Cysteine-scanning Mutagenesis and Disulfide Mapping Studies of the Conserved Domain of the Twin-arginine Translocase TatB Component*

The cytoplasmic membrane protein TatB is an essential component of the Escherichia coli twin-arginine (Tat) protein translocation pathway. Together with the TatC component it forms a complex that functions as a membrane receptor for substrate proteins. Structural predictions suggest that TatB is anchored to the membrane via an N-terminal transmembrane α-helix that precedes an amphipathic α-helical section of the protein. From truncation analysis it is known that both these regions of the protein are essential for function. Here we construct 31 unique cysteine substitutions in the first 42 residues of TatB. Each of the substitutions results in a TatB protein that is competent to support Tat-dependent protein translocation. Oxidant-induced disulfide cross-linking shows that both the N-terminal and amphipathic helices form contacts with at least one other TatB protomer. For the transmembrane helix these contacts are localized to one face of the helix. Molecular modeling and molecular dynamics simulations provide insight into the possible structural basis of the transmembrane helix interactions. Using variants with double cysteine substitutions in the transmembrane helix, we were able to detect cross-links between up to five TatB molecules. Protein purification showed that species containing at least four cross-linked TatB molecules are found in correctly assembled TatBC complexes. Our results suggest that the transmembrane helices of TatB protomers are in the center rather than the periphery of the TatBC complex.

Protein transport across the cytoplasmic membrane of most bacteria proceeds by one of two pathways. The major route of protein translocation is the Sec pathway, to which substrates are targeted by N-terminal signal peptides. Protein transport by the Sec machinery proceeds via an N-to C-terminal threading mechanism (1,2). The Tat pathway operates in parallel to the Sec system but, by contrast, exports folded proteins (3). Substrates that target to the Tat machinery are also synthesized with N-terminal signal peptides, which in this case harbor a consensus SRRXFLK motif (4). The Tat system, which is energized solely by the transmembrane proton electrochemical gradient, is structurally and mechanistically related to the ⌬pH-dependent thylakoid import pathway of chloroplasts (5).
In Escherichia coli, four proteins, TatA, TatB, TatC, and TatE, have been identified that are required for Tat-dependent protein translocation (6 -9). TatC is an integral membrane protein comprising six transmembrane ␣-helices (10,11). TatA, TatB, and TatE are sequence-related proteins that are predicted to share a common structure comprising an N-terminal transmembrane ␣-helix, followed by an adjacent amphipathic helix and a C-terminal region of variable length (6,12) (Fig. 1A). TatA and TatE share an essential function in the export process (6). It is likely that tatE is a cryptic gene duplication of tatA, because expression studies indicate that tatA is expressed at much higher levels than tatE (13). TatB shares some sequence homology with TatA/E (ϳ25% identity at the amino acid level) but is considerably longer. TatB is an essential component of the E. coli Tat machinery (8). TatA and TatB are integral inner membrane proteins. However, upon removal of the N-terminal transmembrane domain, TatA becomes a soluble protein, whereas TatB retains peripheral interactions with the membrane (14). Truncation and point mutational analysis have indicated that only the first 40 or so residues of TatA and the first 101 residues of TatB, comprising the N-terminal membranebound and the adjacent amphipathic helices of each protein, are necessary to support Tat function (15,16).
Both TatA and TatB are found in the cell membrane as high molecular mass complexes. However, the organization and function of the complexes formed by each of these homologous proteins are very different.
TatA forms homo-oligomeric complexes that contain varying numbers of TatA protomers (14,17,18). Electron microscopy of these TatA complexes shows ring-shaped structures of variable diameter that are likely to constitute the protein translocating channels of the Tat system (18).
TatB, by contrast, forms an equimolar complex with TatC (19). This TatBC complex runs as a single ϳ400-kDa species on blue native PAGE when solubilized with digitonin from membranes containing either native or overexpressed levels of Tat proteins (17,20,21). Given the size of the complex it is clear that it must contain multiple copies of both TatB and TatC. An analogous complex is found in the thylakoid Tat system (21). The TatBC complex binds substrate proteins through their twin-arginine signal peptides and thus acts as the substrate receptor complex of the Tat pathway (21)(22)(23)(24). 5 To date the only information on the organization of TatB within the TatBC complex has come from chemical cross-linking studies. Experiments using two types of amine-specific reagents detected homodimeric TatB species (14). This suggests that each TatB protein is in intimate contact with another TatB molecule.
In this study we have used cysteine-scanning mutagenesis to search for residues that are important for TatB function and have then used these cysteine variants to map TatB-TatB interactions. Cross-linking of TatB variants with double cysteine substitutions indicates that the protein forms a higher order multimer within the TatBC complex.
Plasmid pUNITAT1 carries tatA, tatB, and tatC in pQE60 (Qiagen). This construct contains the native stop codon of tatC, and therefore the native, nontagged TatC protein is produced. The tatA, tatB, and tatC genes were amplified by PCR using primers TATA5 (6), TATCBAM (5Ј-GCGCGGATCCTTAT-TCTTCAGTTTTTTCGCTTTCT-3Ј), and MC4100 chromosomal DNA as template. The product was digested with EcoRI and BamHI and cloned into the polylinker region of pQE60 to give plasmid pQEABC. The QuikChange TM system was then used to introduce a 2-bp substitution in the intergenic region between tatA and tatB, such that a PmlI restriction site was introduced. A second round of QuikChange TM was then used to introduce an AflII restriction site between tatB and tatC to coli TatB analyzed in this study. The regions of the TatB polypeptide that were targeted for cysteine substitution in this study are boxed. The predicted secondary structure for the first 90 residues of TatB as determined using the program PSIPRED (60) is shown. Predicted ␣-helical regions are shown as gray cylinders above the amino acid sequence. All other regions have a low predicted propensity for forming defined secondary structure. B, the concept behind disulfide mapping of the TatB interface. For a single cysteine variant in the transmembrane helix of TatB, the cysteine residue in a given TatB molecule will be at the same vertical height in the membrane as the cysteine residue in any other TatB molecule. A major determinant of whether the cysteines in adjacent protomers are close enough to form a disulfide linkage will then be the relative rotations of the helices around their long axes. Only if the cysteines in both protomers are positioned close to the interacting face can a disulfide bond form (i ). In other cases the cysteine residues are too far apart for disulfide bond formation (ii ).
