Efficient secretion of a folded protein domain by a monomeric bacterial autotransporter

Authors

  • Kristen M. Skillman,

    1. Genetics and Biochemistry Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD 20892, USA.
    Search for more papers by this author
    • Present address: Department of Molecular Microbiology, Washington University School of Medicine, St. Louis, MO 63110, USA.

  • Travis J. Barnard,

    1. Genetics and Biochemistry Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD 20892, USA.
    Search for more papers by this author
  • Janine H. Peterson,

    1. Genetics and Biochemistry Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD 20892, USA.
    Search for more papers by this author
  • Rodolfo Ghirlando,

    1. Laboratory of Molecular Biology, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD 20892, USA.
    Search for more papers by this author
  • Harris D. Bernstein

    Corresponding author
    1. Genetics and Biochemistry Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD 20892, USA.
      E-mail harris_bernstein@nih.gov; Tel. (+1) 301 402 4770; Fax (+1) 301 496 9878.
    Search for more papers by this author

E-mail harris_bernstein@nih.gov; Tel. (+1) 301 402 4770; Fax (+1) 301 496 9878.

Summary

Bacterial autotransporters are proteins that contain a small C-terminal ‘β domain’ that facilitates translocation of a large N-terminal ‘passenger domain’ across the outer membrane (OM) by an unknown mechanism. Here we used EspP, an autotransporter produced by Escherichia coli 0157:H7, as a model protein to gain insight into the transport reaction. Initially we found that the passenger domain of a truncated version of EspP (EspPΔ1-851) was translocated efficiently across the OM. Blue Native polyacrylamide gel electrophoresis, analytical ultracentrifugation and other biochemical methods showed that EspPΔ1-851 behaves as a compact monomer and strongly suggest that the channel formed by the β domain is too narrow to accommodate folded polypeptides. Surprisingly, we found that a folded protein domain fused to the N-terminus of EspPΔ1-851 was efficiently translocated across the OM. Further analysis revealed that the passenger domain of wild-type EspP also folds at least partially in the periplasm. To reconcile these data, we propose that the EspP β domain functions primarily to target and anchor the protein and that an external factor transports the passenger domain across the OM.

Introduction

Protein conducting channels (‘translocases’) that catalyse the transport of both folded and unfolded polypeptides across biological membranes have been described. In general, translocases that are comprised of a single protein or a small oligomeric complex form small pores and consequently can only transport substrates in an unfolded or loosely folded conformation. Crystallographic and biochemical studies on the universally conserved Sec complex, which promotes protein translocation across the endoplasmic reticulum and the bacterial inner membrane (IM), have recently provided strong evidence that substrates are transported through a channel formed by a single SecY/Sec61p subunit (van den Berg et al., 2004; Cannon et al., 2005). These results provide an explanation for the long-standing observation that polypeptides must be in an unfolded conformation to travel through the Sec complex (Randall and Hardy, 1986). In contrast, the Tat complex, which transports completely folded proteins across the IM and the thylakoid membrane of chloroplasts, forms large oligomeric rings (Cline and Mori, 2001; Sargent et al., 2001). Likewise, a variety of folded proteins, protein complexes and pili that assemble in the periplasm are transported across the bacterial outer membrane (OM) through large 7–8 nm cylindrical structures formed by secretins (Linderoth et al., 1997; Nouwen et al., 1999). Regardless of their size, however, all translocases presumably provide an aqueous environment through which polypeptide chains can pass (Crowley et al., 1994).

In the case of the bacterial autotransporter (‘type V’) pathway, neither the nature of the transport channel nor the folded status of translocation substrates is well understood. Autotransporters are single polypeptides comprised of an N-terminal ‘passenger’ domain that often exceeds 100 kDa and a C-terminal ∼30 kDa ‘β domain’[reviewed by Henderson et al. (2004)]. After autotransporters are translocated across the IM by the Sec complex, the β domain promotes the transport of the passenger domain across the OM. Presumably following translocation, the passenger domains of some autotransporters are cleaved from the cell surface. It is unclear, however, whether the β domain acts as a translocase for its own passenger domain (as has long been assumed) or if it plays a different role in autotransporter biogenesis. X-ray crystallography has shown that the β domain of the Neisseria meninigitidis autotransporter NalP forms a β barrel structure that is typical of bacterial OM proteins (OMPs) (Oomen et al., 2004). The finding that the C terminus of the NalP passenger domain is packed tightly in an aqueous channel formed by the β barrel is consistent with the possibility that the β domain acts as a monomeric translocase, but implies that the passenger domain would have to be transported in an unfolded or hairpin conformation. A recent biochemical study has provided evidence that the Escherichia coli autotransporter AIDA is also monomeric (Müller et al., 2005). In contrast, analysis of the IgA protease, an autotransporter produced by Neisseria gonorrhoeae, strongly suggested that it forms large oligomeric rings in the OM (Veiga et al., 2002). The results of this study may explain the observation that the IgA protease can promote the translocation of folded protein domains across the OM (Veiga et al., 2004). Moreover, evidence of a protease-resistant intermediate also suggested that the passenger domain of the Shigella flexneri autotransporter IcsA folds at least partially in the periplasm (Brandon and Goldberg, 2001). The apparent discrepancy in the properties of different autotransporters and the fact that the sequence of the β domain is only weakly conserved across the autotransporter superfamily (Loveless and Saier, 1997) raise the possibility that multiple mechanisms are used to promote passenger domain secretion.

