All Four Putative Selectivity Filter Glycine Residues in KtrB Are Essential for High Affinity and Selective K+ Uptake by the KtrAB System from Vibrio alginolyticus*

The subunit KtrB of bacterial Na+-dependent K+-translocating KtrAB systems belongs to a superfamily of K+ transporters. These proteins contain four repeated domains, each composed of two transmembrane helices connected by a putative pore loop (p-loop). The four p-loops harbor a conserved glycine residue at a position equivalent to a glycine selectivity filter residue in K+ channels. We investigated whether these glycines also form a selectivity filter in KtrB. The single residues Gly70, Gly185, Gly290, and Gly402 from p-loops PA to PD of Vibrio alginolyticus KtrB were replaced with alanine, serine, or aspartate. The three alanine variants KtrBA70, KtrBA185, and KtrBA290 maintained a substantial activity in KtrAB-mediated K+ uptake in Escherichia coli. This activity was associated with a decrease in the affinity for K+ by 2 orders of magnitude, with little effect on Vmax. Minor activities were also observed for three other variants: KtrBA402, KtrBS70, and KtrBD185. With all of these variants, the property of Na+ dependence of K+ transport was preserved. Only the four serine variants mediated Na+ uptake, and these variants differed considerably in their K+/Na+ selectivity. Experiments on cloned ktrB in the pBAD18 vector showed that V. alginolyticus KtrB alone was still active in E. coli. It mediated Na+-independent, slow, high affinity, and mutation-specific K+ uptake as well as K+-independent Na+ uptake. These data demonstrate that KtrB contains a selectivity filter for K+ ions and that all four conserved p-loop glycine residues are part of this filter. They also indicate that the role of KtrA lies in conferring velocity and ion coupling to the Ktr complex.

Proteins from the superfamily of K ϩ transporters (termed SKT proteins) 3 are likely to have evolved from simple K ϩ channels of the M 1 PM 2 type (domain or subunit composed of two transmembrane helices (M 1 and M 2 ) connected by pore loop (p-loop) 4 P; prototype KcsA) (see Fig.  1A) (4,5) by multiple gene duplications and gene fusions (1)(2)(3). They occur in all types of organisms except animals and exert functions as diverse as Na ϩ and K ϩ transport (HKT1 from wheat and HKT2 from rice (Oryza sativa) (6 -8) and possibly subunit KtrB (NtpJ) from the bacterial KtrAB system (9, 10)), Na ϩ translocation (HKT1 from Arabidopsis thaliana (11) and HKT1 from rice (8)), putative H ϩ /K ϩ symport (Trk2 from fungi (12) and subunit TrkH from prokaryotic Trk systems (13)), K ϩ /K ϩ symport (Trk1 from fungi (14)), and K ϩ translocation by a P-type ATPase subunit (KdpA from prokaryotes (3)). All SKT proteins contain a 4-fold repeated M 1 PM 2 motif, in which the p-loop folds back from the external medium into the membrane (1)(2)(3)15). (See the model for Vibrio alginolyticus (Va) KtrB in Fig. 1B (1,2,9).) The (M 1 PM 2 ) 4 domain is predicted to cross the membrane eight times, which has been confirmed experimentally for three members of the SKT family (16 -18). The four p-loops from one SKT protein are thought to form a central K ϩ permeation pathway (2) with a structure resembling the K ϩ pore of K ϩ channel tetramers (supplemental Fig. 4) (5, 19 -22). With a few exceptions, each p-loop of the SKT proteins contains a conserved glycine residue at a position structurally equivalent to the central glycine residue of the selectivity filter sequence Thr-Val-Gly-Tyr-Gly from the KcsA channel monomers (see Fig. 1). In analogy to the situation in K ϩ channels (23), these four glycines of the SKT protein KtrB from V. alginolyticus were labeled selectivity filter residues when there was little evidence that they play a role in the specificity of K ϩ transport (9). Since then, it has been confirmed experimentally that several of these conserved glycine residues from Escherichia coli KdpA (16, 24 -27) and the glycine residue from the first p-loop of plant HKT proteins (28) determine cation selectivity.
Inspection of Fig. 1 shows that the four p-loops from VaKtrB exhibit little sequence identity to either the p-loop from KcsA or each other. Hence, the presented folding model for KtrB needs further experimental support. An important issue is the question whether each of the four conserved glycine residues in the putative p-loops P A to P D belongs to the selectivity filter, forming a narrow central K ϩ permeation pathway ("pore") through the membrane (1)(2)(3). Such data are not available for any SKT protein. Here, we report on experiments in which we systematically mutated these four conserved glycine residues in the SKT subunit KtrB of the K ϩ uptake system Ktr from the bacterium V. alginolyticus (29). A few years ago, Ktr was recognized as a separate type of K ϩ uptake system, functionally different from the related Trk system (1,2,29). Ktr is present in many bacteria (1, 29 -32) and appears to always require Na ϩ ions for activity (9,10,32). It is not known whether Na ϩ activates the K ϩ transport process (32) or whether Ktr functions as a K ϩ /Na ϩ symporter (10). VaKtr consists of two types of subunits. KtrA is a peripheral membrane protein presumably bound to the inner side of the cell membrane (29). It belongs to the KTN (K ϩ transport, nucleotide binding) protein/domain family (PFAM2254/COG0569; GOLD genomes online database) (33) and has been proposed to regulate K ϩ transport via KtrB by a change in its binding from NAD ϩ to NADH according to a "ligand-mediated conformational switch" mechanism (33). The second subunit, the integral membrane protein VaKtrB (called NtpJ in Enterococcus hirae) (10,34), is responsible for K ϩ transport across the membrane (2,29).
