The role of tryptophans 371 and 395 in the binding of antibiotics and the transport of sugars by the D-galactose-H+ symport protein (GalP) from Escherichia coli.

The interactions between the D-galactose-H+ symporter (GalP) from Escherichia coli and the inhibitory antibiotics, cytochalasin B and forskolin, and the substrates, D-galactose and H+, have been investigated for the wild-type protein and the mutants Trp-371-->Phe and Trp-395-->Phe, so that the roles of these residues in the structure-activity relationship could be assessed. Neither mutation prevented photolabeling by either [4-3H]cytochalasin B or by 3-[125I]iodo-4-azidophenethyl-amido-7-O-succinyldesacetylforskolin ([125I]APS-forskolin). However, measurements of protein fluorescence show that both residues are in structural domains, the conformations of which are perturbed by the binding of cytochalasin B or forskolin. Moreover, both mutations cause a substantial decrease in the affinity of the inward-facing site of the GalP protein for cytochalasin B, 10- and 43-fold, respectively, but have little effect upon the affinity of this site for forskolin, 0.8- and 2.6-fold reductions, respectively. Both these mutations change the equilibrium between the putative outward- (T1) and inward-facing (T2) conformations, so that the inward-facing form is more favored. They also stabilize a different conformational state, "T3-antibiotic," in which the initial interactions between the protein and antibiotics are tightened. Overall, this has the effect of compensating for the reduction in affinity for cytochalasin B, so that the respective overall Kd values are 0.74- and 3.5-fold that of the wild type, while causing a slight increase, 1.5- and 3.2-fold, respectively, in affinity of the mutants for forskolin. The Trp-371-->Phe mutation causes a 15-fold reduction in the affinity of the inward-facing site for D-galactose, suggesting that this residue forms part of the sugar binding site. In contrast, the Trp-395-->Phe mutation has no effect upon the affinity of the inward-facing site for D-galactose. These effects may be related to the reduction in galactose-H+ symport activity only in the Trp-371-->Phe mutant, although it still effects active transport to the same extent as the Trp395-->Phe mutant. However, there is a 10-20-fold increase in the Km values for energized transport of D-galactose for both mutants.

The efficiency of photolabeling GLUT1 with antibiotics is maximal at 280 nm, suggesting that a tryptophan residue may be activated prior to the covalent attachment of the antibiotic. Moreover, the binding of antibiotics induces a conformational change in GLUT1, where the environment around one or more of the tryptophans must be perturbed. The sites of covalent attachment for cytochalasin B and forskolin in GLUT1 have been localized to putative ␣-helices 10 and 11 by identifying the photolabeled proteolytic fragments from GLUT1 (Cairns et al., 1984(Cairns et al., , 1987Holman et al., 1986;Holman and Rees, 1987;Karim et al., 1987;Wadzinski et al., 1990). The tryptic digestion of the GalP protein, which has been photolabeled with [4-3 H]cytochalasin B, also produces a labeled fragment (M r 17,000 -19,000) of almost identical M r to that produced by digestion of labeled GLUT1 (Cairns et al., 1991). There are two tryptophan residues in GLUT1 that lie within this region, Trp-388 and Trp-412, toward the cytoplasmic ends of putative helices 10 and 11, respectively (Mueckler et al., 1985). Furthermore, cytochalasin B is thought to bind to the inward or cytoplasmic facing site (Deves and Krupka, 1978;Gorga and Lienhard, 1981), so that these tryptophan residues seem likely candidates for forming a dynamic segment of the transporter. The role of these residues has been investigated in several studies in which mutant proteins, produced by site-directed mutagenesis, were tested for their ability to bind antibiotics and transport sugars (Katagiri et al., 1991(Katagiri et al., , 1993Garcia, et al., 1992;Schurmann, et al., 1993). The eukaryotic expression systems used did not provide sufficient amounts of the GLUT1 mutant proteins to allow them to be characterized in detail. The only firm conclusion reached by these studies is that neither Trp-388 nor Trp-412 is essential for photolabeling with either cytochalasin B or forskolin but changing both residues does abolish photolabeling (Inukai et al., 1994).
The alignment of the amino acid sequence of GalP with that of GLUT1 indicates that these tryptophan residues are conserved in GalP and indeed within most members of this extensive family of sugar transport proteins (Griffith et al., 1992). It seems highly likely that these residues, Trp-371 and Trp-395, respectively, in the GalP sequence , play similar roles in the function of the GalP and GLUT proteins. The present investigation has achieved a rigorous characterization of the role(s) of these residues in the binding of antibiotics, D-galactose, and protons, and in the subsequent translocation events. This investigation was only possible because the wildtype and mutant GalP proteins could be produced in high yield, from a strain of E. coli that overexpresses GalP , which is amenable to detailed transient kinetic studies (Walmsley et al., 1994a(Walmsley et al., , 1994bMartin et al., 1995).
In this paper we have determined the precise changes in the kinetic constants caused by mutagenesis of the two highly conserved tryptophan residues in GalP on D-galactose transport. Changes in the inhibition by, and binding of, cytochalasin B and forskolin were rigorously quantified. Also, the effect of these mutations on photolabeling with [4-3 H]cytochalasin B and [ 125 I]APS-forskolin, and on galactose-H ϩ symport activities, were determined. Moreover, we have identified these changes with different conformational states of the protein and with the distribution between inward-and outward-facing forms.
Growth of the GalP-overproducing E. coli strains-The E. coli strains JM1100 (pPER3), JM1100 (pTPM7), and JM1100 (pTPM8), used for the overexpression of the wild-type and mutated GalP protein, were grown on minimal medium according to the method of .
Preparation of Inner Membranes Containing Overexpressed Protein-E. coli cells were disrupted by explosive decompression in a French press at 137.5 megapascals, and inner membranes were prepared essentially as described by Osborn et al. (1972). This procedure yields predominantly inside-out vesicles (Futai, 1978).
