Dual topology of the Escherichia coli TatA protein.

The Escherichia coli Tat system has unusual capacity of translocating folded proteins across the cytoplasmic membrane. The TatA protein is the most abundant known Tat component and consists of a transmembrane segment followed by an amphipathic helix and a hydrophilic C terminus. To study the operation mechanism of the Tat apparatus, we analyzed the topology of TatA. Intriguingly, alkaline phosphatase (PhoA)-positive fusions were obtained at positions Gly-38, Lys-40, Asp-51, and Thr-53, which are all located at the cytoplasmic C terminus of the TatA protein. Interestingly, replacing phoA with uidA at Thr-53 led to positive beta-glucuronidase fusion, implying cytoplasmic location of the TatA C terminus. To further determine cellular localization of the TatA C terminus, we deleted the phoA gene and left 46 exogenous residues, including the tobacco etch virus (Tev) protease cleavage site (Tcs) after Thr-53, yielding TatA(T53)::Tcs. Unlike the PhoA and UidA fusions, which abolished the TatA function, the TatA(T53)::Tcs construct was able to restore the growth of tatA mutants on the minimal trimethlyamine N-oxide media. In vitro and in vivo proteolysis assay showed that the Tcs site of TatA(T53)::Tcs was accessible from both the periplasm and cytoplasm, indicating a dual topology of the TatA C terminus. Importantly, growth conditions seemed to influence the protein level of TatA and the cytoplasmic accessibility of the Tcs site of TatA(T53)::Tcs. A function-linked change of the TatA topology is suggested, and its implication in protein transport is discussed.

The Escherichia coli Tat system has unusual capacity of translocating folded proteins across the cytoplasmic membrane. The TatA protein is the most abundant known Tat component and consists of a transmembrane segment followed by an amphipathic helix and a hydrophilic C terminus. To study the operation mechanism of the Tat apparatus, we analyzed the topology of TatA. Intriguingly, alkaline phosphatase (PhoA)-positive fusions were obtained at positions Gly-38, Lys-40, Asp-51, and Thr-53, which are all located at the cytoplasmic C terminus of the TatA protein. Interestingly, replacing phoA with uidA at Thr-53 led to positive ␤-glucuronidase fusion, implying cytoplasmic location of the TatA C terminus. To further determine cellular localization of the TatA C terminus, we deleted the phoA gene and left 46 exogenous residues, including the tobacco etch virus (Tev) protease cleavage site (Tcs) after Thr-53, yielding TatA T53 ::Tcs. Unlike the PhoA and UidA fusions, which abolished the TatA function, the TatA T53 ::Tcs construct was able to restore the growth of tatA mutants on the minimal trimethlyamine N-oxide media. In vitro and in vivo proteolysis assay showed that the Tcs site of TatA T53 ::Tcs was accessible from both the periplasm and cytoplasm, indicating a dual topology of the TatA C terminus. Importantly, growth conditions seemed to influence the protein level of TatA and the cytoplasmic accessibility of the Tcs site of TatA T53 ::Tcs. A functionlinked change of the TatA topology is suggested, and its implication in protein transport is discussed.
The twin arginine translocation (Tat or ⌬pH) system has unusual capacity to transport folded proteins and enzyme complex across the bacterial cytoplasmic membrane or the chloroplast thylakoid membrane (1)(2)(3). Four functional Tat components, TatA, TatB, TatC, and TatE, have been identified and characterized in Escherichia coli (4). TatA, TatB, and TatE share sequence homology at their N termini, including one transmembrane segment (TMS) 1 and an adjacent amphipathic helix (APH), whereas their C termini vary both in sequence and in length (5). Expression studies suggest that tatE may be a cryptic gene duplication of tatA (6). Functional Tat translocase has been reconstituted in vitro using membrane vesicles derived from cells overproducing TatA, TatB, and TatC proteins (7,8). A large complex of ϳ650 kDa containing TatABC has been purified from the detergent-solubilized E. coli membrane (9,10) and shown to be capable of binding a Tat signal peptide (10). Recently, Alami et al. (11) have reported a hierarchy in targeting of a Tat substrate to the TatBC complex. For the primary interaction TatC is both necessary and sufficient, whereas a subsequent association with TatB likely mediates transfer from TatC to the actual Tat pore (11). Apparently, TatB and TatC are present in a constant 1:1 stoichiometry in the E. coli TatABC complex, whereas the vast majority of the TatA protein does not co-purify with the TatBC core complex (9). On the other hand a TatAB complex of ϳ600 kDa with a TatA to TatB molar ratio of about 15:1 (10,12) and TatA homo-oligomeric complexes (10,13) have been purified from E. coli membranes. The purification of the different Tat complexes suggests that the Tat pathway utilizes multiple, transiently interacting complexes. Consistently, it has been shown that the thylakoidal TatC-TatB complex homologue, cpTatC-Hcf106, serves as precursor receptor and the TatA homologue, Tha4, assembles with cpTatC-Hcf106 complex during the translocation step (2,14,15). Unlike the E. coli Tat system, the thylakoidal Tha4 has not been detected in the thylakoid Tat complex in the absence of ongoing substrate binding or translocation (2,14,15).
