SecA binds to the inner membrane of Escherichia coli through low affinity lipid interactions or with high affinity at SecYEG, the integral domain of preprotein translocase. Upon addition of preprotein and nucleotide, a 30 kDa domain of SecYEG‐bound SecA is protected from proteolysis via membrane insertion. Such protection could result from some combination of insertion into the lipid phase, into a proteinaceous environment or across the membrane. To assess the exposure of SecYEG‐bound SecA to membrane lipids, a radiolabeled, photoactivatable and lipid‐partitioning crosslinker, 3‐trifluoromethyl‐3‐(m‐[125I]iodophenyl) diazirine benzoic acid ester, was incorporated into inner membrane vesicles. The 30 kDa domain of SecYEG‐bound SecA, inserted into the membrane in response to translocation ligands, is 18‐fold less labeled than SecY, which is labeled effectively. In contrast, incorporation of the purified 30 kDa SecA fragment into crosslinker‐containing detergent micelles or addition of detergent to crosslinker‐containing membranes bearing the protease‐protected SecA domain readily allows for labeling of this domain. We propose that the protease‐inaccessible 30 kDa SecA domain is shielded from the fatty acyl membrane phase by membrane‐spanning SecYEG helices and/or is largely exposed to the periplasm.
Pathways of protein targeting and translocation out of the cytoplasm into various compartments have been examined in various eukaryotic organelles, including the mitochondria, nucleus and endoplasmic reticulum, as well as in bacteria (Schatz and Dobberstein, 1996). With their unparalled ease of combining biochemistry and genetics, bacteria offer unique advantages in addressing questions of protein translocation (Wickner et al., 1991; Pugsley, 1993; Arkowitz and Bassilana, 1994; Ito, 1996). Most bacterial protein export occurs post‐translationally (Randall, 1983; Zimmermann and Wickner, 1983). To avoid misfolding or other undesired associations, newly synthesized secretory preproteins may associate with chaperones such as DnaJK, GroELS, trigger factor, SRP and the export‐specific SecB chaperone (Kumamoto, 1989; Luirink et al., 1992; Randall and Hardy, 1995), which maintains these preproteins in a translocation‐competent, loosely folded state (Weiss et al., 1988; Lecker et al., 1989). The SecB–preprotein complex then binds to SecA, the peripheral subunit of preprotein translocase (Hartl et al., 1990), which in turn is bound to the membrane‐embedded domain of preprotein translocase, the SecYEG complex (Brundage et al., 1990). SecA, when in contact with a preprotein, SecYEG and acidic phospholipids, binds ATP, leading to the translocation of 20–30 residues of the preprotein across the membrane. Hydrolysis of the bound ATP then releases the preprotein from SecA, allowing the electrochemical proton gradient (proton motive force) to drive further translocation. Translocating proteins undergo many such cycles of SecA binding, ATP binding to SecA, limited translocation, ATP hydrolysis and preprotein release from SecA, and proton motive force‐driven translocation (Tani et al., 1990; Schiebel et al., 1991). Other proteins, such as SecD and SecF, have also been implicated in preprotein translocation (Matsuyama et al., 1993; Economou et al., 1995). Finally, the translocated protein is cleaved from its leader peptide by the action of leader peptidase, thereby releasing the protein from the outer surface of the inner membrane (Dalbey and Wickner, 1985).
While most, if not all, of the proteins which catalyze preprotein translocation have been identified, many mechanistic questions remain unanswered. One of the most fundamental asks how SecA couples the energy of ATP binding to the translocation of preprotein across the bacterial inner membrane. In the presence of SecYEG, preprotein and bound ATP, SecA (102 kDa) has been shown to insert a 30 kDa domain into the membrane, as detected by inaccessibility of this domain to proteases from the cytoplasmic surface of the membrane (Economou and Wickner, 1994). Since ATP binding energy drives the translocation of 20–30 preprotein aminoacyl residues (Schiebel et al., 1991) and, as shown by cross‐linking techniques, the preprotein crosses in association with SecA throughout its membrane passage (Joly and Wickner, 1993), the 30 kDa domain of SecA may actually form a part of the translocation pathway. It is, therefore, of great interest to define the environment encountered by the inserted 30 kDa domain of SecA. SecA could insert either into the lipid phase of the membrane, into a proteinaceous milieu or largely across the membrane into the aqueous periplasm. Each of these possibilities carry significant implications for the route taken by the preprotein through the membrane. In the present study, we have used a lipid‐soluble, radioiodinated, photoactivatable cross‐linking agent, 3‐trifluoromethyl‐3‐(m‐[125I]iodophenyl) diazirine benzoic acid ester ([125I]TID/BE), to determine whether the membrane‐inserted 30 kDa protease‐protected domain of SecA is exposed to membrane lipids.
