A significant fraction of functional SecA is permanently embedded in the membrane. SecA cycling on and off the membrane is not essential during protein translocation.

SecA has been suggested to cycle on and off the cytoplasmic membrane of Escherichia coli during protein translocation. We have reconstituted 35S-SecA onto SecA-depleted membrane vesicles and followed the fate of the membrane-associated 35S-SecA during protein translocation. Some 35S-SecA was released from the membranes in a translocation-independent manner. However, a significant fraction of 35S-SecA remained on the membranes even after incubation with excess SecA. This fraction of 35S-SecA was shown to be integrated into the membrane and was active in protein translocation, indicating that SecA cycling on and off membrane is not required for protein translocation. Proteolysis experiments did not support the model of SecA insertion and deinsertion during protein translocation; instead, a major 48-kDa domain was found persistently embedded in the membrane regardless of translocation status. Thus, in addition to catalyzing ATP hydrolysis, certain domains of SecA probably play an important structural role in the translocation machinery, perhaps forming part of the protein-conducting channels.

Translocation of most periplasmic and outer membrane proteins across the cytoplasmic membrane of Escherichia coli occurs through the Sec-dependent pathway (1,2), which consists of a distinct set of Sec proteins (3,4). SecA plays a pivotal role in protein translocation. It couples the energy of essential ATP hydrolysis to protein translocation (5)(6)(7)(8) and has been shown to be essential for protein translocation both in vivo (9,10) and in vitro (11)(12)(13)(14). SecA interacts with most components of the translocation machinery. It is believed that SecA binds to the membrane through interactions with SecYEG (15,16) and acidic phospholipids (17)(18)(19)(20)(21), and interacts with precursor proteins by recognizing the positive charge at the NH 2 terminus of the signal peptides (22,23). Binding of precursor proteins to the SecA subunit of the translocase stimulates the translocation ATPase activity of SecA (18). Recently, it has been reported that SecA promotes protein translocation by undergoing ATP-driven cycles of membrane insertion and deinsertion (24) at SecYEG, and that this cycle is regulated by SecD/F (25) and the ATP-binding amino terminus of SecA (26).
While SecYEG are integral membrane proteins (27,28), SecA is considered a peripheral membrane protein since it lacks any predicted transmembrane domains (29), and was found in both cytosolic and membrane fractions (9,30). It is generally believed, therefore, that SecA cycles between the cytoplasm and the membrane during protein translocation (24,31). Consistent with this notion, soluble SecA can restore translocation activity of membranes with depleted or inactivated SecA (11)(12)(13)(14) or impaired SecY (32). On the other hand, extraction experiments with urea or Na 2 CO 3 suggested that a significant fraction of SecA behaves like an integral membrane protein (30). 1 Recently, Kim et al. (33) reported that a strain harboring a plasmid containing the secD secF locus possesses SecA almost entirely as an integral membrane form and displays normal protein translocation as measured by rapid processing of preproteins in vivo (although it is possible that only a small fraction of SecA is active which accounts for the apparent "normal" translocation). Kim et al. (33) suggested that the integral SecA is the catalytically active form in vivo and that, in this state, it might form a part of the reported protein-conducting channels (34). However, it is not clear what fraction of SecA is active and whether this integral SecA remains active and cycles off the membranes as suggested by Economou and Wickner (24).
Here, we present evidence that the integral SecA under the conditions used is indeed active and does not cycle off the membrane during protein translocation. We have determined the fate of the membrane-associated SecA during protein translocation, using membranes reconstituted from 35 S-SecA and SecA-depleted inverted membrane vesicles. There were two forms of 35 S-SecAs on the reconstituted membranes, loosely associated and integral. The loosely associated 35 S-SecA can be replaced by excess nonradioactive SecA, and can be extracted by Na 2 CO 3 or heparin. However, the release of this membraneassociated 35 S-SecA is translocation-independent. On the other hand, a significant fraction of 35 S-SecA remains on the membranes even after incubation with excess nonradioactive SecA. This fraction of 35 S-SecA is integral, since it was found associated with membranes after flotation centrifugation and was also resistant to Na 2 CO 3 extraction. After extraction with Na 2 CO 3 or heparin, the 35 S-SecA remaining on the reconstituted membranes was as active as the SecA remaining on native membranes. Proteolysis revealed that a 48-kDa domain is constantly embedded in the membranes regardless of protein translocation status. Together, these results indicate that a significant fraction of SecA is persistently embedded in the membrane and, as such, is active in protein translocation. Therefore, SecA cycling on and off membrane is not essential for protein translocation.
