Kinetics of Precursor Interactions with the Bacterial Tat Translocase Detected by Real-time FRET*

Background: The Tat machinery transports folded proteins from the bacterial cytoplasm to the periplasm. Results: A Δψ is required for a Tat cargo to move away from the TatBC receptor complex. Conclusion: Cargo migration away from the TatBC complex requires a Δψ-dependent assembly step or conformational change. Significance: Cargo migration from the TatBC receptor is a major rate-limiting step of Tat transport. The Escherichia coli twin-arginine translocation (Tat) system transports fully folded and assembled proteins across the inner membrane into the periplasmic space. Traditionally, in vitro protein translocation studies have been performed using gel-based transport assays. This technique suffers from low time resolution, and often, an inability to distinguish between different steps in a continuously occurring translocation process. To address these limitations, we have developed an in vitro FRET-based assay that reports on an early step in the Tat translocation process in real-time. The natural Tat substrate pre-SufI was labeled with Alexa532 (donor), and the fluorescent protein mCherry (acceptor) was fused to the C terminus of TatB or TatC. The colored Tat proteins were easily visible during purification, enabling identification of a highly active inverted membrane vesicle (IMV) fraction yielding transport rates with NADH almost an order of magnitude faster than previously reported. When pre-SufI was bound to the translocon, FRET was observed for both Tat proteins. FRET was diminished upon addition of nonfluorescent pre-SufI, indicating that the initial binding step is reversible. When the membranes were energized with NADH, the FRET signal was lost after a short delay. These data suggest a model in which a Tat cargo initially associates with the TatBC complex, and an electric field gradient is required for the cargo to proceed to the next stage of transport. This cargo migration away from the TatBC complex requires a significant fraction of the total transport time.

The bacterial twin-arginine translocation (Tat) 2 machinery transports protein cargos from the cytoplasm to the periplasm (1,2). A functional Escherichia coli Tat machinery minimally consists of three membrane proteins, TatA (or TatE), TatB, and TatC (3)(4)(5)(6)(7). TatB and TatC are expressed at approximately equal levels (8), and together act as the receptor for precursor proteins (9,10). TatA is expressed at higher levels than TatB and TatC (8), and can form ring-like structures in vitro (11). A popular model is that cargos are first recruited by TatBC complexes, and are then conveyed across the membrane by a protein conducting channel comprised of TatA oligomers (11,12). After transport, Tat signal peptides are cleaved from precursor proteins by the LepB peptidase (13,14). Recent studies have suggested that precursor binding to the cytoplasmic face of the inner membrane is a functional intermediate in the transport process (15,16). A proton motive force (PMF) is essential for translocon assembly and cargo transport (17)(18)(19). In E. coli, only the membrane potential (⌬) component of the PMF is required for transport (20).
Both TatB and TatC participate in binding and recognition of signal peptides. Bacterial Tat signal peptides contain an (S/T)RRXFLK consensus motif, a hydrophobic domain, and a short polar domain which precedes the signal sequence cleavage site (21)(22)(23). Based on mutant suppression, alanine substitutions and cross-linking experiments, the N-terminal half of TatC interacts with the twin-arginine portion of the signal peptide (24 -27). Cross-linking indicates that TatB interacts with the C-terminal end of the signal peptide (24,27). TatB also makes extensive contacts with the cargo mature domain, presumably through its cytoplasmic domain (28).
It is unclear how the cargo proceeds across the membrane after recognition by the TatBC complex. A severe constraint on possible models is the finding that the mature domain can be efficiently translocated when the signal sequence is crosslinked to TatC near the twin-arginine motif (27). One possibility is that TatA somehow assists with flipping the mature domain from one side of the membrane to the other while the signal sequence remains tethered (27). In this picture, the signal peptide may initially only partially occupy a transmembrane binding pocket before transport, and full binding only occurs during or after transport, consistent with the identification of deep insertion of the signal peptide and distinct translocation intermediates (29 -31). Electron microscopic images of purified TatBC complexes revealed cargo bound within the membrane plane, possibly reflecting cargo in transit (32).
Fluorescence resonance energy transfer (FRET) can provide time-dependent distance information at low protein concentrations. Here we report the use of FRET to probe the binding of a cargo to the TatBC complex, and its migration away from this complex, either by dissociation or by movement along the transport pathway. We found that the TatBC complex has a nanomolar affinity for pre-SufI, and that an electric field gradient is required for migration beyond the initial binding step.
