Mitochondrial ubiquinone–mediated longevity is marked by reduced cytoplasmic mRNA translation

This study uses polysomal RNA sequencing to show that the translational efficiency of specific mRNA changes in long-lived Caenorhabditis elegans with reduced ubiquinone synthesis.


Introduction
Slowing down mitochondrial metabolism is a well-known means of increasing the lifespan of multiple species (Dillin et al, 2002;Copeland et al, 2009;Houtkooper et al, 2013;Breitenbach et al, 2014). In C. elegans, one such model is the clk-1(qm30) mutant that harbors a deletion in the gene encoding a ubiquinone (UQ) biosynthesis enzyme (Felkai et al, 1999). UQ is a redox-active lipid that plays a central role in the electron transport chain and mitochondrial oxidative phosphorylation (OXPHOS) by carrying electrons from complexes I and II to complex III (Stefely & Pagliarini, 2017). In addition to lifespan extension and decreased respiration, clk-1 (qm30) mutants are developmentally delayed and have reduced progeny (Felkai et al, 1999). Mammalian Clk1, known as Mclk-1 or Coq7, is also involved in the lifespan (Liu et al, 2005;Stepanyan et al, 2006;Takahashi et al, 2014). Transgenic mice, rescued from embryonic lethality via the transgenic expression of mouse clk-1, with reduced Clk1 expression live longer and have smaller bodies than WT mice, demonstrating a conserved role for clk-1 in longevity across species (Takahashi et al, 2014). However, it is not fully understood how the reduction of UQ biosynthesis extends the lifespan of these animals.
Until recently, CLK-1 protein was thought to reside exclusively in the mitochondria. However, a distinct nuclear form of CLK-1 has been reported to independently regulate lifespan by mitochondrialnuclear retrograde signaling (Monaghan et al, 2015). Adding back a copy of the clk-1 gene that lacks the mitochondrial targeting signal (MTS) to the clk-1 deletion mutant led to a subtle but significant reduction in the lifespan extension phenotype observed in the full clk-1 deletion mutant (Monaghan et al, 2015). However, the existence and function of a nuclear form of CLK-1 remains controversial. Other studies fail to detect any nuclear localization of the MCLK1/CLK-1 proteins or any biological activity of a C. elegans CLK-1 protein devoid of an MTS .
The importance of communication between stressed mitochondria and the rest of the cell is becoming increasingly appreciated (Quiros et al, 2016). One such reaction to stress is carried out by the mitochondrial unfolded protein response (UPR mt ) (Durieux et al, 2011;Baker et al, 2012;Houtkooper et al, 2013). GCN-2, an eIF2α kinase that modulates cytosolic protein synthesis, is phosphorylated in response to the activated UPR mt in clk-1(qm30) mutants in a manner required for lifespan extension (Baker et al, 2012). Another nonmitochondrial factor, TATA-binding protein-associated factor 4 (TAF-4), is a transcription factor that has been implicated in the extension of clk-1 (qm30) mutant lifespan (Khan et al, 2013). TAF-4 was identified in an RNAi screen for transcription factors required for the lifespan extension of clk-1(qm30) mutants (Khan et al, 2013). Beyond clk-1, TAF-4 was also required for the lifespan extension phenotype of two mutants of the mitochondrial electron transport chain (isp-1 and tpk-1) (Khan et al, transcription complex and is best known for its role in transcriptional silencing in early embryogenesis by becoming sequestered in the cytoplasm because of phosphorylation by OMA-1 (Oocyte MAturation defective) (Walker et al, 2001). The loss of TAF-4 in C. elegans suppresses the lifespan extension phenotype induced by clk-1 mutations; however, the mechanism through which this happens is not known.
Reducing protein synthesis is another well-known means by which eukaryotes extend the lifespan (Hansen et al, 2007;Pan et al, 2007). Caloric restriction, amino acid reduction, and knocking down myriad factors involved in mRNA translation (such as ribosomal proteins, elongation/initiation factors, or tRNA synthetases) all increase the lifespan (Masoro, 2000;Min & Tatar, 2006;Hansen et al, 2007;Pan et al, 2007). The genetic or pharmacological inhibition of the mechanistic target of rapamycin (mTOR)/nutrient sensing pathway or insulin receptor signaling pathway is also marked by reduced mRNA translation rates and a suppression of polyribosome formation (Genolet et al, 2008;Stout et al, 2013). Proteomic analysis of the daf-2 mutants previously revealed a substantial reduction of ribosomal proteins, translation factors, and protein metabolism components coupled to a repression of polyribosome formation (Stout et al, 2013). Moreover, our previous work demonstrated that similarly to daf-2 mutants, mutations in clk-1 worms also result in a strong repression of polyribosome formation (Essers et al, 2015). This led us to hypothesize the existence of unexplored regulatory links between dysfunctional mitochondria and cytoplasmic protein synthesis that are contributing to eukaryotic lifespan extension. In particular, we investigated which subsets of mRNAs are preferentially translated in the longer lived C. elegans with impaired UQ biosynthesis.

