The Periplasmic Folding of a Cysteineless Autotransporter Passenger Domain Interferes with Its Outer Membrane Translocation

Abstract

Autotransporters are single polypeptides consisting of an outer membrane translocation domain mediating the translocation of a passenger domain. The periplasmic folding state of the passenger domain is controversial. By comparisons of passenger domains differing in their folding properties, our results suggest that periplasmic folding of passenger domains interferes with translocation.

Various secretion systems allow gram-negative bacterial pathogens to secrete proteins across the outer membrane (8). One of these systems, the autotransporter pathway, stands out by its apparent simplicity (6, 7, 9, 10). A typical autotransporter, such as AIDA-I, the Escherichia coli adhesin involved in diffuse adherence (1), has a characteristic modular organization (Fig. 1): (i) an N-terminal sec-dependent signal sequence, (ii) a central domain bearing the functional part of the autotransporter, (iii) a 100-amino-acid-long region called the autochaperone region, (iv) a linker region predicted to form an α helix, and (v) a 300-amino-acid-long C-terminal region predicted to form a β barrel in the outer membrane. The structure of the two latter domains has been solved for the autotransporter NalP from Neisseria meningitidis, confirming the predictions (22). This structure constitutes the translocation unit (TU) of the autotransporter. The mechanism of translocation is still poorly understood and remains the subject of various hypotheses (12, 22, 31). In all hypotheses, the TU is thought to insert itself into the outer membrane and mediate the translocation of the rest of the autotransporter, hence its being called the passenger domain (17). Furthermore, although not absolutely necessary for translocation, the autochaperone region was shown to increase the efficiency of translocation (19-21, 33).

FIG. 1.

Organization of AIDA-I and of the MalE-AIDAc fusion. The organization of AIDA-I is represented by showing its signal sequence (SS) in white, the adhesin central domain in light gray, and AIDAc in dark gray. AIDAc has three subdomains: autochaperone, α helix (hatched box), and β barrel. The organization of the chimeric protein MalEβ is represented by showing MalE signal sequence (SS) in white, the mature portion of MalE in light gray, and AIDAc in dark gray. The boxes are not shown at scale. The amino acids of the chimeric protein are indicated at the junction of the different regions. The control protein, MalE*, is also represented, with the C-terminal extension generated by the procedure used to make the construct.

One central question pertains to the folding state of the passenger domain while in transit in the periplasm. Remarkably, autotransporters can mediate the translocation of heterologous polypeptides replacing part of their original passenger domains (15, 16). Using such heterologous passenger domains containing cysteines, it was observed that the formation of disulfide bridges interfered with their translocation (11-13, 28). The authors of these studies therefore argued that there is a “translocation-competent” unfolded state that the passenger domains must adopt in order to be translocated. Other investigators, however, have seen no effect of preventing disulfide bond formation on the export of heterologous passenger domains containing cysteines (5, 14, 15), and some showed that passenger domains can be disulfide bonded in the periplasm and still be translocated (25, 30, 31).

All these studies were performed by using heterologous passenger domains containing cysteines and assessing the effect of preventing disulfide bond formation, through modification of either the protein (using mutants without cysteine) or its environment (using either reducing agents or a dsbA mutant strain). Unfortunately, the latter can have poorly controlled global effects. Moreover, the presence or absence of disulfide bonds cannot always correlate with the presence or absence of secondary or tertiary structures between cysteines (35). Since translocated substrates can be in transient intermediary states, this is peculiarly troublesome in the study of secretion. It has been, for instance, reported that the precursor of OmpA can be translocated through the sec general secretion machinery even with a disulfide bond (29), although it is accepted that the sec machinery exports unfolded polypeptides.

We therefore wanted to address the effect of periplasmic folding of passenger domains of autotransporter on their translocation with a new approach that did not rely on manipulating the oxidation state of disulfide bonds as a surrogate indicator of folding. We decided instead to compare the translocation efficiencies of cysteineless passenger domains carefully characterized and known to differ solely in their folding abilities.

MalE can be translocated by AIDAc to the bacterial surface.