The plasmid pIntein-preSufI-MA was generated by inserting a cysteine-free form of pre-SufI with a 6ϫHis tag into the vector pTYB11 (New England Biochem). Plasmid pIntein-pre-SufI-MA was used as the template for all pre-SufI mutants, as described in supplemental Fig. S1. Plasmids pTatAB cherry C and pTatABC cherry were generated by insertion of the mCherry sequence (plasmid pmCherry, Clontech) into pTatABC (13) immediately after the coding regions of TatB and TatC, respectively (supplemental Figs. S2 and S3). pTatBC and pTatBC cherry were generated by excising the TatA sequence from pTatABC and pTatABC cherry , respectively (supplemental Fig. S4). All mutations were generated via the Quikchange protocol (Stratagene), and all plasmid coding sequences were verified by DNA sequencing. Plasmid pTorA-mCherry-H6 was generated by inserting mCherry along with the TorA signal peptide and 4 amino acids (AQAA) of the TorA mature domain into the pBAD24 vector (supplemental Fig. S5).
Protein Purification and Labeling-All pre-SufI proteins were purified from expression cultures via chitin chromatography as described previously (37) with an on-column cleavage (50 mM dithiothreitol) at 4°C for 20 h (supplemental Fig. S6). Excess dithiothreitol was removed from the eluate by dialysis (10,000 MW cutoff) against phosphate-buffered saline (1ϫ PBS: 137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , and 2 mM KH 2 PO 4, pH 7.5) for 4 h. Ni-NTA chromatography was used for further purification (20). Alexa532 labeling of pre-SufI proteins was performed as described previously (15). After a 4 h dialysis against 1ϫ PBS buffer, excess dye was removed by an additional Ni-NTA purification step (15). The final storage buffer was 20 mM Tris-HCl, 50 mM NaCl, 250 mM imidazole, 50% glycerol, pH 8.0. Purification of spTorA-mCherry-H6 was performed by Ni-NTA chromatography, by the procedure used for pre-SufI (20). The mCherry protein was obtained by purifying the mature protein from a spTorA-mCherry-H6 preparation via Resource Q (GE Healthcare) chromatography.
The protein concentrations of all pre-SufI proteins were quantified by SDS-PAGE using bovine serum albumin as the standard. Spot intensities after Coomassie Blue staining were determined with a phosphorimager (model FX PhosphoImager, Bio-Rad). Alexa532 concentrations were determined by absorbance spectroscopy at 532 nm (⑀ ϭ 81,000 cm Ϫ1 M Ϫ1 ).
Typical dye-to-protein ratios after labeling indicated that 80 -90% of the cysteines were successfully tagged. mCherry concentrations were estimated in 2% SDS based on the mCherry absorbance at 587 nm and its extinction coefficient at this wavelength (38).
Isolation of IMVs-IMVs were isolated from MC4100⌬TatABCDE as described (20), with the following modifications. Cell lysis by French press was performed at ϳ16,000 psi instead of ϳ6,000 psi. In addition, the 2.3 M sucrose cushion was replaced with a 3-step (0.5, 1.5, and 2.3 M) sucrose gradient, which enabled enrichment of a highly active inner membrane fraction. A band in the 0.5 M region was faintly pink when mCherry Tat fusions were expressed (supplemental Fig.  S7), and translucent dark brown when wild type TatABC was expressed. IMV concentrations were determined as the A 280 in 2% SDS (20).
Transport and Membrane Binding Assays-In vitro transport assays and precursor-membrane binding assays were performed as described (15,20), unless otherwise indicated. Gelbased transport kinetics were performed in a heated cuvette to mimic conditions in which the real-time FRET assays were performed. Aliquots (35 l) of the reaction mixture were removed from the cuvette at the indicated time points and quenched with a mixture of 5 M nigericin and 5 M valinomycin on ice. Visualization of protein bands was performed by direct in-gel fluorescence imaging (model FX PhosphorImager, Bio-Rad).
Real-time FRET Assay-Translocation buffer (5 mM MgCl 2 , 50 mM KCl, 200 mM sucrose, 57 g ml Ϫ1 BSA, 25 mM MOPS, 25 mM MES, pH 8.0) (20) and IMVs (A 280 ϭ 2) were added to a heated cuvette (final volume: 800 l) in an SLM-Aminco fluorometer and allowed to equilibrate to 37°C for 5 min. Precursor protein was then added, and equilibration was continued for at least 200 s. Reactions were initiated by addition of competitor precursor or NADH. Excitation and emission wavelengths were 500 nm (4 nm slits) and 550 nm (8 nm slits), respectively.
