The effects of pH on proton sugar symport activity of the lactose permease purified from Escherichia coli.

The lactose permease, which catalyzes galactoside-proton symport into Escherichia coli, has been purified and reconstituted in active form into artificial lipid vesicles. The roles of many detergents and phospholipids in solubilization and stabilization of the activity of the permease have been examined with a view to its eventual crystallization. Initial rates of uptake into reconstituted proteoliposomes determined by rapid mixing techniques proved that the activity of the permease can be comparable to that observed in the intact cell, while the best values for uptake rates obtained with conventional techniques were comparable to those reported for vesicles. The activity of the purified protein has been monitored over time periods of hours to weeks. It is shown that, under the best current conditions, the permease retains full activity for 1 to 2 weeks. Although this is still marginal for its crystallization, future improvements can now be assayed by rather stringent criteria. The mechanism of galactoside transport into reconstituted proteoliposome has been investigated by examining the effects of pH on influx into the vesicles. It is shown that the observed effects are entirely consistent with the predictions of a simple model of proton symport. The apparent increase in rate of uptake that is observed in the presence of a pH gradient is not so much due to an acceleration by a component of the protonmotive force as to the relaxation of inhibition by a product (internal protons) of the symport reaction.

been investigated in such detail in membrane vesicles (3,9). In order to investigate the structure of the permease and the molecular mechanism of the symport process in more detail, it is necessary to purify, to characterize, and to reconstitute the permease in a well defined system. Although activity has been reconstituted, a stable, highly active preparation that would be suitable for high resolution structural studies has not yet been obtained. Newman and colleagues (10)(11)(12) have purified the protein in the presence of octyl glucoside and recovered activity after reconstituting liposomes by removing the detergent through dilution. The galactoside-flux reactions catalyzed by these preparations have been investigated in some detail and shown to be similar to those of the permease in native membrane vesicles (3). Wright and colleagues (13,14) have also solubilized the permease using dodecyl maltoside and reconstituted transport activity after removal of the detergent. This preparation has been physically well characterized, and some of the galactoside-flux reactions have been described. While we were examining various kinds of detergents for solubilization, we found that the permease activity was stabilized in the presence of deoxycholate and could also be reconstituted by removal of detergent by dialysis, Here we describe the details of purification and reconstitution of the permease, and the kinetic characterization of the reconstituted system using rapid mixing techniques to measure initial rates of transport into vesicles. The dependence of uptake on external and internal pH is similar to that already determined for the intact cell (7, 8) and is entirely consistent with the role of the proton as a substrate in galactoside symport. We also demonstrate galactoside-induced proton movements, both uptake from the external medium and release within the vesicle and show that these are compatible with a stoichiometry of 1 for the symport reaction.

MATERIALS AND METHODS
Protein Purification-E. coli "206 (15) was grown in 100-liter batches in minimal medium M9 supplemented with glycerol (0.5% by volume) and casein hydrosylate (0.2% w/v). The cells were harvested by centrifugation, frozen rapidly, and stored at -20 "C. The frozen cell paste (50-100 g) was thawed in 400 ml of 10 mM Tris, 25 mM MgSOI, 1 mM sodium EDTA, 1 mM dithiothreitol, pH 7.8, containing 10 pg/ml bovine pancreatic DNase and cells broken by passage through a French pressure cell (16). The cytoplasmic membranes were collected by centrifugation in an IEC A170 rotor for 90 min at 40,000 rpm at 4 "C and then washed by resuspension in the following sequence of buffers. (i) 10 mM Tris, 10 mM sodium EDTA, 1 mM dithiothreitol, adjusted to pH 8.5 with HCl; (ii) 20 mM Tris, 3 M KCl, 1 mM dithiothreitol adjusted to pH 8.5 with acetic acid; (iii) two washes with 1 mM Tris, 1 mM sodium EDTA, 1 mM dithiothreitol, 10% glycerol, pH 8.5; (iv) 10 mM KPi, 50 mM lactose, 1 mM dithiothreitol, pH 5.8. The final pellet was resuspended in 50 ml of 10 mM KPi, 50 mM lactose, 1 mM dithiothreitol at pH 5.8 and an equal volume of freshly prepared solution of 2.5% jj"octy1 glucoside in 10 mM KPi at pH 5.8 was mixed with the membrane suspension by driving the two solutions simultaneously through a T-shaped plastic connecting piece. The mixture was stirred for 20 min and then centrifuged as before.
