|Home | About | Journals | Submit | Contact Us | Français|
Signal peptides target protein cargos for secretion from the bacterial cytoplasm. These signal peptides contain a tri-partite structure consisting of a central hydrophobic domain (h-domain), and two flanking polar domains. Using a recently developed in vitro transport assay, we report here that a central h-domain position (C17) of the twin arginine translocation (Tat) substrate pre-SufI is especially sensitive to amino acid hydrophobicity. The C17I mutant is transported more efficiently than wildtype, whereas charged substitutions completely block transport. Transport efficiency is well-correlated with Tat translocon binding efficiency. The precursor protein also binds to non-Tat components of the membrane, presumably to the lipids. This lipid-bound precursor can be chased through the Tat translocons under conditions of high proton motive force. Thus, the non-Tat bound form of the precursor is a functional intermediate in the transport cycle. This intermediate appears to directly equilibrate with the translocon-bound form of the precursor.
There are two major protein transport systems in Escherichia coli that translocate proteins from the cytoplasm to the periplasm. The Sec machinery exports the majority of secretory proteins. This transport system translocates unfolded polypeptides in a linear fashion from N- to C-terminus. A minimal system consists of the SecYEG complex, which comprises the transmembrane pore, SecA, an ATPase that guides the signal peptide to and through the SecYEG translocon, and SecB, a chaperone that maintains the precursor in a transport competent configuration (Driessen & Nouwen, 2008, Rusch & Kendall, 2007). In contrast, the twin-arginine translocation (Tat) machinery exports fully-folded and assembled proteins. A minimal system consists of the TatA, TatB and TatC membrane proteins (Sargent, 2007, Natale et al., 2008). A dominant model is that TatBC oligomers act as receptors for the precursor protein, and TatA oligomers form ring-like structures, which provide the pore for passage of the folded cargo proteins (De Leeuw et al., 2001, Gohlke et al., 2005, Dabney-Smith et al., 2006). In E. coli, approximately two thirds of Tat substrates contain metalloprosthetic groups, which are inserted into the proteins in the cytoplasm (Berks et al., 2005, Matos et al., 2008, Papish et al., 2003). Both Sec and Tat transport systems facilitate protein movement across the phospholipid membrane barrier without causing the collapse of transmembrane ion gradients, which are necessary for numerous physiological functions. In fact, a proton motive force (PMF) is the only energy source required for Tat transport. In E. coli, only one component of the PMF, the electric field gradient (Δψ), appears essential for Tat transport – a ΔpH requirement has not been detected (Bageshwar & Musser, 2007). In contrast, ATP is essential for Sec-dependent transport. However, Sec transport efficiency is significantly enhanced by a PMF (Natale et al., 2008, Liang et al., 2009).
A typical signal peptide for both the Sec and Tat systems is comprised of three distinct domains – the n-domain, the h-domain and the c-domain. The n-domain is located at the extreme N-terminal end of the signal peptide, and is usually hydrophilic and often contains positively charged residues. The h-domain follows the n-domain, and is predominantly comprised of hydrophobic residues. The h-domain of Tat signal peptides are somewhat less hydrophobic and generally longer than those found in Sec signal peptides (Berks et al., 2003, Bendtsen et al., 2005, Shanmugham et al., 2006). Numerous studies indicate that hydrophobic residues within the h-domain are important for promoting binding and transport (Cline & Mori, 2001, Gérard & Cline, 2006, Gérard & Cline, 2007, Stanley et al., 2000, Li et al., 2006, Chaddock et al., 1995). The polar c-domain is located at the C-terminal end of the signal peptide. Signal peptidase cleaves the signal peptide from the precursor protein after transport releasing the mature protein (Natale et al., 2008). The major difference between Sec and Tat signal peptides is the presence of a twin-arginine-containing consensus motif (which in bacteria is (S/T)-R-R-x-F-L-K) at the interface between the n- and h-domains of Tat presequences (Berks, 1996). In this consensus motif, the two arginines are almost invariant and virtually essential for efficient Tat transport, though natural exceptions occur in which lysine occurs in place of the first arginine (Molik et al., 2001, Hinsley et al., 2001).
The precise manner in which signal peptides promote cargo transport is not fully understood. As mentioned earlier, SecA and SecB are general and necessary chaperones for secretory proteins targeted to the Sec pathway. No general and necessary chaperones are known for the Tat system, at least not any displaying the broad specificity and essentiality as SecA and SecB within the Sec system. While SlyD and DnaK bind to multiple Tat precursor proteins, and thus could be considered ‘general chaperones’, they are not required for all precursor proteins (Oresnik et al., 2001, Bruser et al., 2003, Graubner et al., 2007). Further, the deletion of SlyD has no effect in vivo (Graubner et al., 2007). Efficient in vitro transport of pre-SufI does not require any chaperones (Bageshwar & Musser, 2007, Holzapfel et al., 2009), suggesting that not all Tat substrates require chaperones. However, specific signal peptide binding chaperones, do exist, e.g., DmsD for DmsA (Oresnik et al., 2001), NapD for NapA (Maillard et al., 2007), HybE for HybO (Dubini & Sargent, 2003), and TorD for TorA (Jack et al., 2004). In some cases, these chaperones appear necessary for a ‘proofreading’ function (Maillard et al., 2007, Jack et al., 2004). The proposed role of this ‘proof-reading’ function is to insure that only completely folded and assembled proteins are allowed to access the Tat export system (Berks et al., 2005, Bendtsen et al., 2005).
