The mitochondrion is the single cellular site of ATP generation via aerobic respiration, and metabolites such as dietary lipids and pyruvate, the metabolic product of glycolysis, are actively transported into mitochondria 79. As the tricarboxylic acid cycle progresses within the mitochondrial matrix, a series of electron-transfer reactions, known collectively as the electron-transport chain, proceeds between large multiprotein complexes and small electron carriers within the inner membrane and matrix 5. The resulting electrochemical gradient generates bioavailable ATP via a rotating inner-membrane ATPase, which couples proton flow down a proton gradient to the catalysis of the phosphorylation of ADP to ATP using inorganic phosphate 80.

The mitochondrion is derived from a symbiotic α-proteobacterium 81, and so the mitochondrial genome is packaged and structured differently from the nuclear genome 82. The sequence of the mitochondrial genome and the translation machinery are also more similar to that of a bacterium than to eukaryotic systems 83, and the mitochondrial transcription machinery is reminiscent of that used by bacteriophages 84. In contrast to the chromatin-based packaging of the nuclear genome, the mitochondrial genome is packaged into non-chromatin nucleoids involving proteins specific to mitochondria, such as Tfam 485. Although the mammalian mitochondrial DNA is small, at around 16.5 kb, it nevertheless encodes 13 protein-coding genes, 22 tRNA genes and 2 rRNA genes, as shown in Figure 186. Unlike nuclear genes, each of which often has multiple dedicated promoters, all mitochondrial genes are expressed together from only three promoters encoded in the regulatory D-loop region 87, which are recognized by the mitochondrial basal transcriptional machinery: the mitochondrial RNA polymerase (Polrmt), and the mitochondrial transcription factors Tfam and Tfb2m 488. The resulting three polycistronic transcripts do not undergo splicing, and are processed by an RNase that excises tRNAs to release the mRNA and rRNA 989 before mRNA translation in the mitochondrial matrix.

The thyroid hormone T₃ is a primary regulator of mammalian mitochondrial biogenesis 19 and can influence mitochondrial function both indirectly and directly. In its indirect role, it binds to members of the nuclear receptor superfamily of transcription factors known as T₃ receptor α and β (T₃Rα and T₃Rβ) to regulate nuclear gene transcription 20 (Figure 3). Nuclear transcriptional targets of these receptors include genes that stimulate mitochondrial biogenesis, such as those encoding the transcription factor nuclear respiratory factor 1 (NRF-1) and the cofactor PGC-1α 2122 as well as the mitochondrial basal transcription factor Tfam 2324.

T₃ also regulates mitochondrial function directly via two pathways: the regulation of nucleotide transport across the inner membrane via a T₃-binding adenine nucleotide transporter (AdNT) 2526; and control of mitochondrial transcription via the mitochondrially localized T₃Rα1 isoform known as p43 (Figure 3; reviewed in 2728, see also 10). Most proteins are imported into mitochondria in an ATP-dependent manner via the protein translocator channel TOM, which recognizes an amino-terminal mitochondrial localization signal that is then cleaved during import. p43, however, is imported into rat liver mitochondria via a different pathway, previously shown for the yeast mitochondrial transcription factor MTF1 15, which is independent of both TOM and mitochondrial ATP levels, and does not result in cleavage of the imported protein.

One potential mechanism by which transcription factors could be accurately sorted among different cellular compartments is revealed when considering the T₃ receptors. It was noticed that a protein construct mimicking a T₃Rβ isoform with a truncated amino terminus (which is the form present in most non-mammalian vertebrates) is specifically imported into isolated rat mitochondria, suggesting a role for the amino-terminal truncation in mitochondrial import 15. p43 is itself a truncated form of the full-length nuclear transcription factor T₃Rα and is translated from an alternative start site in the T₃Rα mRNA 1529. The expression of p43 is, however, regulated independently from the full-length T₃Rα and it shows a distinct tissue-specific pattern of expression 29.

