesterone action via PR-M may be a mechanism for adapting to the increased energy demands of invasive cancer cells. This hypothesis was strengthened by results of a study that correlated increased PR-M expression in leiomyomas with increased mitochondrial membrane potential [108].

Three polymorphisms of the PR (Alu insertion, 331G/Aand, and Val660Leu) have been investigated in the context of cancer susceptibility. In a Chinese cohort, the Alu insertion and Val660Leu polymorphisms were associated with a risk of ovarian cancer, whereas the 331G/A SNP could not be correlated with any cancer risk [213]. In a Caucasian cohort of endometrial cancer patients, the 331G/A SNP could not be correlated with a cancer risk [279], but, in another study, the Alu insertion variant was associated with a higher breast cancer incidence in Mexican women [113].

3.5.3 Androgen receptor in the mitochondria

The androgen receptor (AR), which is activated by ligand-specific binding of the androgenic receptors, testosterone, and dihydrotestosterone [313, 387], is involved in the transcriptional regulation of genes that maintain the male sexual phenotype. Binding of androgens to AR also induces proliferation of prostate cells and initiates prostate cancer; in most prostate cancers, AR is constitutively activated and overexpressed [402]. Solakidis et al. reported the mitochondrial localization of AR in the midsection of spermatozoa (346), and Chaudhary et al. [70] linked the regulation of mitochondrial fission to AR. Mitochondrial fission is initiated upon translocation of the cytoplasmic dynamin-related protein 1 (Drp1) to the mitochondrial outer membrane [408]; impairment of Drp1 results in fragmented mitochondria, which lead to apoptosis [60]. Co-ordination between mitochondrial fission and fusion is essential for optimal cellular proliferation and turnover [19]. In LNCap cells, Drp1 is regulated in an androgen-dependent manner at both the transcriptional and translational levels. The transcript levels of AR and Drp1 are positively correlated, and androgen-mediated Ser-616 phosphorylation of Drp1 reduces mitochondrial fission and promotes cell proliferation [70]. This may be a mechanism utilized by invasive prostate cancer cells to escape apoptosis. It remains to be determined if localization of AR in the mitochondria can be related to mitochondria-mediated survival of cancer cells.

Inherited variations in the AR gene are associated with higher incidences of several types of cancers. The non-coding CAG repeat sequence has been investigated in the context of cancer risk. Generally, repeat lengths <22 are classified as short, whereas those >22 are classified as long. For Mexican men, a lower number of CAG repeats correlates to earlier and more aggressive forms of prostate cancer [118]. For Asian women, there is a link between shorter CAG repeats and a higher risk of developing epithelial ovarian cancer [254]. In contrast, longer CAG repeats are present at a higher frequency in breast cancer patients of the Han Chinese population [82]. Although the CAG repeat length does not affect prostate cancer risk amongst European men, carriers of the SNP rs1204038A allele are more likely to develop this cancer [21]. Koocheckpour et al. performed a comprehensive analysis of the prevalence of germline and somatic AR mutations in AA men suffering from prostate cancer. Compared to Caucasian males, somatic missense AR mutations were more frequent in the AAs. Furthermore, the A-allele of the E213 G/A SNP was more frequent amongst the AA men, although this polymorphism could not be correlated to a higher cancer risk.

3.5.4 Glucocorticoid receptor in the mitochondria

The glucocorticoid receptor (GR), which binds cortisol and other glucocorticoids and is expressed in almost all eukaryotic cells, regulates expression of genes related to inflammation, immunity, metabolism, and development [304]. The mitochondrial localization of GR regulates the expression of mtRNA encoding for the OXPHOS elements [325]. Analysis of human and rodent mtDNA revealed the presence of glucocorticoid-responsive elements (GREs) within the D-loop and within the coding regions of genes such as those for cytochrome oxidase I, ND I, and 12S RNA [152]. As determined by ChiP analysis of mitochondria isolated from hepatocarcinoma cells, the binding of GR was shown in the D-loop and in the ND1 and 12S rRNA sites [294], thus proving a role of GR in transcription of mitochondrial genes.

In the context of tumor physiology, GR induces apoptosis at the mitochondrial level by indirectly upregulating (by repressing miRNA 17-92) the expression of the pro-apoptotic Bcl-2 family member, BIM [266]. BIM, along with other pro-apoptotic proteins such as BAD and BID, causes the release of cytochrome c into the cytosol and triggers caspase-mediated apoptosis [243]. For this reason, glucocorticoids are included in the chemotherapeutic regimens for non-Hodgkin’s leukemia, multiple myeloma, chronic lymphoblastic leukemia (CLL), and ALL [11, 326]. Heideri et al. investigated the GR-mediated apoptotic pathways in the mitochondria, placing them downstream of the p38-MAPK and RUNX1/c-Jun signaling pathways and identifying therapeutic targets in glucocorticoid-resistant leukemias [140]. For Han Chinese, SNPs in the non-coding region of GR were linked to a higher incidence of pediatric ALL [396]. Five SNPs were analyzed, of which the rs41423247 and rs7701443 polymorphisms were significantly associated with a poorer prognosis.

3.5.5 Tumor suppressors in the mitochondria

The tumor suppressor p53 is involved in various cellular functions, including proliferation, senescence, apoptosis, autophagy, metabolism, and differentiation. This is accomplished through transcriptional regulation of an array of target genes and integration of various signaling pathways. In human cancers, this master regulator is the most frequently mutated gene [178]. The regulatory properties of p53 in the cytoplasm have been studied extensively; there is rapid translocation of p53 to the mitochondria (outer membrane) in response to hypoxia, oxidative damage, and DNA damage [9, 238]. Moreover, in p53−/− cancer cells, a p53 fusion protein with a specific mitochondrial targeting sequence induces apoptosis and cell cycle arrest, bypassing the nuclear circuit [238, 257]. There is the translocation of p53 to the mitochondrial matrix [189] wherein p53 forms complexes with the mitochondrial chaperone proteins, Hsp60 and Hsp70 [238]. Of note, p53 is present in mitochondria even in the absence of cellular stress [91], suggesting that endogenous p53 is involved in normal mitochondrial functions. Despite evidence of mitochondrial localization of p53, the mechanism of its entry, in the absence of a mitochondrial targeting sequence [239], is a matter of debate. One hypothesis is that stress-induced binding of Mdm2 to p53 leads to the ubiquitination of p53 and provides a targeting signal to the mitochondria; upon entry, the p53 is deubiquitinated by a mitochondrial ubiquitin-specific protease [239].

