Mitochondrial dysfunction, aging, and the mitochondrial unfolded protein response in Caenorhabditis elegans

Abstract

We review the findings that establish that perturbations of various aspects of mitochondrial function, including oxidative phosphorylation, can promote lifespan extension, with different types of perturbations acting sometimes independently and additively on extending lifespan. We also review the great variety of processes and mechanisms that together form the mitochondrial unfolded protein response. We then explore the relationships between different types of mitochondrial dysfunction-dependent lifespan extension and the mitochondrial unfolded protein response. We conclude that, although several ways that induce extended lifespan through mitochondrial dysfunction require a functional mitochondrial unfolded protein response, there is no clear indication that activation of the mitochondrial unfolded protein response is sufficient to extend lifespan, despite the fact that the mitochondrial unfolded protein response impacts almost every aspect of mitochondrial function. In fact, in some contexts, mitochondrial unfolded protein response activation is deleterious. To explain this pattern, we hypothesize that, although triggered by mitochondrial dysfunction, the lifespan extension observed might not be the result of a change in mitochondrial function.

Introduction

Mutants confirm a link between mitochondria and the aging process

A relationship between the mitochondria and the aging process is a long-standing postulate of aging research because mitochondria are the main source of cellular ATP, a source of reactive oxygen species (ROS), and possess their own DNA with different, possibly more fallible, replication, and repair processes. The mutant approach to the study of aging also quickly identified a link to mitochondrial function. The first 2 long-lived mutants that were found to lead to increased lifespan were age-1 and daf-2, which both act in the same pathway (Dorman et al. 1995), the insulin growth factor-like pathway (Kimura et al. 1997), which is not mitochondrial. However, the third long-lived mutant identified was clk-1 (Wong et al. 1995), which encodes a mitochondrial hydroxylase (Ewbank et al. 1997) that is necessary for ubiquinone (UQ; coenzyme Q) biosynthesis (Jonassen et al. 2001; Miyadera et al. 2001) (Fig. 1). UQ has many functions but its most crucial is as an obligate electron carrier in the mitochondrial respiratory chain (Wang and Hekimi 2016). clk-1 mutants make no endogenous UQ₉ (the subset refers to the length of the isoprenoid sidechain) but are alive because they are able to absorb bacterial UQ₈ (Jonassen et al. 2001). clk-1 mutants nonetheless display unmistakable signs of mitochondrial dysfunction. A particularly relevant early finding was that the mutants lose the ability to uptake an innermembrane potential-dependent mitochondrial dye relatively early in adulthood (Felkai et al. 1999). Thus, clk-1 mutants are long-lived despite early and worsening mitochondrial dysfunction! This was the first indication that mitochondrial dysfunction could lead to increased lifespan.

Fig. 1.

Clk-1 phenotypes are due to a lack of UQ₉. CLK-1 encodes a hydroxylase that functions in the mitochondria to convert demethoxyubiquinone (DMQ₉) into 3-hydroxy ubiquinone (3-hydroxy-UQ₉) which another enzyme converts to ubiquinone (UQ₉) (Ewbank et al. 1997; Jonassen et al. 2001; Miyadera et al. 2001). clk-1 mutants, which lack UQ₉, require UQ derived from their bacterial food source (UQ₈) for survival (Jonassen et al. 2001). Although sufficient for survival, exogenous UQ₈ fails to allow for a wildtype phenotype. Thus, endogenous production of UQ is required. The natural precursor of the aromatic ring of UQ is believed to be 4-hydroxybenzoate (4-HB). However, 2,4-dihydroxybenzoate (2,4-DHB), a hydroxylated analog of 4-hydroxybenzoate (4-HB), can serve as an alternative, unnatural, precursor of UQ synthesis that allows for the biosynthesis of UQ in the absence of CLK-1 (Liu et al. 2017)_. 2,4-DHB rescues all Clk-1 phenotypes, including slow development, behaviors, and aging, suggesting that all Clk-1 phenotypes are indeed due to a lack of UQ₉ (Liu et al. 2017). Created with BioRender.com.

Mitochondrial dysfunction can lead to increased lifespan

A variety of interventions and situations that lead to mitochondrial dysfunction can increase lifespan (Fig. 2).The initial findings with clk-1 were followed by the identification of isp-1(qm150), a mutation in the iron sulfur protein subunit of mitochondrial complex III (Feng et al. 2001). This point mutation leaves a functional electron transport chain (ETC), but one that works at a lower rate and results in low oxygen consumption and an increased lifespan, as well as in pleiotropic phenotypes characterized by slow development, behavior, and reproduction. isp-1 can be partially suppressed by a compensatory point mutation in ctb-1(qm189), the mitochondrially encoded subunit of complex III. The findings with isp-1 mutants, therefore, link longevity not only to mitochondrial function in general, as clk-1 does, but very specifically to ETC function.

Fig. 2.

Mitochondrial dysfunction can lead to longevity. Mutations and RNAi knockdown of several mitochondrial genes can increase longevity. Data from Wong et al. (1995), Feng et al. (2001), Dillin et al. (2002), Lee et al. (2003), Kayser et al. (2004), Yang and Hekimi (2010b), Van Raamsdonk and Hekimi (2011), Houtkooper et al. (2013), Bennett et al. (2014), and Rauthan et al. (2015). Created with BioRender.com.

RNA interference (RNAi) was also used to reveal a link between mitochondrial function and lifespan (Dillin et al. 2002; Lee et al. 2003; Houtkooper et al. 2013). Lee et al. showed that a great variety of genes coding for mitochondrial proteins could be targeted by RNAi knock-down to result in increased lifespan. Dillin et al. showed that the ETC subunits or ATP synthase had to be knocked down during development in order to increase lifespan. Similarly, Houtkooper et al. showed that the lifespan extension can also be obtained by inhibiting translation of mitochondrial proteins during development, by knock-down of mitochondrial ribosomal proteins.

Light was shed on the relationship between long-lived mitochondrial mutants and the longevity effects of RNAi against mitochondrial ETC proteins in a study that also reported the identification of another ETC point mutant, nuo-6(qm200), this time in a subunit of complex I (Yang and Hekimi 2010b). The 2 mutants (isp-1 and nuo-6) were found to be alike in several ways, yet to produce a phenotype that is drastically different from that produced by RNAi against either of the 2 genes. The mutants and RNAi worms are different in anatomy, behavior, stress resistance, gene expression, and mitochondrial function. When the mutations are combined in double mutants they are only slightly additive for most phenotypes and not additive at all for lifespan. In contrast, for lifespan, RNAi against either isp-1 or nuo-6 is fully additive to the mutation in the other gene of the pair, thus allowing for very long lifespans in the RNAi-treated mutants (Fig. 3). This additivity is not the result of having found 2 different ways to affect the same process, thus impairing the process more severely producing a more severe phenotype in this way. Indeed, when the severity of RNAi treatments against the wildtype is increased beyond a certain point, lifespan is not increased further. In fact, even more severe treatments lead only to lethality (Yang and Hekimi 2010b). Thus, these findings very strongly suggest that there are distinct types of mitochondrial dysfunction, but also, more surprisingly, that different types of mitochondrial dysfunction extend lifespan by mechanisms that are sufficiently distinct to be able to act additively.

Fig. 3.

Mitochondrial dysfunction caused by nuo-6(qm200) mutants extends lifespan by a mechanism that is distinct from that of nuo-6(RNAi) knockdown. Data are taken from Yang and Hekimi (2010b). a) Both nuo-6(qm200) and nuo-6(RNAi) mutants extend lifespan, but all other phenotypes are distinct. The phenotype of isp-1(qm150) is very similar to that of nuo-6 (Feng et al. 2001), and RNAi knockdown of other subunits of the mitochondrial ETC exhibit phenotypes similar to nuo-6(RNAi) (Dillin et al. 2002; Lee et al. 2003). b) The longevity phenotypes of nuo-6(qm200) and isp-1(qm150) mutants are not additive to each other suggesting that they extend lifespan by a similar mechanism. c) The longevity phenotype of nuo-6(qm200) is additive to that of isp-1(RNAi) and the longevity phenotype of nuo-6(RNAi) is additive to that of isp-1(qm150). This suggests that the mutants extend lifespan by a mechanism that is distinct from RNAi knockdown of the same genes. See main text for further discussion. Created with BioRender.com.

