Formation of oligomeric rings by XcpQ and PilQ, which are involved in protein transport across the outer membrane of Pseudomonas aeruginosa

Authors

  • Wilbert Bitter,

    1. Department of Molecular Cell Biology and Institute of Biomembranes, Utrecht University, Padualaan 8, 3584 CH Utrecht, The Netherlands.
    Search for more papers by this author
  • Margot Koster,

    1. Department of Molecular Cell Biology and Institute of Biomembranes, Utrecht University, Padualaan 8, 3584 CH Utrecht, The Netherlands.
    Search for more papers by this author
  • Maita Latijnhouwers,

    1. Department of Molecular Cell Biology and Institute of Biomembranes, Utrecht University, Padualaan 8, 3584 CH Utrecht, The Netherlands.
    Search for more papers by this author
  • Hans De Cock,

    1. Department of Molecular Cell Biology and Institute of Biomembranes, Utrecht University, Padualaan 8, 3584 CH Utrecht, The Netherlands.
    Search for more papers by this author
  • Jan Tommassen

    1. Department of Molecular Cell Biology and Institute of Biomembranes, Utrecht University, Padualaan 8, 3584 CH Utrecht, The Netherlands.
    Search for more papers by this author

Wilbert Bitter , E-mail w.bitter@biol.ruu.nl; Tel. (30) 2533011; Fax (30) 2513655.

Abstract

Pseudomonas aeruginosa is able to translocate proteins across both membranes of the cell envelope. Many of these proteins are transported via the type II secretion pathway and adopt their tertiary conformation in the periplasm, which implies the presence of a large transport channel in the outer membrane. The outer membrane protein, XcpQ, which is involved in transport of folded proteins across the outer membrane of P. aeruginosa, was purified as a highly stable homomultimer. Insertion and deletion mutagenesis of xcpQ revealed that the C-terminal part of XcpQ is sufficient for the formation of the multimer. However, linker insertions in the N-terminal part can disturb complex formation completely. Furthermore, complex formation is strictly correlated with lethality, caused by overexpression of xcpQ. Electron microscopic evaluation of the XcpQ multimers revealed large, ring-shaped structures with an apparent central cavity of 95 Å. Purified PilQ, a homologue of XcpQ involved in the biogenesis of type IV pili, formed similar structures. However, the apparent cavity formed by PilQ was somewhat smaller, 53 Å. The size of this cavity could allow for the transport of intact type IV pili.

Introduction

Many Gram-negative bacteria are able to secrete proteins across the cell envelope into the extracellular medium. Various secretion pathways have been described but, currently, three main pathways, which are widely disseminated among Gram-negative bacteria, are being recognized: the type I pathway, the type II or general secretion pathway and the type III or contact secretion pathway (Pugsley, 1993; Salmond and Reeves, 1993). The opportunistic human pathogen Pseudomonas aeruginosa secretes several pathogenicity-related proteins into the environment, such as elastase, lipase, alkaline protease, exoenzyme S and exotoxin A (Liu, 1974; Nicas and Iglewski, 1986). Although all three secretion pathways are present in P. aeruginosa (Tommassen et al., 1992; Yahr et al., 1996), most of the known secreted proteins are transported via the type II pathway. This type of secretion involves two separate translocation steps across the inner and outer membranes (Tommassen et al., 1992; Pugsley, 1993; Salmond and Reeves, 1993). The exoproteins are synthesized with an N-terminal signal sequence, which is removed upon transport across the inner membrane via the Sec machinery (Pugsley, 1993; Salmond and Reeves, 1993). In the periplasm, these proteins adopt a considerable degree of tertiary conformation before translocation across the outer membrane can occur (Pugsley, 1992; Bartoli-German et al., 1994; McIver et al., 1995; Braun et al., 1996). In P. aeruginosa, 12 xcp genes have been identified that are essential for this second translocation step (Filloux et al., 1990; Bally et al., 1991; 1992; Akrim et al., 1993). Detailed knowledge of the mechanism of the type II secretion pathway is lacking. However, only one of the encoded proteins, XcpQ, is located in the outer membrane (Akrim et al., 1993) and is therefore the main candidate to constitute the actual protein translocation channel. XcpQ belongs to a large family of homologous proteins, generally called secretins (Genin and Boucher, 1994; Hardie et al., 1996a). This family includes not only the outer membrane components of type II secretion systems in various bacteria, but also components of the type III secretion pathway, proteins involved in DNA uptake (natural competence) and proteins involved in the biogenesis of type IV pili and of filamentous phages. Recently, several homologues of XcpQ have been shown to form highly stable, multimeric complexes too large to enter the separating gel in standard SDS–PAGE (Newhall et al., 1980; Kazmierczak et al., 1994; Hardie et al., 1996b; Chen et al., 1996). This complex could form the translocation channel.

