Selective SecA Association with Signal Sequences in Ribosome-bound Nascent Chains

The role of SecA in selecting bacterial proteins for export was examined using a heterologous system that lacks endogenous SecA and other bacterial proteins. This approach allowed us to assess the interaction of SecA with ribosome-bound photoreactive nascent chains in the absence of trigger factor, SecB, Ffh (the bacterial protein component of the signal recognition particle), and the SecYEG translocon in the bacterial plasma membrane. In the absence of membranes, SecA photocross-linked efficiently to nascent translocation substrate OmpA in ribosome-nascent chain (RNC) complexes in an interaction that was independent of both ATP and SecB. However, no photocross-linking to a nascent membrane protein that is normally targeted by a signal recognition particle was observed. Modification of the signal sequence revealed that its affinity for SecA and Ffh varied inversely. Gel filtration showed that SecA binds tightly to both translating and non-translating ribosomes. When purified SecA·RNC complexes containing nascent OmpA were exposed to inner membrane vesicles lacking functional SecA, the nascent chains were successfully targeted to SecYEG translocons. However, purified RNCs lacking SecA were unable to target to the same membranes. Taken together, these data strongly suggest that cytosolic SecA participates in the selection of proteins for export by co-translationally binding to the signal sequences of non-membrane proteins and directing those nascent chains to the translocon.

Protein trafficking in bacterial cells is very complex with both Sec-dependent and -independent modes of translocating proteins across the plasma membrane and integrating proteins into the bilayer (1)(2)(3)(4)(5). The distribution of newly synthesized proteins into the appropriate processing pathway is therefore a critical issue for bacterial cells, and they have evolved multiple mechanisms for selecting and properly directing proteins into the correct pathway.
Proteins destined to be translocated or integrated via the Sec pathway are currently thought to be identified and targeted to the SecYEG translocon in two different ways (1-3, 6, 7). Nascent membrane proteins with a hydrophobic (ultimately transmembrane) signal-anchor sequence at their N termini are recognized and bound by the signal recognition particle (SRP), 2 which then interacts with the SRP receptor (FtsY) and GTP to target the ribosome-nascent chain (RNC) complex to the translocon for co-translational integration of the substrate (the nascent membrane protein). Proteins that need to be translocated across the bacterial membrane have a less hydrophobic signal sequence at their N-terminal ends and do not bind to SRP. Instead, such proteins are thought to bind post-translationally to a SecA that is itself bound to a translocon. This is generally thought to occur by the binding of translocation substrates to SecB. However, SecB is a nonspecific chaperone that binds to unfolded cytosolic proteins as well as sorting substrates. SecB does not bind specifically to signal sequences and hence does not function as a selective sorting factor (8,9). Yet SecB does bind tightly and specifically to SecA (10,11) and hence facilitates the association of some substrates with the translocon. SecA does bind signal sequences (12)(13)(14)(15), and this interaction is generally thought to occur at the membrane when the translocation substrate arrives at the translocon-bound SecA (1-3, 6, 7). However, SecA association with RNCs and the signal sequences of translocation substrates has been noted periodically (16 -18).
Another protein, trigger factor (TF), has been shown to act as a sorting factor by associating with nascent cytosolic proteins at the ribosome to ensure that they are not mistakenly directed to the translocon by SRP (2,3). The cell has therefore evolved a sorting factor that does not positively direct the substrate protein into a different compartment but instead simply ensures that the substrate stays in the same compartment in which it was synthesized. This observation is very significant, because it emphasizes how critical it is for the cell to maintain accuracy in protein sorting.
Hence, it is not surprising that the TF and SRP proteins actively scan and select nascent proteins for their substrates during translation. However, it is surprising that there is apparently no protein that recognizes and binds to a signal sequence early in translation to identify and target it to the translocon. SecB only binds to slowly folding proteins late in their translation or after synthesis is complete (10). Therefore SecB cannot select and direct signal sequence-containing proteins to the translocon early in translation. Thus, although essentially every process in the cell, including protein sorting, is mediated or catalyzed by a protein, each protein destined for translocation across the bacterial membrane is apparently synthesized completely before diffusing passively in the cytosol, sometimes in association with SecB, until it encounters and engages an unoccupied SecA⅐SecYEG complex at the membrane. Because no apparent active and positive recognition of the signal sequence by a protein appears to occur to regulate this important aspect of cell metabolism, we decided to re-examine the interactions of signal sequences with potential sorting factors.
One direct approach for detecting the association of a signal sequence with a sorting factor is photocross-linking. A photoreactive probe can be incorporated into a ribosome-bound nascent chain using an in vitro approach we originated (19) that involves translation in the presence of a chemically modified aminoacyl-tRNA, here N ⑀ -(5-azido-2-nitrobenzoyl)-Lys-tRNA (⑀ANB-Lys-tRNA). When illuminated, the probe will react covalently with any protein positioned adjacent to the probe, as was seen when a probe in the signal sequence of a pre-prolactin RNC reacted covalently with the SRP54 subunit of eukaryotic SRP (20,21). This approach has since been used to determine the proximity of nascent chains in bacterial RNCs to various factors postulated to be involved in protein sorting, including SRP and TF (18,(22)(23)(24). Here we have used photocross-linking and other approaches to characterize the association of signal sequences with SecA in the absence of any competing cofactor binding. These experiments strongly suggest that cytosolic SecA is involved in the selection and targeting of nascent translocation substrates to the SecYEG translocon.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-The Escherichia coli strains AD202 (25) and CJ105 (26) were used to prepare wild-type IMVs and SecA ts IMVs, respectively. The E. coli strains BL21/pZ52 (27) and W3110 Iq with plasmid pTRC-OmpA9 (28) (20,30). To prevent incorporation of probes into the N terminus of OmpA, the two lysines in the signal sequence of wild-type OmpA were replaced with Arg, and that construct is here referred to as wild-type OmpA. 4.5 S RNA was transcribed from the above PCR fragment using T7 RNA polymerase. Canine SRP was purified as described previously (31), except that the sucrose density sedimentation was omitted.
