The UPR^mt Protects Caenorhabditis elegans from Mitochondrial Dysfunction by Upregulating Specific Enzymes of the Mevalonate Pathway

Abstract

The mevalonate pathway is the primary target of the cholesterol-lowering drugs statins, some of the most widely prescribed medicines of all time. The pathway’s enzymes not only catalyze the synthesis of cholesterol but also of diverse metabolites such as mitochondrial electron carriers and isoprenyls. Recently, it has been shown that one type of mitochondrial stress response, the UPR^mt, can protect yeast, Caenorhabditis elegans, and cultured human cells from the deleterious effects of mevalonate pathway inhibition by statins. The mechanistic basis for this protection, however, remains unknown. Using C. elegans, we found that the UPR^mt does not directly affect the levels of the statin target HMG-CoA reductase, the rate-controlling enzyme of the mevalonate pathway in mammals. Instead, in C. elegans the UPR^mt upregulates the first dedicated enzyme of the pathway, HMG-CoA synthase (HMGS-1). A targeted RNA interference (RNAi) screen identified two UPR^mt transcription factors, ATFS-1 and DVE-1, as regulators of HMGS-1. A comprehensive analysis of the pathway’s enzymes found that, in addition to HMGS-1, the UPR^mt upregulates enzymes involved with the biosynthesis of electron carriers and geranylgeranylation intermediates. Geranylgeranylation, in turn, is requisite for the full execution of the UPR^mt 3response. Thus, the UPR^mt acts in at least three coordinated, compensatory arms to upregulate specific branches of the mevalonate pathway, thereby alleviating mitochondrial stress. We propose that statin-mediated inhibition of the mevalonate pathway blocks this compensatory system of the UPR^mt and consequentially impedes mitochondrial homeostasis. This effect is likely one of the principal bases for the adverse side effects of statins.

IN eukaryotes, the mevalonate/isoprenoid/cholesterol synthesis pathway, hereafter referred to as the mevalonate pathway, catalyzes the synthesis of metabolites vital for cellular metabolism, growth, and differentiation (Schoenheimer and Breusch 1933; Bloch 1965). The main branch of the pathway converts acetyl-CoA to farnesyl diphosphate, which then serves as a precursor for the metabolism of most subbranches. At each subbranch, the sequential function of one of a series of metabolic enzymes converts the farnesyl diphosphate precursor to a different end product. In mammals, these end products include hemeA and ubiquinones, both of which act as electron carriers for the mitochondrial electron transfer chain (ETC); dolichol, a lipid anchor for N-glycosylation; isoprenyls, which are used for protein prenylation; and cholesterol [reviewed in Goldstein and Brown (1990)]. Mevalonate pathway metabolites are essential for many aspects of cellular and organismal metabolism, and thus the function or malfunction of the pathway has fundamental implications for human health and disease. For example, a loss in isoprenylation can promote apoptosis (Araki et al. 2012), inhibit tumorigenesis through its effects on GTPases (Jiang et al. 2014), or can lead to myopathy (Dirks and Jones 2006). Similarly, while cholesterol is essential for plasma membrane integrity and fluidity, impaired cholesterol homeostasis promotes the onset and progression of many cardiovascular diseases (Wollam and Antebi 2011).

More than four decades since their discovery (Endo et al. 1976), statins remain a leading clinical solution for patients with high serum cholesterol and are now among the most widely used medications in the world. Statins are primarily used to lower the level of cholesterol in the serum of patients through the inhibition of the second enzyme of the mevalonate pathway, HMG-CoA reductase. Statins, however, may have additional beneficial effects, including having anticancer (Chan et al. 2003; Freed-Pastor et al. 2012) and antineurodegenerative properties (Wang et al. 2011). Although tens of millions of people worldwide are currently prescribed with statins, an even more widespread prescription of statins has been tempered because of their adverse side effects. These effects range from mild muscle pain to severe pathological conditions such as depression, insomnia, neuropathy, myopathy, and antiatherogenic syndrome (Golomb and Evans 2008). A growing body of evidence suggests that the inhibition of cholesterol synthesis itself may not be the primary cause of these deleterious side effects. Instead, the inhibition of protein prenylation (Wang et al. 2008; Correale et al. 2014) and an impaired mitochondrial homeostasis (Golomb and Evans 2008; Apostolopoulou et al. 2015) may be among the underlying mechanisms leading to the side effects of statins. One proposed mechanism suggests that reduced levels of the ubiquinone Coenzyme Q₁₀ (CoQ₁₀) cause impaired mitochondrial calcium homeostasis in the muscles of statin-treated humans (Galtier et al. 2012; Sirvent et al. 2012), but this is highly controversial (Apostolopoulou et al. 2015; Auer et al. 2016). At present, the exact underlying mechanism leading to the side effects of statins remains unknown.

Despite the mevalonate pathway’s centrality in human metabolism and its reliance on the sequential activation of >30 enzymes, so far only a handful of these enzymes were identified as being subject to regulation (Sharpe and Brown 2013). For example, milestone discoveries by Joseph L. Goldstein and Michael S. Brown first demonstrated that the levels of cholesterol and cholesterol byproducts regulate the metabolism of the pathway via negative feedback mechanisms [reviewed in Goldstein and Brown (1990)]. Paradigms describing the regulatory mechanisms of the pathway typically focus on the transcriptional and post-translational regulation of the second-most upstream enzyme of the pathway, HMG-CoA reductase (also known as HMGCR1). Because HMGCR1 is the rate-determining enzyme of the pathway, its regulation can determine the level of mevalonate pathway metabolism (Reynolds et al. 1984; Goldstein and Brown 1990; Burg and Espenshade 2011). In addition to regulation through HMGCR1 activity, cholesterol and its derivatives affect the levels of different enzymes of the cholesterol synthesis subbranch, including DHCR7 (Prabhu et al. 2016) and SQEL (Hidaka et al. 1990; Foresti et al. 2013; Zelcer et al. 2014). Additionally, sterol-independent mechanisms of regulation exist (Sharpe and Brown 2013). These include, for example, the p53-dependent activation of enzymes in the pathway during specific cancers (Freed-Pastor et al. 2012) and the post-translational regulation of the first enzyme of the pathway, HMGS-1, in Caenorhabditis elegans (Sapir et al. 2014). The widespread and pleiotropic effects of the different metabolites of this pathway on cellular functions have complicated attempts to piece apart their regulatory mechanisms beyond the control of cholesterol synthesis. Interestingly, like many other invertebrates, C. elegans lacks the cholesterol synthesis subbranch of the pathway (Vinci et al. 2008). The absence of the subbranch for cholesterol synthesis could facilitate the discovery of previously uncharacterized regulatory mechanisms typically masked by this subbranch.

Recently, a growing body of research has suggested an intimate cross-talk between mitochondria and mevalonate pathway metabolism. Studies have found that a key stress response pathway that safeguards the mitochondria, the mitochondrial Unfolded Protein Response (UPR^mt), is capable of affecting mevalonate pathway metabolism during stress. The activation of UPR^mt via ethidium bromide (EtBr) preconditioning protects yeast, C. elegans, and human cells in culture from the inhibitory effects of statins (Rauthan et al. 2013). In addition, both mevalonate pathway metabolism and its downstream subbranch of protein prenylation are proposed to be requisite for the activation of the UPR^mt (Rauthan et al. 2013; Liu et al. 2014; Ranji et al. 2014). Thus, the UPR^mt may upregulate the mevalonate pathway, which in turn may be essential for the proper full execution of the UPR^mt. Nevertheless, the molecular mechanisms underlying this proposed cross-talk between the UPR^mt and the mevalonate pathway remain unknown.

