Caloric restriction controls stationary phase survival through Protein Kinase A (PKA) and cytosolic pH

Abstract Calorie restriction is the only physiological intervention that extends lifespan throughout all kingdoms of life. In the budding yeast Saccharomyces cerevisiae, cytosolic pH (pHc) controls growth and responds to nutrient availability, decreasing upon glucose depletion. We investigated the interactions between glucose availability, pHc and the central nutrient signalling cAMP‐Protein Kinase A (PKA) pathway. Glucose abundance during the growth phase enhanced acidification upon glucose depletion, via modulation of PKA activity. This actively controlled reduction in starvation pHc correlated with reduced stationary phase survival. Whereas changes in PKA activity affected both acidification and survival, targeted manipulation of starvation pHc showed that cytosolic acidification was downstream of PKA and the causal agent of the reduced chronological lifespan. Thus, caloric restriction controls stationary phase survival through PKA and cytosolic pH.

recapitulate most of the glucose-dependent transcriptional response observed in such cultures. Proper PKA inactivation is also required for survival during nutrient-poor conditions. When cultures are subjected to severe carbon starvation during stationary phase, over-activation of the PKA pathway shortens CLS, while mutations that reduce its activity are well known to extend viability (Fabrizio et al., 2003).
The main regulation of PKA kinase activity is by fermentable sugars, and consequently, most research has focused on elucidating the glucose signalling mechanism. The PKA kinase is a heterotetramer composed of two regulatory (Bcy1) and two catalytic subunits (Tpks) in its inactive form. Activation of the kinase occurs when the second messenger cAMP binds to the regulatory subunits, releasing the catalytic subunits, which are encoded by three partially redundant isoenzymes (Conrad et al., 2014;Thevelein & De Winde, 1999).
Therefore, cAMP levels are key for PKA regulation. Glucose addition to de-repressed cultures induces a transient cAMP increase by the activation of adenylate cyclase (Cyr1) via two branches of the pathway: Ras and the G protein-coupled receptor system. Of these two branches, only Ras signalling is essential for PKA activation and growth (Conrad et al., 2014). The concentration of cAMP is downregulated via degradation by the phosphodiesterases Pde1 and Pde2 (Ma, Wera, Dijck, & Thevelein, 1999). While the phosphodiesterases and other regulators of [cAMP] are upstream of PKA, they are PKA targets themselves, contributing to a negative feedback mechanism and the transient nature of the glucose-induced cAMP peak (Vandamme, Castermans, & Thevelein, 2012).
PKA inactivation at diauxic shift is required for proper diauxic transition, post-diauxic growth and stationary phase survival or CLS (Boy-Marcotte et al., 1996;Russell, Bradshaw-Rouse, Markwardt, & Heideman, 1993). However, very little is known about the mechanisms for PKA inactivation when glucose becomes depleted at the diauxic shift. The levels of the inhibitory Bcy1 increase around this time, which was assumed to contribute to PKA inhibition (Winderickx et al., 2003). However, Tpk1 and Tpk2 levels increase in parallel to Bcy1 and PKA may therefore not be inhibited by this additional cAMP/Bcy1 control (Tudisca et al., 2010). Whether changes in the localisation of the Tpks and Bcy1 upon glucose depletion contribute to the inhibition, remains to be stablished (Tudisca et al., 2010).
Changes in cytosolic pH (pH c ) alter the protonation state ratio of all weak acid and basic groups present in the cytosol, thereby potentially affecting most if not all processes occurring inside a cell (Orij, Brul, & Smits, 2011). Recently pH c has been shown to function as a second messenger regulating gene expression (Young et al., 2010), G protein-mediated signalling (Isom et al., 2013), growth (Dechant, Saad, Ibáñez, & Peter, 2014;Orij et al., 2012) and aging (Henderson, Hughes, & Gottschling, 2014) in yeast. In higher organisms, intracellular pH appears to have similar roles and its dysregulation has been linked to cancer progression and neurodegenerative diseases (Harguindey et al., 2017;White, Grillo-Hill, & Barber, 2017).
It is therefore interesting to note that pH c is strongly influenced by nutrient availability. Whereas the pH in the cytosol remains around neutral values during growth on glucose, upon glucose depletion at the end of the growth phase, pH c decreases~1 pH unit (Orij et al., 2012). Imposed abrupt glucose starvation also leads to a strong decrease of pH c (Dechant et al., 2010). A small pH c decrease during the normal growth phase has been shown to act as a growth limiting signal. The signal transduction of this control remains unclear (Orij et al., 2012), but an interaction with regular nutrient signalling is to be expected. Intracellular pH was proposed to control PKA, but different and apparently opposite modes of control have been reported. Intracellular acidification by addition of protonophores at low pH is able to trigger cAMP induction and concomitant activation of PKA, similar to how glucose addition to de-repressed cells activates the pathway (Thevelein & De Winde, 1999). Low intracellular pH triggers cAMP synthesis via Ras (Colombo et al., 1998), but also via direct biochemical regulation, as adenylate cyclase activity increases at acidic pH (Purwin, Nicolay, Scheffers, & Holzer, 1986). Glucose addition to starved cells itself causes a transient cytosolic acidification, in a timescale similar to the cAMP peak (Tarsio, Zheng, Smardon, Martínez-Muñoz, & Kane, 2011). Whether it actually is the glucose-dependent transient acidification that triggers cAMP induction upon glucose readdition remains unclear; the two have been suggested to be independent, although kinetic analyses reveal that the pH c decrease precedes the cAMP response (Thevelein et al., 1987).
In contrast to the previous findings, low pH c has been proposed to inactivate PKA via regulation of the vacuolar H + -ATPase (V-ATPase). Disassembly of V-ATPase responds to pH c perturbations and lack of V-ATPase decreases PKA activity both upon glucose depletion and re-addition (Dechant et al., 2010(Dechant et al., , 2014. Acidification by Pma1 depletion also caused Ras inhibition and a pH-dependent growth arrest (Dechant et al., 2014). Overall, these data support a model in which cytosolic acidification inhibits PKA activity, at least partially via Ras, which appears to contradict the observations that showed a positive effect of acidification on cAMP/PKA activity.
In addition to pH control of PKA, PKA was found to regulate both plasma membrane and vacuolar H + -pumps, Pma1 and V-ATPase (Bond & Forgac, 2008;Souza, Trópia, & Brandão, 2001), which suggests a role of PKA in pH c control. Here, we address this question and show that pH c is controlled by PKA in a glucose concentrationdependent manner and that these changes in pH c are an important determinant of calorie restriction control of CLS.