give plasmid pUNITAT1. This was achieved by substituting the penultimate codon of tatB from CCG (Pro) to CCT (Pro) and by changing the first base of the 2-bp tatB/C intergenic region from A to G. This allows easy replacement of wild type tatA, tatB, or tatC genes. A derivative of tatC in which all four cysteine codons had been substituted for alanines was subsequently introduced into pUNITAT1 as follows. A PCR product covering the tatC gene was generated using primers TATCAFL (5Ј-GCGCGAATTCCTTAAGCATGTCTGTAGAAGATAC-T-3Ј) and TATCBAM, with a cysteine-less derivative of tatC in pBluescript as template (28). The product was digested with AflII and BamHI and cloned into pUNITAT1 digested with the same enzymes to give plasmid pUNITATCC4. Plasmid pUNITATCC4H is identical to pUNITATCC4, except that the stop codon of tatC is absent, and therefore the protein is produced with a C-terminal hexahistidine tag. It was constructed following amplification of the cysteine-less derivative of tatC in pBluescript using primers TATCAFL and TATCH2 (32), digestion with AflII and BglII, and cloning into pUNITAT1 that had been pre-digested with the same enzymes.
A plasmid, pUNITATB, that contains the tatB gene in pBluescript was constructed following amplification of the tatB gene with primers UNIB1 (5Ј-GCGCGAATTCCACGTGTTTGAT-ATCGGTTTTAGCG-3Ј) and UNIB2 (5Ј-GCGCGGATCCC-TTAAGGTTTATCACTCGACGAAGG-3Ј), digestion with EcoRI and BamHI, and cloning into pBluescript that had been previously digested with the same enzymes. Site-specific cysteine mutations were initially introduced into the tatB gene in plasmid pUNITATB using the QuikChange TM method. Each mutation was subsequently subcloned into pUNITATCC4 for cysteine cross-linking and TMAO reductase assays. The following substitutions in the tatB gene were also subcloned into pUNITATCC4H to facilitate purification of TatC-containing complexes: L11C, V12C, and L11C plus V12C. The individual cysteine substitutions are shown in Table 1. For construction of plasmid pFATPL1112, the site-specific mutations for tatB L11C and V12C were introduced into plasmid pFAT586 using the QuikChange TM method.
All clones obtained from PCR-amplified DNA were sequenced to ensure that no undesired mutations had been introduced. The primer sequences for introduction of site specific mutations are available on request.
Cell Growth, Fractionation, and Protein Methods-During all genetic manipulations, and for cysteine cross-linking experiments, E. coli strains were grown aerobically in Luria-Bertani (LB) medium (33). Strains were normally grown at 37°C unless otherwise stated. Concentrations of antibiotics were as described previously (6,30). For induction of plasmid-borne tatABC His alleles, cells were initially grown to an A 600 nm of 0.4 -0.6. Isopropyl thio-␤-D-galactoside (IPTG) was then added to the culture at a final concentration of 2 mM, and growth was continued for a further 6 h before harvesting.
The growth phenotypes of mutants with TMAO as sole respiratory electron acceptor were determined on M9 minimal medium (33) agar plates supplemented with 0.5% glycerol and 0.4% TMAO and incubated in a gas jar under a hydrogen/carbon dioxide atmosphere. The SDS-resistance phenotype of mutants was tested on LB agar plates containing 2% SDS (28). For TMAO reductase assays cells were cultured in modified Cohen and Rickenberg medium (34), supplemented with 0.2% (w/v) glucose and 0.4% (w/v) TMAO. Subcellular fractions for TMAO reductase assays were prepared from small (30 ml) cultures using a cold osmotic shock protocol (35), and TMAO: benzyl viologen oxidoreductase activity was measured as described (36).
Strain DADE (27) containing plasmids pFAT75AH (32) and pREP4 (Qiagen) and cultured without IPTG was used for the protease accessibility experiments. Spheroplasts were prepared as follows. Cells were harvested by centrifugation and then resuspended in ice-cold 33 mM Tris-HCl, pH 8.0, 40% (w/v) sucrose. Na 2 EDTA and lysozyme were added to a final concentration of 5 mM and 0.1 mg/ml, respectively. Cells were incubated on ice for 30 min, and the spheroplasts were harvested at 8000 ϫ g for 15 min. The spheroplasts were resuspended in ice-cold 33 mM Tris-HCl, 40% (w/v) sucrose. Inverted inner membrane vesicles were prepared as described (37).
