Introduction

Mitochondria are ubiquitous organelles in eukaryotic cells whose primary role is to generate energy supplies in the form of ATP through oxidative phosphorylation 1. The oxidative phosphorylation chain is composed of five protein complexes: NADH-ubiquinone oxidoreductase as complex I, succinate-ubiquinone oxidoreductase as complex II, ubiquinone-cytochrome c oxidoreductase as complex III, cytochrome c oxidase as complex IV and ATP synthase as complex V. Oxidative phosphorylation, defined as the oxidation of electron-carriers by oxygen and concomitant ATP production, provides 90% of the cellular chemical energy required in various biological functions. Recent studies have also shown that mitochondria play a central role in apoptosis 2, 3 and cell proliferation 4. Mitochondria are also major factors in modulating calcium signaling 5, 6, which is a universal second messenger.

Over the last 20 years, mitochondrial dysfunction, including that associated with mtDNA mutations, has been identified in human diseases, including seizure, ataxia, cortical blandness, dystonia, exercise intolerance, ophthalmoplegia, optic atrophy, cataracts, diabetes mellitus, short stature, cardiomyopathy, sensorineural hearing loss and kidney failure 7, 8. Large rearrangements or deletions of the mitochondrial genome and about 200 point mutations, including those in genes encoding proteins for subunits of complex I, III, IV and V, rRNAs and tRNAs, have been linked to a variety of clinical disorders 9, 10. Accumulation of mtDNA mutations has also been suggested to play a major role in aging and the development of various age-related degenerative diseases 11. Interestingly, high levels of mtDNA mutations have been found in many tumors and cancer cells 12, 13, 14.

Mitochondrial genome and mitochondrial genetics

The mammalian mitochondrial genome is a double-stranded circular DNA of 16 500 nucleotides 15, 16. It contains 37 genes encoding 13 peptides for the oxidative phosphorylation apparatus, as well as 22 tRNAs and 2 rRNAs essential for protein synthesis within mitochondria. Besides these coding regions, a displacement loop (D-loop) is also present which contains elements regulating mtDNA replication and transcription.

Most mammalian cells contain hundreds or thousands of copies of mitochondrial genomes 1. Since mtDNA is in the proximity of reactive oxygen species (ROS) generation sites (the byproduct of oxidative phosphorylation) and mitochondria have relatively less sophisticated DNA protection or repair systems, mtDNA is therefore vulnerable to high mutation rates 8. As a result, the mtDNA within a cell could be a blend of both wild type and mutant species, a condition called 'heteroplasmy'. The normal situation, in which all mtDNAs are identical, is referred to as 'homoplasmy'. The neutral polymorphisms are most likely homoplasmic, whereas the pathogenic mutations are usually heteroplasmic in nature. It is expected that, due to the multiplicity of mitochondrial genomes in each cell, a threshold of mutant mtDNA must be reached before cellular dysfunction caused by defective mitochondria becomes apparent.

Because mtDNA replication and segregation are not synchronized with nuclear DNA, daughter cells from the same progenitor could have different mtDNA genotypes 8. When the pathogenic threshold is surpassed in certain cells, the phenotype would change. This explains the time-related and tissue-specific variability of clinical features displayed in mtDNA-related disorders.