Results
RNAseq of clk-1(qm30) strains shows major changes in transcription To uncover how the transcriptome of C. elegans changes on deletion of clk-1 and the influence of the clk-1 mitochondrial localization signal, we performed next-generation sequencing of total RNAs (depleted of ribosomal RNA [rRNA]) isolated from worms at the L4 stage that precedes young adulthood. Total RNA pools depleted of rRNA isolated from clk-1(qm30) mutants were compared with (1) clk-1 (qm30) +WT mutants rescued with the WT clk-1 and (2) clk-1(qm30) +nuc mutants with predominantly the nuclear form of clk-1 (Fig 1A) that was described in Monaghan et al (2015). The clk-1(qm30) +WT mutants have WT phenotypes and are rescued with fully functioning clk-1, whereas the clk-1(qm30) +nuc mutants have clk-1 only residing in the nucleus and not in the mitochondria (Monaghan et al, 2015). The principal component analysis (PCA) plot and correlation matrix show the expected clustering of the biological triplicates of each strain (Fig 1B-C). Major changes were observed in the RNA pool of the clk-1(qm30) +WT mutants compared with either clk-1(qm30) or clk-1(qm30) +nuc , with 9,626 and 8,685 genes being differentially expressed, respectively. Far fewer differentially expressed genes were measured when comparing clk-1(qm30) with clk-1(qm30) +nuc , with only 247 genes up-regulated and 571 genes down-regulated. The subtle changes in RNA expression are consistent with the observation that introducing the nuclear form of clk-1 in clk-1(qm30) results in only small phenotypic changes compared with the clk-1(qm30) mutants (Monaghan et al, 2015;Liu et al, 2017).
Transcripts encoding the translation machinery are strongly reduced in clk-1(qm30) We next identified the biological pathways that are altered in clk-1(qm30) worms using The Database for Annotation, Visualization and Integrated Discovery (DAVID) and Gene Ontology (GO) analysis (Huang da et al, 2009a;Huang da et al, 2009b). Among the most up-regulated processes in clk-1(qm30) are metabolic pathways (e.g., oxidation-reduction pathway) and ion transport (Fig 1D). A substantial proportion (25%) of significantly down-regulated processes in clk-1(qm30) mutants is involved in reproduction and development ( Fig 1D). Developmental delay and reduced progeny have been extensively reported as linked to the longevity phenotype of clk-1(qm30) mutants (Larsen et al, 1995;Tissenbaum & Ruvkun, 1998). Remarkably, another 20% of down-regulated GO terms are involved in translation and mRNA processing ( Fig 1D). In total, 38 genes encoding ribosomal proteins, initiation, and elongation factors are down-regulated in the clk-1(qm30) mutants compared with clk-1 (qm30) +WT . In addition, 28 genes specifically involved in biogenesis and processing of ribosomes were down-regulated in the clk-1 (qm30) worms. Similar down-regulation of ribosomal proteins and initiation/elongation factors was observed in the clk-1(qm30) +nuc mutants. These observations suggest that mRNA translation is repressed in the clk-1(qm30)/ +nuc mutants and are in line with the repressed polysome profiles we observed before (Essers et al, 2015).
There is a minor role for nuclear clk-1 in transcriptional changes Because the nuclear form of clk-1 was suggested to associate with the chromatin in the nucleus (Monaghan et al, 2015), we expected changes in the transcriptome in the mutant expressing exclusively the nuclear form of clk-1. However, differentially expressed genes in clk-1(qm30) +nuc were very similar to the differentially expressed genes clk-1(qm30) (Fig 2A). There were 3,720 genes similarly up-regulated in the clk-1(qm30) +nuc as in the total clk-1(qm30) mutants compared with clk-1(qm30) +WT and 4,397 genes similarly down-regulated in both strains compared with the WT control ( Fig 2B). The 292 up-regulated genes and 275 down-regulated genes in the clk-1(qm30) +nuc strain that were not up-or down-regulated in the clk-1(qm30) strain did not lead to significantly clustered GO terms. The 704 genes up-regulated in specifically the clk-1(qm30) strain and not in the clk-1(qm30) +nuc strain were involved in biological processes such as ion transport and oxidation-reduction ( Fig 2C). The 805 genes down-regulated in the clk-1(qm30) strain but not in clk-1(qm30) +nuc were involved in biological processes such as reproduction and response to DNA damage stimuli ( Fig 2C). These processes also appeared when comparing the clk-1(qm30) mutants with WT worms.
