The NH2-terminal half of the Tn10-specified tetracycline efflux protein TetA contains a dimerization domain.

The 43.1-kDa tetracycline-cation/proton antiporter TetA from Tn10 comprises two equal-sized domains, alpha and beta (amino-terminal and carboxyl-terminal halves, respectively). An inactivating mutation in the alpha domain can complement a mutation on a second polypeptide in the beta domain to restore partial tetracycline resistance in bacterial cells, suggesting that intermolecular interactions permit this transport protein to act as a multimer. In the present studies, multimer formation was examined in mixtures of dodecylmaltoside extracts of membranes from Escherichia coli cells containing different TetA derivatives. TetA, TetA alpha, and TetA beta were each fused genetically to a six-histidine carboxyl-terminal tail. The ability of these fusions, immobilized on a nickel affinity column, to bind wild type TetA or other Tet fusions was determined. An interaction between alpha domains on different polypeptides which resulted in multimerization was seen. The binding was specific for Tet protein and did not occur with other membrane proteins or another polyhistidine fusion protein. No alpha-beta interactions were detected by this method, although they are postulated to occur in the intact cell based on the alpha-beta genetic complementations. A dimeric model for TetA having intermolecular alpha-alpha and alpha-beta interactions is presented.

The 43.1-kDa tetracycline-cation/proton antiporter TetA from Tn10 comprises two equal-sized domains, ␣ and ␤ (amino-terminal and carboxyl-terminal halves, respectively). An inactivating mutation in the ␣ domain can complement a mutation on a second polypeptide in the ␤ domain to restore partial tetracycline resistance in bacterial cells, suggesting that intermolecular interactions permit this transport protein to act as a multimer. In the present studies, multimer formation was examined in mixtures of dodecylmaltoside extracts of membranes from Escherichia coli cells containing different TetA derivatives. TetA, TetA ␣ , and TetA ␤ were each fused genetically to a six-histidine carboxyl-terminal tail. The ability of these fusions, immobilized on a nickel affinity column, to bind wild type TetA or other Tet fusions was determined. An interaction between ␣ domains on different polypeptides which resulted in multimerization was seen. The binding was specific for Tet protein and did not occur with other membrane proteins or another polyhistidine fusion protein. No ␣-␤ interactions were detected by this method, although they are postulated to occur in the intact cell based on the ␣-␤ genetic complementations. A dimeric model for TetA having intermolecular ␣-␣ and ␣-␤ interactions is presented.
TetA(B), a cytoplasmic membrane protein encoded by Tn10, is a member of a family of related tetracycline efflux proteins in Gram-negative bacterial cells (1,2). It mediates resistance to tetracyclines by pumping a cation-tetracycline complex across the membrane outwardly in an electroneutral exchange for an inwardly moving proton (3)(4)(5)(6). Experiments with a collection of point mutations had shown that inactivating mutations in the first half of the protein complemented those in the second half in cells containing both polypeptides (7,8). Complementation also occurred with protein fragments (9). However, each half of TetA did not have a unique function completely independent of that of the other half, since full or even half-resistance was rarely restored in complementations, even in cases where the presence of both polypeptides was confirmed. These results suggested that synergistic physical interaction between the two halves of the protein was required for resistance, and that such interaction could occur intermolecularly in a dimeric or higher multimeric state.
Further evidence for the required interaction between the two halves and for dimerization came from Tet protein chime-ras. The sequences of the related tetA genes from the family of tetracycline resistance determinants predicts that each TetA protein has two sets of six putative membrane-spanning ␣-helices separated by a putative large cytoplasmic loop (2, 9 -12). TetA proteins from classes A and C are more closely related (78%) than either are to the class B (Tn10) protein (45%) (1). An "A/C" chimera, containing the first (␣) half from class A and the second (␤) half from class C, was active in expressing tetracycline resistance, whereas a B/C or C/B chimera was not (13). Evidently the ␣ and ␤ halves functioned together only if they were related closely enough. The B/C and C/B chimeras together in the same cell, however, showed about 20% complementation of tetracycline resistance, indicating multimer formation (13). ␣-␤ interaction was also suggested by the ability of the cloned ␣ half to stabilize the cloned ␤ half when both were present on separate polypeptides in the same cell (14). Complementation occurred in this case also.
