The levels of ribonucleotide reductase, thioredoxin, glutaredoxin 1, and GSH are balanced in Escherichia coli K12.

The dithiol forms of thioredoxin and glutaredoxin are hydrogen donors for ribonucleotide reductase. We have determined the intracellular levels of ribonucleotide reductase (RRase), thioredoxin (Trx), glutaredoxin 1 (Grx1), and glutathione (GSH) and the glutathione redox status in new Escherichia coli K12 strains lacking thioredoxin (trxA−), glutaredoxin 1 (grxA−), and/or GSH (gshA−) or overproducing Trx or Grx1 from multicopy plasmids. We propose a regulatory network in which RRase levels are balanced with those of Trx, Grx1, and GSH so that deficiency or overproduction of one component would promote the opposite effect on the others to maintain a balanced supply of deoxyribonucleotides. GSH deficiency strongly increased both Grx1 levels and RRase activity, even more than Trx deficiency. Double gshA−trxA− bacteria were viable, whereas additional deficiency in lipoate synthesis (gshA−trxA−lipA−) caused the inability to grow in minimal medium plates supplemented with acetate plus succinate instead of lipoic acid. Thus, lipoate might be the only substitute of GSH for glutaredoxin reduction in gshA−trxA− cells, although the extremely high Grx1 content (55-fold) of these bacteria suggests that electron transfer from lipoate might be an inefficient reduction mechanism of glutaredoxins. Moreover, the enhanced Grx1 level of gshA−trxA− cells could obviate the need for a large increase in RRase activity, in contrast to grxA−trxA− double mutant cells. Impairment of the sulfate assimilation pathway, leading to very low GSH concentrations, and an oxidized glutathione redox state might explain the inability of grxA−trxA− cells to grow in minimal medium. Restoration of nearly normal levels of both GSH content and redox status cure the growth defect.

The reduction of ribonucleotides to deoxyribonucleotides is catalyzed by the enzyme ribonucleotide reductase (RRase) 1 (Reichard, 1988). Two dithiol-dependent hydrogen donor systems for this enzyme have been discovered separately (Holmgren, 1989). One is the NADPH-dependent thioredoxin system (Holmgren, 1985a), and the other is a glutathione-dependent system where the protein glutaredoxin couples the oxidation of GSH to the reduction of ribonucleotides (Holmgren, 1985b). Thioredoxin and glutaredoxin are also important for the reduction of sulfate to sulfite via 3Ј-phosphoadenosine 5Ј-phosphosulfate (PAPS) reductase, which is the first step in the cysteine biosynthesis and is essential for sulfur metabolism (Tsang, 1981;Kredich, 1987).
A double mutant lacking thioredoxin and glutaredoxin was constructed in Escherichia coli K38 in order to establish the physiological relevance of these two proteins (Russel and Holmgren, 1988). The best growing mutant selected (named A410 strain) was not viable on minimal medium unless a source of reduced sulfur was added. It was concluded that thioredoxin and glutaredoxin were essential for the reduction of sulfate (Russel et al., 1990) but not for deoxyribonucleotide biosynthesis, implying the existence in E. coli of a third unknown hydrogen donor system for ribonucleotide reductase.
To investigate whether a mutator phenotype was associated to the simultaneous deficiency in thioredoxin and glutaredoxin, new double defective bacteria carrying trxA and grx null mutations were isolated in the genetic background of E. coli K12 . In contrast to previous data (Russel and Holmgren, 1988;Russel et al., 1990), the new defective bacteria made colonies of normal size on rich plates and gave rise (at a frequency of 10 Ϫ4 ) to stable derivatives that grew on minimal medium plates. One of these derivatives was saved as E. coli K12 UC647. Strain UC647 shows remarkable physiological properties when compared with its parental wild type cells. Thus, it (i) grows on minimal medium plates; (ii) shows normal mutagenic sensitivity toward a wide variety of DNA-damaging agents; (iii) displays 14% of GSH-dependent and 30% of NADPH-dependent ribonucleotide reduction capacity with CDP as substrate in the presence or the absence of exogenous RRase, respectively; and (iv) has very high levels (Ն20-fold increase) of ribonucleotide reductase activity. These properties indicate (i) that thioredoxin and glutaredoxin might not be essential for sulfate reduction, in contrast to previous data; (ii) that a balanced dNTP synthesis is maintained in the absence of thioredoxin and glutaredoxin, even under exposure to mutagens; and (iii) the existence of a third GSH-dependent hydrogen donor system for ribonucleotide reductase in E. coli.
