Genome-Wide Mechanisms of Lifespan Extension by Dietary Restriction in Yeast

Dietary restriction is arguably the most promising non-pharmacological intervention to extend human life and health span. Yet, only few genetic regulators mediating the cellular response to dietary restriction are known, and the question remains which other transcription factors and regulatory pathways are involved. To gain a comprehensive view of how lifespan extension under dietary restriction is elicited, we screened the chronological lifespan of most gene deletions of Saccharomyces cerevisiae under two dietary regimens, restricted and non-restricted. We identified 472 mutants with enhanced or diminished extension of lifespan by dietary restriction. Functional analysis of such dietary-restriction genes revealed novel processes underlying longevity specifically by dietary restriction. Importantly, this set of genes allowed us to generate a prioritized catalogue of transcription factors orchestrating the dietary-restriction response, which underscored the relevance of cell-cycle arrest control as a key mechanism of chronological longevity in yeast. We show that the transcription factor Ste12 is needed for full lifespan extension and cell-cycle arrest in response to nutrient limitation; linking the pheromone/invasive growth pathway with cell survivorship. Strikingly, STE12 overexpression was sufficient to extend chronological lifespan under non-restricted conditions. Our global picture of the genetic players of longevity by dietary restriction highlights intricate regulatory cross-talks in aging cells.


SUMMARY
We show that the transcription factor Ste12 is needed for full lifespan extension and cell-cycle arrest in response to nutrient limitation; linking the pheromone/invasive growth pathway with cell survivorship. Strikingly, STE12 overexpression was sufficient to extend chronological lifespan under non-restricted conditions. Our global picture of the genetic players of longevity by dietary restriction highlights intricate regulatory cross-talks in aging cells.

INTRODUCTION
Dietary restriction-a reduction in calorie intake without malnutrition, or substitution of the preferred carbon or nitrogen source-extends lifespan in virtually all species studied in the laboratory, from yeast to primates (Masoro 2005;Mair & Dillin 2008).
Dietary restriction has been associated with protection against age-associated disease in mice, including neurodegenerative disorders (Zhu et al. 1999) and cancer (Yamaza et al. 2010), promoting not only a longer lifespan but also healthier aging (Fontana & Partridge 2015). Importantly, this intervention reduces the mortality rate in non-human primates (Colman et al. 2014), and delays the onset of aging-related physiological changes in humans (Holloszy & Fontana 2007), making dietary restriction the most promising intervention targeted to extend human lifespan. Yet, we are still missing a global picture of the genetic architecture of such lifespan response, which is needed to grant a deeper understanding of the genotype-phenotype relationship of aging and longevity (Schleit et al. 2013).
The budding yeast Saccharomyces cerevisiae has been a pivotal model organism in the discovery of the molecular and cellular basis of aging. Two aging models are widely used in this organism: The replicative lifespan of yeast, which refers to the number of times a single yeast cell can divide, and the chronological lifespan (CLS), which is a measure of the viability of a population during stationary phase throughout time. The latter provides a good model for inquiring the molecular changes and damage faced by post-mitotic cells (Longo et al. 2012). 4 In yeast, there is evidence that dietary restriction results in lifespan extension at least through the modulation of the conserved TOR and Ras/cAMP/PKA pathways, which regulate cellular growth and maintenance in response to nutrient availability (Kaeberlein et al. 2005). Depletion of TOR components, Tor1 and Sch9, results in CLS extension (Powers et al. 2006). Under low nutrient conditions, the serine/threonine kinase Rim15 phosphorylates transcription factors Msn2, Msn4, or Gis1, activating a maintenance response (Fabrizio et al. 2004). However, the msn2msn4gis1 triple mutant still shows lifespan extension by DR (Wei et al. 2008). Moreover, transcriptomic evidence and database analysis suggest that a larger number of up-and downstream genes are involved in lifespan extension (Wuttke et al. 2012;Choi et al. 2017); most of these candidates lack direct phenotypical confirmation. These observations suggest that there is yet to be identified an unknown number of regulators mediating lifespan extension by dietary restriction.
Research on aging in yeast has recently taken advantage of genome-wide approaches, enabling a comprehensive description of genes involved in lifespan regulation. For instance, a recent systematic study of replicative lifespan of most viable deletion strains revealed biological processes mediating longevity, such as translation, the SAGA complex, and the TCA cycle (McCormick et al. 2015). In the CLS model, several studies have aimed to estimate in parallel the stationary-phase survival of single-deletion mutants (Powers et al. 2006;Matecic et al. 2010;Fabrizio et al. 2010;Gresham et al. 2011;Garay et al. 2014), showing that autophagy, vacuolar protein sorting, regulation of translation, purine metabolism, 5 chromatin remodeling, and the SWR1 complex are major determinants of stationary-phase survival. However, a direct comparison of the lifespan effects of such gene deletions under nutrient-rich and restricted conditions, that would allow to systematically address the mechanisms of longevity by dietary restriction is still missing.
The aim of this work was to generate a global picture of the underlying genetics of lifespan extension by dietary restriction, by systematically describing gene-diet interactions in yeast. Specifically, we compared the CLS of a collection of 3,718 knockout mutants aged under non-restricted and dietary-restricted media, using a high-resolution parallel phenotyping assay (Garay et al. 2014). These screens revealed 472 genes that influence the lifespan response to dietary restriction. Subsequent analyses uncovered the major biological processes and a comprehensive catalogue of transcription factors that control lifespan extension in a dietary-restriction regime. In particular, we revealed a link between Ste12 the pheromone/invasive growth transcription factor and lifespan regulation, suggesting that Ste12-mediated cell-cycle control is a mechanism of chronological longevity in response to nutrient limitation.