Quantification of Overexpressed Protein-A sample of the membrane preparation (30 g of protein) was subjected to SDS-polyacrylamide gel electrophoresis. After staining with Coomassie Blue, the gels were scanned with a Molecular Dynamics 100A computing densitometer, and the proportion of GalP protein was measured. The levels of expression of the wild-type and mutant proteins were detemined by comparison of the intensities of the bands corresponding to the GalP proteins on Coomassie Blue-stained SDS-polyacrylamide gels. The levels of expression of the wild-type, Trp-371 3 Phe and Trp-395 3 Phe mutant proteins were 50, 66.9, and 68.5% of inner membrane protein, respectively.
Protein Assays-The concentration of protein in the membrane preparations was assayed by the method of Schaffner and Weissman (1973).
Oligonucleotide-directed, Site-specific Mutagenesis-A point mutation was introduced into the DNA coding for the GalP protein of E. coli by oligonucleotide-directed mutagenesis, in which the codon (TGG) for either tryptophan 371 or 395 was replaced by that for phenylalanine (TTC). The single-stranded DNA template for mutagenesis was prepared by subcloning a 2.46-kilobase pair AlwI-AlwI DNA fragment coding for the C-terminal half of GalP from plasmid pPER3, which confers overexpression of GalP, into bacteriophage M13mp18. Mutagenesis was carried out with the Amersham in vitro mutagenesis kit using mutagenic primers (5Ј-GCACAGTACGAAAATCAGCG-3Ј and 5Ј-GTTGGCAATGAAGTTGGTGG-3Ј for the Trp-371 and Trp-395 mutants, respectively) (mismatches are underlined) complementary to the coding strand of galP. The base changes were confirmed by singlestranded dideoxynucleotide sequencing. A 1.561-kilobase pair MluI-NdeI fragment containing the base changes was isolated from the bacteriophage M13 construct and substituted for the wild-type DNA in plasmid pPER3 to give plasmids pTPM7 (Trp-371 3 Phe) and pTPM8 (Trp-395 3 Phe). The substitution was confirmed by subcloning the DNA containing the mutation back into bacteriophage M13mp18 and dideoxynucleotide sequencing. Double-stranded sequencing of the mutant galP genes in the plasmids confirmed that there were no other base changes in the galP gene sequence. The sequencing used reaction kits from Amersham or Applied Biosystems Inc., and the products were separated on a Pharmacia Macrophor apparatus purchased with a grant from SmithKline Beecham plc or on an ABI 373-A automated DNA sequencer operated by the School of Biological Sciences, University of Leeds.
Sugar Transport Measurements-The energized transport of [1-3 H]Dgalactose into E. coli strain JM1100 harboring plasmids pPER3, pTPM7, and pTPM8 was measured as described by Henderson et al. (1977). The transport of [1-3 H]D-galactose was assayed in cell suspensions washed in 150 mM KCl, 5 mM MES, pH 6.5, after growth to an A 680 of 0.6 in rich medium (1% TY: 10 g of yeast extract, 10 g of tryptone, and 5 g of NaCl in 1 liter) containing 20 mM glycerol and 20 g/ml thymine. Transport measurements were always performed on cells that had been grown on rich media, which reduced the level of GalP expression to less than 2% of the total membrane protein. The plasmids were maintained by the addition of 15 g/ml tetracycline. Initial rates of [1-3 H]D-galactose transport were measured at 15 s over an appropriate range of concentrations, from about K m 0.5 to K m 5.
Assay of Sugar-H ϩ Symport-Galactose-promoted pH changes were measured with energy-depleted anaerobic suspensions of intact cells in 150 mM KCl, 2 mM glycylglycine, pH range 6.3-6.5, as detailed by Henderson and Macpherson (1986).
Permeabilization of the Cell Outer Membrane for Inhibition Studies-Cell cultures were grown as described previously for transport measurements and washed twice in the growth volume with 150 mM KCl, 5 mM MES, pH 6.5. The cells were then resuspended to 1 ⁄20th of the growth volume in 200 mM Tris-Cl, pH 8.0, and allowed to equilibrate for 10 min at room temperature. An equal volume of 200 mM Tris-Cl, pH 8.0, 1 mM EDTA was then added, and the cells were left at room temperature for 30 min. 100 volumes of 150 mM KCl, 5 mM MES, pH 6.5, 10 mM MgSO 4 was then added and the cells sedimented by centrifugation. The cells were resuspended to an A 680 of 2.0 in 150 mM KCl, 5 mM MES, pH 6.5, and aliquots (0.5 ml) were assayed for [1-3 H]D-galactose uptake (50 M) after incubation with 80 M cytochalasin B or forskolin at room temperature for 3 min (see sugar transport measurements).
Fluorescence Studies-Fluorescence spectra were measured in a Jasco FP777 spectrofluorimeter. The protein was excited at 297 nm and the fluorescence emission monitored between 300 and 550 nm. Rapid reactions were followed using an Applied-photophysics (London, United Kingdom) spectrofluorimeter, operated at 20°C, as described by Walmsley et al. (1993Walmsley et al. ( and 1994aWalmsley et al. ( and 1994b. Unless stated otherwise, fluorescence measurements were carried out in 50 mM potassium phosphate, 100 mM NaCl, and 1 mM EDTA, buffered to pH 7.4 at 20°C. Routinely the membranes were used at a protein concentrations of 200 g/ml, in the above buffer. Statistical Analyses-The apparent K m 2 and V max 2 values, with their standard deviations, for transport of [1-3 H]D-galactose were obtained by a least squares fit of the unweighted data directly to an hyperbola using ULTRAFIT program (Biosoft) processed on Macintosh personal computers. Generally, the traces for the quench in the intrinsic fluorescence of GalP, when it was mixed with the antibiotics in a stopped-flow apparatus, were fitted to a double-exponential function using the nonlinear least squares regression program supplied with the Applied Photophysics SF.17MV stopped-flow apparatus. The data sets, for the concentration dependence of the rate of binding of the antibiotics, were analyzed by fitting to the appropriate equation using the nonlinear least squares regression program SIGMA PLOT (Jandel Scientific).