Initial attempts to purify the presumed hydrogen donor for RRase from the original mutant strain A410 were not sucess-ful. The newly constructed E. coli K12 UC647 was then used on the assumption that any hydrogen donor might be induced to higher levels in this strain. From UC647 two independent GSH-disulfide oxidoreductases (using hydroxyethyl disulfide as a substrate) were purified to homogeneity and named glutaredoxin 2 (Grx2) and glutaredoxin 3 (Grx3), respectively (Åslund et al., 1994). Grx3 showed a low but significant activity as hydrogen donor for E. coli ribonucleotide reductase, whereas Grx2 was inactive. Therefore, glutaredoxin 3 was considered to be the substitute for thioredoxin (Trx) and the first discovered glutaredoxin (Grx1) in the double mutant strain UC647 (Åslund et al., 1994).
This study investigates the association among the intracellular levels of RRase, their two main hydrogen donors, thioredoxin and glutaredoxin 1, and glutathione and its redox status. To this end new E. coli K12 bacterial strains lacking Trx, Grx1, and/or glutathione or overproducing either Trx or Grx1 from multicopy plasmids have been constructed in the genetic background of UC647.

MATERIALS AND METHODS
Chemicals and Proteins-[ 3 H]CDP was from Amersham Corp. GSSG, GSH, lipoic acid (DL-6,8-thioctic acid), and yeast glutathione reductase were from Sigma. Perchloric acid, orthophosphoric acid, and salts for the mobile phase used in HPLC determinations were from Merck (Darmstadt, Germany). All other chemicals were of analytical reagent grade. Thioredoxin, glutaredoxin, and thioredoxin reductase from E. coli were prepared as described previously (Dyson et al., 1989;Björnberg and Holmgren, 1991;Luthman and Holmgren, 1982). E. coli ribonucleotide reductase was a gift from Prof. Britt-Marie Sjöberg, Stockholm University.
Bacterial Strains and Plasmids-The bacterial strains and plasmids used in this work are listed in Table I. All UC strains are E. coli K12, UC5710 being considered the parental wild type (wt). Strains UC844 (trxA Ϫ ), UC848 (grxA Ϫ ), UC946 (grxA Ϫ trxA Ϫ , unable to grow in minimal medium), and UC827 (spontaneous derivative of UC946 growing in minimal medium) are equivalent to UC518, UC525, UC646, and UC647, respectively  but without the mutagenesis-enhancing plasmid pKM101. The gshA Ϫ bacteria were constructed by P1 (gshA::Tn10kan) transduction and selection for kanamycin resistance and subsequently screened for diamide sensitivity. Strains carrying the gshA::Tn10kan allele showed undetectable levels of glutathione, as determined by HPLC. The triple gshA Ϫ trxA Ϫ lipA Ϫ defective strain (UC954) was constructed by cotransducing to UC859 the lipA2 mutation and the linked Tn10 insertion (zbe279::Tn10). Tet-racycline-resistant transductants were first selected on rich LB medium and then scored for growth on minimal medium plates supplemented with either lipoic acid or acetate plus succinate. The vector plasmid (pBR322) and those carrying the gene encoding the Trx (pBHK8) or the Grx1 (pBR322ECG) proteins have been previously described (Bolívar et al., 1977;Wallace and Kushner, 1984;Höög et al., 1986). Strains overproducing Trx (UC872 and UC919) or Grx1 (UC873 and UC929) were isolated by transformation of UC5710 or UC827 with pBHK8 and pBR322ECG, respectively. Control strains UC871 and UC928 carrying the pBR322 vector were also obtained. Plasmids were isolated using the Magic minipreps DNA purification system. Competent bacteria and transformation were as described by Hanahan (1983). E. coli K38 and derivatives (A307, A407, and A410) were used for comparison with UC E. coli K12 strains.
Enzymatic Activities-Cell-free extracts were prepared from bacteria exponentially growing in LB nutrient medium. RRase activity was determined as described previously  by the formation of [ 3 H]dCDP from [ 3 H]CDP in a reaction mixture containing 30 mM Hepes, pH 7.6, 10 mM MgCl 2 , 1.3 mM ATP, 0.5 mg/ml bovine serum albumin, 0.5 mM [ 3 H]CDP (31,000 cpm/nmol), 2 mM NADPH, 1.2 M E. coli Trx, and 0.2 M E. coli thioredoxin reductase in 0.12-ml final volume. Three different concentrations of bacterial extract were used in the assay, and average activity was calculated by linear regression. All experiments were conducted at least three times. Glutathione reductase and thioredoxin reductase activities were measured as described previously (Mata et al., 1984;Holmgren, 1977). No differences in the activities of these two enzymes were observed among the bacterial strains described in this paper.
