Major pathogenic clonal complexes (cc) of Neisseria meningitidis differ substantially in their point prevalence among healthy carriers. We show that frequently carried pathogenic cc (e.g. sequence type ST-41/44 cc and ST-32 cc) depend on extracellular DNA (eDNA) to initiate in vitro biofilm formation, whereas biofilm formation of cc with low point prevalence (ST-8 cc and ST-11 cc) was eDNA-independent. For initial biofilm formation, a ST-32 cc type strain, but not a ST-11 type strain, utilized eDNA. The release of eDNA was mediated by lytic transglycosylase and cytoplasmic N-acetylmuramyl-l-alanine amidase genes. In late biofilms, outer membrane phospholipase A-dependent autolysis, which was observed in most cc, but not in ST-8 and ST-11 strains, was required for shear force resistance of microcolonies. Taken together, N. meningitidis evolved two different biofilm formation strategies, an eDNA-dependent one yielding shear force resistant microcolonies, and an eDNA-independent one. Based on the experimental findings and previous epidemiological observations, we hypothesize that most meningococcal cc display a settler phenotype, which is eDNA-dependent and results in a stable interaction with the host. On the contrary, spreaders (ST-11 and ST-8 cc) are unable to use eDNA for biofilm formation and might compensate for poor colonization properties by high transmission rates.
Neisseria meningitidis (the meningococcus) is a major cause of morbidity and mortality in children and adolescents (Rosenstein et al., 2001). Healthy individuals frequently carry meningococci in the nasopharynx (Cartwright et al., 1987). The major virulence factor of N. meningitidis is the polysaccharide capsule, of which serogroups A, B, C, W-135 and Y are associated with most cases of disease (Rosenstein et al., 2001). However, a considerable proportion of isolates from healthy carriers does not constitutively express a polysaccharide capsule (Claus et al., 2000; Dolan-livengood et al., 2003; Weber et al., 2006). Nasopharyngeal carriage of N. meningitidis and closely related species induces a type-specific protective immunity to invasive meningococcal disease (Goldschneider et al., 1969). Meningococci appear in tonsillar tissue as cell aggregates (Sim et al., 2000). As these so-called microcolonies are considered the basic unit of bacterial biofilms (Costerton et al., 1995), tonsillar multicellular meningococcal aggregates might resemble biofilm structures. Given that no appropriate animal models for meningococcal carriage exist, in vitro models of N. meningitidis biofilm formation have been established (Yi et al., 2004; Lappann et al., 2006; Neil and Apicella, 2009).
Neisseria meningitidis can be regarded as a metapopulation species comprising multiple clonal complexes (cc), which are defined by multilocus sequence typing (MLST) (Maiden et al., 1998). In Europe, invasive meningococcal disease is mostly caused by the lineages ST-41/44 cc (mostly serogroup B capsule), ST-32 cc (mostly B), ST-11 cc (mostly C) and ST-8 cc (mostly C and B) (Brehony et al., 2007). MLST analyses of isolates from healthy carriers indicate that point prevalences of carriage of the potentially virulent cc differ substantially. In several carriage studies, ST-11 and particularly ST-8 meningococci were underrepresented; however, ST-41/44 cc and ST-32 cc were frequently detected among healthy carriers (Yazdankhah et al., 2004; Claus et al., 2005; Caugant et al., 2007; Maiden et al., 2008). More carriage MLST data are certainly needed to account for regional and temporal fluctuations, especially in outbreak situations such as in the Czech Republic in 1993, when ST-11 was isolated for a short period of time at unusually high frequency (Buckee et al., 2008). Nevertheless, carriage studies led to the observation that the disease association odds ratio is high for ST-11 cc for example, but low for ST-44/41 cc, despite the two lineages showing similar incidences of invasive disease (Yazdankhah et al., 2004). As yet, no pathogenicity factor or biological effector mechanism has been identified to explain this discrepancy (Schoen et al., 2008).
Biofilms, which are surface-attached microbial communities embedded in a self-produced extracellular matrix of polymeric substances, are considered the favoured lifestyle of bacteria in natural and clinical settings (Costerton et al., 1999). Gene expression profiling of Pseudomonas aeruginosa (Whiteley et al., 2001), Staphylococcus aureus (Resch et al., 2005) and Escherichia coli (Schembri et al., 2003) revealed major differences between cells of biofilms and of planktonic cultures. The unique gene expression pattern and the physicochemical properties of biofilms can generate phenotypes that are highly resistant to effectors of the innate and adaptive immune system (Yasuda et al., 1994; Kristian et al., 2008), and to antimicrobial substances (Luppens et al., 2002; Fux et al., 2004).