| PKA promotes cytosolic acidification upon glucose depletion
To study how the PKA pathway is involved in the regulation of pH c , we monitored pH c and OD 600 (optical density at 600 nm) over the course of the progression through all growth phases in a set of mutants previously described to either over-activate or reduce the activity of the pathway. PKA over-activation was achieved by deletion of the PKA regulatory subunit (bcy1Δ), deletion of the phosphodiesterases (pde1Δpde2Δ) and deletion of the Ras regulatory proteins (ira1Δira2Δ). Decreased PKA activity was induced by destabilisation of adenylate cyclase mRNA (DAmP CYR1), overexpression of one of the phosphodiesterases (PDE2 o.e.) and expression of an adenylate cyclase allele with reduced activity (fil1). As previously described (Orij et al., 2012), the wild-types maintained pH c around neutrality during exponential growth, while pH c decreased almost 1 unit when glucose was depleted. Qualitatively, the PKA mutants analysed presented a pH c -time profile similar to their respective parental strain, with a pH c around seven during growth and a strong acidification upon glucose depletion (Figure 1a, Supporting information Figure S1).
Systematic quantitative assessment of pH c during the exponential growth phase ( Figure 1b) and a set time after glucose depletion (Figure 1c) revealed that altered PKA activity did not strongly affect pH c during exponential growth; only the low PKA activity mutant fil1, presented a pH c significantly different from its parental strain. Upon glucose depletion, however, we did observe a clear effect of PKA pathway mutations on pH c (Figure 1c). The strains with overactive PKA, bcy1Δ and ira1Δira2Δ, had a significantly lower pH c than the parental strain whereas the strains with downregulated PKA, fil1 and PDE2 o.e., had a significantly higher pH c than their respective parental strains 10-11 hr after glucose depletion. The strains pde1Δpde2Δ and DAmP CYR1, with respectively overactive and reduced PKA activity fit within this trend, but the difference was not significant.
These results indicate that the PKA pathway affects pH c specifically when glucose is depleted.
Direct activation of PKA with cAMP in wild-type cells also affected pH c upon glucose depletion. We treated BY4741 exponentially growing cultures with cAMP and monitored OD 600 and pH c before and after the treatment (Figure 2a). During growth in the presence of glucose, cultures treated with cAMP did not have an altered pH c , but upon glucose depletion they showed a stronger cytosolic acidification in a dose-responsive manner ( Figure 2b). Addition of cAMP during growth in the strain lacking the phosphodiesterases (pde1Δpde2Δ) caused a stronger effect on pH c upon glucose depletion than in the parental strain, as expected due to reduced cAMP degradation (Supporting information Figure S2). This further supported the idea that PKA activity promotes cytosolic acidification when glucose is depleted.
To test the effect of PKA on pH c independently of growth history, including medium composition (i.e. nutrients available or growth by-products), we performed controlled glucose starvation experiments, in which we washed and resuspended cultures in medium without glucose with various cAMP treatment regimens (Figure 3a Protein Kinase A is essential for growth and genetic manipulations that completely inactivate the pathway are lethal, unless compensated with additional mutations (Broach, 2012). An alternative way to inhibit PKA is the use of strains carrying ATP analogue-sensitive (as) mutations. These point mutations do not affect kinase activity under control conditions but render the catalytic subunit sensitive to the ATP analogue 1NM-PP1 and allow inhibition of the kinase activity by addition of the drug to the media (Stephan, Yeh, Ramachandran, Deminoff, & Herman, 2009;Zaman et al., 2009). In order to evaluate the effect of PKA inhibition on pH c , we performed F I G U R E 1 Genetic manipulation of PKA activity modulates pH c upon glucose depletion. Mutants with overactive PKA activity (green) have reduced pH c and low PKA activity mutants (red) an increased pH c when glucose is depleted from the media. (a) BY4741 (black), bcy1Δ (green) and DAmP CYR1 (red) strains were grown in microplates and OD 600 (filled symbols) and pH c (open symbols) were monitored. A representative example is shown. (b) and (c) summarise pH c during exponential growth (b) and after glucose depletion (c) for the set of PKA mutants analysed (See Materials and Methods for details). Representative OD 600 and pH c curves for each strain can be found in Supporting information Figure S1. Data represent average ± standard deviation (SD) of at least three biological replicates per strain starvation experiments with the strain TPK1 as tpk2Δtpk3Δ, which carries a single analogue-sensitive catalytic subunit of PKA and allows inactivation of PKA kinase activity at any time by 1NM-PP1 addition.
Inactivation of the kinase was induced by incubating the cultures with 2 µM of 1NM-PP1 before the starvation as previously (Aoh, Graves, & Duncan, 2011). This concentration inhibited growth in cultures of TPK1 as tpk2Δtpk3Δ but not in the parental strain (Supporting information Figure S4; and Stephan et al., 2009), suggesting that PKA was specifically inhibited in this mutant. Addition of 1NM-PP1 to wild-type cultures did not affect the pH c in the presence of glucose and did not have additional effects on the acidification upon glucose starvation (Supporting information Figure S5). As with mutants inactivating PKA, and opposite to PKA activation with cAMP, treatment of the TPK1 as tpk2Δtpk3Δ cells with 1NM-PP1 had no effect on pH c in the presence of glucose, but pre-treatment with the inhibitor delayed and significantly reduced the extent of acidification upon glucose starvation (Figure 3c-d). These effects are independent of strain background (Supporting information Figure S6).
PKA promoted pH c decrease after gradual glucose consumption at the end of the growth phase (glucose depletion; Figures 1 and 2), as well as when glucose was removed from growing cultures (glucose starvation; Figure 3), indicating that the mechanism behind the regulation of pH c upon gradual (glucose depletion) versus sudden glucose removal (starvation) is the same. Taken together, our results show that PKA activity, set when glucose is still present, regulates pH c upon glucose starvation. Induction of the PKA pathway in the presence of glucose enhances cytosolic acidification during glucose starvation, whereas a decrease of PKA activity strongly reduces the glucose starvation-induced pH c decrease.

