PilX, a pilus-associated protein essential for bacterial aggregation, is a key to pilus-facilitated attachment of Neisseria meningitidis to human cells


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


The attachment of pathogenic Neisseria species to human cells, in which type IV pili (Tfp) play a key but incompletely defined role, depends on the ability of these bacteria to establish contacts with the target cells but also interbacterial interactions. In an effort to improve our understanding of the molecular mechanisms of N. meningitidis adherence to human cells, we screened a collection of defined mutants for those presenting reduced attachment to a human cell line. Besides underscoring the central role of Tfp in this process, this analysis led to the identification of mutants interrupted in a novel gene termed pilX, that displayed an adherence as impaired as that of a non-piliated mutant but quantitatively and qualitatively unaltered fibres. Moreover, the pilX gene, which encodes a pilin-like protein that copurifies with Tfp fibres, was also found to be essential for bacterial aggregation. We provide here several piece of evidence suggesting that PilX has intrinsic aggregative but no adhesive properties and that the reduced numbers of adherent bacteria seen with a pilX mutant result from the absence of interbacterial interactions. These data extend the current model for Tfp-facilitated adherence of N. meningitidis to human cells by suggesting that the pili lead to an increase in net initial adherence primarily by mediating a cooperation between the bacteria, which is supported by the finding that a major effect on initial adherence could be observed in a wild-type (WT) genetic background after a mechanical removal of the bacterial aggregates.


The attachment of pathogenic bacteria to host cells is an early and crucial event in the onset of an infection. In Gram-negative bacteria, various filamentous surface organelles termed pili or fimbriae are frequently involved in this attachment (Soto and Hultgren, 1999). How these organelles facilitate adhesion is probably best understood for the pili assembled via the chaperone-usher pathway. These fibres are composite structures consisting of a rod of pilin subunits joined to a short fibrillar tip, terminated by an adhesin directly responsible for the attachment to host cells (Soto and Hultgren, 1999). For example, the P pili that are found in the uropathogenic strains of Escherichia coli associated with acute pyelonephritis, are capped by the PapG adhesin (Kuehn et al., 1992), which binds specific glycolipid receptors present on uroepithelial cells (Roberts et al., 1994).

Tfp are a separate and homogeneous class of pili that have been shown by human experimentation to play a pivotal role in the virulence of several human pathogens, including Vibrio cholerae, enteropathogenic strains of E. coli (EPEC) and pathogenic Neisseria species (Herrington et al., 1988; Bieber et al., 1998; Cohen and Cannon, 1999). Tfp are helical structures (Parge et al., 1995), which consist mainly of pilin subunits harbouring a highly conserved sequence signature at their N-termini that are assembled by conserved biogenesis machineries (Strom and Lory, 1993). These fibres participate in a surprising number of functions (Merz and So, 2000). In pathogenic Neisseria species, for example, they not only facilitate bacterial attachment to human cells (Swanson, 1973; Virji et al., 1991), but are also essential for natural transformation (Seifert et al., 1990), a form of cell locomotion known as twitching motility (Merz et al., 2000) and the formation of bacterial aggregates (Swanson et al., 1971). A peculiarity of Tfp-facilitated adhesion is the formation of three-dimensional bacterial microcolonies on the surface of infected cells during an early stage known as localized adherence (Pujol et al., 1997; Nougayrède et al., 2003). Later, Tfp retract and the bacteria come in close contact with the cell surface, typically covering most of it by forming monolayers. The progression to this step, termed diffuse adherence in pathogenic Neisseria species, is under the direct dependence of proteins from the PilT family (Bieber et al., 1998; Pujol et al., 1999).

In all Tfp-harbouring bacteria, a reduced Tfp propensity to form bacterial aggregates is usually accompanied by reduced adherence/colonization. For example, identical single amino acid (aa) substitutions in the pilin subunits of V. cholerae and N. gonorrhoeae, which were found to diminish bacterial aggregation without affecting piliation, led, respectively, to reduced colonization in an animal model of infection (Chiang et al., 1995) and reduced adherence to human cells (Park et al., 2001). The reverse is also true, because pilT or pilT-like mutants, which display an increased aggregation, attach to human cells in higher numbers (Bieber et al., 1998; Pujol et al., 1999). Besides, in pathogenic Neisseria species, Tfp-facilitated adherence is also clearly correlated with the expression of the PilC protein (Rudel et al., 1992; Nassif et al., 1994; Virji et al., 1995). PilC is thought to function as an adhesin localized at the tip of the fibres (Rudel et al., 1995), although other functions (Wolfgang et al., 1998b) and localizations (Rahman et al., 1997) have been proposed for this protein. Therefore, in contrast to other Tfp-harbouring bacteria where fibres are proposed to increase adhesion mostly, if not only, by mediating interbacterial interactions (Hicks et al., 1998; Kirn et al., 2000), in pathogenic Neisseria species Tfp are thought to play mostly a direct role in adherence and the exact contribution of aggregation in this process remains to be determined.

In an effort to improve our understanding of the molecular mechanisms of N. meningitidis adherence to human cells, we screened a genome-wide collection of defined mutants (Geoffroy et al., 2003) for those presenting reduced attachment to a human endothelial cell line. This led to the identification of a protein essential for adherence, named PilX, whose detailed characterization, which is reported here, brought new insights into Tfp-facilitated attachment of pathogenic Neisseria species to human cells.