There is considerable evidence for various functions of p53 in mitochondria, namely mtDNA maintenance and repair, apoptosis, and respiration. For HCT116 cells, there is an association between the mitochondrial DNA polymerase γ (POLG) and p53, with p53 also binding to mtDNA and enhancing the function of POLG [2]. Association of p53 with POLG also improves the latter’s proofreading capacity [14] during mtDNA replication. Similarly, the gap-filling activity of POLG is impaired in mitochondria derived from the livers of p53−/− mice and is restored by introducing a recombinant p53 [2]. Binding of the mitochondrial transcription factor TFAM to oxidized DNA bases is enhanced upon physical interaction with p53 [409]. The transcriptional regulation of mtDNA genes by p53 has, however, not been confirmed, despite evidence of p53 binding sites in the human mitochondrial genome [142]. The rapid, stress-induced translocation of p53 to the outer mitochondrial membrane is a prelude to the action of p53 in mitochondrial outer membrane permeabilization (MOMP), which releases cytochrome c into the cytosol and triggers the apoptotic machinery [323]. Also, the binding of p53 to pro-apoptotic elements such as Bax [417] and pro-caspase-3 [111] has been reported.

In malignant cells, the tumor suppressive actions of p53 are often directed towards subverting the glycolytic pathway and counteracting the Warburg effect. In leukemia cells, p53 induces the expression of TIGAR (Tp53 induced glycolysis and apoptosis regulator), inhibits glycolysis, and increases the production of ROS, ultimately clearing damaged cells via apoptosis [25]. p53 also prevents the uptake of glucose, the substrate for glycolysis, by blocking the expression of the glucose transporters, GLUT1 and GLUT4 [328]. In cancer cells, p53 hinders glycolysis through the use of alternate substrates. Vousden et al. identified mitochondrial phosphate-activated glutaminase (GLS2) as a transcriptional target of p53 [378]. Activation of GLS2 shifts the cellular metabolism from glycolysis to aerobic respiration and glutaminolysis. Consistent with these findings, a dominant-negative mutant of p53 up-regulates the expression of hexokinase II, which, in AS-30D hepatoma cells, docks at the mitochondrial outer membrane and increases glycolytic activity [246, 247]. Similarly, deletion of p53 in human colon cancer cells leads to decreased expression of cytochrome c oxidase and a disruption in mitochondrial activity and structure [420]. For both murine and human cancer cell lines, Matoba et al. reported the synthesis of cytochrome c oxidase (SCO2), an effector necessary for respiratory chain function, as the downstream target of p53 in the regulation of mitochondrial respiration [248]. Disruption of the SCO2 gene in p53+/+ cells also directed the cellular metabolism towards glycolysis in the same way as a loss of p53. Conversely, disruption of the mitochondrial ETC increases the accumulation of p53 and leads to activation of apoptotic pathways. Both loss of complex I [76] and impairment of complex III function [175] lead to the accumulation of p53.

Although the somatic loss of p53 usually accompanies the development of most human cancers, there are germline mutations/SNPs in the familial Li-Fraumeni syndrome, encompassing cancers such as premenopausal breast cancer, bone and soft-tissue sarcomas, adrenal cortical carcinomas, and brain tumors [255]. The hotspot of polymorphisms in p53 resides in codon 172, which encodes its protein binding and interaction domain. The SNP rs2602141, resulting in an arginine-to-leucine substitution, is associated with leukemia, prostate cancer, esophageal cancer, and lung cancer in Asians and Caucasians but not in other ethnic groups [174].

3.5.6 FOXO3 transcription factor in the mitochondria

The forkhead box (FOX) family of proteins consists of 19 sub-families of transcription factors that share a highly conserved DNA-binding domain of approximately 110 amino acids, the forkhead box domain (also known as the winged-helix domain). Within this family, the O subgroup contains four members: FOXO1 (FKHR), FOXO3 (FKHRL1), FOXO4 (AFX), and FOXO6 [77]. The first three are ubiquitously expressed but at various levels, depending on the tissue [77]. FOXO3 localizes to mitochondria [46], which may contribute to the various physiological and pathophysiological functions of this transcription factor. In tumors of AAs, the expression of FOXO3a is lower by 65% as compared to adjacent normal tissues [206].

3.5.7 Aryl hydrocarbon receptor in the mitochondria

The cellular mechanisms that respond to various toxic compounds involve the aryl hydrocarbon receptor (AHR). The AHR is a ligand-activated transcription factor within the Per-Arnt-Sim (PAS) domain superfamily. Exposure to the most potent AHR ligand, 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), is associated with various pathological effects. A portion of the cellular pool of AHR is found in the mitochondria [150], localized to the inter-membrane space of the organelle [150]. The AHR interacts with ATP5α1, a subunit of the ATP synthase complex, and modulates mitochondrial function [360]. An ethnic variability in the allelic distribution of AHR receptor codon 554 and an assessment of variant receptor function have been described [391].

3.6 miRNAs in mitochondria and their role in cancer

MicroRNAs (miRNAs) are small, non-coding RNAs, with 18–22 nucleotides, which regulate expression of mRNA. The miRNAs bind to 3′ untranslated regions (UTRs) in a sequence-specific manner and either inhibit translation or degrade the mRNA. Post-transcriptional regulation by miRNAs is implicated in proliferation, apoptosis, and differentiation; in several human cancers, miRNAs are de-regulated.

One of the first studies to link miRNA-mediated mitochondrial dysfunction with cancer showed that mir-15a induces an efflux of cytochrome c and disrupts the mitochondrial membrane potential [116]. In CLL, mir-15a is frequently deleted [48]. In various human cancers, mir-143, which targets the pro-proliferative factor ERK5 and induces apoptosis through mitochondrial dysfunction, is downregulated [3]. miRNAs are linked to the control of mitochondrial dynamics and mtDNA integrity, factors associated with cancer initiation. The mir-30 family regulates mitochondrial fission and apoptosis through dynamin-related protein 1 (DRP1) and the p53 axis [208]. In some tumors, miR-21 levels are increased and downregulate the expression of PTEN, deactivate PINK1, and prevent the clearing of damaged mitochondria [253, 415]. In hepatocellular carcinoma cells, mir-200a, which targets the mitochondrial transcription factor TFAM [404] and acts as a tumor suppressor [416], is downregulated in breast cancers [404].

