Meningococcal biofilm formation: structure, development and phenotypes in a standardized continuous flow system

Only capsule deficient meningococci form biofilms

Biofilm formation of meningococci under static conditions was analysed on a glass surface using a modified Neisseria defined medium (NDM) as growth medium. Encapsulated meningococcal strains MC58 [serogroup B, sequence type (ST)-32 complex], 2120 (serogroup C, ST-11) and 2594 (serogroup A, ST-5) as well as their unencapsulated derivatives (i.e. MC58siaD-, 2120siaD-, 2594mynB-) were tested for biofilm formation. According to results published recently (Yi et al., 2004), unencapsulated derivatives were very potent biofilm formers, whereas encapsulated parental strains formed no appreciable biofilm. Similar results were obtained with a polystyrene microtitre plate assay using GC-medium. Furthermore, unencapsulated derivatives of nine carrier isolates and three invasive isolates (Table 1) efficiently formed biofilms (data not shown). These findings demonstrate that biofilm formation is a general trait of unencapsulated meningococci. We investigated whether this also held true in a continuous flow system which is described in detail in the following paragraph. For this purpose the apathogenic serogroup Y strain Y2220 was compared with its unencapsulated siaD mutant (Ram et al., 2003; Madico et al., 2006). The wild-type strain did not develop substantial amounts of biofilm mass, in contrast to the unencapsulated derivative, which was highly effective in this respect (Fig. 1B).

Biofilm development of capsule deficient meningococci in a continuous flow cell system

To investigate biofilm development and specific biofilm structures we constructed capsule deficient derivatives of several meningococcal strains harbouring plasmids for the expression of the red-shifted green fluorescent protein (rsGFP). This allowed us to monitor biofilm development in a non-destructive mode in continuous flow chamber cultures by CLSM. The strains were cultivated for up to 120 h on glass surfaces in the continuous flow cell system irrigated with modified NDM. The supplementation of NDM with 5 mM NaHCO₃ and PolyViteX was indispensable to yield persistent and reproducible biofilms in the flow cell system. All capsule deficient strains formed biofilms, and the biofilms of each strain could be maintained for at least 96 h. Figure 1A shows the different biofilm architectures produced by strains MC58siaD-/gfp+, 2120siaD-/gfp+ and 2594mynB-/gfp+. After inoculation, single cells were equally distributed over the glass surface. For reasons of the experimental set-up, these data cannot be shown. Within 1 h after inoculation without medium flow small microcolonies consisting of only a few cells were observed for strain MC58siaD-/gfp+. We suggest that these cell clusters are the product of aggregation and growth immediately after inoculation of the channels. Six hours after inoculation, strains MC58siaD-/gfp+ and 2120siaD-/gfp+ formed dense regularly shaped microcolonies whereas strain 2594mynB-/gfp+ formed a flat biofilm without the presence of these regularly shaped microcolonies. Strain 2594mynB-/gfp+ appeared to cover much more of the substratum than the two other strains. From 1 h to 12 h a consistent increase of biomass and biofilm thickness was visible (Fig. 1). The CLSM images were analysed by the COMSTAT computer program (Heydorn et al., 2000), which comprises 10 different features for quantifying three-dimensional biofilm image stacks. As already seen by visual inspection (Fig. 1A), 2594mynB-/gfp+ covered more than 80% of the substratum, whereas MC58siaD-/gfp+ and 2120siaD-/gfp+ covered less than 40%. In conclusion, unencapsulated meningococci readily formed biofilms under conditions of continuous medium flow. Differences in biofilm architecture between strains became apparent. Therefore, a further set of experiments was designed to elucidate the molecular mechanisms of microcolony formation in meningococcal biofilms.