Mutations in Trp-371 and Trp-395 Impair D-Galactose Transport Activity of GalP
The transport of D-galactose by wild-type GalP, under energized conditions, was characterized by an hyperbolic increase in the initial rate with increasing concentrations of sugar. A least squares fit yielded K m and V max values of 42-59 M and 59 -65 nmol/mg/min, respectively ( Table I).
The substitution of Trp-371 by Phe in the D-galactose-H ϩ symporter (GalP) seriously impaired its ability to transport D-galactose under energized conditions, with more than a 10fold increase in the K m and a 6-fold decrease in the V max , compared with the wild-type (Table I). We can assume that ⌬p is unaffected at the low levels of GalP expression applying in these experiments (see "Materials and Methods"). While the V max term is largely governed by the rate constants for reorientation, the K m term is governed by both the rate constants for translocation and the true affinity of the sugar binding site (Walmsley, 1988;Lowe and Walmsley, 1986). The rate of sugar translocation will be governed by the rate constants for reorientation of the unloaded and sugar-loaded transporter. The increase in the K m for this mutant may simply reflect a reduction in the rate constant for reorientation of the loaded transporter, rather than any change in affinity. An increase in the rate constant for reorientation of the unloaded transporter would lead to a decrease in K m .
In contrast, for the Trp-395 3 Phe mutant, there was about a 20-fold increase in the K m , but the V max was only reduced to about half that of the wild type (Table I). This suggests that this mutation has a more pronounced effect on the affinity of the transporter for sugar, rather than on the rate constants for translocation.
In both cases the data indicate a significant reduction in the specificity (V max /K m ) of the mutant transport proteins for Dgalactose relative to the wild-type transport protein. The specificity of the Trp-371 mutant was reduced 56-fold and that of the Trp-395 mutant 26-fold, relative to the wild type (Table I). However, it was demonstrated that the mutant proteins catalyze energized transport by measuring the maximum accumulation of 0.05 mM D-galactose in the presence and absence of the uncoupler DNP (2 mM). The intracellular concentration of Dgalactose was measured at 4.2, 0.216, and 0.277 mM for the cells expressing wild type Trp-371 3 Phe and Trp-395 3 Phe protein, respectively (using an intracellular volume of 2.059 l/mg dry weight of cells). The concentration of intracellular D-galactose for cells treated with DNP was measured at 0.036 mM, 0.028 mM and 0.036 mM for cells expressing the wild type, Trp-371 3 Phe and Trp-395 3 Phe proteins, respectively, indicating that the cells were completely uncoupled. A comparison of the intracellular D-galactose concentrations of cells treated with and without DNP showed that D-galactose was accumulated 116.67-, 7.71-, and 7.69-fold by cells expressing the wild-type Trp-371 3 Phe and Trp-395 3 Phe proteins, respectively (Fig. 1). The use of the host strain JM1100, which has a GalK phenotype and is therefore unable to metabolize D-galactose, ensured that any effects of DNP on the metabolic enzymes would have no effect on the accumulation of D-galactose We conclude that both mutants are still capable of energized sugar uptake against a concentration gradient, albeit at an efficiency 1% of wild type.

Photolabeling of GalP by Antibiotics Is Not Abolished by the Trp-371 3 Phe and Trp-395 3 Phe Mutations
The wild-type GalP protein was susceptible to photolabeling with [4-3 H]cytochalasin B and [ 125 I]APS-forskolin, which could be significantly inhibited by D-galactose, but not L-galactose, as

comparison of the steady-state kinetic parameters for the uptake of D-galactose mediated by wild-type and mutant GalP proteins
Initial rates of [1-3 H]D-galactose transport were measured at 15 s over an appropriate range of concentrations, from about K m 0.5 to K m 5. The maximal rate of transport (V max ), and the K m values were determined by fitting each data set to the Michaelis-Menten equation. The accumulation ratios were calculated as the maximal uptake of 0.05 mM D-galactose in the presence and absence of the uncoupler DNP (2 mM). shown in Fig. 2 (Cairns et al., 1991;Martin et al., 1994). The hexose D-galactose is the physiological substrate for the GalP protein, which has little, if any, affinity for L-galactose (Cairns et al., 1991;Martin et al., 1994;Walmsley et al., 1994b). This protection by substrate indicated that both cytochalasin B and [ 125 I]APS-forskolin specifically label the GalP protein. Previous studies showed that the wild-type GalP protein is not susceptible to photolabeling with forskolin (Martin et al., 1994), but this antibiotic is a transport inhibitor (Martin, 1993;Martin et al., 1994Martin et al., , 1995. The labeled mutant GalP proteins were found to migrate to identical positions to the wild-type GalP protein on the gels (Fig. 2), confirming their identity.
The Trp-371 mutant GalP protein was less readily photolabeled with [4-3 H]cytochalasin B, but more readily photolabeled with [ 125 I]APS-forskolin than the wild-type protein (Fig. 2); D-galactose was less effective at protecting the mutant against photolabeling, by either inhibitor, than the wild-type protein (Fig. 2).
In the case of the Trp-395 3 Phe mutant the extent of photolabeling with cytochalasin B was moderately less than wild-type, whereas photolabeling with [ 125 I]APS-forskolin was greater for the mutant than the wild type. However, substrate protected the Trp-395 3 Phe mutant against photolabeling by both compounds better than its protection of the Trp-371 3 Phe mutant.
