Direct demonstration of ATP-dependent release of SecA from a translocating preprotein by surface plasmon resonance

Translocase mediates the transport of preproteins across the inner membrane of Escherichia coli. SecA binds with high affinity to the membrane-embedded protein-conducting SecYEG complex and serves as both a receptor for secretory proteins and as an ATP-driven molecular motor. Cycles of ATP binding and hydrolysis by SecA drive the progressive movement of the prepro- tein across the membrane. Surface plasmon resonance allows an online monitoring of protein interactions. Here we report on the kinetic analysis of the interaction between SecA and the membrane-embedded SecYEG complex. Immobilization of membrane vesicles containing overproduced SecYEG on the Biacore Pioneer L1 chip allows the detection of high affinity SecA binding to the SecYEG complex and online monitoring of the translocation of the secretory protein proOmpA. SecA binds tightly to the SecYEG (cid:1) proOmpA complex and is released only upon ATP hydrolysis. The results provide direct evidence for a model in which SecA cycles at the SecYEG complex during translocation. at equilib- rium for different SecA concentrations by Scatchard analysis. Other Biochemical Techniques— Protein concentrations were deter- mined using the Bio-Rad DC protein assay with bovine serum albumin as a standard.


Direct Demonstration of ATP-dependent Release of SecA from a Translocating Preprotein by Surface Plasmon Resonance*
SecA (1) is the central component of the translocase, a multimeric enzyme complex that mediates the translocation of secretory proteins (preproteins) across the inner membrane of Escherichia coli (reviewed in Ref. 2). SecA is a large (102 kDa) homodimeric protein (3,4) that localizes in the cytosol and associates peripherally with the inner membrane. At the membrane, SecA functions as a preprotein receptor and as an ATPdriven molecular motor that threads the preprotein through a pore formed by an oligomeric assembly of the integral SecYEG membrane protein complex (5)(6)(7)(8). SecA binds the SecYEG complex with high affinity. This interaction occurs at least via direct association with SecY (9). SecG is not required for high affinity SecA binding (10), but cross-linking studies indicate that both proteins interact physically during the SecA reaction cycle (11). In addition, SecA associates with low affinity with phospholipids (12).
SecA contains two nucleotide binding domains (NBDs) 1 that are both indispensable for protein translocation (13,14). NBD1 is a high affinity catalytic NBD (K D ADP ϭ 0.15 M) located near the N terminus of the protein. This domain shares significant structural similarity with the DEAD motor domains of helicases (15). The more distally located NBD2 may have a regulatory function instead of being a true ATP hydrolytic site (16,17). The events that take place during ATP-driven protein translocation have been partially resolved. SecYEG-bound SecA binds preproteins via direct recognition of the signal sequence and mature domain (12,18). Translocation of the preprotein is driven by multiple cycles of ATP binding and hydrolysis at SecA (19,20). Each catalytic cycle of SecA involves two distinct translocation steps that utilize the energy of either preprotein or ATP binding to allow the translocation of the preprotein in 20-to 30-amino acid steps (19). Although several studies have addressed the order of events during protein translocation, relatively little is known about the SecA-SecYEG interaction in this process. SecA has been proposed to partition between the cytoplasm and the inner membrane during its reaction cycle (19,21), but other studies question the SecA membrane cycling model as a significant fraction of active SecA remains stably associated with the membrane throughout the translocation reaction (22). Indirect evidence, using SecA bound to a liposome surface instead to SecYEG, suggests that ATP hydrolysis drives the release of the SecA-bound preprotein (19). To address the SecA cycling model directly, we have studied the SecA-SecYEG interaction by surface plasmon resonance (SPR
SPR-SPR measurements were performed using a two-channel BIAlite SPR system (Biacore AB, Uppsala, Sweden). IMVs (2 mg/ml in Buffer A (50 mM Tris-HCl, pH 8.0, 50 mM KCl, 5 mM MgCl 2 , 1 mM dithiothreitol) were passed 11 times through a 100-nm polycarbonate membrane (Avestin, Ottawa, Ontario, Canada) and immobilized on a Pioneer L1 sensor chip (Biacore AB, Uppsala, Sweden). Binding exper-* This work was supported by the Council for Chemical Sciences of the Netherlands Organization for Scientific Research and subsidized by the Dutch Organization for the Advancement of Scientific Research. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ Present address: Institute of Biochemistry, Biocenter, Goethe-Universitä t Frankfurt, Marie Curie Strasse 9, D-60439 Frankfurt, Germany.