Identification of a novel gene required for Tfp-facilitated adherence to human cells but dispensable for Tfp biogenesis

We recently constructed a genome-wide collection of defined N. meningitidis mutants and demonstrated it to be a useful toolbox for functional genomics (Geoffroy et al., 2003). This enabled us to perform a large-scale screen for mutants presenting a reduced adherence to human umbilical vein endothelial cells (HUVECs), which are often used in the analysis of the adherence properties of N. meningitidis (Virji et al., 1991; Nassif et al., 1994; Eugène et al., 2002). The 4548 mutants, ordered in 96 pools, were assayed in duplicate on HUVEC monolayers by signature-tagged mutagenesis (STM). We identified 422 candidates presenting reduced hybridization signals when their respective signature-tags were amplified from the bacteria recovered after a 5 h adhesion assay. In order to exclude the clones where the adhesion defects resulted from phase variation events in genes unlinked to transposon insertions, we transformed into the wild-type (WT) strain each of the 422 selected mutations and tested the transformants individually. We found 32 mutants (0.7% of the analysed mutants) that were actually recovered at reduced numbers after a 5 h infection of HUVECs. Fifteen mutants, harbouring transposon insertions in various housekeeping genes, were discarded because they grew poorly in the infection medium. Finally, 17 mutants carrying transposon insertions in nine different genes displayed reduced binding to HUVECs while being unaffected for growth and were thus considered bona fide adherence mutants (Table 1).

Table 1.  Characteristic features of the adherence mutants identified in this study.
NMANMBNumber of hitsName and putative functionPiliation status
  1. NMA and NMB denote the genes, in the N. meningitidis Z2491 and MC58 genomes, respectively, homologous to the sequences flanking the transposon insertion sites.

  2. Number of hits indicates the number of insertions within a defined gene.

  3. The piliation status was defined by specific immunofluorescence staining of the fibres (see Fig. 1A).

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

065018122pilQ, pilus secretin
065418081pilM, Tfp biogenesis protein
110708871Tfp biogenesis protein
152313091Tfp biogenesis protein
215503334pilG, Tfp biogenesis protein
215603322pilD, prepilin peptidase
215903292pilF, Tfp biogenesis protein
060918472pilC1, pilus-associated protein+
111008902pilX, minor pilin+

The analysis of the genes interrupted in the identified adherence mutants (Table 1), underscored the central role of Tfp in the interaction of N. meningitidis with human cells (Swanson, 1973; Virji et al., 1991). Indeed, we identified mutants in most of the genes previously shown to be essential for fibre biogenesis (pilQ, pilG, pilD and pilF) or pilus-facilitated adherence (pilC1). This prompted us to determine whether the remaining mutants were piliated or not by immunofluorescence (IF) microscopy (Fig. 1A). This showed that three of these genes, NMA0654 (pilM), NMA1107 and NMA1523, encode Tfp biogenesis proteins because the corresponding mutants were non-piliated (Table 1), which was confirmed in a subsequent screen aimed at identifying all the components of the piliation machinery in N. meningitidis (Carbonnelle et al., in press). The only two mutants presenting a reduced adherence that were piliated, besides those interrupted in the extensively studied pilC1 gene, harboured transposon insertions in the same ORF corresponding to NMA1110 in the genome of strain Z2491 (Parkhill et al., 2000), hereafter named pilX. In order to assess at which stage of the adherence process they were affected, the adherence of these mutants to HUVECs (Fig. 2) and Hec-1B epithelial cells (data not shown) was characterized in detail and found to be as low as that of a non-piliated pilD mutant from the very beginning of the adherence assay. These results indicate that pilX encodes a protein dispensable for Tfp biogenesis, but essential for pilus-facilitated adherence of N. meningitidis to human cells.

Figure 1.

Presence and morphology of Tfp in various N. meningitidis strains.
A. Presence or absence of Tfp was monitored by immunofluorescence microscopy after the fibres (green) were labelled with a pilin-specific antibody and the bacteria (red) were stained with ethidium bromide.
B. Fibre morphology in the WT strain and a pilX mutant was assessed by transmission electron microscopy. Fibres were visualized directly after staining with phosphotungstic acid. Magnifications are indicated within parentheses.

Figure 2.

Quantification of the adherence of various N. meningitidis strains to HUVECs. The 0 h points in these graphs represent the sizes of the inocula that were adjusted and counted before the assay. After a 30 min contact during which bacteria adhered to the cells, the cells were incubated up to 5 h and the adherent bacteria were numbered every hour. After 5 h, bacteria had progressed from localized to diffuse adherence. pilXind corresponds to the pilX mutant complemented by an intact copy of pilX placed under the transcriptional control of an IPTG-inducible promoter. Adherence of the pilT and pilX/T mutants were indistinguishable and we therefore display only one curve for the sake of legibility. Error bars indicate standard deviations of the mean (n = 3).

pilX encodes a conserved Tfp-associated pilin-like protein

pilX was predicted to encode a protein of 162 aa with a molecular weight of 18.1 kDa (Fig. 3). It is conserved in the three complete genome sequences of N. meningitidis and that of N. gonorrhoeae, with 85% to 98% identity at the aa level, and presents a conserved genetic organization. In the genome of strain Z2491 (Parkhill et al., 2000), for example, NMA1110 is the last ORF of a potential operon composed of five ORFs. It is followed by a non-coding region more than 1 kb long, which suggests that the phenotype we observed was not because of a polar effect of the inserted transposon on a distal gene expression. This was nevertheless confirmed by introducing in the genome of a pilX mutant an intact copy of pilX placed under the transcriptional control of an IPTG-inducible promoter, which successfully restored adherence (Fig. 2). A match of the predicted protein against domain databases identified a characteristic signal peptide pattern present in all the Tfp prepilins (Strom and Lory, 1993) within the residues 9–29 of the sequence (Fig. 3). Consequently, blastp searches yielded numerous hits with known or putative pilins from various bacterial species, but the region of homology was always limited to the N-terminal part of the molecule that corresponds to the above pattern. It was therefore likely that the first 10 aa of the protein encoded by this ORF could be cleaved by the prepilin peptidase PilD to give a mature protein of 16.9 kDa (Fig. 3), which was predicted to consist like the PilE pilin subunit of a long N-terminal α-helix followed by a globular region (Parge et al., 1995).

