Gain of Function Mutations in Membrane Region M2C2 of KtrB Open a Gate Controlling K+ Transport by the KtrAB System from Vibrio alginolyticus*

KtrB, the K+-translocating subunit of the Na+-dependent bacterial K+ uptake system KtrAB, consists of four M1PM2 domains, in which M1 and M2 are transmembrane helices and P indicates a p-loop that folds back from the external medium into the cell membrane. The transmembrane stretch M2C is, with its 40 residues, unusually long. It consists of three parts, the hydrophobic helices M2C1 and M2C3, which are connected by a nonhelical M2C2 region, containing conserved glycine, alanine, serine, threonine, and lysine residues. Several point mutations in M2C2 led to a huge gain of function of K+ uptake by KtrB from the bacterium Vibrio alginolyticus. This effect was exclusively due to an increase in Vmax for K+ transport. Na+ translocation by KtrB was not affected. Partial to complete deletions of M2C2 also led to enhanced Vmax values for K+ uptake via KtrB. However, several deletion variants also exhibited higher Km values for K+ uptake and at least one deletion variant, KtrBΔ326–328, also transported Na+ faster. The presence of KtrA did not suppress any of these effects. For the deletion variants, this was due to a diminished binding of KtrA to KtrB. PhoA studies indicated that M2C2 forms a flexible structure within the membrane allowing M2C3 to be directed either to the cytoplasm or (artificially) to the periplasm. These data are interpreted to mean (i) that region M2C2 forms a flexible gate controlling K+ translocation at the cytoplasmic side of KtrB, and (ii) that M2C2 is required for the interaction between KtrA and KtrB.

type, like KcsA (10,11) or Kir (12), by multiple gene duplications and gene fusions (7). Whereas the channels form homotetramers from four identical M 1 PM 2 subunits, SKT proteins consist of four covalently linked M 1 PM 2 motifs connected by cytoplasmic loops. The four p-loops (P) are thought to fold back from the external medium to the middle of the membrane, where they form a part of the permeation pathway for K ϩ through the channel center (7)(8)(9)13). Within each p-loop, most SKT proteins contain one conserved glycine residue, which is part of their K ϩ selectivity filter (9, 14 -17). With single conserved glycine residues in SKT proteins, this filter appears to have a simpler structure than in K ϩ channels, in which the filter is formed by the well conserved p-loop sequence TVGYG from each subunit (18).
The SKT-protein KtrB forms the K ϩ -translocating subunit of the Na ϩ -dependent K ϩ -uptake system KtrAB from bacteria (9, 14, 19 -22). KtrA, the other subunit from KtrAB, is located at the cytoplasmic side of the membrane and is a member of the RCK/KTN protein family (1,23). KtrA may regulate K ϩ transport by binding ATP (24,25). It confers velocity, Na ϩ dependence, and K ϩ selectivity to the complex (9). KtrB alone transports K ϩ slowly in a process that is independent of Na ϩ . In addition, it transports Na ϩ with relatively low affinity (K m value of ϳ3 mM Na ϩ (9)). The exact structure of KtrB is unknown, but it has been modeled based on the structure of KcsA (11,13). Most of the KtrB structure was similar to that of KcsA. However, in particular, the C termini from the membrane spans M 2C and M 2D deviated from that of KcsA-M 2 . This may reflect the difference in function between the channel KcsA and the transporter KtrB (13). Subsequent cross-linking studies showed that the external half of KtrB is very similar to that of the KcsA tetramer, whereas its cytoplasmic half deviates. In addition, KtrB may form dimers (26). In their modeling studies, Durell and Guy (13) focused on membrane span M 2C . They divided it into three regions, from M 2C1 to M 2C3 (see Fig. 1A). M 2C1 and M 2C3 can form hydrophobic ␣-helices. However, M 2C2 contains many conserved small and polar residues (Ala, Gly, and Ser, Thr, Lys, respectively; see Fig. 1B). It may form a random coil or ␤-turn structure (13). According to the first Durell and Guy model (13), M 2C1 and M 2C3 span the membrane, and M 2C2 forms a loop that fills the cavity just beneath the p-loops. Alternatively, these authors proposed (13) that M 2C1 and M 2C2 span the membrane (the latter in a coiled conformation) and helix M 2C3 is located within the inner surface of the membrane. The conserved lysine residue ( Lys 325 in VaKtrB from Vibrio alginolyticus; see vertical arrow in Fig. 1B) may form a salt bridge with a conserved aspartate residue from the C terminus of M 2B (VaKtrB Asp 222 ; 13). Region M 2C2 is conserved among the bacterial SKT protein families KtrB, TrkH, and KdpA (7,8,13). All of these require at least one additional subunit for activity. The interaction with the other subunit(s) (i.e. KtrA for KtrB) may modulate the conformation of M 2C2 , thereby allowing the passage of K ϩ through a gate (13).
