The mitochondrial transcription factor TFAM in neurodegeneration: Emerging evidence and mechanisms

Introduction to nuclear transcriptional regulation of mitochondrial content

Cells and tissues can adapt mitochondrial content in response to cellular energetic needs. Nutrient deficiency or mitochondrial dysfunction can trigger an upregulation in the transcription of nuclear encoded mitochondrial genes as a compensatory response called mitochondrial biogenesis [33]. Mitochondrial biogenesis is a process of increasing mitochondrial content that is primarily driven by changes in nuclear transcription. The complexity of this process is beyond the scope of this review, as there are amplification loops, pathway crosstalk, and elements of tissue specificity, but specific transcriptional regulators play a central role (ex. [34]). The PPARγ coactivator-1 (PGC-1) family of transcription co-activators (PGC-1α, PGC-1β, and PPRC1) interact with many transcription factors to regulate nuclear gene expression [35]. Importantly, PGC-1α activates expression of Nuclear Respiratory Factor (NRF)-1 and -2, which together act on many genes whose products are imported into mitochondria [36], including TFAM [37]. Upregulated TFAM transcription is generally concomitant with increased mtDNA, suggesting that TFAM links the nuclear transcription response to mtDNA content, coordinating mitochondrial biogenesis between the two genomes [37, 38].

The role of TFAM in regulating mtDNA replication, transcription, and packaging

TFAM was discovered as a mitochondrial protein that stimulated transcription of model templates in mitochondrial extracts [39]. TFAM is a nuclear-encoded protein that is synthesized in the cytoplasm and imported into the mitochondria [40]. The protein contains two high mobility group (HMG) box domains that insert into the DNA minor groove on the light strand promoter (LSP), heavy strand promoter 1 (HSP1), or non-specific regions of the mtDNA [41–43]. TFAM-binding causes a distortion of the mtDNA and results in DNA bending [42–46]. This distortion enables the specific binding of POLRMT (mitochondrial RNA polymerase) to the start site, where TFAM binds to the N-terminus of POLRMT to recruit TFB2M (mitochondrial Transcription Factor 2) while maintaining an open DNA complex arrangement that enables productive transcription initiation [47–51].

Transcripts are also used in priming DNA synthesis [52], implicating a TFAM function in the initiation of mtDNA replication. In brief, POLRMT synthesizes the nascent transcript from LSP, which either terminates or is post-transcriptionally processed at the conserved sequences blocks (CSB) I, II, or III [52, 53]. The RNA provides exposed 3′-end for the initiation of DNA synthesis by DNA polymerase γ. The nascent DNA frequently pauses, causing a structure termed the D-loop, which has a yet poorly defined function [54]. DNA synthesis that extends beyond the D-loop region has initiated replication. Thus, through the regulation of the LSP transcription initiation, TFAM levels can contribute to mtDNA replication initiation.

TFAM also binds non-specifically to all sequences of the mtDNA, thus coating the mitochondrial genome, providing structural stability and compacting the genome into nucleoid-like structures in vitro [45, 46, 55–58]. TFAM exists almost exclusively in a DNA-bound form [56, 57], as the DNA-free form is rapidly degraded in the mitochondrial matrix by the ATP-dependent mitochondrial Lon protease [59, 60]. TFAM protein abundance is thus tightly correlated with mtDNA content. The notable exception is the case of a POLRMT knockout, which reduces mtDNA content without changing TFAM protein amounts [61]. To summarize, through both its specific and non-specific sequence interactions with mtDNA, TFAM has multiple direct and indirect functions regulating mtDNA copy number to support mitochondrial respiratory function.

TFAM levels and mtDNA copy number

Numerous molecular and genetic models have demonstrated a direct connection between TFAM and mtDNA levels. Heterozygous mutation of TFAM results in ~40% decrease in mtDNA copy number in vivo, whereas the homozygous mutation is embryonically lethal [62]. TFAM heterozygous mice are more prone to metastasis in an intestinal cancer model [63], and TFAM heterozygous cells produce more inflammatory cytokines due to mitochondrial stress signaling [64]. Reciprocally, genetic TFAM overexpression increases mtDNA abundance such that the levels of TFAM are proportionally associated with the levels of mtDNA [65]. TFAM overexpression has also been observed as a protective modifier of diseases in numerous in vivo and in vitro models (reviewed in Campbell [66]). However, there is a limit to the benefit of TFAM overexpression. Titration of TFAM overexpression showed increases in mtDNA levels at low TFAM levels, but decreases as higher levels, suggestive of overcompaction of the genome cause replication inhibition [67, 68].