Biofilm formation of N. meningitidis is a general trait of unencapsulated strains, whereas capsule expression prevents biofilm formation in standard biofilm assays (Yi et al., 2004; Lappann et al., 2006). Using an epithelial cell model, biofilm formation has also been observed for encapsulated meningococci (Brock Neil et al., 2009). Meningococcal biofilms were tolerant of high doses of penicillin, and microcolony formation was dependent on twitching motility (Lappann et al., 2006). The knockout of the two-partner secretion system HprA led to a moderate reduction in biomass of mature biofilms (Neil and Apicella, 2009). However, factors or mechanisms indispensable for meningococcal biofilm formation are currently unknown. Furthermore, an extracellular matrix in which the cells in N. meningitidis biofilms are embedded has not been identified.
Pseudomonas aeruginosa and staphylococci are model organisms in biofilm research. The P. aeruginosa laboratory strain PAO1 forms typical mushroom-like biofilm structures, which consist of a non-motile stalk population and a motile cap population (Klausen et al., 2003). Extracellular DNA (eDNA) was identified as a cell-to-cell or cell-to-substratum connecting component in biofilms of PAO1 as well as in biofilms of clinical isolates (Whitchurch et al., 2002; Nemoto et al., 2003). eDNA was mainly identified at high concentrations at the interface between stalk and cap (Allesen-holm et al., 2006), and has been shown to be involved in cap formation (Barken et al., 2008). eDNA in PAO1 and staphylococcal biofilms originates from chromosomal DNA, and has been suggested to be released by autolysis (Allesen-holm et al., 2006; Qin et al., 2007; Rice et al., 2007).
In this study, we demonstrate that DNase I sensitivity of meningococcal biofilms is a widespread phenomenon in the meningococcal population, and that eDNA is an important structural component of the extracellular matrix of N. meningitidis biofilms. eDNA supported initial biofilm formation. The release of early eDNA could be correlated to murein hydrolase and N-acetylmuramyl-l-alanine amidase activity. eDNA in late biofilms mechanically stabilized biofilm structures. The autolytic activity of outer membrane phospholipase A partly caused late eDNA release. The differential occurrence of DNase sensitivity and autolytic activity among clonal lineages provides hypotheses on the different pattern of host adaptation of meningococcal lineages.
DNase I reduces meningococcal biofilm formation
Unencapsulated meningococci form biofilms in flow chambers and under static conditions (Lappann et al., 2006). Using a static biofilm assay, the impact of DNase I on initial biofilm formation was tested for 51 unencapsulated strains (Table S1) belonging to different clonal complexes (cc), as assigned by MLST. Strains were retrieved from invasive disease and carriage and were selected on the basis of their frequency in the meningococcal carriage (Fig. 1B, data from Claus et al., 2005) and in invasive disease in Germany (Fig. 1C, data from Brehony et al., 2007). The clonal complexes ST-41/44 cc and ST-32 cc were selected for their high prevalence in carriage and disease; ST-11 cc and ST-8 cc were selected because they were predominantly found in invasive disease; ST-23 cc, ST-53 cc, ST-60 cc and ST-198 cc were included as lineages predominantly isolated from healthy carriers. DNase I was added to the bacterial suspensions 10 min prior to biofilm inoculation. DNase I was present at all stages of initial biofilm development, as confirmed by testing the DNA degradation activity of the biofilm supernatants (data not shown). DNase I did not influence meningococcal growth in liquid cultures, and heat-inactivated DNase I did not influence meningococcal biofilm formation (data not shown). Initial biofilm formation of 30 of the 51 strains was significantly reduced (by 75–95%) by DNase I treatment (Fig. 1A, y-axis). All tested strains of ST-23 cc, ST-32 cc, ST-41/44 cc, ST-53 cc, ST-60 and ST-198 cc showed high sensitivity to DNase I during initial biofilm formation. In contrast, initial biofilm formation by strains of ST-8 cc and ST-11 cc was slightly affected by DNase I (Fig. 1A). Hence, DNase sensitivity of meningococcal biofilms is a widespread but not universal phenomenon.
