Type IV pilus biogenesis in Neisseria meningitidis: PilW is involved in a step occurring after pilus assembly, essential for fibre stability and function

Authors


E-mail pelicic@necker.fr; Tel. (+33) 1 40 61 54 82; Fax (+33) 1 40 61 55 92.

Summary

Type IV pili (Tfp) play a critical role in the pathogenic lifestyle of Neisseria meningitidis and N. gonorrhoeae, notably by facilitating bacterial attachment to human cells, but our understanding of their biogenesis, during which the fibres are assembled in the periplasm, then emerge onto the cell surface and are stabilized, remains fragmentary. We therefore sought to identify the genes required for Tfp formation in N. meningitidis by screening a genome-wide collection of mutants for those that were unable to form aggregates, another phenotype mediated by these organelles. Fifteen proteins, of which only seven were previously characterized, were found to be essential for Tfp biogenesis. One novel component, named PilW, was studied in more detail. We found that PilW is an outer-membrane protein necessary for the stabilization of the fibres but not for their assembly or surface localization, because Tfp could be restored on the surface in a pilW mutant by a mutation in the twitching motility gene pilT. However, Tfp-linked properties, including adherence to human cells, were not restored in a pilW/T mutant, which suggests that PilW is also essential for the functionality of the fibres. Together with the finding that PilW is important for the stability of PilQ multimers, our results extend the current model for Tfp biogenesis by suggesting that a multiprotein machinery in the outer-membrane is involved in the terminal stage of Tfp biogenesis during which growing fibres are not only stabilized, but also become perfectly functional.

Introduction

Tfp are surface organelles often found in human bacterial pathogens such as Neisseria meningitidis, N. gonorrhoeae, enteropathogenic Escherichia coli and Vibrio cholerae, where they play an important role in virulence (Mattick, 2002). In pathogenic Neisseria, Tfp facilitate bacterial attachment to target human cells (Swanson, 1973) and they influence the dynamics of the adherence process by powering a form of cell locomotion known as twitching motility and by promoting the formation of bacterial aggregates (Merz and So, 2000). In addition, Tfp are essential for the natural competence of pathogenic Neisseria for DNA transformation, a property which contributes to their virulence by promoting an exquisite genetic adaptability (Merz and So, 2000). Although this surprising number of Tfp-linked properties are either abolished or dramatically diminished in non-piliated mutants, they are still to be fully understood on a molecular level. This situation stems in part from our imperfect understanding of Tfp biogenesis (Tønjum and Koomey, 1997).

Some components of the Tfp biogenesis machinery have now been identified in pathogenic Neisseria, often by virtue of their homology to known components of the model piliated organism Pseudomonas aeruginosa (Lauer et al., 1993). Fibres are mainly constituted of subunits of pilin (PilE) that are synthesized as preproteins, which are packed through internal hydrophobic interactions between conserved N-terminal α helices, leaving hypervariable C-terminal globular regions exposed (Parge et al., 1995). Recently, it has been shown that Tfp biogenesis can be resolved into three genetically dissociable steps (Wolfgang et al., 2000): the assembly of the fibres occurs in the periplasm and is followed by their emergence onto the cell surface and stabilization. Three proteins, the prepilin peptidase PilD, PilF that contains an ATP-binding motif (Freitag et al., 1995), and PilG (Tønjum et al., 1995), are thought to be involved at the step of fibre assembly (Tønjum and Koomey, 1997; Wolfgang et al., 2000). However, the sole component whose role is precisely known is PilD, which cleaves the leader peptide from prepilins, thus allowing their subsequent incorporation into Tfp (Freitag et al., 1995). Then, the assembled fibres emerge onto the cell surface probably through pores formed in the outer-membrane by rings of 12 PilQ subunits (Wolfgang et al., 2000; Collins et al., 2001), whose expression seems to be affected in N. gonorrhoeae by the lipoprotein PilP (Drake et al., 1997). The third step required for Tfp biogenesis, defined as fibre stabilization (Wolfgang et al., 2000), most likely occurs after the assembled fibres have emerged on the surface (Morand et al., 2004) and involves the outer-membrane protein PilC (Jonsson et al., 1991). This step was defined by the finding that the piliation defect in a pilC mutant can be suppressed by a secondary mutation in the pilT gene (Wolfgang et al., 1998a). This indicated that PilC is neither required for the assembly of the fibres, nor for their localization on the bacterial surface. It also suggested that the effects of PilC and PilT on Tfp expression are counteractive (Wolfgang et al., 1998a). Therefore, as the effect of PilT is to antagonize Tfp expression by powering the retraction of the fibres (Wolfgang et al., 1998b; Merz et al., 2000), PilC was proposed to stabilize the Tfp (Wolfgang et al., 2000). Moreover, the balance between the PilC and PilT proteins is thought to regulate Tfp expression and contributes to make it a highly dynamic process (Morand et al., 2004).

It is likely that further analysis of Tfp biogenesis will improve our understanding of the functions mediated by these organelles, which are critical to the virulence of pathogenic Neisseria. We therefore sought to address this problem using a three-step approach consisting of (i) the identification of the complete set of components of the Tfp biogenesis machinery, because as many as 18 proteins are essential for this process in P. aeruginosa (Alm and Mattick, 1997), (ii) the functional characterization of each Pil protein and finally (iii) the fitting of the pieces of the machinery together. In this study, we completed the first step of this approach by performing a genome-wide mutational analysis and we characterized one of the novel Pil components, which brought new insights into N. meningitidis Tfp biogenesis/function.