In this study, we report on the effects of point mutations and deletions in region M 2C2 of KtrB from V. alginolyticus (VaKtrB) on the uptake of K ϩ and Na ϩ by Escherichia coli cells. Many of the changes cause a huge gain of function effect on K ϩ transport, whereas Na ϩ transport remained unaltered. In addition, KtrB-PhoA fusions suggest that region M 2C2 possesses a flexible structure. These results are interpreted to mean that M 2C2 forms a gate for K ϩ permeation through the KtrB protein.

EXPERIMENTAL PROCEDURES
Strains, Plasmids, and Growth Conditions-The strains and plasmids used in this study are listed in the supplemental Table. Plasmids pKtrB G314A to pKtrB S327C containing point mutations in codons from M 2C2 of VaktrB were generated from plasmid pEL903-100 by PCR using the QuikChange mutagenesis kit from Stratagene (La Jolla, CA). Plasmids with deletions in region M 2C of ktrB (pKtrB ⌬314 -328 to pKtrB ⌬326 -328 , and pKtrB ⌬314 -328 Leu 340 -PhoA to pKtrB ⌬326 -328 Leu 340 -PhoA) were generated with the Finnzymes Phusion TM site-directed mutagenesis kit from New England Biolabs (Frankfurt, Germany) from plasmids pEL900-100 and pKtrBL340-PhoA, respectively. KtrAB plasmids with point mutations or deletions in region M 2C2 from ktrB were generated by replacing a 1-kb BlpI-MscI restriction fragment from ktrB in plasmid pIH301 by the same fragment from a ktrB plasmid carrying the desired mutation. Plasmids encoding KtrB-PhoA fusion proteins were generated with the aid of plasmid pPAB404 as described in Ref. 27. For this purpose, ktrB fragments were generated by PCR from plasmid pEL903-100 with a BamHI site in front of the ktrB ribosomal-binding site and a PstI site at the proper position behind the ktrB codon that should be fused to the first phoA codon from pPAB404. The fragments were digested with BamHI and PstI and ligated with pPAB404 opened with the same restriction enzymes.
Cells of E. coli LB2003 (28) containing plasmids with the point mutations in ktrB were grown aerobically in medium K30 (9,29) with 0.2% glycerol as a carbon source. At a value of OD 578 of 0.5 ktrB expression was induced by the addition of 0.02% L-arabinose. After a further 90 min of growth, cells were harvested by centrifugation and prepared for the K ϩ -uptake experiments (protocol 1). Most of the cells of LB2003 containing the M 2C -deletion plasmids were grown at 3 mM K ϩ (K3 medium) containing 0, 0.002, 0.005, or 0.02% L-arabinose. These cells were harvested by centrifugation at a value of OD 578 of 0.5-0.8 and prepared for the K ϩ -uptake experiments (protocol 2). Cells of strain TO114 (30) containing plasmids with the point mutations in ktrB were grown at 115 mM K ϩ according to protocol 1. Cells of strain TO114 containing the M 2C -deletion plasmids were grown aerobically in K3Na0 medium (31) containing 0.4 mM MgSO 4 , 7.6 mM (NH 4 ) 2 SO 4 , 6 M FeSO 4 , 20 mM (NH 4 ) 2 HPO 4 , 10 mM bistrispropane, and 120 mM choline chloride with 0.2% glycerol and 0, 0.002, 0.005, or 0.02% L-arabinose.
Growth tests were done in liquid medium at K ϩ concentrations of 1, 3, 10, 30, and 115 mM using a 1:300 inoculum of cells growing exponentially at 30 mM K ϩ in the absence of L-arabinose. Growth was measured as OD 578 for ϳ8 h and at ϳ24 h after inoculation. The L-arabinose concentrations used for the induction of ktrB are given in the figure legends.
Transport Experiments-Depletion of the cells from alkali cations, net K ϩ uptake by K ϩ -depleted LB2003 cells, and Na ϩ uptake by Na ϩ -depleted TO114 cells were carried out as described in Ref. 9.
Alkaline Phosphatase Reaction-Cells of strain CC181 (32) containing pPAB404-derived plasmids were grown in 5 ml of LB medium up to an OD 578 value of 0.5, harvested by centrifugation, washed with buffer containing 150 mM NaCl and 10 mM TrisCl, pH 8.0, and suspended in 0.5 ml of the same buffer. Alkaline phosphatase activity of the suspension was determined according to Ref. 33 and expressed in nmol⅐min Ϫ1 ⅐mg Ϫ1 cell (dry weight).
