Light-sensing via hydrogen peroxide and a peroxiredoxin

Yeast lacks dedicated photoreceptors; however, blue light still causes pronounced oscillations of the transcription factor Msn2 into and out of the nucleus. Here we show that this poorly understood phenomenon is initiated by a peroxisomal oxidase, which converts light into a hydrogen peroxide (H2O2) signal that is sensed by the peroxiredoxin Tsa1 and transduced to thioredoxin, to counteract PKA-dependent Msn2 phosphorylation. Upon H2O2, the nuclear retention of PKA catalytic subunits, which contributes to delayed Msn2 nuclear concentration, is antagonized in a Tsa1-dependent manner. Conversely, peroxiredoxin hyperoxidation interrupts the H2O2 signal and drives Msn2 oscillations by superimposing on PKA feedback regulation. Our data identify a mechanism by which light could be sensed in all cells lacking dedicated photoreceptors. In particular, the use of H2O2 as a second messenger in signalling is common to Msn2 oscillations and to light-induced entrainment of circadian rhythms and suggests conserved roles for peroxiredoxins in endogenous rhythms.

S unlight is a prerequisite for life by providing a source of heat and energy for primary production. Most organisms exposed to light therefore maintain an ability to adapt their activities to its presence. For example, many organisms, from cyanobacteria to humans 1 , maintain autonomous cycles of activity that correspond to day and night even in complete darkness 2 . The ability to respond to light is a fundamental feature of such circadian clocks 3 . In the yeast Saccharomyces cerevisiae, light alters ultradian metabolic rhythms 4 , which have been proposed to constitute models for circadian clocks 5 . In addition, the Zn-finger transcription factor Msn2 rhythmically shuttles into and out of the nucleus in response to illumination 6 , in spite of the lack of specialized photosensory proteins such as opsins, phytochromes and cryptochromes in this organism 7 . Nuclear localization of Msn2 is inhibited by cyclic AMP-controlled protein kinase A (PKA) 8,9 and stimulated by the phosphatases PP1 and PP2A 10,11 . The Msn2 rhythm was proposed to originate from oscillations in cAMP produced at intermediate light intensities 12 , based on the observation that Msn2 fails to oscillate in mutants deficient in PKA feedback regulation 13 and on the recapitulation of oscillations in a mathematical model of the PKA signalling pathway. A careful analysis of the localization behaviour in single cells indicated that Msn2 localization typically passes through three distinct and successive states in response to illumination: cytoplasmic, nucleo-cytoplasmic oscillatory and more permanently nuclear 14 . Such gradually increased nuclear localization of Msn2 presumably reflects a gradual decrease in PKA-dependent phosphorylation and altered levels of PKA/phosphatase signalling intermediate(s). However, the nature of any such intermediates and the mechanism by which PKA senses light remains unknown.
Typical 2-Cys peroxiredoxins are enzymes that reduce hydrogen peroxide (H 2 O 2 ) via two catalytic cysteines and thioredoxin acting as the hydrogen donor 15,16 . During catalysis, a small proportion of the primary catalytic (peroxidatic) cysteine becomes hyperoxidized to the sulfinylated (Cys-SO 2 H) form, which inactivates the enzyme 17,18 . Peroxiredoxins have received increased attention because of their functions in suppressing malignant tumours 19 , in keeping the genome free of mutations 20 and as conserved executors of the ability of caloric restriction to slow down the rate of ageing 15,16,[21][22][23][24] . In addition, peroxiredoxin hyperoxidation was shown to sustain transcription-independent circadian rhythms in organisms from the three kingdoms of life 1,[25][26][27] . On a similar note, oscillations in peroxiredoxin hyperoxidation were shown to resonate with metabolic tidal rhythms in the marine crustacean Eurydice pulchra and peroxiredoxins have also recently been suggested to modulate yeast ultradian metabolic rhythms 5,28 , indicating conserved functions in maintaining endogenous molecular clocks. The observation that light stimulates H 2 O 2 production in cultured mouse, monkey and humans cells via photoreduction of flavin-containing oxidases such as peroxisomal acyl-coenzyme A (CoA) oxidase 29 and the messenger function of H 2 O 2 in zebrafish, which couples the light signal to the circadian clock 30 , raise the interesting possibility that H 2 O 2 might be a conserved second messenger by which light affects cellular physiology in general and endogenous clocks in particular.
Here we show that in S. cerevisiae a conserved peroxisomal oxidase converts the light impulse into a H 2 O 2 signal that is sensed by the peroxiredoxin Tsa1 and then transduced to thioredoxin to inhibit PKA activity. We propose that peroxiredoxin-mediated H 2 O 2 signalling establishes rhythmic Msn2 nuclear accumulation by superimposing on PKA feedback regulation. In particular, we report that the Tsa1-mediated signal counteracts the nuclear retention of the two most highly expressed of the PKA catalytic subunits, thereby antagonizing a process that contributes to delayed Msn2 nuclear localization in response to H 2 O 2 .