Purification of Antibodies and Enzyme Immunoassay-Purification of anti-Trx and anti-Grx1 antibodies by immunoaffinity chromatography and quantification of Trx and Grx1 protein levels by indirect competitive enzyme immunoassay were carried out as described previously . Antibody titers were 0.183 and 0.531 nmol/ml for anti-Trx and anti-Grx1, respectively.
Glutathione Determination-Bacteria from LB overnight cultures were harvested, washed twice in VB salts, reinoculated in VB medium with casamino acids, and grown to stationary phase in a rotary incubator. Cells were centrifuged at 15,000 ϫ g, and the pellets aliquoted (200 mg/vial) and frozen at Ϫ80°C. Bacteria were thawed, resuspended  Russel and Holmgren (1988), will be referred in the text as grxA to indicate the absence of the classic glutaredoxin (Grx1), because two new additional glutaredoxins (Grx2 and Grx3) have been recently discovered (Åslund et al., 1994). ⌬trxA, grxA::kan zbi::Tn10 (unable to grow in minimal medium) Russel and Holmgren (1988) (5 ml/g) in 1 M HClO 4 containing 2 mM EDTA, and disrupted at room temperature in an ultrasonic bath for 15 min (Alonso-Moraga et al., 1987). The extracts were centrifuged at 31,000 ϫ g for 20 min at 4°C, and the pellets were extracted again with HClO 4 . The protein-free supernatants were filtered through 0.2-m polysulfone filters (Scharlau, Barcelona, Spain), aliquoted, and frozen at Ϫ80°C until used for analysis. For GSSG determination, the extracts were prepared as before, but cells were resuspended at 2 ml/g in HClO 4 and extracted only once. For protein-bound glutathione determination (Rodríguez-Ariza et al., 1994), the pellets from HClO 4 extraction were resuspended in 1 ml of 0.1 M KH 2 PO 4 /0.1 M KOH (1:3, v/v, pH 12.0). To release GSH, 0.1 ml of freshly prepared 0.03 M dithiothreitol solution was added, the mixture being then sonicated for 30 min at room temperature in a ultrasonic bath. Later, acetonitrile (0.5 ml) and 3.5 M HClO 4 (0.1 ml) were supplemented, and the solution was kept on ice 15 min. After centrifugation at 31,000 ϫ g for 20 min, the protein-free supernatant was filtered and frozen at Ϫ80°C. GSH and GSSG were determined by HPLC with electrochemical detection (Rodríguez-Ariza et al., 1994). A liquid chromatograph with a solvent delivery module 126 (Beckman, San Ramon, CA), and a Coulochem II detector (ESA, Bedford, MA) was used. Separation was achieved with a Supelcosil LC-18 (250 ϫ 4.6-mm inner diameter, 5-m particle diameter, 100-Å pore size) reversed-phase column, protected by a Supelguard precolumn, both from Supelco (Bellefonte, PA). Elution was carried out isocratically with 20 mM sodium phosphate buffer, pH 2.7, at 1.5 ml/min. The potential settings of the Coulochem II multidetector were: guard cell, ϩ0.90 V; detector 1, ϩ0.45 V; and detector 2, ϩ0.80 V. Extracts were diluted 20 -50-fold in mobile phase for GSH and protein-glutathione mixed disulfide determinations and not diluted for GSSG determinations.