The present work was designed to determine whether TetA extracted from the cell existed as a multimer. We genetically fused six histidines to the carboxyl terminus of TetA, TetA ␣ , and TetA ␤ of class B. The ability of such a "6H" fusion to bind different Tet protein molecules was measured using Ni 2ϩ affinity chromatography. Table I for summary of plasmids and Fig. 1 for diagrams of protein constructs.
pLY17 (encoding Tet␣-6H, 24 kDa). A 0.6-kilobase EcoRI-XhoI fragment representing the ␤ half of TetA was deleted from pET21b-Tet6. The 5Ј ends were filled in with Klenow DNA polymerase prior to ligation. TetA ␣ was thereby in frame with the polyhistidine tail encoded 3Ј to the XhoI site. Loss of the 0.6-kilobase fragment was confirmed by loss of the ScaI site within it, and by the 6.0 kilobase size of the resulting plasmid. pLY17 was used in combination with pACT7.
pLY22 (encoding Tet ␤-6H, 24 kDa). The same tetA PCR product used to make pET21b-Tet6 was restricted with EcoRI (in the central loop of TetA) and XhoI (at the end of TetA) and cloned into identically restricted pET21b. This put the TetA ␤ domain in-frame with both the upstream "T7 tag" and the downstream polyhistidine tail encoded by pET21b. pLY22 was used in combination with pACT7.
pMalc-Tet1 (encoding MalE-Tet, 86 kDa). A tetA PCR product having BamHI sites on each end was restricted with BamHI and cloned into BamHI-restricted pMAL-C2 (New England BioLabs). This created an in-frame fusion between maltose-binding protein MalE (missing its * This work was supported by Grant AI30646 from the National Institutes of Health. 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. ‡ To whom correspondence should be addressed. Tel.: 617-636-4288; Fax: 617-636-0458. signal sequence) and the (cytoplasmic) amino terminus of the intact TetA, with an intervening 28-amino acid linker containing a factor Xa cleavage site. The fusion protein was regulated by P tac together with the lacI Q gene on the plasmid. Transformants were selected on 20 g/ml tetracycline without IPTG. 2 The strain synthesized several species of fusion protein, the largest and most abundant migrating at 70 kDa. The largest species was probably the intact fusion protein since it reacted with antiMalE, it bound to an amylose column by the MalE domain, and it reacted with antiCt to the carboxyl terminus of TetA. 3 The fusion protein was cleavable between MalE and TetA by factor Xa, as expected. 3 pQEGH12 (encoding 6H-IICB glc ). This plasmid (17) was used in strain ZSC112L (17), which has a glucose transporter ptsG mutation. The fusion protein is regulated by the ptsG promoter and is expressed constitutively in ZSC112L. pRAR1020 (encoding wild type TetA of class B) and pRAR1027 (encoding C/B chimera of TetA). Both have the tet promoter regulated by TetR (13). They were used in strain BC32 (13).
pRKH21 (encoding Tet279-LacZ, approximately 144 kDa). This plasmid (15) in strain RV200 (15) is regulated by P L . It is accompanied by pcI857, a compatible Kan R plasmid encoding the temperature-sensitive cI857 repressor (15). The fusion protein was induced by a shift in temperature from 30 to 42°C. pRKH21 had resulted from a spontaneous fusion between TetA and LacZ (15). Junction sequencing has now been performed (DNA Sequencing Center, Division of Endocrinology, New England Medical Center); the junction is at base pair 836 of TetA, fusing leucine 279 (at the amino terminus of the putative ninth transmembrane helix) of TetA to proline 8 of LacZ.
R222 (encoding wild type TetA of class B). This large, naturally occurring, very low copy number plasmid bears Tn10, which carries the complete class B tet determinant including the tet repressor (1,18). It is compatible with both ori pMB1 and ori p15A plasmids. Expression of TetA was induced by tetracycline.
