Proximity of Periplasmic Loops in the Metal-Tetracycline/H 1 Antiporter of Escherichia coli Observed on Site-directed Chemical Cross-linking*

Our previous study on second-site suppressor mutations of the Tn10-encoded metal-tetracycline/H 1 antiporter suggested that Leu 30 and Ala 354 , located in periplasmic loop 1–2 and 11–12, respectively, are confor-mationally linked to each other (Kawabe, T., and Yamaguchi, A. (1999) FEBS Lett. 457, 169–173). To determine the spatial proximity of these two residues, cross-linking gel-shift assays of the L30C/A354C double mutant were performed after the mutant had been oxidized with Cu 2 1 / o -phenanthroline. The results indicated that Leu 30 and Ala 354 are close to each other but that Gly 62 , which is located in cytoplasmic loop 2–3, and Ala 354 are distant from each other, as a negative control. Then, a single Cys residue was introduced into each of the six periplasmic loop regions (P1-P6), and eleven double mutants were constructed. Of these eleven double Cys mutants, the L30C/A354C and L30C/T235C mutants showed a mobility shift on oxidation, indicating that P1 is spatially close to P4 as well as P6. In contrast,


The
Tn10-encoded metal-tetracycline/H ϩ antiporter (TetA(B)) 1 is a typical bacterial drug export protein (2,3), and its molecular mechanism and molecular structure have been studied as a paradigm of antiporter-type drug exporters including bacterial multidrug exporters (4 -6). TetA(B) belongs to a major facilitator superfamily (MFS) (7); however, the direction of the coupling of substrate transport with protons is opposite to that in the case of MFS symporters such as lactose permease.
The transmembrane helix arrangement in lactose permease has been extensively studied by use of double Cys mutants as an alternative means of three-dimensional structure prediction without x-ray crystallographic analysis (8 -10). It is of interest to find out whether antiporters and symporters are the same with regard to the fundamental molecular construction or whether the difference in the coupling direction reflects the helix arrangement. The twelve-transmembrane structure of TetA(B) has been experimentally confirmed (11), and the exact range of each transmembrane segment was determined on the basis of site-directed chemical modification of cysteine-scanning mutants (12)(13)(14)(15). To evaluate the mutual proximity of transmembrane helices of TetA(B), we introduced a Cys residue into each periplasmic loop on the basis of a Cys-free mutant.
To obtain double Cys combinations, we first looked for a pair for which close proximity was expected. In our previous paper (1), we reported that both G62L, which is in cytoplasmic loop 2-3, and G332S, which is in cytoplasmic loop 10 -11, are suppressed by the second-site mutations of both L30S and A354D, which are located in periplasmic loop 1-2 and loop 11-12, respectively. G62L and G332S mutations both have a similar conformational effect on positions 30 and 354, resulting in the embedding of these residues in the hydrophobic interior of the membrane. Second-site mutations at Leu 30 and Ala 354 prevent the remote conformational distortion caused by the G62L and G332S mutations (1,16). These results indicate the possibility that positions 30 and 354 are spacially close to each other. Thus, the L30C/A354C double mutant was prepared and the cross-linking was examined, with the G62C/A354C double mutant as a negative control. As a result, we could detect intramolecular cross-linking as an SDS-PAGE mobility shift of the double mutant after oxidation; then we introduced a Cys residue into each of the six periplasmic loop regions and constructed double Cys mutants. The results of the cross-linking assay indicate that the two ends, P1 and P6, are close to each other and that one of the central loops, P4, is close to P1. Such a periplasmic loop arrangement is similar to that of lac permease (17), indicating that the MFS transporters may have a general three-dimensional structure despite the difference in the proton-coupling direction.

EXPERIMENTAL PROCEDURES
Single Cys Mutants-Single Cys mutants were constructed by oligonucleotide-directed site-specific mutagenesis according to the method of Kunkel (18). For the mutagenesis, plasmid pCTC377A (12) was used as a template, where the Cys 377 3 Ala mutation was introduced into pCT1183 (19), which carries the 2.45-kilobase pair Tn10 tetA and tetR gene fragments. Mutations were detected as the appearance of a newly introduced restriction site and verified by DNA sequencing. To avoid unexpected additional mutations, a cassette of the mutated fragment * This work was supported by grants-in-aid from the Ministry of Education and the Ministry of Science and Technology of Japan. 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 U.S.C. Section 1734 solely to indicate this fact.
after sequencing was exchanged with the corresponding DNA fragment of the wild-type plasmid. Low copy number mutant plasmids were constructed through exchange of the BglII-BamHI fragment of the cysteine-scanning mutant tetA gene with the corresponding fragment of the low copy number plasmid pLGT2 (20) and used for the construction of double Cys mutants.
