A differential effect of {sigma}S on the expression of the PHO regulon genes of Escherichia coli

The RNA polymerase holoenzyme of Escherichia coli is formed by the subunits α, β, β' and ω that compose the core of the enzyme and a σ subunit. The σ subunit, or σ factor, is responsible for promoter recognition and transcription initiation, while the core enzyme executes transcription elongation. Seven different σ factors have been identified in E. coli. The two most important ones are σ^D (also known as σ⁷⁰) and σ^S (or σ³⁸). σ^D, when associated with the core enzyme (Eσ^D), initiates transcription of the majority of the E. coli genes while Eσ^S recognizes promoters and initiates transcription of genes associated with stationary phase survival and with the response to different stresses, such as osmolarity, pH and temperature shifts (Hengge-Aronis, 2000). More than 70 σ^S-dependent genes have so far been identified and more will probably be found in the future (Hengge-Aronis, 2002a).

Eσ^D and Eσ^S recognize similar promoter sequences (Wise et al., 1996; Espinosa-Urgel et al., 1996; Gaal et al., 2001; Lee & Gralla, 2001) and in vitro studies have shown that many genes are transcribed by both Eσ^D and Eσ^S (Nguyen et al., 1993; Tanaka et al., 1993; Kusano et al., 1996; Colland et al., 2000; Bordes et al., 2000), indicating that there is some overlap in promoter recognition by the two sigma factors. However, despite the similarities in promoter recognition in vitro, the two sigma factors are normally able to distinguish in vivo between σ^D- and σ^S-dependent promoters. No significant differences were found in the consensus sequence of the σ^D- and σ^S- dependent 35 elements (Becker & Hengge-Aronis, 2001; Lee & Gralla, 2001) except that the 35 region in σ^S promoters can be more degenerate than in σ^D promoters (Gaal et al., 2001), suggesting that Eσ^S interacts weakly or not at all with the 35 element. In vitro selection of an optimized σ^S promoter ended with identical consensus elements that agree with those of σ^D-dependent promoters, both in the 10 and 35 positions (Gaal et al., 2001). However, a compilation of 41 σ^S-dependent promoters has led to the consensus CTACACT at positions 13 to 7 (Lee & Gralla, 2001) and another compilation of 56 promoters reached the consensus TG(n)₀₂CYATACT (Lacour et al., 2003). These studies have revealed that over 80 % of the natural σ^S-controlled promoters possess a cytosine at the 13 position (Espinosa-Urgel et al., 1996; Becker & Hengge-Aronis, 2001).

The PHO regulon of E. coli consists of more than 40 genes and operons whose transcription is induced under conditions of inorganic phosphate (Pi) starvation and that are related to the uptake and assimilation of Pi and phosphorylated compounds. The best characterized ones are phoA, phoE, the pst operon and the ugp operon, which encode, respectively, alkaline phosphatase (AP), the anion porin PhoE, the Pi transporter Pst and the glycerol-3-phosphate transporter Ugp. Apart from its role as a Pi-transporter, the Pst system also functions as a negative regulator of the PHO regulon, because most mutations in the pst operon lead to the constitutive synthesis of all PHO genes (Wanner, 1996). The promoters of the PHO genes display one or more consensus regulatory sequences known as PHO-boxes that replace the 35 element. Transcription is regulated by a two-component system that is composed of the proteins PhoB and PhoR. When the concentration of Pi in the medium decreases below a certain level, the sensor protein PhoR auto-phosphorylates and transfers the Pi group to the regulatory protein PhoB, which in turn binds to the PHO-boxes and allows transcription of the PHO genes by interacting with Eσ^D (Wanner, 1996; Makino et al., 1996).

In preliminary experiments we have noticed that in rpoS mutants the expression of AP was considerably stronger than in the wild-type strain, implying that σ^S is involved in the regulation of AP. Here we demonstrate that σ^S negatively affects the expression of phoA, phoB, phoE and ugpB, but not pstS. The competition between σ^S and σ^D for the core RNA polymerase is proposed to explain this differential effect of σ^S on the expression of the PHO genes.

Strains and plasmids.
These are listed in Table 1.