Analysis-Kinetic data were fit with a single exponential plus a linear baseline drift, where a, b, and k ( ϭ 1/) are fit parameters. The data in Fig. 2D were fit to the equation for single site binding when the K D and receptor concentration are of similar magnitude (derived in the supplemental materials), where C, T 0 , and K D are fit parameters.

RESULTS
Experimental Design-Our goal was to examine the interaction between the TatBC receptor complex and a cargo under real-time transport conditions. To this end, we used FRET, with a donor fluorophore (Alexa532) on the cargo and an acceptor fluorophore (mCherry) on one of the Tat proteins. We assumed that binding would result in a decrease in donor fluorescence due to FRET. The fluorescent cargo selected was the natural Tat substrate pre-SufI. Eight different single cysteine mutants of pre-SufI were generated. The different cysteine mutations were spread over the surface of the protein (Fig. 1A). These mutants were labeled with Alexa532 maleimide. The C-terminal cysteine mutant, pre-SufI(479C), serves as our wild type (wt) reference because previous experiments indicated that a C-terminal dye had no effect on transport efficiency (15). The T96C and 479C mutants yielded the highest transport efficiencies (Fig.  1B).
To exclude any endogenous (and therefore unlabeled) Tat proteins from IMV preparations, TatB cherry and TatC cherry were expressed in the Tat deletion strain MC4100⌬TatABCDE. The mCherry domain did not significantly affect transport efficiency ( Fig. 1C) or transport kinetics (Fig. 1D). The fluorescent TatB and TatC proteins allowed us to visually monitor the IMV purification process. Consequently, we identified a minor membrane-containing band within the 0.5 M sucrose region of the sucrose gradient that had the majority of the mCherry protein (supplemental Fig. S7). The MC4100 membranes recovered from this band catalyzed pre-SufI transport about an order of magnitude faster (Fig. 1D) than reported previously for JM109 IMVs (15,20). Further, IMVs prepared with our current protocol were more consistently active. Previously, we were not able to consistently obtain active membranes from MC4100 (20).
The FRET Signal-IMVs isolated from E. coli expressing TatAB cherry C or TatABC cherry exhibited two fluorescence peaks at ϳ610 nm and ϳ641 nm (supplemental Fig. S8, A and B). These peaks both arise from mCherry (supplemental Fig.  S8C). Addition of pre-SufI Alexa532 resulted in a third peak at ϳ550 nm ( Fig. 2A). The decrease in donor fluorescence signal due to FRET was most easily verified by addition of a competitor protein (non-fluorescent cargo), which resulted in an increase in the 550 nm peak, presumably due to replacement of the bound fluorescent protein with the non-fluorescent competitor ( Fig. 2A). No increase in 550 nm emission was observed upon competitor addition if mCherry was not attached to the Tat proteins (supplemental Fig. S8D), indicating that the decrease in 550 nm emission when the fluorescent cargo was bound to the Tat translocon ( Fig. 2A) was indeed due to FRET. The strong mCherry fluorescence emission far from the excitation wavelength of 500 nm indicates a high concentration relative to the cargo. The mCherry concentration was estimated as 322 Ϯ 88 nM (n ϭ 5) under typical assay conditions, over an order of magnitude higher than the pre-SufI concentration of 20 nM ( Fig. 2A). These values are consistent with the amounts of pre-SufI Alexa532 and mCherry needed to approximately reproduce the emission spectra in Fig. 2A in an IMVfree mixture (supplemental Fig. S8D). Any increase in mCherry emission due to FRET was weak and not reliably detected. This is expected if the mCherry acceptor molecules self-quench due to their proximity in a TatBC oligomer (32). Maximizing the FRET Efficiency-Having determined that FRET indeed occurred upon binding of the cargo to the Tat translocon, we next sought to maximize the FRET signal. This was accomplished by individually attaching the donor dye to the 8 locations identified in Fig. 1A, which cover the surface of pre-SufI. In addition, the mCherry acceptor was attached to either TatB or TatC. FRET signal intensity was determined from kinetic experiments in which non-fluorescent competitor cargo was added to IMVs with prebound fluorescent cargo (Fig.  2B). For pre-SufI(T96C), the competitor released the bound cargo with ϭ 24 Ϯ 2 s (Fig. 2B). Control kinetic experiments in the absence of a donor or acceptor fluorophore confirmed the transient FRET signal (supplemental Fig. S9A). We found that TatC cherry yielded stronger FRET signals than TatB cherry (Fig.  2C). Two pre-SufI mutants, G29C and T96C, yielded similarly strong FRET signals when the mCherry domain was attached to TatC (Fig. 2C). However, labeling at G29C reduced transport efficiency (Fig. 1B). To convert the relative FRET signals into reliable distance and/or orientation information, the data in Fig. 2C need to be corrected based on binding affinity. We attempted to do this using our previous binding assay (15), but the errors of these binding measurements and the errors of the FRET measurements were too high to provide meaningful information. Instead, we settled on using TatC cherry and the T96C mutant of pre-SufI for all subsequent experiments since these proteins yielded the strongest FRET signal while retaining transport efficiency (Figs. 1B and 2C).
