Mitochondrial DNA replication is critical for maintaining mtDNA copy number to generate sufficient cellular energy that is required for development and for functional cells. In early development, mtDNA copy number is strictly regulated at different stages, and, as a result, the establishment of the mtDNA set point is required for sequential cell lineage commitment. The failure to establish the mtDNA set point results in incomplete differentiation or embryonic arrest. The regulation of mtDNA copy number during differentiation is closely associated with cellular gene expression, especially with the pluripotency network, and DNA methylation profiles. The findings from cancer research highlight the relationship between mitochondrial function, mtDNA copy number and DNA methylation in regulating differentiation. DNA methylation at exon 2 of DNA polymerase gamma subunit A (POLGA) has been shown to be a key factor, which can be modulated to change the mtDNA copy number and cell fate of differentiating and tumour cells. The present review combines multi-disciplinary data from mitochondria, development, epigenetics and tumorigenesis, which could provide novel insights for further research, especially for developmental disorders and cancers.
Mitochondria are double-membraned organelles primarily responsible for the generation of ATP as well as several other vital cellular functions. Each mitochondrion hosts multiple copies of the mitochondrial genome (mtDNA), which is a circular double-stranded genome encoding key subunits of the electron transfer chain (ETC). There are programmed changes to mtDNA copy number during cellular development and tumorigenesis, which are directly regulated by a group of nuclear-encoded regulators. It has also been demonstrated that changes to mtDNA copy number are associated with the expression of key regulators of cell fate and the DNA methylation status at exon 2 of the catalytic subunit of DNA polymerase gamma (POLGA). These processes are strictly regulated to establish the mtDNA set point, which provides differentiating cells with the potential to increase mtDNA copy number in a cell-specific manner.
Mitochondria and OXPHOS
The mitochondrion is a double-membraned organelle present in nearly all mammalian cells. Mitochondria are associated with a variety of cellular and metabolic activities such as cell signalling, ion homoeostasis, cell differentiation and apoptosis [1,2]. They are primarily described as the powerhouses of cellular metabolism because they are the major generators of cellular energy (ATP) through the aerobic process of oxidative phosphorylation (OXPHOS) [3,4] (Figure 1).
The ETC is located on the inner membrane of the mitochondrion. It comprises five enzyme complexes (Figure 1): Complex I (NADH dehydrogenase; ND), Complex II (succinate dehydrogenase; SD), Complex III (cytochrome c reductase; CYTB), Complex IV (cytochrome c oxidase; COX) and Complex V (ATP synthase). Additionally, coenzyme Q (CoQ) and cytochrome c (CytC) are two electron carriers that are extensively involved . Briefly, Complex I and Complex II accept electrons from the electron donors and pass them on to CoQ. The electrons are then transferred through Complex III, CytC and Complex IV and finally accepted by O2. During this process, a pH gradient across the inner membrane forms due to the protons being pumped out by the complexes. Potential energy stored in the pH gradient is then transformed into biochemical energy, in the form of ATP, through the process of transferring the protons back into the mitochondrial matrix through Complex V . It is noted that the process described above constitutes aerobic respiration. This is a far more efficient process than anaerobic respiration, as it produces 32–36 molecules of ATP per molecule of glucose compared with two molecules of ATP from glycolysis .
The mitochondrial genome
The mitochondrion hosts a genome, mtDNA, whose replication and expression apparatuses are distinct from those of the nuclear genome. mtDNA is a circular double-stranded DNA molecule that is 16.6 kb in size in humans (Figure 2). It consists of a heavy (H) and a light (L) strand. The mitochondrial genome encodes 37 genes which include 22 tRNAs, two rRNAs (12S and 16S rRNA) and 13 polypeptides of the ETC complex enzymes  (Figure 2). These include genes encoding seven subunits of Complex I (ND1–ND6 and ND4L), one Complex III subunit (CYTB), three Complex IV subunits (COX I, II and III) and two Complex V subunits (ATPase6 and ATPase8) (Figure 2). As for the other subunits of ETC complexes, they are nuclear DNA (nDNA)-encoded proteins that are imported into the mitochondrial matrix. Therefore, efficient and functional OXPHOS requires a well-established synergy between the nucleus and the mitochondria to meet the demands of a specialized cell for the production of energy.
The human mitochondrial genome.
