The Molecular Basis of K+ Exclusion by the Escherichia coli Ammonium Channel AmtB*

Background: The Amt family of ammonium channels does not conduct K+. Results: E. coli AmtB variants carrying certain mutations in the conserved twin-histidine element transport K+ against a concentration gradient. Conclusion: The twin-histidine element functions as a filter that prevents K+ conduction. Significance: These findings provide insight into the transport mechanism of Amt channels and the ammonium species (NH3 or NH4+) that serves as their substrate. Members of the Amt family of channels mediate the transport of ammonium. The form of ammonium, NH3 or NH4+, carried by these proteins remains controversial, and the mechanism by which they select against K+ ions is unclear. We describe here a set of Escherichia coli AmtB proteins carrying mutations at the conserved twin-histidine site within the conduction pore that have altered substrate specificity and now transport K+. Subsequent work established that AmtB-mediated K+ uptake occurred against a concentration gradient and was membrane potential-dependent. These findings indicate that the twin-histidine element serves as a filter to prevent K+ conduction and strongly support the notion that Amt proteins transport cations (NH4+ or, in mutant proteins, K+) rather than NH3 gas molecules through their conduction pores.

these proteins remains controversial, and the mechanism by which they select against K ؉ ions is unclear. We describe here a set of Escherichia coli AmtB proteins carrying mutations at the conserved twin-histidine site within the conduction pore that have altered substrate specificity and now transport K ؉ . Subsequent work established that AmtB-mediated K ؉ uptake occurred against a concentration gradient and was membrane potential-dependent. These findings indicate that the twin-histidine element serves as a filter to prevent K ؉ conduction and strongly support the notion that Amt proteins transport cations (NH 4 ؉ or, in mutant proteins, K ؉ ) rather than NH 3 gas molecules through their conduction pores.
Ammonium (used to designate NH 3 and NH 4 ϩ ) is the preferred nitrogen source of many organisms, making the manner by which it traverses biological membranes of great physiological importance. The uncharged form, NH 3 , diffuses across lipid bilayers more than 35 times faster than water (1). When the rate of NH 3 diffusion becomes growth limiting proteinmediated transport of ammonium by the Amt family of channels is required (2)(3)(4)(5). Still subject to debate is the transport mechanism of these proteins and the species of ammonium that serves as their substrate. Based on structural studies and numerous molecular dynamic simulations (6 -11), it has been argued that Amt family members are passive channels that facilitate downhill movement of NH 3 . We hold the opposite view, supported by the research efforts of other groups (3,(12)(13)(14)(15)(16)(17)(18)(19)(20), that these proteins function as active channels that mediate electrogenic ammonium uptake against a concentration gradient. Mechanisms of electrogenic transport that have been suggested include conduction of the membrane-imper-meant protonated form of ammonium, NH 4 ϩ , and the cotransport of NH 3 and H ϩ (2, 3, 5, 12, 13, 16 -21).
Members of the Amt family function as homotrimers (22). Structural studies indicate that these proteins and their only known homologs, the Rhesus (Rh) proteins, have a conduction pore on each monomeric subunit (7,11,21,(23)(24)(25) (Fig.  1). Positioned near the center of each of these otherwise hydrophobic pores is a pair of conserved histidines. It has been proposed that this twin-histidine element plays a critical role in mediating substrate transport (7,11,28). However, recent work by one of us demonstrated that the E. coli AmtB protein can accommodate acidic residues at one or both positions of its His 168 /His 318 twin-histidine pairing while retaining good to excellent ammonium transport activity (14). This finding led us to question why the twin-histidine element is conserved within the Amt family. Here, we report that certain AmtB twin-histidine variants are gain-of-function mutants that conduct K ϩ , a cation having an ionic radius similar to that of NH 4 ϩ but normally not transported by Amt proteins (17,18,29,30). Our analysis shows that the twin-histidine element serves as a substrate selectivity filter and leads us to suggest that Amt channels utilize an electrogenic transport mechanism in which ammonium crosses the phenyl ring constriction as NH 4 ϩ , deprotonates and traverses the conduction pore in close association with its H ϩ , and finally reprotonates prior to entering the cytoplasm.