As one of the central metabolic centers of the cell, mitochondria are sensitive to many external and internal insults, including energy crises (e.g., low levels of ATP) and the accumulation of Reactive Oxygen Species (ROS) (Haynes et al. 2013). The UPR^mt has evolved to protect the cell from these types of stressors (Melber and Haynes 2018). In C. elegans, the UPR^mt has been well characterized and many of its key players are known (Runkel et al. 2013; Munkácsy et al. 2016). In response to mitochondrial or metabolic stress, the bZIP transcription factor ATFS-1 (Haynes et al. 2010) translocates to the nucleus, where it induces a genome-wide transcriptional response (Nargund et al. 2012). ATFS-1 transcriptional targets include many mitochondrial and cytoplasmic-localized chaperones, subunits of the ETC, antioxidants, and mediators of innate immunity (Pellegrino et al. 2014). Importantly, ATFS-1 activation alters the expression of enzymes of diverse metabolic pathways (Nargund et al. 2012), suggesting that it can remodel or even rewire metabolic networks. ATFS-1 works with a group of coregulators to fine-tune the UPR^mt (Haynes et al. 2013). Additional transcription factors that play a role in the UPR^mt response include the homeodomain protein DVE-1 (Haynes et al. 2007), the bZIP protein ZIP-2 (Estes et al. 2010), and the bZIP protein SKN-1 (Munkácsy et al. 2016).

In addition to the link between mevalonate pathway metabolism and the UPR^mt, inhibition of the mevalonate pathway by statins impairs farnesylation, thereby inducing an endoplasmic reticulum (ER) stress response (Morck et al. 2009). This activation of the ER stress response depends upon the activity of two canonical players of the UPR^er, response ire-1 and xbp-1. Thus, statins activate the UPR^er response while blocking UPR^mt activation due to a lack of geranylgeranylation. Nevertheless, the link between statins and the activation of different stress signals deserves additional exploration.

In this work, we used C. elegans to understand better the relationship between mitochondrial function and the mevalonate pathway. To identify the molecular mechanisms that link mitochondrial dysfunction to mevalonate pathway metabolism, we asked whether the mitochondrial stress response directly controls the steady-state expression of enzymes in the mevalonate pathway. Intriguingly, we found that in C. elegans, mitochondrial dysfunction does not affect the levels of HMGR-1 protein, the rate-controlling enzyme of the mevalonate pathway and the primary regulatory point in mammals. Instead, the mitochondrial stress response upregulates the first dedicated enzyme of the pathway, HMGS-1. We characterized the molecular basis for HMGS-1’s upregulation and found that this effect relies upon ATFS-1 and DVE-1, two transcription factors that mediate the UPR^mt in C. elegans. Pathway-level transcriptional analyses identified at least three additional points of direct regulation of mevalonate pathway metabolism by the UPR^mt response. These include: (i) upregulation of an enzyme of the hemeA synthesis subbranch; (ii) upregulation of one enzyme, coq-1, that mediates the first reaction in the subbranch of ubiquinone production; and (iii) upregulation of a geranylgeranyl transferase. Finally, we show that deficiencies in the production of ubiquinone and hemeA induce the UPR^mt, which then upregulates mevalonate pathway enzymes such as hmgs-1. This upregulation is dependent on atfs-1 activity that functions in a compensatory circuit for the perceived mitochondrial dysfunction.

Materials and Methods

Worm strains and maintenance

Unless otherwise stated, C. elegans strains were maintained on NGM (Nematode Growth Medium) plates at 20°, as previously reported (Brenner 1974). The worm strains used in this study and details of strain construction are provided in Supplemental Material, Table S1.

Analysis of deletions and point mutations (dCAPS)

Deletions and point mutations were analyzed by the single-worm PCR method. A list of primers is provided in Table S2. For the analysis of point mutations, the dCAPS method was employed (Neff et al. 1998) using the dCAPS Finder 2.0 software (Neff et al. 2002).

Pharmacological experiments

All pharmacological experiments were conducted using 35-mm diameter plates poured with 2 ml autoclaved NGM solution that was cooled to 55° and mixed with the different pharmacological agents. For chemicals that were dissolved in solvents other than water, the highest concentration of the solvent was used as a vehicle control. After drying the plates for 1 day at room temperature, 150 μl of bacteria (OP-50 or other bacterial strains) was seeded on the plate. After 24 hr at room temperature, the seeded plates were used for experiments or stored in the refrigerator at 4° for a maximum of 2 weeks. To initiate an experiment, ∼100 worms, synchronized to larval stage one (L1) by incubation in S. basal medium (Stiernagle 2006) for 24 hr postbleach, were added to the plates. Unless stated otherwise, all experiments were conducted at 20° and worms were analyzed 72 or 96 hr after the seeding of worms on the plates, which corresponds to a chronological age of day 1 or day 2 of adulthood. Tunicamycin (T7765; Sigma, St. Louis, MO) was dissolved in DMSO to a stock solution of 5 mg/ml that was stored at −20° in the dark until used. The stock solution was diluted 1:1000 to a final concentration of 5 μg/ml in NGM plates and was compared to control NGM plates with DMSO diluted 1:1000. Mevalolactone (M4667; Sigma) was dissolved in water to create a stock solution of 1 M and was used in final concentrations of 1, 10, or 20 mM in the NGM plates.

Feeding RNAi

All double-stranded RNA (dsRNA) bacterial clones were sequenced before use. Feeding was conducted as previously described (Timmons et al. 2001) with the following modifications: (i) Before every experiment, we transformed HT115 (DE3) bacteria with the dsRNA expressing-plasmids of interest. Freshly transformed bacteria (up to 1 week old) were picked and cultured overnight in a water bath that was set to 37° in 2 ml of Luria Broth (LB) containing 50 μl/ml of ampicillin and 12.5 μl/ml of tetracycline. (ii) The next morning, bacterial cultures were diluted 1:100 in LB with 50 μl/ml ampicillin and were grown for 6 hr in a 37° water bath. (iii) IPTG (at a final concentration of 1 mM) and ampicillin (at a final concentration of 50 μl/ml) were spread on 35- or 60-mm NGM plates before seeding the bacteria. After 6 hr of incubation, 1.5 ml of bacterial culture per plate was harvested by spinning down at 4000 rpm. The bacterial pellet was resuspended in 100 μl of LB and spread onto the plate. Plates were incubated for 12 hr in the dark at room temperature, and ∼100 or 300 worms, synchronized to the L1 stage by incubation in S. basal medium for 24 hr postbleach, were added to the 35- or 60-mm RNAi plates, respectively. In case of dsRNA clones that affect worm survival (e.g., dsRNA corresponding to the hmgs-1 gene sequence), the plates were seeded with 600 worms. Unless stated otherwise, all experiments were conducted at 20° and were analyzed 72 or 96 hr after seeding the worms in plates, when the worms were at the chronological age of day 1 or day 2 adults, respectively. For dsRNA mixing experiments, the bacterial clones were grown separately and mixed in 1:1 (volume:volume) ratio before seeding.