Results

Fifteen proteins are essential for Tfp biogenesis in N. meningitidis

In this study, we used an ordered collection of defined mutants that contain non-polar transposon insertions into most of N. meningitidis strain 8013 non-essential genes, which was recently demonstrated to be a useful resource for functional genomics (Geoffroy et al., 2003). Because the ability to form bacterial aggregates is undoubtedly the Tfp-linked phenotype most readily assayable on a genomic-scale, we chose to identify Tfp biogenesis genes by pointing out the non-aggregative (Agg) mutants by phase-contrast microscopy. We identified 39 candidates (0.86% of the analysed mutants) of which we first re-transformed the mutations into the wild-type (WT) strain. We found that the transformants were indeed Agg, excluding the possibility, frequent in Neisseria species, that the observed phenotypes were because of phase variation events in genes unlinked to the engineered mutations. The selected mutants carried transposon insertions in 12 different genes (Table 1). We unambiguously determined whether these mutants were piliated or not, by scoring the presence or absence of Tfp by immunofluorescence (IF) microscopy after staining the fibres with a specific antibody (Fig. 1). Eleven of the identified genes were involved in Tfp biogenesis as the corresponding mutants were non-piliated. Importantly, the identification of non-piliated mutants with transposon insertions into five out of seven genes previously shown to be essential for Tfp biogenesis (pilD, pilE, pilF, pilG and pilQ) underlined the efficiency of our approach. We therefore identified six novel genes essential for Tfp biogenesis (NMB0886, NMB0887, NMB0888, NMB0889, NMB1309 and NMB1808) and one gene dispensable for fibre biogenesis but essential for the formation of multicellular aggregates (NMB0890).

Table 1.  Characteristics of the N. meningitidis Agg mutants identified in this study.
NMBNMANumber of hitsPiliation statusaName and predicted function
  • NMB and NMA denote the genes in MC58 and Z2491 genomes, respectively (Parkhill et al., 2000; Tettelin et al., 2000), which are homologous to the sequences flanking the transposon insertion sites. Number of hits indicates the number of insertions within a defined gene.

  • a

    . Piliation status was defined by specific IF staining (see Fig. 1).

  • b

    . These mutants were obtained by site-directed mutagenesis.

  • +, fibres were detected; –, absence of fibres.

001802641pilE, pilin
032921598pilF, pilus assembly protein
033221562pilD, prepilin peptidase
033321555pilG, pilus assembly protein
077009812b+unknown
088611061Tfp biogenesis protein
088711076Tfp biogenesis protein
088811081Tfp biogenesis protein
088911091Tfp biogenesis protein
089011102+pilX, minor pilin
130915234pilW, pilus stabilization/ maturation protein
180806543pilM, Tfp biogenesis protein
180906532bpilN, Tfp biogenesis protein
181006523bpilO, Tfp biogenesis protein
181106512bpilP, Tfp biogenesis protein
181206505pilQ, pilus secretin
Figure 1.

Presence or absence of Tfp in different strains, as monitored by IF microscopy after the fibres were stained with the 20D9 pilin-specific antibody.

At this stage, missing a few genes involved in Tfp biogenesis was expected because the library we used is predicted to contain mutations in approximately 85% of the N. meningitidis non-essential genes (Geoffroy et al., 2003). This prompted us to complete, in a second step, the previous list of genes by a reverse genetics approach. Interestingly, we noticed that two Tfp biogenesis genes found in the above screen belong to a putative operon of five genes: NMB1808, also named pilM, is the first gene of this putative operon, whereas pilQ is the last one. This suggested that the three intervening genes, pilN, pilO and pilP, might also be essential for Tfp biogenesis, which was documented for pilP (Drake et al., 1997). This was confirmed by generating insertion mutants into each of these genes and demonstrating that they were non-piliated by IF microscopy (Table 1). It should be noted that these mutations were, as above, non-polar and that the observed phenotypes were therefore not because of interferences with the pilQ gene expression, which was nonetheless confirmed by immunobloting (data not shown). Finally, we mined the genomic sequences of N. meningitidis in search of additional genes presenting similarities to known Tfp biogenesis components from P. aeruginosa, which led to the identification of a single candidate, NMB0770, which was homologous to pilZ (Alm et al., 1996). We generated a transposon insertion mutant in this gene and demonstrated, quite unexpectedly, that it was piliated although it was unable to form bacterial aggregates (Table 1).

Taken together, these results demonstrate that 15 proteins are essential for Tfp biogenesis in N. meningitidis, including PilC which could not be isolated in this approach because it is encoded by two alleles in the genome of strain 8013. In addition, we identified two genes (NMB0770 and NMB0890) that are not required for fibre biogenesis, but affect the functionality of Tfp as evidenced by the fact that the corresponding mutants were Agg.

