Altered Substrate Selectivity in a Mutant of an Intrahelical Salt Bridge in UhpT, the Sugar Phosphate Carrier of Escherichia coli *

Site-directed and second site suppressor mutagenesis identify an intrahelical salt bridge in the eleventh transmembrane segment of UhpT, the sugar phosphate carrier ofEscherichia coli. Glucose 6-phosphate (G6P) transport by UhpT is inactivated if cysteine replaces either Asp388 or Lys391 but not if both are replaced. This suggests that Asp388 and Lys391 are involved in an intrahelical salt bridge and that neither is required for normal UhpT function. This interpretation is strengthened by the finding that mutations at Lys391 (K391N, K391Q, and K391T) are recovered as revertants of the inactive D388C variant. Further work shows that although the D388C variant is null for G6P transport, movement of32Pi by homologous Pi/Pi exchange is unaffected. This raises the possibility that this derivative may have latent function, a possibility confirmed by showing that D388C is a gain-of-function mutation in which phosphoenolpyruvate (PEP) is the preferred substrate. Added study of the Pi/Pi exchange shows that in wild type UhpT this partial reaction is readily blocked by G6P but not PEP. By contrast, in the D388C variant, Pi/Piexchange is unaffected by G6P but is inhibited by both PEP and 3-phosphoglycerate. These latter substrates are used by PgtP, a related Pi-linked antiporter, which lacks the Asp388-Lys391 salt bridge but has instead an uncompensated arginine at position 391. For this reason, we conclude that in both UhpT and PgtP position 391 can serve as a determinant of substrate selectivity by acting as a receptor for the anionic carboxyl brought into the translocation pathway by PEP.

In Escherichia coli, transport of hexose phosphates is mediated by the P i -linked antiport carrier, UhpT (1)(2)(3). This well characterized transporter is one of a class of secondary transport systems that together form the Major Facilitator Superfamily (4 -6), the largest known collection of related secondary transporters. Although members of the Major Facilitator Superfamily show great diversity in their substrate specificity and kinetic mechanism, they all share a common structural theme, one characterized by the presence of approximately 12 transmembrane segments thought to transverse the membrane in an ␣-helical conformation. Direct information as to the arrangement of these helices is limited, although several po-tential models have been formulated in specific cases (7)(8)(9).
One approach to the analysis of helix packing in membrane proteins involves identification of interacting charged residues in transmembrane helices. Ordinarily, one expects the presence of an uncompensated electric charge to be highly destabilizing to such structures, due to the low dielectric of the hydrophobic environment. However, this circumstance can be mitigated if oppositely charged amino acids are brought together to form an ion pair or salt bridge (10,11). The idea that such a salt bridge may stabilize secondary structure has both theoretical and experimental support (11)(12)(13). Moreover, intraand interhelical salt bridges have been identified in both globular and in intrinsic membrane proteins (9,(13)(14)(15)(16)(17)(18).
In work described here, we used site-directed mutagenesis and second site suppressor analysis to search for salt bridges in UhpT. This approach led to identification of an intrahelical ion pairing involving Asp 388 and Lys 391 in TM11. 1 Further study indicated that this salt bridge is essential for proper UhpT function and that at least one member of this pair, Lys 391 , can play a direct role in determining substrate specificity.
