Mitochondrial retrograde signalling in neurological disease

Abstract

Neuronal mitochondrial dysfunction causes primary mitochondrial diseases and likely contributes to neurodegenerative diseases including Parkinson's and Alzheimer's disease. Mitochondrial dysfunction has also been documented in neurodevelopmental disorders such as tuberous sclerosis complex and autism spectrum disorder. Only symptomatic treatments exist for neurodevelopmental disorders, while neurodegenerative diseases are largely untreatable. Altered mitochondrial function activates mitochondrial retrograde signalling pathways, which enable signalling to the nucleus to reprogramme nuclear gene expression. In this review, we discuss the role of mitochondrial retrograde signalling in neurological diseases. We summarize how mitochondrial dysfunction contributes to neurodegenerative disease and neurodevelopmental disorders. Mitochondrial signalling mechanisms that have relevance to neurological disease are discussed. We then describe studies documenting retrograde signalling pathways in neurons and glia, and in animal models of neuronal mitochondrial dysfunction and neurological disease. Finally, we suggest how specific retrograde signalling pathways can be targeted to develop novel treatments for neurological diseases.

This article is part of the theme issue ‘Retrograde signalling from endosymbiotic organelles’.

1. Introduction

Mitochondria are ancient organelles present in virtually all eukaryotic cells. Mitochondria house the molecular machinery that generates the majority of cellular ATP and are central to numerous additional metabolic pathways. Mutations in mitochondrial genes were first found to cause human disease in the 1980s. These discoveries led to the characterization of primary mitochondrial diseases. Studies of patient tissue have also implicated mitochondrial dysfunction in sporadic neurodegenerative diseases. More recently, mitochondria have been shown to regulate cell fate decisions during neurodevelopment and have been implicated in neurodevelopmental diseases.

Mitochondrial retrograde signalling pathways have been characterized in proliferating cells but have only recently been described in neurons [1–3]. In this review, we briefly describe the importance of mitochondria in neurological disease, the retrograde signalling pathways that have been identified and how these relate to the nervous system. Finally, we discuss the potential for retrograde signalling pathways to be targeted as a therapeutic strategy for neurological diseases.

2. Mitochondrial dysfunction in neurological disease

(a). Primary mitochondrial diseases

Early-onset mitochondrial diseases are predominantly severe, and often caused by autosomal recessive mutations in nuclear genes. The high energy demand of the nervous system means that mitochondrial dysfunction frequently causes neurological syndromes. Examples include Leigh syndrome, which is associated with mutations in the mitochondrial oxidative phosphorylation (OXPHOS) complexes and their assembly factors [4] and Alpers–Huttenlocher syndrome, caused by mutations in the (nuclear-encoded) mtDNA polymerase [5]. Patients with these diseases can also present pathologies in other tissues with high energy demand, including cardiac and skeletal muscle.

Mitochondrial diseases diagnosed in adulthood are estimated to affect around 1 in 4300 adults, and are predominantly caused by mtDNA mutations (approx. 87% of cases) [6]. There are recognized mitochondrial syndromes that affect the nervous system caused by mtDNA mutations, such as Leber's hereditary optic neuropathy (LHON), myoclonic epilepsy with red-ragged fibres (MERRF) and mitochondrial encephalopathy, lactic acidosis and stroke-like episodes (MELAS), but many patients do not display the well-defined symptoms of these syndromes [7]. Cell- and tissue-specific retrograde signalling mechanisms may contribute to the broad phenotypic heterogeneity in primary mitochondrial disease patients.

(b). Mitochondrial dysfunction in neurodegenerative disease

Mitochondrial dysfunction is also linked to late onset neurodegenerative disease. The association between Parkinson's disease (PD) and mitochondrial dysfunction first came to light in a study of recreational drug users who had developed PD-like symptoms. It was established that they had injected 1-methyl-4-phenyl-1,2,5,6-tetrahydropyridine (MPTP) [8], whose metabolite, MPP⁺, was characterized as an inhibitor of mitochondrial respiratory complex I (NADH dehydrogenase-ubiquinone oxidoreductase) [9]. Subsequently, post-mortems of PD patients revealed deficient complex I activity in the substantia nigra pars compacta (SNpc) [10]. More recently, reduced complex I activity has been observed in the prefrontal cortex in PD patients and in PD patients with dementia [11–13].