Warburg hypothesis and abnormal mitochondria in cancer cells

Cancer cells constitutively upregulate glucose metabolism, even in the presence of abundant oxygen, and synthesize ATP mainly through 'aerobic glycolysis', a metabolic state that is linked to high glucose uptake and lactate production. To explain the fact that cancer cells were high in fermentation and low in respiration, Warburg 17 proposed that cancer originated from a non-neoplastic cell that adopted anaerobic metabolism as a means of survival after injury to its respiratory system, which led to the notion that tumors were initiated by persistent damage to the mitochondria. Since then, changes in the number, shape and function of mitochondria have been reported in various cancers 18. The bioenergetic switch from mitochondrial oxidative phosphorylation to glycolysis has been suggested to be a marker of tumor development or the bioenergetic signature of cancer 19, 20, 21. Furthermore, mitochondrial dysfunction has been shown to initiate critical signaling pathways that regulate cell growth 4, 22. Recent studies suggested that defects in mitochondrial respiration led to elevated levels of NADH, which could subsequently inactivate PTEN through a redox modification mechanism 23. Inhibition of PTEN could activate protein kinase B (Akt) 23, and Akt was shown to enhance glycolysis, possibly through the effects on its key rate-limiting step, phosphorylation of newly acquired glucose by hexokinases 24. Akt also triggers an increase in cell survival 25, which is commonly observed in cancer cells 26. Furthermore, inhibition of oxidative phosphorylation by oligomycin in lung carcinoma was shown to trigger a rapid increase in aerobic glycolysis demonstrating that tumor cells can become glycolytic as a result of suppression of mitochondrial energy production 27. However, when glycolysis was suppressed, tumor cells were unable to sufficiently upregulate mitochondrial oxidative phosphorylation, indicating partial mitochondrial impairment 28. Rapidly growing tumors easily become hypoxic owing to the inability of the local vasculature to supply an adequate amount of oxygen. As a result, tumor cells upregulate the glycolytic pathway by inducing hypoxia-inducible factor 1 (HIF-1) 29. HIF-1 plays an important role in tumorigenesis and will be described in detail in a later section.

mtDNA mutations in cancers

It is interesting to note that, even before DNA sequencing technology was available, abnormal mtDNA was observed in leukemic myeloid cells using electron microscopy 30, 31. Subsequently, mutations in both the non-coding and coding regions of the mtDNA have been identified in various types of human cancers, and the majority of the mutations appeared to be homoplasmic in nature 32, 33. One of the first comprehensive studies of mtDNA in cancer cells demonstrated that among 10 colorectal cancer lines, seven of them exhibited mutations in their mtDNA content 12. The mtDNA mutations were found in rRNA (12S and 16S) genes, subunits of complex I (ND1, ND4L and ND5), complex III (cytochrome b) and complex IV (COXI, COXII and COXIII). A total of 11 out of 12 mutations were nucleotide substitutions, while the remaining mutation was a single base pair insertion. Moreover, all of these mutations were true somatic mutations and did not exist in constitutional mtDNA from the same patient. Similarly, mtDNA mutations within the D-loop control region have been reported as a frequent event in ovarian, gastric and hepatocellular carcinomas 34, 35, 36. Specifically, it was suggested that in the D-loop region, a poly-C stretch (C-tract), termed the D310 region, is more susceptible to oxidative damage and electrophilic attack compared with other regions of mtDNA 37. In another study, it was found that in renal carcinoma, mtDNA harbored disruptive point mutations in eight of nine tumors, seven tumors with complex I genes mutations 38, and one with mutation in a complex III gene.

Increasingly systematic analyses of mtDNA have been performed in various cancers, and in this review, we surveyed 101 papers published between 1998 and 2008, using Medline searches with 'mtDNA mutation cancer/tumor' as keywords. The results of these analyses are summarized in Tables 1, 2, 3, 4. From these reports, we found that the majority of identified mutations (635) were located in the D-loop region (Table 3), as this region can accumulate variances quite easily. In addition, some authors focused solely on the D-loop region when attempting to detect mtDNA mutations. Interestingly, more mutations have been found in genes encoding complex I subunits (Table 1), with 593 mutations reported at the occurrence of 9.3%, compared with genes for other respiratory complexes (Table 2). Similar levels of mutations have been reported in tRNA and rRNA genes, with frequencies of 3.7% and 4.3%, respectively (Table 4), although mutations in tRNA genes are much more prevalent in patients with other mitochondrial diseases. The relatively common 4977 bp deletion mutation was detected in gastric cancer 39, lung cancer 40 and liver cancer cells 41.