Regulation of TFAM stability or function through phosphorylation

Because levels of TFAM protein directly regulate mtDNA abundance, post-translational processes that alter the turnover or stability of TFAM protein could play an important role in regulating mtDNA content. Indeed, the mitochondrial matrix protease Lon selectively degrades TFAM protein and thus impacts mtDNA copy number [59]. Interestingly, the cAMP-dependent Protein kinase A (PKA) phosphorylates TFAM at specific sites that impairs DNA-binding, causing TFAM release and allowing its degradation by Lon [60]. In contrast, extracellular signal regulated protein kinases (ERK1/2) phosphorylate a distinct site on TFAM that reduces promoter binding and transcription, but not non-selective mtDNA binding or mtDNA levels [69]. Besides phosphorylation, several studies concluded that TFAM is also post-translationally modified (PTM) by glycosylation [70], acetylation [71], and ubiquitination [72]. Although the numbers of TFAM PTMs described is growing, many of these have been identified through proteomics without further functional characterization, and the detailed mechanisms still need to be identified. As far as phosphorylation is concerned, both the PKA and ERK1/2 phosphosites act to reduce TFAM DNA-binding and thereby impact stability or function through different mechanisms.

TFAM and mtDNA-mediated inflammation

Recent evidence showed that decreased expression of TFAM (e.g. cells and tissues from TFAM heterozygous (TFAM+/-) mice) resulted in mtDNA stress signaling [64]. In primary TFAM heterozygous cells, the decreased TFAM protein levels caused an alteration mtDNA packaging, as evidenced by large mtDNA structures observed in primary TFAM heterozygous cells, allowing fragmented mtDNA to be released in the cytosol to activate the cyclic GMP-AMP synthase (cGAS) [64]. cGAS detects DNA from a broad range of intracellular viral and bacterial pathogens, also sensing mtDNA released to the cytosol upon mitochondrial stress. This activation results in an increased expression of interferon-stimulated genes and antiviral factors leading to a resistance to viral infections. The addition of TFAM, or other HMG box proteins such as bacterial HU and HMGB1, strongly potentiates cGAS signaling [73]. The mitochondrial genome can also increase innate immune responses, such as through the inflammasome or Toll-like receptor 9 (TLR9)-related signaling (reviewed in West et al. [74]). Thus, any process that reduces TFAM function may activate mtDNA stress signaling, contributing to a pro-inflammatory environment that could alter disease trajectory and tissue function [74].

Overall premise supporting a role for mtDNA and TFAM in neurodegeneration and neuroprotection

MtDNA has a 10–17-fold higher mutation rate than that observed for nuclear DNA, implying that the mitochondrial genome is more vulnerable to accumulate oxidative damage that contributes to replication errors [75]. This vulnerability might be due to the mtDNA’s lack of complex chromatin organization, limited DNA repair activities, and/or proximity to the mitochondrial electron transport chain, a major source of cellular reactive oxygen species. Abnormalities in mtDNA were associated with numerous neurodegenerative diseases (reviewed in [76]). For example, mtDNA mutation, deletion, or depletion, as well as impaired transcription and replication, are observed in human samples from individuals affected by PD [77–81], AD [82–87], ALS [88], as well as in aging human or rat samples [89, 90]. It should be noted that not all investigators agree on whether changes in mtDNA mutation, deletion, or copy number are detected in their respective patient sample sets.

Overexpression analysis of mitochondrial TFAM showed a protective effect on cell survival or function in many models of disease that affect energetically demanding tissues, including type II diabetes [70, 91], heart attack [92], and heart failure [93]. Protective effects of TFAM overexpression is also reported in neuropathological conditions in model systems for aging-related hearing loss [94], memory loss [95], ALS [96], and AD [97, 98]. These novel results imply that TFAM has a protective role in several pathological conditions, but no clear underlying mechanisms have been identified.

Evidence for decreased mtDNA copy number in neurodegenerative disease patients and animal models