Figure 3.

Schematic illustration of the probable features of the PilX protein. PilX presents a canonical signal peptide present in all type IV pilins of which the sequence is shown below the scheme. The arrow indicates the probable site of processing by the prepilin peptidase PilD, which eliminates a leader peptide (LP) corresponding to the first 10 aa of the protein. Like the PilE subunit of the pilus (Parge et al., 1995), the mature PilX protein is predicted to consist of a long N-terminal α-helix followed by a globular region.

To start with the experimental characterization of this protein, we raised a rabbit serum against a purified recombinant protein and used it in immunoblotting experiments. When whole cell lysates were analysed, we detected a ∼17 kDa protein in the WT strain and a pilE mutant, which was absent in pilX mutants (Fig. 4A). Less protein was detected in the pilXind complemented mutants in the absence of IPTG induction, which is consistent with a report showing that transcription is not completely inhibited by the lacIOP regulatory sequences (Long et al., 2003). We next tested whether this protein was cleaved by PilD as suggested by its sequence. We found that the reactive species detected in a pilD mutant indeed presented a slightly higher molecular weight, and probably corresponded to unprocessed protein (Fig. 4A). This result further suggested that the mature pilin-like protein could be associated with the Tfp, which prompted us to assess by immunoblotting whether it copurified with the fibres. Using a classical Tfp purification method in Neisseria species (Wolfgang et al., 1998a), we found the 17 kDa protein in Tfp preparations purified from the WT strain but not from the pilX mutant or non-piliated pilD or pilE mutants (Fig. 4B). Taken together, these results demonstrate that the pilX gene encodes a prepilin that is processed by the prepilin peptidase PilD and associates with N. meningitidis Tfp.

Figure 4.

PilX is a pilin-like protein that copurifies with N. meningitidis Tfp without influencing their anchorage or PilC1 ability to copurify with these organelles.
A. Immunoblot detection of PilX in whole-cell extracts of various strains. Equal amounts of proteins were loaded in each lane as verified by immunoblot detection of the class 4 outer-membrane RmpM protein.
B. Immunoblot detection of PilX and PilC1 in the pilus fractions of various strains obtained from equal numbers of bacteria. The pilus purification procedure was validated by immunoblot detection of the pilin subunit PilE only in the preparations from piliated strains.
C. Immunoblot detection of PilE in crude Tfp preparations obtained from equal numbers of bacteria of the WT strain and a pilX mutant after vortexing steps of 20 and 30 s.

PilX contributes to Tfp-facilitated adherence in an unusual fashion

We ruled out the possibility that the impaired adherence observed for the pilX mutant was because of the expression of a different pilin variant, a phenomenon known to affect adherence (Lambden et al., 1980; Nassif et al., 1993; Virji et al., 1993). We sequenced the gene at the pilE locus in these mutants and found it to be identical to the high-adhesive SB allele expressed in the WT strain (Nassif et al., 1993). Although the relative levels of piliation of a pilX mutant and the WT strain were found to be indistinguishable when tested by IF microscopy (Fig. 1A) and by immunoblotting after Tfp purification (Fig. 4B), we sought to quantify them more precisely. We therefore designed a reliable whole-cell ELISA to measure the amount of fibres produced by a defined strain. This showed that a pilX mutant and the WT strain produced comparable levels of Tfp, the mean calculated ratio WT/pilX being 1.51 ± 0.17. It was possible that the pilX mutant adherence defects were due to subtler morphologic differences, undetectable by IF microscopy. We therefore examined by transmission electron microscopy the fibres produced by a pilX mutant and the WT strain, which did not reveal any noticeable differences in morphology, both strains harbouring bundled Tfp fibres (Fig. 1B).

It has been reported that an impaired adherence can also result from either a lack of intrinsic adhesiveness of the Tfp to human cells (Winther-Larsen et al., 2001) or a pilus anchorage defect (Virji et al., 1995), which prompted us to test the possibility of related phenomena operating in pilX mutants. As observed by IF microscopy, Tfp prepared from a pilX mutant adhered to HUVECs similarly to those prepared from the WT strain (Fig. 5), which showed that PilX does not modulate the intrinsic cell adhesiveness of N. meningitidis Tfp. Besides, the pilX mutant fibres were found to be as susceptible to shearing as WT fibres, because short vortexing steps of 20 s and 30 s released comparable amount of fibres from the two strains (Fig. 4C). This indicated that the fibres in a pilX mutant do not present a gross anchorage defect. Finally, we found that the copurification of PilC1 with the fibres was unchanged in the absence of PilX (Fig. 4B), which demonstrated that the adherence defect in the pilX mutant is not due to a reduced association of the PilC protein with the fibres as reported for example in pilV mutants (Winther-Larsen et al., 2001).

Figure 5.

Intrinsic adhesiveness of the Tfp from various strains to HUVECs. Binding of crude Tfp preparations to confluent monolayers of HUVECs was assessed by immunofluorescence microscopy after the fibres (green) were labelled with a pilin-specific antibody and the cells nuclei (red) were stained with ethidium bromide.