We conclude that neither Trp-371 nor Trp-395 is essential for photoreaction with [4-3 H]cytochalasin B and [ 125 I]APS-forsko-lin, although we cannot rule out the possibility that the Phe residues are susceptible to photolabeling.

Inhibition of D-Galactose Transport into Wild-type and Mutant GalP Strains by Antibiotics
The abilities of (80 M) cytochalasin B or (80 M) forskolin to inhibit the uptake of (50 M) D-galactose by energized cells, from wild-type and mutant GalP strains of E. coli, were determined. The cell walls of the intact cells were rendered permeable to the antibiotics by prior treatment with Tris-EDTA, pH 8 (see "Materials and Methods").
In the case of the wild type, both cytochalasin B and forskolin inhibited the transport of D-galactose, but cytochalasin B was a more potent inhibitor (Fig. 3). The reason for this difference is not clear, since titration of the protein fluorescence of GalP, with these antibiotics, indicates that the GalP protein has a similar overall affinity (K d ) for both of them (Table II).
Interestingly, cytochalasin B and forskolin were more potent inhibitors of the Trp-371 3 Phe mutant mediated D-galactose transport, relative to the wild type. This is most likely due to the greatly impaired D-galactose binding capability of the Trp-371 3 Phe mutant protein (higher K d ) together with the higher affinity (lower K d ) that the mutant has for both cytochalasin B and forskolin compared with the wild-type protein (Table II). Furthermore, forskolin was more potent than cytochalasin B with the Trp-371 3 Phe mutant (Fig. 3), consistent with the higher affinity (lower K d ) of the mutant for forskolin compared with cytochalasin B (Table II).
In contrast, both cytochalasin B and forskolin were less effective inhibitors of the transport of D-galactose mediated by the Trp-395 3 Phe mutant, than by the wild-type, and were much less potent against the Trp-395 3 Phe than against the Trp-371 3 Phe mutant (Fig. 3). The antibiotic forskolin had a significantly greater inhibitor potency than cytochalasin B, with the Trp-395 3 Phe mutant. This probably reflects the substantial difference in the overall affinity (K d ) of the Trp-395 mutant GalP for forskolin and cytochalasin B (Table II).

Fluorescence Studies of the Binding of Antibiotics
Comparisons of the Steady-state Protein Fluorescence Changes Induced by the Binding of Cytochalasin B and Forskolin-The ability to overexpress the wild-type and mutant GalP proteins enabled a much more rigorous analysis of the kinetics of the binding of cytochalasin B and forskolin by monitoring the associated changes in tryptophan fluorescence of GalP (Walmsley et al., 1994a;Martin et al., 1995).
In order to assess the importance of Trp-371 and Trp-395 in the binding of cytochalasin B and forskolin, the intrinsic fluorescence levels of the wild-type and mutant proteins were titrated with these antibiotics in the stopped-flow apparatus. Although the changes in tryptophan fluorescence of the mutant proteins were less than for the wild-type protein, they were sufficient to allow detection of the binding of both cytochalasin B (Figs. 4B and 5B) and forskolin ( Fig. 4D and 5D) to both mutants. In each case, the stopped-flow records were best fitted to a double-exponential function. The total amplitudes of the fluorescence changes were determined from the sum of the amplitudes of the two phases and are plotted as a function of each antibiotic concentration in Figs. 4, B and D, and 5, B and D. A fit of each set of data to a quadratic equation for a second-order binding process (Walmsley et al., 1993) yielded the K d values, and maximal fluorescence changes, for cytochalasin B and forskolin given in Table II. The wild-type protein has a similar overall affinity for cytochalasin B and forskolin, with K d values of about 1.8 -1.9 M (Table II). The binding of cytochalasin B causes a larger quench in the protein fluorescence, of 7.6%, than forskolin, which only causes a 4.2% quench (Table II).
There was a slight increase in the overall affinity of the Trp-371 3 Phe mutant GalP protein for both cytochalasin B and forskolin, which was concomitant with a small decrease in the fluorescence quench, relative to the wild type. However, the differences from the wild type are small.
The Trp-395 3 Phe mutation had a much more pronounced effect, causing a 3.5-fold reduction in the overall affinity for cytochalasin B and 3.2-fold increase in the affinity for forskolin (Table II). Indeed, while the wild-type protein has similar affinities for cytochalasin B and forskolin, the Trp-395 3 Phe mutant has an 11-12-fold difference in affinities for these antibiotics. In addition, the fluorescence quenches with these antibiotics were reduced to about half those for the wild type (Table II).
Comparisons of the maximal fluorescence changes, corrected for the relative levels of expression, indicated that both Trp-371 and Trp-395 contribute to the fluorescence changes associated with the binding of cytochalasin B and forskolin to the GalP protein. Trp-395 contributes much more to the protein fluorescence signal, with cytochalasin B and forskolin, than Trp-371. Thus, although it is likely that neither residue is photolabeled by these antibiotics, they are both reporters of antibiotic binding, indicating that they form part of a dynamic segment, or segments, of the transporter, which might be part of the anti-biotic binding site. Although neither residue is solely responsible for the fluorescence changes associated with the binding of cytochalasin B and forskolin, together, they are probably responsible for most, if not all, of the signals.