Double Cys Mutants-Double Cys mutants were constructed by recombination of a DNA fragment of one single Cys mutant gene with the corresponding fragment of the other single Cys mutant gene. L30Ccontaining double Cys mutants were constructed by exchanging the BglII-EcoRV fragment of the L30C DNA with the corresponding fragment of the other single Cys DNA. When the S92C, S156C, or G62C mutation was combined with the T235C, S296C, or A354C mutation, the BglII-EcoRI fragment of the former mutant was used for recombination.
Disulfide Cross-linking-Crude membranes containing double Cys mutants were prepared by brief sonication of Escherichia coli W3104 cells carrying a mutant plasmid. Disulfide cross-linking was carried out by incubation of the membrane suspension (1.6 mg of protein/ml) in 180 M Cu 2ϩ /o-phenanthroline (CuPh) and 50 mM potassium phosphate buffer (pH 7.5) at 25°C for 30 min. The reaction was terminated by the addition of EDTA (final concentration, 5.7 mM). The samples were mixed with ␤-mercaptoethanol-free SDS sample buffer, followed by SDS-polyacrylamide gel electrophoresis. The TetA(B) bands were visualized by Western blotting using anti-TetA(B) C-terminal antiserum (21). As a control, the sulfhydryl groups of the double Cys mutants were masked by preincubation with 33 mM N-ethylmaleimide (NEM) at 30°C for 10 min prior to the cross-linking reaction. When the effect of This structure is predicted on the basis of the results of site-directed competitive chemical modification (11). The residues subjected to cysteine-introducing mutagenesis in this study are depicted as bold letters.  tetracycline was examined, the membranes were preincubated with 1 mM tetracycline and 50 mM MgSO 4 for 5 min at 30°C prior to the oxidation. Cross-linking by Dimaleimide Derivatives-o-Phenylenedimaleimide (PDM), m-PDM, p-PDM (Wako Pure Chemical Ind., Co., Ltd., Osaka, Japan), and 1,6-bis(maleimido)hexane (BMH) (Pierce, Illinois) were used as cross-linking reagents. A membrane suspension (1.6 mg of protein/ml) was incubated with 0.5 mM cross-linking reagent in 50 mM potassium phosphate buffer (pH 7.5) at 25°C for 30 min. Reactions were terminated by the addition of ␤-mercaptoethanol-free SDS sample buffer. The cross-linking bands were detected as described above. Table I shows the drug resistance of E. coli W3104 cells carrying the plasmid encoding the double Cys mutant tetA(B) gene. All of the double Cys mutants showed significant tetracycline resistance except for G62C/A354C, indicating that these mutants fundamentally maintained the normal conformation.

Drug Resistance Levels of the Double Cys Mutants-
Intramolecular Cross-linking of the L30C/A354C Double Cys Mutant-Leu 30 and Ala 354 are located in putative periplasmic loops P1 and P6 (Fig. 1). The topology shown in Fig. 1 was partially revised from a previous figure (11) on the basis of our recent results relating to the chemical modification of cysteinescanning mutants around transmembrane segment 1 (TM1). 2 In the current topology, Leu 30 is at the middle of P1 whereas Ala 354 is near the boundary between TM11 and P6. The L30C and A354C single mutants and the L30C/A354C double mutant showed the same mobility (34 kDa) as the Cys-free TetA(B) during SDS-PAGE ( Fig. 2A); however, after oxidation of the membranes with Cu 2ϩ /o-phenanthroline, only the L30C/A354C double Cys mutant gave a new band (30 kDa) exhibiting significantly increased mobility in the gel ( Fig. 2A) in addition to the non-oxidized band. The 30-kDa band was not observed when the sulfhydryl groups were masked by preincubation of the membranes with NEM ( Fig. 2A). In addition, when oxidized L30C/A354C membranes were reduced with ␤-mercaptoethanol, the 30-kDa band was not observed (Fig. 2A). The 30:34-kDa band ratio depended on the temperature and the period of oxidation; however, the 34-kDa band remained under all con-FIG. 3. CuPh-catalyzed disulfide cross-linking of double Cys mutants containing the L30C mutation. The L30C/S92C, L30C/ S156C, L30C/T235C, and L30C/S296C double Cys mutants were oxidized with CuPh (-), or their SH groups were masked by NEM prior to oxidation (ϩ).