Binding Affinity-The FRET signal was used to estimate the cargo-TatBC binding affinity. Two titrations were done. The first titration involved adding increasing amounts of fluorescent pre-SufI(T96C) to IMVs. The second titration was identical to the first, except that a large excess of non-fluorescent cargo was preincubated with the IMVs (supplemental Fig. S10). The difference in donor fluorescence intensity between the two titrations (FRET signal) reflects the amount of fluorescent cargo bound to the translocon (Fig. 2D). Based on the mCherry concentration, the Tat receptor concentration is high relative to the apparent K D . Thus, the data in Fig. 2D were fit with Equation 2 ("Experimental Procedures" and supplemental materials), which explicitly includes the receptor concentration as a fit parameter. Unfortunately, the receptor concentration and K D are not uniquely determined from the data. However, the data indicate that the functional receptor concentration is Յ ϳ20 nM and the K D Ϸ 7-23 nM (see Fig. 2D and supplemental materials for a detailed explanation).
FRET Decreases upon Membrane Energization-When fluorescent pre-SufI was prebound to IMVs containing Tat-mCherry fusions and the membranes were subsequently energized by the addition of NADH, the donor fluorescence signal increased (a decrease in FRET). This was observed for membranes containing either TatB cherry or TatC cherry (Fig. 3A). Control experiments in the absence of donor or acceptor fluorophore confirmed the NADH-dependent loss of FRET (supplemental Fig. S9B). Since NADH addition initiates transport across the membrane, these data are consistent with migration of the cargo away from its initial binding site on the TatBC complex as part of the transport process. It is unclear if the observed kinetics reflect migration elsewhere within the translocon (e.g. to the TatA pore) or immediate movement across the membrane. NADH generates a PMF, which is necessary for Tat transport. It was shown earlier that the ⌬ and not the ⌬pH component of the PMF is essential for Tat transport (20). We therefore tested whether the observed FRET signal is sensitive to these two PMF components. We found that the decrease in FRET upon NADH addition requires a ⌬ and not a ⌬pH (Fig. 3B), consistent with the hypothesis that the observed changes in the FRET signal report a transport substep.
A ⌬-dependent Step Precedes Cargo Migration Away from the TatBC Complex-Exponential NADH-dependent FRET changes ( ϭ ϳ30 s) were preceded by a lag phase (Fig. 3A). The duration of the delay was dependent on the batch of IMVs (e.g. compare Fig. 3, A and B), and ranged from ϳ20 to ϳ45 s (n ϭ 9). This delay did not arise from slow formation of the ⌬, as the ⌬ forms within seconds (Fig. 3C). While it takes ϳ25-30 s for the ⌬pH to become fully established (Fig. 3C), it was shown earlier (20) and here (Fig. 3B) that the ⌬pH is not necessary for transport. Therefore, the slow formation of the ⌬pH does not explain the lag phase. Consequently, we reasoned that the lag phase could be explained by a ⌬-dependent conformational  MARCH 30, 2012 • VOLUME 287 • NUMBER 14 change or oligomerization process that is required for the cargo to migrate away from the TatBC binding site. We tested this hypothesis by preincubating the IMVs with NADH and then adding the cargo after a short delay (Fig. 4). The lag phase largely disappeared when the cargo was added 200 s after NADH (Fig. 4C). Under these conditions, slower kinetics ( ϭ ϳ70 -90 s) were observed.