Tfp-linked properties are abolished in a pilW mutant

Unravelling the mechanisms of Tfp biogenesis in pathogenic Neisseria implies that the identification of the components of the Tfp biogenesis machinery is to be followed by the detailed characterization of the corresponding mutants. In this study, we chose to focus on the non-piliated mutant of the gene corresponding to NMB1309 in the MC58 genome (Tettelin et al., 2000), hereafter named pilW. In addition to its apparent inability to form bacterial aggregates, all the functions associated with Tfp were impaired in a pilW mutant, similarly to what was seen when a non-piliated pilD mutant was used as a control. First, using a novel quantification assay that monitors the kinetics of decrease in optical density (OD) occurring in non-agitated liquid cultures upon sedimentation of the multicellular aggregates (Hélaine et al., 2004), we precisely measured the aggregative ability of a pilW mutant, which was confirmed to be nil (Fig. 2). Second, we found that its competence for DNA transformation was essentially abolished, with more than a 1000-fold reduction. Finally, its adhesion to human umbilical vein endothelial cells (HUVECs) was reduced approximately 200-fold (Fig. 3), and the same decrease in adhesion was observed when epithelial cells were used (data not shown).

Figure 2.

Quantification of the aggregative ability of different strains. Aggregation was quantified by measuring the decrease in OD600 that occurs upon sedimentation of bacterial aggregates in standardized non-agitated liquid cultures. The monitored differences in OD600 were only resulting from different aggregative abilities because all the strains grew as well as the WT in the assay medium (data not shown). Aggregation in non-piliated pilW and pilD mutants was indistinguishable and we therefore display only one curve for the sake of legibility. Values are the means of three to six independent experiments.

Figure 3.

Quantification of the adhesion to HUVECs of different strains. The 0 h points in this graph represent the sizes of the inocula. Adherence of non-piliated pilW and pilD mutants was indistinguishable and we therefore display only one curve for the sake of legibility. Error bars indicate standard errors of the means (n = 3).

PilW is a conserved outer-membrane protein

pilW, which is predicted to encode a protein of 253 amino acids (aa) with a molecular weight of 28.4 kDa, was present in all the sequenced pathogenic Neisseria strains, but also in 13 piliated isolates of N. meningitidis, N. gonorrhoeae and N. lactamica that were analysed by PCR using pilW-specific primers (data not shown). The protein was similar, with 27% of identical aa residues over the whole length of the proteins, to PilF of P. aeruginosa, a potential lipoprotein known to be involved in Tfp biogenesis but whose exact function is unknown (Watson et al., 1996). In pathogenic Neisseria, pilW is in the centre of a putative operon of three genes, starting and finishing with genes encoding conserved hypothetical proteins. Mutants in these two genes were present in the library and were thus re-analysed for aggregation and piliation, which demonstrated that the corresponding proteins are not involved in Tfp biogenesis (data not shown). A detailed computer analysis of the predicted PilW protein identified three tetratricopeptide repeats (TPR) domains (CDD entry cd00189), encompassing aa positions 36–69, 70–103 and 141–174 of the protein. These domains are often involved in protein–protein interactions, especially in contacts of TPR-proteins with multiprotein complexes (Blatch and Lassle, 1999). In addition, we found a prokaryotic membrane lipoprotein lipid attachment site (PROSITE entry PS00013), suggesting that the protein encoded by pilW might be anchored in the membrane.

To start with the experimental characterization of the PilW protein, we raised a rabbit serum against a purified recombinant PilW protein and used it in immunoblotting experiments. When whole-cell lysates were used, we detected a ∼28 kDa reactive species in the WT strain, in accord with the molecular weight predicted for PilW, which was absent in pilW mutants (Fig. 4A). This confirmed that the detected protein was indeed encoded by the pilW gene and that its expression was abolished in the mutants. Next, we demonstrated that PilW was localized in the membrane fraction of N. meningitidis (Fig. 4B). In order to refine PilW localization, we separated inner- and outer-membranes using sucrose density gradient centrifugation (Steeghs et al., 2001). The collected fractions were analysed for lactate dehydrogenase (LDH) activity as a marker for the inner-membrane (Steeghs et al., 2001), and for the presence of both lipo-oligosaccharide (LOS) and PilQ as markers for the outer-membrane, by silver staining and immunoblotting respectively. As expected, the LDH activity was essentially detected in the low-density region, corresponding to fractions 6, 7 and 8, whereas LOS and PilQ were mostly localized in the high-density region, corresponding to fractions 2, 3 and 4 (Fig. 5A). The analysis of the different fractions by immunoblotting using the serum against PilW, demonstrated that PilW is localized in the outer-membrane (Fig. 5B). Using the same fractions and a serum against PilC, we found that PilC is also localized in the outer-membrane (Fig. 5C), as previously suggested by immunogold labelling experiments (Rahman et al., 1997). Taken together, these data demonstrate that the pilW gene, which is essential for Tfp biogenesis, encodes an outer-membrane protein conserved in piliated isolates of Neisseria species.

Figure 4.

Immunoblot detection of PilW.
A. Analysis of whole-cell extracts of the WT strain and a pilW mutant. Equal amounts of proteins were loaded.
B. Analysis of cellular fractions of the WT strain. Fractions were resuspended in equal volumes and equal amounts of samples were loaded in each lane. Lanes: 1, whole cells; 2, membranes; 3, cytoplasm; and 4, supernatant. Positions of molecular weight standard proteins are indicated on the right in kDa.

Figure 5.

Analysis of fractions 1–10 collected after separation of N. meningitidis inner- and outer-membranes by sucrose gradient centrifugation. Identical aliquots from each fraction were separated by SDS-PAGE.
A. The sucrose concentration (○) and the LDH activity (□), expressed in international units (IU) l−1, in the fractions are indicated. Silver staining of LOS and immunoblot detection of PilQ.
B. Immunoblot detection of PilW.
C. Immunoblot detection of PilC.