E. coli IMVs were prepared according to Tai et al. (32) with modifications. E. coli AD202 was grown at 37°C in LB medium to A 590 ϭ 1, whereas the SecA ts mutant strain CJ105 was grown at 27°C to A 590 ϭ 1.3 and then incubated at 43°C for another 3.5 h. The sedimented cells from a 1-liter culture were resuspended in 25 ml of buffer A (50 mM HEPES, pH 7.5, 40 mM KOAc, 5 mM MgCl 2 ) containing 250 mM sucrose and 1 mM dithiothreitol. The suspension was passed twice through a French press cell at 10,000 pounds/square inch. After unbroken cells and large debris were removed by sedimentation (6,000 ϫ g, 5 min, 4°C), the cleared lysate was layered over a two-step gradient (1.2 ml each of 0.5 and 1.4 M sucrose in buffer A) and centrifuged in a Beckman SW 41 rotor (37,000 rpm, 3 h, 4°C). The visible interface band, containing mostly crude inner membranes, was collected, diluted, and layered over a four-step sucrose gradient (3 ml of sample, 2 ml each of 0.8, 1.0, 1.2, and 1.4 M sucrose in buffer A). After centrifugation (SW 41, 37,000 rpm, 17 h, 4°C), the visible band at the 1.0 -1.2 M interface was collected and then diluted in buffer A containing 2 mM glutathione, and IMVs were purified by sedimentation (Beckman Ti-50.2 rotor, 40,000 rpm, 1 h, 4°C). The IMV pellet was resuspended in buffer A containing 250 mM sucrose and 2 mM glutathione, and IMVs were stored in small aliquots at Ϫ80°C.
SecA and Ffh-SecA and Ffh were purified from E. coli BL21/pZ52 and BL21(DE3)-pLysE/pDMF6 strains, respectively. Cells in buffer B (25 mM HEPES, pH 7.5, 1 mM dithiothreitol) containing 50 mM KCl and 3 mM MgCl 2 were disrupted using a French press, and the lysates were cleared by centrifugation. SecA was purified by chromatography on 5-ml Heparin HP and MonoQ HR5/5 columns (Pharmacia) using linear gradients of 50 -1250 mM and 50 -850 mM KCl, respectively, in buffer B. Ffh was purified by chromatography on Heparin HP and SP-Sepharose columns (Pharmacia) using linear gradients of 50 -1250 mM and 150 -1050 mM KCl in buffer B, respectively. The Ffh-containing fractions were concentrated using Centricon 10 (Amicon). Purified SecA and Ffh were stored in small aliquots at Ϫ80°C. Samples were photolyzed on ice for 15 min using a 500-watt mercury arc lamp as described previously (33) and then sedimented through a 130-l sucrose cushion (0.5 M sucrose in buffer A) in a Beckman TLA-100 rotor (100,000 rpm, 5 min, 4°C) for those samples containing IMVs. The pellets were resuspended in SDS sample buffer or in immunoprecipitation buffer (see "Immunoprecipitation"). Samples were analyzed by 10 -15% gradient SDS-PAGE (33). In all experiments, photoadducts and nascent chains were detected and quantified using a Bio-Rad FX Molecular Imager.
For the experiments shown in Fig. 3A, some samples were treated with apyrase as described previously (34). After translation in the dark as described above, apyrase (Sigma) was added to a final concentration of 0.1 or 0.2 units/l and incubated for 10 min at 26°C to hydrolyze ATP in the sample. SecA was then added to 0.4 M, after which the samples were incubated for another 10 min before being photolyzed, and then immunoprecipitated by SecA-specific antibodies and analyzed by SDS-PAGE.
Separating Ribosome-bound SecA from Free SecA-Translations (250 l) for the experiments shown in Fig. 6 were conducted as described above, except that 150 pmol of Lys-tRNA replaced the ⑀ANB-Lys-tRNA, SecA was at 0.4 M when present, and procedures were done in the light. RNCs were purified on a Sepharose CL-6B column (50 ϫ 0.7 cm inner diameter) as described previously (35,36). Fractions were analyzed by absorbance at 260 and 280 nm, liquid scintillation counting to determine 14 C counts/min, agarose gel electrophoresis to detect rRNA, and SDS-PAGE. SecA was detected in the 10% SDS-PAGE gels by Western blotting using antibodies specific for SecA and the ECL Plus detection system (Amersham Biosciences).
For the experiments shown in Fig. 7C, SecA⅐RNC complexes were separated from free SecA by sedimentation as described above. After resuspension in a translation mix lacking ribosomes (removed by sedimentation), the SecA⅐RNCs were incubated (20 min, 26°C) in the presence of IMVs. For the ATP-free experiments, the ribosomal resuspension solution was incubated with apyrase (0.7 units/l; Sigma) for 10 min at 26°C before addition to the ribosomal pellet. Following photolysis, samples were immunoprecipitated prior to SDS-PAGE.
Immunoprecipitation-RNC or IMV pellets were resuspended in 55 l of 1% (w/v) SDS, 100 mM Tris-HCl, pH 7.5, and placed at 55°C for a minimum of 30 min. After 5 l were removed for direct analysis by SDS-PAGE, the remaining 50 l were brought up to 500 l with 150 mM NaCl, 50 mM Tris-HCl, pH 7.5, 2% (v/v) Triton X-100, and 0.2% (w/v) SDS. Samples were then pre-cleared by rocking with 40 l of protein A-Sepharose (Sigma) at room temperature for 1 h, after which the Sepharose beads were removed by sedimentation. SecA-or SecY-specific antibodies (generous gifts of Drs. W. Wickner and J.-W. de Gier, respectively) were added to the supernatant, and the samples were rocked overnight at 4°C. Thereafter, 40 l of protein A-Sepharose were added, and the incubation was continued for 2 h. The immunoprecipitate was recovered by centrifugation, washed twice with 150 mM NaCl, 50 mM Tris-HCl, pH 7.5, 2% (v/v) Triton X-100, 0.2% (w/v) SDS, and one time with 150 mM NaCl, and 50 mM Tris-HCl, pH 7.5, prior to SDS-PAGE.