For the disulfide cross-linking experiments, strain DADE-P containing plasmids encoding TatA, TatC, and cysteine-substituted derivatives of TatB were grown overnight at 30°C in 500-ml cultures. For disulfide cross-linking experiments where TatB was expressed in the absence of other Tat proteins, strain DADE harboring plasmid pREP4 was transformed with either pFAT586 (14) or pFATPL1112 (Table 1). 500-ml cultures of each strain were grown until an A 600 of 0.6 was reached, after which expression of TatB was induced by the addition of 1 mM IPTG for a further 2 h. Cells were harvested at 6000 ϫ g for 10 min and washed with 100 ml of Buffer K (50 mM triethanolamine-HCl, pH 7.5, 250 mM sucrose, 1 mM Na 2 EDTA). After washing the cells were resuspended in 15 ml of Buffer K supplemented with DNase I, lysozyme (0.6 mg/ml), and protease inhibitors (Complete EDTA-free protease inhibitor mixture tablets; Roche Applied Science) at half a tablet per sample. Cells were disrupted by passage through a French press twice at 1000 p.s.i. The cell debris was centrifuged twice at 27,000 ϫ g for 15 min, and the total membrane fraction was collected by ultracentrifugation of the supernatant at 206,000 ϫ g for 90 min. Typically, the membrane pellet was resuspended in 1.5 ml of Buffer 1 (50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 10% glycerol),and aliquots were stored at Ϫ80°C until use.
Disulfide cross-linking experiments were carried out at room temperature for 1 h. Oxidized, reduced, and control samples were run simultaneously. A final concentration of 2 mM copper phenanthroline was used in the oxidized sample. After 1 h the reaction was quenched by the addition of 25 mM N-ethylmaleimide, 50 mM Na 2 EDTA, pH 7.5. 10 min after quenching, the reaction volume was made up to 100 l with the addition of SDS loading buffer (final concentration 50 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 0.1% bromphenol blue). Reduced samples contained a final concentration of 10 mM dithiothreitol (DTT) as a reducing agent. After 1 h the reaction volume was made up to 100 l with SDS loading buffer as above. The control sample, containing membrane sample and buffer only, was also made up to 100 l with SDS loading buffer as before after the 1 h of incubation.
For the purification of histidine-tagged Tat complexes, membrane fractions were treated with a freshly prepared 1 mM copper phenanthroline solution at room temperature for 1 h. Subsequently, any unreacted thiols were blocked by the addition of 25 mM N-ethylmaleimide, and the samples were incubated at room temperature for 15 min. Excess cross-FIGURE 2. Assessment of the periplasmic and cytoplasmic exposure of the E. coli TatB protein by protease accessibility measurements. Spheroplasts (A) (in which the periplasmic side of the inner membrane is exposed) and purified inner membrane vesicles (in which the cytoplasmic side of the membrane is exposed) (B) were prepared from ⌬tatABCD⌬tatE strain DADE (27) containing pREP4 and a tatABC-expressing plasmid (pFAT75AH (32)). Where indicated 0.5 mg ml Ϫ1 proteinase K (Prot. K) was added to the membrane preparations either in the presence or absence of 1% (v/v) Triton X-100. Following a 30-min incubation on ice, 1% trichloroacetic acid was added to all samples. The resulting precipitates were resuspended in SDS-PAGE sample buffer and then subjected to SDS-PAGE and immunoblotting using the indicated polyclonal antisera. TatA has been shown previously to be exposed only at the cytoplasmic side of the membrane (38). FtsZ is a cytoplasmic protein. TolC is a periplasmically located peripheral membrane protein. NuoI is a cytoplasmically located peripheral membrane component of NADH:quinone oxidoreductase (complex I). YidC is an inner membrane protein that produces a distinct 40-kDa fragment when digested by proteinase K from the cytoplasmic face of the membrane (44).
linking reagent and N-ethylmaleimide were removed by pelleting the cross-linked membranes at 250,000 ϫ g for 1 h at 4°C. Alternatively, if the samples were to be prepared under reducing conditions, the membranes were instead treated with 1 mM DTT, and the same concentration of DTT was added to all of the purification buffers. Detergent solubilization of the membranes, isolation of histidine-tagged complexes by affinity purification, and analytical size exclusion chromatography of the partially purified complexes followed methods published previously (38,39) with minor variations. For strains expressing TatA, TatC His , and singly cysteine-substituted TatB, membrane fractions were solubilized at a protein con- FIGURE 3. Periplasmic TMAO reductase activities of E. coli tat mutant strains producing cysteine-substituted TatB proteins. Periplasmic TMAO reductase activities were measured from the following: the tat wild type parental strain MC4100 (WT ); the ⌬tatABCD⌬tatE, pcnB derivative, DADE-P (⌬); DADE-P carrying pUNI-TATCC4 (Cys Ϫ ); DADE-P carrying pUNITATCC4 encoding cysteine-substituted TatB proteins with the amino acid position of the substitution shown below each column. 100% activity was taken to be that determined for DADE-P carrying pUNITATCC4 and corresponds to an activity of typically 1.0 mol of benzyl viologen oxidized per min/mg protein. Samples were assayed three times with results not varying by more than 15% from the mean. centration of 5 mg ml Ϫ1 for 1 h at 4°C in 20 mM MOPS, pH 7.2, 200 mM NaCl (Buffer A) containing 1% (v/v) C 12 E 9 (Sigma). Insoluble material was removed by ultracentrifugation at 250,000 ϫ g for 30 min, and the cleared supernatant was applied to a Ni 2ϩ -loaded 5-ml HisTrap HP column (GE Healthcare) equilibrated in Buffer A containing 0.1% (v/v) C 12 E 9 and 30 mM imidazole. The column was washed with 25 ml of Buffer A containing 0.1% (v/v) C 12 E 9 , 120 mM imidazole and then developed in 20 ml of the same buffer with a linear gradient of imidazole to a final concentration of 700 mM. Fractions containing Tat proteins were identified by SDS-PAGE and pooled. The HisTrap pool was concentrated in an Amicon Ultra 4 concentration device with a 100-kDa cut-off filter (Millipore). The concentrated sample was then applied to a Superose-6 HR (1 ϫ 30 cm inner diameter) size exclusion column (GE Healthcare) equilibrated in Buffer A containing 0.1% (v/v) C 12 E 9 . Eluted fractions containing Tat proteins were characterized by SDS-PAGE.
