Mitochondria have a multiprotein machinery, mitochondrial replisome, capable to promote mtDNA replication. All components of mitochondrial replisome in mammals are encoded by nuclear genes, translated on cytosolic ribosomes, and imported into mitochondria. Currently, only major players involved in mtDNA replication are known. They include the DNA polymerase γ holoenzyme, mtSSB, the replicative DNA helicase Twinkle, and the mitochondrial RNA polymerase (POLRMT) (Fig. 4.3) [51–55]. In vitro reconstitution experiments have demonstrated that these replicative proteins represent a minimal mitochondrial replisome capable of replicating the full length of the mtDNA molecule at a rate not significantly different than that found in vivo [56, 57].

The only DNA polymerase found in mitochondria is the DNA polymerase γ (POLγ) [58]. The POLγ holoenzyme is a 245-kDa heterotrimer composed of one catalytic subunit (POLγA also referred to as POLG1) and two accessory subunits (POLγB also called POLG2) [59, 60]. POLγA has three polymerase motifs mediating DNA synthesis, and a 3′  →  5′-exonuclease domain promoting proofreading during replication. POLγB functions as a processivity factor (i.e., it enhances the ability of an enzyme to repeatedly continue its catalytic function without uncoupling from its substrate); one POLγB subunit located close to POLγA enhances POLγA–DNA binding, while the more distantly located POLγB subunit accelerates polymerization rate [61–64]. POLγA levels and activity do not change significantly during development and aging [65]. Moreover, unlike TFAM, the DNA polymerase γ is stably expressed in the absence of mtDNA [66].

Human mtSSB is homologous to the Escherichia coli SSB, and close homologs have been identified in yeast, Drosophila, Xenopus laevis, and mouse [53, 67–70]. mtSSB stabilizes exposed single-stranded mtDNA regions and stimulates primer recognition and processivity of POLγ at the replicative fork, facilitating thereby mtDNA synthesis [68, 71, 72]. Moreover, human mtSSB stimulates Twinkle helicase activity [73]. New evidence suggests direct interactions between mtSSB and other mitochondrial replicative proteins at the replication fork. Lastly, its levels correlate with mtDNA content and are upregulated during early postnatal development of the heart [65].

The acronym Twinkle stands for T7 gp4-like protein with intramitochondrial nucleoid localization. The C-terminal region of Twinkle is homologous to the T7 phage primase/helicase protein gp4 essential for T7 DNA replication [54]. Twinkle exhibits a 5′  →  3′ DNA helicase activity, which is essential in vivo [73–75]. While Twinkle has homology to T7 gp4, human Twinkle lacks critical amino acids observed in related primases and does not have primase activity [54, 76]. Twinkle can form hexa- and heptamers; its linker region, connecting the N- and C-terminal regions, is responsible for this ability [77, 78].

In contrast to the gp4 protein and other replicative helicases, Twinkle prefers a dsDNA substrate with a 5′-loading site and a short 3′-tail to initiate DNA unwinding [73, 74, 79, 80]. However, the functional significance of dsDNA preference of Twinkle is unclear. Intriguingly, the Twinkle helicase motifs have similar organization as a main E. coli recombinase, RecA protein, and a role for Twinkle in recombination has been suggested. Consistent with this hypothesis, its overexpression has led to significant decrease in mtDNA recombination intermediates in human cardiomyocytes [81, 82].

Currently, 31 different mutations in the TWINKLE gene have been identified, resulting in depletion of mtDNA leading to mitochondrial dysfunction (Fig. 4.4) [83]. These TWINKLE mutations are associated with neuromuscular abnormalities. All dominant TWINKLE mutations are responsible for autosomal dominant progressive external ophthalmoplegia (adPEO) characterized by exercise intolerance and muscle weakness, and the ocular muscles are most severely affected [54, 84]. AdPEO patients accumulate deletions in mtDNA in postmitotic tissues [54].

Mitochondrial RNA Polymerase (POLRMT) responsible for mitochondrial transcription is a unique RNA polymerase distinct from the nuclear counterparts and structurally similar to the T7 phage RNA polymerase [55]. POLRMT is not only essential for transcription, but it also synthesizes the RNA primers to prime leading-strand mtDNA synthesis at the origin of heavy-strand replication (O_H) [57, 85, 86]. Transcription from the light-strand promoters (LSP) synthesizes genome length transcripts, which subsequently are processed to yield the individual mRNA and several tRNA species [23, 87]. In mammalian cells, H-strand DNA replication initiation events usually are terminated approximately 600 bp downstream of O_H. The terminated strand remains stably annealed to the parental DNA molecule generating a triplex structure, D-loop, with a displaced parental H-strand [88, 89].