In Klebsiella oxytoca, Neisseria gonorrhoeae and Erwinia spp., a specific lipoprotein is required for the protection of secretins against proteolysis and for their insertion into the outer membrane (Hardie et al., 1996a; 1996b; Drake et al., 1997; Shevchik et al., 1997). This chaperone might remain firmly associated with the secretin complex in the outer membrane (Daefler et al., 1997). However, the existence of such a lipoprotein is not necessarily a general characteristic of type II secretion systems, as it has not been identified thus far in P. aeruginosa and in several other bacteria with a type II secretion system. For the type II secretion system of Erwinia chrysanthemi, it has been shown that OutC and the secretin OutD are the gatekeepers of the species-specific secretion of proteins (Lindeberg et al., 1996). Moreover, it has been suggested recently that OutD itself interacts specifically with the secreted proteins (Shevchik et al., 1997). This would indicate that the secretin is not only the prime candidate for the formation of the actual protein translocation channel but that it also selects the proteins to be secreted. We have now isolated the multimeric complex of XcpQ and shown by electron microscopical examination that this secretin forms large oligomeric rings, which are likely to represent the protein translocation channel.

Results

XcpQ forms a multimeric complex

XcpQ is synthesized in very low amounts by P. aeruginosa, and high-level overexpression of its structural gene is lethal both to P. aeruginosa and to Escherichia coli. Therefore, cells of P. aeruginosa in which xcpQ was only slightly overexpressed by its presence on the low-copy-number plasmids pAK1 or pB28 were used for this study. In these plasmids, the xcpQ gene is under the control of the tac promoter. Induction of xcpQ expression by the addition of IPTG is lethal, but a low amount of XcpQ is formed in the absence of IPTG. This low expression, which is two- to fourfold more than in the wild-type strain, is sufficient to complement the secretion defects in the xcpQ mutant PAN1.

Immunoblot analysis of cell envelope proteins revealed that XcpQ formed, in agreement with the results reported for other secretins, a multimeric complex, which was too large to enter the separating gel in standard SDS–PAGE (Fig. 1). This complex was resistant to denaturation in 8 M urea, in 2% SDS or in several other detergents. However, upon prolonged heating in the presence of sample buffer containing 2% SDS, the complex disintegrated into XcpQ monomers (Fig. 1). Additionally, a number of intermediate-sized, oligomeric forms could be distinguished under the conditions applied. When gradient SDS–PAGE was used, the XcpQ complex did enter the separating gel (Fig. 2A, lane 2). The relative molecular mass (Mr) of the multimeric complex was estimated to be between 600 000 and 750 000. The XcpQ complex shows two or three bands on gradient SDS–PAGE. This multiple banding pattern is probably caused by the partial degradation of some of the subunits within the complex (see below). The electrophoretic mobility and the disintegration pattern observed upon heating in sample buffer were identical for XcpQ complexes originated from wild-type cells, from cells overexpressing XcpQ and from pAK1-containing cells expressing XcpQ as the only Xcp component owing to a deletion of the entire chromosomal xcp gene cluster (results not shown). Therefore, no other identified Xcp protein is part of this complex.

Figure 1.

. Immunoblot showing the XcpQ complex and its disintegration pattern. Cell envelope preparations of P. aeruginosa strain PAO25 containing pAK1 were resuspended in sample buffer containing 2% SDS and incubated for 7 min at varying temperatures. To visualize the intermediate-sized oligomeric forms, large amounts of cell envelope proteins had to be loaded. Subsequently, the proteins were separated on an 8% polyacrylamide gel and transferred to nitrocellulose. XcpQ was visualized with antibodies directed against an N-terminal domain of XcpQ.

Figure 2.

. A. Cell envelopes of xcpQ mutant PAN1 (lane 1), PAO25 (lane 2), phage D3112-resistant mutant PAN2, which showed no production of PilQ (lane 3), and PAN2 containing pB28 (plasmid containing the xcpQ gene; lane 4) were isolated from cells, ultrasonically disrupted in the presence of 1% NP40 to inactivate the endogenous elastase. The protein patterns were analysed on a 3–9% gradient polyacrylamide gel and visualized by silver staining (Ansorge, 1985). Lanes 5 and 6 show the XcpQ complex isolated from cell envelopes of PAO25 and PAN2 carrying pB28 respectively. One of these samples (lane 5) was incubated for 10 min at 95°C to disintegrate the XcpQ complex (lanes 7). A faint background band with an Mr of 59 000 was detected in all lanes, including lanes in which only sample buffer was loaded. B. The purified XcpQ complex (left lane) was incubated for 10 min at 95°C in sample buffer, separated on an 11% polyacrylamide gel, and the proteins were visualized by silver staining. The lane at the right shows molecular weight standard proteins.

