Probing the Structure of the Dimeric KtrB Membrane Protein*

The KtrAB ion transporter is a complex of two proteins, KtrB and KtrA. The integral membrane protein KtrB is expected to adopt the structural architecture typified by the pore domain of potassium channels. Here we show that homo-dimerization of KtrB proteins is most likely a general property of this family of transporters. Using cysteine mutants and bifunctional cross-linkers we define regions of the Bacillus subtilis KtrB molecule that are close to the molecular 2-fold axis and to the dimer interface. Fitting of the cross-linking data to a potassium channel-like model suggests structural similarities between potassium channels and KtrB proteins in the extracellular half of the molecule and differences in the cytoplasmic regions.

The KtrAB ion transporter is a complex of two proteins, KtrB and KtrA. The integral membrane protein KtrB is expected to adopt the structural architecture typified by the pore domain of potassium channels. Here we show that homo-dimerization of KtrB proteins is most likely a general property of this family of transporters. Using cysteine mutants and bifunctional cross-linkers we define regions of the Bacillus subtilis KtrB molecule that are close to the molecular 2-fold axis and to the dimer interface. Fitting of the cross-linking data to a potassium channel-like model suggests structural similarities between potassium channels and KtrB proteins in the extracellular half of the molecule and differences in the cytoplasmic regions.
Regulation of intracellular concentrations of sodium and potassium ions is a characteristic of all living organisms and is achieved by a battery of macromolecules that include ion transporter membrane proteins. The KtrAB transporter family plays a role in these processes by mediating Na ϩ -dependent potassium ion uptake in many prokaryotes (1)(2)(3). At the genetic level this transporter is usually arranged as a two gene operon (3) encoding the KtrB membrane protein and the cytoplasmic KtrA protein. KtrB proteins are 400 -600 residues long and predicted to have 8 transmembrane helices (TMs) 2 (1)(2)(3). KtrA proteins are ϳ220 amino acids long and contain RCK domains that form octameric rings and bind adenine-ribose-phosphatecontaining compounds (4,5). RCK domains are commonly found associated with membrane proteins, including potassium channels, and function as activity regulators.
The KtrAB transporters are part of a large superfamily of ion transport membrane proteins that also includes the plant HKT, the yeast Trk1,2, and the prokaryotic TrK (1,3). Some of the members of the superfamily, including KtrAB, are thought to be either Na ϩ /K ϩ or H ϩ /K ϩ symporters (1,3,6). Sequence homology and hydrophobicity profiles have led to the proposal that all membrane proteins in this superfamily adopt the architecture of the ion pore of potassium channels (1,2). One KtrB polypeptide is thought to contain four repeats of the TM-P loop -TM structural motif observed in potassium channels, such as KcsA (7,8). The repeats, labeled A to D in Fig. 1A, do not have the same sequence, except for a few conserved positions (Fig. 1B), and are connected by relatively large cytoplasmic loops (15-36 residues), which show no similarity to one another. The experimentally determined transmembrane topology of HKT1 from Arabidopsis thaliana revealed 8 TMs that agree well with a KcsA-like architecture (9). Moreover, members of the superfamily tend to have one or more conserved glycine residues in the regions equivalent to the potassium channel selectivity filter sequence GYG (glycine-tyrosineglycine) (1, 2) (Fig. 1B). Mutation of these conserved glycines in either the KtrB or HKT proteins has been shown to cause changes in the transport properties as expected if these residues form part of the transport pathway (2,10,11). Although the evidence described above is compelling, to our knowledge there have been no published biochemical or biophysical studies that directly probe the three-dimensional properties of these proteins, a gap that has limited our understanding of this superfamily.
Recently we provided evidence that the KtrB membrane protein from Bacillus subtilis is a dimer in the membrane and in detergent, that the KtrA regulatory protein is an octamer able to adopt different shapes, and that the transporter complex consists of a KtrB dimer bound to a KtrA octamer (4). In that work, light scattering studies were used to determine the dimeric state of detergent-solubilized KtrB and one amino group cross-linker DSP (dithiobis(succinimidyl propionate)) was utilized to confirm that KtrB was a dimer in the membrane. Here we report substantially expanded cross-linking studies on KtrB with the aim of determining whether the structure is consistent, as has been hypothesized, with the architecture observed in potassium channels. We use several different crosslinkers to confirm the dimer state in B. subtilis and test whether the homo-dimeric organization appears to be a general property of membrane proteins in the KtrAB transporter family. In addition, bifunctional cross-linkers of various lengths are used in conjunction with a series of single-site cysteine mutants to evaluate the relative distances across the dimer interface.
We conclude that the observed cross-linking patterns for the extracellular protein regions fit well with a model of two sideby-side KcsA potassium channel molecules; however, data from the cytoplasmic regions reveal structural differences between KtrB and potassium channels.