T₃ receptors associate with nuclear DNA in a sequence-specific manner via T₃-response elements (T3REs), DNA motifs first recognized in the promoters of T₃-responsive genes. Multiple T3REs have been identified in the mouse mitochondrial genome that confer responsiveness to thyroid hormone in nuclear reporter assays, and p43 binds to these sequences in vitro in EMSAs. These techniques do not, however, address the question of whether p43 binds mitochondrial DNA in vivo under physiological conditions. This was partly addressed in a series of studies utilizing induced-hypothyroid rats, in which physiological T₃ levels were found to regulate mitochondrial gene expression directly in vivo. Mitochondria isolated from the livers of these rats showed that changes in physiological thyroid hormone levels altered the relative levels of mitochondrial mRNA and rRNA, which correlated with altered protein occupation of the mitochondrial D-loop as determined by DNA footprinting 30. The independence of this direct mitochondrial role for T₃ from the well-characterized indirect role was shown when isolated mitochondria from hypothyroid rats were treated with T₃ and the mitochondrial mRNA:rRNA ratio and the pattern of DNA footprinting returned to that of normal rats 30. This indicated that T₃ regulates mitochondria directly and suggested that this pathway may involve a mitochondrial T₃ receptor with similar binding preferences to the nuclear form.

The role of p43 in T₃-mediated regulation of mitochondrial transcription was confirmed using the same in organello system from induced-hypothyroid rats to show that the addition of p43 (translated in vitro to avoid possible contamination by cellular components) stimulated mitochondrial gene transcription in the presence of T₃ 15, whereas T₃ treatment in the absence of p43 did not stimulate mitochondrial gene expression 15. In validation of the mitochondrial role of p43 in vivo, mice overexpressing p43 under the control of a muscle-specific promoter exhibited increased mitochondrial gene expression and mitochondrial biogenesis in muscle, and had increased oxidative metabolism, with body temperature 0.8°C higher than control mice 31.

The direct regulation of mitochondrial transcription by T₃ is complex and highly tissue specific. In organello studies that demonstrate the responsiveness of liver mitochondrial transcription to T₃ also demonstrate that mitochondria from the heart are not regulated in this manner. Rather, T₃ regulation of mitochondria from the hearts of the induced-hypothyroid rats is indirect - via the nucleus - and primarily at the level of regulating mitochondrial DNA copy number 32. This complexity is likely to be shared by other transcription factors with a direct mitochondrial activity, and it may explain why early work did not detect direct binding of a protein to the proposed mitochondrial T3REs 30 despite the requirement of a DNA-binding domain in p43 for the observed mitochondrial function 15. Regardless of this outstanding debate over the location of p43 binding to mitochondrial DNA, evidence is overwhelming that p43 is localized to the mitochondria in rat liver, where it binds to the mitochondrial genome and regulates mitochondrial transcription. A great deal remains to be studied regarding p43 in mitochondria - for example, it is not clear whether this regulatory pathway is conserved within mammals, or in which other tissues it is utilized. Nevertheless, the studies on thyroid hormone and thyroid hormone receptor were the first direct illustration that mitochondrial gene expression is regulated independently from nuclear gene expression and introduced a key model system for the study of nuclear transcription factors in mitochondria.

The transcription factor CREB regulates nuclear gene expression in response to a diverse range of stimuli 3334. CREB is activated by phosphorylation, either by the cyclic-AMP responsive protein kinase A (PKA) or by other kinases, including mitogen-activated protein kinases (MAPKs) and Ca²⁺/calmodulin-dependent kinases (CaMKs) 35. A self-contained CREB pathway exists in mitochondria, which involves PKA 36, cyclic AMP 37 and CREB 38. On stimulation, this pathway induces binding of phosphorylated CREB to cyclic-AMP response elements (CREs) in the mitochondrial DNA D-loop and regulation of mitochondrial gene expression 3940 (Figure 1).