The mitochondrial genome is highly compact with only two non-coding regions and no introns. The main non-coding region contains the displacement loop region (D-loop). It also consists of the origin of replication for the H strand (OH), and the promoters of transcription for both the H and L strands (HSP1, HSP2 and LSP) (Figure 2). The origin of replication for the L strand is located in the other non-coding region that is located two-thirds of the genome downstream of OH, between the genes tRNA-Cys and tRNA-Asn . Genes encoding tRNAs are interspersed between genes encoding rRNAs and the ETC subunits, especially on the L strand . Also, the termination codons are not encoded in some of the mitochondrial genes, but are created post-transcriptionally by the poly(A) tail mechanism , which means that the initial transcript normally ends with either a U or a UA, and polyadenylation can add adenines after these transcripts to form a UAA termination codon.
Another key feature that distinguishes the mitochondrial genome from the nuclear genome is that there are multiple copies of the mitochondrial genome per mitochondrion, and also per cell, as each cell also possesses multiple mitochondria . Indeed, different cell types have different levels of mtDNA copy number depending on their demands for cellular energy . mtDNA copy number could be as high as several thousand in cells such as muscle and neurons, whereas it is much lower in fibroblasts, spleen cells and stem cells . Moreover, the mitochondrial genome can be present in a heteroplasmic manner, which means that both wild-type and mutated forms can coexist in the same cell . It is thought that mtDNA mutations are a constant phenomenon mostly due to the primary targeting of reactive oxygen species (ROS). This is despite the mtDNA-specific replicase DNA polymerase gamma (POLG) having one of the lowest error rates of the eukaryotic polymerases for its exonucleolytic proofreading functions . There is also the presence of diverse repair systems in the mitochondrion including recombinational repair . Nevertheless, the mutations remain mostly silent or have no effect on phenotype, as it has been demonstrated that a mutation level of over 70% is normally required to lead to the onset of one of the many mitochondrial diseases .
In mitochondria, mtDNA molecules are organized into DNA–protein complexes similar to nucleoid structures in prokaryotes which supports the theory that the mitochondrion evolved from a proteobacterial partner that invaded eukaryotes . The mitochondrial nucleoid consists of more than 20 binding proteins, which comprise many of the factors responsible for the transcription and replication of mtDNA. Owing to the limited coding capacity of mtDNA, these factors are nuclear-encoded and imported into the mitochondrial matrix.
mtDNA copy number is strictly controlled by its transcription and replication, which take place in mitochondria. mtDNA transcription and replication are driven by a group of nDNA-encoded mitochondrial specific factors (Figure 1), which includes a group of factors that bind directly with mtDNA, such as POLG and mitochondrial transcription factor A (TFAM), and a group of indirect factors, which regulate the gene expression of those direct factors. The best characterized of the indirect factors are the nuclear respiratory factors (NRF1 and NRF2) and the peroxisome-proliferator-activated receptor γ (PPARγ) co-activator 1α (PGC1α) family (Figure 1). It is noted that the regulation of mtDNA transcription and replication requires the coordinated control of all the factors and is at a complex level. As the related data have been well summarized and reviewed by others [13–18], we are not covering this topic here. However, it is important to point out that the regulation of mtDNA copy number involves numerous factors, which have cell- and developmental-stage-specific expression profiles and are associated with other cellular functions.
The mtDNA set point
In order to meet the very different energy demands of a variety of cell types, mtDNA copy number is strictly regulated during development. This is, perhaps, best exemplified in embryonic stem cells (ESCs), which acquire mtDNA copy number in a cell-specific manner during differentiation through the strict regulation of mtDNA replication. This equips mature cells with the appropriate number of mtDNA copies in order that they are able to perform their specialized functions [19–21] (Figure 3).
Very early on in development, the primordial germ cells, which are the first indicative germ cells, are laid down and they possess approximately 200 copies of mtDNA per cell [22,23]. As these cells mature during the extensive process of oogenesis to form fertilizable oocytes, namely metaphase II oocytes, mtDNA copy number increases significantly to more than 180,000 copies per cell. Indeed, there is a ‘critical threshold’ of mtDNA copy number to enable the metaphase II oocyte to proceed through fertilization [24–27] (Figure 3). Failure to surpass this threshold results in either unsuccessful fertilization or embryos arresting during pre-implantation development. However, those oocytes that mature with insufficient mtDNA copy number, namely mtDNA-deficient oocytes, can be rescued through supplementation with exogenous populations of mtDNA . This improves fertilization rates and successful development to the blastocyst stage , the final stage of pre-implantation development when the embryo first differentiates into two cellular compartments [24,28]. These cellular compartments comprise the inner cell mass (ICM), which gives rise to the embryo proper and then the fetus, and the trophectoderm, which forms the placenta. Indeed, mitochondrial supplementation is being introduced clinically to promote embryo development, especially in older patients when their embryos either fail to fertilize or they arrest very early during development .
Synchronous changes to mtDNA copy number and DNA methylation during early development.