Expression analysis by quantitative PCR (qPCR)

Worms were bleached and synchronized by incubation for 24 hr in S. basal medium at 20°. Synchronized wild-type worms at the L1 stage (∼300 worms/plate, four plates per condition) were spotted onto empty vector and spg-7 RNAi plates. Worms were grown until day 1 of adulthood, then collected and washed with S. basal medium three times and with RNase-free water two more times. Total RNA was extracted using a combination of Trizol/chloroform and an RNeasy kit (Qiagen, Valencia, CA) and was treated with DNase I (New England Biolabs, Beverly, MA) according to the manufacturer’s instructions. The concentration of the messenger RNA (mRNA) was quantified by both Nanodrop and Qubit (Thermo-Fisher Scientific, Waltham, MA), and 0.6 μg of total RNA was used for complementary DNA (cDNA) synthesis using the ProtoScript II Reverse Transcriptase kit (New England Biolabs) with random hexamers as the cDNA primers. qPCR reactions were performed using the Fast SYBR Green Mastermix (Thermo-Fisher Scientific) in Step One Plus (Thermo-Fisher Scientific) or CFX Connects (BioRad, Hercules, CA) machines. Primers were designed using the National Center for Biotechnology Information (NCBI) primer design tool (Primer-Blast) (Ye et al. 2012). All primers span at least one intron–exon junction and have a melting temperature (T_M) ranging from 58 to 62°. All primers were synthesized by IDT and can be found along with their calculated T_M and amplicon size in Table S2. Primer sequences for act-1 or pmp-3 were used for expression normalization (Hoogewijs et al. 2008; Zhang et al. 2012). The ΔΔC_t method was used to calculate the fold change in expression. Results are averages of three biological replicates, and error bars represent SEM. Statistical significances were calculated using the Prism software with unpaired two-tailed Student’s t-tests.

Fluorescence imaging and immunoblotting

To determine the level of expression of GFP, synchronized worms harboring different GFP reporters were mounted on agar pads (2% agar in S. basal) with 3 μl of 10 mM levamisole (16595-80-5; Santa Cruz). Unless otherwise stated, gravid adult worms were visualized at a chronological age of day 1 or 2 of adulthood, i.e., 72 and 96 hr postrelease from L1 arrest, respectively. For each condition, at least 30 worms were analyzed, and at least 10 worms were imaged in a group across all independent experiments. For imaging of groups of worms on slides, 10 worms were collected from the plates and transferred to a 1-μl drop of the levamisole solution. Once the drop was dried out, the worms were oriented head-to-head using a pick and then covered by a coverslip. Levamisole solution was added from the margins of the slide until the entire agar pad was covered with this solution, and the slides were kept for ∼30 min in a humid chamber to reduce the number and size of air bubbles. In every experiment, the same magnification, light intensity, and exposure time were used. Images were captured using a Nikon DSRi-1 camera connected to a Nikon E600 fluorescence compound microscope or using a Nikon SMZ18 fluorescence dissecting microscope connected to a Nikon DS-Fi3 camera. Images were processed using the Adobe Photoshop software according to ethical guidelines for the appropriate use and manipulation of scientific digital images (Cromey 2010).

For immunoblotting, worms were grown on spg-7 or empty vector RNAi plates. At the adult stage, 50 worms were collected, lysed, and the total proteome was separated by SDS-PAGE electrophoresis. Next, the proteins were transferred to nitrocellulose membranes, and probed for the HMGS-1::GFP protein using an anti-GFP antibody followed by stripping and probing with an antiactin antibody.

Determination of the number of nuclei with DVE-1::GFP

Worms expressing the DVE-1::GFP construct were mounted as described above. Nuclei with GFP signal were counted in a Nikon Eclipse 600 microscope using a 100× magnification. The statistical significance of the change in the number of DVE-1::GFP nuclei in each condition was calculated by one-way ANOVA testing using the SPSS software. Normality was tested and confirmed by the Shapiro–Wilk test. Next, the Scheffe post hoc test was included to rule out the effect of multiple tests.

Thrashing assay

Worms were grown until a chronological age of day 1 adult as described above. Groups of five worms were placed in a drop of 20 μl S. basal medium on a glass slide, and after ∼5 min of habituation, the number of thrashes was measured. Movies of the swimming worms, in 30-sec intervals, were captured using Nikon’s SMZ18 dissecting microscope and the Active Presenter software. The captured movies were then played in slow motion for 20 sec to count the number of times the head or the tail cross the midline of the worm, which we defined as one thrash. The statistical significance of the results was analyzed by one-way ANOVA testing using SPSS software. Normality was tested by the Shapiro–Wilk test followed by the Kruskal–Wallis test to analyze data that are not normally distributed.

Generation of a deletion allele in the hmgs-1 locus using the CRISPR-Cas9 technology and rescue experiments

Deletion was generated, upon our request, in the laboratory of Donald G. Moerman as part of the CRISPR-Cas9 Gene Knockout project in C. elegans. An integration cassette, including homology arms in the hmgs-1 gene and a gfp gene under the regulation of a myo-2 promoter, was used. This cassette replaced the sequence of the hmgs-1 gene from the middle of exon two through the middle of exon seven. This replacement removed the majority of the eight total exons of the hmgs-1 gene introducing a frameshift after Serine 64 and a stop codon 10 further amino acids downstream. The replacement was validated by PCR, and worms having this deletion allele were outcrossed four more times to N2 wild-type worms before the experiments. Supplementation of 20 mM mevalolactone to the NGM plates can fully rescue the lethality associated with this hmgs-1 deletion allele providing another indication for the specificity of this allele. To test the recuse potential of the hmgs-1::gfp transgene, this integrated construct was recombined with the hmgs-1 deletion allele. Worms homozygous for the hmgs-1 deletion allele, either with or without the hmgs-1::gfp construct, were grown on plates with 20 mM mevalolactone. These worms, at larval stage four, were transferred to plates without mevalolactone and the number of progeny and livelihood of the worms were determined 96 hr after the transfer.

Data availability

Strains and plasmids are available upon request. Table S1 contains data about all the C. elegans strains used in this study. Table S2 contains data and the individual sequences of all the primers used in this study. The results of the screen conducted to find GTPases that regulate the levels of HMGS-1::GFP are presented in Table S3. Supplemental material available at Figshare: https://doi.org/10.25386/genetics.6021527.

Results

Mitochondrial stress upregulates hmgs-1

To determine whether the mitochondrial stress response affects the metabolism of the mevalonate pathway, we first measured the levels of different mevalonate pathway enzymes in vivo in C. elegans. We started with the first dedicated enzyme in the pathway, HMG-CoA synthase (HMGS-1 in C. elegans or HMGCS1 in mammals). This enzyme undergoes transcriptional regulation in mammals (Gil et al. 1986) and an aging-dependent SUMOylation in C. elegans (Sapir et al. 2014). First, we measured the levels of HMGS-1 protein by using an integrated fosmid in which a green fluorescence protein (gfp) coding sequence was recombined in-frame at the 3′ end of the hmgs-1 gene (Sarov et al. 2012; Sapir et al. 2014). In unstressed conditions, worms carrying this construct (hereafter, HMGS-1::GFP) had a weak GFP signal in the intestine and an even weaker signal in other tissues (Sapir et al. 2014; Figure 1A). In contrast, after using RNAi against the mitochondrial metalloprotease spg-7 (paraplegin) to induce mitochondrial stress (Nargund et al. 2012; Pellegrino et al. 2014), the HMGS-1::GFP signal was strongly upregulated in the intestine (Figure 1B). A similar upregulation of HMGS-1::GFP (Figure 1, C and D) was seen in a strain harboring a mutation in the iron/sulfur protein isp-1 that also results in mitochondrial stress (Nargund et al. 2012; Munkácsy et al. 2016). Under each of these conditions, the mitochondrial stress response upregulated HMGS-1::GFP primarily in the intestine. Double-stranded RNA corresponding to the hmgs-1 gene eliminated the GFP signal in HMGS-1::GFP worms treated with spg-7 RNAi (Figure S1, A and B), demonstrating the specificity of the HMGS-1::GFP construct and the RNAi treatment.

Figure 1.