[125I]TID/BE labels proteins from the lipid phase of the membrane
A photoactivatable cross‐linking reagent embedded in, and restricted to, the hydrophobic interior of the membrane provides a direct method for the detection of interactions between a protein and the lipid core of a membrane. [125I]TID/BE (Figure 1) is an improved, less volatile version of 3‐trifluoromethyl‐3‐(m‐[125I]iodophenyl) diazirine ([125I]TID), a compound widely used to selectively label membrane‐embedded segments of integral proteins (Brunner, 1989, 1993, 1996; Durrer et al., 1995; Weber and Brunner, 1995). To ascertain that [125I]TID/BE can label membrane‐embedded proteins effectively, inner membrane vesicles (IMVs) were prepared from Escherichia coli KM9, incubated with [125I]TID/BE and exposed to UV light (Figure 2). Analysis by SDS–PAGE and autoradiography revealed the labeling of numerous proteins (lane 2). In the absence of irradiation, no labeling was detected (lane 1). Sedimentation of the [125I]TID/BE‐containing IMVs confirmed that >97% of the radiolabel was confined to the membrane pellet. When the remaining ≈3% of radiolabel, found in the supernatant, was incubated with fresh IMVs and subsequently UV irradiated, no protein labeling could be detected by SDS–PAGE and autoradiography (not shown).
To verify that membrane‐embedded translocation‐related proteins could by labeled by [125I]TID/BE in IMVs, membrane vesicles were prepared from cells genetically altered to overexpress either SecYEG or SecDF (Pogliano and Beckwith, 1994; Douville et al., 1995). UV irradiation of SecYEG‐enriched IMVs (lane 4) revealed heavy labeling of several bands as compared with the BL21 background strain (lane 3). The most prominent were of apparent molecular weights of 32, 14 and 12 kDa (lane 4), corresponding to SecY, hemagglutinin (HA)‐tagged SecE and SecG, respectively (Douville et al., 1995). Similarly, bands corresponding to the molecular weights of SecD (60 kDa) and SecF (30 kDa) were prominently labeled in the SecDF‐enriched IMVs (lane 5).
Experiments were performed to verify that SecA could be labeled by [125I]TID/BE. Purified SecA was incubated with a 0.1% SDS–[125I]TID/BE mixture. After sonication, the SecA–SDS–[125I]TID/BE micelles were UV irradiated. Subsequent examination by SDS–PAGE and autoradiography clearly shows cross‐linker labeling of SecA (Figure 3, lane 1). To confirm that the 30 kDa domain of SecA could be labeled by the cross‐linker, purified SecA was first subjected to limited trypsinolysis. Such controlled proteolysis leads to the generation of a defined number of peptide fragments, including one (closed arrow) shown to be identical to the 30 kDa SecA domain protected from proteolysis in the presence of IMVs, preprotein and nucleotide (Price et al., 1996). When the fragments of this limited trypsin treatment were mixed with [125I]TID/BE‐containing SDS micelles, labeling of the fragments, including the 30 kDa fragment, could be observed (lane 2). The prominent 65 kDa fragment (open arrow) was labeled 8.2 times more intensely than the 30 kDa fragment. This ratio, however, simply reflects the relative mass and abundance of the two fragments in the tryptic digest and not preferential labeling of the SecA 65 kDa fragment, since a similar pattern was obtained upon limited proteolysis of metabolically labeled [35S]SecA (lanes 3 and 4), where sulfur‐containing amino acid residues are distributed approximately proportionally in the 65 and 30 kDa SecA fragments (17 and 11 residues, respectively) (Schmidt et al., 1988; Price et al., 1996). Quantitative analysis of the distribution of the 35S label in trypsin‐treated [35S]SecA revealed a 3.7:1 molar ratio in favor of the 65 kDa fragment and there is a mass ratio between the two SecA fragments of 2.2:1, yielding a predicted labeling ratio of 8.1, in good agreement with the observed ratio of [125I]TID/BE labeling.
Incorporation of cross‐linker does not compromise translocation function
To determine whether the cross‐linker affects translocation functions, IMVs pre‐mixed with TID/BE were compared with untreated membranes in terms of their abilities to translocate the 35S‐labeled precursor form of outer membrane protein A (proOmpA) into the vesicular lumen. Translocation reactions used 75 μg/ml proOmpA, in substantial excess over the Km (Crooke et al., 1988b), and thus reflected maximal velocities. Over the course of a 45 min incubation, proOmpA was translocated into the lumen of control IMVs and thus protected from digestion by externally added proteinase K (Figure 4A, lanes 1–5). Translocated proOmpA was partially cleaved by leader peptidase at the periplasmic membrane surface to yield OmpA. Incubation of the membranes with the cross‐linker had no apparent effect on either the rate or extent of [35S]proOmpA translocation (lanes 6–10).
