Basic amino acids in a distinct subset of signal peptides promote interaction with the signal recognition particle.

Previous studies have demonstrated that signal peptides bind to the signal recognition particle (SRP) primarily via hydrophobic interactions with the 54-kDa protein subunit. The crystal structure of the conserved SRP ribonucleoprotein core, however, raised the surprising possibility that electrostatic interactions between basic amino acids in signal peptides and the phosphate backbone of SRP RNA may also play a role in signal sequence recognition. To test this possibility we examined the degree to which basic amino acids in a signal peptide influence the targeting of two Escherichia coli proteins, maltose binding protein and OmpA. Whereas both proteins are normally targeted to the inner membrane by SecB, we found that replacement of their native signal peptides with another moderately hydrophobic but unusually basic signal peptide (DeltaEspP) rerouted them into the SRP pathway. Reduction in either the net positive charge or the hydrophobicity of the DeltaEspP signal peptide decreased the effectiveness of SRP recognition. A high degree of hydrophobicity, however, compensated for the loss of basic residues and restored SRP binding. Taken together, the data suggest that the formation of salt bridges between SRP RNA and basic amino acids facilitates the binding of a distinct subset of signal peptides whose hydrophobicity falls slightly below a threshold level.

Previous studies have demonstrated that signal peptides bind to the signal recognition particle (SRP) primarily via hydrophobic interactions with the 54-kDa protein subunit. The crystal structure of the conserved SRP ribonucleoprotein core, however, raised the surprising possibility that electrostatic interactions between basic amino acids in signal peptides and the phosphate backbone of SRP RNA may also play a role in signal sequence recognition. To test this possibility we examined the degree to which basic amino acids in a signal peptide influence the targeting of two Escherichia coli proteins, maltose binding protein and OmpA. Whereas both proteins are normally targeted to the inner membrane by SecB, we found that replacement of their native signal peptides with another moderately hydrophobic but unusually basic signal peptide (⌬EspP) rerouted them into the SRP pathway. Reduction in either the net positive charge or the hydrophobicity of the ⌬EspP signal peptide decreased the effectiveness of SRP recognition. A high degree of hydrophobicity, however, compensated for the loss of basic residues and restored SRP binding. Taken together, the data suggest that the formation of salt bridges between SRP RNA and basic amino acids facilitates the binding of a distinct subset of signal peptides whose hydrophobicity falls slightly below a threshold level.
The signal recognition particle (SRP) 1 is a ribonucleoprotein complex that targets proteins to the eukaryotic endoplasmic reticulum (ER) as well as the bacterial inner membrane (IM). Although a core domain of SRP is highly conserved throughout evolution, both the size of the particle and its substrate specificity vary considerably (for review, see Ref. 1). Mammalian SRP is a relatively large particle comprised of six polypeptides and a 300-nucleotide RNA. In the first phase of the targeting reaction, the SRP 54-kDa subunit (SRP54) binds to both Nterminal signal sequences and transmembrane segments (TMSs) of integral membrane proteins (which often lack cleaved signal peptides) as they emerge during translation (2)(3)(4). Subsequently the ribosome-nascent chain complex migrates to the ER, where an interaction between SRP54 and a membrane-bound receptor catalyzes release of the nascent chain and its insertion into a protein translocation channel (5)(6)(7). At the other extreme, Escherichia coli SRP consists of only a homolog of SRP54 (Ffh) and an ϳ100 nucleotide RNA (4.5 S RNA) that is closely related to helix VIII of mammalian SRP RNA. Despite the difference in size, bacterial and mammalian SRPs share many biochemical properties (8,9). The substrate specificity of E. coli SRP, however, is more restricted in that it targets primarily inner membrane proteins (IMPs) to the IM (10 -12). Most periplasmic and outer membrane proteins, which contain cleaved signal peptides, are targeted to the membrane by molecular chaperones such as SecB (13). Unlike SRP, chaperones do not recognize signal sequences. Instead, they bind to the mature region of presecretory proteins late during translation or post-translationally to maintain translocation competence and to ensure that signal peptides are accessible to gate open translocation channels (14).