Translation rates are reduced in clk-1(qm30) worms To provide a snapshot of the translational status in the clk-1(qm30) strains, we performed polysome profiling of C. elegans at the L4 stage. Polysome profiling fractionates total cell lysate over a sucrose gradient by density and enables the distinct measurement of 40S small ribosomal subunits, 60S large ribosomal subunits, 80S monosomes (an mRNA molecule with a single ribosome attached), and multiple polysomes (mRNA molecules with multiple ribosomes attached). Polysome profiles of clk-1(qm30) reveal a strong repression of both monosomal and polysomal peaks compared with clk-1(qm30) +WT , suggesting a global repression of mRNA translation (Fig 3). Again, the clk-1(qm30) +nuc mutants show similar yet slightly less dramatic differences compared with WT versus the total clk-1 (qm30) mutant compared with WT (Fig 3). Because the clk-1(qm30)+ nuc does not show striking differences in transcription compared with the total clk-1(qm30) mutant, we focused our remaining studies on comparing clk-1(qm30) with clk-1(qm30) +WT . The polysome profiles, together with the RNAseq results, show that clk-1(qm30) mutants are marked by repressed mRNA translation. Further validation of translational repression in clk-1(qm30) mutants is shown by quantitative PCR (qPCR) analysis, demonstrating that several mRNAs encoding initiation factors and ribosomal proteins are downregulated in clk-1(qm30) mutants compared with WT worms (Fig S1).
Total RNA does not mirror highly translated polysomal RNA In addition to providing a global overview of mRNA translation rates, polysome profiling allows the specific isolation of the RNAs associated with the polyribosomes for comparison with the total RNAs in different clk-1(qm30) strains ( Fig 4A). The PCA plot and correlation matrix show clustering of polysomal and total RNA of the biological triplicates of each strain ( Fig S2). We observed that changes in the polysomal RNA of clk-1(qm30) compared with the polysomal RNA of clk-1(qm30) +WT were not simply mirroring the changes observed in the total mRNA pools of these mutants ( Fig 4B). For instance, there are 434 genes specifically up-regulated in the polysomal RNA pool carrying highly translated messages that are not up-regulated in the total RNA pool of clk-1(qm30) mutants ( Fig 4C). A larger set of 1,657 genes is up-regulated in both polysomal and total RNA, whereas the largest part (2,767 genes) is exclusively upregulated in the total RNA pool ( Fig 4C). A similar pattern is observed with down-regulated genes in clk-1(qm30) mutants, with 466 genes detected in the polysomal RNA pool, 2,574 in both polysomal and total RNA pools, and 2,628 in the strictly total RNA pool ( Fig 4C).
OXPHOS transcripts are highly enriched in polysomal RNA of clk-1 (qm30) strains To determine the functional relevance of the differentially expressed RNAs in total and polysomal RNA pools, we performed Top 50 down-regulated GO terms (left graph) and upregulated GO terms (right graph) in clk-1(qm30) versus clk-1(qm30) +WT (left column) and clk-1(qm30) +nuc versus clk-1(qm30) +WT (right column). Proportions of GO terms associated with translation and mRNA processing are depicted in yellow, development and reproduction in red, metabolism in green, and ion transport in blue. DAVID GO analysis. For instance, ion transport processes are upregulated in the total mRNA of clk-1(qm30) mutants. However, the polysomal expression data indicate that despite this higher abundance in the total pool, mRNAs coding for ion transport are not more highly translated in clk-1(qm30) mutants ( Fig 4D). In contrast, mRNAs involved in metabolic processes such as oxidation-reduction, glycolysis, and lipid metabolism are substantially up-regulated in both the total and the polysomal RNA pools ( Fig  4D). This suggests a metabolic compensation for the reduction of OXPHOS capacity by glycolysis and lipid metabolism in the clk-1 (qm30) mutants. Processes that were exclusively up-regulated in the polysomal RNA pool (and thus exclusively highly translated) in Figure 3. Repressed polysome profiles in the clk-1(qm30) ( ±nuc ) mutants. (A) Representative traces of polysome profiles of clk-1(qm30) strains harvested at the L4 stage, when lysate is normalized to total protein levels of 500 μg. The monosomal peak and polysomal peaks (P1-P6) are indicated. (B) Quantification of polysome peak sizes (AUC). The fold change is represented compared with P1 of the clk-1(qm30) +WT . All peaks of clk-1(qm30) +WT are significantly different from both clk-1(qm30) +nuc and clk-1(qm30). No peaks were significantly different between clk-1(qm30) and clk-1(qm30) +nuc . Error bars represent mean ± SD. Significance was tested with t test and P-values were adjusted to correct for multiple testing using the Holm-Sidak method, with α = 0.05. (A) Fold changes of differentially expressed genes in clk-1(qm30) versus clk-1(qm30) +WT plotted against fold changes of clk-1(qm30) +nuc versus clk-1(qm30) +WT shows a minor role for nuclear clk-1 in transcriptional changes. Colors of the individual data points correspond to the colors of the groups of genes in the Venn diagram in (B). (B) Venn diagram shows that most genes are similarly up-regulated (3720) and down-regulated (4397) in the clk-1(qm30) +nuc and clk-1(qm30) worms when comparing with clk-1 +WT . There were 292 genes exclusively up-regulated in clk-1(qm30) +nuc (blue) and 704 genes exclusively in clk-1(qm30) (green) compared with clk-1 +WT . Furthermore, 175 genes were down-regulated exclusively in clk-1(qm30) +nuc (red) and 805 genes down-regulated exclusively in clk-1(qm30) (orange) compared with clk-1 + WT . Differential expressions in (A) and (B) are with a threshold-adjusted P value < 0.01. (C) Significant Cluster GO Enrichments (threshold Enrichment Score > 3) associated with the 704 genes specifically up-regulated (green) and 805 down-regulated (orange) in clk-1(qm30) strain. There were no significant Cluster GO Enrichments for genes exclusively up or down-regulated in clk-1(qm30) +nuc .