The supernatant was applied to a 100-ml (bed volume) column of DEAE-cellulose (Whatman DE52) that had been pre-equilibrated with elution buffer (10 mM KPi, 1 mM dithiothreitol, 1.25% P-octyl glucoside at pH 5.8), and the column was washed with 200 ml of this buffer after application of the sample. The flow-through fractions having absorbance at 280 nm greater than 0.1 were collected, pooled, and concentrated as rapidly as possible (within 2 h) to approximately l/5 of the initial volume by ultrafiltration using an Amicon Cop. PM-30 filter. The concentrate was applied to a 5-ml (bed volume) column of CM-cellulose, pre-equilibrated with the same buffer as above. After application of the sample, the column was washed with elution buffer at pH 7.2 until absorbance at 280 nm returned to baseline and then the elution buffer was changed for one comprising 50 mM potassium borate, 0.1 M KC1, 0.1% sodium deoxycholate, 1 mM dithiothreitol, 1.5% 0-octyl glucoside, pH 9.
Variations of this procedure that have been used are as follows. (i) The use of cytoplasmic membranes prepared by osmotic lysis and sucrose density gradient centrifugation (17) reduces the amount of contaminants in the DEAE-fractions but is not suitable for large scale preparations or for use of frozen cells. (ii) Replacing the sequential washing by an extraction with 50 mM potassium phosphate buffer, 50 mM lactose, 4% sodium cholate, 1 mM dithiothreitol at pH 7.8 removes many peripheral membrane proteins but also extracts up to 30% of the lactose permease; it also necessitates the addition of at least 0.1 mg/ml phosphatidylethanolamine to the solubilization buffer (11). (iii) Solubilization of the permease from washed or cholateextracted membranes in 10 mM Tris, 10 mM sodium EDTA, 1 mM dithiothreitol, 1% Lubrol-PX, 10% glycerol, pH 7.8. The permease is weakly bound to the DEAE-cellulose column and can be eluted between 0.03 and 0.07 M KC1 in the same buffer. More contaminants result in the DEAE-fraction, but these are removed by the CMcellulose step. These variations make no detectable difference to the final specific activity of the protein.
Reconstitution-The fractions containing the 280 nm absorbance peak that eluted with change in buffers in the CM-cellulose chromatography above were pooled, and the protein concentration was measured. One of the phospholipid mixtures described in Table IV, dissolved in 0.1 M KPi, 10 mM dithiothreitol, 10 mM sodium EDTA, 10% sodium cholate, 1% sodium deoxycholate, pH 9.0, was added to give a protein-to-lipid ratio of between 1:150 and 1:200 (by weight). The pH of the mixture was adjusted to 7.8 by the addition of 1 M KHZPO,, and MgS04 (1 M) was added to give 0.5 mM final concentration. Reconstitution was achieved by dialysis against 1 liter of 50 mM KPi buffer, 0.1 M KC1,0.5 mM MgS04, 0.1 mM dithiothreitol, pH 7.8. The dialysis was performed at room temperature until the onset of turbidity in order to decrease clumping of the proteoliposomes which occurred, particularly when the phospholipid mixture was rich in PE. The buffer was then changed, and dialysis was continued overnight at 4 "C. The proteoliposomes were collected by centrifugation at 10,000 X g for 20 min at 4 "C, resuspended in an equal volume of 150 mM KCl, 50 mM KPi buffer, 0.5 mM MgS04, 0.1 mM dithiothreitol, pH 7.8, and either used immediately or rapidly frozen in liquid N, and stored at -80 "C.