Biochemical studies showed 20 years ago that SRP recognizes the 7-13-amino acid hydrophobic core ("H region") that is a universal feature of signal peptides (15). More recently, crystallographic analysis of mammalian SRP54 and its bacterial homologs revealed the presence of a large hydrophobic groove in the "M domain" that likely represents the signal peptide binding pocket (16,17). Mammalian SRP appears to interact with signal peptides that vary widely in hydrophobicity. In bacteria and the yeast Saccharomyces cerevisiae, which also has multiple targeting pathways; however, SRP discriminates between different targeting signals that vary only slightly in hydrophobicity. In those organisms presecretory proteins that contain moderately hydrophobic signal peptides are bypassed by SRP and targeted by molecular chaperones by default. In E. coli, maltose binding protein (MBP) and OmpA are normally targeted to the IM by SecB, but increasing the net hydrophobicity of their signal peptides reroutes both proteins into the SRP pathway (18). Furthermore, the biogenesis of M13 procoat protein, a small IMP whose insertion normally does not require any targeting factor, becomes SRP-dependent when it contains an unusually hydrophobic signal peptide (19). Likewise, yeast SRP binds preferentially to signal peptides that have a high hydrophobicity index (20). The data suggest that different SRP54 homologs are calibrated to bind to a different range of targeting signals. Indeed the observation that the putative binding pockets of evolutionarily distant M domains differ considerably in size and shape (16,17) might account at least in part for the variation in substrate specificity.
The recent solution of the crystal structure of the E. coli Ffh M domain bound to a fragment of 4.5 S RNA raised the unexpected possibility that SRP RNA may also play a role in substrate recognition (21). The x-ray data show that a portion of the phosphate backbone of 4.5 S RNA lies adjacent to the hydrophobic groove in the Ffh M domain and appears to create an extended signal peptide binding pocket. The structure suggests that electrostatic interactions between the phosphates * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: National Institutes of Health, Bldg. 5, Rm. 201, Bethesda, MD 20892-0538. Tel.: 301-402-4770; Fax: 301-496-9878; E-mail: harris_bernstein@nih.gov. 1 The abbreviations used are: SRP, signal recognition particle; ER, endoplasmic reticulum; HA, influenza hemagglutinin epitope HA.11; IM, inner membrane; IMP, IM protein; IPTG, isopropyl-␤-D-thiogalactopyranoside; MBP, maltose-binding protein; TF, trigger factor; TMS, transmembrane segment. and basic amino acids that often reside at the N terminus ("N region") of signal peptides and that flank TMSs might contribute to substrate recognition. Curiously, biochemical studies have not provided any evidence that SRP interacts with basic amino acids. Mutation of basic amino acids in model signal peptides does not significantly affect recognition by mammalian SRP in cell-free assays (22,23). By contrast, alteration of either the charge of the N region or the distance between basic amino acids and the H region can profoundly affect signal peptide cleavage, interaction with components of the translocation machinery, and translocation into ER vesicles (22)(23)(24)(25)(26). In E. coli, basic amino acids that flank TMSs influence IMP topology but are not required for membrane integration (27). A screen for mutations in the MBP signal sequence that improve export in a secBϪ strain (and that probably reroute MBP into the SRP pathway) yielded changes that increase its hydrophobicity but not its net positive charge (28). None of these results, however, rules out the possibility that interactions between SRP RNA and basic amino acids play a minor role in substrate recognition or that SRP RNA interacts with only a subset of signal peptides. Indeed electrostatic interactions might be expected to make a relatively small contribution to SRP recognition since signal peptides are predominantly hydrophobic and because only about a third of the putative extended binding surface is contributed by the RNA.
In this study we reexamined the role of basic amino acids in the N region of signal peptides in targeting pathway selection. Our experimental strategy was based on the observation that E. coli presecretory proteins can be rerouted into the SRP pathway by increasing the hydrophobicity of their signal sequences (18). We reasoned that if electrostatic interactions between SRP RNA and basic amino acids in signal peptides promote substrate recognition, then the presence of a highly charged signal peptide might likewise alter the targeting of presecretory proteins. Consistent with our hypothesis, we found that replacing the native signal peptides of MBP and OmpA with a moderately hydrophobic, but atypically basic signal peptide derived from the signal peptide of the E. coli autotransporter EspP directed both proteins into the SRP pathway. As expected, SRP recognition required the presence of multiple basic amino acids. Several lines of evidence indicated, however, that basic residues only promote the binding of SRP to a distinct subset of signal peptides that barely escape detection on the basis of hydrophobicity alone.