Insertion mutagenesis of xcpQ

Secretins are supposed to consist of two domains. The C-terminal domain, which is conserved in all members of the superfamily, is supposed to be embedded in the outer membrane, and the N-terminal domain, which only shows conservation within the different subfamilies of secretins, is thought to interact in the periplasm with other components of the secretion systems (Brissette and Russel, 1990; Genin and Boucher, 1994). To determine which domain of XcpQ is required for oligomerization and whether oligomerization and the secretion function can be separated, linker insertion mutagenesis was performed. For this purpose, a modified version of the procedure of Wong et al. (1993) was used. The plasmid, pUWL6, containing xcpQ, was linearized by partial digestion with HaeIII or Bst UI as described in Experimental procedures and ligated with a kanamycin resistance cassette. Plasmid DNA was isolated from all resistant transformants and used to isolate the xcpQ fragments containing the cassette insertion. Subsequently, these fragments were cloned in broad-host-range vector pMMB67HE. In this way, we can ensure that most mutants will have their linker inserted in the xcpQ gene. Subsequently, the kanamycin resistance cassette was removed by BamHI digestion, leaving 12 nucleotide pairs of the cassette. The resulting clones should all contain insertions or small deletions within the xcpQ gene or in the region adjacent to this gene. Six hundred individual clones were used in a triparental mating with the xcpQ mutant strain PAN1 and selected for elastase secretion on LB agar plates containing elastin.

Eight clones that did not show a halo on elastin plates were selected for further characterization (Table 1). Analysis of the culture supernatants of these clones did not reveal the secretion of any protein, normally secreted via the type II pathway (results not shown). Only two of these mutants, E16 and H119, produced normal amounts of the XcpQ multimer (Fig. 3A, lanes 3 and 6 respectively), whereas mutant H105 produced smaller amounts of the complex (Fig. 3A, lane 4). Apparently, the ability of these mutant proteins to oligomerize is not or only partially affected, but, nevertheless, these mutants are totally defective in secretion. One of the mutants, H7, produced an XcpQ monomer with a higher electrophoretic mobility than that of wild-type XcpQ (Fig. 3B, lane 8). Sequence analysis revealed the presence of a large deletion, in addition to the BamHI linker insertion. H7 probably forms a multimer as well, as the monomer was only visible upon heating to 95°C in SDS buffer. However, a distinct oligomeric complex could not be discerned. The other mutant proteins, B7, H115, B1 and H8, were not produced in detectable amounts (Fig. 3). However, after growth of the cells in the presence of IPTG, monomeric forms of these proteins were detected, but they did not form high-molecular-weight complexes (results not shown).

Table 1. . Insertion sites of the 22 linker insertion and deletion mutants, identities of the inserted amino acids, functionality, lethality and complex formation. a. Position 1 is the first amino acid of the XcpQ precursor, 658 amino acids in total. For the insertions, the last amino acid before the insertions is given. For the deletions, the deleted amino acids are given.b. Functionality is based on the results with the elastin plates. ++, elastin degradation by strain PAN1 is already visible after overnight incubation (similar to wild type); +, delayed elastin degradation; −, no elastin degradation; D no elastin degradation by strain PAO25 after overnight incubation.c. Lethality is observed after overnight growth on LB plates containing 1 mM IPTG as an inducer for xcpQ under the control of the tac promoter.d. Complex formation is analysed by immunoblot on 3–9% gradient SDS–PAGE. ++, complex levels similar to wild-type xcpQ located on the same plasmid; + and ±, reduced amounts of complex present.e. The complexes formed by the mutants H40, H7 and H94 showed a smaller Mr compared with the wild-type complex.Thumbnail image of
Figure 3.

. A. Cell envelopes of strain PAN1 containing the xcpQ plasmid pB28 (lane 9) or one of the mutant plasmids with negative, i.e. pB7 (1), pH115 (2), pE16 (3), pH105 (4), pB1 (5), pH119 (6), pH8 (7), pH7 (8), or dominant-negative linker insertions, i.e. pH94 (10), pE1 (11), pB34 (12) were separated on a 3–9% gradient SDS–PAGE. The proteins were blotted onto nitrocellulose and analysed with antiserum directed against XcpQ. B. Cell envelopes of strain PAN1 containing the vector pMMB67HE (lane 10), xcpQ plasmid pB28 (lane 9) or one of the mutant plasmids with negative, i.e. pB7 (1), pH115 (2), pE16 (3), pH105 (4), pB1 (5), pH119 (6), pH8 (7), pH7 (8), or dominant-negative linker insertions, i.e. pH94 (10), pE1 (11), pB34 (12) were separated, after incubation at 95°C in the presence of SDS to disintegrate the complex, on an 11% SDS–acrylamide gel. In this experiment, two background bands are observed with Mrs close to that of the XcpQ monomer. The XcpQ complex and monomer are indicated.