Isolation of Second Site Suppressor Mutants-Strain RK5000 carrying a pTrc(HisC 6 S) derivative encoding the D388C mutation was grown overnight at 37°C in LB broth containing F6P (0.15%) and ampicillin (100 g/ml). Overnight cultures were washed and resuspended in an equal volume of M63 minimal medium salts (24), and 0.1-ml aliquots were plated on M63 minimal plates containing thiamine (2 g/ml), required amino acids (each 50 g/ml), ampicillin (100 g/ml), and F6P (0.15%) as a carbon source. Plates were incubated at 37°C for 4 -5 days, after which colonies arising from the background lawn were picked and restreaked on minimal plates to isolate single colonies. To ensure that growth was due to suppressor mutation(s) in uhpT, candidate plasmids were isolated, and uhpT was excised at its flanking restriction sites by digestion with NcoI and NgoMI. The uhpT gene was reinserted into a parental vector, which was then used to transform strain RK5000 to confirm the original phenotype. In most cases, the phenotype traveled with uhpT, and in these instances the gene was completely sequenced to identify the responsible mutation(s). * This work was supported by National Institutes of Health Grant GM24195. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  Immunoblot Analysis-SDS-polyacrylamide gel electrophoresis was performed using cell extracts without preheating in sample buffer (25). Protein was transferred to nitrocellulose and probed with a peptidedirected rabbit polyclonal antibody reactive to a UhpT C-terminal epitope (26,27). Western blots were developed using chemiluminescence (Amersham Pharmacia Biotech), and after scanning the films, UhpT expression was quantitated by densitometry of digitized images, as described (27,28).
Transport Assays-Cells grown overnight in LB broth plus antibiotic were diluted 200-fold in the same medium and grown at 37°C to a density of 2-5 ϫ 10 8 cells/ml. Cells were harvested by centrifugation, washed twice, and then resuspended in Buffer A (50 mM MOPS/K, 100 mM potassium sulfate, 1 mM magnesium sulfate, pH 7.0) to achieve an OD 660 of 1.4, equivalent to about 2 ϫ 10 9 cells/ml. After equilibration at room temperature, G6P or PEP transport was initiated by adding a one-twentieth volume of labeled substrate (final concentration, 50 M), after which aliquots were removed for filtration on Millipore filters (0.45-m pore size) and washing (twice) with 5 ml of Buffer A lacking magnesium sulfate. Transport of P i was monitored in the same way, using cells pregrown in M63 minimal medium (as above) with 0.2% glucose as carbon source. The high P i content of M63 (24) ensured maximal repression of other P i transporters (Pst and Pit), so that most P i transport occurred via UhpT.
To assay transport (and metabolism) of F6P or PEP in vivo, overnight broth cultures were diluted 1000-fold into M63 salts (above) containing 0.15% F6P or PEP as carbon source. Cultures were placed at 37°C with continuous shaking, and growth was monitored by changes in optical density at 660 nm.

Identification of an Intrahelical Salt
Bridge-Our main objective was to identify possible inter-or intrahelical salt bridges that might exist among or within UhpT transmembrane helices. The topological map of this transporter (Fig. 1), as well as results from earlier studies (28), led us to choose four cationic residues (Lys 82 , Arg 275 , Lys 391 , and Lys 404 ) and four anionic residues (Asp 279 , Glu 305 , Asp 388 , and Asp 400 ) as candidate participants in salt bridges. Our experimental approach to this question then proceeded in two phases. Initially, we replaced each target with cysteine to determine whether normal function was disrupted. For those cases in which a significant deficit was recorded, we then asked whether function was preserved if a residue of like charge was present (Table I). In this screen, only the K404C variant retained partial function, indicating that most of our target residues were essential for G6P transport by UhpT. In most of these latter cases, it appeared that this requirement was specific, because only for position 279 and 388 was it possible to substitute a residue of like charge and recover significant function (Table I).
The second step in our analysis was based on the fact that three pairs of these residues were in a registration (i, iϩ3) or (i, iϩ4) consistent with their participation in a salt bridge in an ␣-helix. Members of the remaining pair, Lys 82 and Glu 305 , were at the center of their respective helices, where they too might interact (7). Accordingly, we generated double mutants in which cysteine residues replaced each target pair (Table II). This test provided clear positive evidence that Asp 388 and Lys 391 might normally interact as an ion pair, because sugar phosphate transport in the D388C/K391C double mutant was about half that found in the wild type protein (Table II); when normalized to UhpT expression, the specific activity of this double mutant was about 75% that of the wild type. 2 In no other case, however, did the specific activity in the double mutants rise above about 1% of the parental protein. Thus, this approach did not provide evidence of interaction for our other targets.