Systematic identification of dietary-restriction genes in yeast
We set out to describe, at the genome-wide level, the way in which dietary restriction influences the lifespan effect of gene mutations. In yeast, extended 6 lifespan is achieved either by limiting the concentration of glucose in the growth medium (Koubova & Guarente 2003) or by using a non-preferred source of nitrogen (Powers et al. 2006;Jiang et al. 2000;Anderson et al. 2003). Here, dietary restriction was achieved by using GABA as the sole source of nitrogen, while nonrestricted medium contained the preferred nitrogen-source glutamine (see Experimental Procedures). Dietary restriction increased half-life from 20.9 ± 0.4 days to 33.7 ± 1.3 days in the WT strain, which represents a 61% extension of the CLS ( Figure 1A).
To identify the genetic determinants of lifespan extension by dietary restriction at the genome-wide level, we measured the CLS of 3,718 gene-knockout strains. We used a high-resolution profiling assay based on the measurement of a relative survival coefficient (s) of each knockout strain aged in co-culture with the WT strain (Garay et al. 2014) under dietary-restricted or non-restricted media ( Figure 1B).
The quantitative nature of our experimental data allowed us to determine whether the gene deletions had a neutral, deleterious, or beneficial effect on the CLS under each condition ( Figure S1; Table S1). We scored 573 significantly short-lived and 254 long-lived single knockout strains in the non-restricted medium (FDR<5%), while dietary restriction resulted in 510 short-lived and 228 long-lived strains.
To validate our large-scale screen, we phenotyped a set of randomly-selected group of knockout strains that showed significant CLS effects. We used a previously-reported assay for quantitative analysis of yeast CLS (Murakami et al. 2008) to obtain viability curves for single-strain cultures as a function of time under the two nutrient conditions (Figure 2A-B). Twelve out of 16 (75%) strains re-tested 7 under non-restricted medium recapitulated the CLS effects observed in the genome-wide screen, while 11 out of 17 (65%) strains tested were consistent with the results under dietary restriction (Figure S2;p<0.05,. We note that this rate of false-positive hits is much lower than in other high-throughput assays that have used pooled deletion strains (Fabrizio et al. 2010;Matecic et al. 2010).
Hence, our profiling approach provides accurate CLS scores for high-throughput profiling under different conditions, which allows scoring the gene-diet interactions influencing lifespan phenotypes.
To gain insight into the genes that mediate the lifespan-extending effects of dietary restriction, we searched for deletion strains that showed differential relative CLS effects when comparing the two diets ( Figure 1C). Specifically, we looked for strains in which the mutant's lifespan relative to the WT was shorter in dietary restriction than in non-restricted medium (diminished lifespan extension). Likewise, we scored those cases in which the relative lifespan was relatively longer under dietary restriction (enhanced lifespan extension). In this way, by comparing the relative lifespans of all strains under non-restricted and dietary-restriction conditions, we were able to identify the set of gene knockouts leading to diminished or enhanced lifespan extension in yeast ( Figure 1D).
We quantitatively defined the relative lifespan extension of each knockout as ln 1 (see Experimental Procedures). The LE of each deletion strain was compared to the distribution of LE in 264 independent WT replicates to obtain a Z-score ( Figure 1E). After filtering out strains that did not show a significant CLS 8 effect in either condition, we obtained a list of 472 gene-knockouts with altered dietary-restriction response (FDR<5% ; Table S2). This comprehensive set, which we termed DR-genes, includes 219 knockouts with diminished longevity (LE<1) and 253 gene-knockouts that displayed enhanced lifespan extension (LE>1).
Together, these results show that many gene-environment interactions underlie longevity by dietary restriction in yeast.