Purification of 35 S-SecA and 35 S-pOmpA-35 S-SecA was purified from BL21 (DE3)/pT7-secA (11). IPTG (0.5 mM) was added to cells grown at 37°C in MinA medium when A 600 reached 0.5. After 10 min, 35 S protein labeling mix was added to 1 mCi/liter and the culture was incubated for additional 2 h. The harvested cells were washed once with DTKM buffer, resuspended in DTKM buffer containing 5 g/ml DNase and 0.5 mM PMSF, and were lysed by passing through a French press at 15,000 p.s.i. The cell lysate was centrifuged at 8,000 rpm for 15 min to remove the debris, followed by another centrifugation at 55,000 rpm for 90 min in a Beckman 70 Ti rotor to give a postribosomal supernatant (S100). The S100 was loaded onto S-Sepharose column equilibrated with 25 mM phosphate buffer, pH 6.4, and was eluted by a NaCl gradient. 35 S-SecA was eluted at 0.3 M NaCl. This 35 S-SecA fraction was concentrated by precipitation with ammonium sulfate. The pellet was then dissolved in 0.15 M ammonium carbonate and passed through a S-300 Sephacryl column. The final SecA product is virtually pure as analyzed by SDS-PAGE and staining/autoradiography. 35 S-pOmpA was purified from BA13/pM103 as follows. An overnight culture grown at 30°C in the MinA medium containing 0.5% glucose was used to inoculate fresh medium at a dilution of 1:100. At cell density of 0.2 A 600 , the culture was rapidly shifted to 42°C to deplete SecA. When the growth of cells was arrested, indicated by a steady A 600 reading over 30 min, IPTG was added to the culture to induce the overproduction of pOmpA. After 10 min, 35 S protein labeling mix was added to 1 mCi/liter and the culture was incubated for an additional 2 h, 35 S-pOmpA accumulated in the cells as inclusion bodies. After cell lysis, 35 S-pOmpA was extracted from the insoluble fraction with 6 M urea. 35 S-pOmpA was purified from this extract using a Q-Sepharose column. The final pOmpA product is virtually pure as analyzed by SDS-PAGE and staining/autoradiography.
Membrane Reconstitution-SecA-depleted membrane vesicles were prepared from CK1801.4 or BA13 cells grown in LinA medium with glucose. SecA depletion was achieved as in the case of BA13/pM103 (above). When the growth of cells was arrested, the cells were harvested and taken through the standard membrane preparation procedures (37,38) with modifications on the last step. The centrifugation time at 37,000 rpm was reduced from 90 min to 30 min to collect only large membrane vesicles to facilitate the recovery of the membrane vesicles by vacuum filtration assay. Membranes from CK1801 strain were prepared by our standard membrane preparation procedures (37). Reconstituted membranes were prepared immediately before use in assays as follows. Membrane vesicles (0.5 mg) were incubated with 50 g of purified 35 S-SecA in 1 ml of translocation buffer containing the energy source at 30°C for 5 min. The mixture was diluted with 2 ml of DE 20 , and membrane vesicles were reisolated by centrifugation at 70,000 rpm for 1 h at 4°C through a sucrose cushion in a TL100.3 rotor. The pellet was suspended in 3 ml of DE 20 , centrifuged again, and resuspended in 150 l of TK buffer. The radioactivity of the reconstituted membrane suspension was counted to calculate the amount of membrane-associ-ated 35 S-SecA. An aliquot of 10 l was used in assays.
In Vitro Translocation of OmpA-Nascent pOmpA was synthesized by in vitro protein translation reaction as described previously (39). Post-translational supernatant was obtained by centrifuging at 90,000 rpm for 20 min in a Beckman TLA 100.2 rotor at 4°C. Translocation reactions (0.1 ml) using the nascent pOmpA contained 85 l of posttranslational supernatant supplemented with 2 l of energy source, 30 g of membrane vesicles, and other constituents as indicated in legends. For kinetic assays, membranes were mixed with the translocation mixture first, and then distributed into individual tubes (100 l each). Translocation assays were carried out at 37°C for 15 min unless otherwise indicated. After incubation, the reaction mixtures were chilled on ice, treated with 0.2 mg/ml proteinase K on ice for 15 min (no difference of translocated OmpA was observed with up to 1 mg/ml proteinase K). Stop solution (0.7 ml) was added to stop the protease digestion. The membranes were recovered by sedimentation at 95,000 rpm for 20 min through a 0.2-ml sucrose cushion, and were analyzed by SDS-PAGE and autoradiography. Translocation efficiency was calculated from the amount of protease-resistant pOmpA plus OmpA, or OmpA alone when the protease treatment was omitted. The radioactivity of pOmpA was corrected for the loss of one [ 35 S]Met residue in OmpA. For translocation with purified pOmpA, 2 g (1-2 l) of purified 35 S-pOmpA in 6 M urea was diluted into 85 l of translocation buffer containing 5 g of SecB (and SecA or other factors where indicated); then, 30 g of membrane vesicles were added. Translocation reactions were performed as above. 35 S-SecA and 35 S-OmpA bands on gel were quantitated either by a Phosphoimager (Fuji BAS-1000) or by scanning densitometry using a PDI Image Analyzing System (Protein Databases Inc., New York, NY). Proteins in solution or in membranes were quantitated by Bradford (40) reagent from Bio-Rad using ␥-globulin as a standard.
Analysis of SecA Release by Filtration Method-The release of SecA from the membrane was also determined with a MultiScreen assay system (Millipore), which consists of a standard 96-well microtiter plate with a sealed microporous membrane bottom and a vacuum manifold. After preliminary tests, we chose a 0.22-m hydrophilic polyvinylidene difluoride, low protein-binding Durapore plate in the experiments. More than 80% of the membrane was retained by the polyvinylidene difluoride filter membrane, whereas less than 5% of free SecA was retained. Reconstituted membrane vesicles were incubated in 50 l of reaction mixture containing 3 g of membrane vesicles and various factors. At the end of incubation, free 35 S-SecA was separated from membrane-associated 35 S-SecA within 1 min by applying vacuum to the system. The filter bottoms containing the membrane-associated 35 S-SecA were then punched out, and counted for radioactivity in a Beckman LS 6500 scintillation counter. All samples were assayed in triplicate.