CREB was first localized to rat brain mitochondria by subcellular fractionation followed by immunoelectron microscopy 38. Despite not having a classical mitochondrial localization signal, the transport of labeled CREB into isolated rat liver mitochondria depends on the mitochondrial translocator TOM, the import route for most proteins into mitochondria 39. The mitochondrial pool of CREB can co-immunoprecipitate with the chaperone protein mtHSP70 40, suggesting a mechanism of targeting to the mitochondria that is dependent on chaperone proteins rather than on a mitochondrial localization signal, as has been shown for p53 41. Once in mitochondria, CREB is regulated by phosphorylation in response to the same stimuli as in the nucleus, and in vitro can bind oligonucleotides bearing the consensus CRE sequence 38. Binding of CREB to the mitochondrial D-loop (Figure 1) has been detected in vivo using ChIP 36 and DNase footprinting, and is dependent on mitochondrial PKA activity 4041. Unlike p43, mitochondrial localization of CREB has been identified in multiple mammalian species and tissues 363839.

An overexpression construct that selectively increases levels of CREB in mitochondria was used to distinguish CREB nuclear and mitochondrial regulatory roles in primary cultured neurons from rat brain 40. These increases in mitochondrial CREB perturbed mitochondrial gene expression without altering the expression levels of CREB's nuclear target, c-fos 40. The mRNAs of mitochondrially encoded NADH dehydrogenase subunits 2, 4 and 5 (Figure 1) were specifically upregulated; conversely, these mRNAs were downregulated on treatment with a dominant-negative form of CREB in the mitochondria 40.

The tumor suppressor protein p53 is a well-known example of a nuclear transcription factor with a role in mitochondria 42. First identified by its transcriptional regulatory function 43, p53 also has non-transcriptional functions, and has been implicated in apoptosis 44, senescence 45, autophagy 46, DNA-damage repair and cell-cycle arrest 47.

In mitochondria, p53 directly regulates apoptosis via protein-protein interactions at the outer membrane, and this function has been reviewed thoroughly elsewhere 4849. However, there is considerable evidence for a second mitochondrial role for p53, in mitochondrial DNA maintenance and in mitochondrial DNA-damage repair. Co-immunoprecipitation of p53 with the mitochondrion-specific transcription and mitochondrial DNA packaging factor Tfam suggests that p53 may regulate DNA-damage repair in mitochondria 50, as it does in the nucleus 5152. In KB human epidermoid cancer cells and in HCT116 adenocarcinoma cells, p53 physically interacts with Tfam, with the effect of enhancing the binding of Tfam to cisplatin-damaged DNA at the expense of oxidized DNA, in a reversal of Tfam's normal binding pattern 50.

p53 also seems to play a role in mitochondrial base-excision repair. In a nucleus-free in vitro system derived from the mitochondria of mouse liver, p53 can stimulate the gap-filling function of the mitochondrial DNA polymerase mtPOLγ 53. A physical interaction between p53 and mtPOLγ in vivo has been detected in HCT116 cells 54, where p53 enhances the replication function of mtPOLγ and interacts with the mitochondrial genome. The observed binding of p53 to the mitochondrial genome was stimulated by, but not dependent on, DNA damage, suggesting that the role of p53 at the mitochondrial DNA may not be confined to the DNA-damage response. Furthermore, in studies comparing mitochondria from p53-deficient cell lines with those from isogenic p53-positive lines, p53 appears to provide an endogenous proofreading function for mtPOLγ during mitochondrial DNA replication 55.

Despite clear localization of p53 to the mitochondrial matrix and a number of direct associations with mitochondrial DNA, evidence that p53 can bind sequence-specifically to regulate expression of mitochondrially encoded genes remains elusive. Sequences from the mouse mitochondrial genome that resemble the nuclear binding motif of p53 confer p53 responsiveness in nuclear reporter assays 56, but there is no evidence that p53 regulates transcription from the mitochondrial genome. Regardless of whether p53 directly regulates mitochondrial transcription, it plays important mitochondrial roles in apoptosis, DNA integrity and response to stress.