Mitochondrial stress upregulates HMGS-1. (A) HMGS-1::GFP protein in worms fed with an empty vector control (EV). (B) spg-7 dsRNA (spg-7i) leads to HMGS-1::GFP upregulation. (C) HMGS-1::GFP protein in worms of an isp-1(+) background. (D) The isp-1(qm150) mutation leads to HMGS-1::GFP upregulation. (E) HMGR-1::GFP in worms fed with an empty vector control. (F) spg-7 dsRNA does not evidently affect the level or distribution of HMGR-1::GFP. (G) Expression of the hmgs-1 promoter fused to gfp in worms grown on an empty vector control. (H) spg-7 dsRNA activates the hmgs-1 promoter. (I) Quantitative PCR (qPCR) analyses of the relative levels of endogenous hmgs-1 and hmgr-1 transcripts in worms fed with spg-7 dsRNA and empty vector control. Error bars represent SE. ** P ≤ 0.01; NS indicates no significant difference using a two-tailed, unpaired Student’s t-test. (J) Western blot analysis showing HMSG-1::GFP upregulation upon mitochondrial stress induced by the atp-3 RNAi. Actin was used as a loading control. The size of the primary band (∼85 kD) is in agreement with the calculated size of HMSG-1::GFP (83.4 kDa), based upon the amino acid sequence of the fusion protein. In all subsequent images, groups of 10 worms were oriented with their anterior side to the left. Unless otherwise stated, in all subsequent experiments, worms were analyzed at the same chronological age (either day 1 or 2 of adulthood). In each experiment, control and tested worms were synchronized and analyzed at the exact same chronological age. isp-1 mutant worms were analyzed 6 days after the L1 stage. Consistent with previous reports (Nargund et al. 2012; Pellegrino et al. 2014), differences in worm size stem from the induction of UPR^mt that results in a smaller body size of the adult worms. Panels A–H are to the same magnification. Bars represent 250 μm and 200 μm (inserts).

A growing body of evidence suggests that mitochondrial stress responses share some molecular and cellular characteristics with other types of organelle stress responses, such as those of the ER (Haynes et al. 2013; Runkel et al. 2013). Therefore, it is possible that other forms of stress such as the ER stress response affect the level of hmgs-1. To address these possibilities, we exposed the HMGS-1::GFP animals to a wide variety of stress-inducing conditions, ranging from starvation to ER stress (UPR^er). Importantly, we found no change in the level of HMGS-1::GFP across any of the other conditions tested. For example, Tunicamycin-induced UPR^er did not alter HMGS-1::GFP distribution (Figure S1, C and D), whereas the same treatment induces the upregulation of a reporter for UPR^er (Figure S1, E and F). This strongly suggests that the upregulation of hmgs-1 is highly specific to mitochondrial stress.

We next aimed to test whether the HMGS-1::GFP protein is active in vivo. To assess this, CRISPR-Cas9 technology was used to generate a loss-of-function allele in the hmgs-1 gene. CAS-9 activity induces the insertion of a cassette replacing the sequence of the hmgs-1 gene from the middle of exon two through the middle of exon seven. This replacement resulted in the removal of the majority of the eight total exons of the hmgs-1 gene (Figure S2, A–C) introducing a frameshift after Serine 64 in the HMGS-1 protein and a stop codon 10 further amino acids downstream. Consistent with HMGS-1 having an essential function in the mevalonate pathway that supports organismal metabolism, growth, and development, animals homozygous for the hmgs-1 deletion allele died as embryos (Figure S2D). This lethality, along with the loss of most of the exons of the hmgs-1 gene including exons that encode for the active site of the enzyme, suggest that this hmgs-1 knockout is a complete loss-of-function allele. To determine the functionality of the HMGS-1::GFP construct in vivo, we introduced this construct in the background of the hmgs-1 knockout allele (Figure S2, B–E). Importantly, the HMGS-1::GFP construct was able to rescue the lethal phenotypes of the hmgs-1 knockout allele, resulting in the generation of worms that were successfully able to reach the adult stage (Figure S2, D and E). This rescue demonstrates the activity of the HMGS-1::GFP protein in vivo and suggests that the expression and distribution of this HMGS-1::GFP protein is similar to the endogenous HMGS-1 enzyme.

Regulation of the mevalonate pathway in mammals relies primarily on the control of the second enzyme of the pathway HMG-CoA reductase (HMGCR1). Thus, we next studied the putative ortholog of HMGCR1 in C. elegans, HMGR-1, by expressing a HMGR-1::GFP fusion protein under the control of the hmgr-1 promoter. Consistent with a previous report (Ranji et al. 2014), HMGR-1 protein was localized in several tissues, including the worm intestine and spermatheca (Figure 1E). Surprisingly, however, we found that spg-7 RNAi was incapable of altering HMGR-1 expression levels or spatial distribution (Figure 1F). This was in stark contrast to the strong upregulation of HMGS-1 during mitochondrial stress (Figure 1, B and D).

HMGS-1 upregulation could stem from several mechanisms, including the induction of hmgs-1 transcription or an inhibition in degradation of the protein. To determine whether an increase in the level of hmgs-1 transcription leads to the observed upregulation of the HMGS-1::GFP protein, we analyzed the promoter activity of hmgs-1 in vivo by expressing the hmgs-1 promoter directly fused to the sequence of the gfp gene. Upon mitochondrial stress, we detected an upregulation of the GFP signal in the intestine of worms harboring this reporter (Figure 1, G and H). Although we cannot rule out some contribution by post-translational modifications, the strong correlation both temporally and spatially between the promoter activity of hmgs-1 and HMGS-1::GFP protein distribution suggests that transcriptional mechanisms account for HMGS-1 upregulation. We then used qPCR to follow the levels of the endogenous transcripts of hmgs-1 and hmgr-1 genes in wild-type worms stressed by the spg-7 dsRNA in comparison to the empty vector control. In agreement with the results of the GFP fusion constructs, we found that hmgs-1 mRNA was upregulated by 2.4-fold in spg-7 RNAi in comparison to the empty vector control (Figure 1I). In contrast, and in accordance with our analysis of the fusion protein in Figure 1F, using qPCR we observed no significant change in the level of hmgr-1 mRNA (Figure 1I). The increase in HMGS-1::GFP protein was confirmed using a western blot analysis (Figure 1J) with a third type of mitochondrial stress, RNAi against the ATP synthase gene atp-3 (Ventura et al. 2009).

Finally, because HMGS-1 protein has not been well characterized in C. elegans, we examined whether this protein is indeed a functional ortholog of the mammalian HMGCS1 enzyme. HMGCS1 is capable of catalyzing the first dedicated biochemical reaction of the mevalonate pathway, i.e., the conversion of Acetyl-CoA and Acetoacetyl-CoA into HMG-CoA (Skaff and Miziorko 2010). To this end, we measured the activity of the C. elegans HMGS-1 enzyme in vitro. We expressed and purified HMGS-1 and human HMGCS1 proteins from bacteria (Figure S3A) and found that both proteins were capable of converting Acetyl-CoA and Acetoacetyl-CoA into HMG-CoA (Figure S3B). We also found that the treatment of C. elegans animals with mevalonate, a metabolite synthesized downstream to HMGS-1, fully rescued the severe phenotypes caused by hmgs-1 RNAi (Figure S4, A–F). The full rescue of hmgs-1 RNAi phenotypes by mevalonate suggests that, in C. elegans, HMGS-1 acts solely in the mevalonate pathway. Taken together, our in vitro and in vivo data indicate that HMGS-1 is a functional ortholog of human HMGCS1, acting as the first dedicated enzyme of the mevalonate pathway. Our results demonstrate that mitochondrial stress upregulates hmgs-1 transcription and consequentially HMGS-1 protein levels as a mitochondrial stress-specific response. Based on these results, we hypothesized that in the intestine, the primary metabolic tissue of C. elegans, the mitochondrial stress response upregulates HMSG-1 as part of a compensatory mechanism to restore mitochondrial homeostasis.