The role of POLRMT as the leading-strand primase responsible for primer synthesis at O_H has been well established. However, the enzyme responsible for initiation of lagging-strand mtDNA synthesis has remained elusive. Emerging evidence suggests that POLRMT can also prime DNA synthesis at the origin of the light-strand replication (O_L) acting as a lagging-strand primase. According to the classical model, upon passage of the leading-strand replication machinery, single-stranded O_L is exposed and forms a stem-loop structure [90, 91]. mtSSB cannot bind to the stem-loop, while POLRMT can initiate RNA primer synthesis from a poly-dT repeat of the O_L stem-loop. After approximately 25-nt primer synthesis is completed, POLRMT is replaced by POLγ to initiate lagging-strand DNA synthesis. Importantly, this process can be reconstituted in vitro, and POLRMT depletion in human cells has resulted in delay in lagging-strand synthesis [57, 92].

Finally, a human ligase III gene has been reported to encode both nuclear and mitochondrial isoforms [93]. Although a role for the ligase III in mtDNA repair has been suggested, its role in mtDNA replication remains to be determined.

A few models of mitochondrial replication in mammals have been proposed, although the precise molecular mechanisms of mtDNA replication are a matter of open debate [83, 94].

According to the strand-displacement model (SDM), first proposed in 1972, leading and lagging strands are synthesized asymmetrically and unidirectionally from separate origins, O_H and O_L, respectively [95–98]. The mitochondrial replication fork initiates the synthesis from a POLRMT-synthesized primer and often arrests at the termination-associated sequence (TAS) generating a D-loop of approximately 650 bp in human cells [88, 89, 98]. When the leading-strand synthesis continues unidirectionally beyond the TAS, it leaves behind a single-stranded region of the parental H-strand. mtSSB binds to this ssDNA preventing nonspecific POLRMT-mediated lagging-strand priming. After traversing approximately 2/3 of the genome, the replication fork exposes the lagging-strand replication initiation site, O_L. Displaced parental H-strand adopts a stem-loop structure, and POLRMT initiates RNA primer synthesis from the poly-dT stretch of this region. POLγ replaces POLRMT and initiates the lagging-strand mtDNA synthesis (Fig. 4.5a).

Based on the finding of double-stranded replication intermediates generated by coupled leading- and lagging-strand synthesis, which were RNase H sensitive, two additional mtDNA replication models have been proposed [99–102]. In the ribonucleotide incorporation throughout the lagging strand (RITOLS) model, the leading strand (H-strand) is replicated unidirectionally from a discrete origin. The lagging-strand synthesis occurs with delay, like in the SDM; however, ssDNA covered by RNA instead of mtSSB generating theta-type, double-stranded RITOLS intermediates sensitive to RNase H (Fig. 4.5b). This RNA may be processed to generate multiple short RNA primers, which can be used to initiate discontinuous lagging-strand replication [102].

According to the strand-coupled model, in contrast to the RITOLS, mtDNA intermediates are fully dsDNA similar to nDNA replication intermediates; hence, this mechanism is also termed as conventional, strand-coupled Okazaki fragment associated (COSCOFA) replication [102]. In the COSCOFA, both leading- (H-strand) and lagging-strand (L-strand) syntheses initiate bidirectionally from multiple origins at a broad initiation zone (Ori-Z) downstream of O_H [103]. The lagging strand is synthesized without delay via most likely Okazaki fragment-associating replication. Both replication forks proceed till they arrest at O_H (Fig. 4.5b). Although the existence of Okazaki fragments during mtDNA replication has not yet been shown, mammalian mitochondria possess enzymatic machinery for their processing [104]. The role of mtSSB in both RITOLS and COSCOFA replications remains to be determined. Different mammalian tissues may exploit different mtDNA replication mechanisms: COSCOFA intermediates have predominantly been found in the heart, skeletal muscle, and in cells amplifying mtDNA after its depletion, while RITOLS intermediates have been found in liver, kidney, and cultured cells [78, 82, 102, 105].