Overproduction and Purification of His 10 -KtrAKtrB complexes-Cells of strain LB2003 containing plasmid pIH301 or one of its derivatives were grown in 1 liter of K30 medium (29) in the presence of 0.2% glycerol (v/v) and 0.02% L-arabinose (w/v) up to an OD 518 value of 1.0 -1.5, harvested by centrifugation, washed once with buffer S containing 600 mM NaCl, 10% glycerol (w/v), and 50 mM TrisCl, pH 8, and resuspended at 10 mg (dry weight)/ml of the buffer S in the presence of 1 mM EDTA and some DNase. Cells were broken by sonication with a Branson 250 Sonifier II cell disruptor (Branson). Subsequently, the suspension was centrifuged for 15 min at 15,000 ϫ g, and its supernatant was centrifuged overnight at 100,000 ϫ g. The pellet was suspended at 10 mg protein/ml of buffer S, and its proteins were solubilized by incubation with 1% ␤-D-dodecylmaltoside (Anatrace) and 1% protease-inhibitor mixture P8849 (Sigma-Aldrich, Steinheim, Germany) for 1 h at 4°C and centrifuged for 45 min at 200,000 ϫ g in a TL100.3 Mini-ultracentrifuge (Beckmann, Frankfurt, Germany). The supernatant was incubated at a concentration of 150 mg of protein/ml packed nickel-nitrilotriacetic acid-agarose (Qiagen, Hilden, Germany) in the presence of 10 mM imidazole in a 10-ml polypropylene column for 1 h. Subsequently, the agarose was washed with 50 volumes of buffer W, containing 200 mM NaCl, 20 mM TrisCl, pH 8, 5 mM ␤-mercaptoethanol, 10% glycerol (w/v), 0.04% ␤-Ddodecylmaltoside (w/v) plus 50 mM imidazole. His-tagged protein was eluted with 3 volumes of buffer W containing 500 mM imidazole. The samples were diluted with an equal volume of twice concentrated sample buffer, containing 4% SDS (w/v), 12% glycerol (v/v), 50 mM TrisCl (pH 6.8), 2% ␤-mercaptoethanol (v/v), and 0.01% Serva blue G (w/v). The proteins in the sample were separated by SDS-PAGE (34) and stained with Coomassie Brilliant Blue.
Detection of Proteins with Antibodies-The amount of Histagged KtrB in cells used for the growth or transport experiments was determined on a Western blot with a monoclonal mouse penta-His antibody from Qiagen (Hilden, Germany) followed by incubation with ImmunoPure antibody goat antimouse IgG labeled with alkaline phosphatase (Pierce) as the secondary antibody and visualization by the reaction with alkaline phosphatase with 4-nitro blue tetrazolium chloride plus 5-bromo-4-chloro-3-indolyl phosphate. For this purpose, an aliquot of the cell suspension containing 50 g of cell protein was spun down and taken up in 50 l of sample buffer. The proteins in the sample were separated by SDS-polyacrylamide electrophoresis (34) and transferred to a nitro-cellulose sheet on which His-tagged KtrB was detected with the alkaline phosphatase reaction.
Alkaline phosphatase in cells used for PhoA activity tests was detected similarly using a monoclonal mouse anti-PhoA antibody (Caltag Laboratories GmbH) and ImmunoPure antibody goat anti-mouse IgG as a secondary antibody followed by visualization with the alkaline phosphatase reaction as described above.
Other Methods-Protein was determined according to Ref. 35. A cell suspension with an OD 578 value of 1.0 was taken to contain 0.3 mg (dry weight)/ml (36). Uptake of K ϩ and Na ϩ by the cells was calculated as described in Ref. 9. Kinetics of cation uptake was determined by the Eadie-Hofstee method (37).

Gain of Function Changes in
Region M 2C2 -Durell and Guy (13) have proposed that region M 2C2 plays an important role in the K ϩ -transport cycle of KtrB. To obtain more information about this region, we changed single amino acids in M 2C2 to cysteines in Cysless VaKtrB containing an N-terminal His tag, as encoded by plasmid pEL903-100 and its derivatives. However, several of these cysteine variants were inactive. This was not unexpected in view of the large degree of conservation of several amino acid residues in M 2C2 (Fig.  1B, residues with a gray to black background). To learn more about these residues, more conservative amino acid changes were engineered in VaKtrB residues at positions Gly 314 , Gly 316 , Gly 321 , Gly 322 , Gly 323 , Lys 325 , and Val 326 (Fig. 2E). Together with the variants KtrB A315C , KtrB S317C , KtrB T318C , KtrB T320C , KtrB I324C , and KtrB S327C , it was investigated whether these constructs influenced growth of E. coli strain LB2003 at low K ϩ concentrations. This strain lacks functional K ϩ uptake systems, and, as a consequence, it grows only at an external [K ϩ ] Ͼ10 mM (28). A plasmid encoding a functional K ϩ uptake system will lower this concentration to 0.1 mM K ϩ (14). In addition, we measured the net uptake of K ϩ and Na ϩ by these strains and determined the produced amounts of these His-tagged KtrB variants by Western blotting with a monoclonal anti-His 5 antibody.