In the absence of a chaperone or after a chaperone has dissociated, signal peptide interactions with the membrane lipids may play a significant role in the translocation process. Based on the finding that thylakoid Tat precursors and E. coli Tat presequences bind to pure lipid membranes, multiple investigators have postulated that signal peptide interactions with the membrane lipids may precede the interactions with translocon components, i.e., the precursor can two-dimensionally search for translocons from a lipid-bound state (Musser & Theg, 2000, Hou et al., 2006, Shanmugham et al., 2006). Consistent with this hypothesis, Tat substrates bind to Tat-deficient bacterial membranes (Bolhuis et al., 2001, Sargent et al., 1999, Brüser et al., 2003). Protease treatment of membrane-bound Tat substrates yields fragments similar in size to the signal sequences, consistent with the hypothesis that an early step of the transport process is that the signal sequence becomes embedded in the membrane lipids and thus protected from proteases (Bruser et al., 2003, Hou et al., 2006). In a review on the membrane bilayer interactions of Tat signal sequences, it was suggested that this lipid interaction step could be crucial for the proof-reading function of the Tat pathway (Bruser & Sanders, 2003).
In our efforts to construct a cysteine-free signal peptide for pre-SufI, we discovered that the hydrophobicity of the residue at position C17, in the middle of the h-domain, strongly influences the membrane binding and transport promoting ability of this signal peptide. Pre-SufI binds to TatABCDE-deficient membranes, indicating the presence of non-Tat binding sites. At a high membrane potential, precursor bound to the non-Tat binding sites of TatABC-containing membranes was transported, suggesting that this non-Tat membrane-bound state is a functional intermediate in the transport process.
As we reported earlier (Bageshwar & Musser, 2007), inverted membrane vesicles (IMVs) constructed from the E. coli strains MC4100 and JM109 are both suitable for measuring in vitro Tat transport activity. However, IMVs from the MC4100 strain yield lower transport efficiencies, on average, and the results are more variable. Thus, we used IMVs from the JM109 strain to determine transport efficiencies. However, there is no currently available TatABCDE deletion strain generated from JM109. Thus, to compare the membrane binding activity of precursors in the presence and absence of Tat translocons, we used IMVs from the MC4100 and MC4100ΔTatABCDE strains, respectively. Both our JM109 and MC4100 strains contained a TatABC overexpression plasmid (pTatABC; see Experimental Procedures). IMVs constructed from the JM109 (pTatABC), MC4100 (pTatABC), and MC4100ΔTatABCDE strains are denoted pTat*, pTat, and ΔTat, respectively. For most of the experiments reported here, we used the natural Tat substrate pre-SufI.
Membrane binding activity was determined by incubating precursors with IMVs for 10 min at 37°C and then centrifuging the samples. The IMV pellets were washed, and then assayed by SDS-PAGE and Western blotting for the presence of precursor protein. TatABC-enriched IMVs from MC4100 and JM109 contained comparable amounts of TatA, TatB and TatC (Fig. S1A) and yielded similar ratios of inside-out vesicles (≥95%; Fig. S1B), in agreement with previous results (Bageshwar & Musser, 2007). These IMVs also exhibited similar binding capacities towards pre-SufI over a range of precursor concentrations (Fig. 1A). These data suggest that precursor interactions with IMVs from JM109 and MC4100 strains are similar, and therefore, that transport efficiency measurements with the JM109 strain can be reasonably interpreted in light of membrane binding interactions with MC4100 strains. Because of the different strains used for binding and transport, some differences (errors) are expected when comparing the absolute amounts of protein bound or transported. For this reason, we compare general trends instead of amounts. The membrane binding activity and transport efficiency of pre-SufI were not affected by up to a 16-fold molar excess of the Sec substrate proOmpA (Fig. 1B). Therefore the Tat pathways is highly selective for the pre-SufI presequence under the conditions described here. More importantly, the membrane binding activity shown towards ΔTat IMVs cannot be explained by the binding of pre-SufI to the Sec transport system.
The membrane binding activity and transport efficiency of pre-SufI depended upon the precursor concentration. Not surprisingly, TatABC-containing membranes had a higher precursor binding capacity than TatABC-deficient membranes. Assuming that the amount of non-Tat-bound precursor protein was identical for both types of membranes, the amount of precursor bound to the Tat transport machinery was estimated by subtracting the amount of precursor bound to Tat-deficient membranes from that bound to Tat-containing membranes. About twice as much wildtype pre-SufI binds to the Tat transport machinery as binds to non-translocon components (Fig. 1C). Thus, a significant fraction of pre-SufI does not bind directly to the Tat machinery. Interestingly, the amount of pre-SufI transported in the presence of NADH was almost indistinguishable from the amount of Tat translocon-bound precursor (Fig. 1C). These data suggest, at least for these conditions, that only pre-SufI bound to the Tat machinery can be chased into the IMV lumen by the addition of NADH. Both translocon binding and transport activity saturated at ~3–4 pmol pre-SufI. To maximize the total amount of bound precursor (total observable signal) and minimize the amount of unbound precursor that needs to be removed before analysis (as a percent of the total), 3.1 pmol (90 nM) pre-SufI was used for most of the subsequent experiments.
To investigate whether pre-SufI binds to a non-Tat proteinaceous receptor, IMVs were treated with trypsin to digest proteins that might act as receptors. This protease digestion treatment was effective since translocon binding activity was eliminated from Tat-containing membranes. Pre-SufI binding to ΔTat membranes was unaffected by trypsinization (Fig. 1D), suggesting that binding was mediated by a non-protein membrane component. For simplicity, we will refer to the pre-SufI that binds to ΔTat IMVs as lipid-bound precursor. We do recognize, however, that trypsinization of IMVs may not have eliminated a non-Tat proteinaceous receptor.