Results
Light is signalled to Msn2 through H 2 O 2 . The transcription factor Msn2 rhythmically shuttles into and out of the nucleus in response to light 6 . To study this phenomenon, we used an experimental setup enabling both single-cell (Fig. 1a-c and Supplementary Movie 1) and population analyses (Fig. 1d) of Msn2 localization dynamics 14 . In short, cells were exposed to blue visible light (l ¼ 450-490 nm) and imaged in a perfusion chamber using epifluorescence microscopy as previously described 14 . At a light intensity corresponding to that on a sunny day (P ¼ 115 mW), Msn2 accumulated into the nucleus and then back to the cytoplasm on average 4.3±3.6 times per hour (average ± s.d., n ¼ 248) and, following an initial lag of B20 min, the proportion of cells exhibiting Msn2 nuclear localization 14 stabilized at around 25% (Fig. 1d). Higher light intensities (P ¼ 200 mW) lead to a higher proportion of cells exhibiting nuclear localization (Supplementary Fig. 1a) and eventually resulted in more sustained nuclear localization 14,31 . In the absence of light, Msn2 did not accumulate into the nucleus 6 .
We hypothesized that H 2 O 2 might serve as the light-induced second messenger, because light stimulates H 2 O 2 production in mammalian cells 29 and because yeast cells lacking the oxidantresponsive transcription factor Yap1 grow poorly in light 4 . To substantiate this, we used the cytosolic genetically encoded H 2 O 2 sensor HyPerRed 32 , which we reasoned would be less sensitive to bleaching upon illumination with blue light than green fluorescent protein (GFP)-based H 2 O 2 probes (for example, HyPer 33 and roGFP2-Tsa2DCR 34 ). Upon the start of illumination, the HyPerRed signal initially stayed constant but already after a couple of minutes decreased steadily at a rate dependent on light intensity ( Supplementary Fig. 1b, Fig. 1b,c), indicating that upon illumination the levels of cytosolic H 2 O 2 increased (Fig. 1e). The Cys199-dependent HyPerRed fluorescence increased as a function of light intensity (Fig. 1e), suggesting that so do the levels of cytosolic H 2 O 2 . In support of H 2 O 2 causing Msn2 nuclear redistribution upon illumination, an increased number of cells lacking the mitochondrial cytochrome c peroxidase Ccp1 exhibited nuclear Msn2 ( Fig. 1d and Supplementary Fig. 1d Fig. 1h-j). Conversely, in cells overproducing Ccp1 (B3-fold, Supplementary Fig. 1k,l), which scavenge H 2 O 2 at an increased rate, light-induced Msn2 nuclear localization was abolished (Fig. 1g), suggesting that cytosolic H 2 O 2 is both necessary and sufficient for the light response of Msn2.
A peroxisomal oxidase converts light into a H 2 O 2 signal. Flavin prosthetic groups play unique roles in redox reactions by their ability to participate in both one-and two-electron transfer processes 35 . In addition, they act as chromophores in dedicated light receptors of the 'light-oxygen-voltage', the 'blue-light sensing using Flavin' and the cryptochrome families 36 . In these proteins, photoexcitation of the flavin group initiates signal transduction through intramolecular conformational changes and/or altered protein-protein interactions 36 . However, upon excitation, flavins also become sensitive to reduction by cellular reducing agents and can subsequently transfer electrons to molecular oxygen, leading to the production of H 2 O 2 (refs 35,37,38). Such reactions of flavin in peroxisomal fattyacyl CoA oxidase were proposed to be the cause of blue light phototoxicity in mammalian cells through H 2 O 2 production 29 . The yeast homologue of peroxisomal acyl-CoA oxidase is Pox1. Strikingly, Pox1 deficiency substantially decreased both cytosolic H 2 O 2 (Fig. 1h) and Msn2 nuclear localization upon illumination ( Fig. 1d and Supplementary Fig. 1m,n). However, both cytosolic H 2 O 2 and Msn2 nuclear localization increased normally upon the addition of exogenous H 2 O 2 to cells lacking Pox1 ( Supplementary   Fig. 1o-r), indicating that peroxisomal acyl-CoA oxidase affects the response to light via H 2 O 2 . These data suggest that a peroxisomal acyl-CoA oxidase can signal light through H 2 O 2 production in an organism lacking dedicated light receptors.