In this study, we mutated each of the four putative selectivity filter glycine residues in VaKtrB at positions 70, 185, 290, and 402 (see Fig. 1B) to alanine, aspartate, or serine. The effects of these changes on K ϩ and Na ϩ uptake by VaKtrAB were determined in E. coli mutants defective in the transport of either K ϩ or Na ϩ ions. A single change of any of these residues affected both the affinity of the transporter for K ϩ and its cation specificity, demonstrating that, in KtrB, all four conserved p-loop glycine residues are part of a K ϩ selectivity filter. Further studies in E. coli on VaKtrB without KtrA revealed an important role for KtrA in the substrate specificity and transport mechanism of the Ktr system.

EXPERIMENTAL PROCEDURES
Strains, Plasmids, and Growth Conditions-The strains and plasmids used in this study are listed in supplemental TABLE 1. Plasmids pNT11 to pNT13, pNT31 to pNT33, and pNT41 to pNT43 (containing mutations in codons 70, 402, and 185 of VaktrB, respectively) were generated from plasmid pKT84 (29) by PCR using the QuikChange mutagenesis kit (Stratagene, La Jolla, CA). Plasmid pEL308 is a derivative of pKT84, encoding C-terminally His-tagged VaktrB. It was constructed with a PCR forward primer containing the intrinsic NsiI site in front of VaktrA and a reverse primer extending the 3Ј-end of VaktrB with a His 6 codon sequence followed by a stop codon and a new AvaI site. The 2-kb PCR fragment was digested with NsiI and AvaI and ligated with the 4.7-kb fragment from pKT84, digested with the same enzymes. For the construction of plasmid pEL901, a SacI site in front of the ribosome-binding site of VaktrB and an SphI site immediately behind its stop codon were introduced by PCR. The product was trimmed with these enzymes and ligated with plasmid pBAD18 (35), digested with the same enzymes. Plasmid pEL903 is a construct similar to pEL901, but encodes VaKtrB with six histidine residues fused to its C-terminal end. Plasmid pEL905 is a pBAD18 derivative containing VaktrAB with a His 6 codon sequence attached to the 3Ј-end of ktrB. It was constructed with a PCR primer pair containing a KpnI site in front of ktrA and an SphI site behind ktrB-His 6 . The 2-kb PCR product was digested with KpnI and SphI and ligated with pBAD18, digested with the same enzymes. Plasmids pEL911 to pEL913 and pEL903-11 to pEL903-13 were generated from plasmids pEL901 and pEL903, respectively, by introducing mutations at codon 70 of ktrB.
Growth Tests at Different K ϩ Concentrations-For plate tests, single colonies of plasmid-containing LB2003 cells were grown on 1.5% agar plates containing K30 medium, 100 g/ml carbenicillin, and one of the carbon sources specified above. Cell material from these colonies was "patched" on a series of plates containing the same medium, but with a K ϩ concentration of 0.1, 0.3, 1, 2, 6, 10, 30, or 115 mM. These plates were incubated at 37°C, and growth was scored after ϳ16 h and checked after 24 and 40 h. Overnight growth tests in liquid minimal medium were done at the K ϩ concentrations listed above using a 1:300 inoculum of cells grown overnight in 30 mM K ϩ .
Depletion from Alkali Cations-Cells grown in liquid K30 medium were harvested during the late exponential phase and depleted of most of their K ϩ ions by a Tris/EDTA treatment (39). K ϩ -depleted, Na ϩloaded, plasmid-containing LB2003 cells were prepared by washing the EDTA-treated cells twice with 200 mM Na-Hepes (pH 7.5), followed by suspension of the cells at 10 -20 mg of cell dry weight/ml of the same medium (39). Cells depleted of K ϩ and Na ϩ (and loaded with triethanolamine) were prepared by washing the Tris/EDTA-treated cells two times with 200 mM Hepes brought to pH 7.3 with concentrated triethanolamine base (triethanolamine/Hepes), followed by suspension at 10 -20 mg of cell dry weight/ml of the same medium (9). Cells were shaken at 20°C before being used for the transport experiment within 2 h after their preparation.
Net K ϩ Uptake by K ϩ -depleted Cells-K ϩ -depleted, plasmid-containing E. coli LB2003 cells were suspended at 1 mg of cell dry weight/ml of 200 mM Na-Hepes (pH 7.3) and shaken at 20°C at 150 rpm. As an energy source, either 10 mM glucose or 20 mM glycerol plus 0.02% L-arabinose was added at t ϭ Ϫ10 min, followed by KCl at the concentrations indicated in the figures at t ϭ 0. At different time points, samples of 1.0 ml were transferred to a 1.5-ml polypropylene centrifuge tube containing 200 l of AR200 silicone oil (Serva, Heidelberg, Germany). The cells were immediately centrifuged for 2 min at 16,000 ϫ g through the oil in an Eppendorf 5414 D centrifuge. The pellet was cut off from the tube and transferred to a tube containing 1 ml of 5% (w/v) trichloroacetic acid. Cations from the cell pellet were released by vigorous vortexing, a freeze/thaw cycle, vortexing, and heating the suspension to ϳ95°C. A volume of 3 ml of 6.25 mM CsCl was mixed with the sample, and precipitated protein was removed by centrifugation. The K ϩ and Na ϩ contents of the supernatants were determined with an ELEX 6361 flame photometer (Eppendorf, Hamburg, Germany).