To gain further insight into protein secretion via the type V pathway, we examined the biogenesis of the E. coli O157:H7 autotransporter EspP. This protein is only distantly related to NalP and the IgA protease and belongs to the class of autotransporters whose passenger domains are cleaved from the cell surface. We obtained several different lines of evidence that the EspP β domain is monomeric and likely resembles the compactly folded NalP β domain. Paradoxically, we found that a folded protein domain fused to the passenger domain of EspP and EspP derivatives was efficiently translocated across the OM. Moreover, we obtained evidence that the native EspP passenger domain is at least partly folded in the periplasm. As it is unlikely that folded polypeptide domains could be accommodated inside the pore formed by the EspP β domain, we propose that the EspP passenger domain is transported across the OM by a novel mechanism that involves an independent translocase.

Results

The biogenesis of a minimal version of EspP is very rapid

To obtain insight into the mechanism of translocation of the EspP passenger domain across the OM, we initially investigated the rate of biogenesis of a truncated version of EspP that contains either a cleavable or non-cleavable 116-amino-acid passenger domain (EspPΔ1-851 and EspP*Δ1-851; Fig. 1). This ‘minimal’∼43 kDa version of EspP secretes its passenger domain normally (Szabady et al., 2005) and was chosen for several reasons. First, we could not accurately monitor the secretion kinetics of the full-length ∼135 kDa protein in pulse-chase experiments because its synthesis requires > 45 s and could not be confined to a logical pulse labelling period. Second, by using a truncated version of EspP we expected to minimize any delays in biogenesis that might simply be attributed to interactions of the large passenger domain with factors in the periplasm. Indeed full-length EspP appears to reach the OM relatively slowly (Szabady et al., 2005; data not shown). Finally, by using EspP*Δ1-851 we could compare the results of subsequent experiments (see below) with those obtained using a nearly identical ∼45 kDa fragment of the IgA protease called C-IgAP (Veiga et al., 2002).

Figure 1.

EspP constructs described in this study. The domain structure of wild-type EspP and derivatives that were used in this study is illustrated. Removal of the signal peptide (SP) during translocation across the IM yields ‘proEspP’. ProEspP is subsequently cleaved between residues 968 and 969 (Brunder et al., 1997) into discrete passenger and β domains following transport of the passenger domain across the OM. All starred derivatives contain two point mutations that block this cleavage. The cholera toxin B subunit (Ctx) was fused to EspP derivatives as indicated. The positions of potential disulphide bonds are shown.

We found that the passenger domain of each EspP derivative was transported across the OM remarkably soon after the completion of protein synthesis. AD202 transformed with a plasmid encoding EspPΔ1-851 or EspP*Δ1-851 under the control of the trc promoter (pJH62 or pKMS3) were grown in minimal medium and the synthesis of the plasmid-borne gene was induced by the addition of 10 µM IPTG. Cells were subjected to pulse-chase labelling and pelleted, and half of each sample was treated with proteinase K to digest the passenger domain population that was exposed on the cell surface. EspP-containing polypeptides were then immunoprecipitated with an antipeptide antiserum that recognizes the C-terminus of the β domain and analysed by SDS-PAGE. Examination of untreated samples showed that more than half of the newly synthesized proEspPΔ1-851 was processed into discrete passenger and β domains within the pulse-labelling period and that processing was complete within 1 min (Fig. 2A, lanes 1–3). Examination of pulse-labelled samples that were treated with proteinase K revealed a ∼33 kDa band that was slightly larger than the cleaved β domain (∼30 kDa) (Fig. 2A, lane 4). The existence of this band, which likely corresponds to a population of EspPΔ1-851 molecules whose passenger domain had been translocated across the OM but not yet cleaved, confirms that translocation precedes cleavage. The observation that more than half of the pulse-labelled EspP*Δ1-851 was also converted to a ∼33 kDa fragment by proteinase K indicated that the biogenesis of the non-cleavable version of the protein was also very fast (Fig. 2B, top panel, lane 4). The rate of passenger domain translocation did not change when cells transformed with a plasmid that encodes espP *Δ1-851 under the control of the weak lac promoter (pKMS4) were grown without IPTG to reduce the expression level by about 20-fold (Fig. 2B, bottom panel). The finding that EspP*Δ1-851 biogenesis was complete within seconds even when the protein was produced at a relatively low level implies that either the translocation-active form of the protein is a monomer or rapid oligomerization is driven by high affinity interactions between individual subunits.

Figure 2.

Rapid biogenesis of a minimal version of EspP.
A. AD202 transformed with pJH62 (Ptrc-espPΔ1-851) were grown in M9 and subjected to pulse-chase labelling after the addition of 10 µM IPTG. Cells were pelleted and proteinase K was added to half of each sample. EspP-containing polypeptides were then immunoprecipitated with the C-terminal antiserum. –PK, no protease added; +PK, proteinase K added. The position of molecular weight markers is shown.
B. AD202 transformed with pKMS3 (Ptrc-espP *Δ1-851) or pKMS4 (Plac-espP *Δ1-851) were processed as described in A after the addition of the indicated amount of IPTG.
C. AD202 containing pKMS3 were grown as in B and pelleted after the addition of 10 µM IPTG. Cell lysates were mixed with SDS-PAGE sample buffer containing either 0.1 or 2% SDS and heated to the indicated temperature prior to electrophoresis. ProEspP*Δ1-851 was detected by Western blot using the C-terminal antiserum.