Isolation of Second Site Suppressor Mutants-To further test the idea of an interaction between Asp 388 and Lys 391 , we isolated revertants of the inactive D388C mutation by screening for growth on F6P as a carbon source. This genetic screen was initiated with the expectation that if Asp 388 and Lys 391 interact so as to eliminate an uncompensated charge, at least some revertants of D388C might map to position 391. We performed this screen in an otherwise cysteine-less background rather than with the wild type protein, because the partially reduced activity of the cysteine-less protein (compare Tables I and III) allowed for a denser background lawn on minimal plates, 2  (30), the results of reporter gene fusions in UhpT and GlpT (31,32), and the accessibility of residues in UhpT to impermeant probes (27). Targets for mutagenesis in this study are shown as either circles or squares, indicating anionic or cationic residues, respectively.
increasing the frequency of revertants. In addition, this cysteine-less parental protein had an N-terminal polyhistidine tag (20) to facilitate later biochemical work.
As noted under "Experimental Procedures," we were able to eliminate revertants arising from mutations outside the UhpT structural gene. Those that remained contained the original D388C mutation along with a second site suppressor at position 391 in which Lys 391 was replaced by one of three uncharged residues, threonine, glutamine, or asparagine.
We characterized these revertants in several ways. For example, phenotypic growth tests in liquid medium confirmed that each of these double mutants utilized F6P (Fig. 2). Indeed, they grew slightly more rapidly than did the parental, cysteineless protein (Fig. 2). By contrast, the host strain, RK5000 (⌬uhpT), and the D388C mutant showed greatly delayed responses; eventual growth of these strains presumably reflects the appearance of adaptive revertants arising after slow growth under strong selective pressure (33,34). We also examined the three suppressor mutants with respect to their expression of UhpT and their capacity to transport G6P (Table III). In the wild type protein, the D388C mutation had reduced UhpT expression nearly 2-fold (not shown), and a similarly reduced expression was found when the D388C mutation was placed in the cysteine-less background (Table III). In both cases, however, G6P transport was reduced to near background levels (Tables I and III). By contrast, in all suppressor mutations, both UhpT expression and G6P transport were elevated to levels resembling that found in the parental protein. It is evident, then, that neither Asp 388 nor Lys 391 are required for normal UhpT function. Equally clear, the presence of one without the other is detrimental. This behavior strongly argues that Asp 388 and Lys 391 form an ion pair in TM11.
Altered Substrate Specificity of the D388C Mutant-UhpT and its relatives comprise a coherent family of transporters and receptors (4,6), each having high specificity for some organic phosphate ester and low affinity toward inorganic phosphate (1). This family has twelve bacterial members, seven of which have been functionally characterized (28). A multiple alignment of the amino acid sequences within this group (28) suggests that some residues within TM11 are highly conserved. This is not true, however, for residues assigned to the intrahelical salt bridge in UhpT, Asp 388 and Lys 391 (Fig. 3). Instead, most family members have either an aliphatic (Ala or Val) or polar (Ser, Thr, or Asn) residue at these positions. However, in one case, the phosphoenolpyruvate transporter of Salmonella typhimurium, we noted the presence of Val 388 and Arg 391 . This was of interest for two reasons. First, the presence of an apparently uncompensated positive charge in the PgtP TM11 contrasted sharply with evidence that this was detrimental to UhpT. Second, because of their acidic carboxyl group, substrates of PgtP (PEP and PGAs) (35,36) carry one more negative charge than do substrates of UhpT. For this reason, we considered it feasible that Arg 391 in PgtP might serve as an internal receptor for the anionic carboxyl on PEP and PGA. And, if so, we reasoned that the D388C variant in UhpT, which no longer transports F6P or G6P, might show an unexpected bias toward transport of PgtP substrates due to the remaining Lys 391 .