To determine the K m for K ϩ uptake by KtrB alone, K ϩ -depleted cells of LB2003/pEL901 were suspended at 0.67 mg/ml in Na-Hepes. At t ϭ 0, three of these suspensions received K ϩ concentrations varying from 22 to 130 M KCl, and the K ϩ concentration remaining in the medium after the centrifugation step was determined for 14 min (see Fig. 8B, E, ‚, and Ⅺ). Immediately afterward, a second assay was started in which the cells received K ϩ concentrations from 40 to 180 M at t ϭ 0 min (see Fig. 8B, •, OE, and f). K ϩ was determined in a mixture of 0.8 ml of the sample supernatant with 0.8 ml of a solution containing 2.5% (w/v) trichloroacetic acid and 10 mM CsCl. The Na ϩ dependence of K ϩ uptake by plasmid-containing E. coli LB2003 cells depleted of K ϩ and Na ϩ was determined with cells suspended in 200 mM triethanolamine/ Hepes (pH 7.3) as described (9). Na ϩ Uptake by Plasmid-containing TO114 Cells-This assay was carried out as described above for K ϩ uptake by E. coli LB2003 cells, except that TO114 cells depleted of K ϩ and Na ϩ were suspended in 200 mM triethanolamine/Hepes (pH 7.3), and then uptake was initiated by the addition of NaCl at t ϭ 0. Na ϩ was measured by flame photometry.
For this purpose, deproteinated cell extracts were prepared according to the protocol described above.
Detection of His-tagged KtrB in E. coli Membranes and Membrane Solubilizates-E. coli LB2003 cells containing plasmid pEL903 or one of its derivatives were grown in 200 ml of K30 medium in the presence of 20 mM glycerol and 0.02% (w/v) L-arabinose. After the suspension had reached OD 578 nm ϭ 1, the cells were isolated by centrifugation and washed once with and subsequently suspended in 10 ml of buffer containing 500 mM NaCl, 50 mM Tris-HCl (pH 7.0), and a 1:100 diluted protease inhibitor mixture for His-tagged proteins (Sigma, Taufkirchen, Germany). Cells were broken by ultrasonification for 3 ϫ 20 s (50% interval) in a Branson B15 cell disruptor using a medium sized tip. Unbroken cells and large cell fragments were removed by centrifugation at 15,000 ϫ g for 15 min. Subsequently, the supernatant was centrifuged in 3-ml portions for 30 min at 250,000 ϫ g in a Beckman TL100 table ultracentrifuge. The pellet (membrane fraction) was dissolved in 0.2 ml of buffer containing 50 mM Tris-HCl (pH 8.0), 600 mM NaCl, and 10% glycerol and stored frozen at Ϫ80°C until used. Proteins from the membrane fraction were solubilized by diluting membranes to 5 mg of protein/ml of the same buffer containing 0.5% (w/v) n-dodecyl ␤-D-maltoside (Ultrol, Merck Bioscience GmbH, Schwalbach/TS, Germany) and the protease inhibitor mixture. The suspension was stirred gently for 1 h at 4°C. Solubilized protein was separated from the residual membrane fraction by centrifugation for 30 min at 230,000 ϫ g in the table ultracentrifuge. The pellet fraction was suspended in the same volume of buffer as used for the solubilization assay. Equal volumes of the membrane fraction, solubilized protein fraction, and residual membrane fraction were subjected to SDS-PAGE. The proteins were blotted onto a nitrocellulose membrane and stained with Ponceau red. Subsequently, anti-pentahistidine monoclonal antibody (Qiagen Inc., Hilden, Germany) was bound to His-tagged KtrB on this membrane. The amount of antibody present on the membrane was detected with an alkaline phosphatase-conjugated secondary antibody.
Calculations and Presentation of Data-Transport parameters for the different types of cells were determined at least twice using different cell preparations. The data given are representative examples. Kinetic parameters for the uptake of K ϩ or Na ϩ by the cells were determined by plotting the data according to the Eadie-Hofstee method (see the sup-plemental material) (40) using the linear regression mode of the GraFit 5 program (Erithacus Software Ltd., Horley, UK). For the calculation of the Na ϩ and K ϩ contents of the cells (in nmol/mg of cell dry weight), we assumed that a cell suspension with OD 578 nm ϭ 1.0 in a Shimadzu UV-120-01 spectrophotometer contains 0.3 mg of cell dry weight/ml, that the external water space of the cells is 1.25 l/mg of cell dry weight for the Tris/EDTA-treated cells centrifuged through silicone oil (39), and that the time required for centrifugation of the majority of the cells through silicone oil is 0.2 min in an Eppendorf 5414 D centrifuge.
The pore structure of the high [K ϩ ] form of the KcsA channel (supplemental Fig. 4A) was generated with the aid of the program Swiss-PdbViewer (41) and changed by hand according to the front cover of Nature (November 1, 2001, Vol. 414) based on the detailed structure and function of the KcsA selectivity filter in that issue (19,20,42). A model for the selectivity filter structure of VaKtrB was generated from that of KcsA by replacing the appropriate amino acids in KcsA (supplemental Fig. 4 (B-D)) using the change residue (homology) option of PdbViewer.

Exchange of Glycine Residues in
VaKtrB-Each of the four conserved glycine residues Gly 70 , Gly 185 , Gly 290 , and Gly 402 from p-loops P A to P D of VaKtrB was replaced with alanine, aspartate, or serine ( Fig. 1), and the effect of these changes on alkali cation transport by the VaKtrAB system was analyzed in E. coli. We chose these substitutions because, of all amino acids, alanine resembles glycine most closely, and for other SKT proteins, the replacement of glycine with aspartate or serine has been reported to produce large effects on the K ϩ selectivity of the transport process (24,28).