EXPERIMENTAL PROCEDURES
KtrB Cloning-KtrB genes from B. subtilis (GI:2635594), Bacillus halodurans (GI:15613161), and Streptococcus pneumoniae (GI:66876889) and KtrA from B. subtilis (GI:2635593) were defined by homology to Vibrio alginolyticus KtrB and KtrA (3). The genes were synthesized from genomic DNA using PCR and cloned into the pET24d expression vector (kanamycin resistance), which adds a C-terminal His tag. KtrB from B. subtilis was also cloned into pET24d with a stop codon before the His tag sequence. C-terminally truncated KtrBs (ending at 435 and 430, and 10 and 15 residue truncations, respectively) were cloned into pET15b (ampicillin resistance) with an N-terminal His tag. Tagging either terminus does not alter the oligomeric state of the molecule. A dicistronic vector pKCeAB containing KtrB and KtrA from B. subtilis was constructed based on the pccKAN vector (ampicillin resistance) (12) for constitutive low level protein expression. QuikChange (Stratagene) was used to introduce single cysteine mutations in the B. subtilis KtrB gene (construct without a tag).
Expression and Purification-Membrane proteins were expressed in the Escherichia coli strain C41(DE3) (13) at 37°C after a 1.5-h induction with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside. Cell pellets from 1 liter of growth were resuspended in 12.5 ml of buffer (50 mM Tris, pH 7.5, 120 mM NaCl, 30 mM KCl, protease inhibitors); after cell lysis protein was extracted with 40 mM dodecyl maltoside, followed by a 45-min spin at 34,500 ϫ g. The supernatant was loaded onto 1 ml of Talon resin cobalt-agarose beads (Clontech) pre-washed with 20 mM Tris, pH 7.5, 120 mM NaCl, 30 mM KCl, 1 mM dodecyl maltoside; loaded beads were washed with same buffer plus 20 mM imidazole. Protein was eluted with the same buffer plus 150 mM imidazole, and the tag was removed by overnight incubation with thrombin while dialyzing against the buffer below.
Radioactivity Uptake Assay-1.1 mg of E. coli polar lipids (Avanti) solubilized in chloroform were dried under a nitrogen stream, resuspended in pentane, and dried again. The lipids were then resuspended in 110 l of high potassium buffer (450 mM KCl, 10 mM Hepes, 4 mM N-methylglucamine, pH 7.4) by bath sonication. The detergent CHAPS was added to a final concentration of 34 mM, and the bath was sonicated and left to stand for 2 h for complete solubilization. Protein (5-10 g) solubilized in dodecyl maltoside was added, and the mixture was incubated for 5 min at room temperature. To form liposomes the mixture was passed through a spin column with 1.5 ml of G-50 beads swollen in high potassium buffer. Liposomes were subsequently passed through a spin column with 1.5-ml G-50 beads swollen with a low potassium buffer (450 mM sorbitol, 10 mM Hepes, 4 mM N-methylglucamine, pH 7.4). The assay was initiated by mixing 80 l of liposome preparation to 100 l of radioactive tracer (50 M KCl in low potassium buffer plus 86 Rb ϩ (at ϳ2000 counts per microliter)). Time points were taken by loading 70 l of reaction mixture into a 1.5-ml Dowex cation exchange column, pre-washed with 2 ml of a solution containing 6% sucrose and 5 mg/ml bovine serum albumin, and equilibrated with 2 ml of 6% sucrose solution; finally, the Dowex matrix was washed with 1.5 ml of 6% sucrose solution. The eluate was collected in a scintillation vial and mixed with 15 ml of scintillation fluid. At the end of each experiment, valinomycin (a potassium and rubidium ionophore) was added to the remainder of liposome preparations, and the mixture was incubated for a couple of minutes; 70 l were then loaded into the Dowex matrix as described above. Normalization of rubidium uptake was done by dividing protein-mediated uptake by total uptake from valinomycin treatment. Polyclonal Antibodies-Polyclonal antibodies against B. subtilis KtrB were obtained by injecting rabbits with purified protein. Serum was collected and pre-purified by incubation with an acetone E. coli powder for 30 min at 4°C; this procedure removes some of the contaminant antibodies that target E. coli proteins. Antibodies were affinity-purified in the presence of detergent on a KtrB-coupled AminoLink Plus agarose gel (Pierce), prepared according to manufacturer instructions.
Membrane Preparations and Cross-linking Experiments-A cell pellet from 100 ml of E. coli culture expressing KtrB was resuspended in buffer and lysed by two passages through an Emulsiflex-C5 cell-cracker (Avestin). Membranes were prepared after a low speed 10-min spin at 10,000 rpm (rotor SS34), 4°C followed by a high speed spin at 50,000 rpm (rotor Ti-70) for 1.5 h at 4°C. Pellets were resuspended in 1 ml of buffer, and total protein concentration was determined by a colorimetric assay (Bio-Rad).
For non-directed cross-linking experiments the proteins were expressed with a C-terminal His tag. The buffer used for membrane preparation was phosphate-buffered saline, pH 7.5. The suspension was diluted to 0.5 mg/ml total protein concentration, cross-linkers (stock at 25 mM in Me 2 SO) were added to membrane aliquots, and the mixture was incubated at room temperature for 30 min. The amino group-reactive cross-linkers used were: disuccinimidyl glutarate (DSG), DSP, 3,3Ј-dithiobis(sulfosuccinimidyl propionate) (DTSSP), and ethyleneglycol bis(succinimidylsuccinate) (EGS) (Pierce). Reactions were stopped by addition of Tris buffer solution (pH 7.5) to a final concentration of 100 mM and incubated for 15 min. Samples were run in a SDS-PAGE, blotted onto polyvinylidene difluoride membrane, and probed with an anti-His tag monoclonal antibody (Qiagen).