Data Analysis-The SPR response prior to injection and the time of injection were adjusted to zero. Binding curves were corrected for changes in the refractive index of the buffer and nonspecific binding to IMVs and the chips surface by the BiacoreX control software 2.0.1 (Biacore AB, Uppsala, Sweden). Corrected binding curves of SecA association and dissociation were quantitatively analyzed using the BIAevalution 2.2.4 software (Biacore AB, Uppsala, Sweden). The data were fitted to a pseudo-first order model of simple 1:1 binding (A ϩ B 7 AB) by non-linear regression. The dissociation rate constant was derived from the equation R t ϭ R 0 e Ϫkoff(tϪt 0 ) where R t is the response in response units (RU) at time t, R 0 is the response at t 0 , and k off is the dissociation rate constant. The association rate constant k on was derived from the equation R t ϭ R eq (1 Ϫ e Ϫ(kon[C]ϩkoff)(tϪt 0 ) , where [C] is the ligand concentration, and R eq is the steady state response level. In addition, the dissociation phase was fitted to a parallel dissociation model (AB1 ϩ AB2 7 A ϩ B1 ϩ B2), using the equation R t ϭ R 1 e Ϫk off1 (tϪt 0 ) ϩ (R 0 Ϫ R 1 )e Ϫk off2 (tϪt 0 ) . The affinity (K D ) was calculated from the ratio k off /k on and from the response levels attained at equilibrium for different SecA concentrations by Scatchard analysis.
Other Biochemical Techniques-Protein concentrations were determined using the Bio-Rad DC protein assay with bovine serum albumin as a standard.

Detection of SecA-SecYEG Interaction by Surface Plasmon
Resonance-To study the interaction between SecA and the SecYEG complex, IMVs containing overexpressed SecYEG (SecYEG ϩ ) were immobilized on a Pioneer L1 sensor chip. The surface of this sensor chip consists of dextran modified with lipophilic anchoring molecules that protrude the lipid bilayer thereby immobilizing the vesicles. The Bialite SPR system contains two flow cells that enable simultaneous monitoring of a sample and a control. A control surface was prepared by loading IMVs containing wild-type (WT) levels of SecYEG onto the second cell. IMV immobilization caused an SPR response between 2500 and 4500 RU (data not shown). Loading levels of the control (WT) and sample (SecYEG ϩ ) cells were similar to allow direct comparison.
The binding of SecA to the immobilized IMVs was followed in time (Fig. 1A). The responses of the WT and SecYEG ϩ IMVs in Fig. 1A include a "bulk" contribution to the SPR signal, i.e. a change in the signal because of changes in the refractive index of the injection buffer. After bulk correction, the SecYEG ϩ response was increased five times compared with the WT response, demonstrating specific SecA-SecYEG interaction. Specific binding of SecA to SecYEG reached saturation of available binding sites at a SecA concentration of about 48 nM (data not shown). The association and dissociation rates of the SecA-SecYEG interaction were determined after correction of the SecYEG ϩ IMV response for the WT IMVs response, which includes the bulk contribution and nonspecific SecA binding (see Fig. 1A and Table I). The association of SecA with the SecYEG complex fitted best to a pseudo-first order model of simple 1:1 binding (A ϩ B 7 AB), with a k on of 2.5 Ϯ 0.4 ϫ 10 Ϫ6 M Ϫ1 s Ϫ1 (see Fig. 1B and Table I). The dissociation phase was more accurately described by a parallel dissociation model (AB1 ϩ AB2 7 A ϩ B1 ϩ B2) with a k off1 of 2.8 Ϯ 0.4 ϫ 10 Ϫ2 s Ϫ1 (27%) and a k off2 of 3.9 Ϯ 0.4 ϫ 10 Ϫ3 s Ϫ1 (73%) (see Fig. 1C and Table I). Parallel dissociation can be caused by the presence of two ligand binding sites with different affinities. However, this phenomenon