Expression of Escherichia coli glutaredoxin 2 is mainly regulated by ppGpp and sigmaS.

Escherichia coli glutaredoxin 2 (Grx2, encoded by grxB) differs greatly from the other two glutaredoxins in structure and catalytic properties. In a wild type strain, levels of Grx2 increased 3-fold in the stationary phase (up to 8 microg/mg). Guanosine-3',5'-tetraphoshate (ppGpp) and sigma(S), which regulate the transcription of genes in the stationary phase, dramatically affected the expression of Grx2. spoTrelA null mutants, lacking ppGpp, had very low levels of Grx2, while overproduction of full-length RelA or valine-induced starvation of isoleucine, both conditions elevating ppGpp levels, resulted in elevation of Grx2. Null mutants for the sigma(S)-specific protease ClpP, which have higher levels of sigma(S), exhibited a 3-fold Grx2 increase. sigma(S) in trans also increased the levels of Grx2. Therefore the stationary phase expression of Grx2 is determined by the sigma(S)-bound form of RNA polymerase in connection with ppGpp, while basal levels should be attributed to sigma(70)-RNA polymerase holoenzyme. Osmotic pressure and cAMP also affected the expression of Grx2, presumably via sigma(S). Furthermore, Grx2 levels were elevated in an oxyR(-) strain. In accordance with the role of Grx2 as a stationary phase protein, null mutants for grxB were shown to lyse under starvation conditions and exhibited a distorted morphology.

In Escherichia coli (E. coli), cytosolic disulfides are kept reduced by the glutaredoxin and thioredoxin systems. In the glutaredoxin system, electrons are transferred from NADPH, to glutathione reductase, glutathione (GSH), and finally to the three glutaredoxins (Grx1, Grx2, and Grx3, encoded by grxA, grxB, and grxC, respectively). In the thioredoxin system the flow of electrons is from NADPH to thioredoxin reductase and finally thioredoxins 1 and 2 (Trx1 and Trx2 encoded by trxA and trxC, respectively). Glutaredoxins catalyze GSH-disulfide oxidoreductions usually via two redox-active cysteine residues separated by two other amino acids (typically CPYC) (1,2). The oxidoreductions are either dithiol reactions to reduce protein disulfides or monothiol reductions of mixed disulfides with glutathione. In comparison, thioredoxins may predominantly reduce protein disulfides. Grx1 and Grx3 are ϳ10-kDa proteins with homologues in most living organisms and have the characteristic thioredoxin/glutaredoxin fold (3). Grx1 can reduce the intracellular disulfides of ribonucleotide reductase (RR) 1 and PAPS reductase, while Grx3 has 5% of the catalytic activity of Grx1 for RR1a (4,5). The larger Grx2 (24.3 kDa) with one N-terminal glutaredoxin domain and a highly helical C-terminal domain (glutathione S-transferases-like structure) (6) cannot reduce RR but has the highest catalytic activity using the mixed disulfide between glutathione and ␤-hydroxyethyl disulfide as substrate (HED assay) (5). E. coli Grx2 has close homologues in Actinobacillus actinomycetemcomitans (87% amino acid identity), Neisseria meningitidis (58%), and Vibrio cholerae (42%), all known pathogens, but it is not a "ubiquitous" protein. All three E. coli glutaredoxins are good electron donors for the reduction of arsenate by arsenate reductase, with Grx2 having a 100-fold higher catalytic efficiency (7). To identify more glutaredoxin functions, null mutants for Grx2 (grxB) were constructed. GrxB Ϫ strains were viable, but with 80% lower levels of general GSH-disulfide oxidoreductase activity and an increased sensitivity to oxidative stress (8).
The glutaredoxin and thioredoxin systems are involved in cellular defense against oxidative stress (9 -11). The transcriptional regulator OxyR regulates major responses to oxidative stress and positively affects transcription of gor (glutathione reductase), katE, katG (catalases), grxA, and trxC (12)(13)(14). Grx1 and Trx1 are both able to reduce, thus deactivate OxyR in vitro, with Grx1 being the preferred reductant in vivo (15,16). Expression of catalases and glutaredoxins is interconnected. All glutaredoxins are highly elevated in catalase deficient strains (katEkatG) (17), whereas catalase HPI is elevated in a null mutant for trxAgrxAgrxBgrxC (8). In the katEkatG null mutants, Grx2 protein levels were determined to be ϳ3-fold higher than in wild type cells, with levels reaching 2% (20 g/mg) of total cell protein in the stationary phase (8,17). Guanosine-3Ј,5Ј-tetraphoshate (ppGpp) and S (RpoS) are two major factors controlling the transcription of genes in stationary phase of growth and are also known to control the transcription of genes involved in the antioxidant response (18,19). E. coli glutathione reductase is positively regulated by S in the stationary phase (14). Trx1 transcription is positively regulated by ppGpp, but for the trxA expression S is not required (20). The levels of Grx2 have been previously shown to be growth phase dependent and highly elevated at the stationary phase of growth (17). This made us interested in further investigating whether ppGpp and S regulate the expression of Grx2.