Next, the effects of DNase I on biofilm formation of strain MC58siaD−/gfp+ (ST-32 cc), the initial biofilm formation of which was DNase-sensitive (Fig. 1A), and strain 2120siaD−/gfp+ (ST-11 cc), the initial biofilm formation of which was DNase-tolerant (Fig. 1A), were assessed in a standardized flow-chamber system. Figure 2A shows the biofilm formation of the green fluorescent strain MC58siaD−/gfp+ after treatment with DNase I at three different time points in the flow cell system. Images were obtained using confocal laser scanning microscopy (CLSM). DNase I present in the inoculum and maintained for further 6 h in the growth medium entirely suppressed biofilm formation of strain MC58siaD−/gfp+ (panel I). Treatment of 6-hour-old flow cell biofilms for a further 6 h with DNase I had a clear suppressive effect on biofilm formation, compared with the untreated control (panel II). In contrast, addition of DNase I to 12-hour-old biofilms of strain MC58siaD−/gfp+ had no visible effect (panel III). Statistical analysis of image series from panels I, II and III using the COMSTAT software (Heydorn et al., 2000) confirmed the suppressive effect of DNase I on initial biofilm formation of MC58siaD−/gfp+, while there was only a minor, but not significant effect on strain 2120siaD−/gfp+ (Fig. 2B). Thus, also in the continuous flow system, DNase I interfered with initial biofilm formation events, including the attachment of single planktonic meningococci to the substratum and to each other. DNase I that was added late after biofilm formation had a smaller effect on biofilms of strain MC58siaD−/gfp+ compared with DNase I added early. In accordance with the static biofilm assay results (Fig. 1A), biofilm formation of strain 2120siaD−/gfp+ was not affected by DNase treatment at any time point (Fig. 2).
Exogenous chromosomal DNA enhances the initial attachment of eDNA-free meningococci
DNase I treatment blocked initial biofilm formation of strain MC58siaD−/gfp+, but not of strain 2120siaD−/gfp+, in the continuous flow system. To test whether highly purified (protein free) exogenous DNA could reverse the effect of DNase I, DNase-treated pre-cultures were incubated with highly purified meningococcal chromosomal DNA. Bacteria were allowed to adhere for 30 min without flow. As expected, DNase I-treated MC58siaD−/gfp+ (Fig. 3B), both with and without subsequent washing (Fig. 3C), adhered poorly to the substratum compared with the non-treated control (Fig. 3A), whereas DNase I treatment only marginally affected adherence in strain 2120siaD−/gfp+ (Fig. 3A–C). DNase I-treated bacteria that were washed twice and then mixed with highly purified chromosomal DNA adhered more efficiently than DNase I-treated cells without subsequent addition of chromosomal DNA (Fig. 3D). Differences among the four experimental groups (Fig. 3A–D) were quantified using COMSTAT software. The quantitative analysis confirmed the differential susceptibilities of strains MC58siaD−/gfp+ and 2120siaD−/gfp+ towards DNase I, and the partial restoration of biofilm formation by exogenous, highly purified DNA (Fig. 3). It is important to note that meningococcal, salmon sperm and E. coli DNA equally performed in this regard (data not shown).
Biofilms of a DNase-sensitive strain incorporate higher amounts of eDNA than biofilms of a DNase-tolerant strain
Strain MC58siaD− (ST-32 cc) biofilms were DNase-sensitive, in contrast to DNase-tolerant biofilms of strain 2120siaD− (ST-11 cc), as demonstrated under static and flow conditions (Figs 1A and 2). To visualize eDNA in the matrix of biofilms of ST-32 cc and ST-11 cc strains, we developed staining procedures for eDNA in meningococcal biofilms using fluorescent dyes. Attempts to use propidium iodide and ethidium bromide were unsuccessful because the dyes penetrated the cell membranes of motile and, thus vital, meningococci. Visualization of eDNA was successful with SYTOX Orange. SYTOX Orange binds to dsDNA (double-stranded DNA) and permeates disintegrated membranes of dead cells, which then appear red upon excitation. Figure 4 shows SYTOX Orange-stained 6-hour-old biofilms of three ST-32 cc and three ST-11 cc strains. In meningococcal biofilms, eDNA was not confined to discrete inter-cellular structures, but was predominantly located at cell surfaces. Live cells devoid of significant amounts of eDNA appeared green due to the expression of green fluorescent protein (GFP). Live cells covered with eDNA appeared yellow or slightly red due to the overlay of the intense green fluorescence of the bacteria and the weaker red fluorescence of the eDNA-bound SYTOX Orange. Vitality of these cells was confirmed by microscopic observation of twitching motility. Dead cells were visible as slightly oversized red cells due to a very strong fluorescence signal of the SYTOX Orange bound to the chromosomal DNA and a moderate signal of remaining GFP. The images were slightly overexposed for the red channel in order to enhance the fluorescence signal from the SYTOX Orange-stained eDNA. Identical microscopic settings were used to display eDNA in biofilms of both strains. The DNase-sensitive biofilms of ST-32 cc strains included a high proportion of eDNA-covered live cells, whereas those were rare in the DNase-tolerant biofilms of ST-11 cc strains. To quantify the cell-bound eDNA, images obtained from four strains each of ST-11 cc and ST-32 cc were subjected to computer-assisted image analysis. The proportionate area of cells covered with eDNA was significantly higher in strains of the ST-32 complex (0.777 versus 0.067, P ≤ 0.001), yet did not differ significantly between strains within a complex (P = 0.70 and P = 0.27 for ST-11 cc and ST-32 cc respectively).