Strains in the NCM3722 background were maintained on LB plates, and strains in the NCM5020 background were maintained on LB plates containing 50 mM KCl. K ϩ -based formulations of N Ϫ C Ϫ minimal medium, Neidhardt's MOPS medium, and Neidhardt's MES medium contain ϳ200, 28, and 20 mM K ϩ , respectively (3,34,35). Na ϩ -based formulations of these defined media were prepared by replacing all K ϩ -containing components with their Na ϩ -containing counterparts.
Growth Assays-For strain growth illustrated in Fig. 2, cells grown in LB medium were diluted 100-fold into either K ϩ -based or Na ϩ -based N Ϫ C Ϫ minimal medium containing 0.4% glucose and 5 mM NH 4 Cl. After overnight incubation, cells were diluted to an A 600 ϭ 0.01 (Ͼ100-fold dilution) into either K ϩ -based or Na ϩ -based N Ϫ C Ϫ medium containing 0.04% glucose and 3 mM glutamine. Glutamine supports good growth as a sole source of nitrogen when cells are cultivated in this medium but its use as a sole nitrogen source elicits the nitrogen limitation response, resulting in the expression of all genes, including amtB, under control of nitrogen regulatory protein C (NtrC) (36 -38). Growth was monitored by changes in absorbance at 600 nm. Strain growth was carried out with aeration at 37°C. All Na ϩ -based N Ϫ C Ϫ medium was supplemented with 1 mM KCl to avoid K ϩ -limiting growth conditions.
For strain growth illustrated in Fig. 4, cells grown in LB medium containing 100 mM KCl were diluted 100-fold into Na ϩ -based N Ϫ C Ϫ medium supplemented with 0.4% glucose, 5 mM NH 4 Cl, and 100 mM KCl. After these cultures reached saturation they were again diluted 100-fold into the same Na ϩbased N Ϫ C Ϫ medium, except with 3 mM glutamine replacing NH 4 Cl as the nitrogen source. Following overnight incubation, cultures were finally diluted 200-fold into Na ϩ -based N Ϫ C Ϫ medium supplemented with 0.4% glucose, 3 mM glutamine, and a mixture of KCl and NaCl (combined concentration of 100 mM). Growth was monitored by changes in absorbance at 600 nm. Strain growth was carried out with aeration at 37°C.
Growth at low NH 3 concentrations (Fig. 3) was performed as previously described (14) with the following differences: cells were first adapted to minimal medium on Na ϩ -based Neidhardt's MOPS medium (pH 7.4) and subsequently acclimated to growth at low pH by inoculation into Na ϩ -based Neidhardt's MES medium (pH 5.5); all adaptive media contained 1 mM KCl; and cells were finally diluted into either K ϩ -based Neidhardt's MES medium (pH 5.5) supplemented with 1 mM KCl, or Na ϩ -based Neidhardt's medium (pH 5.5) supplemented with 0.1 mM KCl and 0.9 mM NaCl.
Expression and accumulation levels of mutant proteins relative to wild-type AmtB have been reported (14). The three AmtB variants studied here were produced in amounts similar to that of the wild-type strain (60 -100%) when glutamine served as the sole nitrogen source ( Fig. 2 and Figs. 4 -7). However, whereas AmtB H168D and AmtB H318D accumulated to 30 -40% wild-type levels under this growth condition, AmtB H168D,H318E was present in amounts Յ5% of its wild-type counterpart. Similar expression and accumulation patterns were found for these three AmtB variants when low NH 3 medium was used for growth studies (Fig. 3).