Biochemical and expression studies show that TatA is the most abundant component of the Tat system, present at an ϳ20-fold molar excess over the TatB and TatC components (6,12). Minimal functional units and key residues of TatA and TatB have been determined by genetic and molecular biology approaches (16 -18). C-terminal truncation analysis revealed that the transmembrane and amphipathic helical regions of TatA and TatB are critical for their function but that the C-terminal domains are not essential for Tat transport activity. Using site-specific mutagenesis to probe the significance of conserved features of the related TatA/B subunits leads to the conclusion that an apparent "hinge" region between the transmembrane segment and an APH is important in both proteins (17,18).
To gain further insight in the TatA function and the operation mechanism of the bacterial Tat translocase, we studied the topology of the Tat proteins. PhoA-positive fusions were obtained with TatA. In addition, PhoA and UidA fusions indicated that the TatA C terminus could be located both in the cytoplasm and in the periplasm. Moreover, we observed the influence of growth media on the intracellular level of TatA and TatC and on the cytoplasmic accessibility of the TatA C terminus. This finding strongly suggests dynamic topology changes of the Tat apparatus, which might be coupled to the translocation of folded proteins across the cytoplasmic membrane.
Construction of tatA-phoA, tatA::Tcs and tatA-uidA Fusions-TatA-PhoA fusions were constructed by random in vitro Tn-PhoA transposition as previously reported (20). Briefly, the transposable element carrying the phoA-neo cassette was amplified by PCR using pmodTap as template and pmodFP (5Ј-attcaggctgcgcaactgt-3Ј) and pmodRP (5Ј-gtcagtgagcgaggaagcggaag-3Ј) as primers. The random in vitro transposition of the transposable element into p8737 was performed by using the EZ::TN transposase according to the manufacturer's instructions (Epicentre, Tebu). The transposition reaction product was then introduced by electroporation into CC118 strain. Blue (PhoA ϩ ) Amp R and Kan R colonies were selected on agar plates containing 5-bromo-4-chloro-3indolyl-phosphate (X-P, 40 g/ml). Plasmids were prepared from these colonies, and Tn-PhoA insertion sites were determined by endonuclease digestion, PCR reaction, and DNA sequencing.
The phoA-neo cassette was removed from pG41 with partial NotI digestion and self-ligation, yielding the plasmid pG41D (see Fig. 1). To construct p9913 (tatA T53 -uidA), the phoA-neo cassette inserted at tatA T53 was replaced by the uidA-neo cassette by two steps of cloning. First, the NotI site in the multiple cloning sites of pG41D (tatA T53 ::Tcs) was removed with a XhoI-SalI double digestion and self-ligation, yielding p9910. The uidA-neo cassette was amplified by PCR using pmodTin as template and pmodFP-pmodRP as primers and cloned into the unique NotI site of p9910, resulting in p9913.
