Mutations in the Glutathione-gated KefC K+ Efflux System of Escherichia coli That Cause Constitutive Activation*

The kefC gene of Escherichia coli encodes a potassium efflux system that is gated by glutathione (GSH) and by GSH adducts. Independently isolatedkefC mutations that result in spontaneous activation of the efflux system have been analyzed. Three mutations affect residues located adjacent to the conserved Rossman fold in the carboxyl-terminal domain. Two mutations lie in a sequence predicted to form a cytoplasmically located loop in the membrane domain of KefC. All of the mutants retain normal regulation by the YabF protein and by GSH adducts. A mutation in the Rossman fold, R416S, alters the normal regulation of KefC by GSH. In contrast to the wild-type protein, which is inactive in the presence of GSH, the R416S mutant is only active in the presence of GSH or its analogue, ophthalmic acid. Other mutations in this region or elsewhere in the protein have their spontaneous activity augmented by depletion of the GSH pool. These data identify a specific role for the carboxyl-terminal domain of KefC in regulating KefC activity and are discussed in the light of recent data that suggest that GSH adducts can bind within a Rossman fold.

The KefB and KefC potassium efflux systems play a major role in protection of Escherichia coli cells against the toxicity of electrophiles. The addition of electrophiles to E. coli cells, or the stimulation of their synthesis in the cytoplasm, elicits the activation of these two efflux systems through the formation of glutathione (GSH) 1 adducts (1)(2)(3). KefB and KefC are maintained in an inactive state by the presence of reduced GSH or its non-sulfhydryl analogue, ophthalmic acid (4 -7). Activation, leading to potassium loss from the cell, occurs when the GSH is conjugated to electrophiles through the sulfhydryl group (1,8). Slower efflux is observed when GSH is removed from cells by mutations affecting the gshA gene (7). KefC is, therefore, a ligand-gated transport system with both negative (GSH) and positive (GSH adducts) effectors. This feature is common in many ion channels, and the organization of the protein is also consistent with a channel-like mechanism.
KefC belongs to a family of membrane proteins that includes Na ϩ /H ϩ antiports and regulatory proteins. 2 One of the significant differences within this group of proteins is the greater complexity of organization of the KefC protein. The aminoterminal domain is similar in hydrophobicity to many other transport proteins; however, there is located within it a potential voltage-gated sequence (RXXX) 5 followed by a series of hydrophobic residues. This is a structural feature similar to that found in many mammalian voltage-gated K ϩ channels (10,12). The protein lacks the distinctive P-loop that would place it in the voltage-gated K ϩ channel superfamily (10). Attached to the carboxyl terminus of the membrane protein via an acidic linker peptide is a "soluble" carboxyl-terminal domain, which contains a Rossman fold that has high sequence similarity to yeast glyceraldehyde-3-phosphate dehydrogenase (11). Finally, genetic evidence supports a homo-oligomeric organization for KefC (13) and the existence of a separate regulatory subunit, YabF for the E. coli KefC system with homologues identified for KefB and the KefX systems of E. coli and Haemophilus influenzae, respectively (10). These features reflect the complexity required to retain appropriate control over KefC activity.
Here we report the analysis of mutations that cause a spontaneous activation of KefC. Five alleles have been identified; two are found in the membrane-located domain of the protein and three are close to the Rossman fold domain. Further analysis of one of the mutations in the Rossman fold motif (R416S) shows that the mutation alters the normal regulation by GSH. These data are consistent with a model in which the carboxylterminal domain of KefC plays a major role in regulating the activity of the protein and may be the location of the GSHbinding site.

MATERIALS AND METHODS
Reagents-All of the chemical reagents were purchased from Sigma or BDH and were of analytical grade when possible. Ophthalmic acid was obtained from Bachem Ltd. (United Kingdom). The chemicals used for preparation of complex growth medium were supplied by Oxoid. Restriction enzymes, Taq DNA polymerase, and universal sequencing primers were supplied by Boehringer Mannheim. The Qiagen plasmid preparation kits were obtained from Qiagen. Other primers designed and used in this study were purchased from Genosys Biotechnologies, Inc. or supplied by Dr. P. Carter. The Wizard PCR Preps DNA purification system was obtained from Promega. The PRISM Ready Reaction DyeDeoxy Terminator cycle sequencing kit was obtained from Applied Biosystems Ltd. The Muta-Gene phagemid in vitro mutagenesis kit, version 2, was purchased from Bio-Rad.