clk-1(qm30) mutants are involved in ATP hydrolysis, protein import into mitochondria, and mitochondrial electron transport ( Fig 4D). More specifically, 50 transcripts encoding proteins involved in mitochondrial OXPHOS were enriched in clk-1(qm30) polysomes (Table 1). This up-regulation does not seem complex-specific because mRNAs coding for proteins of all OXPHOS complexes (I-V) were enriched in clk-1(qm30) polysomes. These data suggest a model whereby proteins involved in ATP production and the import of these proteins into the mitochondria are highly translated despite lower overall expression levels in clk-1(qm30) mutants compared with WT to attempt to overcome their energy deficit.
RNAs coding for ribosomal proteins are substantially reduced in clk-1(qm30) polysomes An even more substantial down-regulation of transcripts involved in translation and ribosome biogenesis processes was revealed in the polysomal RNA pool of clk-1(qm30) deficient worms compared with clk-1(qm30) +WT , confirming the repression of translational machinery proteins (Table 2). In the polysomal RNA pool of clk-1 (qm30) mutants, 62 ribosomal proteins, 16 elongation/initiation factors, 14 amino-acyl tRNA synthetases, and 10 other proteins involved in mRNA processing were down-regulated (Table 2). These results are reminiscent of the proteomic analysis demonstrating a significant reduction of ribosomal proteins and translation factors in the daf-2 insulin receptor mutants (Stout et al, 2013). Furthermore, a subset of these transcripts coding for proteins involved in translation (21 ribosomal proteins, 11 amino-acyl tRNA synthetases, and eight other elongation/initiation factors) was even down-regulated in the total pool of RNA. This indicates that the down-regulation of translation machinery is partially regulated in a transcriptional manner. Interestingly, we observed strong downregulation of fib-1, a gene coding for a protein involved in the regulation and maturation of rRNA, in both total and polysomal RNA ( Fig 4E). The FIB-1 protein resides in the nucleoli where ribosome biogenesis initiates. It was previously shown that small nucleoli are a cellular hallmark of longevity and metabolic health (Buchwalter & Hetzer, 2017;Tiku et al, 2017). A FIB-1::GFP reporter strain for nucleoli revealed reduced FIB-1 levels in other long-lived C. elegans strains (Tiku et al, 2017). In line with this and the RNA-seq data presented here, the clk-1(qm30) worms also show reduced FIB-1 levels compared with WT N2 worms that is accompanied by reduced nucleolar size (Fig 4F and G). Taken together, it appears that an attempt to overcome an energy deficit in clk-1(qm30) mutants is coupled to the suppression of ribosome biogenesis and mRNA translation, which would conserve flagging energy for other critical cell processes.

Differential translational efficiency (TE) of mRNA reveals a role for TOR signaling
Because there are mRNAs that are selectively being translated in the clk-1(qm30) mutant or the WT independently of their abundance in the total mRNA pool, we calculated the TE of individual mRNA transcripts by taking the ratio between total RNA and polysomal RNA (Fig 5A). DAVID GO analysis on transcripts with high or low TE revealed similar biological processes as observed in the groups of genes specifically up-and down-regulated in the polysomal RNA presented in Fig 4D (Fig 5B). The list of genes with significantly different TEs was then analyzed using the "worm longevity pathway" in Kyoto Encyclopedia of Genes and Genomes (Ogata et al, 1999) using Pathview (Luo & Brouwer, 2013), to investigate the overlap in longevity pathways known in C. elegans. Besides hits in the UPR mt and stress resistance, both already described in clk-1(qm30) mutants (Durieux et al, 2011), the results also revealed possible involvement of the mTOR signaling pathway in the clk-1(qm30) mutants. Hence, we also checked the mTOR signaling pathway in Kyoto Encyclopedia of Genes and Genomes Pathview. In C. elegans, homologs of TOR pathway genes all extend the lifespan upon siRNA knockdown (Korta et al, 2012). Interestingly, our RNA-seq data show that let-363 (homologue of TOR) is preferentially translated in WT but not in clk-1(qm30) mutants (Fig 5C and D). Other members of the TOR complex 1 (TORC1), including daf-15 (RAPTOR), clk-2 (TEL2), and R10H10.7 (TTI1), also showed decreased TE in clk-1(qm30) mutants (Fig 5C and D). Moreover, our results show that a key player in autophagy, lgg-1, is preferentially translated in clk-1(qm30) mutants but not in the clk-1(qm30) +WT ( Fig  5C and D).