We next investigated the dependence of pre-SufI membrane binding activity and transport efficiency on pH, ionic strength, and urea concentration (Fig. 2). We first discuss the observed pH effects (Fig. 2A). The effect of buffer pH on the ability of pre-SufI to bind to Tat translocons was significantly different from its effect on pre-SufI interactions with the membrane lipids. The lipid interactions were diminished by increased pH. The data are consistent with the hypothesis that binding is inhibited by a single deprotonation step at an apparent pKa of ~7.5. In contrast, the Tat machinery interactions were optimal at pH ~8, consistent with protonation/deprotonation of the precursor and/or the Tat machinery on either side of this optimum. The transport efficiency also was optimal at pH ~8. This makes intuitive sense, considering that the precursor must bind to the Tat translocon before it can be transported. When transport was energetically driven by NADH addition, the amount of precursor transported was similar to the amount of translocon-bound precursor. This result is consistent with the hypothesis that NADH induces a single round of Tat turnover (Bageshwar & Musser, 2007). However, in the presence of ATP, a larger amount of precursor was transported than was found bound to Tat translocons, suggesting that additional pre-SufI must bind to Tat translocons during ATP-driven transport. This result is consistent with our previous conclusion that ATP maintains a stable, high magnitude PMF across the membrane, and thereby is capable of driving multiple turnover events (Bageshwar & Musser, 2007).
Variations in salt concentrations are typically used to identify the role of ionic interactions, an approach we used here. The precursor-lipid interactions were essentially unaffected by a range of KCl concentrations (0.05–1 M). In comparison, the amount of precursor bound to the Tat machinery decreased by ~35% over this KCl concentration range (Fig. 2B). Thus, both types of pre-SufI binding interactions are largely unaffected by salt concentration, suggesting that they are dominated by non-ionic interactions. In contrast, transport was largely blocked (by ~85%) by high KCl concentrations (Fig. 2B). One possibility is that ionic interactions at later stages of translocon assembly or downstream conformational changes are affected by high salt concentrations. Alternatively, PMF generation or maintenance could be affected. We did not pursue these issues further. However, we did test a mutant in which the double arginine motif in the signal peptide was mutated to KK (pre-SufI-KK). This mutant was shown previously to be transport incompetent (Stanley et al., 2000). As expected, pre-SufI-KK did not transport. Furthermore, pre-SufI-KK did not even bind to Tat translocons (Fig. 2C), consistent with previous reports demonstrating that precursor binding to the TatBC complex is inhibited by the KK mutation (Alami et al., 2003, Gérard & Cline, 2006, Cline & Mori, 2001). However, the lipid-binding activity of pre-SufI-KK was largely unaffected, consistent with membrane binding studies with bacterial and thylakoid precursor proteins (Hou et al., 2006, Bruser et al., 2003).
Addition of urea has been useful as a tool to investigate hydrophobic binding interactions. Urea concentrations up to 2.5 M had virtually no effect on the interactions of pre-SufI with the membrane lipid. In contrast, pre-SufI was completely dissociated from Tat translocons by 2 M urea. Not surprisingly, transport efficiency was also completely blocked by 2 M urea. However, transport efficiency was virtually unaffected by 1 M urea, whereas translocon binding activity was inhibited by ~75% (Fig. 2D). These data suggest that the pre-SufI-translocon binding interactions are weaker in the presence of 1 M urea than in its absence yet still strong enough to promote transport. In other words, binding still occurs and promotes transport, yet the binding affinity is insufficiently strong for the precursor protein to be recovered in a sedimentation assay. Urea causes solvation of hydrophobic groups, thus weakening the hydrophobic effect (Beck et al., 2007, Stumpe & Grubmüller, 2007, Stumpe & Grubmüller, 2008, Hua et al., 2008). Therefore, we speculate that urea interferes with translocon binding interactions by weakening the interaction between the h-domain of the signal peptide and a hydrophobic pocket on the Tat machinery.
We next tested whether the membrane binding and transport characteristics observed for pre-SufI were unique or could be generalized. The first construct that we tested was spSufI-GFP, which consists of the signal peptide from pre-SufI attached to GFP. In earlier work, we showed that spSufI-GFP does not transport in in vitro assays, and does not even compete with pre-SufI for transport (Bageshwar & Musser, 2007). The latter result suggested that spSufI-GFP does not even bind to the Tat translocon. We directly tested this hypothesis here. We found that spSufI-GFP does not bind at all to IMVs, neither to the membrane lipids nor to the Tat translocon (Fig. 3A). These results demonstrate that the signal peptide is not the sole determinant of binding. The model construct spTorA-GFP, which contains the signal peptide from pre-TorA, is commonly used as a model cargo (Bageshwar & Musser, 2007, Santini et al., 2001, Thomas et al., 2001, Barrett et al., 2003, DeLisa et al., 2004). The translocon binding activity of spTorA-GFP saturates at ~4 pmol (Fig. 3A), similar to that observed for pre-SufI (Fig. 1C). These data therefore suggest a similar translocon binding affinity for TorA-GFP and pre-SufI. Next, the binding characteristics of spTorA-GFP were probed using urea and KCl (Fig. 3B). The results can be summarized as follows. Whereas 2 M urea did not affect the interaction of pre-SufI with the lipids, it inhibited the lipid binding interaction of spTorA-GFP by ~60%. The lipid binding interaction of spTorA-GFP was enhanced by 1 M KCl (by ~30%), whereas for pre-SufI it had essentially no effect. Urea and KCl acted synergistically on pre-SufI, causing almost complete dissociation from the Tat machinery and the membrane lipids. However, their effects on spTorA-GFP were largely the same as that observed with urea alone. Increasing the length of the linker between the spTorA-GFP signal peptide from 4 residues to 12 residues (yielding spTorA12-GFP) had essentially no effect on lipid or translocon binding interactions. These data indicate that different Tat precursors have different affinities towards the membrane lipid and the Tat translocons, which are determined by different ionic and hydrophobic interactions. Importantly, however, both binding modes are still apparent and readily distinguishable Thylakoid Tat precursors are also completely dissociated from the membrane surface by a combination of urea and KCl, but urea or KCl alone have different effects on different precursors (Gérard & Cline, 2006, Gérard & Cline, 2007), as we observed here.