Strains, Medium, Chemicals, and ␤-Galactosidase Assays-Unless otherwise specified, the host strain of Escherichia coli was DH5␣ (Life Technologies, Inc.; relevant loci recA, hsdR) and cells were grown in LB (per liter: 10 g of tryptone, 5 g of yeast extract, 10 g of NaCl) at 37°C. Antibiotics were purchased from Sigma, except that 5a,6-anhydrotetracycline was prepared by Mark Nelson of this laboratory. ␤-Galactosidase (LacZ) assays were performed as described (19). The LacZ assay was used to calculate picomoles of Tet279-LacZ (as monomers) assuming 1 pmol of LacZ hydrolyzes 34.9 pmol of ONPG/min at 28°C (see Ref. 19). The specific LacZ activity of Tet279-LacZ was within a factor of two of that expected for LacZ itself, 3 showing that the Tet moiety was not detrimental to the tetramerization of LacZ required for activity. A plasmid bearing the gene for a 6H-LacZ fusion protein cloned into pET14b was provided by Novagen in host BL21(DE3) (see below); this strain was used to prepare a membrane-free cell lysate containing 6H-LacZ.
Complementation Assays-The host strain was BL21(DE3) (Novagen), which bears a chromosomal T7 RNA polymerase gene regulated by the lac repressor. This strain was transformed with pLY22 (Ap R , bearing the Tet␤-6H gene and having a pMB1 origin of replication). A second mutant plasmid known to encode an active class B TetA ␣ domain and which had the compatible p15A origin of replication and encoded Cm R was also introduced; the second plasmid was either pLR1097 (bearing wild type tetA with a deletion in the ␤-domain (9)) or pRAR1032 (bearing the B/C chimeric gene (13)). In these second plasmids the mutant tetA gene was regulated by the tet repressor TetR; non-inhibitory autoclaved chlorotetracycline (10 g/ml) was used for induction (7). Resistance to tetracycline was measured by gradient plates (7) containing the autoclaved chlorotetracycline and 20 M IPTG.
Two different dodecylmaltoside extracts (usually 10 -50 l of each) were combined if desired to allow "mixed multimers" to form. After 30 min of occasional mixing at 4°C, 1/7 volume of 8-fold concentrated column buffer was added (column buffer final concentration was 50 mM sodium phosphate, pH 8.0, 300 mM NaCl, 5 mM imidazole, 0.03% dodecylmaltoside). Small (0.1) ml columns of Ni-NTA (Qiagen) (Ni 2ϩ bound to nitrilotriacetate immobilized on Sepharose CL-6B), were prepared in Pasteur pipettes and washed in column buffer. The samples were loaded onto the columns (50 l every 5-7 min) and washed with column buffer (0.2 ml, 2 min, ϫ6). When desired, an elution in column buffer at indicates a protease factor Xa cleavage site. 40 mM imidazole was then performed at the same rate. Finally, an elution in column buffer at 1 M imidazole (pH adjusted to 8) was done (0.08 ml, 5 min, ϫ3). In some cases eluates were used for dot-blots or assayed for LacZ. Otherwise they were precipitated with trichloroacetic acid (10% trichloroacetic acid, 15-30 min at 4°C, centrifuged 15,000 ϫ g, 10 min), dissolved in sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer (21), and processed by SDS-PAGE (9 or 10%) 0.75-mm thick minigels. Some gels were then electroblotted onto Immobilon P (Millipore) and probed with antiCt or antiTet antiserum, followed by 125 I-Protein A, as described. 1 AntiCt was specific for the carboxyl-terminal 14 amino acids of TetA (22); its reaction with Tet-6H was less than 2% of that with wild type TetA (determined by Molecular Dynamics Computing Densitometer evaluation of x-ray film exposed to immunoblots of SDS-PAGE gels), presumably due to the altered carboxyl terminus of Tet-6H. AntiTet reacted with an epitope between residues 127 and 201 of TetA and reacted equally well with Tet-6H. 1 Some gels were simply stained with Coomassie Brilliant Blue R-250 and dried in a Tut's Tomb frame between two sheets of Ultraclear Cellophane (both from Idea Scientific Co.). Protein bands on stained SDS-PAGE gels were quantitated using the Computing Densitometer; glyceraldehyde-6-phosphate dehydrogenase or ovalbumin served as standards. The total amount of a 6H fusion in extracts was defined as the sum of the amounts in the 40 mM and 1 M imidazole eluates from Ni-NTA.