The core human mitochondrial transcription machinery ­consists of mitochondrial RNA polymerase (POLRMT), an HMG-box transcription factor TFAM, and two mitochondrial transcription factors B1 and B2 (TFB1M and TFB2M, also known as mtTFB1 and mtTFB2, respectively) (Fig. 4.6) [106].

The human POLRMT gene cloned in 1997 encodes a 140-kDa single-subunit protein, containing two large regions characteristic to eukaryotic mtRNA polymerases [55]. The N-terminal part of the protein contains a 41-amino acid targeting sequence and two putative pentatricopeptide repeats (PPR) found in proteins involved in RNA processing in mitochondria and chloroplasts [107, 108]. An alternatively spliced POLRMT isoform missing the N-terminal 262 amino acids, including mitochondrial targeting signal, is located in the nucleus and represents the nuclear RNA polymerase IV; however, its function is not known [109].

The C-terminal portion of POLRMT contains several conserved sequence motifs essential for its catalytic activity and is homologous to the bacteriophage T3/T7 RNA ­polymerase [110]. However, unlike the bacteriophage T7 RNA polymerase, POLRMT is uncapable to initiate transcription on its own and requires assistance of TFAM and TFB1M/TFB2M [106, 111].

TFAM is a multifunctional protein involved in mtDNA packaging, replication, and transcription [38, 39, 112]. Human TFAM was the first mitochondrial transcription factor identified in mammalian cells [113]. TFAM is a member of the high-mobility group (HMG) family of proteins, which are able to bend, wrap, and unwind DNA [114, 115]. TFAM contains two HMG-box DNA-binding domains separated by a linker region and a short C-terminal tail, which is required for specific binding to the promoter region and for interaction with TFB1M/TFB2M and activation of transcription [116, 117]. Like other HMG proteins, TFAM can bind, bend, and unwind mtDNA in a sequence-nonspecific way; for nonspecific binding, its C-terminal part is not essential. However, mtDNA transcription is initiated by the sequence-specific binding of TFAM dimer upstream of the HSP and LSP [118, 119].

TFB1M and TFB2M have originally been identified as homologs of single yeast mitochondrial transcription factor mtTFB, essential for mtDNA transcription in this organism [111]. Like their yeast homolog, TFB1M and TFB2M can assemble into a complex with POLRMT and stimulate transcription in the presence of TFAM in vitro [106]. They also display sequence homology to the bacterial rRNA dimethyltransferases, which promote dimethylation of two adjacent adenine residues at the 3′ end of rRNA during ribosome biogenesis [106, 111]. Consistently, human TFB1M is capable to mediate this reaction in vitro [120]. rRNA dimethyltransferase activity of TFB2M is less efficient than that of TFB1M [121]. Mutations in conserved methyltransferase motifs that affect TFB1M/TFB2M rRNA dimethyltransferase activity do not change their ability to stimulate mtDNA transcription, suggesting that these functions are independent [117].

In the presence of TFB2M, TFB1M appears to be not essential for mtDNA transcription; however, its loss in mice results in embryonic lethality. Moreover, TFB1M defect in the heart has led to mitochondrial deficiency caused mainly by impaired mitochondrial translation [122]. Loss of TFB1M dimethyltransferase activity appears to be responsible for this defect, since dimethylation of 12S rRNA is essential for assembly of the ribosomal subunits [123]. Accordingly, TFB1M silencing has led to significant decrease in 12S rRNA levels, while its overexpression has resulted in 12S rRNA hypermethylation associated with impaired mitochondrial biogenesis and cell death [124].

In contrast, TFB2M displays significantly greater transcriptional stimulatory activity than TFB1M and plays predominant role in mtDNA transcription [106]. Consistently, overexpression of TFB2M but not TFB1M has led to elevated steady-state levels of mtDNA transcripts, while disruption of its rRNA dimethyltransferase domain has had no apparent effects [124, 125]. Thus, emerging evidence suggests that these proteins have diverged but partially overlapping functions: TFB1M predominantly regulates mitochondrial translation, while TFB2M predominantly stimulates mitochondrial transcription.