To test this idea, we exploited the fact that UhpT and PgtP (and other members of this family) each display a P i selfexchange reaction (3,37,38). Thus, if the UhpT D388C deriv-FIG. 2. Growth of UhpT and its derivatives on F6P. RK5000 carrying plasmids encoding the parental, cysteine-less UhpT (q), its D388C derivative (OE), or one of three D388C suppressors, D388C/ K391T (Ⅺ), D388C/K391N (‚), and D388C/K391Q (E), were grown in minimal medium M63 with F6P as carbon source. Strain RK5000 (f) was also examined. Growth was monitored by changes in optical density at 660 nm. The data, from a single experiment, are representative of findings made in three independent trials. a Initial rates of G6P transport were monitored by a 5-min incubation of cells with labeled substrate, as described under "Experimental Procedures." Values shown are means from three independent experiments.
b Expression is given relative to the cysteine-less parental protein (100%). Values are the means from three independent experiments. ative retained this partial reaction, one might justify an extended search for substrates other than G6P. When we examined P i transport in both the D388C mutant and its parental protein, it became clear that despite its null phenotype for G6P transport (Tables I and III), the mutant displayed considerable P i self-exchange (Fig. 4). For example, the reactions had comparable Michaelis constants (K m values of 740 and 550 M, respectively, for parent and mutant), and nearly equivalent maximal velocities (105 and 59 nmol/min/mg protein, respectively). Given that the D388C variant is expressed at only 60% of the parental level (Table III), we concluded that D388C UhpT displays an essentially normal P i self-exchange reaction.
Having found that P i self-exchange is normal in the D388C mutant, we used this reaction as a vehicle to compare substrate preference for parental and mutant proteins. This was essential because UhpT itself does not transport three-carbon organophosphates (29), and the mutant we wished to analyze does not respond to sugar 6-phosphates (Table III). Accordingly, for both wild type and mutant proteins, we ranked putative substrates with respect to their capacity to inhibit P i self-exchange. As expected from earlier work (29), this reaction was strongly inhibited by G6P or the nonmetabolizable 2DG6P in wild type UhpT, whereas PEP and 3-PGA were ineffective inhibitors (Fig. 5A). These findings contrasted sharply with results obtained with the D388C variant, which was strongly inhibited by PEP and 3-PGA but not by G6P and 2DG6P (Fig.  5B). Neither protein responded to glycerol 3-phosphate, a substrate of the related GlpT antiporter (37,38). To quantitate these observations, we calculated for each substrate an inhibition constant (K i ) from linear Dixon plots (40) (Fig. 5C and Table IV). This showed that the D388C mutation causes a 50 -100-fold decrease in the K i value for normal UhpT substrates while at the same time increasing by 20-fold the apparent affinity for PEP or 3-PGA.
These findings strongly suggested that D388C is a gain-offunction mutation in UhpT, and to test this directly we examined this protein with respect to both growth on and transport of PEP. The first test clearly documented that the D388C variant was able to use PEP as sole carbon source, in contrast to its revertants or to the wild type protein (Fig. 6A). Similarly, only the mutant, not its revertants, was able to transport PEP (Fig. 6B). Coupled with the observation that UhpT transports G6P only when the Asp 388 /Lys 391 salt bridge remains intact or when both its elements are removed, these findings lead us to conclude that in both UhpT and PgtP position 391 can serve as a determinant of substrate selectivity. DISCUSSION The study of membrane transport proteins reveals several instances in which a functional or structural role can be assigned to one or more charged residues in a transmembrane helix. The former category is well represented by several cases in E. coli, including two charged residues (Glu 126 and Arg 144 ) thought necessary for substrate binding in the lactose permease, LacY (8), a pair of cationic residues (Arg 46 and Arg 275 ) that may function as substrate recognition elements in UhpT (28), and acidic residues (Asp 51 and Asp 120 ) that help set co-ion specificity for MelB, the Na ϩ /melibiose cotransporter (41). Such residues may also act as stabilizing factors by virtue of their interaction in a salt bridge. Thus, in bacteriorhodopsin, Arg 227 and Asp 96 form an interhelical salt bridge, whereas Arg 82 and Asp 85 are involved in an intrahelical ion pair (14). Interhelical salt bridges in LacY have also been recognized (9,17); indeed, these have been indispensable to development of current models of helix packing in this symporter (9,17). Similarly, it has been proposed that Arg 499 and Glu 503 form an ion pair in the Na ϩ /glucose cotransporter, SGLT1 (42).