PilX mutation and complementation and autoaggregation studies

Recently, PilX has been described as a type IV pilus-associated pilin-like protein that mediates aggregation of meningococcal cells in liquid cultures and on epithelial cells without influencing the adhesive properties to epithelial cells (Helaine et al., 2005). PilX was necessary for bacterial aggregation in static fluid suspensions, and pilX negative mutants did not form microcolonies on epithelial cells. We therefore investigated the biofilm formation of pilX mutant strains in the continuous flow biofilm model. The pilX gene was inactivated in the strain MC58 by insertion of a kanamycin resistance cassette (Fig. 2A). The knock-out was confirmed by polymerase chain reaction (PCR), by Southern blotting, and phenotypically by an aggregation assay (Helaine et al., 2005) (Fig. 2B, C and F), and by reverse transcription (RT)-PCR (data not shown). RT-PCR demonstrated that the pilX mutation did not affect transcription of NMB889 (Fig. 2D), which has recently been shown to be necessary for pilus assembly (Carbonnelle et al., 2005). Furthermore, Western blot analysis using pilin–specific antibodies suggested that pilus subunit expression was not reduced by the PilX mutation (Fig. 2E). We also generated PilX-specific antisera by immunizing rabbits with purified PilX_MC58-GST fusion protein. Only truncated proteins representing the globular domain could be expressed in sufficient amounts. His-tagged fusions could not be expressed in sufficient amounts in E. coli. The GST fusion proteins were not soluble at various temperatures. Antibodies were elicited in the rabbit immunized with the MC58 protein, but not in the rabbit immunized with the 2120 protein (data not shown). Very weak expression of PilX was seen in the MC58 parental strain, whereas in the complemented mutant harbouring the pilX gene under control of a porA promoter, PilX was clearly visible (Fig. 2G). Unfortunately, numerous cross-reactive bands were observed. There was no cross-reactivity of the serum with strains 2120 and 2594. This might be due to extensive antigenic variability (see below).

The pilX knock-out mutant of strain MC58siaD-/gfp+ showed a clear delay in autoaggregation, and complementation in trans of PilX expression under the control of a meningococcal porA promoter restored autoaggregation confirming that PilX was necessary for the phenotype (Fig. 2F). Also strain 2594mynB-/gfp+ exhibited autoaggregation, which was absent in the isogenic pilX mutant strain, again demonstrating the role of PilX. Surprisingly, strain 2120siaD-/gfp+ displayed no interbacterial aggregation. We cannot rule out the possibility that the strain 2120siaD-/gfp+ showed only weak PilX protein expression. However, when MC58siaD-/pilX- was complemented with the pilX gene from strain 2120 by transformation with plasmid pHC34, autoaggregation was also not restored in significant contrast to complementation with the homologous pilX variant (Fig. 3, P = 0.0006). In contrast to strain MC58, strain 2120 could not be transferred to an autoaggregating phenotype by either pilX variant, suggesting that PilX was necessary, but not sufficient for autoaggregation (Fig. 3).

We sequenced pilX genes of a variety of meningococcal isolates and compared the deduced amino acid sequences (Fig. S2). Until now, we identified 35 alleles with an average of 163 amino acids. Thirty-nine segregating sites were identified (0.24 segregating sites per site). These sites are clearly concentrated in proposed globular domain of the protein. Strain 2120 harboured one of six alleles with an eight amino acid insertion (position 107–114), whose function is yet unclear. Preliminary analysis suggest an increase of the dN/dS ratio in the globular domain, and weak signs of recombination at the C-terminus (M. Reinhardt, pers. comm.). The PilX₂₅₉₄ and PilX₂₁₂₀ differ from PilX_MC58 in 11 and 20 amino acids respectively.

PilX mutation and meningococcal biofilm architecture

We next asked whether the diverse effects of meningococcal PilX with regard to autoaggregation might be the reason for differing biofilm architectures. Therefore, biofilm formation by pilX knock-out mutants was recorded by CLSM under continuous flow conditions. The pilX mutation abrogated microcolony formation of the unencapsulated derivative of MC58, resulting in a biofilm with a high surface coverage (1, 4). Microcolony formation by the mutant was restored by pilX complementation in trans. COMSTAT analysis confirmed that biofilms of a pilX knock-out mutant showed increased substratum coverage for MC58siaD-/gfp+ and 2120siaD-/gfp+, but not for 2594mynB-/gfp+ (data shown for strain MC58siaD-/gfp+ in Fig. 4). Biofilm formation for the pilX mutant remained unaffected. Microcolony formation was also observed in strain 2120siaD-/gfp+. Again, the pilX mutation abrogated microcolonies without affecting biofilm formation itself. This finding was interesting, because strain 2120siaD-/gfp+ was incapable of autoaggregation. We conclude that autoaggregation mediated by PilX is not a sufficient explanation for microcolony formation in the biofilm flow system, which was supported by the observation that the autoaggregating strain 2594mynB-/gfp+ did not display microcolonies.