Once fertilization has taken place, mtDNA copy number peaks between fertilization and activation of the embryonic genome [25,31,32], a transcriptional event that enables the embryo to initiate transcription of its own gene products, which means that it is no longer dependent on products carried over from the oocyte. The timing of embryonic genome activation is species-specific, but usually occurs between the two- and eight-cell stages [31,33]. Nevertheless, this mtDNA replication event is usually completed by the two-cell embryo stage. Subsequent mtDNA replication is restricted until the blastocyst stage [25,28,34]. As cell division progresses in an exponential manner in the early embryo, mtDNA copy number is diluted out significantly up to the blastocyst stage. Indeed, this appears to be a key developmental process as it is now evident from studies in human embryos cultured in vitro that they actively secrete, or shed, mtDNA into the surrounding culture medium. To this extent, it has been hypothesized that this is an indicator of better quality embryo development, as it enhances development to the blastocyst stage and promotes implantation . This outcome is supported by the knowledge that too many copies of mtDNA at the blastocyst stage can lead to aneuploidy and failure of the hatched blastocyst to implant .
Pluripotency defines a naïve state that provides the cell with the potential to differentiate into all cell types of the body . Pluripotency is regulated by gene markers that include OCT4, SOX2 and NANOG, which collectively form the pluripotent network, and are expressed in the ICM and undifferentiated embryonic stem cells [38,39]. Indeed, failure to express the markers of the pluripotent network results in loss of pluripotent potential and leads to embryonic death or differentiation [39–41]. Consequently, the trophectodermal cells are the first cells to have undergone differentiation, and they lose their pluripotent potential and express lineage specific markers, such as CDX2 and GATA3 [42–44]. In line with the onset of differentiation, the trophectodermal cells initiate mtDNA replication [25,28,34] (Figure 3). On the other hand, the ICM cells retain low mtDNA copy number and restrict mtDNA replication so that copy number is further reduced with each cell division [19,44,45] (Figure 3).
As the ICM consists of embryonic cells that are pluripotent, the continual dilution of mtDNA copy number contributes to the establishment of the mtDNA set point, which is required to establish lineage commitment . The mtDNA set point is characterized by low mtDNA copy number. This can range from as few as 20 to 200 copies, which is a similar state as that of the primordial germ cells [22,23]. This provides the template for subsequent differentiation events in order that specialized cells acquire the specific numbers of mtDNA copies required to undertake their specialized functions, as is the case for primordial germ cells that differentiated into mature fertilizable oocytes [20,47–49] (Figure 3). Indeed, it is of great importance for a pluripotent cell to establish the mtDNA set point to maintain pluripotency. Partial knockdown of PolgA, the major mtDNA replication factor, in undifferentiated ESCs leads to reduced mtDNA copy number, down-regulation of Oct4 expression and the onset of early differentiation .
There is a close association between pluripotency, mtDNA and mitochondrial function during embryo development, and the consequences of failing to establish the mtDNA set point can be understood from the perspective of induced pluripotency . Induced pluripotent stem cells (iPSCs) are derived from differentiated cells, such as fibroblasts, which undergo reprograming, or dedifferentiation, to re-establish pluripotency. However, the degree of reprogramming that takes place is a key issue. iPSCs may express pluripotent markers; however, they can still have incomplete pluripotent potential due to the carry-over of epigenetic memory from the cells of origin, which leads them to differentiate into certain cell lineages instead of all cell types [50,51]. Furthermore, in some cases, they spontaneously start to express lineage-specific markers, but mtDNA replication is restricted when compared with differentiated ESCs, which eventually leads to failed differentiation . Indeed, it has been shown that iPSCs, which failed to complete the process of reprogramming, attempt to increase mtDNA copy number during differentiation but this then stalls, a situation that we refer to as ‘pseudo-differentiation’ . Thus complete pluripotency requires not only a genetically and epigenetically reprogrammed nucleus, but also the establishment of the mtDNA set point to synchronize the amount of mtDNA available to enable cells to perform their specialized functions when fully differentiated.
Cancer cells help to explain the importance of mtDNA to the process of differentiation. Cancer cells are highly proliferative cells, some of which can be induced to differentiate . Indeed, the differentiated status of a cancer cell is widely used to grade cancers [52,53]. A more differentiated cancer cell is more like a normal somatic cell that is non-cancerous in nature, and it will grow at a slower rate than an undifferentiated cancer cell, as is the case for mature differentiated cells [52,53]. A variety of mitochondrial dysfunctions have been discovered in cancer cells, which are tightly related to mtDNA and highly likely to contribute to their failure to undergo differentiation and their ability to adopt a pseudo-differentiated state [8,54]. As an example, glioblastoma multiforme cells can be induced to initiate the process of differentiation. However, the process stalls and is marked by asynchronous increases in mtDNA copy number and perturbed cell-specific gene expression patterns . Undifferentiated cancer cells share some characteristics associated with pluripotency, including cellular energy mechanisms and the regulation of mtDNA replication when they fail to differentiate [8,21]. We now review some of the related mitochondrial dysfunctions identified in tumorigenesis to bring new insights to the understanding and the importance of mtDNA in the overall picture of development.