Random substitution in the hinge region covering the residues from Gly-21 to Lys-24 of the TatA protein was constructed using gene splicing by an overlap extension protocol (21). Briefly, two DNA fragments were amplified using pG41 as template in separate polymerase chain reactions with the oligonucleotide pairs BglIITatAup (5Ј-gaagatctcgatcccgcgaaattaatacgactc-3Ј) and TatAindown (5Ј-aaaaagcagtacaacgatgacggcaataatcaataa-3Ј) in one reaction, and TatAxup (5Ј-gttgtactgcttt-tt112322433431ctcggctcc-3Ј, where 1, 2, and 3 means 91.6%, respectively, of G, C, and A with 2.77% each of other 3 bases, and 4 represents 91.6% of A with 4.16% of G and 4.16% of C) and SstIITatBdw (5Јtccccgcggcctggcgtagttcatccatcg-3Ј) in another reaction. The two PCR products were then mixed, denatured, and re-annealed and used as template for another PCR by using BglIITatAup and SstIITatBdw as primers. The PCR products were digested by BglII and SstII, cloned into the corresponding sites of pG41, and used to transform CC118. Protein samples were prepared from a dozen of PhoA Ϫ colonies on LB-Amp plates, resolved on SDS-denaturing gels, and analyzed by immunoblot using antisera against PhoA. One of them contained intact TatA T53 -PhoA fusion proteins. DNA sequencing revealed that the plasmid pG41K23I, contained an aa to tt nucleotide alteration leading to an isoleucine substitution for Lys-23.
Enzyme Assays and Bioinformatics-Alkaline phosphatase and ␤-glucuronidase activities were assayed by the hydrolysis of p-nitrophenyl phosphate or p-nitrophenyl glucuronide, respectively. The absorption of the produced p-nitrophenol was measured at 410 nm with a Cary 50 spectrophotometer using a control without extract as the reference blank. One unit of enzyme activity is defined as the release of 1 mol of nitrophenol/min. Specific activity is expressed by unit/mg of protein.
Spheroplast preparation and trypsin sensitivity assay were performed by using the protocol described by Tian and Beckwith (22). In vitro Tev assays were performed on spheroplasts as described by Ehrmann et al. (23). Briefly, cells expressing TatA containing a Tev protease cleavage site were cultivated on LB or M9 media. Harvested cells were washed once in proteolysis buffer (50 mM Tris-HCl, pH8, 0.5 mM EDTA, pH 8) and concentrated to an A 600 ϭ 1 in the same buffer. For proteolysis, 5 mM dithiothreitol and 5 l of Tev protease (50 units) were added to 45 l of cells and incubated at 30°C according to the manufacturer's instructions (Invitrogen). Samples were subjected to SDS-PAGE and Western blotting by using polyclonal antibodies against TatA or PhoA.
Topology prediction of the TatA protein was carried out using the ExPaSy tools available at www.expasy.ch, and secondary structure predication was performed by using the PSIPRED program at bioinf. cs.ucl.ac.uk or the HNN program at npsa-pbil.ibcp.fr.
Cellular Fractionation, Electrophoresis, and Immunoblot-Spheroplasts, cytoplasmic, periplasmic, and membrane fractions were prepared by lysozyme/EDTA/cold osmo-shock and ultracentrifugation, as described previously (24). The protein samples resolved on SDS-denaturing gels by electrophoresis and immobilized onto a polyvinylidene difluoride membrane were analyzed by immunoblot using the ECL ϩ method according to the manufacturer's instructions (Amersham Bioscience). The signals were quantified by using KODAK 1D image analysis software.