Taken together, these results indicate that PilX contributes to Tfp-facilitated adherence in an unusual fashion, i.e. without significantly affecting Tfp production, morphology, intrinsic adhesiveness, anchorage to the bacterial surface or the ability of the PilC1 protein to become associated with the fibres.

Tfp produced by pilX mutants are functional but unable to mediate bacterial aggregation

As noted in the introduction, in addition to promoting adherence to human cells, Tfp produced by pathogenic Neisseria species are also essential for other phenotypes (Merz and So, 2000). It was therefore important to assess whether the fibres produced by a pilX mutant were still capable of promoting these phenotypes. We found that the pilX mutants were competent for natural transformation as opposed to non-piliated mutants, although they presented a reduced competence when compared to the WT strain (Table 2). They also exhibited twitching motility because they were able to crawl over the surface of a glass coverslip to which they attached in contrast to a pilT mutant that was completely non-motile (Merz et al., 2000).

Table 2.  Quantification of the competence for DNA transformation of various N. meningitidis strains. Results are expressed as ratios of the number of rifampicin-resistant transformants, obtained in standardized transformation assays, to the number of total bacteria.
  1. Values are the mean of four independent experiments.

WT(8.86 ± 1.14) × 10−4
pilX(1.12 ± 0.42) × 10−4

Interestingly, we found that pilX mutants were unable to mediate the formation of microcolonies in vitro and are therefore phenotypically identical to non-piliated mutants (Fig. 6A). Given the potential importance of this phenotype in the adherence of pathogenic Neisseria species to human cells (Penn et al., 1980; Park et al., 2001), we developed a method to quantify the aggregative ability of a strain. The formation of bacterial aggregates was followed by measuring the decrease in optical density (OD) that occurs in non-agitated liquid cultures upon sedimentation of the bacterial microcolonies. The kinetics of aggregation that were measured for the various strains were consistent with the aggregative properties that were observed by phase-contrast microscopy. For example, a pilT mutant that is known to display a dramatically increased aggregation (Wolfgang et al., 1998a), with aggregates morphologically distinct from those produced by the WT strain (Fig. 6A), almost completely and rapidly sedimented (Fig. 6B). In contrast, the aggregative ability of a pilX mutant was found to be indistinguishable from that of a non-piliated mutant and therefore nil (Fig. 6B). As expected, aggregation was almost completely restored in the pilXind complemented strain (Fig. 6B). Together, these results indicate that pilX mutants produce Tfp fibres that have selectively lost the ability to promote adherence to human cells and to mediate the formation of bacterial microcolonies.

Figure 6.

Observation and quantification of the aggregative abilities of various N. meningitidis strains. Aggregation was (A) observed by phase-contrast microscopy and (B) quantified by measuring the decrease in absorbance that occurs upon sedimentation of bacterial aggregates in non-agitated liquid cultures. The differences in absorbance were only due to different aggregative abilities because all the strains grew as well as the WT (data not shown). Values are the mean of at least three independent experiments.

PilX possesses aggregative but not adhesive properties

Next, we determined the ability of a pilX mutant to become engaged within bacterial microcolonies formed by a WT strain. In order to differentiate the two strains, one of them was expressing GFP, as previously described in similar experiments performed in V. cholerae (Kirn et al., 2000). In contrast to the aggregates formed during a coculture of WT strain expressing GFP and unlabelled WT strain, which contained similar numbers of fluorescent and non-fluorescent bacteria, the aggregates formed when a fluorescent pilX mutant and an unmarked WT strain were coincubated were completely non-fluorescent (Fig. 7). This suggests that the presence of PilX in each bacterial partner is an absolute requirement for their efficient aggregation. This prompted us to examine the ability of the purified recombinant PilX protein to mediate aggregation of inert latex beads. PilX-coated beads formed macroscopic aggregates that sedimented in a matter of minutes (Fig. 8). Moreover, aggregation was specific to PilX because uncoated beads and beads coated with other proteins all remained in suspension for the duration of the experiment. These results indicate that PilX is able to mediate aggregation of otherwise inert beads, probably through the establishment of PilX-PilX contacts.

Figure 7.

Confocal microscopy observation of the aggregates obtained in cocultures of a strain expressing GFP and an unlabelled strain. Upper panel images show only the bacteria expressing GFP (green), while the lower panels show an overlay of the upper image and a phase-contrast microscopy observation. * indicates which strain is expressing GFP.

Figure 8.

Ability of inert latex beads coated with various proteins to form macroscopic aggregates. Beads were coated with 1, no protein; 2, BSA; 3, purified His6-tagged meningococcal DsbA1 disulphide oxidoreductase; 4, purified His6-tagged PilX.

According to the phenotypic alterations observed in pilX mutants, PilX might be a bifunctional protein also directly mediating adherence to human cells, but this was not supported by the finding that the adhesion of PilX-coated beads to HUVECs was not increased when compared to the adhesion of uncoated beads (data not shown). Alternatively, the reduced numbers of adherent bacteria seen with a pilX mutant might be the consequence of the absence of bacterial aggregation. To test this possibility, we sought to restore a certain level of aggregation in a PilX-negative background, which was accomplished by constructing a pilX/T double mutant. This mutant formed irregularly shaped aggregates similar to those present in a pilT mutant and its aggregative ability was similar to that of  a  pilT mutant  (Fig. 6B).  When  the  adherence  of  the pilX/T mutant to HUVECs was tested, we found that it was almost equivalent to that of a pilT mutant and therefore higher than the adherence of the WT strain (Fig. 2). This indicates that N. meningitidis can adhere to human cells in the absence of PilX when its ability to form aggregates is restored, which suggests that PilX has no intrinsic adhesive properties.