| Calorie restriction controls pH c via PKA
The fact that PKA activity before glucose depletion sets pH c upon glucose starvation reminded us of CR effects on CLS, where glucose levels during growth affect viability after glucose depletion (Murakami et al., 2008). We asked whether CR would also affect starvation pH c , similarly to PKA manipulation. We inoculated yeast in a range of glucose concentrations and monitored OD 600 and pH c during lag phase, growth and after glucose depletion (Supporting information FigureS7), observing a dose-dependent decrease of pH c after glucose depletion as glucose concentration increased (Figure 4a). In controlled starvation experiments, we grew cultures on media containing 2% (control) as well as 1% and 0.5% glucose (CR), starved them for glucose and monitored pH c as well as viability 3 days later.
Decreasing glucose levels did not influence pH c in the presence of glucose, but significantly reduced cytosolic acidification upon starvation ( Figure 4b, d) which correlated with increased viability three days later (Figure 4c).
To confirm that the glucose control of pH c is mediated by PKA, we reasoned that the manipulation of PKA activity should abolish the effects of glucose concentration on pH c . We inhibited PKA using the TPK1 as tpk2Δtpk3Δ mutant, and found that, indeed, starvation pH c now remained high, and became insensitive to the glucose concentration (Figure 4e-f). Complementarily, we analysed starvation pH c in the mutant bcy1Δ which lacks the regulatory subunit of the kinase and therefore has fully active PKA. Cultures lacking BCY1 presented a similarly low starvation pH c after growth at both 2 and 1% glucose ( Figure 4g). We do not show data for 0.5% glucose because bcy1Δ cultures had depleted this low amount of glucose within the time of the experiment (see Experimental Procedures). Therefore, glucose availability during growth and prior glucose depletion regulates starvation pH c via PKA.