Detection of Cellular Location of the TatA C Terminus by
Using PhoA and UidA Fusions-Using an in vitro random Tn-PhoA transposition approach with p8737 ((tatABCD) ϩ ) as a target, a dozen of colonies showing a positive PhoA activity were obtained (20). Further mapping by endonuclease restric- tion enzyme digestion, polymerase chain reaction, and DNAsequencing analysis revealed that four were in TatA after glycine 38 (TatA G38 ), lysine 40 (TatA K40 ), aspartate 51 (TatA D51 ), and threonine 53 (TatA T53 ) (Fig. 1B). The PhoA moiety gains its active form only if it is translocated into the periplasm and is, thus, widely used as a marker for studying membrane protein topology (25,26). Strikingly, the four TatA-PhoA fusions occurred all at the C terminus of TatA, which has been reported to be located in the cytoplasm (13). Therefore, we performed a series of experiments to check such a periplasmic location of the TatA C terminus. Enzyme assay showed that the specific activities of PhoA expressed from the TatA G38 -PhoA, TatA K40 -PhoA, TatA D51 -PhoA, and TatA T53 -PhoA fusions in CC118 (⌬phoA) were 433, 291, 71, and 62, respectively. These values were significantly higher than 1.1, a value obtained with the control plasmid pET22bϩ in CC118. Furthermore, immunoblot analysis revealed that PhoA fusions were completely degraded when the spheroplasts were treated with trypsin ( Fig. 2A1). Therefore, the TatA-PhoA fusions were indeed exposed to the periplasm. In the following experiments we chose to study TatA T53 -PhoA and its derivatives alone for two reasons. First, they gave the strongest signal by immunoblot analysis using anti-TatA antisera. Second, Thr-53 is located downstream of the amphipathic helix that is essential for the TatA function (16). The E. coli TatA protein is anchored in the cytoplasmic membrane via the amino-proximal transmembrane segment (13,27). To obtain a PhoA positive phenotype, the APH must adopt a membrane-span topology to place the TatA C terminus into the periplasm (see Fig. 4C2, see "Discussion"). The hinge region should be crucial for such topology and for the PhoA positive phenotype. To verify this hypothesis we introduced by PCR random substitution in the hinge region of TatA-PhoA fusion expressed from pG41. Altered plasmid pG41 was introduced into CC118 and the transformants obtained were screened for PhoA Ϫ phenotype. Crude extracts were prepared from six PhoA Ϫ colonies. TatA T53 -PhoA fusion protein was detected by immunoblot in one of the strains (data not shown). DNA sequencing analysis revealed that the corresponding plasmid, pG41K23I, encoded TatA T53 -PhoA fusion with a substitution of isoleucine for the lysine at position 23 in the hinge region of TatA (see "Experimental Procedures"). This mutation abolished the PhoA activity but did not affect the stability of the TatA T53 -PhoA fusion protein, indicating that the PhoA moiety is blocked in the cytoplasm. This finding suggests that the lysine 23 is essential for the membrane span topology of the TatA amphipathic helix. Interestingly, it has been recently reported that changing residues in the hinge region of TatA leads to the defect in TMAO reductase export (17,18).
To investigate further the topology of TatA, we replaced phoA by uidA and constructed the TatA T53 -UidA fusion. The uidA gene encodes the ␤-glucuronidase, which is active in the cytoplasm and has been generally used to identify cytoplasmic located segment of membrane proteins (23). The EZ4/p9913 (uidA Ϫ /tatA T53 -uidA ϩ ) displayed a ␤-glucuronidase activity of 86 units, which was 30-fold higher than the basic level observed for the same mutant carrying the control plasmid p8737. Unlike the TatA T53 -PhoA fusion, the TatA T53 -UidA fusion was resistant to trypsin digestion in the spheroplast-trypsin sensitivity assay ( Fig. 2A2 compared with 2A1). These results indicate a cytoplasmic location of the TatA C terminus of the TatA T53 -UidA fusion.
The PhoA-and UidA-positive phenotypes indicated that the TatA C terminus might exist in two different topologies. The requirement of the other Tat proteins for the periplasmic location of the PhoA moiety of the TatA T53 -PhoA fusion was analyzed by measuring the PhoA activity in tat mutants. The ⌬tatB, ⌬tatC, and ⌬tatABCDE mutations resulted in about a 3-fold decrease of the PhoA activity (Table II). Immunoblot analysis revealed that the amount of the TatA T53 -PhoA decreased in the tat mutants. In contrast, these mutations slightly increased the ␤-glucuronidase activity. Together these results would suggest that the presence of the TatB and TatC proteins would be more favorable for a periplasmic location of the TatA C terminus.
Cellular Location of the C Terminus of Functional TatA::Tcs Chimera-The PhoA and UidA fusions indicated a dual topology of the TatA C terminus. Because both TatA-PhoA and TatA-UidA lost the TatA function (data not shown), we constructed another TatA derivative to analyze the topology of TatA under physiological conditions. The phoA and neo genes of pG41 were deleted by NotI digestion (Fig. 1C), and a fragment of 138 bp was left at the Tn5 insertion site (23). The encoded 46 amino acids (see "Experimental Procedures"), containing a tobacco etch virus (Tev) protease cleavage site (Tcs), were thus inserted in-frame after Thr-53 (Fig. 1D), yielding pG41D encoding TatA T53 ::Tcs.