Bacterial Strains, Plasmids, and Bacteriophage-All of the strains and plasmids used in this study are listed in Table I. Phage M13K07 used in this study was purchased from Bio-Rad. P1kc phage was used for transductions.
Media for Growth of Strains-The minimal medium Kx (where x indicates the approximate concentration [mM] of potassium) was used for the culture of cells for potassium transport experiments (14). Solid medium containing less than 1 mM potassium was prepared by repeated * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  washing of the agar in 1 M NaCl to remove contaminating potassium ions in the agar and followed by rinsing with distilled water prior to sterilization. The complex medium LK was prepared as described previously (15). The medium used in the preparation of uracil-containing single-stranded DNA was prepared as described (16). Where required, media were supplemented with ampicillin (25 g ml Ϫ1 ), tetracycline (25 g ml Ϫ1 ), chloramphenicol (12.5 g ml Ϫ1 ), or kanamycin (25 g ml Ϫ1 ).
Growth Conditions-Cells were grown at 37°C in LK complex medium (15) for use in DNA manipulations. For potassium transport experiments, cells were grown in K 120 minimal medium at 37°C. The carbon source for growth was glucose (0.2% w/v). Strains were stored on high potassium medium to prevent the occurrence of suppressor mutations. Supplementation of strains with GSH or ophthalmic acid was achieved by addition of fresh 100 mM stock to a final concentration of 1 mM.
Measurement of Potassium Efflux from Bacterial Strains-Potassium efflux from bacterial strains was recorded as described previously (7,11).
DNA Manipulations-Restriction enzyme digestion and DNA ligation and transformation were carried out using standard protocols (16). Plasmid preparations were made using the Qiagen plasmid preparation kits.
PCR Analysis of kefC Mutants-PCR primers were designed using the published sequence of kefC (11). The primers used to amplify yabF were designed using a sequence obtained from the genome sequencing project (see Ref. 17; accession number P31577 GB.Eco110K). Primer pairs were designed to amplify yabF and kefC in 350 -400-base pair products that overlapped in sequence by approximately 100 base pairs to ensure that there would be no gaps in the sequence. PCR amplification used single colonies of the appropriate strain inoculated into the reaction mix to provide template DNA. Reactions were carried out in a total volume of 100 l (containing each primer at 0.25 M dNTP (200 M each) 2.5 units of Taq DNA polymerase and reaction buffer as supplied by Boehringer Mannheim) and overlaid with approximately 30 l of mineral oil. Once optimized so that only one product was observed by agarose gel electrophoresis, the PCR reaction was purified using the Wizard PCR Preps DNA purification system prior to sequencing. The PRISM Ready Reaction DyeDeoxy Terminator cycle sequencing kit was used to perform cycle sequencing reactions on PCR products or plasmids, using the protocol outlined by the supplier of the kit. Samples were run on an Applied Biosystems 373A DNA sequencer. The cloning junctions of plasmids created in this study were verified using the universal primers supplied by Boehringer Mannheim. All sequencing reactions were performed a minimum of two times on both strands of the template.