Table 1. Strains and plasmids used in this study

Growth media and growth conditions.
The rich medium was LB (Miller, 1992). Medium A is a semi-rich medium that is low in Pi (Levinthal et al., 1962). T-salts medium is a Tris-buffered minimal medium supplemented with 0·4 % glucose (Echols et al., 1961) that contains either 1 mM KH₂PO₄ in the high-Pi minimal medium or 0·1 mM KH₂PO₄ in the low-Pi minimal medium. For the assay of the kinetics of AP induction, cells were grown in a high-Pi minimal medium until they reached an OD₅₄₀ of 0·20·3. They were then washed and resuspended in a minimal low-Pi medium. Samples were taken at 30 min intervals for AP assays. For RNA extraction, a 20 ml sample was taken from bacteria growing in minimal high-Pi medium and a second sample was taken from a culture grown for 2 h in minimal low-Pi medium. For the assays of AP and chloramphenicol acetyl transferase (CAT), cells were grown overnight in medium A, and in medium A supplemented with 1 mM KH₂PO₄.

PCR amplifications.
The rpoS fragment was amplified using genomic DNA extracted from strain MG1655 as template and the oligonucleotides rpoS+ (ATACTGCAGGCAGCAAAGGACAGG) and rpoS (CGTCGCGGCTGAAGCTTACAACAC). Bold letters indicate restriction sites. The DNA fragments used as probes for phoA, phoB, phoE, pstS and ugpB were amplified as above using the oligonucleotides phoA+ (CAGCATTCCTGCAGACGATAC) and phoA (GATCAAGCTTAATGTATTTGTACATGGAGAA), phoB+ (TCAAACACCTCAAGCGCGAG) and phoB (GCTCCAGTGCTTTACGCA), phoE+ (ACCTGGGGGCGTTGTATGAC) and phoE (TTGGTGCGATCTGAGTTGGTAT), pstS+ (CTTCCCTGCGCCGTGTATGC) and pstS (TCAGCGGAGATCAGTTTGGTGTT) and ugpB+ (GACGCGGTGCTGGAGTTCAATA) and ugpB (CCGCCCCTGGGTTTTTCTCATA), respectively.

Plasmid construction.
Plasmids pNP1 and pNP5 were constructed by digesting the rpoS PCR fragment with PstI and HindIII followed by ligation to the same sites of plasmids pKK223-3 and pACT3, respectively. Plasmid pBS11 was constructed by digesting a pst PCR fragment with DraI and BstYI followed by ligation to pKK232-8 digested with SmaI and BamHI.

Enzyme assays.
AP was assayed using p-nitrophenyl-phosphate (p-NPP) as substrate as described by Spira et al. (1995). AP-specific activity is represented by the increase in absorbance at 410 nm min¹ (cell density)¹. Catalase activity was measured qualitatively by mixing 50 µl cells (OD₅₄₀=3·0) with 50 µl 3 % hydrogen peroxide and observing the appearance of bubbles caused by the release of O₂. CAT assays were performed essentially as described by Shaw (1975). Cells were disrupted by sonication and protein concentration was determined by the method of Bradford (1976). The substrate was 5,5'-dithio-bis(2-nitrobenzoic acid) (DTNB) together with acetyl-CoA and chloramphenicol in a total volume of 500 µl. The reaction was started by adding chloramphenicol at a final concentration of 0·1 mM to a cuvette containing 0·4 mg DTNB, 0·5 mM acetyl-CoA and cell extract. The absorbance increase rate at 412 nm was recorded. CAT activity was calculated as nmoles min¹ (mg protein)¹.

RNA extraction and Northern-blot analysis.
RNA was extracted by the guanidine thiocyanate method, as described by Chomczynski & Sacchi (1987). RNA (20 µg) was electrophoresed in a 1 % agarose gel containing 7 % formaldehyde for 3 h. The RNA was transferred to a nylon membrane by capillary action. Probes for phoA, phoB, phoE, pstS and ugpB were synthesized with [α-³²P]dCTP by random primer labelling using the DNA fragments obtained by PCR, as described above. For synthesis of the rpoD probe, a 1·5 kb fragment digested from plasmid pRPOD with BamHI and SacI was used. The labelled probes were hybridized with the membranes at 42 °C for 1620 h and the membranes were exposed to X-ray films.