The synthesis and cellular location of the TatA T53 ::Tcs were analyzed by immunoblot. The anti-TatA polyclonal antisera recognized TatA and, more weakly, an additional nonspecific polypeptide in the urea-washed membrane fractions prepared from the wild type strain carrying the plasmid p8737 ((tatABC) ϩ ) (Fig. 2B, lane 1). As anticipated, the TatA protein was absent from the ⌬tatAE mutant, and the plasmid pG41D specifically directed the synthesis of a polypeptide bigger than TatA and with the size expected for TatA T53 ::Tcs in this mutant (Fig. 2B, lane 2). Both TatA and TatA T53 ::Tcs were detected in the membrane fractions of the ⌬tatB and ⌬tatC mutants carrying the plasmid pG41D (tatA T53 ::Tcs-(tatBC) ϩ ) (Fig. 2B, lanes 3  and 4). Importantly, the plasmid pG41D (tatA T53 ::Tcs-(tatBC) ϩ ) was able to restore the growth defect of the ⌬tatAE, ⌬tatB, ⌬tatC, and ⌬tatABCDE mutants (Fig. 3A1) on minimal TMAO/glycerol liquid media. Therefore, TatA T53 ::Tcs preserves the TatA function, and the insertion of 138-bp in the tatA gene had no effect  on the synthesis of the downstream TatB and TatC proteins, which was confirmed by immunoblot (data not shown). The cellular accessibility of the Tcs site within the physiologically active TatA T53 ::Tcs and of those located at the junction between TatA T53 and PhoA/UidA in the non-functional TatA-PhoA and TatA-UidA fusions was assessed by proteolysis and immunoblot analysis of cells grown in rich LB media. As anticipated, upon the treatment of the spheroplasts with exogenously provided Tev protease, the TatA T53 -PhoA fusion was reduced to the size corresponding to the PhoA protein, implying the separation of the PhoA moieties from TatA T53 (Fig. 2C1,  lane 3, compared with lane 2). Interestingly, when the spheroplasts were subjected to an in vitro Tev treatment the intensity of the TatA T53 ::Tcs band was significantly reduced (Fig. 2C2,  lanes 3 versus 2). Notably, the nonspecific band in the Tevtreated fraction was stronger than that in the non-treated fraction. Therefore, the decrease of the TatA T53 ::Tcs intensity in the Tev-treated fraction could not be due to an uneven loading of the samples. Unlike the TatA T53 -PhoA fusion, TatA T53 ::Tcs was not completely digested by the in vitro Tev treatment. These results would suggest that not all Tev cleavage sites in the TatA T53 ::Tcs are accessible from the periplasm and that the PhoA moiety of the TatA-PhoA fusion blocks the C terminus in the periplasm.
To analyze the cytoplasmic accessibility of the Tev cleavage sites, we co-introduced pMM13 plasmid carrying the tev gene or the control pACYC184 plasmid together with those directing synthesis of TatA derivatives into the DADE strain. The tev gene on pMM13 is under the control of an IPTG-inducible promoter. As anticipated, the Tev cleavage site in the TatA T53 -PhoA chimera was inaccessible from the cytoplasmic side (Fig.  2D1, lane 3 compared with lane 1). A substantial amount of the TatA T53 -UidA fusion was degraded in the presence of pMM13 (tev ϩ ) under leaky conditions (without the IPTG induction), and the fusion protein completely disappeared when the tev gene expression was induced by IPTG (Fig. 2D2, lanes 2 and 3,  respectively). Interestingly, only when the Tev synthesis was induced by IPTG, the TatA T53 ::Tcs band was substantially degraded (Fig. 2D3, lanes 3 versus 2). Together these results showed that the C terminus of TatA is accessible from both the periplasm and the cytoplasm, suggesting an alternation of the TatA C terminus topology that might be important for the Tat function.