Most cases of PD arise spontaneously; however, rare juvenile onset PD can be caused by mutations in mitochondrial quality control proteins PINK1 [14] and Parkin [15], LRRK2 [16], which is known to associate with the outer mitochondrial membrane, and the reactive oxygen species (ROS)-sensitive chaperone DJ-1 [17]. Thus, defects in a variety of mitochondrial-related functions cause monogenic forms of PD.

Alzheimer's disease (AD) is the most common cause of dementia and is characterized by cognitive deterioration and the accumulation of amyloid plaques and neurofibrillary tangles. AD patients have mitochondria with reduced diameter and surface and altered morphology [18], as well as reduced mitochondrial biogenesis [19] and loss of mitochondria in dendritic spines [20]. Deficiencies in complex I, IV (cytochrome c oxidase) and V (ATP synthase) are also reported in AD patients compared to controls [21,22].

Altered mitochondrial structure and function are also a feature of amyotrophic lateral sclerosis (ALS). A recent study of a small number of patients with mixed ALS/frontotemporal lobar degeneration showed altered mitochondrial morphology in patient brain tissue [23], consistent with changes in mitochondrial dynamics, morphology and function observed in cellular and animal models of ALS [24,25]. Mutations in the gene CHCHD10 are associated with frontotemporal dementia-ALS in several European cohorts [26,27]. CHCHD10 encodes a protein localized to the mitochondrial intermembrane space and patient-associated mutations in CHCHD10 cause mitochondrial and synaptic defects [28,29], further highlighting the potential contribution of mitochondrial dysfunction in ALS.

(c). Mitochondrial dysfunction in neurodevelopmental disorders

Williams syndrome (WS) is a developmental disorder caused by a small deletion on chromosome 7. WS patients have alterations in neurodevelopment leading to intellectual disability, increased anxiety, attention deficit hyperactivity disorder and hypersociability [30]. Through a systems biology approach, mitochondria were identified as playing a potential role in WS [31]. A candidate protein, the J-domain-containing family protein DNAJC30, which is within the 7q11.23 chromosomal region deleted in WS, was found to localize to mitochondria where it interacts with ATP synthase. The loss of DNAJC30 inhibited ATP synthase activity. Mice lacking DNAJC30 shared some of the WS neurodevelopmental phenotypes, including altered sociability and anxiety, and cells from WS patients were shown to have mitochondrial defects. Together, these findings suggest that mitochondrial dysfunction contributes to the neurological manifestations of WS.

Tuberous sclerosis complex (TSC) is a multisystem disorder caused by mutations in the genes TSC1 and TSC2. Patients with TSC have benign tumours in multiple organs, including the brain, which can result in epilepsy, autism and intellectual disability [32]. TSC1/2 are negative regulators of the mechanistic target of rapamycin (mTOR) pathway, which controls a wide range of anabolic processes, including mitochondrial biogenesis [33–36]. Neurons derived from patients with TSC2 mutations had increased mitochondria in the cell body, but mitochondria were depleted from the axon [37]. Numbers of presynaptic mitochondria were also reduced in TSC2-deficient neurons and overall, these neurons had reduced mitochondrial function. The inhibition of mTOR using rapamycin is widely beneficial in animal models of TSC and rapamycin reversed many of the mitochondrial defects in TSC2-deficient neurons. Thus, some benefits of mTOR inhibition in TSC may result from reversing the mitochondrial deficit in neurons.

In addition to these rare neurodevelopmental disorders, altered mitochondrial morphology and dynamics have been observed in schizophrenia patients and have been proposed to contribute to pathogenesis [38]. Brain imaging and post-mortem studies have also reported mitochondrial abnormalities in autism spectrum disorder (ASD) [39]. Future studies will likely reveal the contribution of mitochondria to the pathogenesis of a spectrum of neuropsychiatric disorders.