For directed cross-linking experiments KtrB cysteine mutants were expressed without a His tag. The buffer used in these experiments was 20 mM Tris, pH 7.5, 150 mM NaCl, 5 mM EDTA. Membrane preparations were resuspended at 2 mg/ml total protein concentration. Methanethiosulfonate (MTS) bifunctional cross-linkers: 1,3-propanediyl bismethanethiosulfonate (M3M), 1,5-pentanediyl bismethanethiosulfonate (M5M), and 3,6,9,12,15-pentaoxaheptadecane-1,17-diyl bismethanethiosulfonate (M17M), from Toronto Research Chemicals Inc., were dissolved in Me 2 SO at 5 mM and added to the membranes at 50 M concentration. After 1-h incubation, at room temperature or on ice, reactions were stopped by addition of N-ethylmaleimide (NEM) to 10 mM. Stop reagent efficiency was tested by adding 10 mM NEM and M17M crosslinker at roughly the same time to the samples for 1 h. Samples were run via SDS-PAGE, blotted onto polyvinylidene difluoride membrane, and probed with anti-KtrB polyclonal antibody. Before probing the KtrB membrane blot, polyclonal antibodies were further purified by 1-h incubation with a preparation of E. coli membranes not containing KtrB protein blotted on a membrane. Molecular weight standards used in Western blots were MagicMark XP (Invitrogen).
Complementation Assay-Wild-type KtrA and KtrB C-terminal mutants (or wild type) were generated in the pKCeAB dicistronic vector. E. coli TK2420 strain (a gift from W. Epstein, University of Chicago) was transformed with the dicistronic vector and plated onto LBK agar plates (10 g/liter Tryptone, 5 g/liter Yeast extract, 10 g/liter KCl) containing 100 mg/liter ampicillin. Media and growth strategies used were patterned after those reported by Tholema et al. (10). For wild-type KtrAB, mutants and the negative control (an empty pKCe vector), two colonies were picked and grown overnight at 37°C in 10 ml of minimal media containing 30 mM K ϩ (see below). Small inoculums (1:300) were then used to initiate growths in media at eight different K ϩ concentrations: 0.1, 0.3, 1, 2, 6, 10, 30, and 115 mM. After growing overnight, the optical density at 595 nm was measured. The minimal media used consisted of a 69 mM sodium/potassium phosphate buffer, pH ϳ7, plus 8 mM (NH 4 ) 2 S0 4 , 400 M MgSO 4 , 6 M FeSO 4 , 1 mM sodium citrate, 1 mg/liter thiamine HCl, 2 g/liter glucose, 10 mg/liter CaCl 2 , 6 M ferric chloride. A fixed ratio of hydrogen phosphate to dihydrogen phosphate was used, keeping the pH and overall phosphate concentration constant even as the K ϩ :Na ϩ ratios were varied. Phosphate component concentrations (K 2 HPO 4 , KH 2 PO 4 , Na 2 HPO 4 , NaH 2 PO 4 , respectively) used for the various target K ϩ concentrations were: 0. Light Scattering Studies-Samples of KtrB mutants (at 1.5 mg/ml) in which the C-terminal tail was truncated were loaded onto a Superdex 200 size-exclusion column equilibrated with 50 mM Tris (pH 8.5), 120 mM NaCl, 30 mM KCl, 1 mM dithiothreitol, 1 mM dodecyl maltoside. Light scattering, UV absorbance, and refractive index measurements were performed on the eluted peaks. The protein molecular weight was determined in thin slices across the elution peak by combinations of these measurements as described in the literature (14 -18).

Dimerization Is a General Property of KtrB Membrane
Proteins-The predicted solvent-exposed loops of the B. subtilis KtrB membrane protein have a total of 17 lysine residues (14 cytoplasmic and 3 extracellular), all of which are potential reacting sites for amino-group cross-linkers. Amino group cross-linkers have been successfully applied in the determination of the oligomeric state of membrane proteins such as the trimeric glutamate transporters and copper uptake transporters (17,19). In these cases where proteins exist in higher oligomeric states, increasing concentrations of the cross-linking reagent resulted in the unambiguous emergence of a strong final-state (trimer) band well before the monomer band disappeared. If any weak transient dimer band was observed, it was only in the presence of a substantial trimer band and faded before the monomer band disappeared. We have previously shown (4) that addition of the bifunctional amino-group crosslinker DSP (spacer arm, 12 Å) to E. coli membranes containing the B. subtilis KtrB membrane protein results in the appearance of a band on a Western blot that corresponds to a KtrB dimer.
We now demonstrate that other bifunctional amino-group cross-linkers with different spanning lengths and chemical properties: disuccinimidyl glutarate (spacer arm, 7.7 Å), 3,3Јdithiobis(sulfosuccinimidyl propionate) (spacer arm, 12 Å), and ethyleneglycol bis(succinimidylsuccinate) (spacer arm, 16 Å), have the same effect. Western blots ( Fig. 2A) show that increasing concentrations of these compounds result in a reduction in the intensity of a band migrating between 30 and 40 kDa, consistent with a KtrB monomer (expected molecular mass, 49 kDa), and a corresponding increase of a band that migrates between 60 and 80 kDa, equivalent to a dimer. Importantly, even though there are numerous reactive sites on each KtrB molecule, no other discrete bands are detected that may correspond to higher order oligomers. A further increase crosslinker concentration only results in a rise of nonspecific links with a variety of other proteins in the membrane.