We therefore asked the question, how PilX affected microcolony formation, as it was obviously not linked to autoaggregation. Visual inspection of early phase biofilms and short time-lapse movies demonstrated that MC58siaD-/gfp+ and 2120siaD-/gfp+ showed a higher degree of twitching motility compared with both strain 2594mynB-/gfp+ and the pilX knock-out mutants. The data are presented for strain MC58 as time-lapse movies (see Fig. S1). Twitching of strain MC58siaD-/pilX-/gfp+ could be restored by complementation in trans. Thus, the pilX mutation reduced twitching motility of cells in the continuous flow biofilm model. The twitching motility phenotype was associated with microcolony formation (Table 2).

Twitching furthermore resulted in extensive motility of cells even after microcolonies were formed. When we mixed CFP- and YFP-labelled cells of strain MC58siaD-, microcolonies consisted of homogeneously mixed blue and yellow cells (Fig. 5). In contrast, the pilX mutant cells formed a flat biofilm with very distinct zones of blue and yellow cells with no obvious mixing. We therefore suggest that twitching motility results in intensive cell mixture in biofilms.

Why was twitching motility reduced in pilX mutants? Twitching in pathogenic Neisseria is mediated by type IV pili composed of the pilin subunit, which is encoded by the pilE gene. Immunofluorescence analysis of pilX mutants and the respective derivatives complemented in trans using the pilin-specific antibodies SM1 and AD211, respectively, revealed a significant reduction of diplococci with attached pili in the pilX mutants. PilX complementation restored full piliation in all derivatives (Fig. 6). Ten digital microscopic images were taken for each strain, and the ratio of pilated to unpiliated diplococci was determined. Of 459 diplococci 4.36% were piliated in strain MC58siaD-/gfp+, whereas 0.9% of 660 diplococci were piliated in the pilX mutant (two-tailed Fisher's exact test, P = 0.001). PilX complementation of the pilX mutant resulted in a frequency of piliated diplococci of 6.5% of 569 diplococci. Similar results were obtained with the other meningococcal strains [2120siaD-/gfp+, 9.9% of 638; 2120siaD-/gfp+/pilX-, 1.3% of 639; 2120siaD-/gfp+/pilX-/pHC34 (complemented), 15.1% of 602; 2594mynB-/gfp+, 17.5% of 555; 2594mynB-/gfp+/pilX-, 1.2% of 500; 2594mynB-/gfp+/pilX-/pHC35 (complemented), 7.8% of 580]. The reduction of piliated cells was seen both for meningococci taken directly from agar plates and after intermediate growth in broth. We did not observe changes in the proportion of cell-attached pili to unattached pili. Using ELISA, Helaine et al. saw a ratio of piliation of 1.5 if a wild-type strain and the corresponding pilX mutant were compared with regard to pilus formation, indicating a barely measurable effect of PilX on piliation (Helaine et al., 2005). It should be taken into account that the strain used by these authors was encapsulated and belonged to a different genotype. In our hands the pilX mutation resulted in reduced piliation, reduced twitching motility in Neisseria biofilms, and as a result abrogated the formation of distinct microcolonies.

We finally asked the question inasmuch PilX variability contributes to differences in biofilm architecture. We expressed the PilX alleles of strains 2120 and 2594 in trans in the pilX knock-out of MC58siaD-. We clearly observed a complementation of the mutation with regard to microcolony formation irrespective of the pilX allele (Fig. 1B). This supported the idea that pilX irrespective of the allele was needed for pilus assembly, twitching motility and consecutively microcolony formation. We then knocked out the gene encoding the subunit of pili, i.e. pilE, and observed loss of microcolony formation despite of the presence of a functional copy of the pilX gene (Fig. 1B). The mutation was complemented if PilE was expressed from a plasmid (Fig. 1B). This demonstrated the importance of pili for microcolony formation under continuous flow conditions. Of note, pili were not necessary for formation of a biofilm (Fig. 1B).