A peroxiredoxin acts as a receptor of light-induced H 2 O 2 . The peroxiredoxin Tsa1 (see Fig. 2a) is required for the nuclear accumulation of Msn2 in response to H 2 O 2 , but not to other stimuli, for example, NaCl that also trigger this response 39 . Based on our finding that light uses H 2 O 2 as a messenger, we therefore asked whether light-induced Msn2 nuclear localization similarly depended on Tsa1. Indeed, we found that Msn2 remained cytoplasmic upon illuminating Tsa1-deficient cells, as evidenced by a representative single-cell trace (Fig. 2b) and by the strongly decreased fraction of cells displaying Msn2 nuclear localization (Fig. 2c). This was true also at higher light intensities (P ¼ 200 mW, Supplementary Fig. 2a), further underscoring the crucial role of H 2 O 2 in yeast light sensing. Moreover, we found that Tsa1 catalytic cycling was required for signal transduction, as Msn2 remained cytosolic in cells carrying serine substitutions of the Tsa1 catalytic cysteines (Fig. 2a,c,d and Supplementary  Fig. 2b-d). It is known that a small fraction of Tsa1 becomes hyperoxidized to the sulfinic acid form upon each turn of the peroxiredoxin catalytic cycle and such hyperoxidation inactivates enzyme peroxidatic cycling 17,18 (see Fig. 2a). To further test the importance of enzyme cycling, we therefore analysed Msn2 localization in cells expressing tsa1DYF. This mutant lacks the carboxy-terminal YF motif and is therefore resistant to hyperoxidation 40,41 (Supplementary Fig. 2k). The fraction of tsa1DYF cells displaying Msn2 nuclear accumulation increased markedly ( Fig. 2d and Supplementary Fig. 2e). However, this effect was completely negated by serine substitution of C171, which decreased Msn2 nuclear accumulation back to the low levels observed in the tsa1C171S single mutant ( Fig. 2d and Supplementary Fig. 2f). These observations thus further point to the importance of Tsa1 peroxidatic cycling for light signalling to Msn2 and the inhibitory effect of enzyme hyperoxidation on this response. Tsa1 might relay the H 2 O 2 signal produced by light, by oxidizing an Msn2 regulatory target either directly or through the intermediacy of its dedicated reductase, the redundant cytosolic thioredoxins Trx1 and Trx2. As the latter are required for H 2 O 2 signalling to Msn2 (ref. 39), we tested whether they were required also in light signalling. We found that simultaneous inactivation of Trx1 and Trx2 (trx1Dtrx2D) impaired light signalling to Msn2 ( Fig. 2g and Supplementary Fig. 2b,g). Upon reducing peroxiredoxins, thioredoxins become oxidized and are then reduced by thioredoxin reductase (see Fig. 2a). Trx2 was indeed immediately oxidized in a Tsa1-dependent manner upon H 2 O 2 addition (Fig. 2e,f and Supplementary Fig. 2l,m). The other abundant peroxiredoxin Ahp1 is also recycled by the cytosolic thioredoxins but Ahp1 preferentially reduces the model organic peroxide t-butyl hydroperoxide and therefore caused Trx2 oxidation in response to this particular oxidant but not to H 2 O 2 (Fig. 2f). In contrast, Tsa1 preferentially reduces H 2 O 2 and caused Trx2 oxidation in response to H 2 O 2 but not to t-butylhydroperoxide (Fig. 2f) 2h and Supplementary Fig. 2n,o). The TRR1 mutation strongly increased Msn2 nuclear localization in response to light ( Fig. 2g and Supplementary Fig. 2h) and H 2 O 2 (ref. 39). However, this increased response was completely lost upon removing both cytosolic thioredoxins (trr1Dtrx1Dtrx2D, Fig. 2g and Supplementary Fig. 2i). In this mutant, Msn2 nuclear localization was brought back to the low levels observed in the trx1Dtrx2D mutant, establishing that the thioredoxins signal light in their oxidized form. High levels of oxidized Trx2 were maintained in cells lacking both Trr1 and Tsa1 (trr1Dtsa1D, Fig. 2h and Supplementary Fig. 2o), enabling us to investigate whether the lack of response in Tsa1-deficient cells was caused by deficient Trx2 oxidation. Indeed, in trr1Dtsa1D cells, light triggered a vigorous Msn2 response ( Fig. 2i and Supplementary Fig. 2j), suggesting that Tsa1 becomes dispensable for light signalling upon the accumulation of thioredoxins in their oxidized form.

Tsa1-S-S-Tsa1
Tsa1-C48-SOH Tsa1-C48-SOOH Tsa1-C48-SH   Taken together, these results suggest that Tsa1 functions as a specific receptor for H 2 O 2 , but not organic peroxides, and transduces a signal to the cytosolic thioredoxins. This oxidation cue is both necessary and sufficient for Msn2 nuclear localization in response to illumination.

Light inhibits Msn2 phosphorylation by PKA via a H 2 O 2 relay.
Light-induced Msn2 nuclear accumulation was prevented in cells with constitutive PKA activity (bcy1D or pde2D, Fig. 3a), as shown previously 12,14 , and, importantly, the same was true for H 2 O 2 at levels that recapitulated the light response (0.4 mM, Supplementary Fig. 3a,b). However, the Msn2 response to high levels of H 2 O 2 (0.8 mM) was not inhibited by increased PKA activity ( Supplementary Fig. 3b). On the contrary and in agreement with previously published data 39 , the response to high levels of H 2 O 2 was stimulated in cells expressing constitutive PKA activity, indicating that (an)other pathway(s) operates under this condition. At these levels of H 2 O 2, Tsa1 was still required for the Msn2 response 39 (Supplementary Fig. 3c), whereas upon further increased H 2 O 2 (1 mM), it was not ( Supplementary  Fig. 3c), suggesting that the requirement for Tsa1 could be overcome upon increased H 2 O 2 .