Site-directed mutagenesis and DNA sequencing.
Site-directed mutagenesis was performed by the circular mutagenesis method, using double stranded DNA templates and selection with DpnI, as described by Sambrook & Russell (2001). Plasmid pBS11, carrying a pstScat fusion, was used as a template for the PCR reaction. The oligonucleotides pstSmut+ (CTGTCACCTGTTTGTCTTATTTTGCTTCTCGTAGCCAACAAAC) and pstSmut (GTTTGTTGGCTAGGAGCAAAATAAGACAAACAGGTGACAG) contain the desired mutation (underlined). The product of the amplification was treated with DpnI and transformed into strains MG1655 and BS16. Both wild-type and mutated plasmids were sequenced in an automatic sequencer type ABI Prism 3100 Genetic Analyser (Applied Biosystems/Hitachi) to confirm the presence of the point mutation.

Effect of rpoS inactivation on AP synthesis
Overnight-grown cultures of the wild-type strain (MG1655) and its isogenic rpoS : : Tn10 mutant (BS16) grown in excess or limited Pi media were assayed for AP activity. As expected, cells grown in limited Pi expressed high levels of AP (Fig. 1) while cells grown in excess Pi media showed only a basal level of AP activity that did not exceed a specific activity of 0·015 (not shown). The rpoS mutation caused a threefold increase in AP activity (bar b) when compared to the wild-type strain (bar a), indicating that σ^S negatively affected the expression of AP. When a multicopy plasmid that carries the wild-type rpoS⁺ gene under the control of the tac promoter (plasmid pNP1) was introduced into the wild-type strain (bar c) and into the rpoS mutant (bar d), the level of AP activity dropped to approximately half the level of the wild-type parent, suggesting that an excess of σ^S inhibited AP expression. Introducing plasmid pMRG7, which overexpresses rpoD⁺, into the wild-type (bar e) and the rpoS mutant (bar f) increased AP activity by 3·7- and 1·5-fold above the level of their untransformed parents, respectively. The elevated expression of AP in the presence of the multicopy rpoD⁺ plasmid supports a previous observation that phoA transcription is driven by Eσ^D (Makino et al., 1993).

(8K): Fig. 1. Effect of rpoS on AP activity. Cells were grown overnight in medium A and assayed for AP activity. Bars: a, wild-type (strain MG1655); b, rpoS : : Tn10 (strain BS16); c, wild-type transformed with prpoS+ (plasmid pNP1); d, rpoS : : Tn10 transformed with pNP1; e, wild-type transformed with prpoD+ (plasmid pMRG7); f, rpoS : : Tn10 transformed with pMRG7. Bars represent the means±SE of at least four independent experiments.

The concentration of σ^S is known to increase progressively in cells that enter the stationary growth phase and in cells that undergo carbon or Pi-starvation (Hengge-Aronis, 1993; Gentry et al., 1993; Ruiz & Silhavy, 2003). To test at what stage of the Pi-starvation process σ^S affects the expression of AP, exponentially growing cultures of the wild-type strain, rpoS : : Tn10 mutant and its transformant carrying plasmid pNP5 (rpoS⁺ under the control of Ptac and the lacI^q allele that overproduces the Ptac repressor LacI) were suspended in a low-Pi minimal medium and monitored for AP activity for several hours (Fig. 2). All of them entered the Pi-starvation phase between 30 and 60 min as seen by the induction of AP and the subsequent deceleration of the growth rate (insert). The induced enzyme activity of the wild-type strain reached its maximal level at 90 min (Fig. 2, ), while the activity of the rpoS mutant kept rising and reached a threefold increase over the wild-type at 210 min (Fig. 2, ). The data suggest that at the early starvation phase (as of 90 min) the amount of σ^S in the cell was already sufficient to prevent further induction of AP. The pNP5 transformant showed a similar pattern of AP induction even in the absence of the inducer (IPTG), suggesting that the tac promoter was sufficiently leaky to suppress the effect of the rpoS mutation (Fig. 2, ). However, the presence of 100 µM IPTG caused a further inhibition of AP activity from the very beginning of the starvation phase (Fig. 2, •). These results indicate that the negative effect of σ^S on AP synthesis has already begun at the early Pi-starvation phase and that it is stronger when σ^S is overexpressed.