The KtrB protein was overexpressed in E. coli, extracted with dodecyl maltoside and purified by affinity chromatography followed by size-exclusion chromatography, where it migrates as a single peak (Fig. 2B). We have recently shown by light scattering coupled to refractive index and UV absorbance measurements that the protein in this peak forms a homo-dimer (4), corroborating the cross-linking results. Our previous observa-tion that solubilized KtrB is able to specifically bind KtrA (4) demonstrates that the presence of detergent does not cause a loss of the protein native properties.
To determine whether the purified protein retains its functional properties, we reconstituted purified KtrB into liposomes and performed a radioactivity uptake assay. The assay is equivalent to the one used in the characterization of the KcsA potassium channel; briefly, proteoliposomes loaded with 450 mM potassium are exchanged into a solution with a low potassium concentration (50 M). If the protein is a potassium transporter, it will mediate the flow of K ϩ ions, but not their counterions, down the concentration gradient, thereby establishing a difference potential across the membrane. When 86 Rb ϩ , a good substitute for K ϩ , is added to the external solution, it is driven into the proteoliposomes by the electrical gradient. Fig. 2C shows that uptake of rubidium is protein concentration-dependent. A time course (Fig. 2D) shows that uptake by KtrB is slow and that even after 20-min saturation has not been reached. These data were fitted with a double exponential with time constants of ϳ2 min (amplitude, 0.034) and ϳ120 min (amplitude, 0.45). We believe that the apparent initial burst of activity (corresponding to the short time constant exponential) is a systematic feature of the assay and that the activity of KtrB is expressed in the slow rising phase of the data. In equivalent experiments with the KcsA potassium channel, performed by Heginbotham and colleagues (20), saturation is attained within ϳ3 min, and uptake is fitted with single exponential with a time constant of ϳ30 s. Our experiments show that KtrB can behave like a channel; by this we mean that, albeit not very efficiently (slow uptake), K ϩ and Rb ϩ flow through the protein driven solely by their electrochemical gradients. This agrees well with the report by Bakker and colleagues (11), which showed in vivo that the KtrB protein alone (without KtrA) can transport potassium ions independently of sodium. Importantly these functional results demonstrate that detergent solubilization of KtrB preserves the protein's basic functional properties.
To verify that other KtrB proteins are also organized as dimers, we repeated the cross-linking experiments with KtrB proteins from B. halodurans and S. pneumoniae. These membrane proteins were expressed in E. coli, and membrane preparations were exposed to the bifunctional amino group crosslinker DSP. After the reaction, just like with B. subtilis KtrB, a Probing KtrB with Cysteine-reactive Cross-linkers-The discovery of the dimeric nature of KtrB opened up the possibility of using biochemical approaches to gain structural insights about the protein. We used cysteine-reactive bifunctional compounds and a series of single cysteine mutants to identify protein regions close to the dimer interface and to investigate the structure of KtrB. As a constraint, we have assumed that the subunits of the homo-dimeric membrane protein are related by a 2-fold axis of symmetry perpendicular to the plane of the bilayer. Introduction of a single cysteine mutation in a cysteineless KtrB will result in the dyad axis being positioned midway between this cysteine and its counterpart in the other molecule.
Bifunctional MTS reagents of different lengths can be used to estimate the relative separation between the single cysteine mutations in each of the two subunits (Ref. 21, and references within). Under oxidizing conditions the reaction between the MTS compounds and a cysteine is irreversible, and, given enough time, the macromolecule samples many conformations and cross-linking occurs even at positions that seem too far apart. Therefore, distance restraints are usually defined from the rate of formation of the cross-linked protein, or efficiency of cross-link. For practical reasons we have chosen not to measure the rates of dimer cross-linking but to expose each cysteine mutant to a long (M17M, C␣ to C␣ maximum span of ϳ31 Å) and a short (M3M, C␣ to C␣ maximum span of ϳ13 Å) bifunctional reagent for a fixed amount of time. Reaction with the short compound will occur only at a restricted set of mutant sites (Fig. 3A), identifying positions close to the dyad axis (interface). Conversely, the long reagent will react with most or all mutant positions. Local protein structure will also affect the reaction, so it is important to look for trends across several positions.
We obtained a cysteine-less KtrB by mutating the only natural cysteine (Cys-407) to alanine; in this background we prepared 13 single cysteine mutants in the cytoplasmic and extracellular loops (Fig. 3B) guided by sequence analysis algorithms and the experimentally determined topology of the related pro-tein HKT1 from A. thaliana (9). The selected positions are occupied by non-conserved amino acids with polar side chains and are therefore probably non-essential and exposed to the aqueous solution. A mutation that completely wipes out function may have only minimal structural effects and therefore still provide useful information for our purposes. For this reason, we did not test the functional impact of the mutations, but assessed only whether the mutant proteins are still dimeric.