3. Mitochondrial retrograde signalling

Mitochondrial retrograde signalling is the mechanism by which mitochondria communicate their functional status to the nucleus. Upon changes in mitochondrial function, retrograde signalling pathways are activated, leading to changes in nuclear gene expression. As we discuss, mitochondrial retrograde signalling mechanisms are triggered by altered mitochondrial proteostasis, ATP, ROS and Ca²⁺, and activate or interconnect with several well-defined signal transduction pathways (figure 1). These include the mitochondrial unfolded protein response (UPR^mt), the integrated stress response (ISR), mTOR signalling and the endoplasmic reticulum unfolded protein response UPR (ER UPR). Together, these transcriptional responses act to restore mitochondrial function and maintain cellular homeostasis.

Figure 1.

Mitochondrial retrograde signalling pathways in the nervous system. Arrows indicate known targets; dashed arrow indicates that the mechanism of regulation is currently unknown. See text for details.

Mitochondrial retrograde signalling was discovered in Saccharomyces cerevisiae (budding yeast) [40], but has now been extensively characterized in higher model systems. Although clear themes have emerged, so far no single unifying retrograde mechanism has been described [1,3,41]. Here, we highlight mechanisms that have relevance to mitochondrial retrograde signalling in the nervous system.

(a). Mitochondrial generated changes in adenosine triphosphate, reactive oxygen species and Ca²⁺ trigger retrograde signalling

The vital metabolic and homeostatic roles of mitochondria mean that levels of a variety of molecular species are dependent on mitochondrial function. Several of these molecules act as proximal mitochondrial retrograde signals, including ATP, ROS and Ca²⁺. AMP kinase (AMPK)-mediated signalling is highly conserved and can be considered as a retrograde response [42–44]. AMPK is activated in response to a reduced ATP : AMP ratio, increased ROS or cytosolic Ca²⁺ and acts to restore cellular ATP levels, among other roles [44–48].

ROS are key activators of mitochondrial retrograde signalling. ROS responsive pathways regulate the expression of genes that impact mitochondrial function, such as redox homeostasis and mitochondrial biogenesis [49]. The retrograde signalling pathways activated downstream of increased ROS levels may be either protective or detrimental to mitochondrial function. Pathways activated include SIRT3 upregulation, JNK activation, AKT-mediated oestrogen receptor α activation, the stabilization and activation of the transcription factors HIF-1α or Nrf2, and the activation of NF-κB signalling [50–54].

Mitochondria play an important role in cellular Ca²⁺ buffering, and increases in cytosolic Ca²⁺ have been observed in multiple models of mitochondrial stress [55,56]. Increased cytosolic Ca²⁺ leads to the activation of the phosphatase calcineurin and the kinases CaMKK2, CaMKIV and PKC [57]. Activation of these factors in response to mitochondrial stress initiates signalling cascades that regulate the expression of OXPHOS, Ca²⁺ homeostasis and cell proliferation genes [55,56,58–61].

(b). Mitochondrial unfolded protein response and integrated stress response signalling pathways

The UPR^mt is a retrograde signalling pathway activated in response to mitochondrial proteotoxic stress, and acts to restore mitochondrial proteostasis and mitigate metabolic stress. In Caenorhabditis elegans, accumulation of unfolded proteins, defective mitochondrial protein import, oxidative stress, reduced OXPHOS and mtDNA depletion all activate the UPR^mt [41]. Key to the UPR^mt in C. elegans is the transcription factor ATFS-1, which is normally localized to mitochondria, but targeted to the nucleus upon loss of mitochondrial proteostasis [41]. ATFS-1 and the UPR^mt activate a transcriptional response involving the upregulation of genes involved in mitochondrial proteostasis, such as mitochondrial chaperones and proteases.