has been reported to occur frequently in SPR experiments and can also be attributed to sterical hindrance (molecular crowding), mass transport limitations, or non-uniform immobilization of the binding sites (29). The k off1 of the parallel dissociation model yielded an unrealistic (negative) k on ; therefore, the parallel character of the dissociation phase probably does not reflect the existence of two different SecA binding sites. The affinity (K D ) determined by the ratio k off /k on (Table I) was comparable with the SecA binding affinity reported in previous studies (30 -32). The kinetic parameters for SecA binding kinetics were independent of SecA concentration (3-96 nM), and an increased flow rate from 20 to 40 l/min yielded comparable response curves, indicating that the kinetics were not influenced by mass transport limitations. To validate the kinetic analysis of the SecA-SecYEG interaction, the l/min at 25°C. A small fraction of SecA was slowly released. B, fit of the association phase of specific SecA binding to IMVs containing overexpressed SecYEG. The data (gray lines) were fitted (black lines) to a model assuming homogeneous single-site interaction (A ϩ B 7 AB) or a model assuming parallel interaction (A ϩ B1 ϩ B2 7 AB1 ϩ AB2). C, fit of the dissociation phase of specific SecA binding to IMVs containing overexpressed SecYEG. Data (gray lines) were fitted (black lines) to a model assuming homogeneous single-site or parallel interaction. results were subjected to the self-consistency tests proposed by Schuck and Minton (33). The K D determined by Scatchard analysis (3.6 Ϯ 0.2 nM for SecYEG) from the response levels attained at equilibrium for different SecA concentrations was comparable with the k off /k on ratio. The k off calculated using linear regression of the dependence of k obs on [C] (k obs ϭ k on [C] ϩ k off , where [C] is the SecA concentration) was approximately equal to k off determined by fitting to R t ϭ R 0 e Ϫkoff(t Ϫ t 0 ) and the k obs Ͼ k off for all SecA concentrations tested (data not shown). SPR experiments with immobilized vesicles containing SecYE instead or SecYEG (Table I) yielded identical results. This confirms a previous study (10) in which SecG was shown not to contribute to the high affinity binding of SecA to the SecYE complex. Taken together, these data demonstrate that SPR can be used to directly monitor the kinetics of the high affinity interaction between SecA and the SecYEG complex.
Binding of proOmpA to the SecA⅐SecYEG Complex-Se-cYEG-bound SecA binds preproteins either in a complex with the chaperone SecB or as free preprotein, but the exact affinity of the interaction of SecA with preproteins is not known (34). To characterize the interaction of the preprotein proOmpA with SecYEG-bound SecA in more detail by SPR, IMVs were immobilized as described above and saturated with SecA. Subsequently, proOmpA was injected in either the presence or absence of SecB. In the absence of SecB, proOmpA binding resulted in a response of 60 RU (Fig. 2, dashed line) and occurred with an affinity of 14 Ϯ 1 nM (Table II). In the presence of SecB the binding response increased to 120 RU (Fig. 2, solid line). This increased response is probably because of the binding of the larger proOmpA⅐SecB complex, because binding of SecB alone could not be detected under the conditions applied (Fig. 2, dotted line). Biochemical analysis demonstrated that the SecB was normally active (data not shown). The proOmpA⅐SecB complex bound to the SecA⅐SecYEG complex with an affinity of 16 Ϯ 2 nM (Table II), which is nearly identical to the affinity of proOmpA binding in the absence of SecB.