Transport Assays-Strains were grown in LB medium containing 100 mM KCl and, after overnight incubation, diluted 100-fold into Na ϩ -based N Ϫ C Ϫ medium supplemented with 0.4% glucose, 5 mM NH 4 Cl, and 100 mM KCl. After these cultures reached saturation they were again diluted 100-fold into the same Na ϩ -based N Ϫ C Ϫ medium, except with 3 mM glutamine replacing NH 4 Cl as the nitrogen source. Cells cultivated under these conditions were grown to exponential phase (A 600 ϭ 0.3-0.6), harvested by centrifugation, and resuspended in Na ϩ -based N Ϫ C Ϫ medium containing 100 mM NaCl. For determination of cytoplasmic K ϩ pools, cells were collected and processed for inductively coupled plasma atomic emission spectroscopy (ICP-AES) 2 (see below). To analyze K ϩ uptake, resuspended cells were washed twice and finally suspended at an A 600 ϭ 1.0 in Na ϩ -based N Ϫ C Ϫ medium containing 100 mM NaCl, and then treated with 5 mM 2,4-dinitrophenol for 30 min. Treatment with this protonophore results in the rapid loss of cellular K ϩ (39,40). K ϩ -depleted cells were harvested, washed twice with Na ϩ -based N Ϫ C Ϫ medium supplemented with 100 mM NaCl, resuspended at an A 600 ϭ 1.0 in the same medium supplemented with 0.2% glucose, and held on ice until use. To initiate tests of K ϩ transport, cell suspensions were preincubated for 20 min at 37°C prior to the addition of an equal volume of Na ϩ -based N Ϫ C Ϫ medium supplemented with 0.2% glucose and suitable mixtures of KCl, NH 4 Cl, and NaCl (combined concentration of 100 mM). After a 1 min incubation aliquots were collected on Millipore filters (0.45-m pore size; type HAWP), rinsed twice with 5 ml of Na ϩ -based N Ϫ C Ϫ medium supplemented with 100 mM NaCl, and dried. Washed cells (on the filter) were then incubated at room temperature (ϳ25°C) for 1 h in 5 ml of 1 M HNO 3 containing 50 M NaCl to extract K ϩ . Samples were cleared of cell debris by passage through Millipore filters (0.45-m pore size; type HAWP) and analyzed for K ϩ by ICP-AES. To determine whether AmtBmediated K ϩ transport required an energy source (Fig. 7A), glucose was eliminated from all assay buffer preparations. For assays with carbonyl cyanide m-chlorophenylhydrazone (CCCP), this compound or an equal volume of ethanol vehicle was added 5 min prior to the initiation of K ϩ uptake tests. Strain growth, K ϩ depletion, and K ϩ uptake assays were carried out with aeration at 37°C.
Apparent kinetic constants (half-saturation constant values and maximal transport rates) for AmtB-mediated K ϩ uptake were estimated by curve fitting initial transport rates to the Michaelis-Menten equation using the program CurveExpert Professional, version 1.6.2. The concentration range of K ϩ used to approximate these kinetic constants was 0.5-50 mM. Inhibition constant (K i ) values for ammonium inhibition of AmtB-mediated K ϩ transport were determined using linear Dixon plots (41), with the assumption of competitive inhibition between ammonium and K ϩ . For energy requirement assays and K i determinations (Fig. 7), K ϩ was present at twice half-saturation constant values for AmtB H318D and AmtB H168D,H318E in the glnK ϩ background, and 40 mM for AmtB H318D and AmtB H168D,H318E in the glnK background. Transport values measured in the absence of ammonium using these K ϩ concentrations were 180 Ϯ 38 nmol/ml per A 600 per min for AmtB H318D in the glnK ϩ background, 210 Ϯ 51 nmol/ml per A 600 per min for AmtB H168D,H318E in the glnK ϩ background, 140 Ϯ 21 nmol/ml per A 600 per min for AmtB H318D in the glnK background, and 96 Ϯ 6.3 nmol/ml per A 600 per min for AmtB H168D,H318E in the glnK background. The concentration range of ammonium used to approximate K i values was 1 M to 10 mM.
Glucose Determination-Samples from cultures illustrated in Fig. 2A were subjected to filtration (0.22-m pore size; Millipore type Millex-GV) to remove cells. Glucose concentrations in cell-free medium were determined enzymatically using a glucose oxidase/peroxidase-coupled assay monitoring the oxidation of o-dianisidine (Sigma-Aldrich product GAGO-20).