What can we say about the mitochondrial dysfunction associated with long-lived mitochondrial mutants compared with that produced by RNAi against mitochondrial proteins? The consequences of the inhibition by RNAi of the production of 1 or several subunits of large and highly constrained mitochondrial complexes are expected to be very different from those of incorporating mutant subunits with functions altered from single amino acid changes. One expects depressed production of 1 subunit to lead to the assembly of fewer functional mitochondrial complexes, but the assembled complexes should have wildtype function. One also expects such mitochondria to contain unfolded subunits (because these subunits could not be incorporated into a complex), as well as misassembled complexes (Yoneda et al. 2004). Animals treated by RNAi against isp-1 and nuo-6 show the overall phenotype that is typical of other RNAi treatments against subunits of mitochondrial complexes. That is, they are smaller, more transparent, and less fertile than the wildtype (Yang and Hekimi 2010b). However, as mentioned, mutations in these genes produce a very different phenotype. Furthermore, increased expression of hsf-1 (heat shock factor) and hsp-6 and hsp-60 [part of the mitochondrial unfolded protein response (UPRmt)] was observed in nuo-6(RNAi) and isp-1(RNAi) animals but not in the mutants. Similarly, greatly increased expression of the GFP::LGG-1 marker of autophagy was observed in RNAi-treated animals but not in the mutants. The increase in these responses (heat shock, UPRmt, autophagy) in RNAi-treated worms but not mutants is consistent with the hypothesis that normal complex assembly is disrupted after RNAi, but that isp-1 or nuo-6 mutant subunits can be more or less normally incorporated into the irrespective complexes and therefore do not trigger these stress responses, or not to the same degree. Also, while oxygen consumption is lower in the mutants, it is not affected in RNAi-treated animals when normalized to body size, as was observed previously for RNAi knockdown of other subunits (Lee et al. 2003). This suggests that there is a developmental adjustment to reduced mitochondrial function. The fact that this adjustment does not occur in isp-1 and nuo-6 mutants despite very severe losses of, respectively, complex III and complex I activity (Yang and Hekimi 2010b) is one more indication that the cellular consequences of the 2 modes of dysfunction are quite different.

gas-1(fc21) is a mutation in a subunit of mitochondrial complex I that leads to a severely shortened lifespan under standard conditions (Kayser et al. 2004). At first sight it is surprising that gas-1 mutants should be short-lived while their mitochondrial defects and their other phenotypes are strongly reminiscent of those of isp-1 and nuo-6 mutants. The gas-1 mutation was initially identified not for its effect on lifespan but as a mutant with altered sensitivity to volatile anesthetics (Kayser et al. 1999). However, an increased lifespan was also not used for the initial identification of isp-1 and nuo-6. The isp-1 and nuo-6 mutants were identified for having clk-1-like phenotypes that are not maternally rescued, i.e. F2 mutants following mutagenesis that presented with slow development and slow defecation (Feng et al. 2001; Yang and Hekimi 2010b). It seems astonishing that such simple criteria are sufficient to specifically obtain point mutations in mitochondrial respiratory chain subunits. It could be interesting to determine if other short-lived ETC mutations could be obtained by further screens for resistance to volatile anesthetics. However, there is a twist in the gas-1 story. Numerous researchers use the base analogue 5-fluoro-2′-deoxyuridine (FUDR) to prevent DNA replication and thus progeny production to facilitate the manipulations involved in scoring lifespan. FUDR does not affect the lifespan of wildtype worms (Mitchell et al. 1979), but it dramatically lengthens the lifespan of gas-1 mutants in a dose-dependent manner: from a severely shortened lifespan in untreated mutants to a lifespan that is 50% longer than that of wildtype worms and 2 times as long as that of untreated mutants (Van Raamsdonk and Hekimi 2011). Why this happens remains to be investigated but has the potential for a significant insight into the mechanisms by which mitochondrial mutants increase lifespan.

The pathway of lifespan extension of mitochondrial mutants is genetically distinct from other pathways

The 1998 paper in which the longevity by caloric restriction of Eat mutants was described also examined epistatic interaction among mutants affecting lifespan (Lakowski and Hekimi 1998). In particular, it was observed that the short-lived loss-of-function daf-16(m26) mutation (encoding a FOXO transcription factor), which fully suppresses the increased lifespan of insulin/IGF-like receptor daf-2 (Kenyon et al. 1993), was not suppressing the lifespan of the long-lived mutations clk-1(e2519) or eat-2(ad465), suggesting, for the first time, the possibility of the existence of genetically distinct pathways affecting lifespan. An interesting sort of additivity was also observed in clk-1; eat-2 double mutants, again suggesting that different mechanisms are at play in the 2 mutants. Indeed, the double mutant aging curve, which first hugs the eat-2 curve and then the clk-1 curve, suggests that eat-2 prevents early death and clk-1 lengthens the lifespan of the longest living worms. Although these early findings are robust, the picture concerning dietary restriction has become much more complicated by findings that indicate that different types of intervention that reduce the caloric intake of worms might act through different molecular pathways (Greer and Brunet 2009; Honjoh et al. 2009), and that the daf-2/daf-16 pair might exert their influence on lifespan by modulating food intake (Wu et al. 2021).

In addition to the findings with clk-1, it was shown that daf-16 also fails to suppress isp-1 longevity (Feng et al. 2001). That is, although isp-1; daf-16 double mutants do not live as long as isp-1 mutants, they live substantially longer than either the wildtype or daf-16 mutants, which are short-lived. Thus, the mitochondrial dysfunction in isp-1 does not strictly require DAF-16, rather it increases longevity by daf-16-independent mechanisms (by a pathway specific to mitochondrial dysfunction), even though the loss of DAF-16 might somewhat impair its ability to fully express its longevity. Similarly, lifespan extension by knockdown of ETC subunits as well as knockdown of other genes required for mitochondrial function are not suppressed by daf-16 (Dillin et al. 2002; Lee et al. 2003). However, others have made different observations, which they interpret to suggest that daf-16(+) may in fact be required for the longevity of mitochondrial mutants. For example Senchuk et al. (2018) found that mutation of daf-16 strongly suppresses the increased lifespans of both clk-1 and isp-1. How to reconcile these discrepancies is the motivation for the next section, which is an extended note on the interpretation of suppressive interactions.

How interpretable are suppressive interactions to establish mechanisms of aging?

Carefully interpreting suppressive interactions on quantitative phenotypes, especially survival, is difficult (Fig. 4). One difficulty arises from the fact that the organism mostly needs all its genes to function properly. This means that most mutations are likely to shorten survival time under all, or some, environmental or genetic conditions. This is clearly exemplified by the fact that no long-lived mutant could compete efficiently with the wildtype in a natural environment (or even develop and reproduce at all). For example, clk-1 develops slowly, has a small brood size, and dislikes higher temperatures; eat-2 is chronically starved with, among other things, a delayed production of progeny; and daf-2 mutants constitutively become dauer larvae at high temperature. As to daf-16 mutants, they are short-lived by themselves. Thus, 1 risk when observing suppression of a long lifespan is that the combination of 2 mutations that independently impair homeostasis might shorten lifespan in a rather nonspecific way. Another possibility is that modulation of each of the 2 pathways or processes might be promoting longevity by independent downstream mechanisms, but that this cannot be observed when these pathways are codependent, that is, they require each other’s integrity to generate the longevity outcome. Some criteria can be used to overcome this difficulty. For example, it seems more likely that there is a specificity in the suppression when other phenotypes, distinct from the lengthened survival, are suppressed as well. For example, as they should, missense tRNA suppressors of clk-1(e2519) suppress most or all mutant phenotypes (Branicky et al. 2006). daf-16 suppresses all daf-2 phenotypes (Kenyon et al. 1993), and several ced genes that suppress the long lifespan of isp-1 mutants also partially suppress their developmental and behavioral phenotypes (Yee et al. 2014). However, there are some caveats to the analysis of double mutants when only the longevity part of the phenotype of the longevous mutant is suppressed in the double mutant combination (and no other phenotype). On the one hand, it could mean that an interaction that is particularly specific for longevity has been revealed, i.e. the suppressor mutation might point to the mechanism responsible for the longevity. On the other hand, there is the risk that the loss of longevity is due to an additive loss of general homeostasis. In other words, when a double mutant strain is sick in 2 different ways it might just be too sick to express its potential longevity.

Fig. 4.

Suppression analysis is an indispensable tool for elucidating genetic pathways. However, a risk when observing suppression of a long lifespan is that the combination of 2 mutations that independently impair homeostasis might shorten lifespan in a nonspecific way. The specificity and meaning of suppression are more assured when other phenotypes of the longevous mutation are suppressed as well. Examples include the suppression of the dauer phenotype of daf-2 by daf-16 (Riddle et al. 1981; Kenyon et al. 1993) and suppression of the growth and behavioral phenotypes of isp-1 and nuo-6 by mutations in the apoptotic pathway genes ced-9, ced-4, ced-3, and ced-13 (Yee et al. 2014). Finally, the best cases include an understanding how the action of the genes involved fit together functionally. This could include physically proximity in the cell, action in the same signal transduction pathway, or in the same biosynthetic or degradation pathway. For example, the function of DAF-16 as a transcriptional regulator is one of the major outputs of the insulin receptor-like DAF-2, as has been demonstrated by numerous findings. Similarly, CED-13, CED-9, CED-4, and CED-3 are known to be part of a linear cascade of physical interactions, CED-9 and CED-4 are linked to the mitochondria and the vertebrate CED-4 homologue is believed to be sensitive to mitochondrial ROS. Furthermore, epistatic analysis of the longevity phenotype associated with isp-1, nuo-6, and mitochondrial ROS suggests that they act downstream of ced-13 but upstream of the other ced genes, which is consistent with what is known of the CED cascade, the localization of CED-4 and CED-9 on mitochondria, and the likely sensitivity of CED-4 to mitochondrial ROS. Created with BioRender.com.