RESULTS
The Levels of Trx, Grx1, and GSH Are Compensated-The levels of Trx and Grx1 were determined by specific competitive enzyme-linked immunoassays in E. coli strains defective in thioredoxin (trxA Ϫ ), glutaredoxin 1 (grxA Ϫ ), and/or GSH biosynthesis (gshA Ϫ ) (Table II, upper part). A small but consistent increase in Trx was observed in the absence of Grx1 and vice versa. These results suggested the existence of a regulatory network in which the levels of Trx and Grx1 were balanced. This was further addressed by measuring the concentrations of these two proteins in bacteria carrying multicopy plasmids for overproduction of Trx or Grx1 (Table II,  Glutaredoxin activity is coupled to cellular glutathione. Thus, the putative effect of GSH deficiency on Grx1 level was investigated in bacteria carrying a null gshA::Tn10 allele blocking the first step in the GSH biosynthetic pathway (Greenberg and Demple, 1986). An important increase of 5.53-fold in Grx1 was observed in gshA Ϫ single mutant, the increase raising to 54.9-fold when the GSH deficiency was combined with the trxA Ϫ mutation (Table II). Comparatively, the effect of GSH deficiency on Trx level was much less pronounced, as would be expected. To see whether lipoic acid is involved in an alternative pathway in reducing glutaredoxins, we constructed a triple gshA Ϫ trxA Ϫ lipA Ϫ mutant strain by P1 transduction of the lipA2 (lipoate synthesis) mutation to UC859 (gshA::Tn10kan, ⌬trxA). The resulting bacterial strain (UC954) grew on minimal plates supplemented with lipoic acid, but no growth could be observed when replacing the lipoic acid requirement by acetate and succinate. Table III shows the ribonucleotide reductase activity of bacteria containing variable amounts of Trx and Grx1, its two main electron donors. The absence of Trx or Grx1 increased RRase activity of wt cells by a factor of 3.32and 1.60-fold, respectively, whereas bacteria simultaneously lacking Trx and Grx1 exhibited very high levels of RRase activity (ϫ 23.08), as previously reported . The link between these three proteins was further investigated by measuring RRase activity in both wt and grxA Ϫ trxA Ϫ bacteria overproducing Trx or Grx1. Oppositely to the increased RRase activity observed in Trx-or Grx1-defective bacteria, overproduction of these two electron donors decreased RRase activity of wt bacteria to 41 and 33%, respectively. Much more extensive effects were observed in grxA Ϫ trxA Ϫ bacteria, in which Trx or Grx1 overproduction diminished RRase activity to 8 or 12% of the control (grxA Ϫ trxA Ϫ /pBR322), thus showing approximately wt levels. These drastic decreases are explained by the high RRase activity of grxA Ϫ trxA Ϫ /pBR322 bacteria as compared with the corresponding control strain wt/pBR322 (3.83 versus 0.27 nmol dCDP/20 min ϫ mg protein). Deficiency in GSH, electron donor of all glutaredoxins, increased RRase activity to a higher extent than the absence of Grx1, 4.68versus 1.60-fold. Nevertheless, GSH deficiency did not show  H]dCDP, as described under "Materials and Methods." Measurements were carried out in cell-free extracts of bacteria grown in LB nutrient medium. All bacterial strains are derived from UC5710, which is considered the parental wild type. UC827 (able to grow in minimal medium) was the grxA Ϫ trxA Ϫ mutant strain. The overproduced proteins (Trx or Grx 1) are indicated in parentheses. Data show the mean values Ϯ S.D. of at least three independent determinations. The relative values compared with those of wt, wt/pBR322, or grxA Ϫ trxA Ϫ /pBR322 are shown in italics.

Role of Trx and Grx1 in Glutathione Redox
Status-The glutathione content and redox status of trxA Ϫ and/or grxA Ϫ mutant bacteria are presented in Table IV. Glutathione redox status was expressed as the [protein-glutathione mixed disulfide]/[GSH] ratio, because UC bacterial strains showed undetectable levels of GSSG (Ͻ0.1 nmol/g dry weight). The grxA Ϫ trxA Ϫ bacteria unable to grow in minimal medium plates showed very low concentrations of total intracellular GSH (11% of wt level). This decreased GSH concentration was accompanied by a shift toward a more oxidized redox status, which increased 2.47-fold as compared with wild type. Equivalent or even higher effects were observed in the E. coli K38 genetic background included for comparison with UC K12 strains. The UC grxA Ϫ trxA Ϫ derivative strain able to grow in minimal medium plates had recovered nearly wild type levels of both total intracellular GSH concentrations and glutathione redox status. DISCUSSION Here we show that trxA Ϫ or grxA Ϫ mutants have increased levels of Grx1 or Trx, respectively, in agreement with previous results in bacteria carrying the pKM101 plasmid . The inverse relation between the intracellular levels of Trx and Grx1 was further supported by using bacteria overproducing one of these two proteins. Trx concentration seemed to have a higher effect on the Grx1 level than Grx1 on Trx, as expected from the existence of two other glutaredoxins, Grx2 and Grx3 (Åslund et al., 1994). The present paper demonstrates also that bacteria deficient in GSH biosynthesis has an elevated amount (ϫ5.53) of Grx1; this effect was higher than that of Trx deficiency (ϫ2.22), in agreement with the fact that GSH is the physiological hydrogen donor of all known glutaredoxins (Holmgren, 1989;Åslund et al., 1994).