Site-directed Mutagenesis of kefC-Plasmid pKC952 carries the entire kefC gene and 152 base pairs of upstream sequence on a HindIII-BamHI fragment cloned into pHG165 (13). The HindIII-BamHI kefCbearing fragment from pKC952 was cloned into the phagemid vector pTZ19U (18) creating the mutagenesis vector pSM7. Site-directed mutagenesis of kefC borne by pSM7 was carried out using the Muta-Gene phagemid in vitro mutagenesis kit, version 2. The supplied instructions were followed with the exception that single-stranded uracil-containing template DNA was prepared by M13K07 infection of CJ236 and transformed with pSM7 (16). The mutagenic primers were supplied with 5Ј phosphorylation: KefCR416S (GGAGAGCAGTAAGGATCCGGTA-ATCTGCCC), KefCS420A (CACCCCGCTAGCGAGCAGTAATCG), KefCS420A2 (CACCCCGCTAGCGAGCAGTAA), and KefCD264A (GGTTCGATAGCGCTCTCCAGCGC). The primers were designed in such a way that a unique restriction site was incorporated at the site of the mutation to allow subsequent identification of mutant plasmids (BamHI for KefCR416S and NheI for KefCS420A and KefCS420A2). Primer KefCD264A was designed to remove an EcoRV site present in wild-type kefC on creation of the desired mutation. Primer KefCS420A2 was used to create the double mutant R416S,S420A (plasmid pSM21) from the single mutant R416S. These primers were employed in separate in vitro reactions to generate the following: pSM9, bearing kefCR416S; pSM10, carrying kefCS420A; pSM21, carrying

RESULTS
Mutations in the kefC locus that affect potassium retention were selected as isolates unable to grow on K 0.1 medium by ampicillin-enrichment of UV-mutagenized cells of E. coli Frag5. The strains grew normally when incubated in K 10 medium (data not shown). We have previously shown that the K 0.1 negative mutants are gain of function alleles and that they can be inactivated by a null mutation at the same site (20). Therefore, the mutants were transduced to tetracycline resistance with P1 phage grown on strains that carried either kefB::Tn10 or kefC::Tn10 and scored for their ability to grow on K 0.1 medium. Of the twelve independent mutants characterized, seven could be converted to K 0.1 ϩ phenotype by insertions of a kefC::Tn10 and were thus designated kefC alleles. The other five mutations were mapped to the kefB locus. 2 The rate and extent of spontaneous K ϩ efflux were measured for each of the kefC mutants after growth in K 120 medium, filtration, washing, and suspension of cells in K 0 media (Fig. 1). When cells were harvested directly from K 120 medium the K ϩ content of the mutants and the parent strain were similar (data not shown), however, each of the mutants displayed a significantly lower K ϩ pool than the parent at the first time sample (Table II) indicating rapid loss over the first 60 s. K ϩ efflux continued in the mutants for 10 -20 min, whereas the parent retained more than 80% of its K ϩ pool over this period. Five strains exhibited a unique efflux profile (strains TK104 and TK121 and TK116 and TK119 were indistinguishable, and these pairs of strains subsequently proved to be isoallelic; see below).
The location of the mutations was determined by direct sequencing of 300 -400-base pair PCR products of the kefC gene. Only a single base change was observed in each of the mutants with the exception of strain TK120 (kefC120), which carries two changes R416S and S420A (Table II). Mutations were found in both the amino-terminal membrane protein domain (E262K and D264A) and in the carboxyl-terminal domain (R416SϩS420A, V427A, and R516C), which carries the Rossman fold. Strains TK104 and TK121 and strains TK116 and TK119 were found to be isoallelic. yabF is an open reading frame that overlaps the kefC gene and is required for full KefC activity (10). The sequence of yabF was determined in each of the mutants and was found not be changed from that of the parent (data not shown).