Mitochondrial dysfunction in tumorigenesis associated with the mtDNA set point
The Warburg effect
The Warburg hypothesis states that cancer cells primarily rely on aerobic glycolysis instead of OXPHOS for energy production even in the presence of oxygen (Figure 4). In 1924, Warburg initially asserted the hypothesis that the primary cause of cancer was the damage to mitochondria and, subsequently, insufficient metabolic respiration . With the finding of oncogenic mutations as the root cause of cancer , the Warburg effect currently describes the metabolic features of cancer cells. Indeed, not only do cancer cells utilize aerobic glycolysis for energy production but embryonic cells do so in a similar manner . Low mtDNA copy number in cancer and pluripotent cells limits energy production through OXPHOS, and this is thought to aid cellular proliferation . Even though aerobic glycolysis is not as efficient as OXPHOS in generating large amounts of ATP, one explanation is that it incorporates nutrients and metabolizes them into other organic elements (e.g. nucleotides, amino acids and lipids) to support the requirements to build and generate new cells when in a proliferative mode . Although impaired mitochondrial function is not found in most cancers to be the direct cause mediated by the Warburg effect , recent studies have extended the hypothesis to a variety of mitochondrial dysfunctions in cancer cells, which might contribute to or result from the failure to establish the mtDNA set point .
Mitochondrial function and dysfunction in tumour cells and during development.
Pathological DNA mutations
Sporadic nDNA mutations or abnormal expression levels of the nDNA-encoded factors of mtDNA transcription and replication lead to dysfunctional mtDNA transcription and replication, which could result in reduced oxidative capacity and energy production. Once the reduced level affects cellular metabolism, clinical symptoms are often observed. It is known that mutations to POLG could lead to depletion of mtDNA and promote breast cancer . In addition, frequent truncating mutations of Tfam have been demonstrated to induce mtDNA depletion and apoptotic resistance in a mouse model of colorectal cancer . Thus these mutations could perturb the establishment of the mtDNA set point, which then limits the differentiation potential of tumour cells.
The impact of ROS production and mitochondrial metabolites on gene expression
ROS are a group of by-products from cellular metabolism that can readily oxidize other molecules. ROS are frequently generated in the mitochondria . Increased ROS results from cellular stress and can translate the stress into cellular signalling pathways for certain responses . It has been demonstrated that increased ROS could trigger mitophagy and apoptosis to either reduce the number of mitochondria, therefore reducing mtDNA copy number, or trigger cell death dependent on the extent of the increase [64,65]. Low levels of ROS are beneficial for healthy metabolism through post-translational modification of phosphatases and kinases , whereas excessively high levels of ROS generate lethal levels of oxidative stress and lead to cell death by causing damage to DNA , triggering senescence  and apoptosis .
Elevated ROS has been identified in cancer cells that are induced by oncogenes to promote tumorigenesis. However, to maintain the elevated level at a beneficial level rather than at the excessively high level, it is tightly controlled by aberrant gene expression and metabolism in cancer, as cancer cells can produce higher levels of antioxidants to balance the redox level to a level beneficial for tumorigenesis . It has been reported that PGC1α can up-regulate the expression of antioxidant enzymes to maintain redox balance . The tumour-suppressor gene p53 is also known to interact with NRF2 to up-regulate the expression of antioxidants . The oncogene MYC regulates antioxidant levels by promoting glutamine uptake and glutaminolysis . Furthermore, down-regulation of TFAM has recently been found to inhibit the tumorigenesis of non-small-cell lung cancer by activating ROS-mediated c-Jun N-terminal kinase (JNK)/p38 mitogen-activated protein kinase (MAPK) signalling and reduce cellular energy . These findings related to the mtDNA replication and transcription factors provide functional explanations as to how the failure to establish the mtDNA set point blocks the differentiation of cancer cells and promotes tumorigenesis, but this has yet to be tested experimentally.