As just discussed, it is likely that most mutants are stressed by missing some crucial gene function, even if long-lived. One expects these mutants therefore to upregulate stress responses or at least to require the full function of the stress response mechanisms for longevity. Impairment of stress response mechanisms may therefore lead to suppression without necessarily pointing to the process that is responsible for the extended longevity of the mutant. In other words, the activity of the stress response may be necessary but not sufficient. However, it is possible to test whether a particular stress response that is required is in fact the actual cause of the longevity in the long-lived mutant. For this, one has to show that boosting the candidate stress response leads, by itself, to longevity. We discuss specific examples in later sections. However, the approach could be difficult if the simultaneous activation of several stress response mechanisms is required.

Above, we mentioned how the absence of suppression by daf-16 was a first indication of the existence of genetically distinct longevity pathways. It is worth pointing out that failure of suppression by a bona fide suppressor of another long-lived mutant, does not suffer from the caveats we just discussed. Furthermore, when a mutation fails to suppress in 1 environment (e.g. standard growth conditions, on solid agar plates, with live Escherichia coli OP50, etc.), it means that the molecular process behind the long life of the longevous mutant does not require the activity altered by the suppressor mutation. This seems true even if the double mutant (the longevous mutation plus the suppressor mutation) is incapable of having an extended longevity in another environment, in which there might be additional or a different set of stresses. For example daf-16 was initially found not to suppress clk-1 longevity under standard conditions (Lakowski and Hekimi 1998). But daf-16; clk-1 double mutants do not succeed in living long when growing in liquid medium (Vanfleteren and Braeckman 1999). It seems reasonable to suppose that this is because the known stress of liquid medium overwhelms homeostatic processes when worms have lost an important stress response pathway (daf-16) and at the same time possess dysfunctional mitochondria (clk-1).

The UPRmt

The UPRmt is historically considered a transcriptional program activated by various forms of mitochondrial stress or dysfunction including perturbation of mitochondrial chaperones and proteases, but also oxidative phosphorylation (OXPHOS). The pathway was first documented in cultured mammalian cells (Martinus et al. 1996; Zhao et al. 2002), however, the components that regulate the mitochondrial-to-nuclear signal transduction have largely been elucidated in Caenorhabditis elegans via genetic screens searching for genes required for the induction of mitochondrial chaperone genes caused by mitochondrial perturbations (Yoneda et al. 2004; Benedetti et al. 2006; Haynes et al. 2007; Merkwirth et al. 2016; Tian et al. 2016). Most recently, it was demonstrated that the UPRmt is active during normal development and regulates mitochondrial biogenesis in coordination with growth factor-dependent protein synthesis (Shpilka et al. 2021). As such, the UPRmt scales mitochondrial network expansion with cell growth. Of course, the UPRmt program is further increased by perturbations to mitochondrial proteostasis or function presumably to maintain or recover the mitochondrial network in the affected cells. Importantly, all or most of the mitochondrial perturbations that extend lifespan activate the UPRmt, which is also frequently required for the survival or the longevity of long-lived mitochondrial mutants or long-lived RNAi-treated animals (Durieux et al. 2011; Baker et al. 2012; Nargund et al. 2012; Runkel et al. 2013; Wu et al. 2018; Campos et al. 2021). More on this below.

In the following sections, we will first review what is known about the UPRmt, then review findings relating to its involvement in the biology of long-lived mutants, and finally discuss, whether UPRmt activation could be the source of the longevity resulting from mitochondrial dysfunction.

Initial observations and how did we get here?

The UPRmt was initially characterized in cultured mammalian cells where depletion of mtDNAs (Martinus et al. 1996) or targeting of a terminally misfolded protein to the mitochondrial matrix resulted in increased transcription of mitochondrial protein chaperone genes, which are encoded in the nucleus (Zhao et al. 2002). As chaperones promote protein folding and protein complex assembly, the transcriptional pathway known as the UPRmt is considered a pathway that promotes mitochondrial protein homeostasis (proteostasis) (Balch et al. 2008).

To determine the conditions that activate the putative mitochondrial-to-nuclear signaling pathway, 2 transcriptional reporter strains were generated in which the promoters of 2 genes that regulate transcription of the mitochondrial-localized chaperones hsp-60 and hsp-6 were fused to GFP. Consistent with both molecular chaperones being required for worm viability, they were expressed throughout development, but their expression was also increased during diverse mitochondrial perturbations (Yoneda et al. 2004). Mitochondrial biogenesis and assembly of the OXPHOS complexes requires coordination of protein expression from the mitochondrial genome (mtDNA) and the nuclear genome. Those proteins encoded by nuclear genes are translated on cytosolic ribosomes and targeted to mitochondria with mitochondrial targeting sequences (MTS) most commonly located at the amino-terminus. Following import into mitochondria, each protein interacts with molecular chaperones or complex assembly factors to reach the final structure and/or assemble into the appropriate complex (Chacinska et al. 2009).

As expected, inhibition of several mitochondrial-localized protein chaperone genes (hsp-60, hsp-6) as well as genes that encode proteases that degrade damaged or unassembled proteins such as spg-7 (Langer 2000) caused induction of the UPRmt reporter. Assembly of the 5 OXPHOS complexes requires precise stoichiometry of multiple proteins encoded by nuclear genes as well as genes encoded by mtDNA. Impairment of mtDNA synthesis by exposure to ethidium bromide causes strong UPRmt activation due to impaired mtDNA replication and the accumulation of unassembled OXPHOS proteins encoded by the nuclear genome. Similarly, inhibition of mitochondrial ribosomes by RNAi to individual components (Yoneda et al. 2004) or by doxycycline impairs OXPHOS complex assembly and induces the UPRmt (Moullan et al. 2015). Furthermore, impairment of components required for the import of proteins from the cytosol to mitochondria also activated the UPRmt. Combined, these types of perturbations that impair mitochondrial proteostasis are referred to as mitonuclear imbalance owing to the differential accumulation of OXPHOS proteins encoded by the nuclear and mitochondrial genomes (Houtkooper et al. 2013). However, it should be noted that depletion of these same components also perturbs diverse aspects of mitochondrial biology including metabolism.

Regulation of the UPRmt

Using the hsp-60pr::gfp or hsp-6pr::gfp transcriptional reporters, several RNAi screens have been performed to identify components required for the mitochondrial-to-nuclear signaling that occurs during mitochondrial stress or dysfunction (Benedetti et al. 2006; Haynes et al. 2007; Runkel et al. 2013; Liu et al. 2014). The identified components roughly fall into 2 categories; those required for directly mediating transcription and those required to maintain chromatin structure in a state that facilitates transcription of the mitochondrial stress response. Both, chromatin and transcription regulation in response to mitochondrial stress, are mediated by mitochondrial-to-nuclear communication.

Mitochondrial-to-nuclear communication

ATFS-1

The bZIP transcription factor Activating Transcription Factor associated with Stress-1 (ATFS-1) directly mediates the transcriptional response to mitochondrial stress or dysfunction (Haynes et al. 2010). The nuclear accumulation and the subsequent activation of transcription by ATFS-1 is mediated by the function of the mitochondrial network localized throughout the cytosol (Fig. 5). Consistent with ATFS-1 being a nuclear transcription factor, the bZIP protein harbors a nuclear localization sequence (NLS). Importantly, ATFS-1 also harbors an amino-terminal MTS, which directs the protein into the mitochondrial matrix. The majority of ATFS-1 is subsequently degraded by the mitochondrial matrix-localized protease LONP-1 (Nargund et al. 2012).

Fig. 5.

UPRmt regulation via ATFS-1. The bZip protein ATFS-1 harbors both an MTS and an NLS (Nargund et al. 2012). During development, the majority of ATFS-1 is imported into mitochondria where it is degraded. However, perturbations to mitochondria including oxidative phosphorylation inhibition (Durieux et al. 2011), mitonuclear protein imbalance (Houtkooper et al. 2013) or an increase in mitochondrial protein import (Shpilka et al. 2021) reduces the rate ATFS-1 import into mitochondria resulting in accumulation in the cytosol. ATFS-1 then traffics to the nucleus via the NLS where it mediates a transcriptional response to increase components involved in mitochondrial biogenesis, mitochondrial proteostasis, glycolysis, ROS detoxification, and antibacterial innate immunity. Created with BioRender.com.

Mitochondrial protein import requires proteinaceous channels located in both the mitochondrial inner and outer membranes. Import also requires a functional respiratory chain to provide ATP and maintain the proton gradient across the inner membrane (Chacinska et al. 2009). Molecular chaperones located within the mitochondrial matrix are also required to facilitate folding of the incoming polypeptide as well as preventing retrotranslocation. Thus, during OXPHOS dysfunction or perturbations to mitochondrial protein homeostasis (proteostasis), mitochondrial protein import is reduced or completely impaired. Because ATFS-1 harbors an NLS, failure to be imported into mitochondria results in nuclear accumulation of the transcription factor where it binds promoters at a consensus site to induce or limit transcription. As described below, ATFS-1 induces ∼600 mRNAs which include a mitochondrial biogenesis program, mitochondrial molecular chaperones, and proteases, all of the components of the glycolysis pathway, numerous ROS-detoxifying genes as well as a number of genes that encode antibacterial innate immune components (Nargund et al. 2012, 2015).