Protein Translocation Monitored by SPR-To determine whether SPR can be used to monitor the interactions that occur during the protein translocation cycle, experiments were conducted in the presence of ATP. Injection of SecA, proOmpA, and SecB resulted in an SPR response of ϳ200 RU (Fig. 3A, dashed line), which corresponds largely to the sum of the individual binding steps of SecA (100 RU; see Fig. 1) and the proOmpA⅐SecB complex (120 RU; see Fig. 2). When ATP was included in the injection buffer, a significant increase in the SPR response was observed that did not reach saturation at the end of the association phase (Fig. 3A, solid line). To address the possibility that the increased SPR response with time was caused by translocation of injected proOmpA, an experiment was conducted that allowed for only one round of translocation. Oxidation of proOmpA induces the formation of an internal disulfide bridge between Cys 290 and Cys 302 , which causes a translocation arrest in the absence of the proton motive force (19,35). As a consequence, translocated proOmpA accumulates as a translocation intermediate (I 29 ) that occupies the translocation site and blocks it for a second round of translocation (7,19,35). The alkaline carbonate treatment after the immobilization of the vesicles removes the F 1 domain of the F 0 F 1 -ATPase and thus the ability to generate a proton motive force upon the addition of ATP (data not shown). When SecA, proOmpA⅐SecB, and ATP were injected under oxidizing conditions, the SPR signal saturated quickly and reached a much lower level than under reducing conditions (Fig. 3B). The maximal response for the injection of SecA, ATP, and oxidized proOmpA⅐SecB (ϳ200 RU; see Fig. 3B, solid line) was comparable with the response of an injection of SecA and proOmpA⅐SecB without nucleotides (Fig. 3B, dotted line). This is consistent with the observation that oxidized proOmpA blocks the translocation sites for further translocation (7,19,35). Oxidation of a Cys-less proOmpA mutant yielded a response similar to wild-type proOmpA under reducing conditions (data not shown). After reduction of the disulfide bond with dithiothreitol, addition of SecA, proOmpA⅐SecB, and ATP resulted in a further increase in the SRP response consistent with a new cycle of translocation (data not shown). These results demonstrate that the immobilized vesicles are capable of translocation and that SPR can be used to monitor proteinprotein interactions during translocation.
Release of SecA from a SecYEG-trapped Preprotein Is ATPdependent and Does Not Require Completion of Translocation-In the absence of ATP, binding of SecA and proOmpA⅐SecB was reversible, and dissociation occurred immediately at the end of the injection (Fig. 3A, dashed line). In contrast, when the complex was formed under translocation conditions (i.e. with ATP), it remained stably associated with the membrane (Fig. 3A, solid line) and was even resistant against alkaline carbonate treatment (data not shown). This   can be explained in part by translocation of proOmpA, but this reaction is not expected to inhibit dissociation of SecA. Only upon a second injection with ATP, part of the complex dissociated from the membrane, whereas ADP was unable to elicit the release (Fig. 3C). To test whether this dissociation represented SecA that was associated previously with the SecYEG complex, the recovery of high affinity SecA binding sites was monitored by a new injection with SecA. For this purpose, SecA was first injected together with the translocation ligands proOmpA, SecB, and ATP and allowed to dissociate in the presence or absence of ATP. Subsequently, a second injection of SecA was passed over the sensor chip, and the binding responses were compared with that of SecA in the absence of translocation ligands (initial response) (Fig. 3D). When complex dissociation was analyzed in the absence of ATP, a second SecA injection resulted in a response of only 30% of the initial response (Fig.  3D). However, when the complex was allowed to dissociate in the presence of ATP, a second injection of SecA resulted in a response of 55% (Fig. 3D). The increase in SPR response corresponds to the amount of material that dissociated from the membrane upon injection of ATP (Fig. 3C). This suggests that ATP is able to release part of the complex from high affinity SecA binding sites. ADP or the non-hydrolysable ATP analogue AMP-PNP failed to dissociate the complex, and under those conditions no high affinity binding sites for SecA were recovered (data not shown). These results demonstrate that the dissociation of SecA from the membrane after one or more rounds of translocation requires the hydrolysis of ATP. An ATP-dependent release from the membrane has previously been proposed based on the behavior of a protease-protected 30-kDa fragment of SecA (21,36). In these studies, formation of the 30-kDa fragment could also be driven by AMP-PNP in the absence of a preprotein. SPR showed that SecA dissociated readily after it had been bound to the SecYEG complex in the presence of AMP-PNP (Fig. 3E, dotted line). The dissociation could, however, be blocked by a second injection with AMP-PNP (Fig. 3E, solid line). These results show that also a nonhydrolyzable ATP analogue can lock SecA at the membrane, but in contrast to the complex formed in the presence of preprotein and ATP, the AMP-PNP-dependent complex readily dissociates upon depletion of the nucleotide. The ATP injection that is required for the dissociation of SecA could be necessary to complete the translocation of the partially translocated preproteins. Alternatively, SecA could dissociate from the membrane upon ATP hydrolysis irrespective of completion of translocation. To discriminate between these possibilities, the dissociation of SecA from the oxidized proOmpA translocation intermediate was studied. Similar to the complex formed under reducing conditions, the oxidized complex did not dissociate from the membrane at the end of injection (Fig. 3B). Dissociation and regeneration of SecA binding sites could be obtained by addition of ATP (Fig. 3D), implying that completion of translocation is not required for the release of SecA from the translocation site. To analyze the re-association of SecA with translocation sites that are occupied by partially translocated preproteins, binding of SecA to an oxidized proOmpA⅐SecYEG complex was examined. SecA was injected together with the translocation ligands proOmpA⅐SecB and ATP under either oxidizing or reducing conditions. After the end of the injection, complexes were allowed to dissociate in the presence of ATP. Subsequently, a second injection of SecA was passed over the sensor chip, and the binding responses were compared. After preincubation under oxidizing or reducing conditions, the second SecA injection resulted in a similar SPR response of ϳ60 RU (see Fig. 3D and Fig. 4, left panel). The dissociation phase, however, was different for the two conditions. When SecA was bound to the oxidized proOmpA⅐SecYEG complex it remained stably associated with the membrane (Fig. 4, left panel). Again ATP was required for dissociation (Fig. 4, right panel), whereas AMP-PNP and ADP were unable to release SecA.