Strain TK120 carries two mutations located very close together, R416S and S420A. Site-directed mutagenesis was used to separate the mutations, and activity was studied using a construct that gives reduced KefC expression, pkC952 (13). Such a construct was desirable since high level expression of mutant proteins that cause a fast K ϩ leak on the cell could pose a severe growth defect leading to selection for suppressor mutations. Subcloning of the 2.15-kilobase HindIII-BamHI fragment from pkC952 into the mutagenesis vector pTZ19U created a construct, pSM7, that yielded no functional KefC activity (data not shown). This enabled mutations to be introduced into this very high copy number vector without the concern that expression would be toxic to the cell. Mutagenesis was carried out as described under "Materials and Methods," and the resulting clones were used to create separate plasmids carrying the separate R416S and S420A mutations and the two mutations in combination. The DNA was then subcloned into pHG165 to create pSM12 (wild type), pSM14 (R416S), pSM15 (S420A), and pSM30 (R416S,S420A). Plasmids were transformed into strain MJF276 (KefB Ϫ KefC Ϫ ), and the rate of spontaneous K ϩ loss was determined (Fig. 2). As expected, plasmids pSM12 and pkC952, which carry the same sequence, gave similar low rates of K ϩ loss that were slightly faster than those observed in the nontransformed MJF276. Plasmid pSM15, which carries the S420A mutation, introduced a slow leak that was reproducibly faster than the MJF276/pSM12. However, rapid, spontaneous K ϩ efflux was only observed from strains carrying the R416S mutation (MJF276/pSM14 (R416S) and MJF276/pSM30 (R416S,S420A)) (Fig. 2). The rate of K ϩ efflux seen with the combined mutations was greater than that seen with R416S alone, and this is consistent with the slightly enhanced leak seen with S420A alone, i.e. the R416S and S420A mutations are additive with the R416S mutation dominant. The efflux profile for strain TK120 (Fig. 1), which carries the chromosomal R416S ϩ S420A mutations, was similar to that obtained with MJF276/pSM30 (Fig. 2), confirming that the cloned lesions were responsible for the K ϩ leak. Slight differences between the strains may arise due to a slight difference in the genetic background of MJF276 and TK120 and also by the separation in the expression of yabF and kefC in MJF276/ pSM30. This suggestion was supported by analysis of MJF276/ pSM26 (kefCD264A), which carries the D264A mutation in the same plasmid context as the R416S and S420A mutations. The  The missense mutations in the strains were identified by PCR amplification of overlapping regions of kefC and direct sequencing of the products. The parent strain Frag5 was used as a positive control in the PCR-based analysis and has the same kefC sequence as that reported previously (11). The mutations were the result of single base pair mutations that were found on both strands of the template. The resulting codon change is indicated, together with the encoded amino acid alteration. Strain DNA mutation (base) a Codon change Amino acid mutation a Base number 1 is the adenine of the first codon of kefC (ATG). b The K ϩ concentration is that observed at the first time point after suspension of the cells in K 0 medium. spontaneous K ϩ leak seen in MJF276/pSM26 was slower than that seen in TK104 (kefCD264A) (compare Fig. 1 with Fig. 2). Notwithstanding these differences, we conclude that the most important mutation in strain TK120 is the R416S mutation.
YabF is the putative product of the open reading frame 5Ј to kefC and is required in trans for the activity of KefC (10). Since the kefC mutants were identified by their loss of regulation of K ϩ efflux it was possible that the mutations affected the control by YabF. Thus, the K ϩ leak was determined in strains MJF374 (YabF Ϫ ,KefC Ϫ ) 2 and MJF276 (YabF ϩ ,KefC Ϫ ) transformed with pSM12 and pSM14(R416S) (Fig. 3). No significant K ϩ efflux was observed from MJF374/pSM12 or MJF374/pSM14, but normal efflux was observed in the transformed derivatives of MJF276, which possesses YabF activity. Thus, YabF is required for the activity of both the wild-type and the mutated KefC proteins.
We have previously shown that the mutation in TK121 can be partially suppressed by co-expression of the wild-type gene from pkC592 (13), which is equivalent to pSM12 (see "Materials and Methods"). Since the R416S mutation lies in a different domain to TK121 mutation, D264A, we sought to determine whether co-expression of the two KefC proteins would allow either co-dominance or suppression. Plasmids pSM14 (R416S) and pSM12 (wild type) were transformed separately into MJF104 (kefC104, kefB::Tn10), and the spontaneous K ϩ efflux was measured (Fig. 4). Expression of the wild-type kefC gene from a multicopy plasmid, pkC11, or at a reduced expression level from pSM12 partially suppressed the K ϩ loss caused by the kefC104 mutation (D264A). Similarly, transformation of TK120 (kefC120) separately with pSM12, pSM15, or pkC11 suppressed the leak caused by the mutation R416S (data not shown). Thus, both the mutated subunit carrying the D264A mutation and the subunit carrying the combined R416S,S420A mutations can be suppressed by the wild-type gene expressed in either high copy number (pkC11) or low copy number (pSM12). The ability of the KefCS420A subunit to suppress the leak from TK120 in strain TK120/pSM15 suggests that this mutation does not have a major effect on the configuration of the KefC protein, which is consistent with the presence of only a slow spontaneous leak from strain MJF276/pSM15 (Fig. 2). In contrast, the K ϩ leak from MJF104/pSM14, which introduced the R416S mutation, was faster than for either MJF104 or MJF104/pSM12 (Fig. 4), and thus, KefCR416S failed to suppress the D264A lesion. These data indicate that suppres-sion requires wild-type KefC sequences in both the membrane and the carboxyl-terminal domains.