Plasmid Construction-Plasmids pHL36, which contains an HAtagged version of ompA under the control of a trc promoter, and pJH28 and pJH29, which contain malE under the control of a tac promoter, have been described (18,29). To construct plasmid pJH46, the signal peptide of EspP was first amplified by PCR using the oligonucleotides 5Ј-GTTTCCCTTAAAAATGGAGCTCATATGAA-3Ј (EspP1) and 5Ј-GATGTAGAAATTTGAAATATCCATATGTGACGC and E. coli strain EDL933 genomic DNA (ATCC) as a template. The amplified DNA was then cloned into the NdeI site of pJH29. To make plasmid pJH47, the downstream NdeI site was abolished by site-directed mutagenesis using the oligonucleotide 5Ј-CTTTTGCGTCACAGATGAAAATCGA-AGAAGG-3Ј and its complement, and a new NdeI site was introduced in the middle of the EspP signal peptide using the oligonucleotide 5Ј-CA-TCAAGAGCAACTCATATGAAAAAACACAAACGCATACTTGC-3Ј and its complement. A plasmid that lacks the N terminus of the EspP signal peptide (pJH48) was then generated by resealing NdeI-digested pJH47. To construct plasmid pJH50, the EspP signal peptide was amplified with the oligonucleotides EspP1 and 5Ј-TTGAAATATCCATCTCGGC-CGCAAAAGAATATGAGG-3Ј, and the amplified DNA was cloned into the NdeI and EagI sites of pHL36. Subsequently a new NdeI site was introduced into the middle of the signal peptide, and the plasmid was resealed after NdeI digestion as described above to create pJH51. All of the mutant versions of the MBP and ⌬EspP signal peptides were constructed by introducing point mutations into pJH28 and pJH48. Site-directed mutagenesis was performed using the QuikChange mutagenesis kit (Stratagene). DNA encoding the first 94 amino acids of MBP and ⌬EspP-MBP was amplified using the oligonucleotides 5Ј-GTCCGTTTAGGTGTTTTCACGAGGAATTCACCA-3Ј and either 5Ј-TTGAGCGGATCCACCCATGCGGTCGTGTGCCCAGAA-3Ј or 5Ј-TTGAGCGGATCCACCCATGCGGTCGTGTGCCCAGAACATAATG-3Ј and either pJH28 or pJH48 as templates. The amplified DNA was then cloned into the EcoRI and BamHI sites of pGEM-4Z (Promega) to generate pJH56 and pJH57. To equalize the signal in in vitro translations, two amino acids near the C terminus of the ⌬EspP-MBP 94-mer were changed to methionine during the PCR amplification. Plasmid pJH58 was constructed by transferring an NheI-Hind III fragment of pJH42 (29) containing the tig gene into pBAD33.
Protein Export Assays-For most experiments, cells were grown in M9 containing 0.2% glucose. Overnight cultures were washed and diluted into fresh medium at an optical density (OD 550 ) of 0.025. For analysis of OmpA export in SKP1101/SKP1102 and in cells that overproduced TF, M9 supplemented with 0.2% glycerol and all the L-amino acids except methionine and cysteine was used. For Ffh depletion studies, cells were grown overnight in M9 containing 0.2% fructose and 0.2% arabinose, washed in medium lacking arabinose, and then added at OD 550 ϭ 0.005 to medium containing fructose and either arabinose or glucose (0.2%). In general, synthesis of plasmid-borne presecretory proteins was induced by the addition of 50 M isopropyl-␤-D-thiogalactopyranoside (IPTG) at OD 550 ϭ 0.2. For trigger factor (TF) overproduction studies, cultures were divided in half at OD 550 ϭ 0.2, arabinose (0.2%) was added to one portion, and incubation was continued for 30 min before IPTG addition. To analyze protein export at low temperature, cultures were shifted to 22°C at OD 550 ϭ 0.2 and incubated for 40 min before IPTG addition. In experiments involving SKP1101/ SKP1102, cultures were grown at 30°C to OD 550 ϭ 0.1 and then incubated at 42°C for 1.5 h before IPTG was added. In all experiments aliquots were removed from each culture 20 -30 min after IPTG addition. Cells were then pulse-labeled with 30 Ci/ml Tran 35 S-label (Amersham Biosciences) for 30 s and incubated for various chase times. After the chase period proteins were precipitated immediately by the addition of cold 10% trichloroacetic acid. Immunoprecipitations were performed essentially as described (29), and proteins were resolved by SDS-PAGE on 8 -16% minigels (Novex).