RESULTS
Experimental Rationale-Although it is generally best to establish the mechanisms of a process using macromolecular components that originate from the same organism, it is sometimes advantageous to use heterologous components. Such an approach may allow one to characterize an interaction between two components from an organism without interference by homologous components with real or potential overlapping activities (i.e. in the complete absence of other components from that organism that may modulate or compete with the interaction being investigated). Here, by preparing RNCs in a wheat germ translation system that is devoid of SecA, SecB, and TF and contains only a trace amount of wheat germ SRP, we can examine the association of a bacterial signal sequence with purified bacterial SecA, SecB, SRP, or membranes in the absence of any other bacterial proteins and membranes. Because this approach eliminates any chance of interference with ribosome-associated bacterial components that would be present in a bacterial RNC preparation, interpretation of the results is more straightforward. We have therefore chosen to assess the interaction of bacterial signal sequences with SecA and bacterial SRP in a background that completely lacks endogenous bacterial components.
To detect signal sequence association with SecA, SecB, and Ffh, we decided to monitor the extent of photocross-linking to these proteins by probes located adjacent to the signal or signal-anchor sequence in the nascent chain. Although this method only detects proximity and the absence of a cross-link does not necessarily mean the absence of an interaction, the existence of photoadducts between two proteins generally does correlate closely with complex formation between the two proteins. For example, we recently used a fluorescent probe in different nascent chains to characterize their association (or not) with mammalian SRP, and the spectroscopically detected binding of nascent chains to SRP coincided with the observed photocross-linking to SRP54 (37).
Homogeneous samples containing RNCs with a single nascent chain length are created by programming an in vitro translation with a truncated mRNA. Ribosomes initiate normally on such mRNAs and translate until they reach the end of the truncated mRNA. The ribosome retains the resulting peptidyl-tRNA because there is no stop codon thereby forming a stable RNC intermediate in which the length of the nascent chain is determined by the length of the truncated mRNA added to the translation. A photoreactive probe is introduced into the nascent chain during translation by adding ⑀ANB-Lys-tRNA to the translation (20). In this study, probes were incorporated at natural lysine codons located just C-terminal of the hydrophobic core of the signal sequence of OmpA, a protein secreted through the inner membrane of E. coli, or the signal-anchor sequence of leader peptidase (LepB), one of the integral inner membrane proteins of E. coli. A signal-anchor sequence functions as a signal sequence in targeting but is not cleaved and ends up integrated into the bilayer because the longer hydrophobic core constitutes a transmembrane sequence. The presence of a modified ⑀ANB-Lys probe in the nascent chain should have little effect on its affinity for various proteins, because an even larger fluorescent probe had no significant effect on the affinity of the pre-prolactin signal sequence for the mammalian SRP (37). After translation in the dark to create the photoreactive wheat germ RNCs in the presence or absence of purified bacterial proteins, samples were illuminated with UV light to initiate photocross-linking.
Wheat Germ OmpA RNCs Are Targeted to the SecYEG Translocon-To ascertain whether a bacterial signal sequence in a RNC was able to target to the bacterial membrane despite the presence of a wheat germ ribosome, photoreactive RNCs were prepared with a [ 35 S]Met-labeled 85-residue OmpA-nascent chain (here termed OmpA-85). Photoreactive probes were incorporated at positions 24 and 33 in the OmpAnascent chain, although most nascent chains contained only a single probe, because the added ⑀ANB-Lys-tRNA had to compete for incorporation with the full complement of endogenous Lys-tRNA in the translation and only about one in four incorporated lysines was ⑀ANB-Lys (38). After incubation with inverted vesicles of E. coli inner membranes (IMVs) during translation and then photolysis, radioactive bands were observed that had apparent molecular masses consistent with OmpA-85 photoadducts to SecA, TF, and an unidentified 60-kDa protein (Fig. 1A, lane 1). The presumed SecY photoadduct had an apparent molecular mass much lower than predicted, because the electrophoretic mobility of SecY is unusually high (22,39). As expected, formation of these radioactive species was both IMV-dependent ( and 5, respectively). Most important, because the nascent OmpA was adjacent to SecY in the membrane, the OmpA signal sequence targeted the wheat germ RNCs to the bacterial translocon. The wheat germ ribosome in the RNC therefore did not detectably interfere with the initial stages of translocation, including selection of the nascent chain for processing by the SecYEG translocon.
SecA Interacts with Certain Ribosome-bound Nascent Chains in the Absence of Membranes-Although expected, the absence of a SecA or other photoadduct in the absence of added IMVs (Fig. 1A, lanes 3 and 6) was noteworthy, because it demonstrated that no SecA or other OmpAbinding protein was present in the wheat germ translation system. However, if purified SecA was added to the translation incubation in the absence of IMVs, an OmpA-85 photoadduct with SecA was detected by immunoprecipitation (Fig. 1B, lane 3). In fact, nascent OmpA photocross-linking to SecA occurred with nascent chains as short as 60 residues (Fig. 1C, lane 1), which indicates that SecA association with the OmpA signal sequence occurs soon after it emerges from the ribosome. Upon lengthening the OmpA-nascent chain, the extent of its photocross-linking to SecA is reduced (Fig. 1C). The reason for this decline in photoadduct yield is not yet clear. In addition, multiple photoadducts are immunoprecipitated by the antibodies to SecA as the nascent chain lengthens (Fig. 1C). This heterogeneity may be due to nascent chain photocross-linking to SecA at different locations, similar to the differing electrophoretic mobilities reported for different signal sequence photoadducts with Sec61p (40).