(17K): Fig. 2. Kinetics of the effect of rpoS on AP activity during Pi starvation. Exponential-phase cells were suspended in a low-Pi minimal medium and monitored for growth (insert) and AP activity. Wild-type strain (); rpoS : : Tn10 (); rpoS+ plasmid pNP5 in rpoS : : Tn10 without IPTG () and with 100 µM IPTG (•).

rpoS inhibits AP expression in non-starved cells
To test if σ^S is able to inhibit AP expression of non-starved cells, the rpoS : : Tn10 mutation was introduced into a strain that carries a deletion of the entire pst operon (strain NP34). This is a constitutive mutant that produces AP independently of the external Pi concentration. Samples were withdrawn every hour from cultures of this mutant and of its rpoS⁺ parent grown in LB medium (a medium that contains excess Pi). The samples were assayed for growth rate, for AP activity and for catalase activity. Synthesis of catalase (encoded by katE) is strongly dependent on the presence of σ^S (Schellhorn et al., 1998). The insert in Fig. 3(a) shows that both strains grew exponentially for the first 120 min and entered the stationary phase thereafter. Fig. 3(a) shows that during the first 120 min of exponential growth, both strains presented a similar level of constitutive AP activity. Upon entry into the stationary phase, cells growth was drastically reduced and in the rpoS⁺ strain the enhanced formation of Eσ^S polymerase led to the expression of genes related to cell survival, such as catalase, and caused the arrest of AP synthesis. Due to its strong stability (Torriani, 1960) the activity of AP remained constant. The slight and gradual decline of its specific activity [EU (cell density)¹] is due to the gradual increase in the optical density of the stationary cells (see insert). In contrast, in the rpoS cells, no σ^S is formed to compete with σ^D, and many of the Eσ^D-dependent housekeeping genes cease to transcribe, providing excess Eσ^D available for the increased expression of AP that is evident in these cells. The appearance of catalase activity at the onset of the stationary phase (Fig. 3b) testifies to the induction of σ^S. These results demonstrate that σ^S down-regulates AP expression also in the presence of excess Pi and therefore this inhibition is not related to the mechanism of PHO induction by Pi-starvation.

(50K): Fig. 3. Effect of rpoS on AP activity of PHO-constitutive non-starved cells. Exponential growing cells BS7 (Δpst; ) and NP34 (Δpst rpoS : : Tn10; ) were suspended in LB medium and monitored for growth (insert), for AP activity (a) and for catalase activity, which is observed by the bubbled spots (b).

rpoS inhibits the transcription of phoA, phoB, phoE and ugp but not that of pst mRNA
To test whether the effect of σ^S on AP is at the transcriptional level and if the expression of other genes that belong to the PHO regulon are also affected by σ^S, Northern blot analyses were conducted. DNA probes that are specific for the genes phoA, phoB, phoE, pstS and ugpB were hybridized with RNA extracted from Pi-starved and from non-starved wild-type cells, rpoS : : Tn10 mutants and rpoS : : Tn10 mutants transformed with a prpoS⁺ plasmid (pNP1). Fig. 4 shows that, as expected, all probes strongly hybridized with mRNA extracted from Pi-starved cells (lanes 4, 5 and 6), while no signal was detected from hybridization of the probes with mRNA extracted from cells grown in excess Pi (lanes 1, 2 and 3). The signals corresponding to phoA, phoB, phoE and ugpB were significantly stronger in the rpoS mutant (lane 5 as compared to the wild-type lane 4) and were reduced by the overexpression of rpoS (lane 6), indicating that the negative effect on σ^S is at the transcriptional level. In contrast, the signal corresponding to pstS was moderately weaker in the rpoS mutant than in the wild-type and slightly stronger in the presence of the multicopy plasmid that expresses σ^S. These results suggest that σ^S inhibits the transcription of phoA, phoB, phoE and ugp and moderately stimulates the transcription of pstS.

(53K): Fig. 4. Effect of rpoS on the mRNA level of phoA, phoB, phoE, pstS and ugpB. Exponential-phase wild-type cells (strain MG1655), rpoS : : Tn10 mutant (strain BS16) and the rpoS mutant transformed with plasmid pNP1 were resuspended in medium A supplemented or not with Pi and grown for 2 h. RNA was extracted and probed with labelled phoA, phoB, phoE, pstS and ugpB DNA fragments as described in Methods. To confirm that equal amounts of RNA were applied to the gels, the membranes were stripped and rehybridized with a labelled rpoD probe (not shown). Lanes 1 and 4, wild-type; 2 and 5, rpoS : : Tn10; 3 and 6, plasmid pNP1 (prpoS+) in rpoS : : Tn10. Lanes 13, cells grown in excess Pi; 46, cells grown in limited Pi.