PilW is involved in the terminal stage of Tfp biogenesis

As mentioned in the introduction, Tfp biogenesis is thought to occur in three steps: fibres are first assembled in the periplasm, then emerge onto the cell surface and are stabilized (Wolfgang et al., 2000). Based on its localization, it was possible that PilW was involved in a late stage of Tfp biogenesis as the other outer-membrane Tfp biogenesis proteins PilQ and PilC, which are, respectively, involved in the emergence of the fibres onto the cell surface and their subsequent stabilization. This prompted us to determine whether the piliation defect in a pilW mutant might be suppressed by a mutation in the twitching motility pilT gene, as already demonstrated in pilC and pilQ mutants (Wolfgang et al., 1998a; Wolfgang et al., 2000). We found by IF microscopy that the loss of PilT function in a pilW mutant indeed restored Tfp biogenesis (Fig. 1), which is evidence that PilW is not required for the assembly of the fibres. Examination of the Tfp in a pilW/T mutant by immuno-transmisssion electron microscopy (immuno-TEM), using the same pilin-specific antibody as above, showed that the fibres could be immuno-labelled on their entire length and are thus perfectly exposed on the bacterial surface (Fig. 6A). This is similar to what was previously reported in a pilC/T mutant (Wolfgang et al., 1998a) and therefore different from a pilQ/T mutant where the fibres are retained within membrane protrusions because of the absence of PilQ (Wolfgang et al., 2000). This indicates that PilW is involved in the stabilization of the fibres, but not in their emergence on the bacterial surface.

Figure 6.

Analysis of Tfp produced in various strains by TEM.
A. Fibres were immunogold-labelled with the 20D9 pilin-specific antibody, stained with phosphotungstic acid and then visualized. The left and central micrographs are at 40 000× magnification, whereas the right one is at 20 000× magnification.
B. Fibres were visualized directly after staining with phosphotungstic acid. The upper micrographs are at 40 000× magnification except the right one that is at 10 000×. The lower micrographs are all at 100 000× magnification.

As far as can be seen by TEM even at a very high magnification (Fig. 6B), the fibres in the pilW/T mutant and the WT strain are morphologically indistinguishable. Both strains produced few long Tfp that often came out of the cells as large bundles of laterally aggregated fibres. In contrast, the hyper-piliated pilT mutant harboured much more fibres, which also formed bundles but were shorter. Moreover, consistent with IF microscopy observations, we found using a novel whole-cell ELISA that the amount of fibres in the WT strain and the pilW/T mutant were similar, with a ratio of detectable PilE on cell surfaces of 1.18 ± 0.09. However, pilW/T fibres are not identical to WT fibres because they exhibit distinct physico-chemical properties. Indeed, the pilW/T mutant Tfp were found to be more susceptible to shearing. After 2 min of vortexing, virtually no Tfp could be detected on the surface of the pilW/T mutant by IF microscopy, whereas the WT strain still harboured a significant amount of fibres (data not shown). In addition, sheared pilW/T fibres were not efficiently purified by a widely used purification method that relies on dissociation and reassociation of pilin subunits. The yields of Tfp that could be obtained from the pilW/T mutant in this way were reduced approximately 16-fold relative to the WT strain (Fig. 7A). There was, however, no gross difference in the composition of the purified fibres, because equal amounts of Tfp preparations obtained from a WT strain and a pilW/T mutant contained comparable levels of associated PilC (Fig. 7B).

Figure 7.

Immunoblot analysis of Tfp purified from the WT strain (lane 1) and a pilW/T mutant (lane 2).
A. Immunoblot detection of PilE in Tfp preparations obtained from equal amounts of cells.
B. Immunoblot detection of PilC in Tfp preparations that were standardized based on total protein quantification. Positions of molecular weight standard proteins are indicated on the right in kDa.

Taken together, these results indicate that PilW exerts its action on Tfp biogenesis in a late stage occurring after the actual assembly of the fibres and their emergence on the surface. As a result of PilW action, the fibres are stabilized and somehow maturated.

PilW affects the stability of the PilQ multimers

As we localized PilW in the outer-membrane, together with the known Tfp biogenesis proteins PilC and PilQ, it was possible that PilW might exert its action on fibre biogenesis in concert with these proteins. We therefore tested whether the absence of PilW might have a negative effect on the stability of the PilQ homo-multimers enabling the emergence of the fibres onto the cell surface, which are retained in the stacking gel after SDS-PAGE (Drake and Koomey, 1995). Strikingly, while in the WT strain a significant amount of the PilQ protein is indeed retained in the stacking gel, no such high molecular weight species can be detected in a pilW mutant, which displayed a corresponding increase in the monomeric form (Fig. 8). The absence of PilQ multimers in the pilW mutant was not resulting from a reduction in PilQ production because the overall amounts of this protein that could be detected after phenol treatment of the extracts, which leads to a complete dissociation of the multimeric forms (Hardie et al., 1996), were comparable in a pilW mutant and the WT strain (Fig. 8). Mis-localization of PilQ cannot explain this defect either, because this protein was still localized into the outer-membrane in a pilW mutant (data not shown). Immunoblot analysis of other non-piliated mutants, namely pilD and pilP mutants, showed no significative change in the proportion of PilQ retained in the stacking gel (Fig. 8). This indicates that the absence of PilQ multimers in the pilW mutant does not simply result from the absence of Tfp and, most importantly, excludes the possibility that PilW might exert its action through PilP, a pilus biogenesis protein previously reported to affect the expression of PilQ multimers in N. gonorrhoeae (Drake et al., 1997). Moreover, the notion that the roles of PilP and PilW in N. meningitidis Tfp biogenesis are different is supported by the finding that, unlike the pilW/T mutant, a pilP/T mutant was non-piliated. Interestingly, no PilQ multimers could be detected in the pilW/T mutant either (Fig. 8), although it was clearly piliated. This suggests that the absence of fibres in a pilW mutant is not solely the consequence of the apparent absence of PilQ multimers and that the fibres in a pilW/T mutant emerge through a PilQ pore too unstable to be detected or by disrupting the membrane. However, the former possibility is more likely because we confirmed by IF microscopy that the fibres in a pilQ/T mutant, which are retained within membrane protrusions because of the complete absence of PilQ (Wolfgang et al., 2000), emerge on the bacterial surface very inefficiently.