SecA⅐RNC Association Requires a Signal Sequence-The observed OmpA-nascent chain photocross-linking to SecA is signal sequencedependent. When a photoreactive RNC containing an OmpA derivative lacking the signal sequence ( Fig. 2A, Omp-ss(0)) was incubated with purified SecA, no photoadducts to SecA were detected at either of two nascent chain lengths (Fig. 1B, lanes 1 and 2). Under the same conditions, the wild-type OmpA was photocross-linked very efficiently to SecA (Fig. 1B, lane 3). To assess whether the absence of photoadduct was due simply to a reduced SecA-nascent chain affinity, we increased the concentration of SecA in the translation. The yield of photoadduct with wild-type OmpA increased with the SecA concentration until the latter reached 1 M, and the yield then remained constant up to 4 M SecA (data not shown). In contrast, no photoadducts were observed with OmpA-nascent chains lacking the signal sequence even at 4 M SecA (data not shown). Thus, a signal sequence is apparently an obligatory requirement for SecA photocross-linking to, and hence association with, the nascent chain of an OmpA RNC.
A Signal-Anchor Sequence in a RNC Does Not Photocross-link to SecA-To assess further the signal sequence dependence of the SecA interaction with a RNC, we prepared photoreactive RNCs with two different lengths of LepB. Because LepB had been shown earlier to chemically cross-link to SecA during integration into the membrane (41), this choice seemed reasonable for an alternative signal sequence. LepB has a hydrophobic H1 domain at its N terminus that serves as a signal-anchor sequence, and two of the six natural Lys residues in the first 100 amino acids are close to the H1 domain (positions 24 and 30) in positions very similar to the two natural lysine positions in OmpA (24 and 33). Although the ANB probes were positioned in similar locations in OmpA and LepB, no LepB was photocross-linked to SecA (Fig. 2B, lanes  2, 3, 11, and 12). Yet as shown above, a strong photoadduct band was seen with OmpA ( Fig. 2B, lanes 1 and 10). Thus, SecA appears to distinguish between RNCs containing the signal sequence of OmpA or the signal-anchor sequence of LepB.
The observed difference in OmpA and LepB photocross-linking to SecA could be due to the differences in the signal/signal-anchor sequences themselves and/or to the flanking sequences. To address this issue, we constructed a LepB-OmpA hybrid protein in which the signal sequence of OmpA was replaced by the first 19 residues of the LepB signal-anchor sequence ( Fig. 2A, L-Omp). The probes were in the same positions in OmpA and L-Omp, and we found that the L-Omp-nascent chain could be efficiently photocross-linked to SecY in IMVs (data not shown). Although this photocross-linking to SecY showed that the LepB signal-anchor sequence was able to target the L-Omp RNC to the translocon and position itself next to SecY, a L-Omp hybrid nascent chain was only inefficiently photocross-linked to SecA (Fig. 2B, compare lanes 1 and 4, as well as 10 and 13). The 4-fold difference in SecA photocross-linking to L-Omp and OmpA RNCs (Fig. 2E) indicates that the affinity of SecA-nascent chain binding is significantly higher for wild-type OmpA than for the L-Omp chimera. Thus, the identity of the signal sequence dramatically influences the extent of its interaction with SecA in solution.
On the other hand, the fact that the difference in photocross-linking yields was only 4-fold reveals that the SecA interaction with the nascent chain involves both the signal/signal-anchor sequence and its flanking sequence. If SecA binding was mediated solely by the signal sequence, then replacing the OmpA signal sequence with the LepB signal-anchor sequence should have completely eliminated nascent chain photocrosslinking to SecA (cf. Fig. 2B, lanes 11 and 12). Alternatively, if SecA binding was mediated solely by the flanking sequence, then the chimera should have photocross-linked to SecA as efficiently as wild-type OmpA (cf. Fig. 2B, lane 10). However, because the chimeric nascent chain did not behave as either the wild-type OmpA or the wild-type LepB (Fig. 2B,  lanes 4 and 13), it appears that the interaction of the OmpA flanking region with SecA is sufficient to provide significant binding energy. This conclusion is also consistent with the fact that probes positioned in the OmpA flanking sequence at positions 24 and 33 are able to photocrosslink to SecA. Of course, because SecA catalyzes the translocation of polypeptides through the translocon, it is not surprising that it has some affinity for unfolded protein sequences (see Ref. 6).
Signal Sequence Recognition by SecA-What structural elements of the RNC-nascent chain signal sequence are recognized by SecA in solution? Although signal sequences do not have extensive sequence homology, most contain three characteristic regions: a positively charged N-terminal domain (n-region), a hydrophobic core (h-domain), and a polar C-terminal domain (c-region) with the signal peptide cleavage site (42). The OmpA signal sequence has this organization, whereas the H1 domain of LepB lacks a positive charge at the N terminus, has a more hydrophobic core sequence, and also has no signal peptidase cleavage site ( Fig. 2A).
To determine the dependence of SecA association with RNCs in solution on the positive charge at the N terminus of the signal sequence, we  NOVEMBER 11, 2005 • VOLUME 280 • NUMBER 45 prepared a series of L-Omp derivatives that had N-terminal charges ranging from Ϫ2 to ϩ4 (Fig. 2A). SecA-nascent chain photocross-linking was greatest for L-Omp-nascent chains with a ϩ2 charge in the n-region (L-Omp(ϩ2)) (Fig. 2B, lanes 8 and 17, and 2E); in fact, a L-Omp RNC-nascent chain with two positive amino acid residues in the n-region photocross-linked to SecA with the same efficiency as the wild-type OmpA-nascent chain (Fig. 2, B and E). Increasing or decreasing the charge significantly reduced the quantity of SecA photoadduct, even when the charge was reduced by only one (Fig. 2, B and E). Replacing the positive charges with negative charges essentially eliminated SecA photocross-linking to, and presumably association with, the L-Omp signal sequence (Fig. 2, B and E, lanes 5 and 14). These results are consistent with those reported earlier by Akita et al. (12) using a different substrate. Thus, the photocross-linking experiments reveal that the charged domain significantly influences the extent of SecA photocross-linking to the L-Omp signal sequence and hence of SecA binding to the RNC-nascent chain. A ϩ2 positive charge in the L-Omp signal sequence n-region optimizes its interaction with SecA.