Release of SecA from Membranes Is Independent of Protein
Translocation-The existence of SecA in a soluble form as well as in a membrane-associated form suggests that SecA cycles on and off the membrane, presumably during protein translocation. SecA binds to and inserts into membranes under various conditions (11, 15-21, 32, 33). On the other hand, SecA release from membranes has been less characterized. To determine directly whether the membrane-associated SecA can be released in a soluble form during protein translocation, we reconstituted 35 S-SecA onto SecA-depleted E. coli membrane vesicles and followed the fate of the membrane-associated SecA during protein translocation. Since all SecA was radioactive, the amount of SecA on the membrane was easily monitored by counting the radioactivity in the recovered membrane fraction either by liquid scintillation or by SDS-PAGE and autoradiography.
The reconstituted membranes were obtained by incubating purified 35 S-SecA (50 g) with SecA-depleted membranes (0.5 mg) at 30°C for 5 min. About 15 g of 35 S-SecA was associated with the recovered membrane vesicles. We used an unc Ϫ CK1801.4 strain in our experiments to eliminate the potential effects of F 1 -ATPase that has been reported to restore translocation activity of urea-inactivated membranes (14). Membranes from SecA-depleted CK1801.4 strain contained little SecA (data not shown) and were inactive in translocation of nascent proOmpA without exogenous SecA ( Fig. 1, lanes 1 and 2). On the other hand, the reconstituted membranes reached maximum translocation efficiency without additional SecA ( Fig. 1, lanes 3 and 4).
To examine the effects of translocation components on the release of 35 S-SecA from the membranes, reconstituted membranes were incubated either in a post-translational supernatant containing nascent pOmpA or in translocation buffer under various conditions ( Fig. 2A). No significant release of 35 S-SecA was observed when the reconstituted membranes were incubated without nonradioactive SecA ( Fig. 2A, lanes 1-3 and 7-11). Inclusion of 10 g of nonradioactive SecA in the reaction mixture "chased" 50 -60% of 35 S-SecA from the membranes ( Fig. 2A, lanes 4 and 5 and lanes 12 and 13). Such replacement by nonradioactive SecA occurred even on ice, where no translocation took place ( Fig. 2A, lanes 4 and 12). The post-translational supernatant containing nascent precursor proteins had little effects on 35 S-SecA release (compare lanes 1-5 with lanes 8 -13), suggesting that the release process is independent of protein translocation.
The conventional sedimentation method used above includes a dilution step (to reduce the nonspecific binding) and a 15-min centrifugation (to collect the membrane vesicles). The dilution and the centrifugation time may have affected the 35 S-SecA binding and release. To avoid such concerns, a vacuum filtration method was used for rapid separation of the free 35 S-SecA and the membrane-associated 35 S-SecA. After incubation with translocation buffer containing the energy sources in a Multi-Screen assay system (see "Experimental Procedures"), the membranes were quickly collected by applying vacuum to the system. 35 S-SecA remains on the filter only if it binds to the membrane vesicles. First, the kinetics of 35 S-SecA binding and release were examined. Both 35 S-SecA binding and 35 S-SecA release occurred rapidly at 1 min and reached equilibrium within 5 min at 30°C (Fig. 2B). The dissociation constant and the maximum binding for 35 S-SecA were estimated as 72 nM and 82 pmol (as dimer)/1 mg of membrane protein, respectively, close to those obtained with the sedimentation method (data not shown; Ref. 15). Next, the effects of translocation components on the 35 S-SecA release were examined with this filtration method (Fig. 2C). The results were similar with those obtained with the centrifugation method ( Fig. 2A). These data demonstrate that both the centrifugation and the filtration methods can be used to quantitate the amount of membraneassociated SecA. However, to study the release of SecA during protein translocation, both translocation and SecA release must be monitored simultaneously. This can only be achieved with centrifugation followed by SDS-PAGE. Thus, centrifugation was used to collect membranes after translocation in most of the following experiments, repeating some experiments with the filtration method, as needed.
If cycling of SecA between membrane and cytoplasm is an obligatory step for protein translocation, the release of 35 S-SecA from reconstituted membranes would be expected to show some kind of correlation to the protein translocation. However, the results in Fig. 2 suggest that release of SecA is not related to protein translocation. To investigate further the relations between SecA release and protein translocation, we examined the kinetics of release of the membrane-associated 35 S-SecA  [1][2][3][4][5] or 90 l of translocation buffer (ϪpOmpA) with (lanes 9 -13) or without (lanes 7 and 8) the energy source (ATP), followed by incubation for 15 min at 37°C or 0°C as indicated. SecA (20 g) was present in lanes 4, 5, 12, and 13. EDTA (20 mM) was present in lanes 3 and 11. Membranes were isolated by centrifugation and analyzed by SDS-PAGE and autoradiography. The control sample (lane 6) was isolated immediately after mixing without incubation. B, incubations were carried out in 50 l of translocation buffer containing the energy source at 30°C using a MultiScreen assay system (see "Experimental Procedures"). For the binding assay, 14 g of CK1801.4 membranes and 1 g of 35 S-SecA were used. To assay 35 S-SecA release, 14 g of reconstituted membranes were used. At the indicated time, the buffer was quickly removed by applying vacuum to the system. The amount of membrane-associated 35 S-SecA was quantitated by counting the radioactivity retained in the filters. Open circles, the amount of 35 S-SecA bound to the membranes in the binding assay. Closed circles, the amount of 35 S-SecA remaining on membranes in the release assay. C, reconstituted membranes (3 g) were incubated at 37°C for 15 min in 50 l translocation buffer using the MultiScreen assay system with various additions as indicated: ATP (as present in the energy source), EDTA (20 mM), SecA (10 g of nonradioactive SecA). The amount of 35 S-SecA was quantitated as described in B.