In the present study, we used site-directed and second site suppressor mutagenesis as tools to ask if charged residues in the UhpT hydrophobic sector might take part in salt bridges, basing our interpretation on the original analysis of salt bridges in LacY (43). This approach highlighted two residues in TM11, Asp 388 and Lys 391 , as interacting in an intrahelical ion FIG. 4. Kinetics of phosphate exchange. A, rates of P i self-exchange were estimated as described under "Experimental Procedures" for strain RK5000 without (f) and with plasmids encoding cysteine-less UhpT (q) or its D388C derivative (OE). B, the kinetics of P i self-exchange for cysteine-less UhpT (q) and its D388C derivative (OE), presented as recommended by Hofstee (39).

FIG. 3. Conserved residues in TM11.
Multiple alignment of TM11 sequences for P i -linked transporters and receptors with ϳ30% identity to UhpT. Residues conserved in at least five family members are highlighted with shading; charged residues are shown in bold type. The alignment is modified from Fann et al. (28) by addition of GlpT from Rickettsia prowazekii (52).
pair. Thus, sugar phosphate transport was not found after individual replacement of either Asp 388 or Lys 391 , but function was retained when uncharged residues replaced both of them (Tables II and III).
Initially, the null phenotype arising when Asp 388 and Lys 391 were removed individually led us to believe that UhpT function was disrupted due to inappropriate helix packing, but this view did not easily explain why a related antiporter, PgtP (32% identity), contained an apparently uncompensated arginine at position 391. We therefore considered the possibility that lack of function in some of our derivatives was related to altered substrate specificity rather than a change in structure. This seemed more likely when we found that P i self-exchange was unaffected by the D388C mutation (the K391C mutant has not been tested). Subsequently, by using this partial reaction as a screening tool, we showed that D388C is a gain-of-function mutation that biases substrate preference toward PEP and 3-PGA and away from G6P and F6P (Figs. 4 -6 and Table IV).
Two questions are raised by our findings. 1) What is the molecular basis of the new substrate specificity in the D388C variant? 2) Why is there at the same time a discrimination against normal UhpT substrates? No definitive answers can be offered to either question, but a single perspective does address both issues. An early model (44) specified that two electropositive centers in UhpT associate with a pair of negative charges brought into the active site by the anionic substrates, P i and G6P; indeed, we recently identified two required arginines (Arg 46 and Arg 275 ) that may fulfill this function (28). If the configuration that accepts sugar phosphate must achieve an FIG. 6. Growth on and transport of PEP. Growth on (A) and transport of (B) PEP was measured as described under "Experimental Procedures." A, RK5000 without (f) and with plasmids expressing cysteine-less UhpT (q), its D388C derivative (OE), or second site suppressors, D388C/K391T (Ⅺ), D388C/K391N (‚), and D388C/K391Q (E), were grown in M63 with PEP as a carbon source. Growth was monitored as described in the legend of Fig. 2. B, transport of PEP was examined for D388C UhpT (OE) and various controls (f), including RK5000, and RK5000 expressing cysteine-less UhpT and its suppressors, D388C/ K391T, D388C/K391N, and D388C/K391Q.