Proton Uptake-Proteoliposomes were reconstituted in DPPE'I DOPE/DPPG/DOPG (mixture H, Table  IV) and then given two cycles of thawing, rapid freezing, and sonication in medium comprising, for measurement of internal pH changes (proton release), 5 mM potassium phosphate, 0.27 M K&iO4, 1 mM pyranine (24), 0. valinomycin, pH 7.5, or, for measurement of external pH changes (proton uptake), 50 mM potassium phosphate, 0.25 M K2S04, 0.2 mM valinomycin, pH 7.5. To measure proton release, the vesicles (0.02 mg of protein) were diluted 100-fold into 3 ml of medium comprising 50 mM potassium phosphate, 0.25 M &SO4, 0.2 mM valinomycin, pH 7.5, and the change in pyranine fluorescence followed (excitation 460 nm, emission 508 nm (18)). Lactose, dissolved in the dilution buffer, was added to give a final concentration of 5 mM. To measure proton uptake, the vesicles were diluted into medium comprising 5 mM potassium phosphate, 0.25 M KzSO,, and 50 PM 4-methylumbelliferone, 0.2 mM valinomycin at pH 7.5. The change in fluorescence was followed (excitation 340 nm, emission 470 nm) and then lactose, dissolved in dilution buffer, was added to give a final concentration of 5 mM. Fluorescence changes were calibrated by suspending vesicles in media of known pH between 7.0 and 8.0 that included 10 p~ carbonyl cyanide m-chlorophenylhydrazone as protonophore. In an identical series of experiments, the uptake of ["Cllactose (shown by solid circles) was measured at 5-9 intervals by withdrawing 100-pl aliquots from the reaction mixture and treating as for galactoside uptake. All solutions were degassed and flushed with N P before use; the reactions were performed in stirred, sealed cuvettes at 37 "C.
Galactoside Uptake-Proteoliposomes, reconstituted as described, were given two cycles of rapid freezing, thawing, and sonication in the uptake medium described in Table I. Concentrated vesicles (0.02-0.05 mg of protein in 0.2 ml of uptake medium) were then diluted into 500 volumes of dilution medium, and samples were taken at intervals between 2 and 600 s after mixing. Each sample was quenched with an equal volume of 10 mM HgClz in dilution medium, filtered through Millipore filters (0.22-pm pore size), and washed three times with 1 ml of dilution medium containing 1 mM HgC12. Radioactivity was measured by liquid scintillation counting. To measure pH dependence, phosphate buffers were used in the range of pH 6-8. . Dihexanoyl-and dioctanoylphosphatidylethanolamines were synthesized from corresponding phosphatidylcholines by transphosphatidylation using phospholipase D (26). Two-dimensional thin layer chromatography was used to analyze the reaction products and to check the purity of phospholipids used for reconstitution. The solvent systems used were those described in Ref. 26.

RESULTS'
Kinetic Characterization-The galactoside affinity constants, determined either by equilibrium binding or by transport assays using purified protein that had been reconstituted with optimum phospholipid mixture, were similar to those described for native membrane vesicles (Tables I1 and SIV). Isotope exchange at equilibrium and counterflux that the proteoliposomes exhibited (Fig. 1, B and D and Table SIV), as well as the pH dependence of uptake ( Fig. 3 and Table 11), were comparable to those described previously for intact cells (7, 8). The apparent initial rate of uptake, determined in conventional assays, appeared to be considerably less than that observed in the intact cell, although it was in agreement with earlier reports using vesicles (3, 13). When uptake was examined using rapid mixing to initiate the reaction, significantly higher initial rates were measured (Fig. 2) which were closer to those obtained with intact cells (27), although no truly linear phase was observed. Rather, the uptake followed a curve deviating from linearity, even in the first 100 ms, and Portions of this paper (including part of "Results," Tables SI-SIV, and Fig. 1) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press.  To measure pH dependence, phosphate buffers were used in the range of pH 6-8.2 and borate buffers in the range pH 7.8-9.5. No difference was observed between the rates obtained in phosphate and borate buffers at the same pH.