Meningococcal biofilms are more susceptible to rifampin and ciprofloxacin than to penicillin

Following the analysis of the development of meningococcal biofilms, we focused on biofilm phenotypes, the most important of which being the resistance to antibiotics. It is well established that penicillin, albeit being highly efficient in treatment of invasive meningococcal disease, is ineffective for eradication of meningococci from the nasopharynx of healthy carriers (Abramson and Spika, 1985), in contrast to ciprofloxacin (Cuevas et al., 1995) and rifampin (Schwartz et al., 1988). With the newly established biofilm model, we tested the antimicrobial susceptibility of meningococci in biofilms. After 24 h of biofilm growth in flow chambers, 10 times the minimal inhibitory concentration (MIC) of either penicillin, ciprofloxacin, rifampin, was added to the medium. Propidium iodide (PI) in the medium was used for continuous monitoring of the viability of the biofilm cells. The biofilms were recorded with CLSM after 4 h and 24 h to follow the antibiotic killing. Representative data are shown for strain 2120siaD-/gfp+ and 2594mynB-/gfp+ (Fig. 7). After 4 h of antibiotic treatment only the biofilm treated with rifampin was slightly affected. Cells of the apical biofilm layers took up the red PI stain, indicating cell death. The biofilms treated with penicillin or ciprofloxacin as well as the control were completely unaffected. After 24 h of antibiotic treatment the biofilms incubated with rifampin and ciprofloxacin were almost completely dead, whereas in the biofilm treated with penicillin only the upper layers of the biofilm were dead. The deeper layers expressed GFP, and did not take up PI, suggesting that the cells here were alive. The strains displayed identical susceptibility profiles despite of their differing biofilm architectures.

We quantified and confirmed the above results by treating static biofilms with antibiotics. For that purpose we used 24-well cell culture dishes in which a 15 mm round glass cover slip served as the substratum. After 24 h the biofilms were treated with 10 times MIC penicillin, ciprofloxacin and rifampin for another 24 h. Thereafter, the biofilm cells were resuspended, and serial dilutions were plated out on GC agar to determine the colony-forming units (cfu) of surviving bacteria. Without antibiotic treatment 10⁷ cfu well⁻¹ (Fig. 8) were recovered from the biofilm. In contrast to the lack of growth after treatment with ciprofloxacin and rifampin for 24 h, counts of 10⁴−10⁵ cfu were obtained after treatment with penicillin for 24 h.

As a control, we resuspended untreated 24 h old biofilms in medium and added 10 times MIC of penicillin, ciprofloxacin, or rifampin respectively. Subsequently, after additional 24 h shaking, these cultures were plated out. Resuspended biofilm cells were completely sensitive to any of the three drugs (untreated resuspended biofilm cells were viable). Thus, outside the biofilm the cells were sensitive to penicillin as suggested before by MIC determination.

Bacterial strains and culture conditions

Invasive meningococcal wild-type strains were used, i.e. strain MC58 (serogroup B; ST-74) (ST-32 complex) (UK) (kindly provided by E.R. Moxon) (Dunn et al., 1995), strain 2120 (serogroup C; ST-11) (Germany) (Vogel et al., 1998) and strain 2594 (serogroup A; ST-5) (Germany) (kindly provided by Ingrid Erhard). Derivatives of those wild-type strains used in this study are listed in Table 1. GC agar (BD Difco^TM, Heidelberg, Germany) was used for standard cultivation, and was supplemented with PolyViteX (bioMerieux, Nürtingen, Germany). When appropriate, erythromycin (7 μg ml⁻¹), kanamycin (100 μg ml⁻¹), chloramphenicol (7 μg ml⁻¹) and spectinomycin (100 μg ml⁻¹), respectively, were added to the medium. For liquid cultures and biofilm experiments a modification of the NDM (Archibald and DeVoe, 1978) was used. Modified NDM was supplemented with PolyViteX and 5 mM NaHCO₃ to achieve reproducible growth of biofilms under flow conditions. Despite of the fact that pEG2-Ery derivatives proved to be stable for several days of biofilm culture without addition of antibiotics, 7 μg ml⁻¹ erythromycin was added to all biofilm media in this study except for the media used for antimicrobial treatment experiments. All experiments were performed at 37°C. CO₂ enriched atmosphere was used for GC agar cultures only.