A cytosine residue at position 13 is important for σ^S recognition of the pst promoter
The inconsistency between pstS and the other PHO genes with regard to their response to σ^S may reflect sequence differences in their promoters. PHO promoters are devoid of a 35 sequence, carrying one or more PHO-boxes instead. The 10 regions of all known PHO promoters are depicted in Fig. 5. phoA, phoE, phnC, psiE and ugpB all carry a thymine residue at position 13, phoH has an adenine and phoB a guanine residue, while pstS carries a cytosine at that position. Previous reports have suggested that promoters carrying a C residue at position 13 are preferred by Eσ^S (Espinosa-Urgel et al., 1996; Bordes et al., 2000; Gaal et al., 2001; Becker & Hengge-Aronis, 2001; Lee & Gralla, 2001). Therefore, the presence of the 13C residue might confer on the pst operon the ability to be transcribed also by Eσ^S explaining why, in contrast to the other PHO genes, the expression of pstS is reduced rather than induced by the rpoS mutation.

(25K): Fig. 5. DNA sequences of PHO promoters. The PHO-boxes are framed and the 10 sequences are underlined. The 13 position is in bold.

To test this hypothesis, the promoter of pstS, the first gene of the pst operon and which governs the transcription of the entire operon (Aguena et al., 2002), was cloned in plasmid pKK232-8, creating an operon fusion between the pstS promoter and the reporter gene that encodes CAT (cat, plasmid pBS11). The 13 cytosine residue of the pst promoter in this fusion was replaced by a thymine by site-directed mutagenesis (plasmid pNP6). Both plasmids were transformed into the wild-type strain (MG1655) and into its isogenic rpoS : : Tn10 mutant. The transformants were grown overnight under limited Pi concentrations and in excess Pi and were assayed for AP and CAT activity. Fig. 6 depicts the ratios of enzyme activities between the rpoS mutant and the wild-type strain transformants, grown under conditions of Pi-starvation. As before, AP activity was twofold higher in the rpoS mutant than in the wild-type (bar a). CAT activity that represents the wild-type pstS promoter (13C) was 25 % lower in the rpoS mutant when compared to its rpoS⁺ parent (bar c), confirming the lower pstS mRNA level in the absence of σ^S (Fig. 4). In contrast, the modified promoter where the 13C residue was replaced by a T (13T) showed more than a twofold increase in CAT activity in the rpoS mutant (bar d) that was similar to the stimulation of AP activity (bar b). Thus, the C→T transition at the 13 position of the pst promoter caused it to act like the promoters of phoA, phoE and ugp. Under excess Pi the cells showed only low basal levels of both enzymes, indicating that transcription of the mutant pstS is also PhoB-dependent (not shown). These results suggest that the presence of a 13C residue in the pst promoter is important for σ^S recognition of this promoter and below we present an evolutionary rationale to this observation.

(9K): Fig. 6. Effect of rpoS on AP and CAT activities in strains carrying a plasmid with a 13T mutation in the promoter region of pstS. Cells were grown overnight in medium A and assayed for AP and CAT. Bars represent the enzyme levels of the rpoS : : Tn10 mutant (strain BS16) divided by those of the wild-type strain (strain MG1655). Bars a and c, strains carrying the wild-type pst promoter (plasmid pBS11, 13C); b and d, strains carrying the mutated (13T) pst promoter (pNP6). Each bar represents the mean±SE of at least five independent experiments.

Phosphorus limitation is very common in natural environments (Barik et al., 2001; Sundareshwar et al., 2003); hence the PHO regulon of bacteria growing in their natural habitats should be constantly expressed. The activation of the general stress response controlled by σ^S is also frequent in nature (Hengge-Aronis, 2002b). At first sight, the molecular mechanisms used by bacteria to protect themselves against both types of stress should be compatible, if not synergistic. However, during its evolution, E. coli developed two separate and apparently antagonistic mechanisms to cope with each type of stress, the σ^S regulon and the PHO regulon. When cells enter the Pi-starvation phase, the PHO regulon is activated and σ^S starts to accumulate in the cytosol (Gentry et al., 1993; Wanner, 1996; Ruiz & Silhavy, 2003). The simultaneity of these events creates a situation where, on one hand, the PHO genes, whose expression is σ^D-dependent, demand their own transcription and, on the other hand, σ^S accumulates and competes with σ^D for the core RNA polymerase. Thus, in this case there is no apparent benefit from the competition between the sigma factors and therefore it is paradoxical that σ^S, which controls the general cell response to stress, inhibits the expression of genes related to phosphate starvation.