Recently, it has been shown that the mammalian transcription factor ATF5 is functionally related to ATFS-1, and can act downstream of the mammalian UPR^mt [62]. In paraquat-treated HEK 293T cells, a model of mitochondrial stress caused by excessive ROS production, increases in expression of LON protease (LONP) and the mitochondrial chaperones Hsp60 and mtHSP70 were significantly reduced by knockdown of ATF5. Similar ATF5-dependent changes in LONP expression were also observed in HEK 293T cells treated with various OXPHOS inhibitors. Alongside ATF5, the transcription factor CHOP is a key target of the mammalian UPR^mt [63], although how ATF5 and CHOP interrelate is poorly understood.

The ISR is a cellular stress response pathway regulated by the activity of four kinases: protein kinase R (PKR), (PKR)-like endoplasmic reticulum kinase (PERK), general control non-derepressible 2 (GCN2) and haem-regulated inhibitor (HRI) [64]. Once activated, these kinases phosphorylate the translation initiation factor eIF2α. Phosphorylated eIF2α inhibits global protein translation, with the exception of transcripts that contain upstream open reading frames in their 5′UTR, such as the bZIP transcription factors ATF4 and ATF5.

Mitochondrial stress activates the ISR via PERK, GCN2 or PKR, although the specific kinase activated is context-dependent. In multiple cellular models, rotenone treatment activates the ISR via PERK, suggesting mitochondrial dysfunction can also induce endoplasmic reticulum (ER) stress [65,66]. In HeLa cells treated with doxycycline, a mitochondrial translation inhibitor, the ISR was activated via GCN2 [67]. Interestingly, despite an increase in CHOP expression, there was no UPR^mt gene activation in this model. Conversely, in mouse intestinal epithelial cells expressing a mutant form of the mitochondrial enzyme ornithine transcarbamylase (OTC-Δ), induction of the UPR^mt target Hsp60 was dependent on the PKR pathway of the ISR [68].

Recent multi-omics analysis of HeLa cells treated with a variety of mitochondrial stressors, including actinonin and FCCP, identified ATF4 as the main transcriptional regulator of the mitochondrial stress response. Knockdown of the four ISR kinases individually had no impact on ATF4 expression [69]. Knockdown of eIF2α, however, reduced ATF4 levels in these models, suggesting ATF4 is still activated downstream of the ISR. Further research is therefore required to identify how the ISR is regulated in response to mitochondrial stress.

Recent developments in yeast models have revealed additional mechanisms by which mitochondria regulate cytosolic proteostasis. In budding yeast, defects in mitochondrial protein import trigger two post-translational stress response mechanisms termed the mitochondrial precursor overaccumulation stress (mPOS) and the UPR activated by mistargeting of proteins (UPRam). mPOS and UPRam reduce cytosolic translation and increase ubiquitin proteasome activity to counter the accumulation of mis-folded proteins in the cytosol caused by defective mitochondrial import [70,71]. If analogous mechanisms exist in metazoans, they could enable mitochondria to regulate cytosolic translation in concert with the ISR [72].

In summary, mitochondrial dysfunction triggers impaired mitochondrial proteostasis, reduced ATP, increased ROS and increased cytosolic Ca²⁺ leading to activation of intracellular signalling pathways targeting specific transcription factors. These transcription factors communicate and act in synchrony to alter nuclear transcription and restore mitochondrial and cellular homeostasis. These pathways are more diverse in complex organisms, likely due to the requirements of different tissues to respond to and cope with mitochondrial stress.

4. Mitochondrial retrograde signalling in the nervous system

(a). Neuronal mitochondrial retrograde signalling in response to oxidative phosphorylation inhibition

Loss-of-function mutations in subunits of mitochondrial respiratory complex I are the most common cause of the primary mitochondrial diseases Leigh syndrome and Leber hereditary optic neuropathy [4,73]. The inhibition of complex I in dopaminergic neurons using the pharmacological agents rotenone or MPTP (MPP⁺) produces parkinsonian symptoms [74]. The neuronal retrograde response to complex I dysfunction has been investigated in both cellular and in vivo models. A microarray analysis of mouse cortical neurons treated with rotenone for up to 24 h showed changes in expression of genes involved in processes including mitochondrial biogenesis, redox regulation, Ca²⁺ signalling, cytosolic proteostasis and autophagy [75]. Although after 8 h of rotenone treatment, many of the genes involved in mitochondrial biogenesis and proteostasis were upregulated, the majority were downregulated by 24 h of treatment. This biphasic transcriptional response suggests that although acute activation of retrograde signalling may be beneficial, chronic activation may be detrimental to neuronal health. Similarly, in cultured rat neurons treated with rotenone for 7 days, both protective and detrimental transcriptional changes were observed [76].