In human cells, the noncoding region (NCR), also known as D-loop (spanning nucleotides 16,024–16,576), contains the control elements for both mtDNA transcription and replication. Each strand contains the light-strand promoter (LSP), two heavy-strand promoters (HSP1 and HSP2), and the origin of replication for the heavy strand (O_H) (Fig. 4.7) [126, 127]. mtDNA transcription from the LSP and HSP2 generates long polycistronic precursor RNAs, which are subsequently processed by RNase P to produce mature mRNAs, rRNAs, and tRNAs [23, 87, 128, 129]. Transcription from the HSP1 generates a truncated transcript, which encodes 12S and 16S rRNA and two tRNAs [130]. Finally, as it has already been discussed, mtDNA transcription initiated by POLRMT from the LSP generates a primer for mtDNA replication, coupling thereby mitochondrial gene expression and mtDNA maintenance [131].

Transcription initiation is followed by the elongation step. The mitochondrial transcription elongation factor (TEFM) has recently been identified to interact with POLRMT enhancing its processivity. Consistent with an essential role of TEFM in mitochondrial transcription, its deficiency has led to reduced mtDNA transcript levels [132].

It is generally believed that the simultaneous presence of TFAM and the POLRMT–TFB2M heterodimer is required for promoter recognition and initiation of transcription. TFAM homodimer binds specifically to a region upstream of the HSP and LSP [119]. The C-terminal tail of TFAM, required for specific binding to the promoter region, also mediates direct protein–protein interaction with TFB2M, recruiting thereby POLRMT to the transcription initiation sites [57, 117]. In addition, TFAM due to its HMG-box DNA-binding domains mediates bending and unwinding of mtDNA helix. These structural DNA alterations expose the promoter region and may facilitate its recognition and initiation of transcription by the POLRMT–TFB2M complex.

Thus, the core machinery needed for mitochondrial transcription initiation is traditionally regarded as a three-component system composed of POLRMT, TFB2M, and TFAM [131, 133]. However, more recently, it has been demonstrated using a human recombinant mitochondrial transcription system that POLRMT and TFB2M can initiate transcription from the HSP1 and LSP even in the absence of TFAM, albeit with significantly lower efficiency [134]. Importantly, in vitro TFAM-independent transcription from the HSP1 has been equal or higher than from the LSP, while TFAM-dependent transcription has been higher from the LSP than from the HSP1. Moreover, additional cis-acting elements appear to contribute to transcription regulation. Based on these findings, the authors speculate that the amount of TFAM associated with nucleoid may differentially regulate transcription initiation and mtDNA replication [134].

The important role of TFAM in the coordination of mitochondrial transcription and replication has been confirmed using transgenic models. Loss of TFAM in mice results in mtDNA depletion leading to embryonic lethality [135]. Moderately elevated TFAM levels have resulted in elevated amounts of mtDNA transcripts, whereas its excessive overexpression has inhibited both mtDNA transcription and replication [136, 137]. It has recently been suggested that the nucleoid structure is a regulative factor for mitochondrial transcription: more dense mtDNA packaging mediated by excessive TFAM leads to lower accessibility of mtDNA to the transcription apparatus [138].

In addition to regulation of transcription initiation, mitochondrial transcription is also controlled at the termination step. Mitochondrial transcription initiated from the HSP2 generates a full-length RNA transcript, while transcription started from the HSP1 transcribes the 12S and 16S rRNA, tRNA^Val, and tRNA^Phe and terminates at the 3′ end of the tRNA^Leu(UUR) (Fig. 4.8) [139]. Specific binding of the mitochondrial transcription termination factor (mTERF1) to a 28-bp region at the 3′ end of the tRNA^Leu(UUR) promotes termination [140, 141].

Two other members belonging to the mTERF family, mTERF2 and mTERF3, have been identified [142–144]. Highly homologous mTERF proteins contain conserved leucine zipper-like domains, well-known mediators of both protein–protein interactions, and DNA binding. The three-dimensional structure of mTERF3 has recently revealed a novel DNA-binding domain: multiple α-helices separated by loops, which can adopt a half-doughnut shape forming upon binding to DNA a half clamp [145].

Ability of the mTERF proteins to bind to DNA has been confirmed experimentally. mTERF1 has a high affinity to the tRNA^Leu(UUR) region and moderate affinity to the HSP1 and other regions in the D-loop [105]. mTERF2 and mTERF3 also display preferential binding to the HSP [146, 147]. Moreover, mTERF2 has been identified in isolated nucleoid preparations [148].