The effect of TID/BE incorporation on the ability of the membranes to protect a 30 kDa domain of SecA from proteolytic digestion was also examined (Figure 4B). When either control membranes or membranes incorporating cross‐linker were incubated with [125I]SecA, proOmpA and ATP, a 30 kDa domain of SecA was protected from trypsinolysis. In both membrane preparations, omission of either preprotein or nucleotide resulted in digestion of the SecA domain (compare lanes 1 and 6 with lanes 3 and 4, and 8 and 9, respectively). The protected 30 kDa fragment could be generated at enhanced levels using the non‐hydrolyzable ATP analog adenylyl‐imidophosphate (AMPPNP) (lanes 2 and 6), as previously observed (Economou and Wickner, 1994; Economou et al., 1995). Thus the generation of the protease‐protected 30 kDa SecA domain was unaffected by the presence of the cross‐linker and required the same conditions using either membranes incorporating cross‐linker or untreated membranes.
Efficiency of 30 kDa fragment formation
To verify that maximal levels of the 30 kDa protease‐protected fragment were being formed, we determined the efficiency of 30 kDa SecA domain insertion into IMVs. Protease‐protected 30 kDa fragment was formed using [125I]SecA, after incubation in the presence of proOmpA, ATP and increasing amounts of unlabeled SecA. The 30 kDa fragment was assayed densitometrically and expressed as a percentage of the densitometric units formed when only radiolabeled SecA was used. To translate these percentages into picomoles of 30 kDa fragment, it was first necessary to determine the fraction of radioactive label in [125I]SecA that is found in the 30 kDa domain. To do so, a mixture of 125I‐labeled and unlabeled SecA was subjected to limited trysinolysis in solution, leading to the generation of the 30 kDa fragment (see above). Comparison of protein staining of the digestion fragments with the corresponding autoradiographic pattern shows that the radioactivity in [125I]SecA is recovered almost exclusively in the 30 kDa domain, with little in the other SecA fragments of higher molecular weight (compare Figure 5A, lanes 1 and 3 with lanes 2 and 4). To examine more precisely the distribution of radioactive iodine in [125I]SecA, two SecA samples were prepared, one containing unlabeled protein and the other a mixture of unlabeled SecA and a tracer amount of [125I]SecA. The two SecA preparations were divided in half and one portion of each preparation was subjected to limited proteolysis. The untreated and trypsin‐digested samples from each SecA preparation were then separated by SDS–PAGE, electrophoretically transferred to polyvinylidene difluoride (PVDF) membranes and stained with Coomasie brilliant blue. Protein bands corresponding to full‐length SecA and the 30 kDa fragment were excised and either quantitated by amino acid analysis (for those samples in which unlabeled SecA served as starting material) or in terms of radioactive counts (for those samples in which trace amounts of radiolabeled SecA were included in the starting material). Both SecA and the 30 kDa fragments displayed similar specific activities (1186 and 1641 c.p.m./pmol, respectively), confirming that the radioactivity in [125I]SecA is essentially confined to the 30 kDa domain. This observation was used to calculate the amount of 30 kDa fragment protected from protease relative to the starting [125I]SecA. The amount of SecA 30 kDa fragment formed over a range of added SecA concentrations is shown in Figure 5B. At saturation of the ability of the membranes to form the 30 kDa fragment (i.e. using a starting SecA concentration of at least 200 nM), ∼580 pmol of 30 kDa fragment are formed per mg of IMV protein. Scatchard analysis of high affinity SecA binding revealed 490 pmol of SecYEG sites per mg of IMV protein (not shown), a value in good agreement with previous estimates (Hartl et al., 1990; Bassilana and Wickner, 1993; Economou and Wickner, 1994). Therefore, under the conditions employed, essentially all high affinity, SecYEG‐bound SecA generates a 30 kDa protease‐protected fragment.
[125I]TID/BE readily labels the SecYEG complex but not the 30 kDa protease‐protected domain of SecA
Having determined that TID/BE does not interfere with the formation of the 30 kDa SecA fragment, that it could cross‐link to the 30 kDa domain if given access and that maximal levels of the fragment are being produced, experiments were performed to determine whether the protease‐protected SecA 30 kDa domain was exposed to the lipid phase of the membrane and thus accessible to [125I]TID/BE labeling. IMVs were pre‐incubated with [125I]TID/BE and used to prepare the SecA 30 kDa protease‐protected fragment, then UV irradiated. No labeled 30 kDa protease‐protected band could be detected by direct analysis of SDS–PAGE fluorograms of samples containing both proOmpA and ATP or in samples lacking either of these components (Figure 6, lanes 1–3; solid dots mark the position of the SecA 30 kDa domain). Other cross‐linker‐labeled peptide fragments, derived from trypsinized membrane proteins, were however observed. Control experiments using [125I]SecA confirmed the ability of the cross‐linker‐treated membranes to support formation of the 30 kDa fragment in the presence of proOmpA and ATP (Figure 4; Figure 6, lane 4).