Table 1 Complex I mutations
Table 2 Complex III, IV and V mutations
Table 3 D-loop mutations
Table 4 tRNA and rRNA mutations

The direct impact of several mtDNA mutations on tumorigenesis has been tested with the cybrid (cytoplasmic hybrid) system, where mtDNA is singled out for analysis 42. Cybrids carrying a pathogenic mutation at position 8993 or 9176 in the mtDNA ATP synthase subunit 6 gene (ATP6) derived from patients with mitochondrial encephalomyopathy were investigated for tumorigenesis in a nude mouse assay. It was found that the ATP6 mutations conferred an advantage in the early stage of tumor growth 43. In a separate study, the T8993G mutation was introduced into the PC3 prostate cancer cell line, and the resulting mutant cybrids were reported to generate tumors that were seven times larger than the wild-type cybrids 44. Further, as prostate cancer often metastasizes to bone, the above cybrids were co-inoculated in a nude mouse system with bone stromal cells 45. Growth acceleration in cybrids with mtDNA mutation was demonstrated in the bone microenvironment, and this effect was further shown to be likely mediated by upregulation of fibroblast growth factor 1 (FGF-1) and focal adhesion kinase (FAK) 45.

In another investigation, the contribution of mtDNA mutations to tumor cell metastasis was also analyzed 46. It was found that the mtDNA variant, which delivered the highest metastatic potential, contained G13997A and 13885insC mutations in the ND6 gene.

We recently examined the contribution of mtDNA mutations and mitochondrial dysfunction in tumorigenesis using human cell lines carrying a frame-shift mutation in the complex I subunit 5 gene (ND5); the same homoplasmic mutation was also previously identified in a human colorectal cancer cell line 12. With increasing mutant ND5 mtDNA content, respiratory function, including oxygen consumption and ATP generation through oxidative phosphorylation, declined progressively, whereas lactate production and dependence on glucose increased. Both heteroplasmic and homoplasmic mtDNA mutation caused an increased production of mitochondrial ROS. However in cells with heteroplasmic ND5 mutation, the cytosolic ROS level was somewhat reduced, probably due to the upregulation of antioxidant enzymes. As a result, only cells with homoplasmic ND5 mutation exhibited enhanced apoptotic potency. Furthermore, anchorage-dependence and tumor-forming capacity of cells carrying wild type and mutant mtDNA were tested by a growth assay in soft agar and subcutaneous implantation of the cells in nude mice. Surprisingly, the cell line carrying the heteroplasmic ND5 mtDNA mutation showed significantly enhanced tumor growth, whereas tumor formation was inhibited for cells with the homoplasmic form of the same mutation 47.

ROS generation and its role in tumorigenesis

ROS are a collective term, which includes superoxide, hydrogen peroxide and the hydroxyl free radical 48. The mitochondrial electron transport chain is a major source of ROS, as some of the electrons passing to molecular oxygen are instead leaked out of the chain. It has been estimated that generation of these partially reduced oxygen molecules comprises about 2-4% of the oxygen consumed 48. ROS are highly active and can cause damage to different cellular components including mtDNA 49, 50. The damaged mtDNA, if not repaired properly, produces mtDNA mutations, which, in turn, could initiate tumorigenesis and sustain cancer development.