For experiments with melibiose, 6-[3H]melibiose was used as the label at concentrations 0.01-2 mM for influx, 2 mM for efflux and exchange, and 0.05 mM for counterflux. Unlabeled melibiose was used at 2 rnM concentration.

TABLE I1 Kinetic parameters of lactose and melibiose uptake via the purified permease
Permease was reconstituted in E. coli PE + PG and transport assayed at 20 "C. Binding sites were determined by equilibrium binding of thiodigalactoside to the proteoliposomes. The K. and Kb are the Michaelis constants for proton and galactoside, respectively, and K , is the apparent dissociation constant for the substrate proton (7,18). These constants and kat are obtained from pH dependence of sugar uptake shown in Fig. 3. The rates of efflux, counterflux, and exchange are obtained in the same way in the appropriate experiments. In the intact cell, with lactose as substrate, they have the values Kb = 7 X M , K , = 5 X lo-' M , K. = 2.5 X lo-" (7), and,  the rate of uptake decreased quite rapidly thereafter.
The rate of uptake of lactose increased when the pH of the interior of the vesicles was raised above that of the external medium. This was due to an increase in maximum velocity ( V ) and a decrease in the apparent affinity constant for lactose (K,) (Fig. 3). In addition, a decrease in Vwas observed between pH 5.5 and 7.0 when either external or internal pH was lowered. The combination of these effects, attributed to protonation at two types of site (8), gives rise to the complicated pH profiles shown in Fig. 3 and described by Equation 1 in the legend to that figure. In Fig. 3A, where the pH is low and the first type of site is fully protonated, it is the second type of site that largely contributes to the variation in V, whereas in Fig. 3B, where the pH is high and the second type M. G . P. Page, unpublished results. of site (pK, 6.3) is deprotonated, the variation in both V and apparent K , are mostly due to changes in protonation of the first type of site (pK, 9). In Fig. 3C, the complete pH profile is shown for V, which describes the behavior of the ternary complex and hence shows inhibition at low pH values due to protonation at the second type of site. Fig. 3 0 shows the complete pH profile for the specificity constant V/K,,,, which lactose both in the usual uptake medium (Table I)  describes the behavior of the permease at infinitely low lactose concentrations (i.e. the unloaded permease). Here, there is inhibition at low pH values due to protonation at the second type of site and inhibition setting in at higher pH values due to protonation at the first type of site on the internal face of the membrane. Stoichiometry of Uptake-The vesicles could be loaded with the pH-sensitive dye pyranine (18), which allowed the proton uptake during lactose influx to be followed (Fig. 4A). Comparison of the initial rate of proton uptake (i.e. release within the vesicle) with the rate of lactose uptake in the same experiment gave values between 0.50 and 1.1 (mean and standard deviation of six experiments 0.87 2 0.22) for the stoichiometry of proton:lactose symport. Similar experiments using the pH-sensitive dye 4-methylumbelliferone to follow changes in external pH (Fig. 4B) gave a mean value of 0.98 +_ 0.31 for the proton:lactose stoichiometry.