In spite of the pivotal physiological role of GSH, bacteria simultaneously defective in both GSH and thioredoxin were viable, as previously reported (Fuchs et al., 1983). This suggests that glutaredoxins might be reduced independent of GSH. The putative involvement of lipoic acid in glutaredoxin reduction in vivo was investigated because dihydrolipoamide is a good reductant of E. coli Trx in vitro (Holmgren, 1979b). The inability of the gshA Ϫ trxA Ϫ lipA Ϫ triple mutant strain (UC954) to grow on minimal plates supplemented with acetate plus succinate suggests that lipoic acid could be the only substitute of GSH for glutaredoxin reduction in the gshA Ϫ trxA Ϫ bacteria (UC859). The low efficiency of lipoate as hydrogen donor of glutaredoxins would explain the extremely high increase in Grx1 content (ϫ54.9) observed in UC859.
We have previously reported that deficiencies in Trx, Grx1, and particularly in both hydrogen donors result in enhanced ribonucleotide reductase activity . Here we confirm the inverse relation existing between Trx or Grx1 and RRase activity in overproducing bacterial strains. The RRase activity diminished in cells with high intracellular levels of Trx or Grx1, this effect being clearly noticeable in grxA Ϫ trxA Ϫ bacteria where RRase is induced Ͼ20fold. This work demonstrates further that deficiency in GSH biosynthesis (gshA Ϫ bacteria) also increases the activity of ribonucleotide reductase, actually to a larger extent than Trx deficiency. No synergism between RRase activity and the simultaneous deficiency in Trx and GSH was detected in contrast to that reported for Trx and Grx1. This might be explained by the fact that gshA Ϫ trxA Ϫ defective bacteria contained much higher (55-fold) levels of Grx1 than wild type cells. This large increase in Grx1 and conceivably also in other glutaredoxins might exclude the need for inducing RRase.
E. coli trxA Ϫ or grxA Ϫ single mutants exhibited nearly wild type glutathione levels and redox status, including no increases in GSH-protein mixed disulfides. Thus, E. coli must have an additional system to reduce low molecular weight disulfides, as previously suggested from studies with glutathione reductase and thioredoxin or thioredoxin reductase defective bacterial strains (Tuggle and Fuchs, 1985). This proposal is supported further by the nearly normal glutathione redox state of grxA Ϫ trxA Ϫ double mutant selected to grow in minimal medium plates.
The thioredoxin and glutaredoxin 1 defective E. coli K12 strain (UC946) unable to grow in minimal medium plates has very low glutathione content with a markedly oxidized redox status. The glutathione concentration of UC946 could even be overestimated because it was measured in cells grown in minimal medium supplemented with 0.2% casamino acids, thus containing cystine at 3.2 mg/liter. The impairment of sulfate assimilation pathway and the oxidized redox status of glutathione is considered the most reliable explanation for the inability of UC946 to grow in minimal medium, based on the previous proposal that accumulation of PAPS is toxic to the cells (Russel et al., 1990) and that relatively small oxidation of glutathione essentially inactivates the reducing capacity of glutaredoxin (Holmgren, 1979a). The ability to grow in minimal medium plates of E. coli grxA Ϫ trxA Ϫ defective bacteria (UC827) was accompanied by the recovery of nearly wild type levels of both glutathione content and redox status. The functioning of the sulfate reduction pathway in UC827 would require the induction of PAPS reductase and/or of new electron TABLE IV Glutathione content and redox status in trxA Ϫ and grxA Ϫ mutant strains of E. coli GSH and PSSG (nmol/g dry weight) were determined by HPLC with electrochemical detection as described under "Materials and Methods." Measurements were carried out in extracts of bacteria grown in minimal VB medium supplemented with 0.2% casaminoacids. Bacterial strains were derived from UC5710 or K38, which are considered the parental wild types. Data show the mean values Ϯ S.D. of at least three independent determinations. The relative values compared with those of wt are shown in italics. donors for this enzyme, distinct from Trx and Grx1. Glutaredoxin 2 is probably not the electron donor for PAPS reductase, 2 leaving as alternative the induction of Grx3 or expression of a new redoxin from E. coli (Jordan et al., 1996). This would additionally explain the recovery of nearly wild type glutathione redox status in strain UC827, because no differences in glutathione reductase activity were found among the different bacterial strains (data not shown). Now that the E. coli gene (grxC) for Grx3 has been cloned 3 and the Grx2 and Grx3 proteins have been purified to homogeneity (Åslund et al., 1994), it should be possible to quantify in different bacterial constructions the levels of Grx2 and Grx3 by specific competitive enzyme-linked immunoassays and the expression of grxC gene by Northern analysis. These studies will give clues about the roles of glutaredoxins in the physiology of E. coli, thus facilitating the characterization of their function in a physiological system.