Comparisons of the Transient Kinetics of the Binding of Cytochalasin B and Forskolin-Variations in the rate constants for the fast and slow phases in the binding of the antibiotics are informative of the kinetic mechanism (Walmsley et al., 1993(Walmsley et al., , 1994aMartin et al., 1995). We have previously shown for GalP, GLUT1, and AraE sugar transport proteins that the rate of the fast phase increases linearly with the antibiotic (A) concentration, while that of the slow phase is constant. The fast phase was attributed to the formation of the transporter-antibiotic complex (T 2 (A)), while the slow phase was attributed to reorientation of the transporter between inward and outward-facing conformations (T 1 -T 2 ). The following kinetic scheme was proposed for the binding of antibiotics, in which T 1 and T 2 are, respectively, the outward-and inwardfacing forms of the unloaded transporters. In addition, for GLUT1 there was an additional state T 3 (CB) (Walmsley et al., 1994a). T 2 (A) and T 3 (A) are two different conformations of the transporter-antibiotic complex. For such a model the apparent dissociation constant of the transporter-antibiotic complex is given by the following equation, with K 1 , K 2 , and K 3 defined as K 1 ϭ k Ϫ1 /k 1 , K 2 ϭ k Ϫ2 /k 2 , and K 3 ϭ k Ϫ3 /k 3 . The second-order association rate constant (k 2 ) and dissociation rate constant (k Ϫ2 ) for the transporter-antibiotic complexes were determined from the slope and intercept, respectively, of a plot of the antibiotic concentration dependence of the rate of the fast phase, and the dissociation constants for these complexes were calculated from the rate constants (K 2 ϭ k Ϫ2 /k 2 ) ( Table II). As shown in Table II, the wild-type GalP protein was characterized by similar association rates for cytochalasin B and forskolin, but the dissociation of cytochalasin B was significantly slower than forskolin. Thus, the T 2 , or putative inwardfacing transporter, has a higher affinity for cytochalasin B than forskolin.
In the case of wild-type GalP, the measured overall K d for cytochalasin B is 4.6-fold greater than K 2 (Table II), indicating that at least 78.4% of the transporters are in the T 1 (or putative outward-facing) conformation prior to the binding of cytochalasin B (K 1 Ն (K d /K 2 ) Ϫ 1) (Walmsley et al., 1994a). As such, the kinetics of the binding of cytochalasin B do not lend any sup-

TABLE II The kinetic parameters for the binding of cytochalasin B and forskolin to wild type GalP and mutant GalP proteins
The kinetics of the binding of antibiotics to the wild-type and mutant GalP proteins were determined by stopped-flow fluorescence spectroscopy, monitoring the antibiotic induced quench in protein fluorescence. A fit of the amplitude data to a quadratic equation for a second-order binding process, as described under "Results," yielded the maximum percent quench in fluorescence and the overall affinity (K d ). The association (k 2 ) and dissociation (k Ϫ2 ) rate constants were determined from the antibiotic concentration dependence of the rate of signal change (protein fluorescence quench) as described under "Results." The rate constants are numbered so as to be consistent with kinetic Scheme I. In each case, the dissociation constant (K 2 ) for the putative inward-facing transporter (T 2 ) was calculated from the measured values for k 2 and k Ϫ2 (K 2 ϭ k Ϫ2 /k 2 ). port to the existence of the postulated T 3 (CB) conformation for GalP. However, there is kinetic evidence to support the existence of all four postulated states of the human erythrocyte glucose transporter (GLUT1) Lienhard, 1985, 1989;Walmsley, 1986, 1987;Walmsley and Lowe, 1987;Walmsley, 1988;Walmsley et al., 1994a). On the other hand, forskolin is bound 3.5-fold less tightly, to the wild-type GalP protein, in the initial complex (e.g. T 2 (forskolin)), due to its faster rate of dissociation (Table II). Accordingly, for forskolin, the overall K d is only marginally greater than K 2 . However, the binding of forskolin is biphasic, indicating that this process involves at least two steps. Since there is an equilibrium between the T 1 and T 2 conformations in the absence of bound ligand, K 1 and 1/K 3 must have similar values when forskolin is bound (Martin et al., 1995).

Cytochalasin B Forskolin
The Trp-371 3 Phe mutant was also characterized by a concentration-dependent linear increase in the rate of the fast phase of the binding of cytochalasin B (Fig. 4A) and forskolin (Fig. 4C), yielding the kinetic parameters given in Table II. For cytochalasin B, this mutation decreases the association rate constant, while the dissociation rate constant is increased, so that the affinity of the T 2 conformation was reduced by 10.2fold, relative to the wild type. The overall affinity of the mutant for cytochalasin B is similar, or slightly greater, to that of the wild-type (K 2 Ͼ K d for mutant), which indicates that, following the formation of the initial complex, there is a further tightening in the binding of cytochalasin B, as the mutant GalP protein undergoes the T 2 (CB)-T 3 (CB) transition. Thus, the T 3 (antibiotic) state is now detectable in this mutant.
In contrast, the binding of forskolin to the Trp-371 3 Phe mutant GalP protein was characterized by only small changes in the association and dissociation rate constants; these were marginally higher and lower, respectively, than the wild-type values, so that the mutant protein had a higher overall affinity (lower K d ) for forskolin (Fig. 5C, Table II).
More significantly, the Trp-395 3 Phe mutant was characterized by a substantial reduction in the cytochalasin B association rate constant and a moderate increase in the dissociation rate constant, so that the affinity of the T 2 conformation was dramatically reduced by 42.7-fold, relative to the wild-type (Fig. 5A, Table II). Again, the reduction in the affinity of the T 2 conformation was not matched by a similar reduction in the overall affinity for cytochalasin B, indicative of the formation of the T 3 (CB) conformation.
Surprisingly, the Trp-395 3 Phe mutant was characterized by an apparently hyperbolic increase in the rate of the fast phase of binding of forskolin (Fig. 5C). These data provide clear evidence that the transporter-forskolin complex undergoes a further conformational change, the T 2 (forskolin) to T 3 (forskolin) transition. Presumably, for the wild-type protein, this transition occurs at a rate that is too fast to be measured by stopped-flow fluorescence spectroscopy. The data in Fig. 5C were fitted to a hyperbolic function yielding respective minimal and maximal rates of 3.6 (Ϯ 7.0) s Ϫ1 and 28.4 (Ϯ 5.8) s Ϫ1 and an apparent K d of 3.7 (Ϯ 2.7) M. These values will correspond to those for k 3 , k Ϫ3 , and K 2 , respectively, in Scheme I. Using these values and the measured K d , K 1 can be calculated as 0.41 from Equation 1.