Functional classification of dietary-restriction genes
To describe which downstream cellular functions influence lifespan extension by dietary restriction, we sought to classify the 472 DR-genes according to their annotated functional features. We used a kappa statistic approach (Huang et al. 2009) to cluster genes by shared GO terms and mutant phenotypes, as reported in the Saccharomyces Genome Database (see Experimental Procedures). The analysis was performed separately for genes with diminished (LE<1) or enhanced (LE>1) lifespan extension (Figure 3; Table S3). Some clusters recapitulated cellular functions previously related to lifespan regulation, such as autophagy (Meléndez et al. 2003), mitochondrial function (Ocampo et al. 2012;Aerts et al. 2009), and cytosolic translation (Pan et al. 2007;Hansen et al. 2007). Importantly, this classification also identified novel processes such as maintenance of cell-cycle arrest mediated by pheromone and establishment of nucleus localization in the cell. These observations show that our screen was able to independently identify processes that were previously known to influence the dietary response and at the 9 same time allowed the identification of novel genes and processes related to longevity by dietary restriction.
Deletion of genes necessary for the maintenance of cell-cycle arrest resulted in a diminished lifespan extension. Specifically, deletion of pheromone-responsive genes FAR7 and FAR8 had short-lived phenotype under dietary restriction ( Figure   3A). In yeast, these Far proteins prevent cell cycle recovery after pheromone exposure, possibly by inhibiting CLN1-3 (Kemp & Sprague 2003). Moreover, mutations in genes required for processing and correct localization of ribonucleoprotein complexes resulted in a strongly diminished lifespan extension.
While it is well established that ribosomal function is downregulated in response to TOR inhibition or nutrient depletion (Hansen et al. 2007), our screen pointed to specific proteins involved in pre-rRNA processing (Slx9), nucleolar rRNA methyltransferase (Rrp8), nuclear export of pre-ribosomal subunits (Arx1), and translational initiation (Bud27). Likewise, deletion of DYN1-3, JNM1, NUM1, and PAC1, all involved in nuclear movement along microtubules, resulted in diminished lifespan extension, suggesting that microtubule dynamics underlies dietary restriction. In this regard, it is known that certain cellular processes needed for extended longevity, such as autophagy, require intact function of microtubules (Köchl et al. 2006). However, the relationship between nuclear localization and lifespan extension remains unexplored.
We also found clusters with enhanced lifespan extension, such as mitochondrial function ( Figure 3B). While dietary restriction shifts the metabolism towards respiration (Lin et al. 2002), impaired respiration promotes longevity in yeast and nematodes through enhanced retrograde response and activation of anaplerotic pathways (Cristina et al. 2009). Hence, a higher demand for respiration during dietary restriction could lead to the activation of compensatory pathways; similar feedback mechanisms might account for the alleviation of deleterious effects observed in other deletion strains. For example, the short-lived phenotypes of deletion of cell-wall genes CWP1, DSE2, and TIR3 were largely alleviated under dietary restriction. Taken together, these findings suggest that lifespan extension in response to dietary restriction in yeast is a complex phenotype resulting from the interplay of many downstream cellular processes.