Recombinant DNA techniques

Restriction enzymes and DNA-modifying enzymes were purchased from New England Biolabs (Frankfurt/Main, Germany). Chromosomal DNA of N. meningitidis was purified with the Qiagen^® Genomic tips system (Hilden, Germany) according to the manufacturer's instructions. Southern blot hybridizations were performed as described previously with digoxigenin-labelled probes (Roche, Mannheim, Germany) (Hilse et al., 1996). Recombinant plasmids were isolated with QIAprep^® Spin miniprep kit (Qiagen, Hilden, Germany). Transformation of meningococci was performed as described previously. Oligonucleotides were purchased from Sigma-Genosys (Steinheim, Germany). PCR was performed on a thermal cycler obtained from Biometra (Göttingen, Germany). The thermostable Taq DNA polymerase was purchased from New England Biolabs.

Insertional inactivation of meningococcal capsule genes

Plasmids and primers used for construction of meningococcal mutants are listed in Table 3, and Table S1 respectively. For construction of a serogroup A capsule deficient mutant the mynB gene, hypothesized to encode the capsular polymerase of serogroup A meningococci (Swartley et al., 1998), was insertionally inactivated. The mynB gene was amplified with primers NT2 and NT4, and the PCR product was cloned into the vector pCR^®2.1-TOPO (Invitrogen, Karlruhe, Germany), resulting in pNT3. After deletion of a 253 bp HincII fragment (position 477–729 of mynB) from pNT3 the blunt ended HindIII fragment of pTnMax5 (Kahrs et al., 1995) comprising the chloramphenicol resistance cassette was inserted resulting in plasmid pNT5. Subsequently, strain 2594 was transformed with pNT5. The inactivation of the capsule was confirmed by PCR and by ELISA using mAb932 (kindly provided by D. Bitter-Suerbaum, Medical School Hannover, Germany) specific for N. meningitidis serogroup A capsule polysaccharide (data not shown). The capsule deficient mutants of the serogroup C strain 2120, and of the serogroup B strain MC58 have been described previously (Ram et al., 2003; Kurzai et al., 2005).

Labelling of meningococcal strains with fluorescence proteins

The hybrid shuttle vector pEG2 harbours the red-shifted gfp (rs-gfp) gene under the control of a porA promoter (Christodoulides et al., 2000). This plasmid has been modified by replacing the ampicillin resistance cassette (bla) with an erythromycin resistance cassette (ermC), resulting in pEG2-Ery (kindly provided by M. Christodoulides and P. van der Ley, unpublished). The capsule deficient strains 3240 (MC58siaD-), 2517 (2120siaD-) and 2668 (2594mynB-) were transformed with plasmid pEG2-Ery to generate green fluorescent meningococci. For the construction of cyan and yellow labelled meningococci the rs-gfp gene was deleted from plasmid pEG2-Ery by partial restriction with KpnI and subsequent restriction with SacII. After incubation with T4-DNA polymerase the remaining DNA fragment was religated resulting in pAP1. The genes encoding the cyan fluorescent protein (ecfp) and the yellow fluorescent protein (eyfp) were excised from pSM236.1 and pSM236.2 (Lambertsen et al., 2004), respectively, with XbaI and HindIII and cloned between the SpeI and EcoRV sites of pAP1 resulting in pGH53 and pGH52 respectively. Capsule deficient strains MC58siaD-, 2120siaD- and 2594mynB- were transformed with plasmids pGH53 and pGH52. Stable expression of rs-GFP, CFP and YFP was confirmed by fluorescence microscopy (Zeiss LSM510 Confocal Laser Scanning Microscope (Carl Zeiss, Jena, Germany) at 508 nm, 483 nm and 542 nm respectively.