Figure 8.

Immunoblot detection of PilQ monomers and multimers in various strains (upper panels). Equal amounts of protein were loaded in each lane as verified by immunoblot detection of the class 4 outer-membrane RmpM protein (lower panels). Proteins were either directly resuspended in SDS-PAGE sample buffer (left panels) or extracted with phenol before resuspension in SDS-PAGE sample buffer (right panels). Lanes: 1, WT; 2, pilQ mutant; 3, pilW mutant; 4, pilW/T mutant; 5, pilP mutant; and 6, pilD mutant. Positions of molecular weight standard proteins (in kDa) and position of the stacking gels are indicated.

Tfp in a pilW/T mutant are poorly functional

Because of its effect on the physico-chemical properties of the fibres, it was important to test whether PilW was important for Tfp functionality, which was made possible by  the  rescue  of  assembled  Tfp  on  the  surface  of  the pilW/T mutant. Therefore, we measured the ability of the pilW/T mutant to mediate the formation of aggregates and to promote adhesion to human cells. Competence for DNA transformation was not studied because it is known to be abolished in a PilT-negative background (Wolfgang et al., 1998b). We found, by observing a pilW/T mutant by phase-contrast microscopy, that it was able to form multicellular aggregates. However, its aggregative ability was intermediate between those of a non-piliated mutant and the WT strain, and definitely lower than that of a pilT mutant, which hyper-aggregated (Fig. 2). Interestingly, this was also the case for the adhesion of the pilW/T mutant to HUVECs, which was found to be only poorly restored, whereas a pilT mutant adhered to HUVECs even better than a WT strain (Fig. 3). Similar effects on adhesion were observed when epithelial cells were used (data not shown). The pilW/T mutant was therefore phenotypically similar to a pilC/T mutant whose adherence was not restored despite the rescue of Tfp on its surface (Wolfgang et al., 1998a). These results demonstrate that the fibres produced by a pilW/T mutant and the WT strain are not functionally equivalent, showing that PilW is essential for the functionality of the Tfp.

Discussion

The availability of a collection of defined mutants containing mutations into most of its non-essential genes (Geoffroy et al., 2003) prompted us to tackle the problem of Tfp biogenesis in N. meningitidis by a direct genomic-scale phenotypic analysis. The identification of the complete set of genes involved in Tfp biogenesis, and their subsequent detailed characterization is expected to boost our understanding of the properties linked to Tfp in pathogenic Neisseria, which contribute to make them important human pathogens.

Tfp biogenesis genes were identified by searching Agg mutants by phase-contrast microscopy, because the formation of bacterial aggregates is the most easily assayable Tfp-mediated phenotype. Thirty-nine mutants were unable to form aggregates and harboured transposon insertions in 12 genes. As expected, these mutants were non-piliated except those that contained insertions in NMB0890, which encodes a minor pilin involved in the formation of multicellular aggregates (Hélaine et al., 2004). This screening procedure was fully efficient as confirmed by the a posteriori analysis of the mutations in the library, which indicated that all the mutants containing transposon insertions in the above 12 genes were identified (our unpubl. data). Because the collection of mutants we used is not fully exhaustive, we completed the previous list of genes in a second stage, by mining the genome sequences of N. meningitidis and constructing additional mutants. We demonstrated or confirmed that pilN, pilO and pilP, which are probably part of an operon including the above identified pilM and pilQ genes, are essential for Tfp biogenesis. Finally, the only other gene (NMB0770) that was similar to a known Tfp biogenesis gene (pilZ) from P. aeruginosa (Alm et al., 1996), was surprisingly found to be dispensable for fibre biogenesis even though it affected N. meningitidis ability to form multicellular aggregates. This suggests that there are some subtle, but rare, differences between the Tfp biogenesis machineries in these two piliated bacterial species. Nevertheless, the number of genes essential for Tfp biogenesis in N. meningitidis and P. aeruginosa (Alm and Mattick, 1997) and their genomic organizations remain strikingly similar. In summary, we identified eight novel Tfp biogenesis genes, thereby putting the figure of N. meningitidis Pil proteins at 15. Although the possibility that some Tfp biogenesis genes have been missed cannot be excluded (they could be present in several copies that all need to be mutated in order to abolish Tfp expression, as pilC for example, or they may be essential for viability, which would preclude their identification with this approach) it seems likely that we have identified the majority, if not the totality, of N. meningitidis Tfp biogenesis genes.