The Trp-395 3 Phe mutation has caused a large shift in the equilibrium between T 1 and T 2 , so that about 29% of the transporters are in the T 1 (or outward-facing) conformation in the absence of ligands, while stabilizing the T 3 (forskolin) complex (K 3 ϭ ([T 2 (forskolin)]/[T 3 (forskolin)] ϭ 7.7). After equilibration with forskolin, the GalP-forskolin complex is present as 11.3% T 2 (forskolin) and 88.7% T 3 (forskolin). Furthermore, the affinity of the T 2 conformation for forskolin was reduced 2.5-fold. The above data also provide minimal values for k 2 (k 2 Ͼ (k 3 ϩ k Ϫ3 )/K 2 ) and k Ϫ2 (k Ϫ2 Ͼ k 3 ) of 8.6 s Ϫ1 and 28.4 s Ϫ1 , respectively, indicating that the reduction in the affinity of the T 2 conformation is largely attributable to an increase in the dissociation rate constant.
Accordingly, the substitution of Trp-395 for Phe causes a destabilization of the T 1 (or outward-facing) conformation (K 1 Ն 3.63 for the wild-type GalP protein (Walmsley et al., 1994a), K 1 ϭ 0.41 for Trp-395 3 Phe mutant GalP protein), while stabilizing the T 3 (antibiotic) conformation. A value of 1.2 can be calculated for K 3 , in the binding of cytochalasin B to Trp-395 3 Phe mutant protein, since K 1 is known. Thus, at equilibrium, the GalP-cytochalasin B complex is present as 54.5% T 2 (CB) and 45.5% T 3 (CB). This value for K 3 for the T 3 (cytochalasin B) complex compares with a value of 7.7 for the T 3 (forskolin) complex, so that this state is more stable with bound forskolin.

The Binding of D-Galactose to the Trp-371 3 Phe Mutant Is Substantially Reduced
We have shown previously that the affinity of the GalP protein for D-galactose can be determined by measuring the D-galactose inhibition of the apparent rate of binding of cytochalasin B; this technique provides a measure of the affinity of the inward-facing site for the sugar (Walmsley et al., 1993(Walmsley et al., , 1994a(Walmsley et al., , 1994b. When the rate data, determined as a function of both the D-galactose and cytochalasin B, were fitted to an equation for competitive inhibition, this procedure yielded a value of 5.8 mM for wild-type GalP (Walmsley et al., 1993(Walmsley et al., , 1994a). An alternative procedure for determining the K d for D-galactose was also developed during the course of the present studies, which had the advantage of being simpler. The vesicles were equilibrated with cytochalasin B, at a concentration above the K d , and then mixed with varying concentrations of D-galactose in the stopped-flow. The displacement of bound cytochalasin B, presumably from the inward-facing site, by the sugar led to an increase in the protein fluorescence. The signal amplitude increased with the sugar concentration in a hyperbolic manner and a fit of the data to a hyperbolic function indicated a K d for D-galactose of 6.42 mM, after correction for competitive displacement by the cytochalasin B (Table III).
The rate of binding of cytochalasin B to the Trp-371 3 Phe mutant GalP protein was determined as a function of both the D-galactose and cytochalasin B concentrations (Fig. 6a). The data were fitted to an equation for competitive inhibition (Walmsley et al., 1993(Walmsley et al., , 1994a, yielding values for the cytochalasin B association and dissociation rate constants of 2.1 s Ϫ1 and 12.0 s Ϫ1 , respectively, and an apparent K d for D-galactose of 88.7 mM. The association and dissociation rate constants compare well with those determined independently by monitoring only the binding of cytochalasin B (Table II). The K d for D-galactose can also be determined from a fit of the amplitude data (protein fluorescence quench) to an equation for competitive inhibition (Walmsley et al., 1993), yielding a value of 78.6 mM (Fig. 6b). In addition, the K d for D-galactose was also determined by cytochalasin B displacement, yielding a value of 71.7 mM (Table III). When these K d values for the mutant are compared with those for the wild-type, they indicate that the Trp-371 3 Phe mutation causes a 15-fold decrease in affinity of the inward-facing conformation of GalP for D-galactose. This is consistent with the fact that D-galactose affords less protection against photolabeling of the mutant than the wild-type.