Formation of Mixed Multimers between Wild Type TetA and
Tet-6H-DH5␣ cells containing pACT7 plus (i) the vector containing the tet-6H gene (pET21b-Tet6), (ii) the vector (pET21b) alone plus the wild type tetA gene on Tn10 (on naturally occurring plasmid R222), or (iii) pET21b-Tet6 plus R222, were induced with IPTG. The wild type gene in the strains with Tn10 was also induced with tetracycline. Dodecylmaltoside detergent extracts of membranes from all cells were prepared. In one case, extracts from i and ii were mixed in equal volumes for 5 min. The extracts were passed through a Ni-NTA column to bind Tet-6H and co-associated proteins. Bound proteins were eluted with 1 M imidazole, and both the original extracts and the bound proteins were examined for wild type TetA and Tet-6H by immunodot blot using antisera and 125 I-Protein A.
The original extracts loaded onto the Ni-NTA columns were analyzed first. Probing with antiCt antiserum, which reacts with TetA but not with Tet-6H, revealed that the amount of TetA in the two strains bearing Tn10 was similar (Fig. 2,  column A, rows 1 and 3). As expected, the extract from the strain with only Tet-6H (without Tn10) showed a low, probably host cell background, reaction with antiCt (Fig. 2, column A,  row 2). Use of antiTet antiserum, which reacts similarly with both TetA and Tet-6H, showed that cells synthesized less wild type TetA than overproduced Tet-6H (Fig. 2, column B, row 1  versus row 2), as shown before. 1 The samples which bound to the Ni-NTA columns were then analyzed. Use of antiCt demonstrated that little wild type TetA bound to Ni-NTA in the absence of Tet-6H (Fig. 2, column A,  row 4). On the other hand, in the presence of Tet-6H bound to the resin, TetA binding was clearly detectable. This was true whether the two versions of Tet had been synthesized in the same cell (Fig. 2, column A, row 6), or came from separate cells but were mixed together prior to loading onto Ni-NTA (Fig. 2, column A, row 7). If Tet-6H was loaded first, followed by TetA, TetA was still bound (Fig. 2, column A, row 5). Use of antiTet confirmed that similar amounts of Tet-6H were bound to Ni-NTA in all cases (Fig. 2, column B, rows 5-7). From Coomassiestained gels this was estimated to be about 5 g (110 pmol). To evaluate the contribution of the -6H region of Tet-6H to this binding, we mixed a membrane-free cell lysate containing 6H-LacZ with a membrane extract containing TetA. There was excellent binding of 6H-LacZ to Ni-NTA (about 100 g, or 950 pmol of monomers) but no binding of TetA (data not shown). Therefore, the binding of TetA to Tet-6H was not via the polyhistidine region.
Formation of Mixed Multimers of Tet279-LacZ with Tet-6H, but Not with 6H-IICB glc -We also tested Tet279-LacZ, encoded by pRKH21. This fusion contained the ␣ portion of TetA plus the first two putative transmembrane helices and associated extramembrane loops of the ␤ domain. In this case the Tet moiety could be quantitated by LacZ activity. An extract containing 480 pmol of Tet279-LacZ was used alone or mixed with one containing 240 pmol of Tet-6H and passed over a Ni-NTA column. In the absence of Tet-6H, 8.7 pmol of Tet279-LacZ was bound to the column, while 37 pmol was bound in the presence of Tet-6H. These results indicated a possible association between Tet279-LacZ and Tet-6H.