Importantly, mTERF1 appears to bind simultaneously to both the termination region and initiation HSP1 site, resulting in the rDNA looping [130]. Two other mTERF members may also interact with the initiation site (Fig. 4.8) [146]. However, it is currently unclear whether all three mTERF proteins assemble in a complex to bind to the HSP or mTERF1 forms a dimer with either mTERF2 or mTERF3 to promote mtDNA transcription. Depletion of mTERF2 has led to decrease in mitochondrial mRNA levels, while mTERF3 depletion has resulted in stimulation of mitochondrial transcription [146, 147]. Based on these findings, it has been suggested that mTERF2 plays the role of transcription activator, while mTERF3 acts as a transcription repressor. In addition, human mTERF1 contributes to mtDNA replication modulating replication pausing, highlighting further the coordination of mtDNA transcription and replication [105].

A termination site (H2 termination) of the long H2 transcript initiated from the HSP2 has been less characterized. A 22-bp A/T-rich region appears to represent the H2 termination region. ATAD3 and the leucine-rich pentatricopeptide-repeat containing protein (LRPPRC, also known as LPR130) can bind this region contributing to the H2 termination [149]. LRPPRC is essential for OXPHOS components encoded by mtDNA, and its deficiency results in reduced oxygen consumption associated with decrease in mRNA and tRNA levels [149–151]. Importantly, mutated LRPPRC has been found in Leigh syndrome French-Canadian patients displaying cytochrome oxidase deficiency [152].

All ribosomes are composed of two subunits: the small subunit promotes the decoding of mRNAs, whereas the large subunit contains the peptidyl transferase activity mediating peptide bond formation. The forming polypeptide is fed into the polypeptide tunnel, and the newly synthesized polypeptide leaves the ribosome from the polypeptide tunnel exit to be exposed to the hydrophilic milieu. The tunnel exit of ribosomes represents important regulatory site where various factors interact with the ribosome to govern the newly generated proteins [160]. However, in contrast to bacterial and yeast system, very little is currently known about human mitochondrial ribosomal tunnel exit.

The human mitochondrial ribosomes have a lower sedimentation coefficient, i.e., 55S, compared to bacterial 70S and cytosolic 80S counterparts [161]. However, they are larger than bacterial ribosomes based on their particle mass and physical dimensions [155, 162, 163]. The human mitochondrial ribosomes contain almost half the RNA found in bacterial ribosome, with a concomitant increase in the number of ribosomal proteins accounting for a significantly lower rRNA/protein ratio. The small 28S subunit of the human mitochondrial ribosome contains a 12S rRNA and over 30 proteins, whereas the large 39S subunit consists of a 16S rRNA and approximately 48 proteins [164]. These differences result in more porous and open structure with the absence of conventional E-site in mitochondrial ribosomes [161].

In contrast to the mitochondrial rRNA encoded by mtDNA, the mitochondrial ribosomal proteins (MRPs) are the products of nuclear genes [165, 166]. Gene map loci have been reported for the majority of human MRPs [167, 168]. The mapped genes are widely dispersed throughout the nuclear genome, suggesting that they were not transferred all together. Importantly, some of these genes overlap with chromosomal loci associated with various diseases, such as deafness, retinitis pigmentosa, and Usher syndrome [167].

Analysis of amino acid sequences of the majority of human MRPs revealed that most of them are highly homologous to the bacterial ribosomal proteins; although some homologues in small and large subunit are missing accounting for a net mass loss of over 200 kDa, this loss is highly compensated by the acquisition of more than 30 new proteins accounting for a net mass excess of about 870 kDa [161, 166]. Some of these new proteins appear to be multifunctional with extraribosomal functions. One of them, MRPS29 (known also as DAP3), is a GTP-binding protein located in the small subunit; together with MRPS30 (known also as PDCD9), it has previously been implicated in apoptosis [169–171]. Moreover, MRPS29 may mediate the 28S subunit binding to the MIM. Whether their roles for the mitochondrial ribosome are related to their functions in apoptosis remain to be determined. The precise locations of the majority of new MRPs are also unknown.

In contrast to prokaryotic or eukaryotic cytosolic systems, the small 28S subunit is able to bind to mRNA tightly in the absence of initiating factors or tRNA. At least 400 nucleotides of the messenger template are minimally required for efficient binding to the ribosomal subunit. This may explain why the ATPase8 and ND4L transcripts, which are <300 nt, have to be part of a larger message [172].