during translocation of pOmpA using our standard in vitro translation/translocation assay system (37). The reconstituted membranes were incubated at 37°C for different periods in post-translational supernatant containing nascent pOmpA. While the amount of translocated OmpA increased with time, the amount of membrane-associated 35 S-SecA remained essentially unchanged throughout the same period (Fig. 3). A similar release curve was observed even when the incubations were carried out in translocation buffer without pOmpA. These data reinforce the notion that SecA release is independent of protein translocation.
It is possible that the amount of nascent pOmpA in the reaction mixture might be too little to bring a significant amount of 35 S-SecA off the membranes. To amplify the effects of translocation on SecA release, we employed an in vitro translocation system using chemical amount of purified 35 S-pOmpA renaturing from urea solution (41). This in vitro assay system used 2 g of purified 35 S-pOmpA, 5 g of SecB, and 30 g of membranes containing 1 g of 35 S-SecA. If one 35 S-SecA molecule/translocated 35 S-pOmpA molecule cycles off the membranes and the translocation efficiency is 50%, the amount of translocated 35 S-pOmpA (1 g or 28 pmol) should be sufficient to bring all 35 S-SecA (1 g or 5 pmol as a dimer) off the membranes. This may be an underestimate, however, since it has been proposed that one cycle of SecA insertion and deinsertion, therefore, on and off the membranes, moves only 20 -30 amino acid residues (42,43). If this is the case, translocation of one molecule of OmpA would require 12-18 SecA molecules to cycle on and off the membranes. Fig. 4 shows the kinetics of 35 S-SecA release during translocation of purified 35 S-pOmpA. The reconstituted membranes were active in both translocation (45%) and processing (70%) of pOmpA (Fig. 4A, lanes 4 and 5). After translocation, 90% of the OmpA was resistant to proteinase K (0.5 mg/ml) digestion, but was completely degraded by proteinase K in the presence of 1% Triton X-100 (lanes 4 -6 in both panels A and B). Although the release of 35 S-SecA plateaued within 3 min, the translocation of pOmpA increased continuously with time ( Fig. 4C, pOmpA "ϩ", No Chase). After 15 min of translocation, 28 pmol (0.9 g) pOmpA were translocated, but only about 0.3 g (1.5 pmol as dimer) of the membrane-associated 35 S-SecA were released. Furthermore, a similar release curve was observed when no translocation took place (Fig. 4C, pOmpA "Ϫ", No chase). Release of membrane-associated 35 S-SecA increased to 0.6 g when 10-fold excess nonradioactive SecA was added to the reaction mixtures to prevent the rebinding of the released 35 S-SecA (Fig. 4, B and C,  Chase). However, the release curves were identical in the presence or absence of pOmpA (Fig. 4C, Chase). Thus, it is unlikely that the release of SecA is coupled to protein translocation.
We also determined the amount of nonradioactive SecA needed to achieve maximum release of membrane-associated 35 S-SecA during translocation (Fig. 5A). The amount of membrane-associated 35 S-SecA decreased sharply (40%) with increasing amount of nonradioactive SecA from 0 to 10 g (10fold). The maximum release (64%) was reached with 20 g of SecA. No further release was observed thereafter. In these experiments, BSA was used during the SecA reconstitution and chase. When BSA was omitted, a higher level of 35 S-SecA (60%) remained on the membranes after the chase (data not shown). These data suggest that a significant fraction of the membraneassociated 35 S-SecA in the reconstituted membranes cannot be chased. To avoid a possible deleterious effect that high concentrations of SecA (up to 500 g/ml) may have on the translocation assay, we took advantage of the filtration method that allows an efficient recovery of even a small amount of membranes. The amount of membranes was reduced to 1/10 to examine the 35 S-SecA release during protein translocation in the presence of 50-fold excess nonradioactive SecA. As with the centrifugation method (Fig. 4C), similar release curves were obtained in the presence or absence of pOmpA (Fig. 5B). We conclude that release of the membrane-associated SecA is indeed independent of protein translocation. More importantly, a significant fraction of the membrane-associated SecA does not cycle off the membranes.
Since F 1 F 0 -ATPase has been reported to restore the translocation activity of urea-washed membranes (14), we have performed similar experiments with 35 S-SecA-membranes reconstituted from 35 S-SecA and SecA-depleted membranes from BA13 strain, which contains F 1 F 0 -ATPase. Results were similar to those obtained with the reconstituted 35 S-SecA-CK1801.4 membranes (data not shown), indicating that the presence or absence of F 1 F 0 -ATPase has little effect on the preceding observations. Collectively, these data showed that there are two different populations of SecA on the membranes, loosely associated and tightly associated. The former set can be released and replaced by cytosolic SecA in a translocation-independent manner. It is likely that the latter set is permanently integrated into the membranes. Alternatively, this observation could be due to denatured aggregates that co-sediment with the membranes.