RESULTS
K ϩ -dependent Growth Defect in AmtB Twin-histidine Variant Strains-A number of E. coli AmtB mutant proteins carrying acidic residues at the His 168 /His 318 twin-histidine site cause a pronounced growth defect when expressed under nitrogenlimiting conditions in media devoid of ammonium and of high K ϩ content (Fig. 2, A and C). This defect is characterized by a decreased carbon yield relative to both the wild-type and amtBnull strains. We reasoned that the twin-histidine variants exhibiting the growth defect were wasting energy in the futile active transport of K ϩ . Several mutants were selected to test this hypothesis. Two of these, AmtB H168D and AmtB H318D , have near wild-type ammonium transport activity even in the presence of ϳ20 mM K ϩ (Fig. 3 and Ref. 14). Like wild-type AmtB, the AmtB H168D protein also transports the ammonium analog methylammonium (used to designate CH 3 NH 2 and CH 3 NH 3 ϩ ) in the absence of K ϩ , whereas AmtB H318D does not (14). A third mutant, AmtB H168D,H318E , that exhibits neither ammonium nor methylammonium transport activity even at low K ϩ (Fig. 3 and Ref. 14) was also examined. Studies conducted on cells expressing these variants showed that the growth defect was eliminated if Na ϩ replaced most of the K ϩ in the culture medium (1 mM K ϩ supplied to avoid K ϩ -limiting growth conditions) (Fig. 2B).
AmtB Twin-histidine Variants Conduct K ϩ -Having established a link between the external K ϩ concentration and the growth defect observed in certain twin-histidine mutants, we next looked to see whether this defect resulted from AmtBmediated inward K ϩ leakage. A strain lacking the three major K ϩ import systems (Kdp, Trk, and Kup) was constructed for this work. In the absence of these transport systems E. coli possesses only a residual K ϩ uptake activity and requires an elevated concentration of this cation for growth (39,40). We found that the kdp trk kup triple mutant failed to grow appreciably on media containing less than ϳ50 mM K ϩ , and expression of wild-type AmtB had no effect on this growth phenotype ( Fig. 4; compare amtB-null and wild-type strains). Expression of AmtB H168D , AmtB H318D , or AmtB H168D,H318E , on the other hand, reduced the K ϩ requirement to ϳ10, 5, and 1 mM, respectively.
The ability of AmtB to conduct K ϩ was also assayed directly using ICP-AES. Initial experiments indicated that cytoplasmic K ϩ pools of kdp trk kup triple mutants expressing a twin-histidine variant protein, in particular AmtB H318D or AmtB H168D,H318E , were not only higher than their wild-type AmtB-expressing and amtB-null counterparts, but also larger than that of a wild-type AmtB-expressing strain carrying the Kdp, Trk, and Kup systems (Fig. 5). We next depleted each of these strains of their K ϩ to study the net uptake of this cation. As expected, the K ϩ transport rate of a kdp trk kup triple mutant expressing wildtype AmtB was no better than that of an otherwise isogenic amtB-null strain (Fig. 6A). Only a slight improvement (Ͻ2fold) in K ϩ uptake was observed when AmtB H168D was expressed in this background. The initial K ϩ transport rates of wild-type AmtB and AmtB H168D were linearly proportional to the external K ϩ concentration over the range tested (up to 50 mM) and markedly lower than the combined activities of Trk and Kup (Kdp not expressed under the conditions (100 mM K ϩ in the growth medium) used for this assay; Refs. 39, 40).  , and 100 mM (f)). Growth medium isotonicity was maintained by the addition of NaCl such that the total KCl/ NaCl supplement equaled 100 mM. All amtB alleles were carried in the kdp trk kup strain background. The data, from a single experiment, are representative of findings made in three independent trials. AmtB H318D and AmtB H168D,H318E , on the other hand, exhibited considerable K ϩ uptake activity. These two twin-histidine variants were found to have 5-10-fold higher half-saturation constant values (4.9 Ϯ 0.93 mM and 11 Ϯ 1.4 mM for AmtB H318D and AmtB H168D,H318E , respectively) and elevated maximal transport rates (250 Ϯ 6.0 nmol/ml per A 600 per min and 270 Ϯ 22 nmol/ml per A 600 per min for AmtB H318D and AmtB H168D,H318E , respectively) relative to the aggregate kinetic properties of the Trk and Kup systems (half-saturation constant of 1.1 Ϯ 0.07 mM; maximal transport rate of 210 Ϯ 2.1 nmol/ml per A 600 per min; also see Ref. 40). Given that a 1 ml E. coli culture of A 600 ϭ 1 has a total cell volume of ϳ3.6 l (42), we calculate that AmtB H318D and AmtB H168D,H318E concentrate K ϩ 3-10-fold during the first min of the transport reaction when K ϩ is present at 0.5-10 mM in the external medium.