Insertional inactivation and complementation of the pilX and pilE genes

For cloning of the pilX gene, a 2.9 kb DNA fragment was amplified from strain MC58 with primers HC488 and HC489 and ligated into pCR-Script® (Stratagene, Amsterdam, the Netherlands), resulting in plasmid pHC31. Using as a template plasmid pHC31, an inverse PCR was performed with primers HC492 and HC493. The PCR product was restricted with NsiI and ligated with the PstI DNA fragment of pUC4K comprising the kanamycin resistance cassette resulting in plasmid pHC32. The unencapsulated strains expressing GFP, i.e. strains 3349, 3379 and 3380 were transformed with plasmid pHC32 to generate a pilX knock-out. Complementation of the pilX knock-out was achieved by cloning the pilX gene into a derivative of pAP1. Firstly, the erythromycin resistance cassette in pAP1 was excised with SacI and NsiI. The resulting DNA fragment was incubated with T4 DNA polymerase and ligated with the 2.2 kb SmaI restriction fragment of pHP45Ω harbouring the aadA gene conferring spectinomycin resistance to the new plasmid pAP2-1. Subsequently, the pilX gene was amplified from strain MC58 with the primers HC511 and HC512. The PCR product was cloned between the SpeI and the EcoRI restrictions sites of pAP2-1 resulting in plasmid pHC33. The MC58 pilX knock-out mutant was transformed with pHC33. The pilX knock-out and the subsequent complementation of the knock-out were confirmed by PCR and Southern blot hybridization. Additionally, inactivation of the pilX gene was indirectly confirmed by an aggregation assay (Helaine et al., 2005). The pilE gene (NMB0018) with 500 bp upstream and downstream sequences, respectively, of strain MC58 was amplified with primers HC526 and HC527, and cloned into the vector pCR®2.1-Topo® (Invitrogen, Karlsruhe, Germany), resulting in plasmid pHC37. To delete the pilE gene from pHC37 an inverse PCR was performed using primers HC539 and HC540. After restriction of the resulting PCR product with MfeI, it was ligated with the 1.2 kb EcoRI fragment of pUC4K comprising the kanamycin resistance cassette. The resulting plasmid pHC38, in which the pilE sequence was replaced by a kanamycin cassette, was used for transformation of strain 3349. Complementation of the pilE mutation by PilE expression in trans could only be accomplished, if an expression plasmid was introduced first, because pilE mutation abolished natural competence. PilE expression in trans was achieved by transformation of meningococci with the expression plasmid pHC36. For construction of pHC36, the pilE gene of strain MC58 was amplified with primers HC530 and HC531. The PCR product was restricted with SpeI and EcoRI and cloned into the respective restriction sites of pAP2-1 resulting in plasmid pHC36.

RT-PCR and DNA sequencing of NMB0889

Expression of NMB0889 mRNA was tested performing RT-PCR with primers HC475r and HC514 using Qiagen® OneStep RT-PCR kit. RT-PCR parameters were used as described by Claus et al. (2004) with the following exception: Extension at 72°C was performed for 2 min. Automated DNA sequencing was performed with the dye-deoxy terminator cycle method on a Applied Biosystems 377 sequencer.

Western blot

Pilin (PilE) expression was assayed by Western blot analysis using mAB SM1 (anti class I pili, kindly provided by M. Virji; Bristol, UK) and mAB AD211 (anti class II pili, kindly provided by Mark Achtmann; Berlin, Germany). For detection, a goat anti-mouse IgG conjugated with horseradish peroxidase was used.

Immunofluorescence

Immunofluorescence of meningococci was performed on a Zeiss Axio Imager Z1 fluorescence microscope (Carl Zeiss, Jena, Germany) using standard protocols (Hammerschmidt et al., 1996b). For staining class I pili, the monoclonal antibody SM1 was used, for staining of class II pili mAB AD211. Alexa Fluor 488 nm goat anti-mouse IgG antibody (Molecular Probes®) was used for detection.