In addition to their cytotoxic effects, low levels of ROS participate in the regulation of many cellular pathways 51, 52. The interaction of ROS with lipid species and thiol-containing proteins is important in cell growth and differentiation 53, 54. It was suggested that a threshold level of ROS (ROS window) is required for normal/cancer cell functions; above this level, cell death is activated and below it, proliferation is blocked 55, 56, 57. ROS have been shown to be involved in the transmission of survival and proliferation signals associated with tumor promotion and maintenance. For example, H2O2 has been demonstrated to activate the receptor tyrosine kinase 58, 59, Ras-mitogen-activated protein kinase (Ras-MAPK) 60, 61 and phosphatidylinositol 3′ -kinase (PI3K) pathways 58. ROS also mediate the stress signaling pathways involving nuclear factor-kappa B (NF-κB) 62 and the c-Jun NH2-terminal kinase (JNK) 63. The window hypothesis has also been supported by the observation that removal of H2O2 from the cellular environment by catalase blocks cell proliferation via down-regulation of MAPK activity 64. Similarly, ROS are also capable of preventing caspase activation, as in the case of protection of stimulated neutrophils from the toxic effects of oxidative stress 65.

The association of oxidative stress with tumorigenesis has been implicated in the induction of skin cancer by ultraviolet radiation, leukemia by γ-radiation and others, including lung cancer, by smoking. The role of ROS in tumor development has been supported by the demonstration that normal cells exposed to ROS show increased proliferation 54 and expression of growth-related genes 66, 67, 68. Furthermore, a large number of cancer cells are known to produce more ROS than non-cancer cells 69, 70. These observations suggest that ROS stimuli may contribute to cancer initiation, maintenance and development in vivo.

Rapid cell proliferation in cancer results in a surge of oxygen consumption and thus, tumor tissues suffer from hypoxia. The transcription factor, HIF-1 (Hypoxia inducible factor-1), is the key mediator of the hypoxia response through regulating genes involved in metabolism, angiogenesis, cell cycle and apoptosis 71. Transcriptional activation of genes, such as vascular endothelial growth factor and glucose transporter, by HIF1 is among the best-understood examples of regulation of angiogenesis and metabolism during the adaptation to hypoxic conditions 72.

In addition to upregulating the glycolytic pathway, HIF-1 was also shown to inhibit mitochondrial biogenesis and respiration in a renal cell carcinoma model by repression of C-MYC activity 73. Importantly, C-MYC was required for the expression of coactivator, PGC-1β, which is a key regulator of mitochondrial biogenesis 74. Alternatively, HIF-1 downregulates oxidative phosphorylation through activation of pyruvate dehydrogenase kinase 1 (PDK1) 75, 76. PDK1 inactivates the TCA cycle enzyme, pyruvate dehydrogenase, which converts pyruvate to acetyl-CoA. Interestingly, it was also demonstrated that HIF-1 could modulate respiration efficiency in hypoxic cells by regulating complex IV subunit 4 isoforms. Such regulation has important implications in ATP production, oxygen consumption and ROS generation 77.

Emerging evidence has indicated the important role of mitochondrial ROS generation during hypoxic activation of HIF 78. Further, expression of the nuclear-transcribed, mitochondrial-targeted ND2 mutants resulted in enhanced tumor growth, which was accompanied by increased ROS production and HIF-1α induction. These phenotypes were reversible by a complex III inhibitor, ascorbate 79.

Apoptosis, another link between mitochondrial dysfunction and tumorigenesis

Apoptosis is a process whereby a series of proteases, called caspases, are activated through a complex signaling cascade leading to energy-dependent cell death 80. Defects in apoptosis are among the major causes of tumorigenesis 81. Mitochondria play an important role in regulating apoptosis 82. A recent study investigated the effects of mitochondrial respiratory chain modulation on apoptosis 83. It was reported that defects in the respiratory chain could either promote or inhibit cell death, depending on the specific alteration in electron flow 83. The initiation of apoptosis can also occur in the mitochondria through stimulated ROS production. Low levels of ATP and high levels of cytosolic calcium, are usually associated with mitochondrial defects and reported as signals to induce apoptosis 84. Interestingly, in some cases, ROS mediate both pro- and anti-apoptotic effects, depending on their concentrations 85.