miRNAs are implicated in the regulation of the mitochondrial metabolic apparatus. They modulate several enzymes of the Krebs cycle at the transcriptional level. In glioma cells, miR-183 downregulates the levels of isocitrate dehydrogenase-2 (IDH-2) [358]. Succinate dehydrogenase (SDH) is targeted by miR-210, and, in response to oxidative stress, malate dehydrogenase levels in mouse neuronal cells are elevated through reduction in miR-743a levels [106]. Proteins of the ETC are also targets of miRNAs. Over-expression of miR-338 inhibits the expression of the cytochrome oxidase-IV subunit [208], whereas over-expression of miR-181c leads to down-regulation of the cytochrome oxidase-I subunit [84]. In various cancers, including malignant mesothelioma (MM) [365], gastric cancer [109], and breast cancer [421], miR-126 is frequently deleted. In an MM cell line, mir-216 has a tumor suppressor role [363, 364], and its over-expression down-regulates ATP-citrate lyase and increases production of ATP and citrate. Although cancer cells switch to aerobic glycolysis and divert the metabolites into anabolic pathways to support malignant growth, this proliferative program is suspended during metastasis. Invasive cancer cells instead favor OXPHOS and increased ATP production. In breast cancer cells, this is mediated by the transcription co-activator, PGC-1α [200]. In breast cancer tissues, miR-485-3p and miR-485-5p inhibit PGC-1α to regulate mitochondrial respiration and are consequently down-regulated [227]. An indirect route for inducing cancer is through ROS generation and subsequent disruption of the ETC. By inhibiting the antioxidant enzymes, Mn superoxide dismutase (MnSOD), glutathione peroxidase, and theorductase-2, miR-17^* suppresses tumorigenicity of prostate cancer by inducing cancer cell death [394].

Although miRNAs are largely encoded by the nuclear DNA and perform their activities in the cytosol, pre-miRNAs and mature miRNAs are present in human mitochondria [18]. After predicting 33 pre-miRNA and 25 miRNA candidates in the mitochondrial genome by in silico analysis, miR-365, pre-miR302a, and pre-let-7b were co-localized in the mitochondria. Forty-six miRNAs that were at significantly higher levels in the mitochondrial RNA fraction were isolated. Furthermore, putative target sites for these miRNAs – referred to as mitomiRs – were detected in mtDNA. Although the presence of miRNAs in mitochondria is not indicative of their origin, the pre-miRNAs suggest the possibility of post-transcriptional processing of miRNAs in the mitochondria. The functional significance of mitomiRs was underscored by the co-localization of argonaute 2 (Ago2), part of the RNA-induced signaling complex (RISC), with mtDNA-encoded Cox3 mRNA [17]. Further, in the mitochondria of Hela cells, there is an enrichment of 13 nuclear-encoded miRNAs. Regardless of their origin, the mitomiRs contribute to the regulation of genes involved in mitochondrial functions. Thus, in several human cancers, there is de-regulation of mitomiRs, which is associated with mitochondrial dysfunction or mitochondria-associated mechanisms.

3.7 Kinases in mitochondria and their role in cancer

The signaling pathways that regulate and integrate cellular functions are largely dependent on concerted phosphorylation and de-phosphorylation of the pathway mediators and targets. Phosphorylation of proteins occurs via the kinases, and de-phosphorylation depends on the action of phosphatases. Although the regulatory activities of these signaling proteins have been documented for the cytoplasm and the nucleus, especially in the context of steady-state and pathological conditions such as cancer, their specific functions in the mitochondria have only recently been uncovered [216, 309].

3.7.1 Activated protein kinase B (Akt) in mitochondria

Akt, a serine-threonine kinase, is implicated in cellular processes such as apoptosis, metabolism, and proliferation. An early indication of the mitochondrial functions of Akt was the discovery that Akt phosphorylates and inactivates the apoptotic protein BAD, thereby promoting cell survival [86]. A later study implicated Raf-1 in the anti-apoptotic action of Akt, as the expression of a dominant-negative version of the mitochondrial Raf-1 in Akt-expressing cells rendered them susceptible to apoptosis [232]. Subsequently, in response to mitochondrial dysfunction, Akt was detected in the mitochondrial outer and inner membranes [27, 322].

Akt is part of the PI3K-Akt signaling pathway that is active in cancer cells and promotes cell survival by modifying cellular metabolism and inhibiting apoptosis [20]. The functional significance of Akt in mitochondria is highlighted by its role in regulating both glycolysis and OXPHOS. In cancer cells, Akt activates hexokinase II at the outer mitochondrial membrane and promotes glycolysis [120]. However, over-expression of Akt increases mitochondrial respiration in PTEN−/− murine fibroblasts relative to that in wild-type cells [119]. The authors hypothesize that the Akt-mediated increase in OXPHOS could be part of a metabolic adaptation in response to inactivation of the tumor suppressive PTEN. The role of the PI3K-Akt pathway in apoptosis is complex. In a recent study, the PI3K-Akt-mTOR pathway was found to be involved in the chemo-resistance of human seminoma cells through negative regulation of the mitochondrial apoptotic pathway [114]. Furthermore, the activation of HK-II by mitochondrial Akt had an anti-apoptotic effect on the cells [308]. In contrast, a pro-apoptotic role of Akt was seen in HCT11 cells expressing a mutant k-ras oncogene; Akt rapidly translocated into the mitochondria and increased ROS production, leading to cell death [155].

Recently, Buroker reviewed the genetic variants of Akt and their involvement in various cancers [44]. Three rSNPS (rs10157763, rs10927067 and rs2125230) in Akt codon one are associated with an aggressive form of prostate cancer. Other intron one rSNPs (rs4132509, rs12031994, rs2345994) are associated with risk of renal cell carcinoma. All these SNPs reside in the transcription factor binding site and consequently affect downstream signaling pathways.

3.7.2 Pyruvate dehydrogenase kinase 1 in mitochondria

Pyruvate dehydrogenase kinase 1 (PDK1) is involved in the phosphorylation and inactivation of pyruvate dehydrogenase (PDH). PDK1, along with the pyruvate dehydrogenase complex (PDC), resides in the mitochondrial matrix and is involved in balancing the metabolic output of mitochondria between glycolysis, the citric acid cycle, and OXPHOS.