A defined set of transcription factors regulate lifespan extension by dietary restriction
Complex phenotypic responses are frequently coordinated by transcriptional regulation of functionally-related genes. To investigate the transcriptional regulation of longevity by dietary restriction, we analyzed our set of DR-genes using an algorithm to search for the transcriptional regulators of these genes. We used TFrank (Gonçalves et al. 2011), a graph-based approach that takes advantage of available interactions of transcription factors and their targets, to obtain a list of prioritized regulatory players within the yeast regulatory network ( Table 1). This approach allowed us to assess the possible lifespan role of transcription factors that were not included in our genome-wide screen due to gene essentiality or sterility. Transcription factors within the top 5% rank of this analysis included Msn2 and Msn4; these transcription factors are well known regulators of lifespan (Fabrizio et al. 2004). However, most top-ranked transcription factors had not been 11 previously associated to lifespan extension by dietary restriction in yeast.
To directly confirm the role in lifespan regulation of TFRank hits, we generated de novo deletion mutants for seven transcription factors and characterized their CLS

Ste12 is a positive regulator of longevity by dietary restriction and cell-cycle arrest in response to nutrients
Among the top hits of our transcription-factor analysis was Ste12, which acts on downstream genes involved in mating or pseudohyphal growth (Dolan et al. 1989;Roberts & Fink 1994). To further explore the regulatory role of Ste12 in lifespan extension, we characterized the CLS of the ste12 strain under different concentrations of glucose. Lifespan extension was partially abrogated in the ste12 strain under glucose restriction and, to a lesser extent, under high concentration of glucose ( Figure 5A). We also characterized the CLS of this strain using standard aerated culture conditions (Hu et al. 2013), which unambiguously confirmed that Ste12 is a positive regulator of lifespan extension in yeast ( Figure   S3).
Most methods for measuring CLS in yeast-including ours-rely in the ability of stationary-phase cells to re-enter the cell cycle upon transfer to fresh medium.
Given that Ste12 is involved in cell-cycle arrest, we decided to rule out any possible methodological artifact by directly measuring the fraction of dead relative to alive strains in stationary phase. This experiment showed that, under limited glucose concentration, the alive ste12 population decayed faster than the WT under calorie restriction ( Figure 5B; Figure S3). These results indicated that the CLS-effects of the STE12 deletion are maintained regardless of the methodology and nutrient restriction condition used to infer population survivorship in stationary phase.
A bona fide positive regulator of lifespan is expected to increase cell survivorship when over-expressed. We thus investigated whether high STE12 expression causes lifespan extension under non-restricted conditions. To this end, we generated a copper-inducible strain with a GFP fusion to track Ste12 protein levels.
The WT and pCUP1-STE12 strains were aged under varying concentrations of copper sulfate in 2% glucose SC non-restricted medium. We found that the CLS of the STE12-overexpression strain increased readily as a function of copper concentration (Figure 5C), while in the WT strain increasing copper concentrations had no significant effect on CLS nor growth (data not shown). Importantly, GFP signal in the nucleus also increased as a function of copper concentration, 13 confirming that lifespan extension was linked to increased levels of Ste12 in the nucleus ( Figure 4D). Together, these results indicate that Ste12 is a novel positive transcriptional regulator of lifespan in yeast.
Transcriptional targets of Ste12 are known to mediate G1 arrest of the cell-cycle in response to pheromone signal (Dolan & Fields 1990), but the role of such Ste12mediated cell-cycle arrest in aging is unknown. We asked whether Ste12-mediated cell-cycle control in response to nutrient limitation is a cellular mechanism of lifespan extension by dietary restriction in yeast. To this end, we used flow cytometry to monitor the cell-cycle dynamics of yeast populations following nitrogen starvation ( Figure 4E). Our results show that ste12 cells failed to arrest the cell cycle at the same rate as the WT strain. This observation suggests that the transcription factor Ste12 integrates nutrient signaling and regulates downstream genes needed for cell-cycle arrest, which may underlie its beneficial effect on cell survivorship.