taf-4 is required for repressed polysome profiles in clk-1(qm30) TAF-4 is one of few transcription factors reported to be critical for the clk-1(qm30) longevity phenotype (Khan et al, 2013). To test if TAF-4 has a role in polysome formation in clk-1(qm30) worms, we performed RNAi of taf-4 in the clk-1(qm30) mutants. Here, we show that RNAi of taf-4 significantly increases the monosomal and polysomal peaks of clk-1(qm30) compared with the control empty vector HT115 (Fig 6A and C), suggesting this transcription factor plays a direct role in the suppression of polyribosome formation upon the loss of UQ biosynthesis. This increase is not observed in the polysome profiles upon taf-4 RNAi in the clk-1(qm30) +WT ( Fig   Figure 4. Analysis of total and polysomal RNA clk-1(qm30) and clk-1(qm30) +WT . (A) Schematic representation of polysomal fraction (blue) used for RNA isolation and total RNA (green) from C. elegans mutants clk-1(qm30) and clk-1(qm30) +WT that were used for RNAseq. (B) Fold changes of differentially expressed genes in the total RNA of clk-1(qm30) versus clk-1(qm30) +WT plotted against fold changes in polysomal RNA of clk-1(qm30) versus clk-1(qm30) +WT . Colors of the individual data points correspond to the colors of the groups of genes in the Venn diagram in (C). (C) Venn diagram showing 1657 genes were similarly up-regulated and 2574 genes down-regulated in both the polysomal as the total RNA in the clk-1(qm30) strain compared with clk-1 (qm30) +WT . A total of 434 genes were exclusively up-regulated in the polysomal RNA (blue), whereas 2767 genes were exclusively up-regulated in the total RNA (green). Furthermore, 477 genes were exclusively down-regulated in the polysomal RNA (red) and 2628 genes were exclusively down-regulated in the total RNA. Differential expressions (B, C) are with a threshold-adjusted P value < 0.01. (D) Significant Cluster GO Enrichments (threshold Enrichment Score > 3) associated with the genes specifically up and down-regulated in groups of genes in (C) (colors of the bars correspond again to the color of the groups in (C)). (E) Fib-1 is reduced in both the total and polysomal RNA of clk-1(qm30) compared with clk-1(qm30) +WT . (F, G) FIB-1 levels are reduced in clk-1(qm30) worms compared with N2 worms. Error bars represent mean ± SD, significance was tested with t test, ***P < 0.0005. 6B). These data suggest that TAF-4 loss partially restores polyribosome formation in clk-1(qm30) mutants similar to the previously reported attenuation of the longevity phenotype (Khan et al, 2013). Because TAF-4 is a transcription factor, we looked at changes in expression of specific genes involved in translation upon taf-4 RNAi in the clk-1(qm30) mutants. We observed significant upregulation of the translation initiation factor eif-1 when taf-4 is knocked down in clk-1(qm30) worms, suggesting that TAF-4 has a direct role in repressing eif-1 expression (Fig 6D). Eif-1 is important for translation initiation, and yeast eif-1 mutants (sui1 − ) show a reduction of polysomes compared with WT (Yoon & Donahue, 1992). Moreover, the reduction of eif-1 extends lifespan in N2 worms (Wang et al, 2010). Taken all together, these results suggest that the reduction of the translation machinery observed in clk-1(qm30) mutants may at least in part be driven by TAF-4 repression, and may help to explain the previously described link between TAF-4-and clk-1(qm30)-mediated lifespan extension (Khan et al, 2013).

Discussion
This study was initiated following our previous observation that the polysome profiles of clk-1(qm30) mutants revealed the same repression of polysomes as in the daf-2 mutants (Essers et al, 2015). We continued this line of inquiry by isolating and sequencing polysomal and total RNAs from clk-1(qm30) mutants and compared them with clk-1(qm30) +WT worms. These results demonstrated that compared with clk-1(qm30) +WT , very few mRNAs coding for components of the translation machinery such as initiation factors, elongation factors, and ribosomal proteins are present on clk-1 (qm30) polyribosomes. In contrast, the clk-1(qm30) polyribosomes reveal a strong preference for translating RNAs coding for OXPHOS proteins. The reduction in the TE of mRNAs coding for the translation machinery is highly reminiscent of the proteomic landscape we described in the daf-2 mutants, where there is a scarcity of ribosomal proteins (RPs) yet also no appreciable reduction in total RP mRNA expression (Depuydt et al, 2013;Essers et al, 2015). In addition, the small nucleoli in clk-1(qm30) worms confirming reduced ribosome biogenesis is a hallmark of cellular aging that was previously observed in other longevity mutants (Tiku et al, 2017). Taken all together, the results suggest that reduced ribosome biogenesis and shutting down protein synthesis by altering the TE of mRNAs coding for the translation machinery is a common mechanism used by several different systems that drive lifespan extension. Our results go on to suggest that the proposed nuclear form of CLK-1 has a relatively small effect on alleviating the polyribosome reduction observed in the clk-1(qm30) mutants. The subtle differences in lifespan and development previously described (Monaghan et al, 2015) might be explained by reestablishing expression of some reproduction genes in the clk-1(qm30) +nuc mutant, which are well known to associate with lifespan (Hsin & Kenyon, 1999;Monaghan et al, 2015). Because it has been shown that under certain conditions proteins may be imported into the mitochondria without any MTS (Ruan et al, 2017), it is possible that a small proportion of the nuclear CLK-1 enters the mitochondria without an MTS. The present study does not disprove the existence of nuclear form of CLK-1; however, an obvious function of a nuclear CLK-1 is not apparent with these data.