In a recent study, the ‘gatekeeper’ role of PDK1 (and thus cellular metabolism) was elucidated in oncogene-induced senescence [171]. Senescence is commonly defined as a block in cellular proliferation with no cessation in the metabolic activities of cells. Oncogene-induced senescence (OIS) is hypothesized to be a protective mechanism that withdraws the cells from the proliferative pool and thus averts malignant transformation [58]. In human melanoma cells, forced expression of the oncogene BRAF leads to a senescent phenotype with the suppression of PDK1 and simultaneous induction of the PDH-activating enzyme, pyruvate dehydrogenase phosphatase 2 (PDP2). The resulting activation of PDH increases respiration and production of ROS. Upon forced expression of either PDK1 or PDP2, OIS is abolished, leading to the development of BRAF melanomas. Consistent with these results, the RNAi-mediated attenuation of PDK1 and the EGF receptor in glioblastoma cells reversed the Warburg effect towards OXPHOS and inhibited the growth of glioblastomas [371].

3.6.3 HER1 and HER2 kinases in mitochondria

The epidermal growth factor receptor (EGFR) family of receptor tyrosine kinases consists of four members: EGFR/ErbB-1/HER-1, ErbB-2/HER-2/neu, ErbB-3/HER-3, and ErbB-4/HER-4; all except ErbB-3 are associated with tyrosine kinase activity. Ligand-bound, activated EGFR family receptors trigger downstream signaling pathways, including PI3K, MAPK, STAT, phospholipase C, and Ca2+/calceneurin [73]. The HER1 and HER2 kinases, which are involved in cancer initiation, progression, invasiveness, and metastasis, are expressed and/or constitutively activated in various human malignancies, especially in breast cancer [225], lung cancer [117], and glioblastoma multiforme [148].

The mitochondrial localization of HER1 was established by use of immunofluorescence and immune-electron microscopy [412]. The translocation of HER1 into mitochondria is facilitated by EGF-dependent binding to the cytochrome c subunit II (COX2); the resulting phosphorylation of COX2 leads to a decline in CoxII activity, with a resultant drop in cellular ATP levels [96]. The same study also reported a putative mitochondrial localization sequence in HER1. A later study showed that the mitochondrial uptake of HER1 was increased upon exposure to apoptosis inducers and that, in cancer cells, mitochondrial targeting of HER1 was responsible for drug resistance [115, 116]. Together, these findings indicate that mitochondrial HER is involved in modulating metabolism, apoptosis, and cell survival. Further studies are needed to clarify the connection between the mitochondrial localization of HER1 and oncogenesis. The presence of HER2 in mitochondria was reported by Ding et al., who linked HER2 with the regulation of metabolism and resistance to treatment [102]. HER2 possibly interacts with complex IV of the ETC [349].

A dinucleotide CA repeat polymorphism in intron 1 of HER1, ranging from 14 to 21 repeats, has been implicated in the regulation of EGFR expression. Liu et al. evaluated the allelic distribution of this polymorphism in various ethnic groups [222]. The longer allele 20 is higher in Asian individuals; the shorter allele 16 is the most common allele in Caucasians and AAs. In a retrospective study by Nomura et al., both the longer and shorter CA allele was found to be present in East Asian individuals with non-small cell lung carcinoma, with the latter present at a higher frequency [276]. In the same study, two SNPs, G216T and C191A, were present at higher frequencies in lung cancer tissues of patients of Northern European and AA descent compared to East Asians. Another functional SNP, rs4444903, is associated with a higher risk of hepatocellular carcinoma [1]. In Austrian women, a polymorphism in the HER2 codon 655, rs1136201, corresponds to a more aggressive form of breast cancer [384].

Also localized in the mitochondria are the kinases, Abl, ATM, Src, JNK, ERK1/2, P38MAPK, GSK3B, PKA, PKC, and PINK1 [216]. However, their role in cancer disparities is unknown.

3.8 Phosphatases in mitochondria

Both protein and lipid phosphatases localize in the mitochondria [216]. Protein phosphatases include MAP kinase phosphatase (MKP1), Src homology two domain-containing phosphatase 2 (Shp2), and protein tyrosine phosphatases (PTPs). A lipid phosphatase, PTEN, also localizes to the mitochondria. To date, a role for these mitochondrial phosphatases in cancer disparities is not described.

3.9 Oncogenes in mitochondria

The Ras family of proteins is small GTPases that function as binary molecular ‘switches,’ depending on their GTP/GDP binding status [375]. There are four proteins in the Ras family: HRAS, NRAS, KRASA, and KRASB, the latter two being splice variants of KRAS [295]. The Ras proto-oncogenes are expressed in all cell lineages and are mutated in 20–30% of human tumors [39]. Ras-mediated oncogenesis involves the downstream PI3K-Akt pathway and leads to increased cell proliferation, survival, and invasiveness [128]. The localization of all Ras isoforms in the mitochondria has been confirmed [302,306, 310, 390].

Mitochondrial Ras affects metabolism, apoptosis, and mitochondrial biogenesis, which can often be correlated with tumor initiation and/or maintenance. In a murine model of acute myeloid leukemia, NRAS was associated with the anti-apoptotic Bcl-2 protein and protected the leukemic cells from apoptosis [277]. Similarly, mouse fibroblasts lacking NRAS displayed mitochondrial dysfunction and increased apoptosis that could be alleviated upon mitochondrially targeted delivery of NRAS [390]. Cells transformed with the K-ras gene, however, were susceptible to apoptosis and were involved in protein kinase C (PKC)-mediated translocation of KRAS to the outer mitochondrial membrane in association with the Bcl-XL anti-apoptotic protein 30]. In the mitochondria, mutant KRAS also increase ROS via the PI3K-Akt pathway, leading to cell death. In HEK293 cells, translocation of HRAS and KRAS into the mitochondria leads to a reduction of mitochondrial membrane potential and respiration, thus causing a metabolic shift to glycolysis [406]. In glioblastoma cells, RAS-mediated signaling is implicated in the inactivation of pyruvate dehydrogenase, the enzyme that feeds acetyl-CoA into the citric acid cycle for cellular respiration [291]. For human hepatocellular carcinoma, KRAS-induced mitophagy (clearance of diseased mitochondria) and the resulting loss of mitochondria has been linked to early tumorigenesis [180].

A germline SNP, rs61764370, located in the 3′UTR of the KRAS oncogene alters the binding capacity of the let-7 miRNA. Uvirova et al. found, in a Czech cohort, an association of this somatic variant across breast cancers, colorectal cancers, non-small cell lung cancers, and brain tumors [368]. In another study, the rs61764370 SNP was associated with a higher risk of chronic myeloid leukemia in Mexican-Mestizo women [129]. An indication of the role of HRAS polymorphisms in cancer susceptibility came from a study of an Italian gastric cancer cohort. Thirteen rare inheritable variants were detected in the tandem-repeat sequence downstream of the structural part of the HRAS gene and were correlated with a higher incidence of cancer [297]. In a meta-analysis of the prevalence of genetic polymorphisms in breast cancer, rare alleles of HRAS were associated with higher risk [288]. Regarding NRAS, the only relevant genetic variant discovered is the G138R SNP, which is associated with higher colorectal cancer risk among Taiwanese patients [61].