Signalling pathways regulating the transcriptional response to complex I inhibition have been identified in specific neuronal and glial cell types. Complex I inhibition activates the PERK branch of the ER stress response and the ISR in PC12 cells, SH-SY5Y neuroblastoma cells, human dopaminergic neurons, human oligodendroglia and rat retinal cultures [65,66,77–79]. The role that ISR-mediated ATF4 activation plays in response to complex I inhibition differs between models. In oligodendroglia and rat retinal glia, knockdown of PERK and inhibition of ER stress, respectively, significantly reduced cell death after 24 h of rotenone treatment [65,66]. However, in SH-SY5Y cells, knockdown of ATF4 significantly increased cell death after 12 h of rotenone treatment [79]. Further investigation is required to determine whether the different protective and detrimental roles of ISR signalling in these models is dependent on the cell type or the duration of complex I inhibition. Complex I inhibition and dysfunction also activate the inositol requiring enzyme 1 (IRE1) and activating transcription factor 6 (ATF6) ER stress pathways, inhibit mTOR signalling via CaMKII, activate mTOR signalling via alternative pathways and reduce AKT signalling in the nervous system of different cellular and mouse models [60,66,77,80–82].

Succinate dehydrogenase-ubiquinone oxidoreductase (complex II) is a component of both the mitochondrial electron transport chain and the tricarboxylic acid (TCA) cycle. Chronic inhibition of complex II using 3-nitropropionic acid (3-NP) induces striatal degeneration and symptoms of Huntington's disease in rodents and non-human primates [83]. Both detrimental and protective retrograde signalling pathways have been identified in response to complex II inhibition in the nervous system. In rats chronically treated with 3-NP, striatal neurodegeneration is induced by Jun amino-terminal kinase (JNK) signalling and the activation of the transcription factor c-Jun [84]. Increased activity of the Ca²⁺-dependent protease calpain has also been observed in 3-NP-treated rats, and contributes to striatal degeneration [85]. Calpain is known to regulate the activity of multiple signalling cascades including GSK3-β and mTOR-mediated signalling [86]. 3-NP-treated rats also have reduced phosphorylation of the kinases AKT and GSK3-β and the transcription factor CREB, and increased JAK/STAT signalling in the striatum, which contribute to neurodegeneration [87]. Conversely, protective retrograde responses to complex II inhibition have also been observed and include upregulation of NRF2-mediated antioxidant gene expression in cultured astrocytes and neurons [88,89].

Mutations in the complex IV assembly factors SURF1 and LRPPRC inhibit activity of the complex and cause Leigh syndrome [4]. Ubiquitous Surf1 knockdown in Drosophila caused reduced activity of all OXPHOS complexes, pupal lethality, and altered the expression of greater than 5000 genes, reducing the expression of glycolytic, TCA cycle, OXPHOS and mitochondrial protein import genes [90]. Neuronal-specific knockdown of Surf1 in Drosophila produces adult flies with a severe motor deficit. Knockdown of the hypoxia responsive transcription factor HIF-1α or inhibition of Ras–ERK–ETS signalling rescued the motor deficit in these flies, suggesting that both HIF-1α and Ras–ERK–ETS signalling are activated and are detrimental in response to complex IV deficiency [91,92]. HIF-1α is also activated in the brain of SURF1 knockout mice, suggesting that activation of HIF-1α is a conserved response to complex IV inhibition [93].

In SH-SY5Y neuroblastoma cells, knockdown of LRPPRC, which is involved in the translation of complex IV subunits, transiently activates the UPR^mt, as demonstrated by an increase in expression of Hsp60, Hsp70 and the mitochondrial protease CLpP [94]. The expression of these proteins correlated with the imbalance between nuclear and mitochondrial-encoded complex IV subunits, suggesting that the UPR^mt is a protective proteostatic mechanism in this context.