The functional characterization of each component of the piliation machinery is the following step in the analysis of Tfp biogenesis in N. meningitidis. We therefore undertook the characterization of pilW, the homologue of the P. aeruginosa pilF gene whose function was elusive (Watson et al., 1996). First, we found that a pilW mutant was completely devoid of Tfp and presented therefore the same phenotypic defects as any other non-piliated mutant, that is, it was Agg, non-competent for DNA transformation and presented a dramatically decreased adherence to human cells. Then, because Wolfgang et al. recently provided evidence that Tfp biogenesis could be resolved into three steps (Wolfgang et al., 2000), assembly of the fibres followed by their emergence onto the surface and stabilization, we defined the step at which PilW is involved. We demonstrated that PilW, which we localized in the outer-membrane of N. meningitidis together with PilC and PilQ, was specifically involved in Tfp stabilization by finding that the fibres could be restored on the cell surface in a pilW mutant by a loss of PilT function, as previously demonstrated in pilC/T mutants (Wolfgang et al., 1998a). In addition, PilW also strongly affected the stability of the PilQ multimers that enable the emergence of the fibres onto the cell surface, which was not observed in a PilC-negative background (our unpubl. data). Interestingly, this action was not exerted through PilP because we could not find any obvious effect on PilQ multimers in N. meningitidis pilP mutants, unlike what was previously proposed in N. gonorrhoeae (Drake et al., 1997). Nevertheless, it is unlikely that the absence of detectable PilQ multimers in a PilW-negative background might be the cause of the lack of piliation in the pilW mutant, because the pilW/T mutant is perfectly piliated. Rather, these findings are probably an indication that in addition to its role in Tfp biogenesis, PilW interacts with PilQ and thereby stabilizes the multimeric forms of this protein. Together with previous reports, notably the one suggesting that PilQ and PilC interact (Drake et al., 1997), our results extend the current three-step model for Tfp biogenesis (Wolfgang et al., 2000) by suggesting that the two last steps might be performed by a discrete machinery in the outer-membrane, comprising at least PilC, PilQ and PilW. Moreover, it is tempting to speculate that the TPR motifs (Blatch and Lassle, 1999) found in PilW might be responsible for the protein–protein interactions making PilW a piece of this possible outer-membrane multiprotein complex, which could be tested by site-directed mutagenesis.

Another important finding of this work was that although PilW plays no role in the actual assembly or emergence of the fibres, it is essential for their functionality, as indicated by the fact that Tfp-linked phenotypic properties in piliated pilW/T mutants were extremely impaired. This is similar to what has been reported in pilC/T mutants (Wolfgang et al., 1998a), where the impaired adherence was regarded as evidence that the PilC protein, in addition to its role in fibre stabilization, was an adhesin directly responsible for Tfp-facilitated adherence to human cells as previously proposed (Rudel et al., 1995). How can one then explain that a pilW/T mutant is just as deficient in adherence despite the fact it was as piliated as the WT strain and harboured apparently normal fibres on its surface that displayed unaltered amounts of associated PilC? Although PilC copurification with the fibres in the pilW/T mutant was unaltered, we cannot exclude that the observed adherence defect might be the consequence of an impaired function of PilC. It is also possible that PilW, like PilC might be an adhesin essential for fibre biogenesis. Alternatively, the adherence defect in a pilW/T mutant, and maybe in a pilC/T mutant as well, might be the consequence of more general, but subtle, alterations in Tfp structure or stability. This possibility is further supported by the findings that (i) another Tfp-linked property, aggregation, was also extremely impaired in a pilW/T mutant and (ii) the fibres in a pilW/T mutant exhibited altered characteristics. Taken together, our results suggest that the fibres are not only stabilized during the last step of Tfp biogenesis, as was previously thought, but that they are somehow maturated, which is of definite importance for their functionality. The further analysis of the fibres produced by pilW/T and pilC/T mutants might therefore be instrumental in improving our understanding of the mechanisms by which Tfp facilitate the adherence of pathogenic Neisseria to human cells.

In conclusion, the findings here validate the approach we chose for the unravelling of the mechanisms of Tfp biogenesis in N. meningitidis. Eight novel Pil components were identified and the analysis of a single one of them already led to significant findings, which probably have a direct relevance in other bacteria expressing Tfp such as P. aeruginosa. It seems therefore realistic that the analysis of the remaining Pil components will have an important impact on our understanding of the Tfp-linked properties that make the pathogenic Neisseria species successful human pathogens.

Experimental procedures

Bacterial strains and growth conditions

N. meningitidis was grown on GCB agar medium (Difco) containing Kellogg's supplements (Pelicic et al., 2000) and when required, 100 µg ml−1 kanamycin, 5 µg ml−1 rifampicin and 3 µg ml−1 erythromycin. Transformants of E. coli were grown in liquid or solid Luria–Bertani medium (Difco) containing 100 µg ml−1 ampicillin. Growth occurred at 37°C in moist atmosphere containing 5% CO2. The ordered library of 4548 transposition mutants in the piliated N. meningitidis 8013 strain was described previously (Geoffroy et al., 2003). Additional mutants were engineered as follows. A 930 bp PCR product corresponding to the NMB0770 homologue was obtained using primers Nm981F 5′-GACAAATCGCCTTTGC ACCAGTAA-3′ and Nm981R 5′-CCGCTCGCCGTATTTGTC GGGTGG-3′, whereas the pilNOP genes were amplified as a unique 1.8 kb fragment using PilNPF 5′-ATTTTGTTCTG CGGCAGGTGCTGCC-3′ and PilNPR 5′-ATGAACAATTT AATCAAAATCAACC-3′. PCR products were cloned into pCRII-TOPO (Invitrogen), mutated by in vitro transposon mutagenesis and transformed into strain 8013, as described previously (Pelicic et al., 2000). Because all the transposon insertions we obtained in pilN were close to the end of this gene, we constructed the pilN mutant as follows. The previous pilNOP PCR product was cloned into pGEM-T (Promega) and a kanamycin resistance cassette, amplified and made blunt-ended as described previously (Pelicic et al., 2000), was cloned into an XcmI site present in pilN to give vector pYU1. This vector was then transformed into strain 8013 in order to replace the WT allele of pilN by allelic exchange. Because pilP, pilQ, pilT and pilW mutants are non-transformable, the pilP/T, pilQ/T and pilW/T double mutants were constructed by simultaneously transforming the WT strain with chromosomal DNAs extracted from a pilT mutant, where the gene is interrupted by an erythromycin resistance cassette (Pujol et al., 1999), and from pilP, pilQ or pilW mutants respectively.