When introduced into the wild-type strain PAO25, the mutant plasmids that failed to complement the xcpQ mutation in strain PAN1 did not show a dominant-negative phenotype, as evaluated on elastin plates. To investigate whether such dominant-negative mutations can be obtained at all, 300 different clones from the linker insertion bank were introduced into strain PAO25. The transconjugants were analysed for elastase secretion on elastin plates. In total, only three dominant-negative xcpQ mutant alleles were obtained. Restriction enzyme analysis and sequencing revealed that all three mutant plasmids did contain a deletion in the N-terminal domain, in addition to a partial BamHI linker insertion (Table 1). These overlapping deletions had probably arisen from the same illegitimate BamHI digestion (Experimental procedures). The dominant-negative mutations showed severely reduced secretion of proteins in wild-type cells (Fig. 4A). Mutant proteins E1 and B34 did form a complex when expressed in xcpQ mutant strain PAN1 (Fig. 3A, lanes 11 and 12). For deletion mutant H94, this could not be shown on an immunoblot, as the XcpQ-specific antiserum was raised against a domain of the protein that is completely removed by the large deletion in the mutant allele. However, the complex could be observed on a 3–9% gradient SDS–PAGE by silver staining (Fig. 4B, lane 3). The dominant-negative phenotype could be caused by the formation of inactive mixed complexes. To analyse the existence of mixed complexes, XcpQ multimers were isolated from cells expressing both wild-type XcpQ and the H94 mutant protein by sucrose gradient centrifugation (see below). A collection of XcpQ complexes with different Mrs was observed (Fig. 4B, lane 2), probably representing multimers with varying ratios of wild-type and mutant protein. The fact that the mixed complexes did not migrate exactly in between the intact wild-type (Fig. 4B, lane 1) and H94 complex (lane 3) could be caused by partial proteolytic degradation of some of the XcpQ subunits in the mixed complexes. Alternatively, the presence of mutant H94 subunits in the complex could change the conformation of the complex, resulting in a somewhat aberrant migration in SDS–PAGE.

Figure 4.

. A. Proteins from the culture supernatants of strain PAO25 containing plasmid pB28 (lane 1) or one of the dominant-negative linker insertion mutant plasmids pH94 (2), pE1 (3) or pB34 (4) were separated on a 12% polyacrylamide gel and visualized by Coomassie staining. Not all the proteins in the supernatant of PAO25 cultures are secreted via the type II pathway. Elastase, the most prominent protein secreted via this pathway, is indicated. B. Analysis of mixed complexes formed by wild-type XcpQ and mutant protein H94 (lane 2). These mixed complexes were isolated from cells of PAO25 containing pH94 by sucrose gradient centrifugation. As a control, cell envelopes of strain PAO25 (lane 1) and PAN1 containing plasmid pH94 (3) are shown. Proteins were separated on a 3–9% gradient SDS–PAGE and visualized by silver staining. The locations of the intact PilQ and XcpQ complexes and the H94 complex are indicated.

In total, 11 different plasmids that complemented the xcpQ mutation of PAN1 were isolated from the entire linker insertion bank. One of the permissive insertions, E37, was located within the region coding for the signal sequence, whereas the others were dispersed over the entire length of the gene (Table 1). The permissive mutants complemented the xcpQ mutation of strain PAN1 for elastase secretion on elastin plates, although three mutants, E55, E101 and H40, showed a somewhat delayed elastin degradation (Table 1). However, the amount of secreted proteins in the culture supernatants of these different clones varied considerably. Two of the mutants, E55 and E101, did not seem to complement the xcpQ mutation at all in this assay (Fig. 5A, lanes 7 and 8), whereas two other mutants, H40 and B62, showed only a partial reversal of the secretion-negative phenotype (Fig. 5A, lanes 9 and 13). The other mutants restored protein secretion to wild-type levels. Most of the mutant proteins formed XcpQ multimers in similar amounts to the wild type (Fig. 5B). However, except for B62, the mutant proteins that did not fully complement the secretion defect of strain PAN1 (Fig. 5A) also showed strongly reduced amounts of complex (Fig. 5B). This indicates that the complexes of these functional proteins are either not effectively produced or not stably incorporated into the outer membrane. Surprisingly, the mutant H40 produced a complex with a lower Mr than the wild-type complex (Fig. 5B, lane 9). Disintegration of this complex by heating to 95°C in sample buffer resulted in the appearance of two major degradation products (results not shown). This could indicate that the H40 complex still retains some of its functionality with partially degraded subunits.

Figure 5.

. Analysis of permissive linker insertions. A. Proteins from the spent culture medium of xcpQ mutant strain PAN1 containing the vector pMMB67HE (lane 1), xcpQ plasmid pB28 (2) or one of the mutant plasmids with permissive linker insertions, i.e. pE37 (3), pH126 (4), pE8 (5), pH133 (6), pE55 (7), pE101 (8), pH40 (9), pE2 (10), pB65 (11), pH64 (12) or pB62 (13), were separated on a 12% polyacrylamide gel and visualized by Coomassie staining. Elastase, which is the most prominent protein secreted via the type II pathway, is indicated. B. Cell envelopes of strain PAN1 (lane 1) or PAN1 containing the vector pMMB67HE (2) or one of the different mutant plasmids with permissive linker insertions, in the same order as in (A), were separated on a 3–9% gradient SDS–PAGE, blotted onto nitrocellulose and analysed with antiserum directed against XcpQ.