The results of several experiments that were performed with cells that express espP *Δ1-851 or with whole-cell extracts were consistent with the possibility that EspP*Δ1-851 is monomeric. Like many OMPs that fold into a β barrel, EspP*Δ1-851 exhibited marked resistance to SDS-denaturation (Freudl et al., 1986). When cell lysates were mixed directly with SDS-PAGE sample buffer and the mobility of EspP*Δ1-851 was analysed by Western blot, the protein was observed as a compactly folded, rapidly migrating species unless it was first heated (Fig. 2C). In contrast to C-IgAP, however, which generates higher molecular bands on SDS-PAGE (Veiga et al., 2002), EspP*Δ1-851 migrated exclusively as a monomer even when the sample buffer contained only 0.1% SDS. In addition, whereas oligomeric complexes formed in the OM by C-IgAP could be clearly detected by treating cells with a chemical cross-linker, we could not detect any EspP*Δ1-851 complexes using the same methodology (data not shown). Finally, unlike cells that produce large OM channels, cells that expressed espP *Δ1-851 were not sensitive to low concentrations of vancomycin and did not grow on large sugars (Marciano et al., 1999; Fig. S1).

Purified EspP*Δ1-851 is a monomer

Characterization of purified EspP*Δ1-851 provided direct evidence that the protein is a monomer. A derivative of pKMS3 encoding espP *Δ1-851 with a hexahistidine (His) tag at the N-terminus of the passenger domain was expressed using the conditions described above. The presence of the tag did not affect the efficiency of passenger domain translocation (data not shown). The protein was extracted from OM vesicles with dodecyl maltoside (DDM), bound to a Ni-NTA column and eluted with imidazole. This affinity purification step yielded highly enriched EspP*Δ1-851 (Fig. 3A, lane 2). As a final purification step the His-tagged protein was passed over a gel filtration column. The protein eluted as a single peak of ∼137 kDa (Fig. 3B). Single peaks of a similar size were also observed when the protein was extracted and purified in buffers containing octyl-POE, Elugent or Zwittergent 3-14 (Fig. S2A and B and data not shown). In addition, the free β domain generated from the cleavage of EspPΔ1-851 eluted as a single peak of ∼100 kDa (data not shown). Like the EspP*Δ1-851 present in crude cell extracts, the purified protein migrated on SDS-PAGE as a compact species unless heated (Fig. 3A, lanes 3–4). Although the size of a membrane protein complex cannot be accurately determined by gel filtration alone because the contribution of the detergent is difficult to assess, the data strongly suggest that EspP*Δ1-851 is a monomer, dimer or trimer. The results are very striking given that C-IgAP was estimated to form a ∼500 kDa complex by gel filtration (Veiga et al., 2002). To clarify the oligomeric state of EspP*Δ1-851, we subjected the purified protein to Blue Native PAGE. Essentially all of the protein migrated at the molecular weight of a monomer (Fig. 3C). Virtually identical results were obtained when the protein was purified in buffers containing octyl-POE (Fig. S2C). Treatment of purified EspP*Δ1-851 with trypsin also yielded a stable ∼37 kDa fragment that was very similar to the fragment generated by trypsin treatment of intact cells (Fig. 3D). This observation provides additional evidence that purification does not significantly alter the structure of the protein.

Figure 3.

Purified EspP*Δ1-851 is a monomer. BL21 transformed with pTJB2 (Ptrc-his-espP *Δ1-851) were grown in M9 and expression of the plasmid-borne gene was induced by adding 10 µM IPTG.
A. His-tagged EspP*Δ1-851 was solubilized from membranes and purified by Ni-NTA chromatography and gel filtration in buffers containing DDM. Samples of solubilized protein (Sol), the Ni-NTA eluate (Ni) and the gel filtration peak (S-300) were subjected to SDS-PAGE on a 4–12% NuPage gel (Invitrogen) run in MES buffer. The gel filtration peak was unheated (25°C) or heated to 95°C prior to electrophoresis. Proteins were visualized by Coomassie blue staining. The migration of molecular weight standards is shown.
B. Elution profile of the S-300 column plotted as concentration (A280) versus elution volume. The column was calibrated using chymotrypsin, ovalbumin, bovine serum albumin, aldolase, catalase, ferritin and thyroglobulin as standards (□).
C. The peak fractions from the S-300 column were pooled and analysed by Blue Native PAGE.
D. His-tagged EspP*Δ1-851 (∼100 µg ml−1) was untreated or mixed with trypsin (10 µg ml−1) and incubated at 0°C for 20 min (left panel). Proteins were subjected to SDS-PAGE and visualized by Coomassie blue staining. AD202 transformed with pKMS3 were grown and pelleted as described in the legend to Fig. 2C and either untreated or treated with trypsin (10 µg ml−1) at 0°C for 20 min. EspP*Δ1-851 was detected by Western blot with the C-terminal antibody (right panel).