| Effects of cytosolic pH on starvation survival
The fact that PKA actively controls pH c during starvation suggests a functional role of pH c in the adaptation to non-glucose conditions.
To test this hypothesis, we studied the consequences of changes in starvation pH c on viability during starvation. Cytosolic pH in the absence of glucose depends on the pH of the medium (extracellular pH; pH ex ); Therefore, we manipulated starvation pH c by starving cultures for glucose in media with a pH ex in a range from 3-7 and determined viability three days after starvation by colony-forming units counts. Cytosolic pH during glucose starvation decreased with pH ex (Figure 5a). Between pH ex 3-6, starvation survival correlated strongly with pH c , with viability decreasing at lower pH c (Figure 5e, grey bars); at pH ex 7, this correlation collapsed. The pH-dependent Our data suggest that acidification of the cytosol upon glucose depletion reduces CLS. If this is a direct causal relationship, increasing pH c during starvation should improve starvation survival. We attempted to increase pH c by inducing PMA1 overexpression prior the starvation (Henderson et al., 2014), but found no significant effects on starvation pH c (Supporting information Figure S9). Therefore, to induce a high starvation pH c , we manipulated PKA activity in the TPK1 as tpk2Δtpk3Δ strain, using as low a dosage of 1NM-PP1 as possible, which still induced a significant increase in pH c . Starvation of untreated TPK1 as tpk2Δtpk3Δ cultures at a range of pH ex leads to a pH c response similar to wild-type ( Figure 5c). Inhibition of PKA activity prior the starvation prevented the decrease of pH c (Figure 5d), as expected from our previous observations. Under these conditions, the starvation pH c became mostly insensitive to pH ex and stabilized around 6.5 in all starvation conditions. The survival three days after starvation for control TPK1 as tpk2Δtpk3Δ was again similar to wild-type (Figure 5f, grey bars). As hypothesised, 1NM-PP1 treatment abolished not only the pH c response to pH ex but also the loss of viability ( Figure 5f, black bars), supporting the idea that the PKA-controlled reduction in pH c limits starvation viability.