DISCUSSION
The activity of the permease can be monitored at several levels of stringency: retention of binding, retention of exchange activity (including counterflux), and, ultimately, retention of coupled sugar-proton symport (influx and efflux). The substrate dependence of the influx reaction, catalyzed by the purified preparation, was found to be characterized by affinity constants very similar to those reported for uptake into intact cells (7, 18). The maximum velocity obtained in using conventional sampling techniques was found to be considerably less than that reported for the intact cell, but there are a number of problems associated with the accurate determination of rates of uptake into vesicles. Many of these are the result of the small internal volume of the vesicles relative to that of the intact cells. In particular, the small volume may result in a rapid build-up of internal substrate and, consequently, inhibition of further influx, even within the first seconds of uptake. Practically, this effect can be minimized by closely controlling two factors: (i) obtaining vesicles with reproducibly large diameters and (ii) determining the initial rate from points taken as early as possible in the time course of uptake experiments (28). In this respect, it is noteworthy that the maximum velocities obtained using conventional assays (with reaction times of 10 s or more) are at least 5-fold lower than the rate of uptake into intact cells (29,30) and nearly 15-fold lower than rates reported for uptake into well energized cells (31,32); the results obtained are comparable to the rate of lactose uptake into reconstitutedvesicles (standardized by thiodigalactoside binding) reported by Wright and Overath (14) and to that described by Kaback (3) for a vesicle preparation standardized by photoaffinity labeling. However, using rapid mixing and quenched flow to obtain samples between 25 and 100 ms gives estimates of the initial rate that can be nearly 10-fold higher than those obtained in the conventional assays and hence comparable to the rate of uptake into the intact cell. This suggests that product inhibition is indeed contributing quite considerably to rate limitation in the time range of the conventional assay. Of course, using rapid mixing has its own problems, one of which is that the vesicles might be deformed by high pressures developed during flow and mixing, resulting in loss of internal substrate (for example, up to 20% of the trapped pyranine is released during the reaction). Such a loss will lead to an underestimation of uptake, so that the rates obtained still have to be regarded as lower estimates of the real initial rate. Questions as to how much variation of the lateral pressure in the membrane affects the activity of individual molecules remain, as do those of the orientation of the permease molecules in the vesicles and whether this affects the observed uptake activity. The orientation of permease in these preparations has not been assessed; Seckler and Wright (33) have suggested that the permease is randomly oriented in their preparations while Carrasco et al. (34) have claimed that the molecules have native orientation but that the C terminus may be dislocated in some cases. Whatever the orientation, we measure kinetic constants for influx that are similar to those determined in the intact cell where all the permease molecules may be assumed to be uniquely oriented. Simply, this suggests either that the permease is kinetically symmetrical or that we only detect those molecules with normal orientation with respect to the driving force.
The 1actose:proton stoichiometry of uptake has been measured by comparing the initial rate of lactose uptake to the rate of either proton uptake from the medium or proton release within the vesicle, using fluorescent, pH-indicating dyes. The ratio of fluxes was found to be approximately 1, in agreement with previous findings with intact cells using pH- ( Table 11) are not detectable in the range of pH used for these experiments.
sensitive electrodes to measure external pH (4,5,19). Previous estimates of the stoichiometry of transport by reconstituted vesicle systems have been made by comparing the steady state level of galactoside accumulation to the nominal value of the protonmotive force and have yielded stoichiometries of up to 0.7 proton/galactoside (14). Galactoside-induced proton uptake by a reconstituted system was also demonstrated by Foster et al. (12), but the stoichiometry was not reported.