Growth of strain PAN1 containing the different mutant plasmids on medium containing IPTG was lethal for the permissive mutants and the complex-forming negative mutants, but not for the four mutants that did not form XcpQ complexes. Apparently, the lethality of XcpQ is strongly correlated with complex formation and not with its functionality.

Purification of XcpQ

The stability of the multimeric XcpQ complex was exploited to isolate the intact complex from sodium cholate-extracted outer membrane preparations of strain PAO25 containing pB28. The XcpQ complex was solubilized in an SDS/glycerol solution and subsequently purified by sucrose gradient centrifugation. The XcpQ complex specifically concentrated at 26–29% sucrose. In addition to the XcpQ complex and some monomeric XcpQ, this preparation showed only one contaminating protein, with a higher Mr than that of XcpQ (Fig. 2A, lane 5). As P. aeruginosa produces at least one additional secretin under standard growth conditions, PilQ protein, which is involved in the biogenesis of type IV pili, we reasoned that this protein might have co-fractionated with XcpQ. To examine this possibility, cell envelope protein patterns of P. aeruginosa strain PAK and of its pilQ mutant derivative R177 were analysed on a 3–9% gradient acrylamide gel. Silver staining of this gel revealed the absence of a large protein complex in strain R177, which had the same Mr as the contaminating complex in the XcpQ preparation from strain PAO25 (results not shown). Hence, the contaminating high-molecular-weight band in the XcpQ preparation (Fig. 2A, lane 5) most probably represents the oligomeric PilQ complex of strain PAO25. Interestingly, this complex was even more stable than the XcpQ complex, as it remained intact under conditions that led to the complete dissociation of the XcpQ complex (Fig. 2A, lane 7).

To obtain a PilQ-deficient derivative of strain PAO25, spontaneous mutants resistant to the pili-specific phage D3112 were isolated (Experimental procedures). One of the mutants obtained, designated PAN2, did not produce detectable amounts of PilQ on a silver-stained 3–9% gradient polyacrylamide gel (Fig. 2A, lane 3). The XcpQ complex was isolated to homogeneity from cell envelope preparations of PAN2 containing pB28 (Fig. 2A, lane 6). Complete disintegration of the XcpQ complex resulted in the appearance of a major protein with an Mr of 64 000 (Fig. 2B) and a minor band of 45 000 (hardly visible on Fig. 2B). Both bands reacted with antiserum directed against XcpQ (results not shown). This demonstrates that the band of 45 000 is probably a degradation product of XcpQ and that the complex is composed of XcpQ subunits only. The 45 000 band could also be responsible for the multiple banding pattern of the XcpQ complex (Fig. 2A), as the ratio of intact XcpQ subunits and degradation products in the oligomeric complexes could be varying. This pattern could not be changed by using different protease inhibitors during the purification procedure or by using an elastase-negative mutant, PAN6, of P. aeruginosa (results not shown).

Electron microscopic analysis of XcpQ and PilQ

In view of the size and the stability of the XcpQ complex, we reasoned that it might be possible to visualize it by electron microscopy. Indeed, particles with the appearance of rings with an apparent large central cavity (Fig. 6A and B) were observed in the XcpQ preparation by negative-stain electron microscopy. The majority of the particles had the same orientation and showed the same dimensions. These molecules had an average diameter of 198 ± 13 Å (n = 34) with an apparent central pore of 95 ± 10 Å. In addition to the ring-shaped structures, a few large particles with undefined structures were also observed, which could be aggregated XcpQ complexes or aggregated subunits of XcpQ.

Figure 6.

. Negative-stained XcpQ multimeric complexes visualized by electron microscopy. Overall picture (A) and individual images (B) of the XcpQ complex are shown. Bar represents 20 nm.

We expected that all secretins, which are involved in various transport processes across bacterial outer membranes, formed similar large channels. To substantiate this hypothesis, we attempted to isolate PilQ complexes from the xcpQ mutant strain PAN1. However, these attempts were not successful, because of extensive degradation of PilQ during the isolation procedure. As this degradation could be caused by elastase that accumulates in the periplasm of the xcpQ mutant, the structural gene for elastase, lasB, was disrupted (Experimental procedures). The resulting strain, PAN7, could be used to purify the intact PilQ complex by sucrose gradient centrifugation.

Electron microscopic analysis of the PilQ complexes revealed ring-shaped structures similar to those of XcpQ, although with slightly different dimensions (Fig. 7). The average diameter of the PilQ complex is 183 ± 12 Å (n = 16) with an apparent central pore of 53 ± 8 Å. This shows that both of these secretins indeed form similar ring-shaped structures, in which the monomers enclose a large central cavity.

Figure 7.