The second round of SecA binding to the complex formed under reducing conditions was reversible, and dissociation occurred immediately after the end of the injection (Fig. 4, left  panel). In contrast to oxidized proOmpA, reduced proOmpA does not form stable translocation intermediates. Likewise, after reduction of the disulfide bond of oxidized proOmpA with dithiothreitol, a second injection of SecA also resulted in reversible binding (data not shown). Conversion of the intermediate I 29 to full-length translocated proOmpA can be driven by low ATP concentrations once the disulfide bridge is reduced (19). Taken together, these results show that SecA binds with a very high affinity to the SecYEG complex when it contains a partially translocated preprotein and that effective dissociation of this preprotein⅐SecYEG complex requires the hydrolysis of ATP. DISCUSSION The SecA motor of the E. coli preprotein translocase is a highly dynamic protein that in the cell distributes between cytosolic and membrane-associated states (37). The interaction of SecA with the inner membrane has been studied previously by binding experiments using radiolabeled SecA. This demonstrated that the SecA association with the membrane occurs with high affinity with the SecYEG complex (34) and with low affinity with the negatively charged lipids (12). Unfortunately, this discontinuous binding assay provides no information about the kinetics of the interaction and cannot easily be applied when protein translocation conditions prevail. To directly monitor the SecA interaction with SecYEG, we have analyzed this process by SPR. This technique has been used extensively for studies on soluble proteins, but the method has only recently become applicable for studies on the interaction with membranes. So far, these studies only concerned the interaction between a soluble protein or peptide and a lipid bilayer. The results presented here show that SPR can be used to obtain detailed and quantitative information on the interaction between the peripheral SecA protein and its membrane receptor, the SecYEG complex. The affinity for the binding of SecA to SecYEG determined by SPR is comparable with values determined previously (30) by binding experiments with radiolabeled SecA. In addition, SPR could be used to study the interactions that occur during protein translocation.
Based on the formation of a protease-protected 30-kDa fragment, the SecA reaction cycle has been proposed to involve nucleotide-induced membrane insertion and de-insertion steps (21). Because the 30-kDa fragment could be chased by an excess of non-radioactive SecA, SecA was proposed to alternate between a cytosolic and membrane-associated state during its reaction cycle (21). However, other studies failed to demonstrate a translocation-dependent dissociation of membranebound SecA and concluded that membrane cycling is not an essential feature for protein translocation (22). The SPR results presented here demonstrate directly that SecA can dissociate from the activated SecYEG complex during a translocation cycle. In the absence of a preprotein, SecA binding to the SecYEG complex is readily reversible. However, when SecA is actively engaged with a preprotein trapped in the translocation channel, SecA release strictly requires the hydrolysis of ATP. The ATP-dependent dissociation is not related to completion of translocation, because SecA also dissociates in an ATP-dependent manner from partially translocated preproteins that remain trapped because of the presence of disulfide bridge. These data confirm previous suggestions that upon ATP hydrolysis, SecA releases the bound preprotein (19). However, in contrast to our SPR study, this conclusion was based on an experiment with SecA bound to lipid vesicles in the absence of the SecYEG complex under non-translocating conditions.