KtrB K325H , and KtrB K325Q ). Fourth, constructs that neither complemented cell growth nor inhibited it at high K ϩ concentrations (variants KtrB G321C ,KtrB G323C , and KtrB G323S ). Western blots of KtrB samples taken from the growing cells showed that almost all of the variants were present at amounts similar to that of unaltered KtrB. Exceptions with significantly less to no KtrB protein were cells with the variants KtrB G314C and KtrB G323A . Both grew slowly at all K ϩ concentrations (probably due to lack of KtrB protein; Fig. 2E). Fig. 3, panels A-D, show time courses of net K ϩ uptake by some of the variants. Both in the absence and presence of the C90S exchange KtrB transported K ϩ relatively slowly (results not shown and Fig. 3A, respectively). By contrast, almost all of the KtrB variants transported K ϩ considerably faster than did KtrB (Fig. 3B for the KtrB K325C variant). For only two variants transport was somewhat faster than that of KtrB ( Fig. 3C for KtrB K325R and KtrB S325C ), and for two of the variants, KtrB G314A and KtrB T318C it was similar to that of KtrB (results not shown and Fig. 3D, respectively). The kinetic analysis of the transport data showed that His-tagged KtrB transported K ϩ with a 25-fold lower affinity than did KtrB without this tag (K m values for K ϩ uptake of ϳ0.5 and 0.02 mM K ϩ , respectively; Fig.  4B and (9)). More importantly, the stimulation of K ϩ uptake for the variants was due exclusively to an increase in V max , which increased from about 35 nmol⅐min Ϫ1 ⅐mg Ϫ1 (dry weight) of cells for KtrB (Fig. 3A and Fig. 4A, two first columns) to 200 nmol⅐min Ϫ1 ⅐mg Ϫ1 (dry weight) for the KtrB G316S variant (Fig.  4A). By contrast, changes in K m for K ϩ uptake by the different variants were relatively small (Fig. 4B). The observed effects on V max were not due to an increased amount of KtrB variant protein in the cells, as Western blots showed similar amounts of KtrB in cell suspensions used for the K ϩ -uptake experiments (results not shown).
KtrB also transports Na ϩ ions (9). Hence, we tested whether the KtrB variants overactive in K ϩ uptake also translocated Na ϩ faster. For this purpose, we used ktrB-plasmid containing E. coli TO114 cells depleted of both K ϩ and Na ϩ . This strain does not contain Na ϩ efflux systems (30). Hence, net Na ϩ uptake by these cells can be measured after the addition of NaCl to energized cells suspended in a triethanolamine-Hepes medium (9). Supplemental Fig. S1 shows time courses of Na ϩ uptake by KtrB and KtrB G316S -containing cells. The latter is the  A and B) and Na ؉ uptake (C and D) by cells containing KtrB variants with single amino acid changes. K ϩ was determined as outlined in the legend to Fig. 3. For Na ϩ uptake plasmid containing cells of strain TO114 were grown, induced for ktrB expression and depleted of K ϩ and Na ϩ as described under "Experimental Procedures" and in Ref. 9. For the Na ϩ uptake experiment, cells were suspended at 1 mg (dry weight)/ml of medium containing 200 mM triethanolamine Hepes, pH 7.2, 0.2% glycerol, and 0.02% L-arabinose. The suspension was shaken at room temperature. After 10 min, NaCl was added at 1.3 mM, 1.9 mM, 5 mM, or 7 mM. Cell samples of 1 ml were taken at different time points and analyzed for their Na ϩ contents by flame photometry. The nomenclature of the His-tagged KtrB variants with single amino acid changes was as in Fig. 2. most active KtrB variant with respect to K ϩ translocation (Fig.  4A). However, compared with WT-KtrB, the Na ϩ transport of this variant was not changed. The same was true for most of the other KtrB variants. Variant KtrB K325Q shows significantly slower Na ϩ transport than did KtrB (Fig. 4, panels C and D). From these data, we conclude that the gain of function of the different M 2C2 -KtrB variants with single amino acid changes is limited to its K ϩ transport feature. M 2C2 Forms a Flexible Linker within the Membrane-We obtained more information about the topology of M 2C with the technique of PhoA fusions (38). The fusion set behind the last of a cluster of three positive residues, Lys 344 gave a relatively low alkaline-phosphatase activity (Fig. 5, line 1), suggesting a cytoplasmic location of this residue. This result is compatible with the topology model of Fig. 1A and was expected on the basis of the "internally positive rule" for the topology of transmembrane stretches of von Heijne (39). However, an unexpectedly high PhoA activity was observed, when the fusion was set just N-terminal to this topological signal of positive residues, at positions Thr 338 , Phe 339 , or Leu 340 of VaKtrB (Fig. 5, lines 3-5). This result suggests that in these fusions, alkaline phosphatase faces the periplasm. These data indicate that M 2C has a flexible structure and can either cross the membrane or bend back toward the periplasm. A good candidate for the flexible part of M 2C is its middle region M 2C2 with its many glycine residues (Figs. 1 and 5). The result with the deletion of the complete M 2C2 region (residues 314 -328) from the construct in which PhoA is fused behind residue Leu 340 supported this notion because this protein exhibited a low PhoA activity (Fig. 5, line 6). In addition, deletion of the complete region M 2C3 (residues 329 -339) from the Leu 340 -PhoA protein gave the same result, indicating, that region M 2C2 itself cannot fold backward toward the periplasm (Fig. 5, line 7). By introducing smaller deletions into M 2C2 , we then investigated whether specific residues in this region are responsible for the folding back effect of M 2C toward the periplasm. The results argue against this notion, because either residues 314 -320 or residues 321-328 could be deleted without loss of the high PhoA activity of the Leu 340 -PhoA protein (Fig. 5, lines 9 and 14, respectively). However, the slightly larger deletion of 318 -328 did lead to a loss of PhoA activity of the Leu 340 -PhoA construct, indicating that at least parts of M 2C2 have to be present for the folding-back effect.