Reduction in mitochondrial genome content has been observed in multiple human samples (HD, AD, and idiopathic PD (iPD)). In post mortem HD patient lysates, the extent to which mtDNA levels were decreased depended on the grade of disease, showing a progressive decline of 10% (Grade 2), 38% (Grade 3), or 68% (Grade 4) compared to age-matched control subjects [99]. In AD patients samples, which is the most common neurodegenerative disease, mtDNA levels are reduced by 57% in the frontal cortex brain when compared to control patients [82]. Furthermore, AD patients also appeared to have an increased number in somatic mtDNA mutations. These mutations were located in elements involved in mtDNA L-strand transcription and/or H-strand replication and were associated with reductions in the mtDNA L-strand mRNA mt-ND6 (Mitochondrially Encoded NADH: Ubiquinone Oxidoreductase Core Subunit 6) and mtDNA copy number. In micro-dissected pyramidal neurons from the hippocampus of AD patients, mtDNA levels were decreased as were associated markers of mitochondrial biogenesis [100]. iPD patients also showed 27% reduction in mtDNA levels from midbrain sections [101]. Recently, an extensive review of mtDNA sequence and abundance in aged brain provided a compelling evidence for an association between mtDNA copy number reduction to neurodegeneration in AD and Creutzfeldt-Jakob Jacob disease [87]. Interestingly, this study suggested that mtDNA mutation is not a significant contributor to mitochondrial dysfunction in those tissues.

A growing number of studies demonstrated the influence of mtDNA copy number reduction on neurodegeneration in animal models (Table 1). HD is exclusively an inherited disorder caused by a mutation in the Huntingtin protein that results in glutamine expansion in the protein [102]. R6/2 was the first HD mouse model, generated to express exon 1 of the human HD gene with around 150 glutamine codon (CAG) repeats, which causes a progressive neurological phenotype [103]. In the cerebral cortex R6/2 mice, Acevedo-Torres and colleagues found both a significant reduction of mtDNA levels of ~23% and an increase in mtDNA damage relative to controls [104]. They also demonstrated that the reduction of mtDNA is connected to mtDNA damage, which may be an early biomarker for HD neurodegeneration. Another HD mimic-transgenic mouse model is NLS-N171-82Q, which expresses the first 171 amino acids of Huntingtin with an 82-glutamine expansion. This mouse model demonstrated multiple degenerative phenotypes such as losses of motoric function, hypoactivity, and abbreviated life-span [105]. In the study of Chaturvedi et al., striatum mtDNA levels were significantly reduced by 25% in the NLS-N171-82Q HD model relative to the control group [106]. They also found that this HD model exhibited reduced expression of nuclear-encoded genes representing mitochondrial biogenesis markers, such as PGC-1α, and several Complex IV subunits.

Despite the low prevalence of neurodegeneration in normally aged mice [107], aged rodent studies remain relevant for understanding the context of genetic and idiopathic neurodegeneration. Acevedo-Torres et al. compared brain-mtDNA levels in 4-month-old mice with 17- or 24-month-old C57BL/6 mice and found that mtDNA levels were decreased in the 17 and 24-month aged mice (33% and 43% reduction, respectively,) accompanied by increased mtDNA damage in the striatum of the brain [104]. Similarly, Thomas et al. found that aging produced a marked 90% reduction in mtDNA copy number relative to the control group in the mouse brain (and to a lesser degree in the heart) [108]. Not only mtDNA levels, but also mitochondrial biogenesis signaling, is substantially reduced in aging brain and heart compared to the control group. Thus, the study of aging remains fundamental to understanding how mtDNA abundance may impact the development and progression of age-related neurodegenerative diseases.

As a caveat, studies of mtDNA copy number in human brains at various ages have yielded mixed results [109, 110]. Different regions of the brain have also been shown to exhibit significant differences in mtDNA level within a single individual [111], suggesting that sample acquisition may account for some of the variation. Alternatively, uniformity within inbred animal strains may enable the detection of age-related changes, while the genetic diversity among human subjects masks that change. Because none of the human studies are longitudinal in the same individual, we cannot account for changes within an individual during the aging process.

Evidence for TFAM expression changes in neurodegenerative disease models

TFAM protein levels are a central regulator of mtDNA copy number [65], which has multiple associations with neurodegeneration, prompting a review of TFAM levels (mostly protein levels) in association with neuropathological findings in patient samples and animals. AD is the most common neurodegenerative disease whose gross pathology involves cortical and hippocampal shrinkage, enlargement of ventricles, and frequent evidence of extracellular β-amyloid plaques or intracellular Tau neurofibrillary tangles. Sheng et al. showed that relative to control samples, the hippocampus of AD patient brains, but not the cerebellum, contained a 50% reduction in TFAM protein levels, which was concomitant with impaired mitochondrial biogenesis evidenced by decreased NRF1, NRF2 and PGC-1α protein levels [112]. β-amyloid is a cleavage product of the amyloid precursor protein (APP), which mice do not readily make. Many transgenic mice lines have been created to overcome this challenge (reviewed in [113]). A common animal model for study is the APPswe/PS1dE9 double transgenic mouse, which neuronally expresses a chimeric APP (KM670/671NL) and the mutant human presenilin 1 protein harboring a deletion in exon 9. In this mutant mouse model the production of human β-amyloid is increased [114], leading to more visible plaque formation, memory loss, and a moderate behavior phenotype [115]. APPswe/PS1dE9 mice have a 20% to 87.5% reduction in TFAM protein levels in the hippocampus compared to the control group [116, 117]. The levels of mtDNA were not determined in either of these studies and we could not find another AD mouse model that reported levels of TFAM protein.