Growth Tests-Growth of E. coli cells as a function of the K ϩ concentration in the medium can be used as a measure of the K ϩ uptake activity of these cells (11,32,37). We employed this technique either by growing cells in liquid minimal medium of defined K ϩ concentrations (for amino acid replacements in p-loops P A and P D ) (Fig. 2) or by patching cell material on agar plates containing the same medium (data for changes in all four p-loops) (TABLE ONE, fourth column). In strain LB2003, the four K ϩ uptake systems Kdp, TrkH, TrkG, and Kup from E. coli K12 strains are inactive (13,36). Because of this defect, LB2003 cells hardly In B, arrows pointing to the four putative selectivity filter glycine residues indicate the residues that were changed to alanine, aspartate, or serine. grow below 30 mM K ϩ . The empty vector pHG165 (43) used to clone VaktrAB genes (29) did not change this phenotype (Fig. 2 The three alanine variants KtrB A70 , KtrB A185 , and KtrB A290 permitted growth at K ϩ concentrations Ն0.1 mM, indicating that they transport K ϩ well ( Fig. 2A and TABLE ONE, Lines 3-5, respectively). By contrast, cells containing the KtrB A402 variant grew only at Ն10 mM K ϩ , suggesting that K ϩ uptake by this transporter is severely impaired (Fig. 2B, •; and TABLE ONE, Line 6). In addition, cells expressing any of the four aspartate variants were inactive at low K ϩ concentrations (growth at Ն6 -30 mM K ϩ ) (Fig. 2, f; and TABLE ONE, Lines 7-10). Finally, the corresponding serine variants (KtrB S70 , KtrB S185 , and KtrB S290 ) were partially active, with growth at K ϩ concentrations Ն0.3, 3, and 2 mM, respectively ( Fig. 2A, OE; and TABLE ONE, Lines 11-13).
K ϩ Uptake-More information on K ϩ transport can be obtained by measuring net K ϩ uptake by energized, K ϩ -depleted E. coli cells (37,39). For the KtrB position 290 variants (p-loop P C ), such data are already available (9), showing that, at 0.5 mM K ϩ , only the KtrB A290 variant was partially active in K ϩ uptake by the VaKtrAB system (Fig. 3C). Similar transport experiments were done with three other types of KtrB variants (Fig. 3, A, B, and D). The results show that, in decreasing order, the KtrB A185 , KtrB A70 , KtrB S70 , and KtrB D185 variants were still partially active in K ϩ uptake (Fig. 3, A and B) and that, under these conditions, the KtrB D70 variant and all three KtrB position 402 variants exhibited only very low K ϩ uptake activities (Fig. 3, A and D, respectively).
Kinetics of K ϩ Uptake-We then examined which change in the kinetic parameters for K ϩ uptake was responsible for the inhibition of K ϩ uptake by the variants. For the KtrB A290 variant, this effect was almost completely due to an ϳ100-fold increase in the K m for K ϩ (i.e. from ϳ25 M for the wild type to the millimolar range for the variant) (TABLE ONE, Lines 1 and 5, respectively) (9). The kinetic parameters for the other variants were determined as described in the supplemental material. Like KtrB A290 , KtrB A70 and KtrB A185 took up K ϩ with a K m of ϳ1 mM (i.e. a value 50 times higher than that for wild-type KtrAB). By contrast, their V max values (ϳ100 nmol of K ϩ taken up per min/mg of cell dry weight) were only about half that of the wild-type V max value  Growth characteristics and kinetic parameters for the uptake of K ؉ and Na ؉ by the VaKtrAB-containing KtrB variants Growth and K ϩ uptake were determined in plasmid-containing LB2003 cells; Na ϩ uptake was determined in plasmid-containing TO114 cells. WT, wild-type; ND, not determined (because no uptake was detected).

Line
Plasmid a K ϩ V max /Na ϩ V max is the ratio of the V max values for the transport of K ϩ and Na ϩ . For the latter, the background V max for strain TO114 of 110 nmol/min/mg of cell dry weight was substracted. For the cells for which a V max could not be determined (ND), the V max values for the transport of K ϩ and Na ϩ were estimated to be Յ20 and Յ40 nmol taken up per min/mg of cell dry weight, respectively. b K ϩ uptake was strongly biphasic; the slow phase had K m and V max values for K ϩ uptake of 47 mM and 48 nmol/min/mg, respectively.
( Fig. 4A, ‚, •, and E; and TABLE ONE, Lines 1 and 3-5). The cause of the poor growth complementation of cells expressing the KtrB A402 variant ( Fig. 2B and TABLE ONE, Line 6) was that both the K m for K ϩ increased to ϳ4 mM and the V max decreased to ϳ20 nmol of K ϩ taken up per min/mg of cell dry weight (Fig. 4B, f). Other variants for which a kinetic analysis of K ϩ uptake was possible were KtrB S70 and KtrB D185 (Fig. 4B, Ⅺ and छ, respectively). They transported K ϩ with a V max twice that of KtrB A402 and K m values for K ϩ of 2 and 8 mM, respectively. These results agree with those from the growth experiments ( Fig. 2 and TABLE ONE). From the growth and K ϩ transport experiments, we conclude that substitutions of the four putative selectivity filter glycine residues in KtrB are best tolerated in p-loops P A and P B , that even the minor change of any of these four glycine residues to alanine lead to an increase in the K m for K ϩ by a factor of at least 50, that poorer growthcomplementing variants exhibit also decreases in the V max for K ϩ by a factor of 10 or more, and that any substitution of the conserved glycine residue in p-loop P D is hardly tolerated.
Na ϩ Dependence of K ϩ Uptake-K ϩ transport via the Ktr system of V. alginolyticus depends on Na ϩ ions (9). This property was preserved in all of the still active KtrB variants tested. For KtrB A402 and KtrB A290 , this stimulation exhibited a relatively low affinity for Na ϩ (Fig. 5, ࡗ and OE, respectively). By contrast, for KtrB A70 , KtrB S70 , and KtrB A185 , Na ϩ affinity remained almost as high as that of the wild type (Fig. 5, •, f, , and Ⅺ, respectively). This result supports the notion that substitutions of the putative selectivity filter glycine residues in p-loops P A and P B are less severe for transport activity than they are in the C-terminal p-loops of the protein.