To study the translocation mediated by AIDA-I, we decided to fuse the periplasmic maltose binding protein (MalE) to the TU and autochaperone domain of AIDA-I, a polypeptide called AIDAc (27). The DNA fragment coding for AIDAc was amplified by PCR from the genomic DNA of the E. coli strain 2787 (1) with primers bearing recognition sites for the restriction enzymes SacI and XbaI. This fragment was cloned in the commercial pMalp2x vector expressing a MalE-LacZα fusion under the control of the IPTG (isopropyl-β-d-thiogalactopyranoside)-inducible ptac promoter (New England Biolabs). The resulting pMalEAIDAc vector encodes a fusion of MalE (42kDa) to AIDAc (48 kDa), a protein we called MalEβ (Fig. 1). For a control MalE-expressing plasmid, we constructed plasmid pMalEΔlac by cleaving pMalp2x with EcoRI, filling in with the Klenow fragment of E. coli DNA polymerase, and religation. This results in the expression of MalE with a C-terminal extension of 25 amino acids, a protein we designated MalE* (Fig. 1). As a negative control, the empty plasmid pTrc99A (Amersham Biosciences) was used. All plasmids were transformed in ED9, an E. coli strain bearing a nonpolar deletion of the malE gene (araD ΔlacU169 rpsL relA flbB deoC ptsF ΔmalE444). Bacteria harboring the different plasmids were grown on LB agar plates or in LB broth containing 100 μg ml⁻¹ of ampicillin. The bacteria in broth were grown with agitation at 30°C until the optical density at 600 nm (OD₆₀₀) reached 0.4 and induced with 10 μM IPTG for 3 h to minimize a potential toxicity due to overproduction of a membrane protein.

In order to assess the translocation of MalE fused to AIDAc, we first tested if MalE could be accessible to externally added antibodies. Bacteria were grown as described above and fixed at an OD₆₀₀ of 1 in 13 mM sodium phosphate buffer, pH 7.4, containing 10% (wt/vol) paraformaldehyde and 0.08% (wt/vol) glutaraldehyde. Fixed bacteria were pelleted, washed three times with phosphate-buffered saline (PBS), resuspended in 50 mM glucose, 10 mM EDTA, and 20 mM Tris-HCl, pH 7.5, and immobilized on microscope slides coated with poly-l-lysine (1 mg ml⁻¹; Sigma). The slides were then blocked with PBS containing 2% bovine serum albumin. MalE at the bacterial surface was visualized by incubation with rabbit polyclonal anti-MalE antibodies (New England Biolabs) diluted 1/1,000 in PBS-2% bovine serum albumin, followed by incubation with a goat anti-rabbit immunoglobulin G rhodamine red conjugate (Molecular Probes). The slides were washed with PBS and examined with a microscope under phase contrast and epifluorescence.

As seen in Fig. 2A, whereas no signal could be seen with bacteria expressing MalE*, bacteria expressing the MalEβ fusion show a fluorescence signal located around the bacterial cells. It should be noted, however, that many bacteria had only a very weak signal, suggesting that translocation of MalE was significant but not complete.

FIG. 2.

Detection of MalE fused to AIDAc on the surface of whole bacteria. A. ED9 bearing plasmid pMalEΔlac (MalE*) or pMalEAIDAc (MalEβ) was grown at 30°C until reaching an OD₆₀₀ of 0.4 and then induced with 10 μM IPTG for 3 h at 30°C. The cultures were then normalized at equal OD₆₀₀s. The bacteria were fixed on glass slides and labeled with a rabbit antiserum raised against MalE. Goat anti-rabbit serum conjugated to rhodamine red was used as secondary antibodies and the slides were mounted for examination using a phase-contrast microscope equipped with epifluorescence and UV excitation modules. The epifluorescence and corresponding phase-contrast fields are shown in the lower and upper images, respectively. B. ED9 bearing plasmid pMalEΔlac with the MalE31 mutation (MalE31*) or pMalEAIDAc with the MalE31 mutation (MalE31β) was grown, fixed, labeled with anti-MalE antibodies, and processed as described for panel A.