Integral SecA Does Not Cycle Off the Membrane during Protein Translocation-To examine the possibility that the tightly associated 35 S-SecA is simply denatured aggregates that cosediment with the membranes, a flotation experiment was performed. After translocation reaction in the presence of excess nonradioactive SecA, reconstituted membranes were isolated by centrifugation, resuspended in TK buffer, and subjected to the flotation centrifugation. Under the conditions described previously (44), the membranes float, while the soluble proteins and protein aggregates remain at the bottom of the metrizamide gradient according to their densities. Translocated 35 S-OmpA was used to probe the position of membranes. As shown in Fig. 6A, more than 90% of 35 S-SecA and 35 S-OmpA (thus, membranes) were found in the top three fractions, whereas the soluble 35 S-SecA or denatured 35 S-pOmpA aggregates were recovered in the bottom three fractions. These  (44) with a minor modification. The total volume of the metrizamide medium was increased from 0.5 ml to 0.8 ml using a Beckman TL100.2 rotor. After a translocation reaction at 37°C for 15 min in the presence of 20 g of nonradioactive SecA, the reconstituted membranes were recovered by centrifugation at 95,000 rpm for 15 min through the sucrose cushion, and suspended in 100 l of TK buffer. pOmpA aggregates were generated by diluting 100 l of pOmpA (200 g) in 6 M urea with 1 ml of TK buffer, followed by incubation at room temperature for 1 h. The resulting pOmpA aggregates were collected by centrifugation at 95,000 rpm for 15 min, and suspended in 100 l of TK buffer. The suspension of membranes or pOmpA aggregates as well as free soluble SecA solution were brought to a density of 1.29 g/ml with solid metrizamide. One hundred l of each sample was loaded to the bottom of separate centrifuge tubes. A solution of metrizamide in TK buffer of density 1.27 (0.7 ml) was layered over the samples. After flotation centrifugation at 100,000 rpm for 5 h, six fractions (133 l each) were removed from the gradients (top to bottom), and the centrifuge tubes were washed with 30 l of SDS-PAGE sample buffer for a final fraction (fraction 7). Forty l from fractions 1-6 were mixed with 10 l of 5 ϫ sample buffer. All fractions were then analyzed by SDS-PAGE and autoradiography. The amount of 35 S-SecA and 35 S-OmpA in each fraction was given as percentage of the total amount recovered from all fractions. B, reconstituted membranes after incubation with 0.1 M Na 2 CO 3 were isolated, and were incubated in translocation buffer containing the energy source in the absence or presence of excess nonradioactive SecA (I). Alternatively, reconstituted membranes after the chase were isolated, and were incubated in TK buffer or 0. l of translocation buffer containing 5 g of SecA in the absence (pOmpA "Ϫ") or presence (pOmpA "ϩ") of 2 g of proOmpA in the MultiScreen assay system. At the end of the incubation, the buffers were quickly removed by applying vacuum to the system. 35 S-SecA remaining on the membranes were quantitated by counting the radioactivity retained on the filters. data showed that the 35 S-SecA remaining on the membranes after the chase was truly associated with the membranes, since it was retained with the membrane fractions, while soluble proteins and protein aggregates were not.
It is most likely that the 35 S-SecA remaining in the membrane fraction is integrated into the membrane. One criterion for such an integral membrane protein is the resistance to Na 2 CO 3 extraction (45), which was therefore undertaken using treatment with TK buffer as a control. Reconstituted membranes were incubated with excess nonradioactive SecA, reisolated by centrifugation, and then incubated in TK or 0.1 M Na 2 CO 3 on ice for 30 min. After the incubation, membranes were isolated and analyzed by SDS-PAGE. Most of the 35 S-SecA that remained on the membranes after the chase was resistant to extraction with 0.1 M Na 2 CO 3 (Fig. 6B). Similar results were obtained with higher concentrations of Na 2 CO 3 (0.15 and 0.2 M). When the reconstituted membranes were extracted prior to the chase, most of the 35 S-SecA on the Na 2 CO 3 -extracted membranes could not be further chased. These data suggest that the SecA remaining on the membranes after Na 2 CO 3 extraction is resistant to chase and vice versa. Thus, this fraction of SecA is integrated into the membranes and does not cycle off the membrane during protein translocation.
Integral SecA Is Functionally Active in Protein Translocation-Since reconstituted membranes washed with Na 2 CO 3 leaves only integral SecA in the membranes, these membranes were used in translocation assays to examine the translocation activity of the integral SecA. The translocation activities of SecA (molar ratio of translocated pOmpA over SecA dimer) on the Na 2 CO 3 -or TK-treated membranes were compared (Fig.  7A). 35 S-SecA on the reconstituted membranes was found as active as SecA on native membranes derived from the parental CK1801 strain. After Na 2 CO 3 extraction, SecA on both membranes lost about two-thirds of their original activities. Nevertheless, each integral SecA dimer was still capable of translocating, on average, up to 4 molecules of OmpA. These data show that though Na 2 CO 3 treatment does partially inactivate SecA, the SecA retained on the membranes remains active in promoting protein translocation.