| PKA control of starvation pH c
The Protein Kinase A (PKA) pathway is a key regulator of cellular responses, coordinating the balance between growth and stress responses. Recently, pH c has also been shown to regulate growth (Dechant et al., 2014;Orij et al., 2012). More specifically, pH c seems to be a sensor of environmental or metabolic state, connecting carbon source availability with growth regulation. PKA and pH c share a common major input (carbon source availability) and output (growth control). Hence, they presumably interact to ensure a coordinated response. Literature addressed part of this interaction and reported We analysed the effect of manipulation of PKA activity on pH c (Figures 1 and 2). PKA activity did not regulate pH c during growth, but we did observe a strong role of PKA in pH c control upon glucose depletion, which was recently corroborated (Isom et al., 2018). This is remarkable, for two reasons. First, it was usually assumed that starvation acidification was passive, because of the absence of energy and consequently the inactivation of the ATPase Pma1. But we found that Pma1 inhibitor ebselen could further reduce starvation pH c , a clear indication of activity of the pump (Figure 5a-b). The fact that a PKA inactivating mutation resulted in a higher starvation pH c, independently of pH ex (Figure 5d), also shows that starvation pH c is an actively controlled property. Secondly, activity of PKA itself requires cAMP and ATP, both of which are low after glucose depletion (Ashe, Long, & Sachs, 2000;Russell et al., 1993), and the pathway is thought to inactivate upon diauxic shift (Thevelein & De Winde, 1999). Careful assessment of the effect of timing of the cAMP addition showed that it was the activity of PKA as set prior to glucose starvation and not during the starvation itself that controlled the starvation pH c (Figure 3a-b). Together, these results show that the level of PKA activity, controlled by glucose abundance before glucose depletion when the pathway is still active, actively controls pH c in the absence of glucose. A similar role for PKA activity during growth controlling processes after glucose depletion was observed in the delocalisation of trans-Golgi/endosomal adaptors upon acute glucose depletion (Aoh et al., 2011), suggesting a common mechanism for PKA control in the glucose to non-glucose transition. The fact that different initial concentrations of glucose ( Figure 4) or addition of cAMP (Figures 2 and 3) quantitatively set pH c via PKA modulation also shows that PKA is not a mere on/off switch but that it is quantitatively regulated.
How PKA regulates pH c is not known but the main pH c regula- n.s.

F I G U R E 4
The initial concentration of glucose in the media regulates starvation pH c via PKA. (a) pH c after glucose depletion (calculated as in Figure 1c) of BY4741 cultures after growth at the indicated initial concentrations of glucose (glc). A representative growth and pH c curve can be found in Supporting information Figure S7. (b) pH c dynamics of BY4741 cultures starved for glucose after growth at the indicated initial concentrations of glucose (c) Cell viability after three days of glucose starvation as in panel B. (d-g) Cytosolic pH 60 min after glucose starvation for BY4741 (d), TPK1as tpk2Δ tpk3Δ without (e) or with (f) 1NM-PP1 pre-treatment as in Figure 3c-d, and bcy1Δ (g). Data shown are averages ±SD of three biological replicates. Significance was tested using one-way ANOVA with matching with Bonferroni's correction starvation occurs via reversible disassembly of the V 1 and V 0 subunits and seems pH c -dependent (Dechant et al., 2010). This dissociation was impaired in PKA overactive mutants (Bond & Forgac, 2008), although others suggested that PKA is downstream of V-ATPase instead (Dechant et al., 2010(Dechant et al., , 2014.