We next compared the transport efficiencies of spTorA-GFP and spTorA12-GFP under single- and multiple-turnover conditions (Fig. 3C). SpTorA-GFP behaved similar to pre-SufI in that ATP caused an increase in transport efficiency (~2-fold), compared to that observed with NADH. Further, the amount of spTorA-GFP transported with NADH (0.4±0.1 pmol) was comparable with that which bound to the Tat machinery (0.5±0.1 pmol). However, when IMVs were energized with ATP, a significantly higher transport yield was observed (0.8±0.2 pmol) (Fig. 3C). These data imply that additional spTorA-GFP must bind to Tat translocons during ATP driven transport. A similar result was observed for pre-SufI (Fig. 2A). Interestingly, spTorA12-GFP behaved differently than spTorA-GFP under ATP-driven transport conditions in that there was no enhancement in transport efficiency compared with NADH-driven transport. We do not have an explanation for this result.
In order to develop a simpler, non-Western blot-based binding and transport assay, and to better study the kinetics of Tat transport, we sought a fluorescent precursor that was efficiently transported. Since spTorA-GFP was transported less efficiently than pre-SufI, we sought a fluorescent version of pre-SufI, e.g., one that could be prepared by modification of cysteine residues through maleimide chemistry. Wildtype pre-SufI contains two cysteines, C17 and C295 (Fig. 4A). Since labeling experiments indicated that the C295 residue of pre-SufI was not highly reactive towards maleimides, a cysteine was added at the C-terminus to provide an additional labeling site, generating pre-SufI-CCC. The pre-SufI and pre-SufI-CCC proteins exhibited similar membrane binding and transport activities (Fig. 4B). Faced with the task of removing the cysteine in the signal peptide, we constructed the C17S mutant. The transport efficiency of the C17S mutant was ~85% less than that of wildtype pre-SufI. Therefore, we randomly mutageneized the C17S mutant using degenerate primers. Plasmids from 41 colonies were sequenced, yielding 10 new unique mutants. The C17W mutation was generated with specific primers. The membrane binding and transport properties of pre-SufI with 13 different residues at position 17 are summarized in Fig. 4C (raw data in Fig. S2). Every C17 mutant tested displayed a substantially reduced capacity to bind to the membrane lipids. All but three mutants (C17M, C17W and C17I) exhibited reduced translocon binding activity. All but one mutant (C17I) displayed reduced transport activity.
One possible explanation of the C17 mutant data is that the h-domain of the signal peptide strongly influences the partitioning of the pre-SufI signal peptide into the hydrophobic interior of the lipid bilayer and into a hydrophobic groove in the Tat translocon (signal peptide binding site). Such a partitioning is expected to generally follow the predictions of amino acid hydrophobicity according to one of the numerous proposed free energy scales. The transport efficiency of the various C17 mutants displayed a reasonably linear correlation with the free energy change of the mutation, according to cyclohexane-to-water partition coefficients (Fig. 4D, left panel). Strongly hydrophilic, charged substitutions (C17D, C17H and C17K mutations) led to complete inhibition of transport. Such substitutions are expected to adversely affect hydrophobic interactions. Aliphatic substitutions (C17M and C17I) retained good transport activity, consistent with being conservative substitutions. Translocon binding activity was moderately correlated with the cyclohexane-to-water hydrophobicity scale (Fig. 4D, middle panel). The transport activity of pre-SufI was moderately correlated with its translocon binding activity (Fig. 4D, right panel). Variations from linearity likely arise from steric issues. The moderately lower hydrophobicity of the central h-domain of the TorA signal peptide compared with that of pre-SufI (Fig. 4E) may explain, at least in part, the lower transport efficiency of spTorA-GFP (Bageshwar & Musser, 2007). Though highly hydrophobic h-domains can convert Tat signal peptides into Sec targeting domains (Cristóbal et al., 1999), the slight increase in h-domain hydrophobicity of the C17I pre-SufI mutant is apparently insufficient to inhibit Tat targeting.
Based on the data discussed in the previous section, the C17I mutant appeared to be the most useful for an efficient fluorescence-based transport assay: it is the most selective for translocon interactions over lipid interactions (~10:1), and it is the most efficiently transported (110±9% as compared to pre-SufI-CCC). Therefore, we further examined the properties of this mutant. From the membrane binding activity of pre-SufI-ICC measured over a range of concentrations (Fig. 5A), it is clear that the lipid binding activity of pre-SufI-ICC is very low. One potential complication that could arise with a precursor protein that binds to both lipids and translocons is that the amount bound to each is expected to depend on the lipid to translocon ratio. As shown in Fig. S3, this complicating issue disappears for the C17I mutant. We next examined the translocation activity of the C17I mutant over a range of concentrations. The amount of translocated precursor was well correlated with the amount of translocon-bound precursor (Fig. 5B), in agreement with earlier results (Figs. 1C and and4D4D).