Watanabe and Blobel (14) reported that peripheral SecA could be specifically stripped from membrane vesicles by polyanionic heparin, and such heparin-extracted membrane vesicles were as active as control membrane vesicles. The heparinextraction was performed with our reconstituted membranes using TK extraction as a control. The reconstituted membranes lost half the activity and a corresponding amount of SecA by the heparin extraction, such that the translocation activity per SecA molecule on the membranes after TK or heparin extraction was similar (Fig. 7B). Less SecA was extracted by heparin from native membranes derived from CK1801, and so was the loss of the translocation activity of these membranes. Thus, the translocation activities of the native membranes after TK or heparin extraction were also similar when normalized to the amount of SecA remaining on the membranes (Fig. 7B). These data indicate that, unlike Na 2 CO 3 treatment, heparin treatment does not inactivate SecA. Extraction experiment showed that SecA on heparin-extracted membranes was resistant to further Na 2 CO 3 extraction, as were SecY and SecE (Fig. 7C), indicating that SecA in the heparin-extracted membranes is integrated into the membranes. After heparin extraction, the translocation activity per SecA molecule on both the reconstituted membranes and the native membranes was similar, indicating that the integrated 35 S-SecA was as active as the integral SecA in the native membranes. On the other hand, the integral SecA appears to be less active compared with SecA on untreated membranes where 8 molecules of pOmpA were translocated per SecA dimer in 15 min (compare Figs. 7 and 4). This apparent loss of translocation activity is probably due to repeated resuspensions and centrifugations. The translocation activity per molecule of SecA is similar for both heparin-and TK-treated membranes (Fig. 7B), suggesting the integral SecA is at least as active as the whole SecA populations on the membranes. Thus, we conclude that integral SecA is functionally active in protein translocation.
A 48-kDa Domain Is Constantly Embedded in the Membrane Regardless of Protein Translocation Status-Although SecA lacks any predicted transmembrane domains, it does have several relatively hydrophobic regions (residues around 150, 200, and 500). We have performed proteolytic assays to identify possible membrane-embedded SecA domains. Reconstituted membranes were digested with proteinase K (1 mg/ml, ice bath, 15 min) after incubation at 37°C for 15 min under various conditions as indicated (Fig. 8A). When reconstituted membranes were incubated in translocation buffer alone, seven FIG. 7. Translocation activity of membranes after extraction with Na 2 CO 3 or heparin. A, reconstituted membranes and the membranes from its parental CK1801 strain were incubated on ice for 30 min with TK buffer (TK) or 0.1 M Na 2 CO 3 (CO 3 ). The reisolated membranes were incubated in translocation buffer containing 3 g of SecB, 1.3 g of pOmpA, and 5 mM NADH at 37°C for the time indicated. After incubation, the mixtures were treated with proteinase K on ice for 15 min. The membranes were isolated by centrifugation and were analyzed by SDS-PAGE. The bands of 35 S-SecA and 35 S-OmpA was quantitated by Fuji Phosphoimager. The amount of SecA in the native membranes was quantitated from immunoblots. Translocation activity was given as translocated pOmpA (mol)/SecA dimer (mol). B, reconstituted membranes and CK1801 membranes were extracted with 10 mg/ml heparin as described by Watanabe and Blobel (14). Membranes treated with TK buffer were used as control. After the extraction, the membranes were isolated, and were assayed for translocation as described in A. C, reconstituted membranes were extracted by 0.1 M Na 2 CO 3 (CO 3 ) or TK buffer (TK) before (I) and after (II) heparin treatment. After the extractions, the membranes were isolated, and analyzed by SDS-PAGE followed by immunoblots using antibodies against SecA, SecY, and SecE.
SecA fragment bands (95, 90, 66, 48, 36, 28, and 25 kDa) as well as a band equivalent to intact SecA were observed after proteolysis (Fig. 8A, lane 1). Similar results were obtained with shorter or longer incubation at 37°C (data not shown). The amount of the 95-kDa and the intact SecA species varied in different experiments. All bands below 90 kDa were consistently observed throughout more than 20 experiments with different preparations of membranes and SecA. In the presence of ATP, but without pOmpA, the 66-kDa band decreased, while the 25-kDa band increased significantly. The other species increased slightly (Fig. 8A, lane 2). These changes were not observed with AMP-PCP instead of ATP (data not shown). Translocation of pOmpA increased the 66-kDa and the 28-kDa bands about 3-fold, but had little effect on the other bands (Fig.  8A, lane 3). More than half of the three small fragments (36,28,and 25 kDa) and about 30% of the 66-kDa fragment was also found in the supernatant (Fig. 8A, lanes 8 and 9), indicating that these bands were intrinsically resistant to the proteinase K in the presence of ATP and probably not related to protein translocation. The relationship between protein translocation and the amount of the 66-and 48-kDa fragments was further examined. As shown in Fig. 8B, the 66-kDa fragment increased with the increase of proOmpA and concomitant translocated OmpA. However, the amount of the 48-kDa fragment remained the same throughout.