TABLE IV
Inhibition constants for UhpT substrates Inhibition constants (K i ) were determined for each test substrate as an inhibitor of P i self-exchange, using linear Dixon plots (40) as reported earlier (see Fig. 5), with the assumption of competitive interaction between P i and the test substrate. G3P, glycerol 3-phosphate.  5. Substrate specificity of UhpT and its D388C derivative. P i self-exchange was estimated for both cysteine-less UhpT (A) and its D388C derivative (B) in the presence of G6P (q), 2DG6P(E), PEP (OE), or 3-PGA (‚) using 32 P i at 0.1 mM along with the indicated concentrations of competing substrates. In the absence of competing substrate, 32 P i transport varied between 8.5-15.5 nmol/min/mg protein for cysteine-less UhpT, and 6 -11.7 nmol/min/mg protein for its D388C variant, and these values were used to normalize values for the individual trials. C, graphical presentation of the data using the method of Dixon (40): cysteine-less UhpT with G6P (F) or PEP (OE); the D388C variant with G6P (E) or PEP (‚). K i values for these and other test substrates are shown in Table IV. electroneutrality, one might expect that substrates such as PEP and PGA, which carry an additional negative charge relative to G6P, might require the presence of a third positive center before productive binding can occur. The second question is answered in much the same way. That is, when the D388C variant is presented with G6P, failure to achieve an electrostatic balance at the sugar phosphate binding site would be associated with an abortive carrier-substrate complex.
Although such arguments account for our findings related to transport of PEP or G6P, this reasoning does not directly explain why the P i self-exchange is unimpaired in the D388C mutant. At the least one must conclude that TM11 is not involved in binding of P i . This might be expected if the P i self-exchange exploits a different configuration of the protein (44), possibly relying predominantly on factors related to either or both of the two essential arginine residues, one of which lies in TM1 (Arg 46 ), the other in TM7 (Arg 275 ). The absence of either of these inactivates G6P transport (28), but in light of the work reported here, it could be instructive to examine the P i self-exchange reaction as these arginines are removed individually.
The finding of PEP transport by the D388C mutant also leads one to question the stoichiometry and electrical character of the reaction. The kinetic parameters of P i self-exchange by D388C deviate little from its parent (Fig. 4), implying that heterologous P i /3-PGA or P i /PEP antiport by this mutant would retain the exchange stoichiometry shown by wild type UhpT during transport of G6P. In this latter case, early work suggests that UhpT selects monovalent P i to use in an electroneutral exchange with divalent G6P, with an overall 2-for-1 stoichiometry (44). A similar 2-for-1 stoichiometry during heterologous P i /PEP antiport is therefore expected to be electrogenic, carrying negative charge inward, because for our assay conditions (pH 7) PEP or 3-PGA mainly exists as the trivalent anion. Because this reaction would occur against the cellular membrane potential (inside negative), one might also expect, as is found, a relatively low growth yield on PEP (Fig.  6A) and a relatively low rate of PEP transport (Fig. 6B). Although we have not yet studied the electrical character of heterologous exchange in D388C, it is clear that this single mutation (D388C) could have mechanistic impact on more than one attribute of the antiport reaction.
That substrate specificity can be altered by change of a single amino acid residue has been demonstrated for several transporters. In LacY, for example, point mutations lead to a preference for maltose (45), sucrose (46), or malto-oligosaccharides (47). Similarly, point mutants in the melibiose carrier of E. coli can show altered cation and/or sugar specificity (48,49). Most recently efficient transport of lactose was described for a mutant of the E. coli maltose ABC transporter (50). Even among this group, however, our finding that the D388C mutation alters the substrate selectivity of UhpT is unusual. Given that the new preference for PEP or 3-PGA reflects interaction between Lys 391 and the substrate carboxyl, we infer that in wild type UhpT the C1 position of G6P lies on TM11 in the neighborhood of the Asp 388 /Lys 391 ion pair. Of additional interest is the possibility that the approaches described here might let us assign specific functional roles to other charged residues. For example, TM7 contains two charged residues, Arg 275 and Asp 279 (Fig. 3), one of which (Arg 275 ) is known to lie on the translocation pathway (51) and is believed to aid in substrate recognition (28). Arg 275 is conserved among relatives of UhpT (28), but only UhpT has a charged residue at position 279. This situation is reminiscent of that found in TM11, where one of a pair of charges (Asp 388 ) is unique to UhpT, whereas the other (Lys 391 ) is more broadly represented (Fig. 3). Although our tests did not identify the Arg 275 /Asp 279 pair as a simple salt bridge, we should not rule out the possibility that these residues associate such that a more general function of Arg 275 is moderated by Asp 279 to optimize the features specific to UhpT.