To obtain an even more reliable estimate of the size of EspP*Δ1-851, we next analysed the protein by analytical ultracentrifugation. The protein was first purified to homogeneity in a buffer containing octyl-POE using an additional chromatography step. As octyl-POE is buoyantly neutral under the conditions that we used (Locher and Rosenbusch, 1997), the detergent bound to EspP*Δ1-851 did not contribute to its buoyant molecular mass. In sedimentation equilibrium experiments, three different amounts of EspPΔ1-851* were centrifuged for 48 h at three different speeds and the concentration of the protein at 0.001 cm radial intervals was measured. The buoyant mass of the protein under each different condition was calculated by fitting the data to a model describing a single species (Fig. 4A and data not shown). The finding that residuals were small and randomly distributed around zero (Fig. S3) showed that the fit was accurate. A buoyant mass of 12.08 ± 0.58 kDa, which corresponds to an actual mass of 45.80 ± 2.20 kDa, was obtained for all three speeds and all three protein concentrations. The data indicated that EspP*Δ1-851 was monodisperse and had a negligible (> 1 mM) self-association constant that is incompatible with rapid oligomerization of the protein prior to passenger domain translocation. Sedimentation velocity experiments confirmed that the protein behaves as a single monodisperse species that has a sedimentation coefficient (s) = 2.14 × 10−13 s [corresponding to (s20,w) = 2.25S ] (Fig. 4B). Taken together with the results of the Blue Native PAGE analysis, these data provide strong evidence that EspP*Δ1-851 exists solely as a monomer.

Figure 4.

EspP*Δ1-851 is a monodisperse monomer in solution.
A. Sedimentation equilibrium profile of 0.27 (red circles), 0.52 (green circles) and 0.78 (blue circles) A280 units of EspP*Δ1-851 centrifuged at 16 000 r.p.m. at 4°C for 48 h plotted as concentration (A280) versus the radius (r ). The best fit to a single ideal solute is shown (black lines).
B. EspP*Δ1-851 was centrifuged at 56 000 r.p.m. at 20°C. The protein distribution [c(s)] versus s was plotted by analysing the sedimentation velocity data using sedfit.

Folded protein domains associated with EspP are transported efficiently across the OM

In light of the observation that the pore formed by the NalP β domain is only ∼10 Å in diameter, it is unlikely that an EspP monomer could form a channel large enough to transport folded protein segments. Thus, our data predict that if the EspP β domain functions as a translocase, it should only be able to promote the translocation of unfolded passenger domains across the OM. To examine the fate of a folded domain associated with the EspP β domain, we made a chimera in which the cholera toxin B subunit (Ctx), which has been shown to fold and undergo disulphide bond formation in the periplasm (Hardy et al., 1988), was fused to the N-terminus of EspP*Δ1-851 (EspP*Δ1-851-Ctx; Fig. 1). As a control we created EspP*Δ1-851-Ctx(C9S), a derivative in which a cysteine in the Ctx moiety was changed to serine to abolish disulphide bond formation. AD202 were transformed with a plasmid encoding one of the chimeras, and pulse-chase experiments were performed as described above. Radiolabelled cells, however, were divided into three equal portions. One portion was untreated, one portion was treated with proteinase K and the last portion was subjected to pegylation with mPEG-MAL-5000, a reagent that forms covalent adducts with free cysteine residues and produces a mobility shift of > 10 kDa on SDS-PAGE (Lu and Deutsch, 2001). EspP-containing polypeptides were immunoprecipitated with the C-terminal antiserum and analysed by SDS-PAGE. Non-reducing conditions were used to resolve proteins that were subjected to pegylation.

Unexpectedly, we found that the chimeric passenger domain was efficiently translocated across the OM. Essentially all of the proEspP*Δ1-851-Ctx was converted to a 33 kDa fragment by proteinase K, although the protein appeared on the cell surface slightly slower than proEspP*Δ1-851-Ctx(C9S) (Fig. 5A). The finding that the passenger domain could not be immunoprecipitated from protease-treated cells with an anti-Ctx antiserum (presumably because it was degraded) confirmed that the 33 kDa fragment represented complete translocation of the passenger domain across the OM (data not shown). At all time points EspP*Δ1-851-Ctx was completely resistant to pegylation and displayed an increased mobility on SDS-PAGE that likely reflects a folded conformation (Fig. 5B, lanes 1–3). The lack of pegylation was not simply attributed to non-reactivity of the cysteine residues in EspP*Δ1-851-Ctx because substantial amounts of both monopegylated and dipegylated species were observed when cells were grown in the presence of a reducing agent (Fig. S4). By showing that EspP*Δ1-851-Ctx adopts a compact conformation well before passenger domain translocation is complete, the results strongly suggest that the Ctx moiety folds in the periplasm. This conclusion was reinforced by the results of an experiment in which the biogenesis of EspP*Δ1-851-Ctx was slowed down by growing cells at 23°C. Although essentially no translocation of the passenger domain was observed until 1 min after synthesis of the protein, all of the EspP*Δ1-851-Ctx in pulse-labelled cells migrated rapidly on SDS-PAGE (Fig. S5). Moreover, pegylated and dimeric forms of EspP*Δ1-851-Ctx(C9S) were observed on SDS-PAGE and the unmodified protein migrated at the same molecular weight as the fully denatured protein (Fig. 5B, lanes 4–6). Treatment of cells subjected to pulse-chase labelling with chemical cross-linkers strongly suggested that the dimeric form resulted from a time-dependent interaction of the Ctx moieties of EspP*Δ1-851-Ctx(C9S) monomers (data not shown). Consistent with the above results, we also found that the analogous cleavable passenger domain associated with EspPΔ1-851-Ctx (Fig. 1) was effectively secreted into the medium and was isolated exclusively in a folded, rapidly migrating form by native immunoprecipitation (Fig. S6).