Phenotypic analyses

Agg mutants were identified by resuspending mutants individually in the wells of a microtitre plate, which contained 100 µl RPMI 1640 medium supplemented with 10% heat-inactivated fetal calf serum Gold (both from PAA laboratories GmBH, Austria), and visualizing the aggregates after 1–2 h of growth using a phase-contrast microscope. Aggregation was quantified by monitoring the decrease in the OD600 that occurs over the time in non-agitated liquid bacterial cultures upon sedimentation of the multicellular aggregates (Hélaine et al., 2004). In brief, bacteria grown overnight on GCB agar plates were resuspended in RPMI-serum, filtered on 5 µm filters to eliminate large aggregates and adjusted to an OD600 of 0.6. Cells were then grown for 2 h under constant agitation. Finally, the suspensions were incubated at room temperature without shaking and the OD600 of the suspension was measured every 20 min.

Natural competence for DNA transformation was quantified by counting the number of rifampicin-resistant colony-forming units (cfu) obtained after 5 × 107 cfu of the studied strain were transformed (Pelicic et al., 2000) with 2 µg of chromosomal DNA extracted from a mutant of strain 8013 spontaneously resistant to rifampicin.

Adhesion of meningococci to HUVECs or human endometrial adenocarcinoma (Hec1-B) cells was quantified as described previously (Eugène et al., 2002). In brief, 24 well plates were seeded with 105 cells per well and the cell monolayers were infected with 1 ml of bacterial suspension containing approximately 3 × 107 cfu ml−1. After 30 min of contact, unbound bacteria were removed by three washes and the infection was continued for 5 h with washes every hour. Adherent bacteria, recovered by scraping the wells, were counted by plating appropriate dilutions on GCB agar plates.

SDS-PAGE, antisera and immunoblotting

Preparation of protein samples, protein quantification, SDS-PAGE separation, transfer to membranes and immunoblotting were carried out using standard molecular biology techniques (Sambrook and Russell, 2001). When indicated, protein samples were extracted by phenol treatment (Hardie et al., 1996). Briefly, after phenol was added to the samples (v/v), they were vortexed for 2 min. Two volumes of ice-cold acetone were then added and the mixture was vortexed again. After an overnight incubation at −20°C, the proteins were pelleted by centrifugation at 12 000 g for 15 min and the pellet was rinsed with ethanol before being resuspended in SDS-PAGE sample buffer. Detection of immobilized antigens was performed using ECL Plus (Amersham Biosciences). PilC was detected with the rabbit polyclonal serum 18P4 used at 1/2000 dilution (Morand et al., 2001). RmpM was detected using a specific mouse monoclonal serum at 1/5000 dilution (Morand et al., 2004). PilE was detected using the 5C5 mouse monoclonal serum (Marceau et al., 1998), whereas PilQ was detected using a rabbit polyclonal serum kindly donated by T. Tønjum (Tønjum et al., 1998); both sera were used at 1/5000 dilution.

To raise antibodies against PilW, we generated a PCR product using primers Eco5F3 5′-GGATATCAGCACTTCC TACCGCCCC-3′ and XhoR3 5′-GGCTCGAGTTGCAATTCT TCCGAGTA-3′, which contain overhangs corresponding to underlined restriction sites for EcoRV and XhoI respectively. This fragment was first cloned into pCRII-TOPO and then subcloned into pET-20b(+) (Novagen) restricted by EcoRV and XhoI. This introduced a five-histidine tag within the recombinant PilW protein. The protein was expressed in E. coli BL21::DE3 (Novagen), purified using Ni-NTA agarose columns (QIAGEN) and inoculated to a New Zealand rabbit.