. Negative-stained PilQ multimeric complexes visualized by electron microscopy. Bar represents 20 nm.

Discussion

As has been shown for several of its homologues (Newhall et al., 1980; Kazmierczak et al., 1994; Hardie et al., 1996b; Chen et al., 1996), the secretin XcpQ of P. aeruginosa forms large multimeric complexes. Secretins are supposed to consist of two domains. The C-terminal region, corresponding to residues 326–605 of XcpQ, is conserved in all members of the superfamily and is shown to be important for complex formation and located in the outer membrane (Russel and Kazmierczak, 1993; Chen et al., 1996). The N-terminal region, corresponding to residues 53–300, is thought to extend in the periplasm and to interact with other components of the export apparatus (Brissette and Russel, 1990; Genin and Boucher, 1994). The results of our mutant analysis (Table 1) are largely consistent with such a two-domain organization. Several deletions in the N-terminal domain were obtained, which did not interfere with multimerization of the protein. Hence, it appears that the C-terminal domain is sufficient for multimerization. However, we have also isolated three insertions in the N-terminal domain, B7, H115 and B1, which interfered with complex formation. Possibly, incorrectly folded N-terminal domains in these mutant proteins prevent appropriate subunit interactions in the C-terminal domains. Furthermore, in addition to the deletions, several insertions, E16, H105 and H119, were obtained in the N-terminal domain that did not prevent oligomerization, but did inhibit the functioning of the complex. Probably the interaction of these mutant proteins with other Xcp components or with the exoproteins is disturbed, which is consistent with the proposed function of the N-terminal domain of the secretins. In contrast, all insertion mutants in the C-terminal domain that did form multimers, E101, H40, E2, B65, H64 and B62, were functional. Complex formation was also shown to be strongly correlated with lethality of XcpQ. This lethality could be caused by the lack of a specific chaperone in P. aeruginosa that directs the complex to the outer membrane, as has been shown for PulS and the secretin PulD (Hardie et al., 1996b; Daefler et al., 1997).

All outer membrane proteins, of which the crystal structure has been resolved, form β-barrels, composed of antiparallel, amphipathic β-strands (Weiss et al., 1991; Cowan et al., 1992; N. Dekker, personal communication). The prediction of these transmembrane β-strands is rather difficult. However, these strands are relatively well conserved in homologous proteins, whereas the connecting cell surface loops are variable and permissive for mutagenesis. When we screened the sequence of the C-terminal domain of XcpQ, based on secondary structure predictions, the amphipathic character of transmembrane β-strands and the algorithm of Gromiha et al. (1997), for transmembrane β-strands, we found 13 putative transmembrane β-strands (Fig. 8), of which the character was conserved in other secretins. In contrast, hardly any of such sequences were found in the N-terminal domain. Interestingly, the insertion mutant protein, H8, which disturbs the functioning and oligomerization of the protein, is located in a predicted β-strand (Fig. 8). Also, the insertion in mutant protein E101 is located in such a strand, and this protein was very poorly functional. In contrast, all the other mutations in the C-terminal domain are permissive with respect to functioning of the protein, and they were not located in predicted β-strands (Fig. 8). Hence, these results support the putative β-barrel structure of the C-terminal domain of the secretins.

Figure 8.

. Proposed transmembrane β-strands in the C-terminal domain (aa 320 to aa 658) of XcpQ. The potential transmembrane β-strands are shown in bold. The potential periplasmic domains, based on the assumption that the N-terminal domain is located in the periplasm and that the shortest loops are generally also located in the periplasm, are shown in italics. The different linker insertions in this domain are indicated.

The protein translocation channel of the type II secretion systems in Gram-negative bacteria has to be sufficiently large to accommodate the transport of completely folded proteins. Here, we show that the only known outer membrane protein of this secretion system in P. aeruginosa, XcpQ, forms a homomultimeric complex, which appears as a ring-shaped structure with a large apparent central cavity upon electron microscopic evaluation. The size of the cavity formed by XcpQ, i.e. 95 Å, would be sufficient to allow for the transport of completely folded proteins. For example, the maximum diameter of elastase, which is secreted via the type II system, is approximately 60 Å (Thayer et al., 1991). The cavity formed by XcpQ is considerably larger than the apparent pore of 20 Å present in the recently identified protein-conducting Sec61 channel of the endoplasmic reticulum (Hanein et al., 1997). However, the Sec61 channel is involved in the transport of linear polypeptides with, at the most, some secondary structure. As bacterial outer membranes form permeability barriers for harmful compounds, such as antibiotics, and generally only allow the passage of small hydrophilic molecules by diffusion via the porins, the large apparent pores formed by XcpQ complexes must be gated. Possibly, the complex exists in two different conformations. The purified XcpQ complexes seemed to be present in one conformation only. Alternatively, other Xcp proteins could form a plug that closes the channel. The so-called pilin-like Xcp proteins, i.e. XcpT, U, V, W and X (Tommassen et al., 1992; Bleves et al., 1998), would be obvious candidates to fulfil this function by forming some rudimentary pilus-like structure or complex that interacts with XcpQ.