Another important difference between mitochondrial and cytosolic translation is the former is initiated by a much reduced initiation complex, which appears to be composed by two initiation factors, mtIF2 and mtIF3 [173–175]. Mammalian mtIF2 belongs to a family of GTPases and exhibits strong structural and functional homology with E. coli IF-2. mtIF2 promotes fMET-tRNA binding to the 28S subunit in the presence of mRNA and GTP and appears functionally replaces both bacterial initiation factors, IF1 and IF2 [154]. Mitochondrial elongation factors, mtEF-Ts, mtEF-Tu, and mtEF-G1, sharing homology with the corresponding bacterial factors, mediate translation elongation in human mitochondrial ribosomes [176–179].

Similar to bacterial protein synthesis, protein synthesis mediated by mammalian mitochondrial ribosomes is sensitive to chloramphenicol but insensitive to cycloheximide. Chloramphenicol has been widely used to selectively inhibit mitochondrial protein synthesis without affecting the cytoplasmic protein synthesis machinery; conversely, cycloheximide is used to inhibit cytoplasmic protein synthesis with little effect on mitochondrial protein synthesis. The binding site for chloramphenicol can be modified by a mutation in the rRNA gene, resulting in chloramphenicol-resistant (CAP_R) mutants [180].

The standard stop codon UGA encodes Trp, whereas AGA and AGG serve as stop signals for mitochondrial transcripts MTCO1 and MTND6, the remaining 11 open reading frames terminate at UAA or UAG. Bacterial translation system utilizes two release factors (RFs) to terminate transcription: RF1 recognizes UAA/UAG, while RF2 recognizes UGA/UAA to mediate peptidyl-tRNA hydrolysis at corresponding stop codons [181, 182]. Two candidates for the role of human mtRF, related to bacterial RF1, have been identified. While biochemical analysis of mtRF1 has failed to reveal release activity with any codons, mtRF1a recognizes UAA/UAG, but is unable to recognize AGA/AGG [183, 184]. Given that AGA/AGG is directly preceded by a “U” nucleotide, it has recently been suggested that human mitochondria can utilize a −1 frameshift, which converts AGA/AGG in the standard UAA/UAG stop codons. Therefore, only single RF, such as mtRF1a, is needed for translation termination [185].

Finally, recycling of the human mitochondrial ribosome following translation termination is mediated by the mitochondrial ribosome recycling factor (mtRRF) acting in ensemble with mtIF3 and mtEF-G2 [186–189]. Interestingly, human mtRRF contains an N-terminal presequence for its targeting to mitochondria, which is not cleaved after its import to the organelle [186]. However, role of the mtRRF N-terminal region remains to be determined.

mtDNA, similar to its nuclear counterpart, is a target for various damaging agents, including ionizing radiation, chemicals, environmental toxins, and pharmacological drugs. A negative charge on the inner surface of the MIM, generated by Δψ_m, results in tremendous accumulation and concentration (up to 1,000-fold) of lipophilic cations in mitochondria [190]. Many toxic chemicals represent positively charged lipophilic substances and therefore tend to accumulate in mitochondria affecting their function. Another factor contributing to mtDNA vulnerability is the lack of histones able to protect DNA. These factors explain significantly higher (tenfold or more) damaging of mtDNA by alkylating agents compared to that of nDNA [191].

mtDNA is particularly susceptible to oxidative damage caused by ROS generated during OXPHOS [192]. Importantly, in human cells, mtDNA damage is not only more extensive but also persists longer following OS than nDNA damage [193]. Various lesions, including base modifications, abasic sites, and single-strand breaks, are generated upon oxidative DNA damage [194]. Among them 8-oxoguanine (8-oxoG) is one of the most studied DNA lesions. 8-oxoG is a mutagenic lesion: its mispairing with adenine leads to a G–C to T–A transversion upon the subsequent round of replication. It has been demonstrated accumulation of 8-oxoG with age in both nDNA and mtDNA and to a greater extent in the latter [192, 195].

Despite the high vulnerability of mtDNA, it has been initially assumed that DNA repair is absent or very inefficient in mitochondria. Damaged mtDNA has been believed to be degraded, while undamaged copies have been used as templates for mtDNA synthesis. However, more recently, it has been shown the efficient repair of certain types of DNA lesions in mitochondria [196–199]. mtDNA repair mechanisms, including base excision repair (BER), mismatch repair, and recombinational repair, have been extensively investigated in Saccharomyces cerevisiae, and multiple yeast proteins implicated in these pathways have been identified [200–203]. Several mammalian mitochondrial proteins, which may contribute to various DNA repair mechanisms, have subsequently been identified; the growing protein list is summarized in Table 4.2 [198, 199, 204].