Tfp detection, quantification and purification

Pili were detected after IF staining as previously described (Marceau et al., 1998), with minor modifications. Bacteria on coverslips were fixed for 20 min with a solution of 2.5% paraformaldehyde (PFA) in phosphate-buffered saline (PBS). After 5 min of incubation in PBS-0.1 M glycine and 5 min in PBS-0.2% gelatine, Tfp were stained with the anti-PilE 20D9 monoclonal antibody (Pujol et al., 1997) used at 1/1000 dilution in PBS-0.2% gelatine, while the bacteria were stained with ethidium bromide at 1/6000 dilution. The secondary antibody, also used at 1/100 dilution, was a goat anti-mouse antibody labelled with Alexa Fluor 488 (Molecular Probes). Fine structure of the fibres was determined by TEM using a JEOL JEM-100CX microscope operated at 80 kV. A drop of a dense bacterial suspension in PBS was placed on a sheet of Parafilm and the bacteria were adsorbed to formvar-coated grids for 5 min. The bacteria were fixed for 5 min with PBS-1% glutaraldehyde. Grids were then washed twice with water for 5 min, stained during 2 min with 1% phosphotungstic acid, air-dried and viewed. Immunogold labelling of the pili was performed as follows. Bacteria were resuspended in PBS-3% PFA and adsorbed to the grids for 15 min. The grids were then rinsed twice in PBS and placed sequentially onto drops on the following reagents at room temperature: PBS-50 mM NH4Cl (5 min), PBS-5% normal goat serum (5 min), and then the anti-PilE 20D9 monoclonal antibody diluted 1/100 in PBS-0.2% gelatine (for 60 min). After five washes in PBS-0.2% gelatine, the grid was placed for 60 min on a drop of protein A conjugated to 10-nm-diameter gold particles diluted 1/60 in PBS-0.2% gelatine. The grids were then subjected to five washes in PBS-0.2% gelatine, fixed in PBS-1% glutaraldehyde (15 min), and washed twice in distilled water. The grids were then treated with phosphotungstic acid for contrast as above, air-dried and viewed.

Piliation was quantified by a novel whole-cell ELISA assay using the above 20D9 monoclonal antibody (Hélaine et al., 2004). Briefly, bacteria grown on GCB agar plates and propagated in RPMI-serum were resuspended in PBS at an OD600 of 0.1 and 100 µl of serial twofold dilutions were coated in the wells of a microtitre plate before the plates were incubated without covers overnight at 45°C. Coated wells were washed seven times with PBS-0.1% Tween 80 (washing solution). The 20D9 antibody, diluted at 1/1000 in washing solution containing 5% skim milk to block non-specific binding, was then added and the plates were incubated 1 h at 37°C. After seven washes in the same way as above, a peroxidase-linked anti-rabbit IgG (Amersham) diluted 1/1000 in washing solution was added to the wells and the plates were incubated 1 h at 37°C. Finally, after seven washes, we added to the plate 100 µl of solution of O-phenylenediamine (Sigma), freshly prepared following the manufacturer's recommendations. The plates were incubated during 20 min at 37°C and read at 450 nm using a microtitre plate reader. The results are expressed as mean ratios of the OD450 ± standard deviations (n = 3). The validity of this quantitative method was confirmed by the complete absence of background observed when a non-piliated mutant was assayed.

Tfp purification was carried out as follows. Bacteria from two to three Petri dishes were washed twice in 10 ml of 50 mM Tris (pH 8), 150 mM NaCl and their OD600 was standardized. The bacterial pellet was resuspended in 11 ml of 150 mM ethanolamine (pH 10.5) and vortexed for 2 min. Organisms and large debris were removed by centrifugation at 12 000 g for 10 min. Smaller debris were removed by centrifugation at 50 000 g for 50 min, and 1 ml of ammonium sulphate-saturated 150 mM ethanolamine was added slowly to 9 ml of supernatant. Pilus filaments were recovered by centrifugation at 21 000 g for 15 min, resuspended in 200 µl of 50 mM Tris (pH 8) and dialysed for 1 h at 4°C against the same buffer, using 10 000 MWC Slide-A-Lyzer mini dialysis units (Pierce). Relative quantification of the purified fibres was achieved by immunoblotting, by comparing the PilE band intensity in serial dilutions of Tfp preparations made from identical amounts of cells.

Preparation of cellular fractions and separation of N. meningitidis membranes

Bacteria grown in liquid GCB medium were harvested during exponential growth and the cells were collected by centrifugation. The supernatants were filtered using a 0.2 µm filter and concentrated by precipitation with trichloroacetic acid. After washing the pellets with acetone at −20°C overnight, the secreted proteins were resuspended in 50 mM Tris (pH 7.5). The cell pellets were resuspended in 500 µl of 50 mM Tris (pH 7.5), 10 mM MgCl2 and the bacteria were lysed to release the proteins using the FastPROTEIN BLUE protocol (QBIOgene). Protein samples were then recovered and centrifuged for 2 h at 17 000 g. After this centrifugation step, the cytoplasm was localized in the supernatant, whereas the membranes were pelleted. The pellet was washed in 500 µl of 50 mM Tris (pH 7.5), 10 mM MgCl2, pelleted again as above and resuspended in 50 mM Tris (pH 7.5). Equal volumes of the resuspension solution were used in each case.

N. meningitidis inner- and outer-membranes were isolated and separated by sucrose density gradient centrifugation as described (Steeghs et al., 2001). The fractions were tested for LDH activity (Steeghs et al., 2001), as a marker for inner-membrane, and for the presence of PilQ and LOS, as markers for the outer-membrane. LOS detection was carried out by silver staining as previously described (Geoffroy et al., 2003).

Acknowledgements

We thank J.-L. Beretti for advice in protein purification, E. Eugène for help with IF microscopy, C. Fréhel for help with electron microscopy and T. Tønjum (University of Oslo, Norway) for the gracious gift of antimeningococcal PilQ antibody. We are grateful to G. Duménil, T. Pugsley, C. Recchi and J.-M. Reyrat for helpful suggestions and/or critical reading of the manuscript. This work was funded by INSERM and Université Paris V.

Ancillary