In Vitro Translation and Cross-linking-An E. coli translation extract was prepared by first rapidly chilling exponentially growing MRE600. Cells were washed and resuspended in 50 mM triethanolamine-acetic acid (pH 8.0), 50 mM KCl, 15 mM magnesium acetate, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride and passed twice through a French pressure cell at 8000 p.s.i. in a 1:1 (w/v) suspension. The cell lysate was centrifuged at 30,000 ϫ g for 30 min, and the resulting supernatant was then incubated at 37°C for 1 h before freezing. Truncated mRNAs were synthesized by incubating BamHI-digested pJH56 and pJH57 (100 ng/l) and 15 units of SP6 polymerase (Promega) in 40 mM Tris-HCl (pH 7.5), 6 mM MgCl 2 , 2 mM spermidine, 10 mM dithiothreitol, 0.5 mM rNTPs for 1 h at 40°C. In vitro translation reactions (50 l) programmed with these mRNAs were performed essentially as described (31), except that they were incubated at 25°C for 20 min. Reactions were then placed on ice for 5 min and diluted with an equal volume of buffer A (35 mM triethanolamine in acetic acid (pH 8.0), 60 mM potassium acetate, 11 mM magnesium acetate, 1 mM dithiothreitol). Ribosome-nascent chain complexes were collected by centrifugation at 60,000 rpm for 30 min in a TLA100 rotor at 4°C. The pellets were washed, resuspended in 50 l of buffer A, and divided in half. E. coli SRP (50 nM) that had been purified as described (32) was added to one portion. Cross-linking reactions were then performed with 2 mM disuccinimidyl suberate as described (33), and proteins were precipitated with cold acetone. Half of each sample was subjected to SDS-PAGE on 14% minigels. Ffh-containing polypeptides were isolated from the other half by immunoprecipitation and resolved by SDS-PAGE on 8 -16% minigels.

RESULTS
The Highly Basic ⌬EspP Signal Peptide Reroutes E. coli Presecretory Proteins into the SRP Pathway-In considering the hypothesis that basic amino acids in signal peptides play a role in targeting pathway selection, we reasoned that naturally occurring presecretory proteins that contain signal peptides with atypically charged N regions might be SRP substrates. Strikingly, the signal peptides of the serine protease autotransporters of E. coli and Shigella ("SPATEs") are both unusually long and unusually basic. These signal peptides contain a ϳ25amino acid segment that resembles typical signal peptides as well as a ϳ25-amino acid N-terminal extension of unknown function. Previous results indicate that one member of the SPATE family, Hbp, is targeted to the IM by SRP (34). To determine whether the basic residues found in SPATE signal peptides promote SRP recognition, we first replaced the native signal peptides of MBP and OmpA, two proteins that are normally targeted by SecB, with either the complete EspP signal peptide or a truncated version that lacks the N-terminal extension (⌬Esp). We then examined the effect of changing the signal peptide on the targeting of each protein. The EspP signal peptide was chosen as a model because its N region contains four closely spaced basic residues and a histidine, which might also be slightly charged (Fig. 1).