Inhibition of AmtB-mediated K ϩ Transport-We used AmtB H318D and AmtB H168D,H318E as vehicles to analyze the energy requirements of AmtB-mediated K ϩ conduction and the effect ammonium has on this transport reaction. The rate of K ϩ uptake by these two proteins was reduced 90 -95% when the protonophore CCCP was present at a concentration of 10 M (Fig. 7A), indicating that K ϩ uptake by AmtB was dependent on the proton motive force, presumably the membrane potential component, across the cytoplasmic membrane. In support of this conclusion is the finding that neither twin-histidine variant conducted K ϩ after treatment with 2,4-dinitrophenol, a process that both depletes the cytoplasmic K ϩ pool and discharges the proton motive force (39,40), without subsequent addition of an energy source. Competition tests showed that K ϩ transport by these mutant proteins was also reduced in the presence of ammonium (Fig. 7B). The inhibitory effect was far greater for AmtB H318D (K i ϭ 4.9 Ϯ 1.9 M), which conducts ammonium, than for AmtB H168D,H318E (K i ϭ 0.77 Ϯ 0.37 mM), which lacks ammonium uptake activity. Because the GlnK protein quickly inactivates AmtB in the presence of micromolar quantities of external ammonium (43), the impact this compound had on AmtB-mediated K ϩ transport was also examined in a kdp trk kup glnK strain background. Half-saturation constant values for AmtB H318D and AmtB H168D,H318E K ϩ uptake were increased ϳ5-fold when GlnK was not present (Fig. 6B). Similar-sized decreases in ammonium inhibition of K ϩ transport by AmtB H318D (K i ϭ 22 Ϯ 5.3 M) and AmtB H168D,H318E (K i ϭ 2.9 Ϯ 1.6 mM) were observed in the absence of the GlnK protein (Fig. 7B). These results suggest that the uptake of K ϩ is influenced by AmtB-GlnK interaction and, when taken together with the observation that AmtB H318D K ϩ conduction was strongly inhibited by ammonium, provides support for the idea that K ϩ and ammonium share the same transport pathway through AmtB.

DISCUSSION
Controversy remains over the form of ammonium carried by members of the Amt family. Due to the strongly hydrophobic nature of the conduction pore it was proposed that these proteins facilitated diffusion of NH 3 (7,11). A study of AmtB function in a reconstituted system provided evidence for such a transport mechanism (7), but the findings of that work have been criticized for being interpreted incorrectly and could not be corroborated (4,44). Other observations question the plausibility of the facilitated NH 3 diffusion model. Quantitative analyses of ammonium flux have shown that this transport mechanism would not sustain the rates of growth observed under the low external ammonium conditions at which Amt proteins operate (15,44). Also, given that these proteins are active when the internal to external ammonium concentration ratio is Ͼ1 (15), facilitated diffusion would lead to the export rather than import of ammonium without the involvement of a linked energy source. The alternative view of ammonium transport holds that Amt channels actively conduct NH 4 ϩ or cotransport NH 3 and H ϩ . In support of these mechanisms, all plant  members of the Amt family functionally characterized to date, save for those from the distantly related AMT2 subfamily, have been found to mediate electrogenic ammonium uptake (4,17,18,30,45,46). Moreover, Amt proteins from a number of other organisms, including E. coli, concentrate methylammonium in a membrane potential-dependent manner (3,13,16,20). That certain AmtB twin-histidine variants conduct K ϩ against its concentration gradient (Fig. 6) is compatible with these active transport mechanisms. In addition, because K ϩ cannot separate from its charge, our results imply that ammonium migrates through the AmtB pore as either NH 4 ϩ or as a closely associated NH 3 /H ϩ pairing (see below). This is in contrast to a net NH 4 ϩ uptake mechanism that proceeds via a symport reaction in which NH 3 passes through the pore alone while a H ϩ follows a separate unidentified transport pathway.