The dimensions of the resolved structures of outer membrane porins (Weiss et al., 1991; Cowan et al., 1992) and outer membrane phospholipase A (N. Dekker, personal communication) can be used to calculate the number of subunits that could make up an oligomeric ring with the dimensions of XcpQ. If XcpQ contains 12 membrane-spanning β-strands and if each XcpQ subunit forms a cylindrical β-barrel, the complex should contain 14–18 subunits. However, if XcpQ contains 14 membrane-spanning β-strands, 12–15 monomers could make up the complex. Tentatively, the complexes consisting of wild-type XcpQ and the dominant-negative mutant H94 show 12–14 different bands on gradient SDS–PAGE, which is another indication that the number of XcpQ monomers within a secretin complex might be higher than the 10–12 subunits estimated for the complex of the type pIV secretin (Kazmierczak et al., 1994).

P. aeruginosa also produces a second secretin under normal conditions, namely PilQ, which is involved in the biosynthesis of type IV pili. This protein appeared to form a high-molecular-weight complex, which initially co-fractionated with the XcpQ protein. PilQ could be purified from cell envelopes of an XcpQ-deficient strain. This protein complex forms large ring-shaped structures similar to XcpQ, suggesting that all secretins form similar structures. Indeed, the YscC protein of Yersinia enterocolitica, which is involved in type III secretion, is another secretin that was recently shown to form such structures (Koster et al., 1997). The size of the apparent cavity of the PilQ complex, i.e. 53 Å, was somewhat smaller than that of XcpQ, but is in perfect agreement with the size of the P. aeruginosa type IV pili. These pili have been shown to form cylinders with an outer diameter of 52 Å (Folkhard et al., 1981). P. aeruginosa strain PAO25 probably even has the ability to produce a third secretin, namely PscC, a YscC homologue involved in type III protein secretion (Yahr et al., 1996). However, a third protein complex was not detected on gradient SDS–PAGE. This could be because of the specific regulation of the expression of type III secretion systems, which are maximally produced in cation-deficient medium.

Experimental procedures

Bacterial strains, plasmids and growth conditions

The following bacterial strains were used in this study. Strain D40ZQ is a mutant of PAO1 with a deletion of the entire xcp gene cluster (generously provided by M. Bally). For establishing the nature of the contaminating protein in the XcpQ preparation, the P. aeruginosa strain PAK and its pilQ mutant derivative R177 (Martin et al., 1993) were used. For most experiments, P. aeruginosa strain PAO25 (Haas and Holloway, 1976) was used or derivatives of this strain. Strain PAN1 is an xcpQ deletion mutant with a gentamycin resistance cassette (Alexeyev et al., 1995) inserted at the position of xcpQ on the chromosome. This was achieved by cloning the DNA fragment adjacent to xcpQ with the gentamycin omega cassette on the broad-host-range vector, pUR6500HE (Frenken et al., 1993), and introducing this vector into strain PAO25. Subsequently, the cells were cured for this plasmid by the introduction of vector pMMB67HE (Fürste et al., 1986) and simultaneously selected for double cross-overs on 2 μg ml−1 gentamycin. Strain PAN2 is a spontaneous phage D3112-resistant mutant (Darzins and Casadaban, 1989), which does not produce detectable amounts of PilQ on a silver-stained 3–9% gradient polyacrylamide gel. Strains PAN6 and PAN7 are lasB mutants of strain PAN1 and PAN2 respectively. They were isolated according to P. Braun et al. (in preparation). All Pseudomonas strains were grown under aerobic conditions at 37°C in Luria broth (LB). The plasmids pAK1 and pB28 contain the xcpQ gene under the control of the tac promoter in vectors pUR6500HE (Frenken et al., 1993) and pMMB67HE (Fürste et al., 1986) respectively. pUWL6 is a derivative of pUC19, containing the xcpQ gene on a PstI–SmaI fragment of pAX24. Cosmid pAX24 contains the entire xcp gene cluster (Filloux et al., 1990).