From the cultures for which PhoA activity was determined samples were taken in parallel for the detection of the different KtrB-PhoA fusion proteins on a Western blot with anti-PhoA monoclonal antibodies. Supplemental Fig. S2A shows that regardless of the extent of PhoA activity of the constructs Thr 338 -PhoA, Phe 339 -PhoA, Leu 340 -PhoA, Lys 344 -PhoA, and the Leu 340 -PhoA deletion plasmids in Fig. 5, the fusion proteins were present in similar amounts, indicating that the differences in enzyme activity were not due to different amounts of the fusion proteins of the different constructs. However, constructs with a high PhoA activity often showed holo-PhoA as a degradation product from the KtrB-PhoA fusion protein. Apparently, a periplasmic protease splits the fusion proteins at a site close to the connection between KtrB and PhoA.
We tried to complement the KtrB-PhoA data of Fig. 5 with results from KtrB fusions with a protein with a high activity in the cytoplasm. For this purpose, we fused parts of KtrB with either LacZ (27) or green fluorescent protein (40). However, as outlined under supplemental "Results of LacZ and GFP Fusions," both methods gave unsatisfactory results because regardless of the expected location of the fusion site, a relatively high ␤-galactosidase or green fluorescent protein fluorescence was almost always observed due to complete proteolysis of the fusion proteins to holo-LacZ or holo-green fluorescent protein.
Gain of Function Deletions in Region M 2C2 -Because parts of M 2C2 could be deleted without changing the flexibility of this region (Fig. 5), we tested how these deletions affected cell growth and the transport of K ϩ and Na ϩ by the corresponding His-tagged KtrB variants. Remarkably, strains with a ktrB plasmid in which either region M 2C2 or region M 2C3 was deleted grew well at low K ϩ concentrations, and so did all of the other strains carrying plasmids with partial deletions of M 2C2 from ktrB. In addition, we observed that several of these ktrB-M 2C2 deletion plasmids grew at 3 mM K ϩ without induction of ktrB by added L-arabinose. Moreover, like many of the ktrB strains containing point mutations in single M 2C2 codons (Fig. 2D), several of the ⌬⌴ 2C2 strains showed inhibition of growth at high K ϩ concentrations after induction with 0.005 or 0.02% of L-arabi-nose. By contrast, a plasmid containing full-size KtrB did not grow at low K ϩ concentrations without induction, and its growth was not inhibited by high K ϩ concentrations upon full induction. Finally, strain LB2003/pKtrB ⌬326 -328 grew poorly at high L-arabinose concentrations (supplemental Fig. S3). This effect was much smaller in a medium without Na ϩ ions (i.e. the triethanolamine-based medium to grow strain TO114 (9). We will return to this effect below. The results from these growth experiments suggest (i), that at least some of these deletion plasmids encode overactive KtrB variants and (ii) that complementation of growth at low K ϩ concentrations of cells containing these plasmids is not limited by the preservation of the flexible region of M 2C2 in KtrB (Fig. 5). Fig. 6 gives a survey of the kinetics of net K ϩ uptake by K ϩ -depleted cells containing the different ktrB-deletion plasmids. For this purpose ktrB expression was induced with 0.002%, 0.005%, or 0.02% L-arabinose. In parallel experiments the amount of KtrB of the different cells was detected by Western blotting (Fig. 6, bottom). All of the deletion plasmids enhanced K ϩ uptake as indicated by the 5-to 8-fold larger V max values compared with that of KtrB itself. This effect was not due to enhanced