Immunoprecipitation of the SecA 30 kDa domain was performed to examine whether even a low level of labeling of the protease‐protected domain had occurred. Immunoprecipitation of the SecA 30 kDa domain was found to be ∼45% efficient (Figure 6, lanes 4 and 5), and was not compromised either by the presence of [125I]TID/BE or by UV treatment (not shown). When samples containing the 30 kDa domain prepared in UV‐irradiated, [125I]TID/BE‐containing membranes were subjected to immunoprecipitation, only a very faint labeled 30 kDa band was seen (lane 8). As expected, the intensity of the 30 kDa band was lower in samples in which either ATP or proOmpA were omitted (lanes 9 and 10, respectively). The anti‐SecA immunoprecipitations also captured additional labeled bands of higher molecular weights. The presence of these extra bands cannot, however, be correlated to translocation, since their appearance did not require preprotein (compare lanes 8 and 10), SecYEG or physiological temperature (not shown). These bands were captured specifically by the anti‐SecA serum and were not detected in the absence of added SecA, membranes or trypsin (not shown), suggesting that they may originate from SecA molecules which at least partially insert into the lipid phase of the membrane.
Essentially all SecYEG‐bound SecA is able to protect a 30 kDa domain from externally added protease in the presence of preprotein and nucleotide (see above). Since the SecA 30 kDa domain is only very poorly labeled by the phospholipid phase‐restricted cross‐linker, we determined the extent of [125I]TID/BE labeling of the SecYEG complex, as detected by non‐denaturing immunoprecipitation of the SecYEG complex using anti‐SecG antibodies. With its 10 transmembrane domains, SecY offers an ∼23 kDa target for potential labeling by the membrane‐based [125I]TID/BE (213 of 443 amino acid residues; Ito, 1992), a target size similar to the SecA protease‐protected 30 kDa domain. The efficiency of the immunoprecipitation of SecY was calculated using [125I]TID/BE‐labeled IMVs prepared from a SecYEG‐overexpressing strain (Douville et al., 1995) as in Figure 2, lane 5. Comparison of total labeled SecY with immunoprecipitated SecY revealed an efficiency of ∼20%, as determined by densitometric analysis (Figure 6, lanes 6 and 7; asterisks mark the position of SecY, diamonds and triangles mark the positions of HA‐SecE and SecG, respectively). This factor was then used to calculate equivalent autoradiographic exposure times for the SecA 30 kDa domain and SecY labeled in the same [125I]TID/BE‐treated membranes. Upon proportional exposures to correct for differences in immunoprecipitation efficiencies, SecY is seen to be labeled (lane 11) far better than the 30 kDa SecA fragment (lane 8). Indeed, only after an exposure that was 18‐fold longer (corrected for immunoprecipitation efficiencies) did the labeled intensity of the 30 kDa SecA fragment equal that of SecY.
Finally, experiments were performed to determine whether there were inherent differences in the ability of the photoactivated [125I]TID/BE to react with SecY or the 30 kDa SecA domain. Since, as shown above (Figure 5B), 14 pmol of membrane‐inserted SecA forms under the translocation conditions employed (Figure 6, lane 8), this amount of purified SecA 30 kDa domain was mixed with a micellar solution of [125I]TID/BE in SDS. Upon irradiation and subsequent exposure for a time period similar to that required for detection of SecY (taking into account the efficiency of SecY immunoprecipitation), intensive labeling of the 30 kDa SecA domain was observed (Figure 6, lane 12). Labeling of the purified SecA 30 kDa domain was far more than seen in the intact membrane (lane 8) and comparable in intensity with the labeling of 14 pmol of SecY (lane 11). Further confirmation of the inherent ability of the SecA 30 kDa domain to be labeled by [125I]TID/BE, given access to the cross‐linker, came from experiments in which the protease‐protected 30 kDa domain was formed in [125I]TID/BE‐containing IMVs which were then left untreated or incubated with 0.01% SDS prior to UV treatment (Figure 7). Upon such detergent treatment, labeling of the protease‐protected SecA 30 kDa domain was 15‐fold more intense as compared with labeling of the domain in the intact IMVs (compare lanes 1 and 2). In contrast, the detergent treatment did not significantly affect [125I]TID/BE labeling of SecY (compare lanes 3 and 4).
In the presence of preprotein, nucleotide and SecYEG‐containing membranes, SecA is able to protect a 30 kDa domain from proteolytic digestion through insertion into the membrane (Economou and Wickner, 1994). In light of the fact that ATP binding energy is involved in both the membrane insertion of this SecA 30 kDa domain and in the translocation of preprotein segments (Schiebel et al., 1991) and that SecA is adjacent to the membrane‐spanning segment of a translocating preprotein (Joly and Wickner, 1993), it is likely that this protease‐protected 30 kDa SecA domain plays a central role in translocation. In the present study, we have measured the exposure of this inserting domain of SecA to the lipid phase of the membrane.