Consistent with this model, Rauthan et al. (2013), identified several independent mutant strains in which the UPRmt is constitutively activated. Interestingly, these mutants were identified in a mutagenesis screen for strains resistant to relatively high doses of statins used in humans to treat high cholesterol. Surprisingly, all mutations resulted in amino acid substitutions near the N-terminus of ATFS-1 that perturbed the MTS of ATFS-1. Thus, these mutations caused constitutive UPRmt activation by directly impairing mitochondrial import of ATFS-1. While these strains were resistant to high dose statin treatment, they were all developmentally impaired and short-lived relative to wildtype worms in normal laboratory growth conditions (Rauthan et al. 2013). These results suggest that constitutive nuclear activation of ATFS-1 is not sufficient to extend lifespan. However, it should be noted that to protect against mitochondrial stress, ATFS-1 must also accumulate within the mitochondrial matrix. In the matrix, ATFS-1 binds mtDNA at a site similar to the consensus sequence it binds within promoters of nuclear genes. Mitochondrial-localized ATFS-1 limits transcription of mtDNA-encoded OXPHOS transcripts. The mechanism by which constitutive UPR^mt activation provides resistance to high levels of statins remains to be determined. However, it has been proposed to be due to increased outputs of the mevalonate pathway as ATFS-1 activation increases transcription of hmgs-1 which encodes the most upstream component of the mevalonate pathway which increases UQ synthesis as well as geranylgeranylation of multiple regulatory proteins (Rauthan et al. 2013; Oks et al. 2018).

Chromatin remodeling

In coordination with transcriptional adaptations during mitochondrial stress, chromatin structure is also remodeled and also required for UPRmt activation. Early studies demonstrated that activation of the UPRmt transcriptional reporters hsp-6_pr::gfp and hsp-60_pr::gfp were activated when worms were raised on cco-1(RNAi) during the early developmental stages L1, L2, and L3. However, perturbation of OXHPOS from the L4 stage into adulthood did not activate the stress response (Durieux et al. 2011). Importantly, the lifespan extension that occurs during perturbation of OXPHOS by RNAi treatments during the early developmental stage did not occur when perturbed during L4 or beyond. Interestingly, at the end of L4 mitochondrial biogenesis also stops in somatic cells, while accelerating in the germline to promote oocyte production (Tsang and Lemire 2002a).

Independent RNAi screens have revealed many of the chromatin regulators required to maintain or adapt chromatin status compatible with UPRmt activation (Benedetti et al. 2006; Haynes et al. 2007; Tian et al. 2016). During mitochondrial stress, the protein LIN-65 traffics from the cytosol to nuclei in a manner requiring the histone lysine methyl-transferase MET-2, which also resides in the cytosol (Tian et al. 2016). Interestingly, histone methylation by MET-2 occurs in the cytosol before transport to the nucleus where the histone is incorporated into nucleosomes (Fig. 6). It is not entirely clear how LIN-65 trafficking is regulated, but it requires the mitochondrial matrix-localized protease CLPP-1 suggesting inputs directly related to mitochondrial function or stress contribute to LIN-65 function (Tian et al. 2016).

Fig. 6.

UPRmt regulation requires chromatin remodeling. During early adulthood, chromatin is remodeled in somatic cells to limit transcription which impairs UPRmt activation during adulthood. However, perturbation of oxidative phosphorylation during development promotes chromatin remodeling to maintain or allow transcription into adulthood (Durieux et al. 2011; Tian et al. 2016). This requires LIN-65 and the incorporation of di-methylated histone H3 at lysine 9 into nucleosomes (H3K9me2), which requires the histone–lysine N-methyltransferase MET-2 (Tian et al. 2016). Concomitantly, the histone demethylases JMJD-3.1 and JMJD-1.2 remove methyl groups from H3K27me3 which promotes an open chromatin state (Merkwirth et al. 2016). Independently, the histone deacetylase HDA-1 removes acetyl groups from histones. Combined, these action facilitate DNA binding by the homeobox protein DVE-1 (Haynes et al. 2007) along with the coactivator ubiquitin-like protein 5 (UBL-5) (Benedetti et al. 2006) to maintain a transcription-competent chromatin state (Tian et al. 2016). In turn, ATFS-1 accesses UPRmt response elements (Nargund et al. 2015) and activates transcription of UPRmt genes ultimately resulting in increased longevity (Merkwirth et al. 2016; Tian et al. 2016). Created with BioRender.com.

Histone lysine demethylation is also required for UPRmt activation. Two Jumonji C domain lysine demethylases JMJD-1.2 and JMJD-3.1 are required to establish or maintain an open chromatin state that allow DVE-1 and ATFS-1 to bind and initiate transcription (Haynes et al. 2007; Nargund et al. 2015; Merkwirth et al. 2016; Tian et al. 2016). It is important to note that chromatin remodeling occurs independently of ATFS-1. Interestingly, in the absence of ATFS-1, significantly more LIN-65 accumulates in the nucleus perhaps emphasizing that increased mitochondrial stress stimulates LIN-65 trafficking (Tian et al. 2016). While it is unknown how the Jumonji methyltransferases are stimulated by mitochondrial stress, both JMJD-1.2 and JMJD1.3 require alpha-ketoglutarate and iron which are affected by mitochondrial dysfunction further suggesting that chromatin state is intimately linked to mitochondrial function. The histone acetyltransferase CBP-1 (p300) functions downstream of the histone demethylases JMJD-1.2 and JMJD-3.1, but upstream of ATFS-1 to regulate UPRmt activation. In response to mitochondrial stress, CBP-1 mediates histone acetylation at genes induced via the UPRmt promoting ATFS-1 transcription (Li et al. 2021). Surprisingly, the histone deacetylase HDA-1 is also required for UPRmt activation by interacting with DVE-1 (Shao et al. 2020). Lastly, the level of acetyl-CoA regulates chromatin remodeling via the NuRD (Nucleosome remodeling and deacetylase) complex (Zhu et al. 2020). In short, mitochondrial dysfunction reduces tricarboxylic acid cycle activity resulting in less citrate and acetyl-CoA. The ensuing activation of NuRD in collaboration with the nuclear-localized homeobox protein DVE-1 facilitates UPRmt activation.

Concomitantly, DVE-1 interacts with the ubiquitin-like protein UBL-5 and interacts with chromatin regions containing mitochondrial stress response genes (Benedetti et al. 2006; Haynes et al. 2007; Tian et al. 2016; Gao et al. 2019), presumably to maintain chromatin in a state to allow ATFS-1-dependent transcription. It should be noted that both DVE-1 and ATFS-1-dependent transcription are repressed by sumoylation. During mitochondrial dysfunction, the desumoylase ULP-4 removes SUMO from DVE-1 allowing it to bind DNA (Gao et al. 2019). ATFS-1 is also desumoylated, which increases the stability of ATFS-1 in the nucleus promoting UPRmt activation (Gao et al. 2019).

Combined, these findings demonstrate that at least 2 signal transduction pathways between mitochondria and nuclei are required to activate the UPRmt. During development, chromatin is maintained in a state accessible to ATFS-1 allowing UPRmt activation upon mitochondrial stress. However, at the conclusion of the developmental stages, chromatin status is rearranged preventing transcription of UPRmt genes during mitochondrial stress. To maintain the capacity to activate the UPRmt as worms age, chromatin must remain accessible which requires stimulation of histone dimethyltransferase, demethylases and a deacetylase.

UPRmt transcriptional program

Several mRNA profiling and sequencing experiments have been performed to determine the transcripts induced during mitochondrial dysfunction that require UPRmt regulatory components (Nargund et al. 2012, 2015; Lin et al. 2016; Merkwirth et al. 2016; Li et al. 2021; Shpilka et al. 2021). In addition to the well-documented increase in transcripts encoding mitochondrial-localized molecular chaperones and proteases, the UPRmt includes the mitochondrial fission components, nearly all mitochondrial ribosomes components, ROS-detoxifying enzymes, mtDNA replication components, mitochondrial protein import machinery, numerous OXPHOS components, and the UQ biosynthesis components suggesting the pathway mediates a mitochondrial maintenance and recovery pathway. ATFS-1 also induces all components of the glycolysis pathway potentially to generate ATP during mitochondrial perturbation. ATFS-1 binds the promoters of the majority of these genes at a consensus sequence known as a UPRmt element (Nargund et al. 2015).

In addition to increasing transcription of genes required for mitochondrial proteostasis and mitochondrial biogenesis, ATFS-1 also reduces or limits transcription of a number of highly expressed mitochondrial components including multiple TCA cycle and OXPHOS genes. Surprisingly, the same ChIP-sequencing revealed that ATFS-1 accumulates within dysfunctional mitochondria where it binds to a single site within the mtDNA. In the absence of atfs-1, all mtDNA-encoded OXPHOS transcripts are increased suggesting ATFS-1 also limits OXPHOS gene transcription from the mitochondrial genome as well as the nuclear genome. As OXPHOS proteins are among the most highly expressed mitochondrial proteins, their expression levels must be regulated so as not to overwhelm the mitochondrial protein folding capacity, and to promote stoichiometric expression of components encoded by each genome to limit the accumulation of orphaned subunits. Consistent with this model, worms that expressed only mitochondrial-localized ATFS-1, or only nuclear-localized ATFS-1, were unable to assemble the ATP synthase complex during mitochondrial stress and were developmentally impaired (Nargund et al. 2015).