MATERIALS AND METHODS
Chemicals-The sodium salt of cAMP was purchased from Sigma. cAMP solutions were prepared fresh prior to each experiment. All other chemicals were purchased from known commercial sources.
Strains and Plasmids-The bacterial strains and plasmids used in this work are listed in Table I. Unless otherwise stated, E. coli strains are AF1000 (MC4100 relAϩ) derivatives. AF1000 is therefore considered as the wild type. Transductions were done with phage P1 (21).
Growth Experiments and Media-Cells were grown LB liquid medium (22) or LB medium solidified with agar from Merck (15 g/liter) and supplemented (whenever needed) with ampicillin (100 g/ml), kanamycin (50 g/ml), tetracycline (20 g/ml), or chloramphenicol (20 g/ml). For the shake flask experiments, cells were grown in LB medium overnight at 37°C in 10-ml cultures in 100-ml Erlenmeyer flasks. Overnight cultures were centrifuged, resuspended, diluted to 1:200, and grown in 200 ml of glucose ammonia-based mineral salt medium in 1L Erlenmeyer flasks (23).
Overexpression of RelA and S -Transformed cells with plasmids encoding S (pRL40.1) or RelA (pALS10, pALS13 and pALS14) were grown in LB with 100 g/ml ampicillin. Cells were induced with 150 M IPTG for the expression of RelA and 1 mM IPTG for the expression of S . Samples were taken just before induction and 30, 60, and 90 min after induction. RelA expression was induced at a relatively low A 600 (ϳ0.2), while expression of rpoS was induced at late exponential phase, A 600 0.7-0.8.
Treatment of Cells with Hydrogen Peroxide-Cells were grown to late exponential phase, A 600 0.7-0.8 in LB medium. Hydrogen peroxide (5 mM) was then added, and the cells were then further cultivated for 1 h before harvesting.
Valine-induced Isoleucine Starvation-Cells were grown in M9 medium (22) containing BME (basal medium Eagle) vitamin solution (Invitrogen), leucine (50 g/ml), isoleucine (50 g/ml), and glucose (2 g/liter). At A 600 of about 0.2, valine (1 mM final concentration) was added to the cells. Samples were taken at different time points, the bacteria harvested, kept at Ϫ20°C, and analyzed by ELISA after preparation of cell free extracts.
Preparation of Cell-free Extracts-Cells were harvested, resuspended in 50 mM Tris-HCl, pH 8.0, 1 mM EDTA, disrupted at 4°C by sonic disintegration. Phenylmethylsulfonyl fluoride (1 mM) was added, and lysates were centrifuged at 13,000 ϫ g for 30 min at 4°C. The supernatants of the lysates after centrifugation were used as the material for sandwich ELISAs.
Protein Concentration Measurements-Total protein was measured by the method of Bradford (24). Pure protein was determined by measuring A 280 . Antibody concentration was calculated using the relation: antibody (mg/ml) ϭ (A 280 -A 310 )/1.4.
Purification of Antibodies and Enzyme Immunoassay (ELISA)-Purification of antibodies against Grx2 and other redoxins by affinity chromatography and quantification of Grx2 by sandwich enzyme immunoassay were carried out as described previously (17).
Viability Test-The numbers of live and dead cells were determined with a viability kit (LIVE/DEAD BactLight) from Molecular Probes and analyzed using a fluorescence microscope.
Colony and Cell Count-Growth of the cultures was followed by measuring the optical cell density at 500 nm (A 500 ) (relation 1 A 500 ϭ 0.6 A 600 ), by microscopy with a cell-counting chamber (0.02 m depth), and by analysis of the number of colony forming units on LB plates. The stability of the gene marker (grxB::kan) was controlled by replica plating on LB and kanamycin plates.