Intrinsic Na ϩ Uptake Activity of E. coli-When expressed in oocytes and yeast, several plant HKT proteins containing a serine residue instead of glycine in the first pore segment (P A ) transport Na ϩ rather than K ϩ (8,11,28,44,45). We wished to determine whether a similar phenomenon occurs for the VaKtrB serine variants. Net Na ϩ uptake can be determined in E. coli TO114, which lacks all three Na ϩ /H ϩ antiporters (38) and therefore cannot extrude Na ϩ ions well. A disadvantage of strain TO114 is that it contains an intrinsic, not well characterized Na ϩ uptake activity (32,38). Plasmid-containing TO114 cells were depleted of Na ϩ and most of their K ϩ by loading them with triethanolamine (9). The Na ϩ and K ϩ contents of these cells were Ͻ2 and ϳ20 nmol/mg of cell dry weight, respectively. Energized, triethanolamine-  . Kinetics of K ؉ uptake by putative selectivity filter glycine residue KtrB variants. Net K ϩ uptake via KtrAB by plasmid-containing LB2003 cells was determined at four K ϩ concentrations. The kinetics of initial uptake were analyzed according to the Eadie-Hofstee method as described in the supplemental material using cell K ϩ contents at two to four different time points per incubation. A, all data; B, data for variants that transported K ϩ poorly. •; plasmid pNT11, encoding KtrB A70 , K m for K ϩ ϭ 0.8 Ϯ 0.2 mM; E, plasmid pNT41, encoding KtrB A185 , K m for K ϩ ϭ 0.4 Ϯ 0.1 mM; ‚, plasmid pNT21, encoding KtrB A290 , K m for K ϩ ϭ 1.1 Ϯ 0.2 mM; f, plasmid pNT31, encoding KtrB A402 , K m for K ϩ ϭ 4.1 Ϯ 0.9 mM; Ⅺ, plasmid pNT13, encoding KtrB S70 , K m for K ϩ ϭ 1.8 Ϯ 0.2 mM; छ, plasmid pNT42, encoding KtrB D185 , K m for K ϩ ϭ 7.8 Ϯ 1.3 mM. V max values (in nanomoles of K ϩ taken up per min/mg of cell dry weight) are given by the intercepts of the straight lines with the ordinates. FIGURE 5. Na ؉ stimulation of K ؉ uptake by KtrAB and its putative selectivity filter KtrB variants. Energized, triethanolamine-loaded, VaktrAB plasmid-containing E. coli LB2003 cells were assayed for their Na ϩ dependence of K ϩ uptake as described under "Experimental Procedures." 2 mM KCl and 0 -10 mM NaCl were added at t ϭ 0 and 6 min, respectively. Initial rates of K ϩ uptake after NaCl addition at t ϭ 6 min are plotted as a function of the added NaCl concentration. Ⅺ, plasmid pKT84, encoding wild-type KtrAB; •; plasmid pNT11, encoding KtrB A70 ; , pNT41, encoding KtrB A185 ; OE, pNT21, encoding KtrB A290 ; ࡗ, pNT31, encoding KtrB A402 ; f, plasmid pNT13, encoding KtrB S70 . K ϩ uptake rates at 10 mM NaCl were taken as 100% for the different strains. They were 80, 32, 20, 50, 25, and 12 nmol of K ϩ taken up per min and mg of cell dry weight for the wild type, KtrB A70 , KtrB S70 , KtrB A185 , KtrB A290 , and KtrB A402 , respectively. loaded cells containing the empty vector pHG165 or pBAD18 took up Na ϩ in the absence of K ϩ ions (supplemental Fig. 2B). The kinetic parameters for the rapid phase of this Na ϩ uptake activity at 20°C and pH 7.3 were as follows: V max of ϳ110 nmol of Na ϩ taken up/min/mg of cell dry weight and K m for Na ϩ of ϳ6 -10 mM (TABLE ONE, Lines 2 and 16). We conclude that strain TO114 can be used for assaying heterologous Na ϩ uptake activity when this activity significantly exceeds that of this intrinsic E. coli Na ϩ uptake system.
All Four Serine Variants Transport Na ϩ -The Na ϩ uptake activity of KtrAB and its different variants was then determined. Whereas strains carrying wild-type KtrB or any of its Ala and Asp variants took up Na ϩ at a rate similar to that of cells carrying the empty vector (Fig. 6, A-D, Ⅺ, •, f, and E; and TABLE ONE, Lines 1 and 2, respectively), any of the four serine variants took up Na ϩ with an enhanced rate (Fig. 6, OE). Kinetic analysis showed that this stimulation of Na ϩ uptake was due to an increase in the V max (TABLE ONE, compare Lines 11-14 with Lines 1 and 2; and supplemental Fig. 2). The Na ϩ uptake activities of these four KtrB serine variants were all very similar. Because the different serine variants transported K ϩ with very different rates, these variants exhibited a variety of K ϩ /Na ϩ selectivities (TABLE ONE, last column).
VaKtrB Alone Transports K ϩ and Na ϩ -A major difference between bacterial Ktr systems and plant HKT proteins is that the latter transport alkali cations in heterologous systems in the absence of a KtrA-like subunit (6 -8, 11, 28, 44, 45). Up to now, VaKtrB has been considered to be inactive in the absence of other Ktr subunits (29,32). However, cloning of ktr genes from Bacillus subtilis (9,31) into pBAD vectors (35) and subsequent expression of these genes in E. coli showed that B. subtilis KtrB alone mediates slow K ϩ uptake in E. coli (48). To test whether VaKtrB has the same property, its gene was cloned into pBAD18 (35), yielding plasmid pEL901. With the pBAD vector, expression of cloned genes is induced by L-arabinose in the presence of a carbon source other than D-glucose (35). Growth in 0.2% (w/v) glycerol and 0.02% (w/v) L-arabinose, which was optimal, allowed strain LB2003/pEL901 to grow at Ն0.1 mM K ϩ (Fig. 7A, Ⅺ). In the absence of inducer or in the presence of inducer with D-glucose as the carbon source, this strain grew only at K ϩ concentrations Ն30 mM (Fig. 7B, Ⅺ; and data not shown, respectively). Growth at low K ϩ concentrations due to KtrB alone was mutation-specific because, in the presence of glycerol and L-arabinose, cells with the KtrB A70 , KtrB S70 , and KtrB D70 variants grew only at Ն3-10 mM K ϩ (Fig. 7A, •, f, and OE, respectively).