Insertion mutagenesis

Insertion mutagenesis in xcpQ was performed using a modified procedure of Wong et al. (1993). The plasmid pUWL6, containing xcpQ, was linearized by partial digestion with 1 U of HaeIII or Bst UI in the presence of 40 μg ml−1 ethidium bromide at 37°C. There are 53 cleavage sites altogether for these two enzymes in xcpQ. The linearized plasmid was ligated with a kanamycin resistance cassette excised from plasmid pBSL97 with SmaI (Alexeyev et al., 1995). Plasmid DNA was isolated from all, both ampicillin and kanamycin, resistant transformants (93 for HaeIII and 474 for Bst UI) and used to isolate the HindIII/StuI xcpQ fragments containing the cassette insertion. Subsequently, these fragments were cloned in HindIII/Ecl 136II-digested broad-host-range vector pMMB67HE (Fürste et al., 1986). In this way, we could ensure that most mutants contained their linker inserted in the xcpQ gene. The xcpQ alleles in this vector are under the control of the tac promoter. Again, plasmid DNA was isolated from all the transformants (more than 500), and this material was digested with BamHI to remove the kanamycin resistance cassette but leave 12 nucleotide pairs of the cassette (GGGGGATCCCCC). In total, 600 resulting clones were mobilized to strain PAN1 by a modified triparental mating. The acceptor strain PAN1 was mixed with the helper strain containing plasmid pRK2013 and plated on an LB plate. After 2 h of growth at 37°C, the 600 clones were streaked on these plates. The plates were incubated overnight at 37°C, and samples of the streaks were analysed on selection plates, containing 25 μg ml−1 nalidixic acid and 25 μg ml−1 piperacillin. Subsequently, complementation of the xcpQ phenotype was examined on plates containing 1% elastin (Sigma) in a top layer. One of the clones, B28, which contained the BamHI insertion downstream of the xcpQ gene, was used as a positive control. In total, 300 clones were mobilized to strain PAO25, as described above. The clones were analysed on plates containing 1% elastin. After 20 h of growth at 37°C, only three clones showed no degradation of elastin. These dominant-negative clones were analysed by restriction analysis and sequencing. The mutants had probably arisen from an illegitimate BamHI cleavage at the same position, as the 3′ end of the deletion contained neither the 3′ part of the 12 nucleotide pairs nor the remains of a HaeIII or Bst UI cleavage site. At the fusion point, there was a sequence that originally resembled a BamHI palindromic sequence, except for one mismatch, i.e. AGATCC instead of GGATCC.

SDS–PAGE and immunoblotting

Cell envelope preparations of P. aeruginosa were isolated from ultrasonically disrupted cells in the presence of 1% NP40 to inactivate the endogenous elastase (McIver et al., 1995). These preparations were separated on 11% or 3–9% SDS–polyacrylamide gels and, subsequently, the proteins were blotted onto nitrocellulose or the proteins were visualized directly by a silver staining according to Ansorge (1985). Peroxidase activity on blots was visualized by enhanced chemiluminescence (ECL). Secreted proteins were precipitated from spent culture medium with 5% trichloroacetic acid, pelleted and washed with cold acetone.

Antibodies

Polyclonal antiserum directed against XcpQ was raised by using a GST fusion protein (Smith and Johnson, 1988) containing an N-terminal domain of XcpQ (amino acid residues 12–100 of mature XcpQ). This GST fusion protein, ML1, was purified from E. coli cells overexpressing GroE (Stieger and Caspers, 1997) by affinity chromatography on a glutathione agarose column, followed by gel filtration (Sephacryl S-300, 90 cm). The purified ML1 protein was used to immunize rabbits with an intradermal injection (Eurogentech).

Purification of XcpQ and PilQ

XcpQ was isolated from cells of P. aeruginosa strain PAO25 and from mutant PAN2 containing pB28 and grown overnight in LB. Cells were ultrasonically disrupted in the presence of protease inhibitors (Complete; Boehringer Mannheim), and the cell envelope fraction was extracted with 4% sodium cholate in buffer A (100 mM NaCl, 10 mM EDTA, 10 mM Tris, pH 7.5). Subsequently, the XcpQ complex in the pellet fraction was solubilized at 37°C in a buffered glycerol/SDS solution (containing 2% SDS, 20% glycerol and 0.2 M β-mercaptoethanol in buffer A), dialysed against 0.1% SDS in buffer A (buffer B) and concentrated in a microsep 300 (Filtron). The concentrated material was applied to a sucrose gradient (15–35% sucrose in buffer B) and centrifuged for 24 h at 3500 r.p.m. in a SW41 rotor (Beckman). The XcpQ complex specifically concentrated at 26–29% sucrose. Finally, the XcpQ complex was concentrated and sucrose was removed from the solution by using a microsep 300. PilQ was isolated using the same procedure, but with cell envelopes of PAN7 as the starting material.

Electron microscopic analysis of XcpQ and PilQ

XcpQ and PilQ proteins were adsorbed to freshly prepared carbon-coated nickel grids for 1 min at room temperature, washed twice for 2 min with water and stained for 45–60 s with 1% uranyl acetate in water. Proteins were fixed with 2% paraformaldehyde and 0.2% glutaraldehyde in PBS for 5 min directly after adsorption. Proteins were observed in a Philips electron microscope (CM10) operated at 80–100 kV.

Acknowledgements

The authors thank the Phabagen collection (Fraukje van Asma, curator), Marc Bally, Alain Filloux and John Mattick for the gift of strains and phages and Fons Cremers, Henk Pluijgers, Niek Dekker and Carien Dekkers for helpful discussions. This study was supported by EU grant BIO4-CT96-0119.

Ancillary