Unfortunately, the small signal change associated with the binding of cytochalasin B to the Trp-395 3 Phe mutant GalP precluded an analysis of the inhibition, by D-galactose, of the cytochalasin B binding rate, in order to provide a measure of the affinity of this mutant for D-galactose. However, it was possible to obtain a value of the K d for D-galactose by displacement of bound cytochalasin B, which yielded a value of 6.26 mM (Table III). This suggests that the Trp-395 3 Phe mutation produces little, or no change, in the affinity for D-galactose under nonenergized conditions. On the other hand, the K m for the transport of D-galactose by the mutant, under energized conditions, is approximately 20-fold higher than for the wildtype. However, the mutant must still bind protons since its affinity for D-galactose is increased under energized conditions relative to that under nonenergized conditions (K d (inhibition of the binding of cytochalasin B) Ͼ K m (transport measurements)) by 5.6-fold. This difference in the K d and K m values for the Trp-395 3 Phe mutant (5.6-fold) is much lower than for the wild-type (98-fold) or the Trp-371 3 Phe mutant (127-157fold). The fact that D-galactose affords similar levels of protection to photolabeling by cytochalasin B and [ 125 I]APS-forskolin, under nonenergizing conditions, of both the Trp-395 3 Phe mutant and the wild-type GalP proteins is consistent with the similar K d values for D-galactose of these proteins. Sugar-H ϩ symport activity can be assayed by measuring alkaline pH changes associated with the addition of high concentrations of sugar to anaerobic suspensions of de-energized intact bacteria (Karin et al., 1987;Henderson and Macpherson, 1986). Representative traces shown in Fig. 7 illustrate that the addition of 3.3 mM galactose to cells carrying the Trp-395 3

TABLE III A comparison of the dissociation constants for the binding of D-galactose to wild-type and mutant GalP proteins
The binding of D-galactose to wild-type and mutant GalP proteins was monitored as the net increase in protein fluorescence when the sugar was used to displace cytochalasin B bound to the proteins. The proteins were equilibrated with a concentration of cytochalasin B that was sufficient to achieve near saturation of the proteins: wild-type, Trp-371 3 Phe and Trp-395 3 Phe proteins were equilibrated with cytochalasin B to final (mixing chamber) concentrations of 12.5, 5, and 12.5 M, respectively. The vesicles were mixed with D-galactose in a stopped-flow device and the increase in protein fluorescence measured. The signal amplitude increased in a hyperbolic manner with the D-galactose concentration and the apparent K d was determined from an hyperbolic fit of the data. This procedure yielded apparent K d values of 48.7 (Ϯ14.3) mM, 325.9 (Ϯ151.2) mM and 17.9 (Ϯ6.3) mM for the wild-type, Trp371Phe and Trp395Phe proteins respectively. The large error for the Trp-371 3 Phe protein is largely attributable to the low affinity for D-galactose restricting the D-galactose concentration range to an upper limit of 1.5 K d , in order to avoid osmotic effects. The D-galactose concentration was varied over a range that spanned at least 0.5-10 K d for the other two proteins. The apparent K d values were corrected for competitive inhibition by the cytochalasin B using the K d values given in Table II Phe mutation in GalP yielded rates of H ϩ transport (2.62 Ϯ 0.97 (5) nmol H ϩ /min/mg) comparable with those for the wildtype protein transport (3.32 Ϯ 1.54 (8) nmol H ϩ /min/mg). In complete contrast, cells carrying the Trp-371 3 Phe mutation did not show any sugar-H ϩ symport activity ( Fig. 7 and four additional measurements in two separate cultures). This important difference from both the wild-type and the Trp-395 3 Phe mutation is discussed further below. It is possible that much higher concentrations of D-galactose need to be added to reveal sugar-H ϩ symport, given the much reduced affinity of this mutant for sugar (see above), but this could not be achieved in the experimental system used for measuring the pH changes.

DISCUSSION
From the results in this paper we conclude that residues Trp-371 and Trp-395 are located in conformationally dynamic positions within the GalP symporter, with both contributing to the changes in tryptophan fluorescence induced by the binding of antibiotics. Although neither residue is photolabeled by cytochalasin B or forskolin, this does not preclude the possibility that these residues interact directly with the antibiotics, which would be consistent with the observations discussed below.
The mutation of Trp-371 and Trp-395 to phenylalanines reduces the affinities of the initial GalP-cytochalasin B complexes by at least an order of magnitude. However, this is compensated for by a further conformational change in which the binding of the cytochalasin B is tightened, so that the overall affinities of the two mutants for cytochalasin B are similar to that of the wild-type protein.
In contrast, the affinities of the initial GalP-forskolin complexes formed by the two mutant proteins are similar to that of the wild-type protein. But again, there is a further tightening in the interaction between the antibiotic and the transporter, so that forskolin is bound more tightly by the mutants than by the wild-type, indicating that these residues interact differentially with cytochalasin B and forskolin.
The differential interaction of cytochalasin B and forskolin with GalP is best illustrated by the Trp-395 3 Phe mutant. The replacement of Trp-395 by Phe causes about a 43-fold reduction in the affinity of the initial complex for cytochalasin B, but only a 3-fold reduction for forskolin. There is about a 3.5-fold reduction in overall affinity for cytochalasin B, while the overall affinity for forskolin is enhanced by about 3.2-fold. The increase in the overall affinity relative to those of the initial complexes is largely due to destabilization of the outward-facing conformation (Ͼ78% of wild-type GalP molecules are outward facing in the absence of ligands), so that the inward-facing conformation of the Trp-395 3 Phe mutant, to which the antibiotics bind, is favored in the absence of ligands (K 1 ϭ 0.41 and 71% are inward-facing). This is similar to the situation that prevails for human erythrocyte glucose transporters (GLUT1), where the inward-facing conformation progressively predominates at subphysiological temperatures (K 1 ϭ 0.25 and 80% of the transporters face inwards at 20°C; Lowe and Walmsley, 1987;Walmsley, 1988). Furthermore, in both the case of the Trp-395 3 Phe mutant GalP and GLUT1, there is evidence that these proteins undergo a further transition following the binding of cytochalasin B (e.g. in each case K d Ͻ K 2 ): the T 2 (CB) to T 3 (CB) transition that further stabilizes the protein-cytochalasin B interaction.
The nature of the T 2 (antibiotic) to T 3 (antibiotic) transition is not known, but one possibility is that it represents a partial reorientation of the transporter (Martin et al., 1995). The binding of antibiotics may induce the closure of the binding site, by acting as transition-state analogues, which do not allow complete reorientation. In the case of wild-type GalP, there is no kinetic evidence to support the existence of the T 3 (CB) state (Walmsley et al., 1994a) but there is for the T 3 (forskolin) state (Martin et al., 1995). This suggests that the reorientation process may proceed to a smaller extent with cytochalasin B than with forskolin (Martin et al., 1995). This suggestion is consistent with the present studies, which show that the Trp-395 3 Phe mutant binds forskolin more tightly than cytochalasin B (K 3 ϭ 1.2 and 7.7 for cytochalasin B and forskolin, respectively). Moreover, the T 2 to T 3 transition for the Trp-395 3 Phe mutant is at least an order of magnitude faster when cytochalasin B, rather than forskolin, is bound. In the case of wild-type GalP and the Trp-371 3 Phe mutant, this transition is too fast to be measured by stopped-flow fluorescence spectroscopy for both cytochalasin B and forskolin. Clearly, the T 3 (CB) and T 3 (forskolin) states are not equivalent and may be reached by different pathways.