Nonradioactive SecA (20 g) was added to the reaction mixture to determine the membrane-embedded SecA domains (Fig.  8A, lanes 4 -6), since such domains will be resistant to the chase. The proteolytic fragments observed in the absence of ATP were similar with or without SecA chase (lanes 1 and 4). In the presence of ATP, the 66-kDa band decreased significantly, whereas the majority of the 48-and 90-kDa bands remained on the membranes in the presence of excess SecA (lanes 5 and 6  compared with lanes 2 and 3). Addition of proOmpA had little effect (lanes 5 and 6). These data suggest that a SecA domain containing at least the 48-kDa fragment, and perhaps the 90-kDa domain, was present on the membranes all the time regardless of the status of protein translocation. Taken all together, evidence from this work reveals that a significant fraction of SecA is permanently embedded in the membranes and is active in protein translocation. DISCUSSION We have shown that there are two populations of SecA present on reconstituted membranes, loosely associated and tightly associated. Loosely associated SecA probably binds to membrane through ionic interaction between SecA and phospholipids. As a result, exchange of this loosely associated SecA is translocation-independent and such SecA could be stripped away from the membranes by the polyanionic heparin. Binding of SecA to the membrane is a spontaneous process, which can occur even on ice without ATP ( Fig. 2A, and Ref. 15). The tightly associated SecA might become inserted into the membrane. This possibly occurs through interactions with phospholipids, SecYEG (15,(17)(18)(19)(20)(21), or other proteins so that it becomes resistant to extraction by heparin or Na 2 CO 3 . Our data are consistent with evidence reported by Cabelli et al. (30), who showed that in membranes isolated by chromatography in buffer, one-third of the membrane-associated SecA could not be extracted by Na 2 CO 3 . In the current studies, the integral nature of this Na 2 CO 3 -resistant SecA was further confirmed and extended by both its resistance to replacement with excess nonradioactive SecA under translocation condition, and its association with membrane vesicles during flotation centrifugation. The amount of integral SecA is about 10 g (50 pmol of dimers)/1 mg of membrane protein, similar to the amount of SecA present on membranes from the SecA ϩ parental CK1801 and wild type D10 strains. Most of SecA on these native membranes is also resistant to Na 2 CO 3 and urea extraction, 1 presumably because loosely associated SecA had previously been removed by the membrane preparation procedures, which consist of a series of density centrifugation in buffer containing EDTA (38). It has been suggested that SecA cycles on and off the membrane during protein translocation (24,31). However, this cycling model lacks direct experimental evidence. Our data show that a significant fraction of functional SecA is embedded in the membrane during protein translocation; therefore, SecA cycling on and off membrane is not essential in protein translocation.
It appears highly likely that soluble SecA can bind to membranes and a fraction becomes integrated, but that, once integrated, SecA cannot be released into a soluble form. This integral SecA might be modified or rearranged to become permanently embedded in membranes. Several different forms of SecA in the cell have been reported (46). It has been reported (24,25) that a 30-kDa domain of SecA becomes "proteaseinaccessible" during protein translocation and that this 30-kDa fragment could become "protease-accessible" again by chasing it with excess nonradioactive SecA. Based on these data, the  1-6). The upper two-thirds of the centrifuge supernatant were precipitated with 8% trichloroacetic acid, rinsed with 1 ml of ice-cold acetone, dried, dissolved in 30 l of sample buffer, and analyzed by SDS-PAGE and autoradiography (lanes 7-12). BSA were omitted in translocation assays because it has significant protective effects. The positions of molecular size markers as well as the intact SecA are indicated. The 36-kDa fragment in the supernatant fraction appeared as an diffused band because it overlapped with the huge band of proteinase K. B, various amounts of 35 S-proOmpA were added to the translocation buffer containing the energy source. Translocation reaction, proteinase K digestion (0.5 mg/ml), and SDS-PAGE analysis were performed as described in Fig. 4. The amounts of the 66-and 48-kDa SecA fragments as well as translocated OmpA were quantitated. authors proposed a model in which membrane-associated SecA undergoes cycles of membrane insertion and deinsertion during protein translocation. However, our data do not support this model in our system. Instead, we found more than six SecA fragments after proteolysis of the reconstituted membranes (Fig. 8A). Similar proteolytic patterns were also observed when native membranes (detected by immunoblot) or membranes reconstituted from urea-washed membranes and 35 S-SecA were used. 3 The 28-kDa fragment observed here may correspond to the 30-kDa fragment documented by Economou and Wickner (24), since it increased in the presence of proOmpA, although the increase was much smaller (Ͻ3-fold) than reported. However, the 28-kDa fragment only accounts for a small fraction of total proteinase K-resistant SecA fragments on the membranes, and the majority of the 28-kDa fragment is in the supernatant regardless of translocation status (Fig. 8A). The differences in the proteolytic patterns might come from the different experimental conditions. Economou and Wickner (24) reconstituted their membranes by incubating 125 I-SecA with urea-washed membranes on ice in the absence of ATP, and collected membranes after proteolysis by trichloroacetic acid precipitation, which may include soluble protease-resistant SecA fragments. On the other hand, we reconstituted membranes by incubating 35 S-SecA with SecA-depleted membranes at 30°C in the presence of ATP, and collected the membranes by centrifugation. However, even when the reconstitution, translocation, proteolysis, and membrane collection were performed following the described procedures (24), all seven SecA fragments were still observed although the 28-and 25-kDa fragments became more prominent. 3 The other difference is in the labeling of SecA. We used a uniform [ 35 S]Met labeling in vivo, whereas Economou and Wickner (24) labeled their SecA with 125 I in vitro, which labels only surface-accessible tyrosine residues. Since most tyrosine residues will be embedded in the interior of soluble SecA, and the labeling efficiency is calculated as about 1% from the published data (24), it is most likely that only 1 out of the total 25 tyrosine residues in the SecA was labeled. Therefore, only fragments containing this 125 I-labeled tyrosine residue became detectable. The 66-and 48-kDa fragments may not contain the labeled tyrosine residues, thus, missing from their proteolytic patterns. However, we have no satisfactory explanation for the different observation on the 28-kDa/30-kDa fragments. It should be noted that in our hand, the 28-and 25-kDa fragments appear to be intrinsically resistant to proteinase K (1 mg/ml) upon the incubation of SecA with ATP in the presence of membranes.