4. DISPARITIES AT THE NUCLEAR MITOCHONDRIAL PROTEIN LEVEL AND THEIR ROLE IN CANCER

Although most mitochondrial proteins are synthesized in the cytosol by cytoplasmic ribosomes and then translocated into the mitochondria with the help of the mitochondrial translocase complex, several mitochondria-targeted proteins are not exclusively mitochondrial. This dual targeting could involve other organelles, including the endoplasmic reticulum, the nucleus, and the cytosol. The shuttling of proteins between the mitochondria and nucleus can be passive and, especially if the proteins are smaller than 25–30 kDa or selective require the presence of nuclear or mitochondrial localization signals [407]. Examples of these dual-localized proteins are the transcription factors and kinases that have been discussed in previous sections. A small number of primarily mitochondrial proteins can also be found in the nucleus; the functions of the proteins could be different in the two organelles.

Fumarate hydratase (FDH), the enzyme that catalyzes a reversible conversion of fumarate to malate, is a TCA cycle enzyme that is localized and active in the mitochondrial matrix. To elucidate the role of a cytosolic version of FDH, Yogev et al. generated a nuclear fum−/− yeast strain in which a copy of the gene was inserted into the mitochondria genome [407]. Although the mutant cells had a normal TCA cycle, they displayed a higher sensitivity to radiation-induced DNA damage. This effect was reversed when a cytosolic version of FDH (lacking the mitochondrial targeting signal) was introduced into these cells. Furthermore, FDH was translocated into the nucleus of HeLa cells exposed to radiation or hydroxyurea. These findings point to a role of FDH in DNA damage in the nucleus. This hypothesis is strengthened by the increased predisposition to renal cancer and leiomyomatosis following bi-allelic loss of the FDH gene.

The pyruvate dehydrogenase complex (PDC), a large multi-protein complex residing in the mitochondrial matrix, generates acetyl-CoA and feeds it into the TCA cycle. PDC, localized in the nucleus, is involved in the acetyl-CoA synthesis and histone acetylation [351]. Upon mutagenic stress, the nuclear levels of PDC increased, and the mitochondrial levels concomitantly decreased, prompting the authors to postulate that the PDC is translocated from the mitochondria to the nucleus as an intact complex. Further studies are needed to correlate this function of PDC to cancer initiation.

The mitochondrial coenzyme Q biosynthesis protein 7 (COQ7) is a monooxygenase catalyzing the hydroxylation of 5-demethoxyubiquinone to 5-hydroxyubiquinone. Although the presence of a mitochondrial localization signal marks COQ7 as an exclusively mitochondrial protein, Monaghan et al. have recently reported its localization in the nuclei of HeLa cells [267]. This was attributed to a nuclear localization signal that could be cleaved before transfer into the mitochondria. Accumulation of COQ7 in the nucleus is stimulated by ROS and lowered by antioxidants. Moreover, in human cells, deletion of COQ7 increases the levels of ROS and the ROS-responsive genes, SOD2 and NRF2. COQ7 is associated with chromatin and implicated in the transcriptional regulation of TIMM22 (mitochondrial import inner membrane translocase subunit Tim22) [327].

5. DISPARITY AT THE LEVEL OF MITOCHONDRIA-TO-NUCLEUS CROSS TALK AND ITS ROLE IN CANCER

In human cells, mitochondrial dysfunctions lead to mitochondria-to-nucleus retrograde responses. Mitochondria are multi-tasking organelles that integrate pathways of cellular respiration, biosynthesis, apoptosis, and redox potential. Due to the paucity of genes in the mitochondrial genome, most proteins and regulatory RNAs needed for mitochondrial structure and function are encoded by the nuclear genome. Thus, to maintain functioning mitochondria and a healthy cell, constant communication between the nucleus and mitochondria is required. Deregulated expression of nuclear genes encoding the mitochondrial proteins can result in various pathologies. Further, mitochondrial dysfunction arising from cellular/environmental imbalances can signal changes in nuclear gene expression, leading to cellular adaptation. The mito-nuclear cross talk can, therefore, be split into two broad categories: 1) anterograde signaling that involves the regulation of mitochondrial biogenesis and functions through nuclear-encoded transcription factors and other proteins and 2) retrograde signaling that relays metabolic and oxidative changes in the mitochondria to the nucleus and leads to reconfiguration of the nuclear transcriptome.

5.1 Mitochondrial retrograde signaling

Mitochondrial retrograde signaling, first observed in Saccharomyces cerevisiae [214], has been reported in mammalian cell. The main ‘sensors’ that trigger a retrograde response include mtDNA mutations and/or loss in copy numbers, disruption of the OXPHOS machinery, an increase in oxidative stress, and loss of mitochondrial membrane potential. Retrograde signaling causes changes in nuclear gene expression that can lead to metabolic rearrangement (relevant in the context of cancer cells), apoptosis, mitophagy, and DNA repair [28, 45, 122, 161]. Mitochondrial retrograde signaling is linked to various pathological conditions. Shoffner et al. showed that, in the MERRF (myoclonic epilepsy with ragged red fibers) condition, disruption of the ETC and the resulting low production of ATP-triggered retrograde signaling [336]. To compensate for impaired muscle activity, the proliferation of the damaged mitochondria was increased in the muscle cells. A direct association between mitochondrial dysfunction and retrograde signaling and aging has also been found by use of DNA polymerase γ knock-in (Polg^−/−) mice. A knock-in D257A mutation in the exonuclease domain of POLG abolishes its DNA-proofreading capacity and increases the frequency of mtDNA mutations, with the mice displaying signs of premature aging [188, 366]. Mitochondrial retrograde signaling has also been implicated as a causal factor in tumorigenesis [29, 36, 37, 56, 94, 95, 193]. In particular, studies involving murine models and cybrids have established a causal link between mtDNA defects and cancer [169]. Mice heterozygous for TFAM, the regulator of mtDNA replication and transcription, were crossed with mice heterozygous for adenomatous polyposis coli multiple intestinal neoplasias (APCMin^+/−). The double-heterozygote offspring showed an increase in the rates of tumor incidence and growth [392].