Pre-SufI Is Released from TatBC Complexes when Membranes Are
Energized in the Absence of TatA-As a control experiment for the NADH-dependent loss of FRET (Fig. 3A), we repeated these experiments in the absence of TatA. Similar levels of TatB and TatC cherry were recovered in IMVs in the presence and absence of TatA (supplemental Fig. S11). Since TatA is required for transport, our expectation was that the cargo would remain bound to the TatBC complex, and hence, the FRET signal would remain constant. Experiments similar to those in Fig. 3A were performed, where pre-SufI was allowed to bind to IMVs, and NADH was added at time zero. FRET imme- Reactions contained pre-SufI(T96C) Alexa532 (20 nM) and IMVs (A 280 ϭ 2) with TatAB cherry C (blue) or TatABC cherry (red). NADH (4 mM) was added at t ϭ 0 s. B, sensitivity of NADH-dependent fluorescence changes to components of the PMF. The ⌬pH and ⌬ were reduced with nigericin (5 M, blue) and valinomycin (5 M, green), respectively. The control trace (red) contains no ionophores. Conditions as in A with TatABC cherry IMVs (n ϭ 2). C, gradients across membranes of TatABC cherry IMVs (A 280 ϭ 2). The presence of ⌬ (blue) and ⌬pH (red) gradients were determined using 100 nM oxonol VI (EX ϭ 610 nm, EM ϭ 645 nm) and 2.5 M quinacrine (EX ϭ 420 nm, EM ϭ 510 nm), respectively, as described previously (15). NADH (4 mM) was added at t ϭ 50 s. diately decreased ( ϭ 71 s) when the membranes were energized in the absence of TatA (Fig. 5A). This differed from results obtained with IMVs containing TatA (Figs. 3A and 4A), which exhibited a lag-phase and rapid loss of FRET after the lag phase ( ϭ ϳ30 s). Since the Tat machinery is incapable of transport without TatA (or its paralog, TatE) (5), the observed loss of FRET is not due to cargo translocation across the membrane, as confirmed in Fig. 5B. Rather, likely possibilities are that the loss of FRET arises either from a conformational change that results in the mature domain of the bound precursor moving away from the mCherry fluorophore, or from the precursor dissociating from the TatBC complex entirely. The latter appears to be the case since TatBC IMVs pelleted after energization retain ϳ53% less bound pre-SufI than unenergized IMVs (Fig. 5B). Multiple groups have suggested the possibility that the signal peptide can penetrate fairly deeply into the membrane (26,29,31,39). The energy-dependent dissociation of pre-SufI from the receptor complex in the absence of TatA suggests the possibility that the signal sequence may not have penetrated as deeply under these conditions.

DISCUSSION
Tat transport requires both binding of a pre-protein to the Tat proteins as well as movement of the cargo across the membrane. Presumably, these two seemingly conflicting processes are coupled, i.e. the binding reaction gates the transport event. Critical to understanding this process is the development and analysis of assays that can identify and discern the conversion between substeps of transport. To this end, we developed a FRET assay that reports the initial cargo binding step. Our major conclusions are: 1) transport occurs on the minute timescale with a highly purified membrane fraction (Fig. 1D); 2) the cargo binding interaction occurs with an apparent K D Ϸ 7-23 nM (Fig. 2D) and a competitor-induced k off Ϸ 0.042 s Ϫ1 (ϭ1/24 s) (Fig. 2B); 3) the cargo mature domain appears to be nearer to the C terminus of TatC than the C terminus of TatB (Fig. 2C); 4) migration of the cargo from its initial binding site requires a ⌬ (Fig. 3B); 5) a delay in cargo migration from the initial binding site occurs after membrane energization (Figs. 3 and 4), indicating that the ⌬ does not directly promote cargo migration; and 6) TatA increases the affinity of the TatBC receptor complex for the precursor in the presence of a membrane potential (Fig. 5). The implications of these results are now discussed.