The inhibition of respiration in the clk-1(qm30) mutants triggers a retrograde response by up-regulating cell-protective and metabolic genes (Cristina et al, 2009). Our work is in line with these findings and in this study, we confirm the importance of retrograde communication between stressed mitochondria and the rest of the cell. In Drosophila, it was shown that dietary restriction-induced longevity was regulated by activation of the translational repressor  4E-BP, which is the eukaryotic translation initiation factor 4E binding protein (Zid et al, 2009). This resulted in increased ribosomal loading of nuclear-encoded OXPHOS mRNAs. This high TE of mitochondrial proteins was crucial for the dietary restrictioninduced lifespan extension because inhibition of individual mitochondrial subunits of OXPHOS complexes diminished the lifespan extension (Zid et al, 2009). Another study showed that reduction of cytosolic protein synthesis could suppress ageingrelated mitochondrial degeneration in yeast (Wang et al, 2008). This postponed mitochondrial degeneration upon reducing cytoplasmic protein synthesis together with our results suggesting a robust strategy where mitochondria communicate with the cytoplasmic translation machinery to equilibrate cellular energy generated by one compartment and used in another. Interestingly, the mitochondrial dysfunction observed in young Mclk1 +/− mice was shown to paradoxically result in an almost complete protection from the age-dependent loss of mitochondrial function later in life (Lapointe et al, 2009). The increased TE of mitochondrial mRNAs we observed here in the mitochondrial clk-1(qm30) mutants could induce this physiological state that ultimately develops long-term beneficial effects for aging. Our results found that mRNAs coding for OXPHOS complexes I-V substantially enriched on polysomes of clk-1(qm30) worms. Given the marked reduction of total polysomes in these mutants, we can assume that most of the few remaining polysomes are present solely to maintain OXPHOS complexes at viable levels. This translational up-regulation of OXPHOS mRNAs on ribosomes was also recently observed by ribosome profiling in the context of host shutoff during vaccinia virus infection in human cells (Dai et al, 2017). This host shutoff facilitates resource availability for viral evasion and is coupled to a global inhibition of protein synthesis in the host (Bercovich-Kinori et al, 2016). In addition, the authors propose that this translational up-regulation of OXPHOS transcripts is due to their relatively short 59 UTRs, which are known to play a regulatory role in TE (Araujo et al, 2012;Dai et al, 2017). A similar mechanism coupling energy metabolism and translational control is observed in the present study in the context of longevity. Highlighted in grey are the transcripts that were up-regulated in both the polysomal RNA as the total pool of RNA.
Besides OXPHOS transcripts, this study also identified an increased TE for the autophagy gene lgg-1, the C. elegans orthologue of LC3 in humans. Autophagy is an indispensable function in organisms undergoing lifespan extension. In the absence of bec-1, the C. elegans orthologue of the autophagy gene APG6/VPS30/ beclin1, daf-2-mutant worms fail to undergo lifespan extension (Melendez et al, 2003). Reducing bec-1 in worms subject to caloric restriction (eat-2 mutants) or inhibition of the mTOR pathway (reducing let-363 [TOR] or daf-15 [RAPTOR] expression) also has the same effect of preventing lifespan extension (Hansen et al, 2008). These findings are in line with our observation that mTOR is translationally repressed in clk-1(qm30) mutants ( Fig 5D) and fits well with the notion that a reduction of growth can often lead to lifespan extension, in line with current theories proposing hyperfunction as a driving force in aging (Blagosklonny, 2006;Blagosklonny & Hall, 2009). Furthermore, our results suggest The blue data points represent transcripts with high TE in clk-1(qm30) worms being shifted from the total RNA to the highly translated polysomal RNA. The red data points represent transcripts that have low TE in clk-1(qm30) being shifted from polysomal RNA to the total RNA. Differentially translationally regulated genes in red and blue with threshold-adjusted P value < 0.01. (B) Significant Cluster GO Enrichments (threshold Enrichment Score > 3) associated with significantly different TEs (colors of the bars correspond again to the data points in (A)). (C) TE of individual transcripts involved in TOR pathway in clk-1(qm30) +WT (left bar) clk-1(qm30) (middle bar) and their ratio (right bar). Colors of the ratio bars correspond again to the data points in (A). Clk-1(qm30) worms slow reduced TE of let-363, daf-15, clk-2, and R10H10.7 and an increased TE for lgg-1. (D) Schematic overview of transcripts involved in TOR pathway that have altered TE in clk-1(qm30) versus clk-1(qm30) +WT represented in (C) and their involvement in longevity. a novel mechanism for inducing autophagy in systems undergoing lifespan extension by affecting the TE of mRNAs coding for crucial components of the mTOR-autophagy regulatory axis.