Although no hydrophobic stretches of significant length are found in its sequence (Schmidt et al., 1988), the concept of SecA insertion into membranes is well established. SecA exists in both soluble and membrane‐associated forms in the cell (Cabelli et al., 1991). Liposome association of SecA appears to involve a specific and deep insertion into the lipid phase, an event which induces major conformational changes in SecA (Ulbrandt et al., 1992). SecA can also insert into phospholipid monolayers as well as cause aggregation of liposomes in solution, although each of these interactions was inhibited by ATP (Breukink et al., 1992, 1993). While revealing an intimate relationship between SecA and membrane lipids, the physiological significance of these observations is, none the less, unclear, since translocation requires not only acidic phospholipids, but also preprotein, ATP and SecYEG (Brundage et al., 1990; Lill et al., 1990; Hendrick and Wickner, 1991; Douville et al., 1995). Only SecA membrane insertion that leads to the generation of a protease‐protected 30 kDa fragment accurately reflects the requirements for these other translocation components. The focus of this study was to determine whether this insertion occurs into the lipid core of the membrane or, alternatively, into a protein environment within the membrane.
Direct experimental evidence for the interaction of a protein with the lipid phase of a membrane can be obtained using photoactivatable cross‐linking reagents (cf. Brunner, 1993). In the current studies, [125I]TID/BE was used to determine whether the inserted 30 kDa SecA domain is in contact with the lipid phase of the membrane. As with [125I]TID, [125I]TID/BE readily partitions into the lipid phase of the membrane, yet does not escape readily from the apolar membrane interior due to the presence of the benzoate ester group, making it a useful tool for the identification of lipid‐inserted protein domains (Durrer et al., 1996). Incorporation of [125I]TID/BE into IMVs was shown to label membrane‐associated proteins, including the various Sec proteins, effectively and to not interfere with translocation functions of the membranes. Moreover, [125I]TID/BE labels SecA and its 30 kDa membrane‐inserting domain if given access to these polypeptides. Indeed, labeling of the purified SecA 30 kDa domain was then as effective as the labeling of SecY (Figure 6), confirming that SecA does not suffer any inherent handicap in terms of membrane‐based labeling potential.
SecYEG serves as the high affinity binding site of SecA and subsequently is required for SecA membrane insertion and protection of the 30 kDa domain from external protease. Essentially all SecYEG‐bound SecA inserted into the membrane (Figure 5), and effectively all of the SecY polypeptide in the IMVs prepared from E.coli KM9 and used in these experiments is found in translocation‐active SecYEG complexes (Bassilana and Wickner, 1993). Hence, [125I]TID/BE cross‐linker labeling of SecYEG and the 30 kDa protease‐protected SecA domain should be comparable, given equal lipid exposure of the two translocase components. When SecA insertion reactions were performed in [125I]TID/BE‐treated IMVs and labeling of both SecY and the SecA 30 kDa domain were compared in the same experiment, only very weak labeling of the 30 kDa SecA fragment was observed relative to SecY. Quantitative comparison of these labeling intensities revealed that SecY is 18‐fold better labeled than is the SecA 30 kDa domain, suggesting that the SecA region is far less exposed to the membrane lipids than is SecY. The weak labeling of the SecA 30 kDa domain cannot be due to a lack of sensitivity of the detection method, since strong labeling of the purified 30 kDa SecA fragment in [125I]TID/BE‐containing detergent micelles could be detected, using amounts of the purified SecA fragment comparable with the amount of SecY present in the IMVs. Furthermore, upon addition of SDS to [125I]TID/BE‐containing IMVs bearing the membrane‐inserted SecA domain prior to UV treatment, labeling of the protease‐protected SecA 30 kDa domain was readily detected.
In addition to the weak labeling of the protease‐protected SecA 30 kDa domain by [125I]TID/BE, cross‐linking to other SecA fragments was observed (Figure 6). In an attempt to identify these proteolytic fragments, SecA insertion experiments were performed with metabolically labeled [35S]SecA (J.Eichler and W.Wickner, submitted). Use of [35S]SecA addresses the uneven distribution of radiolabel in [125I]SecA (Figure 5), thereby allowing better characterization of the relationship between SecA domains and the membrane. Indeed, [35S]SecA insertion studies served to distinguish between translocation‐related membrane‐inserting SecA domains and lipid‐associated SecA fragments which form in the absence of SecYEG, preprotein or physiological temperature. It is noteworthy that three of the [35S]SecA fragments derived from lipid‐associated molecules (J.Eichler and W.Wickner, submitted) are of similar molecular weight as the non‐translocation‐related bands labeled by [125I]TID/BE in the present study.