The UPR^mt transcription factors ATFS-1 and DVE-1 upregulate hmgs-1

The genetic amenability of C. elegans has led to the identification of a complex network of transcription factors and coregulators that form the basis of the mitochondrial stress response (Figure 2A). These include the transcription factors ATFS-1 (Nargund et al. 2012), DVE-1 (Haynes et al. 2007), SKN-1 (Munkácsy et al. 2016), and ZIP-2 (Pellegrino et al. 2014). Additional proteins that play a major role in the mitochondrial stress response are the transporter HAF-1 and the eIF2α kinase GCN-2, the latter of which regulates protein translation during mitochondrial stress (Baker et al. 2012).

Figure 2.

The UPR^mt, by activating ATFS-1 and DVE-1, upregulates HMGS-1. (A) A scheme demonstrating the molecular details of the mitochondrial stress response in C. elegans. (B) HMGS-1::GFP in isp-1(qm150) mutant worms. We used this background for the targeted screen. (C) hmgs-1 RNAi abolishes the GFP signal, demonstrating the specificity of the GFP construct and the hmgs-1 RNAi. (D–F) Knockdowns of atfs-1 (D), dve-1 (E), and rheb-1 (F) decrease the level of HMGS-1 upregulation. (G) The complete score of the targeted RNAi screen. In each condition tested, n (number of worms) = 100–120 worms. For each dsRNA tested, the results are presented as the percent of the total number of worms monitored. The screen was performed by the analysis of live intact worms under a fluorescence dissecting microscope that enabled the qualitative assignment of worms into the three different categories. Panels B–F are to the same magnification. Bars represent 250 and 500 μm (inserts).

To understand how mitochondrial stress response governs hmgs-1 upregulation, we conducted a targeted genetic screen to identify genes that could affect hmgs-1 expression. First, we used an isp-1(qm150) mutant background to elicit mitochondrial stress and to constitutively upregulate the HMGS-1::GFP protein (Figure 2B). We then systematically tested 36 genes previously characterized as having a role in the mitochondrial stress response (Munkácsy et al. 2016; Tian et al. 2016) for their effects on hmgs-1 expression. We compared the HMGS-1::GFP signal of worms treated with each dsRNA clone with the signal of the empty vector (Figure 2B) or hmgs-1 RNAi (Figure 2C) controls. In these analyses, we found that three dsRNA clones, corresponding to the sequences of the UPR^mt related genes atfs-1, dve-1, and rheb-1, attenuated the upregulation of HMGS-1::GFP in isp-1 mutant worms (Figure 2, D–F).

Previous studies have suggested that zip-2 (Pellegrino et al. 2014) and skn-1 (Munkácsy et al. 2016) may act downstream of atfs-1 in the mediation of some of its responses. Our genetic analyses, however, suggested that these genes were dispensable for the isp-1(qm150)-dependent upregulation of HMGS-1::GFP expression (Figure S5, A–C). Similarly, the knockdown of gcn-2, which plays a role independent of atfs-1 in the mitochondrial stress response, did not affect the upregulation of HMGS-1::GFP in the isp-1 mutant background (Figure S5D).

Additional characterized effectors of mitochondrial or metabolic stress including the daf-16 gene, the dlk-1/pmk-3/sek-3 pathway, and components of the mTOR pathway also did not affect HMGS-1::GFP upregulation (Figure 2G). Moreover, we did not detect a change in HMGS-1::GFP levels when knocking down cep-1, the ortholog of p53 that upregulates HMGCS1 in specific cancers (Freed-Pastor et al. 2012), or sbr-1, the ortholog of the SREBP proteins that mediate the cholesterol-dependent regulation of HMGCS1 (Horton et al. 2002). Because the nature and dynamics of the UPR^mt are somewhat affected by the type of stressor, we also used the spg-7 RNAi as a second, independent method of inducing the UPR^mt (Figure S6, A–C). Consistent with the approach of using the isp-1 mutation, we found that both atfs-1 (Figure S6D) and dve-1 (Figure S6E) were required for the upregulation of the HMGS-1::GFP protein. All other components of the UPR^mt (Figure S6, F–J) tested by the RNAi approach were dispensable for the upregulation of hmgs-1 expression by spg-7 RNAi. Nevertheless, because RNAi typically only partially reduces the expression of the corresponding gene, we cannot completely rule out the possibility that genes found not to alter the levels of the HMGS-1::GFP protein still have a minor role in HMGS-1 regulation.

These results strongly suggest that the molecular mechanisms of HMGS-1 upregulation are extremely context-dependent and highly regulated. Collectively, our screen identified a specific regulation of HMGS-1 by UPR^mt master regulators ATFS-1 and DVE-1 during mitochondrial stress.

Impaired HemeA and ubiquinone synthesis results in UPR^mt activation and a compensatory HMGS-1 upregulation response

Why would HMGS-1 be upregulated during mitochondrial stress? We hypothesized that this upregulation by the UPR^mt may serve as a compensatory response, allowing the cell to increase the synthesis of metabolites required to restore mitochondrial homeostasis. To test this hypothesis, we knocked down different branches of the mevalonate pathway while monitoring the level of UPR^mt induction and HMGS-1::GFP upregulation. We used the promoters of the mitochondrial chaperones hsp-6 and hsp-60 fused to the gfp gene (hsp-6_pr::gfp and hsp-60_pr::gfp, respectively) to assess the level of UPR^mt induction (Yoneda et al. 2004). Consistent with previous reports (Rauthan et al. 2013; Liu et al. 2014; Ranji et al. 2014), knocking down hmgs-1 or other upstream enzymes of the main branch of the mevalonate pathway did not lead to the induction of UPR^mt (Figure S7A). Similarly, RNAi against enzymes of the subbranches that account for transfer RNA (tRNA) modification, protein prenylation, or dolichol synthesis, although often resulting in potent visible phenotypes, did not result in an induction of the UPR^mt or HMGS-1 upregulation (Figure S7, A and B). In stark contrast, however, we found that knocking down an enzyme in the subbranch that facilitates hemeA biosynthesis (cox-15) triggered a potent UPR^mt induction (Figure 3, A and B), which was dependent on atfs-1 activity (Figure 3, C and D).

Figure 3.

Deficiency of HemeA and ubiquinone activates the UPR^mt. (A) The expression pattern of the UPR^mt reporter hsp-60_pr::gfp in worms fed an empty vector control. (B) Knockdown of the cox-15 enzyme, which acts in the synthesis of hemeA, upregulates hsp-60_pr::gfp expression. (C) hsp-60_pr::gfp in the background of an atfs-1(tm4525) loss-of-function mutation while fed an empty vector control. (D) In the background of atfs-1(tm4525), cox-15 RNAi did not upregulate the hsp-60_pr::gfp reporter. (E) The expression of the hsp-6_pr::gfp reporter in wild-type worms. (F) coq-1(ok749) loss activates the UPR^mt. (G) Activation of the UPR^mt, upon coq-1 loss, is dependent upon atfs-1 and dve-1 (H). (I) A thrashing assay revealed that activation of the UPR^mt, via atfs-1, does not protect the worms from the effect of coq-1 loss. n (number of worms) = 20 for each condition. Bars represent SE. Statistical significance was tested by the one-way ANOVA test using the SPSS software; *** P < 0.005; N.S., no significant difference. The details of the statistical analyses can be found in the Material and Methods section. Differences in worm size stem from the effect of the mutant or RNAi backgrounds. Panels A–D and E–H are to the same magnification. The bars represent 500 μm (A–D) and 250 μm (E–H).