We also subjected whole bacterial cells to limited trypsin proteolysis. Bacteria, grown as described above, were harvested by centrifugation, washed, and resuspended in 1 ml of PBS. Trypsin (0.1 mg ml⁻¹ final concentration) or PBS was added to the samples which were incubated for 30 min on ice. The protease inhibitor phenylmethylsulfonyl fluoride (1 mM final concentration) was added to stop the reaction and the samples were boiled for 5 min with an equal volume of twice-concentrated sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer in the presence of β-mercaptoethanol. The samples were then processed for electrophoresis and immunoblotting using polyclonal anti-MalE antibodies.

Figure 3A shows that the 90-kDa band corresponding to the MalEβ fusion protein band disappeared upon treatment with trypsin, suggesting that the MalE polypeptide is accessible to externally added proteases when fused to AIDAc. As expected, the control protein MalE*, expressed by bacteria bearing pMalEΔlac, was not affected. Accessibility of MalE in intact bacteria to both antibodies and proteases therefore strongly suggest that MalE is translocated to the bacterial cell surface when fused to AIDAc.

FIG. 3.

Extracellular protease accessibility of MalE fused to AIDAc. MW, molecular mass. A. ED9 bearing plasmid pTrc99A (−), pMalEΔlac (MalE*), or pMalEAIDAc (MalEβ) was grown as described in the legend to Fig. 2. The bacteria were pelleted and resuspended in PBS in the presence or absence of trypsin (0.1 mg ml⁻¹). After 30 min of incubation on ice, the trypsin was neutralized by the addition of phenylmethylsulfonyl fluoride. The samples were then mixed with loading buffer, separated by SDS-PAGE, and MalE revealed by immunoblotting using a rabbit antiserum against MalE. MalE′ represents a MalE-containing fragment resulting from cleavage of MalE-AIDAc. B. ED9 bearing plasmid pMalEΔlac with the MalE31 mutation (MalE31*) or pMalEAIDAc with the MalE31 mutation (MalE31β) was grown, treated, and processed as described for panel A.

Some MalE is found in the periplasm, cleaved from the MalE-AIDAc fusion.

We were surprised, however, to notice a polypeptide of approximately 42 kDa reacting with the anti-MalE antibodies in bacteria expressing MalEβ (Fig. 3A). We hypothesized that this polypeptide could be some MalE, released in the periplasm by a proteolytic cleavage occurring between MalE and AIDAc. Consistent with this hypothesis, the 42-kDa band was unaffected by trypsin, suggesting that it is not translocated to the bacterial surface.

To confirm this localization, we performed a cellular fractionation of bacteria expressing MalEβ or MalE*. Bacteria were grown as described above and normalized cultures of induced cells, in 10-ml cultures, were harvested. Supernatants were collected, filtered through a 0.2-μm filter, centrifuged at 156,000 × g for one hour and then precipitated with 10% trichloroacetic acid. The harvested bacteria were resuspended in 800 μl of 10 mM Tris-HCl, pH 8, 0.75 mM sucrose, and 100 μl of 4 mg ml⁻¹ lysozyme in 100 mM EDTA, pH 8, was added. After 30 min on ice, the bacteria were centrifuged at 13,200 × g for 20 min and the supernatants collected as the periplasm. The pellets were resuspended in 800 μl of Tris-buffered saline (TBS; 50 mM Tris-HCl, pH 8, 150 mM NaCl) containing a protease inhibitor cocktail (Complete Mini; Roche). The bacteria were lysed by sonication, and the lysates were centrifuged for one hour at 156,000 × g to pellet insoluble material, while the supernatants correspond to the cytoplasmic fractions. The pellets were resuspended in 800 μl of TBS containing 1% Triton X-100 and 10 mM MgCl₂ to solubilize the inner membrane. After 30 min of incubation on ice, the samples were centrifuged again for one hour at 156,000 × g. The pellets, corresponding to the outer membrane, were resuspended in TBS. Amounts of the fractions equivalent to the same volume of unfractionated bacterial culture were then separated by SDS-PAGE and revealed by immunoblotting with anti-MalE antibodies.