Tfp-mediated interbacterial contacts play a primary role in N. meningitidis initial adherence

Because the reduced numbers of adherent bacteria seen with a pilX mutant are probably the consequence of its lack of bacterial aggregation, it was surprising that its adherence was as impaired as that of a non-piliated mutant (Fig. 2). To exclude the possibility that this reduced adherence resulted from subtly compromised fibre integrity, which could not be detected, it was important to estimate the contribution of Tfp-mediated aggregation to adherence in a WT genetic background. Bacterial aggregates were pelleted by a slow centrifugation leaving bacterial suspensions with essentially no aggregates as quantified under a phase-contrast microscope using a counting chamber (Stephens and McGee, 1983). The bacteria in suspension were perfectly piliated as verified by IF microscopy and whole-cell ELISA, and aggregates consequently re-formed slowly, becoming visible again after more than 30 min of incubation (data not shown). Centrifuged or non-centrifuged inocula, containing comparable numbers of colony-forming units (cfu), were used to infect HUVECs and the numbers of adherent bacteria were measured at different timepoints during the first 30 min of infection (Fig. 9). The adherence of the non-centrifuged WT strain, which contained small bacterial aggregates, was 10-fold higher than that of the centrifuged WT strain that did not contain aggregates. Strikingly, the level of adherence of the WT strain that did not contain aggregates was similar during the first 30 min of infection to the one measured for a non-piliated mutant (Fig. 9). These experiments suggest that during the initial adherence of N. meningitidis to human cells, Tfp-mediated interbacterial contacts have a major contribution in the increase of adherent bacteria seen with a piliated strain, probably by leading to the attachment of bacterial aggregates.

Figure 9.

Contribution of the aggregation to the initial adherence of N. meningitidis to HUVECs. Bacteria were left in contact with the cells for short periods of time and then immediately counted. The 0 h points in these graphs represent the sizes of the inocula that were adjusted and counted before the assay. Agg indicates that the aggregates present in the bacterial suspension were removed before the cells were infected. Error bars indicate standard deviations of the mean (n = 3).


Understanding the molecular bases of the attachment of pathogenic Neisseria species to host cells is important because it may lead to the development of effective means of prevention and treatment. For example, it has been shown that systemic vaccination with adhesins associated with pili that are assembled via the chaperone-usher pathway can prevent infection (Langermann et al., 1997). To improve our understanding of N. meningitidis adherence to human cells, we screened a genome-wide collection of defined mutants (Geoffroy et al., 2003) for those presenting an impaired adherence to a human cell line. The finding of 17 mutants that presented important adhesion defects validated this screening procedure. Their detailed analysis confirmed the central role of Tfp in the adhesion of N. meningitidis to human cells, because a majority of these mutants were non-piliated. However, possibly the most interesting mutants were those that were piliated, but that presented an adherence as impaired as that of non-piliated mutants. Besides pilC1, which encodes the PilC protein that has been extensively studied in the past decade (Rudel et al., 1992; Nassif et al., 1994; Virji et al., 1995), the only novel gene interrupted in the mutants of this class was pilX, which was therefore characterized further. We found that pilX encodes a prepilin, conserved in pathogenic Neisseria species, that is cleaved by the prepilin peptidase PilD and copurifies with the pilus fibres. Because of its sequence features typical of type IV pilins, notably a putative long N-terminal α-helix, PilX might be directly incorporated in the fibres in the same way as the major pilus subunit PilE (Parge et al., 1995), being therefore an integral minor component of Neisseria Tfp. However, the impaired adherence of pilX mutants could not be attributed to defined Tfp alterations that were previously associated with such a phenotypic defect (Lambden et al., 1980; Nassif et al., 1993; Virji et al., 1993; Virji et al., 1995; Winther-Larsen et al., 2001). Moreover, this defect is also unlikely to result from the incorrect incorporation into the pilus of a yet undiscovered protein required for adherence, because the pilX fibres did not lose their intrinsic adhesiveness and adherence could be restored in a PilX-negative background as seen in pilX/T mutants. In contrast, the loss of adherence in a pilV mutant, which is indeed due to the improper localization of PilC in the fibres, is not restored by a secondary pilT mutation (Winther-Larsen et al., 2001).

These results suggested that, except for the presence or the absence of PilX, the fibres in a pilX mutant and a WT strain might be indistinguishable. This was supported by the finding that pilX mutants were still competent for DNA transformation and exhibited twitching motility, which indicated that they have lost Tfp-linked phenotypes selectively. However, they were also incapable of forming bacterial aggregates, which again confirmed the close link existing between adherence and aggregation in Tfp-harbouring bacteria. A set of evidence suggested that PilX has intrinsic aggregative properties but that a priori exclude a direct role of PilX as an adhesin. We found that coating inert latex beads with purified PilX was sufficient to confer them the ability to make aggregates, which is likely to be a consequence of the establishment of PilX–PilX interactions, an hypothesis supported by the inability of pilX mutants to become engaged within bacterial microcolonies formed by a WT strain. On the other hand, evidence that PilX has no intrinsic adhesive properties came from the finding that adherence could be restored in a PilX-negative background by a loss of function mutation in the pilT gene that is responsible for the retraction of the fibres (Merz et al., 2000). Interestingly, this was accompanied by a rescue of the aggregative abilities of the mutant. These findings are consistent with the following scenario. In a WT genetic background, where the formation of the antiparallel electrostatic fibre–fibre contacts required for aggregation (Craig et al., 2004) is necessarily counteracted by the continuous PilT-driven retraction of the pili, the establishment of homotypic contacts between PilX proteins present in the fibres of different cells might be essential for the stabilization of N. meningitidis aggregates. In a PilT-negative background, however, the complete absence of retraction of the fibres lessens the necessity for this stabilizing role of PilX. Nevertheless, even in a PilT-negative background, it should be noted that PilX seems to have a small contribution to aggregation because the aggregative ability of a pilX/T mutant was slightly lower than that of a pilT mutant. Remarkably, the pathogenic Neisseria species seem to have evolved a set of three minor pilins, each associated with a particular Tfp-mediated phenotype. ComP is thus necessary for competence for DNA transformation (Wolfgang et al., 1999), PilV affects adherence to human cells (Winther-Larsen et al., 2001) and PilX is essential for bacterial aggregation.