Figure 5.

A folded domain fused to EspP*Δ1-851 is efficiently translocated across the OM.
A. AD202 transformed with pKMS5 (Ptrc-espP*Δ1-851-ctx) or pJH78 [Ptrc-espP *Δ1-851-ctx(C9S)] were processed as described in the legend to Fig. 2A. EspP-containing polypeptides were immunoprecipitated with the C-terminal antiserum. –PK, no protease added; +PK, proteinase K added. The position of molecular weight markers is shown.
B. A portion of the radiolabelled cells from the experiment shown in A were subjected to pegylation. EspP-containing polypeptides were immunoprecipitated with the C-terminal antiserum and analysed by SDS-PAGE on a 10% NuPage gel run in MOPS buffer under non-reducing conditions.

As there are two cysteine residues near the middle of the EspP sequence (Fig. 1), our observations on EspP*Δ1-851-Ctx raised the possibility that the native EspP passenger domain might also fold at least partially prior to its translocation across the OM. To test this possibility, we transformed a dsbA-strain that cannot form disulphide bonds in the periplasm (HDB121) and an isogenic dsbA+ strain (HDB120) with a plasmid encoding wild-type espP under the control of the trc promoter. Cells were subjected to pulse-chase labelling as described above after the addition of 100 µM IPTG to match the level of protein synthesis observed upon addition of 10 µM IPTG to cells that produce EspPΔ1-851 derivatives. Cells were pelleted and divided in half, and one half was treated with proteinase K. Proteins present in the culture medium were then subjected to pegylation, and EspP-containing polypeptides were immunoprecipitated from all of the samples using an antipeptide antiserum directed against the N-terminus of proEspP. Examination of untreated cells showed that proEspP was processed into discrete passenger and β domains with essentially the same efficiency and kinetics in both strains (Fig. 6A, lanes 1–4). Consistent with this observation, examination of proteinase K-treated cells showed that the rate of passenger domain translocation across the OM was nearly identical (Fig. 6A, lanes 5–8). Interestingly, the cleaved passenger domain produced by HDB120 was completely resistant to pegylation, but both mono- and dipegylated forms of the protein secreted by HDB121 were observed (Fig. 6B). The data imply that the native EspP passenger domain folds sufficiently well in the periplasm to undergo disulphide bonding prior to its transport across the OM. It is particularly noteworthy that the cysteine residues in the passenger domain secreted by the dsbA-strain remain accessible to pegylation. This observation indicates that disulphide bonding does not occur spontaneously, but instead requires a substantial degree of folding in the periplasm. Finally, experiments in which we fused Ctx to the N-terminus of the full-length EspP passenger domain to create EspP-Ctx (Fig. 1) provided evidence that multiple-folded domains can be transported across the OM (Fig. S7).

Figure 6.

A folded domain associated with full-length EspP is efficiently translocated across the OM.
A. HDB120 (dsbA+) and HDB121 (dsbA–) transformed with pRLS6 (Ptrc-espP) were processed as described in the legend to Fig. 2A except that 100 µM IPTG was added. EspP-containing polypeptides were then immunoprecipitated with the N-terminal antiserum. –PK, no protease added; +PK, proteinase K added.
B. The EspP passenger domain present in the culture medium after a 10 min chase in the experiment in A was subjected to pegylation and immunoprecipitated. The position of molecular weight markers is shown.

Discussion

In this report we describe evidence that challenges the conventional view of the autotransporter β domain as an autocatalytic protein translocase. Initially we analysed the quaternary structure of EspP using a ‘minimal’ version of EspP containing a non-cleavable passenger domain (EspP*Δ1-851). The finding that passenger domain translocation precedes cleavage confirmed that EspPΔ1-851* represents an intermediate in protein biogenesis. Analysis of the secretion kinetics and biochemical properties of EspP*Δ1-851 indicated that the protein is a monomer. The data clearly showed that the oligomeric states of purified EspP*Δ1-851 and a similar ‘minimal’ version of the IgA protease (C-IgAP; Veiga et al., 2002) are very different. Furthermore, the finding that EspP*Δ1-851 was highly resistant to SDS-denaturation and proteolytic digestion indicated that the β domain likely folds into a compact β barrel. Taken together, the data strongly suggest that the overall structure of the EspP β domain is similar to that of NalP. Based on this analogy, it is likely that only a polypeptide with no tertiary structure (or perhaps a fully extended polypeptide in a hairpin conformation) can fit inside the pore formed by the EspP β domain. In light of the predicted constraints on passenger domain translocation, the observations that a passenger domain containing a small disulphide-bonded moiety (Ctx) was efficiently transported across the OM and that the native EspP passenger domain contains a disulphide bond present an apparent paradox. The disulphide bonding of Ctx is indicative of a completely folded state because the cysteine residues are located at opposite ends of the Ctx moiety. Although we cannot assess the overall folded state of the EspP passenger domain prior to translocation, the formation of a disulphide bond would create a hairpin projecting from the main polypeptide chain that would jam a narrow channel. The findings that the Ctx moiety folded prior to its translocation across the OM and that disulphide bond formation in the EspP passenger domain was DsbA-dependent confirmed that folding was initiated in the periplasm. Although we cannot completely exclude the possibility that the Ctx moiety or the EspP passenger domain unfold immediately before translocation and then refold outside the cell, this scenario is unlikely because intramolecular disulphide bonds do not appear to form readily outside the periplasm.