The reconstituted system lends itself to the facile manipulation of the intravesicular compartment. Thus, the effects of internal pH on transport activity could be investigated in a more readily controlled manner than possible in the intact cell and without the complication of the outer membrane and periplasm of the intact cell. The regulation of activity by proton concentration on either side of the membrane can be interpreted in terms of two types of proton binding sites exposed on each face of the membrane. The first type of site, which appears to represent binding of the symported proton, is involved in substrate binding, for protonation of this site on the external face (apparent pK, 8.9-9.2) increases the affinity of the permease for galactoside. Binding of the proton to this type of site exposed to the internal face of the membrane decreases the apparent affinity of the permease for external galactoside through competitive (product) inhibition. The second type of site, which has a much lower affinity for protons (pK, 6.3), appears to affect only the maximum velocity of the permease, and protonation of this site causes a marked decreaser in turnover rate between pH values of 7.5 and 5.5 on either side of the membrane. These findings appear relevant to the understanding of the mechanism of the symport reaction. Firstly, they confirm the prediction of steady state kinetic analysis that the internal protons, as a product of the reaction, should compete with influx and confirm the pH dependence described for the intact cell (7, 8). Secondly, it has previously been reported that the ApH component of the protonmotive force alters the kinetic properties of the carrier in a manner that is dependent on the square of the magnitude of ApH (35). The possibility of manipulating the pH on both sides of the membrane afforded by the reconstituted system reveals that this effect is due to rate-limiting product inhibition by internal protons. In conclusion, it appears from this kinetic characterization that the purified and reconstituted permease can have essentially the same activity as it does in its native environment in the intact cell, provided that the limitations imposed by the nature (see Miniprint Section) and the size of the vesicles are taken into consideration.
It is clear that the nature of the phospholipid environment in which the permease is reconstituted greatly affects its activity, and the specificity for PE-rich lipid mixtures has already been reported (10,36). For reasons unknown to us, we were unable to obtain full activity when PS was the major lipid species (unlike Chen and Wilson (36), who reported that PS was as effective as PE). From our results, it appeared that equilibrium binding was not greatly affected by the phospholipid environment, but reactions involving net turnover of the permease were dependent on both the nature of the phospholipid head group and the fluidity of the lipid core. Thus, it appears that the permease can be incorporated in its native state into a variety of lipid bilayers, but either is not able to turn over so rapidly in some environments as it does in others or is not as tightly coupled in some membranes as it is in others. Comparison of the initial rates of lactose uptake with those of the lactose-induced proton uptake suggests that coupling is as tight in a11 the reconstituted systems as it is in the native membranes and that the low velocities (and corresponding low accumulation ratios) are due to smaller intrin-sic rate constants for certain steps in the transport cycle in these preparations. In particular, the steps that occur after galactoside dissociation (i.e. release of proton and return of the unloaded permease) appear to be most strongly affected by the phospholipid composition. It is already known that the turnover number of permease has a much higher temperature dependence than binding, which is attributed to an effect of the fluidity of the phospholipid bilayer on one or both of these steps (18, 37).
For purification and for many methods of structural investigation, it is desirable to have the permease in monodisperse solution in the presence of detergent and for this reason the behavior of the protein with a number of detergents has been examined. In the interaction with surfactants used during reconstitution, it appears that both the hydrophilic and hydrophobic parts of the detergent play a role in maintaining activity. Detergents with short alkyl substituents, such as octyl glucoside and dihexanoylphosphatidylethanolamine, appear to be unable to support high concentrations of protein in solution. This may be because these detergents form small, very dynamic micelles that might allow the protein molecules to come into contact very frequently. The detergents with larger alkyl substituents (for example dodecyl maltoside) form larger, less dynamic micelles and apparently do not perturb the protein to such a great extent as the smaller detergents (for example, octyl glucoside will not support substrate binding while dodecyl maltoside does). The head group effect is not easily rationalized, although, comparing the decyl-or octyl-substituted detergents, it is clear there is some difference between polyoxyethylene, sugars, and other polar substituents in their interaction with the protein, manifested in the stability of the active functions.
Before embarking on a structural or mechanistic investigation of the lactose permease at the molecular level, it is essential to determine whether the isolated preparation has retained its native configuration. One of the most sensitive, and by far the most significant of the criteria that can be used, is that the activity of the preparations should correspond as nearly as possible to that of the protein in its native environment. With full expression of activity, one can be confident that the protein has retained its native structure during solubilization and purification. The protein concentrations that could be achieved are not yet high enough for high resolution structural studies, but they are sufficient first steps in this direction either by electron microscopy (38) or the examination of topography by chemical modification, as is described in the accompanying paper (44).