| The importance of pH c in starvation survival
The active control of cytoplasmic acidification by PKA suggests that this low pH c during starvation is beneficial. What we found, however, was a strong correlation between an acidic starvation pH c and decreased viability after three days below neutral pH ex (Figure 5a-b, e). Low viability at neutral/alkaline starvation pH ex has been associated with lack of enzyme aggregate formation (Petrovska et al., 2014). At neutral pH ex , factors other than pH c may contribute to the viability loss, as suggested by the lack of effect of ebselen (Figure 5e).
For instance, at pH ex 7, pH c is lower than pH ex , reversing the normal proton gradient. Similar to what occurs during growth at alkaline pH ex , this may interfere with nutrient import, which is often coupled to H + symport (reviewed in Ariño, 2010).
It seems remarkable that pH c changes so small might have such strong effects, as we understand that a change of 0.2 pH units will change the protonation state of one group by a factor of only 1.6. It should be noted, however, that pH sensing by macromolecules is already complex; several accessible or hidden amino acids together may form a pH sensing network (Isom et al., 2013), and complexity is added when considering receptor-ligand or enzyme-substrate interactions where both partners can be pH sensing. Phosphorylation, with a pK a completely in the physiological range, adds another layer of sensitivity (Young et al., 2010).
In good agreement with low pH c limiting starvation survival, blocking cytosolic acidification upon glucose depletion by inhibition of PKA resulted in fully retained viability upon starvation (Figure 5c-

f). This PKA inactivation likely activated the transcription factors
Msn2/4 (Görner et al., 1998), so we cannot exclude that this PKA inactivation also affected viability through Msn2/4 and the induction of stress resistance genes (Fabrizio et al., 2003).
The ultimate experiment to show that pH c is the main determinant of survival downstream of PKA is to rescue the low viability of a high PKA mutant by increasing pH c . We attempted to increase pH c in a number of ways. All our attempts at restoring starvation pH c independently from PKA failed, showing all the more how robustly this reduced pH c is controlled. We increased pH ex in the PKA overactive mutant ira1Δira2Δ. This too failed to fully compensate for the strong acidification, and the pH c of ira1Δira2Δ remained lower than wild-type for all the tested pH ex (Supporting information Figure S10A-B). The effects on viability of the small pH c increase from pH ex 5 to 6 were rather limited, but increased survival, although the low pH c in ira1Δira2Δ caused a lower survival than in the parental strain at any pH ex (Supporting information Figure S10C-D).
We conclude that PKA-controlled pH c is a major factor in the control of starvation survival. Yeast cultures naturally experience such severe carbon starvation conditions in stationary phase.
Although the gradual transition into stationary phase likely involves additional adaptations compared to sudden starvation, several observations support the idea that a low pH c regulates stationary phase survival (CLS) as well. First, growth on media at a low pH ex , which enhances cytosolic acidification after glucose depletion (Supporting information Figure S8B), shortens CLS too ( Figure S8C and Burtner, Murakami, Kennedy, & Kaeberlein, 2009;Fabrizio et al., 2004). Second, addition of acetic acid reduces CLS only at low pH ex (Burtner et al., 2009), conditions in which the acid is protonated and therefore can diffuse through the plasma membrane and acidify the cytosol (Kane, 2016). Thus, our observations together with previous literature support that cytosolic pH rather than acidic media reduces CLS.
Glucose abundance promotes CLS via regulation of the PKA signalling pathway (Fabrizio et al., 2003(Fabrizio et al., , 2004Murakami et al., 2008;Wei et al., 2008). Here, we show that glucose abundance promotes cytosolic acidification via PKA and that this enhanced acidification limits CLS. Thus, taken together our data support a model in which increased glucose availability, via quantitative activation of PKA, promotes the decrease of pH c upon starvation, which is one of the mechanisms decreasing viability upon CLS ( Figure 6). This is remarkable: Why would yeast actively decrease pH c if it reduces survival? An explanation for this apparent paradox would be that acidification also has a positive impact on physiology besides the detrimental aspect uncovered here. For instance, acidificationtriggered protein aggregation has been proposed to inactivate enzymes and protect them from damage, facilitating growth resumption when glucose is again available (Petrovska et al., 2014). Also, the key pH c regulator Pma1 is estimated to consume around 20% of cellular ATP under normal conditions (Ambesi, Miranda, Petrov, & Slayman, 2000). Hence, inhibition of the ATP-dependent H + -pumps may reduce energy expenditure at the cost of decreasing pH c . An alternative is that the unfavourable pH c decrease may be a consequence of the favourable PKA-dependent fast fermentative growth programme. PKA activity may have evolved to balance between fast growth and survival, with dysregulations affecting both (Zakrzewska et al., 2011;Zaman et al., 2009). Such a growth-versus-survival trade-off is similar to the model of quasi-programmed aging, according to which the age-related death would be an aftermath of the growth programme (Arlia-Ciommo, Piano, Leonov, Svistkova, & Titorenko, 2014). Alternatively, the death of a fraction of the population may be beneficial too, as it releases nutrients that can be used for growth by the remaining (genetically identical) more adapted cell subpopulation (Fabrizio et al., 2004). What causes cells to lose either metabolic activity or the capacity to resume growth after intracellular acidification is yet undetermined and should be addressed in subsequent studies.