Ste12 mediates lifespan extension in association to Tec1
To gain further insight on the cellular pathways of lifespan extension by dietary restriction, we asked whether the lifespan effect of STE12 was linked to Tec1, a transcription factor of the invasive-growth pathway. Ste12 and Tec1 form a heterodimer that activates filamentous-growth genes (Madhani et al. 1999).
Moreover, Tec1 has also been associated to lifespan regulation in response to rapamycin (Brückner et al. 2011). We generated a tec1ste12 double-deletion mutant and found an epistatic interaction between these genes under glucose 14 restriction; the CLS phenotypes of the ste12tec1double deletion was not the additive combination of the single-deletion phenotypes ( Figure 6A). This genetic interaction suggests that Tec1 acts in concert with Ste12 to promote longevity by dietary restriction.
Finally, we asked whether the effect of Ste12 is associated to elements of the pheromone-responsive MAPK pathway. Deletion of the pheromone-receptor genes STE2 and STE3 did not show a lifespan-extension defect in response to glucose restriction ( Figure 6B). Altogether, these results indicated that Ste12 regulates lifespan extension in response to nutrients through functional association to Tec1 and the invasive-growth pathway, but not under the control of upstream pheromone-pathway components.

DISCUSSION
"The intrinsic nature of the ageing process is essentially one of systems degradation" (Kirkwood 2008). With a growing number of genetic aging factors in hand, the next great challenge is to understand how the different mechanisms underlying aging and longevity are integrated to one another and to the environment. In this study, we have adapted the chronological-aging paradigm in We have provided compelling evidence that STE12 is a positive regulator of longevity by dietary restriction. Not only did the ste12 strain show diminished extension by dietary restriction measured both by outgrowth and dead/alive assays, but also STE12 overexpression was sufficient to extend lifespan under a non-restricted diet. Importantly, the STE12 phenotypes in our screening settings were confirmed under standard chronological-aging conditions. Ste12 is a transcription factor downstream of two cell-differentiation programs regulated by MAPK pathways, namely mating and invasive growth (Dolan et al. 1989;Madhani et al. 1999). We also found that deletion of STE12 results in a failure to arrest the cell cycle upon nutrient starvation. In the presence of pheromone, the cell cycle is We have shown that deletions of STE12 and TEC1 interact in an epistatic manner, which suggests that these transcription factors act in concert to promote yeast lifespan. Ste12 is associated to Tec1 during pseudohyphal growth (Chou et al. 2006). In addition to controlling cellular development in response to stimuli, Tec1 is needed for full lifespan extension in response to the Tor1-inhibiting drug rapamycin (Brückner et al. 2011), suggesting that Tec1 acts downstream of the TOR pathway. Also, transcription analyses have shown that Ste12 is a regulatory hub during stationary phase under rapamycin treatment (Wanichthanarak et al. 2015), further strengthening the idea that Ste12 and Tec1 link TOR and MAPK-signaling pathways. However, Tec1 promotes cell-cycle progression by activation of Cln1 (Madhani et al. 1999), in conflict with the fact that the cell-cycle is arrested in response to nutrient limitation. It is thus likely that the players upstream Ste12 and Tec1 are involved in an intricate signaling response leading to lifespan extension.
It remains to be addressed whether the role on aging of the set of transcription factors identified in this study is conserved. For instance, STE12 has no clear homolog in animals, however, transcriptional networks can be rewired through evolution, leading to changes in the regulation exerted by specific regulators, while the downstream targets remain associated (Sorrells et al. 2015). In addition, the yeast three-kinase module regulating MAPK pheromone and invasive-growth pathways are conserved in other organisms (Widmann et al. 1999). In particular, KSS1 and FUS3 are key members of the MAPK pathway that regulates cell differentiation programs in yeast (Bardwell 2004), while their mammalian counterpart MAPK1 has been reported to regulate cell-fate determination (Chaman et al. 2015). Also, the MAPK1/ERK pathway is central to the development of several age-associated diseases in mammals (Carlson et al. 2008). Thus, the study of targets downstream the MAPK pathway in yeast might bring important insights into the regulation of aging in other eukaryotes, including humans.