Together, these studies demonstrate that different OXPHOS complexes regulate unique patterns of retrograde signals within the nervous system (figure 1). These signals elicit transcriptional responses that have differential effects on neuronal and glial function. Many neurological diseases display defective activity in multiple OXPHOS complexes; comprehending how the different retrograde networks are inter-regulated is therefore required to fully understand the contribution of mitochondrial dysfunction to these different diseases.

(b). Mitochondrial retrograde signalling in neurodegenerative disease

In addition to impaired OXPHOS, other aspects of mitochondrial organelle biology, such as mitochondrial fission/fusion, mitophagy and mitochondrial proteostasis, are compromised in neurodegenerative disease and can activate retrograde signalling (figure 2).

Figure 2.

Overview of the retrograde signalling pathways implicated in neurodevelopmental disorders and neurodegenerative diseases. TSC, tuberous sclerosis complex; WS, Williams syndrome; ASD, autism spectrum disorder; ALS, amyotrophic lateral sclerosis. See text for details.

Mutations in the mitochondrial quality control genes PINK1 or Parkin cause juvenile and early-onset forms of PD. In Drosophila, mutations in PINK1 or Parkin activate the PERK branch of the ER stress response pathway [95]. PERK inhibition prevented dopaminergic neuron loss in these fly models, suggesting that this pathway contributes to disease pathogenesis. Knockdown of HIF-1α, Ras–ERK–ETS signalling or the Hsp90 chaperone TRAP1 are also protective in Drosophila PINK1 or Parkin mutants [91,92,96]. Thus, multiple detrimental signalling responses are activated in response to the loss of PINK1 or Parkin expression. Interestingly, activation of HIF-1α has also been observed in PINK1 knockout mice and primary neurons, where it was responsible for upregulating glycolytic genes, indicating that HIF-1α activation is a conserved response to the loss of PINK1 in neurons [97].

Given the multifactorial nature of AD aetiology, isolating the signalling events that are caused by mitochondrial dysfunction is challenging. However, current evidence is consistent with a role for retrograde signalling in AD. The UPR^mt transcriptional response was observed in AD patient brain tissue, suggesting activation of this mitochondrial stress response pathway [98]. Both sporadic and familial AD patients had increased mRNA expression of the UPR^mt genes Hsp10, Hsp60 and CLpP in the prefrontal cortex. Activation of the ISR, altered AKT and mTOR signalling, and reduced levels of HIF-1α have also been documented in AD patients [99–101].

In models of ALS, a potential role for the UPR^mt has been uncovered in response to mutant SOD1, TDP-43 and FUS expression. In G93A-SOD1 mutant mice, age- and sex-dependent increases in the expression of CHOP, Hsp60 and the mitochondrial serine protease Omi were observed [102]. Expression of A315T-mutant TDP-43 or P525 L-mutant FUS in HEK293 cells or Drosophila increased mRNA levels of Hsp60, mtHsp70 and LonP [23,103]. In HEK293 cells, these increases were associated with an upregulation of ATF5. Interestingly, the temporal regulation of UPR^mt genes differs between A315T-mutant TDP-43 and P525 L-mutant FUS cellular models, suggesting that independent retrograde signalling pathways are activated. Surprisingly, in Drosophila expressing P525 L-mutant FUS, knockdown of various UPR^mt genes in the eye reduced retinal degeneration, suggesting that activation of the UPR^mt contributes to pathology [103].

Determining whether these signalling mechanisms activated in neurodegenerative diseases are a direct response to mitochondrial dysfunction and their potential as therapeutic targets is a future challenge.

(c). Mitochondrial retrograde signalling in neurodevelopmental disorders

The understanding that mitochondria play a causative role in neurodevelopmental disorders is a recent discovery. Although it is still early days, excitingly, several studies have already shown that retrograde signalling is activated in neurodevelopmental disorder models (figure 2).