synthesis of KtrB by the deletion plasmid-containing cells. To the contrary, most of these cells synthesized considerably less KtrB than did cells with the intact ktrB gene (Fig.  6, bottom). Hence, the effect of the deletions on V max appears to be larger than a factor of 5-8. Especially strains with deletions of ktrB codons 329 -339 (⌬M 2C3 ), 326 -328, and 324 -328 produced very little KtrB protein. Compared with KtrB, the two small deletions ⌬314 -320 and ⌬326 -328 hardly affected the K m values for net K ϩ uptake by the cells (values of 0.6, 0.8, and 0.6 mM, respectively). All of the other ⌬M 2C2 variants exhibited elevated K m values for K ϩ uptake with those of KtrB ⌬314 -323 FIGURE 6. Kinetics of K ؉ uptake by strains containing KtrB with deletions in region M 2C . Plasmid-containing cells of strain LB2003 were grown at 3 mM K ϩ according to protocol 2, except for strain LB2003/pEL903-100, which was grown at 10 mM K ϩ , and for strain LB2003/pKtrB ⌬326 -328 , which was grown at 30 mM K ϩ according to protocol 1. The L-arabinose concentration used is indicated in the figure. The bottom panel gives the relative amounts of KtrB variants present in the cell suspensions used for the uptake assays. Histagged KtrB variants labeled with KtrB, ⌬314 -328, ⌬329 -339, ⌬314 -317, ⌬314 -320, ⌬314 -323, ⌬326 -328, ⌬324 -328, and ⌬318 -329 are encoded by plasmid pEL903-100 and its derivatives pKtrB ⌬314 -328 , pKtrB ⌬329 -339 , pKtrB ⌬314 -317 , pKtrB ⌬314 -320 , pKtrB ⌬314 -323 , pKtrB ⌬326 -328 , pKtrB ⌬324 -328 , and pKtrB ⌬318 -329 , respectively. FIGURE 7. Na ؉ uptake by strains containing KtrB with deletions in region M 2C2 . Plasmid-containing cells of strain TO114 were grown at 3 mM K ϩ and 0.02% L-arabinose in the triethanolamine-Hepes based medium described under the "Experimental Procedures." Na ϩ uptake after the addition of 3 mM NaCl to the cell suspension was determined as outlined in the legend to Fig. 4. The ordinate gives the initial rate of Na ϩ uptake after the addition of NaCl to the cell suspension. The bottom panel gives the relative amounts of KtrB variants present in the different suspensions used for the uptake experiments. Labeling of the KtrB variants is as outlined in the legend to Fig. 6. and KtrB with the complete M 2C2 deletion (KtrB ⌬314 -328 ) around 5 mM. By contrast, the complete deletion of M 2C3 (KtrB ⌬329 -339 ) increased the K m value to only ϳ1 mM (Fig. 6).
The KtrB ⌬326 -328 Variant Transports Na ϩ Faster-Because deletions in M 2C2 and M 2C3 might also increase permeation pathways through KtrB for cations other than K ϩ , we measured Na ϩ -uptake by some of the strains (Fig. 7). For this purpose, we focused on strains with the complete deletions of M 2C2 and M 2C3 , as well as on those with the small deletions ⌬314 -320 and ⌬326 -328. The latter two transported K ϩ with relatively low K m values, similar to that of KtrB (Fig. 6). For KtrB ⌬326 -328 , Na ϩ transport was significantly faster than that of the KtrB control (Fig. 7). Kinetic analysis showed that this effect was due to a higher V max and a lower K m value than that of KtrB. For the other deletion variants, Na ϩ uptake was similar to or somewhat lower than that of KtrB (Fig. 7). However, all ktrB-deletion strains produced less KtrB than did cells with intact KtrB (Fig. 7,  bottom). Hence, it might be that also KtrB-deletion variants other than KtrB ⌬326 -328 transport Na ϩ faster than does KtrB. In conclusion, all of the deletions in M 2C2 and M 2C3 cause a gain of function of K ϩ transport via KtrB (Fig. 6), and the KtrB ⌬326 -328 variant transports Na ϩ also faster (Fig. 8).