Mutations to mtDNA
The mitochondrial genome encodes several subunits of the ETC complexes and other functional RNAs. Therefore, reduced function in the ETC complexes resulting from homoplasmic mtDNA mutations can lead to OXPHOS inefficiency. Mutations in the mtDNA-encoded ND genes and COX1 have been identified in primary tumours and it is suggested that they promote tumour progression [74–78]. This will contribute to the inefficiency within the ETC, as more electrons are passed through the complexes resulting in the production of more ROS that causes further mtDNA damage or mutation  (Figure 4). Nevertheless, mechanistic evidence is needed to determine whether mtDNA mutations are a part of tumorigenesis or just a part of mitochondrial dysfunction in cancer. Even though the levels of mtDNA mutations are not directly related to mtDNA copy number, they affect the cell's respiratory capacity. If the levels of mtDNA mutations affect the function of the mtDNA-encoded ETC subunits, it therefore suggests that the establishment of the mtDNA set point is disturbed functionally. Consequently, mtDNA copy number should be assessed in the context of mitochondrial function and whether a cell's set point is modulated to compensate for this.
The control of the number of mitochondria
Mitochondrial biogenesis is activated through environmental stimuli or when cells are under redox stress [79–82]. Mitochondrial biogenesis is a complicated procedure requiring mtDNA replication and synthesis of ETC subunits [81,82]. In early differentiation, mitochondrial maturation also involves the formation of cristae and a higher membrane potential for normal mitochondrial functions [83,84] (Figure 4). The Sirtuin family, especially Sirtuin 1 and 3 (encoded by SIRT1/SIRT3) have been shown to be involved in mitochondrial biogenesis, in mitochondrial transcription as the target of PGC1α, and have a role in suppressing ROS production [85,86]. Moreover, the ETC subunits, as well as other mitochondrial structural proteins, are encoded by both the chromosomal and mitochondrial genomes. Thus the process of mitochondrial biogenesis is highly dependent on the ordered and effective co-operation between the hierarchical network of the factors from both the nucleus and the mitochondria. In cancer, some oncogenes and tumour suppressors have been found to affect this network to alter the rate of biogenesis. The tumour suppressor p53 has been demonstrated to repress the transcription of the main initiator of mitochondrial biogenesis, PGC1α, when telomere function is compromised , and therefore causes deregulation of the downstream pathway through cofactors of PGC1α, namely PPARγ, NRF1/2 and oestrogen-related receptor (ERR) α/β and, extensively, to key mitochondrial regulators, such as TFAM and POLGA . In addition, the oncogene MYC could promote mitochondrial biogenesis by up-regulating expression of PGC1β [88–90] (Figure 1). The cell proliferation factor signal transducer and activator of transcription 3 (STAT3) also directly binds to the D-loop region of mtDNA to regulate its transcription. More importantly, STAT3 promotes mitochondrial transcription to optimize energy production which is used to promote the proliferation of ESCs . STAT3 also targets hypoxia-induced factor 1α (HIF1α), the master transcription regulator of the genes that respond to hypoxia, which also interacts with PGC1α to promote mitochondrial biogenesis [92,93]. Increased mitochondrial biogenesis in cancer cells is probably a response to offset lower energy production resulting from the Warburg effect and to balance out the elevated levels of ROS.
On the other hand, the number of mitochondria is also controlled by mitochondria-specific autophagy, known as mitophagy. Mitophagy describes the process of degradation of mitochondria by the autophagosome (Figure 4). Mitophagy is triggered when the mitochondrial membrane is depolarized, which, in most cases, results from mitochondrial fission (discussed below) [64,94]. Mitophagy has tumour-suppressor function as the loss of the key regulators of mitophagy, parkin and BNIP3/NIX, promotes tumorigenesis dependent on the type and the stage of the cancers [95–97]. BNIP3/NIX can be induced by hypoxia as the tumour cell microenvironment involves the regulation of p53 . In ESCs, cellular function and pluripotency are also sensitive to the microenvironment, especially in tissue culture. It has been shown that the lack of a growth factor, Gfer, can trigger mitophagy and loss of pluripotency, which could result in reduced mtDNA copy number and failure to establish the mtDNA set point .