The fact that the 30 kDa protease‐protected domain is labeled, albeit weakly, supports the notion that protease protection is the result of membrane insertion and not due to some internal refolding of the SecA protein upon its binding to SecYEG in the presence of preprotein and nucleotide. If, therefore, the inserted SecA 30 kDa domain is mostly inaccessible from the membrane lipids, it must either be largely surrounded by a protein sleeve or reside largely across the membrane in the periplasm, with only a small proportion of the domain lying within the plane of the membrane. Given the presumably high energy barrier to insertion of the largely hydrophilic SecA 30 kDa domain into the hydrophobic environment of the lipid interior of the membrane, it is reasonable that the protease‐protected domain of SecA inserts into or through a proteinaceous milieu. Tokuda and co‐workers recently have proposed a model in which SecA insertion is coupled to an inversion of SecG topology in the plane of the membrane, detected as a reversible inability to detect an antigen located at the COOH‐terminus of the protein (Nishiyama et al., 1996). This SecG membrane inversion only occurred under conditions similar to those which lead to the generation of the protease‐protected SecA 30 kDa domain. Exchange of ATP for its non‐hydrolyzable analog AMPPNP during the course of a reaction prevented the recovery of native SecG topology; use of AMPPNP also prevents de‐insertion of inserted SecA (Economou and Wickner, 1994; Economou et al., 1995). In light of similarities between SecA insertion and SecG topological inversion, it is tempting to speculate that SecG participates in the protein envelope that serves to mask the inserted 30 kDa domain of SecA from membrane lipids and concomitant [125I]TID/BE labeling, as described in the present study. This is reminiscent of the topology proposed for the membrane proteins of the histidine permease from Salmonella typhimurium, where it is hypothesized that the hydrophobic integral membrane proteins HisQ and HisM form a protein channel in which hydrophilic HisP subunits, shielded from membrane lipids, are thought to reside (Kerppola et al., 1991). Studies examining [125I]TID/BE labeling of the protease‐protected SecA 30 kDa domain in IMVs prepared from a modified KM9 strain lacking SecG did not, however, show an increase in cross‐linking to the SecA fragment (F.Duong and J.Eichler, unpublished data). Furthermore, the weak, yet detectable, cross‐linker labeling of the inserted SecA 30 kDa domain suggests that protection of this region from the lipid phase of the membrane by SecG or other proteins is not absolute and that some small proportion of the inserted domain is in contact with the lipids.
Closer inspection of the labeling pattern of the three components of SecYEG (Figure 2; Figure 6, lane 6) reveals that SecG was labeled more weakly than SecE, despite their similar predicted transmembrane contents (6 kDa in the case of SecE and 4 kDa in the case of SecG) (Schatz et al., 1989; Nishiyama et al., 1996). Analogous results were obtained using purified SecYEG reconstituted into proteoliposomes (not shown). This uneven distribution of label may reflect partial shielding of this translocase subunit from membrane lipids. This hypothesis may be relevant, considering the inversion of SecG membrane topology in response to the cycle of SecA insertion and de‐insertion (Nishiyama et al., 1996). The energetic cost of SecG ‘flipping’ across the plane of the membrane could be lower in a proteinaceous rather than in a hydrophobic lipaceous environment.
As a first step in defining the environment encountered by the membrane‐inserting SecA 30 kDa domain, we have examined the lipid exposure of the protease‐protected SecA 30 kDa domain using a lipid‐restricted, carbene‐generating and radiolabeled cross‐linker. Use of such probes to examine lateral accessibility of a protein to the fatty acyl phase of the membrane is strictly parallel to the use of water‐soluble probes to measure accessibility of a protein to the membrane surface. When proteins are inaccessible for labeling, positive controls are required to demonstrate that other, label‐accessible proteins can be readily labeled in intact membranes and that the inaccessible proteins can react with the probe upon disruption of membrane integrity. These are the classical criteria of membrane protein topology (Bretscher, 1971), previously used to study protein translocation and SecA membrane insertion/de‐insertion. We have now applied the same principles, using [125I]TID/BE, to study accessibility of the protease‐protected 30 kDa domain to the lipid phase of the membrane. Little of the SecA 30 kDa membrane‐inserted domain is exposed to the cross‐linker from the fatty acyl phase of the membrane, though it is readily labeled in detergent‐mixed micelles. We cannot discount the possibility that the presence of detergent modifies the conformation of the ‘native’ protease‐protected SecA 30 kDa domain in such a way as as to make it more amenable to labeling by the cross‐linker. The carbene‐generating [125I]TID/BE cross‐linker, however, is believed to be highly reactive without expressing sequence preferences (Brunner, 1996). SecA contains no apolar stretches and as such should not be labeled differentially at different sites by [125I]TID/BE (Schmidt et al., 1988). With the sequence of the SecA 30 kDa domain recently defined (Price et al., 1996), topological studies of the environment encountered by this SecA domain within the membrane can continue. Such investigations could make use of a variety of experimental techniques, including spin‐labeling, fluorescence quenching, modification of specific amino acid residues to introduce cross‐linking entities or in vitro translation to incorporate cross‐linker‐bearing amino acids into the SecA 30 kDa domain, and will lead to a more precise description of the immediate surroundings of the protease‐protected SecA 30 kDa domain.