The KefC K ϩ efflux system is maintained in a closed state by GSH or its non-sulfhydryl analogue, ophthalmic acid (6,7). Cells that lack GSH leak K ϩ via KefB and KefC, and this phenotype is reminiscent of those observed with the mutants described here, which raises the possibility that the mutations affect the GSH-binding site. Thus, we sought to determine whether the K ϩ leak observed with the mutations was exacerbated in a GSH-deficient mutant. Strain MJF335 (KefB Ϫ ,KefC Ϫ ,GshA Ϫ ) was transformed separately with pSM14 (R416S), pSM26 (D264A), pSM30 (R416S,S420A), or pSM12 (wild type), and the spontaneous K ϩ leak was determined (Fig. 5, A-C). The rate of K ϩ efflux from MJF335/pSM12 was enhanced over that seen in the isogenic GSH-sufficient strain MJF276/pSM12 and could be inhibited by inclusion of either GSH or ophthalmic acid in the growth medium used to prepare the cells (Fig. 5A). Similarly, when plasmid pSM26 (D264A) was introduced into strain MJF335 the leak was increased compared with that seen in MJF276/pSM26 to the extent that over 60% of the K ϩ pool was released in the first 60 s after suspension of the cells into K 0 medium (Fig. 5B).
GSH or ophthalmic acid reduced the K ϩ leak to that seen in MJF276/pSM26 (Fig. 5B). Thus, the leak caused by the absence of glutathione was additional to the spontaneous leak caused by the D264A mutation. In contrast it was observed that the leak of K ϩ was slower from MJF335/pSM14 (R416S) than from MJF276/pSM14 and was enhanced by growth in the presence of GSH (Fig. 5C). Ophthalmic acid also restored spontaneous K ϩ efflux to strain MJF335/pSM14. Similar data were obtained with MJF335/pSM30 (R416S,S420A) (data not shown). These data clearly show that the spontaneous activity of KefCR416S is dependent upon the presence of GSH or its analogue, which is an inversion of the normal phenotype.
The above analysis was confirmed and extended by analysis of the chromosomal mutations affecting KefC. Each of the strains was transduced to kanamycin resistance using a gshA::Tn10(Kan) insertion mutant and the K ϩ leak determined in the presence and absence of GSH (Table III). For all of the mutants except TK120 (R416S,S420A) the rate of K ϩ was faster in a GSH-deficient derivative than in the GSH ϩ parent, and the leak could be reduced by growth of the cells with GSH (Table III). As observed above for MJF335/pSM14 and MJF335/ pSM30, the GSH-deficient derivative of strain TK120 (R416S,S420A) exhibited a slower K ϩ leak than did TK120 itself (Table III), and the fast leak could be restored by growing the strain in the presence of GSH (data not shown). Thus, the reversal of the normal GSH-gating is specific to the R416S mutation.
GSH adducts activate the KefC system (7). Since the phenotype of the KefC R416S derivative is an altered interaction with GSH the effect of the mutation on activation by the GSH adduct, N-ethylsuccinimidyl-S-GSH, formed by the reaction of N-ethylmaleimide with GSH (7) was investigated. The rate of N-ethylmaleimide-elicited efflux were similar for MJF276/ pSM12, MJF276/pSM14, MJF276/pSM26, and MJF276/pSM30 (data not shown). It is clear that the mutation does not substantially affect the activation of KefC by GSH adducts. DISCUSSION The data presented here identify residues within KefC that when mutated alter the regulation of the efflux system such that it exhibits a spontaneous K ϩ leak. The mutations map to two regions of the protein: a well-defined region in the aminoterminal hydrophobic domain, the HALESDIE sequence; and the carboxyl-terminal hydrophilic domain. All of the residues that cause a spontaneous K ϩ leak are highly conserved across the four members of the KefB/C family of proteins identified to date but are not as well conserved in the other members of the wider family that includes the NapA Na ϩ /H ϩ antiport family and the RosB regulatory protein. 2 In addition we show that one of the mutations, R416S, which is in the carboxyl-terminal domain, specifically alters the regulation of the system by GSH. In this mutant, K ϩ efflux is entirely dependent upon the presence of GSH or its analogue, ophthalmic acid. In the presence of either peptide the efflux system is active, but in its absence the K ϩ leak is much reduced. Neither mutations in the HALESDIE sequence nor other mutations in the carboxylterminal domain have this effect on the phenotype since the absence of GSH increases the K ϩ leak in these strains.