Mitochondrial dysfunction as an indicator of pathogen infection

It is well appreciated that mitochondrial dysfunction serves as an indicator of pathogen infection. Considerable work in mammals has demonstrated that release, or extrusion of mtDNA or double-stranded RNA from the mitochondrial matrix serves as an early indicator of viral infection resulting in activation of diverse antiviral responses (West and Shadel 2017; Dhir et al. 2018; Tigano et al. 2021). Some of the original findings demonstrating that mitochondria serve as a sentinel of pathogen infection were found in C. elegans. Bacteria are the primary food source of C. elegans. Many species of bacteria produce toxins that perturb OXPHOS in other bacteria. Due to the homology of the OXPHOS systems between prokaryotes and eukaryotes, bacterial toxins often perturb OXPHOS in metazoans. For example, species of Streptomyces produce antimycin a and oligomycin which are complex III and ATP synthase inhibitors, respectively. The OXPHOS inhibitors are utilized to outcompete other bacteria. Considerable work has suggested that metazoans monitor intracellular functions as indicators of pathogen infection including translation efficiency, proteasome activity, and mitochondrial function (Dunbar et al. 2012; McEwan et al. 2012; Melo and Ruvkun 2012; Kirienko et al. 2013; Liu et al. 2014). In turn, signaling pathways induce xenobiotic detoxifying genes and antimicrobial genes to limit bacterial infection and eliminate toxic compounds.

In addition to promoting mitochondrial function, the UPRmt also induces transcription of a number of antibacterial innate immune genes and xenobiotic detoxifying genes (Liu et al. 2014; Pellegrino et al. 2014). Pathogenic species of Pseudomonas aeruginosa perturb OXPHOS and activate the UPRmt (Liu et al. 2014; Pellegrino et al. 2014; Deng et al. 2019). One such molecule is antimicrobial peptide CNC-4 which is potent against diverse bacterial species (Sapkota et al. 2021). Worms lacking atfs-1 die more quickly than wildtype worms upon exposure to P. aeruginosa, and constitutive UPRmt activation conferred by mutations in the MTS of ATFS-1 (Rauthan et al. 2013) promotes clearance of the pathogen from the intestinal lumen and prolong survival. These studies suggest a model that mitochondrial dysfunction or perturbations are an indicator of infection that initiates an ATFS-1-dependent innate immune response (Fig. 7).

Fig. 7.

The UPRmt mediates an antibacterial immune response. In response to mitochondrial dysfunction including that caused by the infectious bacteria Pseudomonas aeruginosa, ATFS-1 induces transcription of genes that encode antibacterial proteins, xenobiotic detoxifying proteins (Melo and Ruvkun 2012) along with components that promote recovery of the mitochondrial network (Liu et al. 2014; Pellegrino et al. 2014; Deng et al. 2019). The transcriptional response promotes survival upon P. aeruginosa infection as well as limiting accumulation of the pathogen within the intestinal lumen. During infection P. aeruginosa secretes the molecule phenazine to facilitate redox reactions by shuttling electrons from the hypoxic environment within the biofilm to oxygen located outside the biofilm. Phenazines also perturb the ETC in mitochondria resulting in UPRmt activation. Created with BioRender.com.

Pseudomonas aeruginosa secretes several molecules that could potentially perturb OXPHOS during infection including the complex IV inhibitor cyanide, electron shuttles known as phenazines, or pyoverdines which function as iron chelators (Kirienko et al. 2013). Surprisingly, upon Pseudomonas infection, the bZIP protein ZIP-3 accumulates and inhibits ATFS-1. Inhibition of ZIP-3 via gene deletion or RNAi results in enhanced ATFS-1 activity, which reduces Pseudomonas accumulation within the intestine and prolongs survival. An important finding in this study is that phenazines are a major activator of the UPRmt. Phenazines are electron shuttles secreted by P. aeruginosa upon biofilm formation to maintain redox balance when oxygen is depleted (Dietrich et al. 2006, 2008). The intrinsic electron shuttling activity of phenazines is sufficient to perturb host OXPHOS and activate the UPRmt. Interestingly, exposure to Pseudomonas unable to produce phenazines was considerably more toxic to C. elegans as they were unable to detect the pathogen and activate the UPRmt (Deng et al. 2019).

Paracrine UPRmt regulators and mitokines

As described above mitochondrial dysfunction activates the UPRmt via mitochondrial-to-nuclear communication. However, multiple signal transduction pathways have been identified by which mitochondrial stress and UPRmt activation in neurons can lead to activation of the UPRmt in neighboring cells via paracrine signaling. That perturbation of mitochondrial function in diverse neuronal types activates the UPRmt in intestinal cells has been demonstrated by several groups (Durieux et al. 2011; Shao et al. 2016; Tian et al. 2016). Signal transduction from neurons to intestinal cells required the secreted neuronal peptide FLP-2 as well as the canonical UPRmt regulators located in neurons. FLP-2 was shown to be required to propagate the signal between neurons, however, the mechanism by which the mitochondrial stress signal regulates intestinal cells remains to be elucidated.

Separate studies have shown that in addition to direct perturbations to neuronal mitochondria, expression of aggregation prone polyglutamine proteins in neurons also results in UPRmt activation in neurons but also in intestinal cells indicative of paracrine signaling. Polyglutamine aggregates interact with mitochondria and disrupt OXPHOS initiating a canonical UPRmt in the affected neurons (Berendzen et al. 2016). Subsequently, dense core vesicle release and serotonin secretion are required to propagate UPRmt activation to the intestinal cells. Cell nonautonomous UPRmt activation also requires the retromer complex and the Wnt ligand EGL-20 (Zhang et al. 2018). Impressively, overexpression of EGL-20 in serotonergic neurons is sufficient to induce the UPRmt in diverse tissues. The downstream components in the Wnt pathway including the intestine-localized Frizzled Receptor and transcription factor Beta-catenin are also required to receive and propagate the Wnt signal leading to UPRmt activation. Ultimately, DVE-1 and ATFS-1 are required to activate the UPRmt suggesting the Wnt pathway interacts with the chromatin or transcriptional regulators of the UPRmt, however, the precise mechanisms remain to be elucidated. Interestingly, recent work indicates that noncell autonomous UPRmt signaling can also originate in the germline and propagate to somatic cells to adapt proteostasis (Calculli et al. 2021) suggesting diverse cell-to-cell signaling networks function to coordinate mitochondrial biology in the intestine (Fig. 8). Combined, these studies indicate that paracrine UPRmt regulation promotes or coordinates metabolism throughout the organism to protect or prepare the organism for impending mitochondrial challenges.

Fig. 8.

Mitochondrial perturbation within neurons promotes UPRmt activation in intestinal cells via cell nonautonomous signaling. Neuron-specific OXPHOS perturbations caused by cco-1 (RNAi) or expression of polyglutamine results in UPRmt activation within neurons, but also in intestinal cells via cell nonautonomous signaling (Durieux et al. 2011; Shen et al. 2022). Neuron-to-intestinal cell communication requires secretion of serotonin (Berendzen et al. 2016), the peptide FLP-2 (Shao et al. 2016), and the Wnt ligand EGL-20 (Zhang et al. 2018). In turn, the Wnt ligand engages the Frizzled receptor on the surface of intestinal cells leading to beta-catenin activation which engages DVE-1 and ATFS-1 to mediate UPRmt activation (Zhang et al. 2018). Importantly, stimulation of the cell nonautonomous UPRmt is sufficient to extend lifespan (Durieux et al. 2011), but also promotes increased mtDNA accumulation within oocytes which increases longevity of subsequent generations (Zhang et al. 2021). Created with BioRender.com.

Neuronal transmission of UPRmt signaling via EGL-20 also promotes UPRmt activation in subsequent generations, potentially to prepare the subsequent generation to survive anticipated mitochondrial perturbations such as environmental toxins (Zhang et al. 2021). The communication between neurons and the germline by EGL-20 results in increased mitochondrial genomes (mtDNAs) in subsequent generations consistent with the UPRmt driving a mitochondrial biogenesis program. The increase in germ cell mtDNAs is proposed to cause mito-nuclear imbalance (Houtkooper et al. 2013) that perpetuates in a transmissible increase in longevity that require atfs-1 (Zhang et al. 2021).