For the TatB L11C plus V12C variant expressed from plasmid pUNIBPL1112H, the purification protocol was the same as for the single cysteine variants with the exception that the membranes were solubilized with 2% digitonin (Merck) rather than C 12 E 9 , and the C 12 E 9 in the column buffers was replaced with 0.1% (w/v) digitonin.
SDS-PAGE and immunoblotting analyses were as described (40,41), and immunoreactive bands were visualized with the ECL detection system (Amersham Biosciences). Bands were quantified using a Fuji BAS-100 PhosphorImager and the soft-

RESULTS AND DISCUSSION
TatB Is Not Exposed at the Periplasmic Side of the Cell Membrane-At the outset of this study, the topological organization of TatB in the cytoplasmic membrane had not been definitively established. We therefore proceeded to map the exposure of TatB at either side of the membrane using a protease accessibility assay. In agreement with a gene fusion study (19) and experiments on the plant thylakoid TatB ortholog Hcf106 (47), TatB was found to be exposed at the cytoplasmic (cis) side of the membrane (Fig. 2). By contrast, TatB was not accessible to protease in spheroplasts (Fig. 2) demonstrating that no portion of TatB is exposed at the periplasmic (trans) side of the membrane. The gross topological organization of TatB is thus identical to that determined previously for the homologous TatA protein (38) (Fig. 2).
Construction, Expression, and Activity of Single Cysteine Variants of TatB-To date there have only been very limited investigations aimed at identifying specific amino acid residues that are important for TatB function (7, 48 -50). We therefore undertook a scanning mutagenesis study in an attempt to systematically identify such residues. Truncation experiments as well as sequence conservation indicate that the essential por-tions of TatB reside within approximately the first 50 amino acids (15). We have therefore confined our scanning mutagenesis analysis to this key region of the protein. Each targeted residue was substituted with cysteine to allow subsequent disulfide mapping studies (below).
The E. coli TatA and TatB proteins do not contain cysteine. However, four cysteine residues are found in TatC. To avoid potential complications in the subsequent disulfide mapping studies of TatB arising from the presence of cysteine residues in TatC, we constructed a modular cloning vector pUNITATCC4, which allows co-expression of tatAB with a cysteine codonsubstituted version of tatC. This tatC allele, in which all four TatC cysteines have been replaced with alanine residues, has been shown previously not to significantly affect Tat transport activity (28). Consistent with this earlier observation, pUNITATCC4 was able to restore an approximately wild type level of the Tat substrate protein TMAO reductase (TorA) to the periplasmic fraction of a strain lacking all genes encoding known Tat components (Fig. 3). The cysteine mutations The percentage next to each residue is the percentage of cross-linked dimer (expressed as a percentage of monomer plus dimer) formed after oxidation at room temperature for 1 h. A, only those residues that show periodicity of cross-linking when substituted for cysteine are included. The TatB strongly interacting face is indicated with a solid line. The dashed line indicates that the interacting face may extend as far as Leu-9. Although the level of dimerization at position 16 is very low, this may be partly explained by the substitution of glycine by much bulkier cysteine that could affect the inter-helical packing.
required for the scanning mutagenesis experiments were introduced into the tatB gene in pUNITATCC4.
The first 20 amino acids of TatB are predicted to form a transmembrane ␣-helix with the N terminus located at the periplasmic side of the membrane (Figs. 1A and 2). Nineteen unique cysteine mutations covering TatB residues 2-20 were constructed. None of the mutations blocked the export of the Tat substrate TMAO reductase (Fig. 3). Indeed, most of the mutations gave periplasmic TMAO reductase levels that were close to wild type. Strikingly, replacement of the highly conserved glutamate residue at position 8 with cysteine did not affect the measured Tat transport activity. This observation is consistent with previous reports that substitution of this residue in the E. coli protein with alanine was also well tolerated (48,49). The lowest periplasmic TMAO reductase activity was seen with the substitution of the isoleucine at residue 4. However, even in this case periplasmic TorA activity was greater than 60% of the unsubstituted control.
The 30 or so residues immediately following the transmembrane segment are strongly predicted to form an amphipathic ␣-helix. We targeted 12 adjacent residues toward the center of this segment, covering three predicted turns of the ␣-helix, for substitution by cysteine (Fig. 1A). None of these substitutions resulted in an inactive TatB protein because all these variants supported the export of TMAO reductase (Fig. 3). However, the two neighboring residues Leu-39 and Arg-40 showed 50% or greater reduction in periplasmic TMAO reductase activity when substituted with cysteine. Mutations located at the equivalent positions in TatA have been shown previously to have a drastic effect on transport (16,48).

Disulfide Mapping Studies of TatB Using Single Cysteine
Variants-Studies using amine-specific cross-linking reagents have shown that pairs of TatB molecules are in intimate contact in their native membrane environment (14). The availability of a panel of single cysteine variants has enabled us to apply sitespecific disulfide cross-linking to probe the interactions between TatB molecules in more detail. In this approach a disulfide cross-link is formed under oxidizing conditions between two introduced cysteine residues provided that the ␤-carbons of the two cysteines are in close proximity (i.e. about 4 Å or less). This method is therefore potentially able to identify sites involved in protein-protein interactions. It is a particularly powerful tool for mapping homo-oligomeric interactions in integral membrane proteins because a cysteine residue in the transmembrane helix of one protomer should be at the same vertical height in the membrane as the same cysteine residue in another protomer. Whether the two cysteine residues can approach closely enough to form a disulfide bond will therefore depend on the relative arrangement of the helices within the plane of the membrane (Fig. 1B). Because all the single cysteine variants that we have analyzed in this study retain Tat transport activity (above), it can be inferred that the substitutions do not interfere with the structure or structural organization of TatB. It is therefore reasonable to use of our panel of single cysteine variants in disulfide mapping experiments to probe TatB-TatB interactions.