Levels of Grx2 Are Directly Up-regulated by ppGpp-ppGpp
is the positive regulator of the stringent response and positively affects the general starvation response (18,25) occurring in stationary phase. To examine whether Grx2 expression is affected by ppGpp, we measured levels of Grx2 in strains with altered ppGpp content. In a spoTrelA double null mutant, levels of Grx2 at the exponential phase were less than one-third compared with wild type, and no increase was seen at the stationary phase of growth ( Fig. 1). Since SpoT and RelA are known to be the only cellular gene products accounting for the synthesis of ppGpp, and spoTrelA null mutants are therefore devoid of ppGpp, these results suggest that grxB transcription is regulated by ppGpp. To examine this possibility further, we measured the levels of Grx2 in strains containing plasmids encoding RelA (Fig. 2). A truncated form of RelA (1-133 amino acids), known not to elevate ppGpp levels, had no effect on the expression of Grx2. In contrast, the full-length form (733 amino acids) resulted in a clear elevation of Grx2. A truncated form of RelA-(1-455), which is known to positively regulate ppGpp (26), also increased the levels of Grx2 (Fig. 2). Induction of RelA in a spoTrelAϪ strain was followed by a modest increase in Grx2 expression, probably due to the low levels of ppGpp, even after induction (data not shown). To imitate isoleucine starvation leading to RelA-dependent accumulation of ppGpp, valine was added to the culture medium of growing cells. The addition of valine led to up-regulation of Grx2 levels (Fig. 3). These data suggest that expression of Grx2 is positively regulated by ppGpp levels.
Levels of Grx2 Are Controlled by S -Apart from ppGpp, the rpoS-encoded sigma factor S (RpoS or S ) regulates the transcription of different genes in the stationary phase (19). ppGpp up-regulates RpoS levels (27), and S -dependent promoters require ppGpp for their induction (28). To examine whether S affects the expression of Grx2, we measured Grx2 levels in null mutants for genes affecting total S content. Cells lacking S -specific ClpP protease (clpP null mutants) have higher S levels (29). In this genetic background Grx2 levels were significantly increased during all growth phases (Fig. 1), suggesting that S may up-regulate the levels of Grx2. The effect was more pronounced in the stationary phase where the Grx2 levels were twice as high as in the wild type. Thus the growth phase-dependent regulation is maintained in the clpP mutant. To ex- clude that the observed effect was due to Grx2 itself being a substrate for ClpP protease, levels of Grx2 were measured in an rpoSclpP null mutant (Fig. 1). Levels of Grx2 were the same at the exponential phase of growth but decreased at the stationary phase, suggesting that Grx2 is not a substrate for ClpP protease, but its expression is rather dependent on S . The high levels of Grx2 at the exponential phase of the rpoSclpP and rpoS null mutants suggest that transcription of Grx2 is also regulated by 70 (Fig. 1). Even under 70 transcriptional control, levels of Grx2 were much lower in the relA spoT mutant, presumably because of the very low ppGpp levels in that strain.
To further investigate whether transcription of grxB is controlled by S , we measured Grx2 levels in strains transformed with a plasmid encoding rpoS under an inducible promoter. In a rpoS Ϫ strain, induction of S led to very modest increases in Grx2 levels, presumably due to the already high base levels of Grx2 (data not shown). Clear up-regulation of Grx2 upon induction of S was observed in the wild type strain (Fig. 4). Levels of Grx2 remained very low when S expression was induced in a spoTrelAϪ strain (Fig. 4), presumably due to the low levels of ppGpp in the particular strain. S -dependent promoters require ppGpp for induction of their regulated genes (28). In conclusion, ppGpp, S , and 70 can regulate the expression of Grx2. ppGpp is the prerequisite for both 70 and S , which themselves can control the expression of Grx2 in the exponential and stationary phase, respectively. Regulation of Trx1-Trx1 levels were lower in the spoTrelAϪ strain, but showed no increase in the clpP null mutant (Fig.  5A). These results are consistent with Trx1 being regulated by ppGpp, but not by S (20). However, the ppGpp regulation of Trx1 differed from that for Grx2. Levels of Trx1 in the null mutant for spoTrelA were very low only at the stationary phase of growth. In comparison, levels of Grx2 remain very low at all growth stages (Fig. 1). This indicates that ppGpp regulates the expression of Trx1 only at the stationary phase. Levels of Trx1 were increased after induction of RelA in a wild type strain (Fig. 5B), but after induction of S , no effect was seen (data not shown). These data confirm previous findings on the regulation of Trx1 in stationary phase by ppGpp (20).