The endogenous cysteine within the mature domain of pre-SufI-ICC was replaced with alanine (C295A), yielding a pre-SufI mutant (pre-SufI-IAC) with a single cysteine at the C-terminus, immediately downstream of 6xHis-tag. The transport efficiency of pre-SufI-IAC was indistinguishable from that of pre-SufI-ICC (Fig. 6A). The cysteine of pre-SufI-IAC was labeled with Atto565 maleimide, yielding pre-SufI-IACatto. Unlabeled and Atto565-labeled pre-SufI-IAC showed indistinguishable transport efficiencies and binding affinities towards membrane lipids and Tat translocons (Fig. 6B). The transport kinetics of fluorescent and non-fluorescent pre-SufI-IAC, as monitored by Western blotting and direct in-gel fluorescence imaging, were indistinguishable (Fig. 6C). Thus, pre-SufI-IACatto is an efficiently transported fluorescent Tat substrate that behaves identically to the unlabeled protein.
A major question that arises from the data presented thus far is whether lipid-bound precursors reside in a dead-end state and must dissociate from the membrane before they can possibly be translocated, or whether they can proceed from the lipid-bound state directly to the translocon-bound state for transport. The latter scenario was first suggested for the thylakoid Tat machinery based on the behavior of a transport incompetent precursor (Musser & Theg, 2000). When avidin was added to an import reaction containing biotinylated and unbiotinylated precursor, the biotinylated precursor was not translocated yet it still bound to thylakoid membranes and competitively inhibited transport of the unbiotinylated precursor. One explanation is that the lipid-bound state of the precursor protein is an intermediate in the transport process. If so, lipid-bound avidin-precursor complexes might be sufficient to inhibit transport of unbiotinylated precursor. What was unclear in these experiments, however, was whether the avidin-precursor complex could bind to the Tat machinery. We sought here to test this hypothesis.
The original basic principle behind the experiments with biotinylated precursor protein was to generate a translocation intermediate, i.e., a precursor undergoing transport stuck within the translocon. Addition of avidin to a biotinylated protein results in a very strong, non-covalent interaction between the biotin moiety and the avidin. The resultant precursor-avidin complex is much different in size and shape than the precursor alone. Such characteristics might cause the precursor to become stuck at an intermediate state of transport. Biotinylated pre-SufI-IAC (pre-SufI-IACbiotin) was generated by reaction of pre-SufI-IAC with biotin maleimide. Avidin blocked transport of pre-SufI-IACbiotin when the transport reaction was initiated with NADH, unless free biotin was pre-mixed with the avidin (Fig. 7A). These data support the hypothesis that avidin blocks transport by binding to the biotin moiety on the precursor protein. The pre-SufI-IACbiotin-avidin complex competitively inhibited the transport of unbiotinylated pre-SufI-IAC in the presence and absence of avidin (Fig. 7B). These results (Fig. 7A and 7B) are qualitatively identical to those reported earlier for the thylakoid Tat complex (Musser & Theg, 2000). Interestingly, the pre-SufI-IACbiotin-avidin complex binds almost as strongly to the Tat machinery as pre-SufI-IACbiotin alone, and it does not matter whether the precursor is incubated first with IMVs or first with avidin (Fig. 7C). Thus, the avidin moiety does not prevent pre-SufI-IACbiotin from interacting with the Tat machinery. These data therefore indicate that avidin blocks transport of pre-SufI-IACbiotin at some step after the initial translocon binding step. Somewhat surprisingly, avidin enhanced the ability of pre-SufI-IACbiotin to bind to the membrane lipids (Fig. 7C). This result can be explained by a non-specific interaction of avidin with the membrane lipids.
To probe the strength of the interaction between pre-SufI and the Tat translocon, a near saturating concentration (90 nM, 3.1 pmol) of both pre-SufI-IACbiotin and pre-SufI-IACatto were incubated with IMVs in the presence of avidin. However, we varied the order of addition of the two precursor proteins. If the translocon-bound form of the precursor was a strong interaction, implying a slow off-rate, pre-incubation of IMVs with pre-SufI-IACatto should lead to translocon-bound precursor that can be chased into the lumen of the IMVs after NADH addition even in the presence of precursor-avidin complexes. However, this was not seen (Fig. 7D, lane 9). Instead, the transport of fluorescent pre-SufI was strongly inhibited by precursor-avidin complexes, whether these complexes were added before, after, or at the same time as the fluorescent pre-SufI (Fig. 7D, lanes 7–9). These data suggest that a translocon-bound precursor molecule can relatively rapidly exchange with another precursor molecule, irrespective of whether either molecule is attached to an avidin molecule through a biotin linkage. Note that the precursor-avidin complexes are more inhibitory at a higher total precursor concentration (compare Fig. 7B and 7D), for unknown reasons.
We next sought to more directly investigate whether the lipid- and translocon-bound forms of pre-SufI could interconvert. In multiple ATP-driven transport experiments (Figs. 2A and and3C),3C), the transport yield was larger than the amount of precursor protein bound to Tat translocons. This extra protein had to come from the soluble (aqueous) phase, from the pool of lipid-bound precursor protein, or from both precursor reservoirs. To eliminate the soluble precursor protein, pre-SufI was pre-incubated with pTat* IMVs and the mixture was then centrifuged (Fig. 8). The recovered IMVs contained precursor protein strongly bound to the membrane since essentially no pre-SufI was recovered in the supernatant upon re-centrifugation (Fig. 8, lane 6). Urea dissociated ~3/4 of the membrane-bound pre-SufI (Fig. 8, compare lanes 7 and 5), consistent with the hypothesis that the majority of the precursor was bound to Tat translocons (compare with Fig. 3B). When NADH was added, ~1/3 of the initially recovered precursor was transported (Fig. 8, compare lanes 10 and 4). However, when ATP was added almost all of the initially recovered precursor was transported (Fig. 8, compare lanes 11 and 4). Since significantly more precursor was transported upon ATP addition than was initially bound to translocon components (Fig. 8, compare lanes 11 and 7), additional precursor must have been recruited to the translocons during the incubation period. Since there was essentially no soluble precursor present in the reaction (Fig. 8, lane 6), the additional precursor must have come from the lipid-bound precursor reservoir. Thus, these data suggest that the lipid- and translocon-bound forms of pre-SufI can directly interconvert, i.e., that the lipid-bound form of the precursor can laterally diffuse to the translocation machinery, bind to it, and undergo transport.