KtrA Does Not Suppress the Gain of Function Phenotype Exerted by Point and Deletion Mutations in ktrB-Because one of the functions of KtrA is the control of the transport cycle via KtrB (9), we tested how the presence of KtrA affects the gain of function effects described above. For this purpose, we used the pIH301-encoded KtrAB construct, in which KtrA carries an N-terminal extension with 10 histidine residues. In this plasmid, we engineered the ktrB G316S , ktrB T318C , and ktrB K325Q mutations as well as the ⌬314 -328 (⌬M 2C2 ), ⌬318 -329, and ⌬326 -328 deletions. In growth complementation tests with strain LB2003, the presence of KtrA did not influence the effect of growth of any of these changes observed for KtrB alone, indicating that KtrA does not suppress the effects of these mutations and deletions in region M 2C2 from KtrB. For the changes G316S, T318C, and K325Q and the deletion ⌬314 -328 (⌬M 2C2 ), we tested net K ϩ uptake by K ϩ -depleted LB2003 cells and came to the same conclusion, i.e. that we did not observe an effect of KtrA on K ϩ transport by the KtrB variants (data not shown). M 2C2 Deletions Abolish the Binding of KtrB to His-tagged KtrA-We attempted to purify KtrAB complexes with the following changes in KtrB: G316S, K325Q, ⌬314 -328 (⌬M 2C2 ), ⌬318 -329, and ⌬326 -328 by affinity chromatography on nickel-nitrilotriacetic acid-agarose. Whereas the two point mutants yielded normal complexes, the three KtrB-deletion variants did not bind to His-tagged KtrA on the affinity column (Fig. 8, and results not shown for the His 10 -KtrAB K325Q variant). We conclude that at least the KtrB-M 2C2 residues 326 -328 are essential for the interaction between KtrA and KtrB.

DISCUSSION
The transmembrane stretch M 2C from KtrB is, with its 40 amino acid residues, much longer than most transmembrane helices. It is difficult to predict a structure for its middle part M 2C2 with its conserved Gly, Ala, Ser, Thr, and Lys residues (Fig. 1B). Durell and Guy (13) have proposed that M 2C2 forms either a loop within the membrane connecting the apolar membrane spanning stretches M 2C1 and M 2C3 with each other, or that M 2C1 and M 2C2 span the membrane together (the latter in a coiled conformation) and that helix M 2C3 is located within the inner surface of the membrane. Our results do not allow us to distinguish between the two models. Because of the observed flexibility of M 2C2 (Fig. 5), we favor the first model. We assume that a flexible M 2C2 loop fills the cavity just below the p-loop and that VaKtrB M 2C2 -residue Lys 325 forms a salt bridge with VaKtrB residue Asp 222 from membrane span M 2B (see the model in Fig. 9) (13). By electric repulsion, K ϩ may weaken this salt bridge during the transport cycle and thereby open the M 2C2 gate, allowing K ϩ to permeate to the cytoplasm (Fig. 9). The gain of function mutations characterized in this article (Figs. 2-4 and 6) are proposed to change the structure of M 2C2 in such a manner that K ϩ can slip through the gate without having to open it. With the point mutations, this slip-through effect was specific for K ϩ , whereas with at least some deletions (in particular that of VaKtrB ⌬326 -328 ) the rate of Na ϩ transport through KtrB increased also (Fig. 7). The latter was not unexpected, since even small deletions may change the structure of a protein fundamentally and thereby alter its substrate (in the case of KtrB, transport) specificity. More importantly, either M 2C2 or M 2C3 could be deleted without KtrB losing its K ϩ transport function. Apparently, both regions are not essential. We assume that the K ϩ transport feature of KtrB is determined by or close at its selectivity filter at the periplasmic half of the protein, and that like in KcsA (11), most of its cytoplasmic half functions as a K ϩ pore. The removal of M 2C2 or M 2C3 does not change the pore structure in such a manner that K ϩ cannot permeate through it any longer.
Only those SKT proteins that require additional subunits for activity contain an M 2C2 region. Because KtrA controls the K ϩ transport cycle in Ktr systems via KtrB (9), we determined whether this control by KtrA still exists for the KtrB-M 2C2 variants. For both the point mutations and deletions in M 2C2 , KtrA did not suppress the effect of these mutations (see "Results"), indicating that in these situations KtrA lost its control of the K ϩ transport cycle. With the deletion variants, the explanation for this effect is simple because KtrA and KtrB did not interact with each other (Fig. 8), suggesting that the region M 2C2 is required for the binding of the two subunits to each other. Some of the KtrB variants with a single amino acid change in M 2C2 still formed KtrAB complexes (Fig. 8). In these constructs, the point mutations appear to open the M 2C2 gate permanently without KtrA being able to close it during the transport cycle. In this respect, it is important to establish whether the observed gain of function effects render K ϩ transport via the KtrB variants maximally equal to that of KtrB in the KtrAB complex or that K ϩ transport by the variants exceeds that of the latter. KtrB alone and KtrAB transport K ϩ with V max values of ϳ25 and 200 nmol⅐min Ϫ1 ⅐mg Ϫ1 of cell (dry weight), respectively (9). The V max values of some of the KtrB variants were significantly higher than that of KtrAB (Figs. 4 and 6). Moreover, several of the KtrB-deletion variants produce considerably less KtrB protein than does unaltered KtrB (Fig. 6). Finally, many of these KtrB variants transport K ϩ with considerable rates without induction of ktrB transcription by L-arabinose, a condition at which we could not detect any KtrB in the cells at all (Results not shown). All of these results suggest that several of the variant KtrB proteins are more active in K ϩ transport than is unaltered KtrB in the KtrAB complex. However it is difficult to quantify this effect. Comparison of the amounts of either unaltered KtrB or of the KtrB-deletion variants in the immuno-blots at the different L-arabinose concentrations with the observed V max values for K ϩ transport by these cells (Fig. 6) show that transport levels off at sub-maximal amounts of KtrB. In addition, in attempts to isolate some of the deletion variant KtrB proteins we observed that in contrast to unaltered KtrB protein they have a strong tendency to aggregate. When this also occurs within the cell membrane, the amount of active KtrB will be lower than expected on the basis of the immuno-blot data, increasing the gain of function effect per active KtrB molecule.