Finally, we found an unexpected role for TAF-4 as a repressor of polyribosome formation when UQ biosynthesis is compromised. Although TAF-4 is essential for all RNA transcriptions in the early embryo in C. elegans (Walker et al, 2001), this requirement is different in later developmental stages, and worms exposed to taf-4 RNAi starting from L1 throughout their remaining life grow normally (Khan et al, 2013). The authors propose a role for TAF-4 in the retrograde response model of longevity in mutants with dysfunctional mitochondria. In this model, dysfunctional mitochondria signal via cAMP response element-binding proteins to activate TAF-4 in a way that increases life-extending gene expression. Here, we show that one of the consequences of TAF-4 loss in clk-1(qm30) mutants, and perhaps also in isp-1 and tpk-1 mutants, is a failure of the system that shuts down cytosolic mRNA translation in response to mitochondria impairment. This unexpected ability of TAF-4 to control protein synthesis, by regulating the expression of eif-1, will be very interesting for further study.
This study demonstrates that molecular pathways of long-lived clk-1(qm30) mutants with dysfunctional mitochondria converge with those of other distinctly different long-lived mutants to reduce cytoplasmic mRNA translation. It also reveals a striking change in the Figure 6. RNAi of taf-4 partially restores repressed peaks in clk-1(qm30) mutants. (A) Representative traces of polysome profiles of clk-1(qm30) worms fed HT115 bacteria transformed with the empty vector (EV) or expressing taf-4 RNAi when harvested at the L4 stage, when lysate is normalized to total protein levels of 500 μg. The monosomal peak and polysomal peaks (P1-P5) are indicated. (B) Representative traces of polysome profiles of clk-1(qm30) +WT worms fed HT115 bacteria transformed with the EV or expressing taf-4 RNAi, when lysate is normalized to total protein levels of 500 μg. The monosomal peak and polysomal peaks are indicated. (C) Quantification of polysome peak sizes (AUC). The fold change is represented compared with P1 of the clk-1(qm30) +WT fed with HT115 bacteria. Polysomal peaks P1 and P2 were significantly different between clk-1(qm30) +WT fed with HT115 and taf-4 RNAi bacteria. No peaks were significantly different between clk-1(qm30) +WT fed with HT115 and taf-4 RNAi bacteria. Error bars represent mean ± SD. Significance was tested with t test and P-values were adjusted to correct for multiple testing using the Holm-Sidak method, with α = 0.05. (D) Relative expression levels determined by qRT-PCR in clk-1(qm30) (+WT) fed with either the EV or taf-4 RNAi bacteria. Expression levels were normalized using the geometrical mean of reference genes Y45F10D.4, tba-1, and csq-1. Significance was tested using one-way ANOVA with Sidak's multiple comparisons test. Error bars represent mean ± SEM. **P < 0.01, significance indicated only between clk-1(qm30) EV and clk-1(qm30) taf-4 RNAi.
transcripts that the polyribosomes prefer to translate when mitochondria are compromised. Ribosomal proteins and other protein parts of the translation machinery are very marginally translated, whereas the mitochondrial OXPHOS genes are still robustly translated despite a substantial reduction in the number of polysomes. We propose that this mechanism is likely recapitulated in many more models of lifespan extension and reaffirms studies that aim to extend human lifespan by reducing protein synthesis.

Polysome profiling
Gradients of 17-50% sucrose (11 ml) in gradient buffer (110 mM KAc, 20 mM MgAc2, and 10 mM Hepes, pH 7.6) were prepared on the day before use in thin-walled, 13.2-ml, polyallomer 14 89-mm centrifuge tubes (Beckman Coulter). Nematodes were lysed in 500 ml polysome lysis buffer (gradient buffer containing 100 mM KCl, 10 mM MgCl 2 , 0.1% NP-40, 2 mM DTT, and RNaseOUT [Thermo Fisher Scientific]) using a Dounce homogenizer. The samples were centrifuged at 3,500 g for 10 min to remove debris and the supernatant was subjected to bicinchoninic acid assay (BCA) protein assay. In all, 500 mg of total protein for each sample was loaded on top of the sucrose gradients. The gradients were ultracentrifuged for 2 h at 40,000 rpm in a SW41Ti rotor (Beckman-Coulter). The gradients were displaced into a UA6 absorbance reader (Teledyne ISCO) using a syringe pump (Brandel) containing 60% sucrose. Absorbance was recorded at an optical density of 254 nm. All steps of this assay were performed at 4°C or on ice and all chemicals came from Sigma-Aldrich unless stated otherwise. Polysome peaks were quantified by measuring the area under the curve (AUC) in ImageJ. Significance was tested with t test, and P-values were adjusted to correct for multiple testing using the Holm-Sidak method, with α = 0.05.