Because both hemeA and ubiquinone serve as carriers of electrons in the ETC, their loss may be indicative of a shared response. To test this, we used loss-of-function mutations of coq-1 and coq-2, two genes encoding enzymes of the ubiquinone synthesis pathway, to block ubiquinone biosynthesis. We found that, while worms homozygous for coq-1 or coq-2 loss-of-function mutation arrested at the young adult stage, they exhibited a strong induction of the hsp-6_pr::gfp reporter (Figure 3, E and F and Figure S8, A and B). Interestingly, in coq-1 mutant animals, hsp-6_pr::gfp upregulation was dependent on atfs-1 (Figure 3G) and dve-1 (Figure 3H), whereas in the coq-2 mutant background this upregulation was dependent on atfs-1 (Figure S8C) but only marginally on dve-1 (Figure S8D). The effect of coq-2 was independent of the bacterial strains used as a food source and occurred equally in HT115 and OP50 strains (Figure S9, A–D).

If a feedback loop mechanism exists, a reduction in the level of specific metabolites of the pathway will activate a system that will upregulate enzymes responsible for the synthesis of the deficient metabolites. To examine this possibility, we analyzed the levels of hmgs-1 during reduced ubiquinone production by measuring the levels of HMGS-1::GFP in coq-2 loss-of-function mutant animals. In consistence with this model, we found that the loss of a functional ubiquinone synthesis pathway leads to HMGS-1::GFP upregulation (Figure S10, A and B) and that this upregulation is partly dependent on atfs-1 (Figure S10C). Although ATFS-1 has many different targets (Nargund et al. 2012), the upregulation of HMGS-1::GFP has a mechanistic rationale as a compensatory response in conditions in which the production of ubiquinone and hemeA is attenuated. Taken together, our results support a model of a compensatory mechanism that has evolved to cope with low levels of ubiquinone or hemeA electron carriers by inducing the UPR^mt response to upregulate the HMGS-1 enzyme.

To understand the physiological significance of UPR^mt activation following the block in ubiquinone synthesis, we measured worm thrashing as a functional readout for animal health. We found that a lack of ubiquinone synthesis significantly reduced the number of thrashes, presumably because of a reduction in respiration or in other mitochondrial functions that impede normal muscle activity. Although atfs-1 RNAi was efficient in blocking the UPR^mt signal in the coq-1 mutant background (Figure 3G), it did not significantly affect the level of worm thrashing (Figure 3I). This further supports a model in which the UPR^mt directly upregulates enzymes of the mevalonate pathway. Because the coq-1 allele used is a presumable null allele (Rodríguez-Aguilera et al. 2005), any upregulation of enzymes, either up or downstream of coq-1, would not be able to rescue the phenotypes of coq-1 deficiency.

UPR^mt requires the metabolism of the mevalonate pathway main branch and geranylgeranylation

The above experiments raise a conundrum: the deficiency in hemeA and ubiquinone production strongly elicits the UPR^mt (Figure 3, B and F), but the inactivation of the main branch, which presumably also results in hemeA and ubiquinone deficiency, fails to induce the UPR^mt (Figure S7A). It is possible, therefore, that some products of the mevalonate pathway are required for the UPR^mt. This hypothesis is in agreement with recent literature suggesting the necessity of a functional mevalonate pathway for UPR^mt activation (Rauthan et al. 2013; Liu et al. 2014; Ranji et al. 2014).

To test this hypothesis, we knocked down different branches of the pathway in the background of a constitutively activated UPR^mt response. We used the isp-1(qm150) mutation to induce UPR^mt and the hsp-6_pr::gfp reporter as the readout for the level of the UPR^mt signaling (Figure 4A). As a positive control, we used atfs-1 and dve-1 RNAis to reduce the expression of the UPR^mt reporter, as previously established (Figure 4, B and C). Next, we tested whether reduced expression of enzymes in the main branch of the pathway could affect the UPR^mt. We found that knocking down hmgs-1 or hmgr-1 attenuated the activation of the UPR^mt response in the isp-1 mutant background (Figure 4, D and E). In contrast, hmgs-1 RNAi did not attenuate the ER stress response (Figure S11, A–C), suggesting that the requirement of a functional mevalonate pathway is specific for the UPR^mt. Importantly, we were able to chemically rescue, in a concentration-dependent manner, the block in UPR^mt activation caused by hmgs-1 RNAi by supplementing mevalonate exogenously (Figure S12, A–F). This rescue directly suggests that one of the metabolites produced by the pathway downstream of mevalonate itself is required for UPR^mt activation.

Figure 4.

Activation of the UPR^mt requires geranylgeranylation. (A–F) The results of a targeted screen for the requirement of different branches of the mevalonate pathway for UPR^mt signaling. (A) The isp-1(qm150), hsp-6_pr::gfp, worm strain that was used in the screen. (B) Knockdowns of atfs-1 and dve-1 (C) were used as positive controls. (D) Knockdown of hmgs-1 or hmgr-1 (E) inhibit the UPR^mt. (F) Knockdown of the geranylgeranyltransferase ggtb-1 gene suppresses the UPR^mt. (G–J) A constitutively activated ATFS-1 protein in the background of the UPR^mt reporter hsp-60_pr::gfp. (G) An empty vector control. (H) atfs-1 RNAi reduces the level of UPR^mt activation in most tissues. The expression in the pharynx may stem from the resistance of this tissue to RNAi knockdown (Kumsta and Hansen 2012). (I) Knockdowns of hmgs-1 or ggtb-1 (J), although resulting in potent visible phenotypes, do not suppress the upregulation of hsp-60_pr::gfp by the activated ATFS-1 protein. Differences in worm size stem from the effect of the different RNAis. The atfs-1 and the isp-1 mutant worms were analyzed 4 and 6 days after the L1 stage, respectively. Panels A–F and G–J are to the same magnification. The bars represent 500 μm.

Next, we tested whether reduced expression of enzymes in any of the other subbranches of the mevalonate pathway was sufficient to inhibit UPR^mt induction. We found that knocking down the ggtb-1 gene, which encodes a type II geranylgeranyltransferase, inhibits UPR^mt induction (Figure 4F). In contrast, knocking down other subbranches of the mevalonate pathway did not attenuate the activation of the UPR^mt response (Figure S13).

The UPR^mt relies on the sequential activation of several proteins, including ATFS-1, which plays a cardinal role in the UPR^mt response. Therefore, impaired mevalonate pathway metabolism concomitant with a block of geranylgeranylation may impede ATFS-1 activation. To test the possible connection between mevalonate pathway metabolism and ATFS-1 activation, we used a strain harboring a constitutively activated allele of atfs-1 (atfs-1(et15)). In this mutant background, ATFS-1 protein translocates to the nucleus, where it exerts its activity constitutively even in the absence of mitochondrial stress (Rauthan et al. 2013; Pellegrino et al. 2014) (Figure 4G). As a control, we used the atfs-1 RNAi that blocked the UPR^mt signal almost completely in most of the tissues of these animals (Figure 4H). We found that, in this background, knockdown of hmgs-1 or ggtb-1 did not affect the level of UPR^mt induction (Figure 4, I and J), indicating that the requirement for mevalonate pathway metabolism and geranylgeranylation lies upstream or in parallel to ATFS-1 activation.

Taken together, our results suggest that the reliance of the UPR^mt response on a functional mevalonate pathway metabolism stems from the requirement for functional geranylgeranylation. Furthermore, our genetic analyses reveal that geranylgeranylation is required upstream or in parallel to ATFS-1 activation.