| PKA and pH c interaction: feedback loops upon nutritional transitions
Active protein kinase A is well known to promote fermentable growth and limit starvation survival. We now describe the role of PKA controlling pH c specifically upon glucose depletion. It appears, from literature (Aoh et al., 2013) and our current work, that PKA activity has different effects upon glucose addition than at glucose depletion. This may seem remarkable, but is much less so in the light of the kinase targets present. Upon both glucose starvation or addition, global changes of protein expression occur (Boy-Marcotte et al., 1996;Radonjic et al., 2005) and both catalytic and regulatory subunits of PKA itself change localisation (Tudisca et al., 2010). The completely different set of targets dictates that the role of PKA during glucose depletion is not simply the reversal of that during addition. This is one of the reasons why comparison of mutants may give ambiguous results, because here too the history is different, and therefore, the network present is not the same. The only way to study such densely feedbacked networks is through carefully times  During growth, glucose levels activate PKA quantitatively. When glucose is depleted from the media, pH c decreases according to the level of PKA activity. Orange represents high glucose concentrations, which, via induction of high PKA activity, trigger a low pH c upon glucose depletion and limit survival. Green represents low glucose, which leads to lower PKA activity and thus increases pH c upon glucose depletion and survival et al., 2012). By analysing the effects of PKA on pH c , we identified a new role for pH c in a nutritional transition. In the presence of glucose, pH c limits growth rate, while in its absence it limits survival, defined as the number of cells able to form colonies after 48 hr under optimal growth conditions. Therefore, upon glucose addition a pH c decrease contributes to the activation of PKA (Colombo et al., 1998;Thevelein & De Winde, 1999), during growth, acidification gradually reduces growth rate possibly through PKA (Dechant et al., 2010(Dechant et al., , 2014, whereas upon glucose depletion it is PKA activity that causes most of the pH c decrease. Our work establishes the mutual interaction between PKA and pH c as a major control node of nutritional transitions.

| Yeast strains and plasmids
Strains used in this work are listed in Supporting information  Primers used for strain generation and plasmid verification can be found in Supporting information Table S3.