Influence of Growth Conditions on the Level of TatA and on Its C-terminal Location-The finding of both the periplasmic and cytoplasmic location of the TatA C terminus and the TatB and TatC dependence of the periplasmic localization raised the question of whether TatA undergoes a topology change to perform its function. We sought to analyze the cellular location of the TatA C terminus under different growth conditions. The Tat system is essential for anaerobic growth on TMAO/glycerol minimal media. In this case, TMAO is reduced by the periplasmic TMAO reductase (TorA) and the membrane-bound dimethyl sulfoxide (Me 2 SO) reductase (DmsABC) and serves as a terminal electron acceptor in anaerobic respiration (28). Because tat mutations lead to mislocation of these two enzymes, tat mutants are incapable of anaerobic growth with TMAO as the sole electron acceptor unless they are complemented in FIG. 3. Influence of growth conditions on the level and cytoplasmic side accessibility of the TatA T53 ::Tcs C terminus. DADE (⌬tatABCDE) mutant carrying pET22bϩ (squares) or pG41D (tatA T53 ::Tcs) and pMM13 (tev) (circles and triangles) were incubated in anaerobiosis on minimal TMAO/glycerol media (filled symbols) or minimal fumarate/glycerol media (open symbols) at 37°C. IPTG was added in the cultures indicated by triangles. Cultures were inoculated from LB-pre-cultures, and optical density (A 600 ) was measured at times indicated (panel A). Twenty micrograms of crude extracts prepared from the corresponding cultures were resolved on SDSdenaturing polyacrylamide (15%) gels, and the TatA T53 ::Tcs protein was visualized by immunoblot using antisera against TatA (panel B). Twenty micrograms of crude extracts prepared from DADE (panel C, lanes 3 and 4) or MC4100 (other lanes) grown on either TMAO/glycerol (T) or fumarate/glycerol (F) minimal media were resolved on SDS-denaturing polyacrylamide gels, and the TatA and TatC proteins were visualized by immunoblot using antisera against TatA (lanes 1-3) or TatC (lanes 4 -5). The nonspecific bands recognized by the anti-TatA or anti-TatC antisera are indicated by NS. trans by functional tat genes (29). As shown in Fig. 3A1, when inoculated from fresh LB pre-cultures into minimal TMAO media, DADE (⌬tatABCDE) exhibited a residual growth for about 5 h until A 600 ϭ 0.2 under the conditions used. Almost no further growth was observed from then on (Fig. 3A1, filled  square). The introduction of pG41D (tatA T53 ::Tcs-(tatBCD) ϩ ) restored the growth of DADE on the TMAO/glycerol minimal media (Fig. 3A1, filled circle). The presence of pMM13 (tev ϩ ) under non-induced conditions did not interfere with the restoration of the DADE/pG41D growth (Fig. 3A1, filled circle). Remarkably, the expression of the tev gene induced by IPTG completely abolished the restoration of the DADE/pG41D growth (Fig. 3A1, filled triangle). Immunoblot analysis revealed that the TatA::Tcs was detected only in the absence of Tev protease (Fig. 3B, lane 2). Together these results showed that the Tev cleavage site of TatA::Tcs was accessible from the cytoplasm, and the digestion of TatA::Tcs by Tev abolished the Tat function.
Fumarate reductase is a membrane-bound flavoenzyme that catalyzes the final step in anaerobic respiration when fumarate is the terminal electron acceptor (28). The membrane assembly of fumarate reductase is Tat-independent (30). As anticipated, DADE (⌬tatABCD) grew equally well in the absence as in the presence of the tatA T53 ::Tcs-(tatBCD) ϩ genes on the fumarate/ glycerol minimal media (Fig. 3A2). Moreover, expression of the tev gene had no effect on the growth (Fig. 3A2, open triangle). Intriguingly, TatA::Tcs was detected at a much lower level compared with the cells grown on the TMAO/glycerol minimal media. Most importantly, it was resistant to the cytoplasmic Tev proteolysis (Fig. 3B, lanes 4 and 5), which was in remarked contrast to the TatA::Tcs found in cells grown on the TMAO/ glycerol minimal media (Fig. 3B, lanes 2 and 3).