Among studies of cell death resistance due to mitochondrial dysfunction in cancer cells, it was reported that mitochondrial respiration defects led to activation of the Akt survival pathway. As mentioned earlier, this up-regulation of Akt was suggested to result from an increase in NADH, the substrate of respiratory complex I, which then inactivates PTEN through a redox modification mechanism 23. In another study, modulation of mitochondrial function by up-regulation of mitochondrial chaperones has been implicated in the survival of cancer cells 86, 87. Heat shock protein 90 and its mitochondrial-related molecule, TRAP-1, were suggested to interact with cyclophilin D to inhibit cell death 87, whereas Hsp60 was shown to orchestrate a broad cell survival program centered on stabilization of mitochondria to restrain p53 function 86.

Interestingly, it was also reported that the molecular mechanism through which ATP6 mutations at positions 8993 and 9176 promote tumorigenesis is by preventing apoptosis 43, although the details of such a signaling pathway remain unclear.

Retrograde regulation and other mitochondrial signaling mechanisms in cancer cells

Retrograde regulation is a communication pathway from the mitochondria to the nucleus that is used to describe the cellular response to the changes in the functional state of mitochondria 88. The first evidence of altered nuclear gene expression in response to mitochondrial dysfunction in mammalian cells came from studies showing increased mRNA levels coding for various mitochondrial proteins in several types of mtDNA-less (ρ0) cells 89, 90. One of the mechanisms suggested to play a role in the retrograde response was mitochondrial stress, which is supported by changes in mitochondrial membrane potential and elevation of calcium levels 88. Using ρ0 human osteosarcoma 143B cells and cybrid cell lines carrying mutated mitochondrial tRNAs, it was shown that respiratory deficiency and the associated calcium increase induced the activation of CaMKIV (calcium/calmodulin kinase IV). The activation of CaMKIV in turn activated CREB (cAMP-responsive element-binding protein) and Egr1 (early growth response gene-1) through PKC-mediated phosphorylation 22, 91. In a recent report, such mitochondrial-nuclear communication was further divided into two different pathways: one caused by a reduction in respiration and another, named intergenomic signaling, which requires mtDNA 92. Using DNA microarrays in the budding yeast Saccharomyces cerevisiae, it was shown that intergenomic signaling functions in coordinating mitochondrial and nuclear gene expression.

In a Drosophila system, it was shown that mitochondrial dysfunction activated the production of both AMP and ROS, the former stimulating AMPK and p53, and causing the loss of cyclin E, and the latter turning on JNK, FOXO and other G1-S cell cycle checkpoint molecules. These findings demonstrate mitochondrial retrograde regulation of cell cycle progression via AMP and ROS at sublethal concentrations through independent signaling molecules 93.

In tumor cells, retrograde signaling has also been demonstrated as a pathway that links mitochondrial dysfunction to oncogenic events. In paraganglioma, mutations in the mitochondrial tumor suppressor, succinate dehydrogenase (SDH) result in the accumulation of succinate, which was shown to inhibit HIF-α prolyl hydroxylases, leading to the stabilization and activation of HIF-1α 94. Thus, succinate was suggested as a retrograde linkage between abnormal mitochondrial metabolism and oncogenesis. In an early study, cytoplasts (cells depleted of nuclei) from tumor cells were shown to transfer tumorigenic properties when fused with nuclei from normal cells, indicating that cytoplasmic factors can induce malignant phenotypes 95. Depletion and partial depletion of mtDNA by a mitochondrial-specific ionophore, carbonyl cyanide m-chlorophenyl hydrazone, induced increasingly invasive behavior in C2C12 rhabdomyoblasts and A549 human lung carcinoma cell 96, 97, 98. In such cases, a number of genes involved in Ca2+ response, glucose metabolism, oncogenesis and apoptosis were upregulated 99. Further, in such systems, it was shown that calcineurin-mediated activation of the insulin-like growth facor-1 (IGF-1) receptor pathway and metabolic shift to the glycolytic pathway, provided a survival advantage to cells under mitochondrial stress caused by mtDNA depletion 100. Recently, it has been shown that cells treated with a human carcinogen, dioxin, displayed resistance to apoptosis, increased expression of the tumor marker, cathepsin L, and a high degree of invasiveness, which are linked to the triggering of a signaling pathway that promotes tumor progression in vivo through directly targeting mitochondrial transcription and induction of mitochondrial stress signaling 101.