As far as we are aware, the binding experiments reported here provide the first estimate of the binding affinity of a signal peptide for functional E. coli Tat translocons. The estimated K D of 7-23 nM indicates a fairly strong, highly specific interaction, as is reasonably expected for a selective process. From this measured K D and the competitor-induced k off Ϸ 0.042 s Ϫ1 (Fig.  2B), the apparent k on is 10 6 -10 7 M Ϫ1 s Ϫ1 . This second order rate constant is of the magnitude expected for enzymatic interactions, where orientation of the reactants is clearly important, and is complicated by the fact that it likely reflects both binding to the membrane lipids (15) as well as lateral diffusion to the Tat translocon. Whereas the precursor can apparently exchange between lipid-and translocon-bound forms on the tens of seconds timescale, release from the membrane surface is significantly slower (15).
It is unclear whether the signal peptide binding interaction was broken at the point at which the FRET interaction was lost under transport conditions, or whether there is sufficient flexibility between the mature domain and signal peptide such that large movements of the mature domain away from the C termini of TatB and TatC can occur without dissociation of the signal peptide. The latter is consistent with the finding that the thylakoid Tat machinery can transport cargos covalently linked to TatC via their signal peptide (27). In contrast, the NADHdependent release of the cargo from membranes containing  (20 nM) and IMVs with TatBC cherry (red) or TatBC (blue) (A 280 ϭ 2). The TatBC cherry data were fit with a single exponential plus a linear baseline drift, yielding ϭ 71 s (the baseline drift accounts for 13% of the total fluorescent change). B, membrane binding and cargo transport in the presence and absence of TatA. Pre-SufI(T96C) Alexa532 (20 nM) was allowed to bind to membranes (TatBC or TatABC IMVs, A 280 ϭ 2) for 10 min at 37°C, respectively. After a 10 min binding period, control samples were then centrifuged (16,000 ϫ g, 45 min, room temperature), and the pellet fractions were analyzed by SDS-PAGE (lanes 1-2). Resuspended membranes were energized with 4 mM NADH for 20 min at 37°C (lanes 3-6). Proteinase K (PK, 0.7 mg/ml) was added to half the reactions and all reactions were immediately centrifuged (16,000 ϫ g, 45 min, room temperature). The resultant membrane pellets were resuspended in translocation buffer containing 68 mM PMSF and resolved via SDS-PAGE. The amounts of recovered precursor (p, blue) and mature (m, red) proteins are shown above the sample gel, as indicated (n ϭ 3). Less pre-SufI was recovered with membranes lacking TatA (compare lanes 3 and 4), indicating that the TatBC affinity for pre-SufI is lower without TatA in the presence of a PMF. Matured protein was protease protected (compare lanes 4 and 6), indicating transport. The average amount of protein initially recovered with TatABC membranes (lane 2) was set to 100%.
TatBC complexes but no TatA (Fig. 5) indicates a weaker membrane binding affinity in the presence of a ⌬ under these conditions, and thus clear dissociation of the signal peptide from the TatBC receptor complex under these conditions.
The analysis of the data in Fig. 2D indicates that the functional TatBC receptor concentration is Յϳ20 nM (supplemental materials). Thus, the functional binding site concentration is significantly lower than the number of TatC molecules (322 Ϯ 88 nM) calculated based on the mCherry concentration. Note that TatABC cherry IMVs contained only full-length TatC cherry (supplemental Fig. S11), indicating that the mCherry fluorescence signal was not contaminated by degradation products of the fusion protein. One possibility is that most TatC molecules are inaccessible or non-functional, possibly a consequence of overexpression. Tarry et al. (32) found that cargos bind to only a few available binding sites in TatBC oligomers, which they estimated to be heptamers. According to our data, active heptamers could account for ϳ140 nM ( ϭ7 ϫ 20 nM) of the TatC molecules present. The remaining ϳ180 nM of the TatC molecules present may then be inactive or inaccessible. However, a recent report with covalent TatC dimers suggests that TatBC complexes contain an even number of TatC molecules (40). Thus, the oligomerization state of active TatBC complexes remains uncertain.
For all the pre-SufI single cysteine mutants tested, FRET signals were higher with TatC cherry than with TatB cherry . This is seemingly at odds with a recent crosslinking study, which suggested that pre-SufI binds more closely to TatB than to TatC (28). However, the observed FRET signal arises from the proximity of the pre-SufI mature domain to the mCherry protein, which was at the C termini of the Tat proteins. Our data do not exclude the possibility that the dye molecules on the pre-SufI mutants are closer to the amphipathic region of TatB than to the TatB C terminus. Consequently, the models of Maurer et al. (28) are entirely consistent with our FRET data.