A first K ϩ uptake experiment showed that cells containing KtrB alone transported K ϩ slowly (Fig. 8A, Ⅺ) and that K ϩ uptake was barely detectable in non-induced cells or cells containing the KtrB A70 variant (छ and •, respectively). A subsequent experiment on the kinetics of K ϩ uptake by KtrB alone showed that the K m for K ϩ was as low as that of the complete VaKtrAB system (i.e. Յ25 M). The initial K ϩ uptake by KtrB alone remained linear with time for several minutes (Fig. 8, A and B). During K ϩ uptake at low K ϩ , the concentration of this cation could be determined very accurately in the medium (Fig. 8B). Hence, the kinetic parameters for K ϩ uptake were determined from the decrease with time in the K ϩ concentration in the medium, giving a K m for K ϩ of 16 Ϯ 3 M and a V max of 15 Ϯ 1 nmol of K ϩ taken up/min/mg of cell dry weight (Fig. 8B, inset). A second experiment gave similar results (data not shown). We conclude that KtrB alone preserves its high affinity for K ϩ . The data also suggest that KtrB alone transports K ϩ much more slowly than does the complete VaKtrAB system. To obtain more information on this point, we compared KtrB amounts in different KtrB-His 6 constructs by staining the protein with anti-pentahistidine monoclonal antibody (supplemental Fig. 3). The results showed that, in the membrane fraction of LB2003 cells, plasmid-encoded KtrB from KtrAB-His 6 in pKT84-like pEL308 was present at an ϳ8-fold smaller amount than was plasmid-encoded KtrB from either KtrB-His 6 or KtrAB-His 6 in the pBAD18 system in that fraction (strains LB2003/pEL903 and LB2003/ pEL905 grown in 0.02% (w/v) L-arabinose). Taken together, these results indicate that VaKtrAB transports K ϩ with a larger turnover number by 2 orders of magnitude compared with KtrB alone.  In contrast to the complete KtrAB system, KtrB alone transported Na ϩ (Fig. 8C). This property was neither preserved in any of the three KtrB position 70 variants nor due to the vector plasmid pBAD18 (Fig.  8C) and occurred in the absence of K ϩ ions. Na ϩ uptake by KtrB alone was characterized by a relatively high K m for Na ϩ (ϳ9 mM) and a high initial rate (V max ϳ 170 nmol/min/mg of cell dry weight) (TABLE ONE,  Line 15).
Additional studies on KtrB alone indicated that it had lost its Na ϩ dependence of K ϩ uptake (Fig. 8D). Concomitantly, the cells took up Na ϩ . At t ϭ 1.5 min (i.e. 2.5 min after the addition of 5 mM NaCl), the cells contained 110 nmol of Na ϩ /mg of cell dry weight. This amount decreased gradually with time to 50 nmol of Na ϩ /mg of cell dry weight at t ϭ 9.5 min (data not shown). It is unlikely that the absence of stimulation of K ϩ uptake by Na ϩ is due to the slow K ϩ transport rate mediated by KtrB alone because KtrAB with the KtrB A402 variant transported K ϩ at a similar rate and nevertheless did so in an Na ϩ -dependent manner (Fig. 5). Rather, the results on the uptake of K ϩ and Na ϩ shown in Fig. 8D demonstrate that KtrB alone transports these cations in an uncontrolled manner.
Finally and in contrast to the complete VaKtrAB system, cells expressing KtrB alone at 115 mM K ϩ in the medium tended to lose KtrB activity during multiple transfers. This effect did not occur in growth medium containing Յ30 mM K ϩ , supporting the notion that K ϩ transport via KtrB alone is uncontrolled and can lead to poisonous overloading of the cells with K ϩ .
A comparison of the K ϩ transport properties of VaKtrB with those of the complete VaKtrAB system indicates that the role of VaKtrA lies in conferring both a high rate and Na ϩ dependence to the KtrAB complex. In addition, the properties of KtrB alone, with its features of K ϩ as well as Na ϩ transport, are reminiscent of plant HKT systems with a glycine residue in p-loop P A (7,8,28,45).