RNA sequencing
Total/polysomal RNA isolation For isolation of total RNA, worms were homogenized with a 5-mm steel bead using a TissueLyser II (Qiagen) for 5 min at a frequency of 30 times/s in the presence of TRIzol (Invitrogen), and the isolation was continued according to the manufacturer's protocol. Polysomal fractions from two experiments were pooled and mRNA was extracted using TRIzol LS (Invitrogen) according to the manufacturer's protocol. Contaminating genomic DNA was removed using RNase-Free DNase (Qiagen) and samples were cleaned up with the RNeasy MinElute Cleanup kit (Qiagen) and sequenced by GenomeScan BV at a 20 M read depth.

Library preparation
The samples were processed for Illumina using the NEBNext Ultra Directional RNA Library Prep kit (NEB: New England Biolabs [NEB] #E7420) according to the manufacturer's protocol. Briefly, rRNA was depleted from total RNA using the rRNA depletion kit (NEB #E6310). After fragmentation of the rRNA-reduced RNA, a cDNA synthesis was performed to ligate with the sequencing adapters and PCR amplification of the resulting product. Quality and yield after sample preparation was measured with the Fragment Analyzer. Size of the resulting products was consistent with the expected size distribution (a broad peak between 300 and 500 bp). Clustering and DNA sequencing using the Illumina cBot and HiSeq 4000 was performed according to the manufacturer's protocol with a concentration of 3.0 nM of DNA. HiSeq control software HCS v3.4.0, image analysis, base calling, and quality check were performed with the Illumina data analysis pipeline RTA v2.7.7 and Bcl2fastq v2.17. RNA-seq data have been deposited in the ArrayExpress database at European Molecular Biology Laboratory: European Bioinformatics Institute under accession number: E-MTAB-7166.

Bioinformatics analysis
Mapping of the reads The quality of the reads in the fastq files was confirmed using FastQC version 0.11.4 (http://www.bioinformatics.babraham.ac.uk/ projects/fastqc). FastQC performs multiple quality tests and provides the user with warnings about possible problems with the raw data. Paired end reads were mapped to the C. elegans genome using HISAT2 version 2.1.0 (Kim et al, 2015). The WBcel235 genome assembly was downloaded from WormBase and used as reference genome for the mapping. After successful mapping, counts per gene were extracted from the .SAM files with HTSeq version 0.9.1 (Anders et al, 2015). The -stranded=reverse setting was used, conforming with the NEB Ultra Directional RNA Library Prep kit procedure.

Statistical analysis
The statistical analysis was performed in R using DESeq2 version 1.16.1 (Love et al, 2014). DESeq2 models the raw counts with the negative binomial distribution and normalizes between samples by calculating size factors using the median ratio method (Anders & Huber, 2010). The statistical test is based on a generalized linear model (GLM) using the negative binomial model as an error distribution model. To more accurately model dispersion parameters, the trend of dispersion to abundance is taken into account using an empirical Bayes procedure. The significance of the GLM coefficients is determined using a Wald test. The P-values are adjusted to correct for multiple testing. In the differential translational regulation analysis, the fitted GLM took the following form: logðμ i Þ = β 0 + β 1 X sample + β 2 X type + β 3 X sample X type : (1) Here, the Wald test was performed on the interaction parameter 3 to infer significance. The log2 fold change calculated in this case corresponds to the following: log2 fold change = log 2 strain A polysomal strain A total strain B polysomal strain B total : (2) where in Equation (1), the sample refers to the strain and type to polysomal or total RNA. Experimental and statistical normalizations were performed for each condition individually, so counts for a specific gene in the polysomal dataset can be possibly higher than in the total RNA dataset, even though the first is technically a subset of the latter. Still, we can test if genes are highly translated in one strain when it is not in the other.

Fluorescence microscopy imaging
For imaging, synchronized L4 stage N2 and clk-1(qm30) worms carrying FIB-1::GFP were placed on top of a 2% agarose pad on slides and anaesthetized using 1 mM levamisole. Nucleoli in the pharynx were imaged using 100× magnification (PLAN APO 100×/1.40/OIL Phaco3) with a Leica DM6 microscope and Leica LAS-X software. Total and average nucleolar area per worm was quantified using ImageJ.
qPCR of total and polysomal RNA cDNA synthesis was performed with 1 μg of total RNA using the QuantiTect Reverse Transcription Kit (Qiagen). LightCycler 480 SYBR Green I Master (Roche) was used for qPCR analysis, primers are listed in Table 3. Data were analyzed with LightCycler 480 software release 1.5 and LinRegPCR version 2015.3. Relative gene expression was normalized to the geometric mean of reference genes Y45F10D.4, tba-1, and csq-1.