Although the EspP signal peptide did not alter the targeting pathway of either MBP or OmpA (data not shown), we found that the ⌬EspP signal peptide eliminated the SecB requirement for export. Initially MC4100 (secBϩ) and HDB55 (secBϪ) were transformed with plasmids encoding MBP or OmpA or a derivative containing the ⌬EspP signal peptide ⌬EspP-MBP or ⌬EspP-OmpA under the control of the trc promoter. The plasmid-borne versions of OmpA were HA-tagged to distinguish them from endogenous OmpA. The synthesis of plasmid-borne proteins was induced by the addition of IPTG, and export was examined in pulse-chase labeling experiments. Radiolabeled proteins were immunoprecipitated, and export was assessed by comparing the relative amounts of precursor and mature forms of MBP or OmpA at each time point. Consistent with previous results, the wild-type proteins were exported much less efficiently in the secBϪ strain than in MC4100 ( Fig. 2A, lanes  1-6). By contrast, ⌬EspP-MBP or ⌬EspP-OmpA was exported equally well in both strains ( Fig. 2A, lanes 7-12). These results imply that the presence of the highly basic signal peptide reroutes the proteins from the SecB pathway to another targeting pathway or abolishes the need for a targeting factor altogether.
Further investigation indicated that the ⌬EspP signal peptide directs presecretory proteins into the SRP targeting pathway. To test the effect of depleting SRP on the export of proteins containing the ⌬EspP signal peptide, isogenic secBϩ and secBϪ strains in which ffh is under the control of the araBAD promoter (HDB51 and HDB52, respectively) were transformed with a plasmid encoding MBP or ⌬EspP-MBP and grown in medium supplemented with arabinose. Ffh was then depleted from half of the cells by switching the carbon source to glucose, and protein export was assayed as described above. Ffh depletion did not measurably affect the export of ⌬EspP-MBP in HDB51 but caused a significant export defect in the secBϪ strain (Fig. 2B, lane 4). The results suggest that ⌬EspP-MBP is targeted by SRP in wild-type E. coli but can also be targeted effectively by molecular chaperones when the SRP pathway is impaired. Indeed given that the ⌬EspP signal peptide is only moderately hydrophobic, this interpretation of the data is consistent with other results showing that SRP dependence correlates with an unusual degree of signal peptide hydrophobicity (see below and Ref. 18). We next obtained direct evidence that SRP can interact with the ⌬EspP signal peptide in chemical cross-linking experiments. Cell-free translation reactions were programmed with mRNAs that encode the first 94 amino acids of MBP or ⌬EspP-MBP, radioactive nascent chains were synthesized, and the homobifunctional cross-linker disuccinimidyl suberate was added to isolated ribosome-nascent chain complexes in the presence or absence of 50 nM E. coli SRP. When ⌬EspP-MBP (but not wild-type MBP) nascent chains were synthesized, a prominent radiolabeled band of ϳ55 kDa (the combined molecular mass of Ffh and the nascent chain) was observed in the presence of SRP (Fig. 3A, lanes 1-4). Immunoprecipitation with an anti-Ffh antiserum confirmed that the band corresponded to a cross-linked complex of Ffh and the nascent chain (Fig. 3B, lane 4). Ffh was cross-linked to the ⌬EspP signal peptide considerably less efficiently than to the highly hydrophobic MBP*1 signal peptide (data not shown), but the reason for this discrepancy is unclear.
The results of a different set of experiments strongly suggested that SRP also targets ⌬EspP-OmpA to the IM. Presumably because SecB targets wild-type OmpA post-translationally, a variable amount of pro-OmpA was reproducibly observed in pulse-labeled MC4100 and related secBϩ strains (Figs. 2, A and C, lane 1). This effect was particularly pronounced when cells were grown at 22°C (Fig. 2C, lane 1, top  panel). Interestingly, the precursor form of ⌬EspP-OmpA was not observed in pulse-labeled MC4100 (Fig. 2A, lane 7; Fig. 2C,  lane 3, top panel). When a strain harboring an ffh Ts mutation (SKP1101) and an isogenic ffhϩ strain (SKP1102) were shifted to 42°C, however, the ⌬EspP-OmpA precursor was observed in the mutant strain (Fig. 2C, lane 3, bottom panel). These results suggest that ⌬EspP-OmpA is targeted rapidly to the IM by the co-translational SRP pathway in wild-type cells but routed by default into a slower post-translational pathway when SRP function is impaired. We obtained further evidence that SRP targets ⌬EspP-OmpA to the IM in experiments in which we overproduced TF, a chaperone that binds promiscuously to nascent polypeptides early in biosynthesis. Previous work showed that TF overproduction strongly retards the export of OmpA, ␤-lactamase, and alkaline phosphatase (a protein that does not require a chaperone for export) but does not affect the biogenesis of proteins targeted by SRP (29). This effect can be explained by the observation that the binding of SRP and TF to nascent polypeptides is mutually exclusive (35). We transformed HDB37 (MC4100 araϩ) with plasmids expressing the TF gene under the control of the araBAD promoter and either OmpA or ⌬EspP-OmpA. As expected, the addition of arabinose greatly delayed the export of OmpA (Fig. 2D, lanes 1-4). TF overproduction, however, only very slightly affected the export of ⌬EspP-OmpA (Fig. 2D, lanes 5-8). Taken together with the results described above these data provide strong evidence that the presence of the ⌬EspP signal peptide routes presecretory proteins into the SRP pathway.