Although the above data were consistent with the possibility that the lack of aggregation in pilX mutants was the main, if not only, reason for their decreased net adherence, the magnitude of the decrease in adherence that was observed was intriguing because it suggested that the contribution of aggregation in Neisseria Tfp-mediated adherence to human cells was more important than previously thought (Penn et al., 1980; Park et al., 2001). Therefore, we sought to quantify the contribution of aggregation to the adherence process in an unaltered WT genetic background. Strikingly, when bacterial aggregates were eliminated by a simple centrifugation step before the infection assay (Stephens and McGee, 1983), the net adherences of the WT strain and a non-piliated mutant were identical. After 2 min of contact, 0.1% of the bacteria present in the inoculum attached to the cells, a number reaching 1% after a 30 min contact. In contrast, when non-centrifuged inocula were used, while the adherence of a non-piliated mutant was unchanged, that of a WT strain was dramatically increased with 1% and 8% of the bacteria attached to the cells, after 2 min and 30 min contact respectively. This suggests that the main role of Tfp in the initial attachment of N. meningitidis to human cells might be similar to what has been proposed in other Tfp-harbouring bacteria such as V. cholerae and EPEC (Hicks et al., 1998; Kirn et al., 2000), i.e. to increase the number of adherent bacteria mostly by mediating a cooperation between them, leading to the attachment of bacterial aggregates instead of single cells. Moreover, it is possible to surmise that the impaired aggregation and adhesion previously reported for isogenic N. meningitidis strains containing particular pilin variants (Penn et al., 1980; Marceau et al., 1995) or N. gonorrhoeae mutants containing single aa substitutions in PilE (Park et al., 2001), might be because of an impaired incorporation of PilX in the fibres. Interestingly, such a possibility was envisioned by Park et al. which stated in their manuscript that the reduced aggregation they observed ‘in the PilE mutants might reflect the inabilities of those pilins to interact with a pilin-like protein or promote incorporation of a pilin-like protein into the fibres’ (Park et al., 2001).

Tfp-facilitated adherence in N. meningitidis might thus be more compatible with the models currently favoured in V. cholerae and EPEC (Hicks et al., 1998; Kirn et al., 2000). The first contact between N. meningitidis and human cells is probably mediated by a surface molecule, which could be the fraction of PilC that was localized in the outer-membrane by immunogold labelling experiments (Rahman et al., 1997) and by immunoblotting after separation of the membranes (Carbonnelle et al., in press). Simultaneously, Tfp increase the overall number of adherent bacteria by promoting the attachment of bacterial aggregates. Nevertheless, it is also possible that adherent bacteria recruit unattached bacteria or bacterial aggregates, a possibility that is supported by the finding that pilU mutants of N. gonorrhoeae, PilU being a protein of unknown function, probably adhere as single bacteria because they present a reduced aggregation in vitro, but display an increased adhesion most likely resulting from the formation of large bacterial aggregates directly on the cells (Park et al., 2002). Consequently, typical three-dimensional bacterial microcolonies can be seen on the cell surface. Tfp are then retracted by the PilT protein, ultimately disappear and all the bacteria come in direct contact with the cell surface (Pujol et al., 1999). This model, of course does not exclude interactions between Tfp and the cells during the adherence process. While these contacts, possibly involving the Tfp-associated fraction of PilC and CD46 as a specific receptor (Källström et al., 1997), do not seem to increase net initial adherence as suggested here, they clearly play a pivotal role in the dynamics of the adherence process because they are necessary for the progression from localized to diffuse adherence. This is likely to result from a cascade of signalling events induced by the contact of Tfp with the cells, including the rapid and localized formation of cortical plaques enriched in numerous cellular components (Merz and So, 1997) and the increase in cytosolic free calcium known to control many cellular responses (Källström et al., 1998).

In conclusion, the findings here suggest that the model for Tfp-facilitated adherence favoured in most Tfp-harbouring bacteria might be applicable to N. meningitidis as well, at least in part, which is consistent with the extreme conservation of the overall structure and mode of assembly of these filamentous organelles. Accordingly, the main role of Tfp during initial adherence could be to increase net attachment by mediating a cooperation between the bacteria, explaining the presence of bacterial microcolonies on the cell surface at early stages during adherence. It would be interesting, to determine whether minor pilins in other Tfp-harbouring bacteria might have a role in aggregation equivalent to that of N. meningitidis PilX.