Selective SecA Association with Signal Sequences
However, positive charges in the n-region are not sufficient by themselves to elicit SecA binding. The introduction of two positively charged amino acid residues at the N terminus of the mature OmpA sequence (i.e. lacking a signal sequence) did not elicit photocross-linking to SecA (Fig. 2C, lanes 7 and 15, and 2E).
In contrast, SecA association with a RNC signal sequence is absolutely dependent upon the presence of a stretch of nonpolar amino acids in the h-domain. Mutated proteins with a significantly reduced h-domain length did not photocross-link to SecA, even when the n-region was positively charged (Fig. 2C, lanes 5, 6, 13, and 14). Systematically decreasing the length of the h-domain from 15 to 7 amino acid residues reduced the extent of SecA-nascent chain photoadduct formation and finally resulted in the disappearance of SecA⅐RNC photocross-linking when the h-domain reached 7-9 amino acids in length (Fig. 2, C and E). The hydrophobic core of the signal sequence is therefore a primary recognition element for SecA binding.
Furthermore, the h-domain must consist of a contiguous sequence of hydrophobic amino acids. If a positively charged Arg is positioned in the middle of the OmpA signal sequence h-domain, photocross-linking to SecA disappears (Fig. 2D, compare lanes 1 and 2). The introduction of polar amino acid residues into the hydrophobic core of the signal sequence inhibits the translocation of many secretory proteins through the bacterial membrane (43), and these mutations have been thought to disrupt the interaction of the secretory protein with the SecYEG translocon. Yet the results shown in Fig. 2D suggest that charged residues in the h-domain may also, or instead, interfere with SecA binding to the nascent chain prior to its insertion into the SecYEG translocon.
Taken together, these data demonstrate that SecA photocross-linking to the RNC-nascent chain depends upon both the hydrophobicity of the h-domain and the charge of the n-region in the bacterial signal sequence. Consistent with these results, a very recent study of the conformation of a signal peptide bound to SecA identifies direct interactions between SecA and these regions of the signal sequence (44). SecA recognition of a RNC-bound signal sequence therefore involves at least two different elements of its structure, specifically its hydrophobic core and the electrostatic charge at its N terminus, in addition to its flanking sequence.
Early SecA-nascent Chain Interactions Are Independent of Both ATP and SecB-Because SecA is an ATPase (3,6), SecA may need to be in the ATP-bound conformation to bind to the RNC-nascent chain. RNCs were therefore prepared in the presence of ATP and then treated either with apyrase to hydrolyze ATP (34) or with sodium azide to block ATP binding to SecA (45) prior to SecA addition. Neither treatment significantly affected the yield of SecA-nascent chain photoadduct (Fig. 3A). Thus, SecA binding to the signal sequence in a RNC-nascent chain occurs in the absence of ATP and is not ATP-dependent.
To determine whether SecB enhances or interferes with SecA binding to the nascent chain early in translation, increasing concentrations of SecB (up to 25 M monomer) were incubated with photoreactive OmpA-85 RNCs in the absence of IMVs prior to photolysis, and the extent of SecA-nascent chain photocross-linking was not affected by SecB. We also observed no photocross-linking of SecB to ribosomebound OmpA-nascent chains as long as 165 residues (Fig. 3B and data not shown), yet these nascent chains could photocross-link to SecB after being released from the ribosome by puromycin (data not shown), as has been reported previously (23). Thus, SecB does not influence the SecA-nascent chain association early in translation.
SecA and Canine SRP Compete for Nascent Chain Binding-The bacterial Lep signal-anchor sequence has been shown previously to target a RNC to the eukaryotic translocon in a SRP-dependent manner (46). Because SecA binds to the Lep signal sequence when it has a ϩ2 charge in the n-region (Fig. 2E), we assessed whether SecA and canine SRP would compete for binding to nascent L-Omp(ϩ2)-85. As shown in Fig.  4, the addition of canine SRP reduced the extent of SecA photocrosslinking to the nascent chain, thereby demonstrating that SecA and canine SRP compete for binding to an exposed RNC signal sequence. Do SecA and SRP also compete for binding to the same site on the ribosome? Because canine SRP binds to RNCs and even non-translating ribosomes with very high affinity (37) and because SecA also binds to ribosomes, the SecA and SRP binding sites may overlap or coincide.
Nascent Chain Photocross-linking to Ffh-Based on the extent of nascent chain photocross-linking (Fig. 2E), SecA interacts poorly with signal-anchor sequences but associates strongly with signal sequences. In contrast, bacterial SRP (Ffh ϩ 4.5 S RNA) is thought to preferentially associate with membrane protein signal-anchor sequences (23,(47)(48)(49)(50)(51)(52). We therefore compared the relative efficiencies of nascent chain photocross-linking to Ffh using some of the same constructs that we had used with SecA. As expected, essentially no photocross-linking to Ffh was observed for an Omp-ss(0) RNC lacking a signal sequence (Fig. 5). Hybrid L-Omp-nascent chains with the Lep signal-anchor sequence photocross-linked very efficiently to Ffh, whereas OmpA and L-Omp(nϩ2) photocross-linking to Ffh were less efficient, as shown in Fig. 5. A comparison of the data in Figs. 2E and 5 reveals that SecA and Ffh are notably complementary in their recognition of signal sequences. Although SecA associates more tightly with the n ϭ ϩ2 version of L-Omp than the n ϭ 0 version, Ffh associates more tightly with the latter than with the former. Thus, bacterial SRP binds preferentially to membrane protein signal-anchor sequences, as others have noted (23,(47)(48)(49)(50)(51)(52). But SecA binds more tightly to the nascent chain signal sequences that are poorly recognized by Ffh.
SecA Binds to Ribosomes in the Absence of a Nascent Chain-Factors involved in nascent protein sorting often bind to the ribosome so as to interact with the appropriate nascent chain as it emerges from the ribosomal nascent chain tunnel (18,24,(53)(54)(55)(56). In fact, canine SRP binds significantly more tightly to translating than to non-translating ribosomes, and the affinity of SRP binding to the ribosome increases tremendously (but differentially) when SRP is exposed to different nascent chain signal sequences (37). Does SecA bind to a ribosome or RNC, and if so, is such binding nascent chain-dependent?