Expression of Grx1 showed a slight up-regulation in the spoTrelAϪ strain. No effect was seen for Grx3 and Trx2 in the strains examined for Grx2 levels (data not shown). It is therefore unlikely that S or ppGpp regulates Grx3 and Trx2.
cAMP Down-regulates Levels of Grx2 at the Exponential Phase-cAMP, as part of the cAMP-cAMP catabolite repressor protein complex (cAMP⅐CRP), down-regulates the expression of S (30). Therefore, null mutants for cya (encoding adenylate cyclase) have higher levels of S than the wild type even at the exponential phase of growth. Grx2 levels were also slightly up-regulated at the exponential phase in strains lacking cAMP (cya Ϫ strain, Fig. 1). Addition of cAMP in a null mutant for the cya gene resulted in a down-regulation of Grx2 at the exponen- tial phase, while no significant effect was observed when cAMP was added at the stationary phase of growth in the particular strain (Fig. 6). Addition of cAMP to cya Ϫ strains in the exponential phase of growth is known to lead to degradation of S (31). The lack of CRP consensus binding sites at the promoter of grxB together with the previous findings on the regulation of Grx2 expression by S suggest that the effects of cAMP⅐CRP on the levels of Grx2 are most likely due to changes of the S levels.
Grx2 Levels Increase in Hyperosmotic Environments-A clear up-regulation of Grx2 levels occurred after treatment with NaCl (Fig. 7). As stability and translation rates of S are increased in hyperosmotic conditions (32), it is likely that the increases in Grx2 levels were a consequence of the elevated S .
OxyR Does Not Regulate the Expression of Grx2 and Grx3-OxyR regulates the transcription of a large number of genes in response to oxidative stress (33). Levels of Grx2 and Grx3 showed no significant change before and after treatment with hydrogen peroxide in both the wild type and the oxyR null mutant (Fig. 8). However, the steady-state levels of Grx2 were significantly elevated in the oxyR null mutant (Fig. 8B). These data suggest that transcription of grxB is independent of OxyR but dependent on an alternative backup system. Levels of Grx1, known to be regulated by OxyR, showed an increase of up to 6-fold in the wild type, but remained stable in the oxyR Ϫ strain. The same was observed for Trx2, but with a 2-fold increase.
grxB Ϫ Phenotype-MC4100grxB null mutant had the same maximum specific growth rate as MC4100 during growth on mineral salt medium (data not shown). During prolonged cultivation of the grxB null mutant in the stationary phase on glucose ammonia-based mineral salt medium, the measured A 500 remained approximately constant and was identical to wild type strain (Fig. 9). However, the colony count on agar plates of the grxB null mutant was significantly lower than that of the wild type strain (Fig. 9), as well as the number of cells determined by microscopic analysis (data not shown). A large number of cells in the culture of the grxB null mutant were much bigger than the cells of the wild type culture. Analysis of the cell viability with a fluorescence test kit showed a higher amount of damaged cells in the grxB null mutant compared with the wild type (data not shown). We explain these observations by a higher death rate of grxB null mutant. The dead cells would serve as a substrate for the surviving population, causing cell division and growth (large cells). In contrast, wild type cells after glucose starvation become smaller and also survive a long time of cultivation under starvation conditions. The hypothesis of an increased lysis of the grxB null mutant is supported by analysis of the total protein content in the cultivation medium, which was increased in the culture of MC4100grxB Ϫ but not in the culture of MC4100 (Fig. 10). DISCUSSION The glutaredoxin and thioredoxin systems are envisaged as antioxidants in normal cell metabolism, in view of their ability to reduce cytosolic disulfides. Stationary phase cells exhibit increased amount of intracellular disulfides and carbonylated proteins compared with exponential phase cells (34). One would therefore expect that the thioredoxin and glutaredoxin system would adapt to the conditions of the stationary phase by providing more reducing equivalents. Levels of GSH (35) and glutathione reductase (14) are elevated at the stationary phase as are other antioxidant systems including HPII (36). From the two thioredoxins and the three glutaredoxins of the thioredoxin and glutaredoxin systems, it is only Grx2 and Trx1 that are elevated in the stationary phase (17). Grx1 is down-regulated, whereas Grx3 and Trx2 maintain the same levels irrespective of the cell growth phase (17). These data point to Trx1 and Grx2 as the most potential antioxidant components of the five presently known dithiol thioredoxins and glutaredoxins at the stationary phase of growth. We wanted to investigate the regulation of Grx2 during the transient from exponential growth to starvation. There are three major responses of E. coli entering the stationary phase: the stringent response with ppGpp as the cellular alarmone, the general starvation response with S as the transcriptional regulator, and the cAMP/CRP-connected specific response to glucose starvation (37). Our results suggest that the levels of Grx2 are regulated in the stationary phase by ppGpp in combination with S . Grx2 expression also seems to be regulated by 70 ; Levels of Grx2 in an rpoS Ϫ strain were almost higher than in the wild type rather than lower as one would expect if RpoS was the exclusive factor for Grx2. Therefore, there are two options for the regulation of Grx2 by ppGpp. ppGpp can be either a direct effector by binding to the 70 -RNA polymerase holoenzyme (exponential phase), or it may influence the grxB transcription indirectly by induction of S (stationary phase). Our data suggest that the exclusive induction of S is not sufficient for grxB expression. ppGpp is additionally required as a positive regulator of S -dependent expression of Grx2. A similar case has been described for the regulation of the uspB gene expression (38).