In a second experiment we included an extract containing 6H-IICB glc as a control to see if Tet would stick nonspecifically to another membrane protein immobilized on the Ni-NTA column. 6H-IICB glc is an E. coli membrane protein of 8 putative transmembrane segments which transports glucose as part of the phosphotransferase system and which has a polyhistidine head at the amino terminus (17). As a final control to measure background binding of Tet279-LacZ, we also included an extract from cells containing no polyhistidine fusion protein. We found that the background binding of Tet-279-LacZ (Table II) was 1.5% of that applied, similar to that in the first experiment (1.8%). Even though three times as much 6H-IICB glc as Tet-6H was applied and bound to the Ni-NTA column, only one-tenth as much net Tet279-LacZ was bound to it as to Tet-6H (Table  II). These results showed that the binding of Tet279-LacZ was specific for Tet-6H and did not occur with an unrelated integral membrane protein. The results also showed that the Tet-Tet interaction might be between the ␣ domains of the two different polypeptides, or between an ␣ and a ␤ domain, but possibly not between the ␤ domains, since Tet279-LacZ had the entire ␣ domain but only the first two helices and associated  (4), wild type TetA applied after Tet-6H (5), wild type TetA and Tet-6H from the same cell (6), or wild type TetA plus Tet-6H premixed before application (7). Blots were probed with antiCt (A) or antiTet antiserum (B). loops of the ␤ region.

␤-␤ Interactions Were Not Required for Multimer Formation-In
Tet␣-6H the ␤ domain is completely absent. If indeed ␤-␤ interaction was not required for Tet-Tet binding, Tet␣-6H should be able to bind a Tet protein containing both ␣ and ␤ domains. As its prospective partner, we used the MalE-Tet fusion in which the entire TetA protein was fused genetically via a linking region to the carboxyl terminus of MalE. The MalE-Tet fusion could be detected by its large size (migrating at 70 kDa) on Coomassie-stained SDS-PAGE. We again included 6H-IICB glc as a negative control. Dodecylmaltoside extracts containing MalE-Tet were mixed with extracts containing no fusion, IICB glc , or Tet␣-6H and passed over a Ni-NTA column. Sequential elutions were done at 40 mM and 1 M imidazole.
␣-␤ Interactions Did Not Contribute to Multimer Formation-The genetic complementation which occurred between ␣ and ␤ domains on different polypeptides (7-9, 13, 14) had suggested that these two domains could interact physically. We looked for such an ␣-␤ interaction with Tet␣-6H. We mixed extracts containing either B/B (that is, wild type TetA from the class B tetracycline resistance determinant) or C/B (a chimera containing the ␣ domain from class C, the ␤ domain from class B) (see Fig. 1) with an extract containing Tet␣-6H. A MalE-Tet/Tet␣-6H mixture was included as a positive control. Since B/B and C/B were not made by cells in sufficient quantities to be detected by Coomassie-stained SDS-PAGE, all proteins bound to Tet␣-6H were detected by immunoblot of SDS-PAGE gels. The results are shown in Fig. 4. Binding of Tet␣-6H to Ni-NTA was verified using antiTet (Fig. 4, lane 1Љ), as was binding of MalE-Tet to Tet␣-6H (data not shown). Using an-tiCt, binding of B/B to Tet␣-6H was seen (Fig. 4, lane 2Ј), as expected from the earlier experiments. However, little binding of C/B was seen (Fig. 4, lane 3Ј), even though much more C/B than B/B had been applied to the Ni-NTA columns, (Fig. 4, lane  3 versus lane 2). These results suggested that the ␣ domains of classes B and C interacted poorly. They also unexpectedly implied that there was little interaction between the ␤ domain of class B (on C/B) with the ␣ domain of class B (on Tet␣-6H).
Use of Tet␤-6H to Confirm Lack of ␣-␤ and ␤-␤ Interactions: Observation of ␣-␣ Interactions-The fact that C/B did not bind to Tet␣-6H suggested that the ␣ and ␤ domains of class B did not interact, despite genetic data to the contrary. It was possible that the C/B protein was for some reason in a nonbinding conformation after extraction. Therefore, we constructed a polyhistidine fusion having only the ␤ domain of class B for binding studies. This fusion was designated Tet␤-6H.
The Tet␤-6H protein was identified on Coomassie-stained SDS-PAGE gels of Ni-NTA-bound protein as a band migrating slightly more slowly than Tet␣-6H and not present in fusionless host cells (data not shown). Quantification of these bands indicated that cells containing pLY22, encoding Tet␤-6H, produced only about 2% as much fusion protein as did cells bearing pLY17 (encoding Tet␣-6H).