When then the activity of some of the KtrB variants exceeds that of KtrB in the KtrAB complex, one might expect that the loss of control due to the release of a gate for K ϩ transport within KtrB would give the protein K ϩ channel like properties. However, the K m values for K ϩ transport of almost all of the his-tagged KtrB variants did not increase in comparison to that of unaltered his-tagged KtrB (Fig. 4B). In addition, only in a few of the deletion mutants these K m values became with 5 mM significantly higher than the 0.7 mM of wild type (Fig. 6). If KtrB were to become a K ϩ channel, one would expect K ϩ -translocation to be controlled by diffusion and as a consequence to occur with very low affinity. Since this was not observed, we conclude that the changes introduced into M 2C2 do not transform KtrB into a K ϩ channel, but rather give it some channel-like properties.
Gadsby discusses several examples of similar transitions between (active) transporters and channels (41,42), in particular for the Cl Ϫ transporter from E. coli (43). He depicts transporters with two gates, one at each side of the membrane, controlling the uptake and release of the transported compound at the two membrane interfaces. Our conclusion that the changes in M 2C2 do not transform KtrB into a K ϩ channel may suggest that these changes open only one of the two gates in KtrB, presumably the one at the cytoplasmic side of the membrane. At present no information is available, whether KtrB also contains a gate within its periplasmic half and how such a gate controls the K ϩ -transport cycle.
The function of Na ϩ for K ϩ transport via the KtrAB complex is unknown. Since KtrB alone translocates Na ϩ (9), KtrAB might be a Na ϩ /K ϩ -symporter. However, no direct evidence exists in support of this notion. Our results suggest that the permeability pathway of the cations K ϩ and Na ϩ through KtrB differ from each other. This is concluded from the fact that the variants with the point mutations in M 2C2 were specific for the increase of K ϩ transport with no effect on that of Na ϩ (Figs. 3 and 4). Hence, these mutations cause subtle changes in the KtrB structure leading to the opening of a gate for K ϩ , but not for Na ϩ . This notion appears to contradict a model according to which K ϩ and Na ϩ file behind each other through the same central pathway of KtrB during the transport cycle. The small KtrB ⌬326 -328 -deletion caused a significant stimulation of the transport of both K ϩ and Na ϩ (Figs. 6 and 7). Hence this part of M 2C2 may play a role in Na ϩ transport via KtrB as well.
Our fusion studies of KtrB with alkaline phosphatase PhoA showed that M 2C2 possesses a flexible structure either allowing M 2C3 to reach the cytoplasm or to bend back artificially toward the periplasm (Fig. 5 and supplemental Fig. S2B). Such a flexible structure of M 2C2 is in accordance with both Durell and Guy models (13). The PhoA technique is normally used to determine whether a residue of an integral membrane protein is located at the inner or outer side of the bacterial membrane (38). It will have to be determined whether this technique can also be used for the determination of flexible structures within the membrane span of proteins other than KtrB. In conclusion, region M 2C2 from membrane span M 2C of KtrB forms a flexible FIGURE 9. Model explaining how M 2C2 functions as a gate for K ؉ transport through KtrB. Region M 2C2 is proposed to form a gate, which in its closed form prevents K ϩ translocation from the selectivity filter region (indicated by the two glycine residues) to the cytoplasm. A salt bridge between VaKtrB-M 2C2 residue Lys 325 and Asp 222 from membrane span M 2B (13) will keep the gate closed. K ϩ permeates from the periplasm into KtrB, where it is dehydrated to the enter the selectivity filter region (11). From this position, K ϩ opens the gate by electrostatic repulsion of residue Lys 325 , leading to a weakening of the Asp 222 -Lys 325 salt bridge and thereby allowing K ϩ to move to the cytoplasm, followed by closing of the M 2C2 gate. The gain of function amino acid changes in M 2C2 allow K ϩ to slip through the gate. The presence of a second gate for K ϩ at the periplasmic site of KtrB (42) and a role for Na ϩ in the K ϩ transport cycle are not included in this simplified model. structure within the membrane that functions as a gate for K ϩ translocation via the KtrB protein of V. alginolyticus.