PKA excludes Msn2 from the nucleus by both inhibiting its nuclear import and by stimulating its nuclear export through phosphorylation of serine residues in the nuclear localization (NLS) and nuclear export (NES) sequences 9,43,44 . We found that a fusion construct containing a heterologous NES and the Msn2 NLS (amino acids 576-704), which becomes dephosphorylated and concentrates into the nucleus upon glucose starvation 43 , neither responded to light nor H 2 O 2 (0.4 mM, Supplementary  Fig. 3d), suggesting that these cues regulate Msn2 at the level of nuclear export. In agreement with this, alanine substitution of one or both of the PKA-cognate NES residues S288 and S304 enhanced light-induced nuclear localization of Msn2 in wild-type cells and was able to restore nuclear localization in cells with constitutive PKA activity (bcy1D or pde2D, Fig. 3a-c and Supplementary Fig. 3e). These NES mutations also restored the Msn2 light (Fig. 3a-c and Supplementary Fig. 3f,g) and H 2 O 2 responses ( Supplementary Fig. 3h,i) in cells lacking Tsa1, suggesting that light inhibits Msn2 nuclear export by counteracting the phosphorylation of these serine residues. The Tsa1-dependent regulatory function on these two serine residues in response to light and H 2 O 2 was further demonstrated by their rapid, dose-dependent and transient dephosphorylation upon exposure to H 2 O 2 in wild-type cells, but not in cells lacking TSA1, as evidenced by both quantitative phosphoproteomic mass spectrometry (MS) analyses of SILAC (stable isotope labelling with amino acids in cell culture)-labelled purified Msn2 treatment, as the msn2S288AS304A mutant still responded to illumination (Fig. 3c), whereas a mutant in which all six relevant NES and NLS serine residues 9,43 were substituted for alanine residues displayed permanent, light-independent nuclear accumulation ( Supplementary Fig. 3o). In support of the role of Tsa1-dependent light signalling in counteracting PKA-dependent inhibition of Msn2, overexpression of the cAMP phosphodiesterase PDE2, which decreases PKA activity 45 , restored the defective light response of the tsa1D mutant ( Fig. 4a and Supplementary Fig. 4a-d), but did not restore the oxidation of thioredoxins ( Supplementary Fig. 4g,h). In addition, thioredoxin oxidation was unaffected upon the addition of H 2 O 2 in Tsa1-proficient cells overexpressing PDE2, further supporting Msn2 signal control downstream of the thioredoxins upon Pde2 overproduction (Supplementary Fig. 4e,f). Conversely, the increased light response in cells carrying the tsa1DYF mutant could be mitigated by overexpressing a PKA catalytic subunit Tpk1 (Fig. 4b and Supplementary Fig. 4i-m). Taken together, our data establish that the light-induced H 2 O 2 signal transduced by Tsa1 primarily relieves Msn2 from inhibitory phosphorylation of serine residues in the nuclear export sequence, allowing Msn2 to accumulate in the nucleus.
Intermediate PKA activity recapitulates Msn2 oscillations. The use of a cell-permeable substrate analogue inhibitor (1-NM-PP1) in cells carrying substrate analogue-sensitive PKA catalytic subunits (tpk1M164Gtpk2M147Gtpk3M165G 46 ) provided a tool to further dissect the importance of PKA inhibition in light-induced Msn2 nuclear localization. Partial inhibition of PKA recapitulated intermediate and oscillatory Msn2 nuclear localization (Fig. 4c,d and Supplementary Fig. 4n), whereas a more drastic inhibition led to instant and permanent Msn2 nuclear localization (Fig. 4c and Supplementary Fig. 4o,p). The Msn2 response to partial PKA inhibition was, however, significantly faster than the response to both illumination and H 2 O 2 (compare the response times to a level of inhibitor that recapitulated intermediate and oscillatory Msn2 nuclear localization (100 nM), 115 mW light or 0.4 mM H 2 O 2 , Fig. 4e), suggesting that the response delay reflects the time needed for H 2 O 2 signalling to PKA.