This fractionation confirmed that, in bacteria expressing MalEβ, the 42-kDa band was recovered in the periplasmic fraction (Fig. 4A). The 42-kDa periplasmic band could be a degradation product bearing an incomplete MalE polypeptide. We reasoned that if full-length MalE was released in the periplasm, bacteria expressing MalEβ should be able to complement the malE mutation of ED9. Indeed, when bacteria expressing MalEβ were grown for 48 h at 37°C on MacConkey agar plates containing 2% maltose (wt/vol), 100 μg ml⁻¹ ampicillin, and 10 μM IPTG, it was clear that the bacteria were able to utilize maltose (Fig. 4B). Overall, these experiments therefore suggest that some MalE is cleaved from the MalE-AIDAc fusion and accumulates in the periplasm as a fragment we designated MalE′.

FIG. 4.

Release of a MalE-containing fragment in the periplasm of bacteria expressing a MalE-AIDAc fusion. MW, molecular mass. A. ED9 bearing plasmid pTrc99A (−), pMalEΔlac (MalE*) or pMalEAIDAc (MalEβ) was grown as described in the legend to Fig. 2. A subcellular fractionation was performed as described in the text, and the following fractions were collected: culture supernatant (S), cytoplasm (CYT), periplasm (PER), inner membrane (IM), and outer membrane (OM). Total extracts of whole cells (WC) were obtained from an aliquot of the cultures taken prior to the fractionation procedure. Comparable amounts of the samples were separated by SDS-PAGE, and MalE was revealed by immunoblotting using a rabbit antiserum against MalE. B. ED9 bacteria, bearing plasmid pTrc99A (−), pMalEΔlac (MalE*), or pMalEAIDAc (MalEβ), were grown overnight at 30°C. The overnight cultures were normalized and spotted on MacConkey agar plates containing 2% maltose. The plates were incubated at 37°C for 48 h. We ascertained that the presence of maltose did not affect translocation of MalE by AIDAc (data not shown), and therefore, we interpret the results as being representative of the situation of bacteria grown in LB.

The fractionation also revealed a band the size of MalE in the culture supernatant of bacteria expressing MalEβ. The passenger domain of AIDA-I is normally cleaved from AIDAc after translocation and remains noncovalently attached to the outer membrane (27). This result thus suggests that MalE is translocated by AIDAc. For unknown reasons, however, the cleavage is less efficient with the MalE-AIDAc fusion and the released MalE does not remain noncovalently linked to the outer membrane. Lastly, the cytoplasmic fractions of bacteria expressing MalE* or MalEβ reveal bands the size of MalE* or MalE′ and MalEβ, respectively. These bands are either cytoplasmic precursors of the exported polypeptides or periplasmic contaminants released during lysis, because the periplasmic extraction was incomplete.

Translocation of a MalE mutant which is defective for its folding ability.

We hypothesized that the appearance of MalE′ could be due to the fact that periplasmic folding of MalE interferes with its translocation by AIDAc. To test this hypothesis, we used MalE31, a previously characterized mutant of MalE with a folding defect (2). MalE31 bears two amino acids substitutions: Gly→Asp and Ile→Pro at positions 32 and 33 of the mature protein, respectively. It was shown that the malE31 mutation decreases the in vitro refolding rate of MalE by more than 25-fold and that, when overproduced, this protein variant is degraded and forms inclusion bodies in the periplasm of E. coli (2-4). However, MalE31 can be completely refolded into a wild-type structure when purified and refolded in vitro (24). Thus, the mutation influences only the intrinsic folding ability of MalE.

The mutations resulting in MalE31 were introduced by site-directed mutagenesis in the pMalEAIDAc and pMalEΔlac plasmids using the QuikChange II site-directed mutagenesis kit (Stratagene). We showed by immunofluorescence that fixed whole bacterial cells expressing MalE31β could be labeled with anti-MalE antibodies, whereas bacteria expressing MalE31* were not, similarly to what was observed with wild-type MalE (Fig. 2B). This suggested that MalE31 fused to AIDAc escaped degradation and aggregation in the periplasm and was translocated to the bacterial surface. We also showed that MalE31 was accessible to externally added proteins by performing a limited trypsin proteolysis on whole cells (Fig. 3B). Figure 3B further shows that, contrary to what was seen with MalEβ, no band of 42 kDa could be seen in bacteria expressing MalE31β. Since a protein fragment containing MalE31 did not readily accumulate in the periplasm, our experiments suggest that MalE31 was fully translocated when fused to AIDAc, contrary to what was observed with wild-type MalE. It should be noted that, consistent with this hypothesis, more of the bacteria expressing MalE31β could be labeled by the anti-MalE antibodies than bacteria expressing MalEβ (compare Fig. 2A and B). Overall, these results suggest that translocation of MalE31 by AIDAc was more complete than that of MalE and therefore that the folding of wild-type MalE interfered with its translocation.