SecA association with ribosomes was evaluated by including SecA and [ 14 C]Lys-tRNA in a wheat germ translation and then using gel filtration chromatography to separate SecA bound to ribosomes from free SecA. Ribosome elution was detected by absorbance at 260 and 280 nm (Fig. 6, A-D), [ 14 C]Lys incorporation into nascent chains (Fig. 6, A-D), and agarose gel electrophoresis to detect ribosomal RNAs (data not shown), whereas SecA elution was detected by Western blotting (Fig. 6,   FIGURE 3. SecA-RNC photocross-linking is unaffected by ATP or SecB. A, parallel L-Omp(ϩ2)-85 RNC samples were prepared and then incubated (10 min, 26°C) with apyrase or sodium azide as indicated. After the addition of SecA to 0.4 M and photolysis, the samples were immunoprecipitated with antibodies to SecA and analyzed as described in the legend to in Fig. 1. B, parallel L-Omp(ϩ2)-85 RNC samples were prepared and incubated with 0.4 M SecA (lanes 6 -15) and/or with 0.8 M (lanes  2, 7, and 12), 1.6 M (lanes 3, 8, and 13), 4.0 M (lanes  4, 9, and 14), 25 M (lanes 5, 10, and 15), or no (lanes  1, 6, and 11) SecB. After photolysis, the samples were loaded directly on the gel (lanes 1-10) or were immunoprecipitated with antibodies to SecA before loading (lanes 11-15). Photoadducts to SecA are indicated by a closed circle, peptidyl-tRNA by pt, and nascent chains by nc. M r , apparent molecular mass.

E-H).
Translations contained L-Omp(ϩ2)-85 mRNA (Fig. 6, A and E), Omp-ss(0) mRNA (Fig. 6, B and F), or no mRNA (Fig. 6, C and G). A fourth sample lacked both mRNA and ribosomes after the latter were removed from the wheat germ extract by sedimentation prior to its being added to the sample (Fig. 6, D and H). The coincidence of A 260 , A 280 , rRNA, and nascent chain [ 14 C]Lys (in the samples with mRNA) showed that the ribosomes eluted in fractions 12-14 (Fig. 6, A-C), the void volume of this chromatography system. Western blot analysis of the elution of SecA revealed that significant binding of SecA to ribosomes occurs (Fig. 6E) when the RNCs contain a signal sequence to which SecA binds (Fig. 2E). Yet surprisingly, SecA binding to ribosomes did not disappear when the RNCs either lacked a signal sequence (Fig.  6F) or had no nascent chain (Fig. 6G); SecA was absent in fractions 12-14 only when the sample lacked any ribosomes (Fig. 6H). SecA  . Nascent chain photocross-linking to Ffh. RNCs were prepared in the presence or absence of 250 nM E. coli SRP, photolyzed, and analyzed as in Fig. 1 using ⑀ANB-Lys-tRNA and various truncated mRNAs as follows: OmpA-85, L-Omp, L-Omp(ϩ2), and Omp-ss(0). OmpA derivatives were truncated at the site corresponding to codon 85 in wild-type ompA gene. The photocross-linking efficiencies of these nascent chains are shown relative to that of L-Omp (taken as 100%) taking into account the number of methionines in each nascent chain and the total nascent chains synthesized. Average data and standard deviations from at least three independent experiments are shown. bound to ribosomes was well separated from free SecA, because the latter eluted in fractions 18 -28 and was not detected in fractions 12-14 (Fig. 6H). Thus, SecA binds to ribosomes whether or not the ribosome is translating, and also whether or not the nascent chain contains a signal sequence. But the binding affinity of SecA for RNCs increases dramatically when the nascent chain contains a signal sequence that is recognized by SecA (Fig. 6, compare E with F). These properties are therefore very similar to those of canine SRP, an acknowledged targeting factor (37).
In the absence of ribosomes, a large fraction of the SecA eluted from the gel filtration column earlier than monomeric SecA, suggesting that this SecA was part of a multimeric complex that was larger than monomeric SecA, and may be a SecA dimer (Fig. 6H). It is interesting that the presence of ribosomes appears to prevent the formation of this larger SecA-containing complex, but we have not characterized this observation further.
The same analyses were performed with translating and non-translating bacterial ribosomes, and the results were the same (data not shown). SecA therefore binds to a ribosomal site in both bacterial and eukaryotic ribosomes, most likely to a conserved site that is located close to the nascent chain exit site, because SecA associates with short nascent chains (Fig. 1C).
Purified SecA⅐RNCs Target Secretory Proteins to Translocons-To obtain IMVs that lacked any functional SecA, we prepared IMVs from SecA ts cells (E. coli strain CJ105) that had been incubated at an elevated temperature to inactivate the SecA (26). The absence of functional SecA in these IMVs was shown in two different ways.
OmpA-132 was translated in the presence of IMVs, and the extent of successful translocation across the bacterial membrane was indicated by the amount of radioactive OmpA protected from digestion by added protease. As shown in Fig. 7A, OmpA was translocated into wild-type IMVs (and hence protected from protease) either with or without added SecA (lanes 3 and 4), whereas IMVs prepared from the SecA ts strain could only translocate OmpA if functional SecA was added to the sample (compare lanes 5 and 6). Thus, the IMVs prepared from the SecA ts strain were completely deficient in functional SecA (Fig. 7A, lane 6) but were otherwise completely functional in translocation despite the exposure to high temperature (Fig. 7A, lane 5).
To examine the initial stages of nascent chain interaction with the translocon, photoreactive OmpA-85 RNCs were translated in the presence of either wild-type IMVs or SecA ts IMVs with or without added SecA. As shown in Fig. 7B, OmpA-85-nascent chains photocross-linked to SecY in wild-type IMVs whether or not additional SecA was present (lanes 1, 2), whereas SecY-OmpA-85 photoadducts were observed with SecA ts IMVs only when functional SecA was added to the sample (lanes 3, 4). Because OmpA-85 photocross-links to SecY only when the nascent chain is in close proximity to the translocon protein, the residual SecA present in the purified wild-type IMVs was sufficient to effect targeting (Fig. 7B, lane 2). However, RNCs were unable to target to SecA ts IMVs in the absence of added SecA (Fig. 7B, lane 4), thereby demonstrating that the IMVs purified from the SecA ts strain had no SecA activity.