The major finding of this study is two distinct membrane binding modes for bacterial Tat precursor proteins. The first of these binding interactions is the expected binding interaction with the TatABC proteins. The second membrane binding interaction is not mediated by any of the Tat proteins, since this interaction was detectable in a ΔTatABCDE strain (Fig. 1C). Due to the known affinity of Tat precursors and signal peptides for pure lipid membranes (Musser & Theg, 2000, Shanmugham et al., 2006, Hou et al., 2006), the retention of the non-Tat binding interaction in trypsinized membranes (Fig. 1D), and the dramatic effects on binding of single site mutations in the signal peptide (Fig. 4C), we ascribe the second binding mode to a direct interaction between the membrane lipids and the signal peptide. Whether this interaction is an adsorption to the membrane surface, or a partial or full penetration of the membrane bilayer by the signal peptide requires further investigation. Though it has been suggested that a membrane ATPase promotes the integration of the signal sequence into the membrane bilayer (Brüser et al., 2003), we find that the lipid binding interaction is ATP-independent and occurs after shaving the membrane with protease (Fig. 1D), suggesting that an ATPase is not involved. The lipid-bound precursor can be chased into the IMV lumen in the presence of ATP (Fig. 8), indicating that the lipid-bound precursor remains translocation competent. We consider it unlikely that an ATPase is involved in the transfer of the precursor from the lipid to the translocon, and surmise that this ATP effect results simply from a longer-lived, higher magnitude PMF (Bageshwar & Musser, 2007). The lipid-bound precursor state we identified corresponds well with the Tm-1 state identified in thylakoid binding studies (Hou et al., 2006). A model of the Tat transport cycle as discussed in the Introduction and consistent with the data presented here is shown in Fig. 9.
The two distinct membrane binding modes for Tat precursor proteins are distinguishable by their differential susceptibility to urea and KCl. Since urea reduces the binding affinity of precursors for Tat translocons, the translocon binding interactions appear to be mediated, at least in part, by hydrophobic interactions. For the bacterial Tat precursors examined, the data are consistent with the hypothesis that the h-domains of the signal peptides bind in a hydrophobic groove of the TatBC complex (Holzapfel et al., 2007, McDevitt et al., 2006). Mutation of the double arginine motif to KK led to the complete loss of translocon binding (Fig. 2C). Since this is a conservative mutation, the translocon binding site is sufficiently selective that it can distinguish between lysine and arginine, perhaps by recognizing the guanidinium group on the latter. However, the charge of the side chain does not dominate the binding affinity (although it may certainly contribute), since high salt concentrations have little affect on the translocon binding interaction (Fig. 3B). Interestingly, both KK and AA versions of pre-SufI co-immunoprecipitated with the TatBC complex when expressed under in vivo conditions (McDevitt et al., 2006). In light of our data, these results can be explained by binding of the pre-SufI mutants to lipid within the detergent micelles containing the purified TatBC complexes. The strong binding of the translocation incompetent KK mutant of the reduced TorA-PhoA protein to Tat-containing IMVs (Panahandeh et al., 2008) can be explained, at least in part, by binding to the membrane lipids.
The finding that the bacterial Tat precursor pre-SufI remains strongly bound to the IMV surface once bound (Fig. 8) agrees with previous results on the thylakoid Tat system, which also indicated strong membrane binding (Musser & Theg, 2000, Hou et al., 2006). However, we found here that the lipid- and translocon-bound forms of the precursor appear readily exchangeable (Fig. 7D). This conclusion may help explain why it has been so difficult to generate a translocation intermediate for the Tat machinery – the precursor simply does not bind tightly to the translocon. The tight membrane binding interaction must therefore be provided by the non-Tat binding interaction. Our previous result that a precursor could compete for transport despite being added ~5 min after initiation of transport with NADH (Bageshwar & Musser, 2007) is difficult to comprehend in the context of a strong, stable interaction of the signal sequence with the Tat translocon. However, this result is much easier to comprehend now, knowing that the membrane-bound forms of the precursor rapidly interconvert. Precursor proteins could bind and dissociate from the translocon on a relatively rapid timescale, and a translocon interaction may only occasionally lead to cargo transport.
An important question is whether the lipid-bound form of the precursor protein is a necessary intermediate within the transport cycle. Here we have showed that this state exists under in vitro conditions and that it represents a pool of transport competent precursor protein. In vivo, one role of precursor-specific chaperone proteins, such as DmsD, NapD, HybE, and TorD (Maillard et al., 2007, Oresnik et al., 2001, Dubini & Sargent, 2003, Jack et al., 2004) may be to minimize lipid binding reactions. It is unknown how deeply the signal sequence penetrates into the lipid bilayer. If the signal sequence binds to the translocon from a position deeply embedded within the lipid bilayer, it is difficult to envision how a signal sequence could also find the translocon directly from the aqueous phase. Thus, if the signal sequence embeds deeply within the lipid bilayer, the lipid-bound state would likely be a necessary intermediate for transport. On the other hand, the signal sequence may not penetrate deeply within the lipid bilayer and may have easy access to the translocon binding site from the cytoplasm. However, the conclusion that the lipid- and translocon-bound forms of the precursor easily interconvert suggests that the lipid-bound state will be accessed during the normal transport process even if the soluble precursor binds directly to the translocon. Future work is needed to address these important issues. However, it is clear at this juncture that the lipid-bound form of Tat precursor proteins are important functional intermediates that should be acknowledged in models of the transport cycle.