Mitochondrial dysfunction is increasingly recognized as a pathological feature of ASD, TSC and WS. Recent evidence suggests a role for mitochondrial retrograde signalling in the pathogenesis of ASD. Knockdown of IMMP2L, a mitochondrial protease associated with ASD, in astrocytes, led to altered expression of 38 genes, including neurodevelopmental transcription factors [104]. The expression of transcription factors such as JARID2, ZFXH3 and SATB2, involved in transcriptional regulation in the brain, were mis-regulated in IMMP2L knockdown astrocytes. These transcription factors are thus potential targets of mitochondrial retrograde signalling in ASD.

Mutations in DNM1L, encoding the mitochondrial fission protein dynamin-1-like protein (DRP1), are associated with the development of severe neurological defects including epileptic encephalopathy, developmental delay and optic atrophy during infancy or childhood [105,106]. Knockout of DNM1L in the forebrain of mice led to the induction of the ISR and upregulation of the cytokine FGF21 [107]. However, there was no change in the mRNA levels of UPR^mt markers Hsp10, CLpP or Ymel1, and no increase in Hsp60 protein levels in the brains of DNM1L knockout mice. Future investigation is therefore required to identify the cellular changes mediated by the ISR in response to the loss of DRP1.

Impaired mitochondrial dynamics are also associated with TSC, WS and schizophrenia [31,37,108]. Interestingly, disrupting mitochondrial dynamics during neurodevelopment impacts stem cell fate and neurogenesis via mitochondrial retrograde signalling. Knockout of the mitochondrial fusion genes MFN1/2 or knockdown of the mitochondrial fusion protein OPA1 led to a reduction in neural stem cell self-renewal and increased commitment to a progenitor state in the developing mouse cortex [109]. Conversely, the loss of DRP1 led to an increase in neural stem cell self-renewal. The loss of MFN1/2 caused increased ROS and activated an NRF2-mediated retrograde signalling mechanism, leading to upregulation of differentiation, Notch signalling and redox genes. The loss of NRF2 reduced the number of embryonic neural stem cells and the expression of pro-neural transcription factors. Inducible knockout of MFN1/2 in the adult neurogenic niche caused depletion of neural stem cells, learning and memory deficits. Whether mitochondrial retrograde signalling via NRF2 is implicated in neurodevelopmental diseases associated with defective mitochondrial fusion is yet to be explored.

5. Targeting mitochondrial retrograde signalling as a therapy in neurological disease

There is now good evidence that targeting mitochondrial retrograde signalling mechanisms can improve neuronal function and prevent neurodegeneration [3]. Targeting retrograde signalling via knockdown of HIF-1α restores neuronal function and lifespan in Drosophila models of mitochondrial disease and PD [91]. In a mouse model of PD, knockout of the UPR^mt/ISR target CHOP was neuroprotective, while in a mouse model of hearing loss, inhibition of retrograde signalling by reducing the expression of the transcription factor E2F2 improved hearing [110,111]. These pioneering studies indicate that manipulation of mitochondrial retrograde signalling can provide functional improvement in animal models of neurological disease.

Below we describe two signalling pathways, the ER UPR and mTOR signalling, that have strong links to mitochondrial dysfunction and can be targeted pharmacologically. We discuss how these pathways could be exploited to target mitochondrial retrograde signalling as a treatment for neurological disease.

A signalling pathway with clear therapeutic promise is the ER UPR. The ER UPR is activated by loss of proteostasis in the ER and consists of three pathways that direct a concerted transcriptional response: the PERK pathway, the IRE1 pathway and the ATF6 pathway. Activation of the three branches of the ER UPR leads to the upregulation of the transcription factors ATF4, spliced X-box binding protein 1 (XBP1) and the DNA-binding fragment of ATF6 (ATF6f) [112]. These transcription factors regulate the expression of genes involved in ER protein degradation, protein folding and lipid synthesis, and work together to reduce ER stress or trigger apoptosis.