VaKtrB and Its Position 70 Variants Are Made in Comparable Amounts and with Comparable Detergent Solubilities-
The significance of the results from the transport assays reported here relies on the assumption that equal amounts of native and variant transport proteins are present in the membrane and that they fold similarly in this phase. Our observation that many VaKtrAB systems with different KtrB variants show V max values for K ϩ transport similar to that of the wild type (TABLE ONE, fifth column) suggests but does not prove that these conditions are fulfilled. To obtain more information on this point, VaK-trB and its three position 70 variants were His-tagged at their C termini using the pBAD18 system. Growth and transport experiments demonstrated that His-tagged VaKtrB and its variants exhibited growth characteristics and K ϩ uptake characteristics similar to those of their non His-tagged counterparts (data not shown and Figs. 7 and 8A for the latter, respectively), making a comparison between these two types of VaKtrB proteins meaningful. Eadie-Hofstee plot of the kinetics of K ϩ uptake from the data in B. For this analysis, the differences in K ϩ concentrations in the medium between the following sampling times were taken at an average external K ϩ concentration as indicated in parentheses: f, t ϭ 2 and 8 min (120 M K ϩ ); Ⅺ, t ϭ 3.5 and 8 min (80 M K ϩ ); OE, t ϭ 2 and 5 min (55 M K ϩ ) and t ϭ 5 and 8 min (19 M K ϩ ); ‚, t ϭ 2 and 3.5 min (31 M K ϩ ) and t ϭ 3.5 and 5 min (15 M K ϩ ); •, t ϭ 2 and 3.5 min (11 M K ϩ ). C, Na ϩ uptake by plasmid-containing E. coli TO114 cells was determined as described in the legend to Fig. 6, except that cells were grown in the presence of 0.2% glycerol plus 0.02% L-arabinose, and cells were energized with these compounds in the transport assay at t ϭ Ϫ10 min. E, plasmid pBAD18 (vector); Ⅺ, plasmid pEL901, encoding wild-type VaKtrB; •, plasmid pEL911, encoding KtrB A70 ; f, plasmid pEL912, encoding KtrB D70 ; OE, plasmid pEL913, encoding KtrB S70 . D, K ϩ uptake by KtrB alone is independent of Na ϩ ions. LB2003/pEL901 cells were depleted of K ϩ and Na ϩ and assayed for Na ϩ dependence of K ϩ uptake in 200 mM triethanolamine/Hepes (pH 7.3). At t ϭ Ϫ1 and 0 min, NaCl and 1 mM KCl were added, respectively. Ⅺ, control (no NaCl added); •, 2 mM NaCl; f, 5 mM NaCl. FIGURE 8. KtrB alone transports K ؉ and Na ؉ . Cells were grown in 0.2% glycerol in the absence or presence of 0.02% L-arabinose and depleted of K ϩ (A and B) or both K ϩ and Na ϩ (C and D) as described under "Experimental Procedures." A, K ϩ uptake by plasmidcontaining E. coli LB2003 cells was determined as described in the legend to Fig. 3, except that, at t ϭ Ϫ10 min, cells were energized with 20 mM glycerol plus 0.02% L-arabinose. Ⅺ, plasmid pEL901, encoding wild-type VaKtrB, induced cells; छ, plasmid pEL901, non-induced control cells; •, plasmid pEL911, encoding KtrB A70 , induced cells. B, shown are the kinetics of K ϩ uptake by KtrB alone. Expression of ktrB by LB2003/pEL901 cells was induced with L-arabinose. K ϩ uptake was calculated from the K ϩ concentration remaining in the medium as described under "Experimental Procedures." At t ϭ 0 min, KCl was added to the suspension at the following concentrations: assay 1, 22 M (E), 67 Membranes were prepared from the different VaktrB-expressing cells, and we examined to what extent the nondenaturing detergent n-dodecyl ␤-D-maltoside solubilized the corresponding His-tagged KtrB proteins using anti-pentahistidine monoclonal antibody to detect the Western-blotted protein fractions (Fig. 9). The results show that comparable amounts of VaKtrB and its position 70 variants were present in the membrane fraction and that, for all four VaKtrB proteins, the majority was solubilized from the membrane fraction by 0.5% (w/v) n-dodecyl ␤-D-maltoside. Because this detergent does not solubilize aggregated or denatured membrane proteins, we conclude that the large differences in growth parameters and transport rates of cells containing the different VaKtrB variants are not caused either by different extents of insertion into the membrane or by denaturation or aggregation of variants.

DISCUSSION
Before structures of K ϩ channels were known, two criteria were developed to determine whether an amino acid residue functionally belongs to the K ϩ selectivity filter of a K ϩ channel. First, changes of the residue should lead to a decrease in the K ϩ /Na ϩ selectivity of the channel, and/or second, such a change should inhibit K ϩ channel activity (23). Subsequent high resolution structures of K ϩ channels confirmed that the K ϩ selectivity filter residues identified according to these criteria (23) are indeed part of such a filter (5, 19 -22, 42). The four VaKtrB glycine residues Gly 70 , Gly 185 , Gly 290 , and Gly 402 studied here are the only four residues universally conserved within the KtrB subfamily (1). Modeling based on the structure of the KcsA channel (5, 19) positions these four residues within a putative K ϩ selectivity filter ( Fig. 1 and supplemental Fig. 4) (1)(2)(3)9). It was the aim of this study to determine whether, according to the criteria of Ref. 23, these glycine residues exert this selectivity filter function. Our results confirm this notion because (i) the replacement of any of these glycine residues with serine led to a decrease in the K ϩ /Na ϩ selectivity of the system by a factor of up to at least 35 (Figs. 3,4, and 6 and TABLE ONE, last column), and (ii) the minor change of these residues to alanine either abolished K ϩ transport almost completely (residue 402) or reduced the affinity of these proteins for K ϩ by 2 orders of magnitude (residues 70, 185, and 290) (Figs. 2-4 and TABLE ONE). These results are compatible with (but do not prove) a model in which these residues from each of the four KtrB p-loops are symmetrically arranged around a central K ϩ pore (supplemental Fig. 4), as has been proposed by Durell et al. (1)(2)(3).
Previous work with strain TO114 showed that amine-loaded E. coli cells take up Na ϩ (32,38). We have extended these observations by showing that this Na ϩ uptake activity exhibited a high rate and low affinity ( Fig. 6 and TABLE ONE). Booth (46) predicted that E. coli must contain an electrogenic Na ϩ uptake system, and here we have identified such an activity for the first time. At present, it is not known whether the rapid Na ϩ uptake observed by us occurs via an Na ϩ channel or an Na ϩ transporter. An Na ϩ channel has been described for the bacterium Bacillus halodurans (47), but it is not known whether E. coli also contains such a channel.
Any Na ϩ uptake to be detected for a heterologously expressed protein in E. coli TO114 must occur with a rate that significantly exceeds that of the already high intrinsic Na ϩ uptake activity of these cells. Another disadvantage of the use of strain TO114 is that its Na ϩ uptake can be assayed only in the absence of K ϩ ions (49). Despite these limitations, we detected two types of Na ϩ transport activities for VaKtr: first, Na ϩ uptake by the complete KtrAB system containing any of the four KtrB serine variants (Fig. 6), and second, Na ϩ uptake via VaKtrB in the absence of VaKtrA (Fig. 8C). The latter is remarkable because such an activity is not observed for the complete VaKtrAB system. Compared with the latter, KtrB alone loses its Na ϩ dependence of K ϩ transport.