To explain the data, we propose that the EspP passenger domain is transported across the OM by a novel mechanism that involves an external translocase. In this model the β domain functions primarily to target the protein and anchor it in the OM. Given that the C-terminus of the passenger domain is hydrophilic, however, it is likely to be situated in the middle of the pore formed by the β domain (like the C-terminus of the NalP passenger domain) upon integration of the protein into the OM. Consistent with this prediction, experiments involving a variety of engineered protease sites and proteases all show that the EspP β domain protects a small fragment of the passenger domain (K.M. Skillman, T.J. Barnard and J.H. Peterson, unpubl. data). To account for the incorporation of the passenger domain into the β domain pore, we propose that the C-terminus of the passenger domain is enveloped by the β domain as it acquires tertiary structure in the periplasm. Subsequently the β domain is integrated into the OM and the bulk of the passenger domain is translocated across the OM by the same or two different translocases. Recent results have indicated that the integration of β barrel proteins (including autotransporters) into the OM is catalysed by the Omp85/YaeT complex (Voulhoux et al., 2003; Wu et al., 2005). As some OM proteins contain substantial extracellular loops (> 50 residues), it is conceivable that this complex acts as both a generic insertase and translocase that has a dual role in autotransporter biogenesis. By postulating that the β domain simply targets the passenger domain to the OM and indirectly promotes its translocation, our model may account for the surprising sequence diversity of the β domain within the autotransporter superfamily.

Our data can also be explained by two alternative (albeit much less probable) models. One possibility is that the EspP β domain is initially integrated into the OM in an expanded conformation, or that it expands during passenger domain translocation to accommodate small folded domains. It is difficult to imagine, however, how a small β barrel could form an expanded pore without losing its aqueous character. Moreover, it is not clear how the extensive hydrogen bond network that maintains the compact conformation of a β barrel could be disrupted in a lipid environment. It is also formally possible that EspP subunits integrate into the OM in an altered conformation that leads to the formation of transient oligomeric complexes. In this model the subunits would rapidly dissociate and adopt a stable β barrel conformation following passenger domain translocation. Once again it is difficult to imagine how the β domain could alternately expose hydrophilic segments during oligomerization to create an aqueous channel and then tuck them into the middle of a monomeric β barrel upon dissociation. Furthermore, we have not observed any changes in the protease sensitivities of EspP or EspP derivatives in intact cells or spheroplasts during pulse-chase experiments that would suggest the existence of multiple conformational states. In addition, the evidence that Ctx domains mediate a time-dependent association of EspP*Δ1-851-Ctx protomers is difficult to reconcile with a β domain oligomerization-disassembly model. Indeed based on a variety of theoretical and experimental considerations, we surmise that the passenger domains of all autotransporters may be transported across the OM in the context of a monomer and that oligomerization of proteins like the IgA protease is an independent event.

Our results also suggest that the secretion of folded polypeptide domains associated with EspP may be limited. We found that under certain conditions (e.g. incubation at low temperature, increased level of synthesis, growth in rich medium) the biogenesis of the Ctx fusions was partially impaired (Figs S5A and S7A; J.H. Peterson, unpubl. data). In addition, a chimeric passenger domain containing maltose binding protein fused to the N-terminus of EspP*Δ1-851 was not translocated across the OM (J.H. Peterson, unpublished data). Moreover, it was previously shown that modification of the EspP signal peptide causes the protein to fold into a conformation that is incompatible with passenger domain translocation (Szabady et al., 2005). It is important to note that in all of these experiments either the entire passenger domain or the entire EspP derivative appeared to be trapped in the periplasm. We have never observed a partially translocated intermediate in which completion of passenger domain translocation was hindered by formation of a folded polypeptide segment. The data suggest that translocation may be limited more by the tendency of a polypeptide to undergo non-productive side reactions such as aggregation or inappropriate interaction with the β domain than by its size or shape. In light of the prediction that many autotransporter passenger domains contain long β helical segments, it is conceivable that this structural feature has been conserved because it helps to maintain a transport-competent conformation during transit through the periplasm.

Experimental procedures

Growth conditions and bacterial strains

Unless otherwise noted, all cultures were grown at 37°C in M9 minimal medium containing 0.2% glycerol, all the amino acids except methionine and cysteine (40 µg ml−1) and ampicillin (100 µg ml−1). The strains used in this study were BL21 (Studier et al., 1990), AD202 (MC4100 ompT::kan; Akiyama and Ito, 1990), HDB120 (AD202 zih-35::Tn10) and HDB121 (AD202 zih-35::Tn10 dsbA::kan).