UPRmt-regulated mitochondrial biogenesis

Components of the insulin signaling pathway as well as mTOR and S6 kinase are required for ATFS-1-dependent UPRmt activation (Baker et al. 2012; Gatsi et al. 2014). Recent work from our group indicates that ATFS-1 mediates a pathway during worm development that coordinates mitochondrial biogenesis or expansion with cell growth (Shpilka et al. 2021). It was recently shown that ATFS-1 harbors a relatively weak MTS compared with OXPHOS proteins (Melo and Ruvkun 2012), which is required for UPRmt activation during mitochondrial dysfunction (Rolland et al. 2019). In the absence of atfs-1, worms develop slowly and have less mitochondria. Interestingly, the mitochondria that do accumulate are quite small and dysfunctional. Similar to UPRmt activation during mitochondrial stress (Rolland et al. 2019), the atfs-1-dependent increase in mitochondrial biogenesis also requires the relatively weak MTS of ATFS-1. If the strength of the MTS was increased to that similar to OXPHOS proteins, ATFS-1 accumulated in mitochondria and mitochondrial biogenesis was impaired. S6 kinase which regulates the rate of protein synthesis during growth by phosphorylating a ribosomal subunit is also required for atfs-1-dependent mitochondrial biogenesis. Surprisingly, in the absence of S6 kinase, ATFS-1 fails to accumulate in nuclei, instead accumulating within mitochondria. These findings suggest a model in which the high quantity of OXPHOS and TCA cycle proteins being produced during development exclude or outcompete ATFS-1 for import into mitochondria resulting in ATFS-1 accumulating within nuclei. ATFS-1 drives a mitochondrial biogenesis transcriptional program until import capacity of the mitochondrial population can import and degrade ATFS-1. Consistent with this model, high expression a single protein (green fluorescent protein) with a strong MTS in muscle cells was sufficient to increase atfs-1-dependent transcription and mitochondrial biogenesis (Shpilka et al. 2021). Mitochondrial perturbations can further increase the mitochondrial biogenesis pathway potentially to promote mitochondrial network recovery.

Consequences of prolonged UPRmt activation

In addition to the ∼100 OXPHOS proteins and assembly factors encoded by the nucleus, in C. elegans 12 OXPHOS proteins are encoded by mtDNA, which is located within the mitochondrial matrix. Tissue-specific isolation of individual mitochondria demonstrated that neuronal and intestinal cell mitochondria harbor between 1 and 2 mtDNAs, while mitochondria in the germline harbor between 3 and 4 mtDNAs (Ahier et al. 2018). Mutations or deletions within mtDNA accumulate during aging but are also the cause of inherited mitochondrial diseases. However, a mutation in a single mtDNA has little effect on total cellular mitochondrial function as 100 s of mtDNAs exist in most cells. However, if the mutant genome accumulates to 60 % of the cellular population a decline in mitochondrial function ensues resulting in pathology. The events that allow or promote the clonal amplification of mutant mtDNAs to accumulate at the expense of wildtype mtDNAs remain unclear. However, it has been proposed that mutant mtDNAs may have a replicative advantage (Stewart and Chinnery 2021).

Several worm strains have been identified that harbor a mixture of mutant mtDNAs and wildtype mtDNAs, providing insight into the events that promote deleterious heteroplasmy. The most studied strain (uaDf5) has a combination of wildtype mtDNAs and a mtDNA with a 3.1-kb deletion that lacks 4 OXPHOS proteins and 7 genes that encode tRNAs required for protein synthesis within the mitochondrial matrix (Tsang and Lemire 2002b). In general, the deleterious mtDNA (ΔmtDNA) is stably inherited and maintained at ∼60% of the total mtDNA from worm lysate, although considerable variability has been observed (Tsang and Lemire 2002b; Gitschlag et al. 2016). The underlying mechanisms by which clonal deleterious mtDNAs (ΔmtDNAs) accumulate to relatively high levels relative to wildtype mtDNAs remains unclear (Pereira et al. 2021).

As expected, the heteroplasmic strain has a mild impairment in OXPHOS, which causes constitutive UPRmt activation (Gitschlag et al. 2016; Lin et al. 2016). However, this strain is not long-lived (Liau et al. 2007). This, like the complex I gas-1 mutation discussed above, is in contrast to the increased longevity caused by mutations in nuclear-encoded OXPHOS proteins, or RNAi knockdowns of such proteins that prevent correct ETC assembly. Interestingly, inhibition of UPRmt components does not inhibit, or slow, development of the heteroplasmic strain. Again, unlike the situation for mutant strains with lesions in nuclear OXPHOS genes. Surprisingly, inhibition of UPRmt signaling components such as ATFS-1 causes a loss of ΔmtDNAs suggesting the UPRmt is promoting deleterious heteroplasmy (Gitschlag et al. 2016; Lin et al. 2016). As ATFS-1 promotes mitochondrial biogenesis, we proposed that the UPRmt maintains ΔmtDNAs in an attempt to recover mitochondrial function (Fig. 9). Consistent with this model, inhibition of the mtDNA replicative polymerase polg-1 also resulted in preferential reduction of the ΔmtDNA, as did inhibition of the mitochondrial fission component drp-1 (Lin et al. 2016), which is also required for mtDNA replication and mitochondrial biogenesis (Lewis et al. 2016). Alternatively, the Patel lab proposed that increased mitochondrial proteostasis conferred by UPRmt activation protects those mitochondria harboring deleterious mtDNAs by preventing their degradation by autophagy (Gitschlag et al. 2016). The group has also shown that signaling pathways affected by nutrient status also promote ΔmtDNA maintenance (Gitschlag et al. 2020). In summary, it is clear that the UPRmt is required to maintain ΔmtDNAs in worm models of deleterious heteroplasmy, however, the precise mechanism by which the nuclear transcriptional response mediated by ATFS-1 maintains ΔmtDNAs remains to be determined.

Fig. 9.

The UPRmt maintains deleterious mtDNA heteroplasmy by promoting replication within dysfunctional mitochondria. The bZIP protein ATFS-1 is required to maintain the mutant mtDNA population in a condition known as deleterious mtDNA heteroplasmy (Tsang and Lemire 2002b; Gitschlag et al. 2016; Lin et al. 2016). ATFS-1 is imported into functional mitochondria where it is degraded by the protease LONP-1 (Nargund et al. 2012). However, ATFS-1 accumulates within mitochondria with perturbed oxidative phosphorylation where it avoids degradation and binds mtDNA along with the replicative polymerase POLG-1, thus promoting mtDNA replication in dysfunctional mitochondria. In heteroplasmic worms that harbor a mixture of mutant and wildtype mtDNAs, ATFS-1, and POLG accumulate in dysfunctional mitochondria and primarily interact with mutant mtDNAs providing a replicative advantage to the mutant mtDNA (Yang et al. 2022). Created with BioRender.com.

Deleterious UPRmt in relation to the mitochondrial permeability transition pore

Deleterious consequences of UPRmt activation have also been linked to the formation of the mitochondrial permeability transition pore (mPTP) (Angeli et al. 2021). A number of studies suggest, but might not have definitely demonstrated, that the mPTP in the inner membrane is created by the F-ATP synthase (Bernardi and Lisa 2015). The loss of the putative pore-forming component of F-ATP synthase extends adult lifespan, thus the mPTP might normally promote aging in C. elegans. In C. elegans it was found that loss of OSCP/ATP-3 during adulthood leads to initiation of the mPTP, activation of the UPRmt and a shortened lifespan. ATFS-1 helps drive the reduction of lifespan, suggesting that the UPRmt program can promote aging during adulthood. A particularly surprising finding is that OSCP inhibition can elicit UPRmt activation during adulthood, which is unlike other perturbations of most OXPHOS components which must be impaired prior to the L4 stage (Durieux et al.), however, the underlying mechanism remain to be determined. It will be interesting to determine the atfs-1-dependent processes that shorten lifespan upon mPTP pore opening.

A function of CLK-1 in UPRmt?

As mentioned above, clk-1 mutants were the first mutants to be identified in which longevity correlates with mitochondrial defects. Despite this, the mechanisms by which the loss of CLK-1 extends longevity remain uncertain beyond an association with mitochondrial dysfunction. CLK-1 is a mitochondrial hydroxylase that is necessary for the biosynthesis of UQ (Coenzyme Q) (Ewbank et al. 1997; Jonassen et al. 2001; Miyadera et al. 2001). UQ is necessary for mitochondrial electron transport and other redox functions (Wang and Hekimi 2016). As we will review further below, the normal development of these mutants, like that of other long-lived mutants, requires the activation of the UPRmt (Baker et al. 2012).

However, a recent study suggested a much more intimate relationship between the UPRmt and CLK-1 (Monaghan et al. 2015). The study sought to demonstrate that the extended lifespan of clk-1 mutants might not be related to the gene function in UQ biosynthesis but to a moonlighting function in the nucleus related to the UPRmt. This interpretation was based in large part on finding vertebrate CLK-1 homologue in association with chromatin in in vitro pull-down experiments, and on the identification of what appeared like a cryptic nuclear localization signal in the CLK-1 protein. This was further supported by observations suggesting that a C. elegans CLK-1 protein devoid of its normal mitochondrial localization signal might partially rescue the extended mutant lifespan and the altered expression of genes involved in mitochondrial quality control. However, this interesting but surprising claim has not been confirmed by other studies. For example, Molenaars et al. scored changes in gene expression between the clk-1 knockout strain (qm30) and the strain lacking only the MTS, and failed to find significant differences (Molenaars et al. 2018).