Even though various nuclear targets of stress-initiated retrograde signaling have been identified, the complete genome-wide rearrangement remains to be characterized. The transcriptional profiles of cells responding to mitochondrial retrograde signaling have been established and compared that to those of the ‘resting’ state. To determine changes mediated by retrograde signaling, most of these investigations involved use of Rho0 or mtDNA-depleted versions of the parental cells [28, 29, 94, 95, 97, 122–125, 231, 258]. The consensus is that the genes affected by retrograde signaling control diverse functions, in particular, mitochondrial biogenesis, metabolism, cell adhesion, metastasis, and apoptosis. In Rho0 cells, Guha et al. found higher expression of the general transcription factor TFIIH and the hematopoietic-specific AML1 (RUNX1). Delsite et al. observed that Rho0 cells display reduced expression of cytochrome p450 (metabolism), CDK inhibitor p19 (cell differentiation), and signaling molecules (PKCγ and protein tyrosine phosphatase C) [94]. For Rho0 breast epithelial cells, Kulawiec et al. reported an alternate regulation of the p53-mediated signaling, leading to increases in DNA breaks and chromosomal rearrangements [192]. In osteosarcoma cells, there is up-regulation of HIF1α, which confers metabolic advantages to malignant cells. The retrograde signaling activates the JNK and PGC1α transcription factors that initiate pro-survival and mitochondrial biogenesis pathways [156].

In tumor cells, the metabolic switch and retrograde signaling often re-program the genetic components of cellular respiration. In several tumor cell lines, there is activation of the Akt transcription factor in response to mitochondrial respiratory stress [122, 286]. ROS-mediated disruption of complex I of the ETC chain is linked with Akt activation [330]. Moreover, Akt phosphorylates a transcriptional co-activator, hnRNAP2 (heterogeneous ribonucleoprotein A), which is overexpressed during retrograde signaling. Implicated in the post-transcriptional processing of mRNA [242, 385], hnRNPA2 has been considered, in several clinical studies, as an early biomarker for lung epithelial carcinoma and metastasis [362, 419]. There is over-expression of hnRNAP2 in cancers of the brain, colon, breast, cervix, and ovary [285]. Guha et al. confirmed that hnRNPA2 is up-regulated in response to low mtDNA and mitochondrial stress signaling; it acts as a proto-oncogene by increasing the proliferation, survival, and invasiveness of cancer cells. Silencing of hnRNPA2 leads to apoptosis [123].

The regulatory action of hnRNPA2 in enabling cells to adapt to a malignant condition (such as through a metabolic shift) occurs by two mechanisms. It acts as a co-activator of transcription factors such as Akt and IGHR1 [122, 123], and it introduces alternative splicing as a means of regulation. In breast cancer cells, the hnRNPA 1 and two proteins modulate expression of the pyruvate kinase gene by alternative splicing, an action that leads to expression of pyruvate kinase isoform M (PKM) [67, 86]. This embryonic isoform, which is re-expressed in cancers, promotes aerobic glycolysis, as opposed to the adult isoform, PKM1, which promotes OXPHOS [72]. In breast and colorectal epithelial cells, alternative splicing of the GTPase Rac1b by hnRNPA1 correlates with cancerous transformation [287]. Furthermore, in hepatocellular cancer cells, hnRNPA2 binds to the telomeric repeat TTAGGGn sequence and human telomerase (hTERT), probably to maintain telomere length and indirectly affect the checkpoints for cell cycle arrest [262]. This is an indication of an epigenetic link between mitochondrial dysfunction (and retrograde signaling) and nuclear gene expression.

There are several mechanisms of mitochondrial retrograde signaling. The most widely studied are a) ROS-mediated, b) the mitochondrial unfolding protein response, c) Ca2+ gradient and calcineurin, d) NAD⁺/NADH-mediated, e) ATP-mediated, AMP-activated protein kinase signaling, and f) inter-genomic mipigenetic mechanisms. These may respond to different stimuli, but they often share various signaling molecules and converge on the common goal of nuclear genome regulation.

5.2 Mitochondrial ROS-mediated signaling

Inefficient electron transport through the ETC complexes leads to formation of free radicals or ROS, which, upon exceeding the capacity of the antioxidant enzymes, leads to damage of the ETC and respiratory stress [181, 399]. The production of ROS in mitochondria is enhanced by various stresses, including hypoxia, chemotherapy, radiation, and other pathophysiological injuries [110]. Neurological disorders such as the NARP (neuropathy, ataxia, and retinitis pigmentosa) syndrome are ascribed to the action of ROS on mitochondrial biogenesis [389]. Mitochondrial ROS production and ROS-mediated signaling are the main causes of lung fibrosis following alveolar injury [418]. Apart from inducing retrograde signaling, ROS also acts indirectly by oxidatively damaging mtDNA, as the proximity of mtDNA to the ETC makes it particularly susceptible to mutations.

ROS-mediated signaling is implicated in tumor initiation and progression. Formentini et al. demonstrated that human colorectal cancer cells overexpress ATPase inhibitory factor 1 (IFI1), which depolarizes the mitochondrial membrane and increases ROS production, which, in turn, activates NFkB and pro-survival genes [110]. The IFI1-ROS nexus is also responsible for the metabolic switch of cancer cells to aerobic glycolysis [319]. Increased production of ROS activates HIF1α and HIF2α, which, in melanomas, turn on expression of the Src oncogene and promote metastasis [136]. With cybrid technology, Ishikawa et al. showed that ROS-induced mutations in the mtDNA-encoded NADH dehydrogenase subunit 6 (ND6) increased the metastatic potential of tumor cells engrafted in mice [154]. In cancer cells, ROS-mediated deregulation of mitochondrial redox leads to activation of the Akt proliferation and survival pathways [286]. Singh’s group has demonstrated that retrograde mitochondria-to-nucleus signaling involves regulation of NADPH oxidase (Nox1) and that this enzyme is over-expressed in most breast and ovarian tumors [58].