Plasmid construction

Plasmids pJH62 and pRLS6, encoding espPΔ1-851 and espP under the control of the trc promoter, have been described (Szabady et al., 2005). To construct pKMS3, which encodes espP*Δ1-851 under the control of the trc promoter, the EspP β domain and the C-terminal 116 amino acids of the EspP passenger domain were amplified by polymerase chain reaction (PCR) as described using pRLS11 as a template and cloned into the EagI and HindIII sites of pHL36 (Szabady et al., 2005). A hexahistidine tag was placed at the N-terminus of proEspPΔ1-851* by inserting the annealed oligonucleotides 5′-GGCC(CACCAT)3CT-3′ and 5′-GGC CAG(ATGGTG)3-3′ into the EagI site of pKMS3 to generate pTJB2. A plasmid containing espP *Δ1-851 under the control of the lac promoter (pKMS4) was made by cloning the 1.2 kb EcoRI-HindIII fragment from pKMS3 into pJN3.2 (Qi et al., 2002). The gene encoding the cholera toxin B subunit (Ctx) was amplified by PCR using the oligonucleotides 5′-CAGTTTTACTATCTTCAGCGGCCGCACATGGAACACCTC-3′ and 5′-GGGCTTTTTTATATCTTAATCGGCCGTACTAAT TGCGGC-3′ and genomic DNA from the Vibrio cholerae Pacini O139 serovar strain MO45 (ATCC). To construct plasmids pJH74, pJH75 and pKMS5, the amplified DNA was digested with EagI and cloned into the cognate site of pJH62, pRLS6 and pKMS3 respectively. A point mutation (C9S) was introduced into the Ctx moiety in plasmids pJH74 and pKMS5 using the QuikChange mutagenesis kit (Stratagene) to generate pJH77 and pJH78.

Purification and characterization of EspP*Δ1-851

Saturated 5 ml cultures of BL21 transformed with pTJB2 were washed, inoculated into 250 ml of fresh medium at an optical density (OD550) of 0.01 and grown overnight. The cells were then washed and inoculated into 10 flasks containing 1 l medium at OD550 = 0.02 and grown to OD550 of ∼0.2. Expression of the plasmid-borne gene was induced by the addition of 10 µM IPTG. After 2 h the cells were pelleted and frozen in liquid nitrogen. Thawed cells (10–15 g) were resuspended in 100 ml cold lysis buffer (50 mM Tris-HCl pH 7.5, 200 mM NaCl, 10 mM MgCl2, 10 µg ml−1 DNAse I, 100 µg ml−1 AEBSF) and then disrupted using an Emulsiflex-C5 homogenizer (Avestin). Membranes were pelleted (4°C, 1 h, 235 000 g), dispersed using a Dounce homogenizer and stirred overnight at 4°C in buffer B (50 mM K2HPO4 pH 7.5, 200 mM NaCl, 20 mM imidazole) containing 2.0% DDM (Anatrace) or 2.0% octyl-polyoxyethylene (octyl-POE, Bachem). Insoluble material was then removed by centrifugation (4°C, 1 h, 370 500 g) and soluble protein was applied to a 15 ml Ni-NTA column (Qiagen) equilibrated with buffer C (50 mM K2HPO4 pH 7.5, 200 mM NaCl, 10% glycerol) containing 0.1% DDM or 1.0% octyl-POE. The column was washed with buffer C containing 0–50 mM imidazole and EspP*Δ1-851 was eluted with buffer C containing 250 mM imidazole. The peak fractions from the Ni-NTA column were applied to a HiPrep Sephacryl S-300 column (Amersham) equilibrated with buffer D (20 mM Tris-HCl pH 7.5, 200 mM NaCl, 0.02% sodium azide, 0.5 mM EDTA) containing either 0.1% DDM or 1.0% octyl-POE.

The peak fractions from the S-300 column were analysed on 5–20% Blue Native gels as described (Schagger and von Jagow, 1991). For protein samples containing DDM, 0.002% SERVA Blue G (Coomassie, SERVA) was added to the cathode buffer instead of the sample. Because His-tagged EspP*Δ1-851 would not migrate into the gel in the presence of octyl-POE, samples containing this detergent were exchanged into 750 mM aminocaproic acid, 50 mM Bis-Tris pH 7.0, 1.0% octyl-POE using a PD-10 Sephadex G-25 column (Amersham). Coomassie dissolved in 500 mM aminocaproic acid was added to samples to a final detergent to Coomassie ratio of 4:1 (w/w) and the concentration of dye in the cathode buffer was increased to 0.02%. To allow visualization of protein bands during electrophoresis, the cathode buffer was replaced with fresh buffer containing 0.002% Coomassie after the dye front had moved approximately one-third of the distance to the bottom of the gel.