Materials and methods
SecA (Cunningham et al., 1989), SecB (Weiss et al., 1988) and proOmpA (Crooke et al., 1988a) were prepared as described. AMPPNP, ATP, bovine serum albumin (BSA), dithiothreitol (DTT), soybean trypsin inhibitor and trypsin were from Sigma (St Louis, MO), creatine kinase, creatine phosphate and proteinase K from Boehringher‐Mannheim (Indianapolis, IN), protein A–Sepharose beads from Pharmacia (Uppsala, Sweden), ‘Reactivials’ from Pierce (Rockford, IL), silica gel 60 TLC plates from Merck (Rahway, NJ) and [125I]Na from Amersham (Arlington Heights, IL). [35S]proOmpA (Crooke and Wickner, 1987) and [125I]SecA (Economou and Wickner, 1994) were prepared as described.
Inner membrane vesicles
IMVs were prepared as previously described (Douville et al., 1995) from E.coli strains KM9 (unc‐::tn10, relA1, spoT1, metB1, Klionsky et al., 1984), BL21/pHAsecEYG, an overproducer of SecYEG (Douville et al., 1995), or BL21/pGAP1, an overproducer of SecDF (Pogliano and Beckwith, 1994). The membranes were treated with 6 M urea (30 min, 0°C) to remove and inactivate endogenous SecA (Cunningham et al., 1989).
The cross‐linking agent [125I]TID/BE was prepared by radioiododestannylation of the tin‐based precursor 4′‐(3‐trifluoromethyl‐3H‐diazirin‐3‐yl)‐2′‐tributylstannylbenzyl benzoate (TTD/BE) (Durrer et al., 1995; Weber and Brunner, 1995). Briefly, 50 nmol of TTD/BE was added to a 1 ml ‘Reactivial’ and dried under nitrogen. The dried material was dissolved in 35 μl of acetic acid to which 10 mCi of [125I]Na were added. Iodination was initiated upon addition of 10 μl of peracetic acid and allowed to continue for 2 min at room temperature. The reaction was quenched by adding 10 μl of NaI (100 mM) and 200 μl of sodium bisulfite [10% (w/w)]. The labeled compound was extracted into 200 μl of ethyl acetate by vortexing. The lower phase was removed and discarded, while the ethyl acetate upper phase was evaporated under nitrogen and the residue dissolved in 200 μl of diethyl ether. The solution was then applied to a silica gel 60 TLC plate and developed with ether:hexane (1:9). The position of TID/BE prepared with non‐radioactive iodine was examined under a 256 nm UV light and served as a reference point for the migration of the radiolabeled compound. [125I]TID/BE was scraped from the TLC plate, dissolved in 1 ml of ethanol and filtered. The efficiency of incorporation of 125I was 25–40%. [125I]TID/BE was stored at −20°C and used within a month.
Translocation of proOmpA
The ability of IMVs [in some cases pre‐treated with up to 10% (v/v) TID/BE] to translocate proOmpA was determined by incubating 5 μg of IMVs with 2.8 μg of SecA, 4 μg of SecB, 20 μg of BSA and buffer B (50 mM KCl, 1 mM DTT, 50 mM Tris–HCl, pH 7.9) in a volume of 95 μl. To this mixture were added 5 μl of [35S]proOmpA [1.5 mg/ml, 124 000 c.p.m. in urea buffer (6 M urea, 1 mM DTT, 50 mM Tris–HCl, pH 7.9)]. Incubation at 37°C (0–45 min) was terminated by transfer to ice. Samples were digested with 10 μl of proteinase K (10 mg/ml) for 15 min. Proteolysis was terminated by addition of 15% trichloroacetic acid (TCA). After 30 min on ice, the samples were centrifuged (10 000 r.p.m., 15 min, 4°C, microfuge), resuspended in 1 ml of cold acetone and once again centrifuged. Pellets were dried (37°C, 5 min), dissolved in 30 μl of sample buffer containing β‐mercaptoethanol (Brundage et al., 1992), heated (95°C, 3 min) and examined by SDS–PAGE and fluorography.
Formation of the SecA 30 kDa domain
For formation of a radiochemical quantity of the 30 kDa protease‐protected SecA domain, 10 μl of 10× buffer B, 3 μg of SecB, water, 2 μl of [125I]SecA (100 000 c.p.m./μl), 5 μg of IMVs, 1 μl of proOmpA (1.5 mg/ml in urea buffer) and 5 μl of ATP or AMPPNP (5 mM final concentration) were added in sequence to a final reaction volume of 0.1 ml. The reaction was incubated at 37°C for 30 min for reactions with ATP, or room temperature for 5 min for reactions with AMPPNP. After transfer to ice, the samples were digested with 10 μl of trypsin (10 mg/ml) for 15 min. The samples were then TCA precipitated, acetone ‘washed’, dissolved in sample buffer and examined by SDS–PAGE as above, using either autoradiography or phosphorimaging. Phosphorimaging was performed with a Molecular Dynamics 445SI phosphorimaging apparatus, using IPLabH software (Molecular Dynamics, Sunnyvale, CA).