TFAM protein levels were also measured in HD patient samples and animal models, mouse aging, and iPD patients. Kim et al. examined the TFAM protein levels in Grade 2, 3, and 4 of HD patients, which were reduced by 25, 32, and 41%, respectively, compared to age-matched control subjects [99]. In HD patients, the reduction of TFAM protein levels was accompanied by reduced nuclear mitochondria biogenesis markers, dampened number of mitochondria, altered mitochondrial morphogenesis, increased mitochondrial fission, and reduced mitochondrial fusion. In the HD transgenic mouse NLS-N171-82Q, TFAM showed a 25% reduction of TFAM mRNA levels compared to the wild-type mice along with reduced nuclear mitochondrial biogenesis transcription signatures and mtDNA [106]. Aged mice (21 months) had 60% less TFAM expression in brain preparations relative to that of 5-month old control mice [108]. Moreover, iPD patients had around 37.5% reduction in TFAM protein levels and loss of the respiratory chain complexes in individual dopaminergic neurons [101]. IPD patients also showed a significant reduction in Complex I and II subunits compared to the age-matched control group. These changes are correlated with increased ERK1/2 activation in human iPD midbrain neurons [118], which drives autophagic turnover of mitochondria [119, 120] in concert with suppression of TFAM-mediated mtDNA transcription [69] and mitochondrial protein synthesis [121]in cellular models of iPD. The collection of evidence is still indirect, but point to an alteration in mitochondrial biogenesis that reduces TFAM protein levels in some patient samples and many animal models. Whether evidence supports a central role for TFAM levels in the susceptibility to disease will be discussed in a later section.

The TFAM genetics in neurodegenerative diseases

There are several reports illustrating the association of TFAM gene mutations or polymorphisms, mostly from two TFAM mutations (rs1937 and rs2306604), and the development of PD, dementia, AD, and HD (Table 2). TFAM SNP rs2306604 is located in intron 4, but which allele is associated with each disease is still disputed. According to Gaweda-Walerych et al., the G allele is associated with PD risk [122], whereas Gatt et al. reported that the A allele is associated with PD in males [123]. Günther et al. [124] and Belin et al. [125] also reported that rs2306604A is connected with a moderate risk factor for AD. There is also a report arguing that there is no association between rs2306604 and late-onset AD [126]. For TFAM rs1937, Günther et al. observed that the G allele moderately increases risk for AD [124], while Zhang et al. reported that the C allele provides a protective role against late-onset of AD [126]. Although there are continuously new reports about TFAM genetics on the risk of neurodegeneration, it is not yet possible to conclude which allele is a specific major risk of neurodegeneration due a lack of consensus among studies. It also remains unknown whether or not either of these alleles are able to impact TFAM protein levels or function.

In vivo models of TFAM deficiency and its neurodegenerative phenotypes

A growing number of studies demonstrated the ability of ‘loss-of-function’ or ‘gain-of-function’ manipulations of TFAM expression to modify neurodegenerative phenotypes (Table 3). TFAM knockout is embryonic-lethal; therefore, several groups utilized neuron-specific knockout as ‘loss-of-function’ models of TFAM in mice. The first of such TFAM models was the mitochondrial late-onset neurodegeneration (MILON) mouse, which used CaMKIIa-CRE to disrupt TFAM in the forebrain, including the CA1 pyramidal cell layer of the hippocampus [127]. The MILON mouse brain showed a reduction of ~50% in mtDNA levels, accompanied by progressive nerve cell loss, apoptosis, and gliosis in the neocortex and hippocampus. Also, MILON mice were considerably more vulnerable to an excitotoxic challenge. The results demonstrated that a complete loss of TFAM would render neurons more vulnerable to stress, leading to neurodegenerative phenotypes.