Inhibition of the UPR^mt does not stem from RHEB-1 inactivation

Geranylgeranylation is requisite for the activation of many small GTPases (Cherfils and Zeghouf 2013), underscoring the possibility of the involvement of a small GTPase in the activation of the UPR^mt response. A growing body of evidence suggests an evolutionarily conserved role for GTPases as facilitators of many stress responses that coordinate cellular metabolism (Bar et al. 2016). It is plausible, therefore, that in C. elegans, the activation of the UPR^mt relies on the function of a specific GTPase. Of note, in our targeted screen, apart from ATFS-1 and DVE-1, the only other gene whose RNAi altered the levels of HMGS-1 is the small GTPase rheb-1 (Figure 2F), which was previously identified as regulator of the UPR^mt (Haynes et al. 2007). Based on these two indicators, it is possible that impaired mevalonate pathway metabolism and loss of geranylgeranylation block RHEB-1 activation and consequentially the UPR^mt response.

To test this model, we examined whether mevalonate pathway metabolism is requisite for RHEB-1 activation in the context of UPR^mt. Similar to the hmgs-1 and ggtb-1 knockdown results, we found that rheb-1 knockdown somewhat modulates UPR^mt activation (Figure S14, A and B). This effect of rheb-1 takes place upstream or in parallel to atfs-1 activation (Figure S14, C–F). Next, we used the DVE-1::GFP reporter, previously shown to be a readout for RHEB-1 protein function (Haynes et al. 2007), to determine the level of RHEB-1 activation. In wild-type worms, DVE-1::GFP accumulates for unclear reasons at a few nuclei at the anterior and posterior sides of the worm (Figure S15, A and A’). As previously reported (Haynes et al. 2007), we found that RHEB-1 loss leads to an increased number of DVE-1::GFP positive nuclei at the posterior side of the worms (Figure S15, B and B’). The knockdowns of hmgs-1 and ggbt-1 result in severe phenotypes that demonstrate the potency of these RNAis, but these conditions did not alter the number of DVE-1::GFP nuclei in comparison to the control EV RNAi (Figure S15, A and C–E). Although it is possible that the distribution of DVE-1::GFP reporter represents only one aspect of RHEB-1 activation, this result suggests that impaired prenylation does not affect at least one function of RHEB-1, i.e., restricting DVE-1::GFP expression.

Because RHEB-1 did not appear to be the missing link between geranylgeranylation and the UPR^mt, we screened the family of small GTPases in C. elegans for suppressors of the UPR^mt response. In the background of a constitutively activated UPR^mt, i.e., the isp-1(qm150), hsp-6_pr::gfp background, we tested 49 out of the 53 predicted small GTPases encoded by the C. elegans genome (Reiner and Lundquist 2016). To rule out a possible effect of the method used, we used an additional approach to induce the UPR^mt, i.e., the spg-7 RNAi. Using these two approaches, we did not identify in our screen any gene, other than rheb-1, as being a requirement for UPR^mt activation (Table S3).

Mitochondrial stress induces a coordinated upregulation of three different subbranches

Our pathway-level genetic analyses demonstrated that proper mitochondrial function requires hemeA and ubiquinone biosynthesis. Moreover, we found that geranylgeranylation plays a critical role in the execution of the UPR^mt response. These results highlight the possibility that, along with HMGS-1 of the main branch, UPR^mt upregulates the levels of additional enzymes to increase the metabolic flux of selected subbranches. To address this possibility, we first examined available whole-genome expression data related to the UPR^mt (Nargund et al. 2012). Highly consistent with our genetic analyses, these expression studies identified enzymes of the hemeA, ubiquinone, and geranylgeranylation subbranches as enzymes that are potentially upregulated upon mitochondrial stress (Figure 5A). To examine this possibility, at a single gene resolution, we measured the levels of endogenous transcripts of different enzymes of the pathway, comparing N2 wild-type worms fed with either the spg-7 dsRNA or an empty vector control. In agreement with both our genetic analysis and the whole-genome studies, we found that mitochondrial stress specifically upregulates enzymes of the three subbranches (Figure 5A and Figure S16). These include an enzyme that catalyzes the biosynthesis of a metabolite essential for the hemeA biosynthesis (Y46G5A.2), the first enzyme of the ubiquinone synthesis subbranch (coq-1), and one enzyme that mediates the geranylgeranylation of proteins (M57.2). Next, we analyzed the expression and distribution of selected proteins encoded by the upregulated genes in vivo during mitochondrial stress. All of the upregulated genes we identified, except hmgs-1, lie within operons and therefore cannot be fused to a gfp sequence using standard cloning or fusion PCR methods. Nevertheless, one of these genes, coq-1, is available with a gfp sequence at its 3′ end in the TransgeneOme library of fosmids (Sarov et al. 2012). Therefore, we obtained and validated the integrity of the gfp-tagged coq-1 gene in the fosmid. Next, we generated transgenic worms harboring the coq-1::gfp fosmid. COQ-1::GFP expressing worms exhibited a weak, intestinally enriched GFP signal (Figure 5B) that is abolished by coq-1 RNAi (Figure S17, A and B). Consistent with our qPCR results, the COQ-1::GFP protein was upregulated during mitochondrial stress (Figure 5, C and D). Moreover, COQ-1::GFP upregulation was dependent on atfs-1 activity (Figure 5E) and the GFP signal of this construct was specific to the coq-1 gene (Figure S17, A–D). These results show that in addition to regulating HMGS-1, an enzyme in the main branch of the pathway, the UPR^mt, via afts-1 activation, can upregulate enzymes in specific subbranches. We propose that the coordinated upregulation of HMGS-1 along with enzymes of specific subbranches is a genetic program that has evolved to facilitate a protection from mitochondrial dysfunction.

Figure 5.

The mitochondrial stress response upregulates multiple branches of the mevalonate pathway. (A) Quantitative PCR (qPCR) analyses of the relative transcription levels of enzymes from different branches of the mevalonate pathway. The pathway metabolites are labeled in black. Enzymes labeled in green were analyzed by qPCR; enzymes in gray were not analyzed. Gray-blue arrows represent the flow of metabolites from the main branch to the end products. N2 worms grown on bacteria transformed with the empty vector, and spg-7 dsRNA plasmids were used for experiments. Dark-orange arrows mark enzymes that were suggested to be upregulated upon mitochondrial stress in a whole transcriptomic analysis (Nargund et al. 2012). Error bars represent SEM. * P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001 using two-tailed, unpaired Student’s t-tests. (B–E) The levels and distribution of a COQ-1::GFP reporter. (B) The distribution of the COQ-1::GFP protein in worms fed with an empty vector control. The COQ-1::GFP protein is enriched in the worm’s intestine. (C) RNAi for spg-7 upregulates COQ-1::GFP levels primarily in the intestine. (D) Dilution of the spg-7 dsRNA by half did not evidently affect its ability to upregulate the COQ-1::GFP protein. (E) atfs-1 RNAi blocks the upregulation of COQ-1::GFP in the spg-7 RNAi background. Panels B–E are to the same magnification. Bars represent 200 μm.

Discussion

We have found that mitochondrial dysfunction activates the UPR^mt signaling cascade to remodel cellular metabolism through the coordinated control of critical regulatory points in the mevalonate pathway. Three independent approaches, i.e., whole transcriptome expression profiling (Nargund et al. 2012), RNAi-based functional assays, and targeted qPCR analyses, demonstrated an upregulation in mevalonate pathway enzymes. These included the main branch enzyme HMGS-1, enzymes responsible for the synthesis of the ETC electron carriers ubiquinone and hemeA, and an enzyme that facilitates geranylgeranylation of proteins. The upregulation of the HMGS-1 enzyme relies on the activity of two transcription factors, ATFS-1 and DVE-1, previously shown to play a critical role in the mitochondrial stress response. We propose that this genetic circuit has evolved to upregulate the synthesis of electron carriers that could alleviate mitochondrial dysfunction, while in parallel enhancing geranylgeranylation, which is required for the full activation of the UPR^mt response.