A model that has emerged for the E. coli Tat system and the pre-SufI cargo is that the cargo first binds to the membrane lipids. From there, the cargo laterally diffuses to the TatBC complex. Translocation across the membrane, presumably with the assistance of TatA oligomers, requires a ⌬ (15). From studies on the thylakoid Tat system, the oligomerization of a TatBC complex with a TatA complex requires a PMF (41). It seems safe to deduce, then, that such an oligomerization process is responsible for the lag phase observed in Fig. 3A. Loss of FRET is fairly rapid after this lag phase, possibly indicating that migration of the cargo mature domain from TatBC is fairly quick after TatABC oligomerization is complete. Fig. 4C indicates that migration after binding occurs essentially immediately if the cargo is added ϳ200 s after ⌬ generation has been initiated, possibly indicating that TatABC oligomerization occurred before the cargo bound to TatBC. The cargo migration away from TatBC is ϳ2.6-fold slower under these conditions, likely due to the fact that the detectable ⌬ had already collapsed ( Fig. 3C; see also Ref. 15) when the cargo was added. Receptor binding and migration of the mature domain away from the TatBC complex occurs concurrently in Fig. 4C. The absence of a detectable increase in FRET at the beginning part of the kinetics in Fig. 4C indicates a rapid binding step ( Յ ϳ3 s), which includes both lipid binding and diffusion to the Tat translocon. Our data are therefore consistent with the hypothesis that TatBC and TatA oligomerization occurs in the presence of a ⌬, with or without the cargo. Further, the ⌬ makes the translocation system receptive to cargo movement from the receptor binding site (e.g. by a conformational change, or the oligomerization itself), but does not appear to directly drive it energetically. We expect that under in vivo conditions where a large ⌬ is consistently maintained, all three Tat components can and do oligomerize and cargo transport occurs on the subminute timescale.
According to the oligomerization model discussed in the previous paragraph, TatA and TatBC form separate oligomers in the absence of a cargo and a PMF (8,11,40,42). Thus, our expectation was that the stability of the TatBC-precursor complex should be unaffected by the absence of TatA in the presence or absence of a ⌬. We were therefore surprised that pre-SufI was released from IMVs upon addition of NADH in the absence of TatA (Fig. 5B). We interpret these data to imply that the receptor complex has a weaker affinity for pre-SufI in the absence of TatA when a ⌬ is present. This appears inconsistent with an oligomerization model in which TatA is not part of the receptor complex. However, an alternative, revised model is that at least some TatA is part of the receptor complex, consistent with earlier results (8,42). This picture is also consistent with a recent study, which reports some cross-linking of pre-SufI to TatA in the absence of a PMF (43). In light of these data suggesting that TatA is part of the TatBC receptor complex, it is not at all surprising that the absence of TatA affects the interaction between TatBC and pre-SufI. According to our FRET assay, a weaker receptor complex binding affinity in the absence of TatA is only apparent in the presence of a PMF. TatA is necessary to keep pre-SufI bound to the TatBC complex in the presence of a PMF, eventually leading to transport (compare Figs. 4A and 5A).
In summary, we have isolated membranes exhibiting rapid Tat translocation activity. For TatABC cherry IMVs, the overall translocation process is characterized by ϭ ϳ80 s (Fig. 1D). When a ⌬ is generated across these IMV membranes with bound precursor protein, migration away from the TatBC binding site occurs with ϭ ϳ30 s after a significant (20 -45 s) lag phase (Fig. 4A). Migration away from the TatBC binding site occurs more slowly, but with no lag phase, if the precursor is added to previously energized membranes ( ϭ ϳ90 s, Fig. 4C), suggesting that the effect of the ⌬ slowly wears off or that transport in the presence of a weak, undetectable ⌬ (15) occurs more slowly. Binding to the membrane surface and diffusion to the TatBC complex occurs rapidly ( Յ ϳ3 s). Therefore, the migration process in which the cargo moves away from the TatBC complex requires a significant fraction (ϳ40%) of the total transport time. It is unclear whether the migration step itself is rate-limiting or whether an upstream kinetic step limits the migration rate. The gel-based transport assay has poor time resolution, and, since it is not a real-time assay, it is not clear whether any transport substeps can occur after quenching the reaction. Consequently, it remains unclear whether the cargo proceeds directly across the membrane bilayer upon leaving the TatBC complex or whether there is an additional kinetically significant intermediate, e.g. cargo within the TatA pore. Further experiments are needed to clarify these downstream events.