The much higher value of the overall K d for D-galactose (0.67 mM, Walmsley et al. 1994b) compared with K m (0.06 mM, Table  I) implies that energization increases the affinity of the sugar binding site (Cairns et al., 1991;Walmsley et al., 1994). The affinity of the inward-facing site for D-galactose was determined to be 5.8 mM by the inhibition of the rate of binding of cytochalasin B (Walmsley et al., 1994b). The affinity of this site was also determined by D-galactose titration of the fluorescence of 8-anilino-1-napthalenesulfonate bound to GalP, yielding a K d of 7.2 mM, from which the affinity of the outward-facing site was calculated to be 13.8 mM (Walmsley et al., 1994b). A comparison of the K d values obtained for D-galactose binding to the mutants with that for the wild-type will be informative of any changes in the sugar binding site. On the other hand, a comparison with the K m values for transport should be informative as to whether these mutations have affected the communication between the proton and sugar binding sites. The K d for the binding of D-galactose to the inward-facing site of the Trp-371 3 Phe mutant GalP was 11-15-fold higher than for the wildtype, suggesting that this residue may be involved directly in the interaction with the internal sugar. The reduction in the affinity of the sugar binding site probably accounts for the increase in the K m for transport, which is about 10-fold higher than that for the wild type. In contrast, the K d for the binding of D-galactose to the inward-facing site of the Trp-395 3 Phe mutant GalP was similar to that of the wild-type, suggesting that this residue does not interact directly with the sugar. However, the transport K m was increased about 20-fold relative to the wild type, so that it is only about 5.6-fold lower than the K d for sugar binding to the inward-facing site. This might have indicated that the primary effect of the Trp-395 3 Phe mutation was to decrease the coupling between the putative proton and sugar binding sites, but its sugar-H ϩ symport activity was hardly diminished, while that of the Trp-371 3 Phe mutant was apparently abolished. However, both mutants were still able to accumulate galactose to about the same extent, in an uncoupler-sensitive manner. This implies that the Trp-371 3 Phe mutant is not fundamentally impaired in H ϩtranslocation, although its apparent affinity for protons may be diminished by changes in its binding of sugars. These intriguing observations are the subject of further investigations, aimed at determining the site(s) of H ϩ binding, the participation of H ϩ in the discrete kinetic events we have identified, and its possible pathway during translocation through the GalP protein. Garcia et al. (1992) have mutated the corresponding tryptophan residues in GLUT1 (Trp-388 and Trp-412) to leucine and glycine residues. Essentially, they found that the GLUT1 Trp-388 mutants had a decreased sensitivity of transport to cytochalasin B, while the sensitivity of the Trp-412 mutants was unaltered. This contrasts with GalP, where the Trp-371 mutant-mediated D-galactose transport is more sensitive to cytochalasin B inhibition than the wild-type, which is more sensitive than the Trp-395 mutant. For GLUT1, the Trp-412 mutant, but not the Trp-388 mutant, was shown to have reduced transport activity. Both of the Trp mutants of GalP have reduced transport activities, with the Trp-371 mutant having the lowest specificity constant. On the other hand, Katagiri et al. (1991 and have reported that both the Trp-388 3 Leu and Trp-412 3 Leu mutants of GLUT1 have specificities (V max / K m ) that are reduced to 20 and 15% of the wild type. However, the Trp-388 3 Leu mutation was reported to cause a decrease in the specificity constant (V max /K m ) for zero-trans uptake that was not compensated for by a corresponding change in the specificity constant for zero-trans efflux (Katagiri et al., 1993); this contradicts the Haldane relationship for a passive transport system and is impossible (V max /K m (influx) ϭ V max /K m (efflux); Lowe and Walmsley, 1987). Both of the mutants had reduced levels of photolabeling with antibiotics (Katagiri et al., 1991(Katagiri et al., , 1993. Schurmann et al. (1993) have reported that the Trp-388 3 Leu mutation decreases both the photoaffinity labeling of GLUT1, with [ 125 I]APS-forskolin, and the glucose FIG. 7. A D-galactose-promoted alkaline pH change is demonstrated by the Trp-395 3 Phe mutant of GalP but not by the Trp-371 3 Phe mutant. E. coli strain JM1100 expressing either the wild-type, Trp-371 3 Phe, or Trp-395 3 Phe mutant GalP proteins were grown overnight on rich media containing 20 mM glycerol as described under "Materials and Methods." D-Galactose (3.3 mM) was added to an anaerobic suspension of cells of the indicated strain (17.3 mg, dry mass) in 3.7 ml of 150 mM KCl, 2 mM glycylglycine, pH range 6.3-6.5. The pH was measured as described by Henderson and Macpherson (1986). An upward deflection represents an alkaline pH change. transport activity; while the Trp-412 3 Leu mutation does not affect the photolabeling but abolishes transport activity. These later studies are more consistent with those reported herein on GalP, in which the transport activity (V max /K m ) for the Trp-371 3 Phe and Trp-395 3 Phe mutants is reduced to 1.7 and 3.8% of the wild-type level, respectively. However, in contrast to GLUT1, the low specificity constant for the Trp-395 3 Phe mutant was shown to be largely due to a substantial increase in the K m for transport under energizing conditions.