In work presented here, we have described a set of AmtB twin-histidine variants that have altered substrate specificity and now conduct K ϩ . The spectrum of compounds (methylammonium, ammonium, and K ϩ ) carried by each mutant depends on the position and number of acidic residues present in the conduction pore. Thus, AmtB H168D transports all three compounds, AmtB H318D conducts both ammonium and K ϩ but no longer exhibits methylammonium uptake activity, and AmtB H168D,H318E carries only K ϩ (Figs. 3, 4 , and 6; and Ref. 14). These findings lead us to conclude that the His 168 /His 318 twinhistidine element serves as a substrate selectivity filter that prevents K ϩ transport. How might this element enable members of the Amt family to discriminate between K ϩ and ammonium? If, as our results predict, ammonium crosses the phenyl ring constriction and enters the conduction pore as NH 4 ϩ its charge will need to be masked to migrate across this hydrophobic environment. By accepting a H ϩ from NH 4 ϩ and only transferring it back just prior to substrate release into the cytoplasm, a transport mechanism shown to be plausible in a recent simulation study (47), the twin-histidine element allows the H ϩ to remain delocalized as it moves in parallel with NH 3 through the conduction pore. The transport of K ϩ is prohibited because, unlike NH 4 ϩ and CH 3 NH 3 ϩ , this cation is incapable of separating from its charge. Certain acidic amino acid substitutions of the twinhistidine element reduce the need for such charge separation by increasing pore hydrophilicity and, as a consequence, allow K ϩ to be carried. It is not clear why these mutations would also cause progressive decreases in CH 3 NH 3 ϩ and then NH 4 ϩ transport. However, the underlying trend suggests that this behavior results from the differences in the manner by which the introduced acidic residues handle H ϩ and K ϩ as well as changes to the hydrophobic character of the pore.
The strong preference of Amt family members for ammonium over K ϩ (3,17,18,29,30,48) implies that this selectivity is physiologically important. Our current analysis highlights why these proteins do not carry K ϩ . The inward leakage of this cation through AmtB, although able to substitute for the K ϩ uptake activities of the Trk and Kup systems, is associated with a large energy cost (Fig. 2, A and C). We suggest that this cost is incurred by the action of two processes. First, K ϩ movement across the cytoplasmic membrane is known to play an important role in regulating pH homeostasis (39,49,50). Given the elevated K ϩ pools found in cells expressing AmtB twin-histi-dine variants, substantial energy reserves would likely be used to preserve normal intracellular pH. Similarly, energy would be required to maintain the membrane potential that would otherwise be dissipated by illicit AmtB-mediated K ϩ conduction. These energy expenditures can prove problematic because members of the Amt family oftentimes need to operate in nutrient-poor environments having high K ϩ levels. For instance, the two Amt proteins of Nitrosopumilus maritimus, a member of a group of ammonium-oxidizing marine archaea that are key intermediates in the global nitrogen cycle, are likely active in the open ocean where ammonium and K ϩ concentrations are Ͻ1 M and ϳ10 mM, respectively (51)(52)(53). Growth and survival under such energy-and nitrogen-limited conditions would be severely compromised if Amt proteins did not discriminate against K ϩ .