One of the established roles of geranylgeranylation is to anchor one type of G-proteins, the small GTPases, to the membranes of different cellular compartments (Vögler et al. 2008). Several lines of evidence suggest that small GTPases play a role in the regulation of various stress responses (Bar et al. 2016). Therefore, we speculated that the inhibition of UPR^mt by a block in geranylgeranylation (Figure 4, D–F) stems from the inactivation of a small GTPase that is required for the UPR^mt. Our targeted screen, however, did not result in the identification of such a small GTPase. One possible interpretation of these results is that our screen either lacked the small GTPase essential for UPR^mt execution or, due to the partial effect of RNAi knockdown, missed this small GTPase. Alternatively, it is plausible that a functional overlap between two or more small GTPase proteins masks a requirement of a single small GTPase for the activation of UPR^mt. This redundancy, however, does not hold in treatment with statins that impair the activity of most, if not all, the small GTPases that need to undergo geranylgeranylation for their activity. Another possibility is that geranylgeranylation of a protein that is not a small GTPase is necessary for UPR^mt execution. For example, the lack of geranylgeranylation of the nuclear lamina protein lamin A, that normally is subjected to this type of modification (Davies et al. 2009), may block the translocation of proteins, including ATFS-1, into the nucleus.

The mevalonate pathway accounts for the synthesis of several biomolecules that play a role in diverse cellular processes, including the regulation of membrane fluidity, respiration, and post-translational modification of proteins. So far, most studies in mammals have been focused on the regulation of HMGCR1 (Brown and Goldstein 1980; Burg and Espenshade 2011) or the control of enzymes in the cholesterol synthesis subbranch (Gill et al. 2011; Foresti et al. 2013; Prabhu et al. 2016). Because HMGCR1 is the proposed rate-determining enzyme of the pathway in mammals, it might be regulated in other cases that require remolding of mevalonate pathway metabolism, for example in mitochondrial dysfunction. The lack of regulation of C. elegans HMGR-1 in mitochondrial stress may represent a nematode-specific property of the HMGR-1 regulatory network. Another possibility, however, is that we have discovered an evolutionarily conserved mechanism by which the UPR^mt controls the HMGS-1/HMGCS1 protein rather than controlling the HMGR-1/HMGCR1 enzyme. The molecular details of the UPR^mt in mammals are not well characterized (Haynes et al. 2013). Nevertheless, a recent study of combined transcriptomic and proteomic analyses of human cells exposed to EtBr has identified HMGCS1 but not HMGCR1 among the list of a 100 proteins whose levels are mostly affected by the UPR^mt (Bao et al. 2016). In contrast to our findings in C. elegans, HMGCS1 was not upregulated but downregulated upon UPR^mt activation. While HMGCS1 and HMGS-1 seem to be regulated in opposite directions, these two observations suggest an evolutionarily conserved mechanism in which the UPR^mt governs the levels of the first dedicated enzyme of the pathway, HMGS-1/HMGCS1. We thus propose that the regulation of HMGCS1/HMGS-1 levels plays a central role in the regulation of mevalonate pathway flux in normal physiology, in stress conditions, and probably in diseases of mitochondrial dysfunction.

In mammals, SREBP-1 upregulates HMGCS1 under conditions of low cholesterol (Horton et al. 2003), whereas p53 upregulates HMGCS1 during tumorigenesis (Freed-Pastor et al. 2012). We did not find a role for the respective C. elegans orthologs sbp-1 and cep-1 in the upregulation of HMGS-1 upon UPR^mt (Figure 2G). Our results suggest a context-dependent regulation of HMGCS1/HMGS-1 proteins by transcription factors dedicated to different physiological conditions. Thus, we predict that in humans the regulation of HMGCS1 in mitochondrial stress does not rely on the activity of SREBPs or p53. Instead, we propose that HMGCS1 is regulated by transcription factors dedicated to the UPR^mt response such as ATF4 (Bao et al. 2016) or ATF5 (Fiorese et al. 2016).

In mammals and in C. elegans, whole proteome studies have identified specific residues of HMGCS1/HMGS-1 proteins as undergoing phosphorylation (Van Hoof et al. 2009), acetylation (Choudhary et al. 2009), ubiquitination (Kim et al. 2011; Wagner et al. 2011, 2012), or SUMOylation (Sapir et al. 2014). These findings highlight the possibility that HMGCS1/HMGS-1 undergoes a complex post-translational regulation. In support of these findings, we have identified SIRT2.2 and AMPK-1 in a proteomic screen for HMGS-1 interactors in C. elegans (Sapir et al. 2014). The biological significance of these post-translational modifications and the relative functional weight of transcriptional vs. post-translational regulation of HMGCS1/HMGS-1 proteins will be important topics for future investigations.

Aside from upregulating HMGS-1 in the main trunk of the pathway, we found that C. elegans’ UPR^mt also upregulates enzymes that play a role in geranylgeranylation and the synthesis of electron carriers. The identification of M57.2, the ortholog of human RABGGTA (Rab geranylgeranyltransferase α subunit), as a gene upregulated by spg-7 RNAi (Figure 5A) suggests that the cellular protection of UPR^mt involves an increase in geranylgeranylation. Because UPR^mt activation requires a functional geranylgeranylation subbranch (Figure 4F), M57.2 upregulation may facilitate a stronger UPR^mt response in cases of mitochondrial dysfunction. A functional mevalonate pathway was shown to be requisite for the activity of miRNAs in C. elegans (Shi and Ruvkun 2012), raising the possibility that the block of UPR^mt stems from the dysregulation of miRNAs. Nevertheless, this requirement of mevalonate pathway metabolism for the activation of miRNAs was mapped to the dolichol synthesis subbranch (Shi and Ruvkun 2012), which, unlike the geranylgeranyl subbranch (Figure 4F), does not alter the pattern of UPR^mt activation (Figure S13). Importantly, knockdown of the M57.2 gene results in the activation of the UPR^er (Morck et al. 2009). Thus, by controlling the levels of M57.2 and consequentially the level of geranylgeranylation, the UPR^mt response can positively regulate its own activation while inhibiting the activation of the UPR^er response.

The upregulation of enzymes that are responsible for the synthesis of ubiquinone and hemeA electron carriers may be part of a compensatory mechanism that attempts to restore mitochondrial hemostasis. In yeast, a growing body of evidence suggests that mitochondrial malfunction results in the activation of transcription factors that increase the level of ubiquinone-synthesizing enzymes [review in González-Mariscal et al. (2014)] In humans, several endogenous or environmental conditions lead to ubiquinone deficiency or overproduction. However, beyond the activity of PPARα, the molecular mechanisms that control these effects remain largely unknown (Bentinger et al. 2010). Because CoQ10 is the end product of the subbranch that synthesizes ubiquinone in humans, CoQ10 supplementation is a suggested treatment to circumvent the adverse side effects of statins. Our study suggests that the upregulation of enzymes from the subbranches that produce ubiquinones, hemeA, and geranylgeranyl moieties represents a compensatory mechanism that is activated when the levels of these metabolites are reduced. Based on these data, we propose that to overcome the adverse effects of statins, a clinical approach should focus on the activation of geranylgeranylation and the upregulation of hemeA synthesis, along with supplementation with CoQ10. Supporting this hypothesis is a report that suggests that supplementation with either mevalonate or geranylgeranyl pyrophosphate, but not with other metabolites of the pathway, can mitigate the antitumorigenic effects of statins in culture (Jiang et al. 2014). This report also suggests that geranylgeranyl-transferase II inhibition slows cancer growth, implying that impaired geranylgeranylation is a mechanism underlying the antitumorigenic activity of statins. Our results show, however, that blocking of geranylgeranylation inhibits the UPR^mt and potentially blocks the compensatory mechanism of ATFS-1 activation and mevalonate pathway upregulation.