Oxidation-induced cross-linking analysis was carried out on membranes isolated from a ⌬tatABCD⌬tatE strain co-expressing single TatB cysteine substitution variants together with TatA and cysteine-less TatC from plasmid pUNITATCC4. Observed cross-links were thus derived from TatB protein in its native membrane environment and in the presence of its TatC partner protein to which it is presumably bound (and see below).
The results obtained for cysteine substitutions in the putative transmembrane segment of TatB (residues 2-20, inclusive) are shown in Fig. 4. Upon oxidation many of the cysteine variants exhibited a TatB immunoreactive band migrating at twice the apparent molecular mass of the TatB monomer. This dimer was not present, or was significantly reduced, when membranes were treated with the reductant DTT indicating that it arose from formation of a disulfide bridge. A number of additional bands that migrate with a molecular weight between that of the monomer and the dimer were also observed under conditions that induced TatB dimer formation. Because these bands crossreact with the TatB antiserum and are only observed for those cysteine substitutions that also give rise to TatB dimer species, it is likely that these are different conformers of the TatB dimer. A similar phenomenon has been reported for other proteins (51).
Residues 2-8 exhibited similar patterns of cross-linking (Fig.  4). Significant cross-linking was observed at each successive position. This shows that all these residues are able to come into close proximity to the same residue in a neighboring protomer and suggests that this portion of the protein is relatively mobile. A second shared feature of the cross-linking at these positions is that the cross-links were already present in the membranes as prepared and were not significantly enhanced by the addition of the cross-linking catalyst Cu(II)phenanthroline (CuP). We considered the possibility that these residues are exposed at the periplasmic face of the membrane and are oxidized in vivo by the periplasmically located disulfide bond machinery. However, the pattern and intensity of cross-linking we observed with these TatB derivatives were not affected when repeated in strain GB100 that lacks a functional DsbA protein (data not shown).
In contrast to the behavior of the position 2-8 cysteine variants, the cysteine variants at positions 9 -20 exhibited no crosslinking in the absence of the catalyst CuP (Fig. 4). Upon CuP addition cysteine variants at positions 9 -17 showed positionspecific variations in cross-linking (Fig. 4). These observations indicate that this portion of the TatB protein has a defined structure. They also suggest that the transmembrane helices of TatB protomers are in close proximity to each other. If the positions of the strongest disulfide cross-links are plotted on a helical wheel, it can be seen that they fall on one face of a helix (Fig. 5A). This supports the proposal that the transmembrane region of TatB is helical. It also identifies the face of the helix that is involved in symmetrical protomer-protomer interactions with other TatB transmembrane helices. This side of the helix shows greater amino acid sequence conservation than the noninteracting side, as would be expected of the helix face involved in specific protein-protein interactions rather than nonspecific protein-phospholipid contacts. If the helical structure extends beyond residues 9 -17, then the interacting face contains almost invariant Glu-8, which is the only potentially charged residue in the TatB transmembrane helix. The periodicity of the cross-linking breaks down toward the end of the transmembrane region where residues 18 -20 all exhibit substantive cross-links and thus apparently increased conformational freedom.
The oxidation-induced cross-linking results for cysteine substitutions in the amphipathic segment of TatB (residues 31-42, inclusive) are shown in Fig. 6. These residues show positionspecific cross-linking suggesting that this portion of the TatB protein has a defined structure. It also suggests that the amphipathic helices of TatB protomers are in close proximity to each other. More detailed structural interpretation is complicated by the fact that the cross-linking observed at many positions is catalyst-independent. This type of cross-linking becomes more prominent as one moves down the helix suggesting that the secondary structure, mobility, or environment of residues 37-42 differs from that of the immediately preceding helical region. Catalyst-independent formation of the representative R40C dimer was unaffected in a dsbA strain (data not shown) arguing against the possibility that this region of the protein is cross-linked because of exposure to the periplasmic disulfide bond-forming machinery. This result is consistent with the cytoplasmic location of the extramembranous domain of TatB deduced above from protease protection experiments (Fig. 2).
It is noteworthy that all the positions in the amphipathic segment that show significant catalyst-induced cross-linking (Val-32, Trp-35, Ile-36, and Leu-39) map on the hydrophobic side of the helix (Fig. 5B). A possible explanation for this observation is that the residues showing CuP-independent cross-linking are exposed to water-soluble oxidants in the cell extract, whereas residues showing cross-linking induced by the lipophilic CuP catalyst are buried in the bilayer. In this model the amphipathic helix would lie along the surface of the bilayer. Application of this accessibility argument to the CuP-independent cross-linking at the N terminus of the transmembrane helix would suggest that the side chains of the residues in this region can reach the polar head group region of the phospholipids or the aqueous phase outside the bilayer.

Disulfide Cross-linking of a TatB Dimer Does Not Affect the Composition or Stability of the TatBC
Complex-To draw conclusions about the structural significance of the disulfide cross-links between TatB molecules that we had observed, it was necessary to establish that the cross-linked proteins remain within the physiologically relevant TatBC complex. To this end we examined whether two representative crosslinked variants co-purified with TatC. Transmembrane helix variants L11C and V12C were chosen for this analysis because they show high levels of catalyst-induced cross-linking. In each case the variant was co-overexpressed with both TatA and a hexahistidinetagged version of the cysteineless variant of TatC. Membranes containing the overexpressed Tat proteins were subject to catalyst-induced cross-linking, solubilized in the detergent C 12 E 9 , and subject to Ni(II) affinity chromatography. Because the TatC component alone had a hexahistidine tag, only those TatB proteins that remained associated with TatC were retained by the column. The affinitypurified TatBC His complexes were further characterized by analytical size exclusion chromatography (Fig. 7). We first confirmed that removal of the cysteine residues from TatC did not affect the chromatographic properties or stability of the TatBC His complex. We found that the cysteineless complex had a composition and size exclusion column profile that was indistinguishable from the previously characterized parental complex (39) (data not shown).