Experimental procedures

Bacterial strains and growth conditions

N. meningitidis was grown on GCB agar plates (Difco) containing Kellogg's supplements and, when required, 100 µg ml−1 kanamycin, 5 µg ml−1 rifampicin, 5 µg ml−1 chloramphenicol and 3 µg ml−1 erythromycin. E. coli transformants were grown in liquid or solid Luria–Bertani medium (Difco) containing 100 µg ml−1 ampicillin. Bacteria were grown at 37°C in moist atmosphere containing 5% CO2. The library of 4548 defined signature-tagged transposition mutants in a highly adhesive variant of the piliated N. meningitidis 8013 strain has been described elsewhere (Geoffroy et al., 2003). In order to eliminate potential secondary variations, all the experiments with the pilX mutant presented in this study were carried out after the pilX mutant allele was PCR amplified using primers 1110F 5′-AAGCGTCAGC AAAATGCCACGTTAT-3′ and 1110R 5′-CGATTAGAGAAGG CTTCACAGCCGA-3′ and retransformed as previously described into the WT strain. To complement the pilX mutants, the WT pilX allele was amplified using primers 1110-IndF 5′-CCTTAATTAAGGAGTAATTTTATGATGAGTAATAAAA TGGAACA-3′ and 1110-IndRbis 5′-CCTTAATTAACTATTTTT TACGATTAGAGAAAGC-3′, which contained overhangs with underlined restriction sites for PacI. This PCR fragment was restricted with PacI and cloned into PacI-cut pGCC4 vector, adjacent to lacIOP regulatory sequences (Mehr et al., 2000). This placed pilX under the transcriptional control of an IPTG-inducible promoter within a DNA fragment corresponding to an intragenic region of the gonococcal chromosome conserved in N. meningitidis. The pilXind allele was then introduced into the chromosome of a pilX mutant by homologous recombination. The double pilX/T mutant was obtained by transforming into the pilX mutant, a pilT mutation where the gene is interrupted by an erythromycin resistance cassette (Pujol et al., 1999). Strains expressing the green fluorescent protein (GFP) were obtained by transformation with the pAM239 plasmid where the gfp gene is under the control of an IPTG-inducible promoter (Solomon et al., 2000).

Mutants were tested for growth defects in RPMI 1640 medium supplemented with 10% heat-inactivated fetal calf serum Gold (both from PAA laboratories GmBH, Austria). Bacteria grown on GCB agar plates were incubated for 2 h at 37°C in prewarmed RPMI-serum under constant agitation. The suspension was then adjusted to an OD600 of 0.01 and it was incubated at 37°C under constant agitation. The OD600 was then measured every hour during 8–9 h and the data were plotted on a graph for comparisons.

Adherence assays

Adhesion of meningococci to HUVECs was performed as described previously (Eugène et al., 2002), with minor modifications. The 24 well plates were seeded with 105 cells per well, which attached to the bottom of the wells after 16 h of incubation at 37°C in a moist atmosphere containing 5% CO2. Before the assay, bacteria grown on GCB agar plates were adjusted to a specific OD600 and then incubated for 2 h at 37°C in prewarmed RPMI-serum. To count the cfu in the inoculum, 5–6 serial 10-fold dilutions were performed, each dilution being vortexed 3–5 s at maximal speed. Because these vortexing steps are enough to dissociate bacterial aggregates, as confirmed by direct counting of the bacterial aggregates using a counting chamber (Stephens and McGee, 1983), the counted cfu clearly consist of single bacteria. Cells were infected with 1 ml of bacterial suspension in RPMI-serum. Usually, after 30 min of contact, unbound bacteria were removed by three washes with 1 ml of RPMI-serum and the infection was continued for 5 h with washes every hour, a time sufficient for meningococci to progress from initial localized adherence to late diffuse adherence (Pujol et al., 1999). Adherent bacteria, recovered by scraping the wells, were counted as above by plating appropriate dilutions on GCB agar plates. When we monitored the initial adhesion, unbound bacteria were removed after 2, 10, 20 and 30 min of contact and the adherent bacteria were plated for counting as above. Where indicated, aggregates were removed from bacterial suspensions before the assay by centrifugation at 200 g for 1 min (Stephens and McGee, 1983) and the suspensions were then re-adjusted to a specific OD600.

When pools of mutants were assayed in the STM screen, clones grown on GCB agar plates were resuspended individually in RPMI-serum in the wells of a microtitre plate, OD600 was measured on a microtitre plate reader and equivalent amounts of each mutant were pooled. The OD600 of the pool was adjusted at 0.1, which corresponded approximately to 2 × 106 cfu of each mutant, and the suspension was incubated during 2 h at 37°C. One ml of this suspension was used to infect HUVECs in duplicate as above. The adherent bacteria were then recovered and plated. More than 104 colonies were collected after overnight incubation and their chromosomal DNA was prepared using the Wizard Genomic DNA Purification kit (Promega). Signature-tags were amplified, labelled and hybridized as described previously (Geoffroy et al., 2003).

To assay the intrinsic adhesiveness of Tfp to HUVECs, we used crude Tfp preparations obtained as follows. Bacteria from one Petri dish were resuspended in 1 ml of RMPI-serum. Fibres were sheared off by vigorous vortexing for 1–2 min, separated from bacteria and large debris by centrifugation at 6000 g for 20 min and incubated for 2 h with HUVECs grown on coverslips (Winther-Larsen et al., 2001). Unbound fibres were removed by three washes and the cells were fixed with 2.5% paraformaldehyde in phosphate-buffered saline (PBS). Coverslips were then treated for IF as described below.