5.3 Mitochondrial unfolded protein in signaling

The mitochondrial unfolded protein response (UPR) is a type of retrograde signaling that is initiated upon the efflux of peptides from the damaged mitochondrial matrix proteins to the cytosol [182, 228]. These peptides can then translocate into the nucleus and activate nuclear transcription factors. By maintaining an equilibrium between folded and unfolded proteins, the UPR is responsible for mitochondrial integrity and biogenesis. Mitochondria have molecular chaperones to facilitate protein folding and proteases that degrade misfolded proteins. Increased accumulation of unfolded, misfolded, and unassembled proteins due to damage to the protein-folding machinery can lead to organelle dysfunction. Due to the potential sensory function of damaged mitochondrial peptides, the mtUPR is recognized for its role in synchronizing expression of mitochondrial and nuclear genes [168]. Currently, two pathways of mtUPR are postulated in mammalian cells. In the first model [316], accumulation of unfolded proteins in the mitochondrial matrix with the help of heat-shock protein (HSP) chaperones leads to their cleavage by Clp proteases and subsequent efflux of the peptides. The damaged peptides then activate JNK-mediated stress signaling. In the second model, accumulation of unfolded proteins in the inter-membrane space, mtUPR is mediated by the estrogen receptor. ERα-mediated mtUPR upregulates the transcriptional activators, NRF1 and NRF2, which control mitochondrial biogenesis [280].

In various patho-physiological conditions, such as neurodegenerative disorders and aging, the mtUPR acts as a rapid stress response. The accumulation of amyloid beta protein in the mitochondria is a characteristic of Alzheimer’s disease (AD) [57]. Although amyloid β protein is cleaved by the mitochondrial HTRA2 chaperone/Omi serine protease complex, a direct role of mtUPR in the pathology of AD has not been demonstrated [283]. There is an accumulation of α-synuclein in the brain mitochondria of Parkinson’s disease patients (PD) [101]. The added presence of HSP60 in the brains of these patients is an indication of the involvement of mtUPR in PD [100, 330]. mtUPR is implicated in the regulation of the aging of hematopoietic stem cells (HSCs) [265]. Activation of mtUPR in aged HSCs results in an induction of the NAD+-dependent deacetylase, SIRT7. SIRT7 binds to the transcription factor, NRF1, which attenuates mitochondrial translation, leads to HSC quiescence and prevents differentiation.

5.4 Mitochondrial Ca2+ signaling

Mitochondria maintain a Ca2+ gradient across the inner membrane by use of gated Ca2+ channels and Na+/H+-Ca2+ exchangers [295]. The uptake and release of Ca2+ regulate various physiological processes, including ATP production. Disruption of mitochondrial Ca2+ signaling triggers the opening of permeability transition pores (mPTPs) and initiates the apoptotic cascade [131, 354]. Since the uptake of Ca2+ depends on the mitochondrial membrane potential, any disruption in the latter due to low mtDNA copy numbers, the breakdown of mitochondrial respiration, or treatment with mitochondrial ionophores disrupts the uptake of Ca2+. The leakage of Ca2+ into the cytosol leads to activation of the phosphatase, calcineurin (Cn) [6, 124]. Ca2+/Cn-mediated retrograde signaling culminates in changes in the expression of an array of kinases, phosphatases, and nuclear transcription factors that affect Ca2+ storage, mitochondrial biogenesis, mtDNA transcription, cell survival, and apoptosis [6, 29, 124].

The signaling pathways activated by the Ca2+/Cn relay in mouse skeletal myoblast C2C12 cells have been studied by Guha et al. Increased expression of Cn leads to activation of calmodulin-dependent protein kinase (CAMK), which activates the downstream targets, Akt Ca2+/Cn-dependent transcription factor (NFAT), cAMP response element binding protein (CREB), CCAAT enhancer binding protein delta (CEBPδ), and NFκB, which are then translocated into the nucleus.

5.5 Mitochondrial signaling mediated by NAD+/NADH

Nicotinamide adenine dinucleotide (NAD) in its oxidized (NAD+, electron acceptor) and reduced (NADH, electron donor) forms is a metabolite that is localized to mitochondria, cytoplasm, and in the nucleus. As a coenzyme, NAD regulates electron transport and the TCA cycle in the mitochondria. It also regulates glycolysis [53, 318, 406]. NAD+ serves as a substrate for two families of proteins involved in response to the DNA damage. These proteins include the PARP family and the sirtuin family of NAD+-dependent deacetylases (SIRTs) [278, 318].

Of the 17 members of PARP family, PARP1, 2, and 3 functions in DNA repair. PARP1 and PARP2 are involved in base-excision repair (BER). However, PARP3 is involved in signaling double-strand breaks (DSB). The PARPs, when sensing DSBs in the nuclear DNA, synthesize poly(ADP-ribose). PARPs utilize NAD+ as a substrate and transfer the ADP-ribose moiety of NAD+ to acceptor proteins. PARP1 and PARP2 are then auto-PARylated, and further recruit and PARylate other repair enzymes [34, 47, 315, 318]. PARP1, which localize to the mitochondria [311], PARylate POLG and EXOG and, in contrast to its function in the nucleus, inhibit BER [355].

The sirtuin (SIRT) proteins also use NAD+ as a substrate. SIRTs deacetylate proteins by the removal of acetyl groups from lysine residues [256]. The SIRT family contains seven members. SIRT1, SIRT6, and SIRT7 localize in the nucleus; SIRT2 in the cytoplasm; and SIRT3, SIRT4, and SIRT5 in the mitochondria [256, 278]. SIRT1 regulates mitochondrial biogenesis. It deacetylates peroxisome proliferator-activated receptor-γ co-activator 1α (PGC-1α). An increase in Ca2+ levels in the mitochondria influences the NAD+/NADH ratio in the cytoplasm by causing an efflux of NADH from the mitochondria to the cytosol [240]. This change influences the NAD+-consuming SIRT enzymes and provides a link between Ca2+ fluctuations, NAD+ levels, DNA damage and repair, and mitochondrial dysfunction [240].

5.6 ATP-mediated signaling

AMP-activated protein kinase (AMPK), a sensor of intracellular energy (ATP:AMP ratio) levels [334], is a serine/threonine kinase that triggers orchestrated responses to restore ATP levels and maintain energy balance [51, 334]. p53 is a target of AMPK phosphorylation, and this phosphorylation triggers the accumulation of mitochondrial p53 that promotes apoptosis via the Bak-Bcl-xL complex [166]. In response to oxidative stress, the ataxia telangiectasia mutated (ATM) kinase phosphorylates AMPK [333]. Phosphorylated AMPK, in turn, phosphorylates SIRT1 and PGC-1α, which are involved in mitochondrial nuclear signaling responses [144, 158, 314]. Mitochondrial SIRT4 also contributes to mito-nuclear crosstalk via AMPK-mediated signaling [181].