The dynamics of mitochondrial fission and fusion happen at a balanced rate in normal cells, which closely responds to hypoxia, energy metabolism, apoptosis and other cellular stress  (Figure 4). Mitochondrial fission is driven by the GTPase dynamin-related protein 1 (DRP1) with co-factors including MFF1 and FIS1. Mitochondrial fusion requires the GTPases mitofusin 1/2 located on the outer membrane and OPA1 on the inner membrane of mitochondria . Mitochondrial fission is important for mitosis to distribute the organelles evenly into daughter cells, whereas fusion has been found to promote OXPHOS and cell growth probably by increasing the density of the ETC complexes and improving the formation of the ETC supercomplexes . Evidence of increased fission rate has been reported in cancer cells and has been related to the increases in ROS and mitophagy, which again could lead to the reduction in mtDNA copy number . Moreover, it has been noted that mitochondrial dynamics might play vital roles in apoptosis, as observed in interactions between the key regulators of fusion/fission and apoptosis including oncogenic BCL-2 family members [101,102]. As apoptotic evasion is one of the hallmarks of tumorigenesis, these findings have shown the potential contribution of mitochondrial dynamics to tumorigenesis. However, direct experimental results need to be analysed to determine the cellular mechanisms in which the abnormal mitochondrial dynamics identified in tumorigenesis contribute to or resulted from the abnormal regulation of mtDNA copy number and stalled differentiation.
In summary, due to the fundamental role of mitochondria in cellular metabolism and bioenergetics, it is not surprising to see such a variety of mitochondrial dysfunctions associated with tumorigenesis (Figure 4). However, most of the explanations are not clear enough to clarify how mitochondrial dysfunction mechanistically relates to tumorigenesis and failure to complete differentiation. As it has been shown that, by resetting mtDNA copy number and thus the mtDNA set point, cellular differentiation could be promoted , these dysfunctions probably result from or contribute to the failure to establish the mtDNA set point. As reviewed above, the interactions between oncogenes/tumour suppressors and mitochondrial replication and transcription factors are present at multiple levels, which result in a variety of metabolic changes during tumorigenesis. Cancers derived from different cell types exhibit different mitochondrial dysfunctions with different metabolic pathways being perturbed. Moreover, the mtDNA set point is essential to determining cell fate. These findings highlight that the regulation of mtDNA copy number co-operates tightly with other metabolic pathways to regulate various cellular functions, which, if perturbed, could lead to tumorigenesis.
The effect of DNA methylation on mtDNA copy number during differentiation
What is DNA methylation?
DNA methylation is the only known covalent modification of DNA . Typically, DNA methylation occurs at 5΄ of the pyrimidine ring of cytosine (5 mC) followed by guanine (CpG dinucleotide) in mammalian DNA (Figure 5). It is known that over 28 million CpGs exist in the human genome. In plants, it can also occur at non-CpG cytosines . There is also emerging evidence showing that non-CpG DNA methylation can occur at CpA and CpT dinucleotides in mammalian cells, such as embryonic and neuronal cells [105,106].
DNA methylation and demethylation mechanisms.
Methyl groups are added to cytosines by DNA methyltransferase (DNMT) enzymes  (Figure 5). There are two types of DNMT enzymes either working to maintain DNA methylation during DNA replication or initiating DNA methylation at a novel DNA site. Maintenance of DNA methylation is carried out by DNMT1 along with DNA replication, whereas DNMT3A and DNMT3B are believed to function as the de novo methylation DNMT enzymes adding methyl groups at new cytosine sites  (Figure 5). 5 mC can be demethylated to 5-hydroxymethylcytosine (5 hmC) by ten-eleven translocation methylcytosine dioxygenase (TET; Figure 5) . The TET family, consisting of TET1, TET2 and TET3, is known to oxidize 5 mC to 5 hmC [109,110].
CpG dinucleotides usually appear in repeat clusters termed CpG islands (CGIs). The widely accepted definition of CGI was given by Gardiner-Garden and Frommer in 1987 as a 200-bp sequence of DNA with a CG content of more than 50% and the observed/expected CpG ratio (O/E ratio) is more than 0.6 . Where the O/E ratio is calculated by the formula :
CGIs contain high CpG components, which could be 5-fold higher than the other regions. The estimated number of CGIs is 30 000 in the human genome. On average, CGIs are ∼1 kb in length with 7% being CpG dinucleotides [111,112].
Chemical agents can also inhibit DNA methylation or induce DNA demethylation. They have been widely used in DNA methylation research. 5-Azacytidine is a chemical analogue of cytosine that inhibits DNA methylation at low concentrations . It blocks DNA methyltransferease, DNMT1, which causes global DNA demethylation  (Figure 5). Recent research has revealed that vitamin C can also induce DNA demethylation by enhancing the activity of the TET1 enzymes and thus improving the conversion of 5 mC into the DNA demethylated state of 5 hmC  (Figure 5).
In mammalian cells, the methylome is mainly composed of genomic regions of low CpG densities, whereas regions of higher CpG densities, such as CGIs, are mainly unmethylated. DNA methylation profiles vary significantly between different tissue types reflecting different gene expression profiles and biological functions . DNA regions with a different methylation status can result in different biological consequences.