ditions. On the basis of the mobility of the 30-kDa band and disappearance of the band on masking of the SH groups or under reducing conditions, it is concluded that the 30-kDa band of the L30C/A354C double Cys mutant observed under oxidizing conditions is due to the conformational shrinkage caused by intramolecular cross-linking. These results indicate that positions 30 (P1) and 354 (P6) are spatially close to each other.
As a negative control, the G62C/A354C double Cys mutant was oxidized under the same conditions as the L30C/A354C double Cys mutant. In our previous study, it was confirmed that Gly 62 is located on the cytoplasmic surface of the membrane (16). As expected, the G62C/A354C double Cys mutant showed no mobility shift under oxidizing conditions (Fig. 2B).
Cross-linking of Double Cys Mutants Containing the L30C Mutation-Combinations of the L30C mutation and a Cys mutation in each periplasmic loop region were constructed. Of the four double Cys mutants, only L30C/T235C produced a 30-kDa band like the L30C/A354C mutant. Masking by NEM and reduction with ␤-mercaptoethanol completely prevented this mobility shift (Fig. 3). Thus, it is concluded that positions 30 (P1) and 235 (P4) are close to each other. The other three combinations of L30C with S92C, S156C, and S296C did not produce a 30-kDa band (Fig. 3). However, in the case of the L30C/S92C and L30C/S296C mutants, a 64-kDa band corresponding to a dimer appeared under oxidizing conditions (Fig.  3). The 64-kDa band disappeared with masking by NEM. The 64-kDa band was also observed for the S92C and S296C single Cys mutants but not for the L30C single Cys mutant (data not shown). Therefore, the 64-kDa band clearly represents the dimer formation due to Cys 92 -Cys 92 or Cys 296 -Cys 296 intermolecular cross-linking, suggesting that positions 92 and 296 are located at the periphery of the transmembrane helix bundle. On the other hand, the L30C/T235C mutant did not produce a 64-kDa band. The L30C/S156C mutant produced a 71-kDa band, the appearance of which was independent of oxidation.
Cross-linking of Other Double Cys Mutants-The double Cys mutants containing S156C combined with L30C (Fig. 3), T235C (Fig. 4A), and A354C (Fig. 4A) all did not produce a 30-kDa band under oxidizing conditions. The S156C/S296C double mutant produced an oxidation-dependent 64-kDa band that drastically decreased with masking of SH groups by NEM (Fig. 4A). The S156C single Cys mutant did not produce a 64-kDa band. This finding also supported the Cys 296 -Cys 296 intermolecular cross-linking. The S156C/T235C mutant gave an unidentified and oxidation-independent 71-kDa band similar to the L30C/ S156C mutant.
The double Cys mutants containing S92C combined with L30C (Fig. 3), T235C, S296C, and A354C (Fig. 4B) all did not produce a 30-kDa band under oxidizing conditions. On the other hand, all of these double Cys mutants produced an oxidation-dependent 64-kDa band, confirming the Cys 92 -Cys 92 intermolecular cross-linking. In the case of the S92C/T235C mutant, the 64-kDa band was very thin but significant. The S92C/ S296C double Cys mutant produced 98 and 118-kDa bands corresponding to a trimer and a tetramer, respectively, in addition to the 64-kDa band, indicating that both Cys 92 and Cys 296 in this mutant can undergo intermolecular cross-linking and confirming the location of these positions at the periphery of the helix bundle.