Flow chamber experiments

The experiments were performed at the Center for Biomedical Microbiology of the Danish Technical University. Biofilms were cultivated at 37°C in three-channel flow cells with individual channel dimensions of 1 × 4 × 40 mm. Stable growth temperature was maintained by using a culture room at 37°C. The flow system was assembled and prepared as previously described (Christensen et al., 1999). A 24 × 50 mm glass cover slip (Knittel Gläser, Braunschweig, Germany) was used as substratum for biofilm growth. Before each experiment, the system was sterilized by pumping a 0.5% (wt/vol) hypochlorite solution into the system and leaving it there for 4 h. The system was consecutively flushed with 2 l of sterile water. Thereafter, the flow chamber was rinsed with modified NDM overnight at 37°C. Inocula were prepared as follows: meningococcal cultures on GC agar plates grown for less than 8 h were resuspended in modified NDM, diluted to a density of 10⁸ cells ml⁻¹, and incubated at 37°C for 5–6 h shaking. Two hundred microlitres portions of these cultures with 1.0 × 10⁸ cells ml⁻¹ were injected into each channel of the flow cell. The flow cell was turned upside down for 1 h to allow for adhesion of cells to the glass surface without flow. After 1 h the flow was resumed. During growth of biofilms the medium was pumped through the flow cells at a constant rate of 0.2 mm s⁻¹ by using a peristaltic pump (model 205S; Watson Marlow, Calmouth, UK).

Microscopy and image analysis

All microscopic observations and image acquisitions were performed on a Zeiss LSM510 Confocal Laser Scanning Microscope (Carl Zeiss, Jena, Germany). Images were obtained using a 40×/1.3 Plan-Neofluar oil objective. Multi-channel simulated fluorescence projection images, and vertical and horizontal cross sections through the biofilm were generated by using the IMARIS software package (Bitplan AG, Zürich, Switzerland) running on an Indigo2 workstation (Silicon Graphics, Mountain View, USA). Images were further processed for display by using Photoshop (Adobe, Mountain View, USA). Biofilm development was quantified with the COMSTAT computer program (Heydorn et al., 2000). Strains were grown in three separate channels. From each channel six image stacks were acquired randomly down through the channel at different time points. The image stacks for each time point were gained from independent experiments, because the experimental procedure of image acquisition for COMSTAT analysis strongly interferes with meningococcal biofilm development at early time points.

Antibiotic treatment

The MIC of penicillin G, ciprofloxacin and rifampin (Sigma-Aldrich, München, Germany) was determined for the strains 3349 (MC58siaD-/gfp+), 3379 (2120siaD-/gfp+) and 3380 (2594mynB-/gfp+) using Etest^® (AB Biodisk, Solna, Sweden). Meningococci grown on Columbia agar were harvested with a sterile cotton swab and resuspended in 0.9% NaCl solution. The optical density of the bacterial suspension was adjusted to McFarland 0.5 using a Vitek Densicheck (bioMerieux, Marcy l'Etoile, France). Bacteria were streaked out on Mueller-Hinton agar supplemented with 5% sheep blood to achieve subconfluent growth after incubation at 37°C for 18–24 h in the presence of an Etest strip. Etest results were evaluated according to the manufacturers instructions. For strains MC58siaD-, 2120siaD- and 2594mynB-, the penicillin MICs (μg ml⁻¹) were 0.032, 0.047 and 0.047, respectively, the ciprofloxacin MICs were 0.003 for all strains, the rifampin MICs were 0.004, 0.004 and 0.012 respectively. Antibiotic susceptibility was assessed for biofilms grown under flow and static conditions as well as for planktonic cells. Biofilm susceptibility to penicillin, ciprofloxacin and rifampin in the biofilm flow system was tested by supplementing the modified NDM with 10 times MIC of the respective antibiotics. Twenty-four-hour-old biofilms were treated with antibiotics under constant flow conditions for 4 h, and 24 h respectively. Bacterial viability in biofilms was estimated by adding with 1 μM propidium iodide (PI) to the medium, which selectively stains dead cells. PI in conjunction with GFP has been successfully used as a dead cell stain for P. aeruginosa (Hentzer et al., 2001). Live GFP-stained and dead PI-stained bacteria were visualized by CLSM. Antibiotic susceptibility in a static biofilm model was assessed in 24-well cell culture dishes on round 15 mm glass cover slips. One millilitre with 10⁸ cells ml⁻¹ of bacterial suspension was seeded per well to initiate biofilm growth. The medium of a 24 h old biofilm was replaced by fresh medium containing 10 times MIC of penicillin, ciprofloxacin and rifampin respectively. After another 24 h the biofilm was rinsed twice with medium without antibiotics. Then the biofilms were harvested with cotton swabs and vortexed for 1 min to disintegrate the biofilm and resuspend the bacteria. The quantity of viable biofilm cells was determined by serial dilutions and plating on GC agar. In the same set of experiments, the antibiotic susceptibility of planktonic cells was tested. Static biofilms were grown as described above. The biofilm medium of a 24 h old biofilm was replaced by fresh medium containing 10 times MIC of penicillin, ciprofloxacin and rifampin respectively. Then the biofilm was harvested with cotton swabs to disintegrate the biofilm and vortexed for 1 min to resuspend the biofilm bacteria. These planktonic cultures were incubated for 24 h at 37°C and 200 r.p.m. Then the planktonic cells were washed twice in antibiotic-free medium and plated on GC agar to determine viable cells.