The SPR observations also show resemblance to the behavior of the SecA proteolytic 30-kDa fragment that can be chased by an excess of non-radioactive SecA in the presence of ATP (21), although this experiment does not demonstrate net release of SecA from the translocation site. In the absence of a preprotein, the 30-kDa fragment of SecA could be formed in the presence of non-hydrolyzable ATP analogs, and formation of the 30-kDa SecA fragment was attributed to the binding of ATP during the translocation reaction (36). The requirements for the formation of a stable SecA⅐SecYEG complex that was detected by SPR are different. Various reports suggest that the 30-kDa fragment formed in the presence of AMP-PNP may differ from the one formed under translocation conditions (36,39). The SPR experiments demonstrate that SecA readily dissociated form the SecYEG complex after injection in the presence of AMP-PNP, unless the nucleotide was also present during the dissociation. A SecA⅐SecYEG complex that remained stable after depletion of nucleotides could only be formed under translocation conditions or when SecA re-bound to the SecYEG complex containing a partially translocated preprotein. This suggests that stable membrane association of SecA under translocation conditions is dependent on the presence of a preprotein rather than the binding of ATP. The presence of the preproteins is, however, not sufficient as a dissociable SecA⅐preprotein⅐SecYEG complex is formed when SecA and proOmpA are injected in the absence of ATP (Fig. 3, A and C). Apparently, stable membrane association of SecA does not result from the binding of ATP directly but needs to be preceded by an ATP-dependent preprotein translocation step. A possible explanation for this phenomenon is that SecA interacts with high affinity with the exposed non-translocated polypeptide domains of the preprotein. This may anchor SecA more stable to the SecYEG pore complex such that ATP hydrolysis is needed to dissociate the SecA-preprotein interaction (19). Alternative explanations need to be considered, as well. SecA has been proposed to be subjected to changes in its quaternary structure during its reaction cycle (40 -42). Differences in the oligomeric state of SecA could affect the affinity of the SecA-SecYEG interaction, but in our SPR experiments we cannot discriminate between a monomeric and dimeric state of SecA at the translocase. Because of the presence of a trapped translocation intermediate, the SecYEG complex may exist in a different conformational state that permits a tight interaction with SecA. Biochemical and electron microscopy studies have indicated that the translocation pore is formed by an oligomeric assembly of SecYEG complexes (7). In a three-dimensional structure reconstituted from two-dimensional SecYEG crystals the SecYEG complex is dimeric and displays a deep cavity at the dimer interface (38). Because this cavity is closed at the periplasmic side of the membrane, the structure was proposed to represent a "closed state" of the translocation channel (38). The structure would convert into an "open state" when the complex is actively involved in translocation, but it is not clear whether the open state reflects a different conformational or oligomeric state. In this respect, electron microscopy suggests that SecYEG dimers tetramerize upon activation of SecA by AMP-PNP or ATP and a preprotein (7). Although the oligomeric state of the active SecYEG pore complex remains a topic of debate, both models propose that the protein-conducting pore needs to be activated at the initiation of translocation. The ATP-dependent step that is required for the stable membrane association of SecA could be related to this activation of the translocation pore. In such a scenario, SecA binds to the inactive SecYEG complex in a reversible manner. Upon binding of a preprotein and ATP to SecYEGbound SecA the SecYEG complex is activated, and this event locks SecA at the translocation site. When SecA is released from the membrane after the hydrolysis of ATP, the trapped translocation intermediate can stabilize the pore in the activated state. As a consequence, in a second reaction cycle SecA can bind tightly to the SecYEG complex in the absence of ATP. In a non-arrested reaction cycle, re-binding of SecA to the translocation intermediate results in a translocation progress (19,20).
In conclusion, this study demonstrates for the first time that SPR can be used to monitor the association and dissociation kinetics of a peripheral protein component with its membrane receptor embedded in its natural lipid environment. The SPR results confirm a translocation model in which SecA alternates between a free (cytosolic) and membrane-associated state. The study further demonstrates that during the translocation reaction, SecA is tightly bound to the activated SecYEG complex but can be released upon the hydrolysis of ATP.