Analytical ultracentrifugation

For analytical ultracentrifugation experiments, cell membranes isolated as describe above were solubilized overnight in buffer B containing 5% Elugent (Calbiochem). Soluble protein was applied to a Ni-NTA column equilibrated with buffer C containing 0.1% DDM. The column was washed and His-tagged EspP*Δ1-851 was eluted as described above. The peak fractions were collected and dialysed overnight against 2 l buffer E (50 mM Tris-HCl pH 8.0, 0.5 mM EDTA). The protein was then applied to a 10 ml Q-Sepharose column (Amersham) equilibrated with buffer E containing 1.0% octyl-POE and eluted using a gradient of 0–0.5 M NaCl in buffer E. The peak fractions were pooled, concentrated and purified by gel filtration in buffer D containing 1.0% octyl-POE as described above.

Analytical ultracentrifugation experiments were conducted in a Beckman Optima XL-A centrifuge. For sedimentation equilibrium analysis, His-tagged EspP*Δ1-851 in buffer D containing 1.0% octyl-POE was centrifuged at 12 000, 16 000 and 20 000 r.p.m. at 4°C. Data were acquired as an average of 16 absorbance (A280) measurements. Equilibrium was achieved within 48 h. A buoyant mass [M1(1 − ν1ρ)] was calculated using Optima XL-A data analysis software running under Microcal Origin 6.0 by fitting the data from each scan to the equation Ar = Ao,1exp[HM1(1 − ν1ρ)(r2 − inline image)] + E where Ao,1 is the absorbance at a reference point ro, Ar is the absorbance at a given radial position r, H is ω2/2RT, and E is a small baseline correction. The density of the buffer (ρ) was measured on a Mettler Toledo DE51 density meter at 20°C and corrected to 4°C using standard tables. The partial specific volume (ν1) was calculated from the amino acid composition of His-tagged EspPΔ1-851* using previous described consensus data (Perkins, 1986). For sedimentation velocity analysis, His-tagged EspPΔ1-851* in buffer D containing 1.0% octyl-POE was centrifuged at 20.0°C at 56 000 r.p.m. Scans were performed at 3.0 min intervals and the data were analysed using sedfit (Schuck, 2000). Sedimentation coefficients (s) were corrected to s20,w using the buffer density that was determined experimentally and a value for the buffer viscosity that was calculated using sednterp (http://www.jphilo.mailway.com).

Analysis of EspP biogenesis in vivo

Overnight cultures were washed and diluted into fresh medium at OD550 = 0.02. When the cultures reached OD550 = 0.2, synthesis of EspP derivatives was induced by the addition of IPTG at a concentration of 10 µM (for EspPΔ1-851 derivatives) or 100 µM (for full-length EspP derivatives). Cells were subjected to pulse-chase labelling as previously described (Ulbrandt et al., 1997) 30 min after the addition of IPTG. In experiments that did not involve further manipulations, aliquots from each culture were immediately subjected to trichloroacetic acid (TCA) precipitation. To assess passenger domain translocation, aliquots of radiolabelled cultures were poured over ice and all further steps were conducted at 0–4°C. Cells were collected by centrifugation (5 min, 3000 g) and resuspended in M9 salts. Samples were divided in half, one portion was treated with proteinase K as described (Ulbrandt et al., 1997), and proteins were collected by TCA precipitation. In some experiments, cells were resuspended in M9 salts, pH 6.8 and divided into three equal portions. One portion was treated with proteinase K and one portion was untreated as described above. An equal volume of 40 mM methoxy-polyethylene glycol maleimide (mPEG-MAL-5000; Nektar Therapeutics) in M9 salts, pH 6.5 was added to the third portion and proteins were collected by TCA precipitation after a 30 min incubation at 4°C. In experiments where secreted polypeptides were analysed, culture aliquots were chilled after the indicated chase period, cells were pelleted and supernatants were clarified by centrifugation (10 min, 20 800 g). In some experiments, supernatants were incubated with an equal volume of M9 salts, pH 6.5 containing 40 mM mPEG-MAL-5000 for 30 min at 4°C and proteins were then TCA-precipitated.

Trichloroacetic acid precipitates were solubilized and immunoprecipitations using polyclonal rabbit antisera generated against N and C terminal peptides of EspP (Szabady et al., 2005) or the cholera toxin B subunit (Immunology Consultants Laboratory) were performed as described (Ulbrandt et al., 1997). The anticholera toxin antiserum recognizes both native and non-native forms of the protein. For native immunoprecipitations, samples were incubated with appropriate antisera overnight at 0°C in the presence of 1 mM PMSF and 100 µg ml−1 casein.

In experiments that did not involve radiolabelling, cells were pelleted 30 min after the addition of IPTG (4°C, 2500 g), resuspended at 10 OD ml−1 in M9 salts and sonicated. Unbroken cells were then removed by centrifugation (4°C, 2500 g).

SDS-PAGE and Western blotting

Except where noted, proteins were heated to 95°C and resolved by SDS-PAGE under reducing conditions on 8–16% minigels (Invitrogen). Western blots were probed with horseradish peroxidase-linked protein A (Amersham) and antibody–antigen complexes were detected using the SuperSignal Pico Chemiluminescence kit (Pierce).

Acknowledgements

We thank Susan Buchanan for many stimulating conversations and Yihong Ye for helpful comments on the manuscript. We also thank Eliana Bonifacino for performing preliminary experiments. This research was supported by the Intramural Research Program of the National Institute of Diabetes and Digestive and Kidney Diseases.

Ancillary