BER, the major mtDNA repair mechanism, targets oxidatively modified bases, including 8-oxoG and thymine glycol. A number of enzymatic activities promoting this repair pathway have been identified in human mitochondria (Table 4.2). It is initiated with recognition followed by removal of a damaged or inappropriate base by damage-specific DNA glycosylases, including uracil DNA glycosylase (UDG), 8-oxoG DNA glycosylase/AP lyase (OGG1, a homologue of E. coli MutM), thymine glycol glycosylase (NTH), and MYH glycosylase (homologue of E. coli MutY glycosylase). These enzymes cleave the N-glycosylic bond between the base and the sugar leaving an abasic site. This abasic site is recognized by apurinic/apyrimidinic nucleases, such as APE1 or APE2, which generate a 3′-OH and 5′-phosphate DNA break. POLγ extends the formed 3′-OH moiety, and DNA ligase, such as LIG3, joins DNA ends completing repair. In contrast to the short-patch and long-patch BER present in the nucleus, only the short-patch pathway, replacing a single damaged base, has been found in mitochondria [205]. Moreover, it has been demonstrated that mitochondrial proteins involved in BER are not freely soluble in the mitochondrial matrix, but are rather associated with the MIM [206].

In addition to BER, it has been reported that rat mitochondrial lysates can promote removal of GT and GG mismatches [207]. Moreover, proteomic analysis of mouse mitochondria has revealed the presence of Mlh1 and Msh5, mammalian homologues of E. coli mismatch repair proteins MutL and MutS, respectively [10]. These data suggest that mitochondria possess mismatch repair capacity; however, the role of this pathway in mtDNA maintenance remains to be determined.

Mitochondrial BER pathway can efficiently remove damaged bases; however, DNA double-strand breaks and cross-links can be repaired via homologous recombination and/or nonhomologous end joining mechanisms. Although emerging evidence suggests that human mitochondria possess homologous recombination activity, which has further been confirmed by proteomic identification of several proteins involved in this pathway, the extent and importance of mtDNA recombination repair remain a matter of open debate [199, 208–211]. Interestingly, in adult human heart, mtDNA is organized in complex catenated networks containing abundant recombination intermediates, Holliday junctions [82]. Similar mtDNA organization with multiple recombination junctions has been found in human and mouse brain but not in other tissues. Moreover, cardiac mtDNA recombination intermediates appear to correlate with mtDNA copy number [212]. Overexpression of TFAM and Twinkle in mice has led to elevated mtDNA copy number associated with a significant increase in recombination intermediates in the heart, brain, and skeletal muscles [82]. Based on these recent findings, it has been hypothesized that recombination mechanisms may play particularly an active role in mtDNA maintenance in the heart. In the myocardium, ROS originating from the highly active OXPHOS represent an abundant source of mtDNA damage [213]. Oxidatively modified nucleotides can be repaired via BER mechanisms; however, they can also lead to double-strand breaks, which require recombinational repair [214]. In addition to a protective role against oxidative mtDNA damage, ROS-induced recombination-dependent mtDNA replication associated with accelerated mitochondrial biogenesis may represent a myocardial response to increased energy demand.

Modulation of mtDNA repair has been more readily addressed in yeast, whereas in humans, such studies are very limited. Targeting of overexpressed OGG1, a glycosylase that removes 8-oxoG, in HeLa mitochondria has significantly enhanced the resistance to OS [215–217]. However, Ogg1-deficient mice, although accumulate 10- to 20-fold higher 8-oxoG levels in mtDNA, display no mitochondrial dysfunction and develop and age normally [218–220]. Similarly, UDG knockout mice deficient in both mitochondrial and nuclear activities show no apparent mitochondrial defects under normal conditions, although they have increased postischemic brain injury [221]. Consistently, inhibition of S. cerevisiae UDG has not resulted in increased mutagenesis of mtDNA [222].

Growing evidence suggests a role for mtDNA repair in the pathogenesis of neurodegenerative diseases and carcinogenesis; however, the exact contribution of mtDNA repair pathways to these pathologies remains to be determined [204]. Lastly, age-dependent accumulation of oxidative damage in mtDNA in cardiomyocytes leading to various mutations and the role of mtDNA repair mechanisms in this process will be discussed in Chap. 13.