E. coli strains MC4100, MC4100ΔTatABCDE, JM109 and BL21(λDE3) have been described earlier (Casadaban & Cohen, 1979, Yanisch-Perron et al., 1985, Studier et al., 1990, Wexler et al., 2000). Overexpression cultures were grown in Luria-Bertani (LB) medium at 37°C supplemented with the appropriate antibiotics (Sambrook & Russell, 2001). Plasmids pET-SufI, pSufI-GFP and pTorA-GFP have been described earlier (Yahr & Wickner, 2001, Bageshwar & Musser, 2007). The protein constructs used here and the primers used to generate the necessary expression plasmids are provided in Tables S1 and S2. Plasmid pTorA12-GFP, which encodes 12 amino acid residues of the mature TorA protein between the TorA signal peptide and the GFP domain, was generated by inverse PCR using Pfu Turbo DNA polymerase (Stratagene), TorA12-F and TorA12-R as the primers, and pTorA-GFP as the template. All single amino acid changes were generated by QuikChange site-directed mutagenesis (Stratagene). Plasmid pSufI-CCC was constructed using primers 3′-Cys-F and 3′-Cys-R and the pET-SufI template. This plasmid encodes for pre-SufI-CCC, which contains an additional cysteine at the C-terminus of pre-SufI. Plasmid pSufI-SCC, which was generated using primers SufIC17S-F and SufIC17S-R and the pSufI-CCC template, encodes for pre-SufI-SCC, which has a C17S mutation in the signal peptide of pre-SufI. Position C17 in pre-SufI was randomly mutagenized by randomizing the codon (equal base probability at each position) using primers SufIC17N-F and SufIC17N-R and the pSufI-SCC template. A total of 41 clones were sequenced to obtain 10 mutants. The resultant plasmids and encoded proteins are designated pSufI-XCC and pre-SufI-XCC, respectively, where the identity of the C17 mutation in the pre-SufI signal peptide is identified by the X. Plasmid pSufI-WCC was generated using primers SufIC17W-F and SufIC17W-R and the pSufI-SCC template. Plasmid pSufI-IAC, which encodes the C295A mutation in the mature domain of SufI, was generated using primers SufIC295A-F and SufIC295A-R and the pSufI-ICC template. Plasmid pSufI-KK-CCC, which encodes KK instead of the double-arginine motif, was generated using primers SufI-KK-F and SufI-KK-R (both primers are flanked by NdeI restriction endonuclease sites) and the pSufI-CCC template by inverse PCR using Pfu Turbo DNA polymerase. The PCR product was cleaved with NdeI and self-ligated to generate pSufI-KK-CCC, which encodes pre-SufI-KK-CCC. The coding region of all plasmid constructs was confirmed by DNA sequencing.
The precursor proteins spTorA-GFP, spTorA12-GFP and spSufI-GFP were expressed from plasmids pTorA-GFP, pTorA12-GFP and pSufI-GFP, respectively, purified under denaturing conditions, and folded by removing urea by dialysis (Bageshwar & Musser, 2007). Pre-SufI was expressed and purified by Ni-NTA chromatography (Bageshwar & Musser, 2007). The pre-SufI mutants were purified in identical fashion. Plasmid pTatABC (Yahr & Wickner, 2001) was used for overexpression of TatA, TatB and TatC under the control of the arabinose promoter (Bageshwar & Musser, 2007). SecB and proOmpA-HisC were purified as described (Liang et al., 2009).
The protein concentrations of all Tat substrates were quantified by SDS-PAGE using bovine serum albumin (BSA) as the standard. Spot intensities after Coomassie Blue staining were determined with a PhosphorImager (model FX; Bio-Rad Laboratories). SecB and proOmpA concentrations were determined by the BCA method (Pierce) using BSA as a standard.
IMVs were obtained from E. coli strains MC4100, MC4100ΔTatABCDE and JM109, as described (Bageshwar & Musser, 2007). The MC4100 and JM109 strains included the plasmid pTatABC for overexpression of the Tat machinery, which was induced by 0.7% arabinose. The in vitro Tat transport assay was performed as described (Bageshwar & Musser, 2007), with minor modifications. These modifications were a slightly increased pH (pH 8), a decreased IMV concentration (A280 = 2) and transport was usually initiated by NADH addition instead of by IMV addition. The total protein in IMV preparations was quantified as the A280 in 2% SDS. Typical IMV stock solutions had an A280 ≈ 50–60. The inside-out percentage of the IMV preparation was identified as described earlier (Fig. S1) (Bageshwar & Musser, 2007). Most standard transport assays were performed with 90 nM pre-SufI (3.1 pmol in 35 μL). This pre-SufI concentration yields lower overall transport efficiencies than we reported earlier, but the Western blots are cleaner. Variations on these conditions are indicated in the text or figure captions. Precursor transport was initiated by 4 mM NADH addition, unless otherwise noted. When 4 mM ATP was used, a regenerating system was included (7.1 mM phosphocreatine, 0.29 mg/mL creatine kinase).