Furthermore, to test the hypothesis (Liu et al. 2017) took a different approach, which avoided, or by-passed, the question whether or not CLK-1 could be found in the nucleus and whether it had a UQ biosynthesis independent function there. In the absence of the function of the CLK-1 hydroxylase, cells accumulate the biosynthetic intermediate demethoxyubiquinone (DMQ) instead of UQ. The natural precursor of the aromatic ring of UQ is 4-hydroxybenzoate (4-HB). After attachment of the isoprenoid sidechain, a series of enzymatic reaction on this aromatic precursor leads to UQ, including a hydroxylation at position 2 of the ring carried out by CLK-1. A yeast study showed that treatment of a coq7 (clk-1 ortholog in yeast) null mutant with 2,4-dihydroxybenzoic acid (2,4-DHB) results in restoration of UQ₆ biosynthesis (Xie et al. 2012). Wang et al. (2015) found that this restoration by treatment with 2,4-DHB could also be obtained in CLK-1-deficient mouse cells and in actual Mclk1 KO mice. Apparently, the enzymes upstream of CLK-1 in the biosynthetic pathways are sufficiently promiscuous to function with 2,4-DHB and 2,4-DHB -dependent intermediates. clk-1 worm mutants were treated with 2,4-DHB and it was found that UQ biosynthesis was indeed partially restored, and that ALL phenotypes of the mutants were rescued. As there is not any CLK-1 protein present in these animals, these findings indicate that all phenotypes are due to the deficient UQ synthesis that can be restored with 2,4-DHB treatment and NONE could be ascribed to any putative alternative function of the protein. Thus, the longevity of clk-1 mutants appears to be intimately tied to its defect in UQ synthesis, although the mechanism by which a lack of UQ leads to longevity is still unknown.

Suppression of the long lifespan of mitochondrial mutants: the apoptotic signaling pathway and mtROS

The genetic bases of apoptosis were famously first investigated in worms (Hedgecock et al. 1983). In worms, apoptosis in the soma is purely developmental, that is, it serves to eliminate cells that are not needed or to change the complement of cells in the 2 sexes. In the germline, however, apoptosis is a homeostatic mechanism that promotes germline health, including by eliminating damaged germ cells (Gartner et al. 2008). This last observation suggests that there is a role for apoptosis in aging, as something has to explain the immortality of the germline, which might arise from the efficient elimination of nuclei with damaged DNA.

The extended lifespan of the long-lived mitochondrial ETC mutants, isp-1 and nuo-6, but not that of clk-1 mutants, is partially suppressed by mutations that regulate apoptosis, despite the fact that these mutants do not affect somatic apoptosis, nor mitochondrial function (Yee et al. 2014; Rauthan et al. 2015). Most of the core apoptotic molecules are involved (CED-3, CED-4, and CED-9), but not the BH3-only protein EGL-1, which is upstream of all worm apoptosis. Instead, the unrelated BH3-only protein CED-13, which is involved in physiological germline apoptosis (Schumacher et al. 2005), is required.

Previous findings, described above, indicated that the phenotype of isp-1 and nuo-6 mutants was very different from that of animals in which these same genes were knocked down by RNAi (Yang and Hekimi 2010b). Furthermore, the lifespan extension produced by the mutation and knockdowns were additive. Consistent with these observations, it was found that the apoptotic pathway is not required for RNAi against isp-1 and nuo-6, nor for other long-lived mutants, not even clk-1. In fact, loss of CED-4 even further extends the lifespan of clk-1 mutants (Yee et al. 2014).

This very specific effect of the apoptotic pathway on lifespan does not appear to be mediated by an effect on the mitochondria. Indeed, neither the low oxygen consumption nor the low ATP levels characteristic of isp-1 and nuo-6 mutants are rescued by loss of apoptotic signaling (Yee et al. 2014). It was therefore surprising that loss of the apoptotic pathway signaling substantially suppressed the hypomorphic mutant phenotypes that one would have expected to be due to low mitochondrial function, such as slow development, growth, and behaviors. These observations suggest that the hypomorphic phenotypes are not, or only partially, the direct result of mitochondrial defects but result from the reaction of the cell to the presence of mitochondrial defects. This is consistent with a study showing that loss of CED-4 suppresses the extended lifespan of isp-1 and nduf-7(et19), another, not previously mentioned, long-lived complex I mutant, but not the UPRmt activation observed in these mutants (Rauthan et al. 2015).

One of the ways in which the ETC defects in isp-1 and nuo-6 manifest themselves is as an increase in mitochondrial superoxide generation (Yang and Hekimi 2010a). Furthermore, the longevity of nuo-6 is suppressed by antioxidant treatment (Yang and Hekimi 2010a), and there is mixed evidence for this in the case of isp-1 (Yang and Hekimi 2010a; Desjardins et al. 2017). It is possible therefore that ROS are carrying the message of mitochondrial dysfunction to the rest of the cells to induce changes that ultimately extend lifespan. To test this possibility, wildtype worms were subjected to very low levels (0.1 mM) of the mitochondrial superoxide generator paraquat (PQ) (Yang and Hekimi 2010a). These levels of PQ do not appear to damage the mitochondria as established by intact levels of oxygen consumption, nor did they increase any aspect of ROS detoxification or repair as established by a lack of change in the expression of relevant genes following treatment (Yee et al. 2014). However, this small amount of PQ was sufficient to increase lifespan substantially (Yang and Hekimi 2010a; Van Raamsdonk and Hekimi 2012). The effect of PQ is also suppressed by loss of the 2 CED proteins that are physically associated with the mitochondria (CED-9 and CED-4) as well as by loss of CED-3, which is the downstream effector of the system. In contrast, PQ treatment can by-pass the requirement for CED-13, which acts upstream of the mitochondria-associated proteins (Yee et al. 2014).

The above paragraphs point to a high degree of specificity in the effects of the loss of the apoptotic pathway on the longevity of isp-1 and nuo-6: (1) the apoptotic pathway implicated in longevity is an alternative pathway that is regulated by CED-13 instead of the canonical EGL-1; (2) apoptosis is not affected by the mitochondrial mutations, nor is mitochondrial function affected by the ced mutations, suggesting no strong functional codependency of mitochondrial respiration and apoptosis; (3) loss of the CED-13 and CED-4, but not EGL-1, suppress other mutant phenotypes in addition to longevity; (4) loss of the pathway suppresses a treatment (PQ) that mimics a consequence of the mutations that is involved in their longevity (increased mtROS); and (5) epistatic genetic interactions for the longevity phenotype suggest a pathway of actual physical interactions that is consistent with what was already known (e.g. CED-13 acts upstream of the mitochondria, CED-9 and CED-4 are associated with mitochondria, the vertebrate CED-4 homologue is sensitive to mtROS, PQ acts in the mitochondria thus downstream of CED-13).

Further studies with PQ have also implicated other signal transduction pathways in its action on lifespan, such as roles for AMPK and HIF-1 (Hwang et al. 2014). Finally, based on an older observation demonstrating a role for clk-1-dependent ROS on RAS signaling (Shibata et al. 2003), a recent paper, although not focussed on aging per se, demonstrated a role for damage-independent low-level PQ, the antioxidant N-acetyl cysteine, superoxide dismutases (SODs), and ROS generated by mitochondrial mutations, including clk-1, isp-1, and nuo-6, in modulating signaling by the RAS pathway, as well as by the ROS-generating NOX signaling pathway (Kramer-Drauberg et al. 2020). In this study, the effects of PQ and excess mitochondrial superoxide due to the loss of mitochondrial SOD-2 were shown to be mediated by cytoplasmic conversion of the mitochondria superoxide into hydrogen peroxide by the cytoplasmic SOD (SOD-1). Likely the superoxide exits mitochondria through specialized channels (Han et al. 2003; Lustgarten et al. 2012).

Inducing longevity by hormesis

A set of studies, in particular by Michael Ristow and his group, have focused on treatments that produce enough mtROS to induce damage and thus engage stress response pathways that protect from ROS and likely other stresses. Some interventions based on complex I inhibition that have been shown to increase lifespan via mtROS have also been shown to increase ROS defenses (Schmeisser et al. 2013). These effects are dependent on the production of mtROS as well as signaling through PMK-1 and SKN-1 (the C. elegans p38 MAP kinase and Nrf2, respectively) and involves the upregulation of ROS-detoxifying enzymes such as SOD-1 and SOD-2 as well as resistance to oxidative stress. For example, Schulz et al. showed that inhibition of glycolysis with the inhibitor 2-deoxy-d-glucose (DOG) leads to increased lifespan (Schulz et al. 2007). As for complex I inhibition, this effect is dependent on an increase in ROS. However, in this case the increased ROS is the result of increased respiration rather than a mitochondrial defect, and AAK-2, the homologue of AMPK, is required for the increase in respiration. The lifespan extension brought about by the acute reduction of insulin signaling might also be in part achieved by mtROS resulting from increased respiration (Zarse et al. 2012). How much these mechanisms are involved in the longevity of longevous mitochondrial mutants is poorly studied. But gene expression studies in longevous mutants have failed to demonstrate any dramatic increases in ROS detoxification or ROS damage repair (Yee et al. 2014). How much the UPRmt might be required for these hormetic mechanisms of lifespan extension is not known and this might be important to explore.