| Growth curves
In growth curve experiments, yeast cultures were pre-grown overnight in Synthetic Complete medium in glass tubes until glucose depletion and diluted to OD 600 2 (~1:10 dilution) in low fluorescence medium in microtitre plates. Growth (OD 600 ) and pH c were then DOLZ-EDO ET AL.
| 9 of 12 monitored for 18-30 hr every 10 min. Fluorescence microscopy inspection was used to check that the pHluorin signal remained cytosolic at the end of the growth curves.
To better compare the pH c profiles for the set of PKA mutants analysed, we summarised the information of the pH c profiles by calculating the pH c during growth and the pH c after glucose depletion.
For such calculations, we took into account that the different strains presented different lag phases and/or growth rates and therefore depleted glucose at different times. We defined the pH c during growth as the average pH c during the hour after the population had undergone two OD 600 doublings. Cytosolic pH after glucose depletion was defined as the average pH c measured between 10 to 11 hr after the moment of glucose depletion, set as the time-specific growth rate (µ) decreased below 0.02 per hr. Growth rates were calculated as reported previously (Orij et al., 2012).

| Starvation experiments
For starvation experiments, yeast cultures were grown overnight to exponential phase (OD 600~5 -10,~0.25-0.5 of maximal OD 600 ) in Erlenmeyer flasks in low fluorescence medium. Pre-treatments were also performed in flasks to ensure proper aeration of the cultures.
Cultures were aliquoted in 1.5 ml tubes, washed twice with fresh low fluorescence medium without glucose, resuspended in this medium and transferred to microtitre plates. Cytosolic pH was then monitored for 1 hr every 5 min. Note that because of the washing and preparation time, pH c is not measured immediately after the starvation and the first minutes of the pH c drop are not determined.
In Figure 4b-g, cultures exponentially growing overnight on different starting concentrations of glucose were diluted in fresh medium with the same concentration of glucose and grown for at least one doubling before the experiment. In that way, we ensured that the cultures were not near glucose depletion at the moment of the experiment.

| Starvation at different external pH (pH ex ) and viability assay
To assess viability after glucose starvation, we followed the starvation protocol detailed above with the following variations. Starvation media were adjusted to various pH (pH ex ) by adding buffers except for pH ex 3, for which pH was set with HCl. Tartaric acid 25 mM was used to buffer at pH 4, 25 mM sodium citrate for pH 5, 50 mM MES for pH 6 and 100 mM MOPS for pH 7. The actual pH of the media after buffer addition was as follows: 4.4 for pH ex 4, 5.3 for pH ex 5, 6.3 for pH ex 6. For pH ex 7, two different buffer stocks at pH 7 and 6.8 were used, leading to small differences in viability among the experiments (Figure 5e vs. f). After the washes, the starved cultures were transferred to glass tubes. Tubes were subsequently incubated in a rotating wheel at 30°C. Samples were taken to monitor pH c and assess viability. To evaluate survival, samples were serially diluted and plated on YPD immediately after the transfer to starvation conditions and three days later. Plates were incubated at 30°C for about 48 hr, and the number of colony-forming units was determined. The number of colony-forming units immediately after the transfer to starvation conditions was considered as the reference point (100%) to determine the percentage of viable cells at day 3.

| Statistical analysis
Biological replicates were performed on different days. Statistical analysis was performed with GraphPad Prism 6 software. Unless otherwise indicated, significance between conditions was evaluated by using ANOVA with Bonferroni's multiple comparison correction of the p-values. Gaussian distribution of the data was assumed.
Biological variability was high in starvation experiments and affected the absolute pH c values but not the differences between pH c under different treatments. This suggests that additional factors yet to be identified affect absolute pH c under these conditions. To exclude these confounding factors from our statistical analysis, we performed paired comparisons in these set of data (Figures 3 and 4c-g).
In the figures, significance is indicated as follows: n.s., not significant; *p-value ≤ 0.05.
We also want to thank Dr. A. Ullah and R.L. McIntyre for their contribution to the preliminary work and to Dr. A. Zakrzewska for critical reading of the manuscript.

CONFLI CT OF INTEREST
None Declared.