Use of the dopamine transporter promoter to drive CRE-mediated TFAM knockout in dopaminergic neurons resulted in a Parkinson’s like neurodegenerative phenotype [128]. These mice, called MitoPark mice, appear healthy at birth, but develop progressive degeneration of dopamine neurons as adults accompanied by depletion of striatal dopamine and impaired spontaneous locomotion. Pathological analysis revealed intraneuronal inclusions, and eventually neuronal cell death [129]. Additional studies showed that TFAM depletion in the MitoPark mice led to fragmentation of the mitochondrial network, formation of large mitochondrial aggregates, and impaired transport of mitochondria to the nerve terminals in the striatum [130]. MitoPark mice also exhibit functional changes in the substantia nigra pars compacta dopamine neurons, with slow progression of Parkinsonian phenotype and motor impairment [131]. Furthermore, hyperpolarization-activated cyclic nucleotide-gated ion channel function and release of dopamine were reduced in young, presympotomatic MitoPark brain slices, indicating that altered nigrostriatal function precedes behavioral Parkinsonian symptoms. Whether partial depletion of TFAM, such as conditional induction of TFAM heterozygosity, can modify other models of neurodegenerative diseases still needs to be established.

Evidence for TFAM overexpression or TFAM enzyme replacement therapy in addressing neurodegenerative disease models

From the above discussion, it is evident that TFAM and mitochondrial genome levels frequently decline in both patient samples and animal models of neurodegeneration, and genetic TFAM deficiency can cause neurodegeneration. However, these results could also be explained by indirect effects of decreased mitochondrial biogenesis and mitochondrial respiratory dysfunction, respectively. TFAM overexpression (OE) has been used as a tool to demonstrate the importance of TFAM in preserving a healthy mitochondrial genome. Indeed, TFAM OE improved the phenotypes in models with mitochondrial disorders [132, 133]. Interestingly, TFAM OE in wild-type mice increased the mtDNA content but did not increase mitochondrial respiratory function [65]. The use of TFAM OE thus allows the assessment of direct effects of TFAM on mtDNA in aging-related disorders and in neurodegenerative disease models (Table 3). Notably, whole body TFAM OE improved hippocampal long-term potentiation as well as motor learning memory in mice at 24 months of age, associated with a reduction of lipid peroxidation and improved Complex I and IV activities in the brain [95]. Furthermore, the Interleukin 1 beta (IL-1β), which is an important mediator of the inflammatory response and mtDNA damage, was decreased in microglia [95]. In the 3xTg mouse model for Alzheimer’s disease (PSEN1 PS1M146V, APPswe, and MAPT P301L triple transgenic), TFAM OE improved cognitive function and reduced both mtDNA oxidative damage and Aβ accumulation [98]. In a cell-based model of Aβ toxicity, human SH-SY5Y neuroblastoma cell lines showed increased oxidative stress and decreased mitochondrial respiratory function, both of which were significantly improved when TFAM was overexpressed [97]. Using a cell-based model of toxicant-induced Parkinson’s disease, the inhibition of mitochondrial respiratory function by exposure to 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) was significantly reduced when TFAM was overexpressed in human SH-SY5Y cells [134]. In light of the number of studies showing that TFAM protects against diseases with oxidative stress and mitochondrial components (see [32] and above references), any intervention causing increase of TFAM levels, and therefore mtDNA stability and enhanced mitochondrial biogenesis, could be a viable approach to slow the progression of neurodegenerative diseases.

Non-genetic approaches, such as recombinant protein treatment, have also been used to test TFAM-mediated protection in neurodegenerative models. Recombinant human TFAM (rhTFAM) treatment can stimulate mitochondrial biogenesis in human cells derived from sporadic Parkinson’s disease patients [135]. PD cybrids, which carry the mtDNA of PD patients but have a cell line nucleus, showed a significant reduction in the electronic transport chain of Complex I, increased oxidative stress, respiratory impairment, and reduced mtDNA levels. These cells also spontaneously formed Lewy body inclusions, Treatment with TFAM complexed with mtDNA increased mitochondrial function as evidenced by induction of mtDNA, mtRNA, TFAM, and respiratory chain protein levels. In animal models of aging, treatment of 20-month aged mice with rhTFAM stimulated mitochondrial biogenesis and mtDNA gene expression in the absence of any apparent systemic toxicity [108]. Increases in mitochondrial oxidative metabolism were mirrored by improvements in spatial learning and memory in aged mice. In a human cell culture model of Alzheimer’s disease (PS1 P117L), cells treated with rhTFAM (but not a control form that cannot be imported into mitochondria) reduced mitochondrial reactive oxygen species and consequently, reduced mtDNA oxidative damage and prevented Aβ accumulation [98]. The protective effect of overexpressed TFAM is expected to be largely dependent on its role in maintaining the mtDNA nucleoid formation and mtDNA copy number, or in regulating mtRNA synthesis and mitochondrially-encoded protein synthesis. These findings support further investigations for a beneficial effect of rhTFAM in human aging and neurodegeneration