The TatA::Tcs was synthesized from the plasmid bearing tatA T53 ::Tcs gene. We sought to check if the level of TatA and TatC from chromosomal copy of the tatA and tatC genes was also influenced by the growth conditions (Fig. 3C). In these cases differences were observed, but they were much less important than that observed for the TatA::Tcs. To gain more accurate comparison data, the immunoblot results were quantified by using KODAK 1D image analysis software. The nonspecific bands recognized by anti-TatA or anti-TatC sera were used as internal controls. The ratios of TatA or TatC signal over the signal of the corresponding nonspecific band were compared. We repeatedly observed about a 1.5-and 2-fold increase of the TatA and TatC signals, respectively, in the cells grown on the TMAO/glycerol minimal media comparing with those grown on the fumarate/glycerol minimal media (Fig. 3C, lane 1  versus lane 2 and lane 5 versus lane 6). Recently, genome-wide analysis using DNA array techniques has revealed that TMAO modulates the expression of eight transcriptional units via the TorS/TorR two-component system in E. coli (31). The tat operon is not among these transcriptional units. Consistently, the 126-bp fragment upstream the tat operon of the p8737 does not contain the highly conserved TTCAnA motif of the TorR binding box (31,32). Taken together, these results might suggest a fast turnover of these proteins in the order of TatA::Tcs Ͼ TatC Ͼ TatA, whereas their synthesis remained unchanged when cells were grown on the fumarate/glycerol minimal media. Alternatively, the interaction between the TMAO reductase and the Tat proteins might protect the latter from degradation. DISCUSSION The E. coli TatA protein is anchored in the cytoplasmic membrane via the amino-proximal TMS, and the membrane association stabilizes the adjacent APH (13,27). These two domains are critical for the TatA function (16). When expressed without the N-terminal TMS domain the truncated TatA is largely unstructured in aqueous solution but is able to insert into phospholipid monolayers and interacts with membrane bilayers (13). There is no charged residue in front of the TMS or in the TMS of TatA (Fig. 4A), but a positively charged Lys-Lys (KK) cluster is located in the hinge region that connects TMS with APH. Application of the "Positive-Inside" rule (26) would suggest that TMS has a N out -C in topology (Fig. 4C). This topology is supported by reported cytoplasmic location of the TatA C terminus (13,33,34) by the TatA-UidA fusion and by the cytoplasmic accessibility of the Tev cleavage site in TatA T53 ::Tcs. Intriguingly, another double lysines (KK) motif is found at the end of the amphipathic helix. The two KK clusters might interact with the negatively charged phospholipid head groups either on the same side (Fig. 4C1) or at the opposite sides (Fig. 4C2) of the membrane bilayers. Using the PhoA, UidA fusions, and the Tcs insertion we observed that the Tcs site after Thr-53 in the hydrophilic domain of TatA could be exposed to both the periplasmic and cytoplasmic side of the membrane, implying a dual topology of the APH of TatA. When cells were grown on fumarate/glycerol minimal media, the C terminus was not accessible from the cytoplasm. In this case APH must have the membrane-spanning topology. Because tat mutants grow normally on fumarate/glycerol minimal media, the membrane-spanning topology of the amphipathic helix might reflect a resting state or less required state of TatA. The Tat function is essential for bacterial growth on TMAO/glycerol minimal media. Under these conditions the TatA C terminus was accessible from the cytoplasm (Fig. 3); hence, the amphipathic helix must lay at the cytoplasmic side of the inner membrane. Because the C terminus of TatA is not essential for its function (18), the dual topology of the amphipathic domain but not the cellular location of the C terminus is required for the Tat function. We observed that mutation in the hinge region affected the C terminus-out topology of TatA. Interestingly, it has been reported that the hinge region and the hydrophobicity of the amphipathic helix are crucial for the Tat function (17,18).