5.7 Mito-nuclear mechanisms underlying tumorigenesis

The mipigenetic mechanism is defined as mito-nuclear intergenomic cross-talk at epigenetic and genetic levels involving gene expression and genomic instability [339]. Epigenetic and genetic mechanisms regulate gene expression. Our studies have identified an effect on the epigenetic landscape of the nuclear genome as a consequence of mitochondrial dysfunction. In particular, we demonstrated reversible and irreversible changes in genomic DNA methylation profiles of the nuclear genome [258].

5.7.1 Epigenetic changes in nuclear genome

Perturbed DNA methylation profiles of certain loci within the human nuclear genome are associated with depleted mtDNA. Depletion of the mitochondrial genome in two different cell types resulted in the aberrant methylation of promoter CpG islands (high CG-rich regions) that were unmethylated in the parental cell line [258]. Repletion of mtDNA resulted in the partial re-establishment of methylation profiles back to their original parental state. These results provide an insight into the role of mitochondria in establishing or maintaining nuclear DNA methylation at the genomic level [258].

There is an interdependent relationship between mitochondria and the nuclear genome, including DNA methylation in both the nucleus and the mitochondria [258]. Mammalian mitochondria have mitochondrial DNA methyltransferase 1 (mtDNMT1) activity, 5-methylcytosine (5mC), and 5-hydroxymethylcytosine (5hmC). Shock et al. showed that translocation of nuclear DNMT1 to the mitochondrial matrix is regulated by expression of a conserved mitochondria-targeting sequence, upstream of the gene’s transcription start site within the nuclear-encoded gene [335]. Alterations in mtDNMT1 affect transcription from the light and heavy strands of mtDNA, suggesting a correlation between 5hmC- and 5mC-mediated transcriptional regulation of mtDNA by a nuclear-encoded gene. These findings provide evidence implicating epigenetic regulation of the mitochondrial genome by nuclear-encoded, translocated mtDNMT1 relative to mitochondrial dysfunction. Takasugi et al. described epigenetic regulatory mechanisms affecting the transcriptional control of nuclear DNA-encoded mitochondrial proteins [356]. They used a systematic approach to identifying the presence of differentially methylated regions within nuclear-encoded mitochondrial proteins in a tissue-dependent manner and demonstrated the presence of differentially methylated regions in 636 of 899 nuclear-encoded mitochondrial genes that have functional roles related to the mitochondria.

5.7.2 Genetic changes in the nuclear genome

We have demonstrated that mtDNA dysfunction in human cells to leads a reduced rate of repair of oxidative DNA damage and to increased frequencies of mutation within the nuclear genome [95]. The mitochondria-mediated mutator phenotype is suppressed by inactivating subunits of error-prone DNA repair. Error-prone repair polymerases are involved in the bypass of several types of DNA lesions that have the potential to inhibit chromosome replication [299]. Thus, mitochondrial dysfunction limits or decreases nuclear DNA repair, resulting in unrepaired DNA lesions, which are subsequently converted into mutations by error-prone repair. Increased error-prone bypass of DNA lesions may be characteristic of imbalanced pools of deoxyribose nucleoside triphosphates (dNTPs). Mitochondrial dysfunction induces an imbalanced pool of dNTPs [97]. nDNA mutations may activate proto-oncogenes and/or inactivate tumor suppressor genes, leading to genomic instability, which is involved in the development of cancers.

The functional state of mitochondria in human cells is monitored by the mitochondria damage checkpoint (mitocheckpoint) [258, 343]. The mitocheckpoint coordinates and maintains the balance between apoptotic and anti-apoptotic signals. When mitochondrial dysfunction occurs, the mitocheckpoint is activated to restore normal mitochondrial function and avoid production of mitochondria-defective cells. This response induces changes to the nuclear epigenome (Figure 3) [258].

Figure 3. Differences in mipigenetic mechanisms may contribute to cancer disparity.

Schematic depiction of mechanisms probably involved in cancers induced by mitochondrial dysfunction. Mitochondrial dysfunction or suboptimal function can arise due to mutations or variants in genes involved in nuclear genes encoding the mitochondrial proteome or mitochondrial genes encoding the OXPHOS. See text and Figure 2 for variants described in AAs, Caucasian Americans, and other ethnic populations. A transient mitochondrial dysfunction turns on the mitocheckpoint control to restore the normal mitochondrial function. Restoration involves intergenomic mipigenetic cross talk at epigenetic (e.g., DNA methylation) and genetic levels (gene expression). A severe mitochondrial dysfunction may result in cellular senescence, leading to apoptosis and mitochondrial diseases. A persistent mitochondrial dysfunction arising due to mutations in either nuclear or mitochondrial genes can cause senescence and result in genetic instability in the nuclear genome as well as resistance to apoptosis, underlying tumorigenic cell transformation, and development of cancer.

If mitochondria are severely dysfunctional, they can trigger senescence resulting in apoptosis and mitochondrial diseases (Figure 3). If mitochondrial dysfunction is persistent and defective, mitochondria accumulate in the cell, leading to instability of the nuclear genome [97, 299]. Nuclear genome instability causes cells to acquire new functions, such as resistance to apoptosis and cellular transformation [281, 282], migration, and invasive characteristics involved in tumorigenesis (Figure 3) [191, 258]. Cellular senescence provides a barrier to tumorigenesis [293]. Our studies suggest that a mitochondrial defect leads to cellular senescence and that senescence bypass is induced due to the instability of the nuclear genome [282]. We propose that mitochondrial dysfunction-induced cellular senescence acts as an additional mitocheckpoint mechanism that suppresses tumor development (Figure 3).

6. CONCLUSIONS AND FUTURE PERSPECTIVE

We have presented a list of genetic factors that relate to mitochondrial function. We suggest that variants or mutations in either mitochondrial or nuclear genome-encoded mitochondrial proteins contribute to mitochondrial dysfunction or to suboptimal mitochondrial function, which contributes to cancer diversity and tumor aggressiveness related to racial disparities. Ethnic differences, either at the level of expression or genetic variations in mitochondria-localized transcription factors, kinases and phosphatases, and tumor suppressors and oncogenes may underlie susceptibility to cancer and aggressiveness of cancers frequently observed in AAs and other populations. Differences in mitochondrial function may alter the mipigenetic cross talk between mitochondria and the nucleus and contribute to cancer disparities. Agents that restore mitochondrial function to optimal levels should permit sensitivity to anticancer agents, including radiation treatment of aggressive tumors, thereby reducing racial disparities.