Tsa1 governs Msn2 through PKA cellular localization. So far, our data indicate that light signalling prevents PKA-dependent inhibition of Msn2 nuclear export by counteracting the phosphorylation of NES residues. To further decipher how the light pathway modulates PKA activity, we monitored PKA subcellular location, using GFP fusions of its catalytic subunits. In agreement with published data 47,48 , PKA catalytic subunits concentrated in the nucleus in the absence of H 2 O 2 (Fig. 4f-h and Supplementary Fig. 4q-s), in a manner dependent upon the PKA regulatory subunit Bcy1 (Fig. 4f-h and Supplementary Fig. 4r). In contrast, upon H 2 O 2 treatment the two most highly expressed PKA catalytic subunits Tpk1 and Tpk2, but not Bcy1 and Tpk3, rapidly redistributed from the nucleus into small cytoplasmic foci (Fig. 4f-h and Supplementary Fig. 4q,r). Such nucleus to cytoplasm redistribution of Tpk1 and Tpk2 in response to H 2 O 2 depended on Tsa1-mediated signal transduction, as Tpk1 and Tpk2 remained concentrated in the nucleus in cells lacking Tsa1 or carrying a serine substitution of the catalytic Cys48 (Fig. 4g,h). The constitutively cytoplasmic localization of Tpk1 and Tpk2 in cells lacking Bcy1 (Fig. 4f-h) offered a possibility to further test the impact of nucleus to cytoplasm relocalization of PKA catalytic subunits upon the addition of H 2 O 2 on Msn2 nuclear localization. In cells lacking Bcy1, Msn2 does not localize to the nucleus because of increased PKA-dependent phosphorylation (Fig. 3a). This inhibition can, however, be overcome by substituting Msn2 serines relevant for nuclear transport by alanines (Fig. 3c) and, in fact, Msn2 concentrated much more rapidly in the nucleus of bcy1D msn2S288AS304A cells (Figs 3c and 4i), suggesting that Bcy1-dependent nuclear concentration of PKA catalytic subunits delays the Msn2 response upon illumination and H 2 O 2 addition. We conclude that the peroxiredoxin Tsa1 regulates Msn2 at the level of the subcellular localization of the PKA catalytic subunits, and that nuclear concentration of PKA is required for the delayed Msn2 response upon H 2 O 2 .

Discussion
For some time it has been known that the S. cerevisiae transcription factor Msn2 rhythmically shuttles into and out of the nucleus in response to light 6 , by a mechanism involving the modulation of PKA-dependent phosphorylation, which prevents Msn2 accumulation in the nucleus 12 . However, it is not known by which mechanism PKA actually detects the presence or absence of light. We have here identified this light-sensing pathway and showed that it comprises a conserved peroxisomal oxidase, Pox1, the antioxidant enzyme peroxiredoxin Tsa1 and its cognate oxidoreductases, the redundant cytosolic thioredoxins Trx1 and Trx2 (Fig. 5) ARTICLE counteracting the nuclear retention of the two most highly expressed catalytic subunits of PKA 48 . Tpk2 has previously been shown to be transported out of the nucleus and into small cytoplasmic messenger RNA processing bodies upon nutrient starvation 48 and in cells lacking Igo1 and Igo2 (ref. 49), which regulate the transition to the stationary phase through the 5 0 -3 0 mRNA decay pathway 50 , suggesting that the subcellular localization of PKA is altered also in response to other stimuli that modulate PKA activity. The almost constitutive nuclear localization of Msn2 in a mutant in which neither Tpk1 nor Tpk2 concentrate in the nucleus (bcy1D , Figs 3c and 4g,h) and the lack of Msn2 response in mutants not transporting the PKA catalytic subunits out of the nucleus upon the addition of H 2 O 2 (tsa1D and tsa1C48S, Fig. 4g,h) Fig. 4u,v). Importantly, the Msn2 response was still postponed in tsa1DYF mutant cells (Fig. 2d), suggesting that the slow nuclear exclusion of the PKA catalytic subunits is important for the delayed Msn2 nuclear accumulation in response to H 2 O 2 and light. The additional target(s) of the Tsa1-mediated pathway could be PKA, a regulatory protein in the PKA pathway, or possibly the phosphatases PP1 and PP2A, which also regulate Msn2 nuclear accumulation 10,11 . In fact, PP1 was suggested to modulate glucose starvation-induced Msn2 nuclear localization by targeting PKA-cognate serine residues involved in Msn2 nuclear transport 9 .
What is the origin of the rhythmic nature of the oscillations in Msn2 nucleo-cytoplasmic shuttling? One possibility is that the light signal diminishes PKA activity to levels at which negative feedback regulation of the pathway periodically releases upstream stimulatory modules from feedback inhibition, leading to oscillatory Msn2 shuttling into and out of the nucleus 12 . The fact that partial inactivation of PKA activity recapitulated Msn2 oscillations in the absence of light (Fig. 4c,d) point to the importance of sensitive and partial reduction of PKA activity for the oscillatory behaviour. However, a role of the phosphatases cannot be ruled out, based on the fact that deficient Msn2 NES phosphorylation fully restored the lack of light response in cells lacking Tsa1 (Fig. 3c), whereas overproduction of the cAMP phosphodiesterase only partly suppressed this defect (Fig. 4a).