18
Our genome-wide screens provide a much-needed comprehensive picture of the mechanisms of lifespan extension by dietary restriction, underscoring links between nutrient sensing, the cell-cycle arrest machinery, and longevity in yeast.
Our approach can be readily applied to other genetic, environmental, or pharmacological perturbations, providing a systematic framework to describe aging networks in a simple tractable system. Other cross-talks among downstream cellular processes and their transcriptional regulators may remain to be uncovered, which will shed further light to the genetic wiring of aging cells. conditions. SC medium used for DR based on glucose concentration was 0.17% yeast nitrogen base (YNB) without amino acids, 0.5% or 2% glucose, and 0.2% amino acid supplement mix.

EXPERIMENTAL PROCEDURES
All outgrowth cultures were performed in low-fluorescence medium (YNB-lf).
Nitrogen-starvation medium for cell-cycle progression experiments was 2% glucose and 0.17% YNB without amino acids and ammonium sulfate.  (Garay et al. 2014). Data acquisition and initial processing have been described. In brief, five days after inoculation, 5 µl outgrowth cultures were inoculated every other day into 150 µl of fresh low-fluorescence medium; absorbance at 600nm (OD 600 ) and fluorescence (RFP and CFP) measurements were taken every 150 min throughout 14 hrs. An apparent survival coefficient, s, and its standard error, σ s , were obtained from the slope of the linear regression (Robustfit, Matlab) of the log of the ratio of RFP to CFP signal at a fixed interpolation time point in the outgrowth culture (10 hrs), and the number of days in stationary phase, as previously described (Garay et al. 2014).

Scoring CLS phenotypes and lifespan extension coefficients. Short-and long-
lived knockouts under NR or DR were determined by assigning a Z-score to each mutant's s coefficient; the distribution's mean and standard deviation were from the measurement of 264 WT RFP /WT CFP independent co-cultures under either condition.
Two-tailed p-values were obtained from each Z-score to compute a false-discovery rate (FDR); we assigned significant phenotypes using a q<0.05 cutoff. Viability data points relative to T 0 were used to plot a survival curve, which was fitted to an exponential decay model ( ) where N 0 is the percentage of viability at T 0 , T is time in days, and r is the rate of death. For validation of CLS effects, mutants were taken from the RFP-tagged deletion collection (or generated 22 de novo, when indicated) and viability was assayed to calculate death rates in at least 7 experimental replicates, which were compared to replicates of a WT strain; significant CLS effects were considered using a p<0.05 cutoff (T-test).
CLS assay in standard aeration conditions. Pre-inoculums from three different colonies of each strain were set in 5 mL SC medium for 24 hours, these were 23 Kappa values were used to build a matrix that represented the agreement between each gene-pair. Gene-pairs that showed kappa>0.35 were regarded as likely similar and thus used as cluster seeds to form larger groups of genes; groups sharing more than 50% of their genes were merged in subsequent iterative steps. Clusters with at least four elements were manually named by inspection in the SGD and GO enrichment. Network representation was created using Cytoscape; edges between nodes represent kappa agreement above the established threshold (kappa>0.35).
Alive/dead staining assay. We used the same scheme of 96 deep-well plates             Relative death rate (NR)