Membrane preparations of each mutant were exposed for 1 h to the reagents M3M and M17M, with a C␣ to C␣ span of ϳ13 Å and ϳ31 Å (22-25), respectively. The reactions were terminated with the addition of NEM and analyzed by Western blot. As a control, the efficiency of the stop reagent was demonstrated by simultaneously adding NEM and M17M to the samples, incubating 1 h then Western blotting (see supplemental data). Exposure to the M3M reagent (Fig. 4A) resulted in the formation of strong dimer bands at only a few positions: S159C, N218C, E291C, E335C, and G440C. The apparent weak dimer band at position 252 was not reproduced in other equivalent blots, therefore M3M cross-linking at position 252 was not considered further.
The level of protein expression for the mutants was monitored by Western blot analysis of samples not exposed to crosslinkers (Fig. 4B); in this case samples were incubated as before but without reagents. NEM was added after 1 h. Surprisingly, mutants S159C and G440C each showed intense dimer bands that must result from the spontaneous formation of disulfide bonds at those positions between the two subunits. These dimer bands disappear if the samples are incubated with reducing reagent prior to running the SDS-PAGE (data not shown). The band at N119C was disregarded due to its very weak intensity. This blot also shows that, with the exception of T73C, mutants expressed at similar levels. Despite its relatively low expression, the T73C mutant still cross-links well with the long compounds (Fig. 4C, M17M blot), and results obtained at this position are included in our analysis below.
Exposure to the longest reagent M17M (Fig. 4C) resulted in strong dimer cross-linking for almost all mutants (except for T15C), demonstrating that the large majority of cysteines introduced are accessible and reactive to the MTS reagents. No reaction was observed at T15C, probably because the side chain is not accessible, and we did not consider this position further. As expected, the cysteineless KtrB control shows no dimer crosslinked band after exposure to M17M. There are a large number of higher molecular weight bands in the cysteine mutant lanes relative to the cysteineless protein. Because all of the KtrB variants used here contain just one cysteine residue per monomer then the higher molecular weight bands must result from cross-linking of KtrB with other protein components in the membrane. To account for the possibility that we are observing interdimer and not intra-dimer cross-linking, we also compared the effects of two different temperatures, room temperature and 0°C, on the M17M reaction pattern (Fig. 4D). The low temperature was set below a phase transition of the E. coli lipid bilayer so that lateral diffusion within the membrane was decreased, resulting in a reduction of collisions between dimers (22,26). It has been reported that cross-linking easily occurs in the monomeric LacY, giving rise to an apparent dimer. Equivalent temperature experiments resulted in a reduction of the LacY dimer band by Ͼ90% (22). Fig. 4D shows that there are no discernible temperature effects with KtrB at positions 119, 159, 218, 291, 335, 374, and 440. We conclude that M3M and spontaneous disulfide cross-links at S159C, N218C, E291C, E335C, and G440C represent reactions within a dimer molecule and define regions close to the molecular 2-fold axis. Conversely, we observe a strong temperature effect at position 403 (Fig. 4D), where almost no dimer band is present at 0°C, indicating that only inter-dimer cross-linking is occurring. Smaller temperature effects were observed at positions 43, 73, 190, and 252, where at 0°C dimer bands are slightly weaker and the monomer band is slightly stronger than at room temperature. It cannot be ruled out that temperature has an impact either on the rate of the chemical reactions or on the flexibility of the protein and therefore on the efficiency of cross-link. However, if either one of these effects played a significant role in the temperature dependence of the crosslinks, we would expect to observe a decrease in the intensity of the dimer bands relative to the monomer bands for most of the mutants at the lower temperature. Because this is not the case, we favor the explanation that the weakly temperature-dependent dimer bands result from a mix of both inter-and intra-dimer crosslinking. Based on the relative intensities of the dimer bands at the two different temperatures, we estimate that roughly half of the observed cross-linking at these positions is intra-molecular. These temperature dependence experiments also aid in the analysis of observed cross-linking by identifying regions sensitive to dimer collisions and therefore probably close to the periphery of the dimer. In addition, the presence of dimer bands at 0°C for almost all mutant positions (Cys-430 is the exception) demonstrates that the mutations do not drastically alter the dimeric nature of KtrB.
The cross-linking results are summarized in Fig. 4E. All cysteine residues introduced in the short extracellular loops are within a few residues of the nearest TM and therefore serve as positional restraints for the nearby transmembrane segment. In the cytoplasmic loops, with the exception of the mutant in the loop B2-C1 (N218C), cysteine mutants were introduced in the center of the loops. Based on the observed cross-linking patterns, we propose that the regions close to the molecular dyad axis, positioned at or near the dimer interface, are as follows: on the extracellular side, the ends of TM-B1 (disulfide bond at the external position S159C) and TM-C2 (M3M cross-link at the external site E291C); on the cytoplasmic side, the end of TM-B2 (M3M cross-link at N218C), the loop between TM-C2 and TM-D1 (M3M cross-link at E335C) and the C terminus (disulfide bond at G440C). Regions determined to be close to the periphery of the molecule, as determined from the temperature . D, temperature dependence of M17M cross-linking reaction. Tested mutants are indicated above each pair of lanes. "0" and "rt" labels correspond to reactions performed at 0°C or room temperature, respectively. "ϩ" and "ߜ" symbols mark the mutants where reaction is temperature-dependent or independent, respectively. "ϩϩϩ" indicates the large difference between the two reaction conditions at position 403. dependence of M17M cross-linking (Fig. 4D), are also marked in Fig. 4E. Strikingly, none of the cytoplasmic positions tested were temperature-dependent.