The first implication of these results is that the carboxylterminal domain plays a significant role in regulating the closed state of the KefC system. An obvious hypothesis would be that the two regions identified in this study interact directly to effect the closure of KefC, in a way analogous to the interaction of ␤ subunits of Shaker channels with the S4-S5 linker peptide (12,21). The prediction from positive charge distribu- a Strains deficient in synthesis of GSH were created for each strain carrying a kefC mutation by transduction to kanamycin resistance using Frag56 (gshA::kan) as donor.
b Supplementation of the growth medium with 1 mM GSH ensured that the cells were GSH-replete prior to harvest for measurement of spontaneous K loss.
c Potassium loss was measured for 20 min, and the pattern observed at 3 min was maintained throughout this period of incubation. tion and hydropathy plots is that the HALESDIE sequence of KefC is located at the cytoplasmic face of the membrane as is the carboxyl-terminal domain, which would place the two in the right location to interact. The HALESDIE sequence is part of a strongly conserved region of the KefB/C family of proteins. 2 The sequence can be represented as RHELEXDIEPFK (where X ϭ Ala, Ser, or Thr) and therefore, has a strongly acidic nature although the D264 is not conserved in KefB (A262) and E266 is substituted in KefX (A263) of Myxococcus xanthus. However, in all four members of the KefB/C family there are always three negatively charged amino acids in this sequence that is flanked by conserved positively charged groups. Thus, the mutant analysis has clearly defined a region of KefC that is important for regulation of activity.
The R416S, V427A, and R516C mutations of the carboxylterminal domain affect conserved residues but have different effects on KefC activity. Although all confer a K ϩ leak upon the KefC protein, implying loss of regulation, the effect of removing GSH is to convert the protein to the closed state in the case of R416S and to a more active state with R416C and V427A. R416 falls immediately after the putative ␤-␣-␤ (Rossman) fold of the carboxyl-terminal domain and is conserved in glyceraldehyde-3-phosphate dehydrogenase and in the carboxyl-terminal Rossman fold of TrkA, the regulator of K ϩ uptake in E. coli (22). Analysis of the sequences of the carboxyl-terminal domain in the four members of the KefB/C family shows that only the first 120 amino acids is conserved and therefore, that this region may be important for regulation of activity. The difference in the phenotype between the R416S and the V427A and R516C mutations suggests that the region of the ␤-␣-␤ fold may be critical for controlling the activity of this domain as a regulator of KefC opening and closing. Some support for this hypothesis is the recent observation that GSH adducts bind in the Rossman fold of glyceraldehyde-3-phosphate dehydrogenase from human erythrocytes (23). Membrane-associated glyceraldehyde-3-phosphate dehydrogenase bound various GSH adducts and was photoaffinity-labeled by S-(p-nitrobenzyl)-GSH-125 I-4azidosalicylic acid, and this could be competitively inhibited by low concentrations of NAD ϩ and ATP (23). The residues identified by covalent linking were not in the GXGXXG motif that forms the core of the Rossman fold but at a slightly more distant site within the predicted binding site for NAD ϩ (23,24). The Rossman fold domain of glyceraldehyde-3-phosphate dehydrogenase has also been implicated in the binding of RNA (9,25), and it has been suggested that the fold is particularly suited to binding of both small and large molecules. These observations would be consistent with an important role for the carboxyl-terminal domain of KefC in regulating channel activity and suggest that this domain might contain the binding site for GSH adducts.