SecA has been proposed to function as a dimer (47,48). It is possible that integral SecA exists as a dimer in the membranes, since it is active in protein translocation. The presence of the constant 90-kDa and the 48-kDa SecA fragments (Fig. 8A) suggests that integral SecA is deeply embedded in the membrane. Thus, it is likely that there is a fraction of SecA dimer deeply embedded in the membranes that give rise to 90-and 48-kDa fragments upon protease digestion. These domains of SecA may form a part of the protein-conducting channels (33,34). Indeed, SecA has been shown to be the nearest neighbor of preproteins during translocation (49). Therefore, SecA on the membranes may play a structural role as well as a catalytic role for protein translocation across the inner membrane. The structural role of the SecA may also provide an explanation for the findings that SecY-deficient membranes and SecE-deficient membranes are active in protein translocation in vitro (14,50,51). 4,5 Watanabe and Blobel (14) have reported that the urea-or heparin-extracted membrane vesicles, which contained only membrane-integral SecA, were fully active in protein translocation in the presence of F 1 -ATPase. In addition, they showed that reconstituted proteoliposomes from detergent extracts of heparin-extracted membrane vesicles containing only integral SecA but no F 1 -ATPase were fully active in protein translocation. The significance of this finding, however, has been questioned (33), since the proteoliposomes also lack SecY, an "indispensable" component of the translocation machinery (52). To address these concerns, we carried out heparin extraction with SecA reconstituted membranes and native membranes and confirmed that the extracted membranes are active (Fig. 7). The amount of SecY is not affected by the extraction and there is no F 1 -ATPase in the membranes. Thus, the translocation activity of these membranes is indeed due to the integral SecA. F 1 forms complex with F 0 on the wild-type membrane, blocking the proton flow through the F 0 channel (53). When F 1 is removed, the membranes become permeable to protons (54), possibly resulting in the collapse of proton motive force and loss of translocation activity (55). Restoration of translocation activity of urea-treated membranes by adding back soluble F 1 -ATPase (14) may occur through restoration of the proton motive force, which helps protein translocation under certain conditions (35,42,56,57). Since we used unc Ϫ membranes, the Na 2 CO 3 and the heparin treatments do not disrupt the proton motive force due to the F 1 F 0 ATPase. Therefore, F 1 -ATPase is not required for the translocation activity of the reconstituted membranes in our system. Watanabe and Blobel (14) also found that heparintreated membranes containing less SecA are as active as the control membranes. However, we found that the translocation activity levels of heparin-treated membranes are lower than the control membranes, but are the same when normalized to the remaining SecA. This difference probably comes from the amount of precursor proteins used in the in vitro systems. The amount of SecA remaining on the membranes after heparin extraction may be sufficient for maximum translocation of the nascent precursor proteins whose amount is usually small. On the other hand, when purified precursors were used, there are usually more precursor molecules than SecA in the in vitro system. As a result, although adding soluble SecA to our reconstituted membranes did not enhance translocation of the nascent proOmpA (Fig. 1), it did increase the translocation efficiency when purified proOmpA was used (Fig. 4). Thus, our data support and extend these previous findings that integral SecA is active. Moreover, since each integral SecA molecule can support several cycles of OmpA translocation, SecA cycling on and off the membrane is clearly not an obligatory step for protein translocation.
Whether or not cytosolic SecA plays an essential role in protein translocation is controversial. SecA has been reported to interact with the precursor proteins (22,23,58). Additionally, soluble SecA/SecB complex has been detected in vivo and was shown to be involved in targeting precursor proteins to the membrane (59). On the other hand, proOmpA or proOmpA/ SecB complex has been shown to bind to SecA at membranes (15) and the SecA/SecB cascade is not essential for protein translocation, since SecB only binds to a subset of preproteins. It is also known that soluble SecA is not required for translocation if sufficient SecA is bound to the membranes (11,14). Recently, Kim et al. (33) have shown that a strain with all its SecA on the membrane displayed normal translocation of preproteins in vivo. Our in vitro data shows that soluble SecA exchanges with the loosely bound SecA regardless of protein translocation status (Figs. 3 and 4). What then, is the function of the cytosolic or free SecA? One possible function of cytosolic SecA is to shuttle the SecB/precursor from cytosol to membranes. Another known function for the cytosolic SecA is in the repression of its own synthesis (60). SecA has been shown to compete with ribosomes for the ribosome binding site of its own mRNA (61). It has been suggested that the accumulation of preproteins when translocation is impaired could be sensed by soluble SecA, leading to the derepression of SecA synthesis (62,63). The increased soluble SecA levels, in turn, could compensate for the impaired Sec components, such as mutant SecY (32,62). Furthermore, SecA shares sequence homology with RNA helicase (64) and binds to a variety of preprotein mRNAs (65). Therefore, soluble SecA may play a more important role in the regulation of protein expression.