SRP Recognizes the ⌬EspP Signal Peptide on the Basis of Both Charge and Hydrophobicity-We next wished to deter-mine whether the basic amino acids in the N region of the ⌬EspP signal peptide are required for SRP binding. To this end we mutagenized the basic residues in various combinations to glutamine. Mutants that contained glutamine in place of the first two lysines and the histidine (⌬EspP(Ϫ3)), the lysine and arginine adjacent to the H region (EspP(Ϫ2)), and all five of the charged and partially charged residues (⌬EspP(Ϫ5)) were produced (see Fig. 1). MC4100 and HDB55 were transformed with plasmids that encode the modified versions of ⌬EspP-MBP, and export was assessed as described above. Decreasing the net charge of the N region of the signal peptide did not affect export in MC4100 but led to progressively severe export defects in the secBϪ strain (Fig. 4, top three panels). Interestingly, mutation of either the first two or the last two charged amino acids in the ⌬EspP signal peptide partially restored the SecB requirement. These results imply that complete rerouting of MBP into the SRP pathway requires the presence of basic amino acids at multiple positions within the N region of the ⌬EspP signal peptide.
In considering the features of a signal peptide that promote SRP binding, we were struck by the fact that the H regions of the ⌬EspP and MBP signal peptides are curiously similar in sequence (see Fig. 1). Five amino acids in the respective H regions are identical, and two others are closely related. Although both H regions contain seven large and two small hydrophobic amino acids, a calculation based on a standard hydropathy scale (36) indicates that the ⌬EspP H region has a higher average hydrophobicity. We conjectured that this rela-tively small difference between the two signal peptides might help to explain their differential ability to interact with SRP. To test this possibility we attached versions of the ⌬EspP signal peptide that contain single point mutations (F12A and L15T) to MBP. These mutations were chosen because they introduced the less hydrophobic amino acids found at specific positions in the MBP signal peptide into the ⌬EspP signal peptide. The single amino acid substitutions had no effect on MBP biogenesis in MC4100 but created export defects in  4. Interaction of SRP with the ⌬EspP signal peptide requires a minimum level of charge and hydrophobicity. MC4100 and HDB55 were transformed with a plasmid that produces the indicated variant of ⌬ EspP-MBP. After IPTG was added, protein export was analyzed by pulse-chase labeling and immunoprecipitation with an anti-MBP serum. The length of the chase is shown. p, precursor; m, mature.
HDB55 that were at least as severe as those produced by the ⌬EspP(Ϫ5) mutant (Fig. 4, bottom two panels). Indeed the export of MBP containing the ⌬EspP(L15T) signal peptide showed essentially the same degree of SecBϪ dependence as wild-type MBP (compare Figs. 4 and 2A). These results strongly suggest that the H region of the ⌬EspP signal peptide barely surpasses a threshold level of hydrophobicity that is essential for SRP recognition. Moreover, by showing that targeting pathway selection is far more sensitive to small changes in the H region than to neutralization of the entire N region, the data suggest that signal peptide hydrophobicity is the primary parameter that governs SRP binding.