Retrograde signaling induces the expression of a number of tumor-specific marker genes, such as extracellular matrix protease, TGF-β and epiregulin, as well as other genes that control cell growth and proliferation, such as PKC, JNK/MAPK, CREB and NF-κB 88. A proteomics approach was also used to gain insight into the nuclear gene targets of mitochondrial stress signaling 102, 103. In one study, the potential role for one of the identified retrograde response proteins, UQCRC1 (encoding complex III subunit core protein 1), was analyzed, and it was found to be highly expressed in breast and ovarian tumors 102.

Mitochondrial defects and genome instability

Since mitochondria are the major source of cellular ATP production, it is likely that mitochondrial dysfunction leads to a reduction in ATP levels that may affect the ATP-dependent pathways involved in transcription, DNA replication, DNA repair and DNA recombination. Mitochondria are also involved in the biosynthesis of deoxyribose nucleoside triphosphates (dNTP) 104, 105. Taken together, it is conceivable that mitochondrial deficiency could lead to mutagenesis in the nuclear genome. In yeast, it was reported that mitochondrial dysfunction caused by respiration inhibition, mtDNA depletion or mtDNA deletion resulted in a twofold to threefold increase in the nuclear DNA mutation frequency 106. In human cell lines depleted of mtDNA, it was reported that dNTP pools were affected, and in particular, a threefold reduction in dTTP pools was detected 107. Since imbalanced dNTP pools had been shown to be mutagenic 108, a molecular mechanism linking mitochondrial dysfunction to nuclear genome instability was proposed 107. Interestingly, disruption of mitochondrial function in mouse zygotes led to telomere attrition, telomere loss, and chromosome fusion and breakage, mediated by alterations in ROS production 109.

Significance of investigation of mtDNA mutations in cancer

Despite tremendous progress in identifying and characterizing nuclear oncogenes, tumor suppressor genes and their roles in cancer development, there are still many aspects of tumorigenesis that cannot be explained. The role of mitochondria, specifically mtDNA mutations, remains largely unclear. Although evidence suggests that some mtDNA mutations do play a role in certain stages of cancer development, there are still multiple potential pitfalls in such investigations 110. Special caution and general guidelines should be followed in this very important yet complicated line of research 111.

Based on our recent results 47 and studies from other labs, we propose that mtDNA mutations could function in cancer development as follows: in the initial stage, cancer cells are very mutagenic either because of a carcinogenic insult or due to the compromised repair mechanism, and mtDNA is more likely to be mutated at this stage. Because of the replicative advantage of mutant mtDNA molecules, such as that previously described for mtDNA carrying the mutation associated with the mitochondrial encephalomyopathy, mtDNA mutations are enriched to a certain level of heteroplasmy which would enhance tumor progression due to either the elevated ROS generation, which in turn activates the oncogenic pathways, or the increase in genome instability, or both. However, after transformation, it may become more important to have a functional respiratory chain than an inhibited one to sustain rapid cell proliferation. In some cases, the mutant mtDNAs causing severe mitochondrial defects are selected against and diluted out; in other cases, residual mutant mtDNA might escape the selection. In late stages of cancer, the cells are progressively adapted to a glycolytic metabolism because of the hypoxic environment. This may lead to the selection of cells in which the mutations make them mitochondrial function-independent and, therefore, cells with homoplasmic mtDNA mutations may become predominant in such tumors. If this hypothesis is correct, the involvement of mtDNA mutations may, in fact, be much more prevalent in early stage cancers than originally thought.