Aggregation assay

To quantify bacterial aggregation, we used a modification of an assay that was designed for N. meningitidis (Helaine et al., 2005). Bacteria grown for less than 10 h on GC agar were thoroughly resuspended in modified NDM or p.p.m., respectively, and adjusted to an OD₆₀₀ of 0.1 in 10 ml. Bacteria were then grown in glass test tubes under constant agitation at 200 r.p.m. until OD₆₀₀ of 0.9 was reached. Subsequently, the suspensions were incubated at room temperature without shaking and the OD₆₀₀ of the suspensions close to the liquid air interface was measured at 30 min intervals.

Crystal violet assay

Biofilm growth in a static model was assessed in modified NDM in 24-well cell culture dishes on round 15 mm glass cover slips. One millilitre with 1.0 × 10⁸ cells ml⁻¹ was seeded per well to initiate biofilm growth. After 24 h, the biofilm was washed with 1 ml of distilled water per well. To stain the adherent bacteria, 250 μl of 0.05% Crystal Violet (CV) (Difco, Augsburg, Germany) were added for 10 min. After two washes with each 1 ml of distilled water, the CV stain of the biofilm was dissolved in 96% ethanol and quantified by measuring optical density at 570 nm.

Production of rabbit antibodies

For expression of a PilX fusion with glutathione S-transferase (GST), the GST Gene Fusion System (Amersham Biosciences Europe, Freiburg, Germany) was used. For this purpose, the pilX gene of strains MC58 and 2120 was amplified by PCR omitting the 5′-sequence encoding the first 62 residues of the PilX preprotein. For strain MC58, the primers ML20 and ML27 were used, for strain 2120 primers ML22 and ML27. Primers contained a restriction site for BamHI. PCR products were restricted with BamHI and cloned into the BamHI restricted plasmid pGEX-3X. The proteins were expressed in E. coli DH5α at 27°C by induction with 1 mM isopropyl-beta-d-thiogalactopyranoside (IPTG). Bacteria were lysed using the BugBuster^TM Master Mix (Novagen, Darmstadt, Germany). The insoluble proteins were purified by separation using sodium dodecyl sulphate (SDS)-polyacrylamide gel electrophoresis, and excision of coomassie blue stained bands, which were consecutively elctroeluted using the BIOTRAP system (Schleicher und Schuell, Dassel, Germany). Rabbits were immunized by immunoGlobe Antikörpertechnik (Himmelstadt, Germany).

Statistical analysis

Substratum coverage was calculated by the COMSTAT computer program (Heydorn et al., 2000). Structural biofilm parameters were evaluated with Microsoft Excel. Statistical evaluation was performed using two-tailed paired Student's t-test. COMSTAT data of six images obtained from a single channel were combined to a single mean value, which served as input for the t-test.