One could argue that the comparison of the folding kinetics of MalE and MalE31 is no longer valid when they are fused to AIDAc. To address this issue, we used plasmids allowing the expression of the alkaline phosphatase (PhoA) fused to the C terminus of MalE or MalE31, respectively (J. M. Betton, unpublished data). The resulting chimeras were soluble or aggregated, respectively (Fig. 5), and the fate of MalE and its mutant had little influence on that of its fusion partner, since PhoA was observed to be active in both cases (Betton, unpublished data). This shows that the foldings of the fusion partners are unaffected by one another.

FIG. 5.

Fusion of the alkaline phosphatase PhoA has no effect on the cellular localization of MalE or MalE31. ED9 bearing an empty control plasmid pTrc99A (−) or a plasmid allowing expression of a MalE-PhoA fusion (MalEphoA) or a MalE31-PhoA fusion (MalE31phoA), was grown as described in the legend to Fig. 2. The cultures were then normalized at equal OD₆₀₀s. Whole cells (WC) and periplasmic fractions (P) were obtained as described in the legend to Fig. 4. Aggregate fractions (A) were obtained by osmotic lysis of spheroplasts and extraction of the insoluble fractions of the lysates with 6 M urea. Samples of these fractions were resolved by SDS-PAGE and proteins revealed by immunoblotting using a rabbit antiserum against MalE. MW, molecular mass.

It could also be argued that the cleavage of MalE′ is unrelated to the translocation event directed by AIDAc. Again we find that this is unlikely, since there is no more cleavage between MalE and PhoA than between MalE31 and PhoA (Fig. 5).

We therefore conclude that the cleavage occurring between MalE and AIDAc in the MalE-AIDAc fusion is most likely due to the periplasmic folding of MalE interfering with the translocation mediated by AIDAc.

Our results do not preclude the possibility that some kind of folding is tolerated by autotransporters or even that some folding might be necessary for translocation. They strictly indicate that the nature or the result of the folding of MalE is incompatible with transport. The simplest explanation for this interference is that the size of the folded polypeptide blocks the translocation conduit of the autotransporter. It has been proposed that this conduit is either a pore of 20 Å in diameter, formed upon oligomerization of multiple copies of a TU (32), or the lumen of the β barrel of a single TU, with a diameter around 10 Å (18, 22, 25). Folded MalE is an ellipsoid of 30 by 40 by 65 Å (26), a size that is indeed incompatible with the dimensions of these channels. In that case, our experiments suggest that native passenger domains have to remain partially unfolded in order to get translocated through these narrow channels. This interpretation might be too simplistic, and the nature of the folding of MalE could also be an issue. As pointed out by others (22, 25), the view that the TU is solely responsible for translocation should be revised in light of the recent discovery of an extracytoplasmic machinery centered around Omp85 and involved in the insertion of proteins into the outer membrane, including autotransporters (23, 34, 36). The folding of MalE might cause it to be rejected by the Omp85 machinery and actively sorted to the periplasm. Furthermore, because autotransporters have been shown to be able to display heterologous polypeptides, we assume that passenger domains have no role in their own translocation. Yet we know that intramolecular interactions, such as the one mediated by the autochaperone region, are important for translocation. The folding of MalE could alter these critical interactions and thus interfere with translocation.

In conclusion, our results show, by a new approach, that periplasmic folding can interfere with translocation by autotransporters. This suggests that, in order to be translocated, native autotransporter passenger domains have to remain at least partially unfolded or that they transiently interact in a specific way with their own translocation domain or an external translocation apparatus. More work is required to sort between these hypotheses.