We obtained additional evidence that SRP recognition requires a minimum level of signal peptide hydrophobicity in experiments in which we increased the net positive charge of the N region of the wild-type MBP signal peptide. Our mutagenesis strategy involved changing 2 or 3 amino acids to arginine or lysine. The most highly charged signal peptide variant (MBP(ϩ3)) contains a stretch of five consecutive basic amino acids that is nearly identical in sequence and location to the basic motif found in the ⌬EspP signal peptide (see Fig. 1). Pulse-chase experiments conducted in MC4100 and HDB55 cells showed that attachment of a signal peptide containing two extra positive charges (MBP(ϩ2)) to MBP had no effect on the rate of export or the SecB requirement (Fig. 5, top two panels). The export of MBP containing the MBP(ϩ3) signal peptide was then analyzed at 22°C as well as 37°C since electrostatic interactions are likely to be stronger at low temperature. Remarkably, attachment of the MBP(ϩ3) signal peptide appeared to actually increase dependence on SecB at both temperatures and slow export at 22°C (Fig. 5, last three panels). These data confirm that a high net positive charge of a signal peptide does not promote SRP binding if the hydrophobicity falls even slightly below a sharply defined threshold.
A High Degree of Signal Peptide Hydrophobicity Is Sufficient to Promote SRP Recognition-Given that the targeting of ⌬EspP-MBP appeared to be more sensitive to changes in hydrophobicity than net positive charge, we hypothesized that basic amino acids might be superfluous for SRP recognition provided that a signal peptide is sufficiently hydrophobic. To test this idea, we first systematically increased the hydrophobicity of the ⌬EspP(Ϫ5) signal peptide by mutating C11 and G14 and increasing the length of the H region and examined the export of MBPs containing the mutant signal peptides. Although some of the mutations slightly delayed MBP export in MC4100 (Fig. 6A, lanes 1-3), it was clear that increases in the overall hydrophobicity and length of the H region progressively reduced the SecBϪ dependence of export (Fig. 6A, lanes 4 -6). Elevating the hydrophobicity of the signal peptide concomitantly increased the severity of export defects in HDB51 after Ffh depletion (Fig. 6B, lanes 3 and 4). This enhanced SRP dependence likely reflects protein aggregation in the absence of a co-translational targeting mechanism. A signal peptide that contained leucines in place of Cys-11 and Gly-14 (⌬EspP*2(Ϫ5)) appeared to confer partial dependence on both the SecB and SRP pathways and therefore probably interacts with SRP only marginally. The data demonstrate that a high degree of signal peptide hydrophobicity is sufficient to route a presecretory protein into the SRP pathway. To corroborate this conclusion, we subsequently reexamined the export of MBP*1, an MBP derivative containing three amino acid substitutions that increase the hydrophobicity of the signal peptide (Fig. 1). Previous studies showed that MBP*1 is targeted to the IM by SRP (18). Because the MBP*1 signal peptide is nearly as hydrophobic as the most hydrophobic ⌬EspP signal peptide derivatives described above, we surmised that the three basic amino acids in the N region might be dispensable for SRP recognition. Consistent with this prediction, we found that like MBP*1, MBP*1(Ϫ3) was exported efficiently from secBϪ cells (Fig. 6A, bottom panel). Moreover, both proteins showed similar export defects in cells that lack Ffh (Fig. 6B, bottom panel). Taken together the results provide additional evidence that SRP recognizes signal peptides primarily on the basis of hydrophobicity.

DISCUSSION
In this report we describe evidence that basic amino acids in the N region of signal peptides can play a significant role in promoting signal peptide recognition by SRP. Initially we found that the unusually basic ⌬EspP signal peptide suppresses the SecB requirement in the export of MBP and OmpA under physiological conditions and that this effect was dependent on the presence of multiple basic amino acids in the N region. Taken together, several observations strongly suggest that the elimination of the SecB requirement was due to a rerouting of the proteins into the SRP pathway. First, the export of ⌬EspP-MBP was inhibited by Ffh depletion in secBϪ cells. The simplest interpretation of this result is that the protein can be targeted effectively by both SRP and chaperonebased pathways and that export defects are detected only when multiple pathways are impaired. Because SRP acts at a very early stage of protein biosynthesis, this explanation implies that it provides the primary targeting pathway for ⌬EspP-MBP. Second, cross-linking experiments showed directly that SRP can interact with the ⌬EspP signal peptide. Third, the presence of the ⌬EspP signal peptide accelerated OmpA export except when the SRP pathway was impaired. Fourth, the ⌬EspP signal peptide prevented the delay in OmpA export that is associated with TF overproduction. Based on previous studies (29,35), the most likely explanation of this result is that interaction with SRP prevents the binding of TF to the mature region of ⌬EspP-OmpA. Finally, several experiments showed that the presence of a highly basic N region is necessary but not sufficient to explain the strong effect that the ⌬EspP signal peptide exerts on targeting pathway selection. The data strongly suggest that the basic amino acids in the ⌬EspP signal peptide contribute to eliminating the SecB requirement by promoting a specific macromolecular interaction rather than by affecting the folding of presecretory proteins.