To increase the number of restraints for the ends of TM segments in the cytoplasmic regions we generated four more mutants (G125C just before TM-B1, S225C before TM-C1, P346C just before TM-D1, and T430C just after TM-D2) and performed M3M and M17M exposures, as well as spontaneous cross-linking. In the last TM-P loop -TM repeat, both Cys-346 and Cys-430 formed M3M-mediated cross-links, with Cys-346 proving more efficient at forming a disulfide bond. The M17M blot demonstrates that the cross-links at Cys-346 and Cys-430 are not temperature-dependent and, therefore, occur within the dimer molecule (Fig. 4G). The absence of cross-linking at cytoplasmic positions Cys-125 and Cys-225 is clarified by a lack of reactivity even after exposure to M17M, indicating that these positions are not accessible to the reagents. The same was found for a new mutation made on the extracellular side of the molecule, A164C, located just after Cys-159 and before the pore helix of repeat B (Fig. 4, F and G).
Properties of the C Terminus-We have interpreted the spontaneous formation of direct disulfide bonds, or of cross-links using the short M3M molecule, as identifying regions that are close to the molecular dyad axis. Because these residues are presumably located at or near the dimer interface, sequence perturbations in these regions could possibly alter the oligomeric state of the membrane protein and impact the ability of the transporter to function.
To test this interpretation we created several mutations in the C-terminal end of KtrB, a region that contains one of the sites of spontaneous disulfide bond formation, at G440C, just five residues before the end of the protein. This region displays a substantial degree of sequence conservation (Fig. 5A). A truncation mutant (residues 1-435) of KtrB with the last 10 residues removed yields two peaks in a size-exclusion chromatography elution profile (Fig. 5B), clearly different from the single peak profile of the wild-type protein (residues 1-445) (Fig. 2B). Three-detector light scattering analysis reveals the first peak to be a dimer (105.1-kDa molecular mass) like that observed for the wild-type protein, whereas the new second peak is a monomer (47.9 kDa). The expected molecular mass is 48.6 kDa. A similar result was also obtained using a mutant with the last 15 residues truncated (data not shown). Hence, removal of the final 2% of the protein sequence, a region able to form spontaneous disulfide bonds with its counterpart in the dimer and presumably located near the interface, results in significant disruption of the dimer. These data demonstrate that the C-terminal end of KtrB is important for dimer stability, probably by establishing contacts across the interface, and reinforce our molecular interpretation of the cross-linking data in general.
To investigate the functional importance of the C-terminal tail of KtrB we tested deletion mutants, point mutants, and wild-type KtrB using an in vivo complementation assay. We made use of the TK2420 E. coli strain in which a series of proteins involved in potassium ion uptake have been disabled and therefore cells require additional K ϩ in the medium for optimal growth. The presence of a functional KtrAB transporter aids K ϩ uptake and allows these cells to grow in low K ϩ medium.
Results from the complementation assay are presented in Fig.  5C. Others have shown that complementation is observed only when KtrB and KtrA are expressed together in cells (2,3). Under the conditions of this assay, co-expression of wild-type KtrA and wild-type KtrB allowed the cells to achieve saturated growth in the presence of 1 mM K ϩ (Fig. 5C). As a negative control, cells containing no KtrAB transporter show no detectable growth at 10 mM K ϩ or below, only slight growth at 30 mM K ϩ , but hearty growth indistinguishable from cells containing wild-type KtrAB at 115 mM K ϩ . Cells co-expressing either of the C-tail deletion mutants of KtrB (missing the last 10 or 15 subtilis KtrB in which the C terminus is truncated, missing the last 10 residues. UV monitoring at 280 nm is shown by a thin continuous line. Protein molecular weight was determined across each peak by the three monitor light-scattering techniques and is displayed by the thick lines. C, plot of bacterial culture optical density as a function of K ϩ concentration. K ϩ -uptake-deficient TK2420 E. coli cells were transformed with a dicistronic vector encoding wildtype KtrA and either wild-type or mutant KtrB from B. subtilis. Mutations in KtrB are indicated. Cells were grown in minimal media with different K ϩ concentrations; optical density was measured after ϳ14 h. Control data were acquired on vector-only transformants. Results are averages of measurements from two different colonies. "⌬10C" and "⌬15C" represent KtrB mutants with the last 10 or 15 residues removed, respectively. "G445X" indicates a single-residue truncation of the terminal glycine. Certain lines are highlighted to aid visualization. residues, respectively) with wild-type KtrA required 30 mM K ϩ to reach saturation and showed no detectable growth at 10 mM K ϩ or below. The substantial reduction in the apparent activity of these mutants seems to be consistent with their decreased dimer stability. It must be noted, however, that we could not evaluate whether any of the mutants tested expressed at normal levels, because the amount of KtrB protein (including wild type) produced in the complementation assay was too low to be detected on a Western blot. Point mutations at a conserved position (Y437A) and a non-conserved position (G440A) of the C-terminal tail display growth like that of wild-type KtrB, reaching saturation at 1 mM K ϩ . A single-residue deletion of the terminal glycine (G445X), the most conserved feature of the C-tail (Fig. 5A), results in a very slight reduction in apparent KtrAB activity with cells reaching saturation at 2 mM K ϩ .