The E. coli TatA and TatB proteins share homology in the transmembrane segments and the adjacent amphipathic helices, but their C termini are different both in length and in sequence (5). Recently, Lee et al. (16) report that, like TatA, the transmembrane segment and the amphipathic helix are essential for the TatB function. Does TatB also change its topology during protein export through the Tat apparatus? The amphipathic helix of TatB is almost twice longer than that of TatA and does not contain double basic residues at the ends of the amphipathic helix (Fig. 4). Interestingly, using a random Tn-PhoA insertion approach with the tatABCD genes as target, we obtained a dozen of PhoA ϩ fusions only in tatA and tatC but not in tatB or tatD (20) (this study). These results would suggest that the C terminus of TatB is not accessible from the periplasm. Bolhuis et al. (9) have constructed a TatB-TatC fusion protein by connecting the C terminus of TatB to the N terminus of TatC. Because the N terminus of TatC is located in the cytoplasm (20,35,36), the fusion should prohibit the C terminus of TatB changing the cellular location. Interestingly, the TatB-TatC fusion protein functionally complements a null mutant lacking both TatB and TatC in E. coli. Therefore, TatB might not change its topology to perform Tat function.
The TatC proteins are highly conserved within various organisms (1). This group of proteins contains six predicted membrane-spanning domains with N and C termini exposed to the cytoplasmic or stromal side of membrane (9,20,37,38). Strikingly, using a PhoA and UidA protein fusion approach we recently showed that the predicted fourth and fifth transmembrane helices and the predicted second cytoplasmic loop between the two helices of E. coli TatC protein are located in the periplasm (20). This topology might reflect an operational state of TatC, which might also change its topology in the protein translocation process.
Based on the increasing biochemical and genetic evidence, Mori and Cline (2,14,15) have recently proposed a cyclical assembly model for the components of the thylakoidal Tat pathway. In the resting state, with no precursor and no thylakoidal pH gradient, the cpTatC-Hcf106 (TatC-TatB) and Tha4 (TatA) complexes are separately present in the membrane (resting state). Precursor binds the cpTatC-Hcf106 complex. This step could be reversible or weak for most genuine precursor (binding step). The pH gradient triggers Tha4 recruitment to the precursor-bound cpTatC-Hcf106 complex and causes a rearrangement of components to form an active translocon (assembly step). Upon completion of protein translocation, Tha4 is released, and cpTatC and Hcf106 reform the cpTatC-Hcf106 complex (disassembly step). To this cyclical assembly and flexible channel model, we propose topology changes of TatA and TatC for the mechanism of folded protein translocation (Fig. 5). After recruitment of TatA to the precursor-bound TatC-TatB complex, the membrane-spanning amphipathic helix of TatA might move into the cytoplasm, and the fourth and fifth helixes of TatC might move from the periplasm into the inner membrane (Fig. 5, B-C). The TatA⅐TatB⅐TatC-substrate complexes undergo a rearrangement, and multiple amphipathic helices of TatA change back to the membrane-spanning topology and form a hydrophilic cavity to accommodate the folded substrate. Berks et al. (4) suggest that multiples of TatA assemble around a core TatC to form the channel (4). The stoichiometry between TatA and TatB⅐TatC complex might vary depending on the size of substrate to be exported. Expansion and contraction of the hydrophilic cavity would be accomplished by adding or removing TatA monomers. The fourth and fifth helices of TatC and the loop region between them might move back to the periplasm, driving the translocation of the substrate across the membrane and in parallel with the disassembly of the Tat translocase complex.
The salient feature that we proposed for the mechanism of folded protein translocation across the cytoplasmic membrane is the change of TatA and TatC topology. It is unusual but not unique to the Tat pathway. In E. coli the functional minimal Sec translocase comprises a membrane-embedded complex of SecY and SecE and a peripheral membrane domain, SecA (39,40). SecA delivers the preprotein to the protein-conducting channel by undergoing ATP-driven cycles of membrane insertion and deinsertion (41). SecG possesses two strongly hydrophobic regions with a weakly hydrophobic one between them (42). The SecG protein shows an unusual property of inverting its orientation in the membrane, which is tightly coupled to the SecG function and linked with the insertion-deinsertion cycle of SecA (43). Another example is from the study of the voltagedependent gating of the colicin A channel, which involves a substantial structural rearrangement resulting in the transfer of about 35% of the 200 residues in its pore-forming domain across the membrane (44). Intriguingly, such a structural rearrangement of colicin A is capable of moving exogenous proteins of up to 134 residues across lipid bilayers and presenting them in a functional form to the trans solution (45). A basic mechanism of membrane insertion and topology reversion thus may be widely used by apparently distinct systems in protein translocation.