Suppression of the long lifespan of mitochondrial mutants: immune signaling

Given the links between mitochondrial dysfunction, UPRmt, and innate immunity, one might expect that mitochondrial mutants would have altered innate immunity. A recent paper showed that isp-1 and nuo-6 mutants upregulate a substantial number of innate immunity genes and show enhanced resistance to the bacterial pathogen P. aeruginosa strain PA14 compared with wildtype worms (Campos et al. 2021). Activation of the UPRmt is required for the upregulation of the immune response. Loss of atfs-1 prevents the upregulation of immunity genes and the enhanced immune response in isp-1 and nuo-6 mutants whereas an atfs-1(gf) mutation is sufficient to upregulate immunity genes (Pellegrino et al. 2014). Although the immune response in isp-1 and nuo-6 does not appear to involve activation of p38 itself, it is also dependent on genes in the p38-mediated innate immune signaling pathway such as nsy-1, sek-1, pmk-1, and atf-7. Loss of nsy-1, sek-1, pmk-1, and constitutive activation of atf-7, prevent the upregulation of immunity genes and the enhanced immune response and, like the loss of atfs-1, prevent the lifespan extension of isp-1 and nuo-6. There is, however, no evidence that upregulation of the immune response is by itself sufficient to increase lifespan. Thus, mutants with impaired mitochondrial function specifically may require an intact p38-mediated innate immune signaling pathway for their longevity. Consistent with this, it has been shown that under different conditions, it is downregulation of the immune response genes that is required for longevity. Longevity conferred by dietary restriction also depends upon modulation of the p38-mediated innate immune signaling pathway. However, rather strikingly, the pathway must be downregulated, rather than upregulated for dietary restriction to increase lifespan (Wu et al. 2019).

Requirement for the UPRmt to support the development of mitochondrial mutants but not their extended lifespan

Several studies have documented that the development of mitochondrial mutants such as clk-1 and isp-1 require a functional UPRmt (Durieux et al. 2011; Baker et al. 2012; Nargund et al. 2012) (Fig. 10). These studies also observed an induction of the expression of the hsp-6 and hsp-60 reporter genes in these mutants. Furthermore, although the loss of atfs-1 (and that of dve-1 and ubl-5) from conception prevents the normal development of isp-1 and clk-1, and suppresses the long lifespan of nuo-6 (Wu et al. 2018), the loss of ATFS-1 during adulthood via RNAi does not suppress the longevity of either clk-1, isp-1, or nuo-6 (Wu et al. 2018). These results with atfs-1 are also consistent with the failure of gain-on-function atfs-1 alleles to induce extended lifespan (Rauthan et al. 2013; Bennett et al. 2014). However, the interpretation of this last point is complicated as the increased nuclear function in the atfs-1(et18) strain is not without consequences. The mutation that perturbs import of ATFS-1 into mitochondria and increases nuclear activity, also impairs the accumulation of ATFS-1 within mitochondria which mediates transcription of OXPHOS components from mtDNA.

Fig. 10.

UPRmt activation can be decoupled from longevity. Mutation and RNAi knockdown of mitochondrial genes that activate the UPRmt can either shorten or lengthen lifespan. This suggests that UPRmt activation is not sufficient for the longevity of mitochondrial mutants/RNAi knockdowns, even though the UPRmt may be required for the viability and/or longevity of some of the mutants and RNAi knockdowns. See main text for further discussion. Note that the timing of the RNAi treatment is not indicated except for atp-3, as RNAi during development activates the UPRmt and extends lifespan whereas RNAi during adulthood activates the UPRmt but shortens lifespan (Angeli et al. 2021). Data from Wong et al. (1995), Feng et al. (2001), Kayser et al. (2004), Liau et al. (2007), Durieux et al. (2011), Baker et al. (2012), Nargund et al. (2012), Pujol et al. (2013), Rauthan et al. (2013), Bennett et al. (2014), Rauthan et al. (2015), Ren et al. (2015), Gitschlag et al. (2016), Lin et al. (2016), Angeli et al. (2021), and de la Cruz-Ruiz et al. (2021). Created with BioRender.com.

Uncoupling the UPRmt from lifespan by RNAi knock-down

Bennett et al. identified 34 RNAi clones that reproducibly induce expression of the hsp-6p::gfp reporter. The targeted genes act in processes whose impairment is expected to induce the UPRmt (see above). The genes are involved in mitochondrial protein import, fat storage, sugar metabolism and other aspects of mitochondrial biology such as mitochondrial fission, protein quality control, and ion transport (Bennett et al. 2014). Nineteen of these clones were tested for effects on lifespan. Ten of them were found to increase lifespan and 6 were found to decrease lifespan. Thus, knockdown of some mitochondrial proteins produce lifespan increase without stimulation of the UPRmt. These might still require functional UPRmt processes to increase lifespan because they damage the mitochondria, but the stimulation of the UPRmt as measured by the reporter expression does not correlate with longevity. In the same study, the authors could also not observe a requirement of ATFS-1 for the longevity produced by RNAi against cco-1 (a subunit of cytochrome c oxidase). However, the allele used, atfs-1(tm4525), is a known hypomorph, which could have affected the results (Nargund et al. 2015). Thus, even animals that sustain mitochondrial proteostasis stress, as would be expected from loss of a subunit of a macromolecular complex like cytochrome c oxidase, might not always require the full UPRmt response for longevity.

These findings are reminiscent of the interactions of isp-1 and nuo-6 mutations with RNAi against these genes. As reviewed above, the UPRmt is required for the development and/or the longevity of these 2 mutants and their longevity is not additive: isp-1; nuo-6 double mutant live as long, but not longer, than the single mutants (Yang and Hekimi 2010b). Both genes encode subunit of respiratory chain complexes and are thus expected to require the UPRmt for longevity conferred by RNAi against them. Yet, despite these considerations, the observation is that the effects on lifespan of the mutations and the RNAi treatments are additive: nuo-6 mutants treated by RNAi against isp-1, as well as isp-1 mutants treated by RNAi against nuo-6, produce very long additive lifespans, much longer than those obtained with only 1 intervention at a time (Yang and Hekimi 2010b). This suggests that the mutations and RNAi against the same proteins are affecting lifespan by separate mechanisms, not both by stimulation of the UPRmt. Importantly, this conclusion is not inconsistent with the notion that both mechanisms can produce longevity only when an intact UPRmt process keeps the worms alive by compensating for damaged mitochondria.

Interestingly, as discussed above, overexpression of epigenetic regulators of the UPRmt can increase lifespan in a UPRmt-dependent manner (Merkwirth et al. 2016). However, it is possible that these regulators regulate more than just the UPRmt. It is therefore difficult to conclude that at this time that they increase lifespan by upregulating the UPRmt.

Concluding remarks

How to reconcile all the findings that we reviewed in a model of the relationship of the UPRmt with the lifespan extension obtained by targeting mitochondria? First of all, it is clear that mitochondrial impairment is not sufficient by itself to increase lifespan under standard conditions. As for the UPRmt, the evidence is not simple. On the one hand, atfs-1 gain-of-function alleles do not increase lifespan, but we cannot expect them to perfectly mimic all aspects of UPRmt activation. For example, import of ATFS-1 into mitochondria is required to increase mtDNA content during OXPHOS dysfunction (Yang et al. 2022), which is likely dysregulated in the atfs-1 gain-of-function alleles as they harbor a dysfunctional MTS. On the other hand, as noted, overexpression of epigenetic regulators might extend lifespan by acting on more processes than just the UPRmt. Secondly, there seem to be several mechanistically distinct ways in which altered mitochondrial function can affect lifespan. Some require the UPRmt, and some do not, some stimulate the UPRmt, and some do not, and some are additive to each other for lifespan. Third, the mitochondrial UPRmt appears frequently to be required during development for subsequent increases in lifespan. This may not be surprising considering all the way the UPRmt supports mitochondrial function during development. Damage sustained during development might prevent subsequent lifespan increases, explaining how constitutively impairing the UPRmt suppresses longevity.

As the present review attests, the UPRmt subsumes many important processes that allow for unimpaired mitochondrial function. Furthermore, it seems self-evident that well-functioning mitochondria are crucial for the health of the cell. This very naturally leads to a model in which damage to mitochondrial function by mutation or RNAi interference would lead to the activation of the UPRmt, which in turn could increase lifespan by modifying mitochondrial function in a beneficial way. Unfortunately, the data does not strongly support such a model. Rather, based again of all we reviewed, it appears more likely, that the UPRmt is required for the longevity of animals with impaired mitochondrial function because the UPRmt is required for development of these animals and might also be required for keeping their damaged mitochondria functional. To date, there are no clear-cut findings that indicate that stimulation of the UPRmt is sufficient for extending lifespan. Yet, all we know about the very broad and continuous functions of the UPRmt throughout life suggests that, if the mechanism that allows for the extended longevity of the mutants with impaired mitochondrial function acted in mitochondria, the UPRmt should always be necessary for longevity. The apparent paradox therefore is that we expect the UPRmt to be the source of mitochondrial dysfunction-dependent longevity, but there is no clear-cut indication that it is. A possible explanation for this apparent paradox could be that mitochondria that are damaged by pro-longevity interventions generate a signal that exits the mitochondria and acts outside the mitochondria and not particularly on mitochondrial function. As we have seen mtROS could be such a signal as ROS are bona fide signal transduction modulators. It remains to be discovered which changes in cellular physiology that results from the action of mtROS leads to increased longevity.