Although we found that signal peptide charge can influence targeting pathway selection in E. coli, our results clearly show that signal peptide hydrophobicity is the primary criterion for SRP recognition. We found that SRP recognizes signal peptides that are devoid of basic amino acids provided that they are atypically hydrophobic. Furthermore, single point mutations that slightly change the hydrophobicity of the H region profoundly affect SRP recognition (see also Ref. 18), whereas mutations that alter the charge of the N region have much smaller effects. Indeed one of our most intriguing observations is that a threshold level of signal peptide hydrophobicity is absolutely essential for SRP recognition. Taken together with the finding that a 20-fold overproduction of SRP does not alter the targeting of MBP (18), this observation suggests that SRP has dra-matically different affinities for signal peptides that vary only slightly in hydrophobicity. Given that SRP binds to a diverse range of substrates, such an exquisite degree of specificity seems surprising. The ability of E. coli SRP to interact with signal peptides may be very limited, however, because it is probably designed to interact primarily with the extended stretches of hydrophobic residues found in the TMSs of IMPs. In this regard it is interesting to note that SRP recognizes the MBP*1 signal peptide but not ⌬EspP*1(Ϫ5) signal peptide. The H domain of the former peptide is longer but has a lower average hydrophobicity. Indeed it makes sense that the number of hydrophobic amino acids in a targeting signal would be an important factor in SRP recognition since few TMSs have an average hydrophobicity equivalent to that of signal peptides such as MBP*1.
Our results also imply that basic residues promote the binding of SRP to only a subset of signal peptides whose hydrophobicity falls slightly below a critical level. The contribution of signal peptide charge to SRP recognition may not have been detected in previous studies precisely because it was not significant for the recognition of the small number of model signal peptides that were examined. Our data predict that basic residues promote the recognition of only relatively few naturally occurring signal peptides in E. coli because the hydrophobicity threshold for SRP interaction is set extremely high. If the threshold is set closer to the hydrophobicity of an average signal peptide in other species due to differences in the structure of SRP54/Ffh or the interaction of SRP with the translation machinery, however, then the composition of the N region may be relevant for the binding of a much greater number of substrates.
In light of the crystallographic analysis of the SRP ribonucleoprotein core (21), it is very likely that basic residues in signal peptides promote SRP binding by forming electrostatic interactions with the phosphate backbone of SRP RNA. It is doubtful that basic residues form salt bridges with SRP54/Ffh because the protein does not have any significant negatively charged surfaces (16,37). We cannot completely exclude the possibility, however, that basic amino acids in signal peptides facilitate SRP binding by an indirect mechanism, perhaps by affecting the length or ␣-helical structure of the H region. Given that arginine-and lysine-to-glutamine substitutions perturb the interaction of the ⌬EspP signal peptide with SRP significantly but presumably alter its biophysical properties only very minimally (38), this scenario seems unlikely. A simple model that emerges from our data is that electrostatic interactions involving SRP RNA help to stabilize the binding of signal peptides that bind to SRP54/Ffh with only moderate affinity. Factors such as the size and hydrophobicity of the signal peptide binding pocket in the M domain and the relative position of the M domain with respect to the phosphate backbone of SRP RNA may influence the range of substrates that are effectively engaged via these stabilizing interactions. Indeed a two-part binding surface that has the capacity to form two distinct types of chemical bonds with potential ligands may have evolved to fine tune the limits of SRP recognition to meet the needs of different organisms.