Modeling the Cross-linking Data-If as has been suggested the KtrB molecule adopts an overall KcsA-like structure, then a model created by placing two KcsA molecules side by side, while maintaining 2-fold symmetry, should agree with our cross-linking data. Our experiments cannot distinguish between two possible arrangements of a dimeric KtrB where: 1) each polypeptide forms a separate channel-like subunit or 2) the polypeptides are involved in domain swapping and participate in both subunits. However, the observation that modest C-tail truncations result in stable monomers argues against any type of extensive domain swapping. In either case our experiments allow us to define the regions close to the molecular 2-fold axis and to probe the overall KtrB structure.
Certain considerations must be taken into account when utilizing cross-linking data for structural probing. First, results should only be interpreted as yielding relative distance restraints. That is, detection of cross-linking with a reagent that spans ϳ13 Å cannot be strictly interpreted as meaning the cross-linked cysteine residues are separated by that distance or less, but only that they are generally closer to the molecular dyad axis than other residues that require a longer linker to react. Due to thermal motion and the long reaction time (1 h), residues that are on average further apart than the spanning distance of the reagent will eventually sample spatial positions that result in cross-linking. Second, local protein structure may sterically hinder the ability to form a cross-link, requiring a longer linker than would be required for two residues the same distance apart in a different part of the protein.
The length of the cytoplasmic loops connecting the TM-P loop -TM repeats of KtrB could, in theory, permit any of six possible different dispositions of these internal structural units around the ion pathway of a given monomer. Fig. 6A shows the four possible arrangements where repeats B and C are side by side within a monomer.
Our cross-linking studies of KtrB indicate that the extracellular ends of both TM-B1 (S159C mutant) and TM-C2 (E291C mutant) are located close to the 2-fold axis of the dimer (Fig.  4E). These correspond to the first and second helices of repeats B and C, respectively. These two positions, one of which is able to form a disulfide bond, serve as anchor points for the interface of our dimer model. The most reasonable fit to the extracellular cross-linking data of KtrB when mapped onto a three-dimensional dimer model (two side-by-side KcsA molecules) is shown in Fig. 6B. In order for the extracellular ends of both TM-B1 and TM-C2 to be positioned near the dimer 2-fold axis, they must be located adjacent to each other within the KtrB monomer structure. This requirement is satisfied in only two of the possible types of monomer arrangements (I and III) depicted in Fig.  6A. In Fig. 6B we show a dimer model of KtrB consisting of type-III disposition monomers. It is evident from the extracellular side of the dimer model why some of the tested positions display temperature-dependent cross-linking reactions. These positions are located close to the periphery of the model and away from the dimer interface and are therefore sensitive to collisions with other dimer molecules in the membrane. The straightforward match between the extracellular data and this KcsA-based model provides support to the notion that the KtrB protein adopts the architecture present in potassium channels.
Cross-linking data on the cytoplasmic side of KtrB indicate proximity to the molecular dyad axis at the end of TM-B2 (N218C), the loop C2-D1 (E335C), the end of TM-D1 (P346C), the end of TM-D2 (T430C) and C terminus (G440C) (Fig. 4, E and F). Modeling of data from positions in the middle of the cytoplasmic loops of KtrB poses a problem because equivalent regions do not exist in KcsA, which is formed by four polypeptides. Nevertheless, the cross-links that restrain the positions of the cytoplasmic ends of TM-B2, TM-D1, and TM-D2 show that in a KcsA-based model (Fig. 6C) these helices are further from the dimer 2-fold axis than are the cross-linked extracellular regions (compare with Fig. 6B). These apparently longer distances indicate that our KcsA-based model, as is, does not match well with the observed data for the cytoplasmic half of the KtrB dimer.
In Fig. 6 (B and C) we have depicted a dimer composed of type-III disposition monomers (Fig. 6A), because it matches the extracellular cross-linking data, both in position and temperature dependence, of KtrB. By comparison, a dimer of type-I disposition monomers would equally match the extracellular data but would result in even longer apparent cross-linking distances on the cytoplasmic side. In fact, none of the possible alternative dispositions explain the cytoplasmic data in a straightforward fashion. Moreover, simply modeling a symmetric opening of a gating motion in each pore, as seen in some K ϩ channels (27), does not help.
Additional Testing of the Model-To further test the dimer model shown in Fig. 6 (B and C), we exposed single cysteine mutants of KtrB to M5M, a bifunctional cysteine cross-linker that is slightly longer (C␣ to C␣ maximum span, ϳ16 Å) than M3M. If the KcsA-based model is correct, the extracellular cross-linking pattern would now be expected to expand outward to include transmembrane helices located further away from the dyad axis, such as TMs B2, C1, D1, or A2. Exposure to M5M was performed with all mutants except for 159C and 440C, which spontaneously form a disulfide bond. As with M3M, cross-linking was detected at mutant positions Cys-218, Cys-291, Cys-335, Cys-346, and Cys-430 (Fig. 6D). As anticipated, the reaction pattern now expands to the extracellular ends of TM-C1 and TM-D1 with new dimer bands detected at positions Cys-252 and Cys-374, respectively. We have shown that Cys-374 is not temperature-dependent (Fig. 4D), indicating that the M5M band at this position corresponds to intra-dimer cross-linking. Cross-linking at Cys-252 was shown to be weakly temperature-dependent, probably resulting from a mix of inter-and intra-dimer reactions. If, as previously discussed, half of the intensity of a dimer band formed at Cys-252 is considered to result from intra-dimer cross-linking, then this position can be considered for our model. A much weaker band was also observed at Cys-403, a position shown to be highly temperature-dependent, indicating a location on the periphery of the KtrB dimer involved in inter-dimer reac-tions. The weak M5M band at this position was, therefore, not considered further. Overall, the M5M results fit the cross-linking pattern defined by M3M and spontaneous disulfide bond formation and support the model in Fig. 6B. The lack of M5M cross-linking at the extracellular end of TM-B2 may simply reflect local structure hindrance to a crosslinker of this length. All the crosslinking results are summarized in Fig. 6E.