Biological function of DNA methylation in different gene regions
DNA methylation plays vital roles in embryonic development, maintenance of pluripotency, genomic imprinting and X chromosome inactivation by regulating transcription and stabilizing chromosomes and chromatin structure [117–119]. It is a dispensable epigenetic system in the regulation of gene expression. Most often, DNA methylation is best characterized by silencing transcription which occurs at CpG islands in gene promoter regions. Through access to high-throughput methylome datasets, it has been suggested that DNA methylation seems to occur at intragenic (within gene-coding) regions, and at regulatory intergenic (between gene-coding) regions, which appear to regulate gene expression patterns differently.
More than 50% of human genes possess CGIs in their promoter regions . As mentioned above, most of them remain unmethylated in somatic cells. Some suppressed genes have been shown with methylated CGIs in their promoter regions . Hypermethylated promoters have been well characterized to block binding of transcription factors and to silence gene expression. Aberrant methylation of promoter-located CGIs, especially hypermethylated promoters, plays important roles in repressing tumour-suppressor genes [121,122]. Interestingly, nearly all of the known mitochondrial replication and transcription factors have a CGI in their promoter regions, which indicates that their gene expression is probably affected by the DNA methylation status across their CGI regions. DNA hypermethylation across the promoters of PPARGC1A (encoding PGC1α) has been shown to be mediated by DNMT3B and leads to the reduction of mitochondrial content in muscle cells in diabetic patients . Methylation in the promoter region of TFAM also has an impact on mitochondrial function in insulin resistance in peripheral leucocytes .
Methylome datasets of single-base resolution have shown that fewer CGIs exist but more DNA methylation occurs in intragenic regions [125,126]. Intragenic methylation also presents in a strong tissue-specific manner. However, the function of intragenic methylation shows diversity and has also been referred to as the DNA methylation paradox . It is believed that high DNA methylation is important for blocking transcription of the repetitive DNA elements and suppressing initiation from spurious promoters within gene regions [122,128]. However, a non-monotonic relationship is observed between intragenic methylation and gene expression . Both positive and negative correlations exist between DNA methylation at exons and gene expression in different cell types . Evidence has also shown that DNA methylation across exons is higher than in introns . The transition of DNA methylation at exon/intron boundaries leads to a proposed function for maintaining transcriptional efficiency by regulating splicing and elongation . Moreover, the dynamics of DNA methylation across gene regions might also be related to the transcriptional unit including the accessibility of chromatin and the relative density of the nucleosomes, DNMT1 and human RNA polymerase II (POL II) . In an inactive gene region, chromatin is closed and dense nucleosomes block DNMT1 from mediating DNA methylation, whereas when chromatin partially opens, DNMT1 methylates gene regions along with POL II to execute transcription; a completely opened chromatin is fully occupied with POL II for transcription, which interferes with the access of DNMT1.
Overall, there is still a question mark over the relationship between methylome patterns and transcriptional profiles. When talking about DNA methylation patterns, inter- (between gene-coding regions) and intra- (within gene-coding regions) genetically, it is necessary to take gene expression and other regulatory mechanisms into consideration and investigate them all thoroughly based on the particular cell types and their corresponding cell functions.
The effect of DNA methylation on mtDNA copy number during differentiation
During cell differentiation and embryonic development, the DNA methylation profile is re-established. After fertilization, TET enzymes become activate and erase the parental DNA methylation levels significantly  (Figure 3). By the blastocyst stage, the extensively demethylated genome of the ICM starts to be re-methylated by DNMT3a and DNMT3b up to their specific levels and maintained by DNMT1 afterwards during DNA replication  (Figure 3). The changes that occur during differentiation are believed to be predominantly responsible for transcriptional changes specified for different cell fates. It has been reported that in vivo liver differentiation in human ESCs is characterized by DNA demethylation over time . In a DNA methylation study targeting genome-wide promoter regions of mouse ESCs, it was reported that the DNA–protein interactions between the promoter regions and OCT4, NANOG, Polycomb proteins and histone modifications (H3K27 trimethylation) are correlated with the DNA methylation status of the cell . Increasing studies on DNA methylation in embryonic cells have provided insights into how developmental processes are modulated by DNA methylation and more and more cell lineages are characterized by their DNA methylation signatures.