The functionality of Tet␤-6H encoded by pLY22 was assayed in vivo by the ability to complement TetA having a mutated ␤ domain encoded on a compatible plasmid. Two different compatible mutant plasmids were tested in trans with pLY22 (see "Experimental Procedures"). No plasmid offered tetracycline resistance alone (minimal inhibitory concentration of tetracycline Ͻ0.2 g/ml). pLY22 complemented both mutant plasmids to give tetracycline resistance (minimal inhibitory concentration Ͼ10 g/ml). Therefore, the Tet␤-6H domain was functional, at least in the intact cell expressing a complementing Tet protein.
Biochemical studies were then performed. Extracts of cells containing Tet␤-6H or Tet␣-6H were loaded onto Ni-NTA columns at volumes which contained approximately equal amounts of each fusion protein. A volume of a host extract identical to the volume used for Tet␤-6H was also loaded onto a column as a control. Then an extract containing Tet279-LacZ (or MalE-Tet in one case) was passed over the columns. Binding of Tet279-LacZ to the host extract column was considered as background. The molar ratio of Tet279-LacZ to 6H fusion applied to the column was about 2. The net molar ratio eluting at 1 M imidazole was about 0.038 for Tet␣-6H but only 0.002 for Tet␤-6H. Tet␤-6H also bound no MalE-Tet observable on SDS-PAGE even though the MalE-Tet Ϭ Tet␤-6H molar ratio applied to Ni-NTA was about 7 (data not shown). MalE-Tet was bound to Tet␣-6H in the presence of Tet␤-6H extract, as expected, although the required large volume of Tet␤-6H extract increased background bands on SDS-PAGE, making quantification difficult (data not shown). These results showed that Tet␤-6H was neither able to bind Tet containing both ␣ and ␤ domains (in MalE-Tet), nor able to bind the ␣ domain in Tet279-LacZ. Our earlier failure to see ␣-␤ interaction in extracts was therefore confirmed, as was the absence of ␤-␤ interaction. The results also suggested that ␣-␣ interactions between Tet279-LacZ and Tet␣-6H were responsible for Tet multimerization in this assay.
Comparison of ␣ Binding to ␣ with ␣ Binding to ␣ϩ␤ in Multimer Formation-It appeared that binding between different Tet proteins in extracts could occur solely by ␣-␣ interactions. To see whether ␣-␣ interactions might be fortified by additional ␣-␤ ones, we compared binding of Tet279-LacZ to Tet␣-6H with that to Tet-6H. In this experiment the amount of Tet␣-6H recovered from cells and applied to the Ni-NTA column was five times the amount of Tet-6H. 180 pmol of Tet279-LacZ was mixed with Tet␣-6H or Tet-6H and loaded onto a Ni-NTA column. Summation of values for elutions at 40 mM and at 1 M imidazole gave 37 pmol of Tet279-LacZ co-eluting with the 870 pmol of Tet␣-6H bound to the column, while 8.8 pmol of Tet279-LacZ co-eluted with the 170 pmol of Tet-6H bound. On a molar basis, Tet-6H bound about the same amount of Tet279-LacZ as did Tet␣-6H. Therefore, the additional ␣-␤ interactions in Tet-6H were not helpful to the association. We concluded that, except for a possible role for the bit of the ␤ domain in Tet279-LacZ, the in vitro binding of one Tet polypeptide to another seen using polyhistidine fusions and Ni-NTA columns must occur by interactions between two or more ␣ domains. DISCUSSION We report here initial biochemical studies on the quaternary structure of the tetracycline-cation/proton antiporter TetA. From genetic data described earlier we had expected that TetA protein was capable of functioning in vivo as a dimer or other multimer. We had also imagined that the interaction would be between the ␣ and ␤ domains. Earlier we had found that a small proportion of either the B/B protein or the C/B chimeric protein could be cross-linked into a immunoreactive band having the molecular weight of a dimer, but that little coimmunoprecipitation of one Tet polypeptide by antibody specific for another occurred, with or without cross-linking. 3 In the present work we explored another biochemical method to test the multimer hypothesis. Immobilized Ni 2ϩ can be used to bind proteins having a polyhistidine region (23). By the use of TetA-polyhistidine fusion proteins, we were able to clearly show specific association between two distinguishable Tet protein molecules from cell membrane extracts. These heteromultimers between two Tet species formed simply upon mixing a dodecylmaltoside extract containing one Tet species with an extract containing the other. Apparently, in the mixtures the original homomultimers have readily dissociated (within minutes) into subunits, followed by rapid association with a heterologous subunit into a multimer which was stable enough to detect. Presumably, the rates of both association and dissociation are high, while the former exceeds the latter to account for multimer stability on Ni-NTA. Binding did not occur between TetA and another polyhistidine fusion of an integral membrane transport protein, 6H-IICB glc , nor did other cell membrane proteins associate with Tet-6H to any notable extent, as was evident by its purity following Ni-NTA chromatography. 1 Therefore, we believe the Tet-Tet interactions to be specific.