We next characterized the cross-linked TatB single cysteine variants. The TatBC His complexes of both variants eluted at the same position as the parental complex on a size exclusion column (Fig. 7). Significantly, for both variants a cross-linked TatB species was seen to co-migrate with the TatC His protein and monomeric TatB. These observations confirm that the cysteine cross-linking occurs between TatB proteins within TatBC complexes and that cross-linking of TatB via its transmembrane helix does not detectably perturb the stability of the TatBC complex.
Cross-linking Analysis of TatB Proteins Carrying Double Cysteine Substitutions in the Transmembrane Segment-The disulfide cross-linking data suggest that the transmembrane helices of TatB molecules interact with each other through a single face. The most straightforward explanation for this observation is that TatB molecules form homodimeric substructures within the TatBC complex in which the interacting sides of the two transmembrane helices face each other. This model can be tested using variants in which not one but two residues on the interacting face are substituted with cysteine. If the local organization of TatB really is a dimer, then upon induction of disulfide bond formation the only cross-linked species observed should still be a dimer even though each protomer is now potentially capable of making more than one disulfide interaction. By contrast, if the interacting face of the transmembrane helix is involved in interactions with more than one other TatB protomer, then cross-linked species higher than a dimer should be observed.
To carry out this test, we constructed TatB variants expressing the three possible pairwise combinations of three strongly cross-linking variants from the middle of the transmembrane helix, namely L11C, V12C, and I15C. In each case the doubly cysteine-substituted TatB protein was competent to support the Tat-dependent export of TorA (Fig. 3). Oxidation of membranes co-expressing TatA and cysteine-less TatC together with any of the three double cysteine substitutions of TatB resulted in the formation of further TatB species that migrated more slowly than the dimer form (Fig. 8). For the L11C plus I15C and the V12C plus I15C variants, a single additional product is found, which we assign as a TatB trimer (Fig. 8A). For the L11C plus V12C variant of TatB, at least two further cross-linked products are detected, which we assign as TatB tetramers and pentamers (Fig. 8, A and B). Note that the exact oligomeric state of the products cannot be trivially established by a consideration of apparent molecular weights of the species because we found these to differ markedly in different gel systems (e.g. compare Figs. 8 and 9) and because the TatB monomer shows anomalous migration under SDS-PAGE (6). Nevertheless, it is difficult to envisage how combinations of nonquantitative cross-links can increase oligomeric size by more than one protomer at a time.
To test whether the formation of the multimeric TatB crosslinked species was dependent upon the presence of other Tat proteins, we expressed a C-terminally His-tagged version of the L11C plus V12C TatB variant from a clone that lacked other tat genes and in a strain devoid of all known Tat components. The membranes containing this double cysteine variant of TatB formed an equivalent ladder of oxidant-induced TatB species (Fig. 8C) to that observed in the presence of other Tat components (Fig. 8, A and B). This result indicates that formation of the TatB oligomer does not involve TatC or TatA. It should be noted that a TatB dimer has been detected previously using an amine-specific cross-linker using membranes from a strain devoid of all other Tat components (14).
An important question is whether the multiply cross-linked form of TatB detected with the L11C plus V12C variant reflects the organization of TatB within the TatB complex. It could, alternatively, be associated with TatB protomers that are not associated with TatC (we have established that the cross-linking pattern is not dependent on TatC; see above) or arise from interactions between TatB molecules in different TatBC com-plexes. To distinguish between these possibilities, we investigated whether the cross-linked TatB variant co-purifies with TatC and, if it does, whether the apparent molecular mass of the complexes gives any indication of inter-complex association.