The PERK pathway is chronically activated in patients with and/or animal models of dementia, PD, prion disease and ALS, and pharmacological inhibition of the PERK pathway significantly reduces neurodegeneration in animal models of these diseases [113]. The ER UPR is activated by loss of proteostasis in the ER lumen. However, the mis-folded and aggregated proteins in AD, PD and ALS are either cytoplasmic or extracellular and whether they cause proteotoxic stress in the ER lumen of neurons is unclear. Thus, protein aggregation may not be the cause of ER UPR activation in neurodegenerative disease. Given that mitochondrial dysfunction activates eIF2α and ATF4, as well as the IRE1 and ATF6 UPR pathways [69,77,80,114], and that mitochondrial dysfunction is evident in most neurodegenerative diseases, it is conceivable that ER UPR activation is due to mitochondrial dysfunction. Mitochondria and the ER are physically and functionally linked, a coupling that is vital for maintenance of multiple cellular functions. ER–mitochondria contacts, or mitochondria-associated membranes (MAMS), are responsible for regulating cellular Ca²⁺ distribution, lipid synthesis, mitochondrial fission and fusion, and autophagosome formation [115]. Alterations in processes mediated by ER–mitochondria coupling, including the number of MAMS and cellular Ca²⁺ distribution, lead to ER stress and activation of the ER UPR [116–118]. We propose that the activation of the ER UPR observed in many neurodegenerative diseases is, at least in part, a result of perturbed mitochondrial function leading to failure of shared ER–mitochondrial functions and ER stress (figure 1). A recent study in a Drosophila model of mitochondrial stress supports this model [114]. Thus, activation of the ER UPR in neurodegenerative disease models may be a mitochondrial retrograde response. Pharmacological agents that modulate the ER UPR and are beneficial in neurodegenerative models may therefore be targeting mitochondrial dysfunction as the root cause.

mTOR signalling is a key regulator of cellular anabolic processes including protein synthesis, lipid metabolism, mitochondrial biogenesis and autophagy. The importance of mTOR signalling in neurological disease has become apparent with the demonstration that rapamycin treatment alleviates pathology in both neurodevelopmental and neurodegenerative diseases associated with mTOR activation [119]. In the Ndufs4 knockout mouse model of Leigh syndrome, which causes encephalopathy and death around seven weeks of age, rapamycin dramatically improved locomotor performance and extended lifespan [81]. mTOR signalling was activated in the brain in Ndufs4 mutant mice [81], but the mechanism was not determined. Recently, mTOR signalling was also shown to be activated in a mouse model of mitochondrial myopathy, using mice carrying a dominant mutation in the replicative helicase Twinkle, which accumulate deletions in mtDNA in muscle [120]. In this model, ATF4 was also activated. Based on previous studies, it was proposed that ATF4 expression was increased downstream of mTOR activation [121], and this could potentially occur via the translation initiation factor eIF4E and its regulator 4E-BP [122], which is a direct substrate of mTOR complex I. It is currently unclear how mitochondrial dysfunction activates mTOR, but if this is via a retrograde signalling mechanism, then targeting this mechanism could have potential as a therapy for a range of neurological diseases. Thus, it will be important in future to establish the mechanism by which mTOR signalling is activated by mitochondrial dysfunction in the nervous system (figure 1).

6. Conclusion

In summary, we are now beginning to understand mitochondrial retrograde signalling mechanisms in the nervous system and their role in pathology. Applying this understanding in the context of the exquisite complexity of the brain will be challenging, but has the potential to reveal new insight that is applicable to a broad spectrum of neurological diseases.

Data accessibility

This article has no additional data.

Authors' contributions

L.G., R.J.H. and J.M.B. wrote the manuscript. J.M.B. edited the manuscript.

Competing interests

We declare we have no competing interests.

Funding

Work in the Bateman lab is funded by Alzheimer's Research UK (grant no. ARUK-IRG2017A-2). L.G. is supported by the UK Medical Research Council (grant no. MR/N013700/1) and King's College London MRC Doctoral Training Partnership in Biomedical Sciences; R.J.H. was funded by a PhD studentship from the Guy's and St Thomas’ charity.