Stat3 was first detected in mitochondria as a result of its functional association with GRIM-19, a subunit of the respiratory electron-transport chain NADH dehydrogenase (Complex I), which functions in the transfer of electrons from NADH to the respiratory chain 5758. In the nucleus, Stat3 mediates the transcriptional response to growth factors such as interleukin-6 and epithelial growth factor 59. Differences between Stat3 function in the mitochondria and nucleus are exemplified by the fact that the mitochondrial pool of Stat3 mediates oncogenic transformation by the small GTPase H-Ras, a process that is mechanistically distinct from how nuclear Stat3 supports oncogenic transformation by the viral oncogene v-Src 60. Both Stat3-knockdown cell lines and Stat3-knockout mice show disrupted electron-transport chain function 61, which suggests that Stat3 directly regulates mitochondrial function via its effects on the electron-transport chain. Engineered Stat3 mutant proteins have shown that the nuclear role and the mitochondrial role can be functionally isolated 61.

The estrogen receptor was first found to localize to mitochondria of rabbit uterus and ovary in 2001 62. This receptor regulates gene expression by binding to estrogen-response elements (EREs) in gene promoters following the binding of the steroid hormone estrogen to the receptor 63. Nuclear targets include NRF-1, which, as noted earlier, encodes a transcription factor that stimulates mitochondrial biogenesis and is a transcriptional regulator of genes encoding the mitochondrial basal transcription machinery 64 (Figure 2). The indirect regulation of mitochondrial function by the actions of the estrogen receptor has been reviewed elsewhere 6566.

There is evidence that estrogen acts directly in mitochondria by two pathways: one utilizing the receptor and the other independent of it 276768. The presence of the estrogen receptor in mitochondria is well established; both isoforms, ERα and ERβ, localize to mitochondria in diverse cell lines and tissues, yet their functions remain contentious. EMSAs suggest that ERβ may bind directly to the D-loop of the mitochondrial genome in MCF-7 breast cancer cells (Figure 1). This binding was stimulated by treatment of the cells with estrogen and inhibited by treatment with ERβ-specific antibodies 69. It has not, however, been shown that isolated mitochondria respond to estrogen treatment by altering gene expression in an ERβ-dependent fashion.

Other putative functions for the mitochondrial estrogen receptor have focused on protein-protein interactions identified using a bacterial two-hybrid screen. This screen revealed that ERα can interact stably and reproducibly with the mitochondrial protein 17β-hydroxysteroid dehydrogenase type 10, suggesting a role for mitochondrial ERα in regulating cellular steroid metabolism and response 70.

Since cancer is in part a metabolic disease 71, and altered mitochondrial DNA sequence and transcription levels have been observed in both primary tumors and in cancer cell lines 7273, a mitochondrial role for the estrogen receptor would be relevant to both estrogen receptor biology and the study of hormone-sensitive breast cancer. As with other nuclear factors, however, the indirect action of the estrogen receptor on mitochondrial gene expression is a confounding factor that complicates investigation. Furthermore, as noted above, estrogen seems to regulate mitochondria directly, even in the absence of its receptor 68. The disagreement over the role of mitochondrial estrogen receptors could be, in part, due to cell-specific functions. Nevertheless, mitochondria are clearly an important target of estrogen hormone action.

Although only a small number of nuclear transcription factors have had a mitochondrial role validated, either in a nucleus-free in organello system or by detection of binding to the mitochondrial genome, there are a number of nuclear transcription factors that have been localized to mitochondria but where the mitochondrial role remains understudied. The glucocorticoid receptor 74, the heterodimeric transcription factor AP-1 75, and the peroxisome proliferator-activated receptor γ (PPARγ) 76 have all been detected in mammalian mitochondria, and there is some evidence for the glucocorticoid receptor 77 and AP-1 78 binding to the mitochondrial genome to potentially regulate gene expression.