As mentioned before, mtDNA copy number is also reset to establish the mtDNA set point starting at the blastocyst stage leading towards specified mtDNA copy number for different cell lineages (Figure 3). It has been questioned whether there is a relationship between the coinciding changes in DNA methylation and mtDNA copy number. A recent study revealed that the levels of DNA methylation increased at exon 2 of PolgA in mouse ESCs as they underwent differentiation into neurons . Furthermore, the level of DNA methylation varies at exon 2 of PolgA in a tissue-specific manner. Statistical analysis suggested that PolgA expression was negatively correlated with its intragenic DNA methylation at exon 2 and the corresponding mtDNA copy number, which indicates that changes to DNA methylation could modulate mtDNA copy number by regulating the gene expression of the primary mtDNA replication factor POLGA. Similar findings are also observed in the differentiation process from human ESCs to neural stem cells and finally into astrocytes . High DNA methylation was observed in pluripotent and multipotent cells that correlated with low POLGA expression and low mtDNA copy number. In terminally differentiated astrocytes, however, DNA methylation at exon 2 of POLGA decreased with higher mtDNA copy number. Interestingly, cancer cell lines, such as the glioblastoma multiforme HSR-GBM1 cells, which failed to differentiate, are unable to demethylate exon 2 of POLGA . However, in the presence of DNA demethylation agents such as 5-Aza, the high DNA methylation status at exon 2 of POLGA was decreased promoting mtDNA replication and cell differentiation. These findings highlight the important role of DNA methylation in modulating mtDNA copy number, both of which act as vital determinants in cell differentiation and development (Figure 6). Likewise, partial depletion of mtDNA in cancer cells changes the methylation status at exon 2 of POLGA and affects the tumorigenic capacity of these cells. Nevertheless, when tumours form, mtDNA copy number is reinstated to original levels, suggesting that cells do, indeed, have an ‘in built’ set point [21,135].
The mtDNA set point is critical for differentiation of pluripotent cells and tumour cells.
Interestingly, the mtDNA set point can be modulated by mtDNA genotype. In cells that have had their mtDNA replaced with mtDNA from different genotypes, either in tumour cell  or ESC  models, not only are chromosomal gene expression patterns changed, but mtDNA copy number is modulated in response along with DNA methylation patterns at exon 2 of POLGA . This indicates that the establishment of the mtDNA set point is a two way process where interactions between the nuclear and mitochondrial genomes reflect adaptation to the environment and cellular phenotype.
mtDNA is a critical part of the cellular energy-production system. Nevertheless, its impact is no longer limited to powering cellular function. mtDNA copy number changes during cellular development which affects fertilization rates and embryo viability. The establishment of the mtDNA set point is essential for the potential of a cell to complete differentiation. During differentiation, mtDNA copy number increases in a lineage-specific manner as dictated by a cell's specialized cellular functions. The changes to mtDNA copy number are determined not only by the expression of mtDNA transcription and replication factors but also by stage-specific markers, as is the case during pluripotency, which it is very tightly regulated. The synchronous changes to DNA methylation during development, especially at exon 2 of POLGA, provide insights into our understanding of the cross-talk between the nucleus and the mitochondria; in other words, the changes to gene expression profiles and mtDNA copy number. These findings can be combined further to understand tumorigenesis and developmental disorders. Aberrant DNA methylation profiles found in early development and tumorigenesis interfere with the expression of POLGA, which destabilizes the mtDNA set point and the expansion of mtDNA copy number during differentiation. As a great variety of genes and regulators are involved in regulating mtDNA as well as general mitochondrial function, it is highly likely that similar mechanisms exist that have an impact on development and tumorigenesis.
5-Aza, 5-azacitidine; CGI, CpG island; CoQ, coenzyme Q; COX, cytochrome c oxidase; CytB, cytochrome c reductase; CytC, cytochrome c; D-loop, displacement loop; DNMT, DNA methyltransferase; ERR, oestrogen-related receptor; ESC, embryonic stem cell; ETC, electron transfer chain; HIF1α, hypoxia-induced factor 1α; 5 hmC, 5-hydroxymethylcytosine; 5 mC, 5-methylcytosine; ICM, inner cell mass; iPSC, induced pluripotent stem cell; ND, NADH dehydrogenase; nDNA, nuclear DNA; NRF, nuclear respiratory factor; O/E ratio, observed/expected CpG ratio; OXPHOS, oxidative phosphorylation; PGC1α, peroxisome-proliferator-activated receptor γ co-activator 1α; POLG, DNA polymerase gamma; POLGA, POLG subunit A; POL II, RNA polymerase II; PPAR, peroxisome-proliferator-activated receptor; ROS, reactive oxygen species; SIRT, sirtuin; STAT3, signal transducer and activator of transcription 3; TET, ten-eleven translocation methylcytosine dioxygenase; TFAM, mitochondrial transcription factor A.
This work was funded by the Australian National Health and Medical Research Council [grant number GNT1041471] and the Victorian Government's Operational Infrastructure Support Program.
The Authors declare that there are no competing interests associated with the manuscript.