The role of a peroxisomal oxidase in light-induced H 2 O 2 production in yeast reported here, as well as in mouse, monkey and human cell lines 29 , together with the messenger function of H 2 O 2 in zebrafish, which couples the light signal to the circadian clock 30 , indicates that H 2 O 2 might be a conserved second messenger by which light affects cellular physiology in general and circadian clocks in particular. Interestingly, recent data in flies support a role for H 2 O 2 in signalling ultraviolet light independent of normal light-sensing pathways through two specific isoforms of the transient receptor potential A1 membrane channel 51 . Our data support a receptor role for peroxiredoxins in the regulation of light-induced ultradian rhythms in yeast and given the high degree of conservation of all components of the pathway, probably also in other organisms. The exquisite sensitivity of peroxiredoxin catalytic cysteines to H 2 O 2 makes them ideally suited both as receptors to relay light-induced H 2 O 2 signals, as well H 2 O 2 scavengers to terminate them 52 . By identifying a peroxiredoxin as a light-receptor of cAMP-PKA, which constitutes a pleiotropic nutrient-responsive pathway regulating, for example, core circadian clock mechanisms 53 and ageing 15 , our data contribute to the understanding of the roles of light, H 2 O 2 and peroxiredoxins in the control of endogenous rhythms in general. The data also suggest that peroxiredoxin hyperoxidation and reactivation play a critical role in modulating endogenous rhythmicity by turning off and on H 2 O 2 -mediated signal transduction, respectively. On a final note, it is well established that aged organisms display less rhythmical circadian clocks 54 and conversely that behaviourally arrhythmic animals exhibit accelerated ageing 55 . Given the roles of peroxiredoxins in both stimulating longevity [21][22][23][24] and in controlling biological rhythms 1,25,26 , further studies to answer the intriguing question of how the control of these processes are linked through peroxiredoxins are warranted.
Plasmids. The plasmids used in this study are listed in Supplementary Table 2. Msn2-GFP was expressed from the plasmid pADH1-MSN2-GFP 8 based on the yCPlac111 backbone containing a LEU2 marker and MSN2 expression controlled by the ADH1 promoter. pHR1668 and pWR78 derive from pAMG5 (ref. 8). The MSN2 sequence encoding amino acids 1-264 was integrated using standard cloning techniques. The S304A point mutations in plasmids pWR76 and pWR78 were constructed using a QuikChange Mutagenesis Kit II (Stratagene) according to the manufacturer's recommendations on template plasmids pAMG and pHR1668, respectively. The plasmid pHyPerRedSc was constructed by synthesizing the HyPerRed H 2 O 2 sensor 33 , codon optimized for yeast (sequence available upon request) and subcloning it into the vector pRS416-GPD 62 using BamHI and EcoRI sites inserted upstream and downstream of the HyPerRedSc open reading frame, respectively (GenScript). C199S point mutagenesis was performed by changing a TGT codon into a TCT in the HyPerRedSc coding sequence (positions 361-363, GenScript). Plasmid pRS315-TRX2-ProteinA was constructed in a two-step procedure. First the TRX2 open reading frame and 300 bp upstream and 164 bp downstream flanked by BamHI and PmlI sites was PCR amplified and cloned into plasmid pRS315 (ref. 8). Second, a Protein A tag was incorporated at the TRX2 C terminus through PCR amplification of the Protein A tag and a stop codon flanked by AlfI and SnaBI sites using plasmid pBS1479 as a template 63 . Plasmid pWR386 was constructed similarly as plasmid pGGM2-ZF described previously 37 , through subcloning of (1) the GAL4 promoter flanked by SacII and SalI sites PCR amplified from genomic DNA into SacII and SalI sites in plasmid pAMG and (2) a GST-tag digested from plasmid pGEX-5X-3 using SalI-XhoI and subcloned into the SalI site of pAMG. pKP7 was constructed by first subcloning an MSN2 fragment containing S288A, S582A, S620A, S625A and S633A mutations from pCUP1MSN2A5-GFP 64 into pWR386 and then by introducing the S304A mutation using the QuikChange Mutagenesis Kit II (Stratagene) according to the manufacturer's recommendations.
Growth conditions. Strains were grown in the dark at 30°C in buffered synthetic defined medium (Yeast Nitrogen Base (Formedium) with 0.5% ammonium sulfate, 2% glucose and complete supplement mixture without appropriate amino acids, 1% succinic acid and 0.6% NaOH to buffer the medium to pH 5.8). Overnight cultures were inoculated the day before the experiment and grown to an OD 600 of 0.4-0.5 in the morning the day of the microscopy experiment.
Fluorescence microscopy. Images were acquired using automated epifluorescence microscopes (TE2000E-PFS, Nikon Instruments or Olympus IX81) equipped with Â 60 oil-immersion objectives (numerical aperture 1.4, Plan Apochromatic, Nikon Instruments or PlanApoN Â 60/1.42 Oil, Olympus) and electron-multiplying charge-coupled device cameras (iXon DU-885-CS0-#VP, Andor Technology or a 12-bit Hamatsu camera). The yeast cells were kept in a heated perfusion chamber (FCS2, Bioptechs Inc.) at 28°C to avoid heat-induced stress responses. The objective was heated to 26.2°C (according to the manufacturer's instructions) to maintain a stable temperature in the perfusion chamber. The cover glasses was precoated for 1.5 h with protein concanavalin A, 0.5 mg ml À 1 in 0.01 M PBS, to immobilize yeast cells on the surface. The coating did not induce any stress response. By combining neutral density filters, the illumination intensity was in most experiments set to 115 mW in the microscope field of view, unless otherwise noted. GFP fluorescence imaging was accomplished with a mercury lamp, an excitation filter (470 ± 20 nm), a dichroic mirror (505 nm) and an emission filter (515 ± 15 nm) (Chroma Technology) using the Nikon microscope. On the Olympus microscope, GFP was imaged using a quadruple emission filter at 521 ± 17 nm and an excitation bandpass filter at 485 ± 10 nm. The microscope setups were controlled with the NIS-Elements software (Nikon) and the Xcellence software (Olympus), respectively. HyPerRedSc fluorescent imaging was accomplished in the Olympus system using a quadruple filter set with 606 ± 25 nm for detection and an excitation bandpass filter at 560±12.5 nm was used for imaging.