The SecA ts IMVs were therefore used to assess nascent chain targeting to the SecYEG complex using preformed SecA⅐RNC complexes. Upon incubation with IMVs from SecA ts cells that lacked functional SecA, the SecA⅐RNC complexes were able to target to the translocon and photocross-link to SecY (Fig. 7C, lane 3). However, no SecY-OmpA-85 photoadduct was detected when photoreactive RNCs lacking SecA were incubated with the SecA ts IMVs (Fig. 7C, lane 4). Thus, the SecA that was stably bound to RNCs (Fig. 6) effected proper targeting of the OmpA-85 RNC to the bacterial translocon, a result that strongly suggests that SecA is directly involved in targeting nascent chains cotranslationally to the SecYEG translocon.
In contrast to SecA binding to the signal sequence, which is ATPindependent (Fig. 3A), the SecA-dependent (Fig. 7C) targeting of RNCs to the translocon is ATP-dependent. When SecA-bound RNCs are exposed to SecA-depleted IMVs in the presence of ATP, the RNCs are targeted to the translocon, as evidenced by nascent chain photocrosslinking to SecY (Fig. 7C, lane 6). However, when the purified RNC⅐SecA complexes were resuspended into an apyrase-treated solution containing the same IMVs, no nascent chain photocross-linking to SecY was observed (Fig. 7C, lane 7). FIGURE 7. Purified SecA⅐RNC targets a nascent secretory protein to a translocon. A, an mRNA encoding an OmpA derivative with a stop codon replacing residue 133 was translated to create 132-residue OmpA polypeptides that were released from the ribosome in the presence or absence of wild-type (wt) or SecA ts IMVs and/or 1 M purified SecA as indicated. After translation, proteinase K was added to 100 g/ml in each sample and incubated on ice for 15 min (32) prior to IMV sedimentation and SDS-PAGE analysis. Positions of the OmpA-132 precursor (p) and mature (m) forms are shown. B, OmpA-85 mRNA was translated (20 min, 26°C) in the presence of ⑀ANB-Lys-tRNA and of 1 M SecA as indicated; wild-type IMVs or SecA ts IMVs were then added as indicated, and samples were incubated for another 20 min. After photolysis, IMVs were pelleted and immunoprecipitated with antibodies specific for SecY, and OmpA-85-SecY photoadducts (X) were identified as described in the legend to Fig. 1. C, translations (30 min, 26°C) contained ⑀ANB-Lys-tRNA and either 0 or 1 M SecA as indicated. The resulting RNCs were separated from free SecA by sedimentation in the dark, and the RNC pellets were solubilized in the same solvent (treated or not with apyrase as indicated) that contained either wild-type IMVs or SecA ts IMVs, as indicated, but no ribosomes or SecA. After a 20-min incubation, SecY-OmpA-85 photoadducts were detected as in B. M r , apparent molecular mass.

DISCUSSION
SecA in solution exhibits most of the same functional properties as mammalian SRP, the cytosolic component that identifies and directs nascent secretory and membrane proteins to the translocon in the endoplasmic reticulum membrane (57). Specifically, SecA binds and photocross-links to ribosome-bound nascent chains with an appropriate signal sequence but not to nascent chains lacking a signal sequence (Fig. 1). This SecA association with RNCs occurs early in translation, when the nascent chains are short (Fig. 1C), and is unaffected by the presence of either ATP or SecB (Fig. 3). SecA binds tightly to ribosomes whether or not they are translating an mRNA (Fig. 6), and the SecA binding site on ribosomes either coincides with or overlaps the SRP binding site (Fig. 4). Finally, SecA is required to target RNCs to SecYEG translocons in the membrane (Fig. 7). These combined data are therefore consistent with SecA in the cytosol functioning to select and target nascent secretory proteins in RNCs to translocons in the bacterial plasma membrane.
Yet, although mammalian SRP is released from the ribosome and membrane and recycled after the nascent chain is transferred to the eukaryotic translocon (57), SecA is intimately and actively involved in the translocation process at the SecYEG translocon (1-3, 6, 7). Because 50 -75% of the SecA in the cell are found in the cytosol and are not membrane-bound (16,58), the total cellular concentration of SecA is apparently sufficient to allow it to function both in events at the membrane and in selecting RNCs and targeting them to the translocon. Another functional difference between SecA and mammalian SRP is that the latter targets both nascent secretory and membrane proteins to the endoplasmic reticulum translocon (57). In contrast, SecA does not bind or photocross-link to nascent LepB, a bacterial membrane protein (Fig. 2, B and E); instead, signal-anchor sequences of membrane proteins preferentially associate with Ffh (23,(47)(48)(49)(50)(51)(52). In fact, a systematic variation of the number of hydrophobic residues and N-terminal positive charges in the LepB signal-anchor sequence reveals that the nascent chains bound most tightly by SecA are the nascent chains bound most weakly by Ffh (Fig. 5), an observation that strongly suggests that these two sorting factors function in a complementary manner in the cell.