As to the S92C/S296C double Cys mutant, several crosslinking reagents for SH groups were tested. As shown in Fig.  4C, all of the cross-linking reagents tested here caused the formation of dimers, trimers, and tetramers; however, the degree of the multimer formation was in the following order: p-PDM Ͼ BMH Ͼ m-PDM Ͼ o-PDM. NEM pretreatment prevented the multimer formation with cross-linking reagents. These results supported the hypothesis that positions 92 and 296 are located at the periphery of the periplasmic helix bundle of TetA(B).
Effect of Tetracycline on Cross-linking-In our previous study, we found that conformational changes in the cytoplasmic and periplasmic loop regions were induced in the presence of tetracycline (22). Thus, the effect of tetracycline on the crosslinking was examined. Among the eleven periplasmic double Cys mutants, only the cross-linking of the L30C/T235C mutant was significantly prevented in the presence of tetracycline (Fig.  5), suggesting that the distance between P1 and P4 and/or the orientation of the side chains could be changed by tetracycline. DISCUSSION In this study, we showed that intramolecular disulfide crosslinking could be detected as a mobility shift during SDS-PAGE of double Cys mutants under oxidizing conditions. On the basis of this mobility shift, we investigated the proximity of the periplasmic loop regions. As a result, it was revealed that P1 was close to P6 and P4. The P2 and P5 loops seem to be located on the outside surface of the helix bundle. These results indicate that in the periplasm, the terminal loops are proximal to each other in addition to one of the central loops, and the second two loops from the terminal are at the periphery of the helix bundle. Tetracycline caused alteration of the distance between and/or orientation of P1 and P4 but did not affect the distance between P1 and P6. Our site-directed chemical modification studies on the cysteine-scanning mutants of TetA(B) 3 exhibited that helices 3, 6, 9 and 12 are completely embedded in the hydrophobic interior of the membrane whereas helices 2, 5, 8, and 11 face a water-filled channel throughout their length. On the other hand, helices 1 and 7 face the channel only at their periplasmic halves whereas helices 4 and 10 face the channel at their cytoplasmic halves; that is, these four helices tilt with respect to the channel. On the basis of the results of site-directed chemical modification studies and the current results of cross-linking studies, we presented a model of the helix arrangement of TetA(B) on the periplasmic surface of the membrane as shown in Fig. 6.
In this study, we examined the intramolecular cross-linking on the basis of a mobility shift during SDS-PAGE caused by a cross-linking-induced conformational change of TetA(B). This method is remarkable because it can be performed with intact membrane proteins retaining full transport activity; however, when the cross-linking-induced conformational change is small, the cross-linking may be missed in the gel shift assay. Split proteins (9) or protease digestion of protease-site-introduced proteins (17) has been generally used to confirm intramolecular cross-linking. Unfortunately, because split TetA(B) proteins exhibit reduced transport activity (23) and the efficiency of the membrane assembly by the Cys mutants of the split TetA(B) was very low, we could not use split TetA(B) proteins. Similarly, protease-site-introduced TetA(B) proteins could not be used in this experiment. Therefore, further investigation is required to determine whether the Cys residues in the double Cys mutants showing no mobility shift during oxidation are really distant from each other. However, the prox-imity of the Cys residues in the double Cys mutants showing a mobility shift is beyond doubt.
Cross-linking studies on membrane transporters have been extensively performed by Kaback et al. (8), with lactose permease, and were revealed to be very useful for estimating the helix arrangements of polytopic membrane proteins. For periplasmic loops of lactose permease, Sun and Kaback (17) reported that P1 is close to P4 and P6, and P2 and P5 are at the periphery. The cross-linking between P1 and P4 was affected by the ligand-induced conformational change (24). The results of the cross-linking studies on the periplasmic loops are strikingly similar to the current results for TetA(B). As for the transmembrane segments of lactose permease, TM1 is close to TM5, 7, and 11 (10,25). TM5 is also close to TM7 and 8 at the periplasmic side (27). If cross-linking can be formed across the water-filled channel, our model for helix packing of TetA(B) was perfectly consistent with the cross-linking results for lactose permease. Goswitz and Brooker (26) presented a general model for the transmembrane helix arrangement of MFS transporters. When considering the tilting of helices 1, 4, 7, and 10, our model is consistent with the Goswitz and Brooker model (26), including the order of each helix arrangement. Our current study strongly supports the idea that MFS transporters have a common three-dimensional structure.