For precursor-membrane binding studies, IMVs (typically A280 = 2) and precursor (typically 90 nM) were incubated in a 35 μL reaction volume of High-BSA translocation buffer, pH 8.0 (HB-TB). HB-TB contains a 10-fold higher concentration of BSA (570 μg/mL) than translocation buffer (TB; 5 mM MgCl2, 50 mM KCl, 200 mM sucrose, 57 μg/mL BSA, 25 mM MOPS, 25 mM MES, pH 8.0). The high BSA concentration was used to help prevent non-specific binding. Protein LoBind microfuge tubes (1.5 mL, Eppendorf) were effective for eliminating non-specific binding to the walls of the reaction vessel. The selectivity of the binding reaction conditions is illustrated by the complete absence of precursor in the minus IMV control lanes of Figs. 2C, ,3B,3B, and and7C.7C. After a 10 min incubation at 37°C, the binding reactions were centrifuged at 16,200 g at 4°C for 30 min to sediment the IMVs. Supernatants were aspirated away. Pellets were washed with 200 μL HB-TB (without resuspension) and re-centrifuged under the same conditions to remove any residual supernatant and any residual precursor bound non-specifically to the reaction vessel. The wash supernatant was removed by aspiration. When monitoring the effect of pH, urea, KCl, and urea + KCl on the binding efficiency of precursor proteins to IMVs, pellets of sedimented IMVs were also washed with 200 μL HB-TB. The pellets were then suspended in 2X Gel Buffer (4% SDS, 10% glycerol, 0.04% bromophenol blue, 0.4% β-ME, 10 M urea, and 200 mM Tris, pH 6.8), and incubated in a boiling water bath for 10 min. Samples were pulse centrifuged at 16,000 g, and then were resolved by 8% SDS-PAGE for pre-SufI and 10% SDS-PAGE for spSufI-GFP, spTorA-GFP and spTorA12-GFP with known standards. Gels were electroblotted onto PVDF membranes and immunoblotted using appropriate antibodies, as described earlier (Bageshwar & Musser, 2007). To rigorously remove soluble pre-SufI (Fig. 8), IMVs were preincubated with the precursor protein as described earlier and sedimented three times through a 1.0 mL 0.7 M sucrose cushion on top of a 10 μL 2.2 M sucrose cushion (both cushions were in TB buffer). IMVs were collected from the cushion interface.
IMVs were treated with trypsin to digest proteins accessible to the aqueous environment as described earlier (Lorence et al., 1988), with minor modifications. IMVs were diluted with TB buffer (pH 8.0) devoid of BSA to A280 = 10 (1 mL final volume), and treated with 0.5 mg/mL trypsin (Sigma) for 200 min at room temperature. Digestions were quenched with 50 μL of 100 mg/mL egg white trypsin inhibitor (Sigma). The trypsinized IMVs were sedimented and washed once with HB-TB, as described in the previous section.
We tested six different hydrophobicity scales (Pace, 1995, Kyte & Doolittle, 1982, Wimley & White, 1996) to fit the transport efficiency data in Fig. 4D. The cyclohexane-to-water partition coefficient scale (Radzicka et al., 1988) yielded the best correlation with our data. However, this hydrophobicity scale does not include a value for proline, so this data point is not shown. Because there is no value for proline and proline is present in the pre-SufI signal peptide, the Kyte and Doolittle scale (Kyte & Doolittle, 1982) was used for Fig. 4E. The Kyte and Doolittle scale also yielded a reasonable correlation with the transport efficiency data.
The pre-SufI-IAC precursor protein was labeled with biotin and Atto565 at the single C-terminal cysteine as follows. Pre-SufI-IAC was incubated with a 10-fold molar excess of tris[2-carboxyethylphosphine] hydrochloride for 10 min (to reduce any disulfides), and then with a 10-fold molar excess of Atto565 maleimide (Sigma) or with 40-fold molar excess of N-(3-maleimidyl propionyl)biocytin (Invitrogen) for 15 min (dark, room temperature). Reactions were quenched with 10 mM β-ME, and purified by Ni-NTA chromatography. Labeled proteins were mixed with 1 mL of Ni-NTA Superflow resin (Qiagen) preequilibrated with 10 mM Tris, 250 mM NaCl, 20 mM imidazole, pH 8.0. The resin was then loaded onto a 10×1 cm column and sequentially washed with: (1) 50 mL of 10 mM Tris-HCl, 1 M NaCl, 20 mM imidazole, pH 8.0; (2) 20 mL of 100 mM NaCl and 10 mM imidazole, pH 8.0; and (3) 5 mL of 100 mM NaCl, 10 mM imidazole, 50% glycerol, pH 8.0. The labeled precursor was eluted (0.2 mL fractions) with 100 mM NaCl, 250 mM imidazole, 50% glycerol, pH 8.0, and stored at −80°C. Biotinylated protein was detected on blots with avidin-HRP and chemilumescence (Harlow & Lane, 1999). Fluorescent proteins were detected by direct in-gel fluorescence imaging using a model FX PhosphorImager (Bio-Rad Laboratories).
All errors are standard deviations. Comparisons of different conditions within each figure panel utilized the same IMV preparation(s). Values compared between panels may be different due to different IMV preparations.
We thank T.L. Yahr for pET-SufI, pTatABC, and the TatA, TatB, TatC, and SufI antibodies; T.L. Yahr and W. Wickner for ptrcOmp9; A. Driessen for pHKSB366; C. Robinson for pJDT1; T. Palmer for MC4100 and MC4100ΔTatABCDE; and Meng Chen and Kelly Soltysiak for technical assistance. This work was supported by the National Institutes of Health (GM065534) and the Welch Foundation (BE-1541).