Hence, it appears that KtrB alone mediates only non-coupled K ϩ and Na ϩ fluxes, suggesting that one function of KtrA is to confer coupling of ion fluxes to the Ktr complex.
Previously, it was reported that the Ktr system (described as NtpJ) from E. hirae transports Na ϩ at pH 10 (10). Our result of Na ϩ transport by VaKtrB alone was obtained at pH 7.3, indicating that, in principle, KtrB also transports Na ϩ ions at neutral pH. Like VaKtrAB, the complete Ktr system from Synechocystis sp. PCC 6803 does not transport Na ϩ in E. coli TO114 (32). This observation has been taken to mean that the role of Na ϩ lies in activation of K ϩ transport by the Ktr system (32). However, our results both with KtrB alone (Fig. 8C) and with the KtrB serine variants (Fig. 6) as well as those with Ktr from E. hirae at pH 10 (10) all demonstrate that, in principle, Ktr can translocate Na ϩ ions. Because one role of KtrA appears to be coupling of K ϩ and Na ϩ fluxes in the complex (see above), one may speculate that, with the complete system, Na ϩ transport occurs only in the presence of K ϩ , a condition that (as outlined above) cannot be tested at present. Clearly, more work has to be done before it can be decided whether the function of Na ϩ in Ktr systems is to activate K ϩ transport (32) or whether these systems symport K ϩ with Na ϩ .
A comparison between K ϩ transport by the complete VaKtrAB system and that by VaKtrB alone allows a second conclusion about the function of KtrA. KtrB alone and the complete system exhibited the same high affinity for K ϩ , but had a different turnover number by 2 orders of magnitude (Fig. 8B, inset; and supplemental Fig. 3) (9), indicating that KtrA confers velocity to the system. Roosild et al. (33) have (non-induced controls), ktrB expression was induced with 0.02% (w/v) L-arabinose during growth. Samples containing equivalent cell amounts from the membrane fraction (M; lanes 2, 5, 8, 12, and 15), the solubilizate (S; lanes 3, 6, 9, 13, and 16), and the nonsolubilized remainder of the membrane fraction (P; lanes 4, 7, 10, 14, and 17) were subjected to SDS-PAGE, and the proteins were transferred to a nitrocellulose membrane. A, immunodetection of His-tagged KtrB with anti-pentahistidine monoclonal antibody; B, protein patterns on the nitrocellulose sheet as detected with Ponceau red. In B, lanes 1 and 11 show prestained standard proteins with relative molecular masses (top to bottom) of 122, 75, 47, 33, 24, and 20 kDa. In B, the expected position of KtrB is indicated by the arrow. WT, wild type.
proposed that KtrA-like proteins/domains modulate the activity of their corresponding transporters by a ligand-mediated conformational switch. Our results agree with such a mechanism in the sense that, in the absence of KtrA, KtrB transports K ϩ at a low rate.
Two results of this study demonstrate a similar function of bacterial Ktr and plant HKT systems. First, VaKtrB alone translocated both K ϩ and Na ϩ ions (Fig. 8), resembling the translocation properties of HKT proteins containing a glycine residue in p-loop P A (6 -8, 28). Second, in the complete VaKtrAB system, only the single serine versions of the selectivity filter variants translocated Na ϩ ions across the membrane. This agrees with observations that plant HKT proteins carrying a serine residue at the filter position in p-loop P A exhibit K ϩ -independent Na ϩ translocation in heterologous expression systems (8,11,28). These similarities in function of bacterial Ktr and plant HKT systems suggest that KtrB may serve as a model system for plant HKT proteins. Our results with KtrB predict that serine replacement of the selectivity filter glycine residues in p-loop P B , P C , or P D of HKT1 from wheat or HKT2 from rice will also lead to K ϩ -independent Na ϩ translocation.
With its one conserved glycine residue per p-loop, the selectivity filter of KtrB appears to be much simpler than that of the K ϩ channel KcsA. In the latter, the central pore is formed by the selectivity filter sequence Thr-Val-Gly-Tyr-Gly from its four subunits. The backbone carbonyl oxygens of these residues and the oxygen side chain atoms of the threonine residues form four K ϩ -binding sites within the pore (supplemental Fig. 4A) (19,20). At low K ϩ concentrations, either sites 1 and 3 or sites 2 and 4 are occupied by dehydrated K ϩ ions, whereas the two other sites contain a water molecule (19). Although experimental evidence is lacking in support of the following notion, a comparison of the selectivity filter sequence of KcsA with those of VaKtrB suggests that KtrB may contain only two K ϩ -binding sites, equivalent to sites 2 and 3 in KcsA (supplemental Fig. 4 (A-C). A feasible mechanism for the K ϩ transport process of KtrB would be that dehydrated K ϩ binds first to the external site (site 2 in KcsA) and that the internal site (site 3 in KcsA) contains a water molecule. In a second step, Na ϩ modulates the transfer of K ϩ from the external to the internal site, where K ϩ is released. Na ϩ translocation by KtrB may occur via either the central pore or another part of the protein. Our observation that the affinity for the Na ϩ dependence of K ϩ transport is sensitive only to changes in p-loop P C and P D residues Gly 290 and Gly 402 (Fig. 5) suggests that Na ϩ translocation occurs through this part of the protein, but other explanations are also possible. Finally, our observation that the four single serine variants all transported Na ϩ (Fig. 6) can be explained by assuming that the hydroxyl side chains of these single serine residues modulate the external binding site in the selectivity filter of KtrB (site 2 in KcsA) in such a way that, in comparison with K ϩ , the smaller Na ϩ ion fits better into this pocket (supplemental Fig. 4D).