Illumination of cells. Illumination of cells was accomplished simultaneously with GFP excitation as previously described 14 . Cells were continuously illuminated with blue light (450-490 nm) and fluorescence images were typically acquired every 4 s. Bright-field images acquired once every 60 s with a small focus offset were used to identify cells in the CellStat software 14 . After each Msn2-GFP imaging experiment, non-illuminated cells were imaged to ensure that only cells that had been exposed to light exhibited nuclear Msn2p localization.
H 2 O 2 treatment. Growth medium containing H 2 O 2 at the desired concentration was injected into the perfusion chamber at the start of the experiment. Bright-field and fluorescence images were captured every 30 s at a low intensity (26 mW) to avoid light-induced stress ( Supplementary Fig. 1c). The H 2 O 2 solution was made fresh before the start of each experiment.
Image and signal analysis. Image and signal analysis was performed using the CellStat software 14 . To quantify the degree of Msn2 nuclear localization, the signal intensity in the nucleus was compared with that in the cytosol 14 . Signal analysis was implemented using routine MATLAB commands. The nucleocytoplasmic localization trajectories were ordered in MATLAB by means of non-metric multidimensional scaling from low responders to high responders 65 Fig. 1b,c, Fig. 1f) and bleaching-sensitive HyPerRedScC199S fluorescence (Supplementary Fig. 1b,c,f) at the corresponding time point. Graphs presented represent the average bleaching-corrected HyPerRed fluorescence (Fig. 1c,h and Supplementary Fig. 1g).
Ccp1 immunoblot analysis. Levels of Ccp1 protein were determined using immunoblot analysis as previously described 22 , the Odyssey Infrared Imaging system (LICOR Biosciences) and rabbit anti-Ccp1 serum 66 (dilution 1:4,000) kindly provided Dr T. Tatsuta. Simultaneous detection of the Pgk1 protein using a mouse anti-Phosphoglycerate Kinase Monoclonal Antibody (22C5D8, Novex part no 459250, Thermo Scientific, dilution 1:10,000) served as a loading control.
Ratios (H 2 O 2 -treated/untreated) of Msn2 phosphopeptides were obtained by comparing the MS1 peak intensities of SILAC-labelled peptides from H 2 O 2 -treated cells grown in the presence of a light-isotope reagent ( 12 C-arginine and -lysine) with the peak intensities of untreated cells labelled with a heavy-isotope reagent ( 13 C-arginine and -lysine). To quantify phosphorylation in untreated tsa1D cells (Supplementary Tables 1 and 2), peak intensities were compared between tsa1D cells grown in the presence of a light-isotope reagent ( 12 C-arginine or -lysine) and wild-type cells labelled with a heavy-isotope reagent ( 13 C-arginine or -lysine).
Samples were subjected to reversed-phase nano-HPLC (Ultimate 3000, Dionex, Sunnyvale, CA, USA) using the following setup: peptides were loaded onto a precolumn (PepMAP C18, 0.3 Â 5 mm, Dionex), desalted for 30 min with 0.1% trifluoroacetic acid (20 ml min À 1 ), and eluted onto an analytical column (PepMAP C18, 75 mm Â 150 mm, Dionex) for 1 h by means of an organic solvent gradient (275 ml min À 1 ), as described elsewhere 67 . The nano-HPLC was directly coupled to an LTQ Orbitrap Velos mass spectrometer (Thermo Scientific) via a nanoelectrospray ion source (Proxeon). The capillary temperature was set to 200°C. The source voltage on the metal-coated nanoelectrospray ion emitters (New Objective) was 1.5kV. The 12 most abundant ions were subjected to MS/MS analysis and monoisotopic precursor selection was enabled. For some of the experiments (indicated as set I in the material available online at proteomeXchange.org, see below), MS3 analysis was performed; for the other experiments (set II), multistage activation was triggered at neutral losses of 98, 49 and 32.6. Fragmented precursors were excluded from further fragmentation for 30 (respect. 60) s (with 5 p.p.m. accuracy) and singly charged peptides were generally excluded from MS/MS analysis.
Peptide identification and SILAC quantification were performed using the SEQUEST algorithm in the Proteome Discoverer software package (ver. 1.4.0.288, Thermo Fisher Scientific) with settings similar to those described elsewhere 67 . In short, phosphorylation of Ser/Thr/Tyr, neutral loss of water from Ser/Thr, oxidation of Met and 13 C 6 Lys/Arg were set as variable modifications. In addition, carbamidomethylation of Cys was set as a static modification. Spectra were searched against the Saccharomyces Genomic Database (6717 entries, 03 February 2011) including contaminants with tryptic specificity, allowing two missed cleavages, a precursor mass tolerance of 3 p.p.m. and a fragment ion tolerance of 0.8 Da. Quantification settings were the default values for the precursor ion quantifier and the event detector was set to 2 p.p.m. The results were filtered at the XCorr values to a false discovery rate of 1% on the peptide level, as described elsewhere 68 . Additional manual validation of peak assignments was performed in the Xcalibur Qual Browser (Thermo Scientific) using a 2 p.p.m. mass window centred on the exact precursor mass. Light/heavy (L/H) ratios of each SILAC experiment were normalized and Arg-Pro conversion was taken into account as described elsewhere 44,67 .