For formation of larger chemical amounts of the 30 kDa fragment, 10 μl of 10× buffer B, 12 μg of SecB, water, 7.8 μg of SecA, an ATP‐regenerating system [1 μg of creatine kinase and 2 μl of phosphocreatine (10 mM final concentration)], 24 μg of IMVs pre‐mixed with 0.5 μl of [125I]TID/BE (2.97×106 c.p.m./μl), 18 μg of proOmpA (in urea buffer) and 5 μl of ATP or AMPPNP (5 mM final concentration) were added in sequence to a final reaction volume of 0.1 ml and the reaction performed as described above. After transfer to ice, the samples were irradiated for 2 min using a 450 W water‐cooled mercury arc lamp (Ace Glass, Vineyard, NJ), from a distance of 10 cm. The samples were then treated with trypsin and processed for SDS–PAGE and autoradiography or phosphorimaging.
[125I]TID/BE labeling of SecA and of its 30 kDa fragment in SDS micelles
The 30 kDa protease‐protected SecA domain was generated in solution as described by Price et al. (1996). Briefly, a solution of SecA (550 μg/ml) was treated with trypsin (6.2 μg/ml) on ice for 5 min, then treated with a 20‐fold excess of soybean trypsin inhibitor. In some cases, [35S]SecA (100 000 c.p.m.) was included in the trypsinolysis reaction. The protein fragments were then TCA precipitated, acetone ‘washed’, heated in sample buffer (3 min, 95°C) and examined by SDS–PAGE as described above. After identification of the 30 kDa fragment using the Reversi‐stain system (Diversified Biotech, Boston, MA), the band was excised and electroeluted using a Schleicher and Schuell Elutrap apparatus.
For cross‐linking studies, SecA, trypsinated SecA or the electroeluted 30 kDa fragment were sonicated with SDS [0.1% (w/v) final concentration] pre‐mixed with 0.5–2 μl of [125I]TID/BE (2.97×106 c.p.m./μl), UV treated as above and examined by autoradiography or phosphorimaging.
Immunoprecipitation of the SecA 30 kDa domain or SecYEG from [125I]TID/BE‐containing IMVs was performed as follows: 100 μl of SecA insertion reactions (see above), in which proteolysis had been terminated by addition of 10 mg/ml soybean trypsin inhibitor, were incubated with 100 μl of protein A–Sepharose beads with pre‐bound anti‐SecA antibodies and 1 ml of RIPA buffer (150 mM NaCl, 1% NP‐40, 0.5% deoxycholate, 0.1% SDS, 50 mM Tris–HCl, pH 8) (Harlow and Lane, 1988) or anti‐SecG antibodies and 1 ml of non‐denaturing immunoprecipitation buffer (150 mM NaCl, 1.25% β‐octylglucoside, 1.5 mg/ml E.coli phospholipids, 40% glycerol, 50 mM Tris–HCl pH 8) (Brundage et al., 1992). For SecY immunoprecipitations, the samples were not exposed to trypsin. After 60–120 min of rocking at 4°C, the beads were collected in a microfuge and resuspended in 1 ml of their respective buffers three times. The supernatants were removed, 35 μl of sample buffer added and the samples heated (3 min, 95°C for the anti‐SecA beads; 10 min, 37°C for anti‐SecG beads). After centrifugation, the supernatants were transferred to new tubes, re‐centrifuged and examined by SDS–PAGE using ‘15%’ or ‘high Tris’ gels and autoradiography or phosphorimaging.
Protein concentrations were determined using Bradford reagent (Bio‐Rad, Hercules, CA) with BSA as a standard. SDS–PAGE was performed using either ‘15%’ gels or ‘high Tris’ gels (Douville et al., 1995). Electrophoretic transfer of proteins to either nitrocellulose (Bio‐Rad, Hercules, CA) or PVDF (Millipore, Bedford, MA) was performed in a Genie electrophoretic blotter (Idea Scientific, Minneapolis, MN) at 250 mA for 1 h. Amino acid analysis was performed using an Applied Biosystems model 477A apparatus. Densitometry was performed using a Silverscan III scanner (LaCie Ltd, Beaverton, OR) and IPLabH software. Scatchard analysis of SecA binding to IMVs was performed as previously described (Hartl et al., 1990). The preparation of [35S]SecA is described elsewhere (J.Eichler and W.Wickner, submitted).
Note added in proof
Using [35S]SecA, we have recently observed that, in addition to the 30 kDa SecA domain described here, an N‐terminal 65 kDa domain of SecA is also protected from proteolysis during translocation (J.Eichler and W.Wickner, submitted). Like the 30 kDa domain, the protease‐protected SecA 65 kDa domain is not readily accessible to the membrane lipid phase, as determined by a lack of [125I]TID/BE labeling.
We thank Peter Durrer and Roland Graf for their assistance and hospitality and Franck Duong and Al Price for critical reading of the manuscript. This work was supported by grants from the National Institutes of General Medical Sciences (W.W.) and the Swiss National Science Foundation (J.B.). J.E. was supported by a long‐term fellowship from the Human Frontiers Science Program Organization.
- Copyright © 1997 European Molecular Biology Organization