DISCUSSION
In a previous study (4) focused on the structural and functional properties of the KtrA regulatory protein, we established that the KtrAB complex is composed of a KtrA octamer and a KtrB dimer. Here we have explored the structural properties of the KtrB protein, provided additional data supporting the dimeric organization of B. subtilis KtrB, and show that homo-dimerization is probably a general property of the KtrB membrane protein family. We have identified several regions of KtrB that react as if close to the molecular dyad axis and have bolstered our interpretation of the cross-linking data by demonstrating that removal of one of these regions promotes the breakdown of the dimer into monomers. In light of the hypothesis that KtrB may be similar in structure to potassium channels, we have mapped the crosslinking data onto a model consisting of two side-by-side KcsA channels. Interestingly, the conserved nature of the KtrB dimer organization does not seem to be reflected by sequence conservation along peripheral transmembrane regions (data not shown).
The dimeric organization of these proteins is intriguing in light of the proposal that they adopt the architecture seen in the pore domain of potassium channels. It implies that two ion transport pathways exist side by side and raises the question of whether the pores are structurally independent, that is, formed by one polypeptide chain wrapped around a central pore axis or created out of different regions of two polypeptides. Our C-terminal tail truncation data shed some light on this question. The observation that the removal of as little as 10 residues from the C terminus of KtrB causes the dimer to break down into stable monomers suggests that there is no extensive domain swapping. We believe that our data are best explained by two side-by-side KtrB monomers. Whether the two ion pores are functionally independent is unknown.
The revelation that KtrB forms homo-dimers has provided us with the opportunity to probe the structure of the protein in a simple fashion, using cysteine mutagenesis and directed cross-linking reactions. With this approach we have established that the regions close to the molecular dyad axis and therefore probably close to or participating in the dimer interface include the extracellular ends of TM-B1, TM-C2, the cytoplasmic regions of TM-B2, TM-D1, and TM-D2 as well as the loop C2-D1 and the C-terminal tail (Fig. 6E). The cross-linking results have also allowed us to probe the proposed structural relationship between the KtrB transporter and the KcsA potassium channel. The KcsA-based dimer model presented in Fig.  6B explains well the cross-links (disulfide bonds or short M3Mmediated links) observed in the extracellular half of KtrB by placing the ends of TM-B1 and TM-C2 near the 2-fold axis. The use of M5M, an MTS reagent slightly longer than M3M, expands the interaction radius in a manner consistent with this interface. The temperature-dependent nature of cross-linking observed at some positions is also explained by this model, because these regions lie on the extracellular periphery of the dimer where collisions between KtrB dimers are more likely to occur.
Our cross-linking data support the view that KcsA serves as a good model for the extracellular half of KtrB but that adjustments are required on the cytoplasmic half of the model. This is not surprising, because even among potassium channels the structural core that includes the selectivity filter is located on the extracellular side and tends to remain predominantly unchanged, whereas the cytoplasmic half can undergo major conformational adjustments during gating and is therefore structurally more adaptable.
Either of the following possible adjustments to the model could remedy the apparent too-long cross-links of the cytoplasmic region. First, the axes of the KtrB monomers might be tilted somewhat with respect to the membrane plane, with cytoplasmic loop interactions across the dimer interface drawing the cytoplasmic ends of the subunits closer together than in the present model while leaving the spacing of the extracellular halves unchanged. The resulting ion pathways would no longer be perpendicular to the membrane or parallel to each other. The somewhat inverted-teepee shape of KcsA, wider on the extracellular side than on the cytoplasmic side, could easily accommodate a modest inward rotation, and the effect would be to shorten all of the cytoplasmic cross-link vectors depicted in Fig. 6C. A second possible explanation is that the KtrB sequence asymmetry is manifested in structural asymmetry, especially in the cytoplasmic half. Because KtrB contains one or more glycine residues in the regions where helical bending has been shown to occur during potassium channel gating (27) (Fig.  1A), it is possible that the cytoplasmic ends of some KtrB innerhelices are bent toward the dyad axis. However, the trajectories or bending of the four inner-helices of KtrB do not need to be like one another, making this scenario difficult to model.
In conclusion, the KtrB membrane protein is a homo-dimer and we suggest the extracellular half of its constituent mono-mers resemble KcsA in structure. By contrast, the cytoplasmic half of the KtrB transporter appears to differ from our simple KcsA-based model and to contain features that are specific to the transporter family.