The TatB L11C plus V12C variant was co-expressed with TatA and TatC His . Membranes from these cells were treated FIGURE 9. Disulfide-linked TatB multimers are found within TatBC complexes. A and B, the TatB variant containing both L11C and V12C substitutions was co-expressed with TatAC His from plasmid pUNIPLB1112H in the ⌬tatABCD⌬tatE strain DADE. Membranes from this strain were either subject to oxidant-induced cross-linking (A) or purified in the presence of 2 mM DTT (B). In the control experiment (C ), wild type TatB was co-expressed with TatA-C His from plasmid pUNITATCC4H in strain DADE and processed without addition of oxidation catalyst or reductant. In each experiment the membrane sample of interest was solubilized in digitonin and TatBC His complexes and then isolated by Ni(II) affinity chromatography. The affinity-purified material was further characterized by size exclusion chromatography. The eluted fractions from the size exclusion column were subject to nonreducing SDS-PAGE and immunoblotting with TatB antiserum (A and B, left-hand panels) or with His tag-specific antisera to reveal TatC His (A and B, right-hand panels), or stained with Coomassie Blue (C ). The migration positions of molecular weight standards are indicated, where appropriate, to the left of each gel. The migration positions of the Tat proteins are given at the right-hand side of each gel with ** indicating putative TatB dimers, *** indicating trimers, and **** indicating tetramers. Longer exposures of the TatB immunoblot in A were taken to more clearly reveal the higher order TatB oligomers. Panels for these ϫ2.5 and ϫ15 exposures are labeled as such and are positioned above the main TatB immunoblot. The numbers of the individual fractions from the size exclusion column that were analyzed are given along the bottom of each gel together with the elution positions of the standard proteins ferritin (dimer, 880 kDa; monomer, 443 kDa) and thyroglobulin (669 kDa) as well as the Tat-BC His peak. either with oxidation catalyst (Fig. 9A) to induce the formation of cross-linked TatB oligomers or, as a control, with dithiothreitol (Fig. 9B) to maintain the cysteines in a reduced state. The Tat proteins were extracted from the membrane with digitonin, and the TatC His -containing fractions were isolated by nickel affinity chromatography. The purified material was further analyzed by size exclusion chromatography. Analysis of the cross-linked sample showed that high order oligomers of TatB, up to at least the level of the putative tetramer, co-purify with TatC His (Fig. 9A). Thus the TatB tetramer observed by crosslinking is able to associate with TatC. Size exclusion chromatography was used to determine the apparent molecular mass of the TatBC complexes containing cross-linked TatB molecules (Fig. 9A). The elution peak for complexes containing the TatB dimer species is in the same fraction as the peak for the 669-kDa molecular mass marker (Fig. 9A). This peak elution position is the same as that determined for TatC (Fig. 9A), for the TatB variant run under reducing conditions (Fig. 9B), and for the wild type TatBC complex (Fig. 9C). Thus, TatB dimerization does not detectably perturb the structure of the TatBC complex, a result that is consistent with the analysis of single cysteine variants described above. Interpretation of the chromatographic behavior of the putative TatB trimer and tetramer species is less straightforward. Both species appear to be associated with complexes that elute one fraction before the wild type TatBC complex (Fig. 9A). Nevertheless, the apparent molecular mass of these complexes (Ͻ880 kDa) is still substantially smaller than that expected if the cross-links are formed between TatBC complexes given that the minimal dimer should have an apparent molecular mass of around 1,340 kDa. On this basis we conclude that the observed higher order TatB oligomers arise from interactions within, rather than between, TatBC complexes. We cannot exclude the possibility that the higher order cross-links occur between TatB protomers located within TatBC complexes and free TatB molecules in the membrane. Nevertheless, on balance the purification data indicate that the TatBC complex contains at least four TatB protomers that directly contact each other through their transmembrane domains.
Rather little is currently known about the organization of the TatB and TatC proteins within the E. coli TatBC complex. It is reasonably well established that TatB and TatC are present at an equimolar ratio (19). However, the number of copies of each subunit in the complex is less clear. The apparent molecular mass of the TatBC complex has been variously estimated by gel permeation chromatography to be around 600 kDa (19) (Fig. 9) and by blue native PAGE to be between 370 and 440 kDa (17,20). Both techniques are likely to significantly overestimate the mass of the polypeptide in the complex. Nevertheless, they allow us to put an upper limit of 12 on the number of TatB molecules in the TatBC complex. Our site-specific disulfide cross-linking data provide significant additional information on the organization of TatB with the TatBC complex. In particular, the observation with the L11C plus V12C variant that TatB associates with at least a tetramer through its transmembrane helix puts severe constraints on the organization of TatB. If we make the reasonable assumptions that each TatB or TatC molecule within the complex makes the same protein-protein interactions as any other TatB or TatC molecule, then the observation of TatB tetramer structures within the TatBC complex can only be explained by schemes in which the TatB protomers are located internal to the TatC protomers. The two general types of arrangement that could achieve this are illustrated in Fig. 10, A  and B. An alternative model in which the TatB protomers are located on the outside of the TatBC complex (Fig. 10C) is not compatible with the cross-linking data for two reasons. First, the TatB helices cannot be aligned in such a way as to allow the formation of the disulfide-linked dimers that are observed with the single cysteine variants. Second, given both that the maximum center-to-center distance between two TatB helices that is compatible with disulfide cross-link is about 15 Å, and that the TatBC complex has a radius of around 60 Å (52), then the highest estimate of the number of TatB protomers in the TatBC complex (twelve) is still only about half the number of protomers that would be required to allow the observed higher order disulfide-linked TatB oligomers.
Molecular Modeling and Molecular Dynamics Studies of the TatB Transmembrane Helix-To gain further insight into the structure and interactions of the TatB transmembrane helix, and to aid in interpretation of the cysteine FIGURE 10. Models for the subunit organization within the E. coli TatBC complex. The models are viewed looking down onto the membrane surface and therefore show the possible organization in the plane of the membrane of TatB transmembrane helices (circles) and TatC protomers (shaded boxes occupying roughly the area of the six transmembrane helices found in TatC). The models incorporate the known equimolar ratio of TatB to TatC protomers and are based on the assumption that proteins of the same type will exhibit identical interactions with other proteins. A and B, models consistent with the observations that TatB can be cross-linked to a dimeric state by single cysteine substitutions at residues 11 or 12 and that TatB can be cross-linked to at least a tetrameric state in variants containing both such substitutions. The approximate positions of TatB residues 11 and 12, which lie within the transmembrane helix, are indicated by the unfilled and filled circles, respectively, within the TatB helix. In model A 11 to 11 and 12 to 12 cross-links could be formed with the protomer at the opposite side of the bundle instead of, or as well as with, the adjacent protomers. C, a model in which the TatB protomers are on the outer face of the complex. We argue that this arrangement is unlikely. The relative orientations of the TatB helices do not allow the observed disulfide-linked dimers formed by single cysteine variants. In addition, there are unlikely to be sufficient copies of TatB in the complex to enable individual TatB protomers to come within disulfide-bonding distance of each other. mutagenesis and cross-linking studies, we undertook molecular modeling and molecular dynamics studies of this region of the protein in a palmityloleoyl phosphatidylethanolamine (POPE) bilayer. Simulation of the whole proposed TatB tetramer proved too computationally demanding to be feasible. Instead we carried out a more tractable dimer simulation. We anticipated that this simulation would give an indication of the sort of structural features of the helix that might be important for subunit interactions in the tetramer. Details of the simulation methods are given in the Supplemental Material, and a snapshot of the final dimer model in a bilayer environment is shown in Fig. 11. In the simulation the interhelical interface approximates that inferred in the disulfidemapping experiments.