The probability of phosphorylation site localization was calculated using phosphoRS 2.0 software 69 implemented in Proteome Discoverer. A phosphorylation site probability of 75% or higher was considered confidently localized. All identified peptide spectrum matches of Msn2 can be found in Supplementary Data 2. All raw MS data files have been deposited at the ProteomeXchange Consortium 70 (http://www.proteomexchange.org) via the PRoteomics IDEntifications partner repository with the dataset identifier PXD001853. Peptide entries in Supplementary Data 2 were filtered for the presence of an L/H ratio and phosphopeptides with a confidence in phosphorylation site allocation other than 'high' (high ¼ phosphoRS 2.0 probability of Z75%) were removed. The remaining peptide entries were grouped according to their biological samples (all.raw files of the same sample) and designated as replicates of an experimental condition. Phosphopeptides spanning the same set of phosphorylated residues were further grouped into and designated as a phosphorylation site group (PSG). Unphosphorylated counter peptides (amino acid sequence must harbour at least one of the phosphorylated amino acid residues of the corresponding PSG) were additionally grouped. PSG and counter-peptide ratios were log 2 transformed and averaged, and a confidence interval of ± 1 s.d. was calculated. Resulting log 2 values were reversed and are summarized in Supplementary Data 1 (see worksheets 'n ¼ 2 (biological replicates)' (hidden columns) and 'n ¼ 1 (biological replicates)'). Log 2 averaged ratios of all individual biological replicates of an experimental condition were further used to calculate the average ratio and confidence interval of the corresponding experimental condition. The resulting log 2 values were reversed and are summarized in Supplementary Data 1.
Immunoprecipitation of Msn2 and NES phosphorylation analyses. Cells were harvested using filtration at the indicated time points following the addition of H 2 O 2 . Cells were lysed using 10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.5 mM EDTA, 0.5% Triton X-100, protease inhibitor cocktail (Roche), 1 mM sodium vanadate and 1 mM NaF. Cell lysates were incubated for 1 h at 4°C with anti-GFP beads (GFP-Trap_A, Chromotek). Beads were washed with lysis buffer and eluted in loading buffer. Total amounts of the Msn2NES-GFP fusion protein expressed from pAMG9 (ref. 8) were visualized using anti-GFP (Roche Molecular Biochemicals, dilution 1:10,000) and PKA-dependent phosphorylation using a phospho-PKA substrate antibody (number 9624, Cell Signaling, dilution 1:1,000). A representative uncropped blot picture of the anti-GFP and anti-phospho PKA immunoblot analysis of Msn2-NES phosphorylation can be seen in Supplementary  Fig. 3n. The detection of PKA-specific phosphorylation by the phospho-PKA substrate antibody was verified by complete loss of antibody reactivity on total yeast extracts prepared from the strain YTpkCMx3 upon treatment with 500 nM 1-nM-PP1.
Statistical analyses. To assess statistically significant Msn2 localization differences in mutant strains and differences in various treatments, analysis of variance and Tukey's tests were performed on data obtained from the 685 time points between 10 and 60 min following the start of the treatment. Analysis of variance confirmed similar variance between groups compared. Tukey's tests were suitable, because such Msn2 localization data mostly conformed well to a normal distribution. A P-value of o0.05 was considered to indicate a statistically significant difference and, for all experiments, percentages of data points expressing significant differences were calculated (see Supplementary Note 1). Values for the time Msn2 spent in the nucleus or the time to first Msn2 nuclear localization were less clearly normally distributed and were instead tested for statistical significance using two-sided Mann-Whitney U-test 14 (see Supplementary Note 1).
To assess statistically significant differences between HyPerRed fluorescence in different strains or between HyPerRed and HyPerRedC199S expressing strains, non-bleaching corrected fluorescence values were compared ( Supplementary  Fig. 1b,c,f) using two-sided Mann-Whitney U-tests (see Supplementary Note 1). This test was chosen because the distribution of HyPerRedSc or HyPerRedScC199S fluorescence was skewed towards highly fluorescing cells.
Differences in the amount of Msn2-NES phosphorylation (Fig. 3h) were analysed for statistical significance using two-sided Student's T-tests (see Supplementary Note 1).
Data availability. All raw MS data files have been deposited at the ProteomeXchangeConsortium (http://proteomexchange.org) via the Proteomics IDEntifications partner repository with the dataset identifier PXD001853. The data that support the findings of this study are available from the corresponding author on request. The MATLAB code used in signal analysis is available from the corresponding author upon request.