In E. coli, σ^S and the alarmone guanosine tetraphosphate (ppGpp) are the key factors that promote the transition from growth proliferation to stasis, where proteins related to protection against the deleterious effects of oxidation are expressed. This led to the suggestion that there is a trade-off between bacterial survival and proliferation, such that the expression of genes encoding proteins involved in cell growth is inhibited by factors that promote survival and vice-versa (Nyström, 2003). The PHO regulon genes, whose function is to acquire and assimilate Pi in order to restore cell growth, are inhibited by σ^S whose main concern is with the expression of genes related to cell survival. The competition between σ^S and σ^D for the core RNA polymerase inhibits the σ^D-transcribed PHO genes either directly or indirectly via the σ^S-promoted inhibition of the positive regulator PhoB.

Unlike all other PHO genes tested, pstS was somewhat stimulated by the induction of σ^S. In addition to its role in Pi transport the pst operon also serves as a negative regulator of the PHO genes (Wanner, 1996). Moreover, its promoter carries a feature shared by many σ^S-promoters (Hengge-Aronis, 2000), namely, the presence of a functional IHF binding site that helps elevate its expression and thereby reduce the expression of AP (Spira & Yagil, 1999). The pst promoter is the only one of eight known PHO promoters that possesses a cytosine residue at the 13 position (Fig. 5). Our results suggest that pst may be transcribed in vivo by both Eσ^D and Eσ^S, and that the other PHO genes are transcribed only by Eσ^D. Is there a teleological reason for the differential behaviour of pstS in relation to the other PHO genes? Being a negative regulator of PHO, an increase in Pst expression would reduce the transcription of the other PHO genes that are driven by σ^D, thereby providing more RNA polymerase core enzyme to interact with σ^S. As a result, σ^S-dependent genes that are important to bacterial survival during stress could be more readily transcribed. In such a trade-off way, the controlled repression of the PHO genes by Pst might be beneficial for cell survival during prolonged Pi starvation periods. The negative effect of rpoS on gene expression as a result of σ^S competition against σ^D was already reported for other σ^D-transcribed systems. These include the glucose transport-related genes mal and mgl (Notley-McRobb et al., 2002), the type 1 fimbrial genes fimA and fimB (Dove et al., 1997), ompF (Pratt et al., 1996), the stress-induced gene uspA (Farewell et al., 1998) and several other genes that were found to be hyperexpressed in the absence of a functional σ^S (Xu & Johnson, 1995; Farewell et al., 1998).

Ruiz & Silhavy (2003) have recently shown that in a pstS mutant that causes PHO constitutivity σ^S is already strongly expressed in the exponential growth phase. The results shown in Fig. 3(b), where σ^S-dependent catalase activity was induced only upon entry into the stationary phase, suggest that even if σ^S is expressed at high levels in the exponential phase in PHO-constitutive mutants, it is not able to induce the synthesis of σ^S-dependent promoters. Also, there was no significant difference in the level of AP between the wild-type and the rpoS mutant during the exponential phase (Fig. 3a). Kvint et al. (2000) have demonstrated that σ^S-dependent promoters require ppGpp for induction in the stationary phase, but PHO-constitutive mutants present a low level of ppGpp in the exponential phase of growth (Spira et al., 1995). Thus, if σ^S is present in the exponential phase, it is probably inactive.

In conclusion, we have shown that σ^S negatively affects the expression of several PHO genes, but not that of the pst operon. We suggest that this effect is due to a competition between σ^S and σ^D for the core RNA polymerase. Since the PHO genes are transcribed by Eσ^D, accumulation of σ^S in the cytosol during the starvation phase reduces their transcription. In contrast, pst, which is also a negative regulator of PHO, may be transcribed by both σ^S and σ^D. Through this mechanism the PHO regulon has evolved to maintain a trade-off balance between cell nutrition and cell survival during severe Pi-starvation stress.

This work was supported by the Fundação de Amparo a Pesquisa do Estado de São Paulo (FAPESP). E. Y. was supported by the Kurt Lion Foundation. We thank Barry Wanner, Mike Cashel, Regine Hengge-Aronis and Kozo Makino for supplying strains.