Unexpectedly the crucial interaction in formation of Tet multimers in vitro appeared to be between two (or more) ␣ domains, rather than between an ␣ and a ␤ domain. However, in intact cells, besides the genetic data there are also physical indications of ␣-␤ interaction. The amount of a polypeptide comprising the ␤ half of TetA in whole cells was increased 1.5-fold or more by the presence of the ␣ half polypeptide, suggesting a physical interaction of the two (14). We have observed that the amount of full-length B/C chimera in cells (normally very low) increased notably if the C/B chimera was present in the same cell 3 ; a simple explanation for those results could be that the C/B protein formed a multimer with the B/C protein via same-class ␣-␤ interactions and stabilized it, although other explanations are possible. The fact that in the present work we did not see ␣-␤ interactions after the Tet protein had been extracted may mean that the ␤ domain for the C/B and Tet␤-6H constructs did not have native binding properties in our extracts or under our assay conditions. However, recent circular dichroism studies on purified full-length Tet-6H, at least, show that both ␣ and ␤ domains in that polypeptide do have approximately the expected ␣-helical content. 3 A TetA dimer may be held together both by ␣-␣ interactions (seen in the present study for proteins extracted from membranes by dodecylmaltoside) and by ␣-␤ interactions (not apparent using extracts, but inferred from genetic and biochemical studies in whole cells). A model in which both ␣-␣ and ␣-␤ interactions occur within a TetA dimer is shown in Fig. 5. During complementation of B/C with C/B in vivo, the ␣-␣ interactions would presumably not occur, but the ␣-␤ ones would. Two active sites/wild type dimer, or one/ complementing dimer, would be expected. Our model might explain why Tet␣-6H was found in cells at high concentrations similar to those of the full-length fusion Tet-6H, while the amount of Tet␤-6H was 50-fold lower, since the model allows ␣ to bind to ␣ (or to ␤), and such associations may prevent degradation. Absence of selfassociation for ␤, as modeled, would lead to degradation of ␤ when alone in a cell.
The proposed structure of the dimer differs from that proposed for a monomer both because of the additional ␣-␣ interactions and the altered topology of the central loop (Fig. 5). A monomer of TetA has both the domains (␣ and ␤) required for activity, and we cannot discount the possibility that a complex consisting of only one ␣ and one ␤ domain is capable of functioning. On the other hand, even when these two domains are tethered together in a normal monomer, considerable interaction with other such monomers must be allowed in vivo, since intermolecular complementation can occur. Self-association of monomers into dimers might be favored in the two-dimensional FIG. 5. Model of possible Tet dimer and monomer. The plane of the page represents that of the membrane surface. Hypothetical active site is denoted by an ϫ. A ribbon representing the large cytoplasmic loop connects the ␣ and ␤ domains within a single polypeptide strain. This loop is located in the cytoplasm above the membrane surface. membrane bilayer even more than the considerable degree seen here in detergent extracts.
Multimerization provides possibilities for scaffolding, interfaces, and allostery. Some other membrane transport proteins of the same superfamily (24) as TetA are known to occur as multimers, including the facilitated glucose transporter GLUT1 (17,25), the erythrocyte anion exchanger Band 3 (26), and the Na ϩ /glucose cotransporter (27). The relationship between these multimerizations and function is uncertain (25,28), and at least one example exists (the lactose permease, LacY) in which the transporter almost certainly functions as a monomer (29). Our results strengthen the concept that the mechanism of action of TetA involves a multimeric state.