SDS-PAGE, antisera and immunoblotting

Preparation of protein samples, SDS-PAGE separation, transfer to membranes and immunoblotting were performed using standard molecular biology techniques (Sambrook et al., 1989). Proteins were quantified using the Micro BCA protein assay reagent kit (Pierce), as suggested by the manufacturer. Detection of immobilized antigens was performed by chemiluminescence using ECL or ECL Plus detection reagents (Amersham). PilC1 was detected using the rabbit polyclonal serum 18P4 (Morand et al., 2001), at 1/2000 dilution. PilE was detected using the 5C5 mouse monoclonal serum (Marceau et al., 1998), whereas RmpM was detected using a specific mouse monoclonal serum (Morand et al., 2004), both sera were used at 1/5000 dilution.

To raise antibodies against PilX, we amplified a PCR product corresponding to the pilX gene devoid of its N-terminal sequence encoding the first 26 residues of the PilX preprotein. We used primers Nde2 5′-CCATATGAGCGTCATTGCC ATACCTTC-3′ and Bam 5′-GGGATCCTTTTTTACGATTAGA GAAGG-3′, which contained overhangs corresponding to underlined restriction sites for NdeI and BamHI respectively. This fragment was first cloned into pCRII-TOPO (Invitrogen) and then subcloned into pET-14b (Novagen) restricted by NdeI and BamHI. This introduced a six-histidine tag at the N-terminus of the recombinant PilX protein. The protein was expressed in E. coli BL21::DE3, purified using Ni-NTA agarose (QIAGEN) and injected into a New Zealand rabbit.

Tfp detection, quantification and purification

Tfp were detected using IF staining as previously described (Marceau et al., 1998), with minor modifications. Briefly, bacteria grown on GCB agar plates and propagated 2 h in RPMI-serum were fixed on coverslips for 20 min with a solution of PBS-2.5% paraformaldehyde. After 5 min of incubation with PBS-0.1 M glycine, samples were incubated with PBS-0.2% gelatine. Tfp were stained with the 20D9 monoclonal antibody that is specific for the PilE hypervariable region present in strain 8013 (Pujol et al., 1997), used at 1/1000 dilution, while the bacteria were stained with an ethidium bromide solution at 1/6000 dilution. The secondary antibody, used at 1/100 dilution, was a goat anti-mouse antibody labelled with Alexa Fluor 488 (Molecular Probes). Samples were observed using a Zeiss 510 confocal microscope. The fine structure of the fibres was determined by transmission electron microscopy 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.

To quantify piliation, we designed a whole-cell ELISA using the above 20D9 monoclonal antibody. Bacteria grown on GCB agar plates and propagated in RPMI-serum were resuspended in PBS at an OD600 of 0.1. One hundred µl of serial twofold dilutions were coated in the wells of a microtitre plate and the plates were incubated without covers overnight at 37°C. Coated plates 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 bacteria, 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 supplementary 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. Results are expressed as mean ratios ± 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 and by the increase in piliation, estimated at 5.12 ± 0.84 fold, observed for a pilT mutant.

Pilus purification was carried out as follows. Bacteria from two to three Petri dishes inoculated with 108 cfu the previous day 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).

Quantification of competence for DNA transformation

Natural competence for DNA transformation was quantified by counting the number of rifampicin-resistant cfu obtained after transforming 5 × 107 cfu of the studied strain with 2 µg of chromosomal DNA extracted from a mutant of strain 8013 spontaneously resistant to rifampicin.

Monitoring twitching motility

Twitching motility was monitored as described (Merz et al., 2000) by resupending bacteria grown on GCB agar plates in phenol-red free DMEM containing l-glutamine and pyruvate (Gibco BRL) and supplemented with 1% bovine serum albumin (BSA) fraction V (Sigma). Bacteria were allow to attach to small glass coverslips during 1 min, which were then mounted on two parallel lines of silicon grease on large coverslips. Unattached bacteria were washed away by carefully adding fluid and the remaining bacteria were observed under a phase-contrast microscope for short directed movements. The validity of this method was confirmed by the complete absence of movement observed when a pilT mutant was assayed.

Aggregation monitoring and quantification

Aggregation was monitored by resuspending bacteria grown on GCB agar plates individually in the wells of a 24 well plate, which contained 1 ml of RPMI-serum, and visualizing the aggregates after 1–2 h of growth using a phase-contrast microscope. Aggregation of beads coated with various proteins was monitored in the same way. Ten µl of a 10% solution of latex beads (Sigma) were coated with 10 µg of protein in 1 ml of PBS, during a 10 min incubation at room temperature under constant rotatory agitation. The proteins we used for coating were pure recombinant His6-tagged PilX, His6-tagged meningococcal DsbA1 disulphide oxidoreductase (Tinsley et al., 2004) and BSA. Coaggregation experiments were carried out using mixtures of unlabelled and GFP-expressing bacteria, adjusted at an OD600 of 0.05. This culture was gently agitated for 2 h before gfp transcription was induced during 3 h by the addition of 100 mM IPTG. Samples were then examined by fluorescence microscopy using a confocal microscope Zeiss 510.

To quantify aggregation, we designed an assay based on a method previously used in Myxococcus xanthus (Shimkets, 1986), which relies on the measure of the changes in optical density that occur in a non-shaking culture upon sedimentation of multicellular aggregates. 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 in 30 ml of RPMI-serum. Cells were then grown for 2 h under constant agitation. Finally, these suspensions were incubated at room temperature without shaking, and the OD600 of the supernatant was measured at 20 min intervals.


We thank G. Duménil for GFP-expressing N. meningitidis strains, E. Eugène for help with immunofluorescence microscopy and C. Fréhel for help with electron microscopy. We are grateful to G. Duménil, C. Recchi, J.-M. Reyrat and M.P. Sheetz for helpful suggestions and/or critical reading of the manuscript. This work was funded by INSERM and Université Paris V.