Analysis of the simulation shows that the dimer model is stable in the bilayer environment. The peptide dimer spans most of the width of the membrane with residues 4 -19 maintaining a helical conformation. Substantially greater structural fluctuations are found at the termini of the TatB peptides than in their helical core as judged by the root mean square fluctuations of the C-␣ atoms from their average value. These fluctuations can best be described as partial unwinding of the helix ends. These observations are consistent with the results of the disulfide mapping experiments, which suggested that the transmembrane region of TatB has a defined helical core structure but has greater conformational flexibility near the membrane surface. Inspection of the simulation suggests that the flexibility of the helix ends may arise from the increased number of possible interactions between the amino acid side chains in these regions with water and the phospholipid head groups.
The lowest root mean square fluctuations of the C-␣ atoms from their average values over the duration of the simulation were seen for residues Ser-7 and Glu-8. This is because the (protonated) side chain of Glu-8 on one peptide is hydrogen-bonded to the side chain of either Glu-8 or Ser-7 on the other helix throughout the course of the simulation. The side chains of the Ser-7 and Glu-8 residues that were not involved in this interaction tended to hydrogen bond with the carbonyl oxygens of the POPE glycerol backbone. Interactions between polar residues have been shown previously by both experimentation (53,54) and by MD simulations (55) to drive transmembrane helix association. Indeed a specific Glu/Serdriven experimental dimerization has been reported (56). Glu-8 is highly conserved in TatB sequences (5,57) and has been suggested to have a role in linking proton movements to the transport process (5,8). Surprisingly, substitution of Glu-8 in the E. coli TatB protein by either cysteine (this work), alanine (48,49), or glutamine (49) did not affect Tat transport activity, although a Glu to Gln substitution in the chloroplast TatB homolog affected association of the protein with TatC (58). These data suggest that Glu-8 may have a structural rather than mechanistic function in Tat transport as indicated by our modeling studies.
The helices in the dimer model extend further toward the N terminus of the peptide (a helix starts at position 4) than is supported by the cross-linking data (a helix is only obvious from position 9). A possible explanation for this discrepancy is that cysteine substitutions at Ser-7 and Glu-8 disrupt the interactions between these residues that stabilize the helical dimer resulting in increased disorder at the helix end in the engineered variants.
In the bilayer TatB dimer simulation residues Leu-11 and Ile-15 show particularly strong interactions across the protomer interface (Fig. 11). In the proteobacterial TatB sequences Leu-11 is highly conserved, whereas one of the ␤-branched amino acids isoleucine or valine is normally found at position 15. It has been argued that ␤-branched amino acids are strongly favored at interfaces between helices because such amino acids have only one rotamer when in an ␣-helix, and this reduces the entropic cost of helix association (59).
In conclusion, cysteine-scanning mutagenesis in the essential region of the E. coli TatB protein failed to identify FIGURE 11. Snapshot of a simulated E. coli TatB transmembrane helix dimer structure in a phospholipid bilayer. The protein model contains the first 25 amino acids of TatB protein. Two copies of this peptide were subjected to molecular modeling and molecular dynamics simulations in a POPE bilayer as described in the Supplemental Material. A snapshot at the end of a 30-ns simulation is shown. The two peptide backbones are shown in ribbon representation within a surface representation of the dimer model. The residue pairs Leu-11-Leu-11 and Ile-15-Ile-15 that are inferred to be in close proximity from the disulfide mapping studies are depicted in stick form. Also depicted are the interpeptide Ser-7-Glu-8 pairs that are usually hydrogenbonded during the simulation and the Pro-22 residue that has been suggested previously to be involved in providing a hinge or kink between the transmembrane and amphipathic helices of TatB. Representative POPE molecules are shown in stick form to give an indication of the position of the phospholipid bilayer It should be noted, however, that for reasons of clarity the POPE molecules depicted are not necessarily those that are most closely packed to the TatB dimer. On the basis of the protease accessibility experiments shown in Fig. 2, the aqueous compartment at the top of the figure can be inferred to be the periplasm and the aqueous compartment at the bottom of the figure the cytoplasm. This figure was prepared in VMD (25).
individual amino acid residues that are important for TatB function. This suggests that it is the overall structural fold of TatB that is important for its role in Tat transport. This would argue against TatB forming a site for the sequencespecific recognition of the twin-arginine motif of Tat signal peptides, an inference that is in agreement with photoaffinity cross-linking studies (23). For the same reason it is unlikely that TatB provides protonatable residues involved in transducing the proton electrochemical gradient into the mechanical work required for Tat transport.
The periodicity of cross-linking seen in scanning disulfide cross-linking studies and molecular dynamics simulations both support the idea that the transmembrane N-terminal region of TatB contains a helical core. The cross-linking studies show that TatB protomers within the TatBC complex interact through this helix. These interactions occur on one face of the helix and involve groups of at least four protomers. Given this arrangement it is most likely that the TatB transmembrane helices are located in the interior of the TatBC complex.
Comparison of the results of this study of TatB with analogous scanning mutagenesis and disulfide cross-linking studies of the homologous TatA protein should be valuable in the future in identifying structural features underlying the different functions of the two proteins.