The observation that SecA in solution binds to the ribosome near its nascent chain exit site and appears to function as a sorting factor exacerbates an already-crowded environment at the exit site. When RNCs containing a photoreactive probe in the signal sequence of OmpA were exposed simultaneously to SRP, TF, and SecA, a fraction of the nascent chains were photocross-linked to each of these components (18). One interpretation of these results is that the nascent chain is proximal to multiple cofactors when it emerges from the ribosomal nascent chain FIGURE 8. A possible mechanism for sorting proteins in bacterial cells. Bacterial TF (yellow), SRP (pink), and SecA (blue) are each proposed to bind non-covalently to free translating or non-translating ribosomes (gray). The cellular distribution of sorting factor⅐ribosome complexes at any given time is dictated by their relative affinities and the cellular concentrations of sorting factors and ribosomes. Although not depicted for clarity, a single ribosome may simultaneously bind two or more sorting factors. As nascent chains emerge from the ribosome, the sorting factors continue to associate and dissociate from the ribosome until a nascent chain sequence (e.g. a signal sequence (SS, red) or signal-anchor sequence (SA, dark blue)) is exposed that binds with high affinity to its cognate sorting factor. This interaction stabilizes the RNC⅐sorting factor complex and thereby directs it into the proper pathway. SecB (orange), YidC (lavender), and FtsY (green). C, chaperones involved in the folding of cytoplasmic proteins, are not shown. tunnel, because these cofactors are simultaneously bound to the ribosome at sites that surround the exit site. Other possibilities are that only one or a limited number of the cofactors bind near the exit site and that the distribution of photoadducts reflects the concentrations of the cofactors and their relative affinities for ribosomal and nascent chain binding sites. SRP and TF have now been shown to bind simultaneously to the ribosome (55, 56), but it seems unlikely (although not impossible) that the limited space around the exit site could also accommodate a third molecule, especially the large SecA. Consistent with that view, the binding sites of SecA and mammalian SRP apparently overlap on wheat germ ribosomes (Fig. 4).
It therefore seems reasonable to focus on the dynamics of sorting factor interactions with the ribosome, in which the presence of a sorting factor at the exit site is dictated by its affinity (i.e. its association and dissociation rates) for the ribosome and its cellular concentration. In this situation, the frequency and duration of occupancy of a site near the exit site by the various factors would depend on their relative affinities and concentrations. Such a scheme is depicted in Fig. 8, where SecA, TF, and SRP each establish a binding equilibrium with free ribosomes that then determines the extent of nascent chain exposure to each of the sorting factors. Of course, two (or more?) of these factors could bind simultaneously to the ribosome; Fig. 8 shows them binding separately simply to avoid complicating the schematic. A particular pathway is selected when a sequence in an emerging nascent chain is recognized by and binds to its cognate sorting factor. The increased affinity arising from the binding of a sorting factor to its cognate nascent chain sequence (cf. the 20 -160-fold lower K d values for SRP⅐RNC complexes when the nascent chain contains a signal sequence (37)) would greatly slow the dissociation of the sorting factor from the ribosome, thereby directing the RNC into the appropriate pathway. If a non-cognate sorting factor happens to be bound at the exit site when a sorting sequence of the nascent chain (a signal sequence, a signal-anchor sequence, or neither) first emerges, the nascent chain would not bind to a sorting factor until the cognate factor bound to that RNC.
This model is at odds with the current view of protein sorting in bacteria. Although SRP and TF have been shown to interact directly with nascent cytosolic and membrane proteins co-translationally to direct them into the appropriate sorting pathway, secretory proteins are thought to be fully translated and released from the ribosome before diffusing post-translationally and passively in the cell until they encounter and bind to an unoccupied SecYEG translocon and SecA at the membrane (1-3, 6, 7). Some secretory proteins associate with SecB, but this interaction does not depend upon SecB binding to the signal sequence (8 -10). Although SecA is known to bind to signal sequences (12)(13)(14)(15), why has it not been previously thought to function in the sorting process prior to RNC arrival at the translocon?
One possible explanation is related to the competition for RNC binding described in the above model and depicted in Fig. 8. If TF, SRP, and SecA each bind to RNCs, then the amount of a particular sorting factor⅐RNC complex present in an in vitro experiment will be dictated by the relative concentrations and affinities of the sorting factors, as well as the presence or absence of a cognate sorting sequence in the nascent chain. It was for this reason that we chose to examine nascent chain interactions with SecA using an otherwise eukaryotic system that lacks TF, Ffh, and SecA. The association of SecA with RNCs could then be examined without interference from unknown amounts of Ffh and TF that co-purify with the ribosomes, IMVs, and other components. As has been noted previously (2), it is important to assess the functional role of a sorting factor in the complete absence of species that might compete with or modulate its activity. Because studies focusing on TF or Ffh usually add an excess of those proteins to the in vitro incubations, any effects that SecA might elicit could be overwhelmed and hence invisible.
There are also at least two other possible explanations for cytosolic SecA not being recognized as a sorting factor for secretory protein RNCs. The first is that SecA has been shown by multiple laboratories to play an active and critical role at the membrane in facilitating protein export by binding to SecYEG and catalyzing protein transport through the translocon by an ATP-dependent mechanism (1-3, 6, 7). Translocon-bound SecA binds tightly to the signal sequence of a protein destined to be translocated through the plasma membrane (1-3, 6, 7), as well as to the SecB that binds to many transport substrates and maintains them in a translocation-competent state (10,11). For this reason, most experiments that address the mechanisms of protein export in bacteria have, quite reasonably, focused on membrane-bound SecA.
The second is that bacterial secretory proteins can be translocated post-translationally into IMVs containing SecA (2). This has been interpreted to mean that no sorting factors are required for post-translational translocation, because as noted above, SecA must act at the membrane to effect translocation. Yet the fact that 50 -75% of cellular SecA is found in the cytosol (16,58) suggests that it does not function solely at the membrane. Instead, there may be two alternatives for secretory protein targeting to the translocon, one that occurs post-translationally and without any sorting factor assistance and one that is mediated cotranslationally by SecA. The choice of pathway is then presumably regulated by the physiological state of the cell.
Virtually every cellular process is regulated and/or catalyzed by a protein cofactor or enzyme, a statement that is also true for protein sorting, as evidenced by the actions of SRP and TF. Given the importance to the cell of accurate and timely protein sorting, it is therefore surprising that no bacterial protein has so far been identified as a positive sorting factor that actively directs secretory proteins to the SecYEG translocon. However, the experiments presented here in the absence of competing sorting factors strongly suggest that cytosolic SecA selects secretory protein RNCs by associating with RNC signal sequences that are not recognized or are recognized only poorly by Ffh and TF, and SecA then targets the nascent chains to SecYEG translocons for export.