Grx2 expression was inhibited by cAMP at the exponential phase. As the promoter of Grx2 lacks a binding site for the cAMP-CRP complex, it is very unlikely that the protein is regulated by cAMP directly. cAMP is known to function as a negative regulator for the expression of S , so the effect on Grx2 is more likely to be indirect through down-regulation of S . Up-regulation of S could potentially explain the higher levels of Grx2 after hyperosmotic treatment (32) and the increased transcription of the gene under acidic conditions (39). However, hyperosmotic treatment of S -regulated genes could also be independent of S (40,41). The expression profile of Grx2 seems thus similar to that of OsmY, another stationary phase protein osmotically regulated, requiring E S for its expression, with basal levels transcribed by E 70 . OsmY, however, contains cAMP-CMP binding sites at its promoter, which inhibit the initiation of its transcription by the complex (42,43).
Grx2 is involved in the response to oxidative stress, as Grx2 levels were significantly up-regulated in catalase-deficient strains, and null mutants for grxB have high levels of carbony-  lated proteins in their cytosols (8). In this work, levels of Grx2 were highly elevated in oxyR null mutants, although levels of Grx2 and Grx3 are known to be down-regulated after exposure to hydrogen peroxide (17). These findings suggest that Grx2 is not regulated by the antioxidant defense orchestrated by OxyR but from an alternative system. Another potential antioxidant not regulated by OxyR is the thioredoxin-like protein NrdH, which is also up-regulated in an oxyR Ϫ strain (44). A global analysis of E. coli transcripts up-regulated by hydrogen peroxide did not include Grx2 (Grx3 or Trx1), but Grx1 and Trx2 (45). Therefore, the regulation of Grx2 suggests that the protein is mainly involved in the response to stationary phase conditions.
The elevation of Grx2 levels to up to 1% of total cell protein (17) combined with the high GSH levels of the stationary phase (35), and the distorted morphology of the grxB Ϫ cells at the stationary phase imply a vital yet unknown function for Grx2. The three-dimensional structure of the molecule may provide some hints. Grx2 shows close structural similarities to the glutathione S-transferase family of proteins (6), the human glutathione transferase omega 1 (GSTO1) (46), the mouse glutathione transferase theta-like stress response protein (p28) (47), and the human chloride intracellular channel 1 (CLIC1) (48). Common structural characteristics are a two-domain structure, the first domain being a thioredoxin fold domain in which the active site residues, and the second domain an ␣-helical structure. All these proteins are detoxifying or stress response proteins. The similarities to CLIC1, belonging to a family of proteins involved with forming chloride channels in intracellular membranes or are chloride channel modulators (48), provides the interesting possibility that Grx2 may be involved in osmoregulation, especially after the up-regulation of Grx2 by hyperosmosis. Another potential function for the up-regulation of Grx2 at the stationary phase could be the alleviation of the host responses to E. coli. Grx2 prevents apoptosis in rat neurons by activating the binding activity of NF-B via Ref-1 (49). Grx2 can actually affect both the Ras/ phosphoinotiside 3-kinase/Akt/NF-B and the JNK1/2/AP1 cascades (50). When growing in LB 10% of total Grx2 can be detected extracellularly after 24 h of growth (data not shown). It is therefore tempting to envisage an anti-inflammatory role of this protein in relation to symbiosis with the host. The up-regulation of Grx2 transcription under acidic conditions (39) also points to a host defense adaptation. Clearly, more work is needed to elucidate the role of E. coli Grx2.