We therefore conclude that DNase-sensitive biofilms harbour significantly more eDNA than biofilms of DNase-tolerant strains. However, both the DNase-sensitive and the DNase-tolerant biofilms included similar sized fractions of dead cells, indicating similar degrees of cell death and DNA release in early biofilms. The dead cells within the biofilms of both strains were randomly distributed throughout the cross-sections of the early biofilms, indicating a stochastic process of cell death rather than the death of a distinct subpopulation of cells induced by, for example, microenvironmental conditions.
Biofilm formation in MltA, MltB and AmpD mutants
eDNA was essential for initial biofilm formation of strain MC58siaD− and was also released during biofilm formation. The DNA released by the small fraction of lysed cells is considered sufficient to ensure initial biofilm formation in S. aureus and P. aeruginosa (Allesen-holm et al., 2006; Rice et al., 2007). We hypothesized that factors that have been demonstrated to be involved in cell separation, cell wall recycling and autolysis in E. coli (Park, 1995; Bramhill, 1997; Firczuk and Bochtler, 2007; Vollmer et al., 2008) might contribute to early DNA release and initial biofilm formation in meningococci. To test this hypothesis, knockout mutants of the membrane-bound lytic transglycosylase A (MltA), the membrane-bound lytic transglycosylase B (MltB) and the cytoplasmic N-acetylmuramyl-l-alanine amidase (AmpD) of the green fluorescent strains MC58siaD−/gfp+ (ST-32 cc) and 2120siaD−/gfp+ (ST-11) were tested for biofilm formation under flow conditions. Please note that for reasons of clarity in this paragraph wild-type and mutant strains were not termed ‘gfp+’, despite of GFP expression. Before inoculation, cells from agar plates were subjected to three growth cycles (passages) in liquid medium. Twelve hours after inoculation with bacteria of the early exponential growth phase of passage 3, biofilm formation was monitored with CLSM. The mltA, mltB and ampD mutants of strain MC58siaD− were largely deficient in biofilm formation, in contrast to strain 2120siaD− (Fig. 5A). Of note, the mltA and mltB mutants were not impaired in growth in liquid medium (data not shown). It is further noteworthy that mltA mutants of both strains exhibited a prominent cell separation defect leading to the formation of grape-like aggregates (data not shown), which confirmed previous findings in N. meningitidis (Adu-bobie et al., 2004). The formation of dense aggregates likely resulted in an underestimation of cell mass in the mltA mutants compared with the wild-type because cell mass was extrapolated from optical density values. Biofilm formation was largely restored in mltA, mltB, and ampD mutants of strain MC58siaD− when stationary phase bacteria of passage 3 were used as the inoculum (Fig. 5B). As biofilm formation of strain MC58siaD− was impaired by DNase I treatment (Figs 1A and 2), and eDNA could be visualized in early MC58siaD− biofilms (Fig. 4), we hypothesized that biofilm formation of the mutants of strain MC58siaD− was reduced owing to reduced DNA release during early exponential growth, which was necessary for strain MC58siaD− biofilm formation. 2120siaD− biofilm formation was not impaired in the mutants because 2120siaD− biofilm formation is independent of eDNA, as shown above. The biofilm-deficient phenotypes of mltA, mltB, and ampD mutants of strain MC58siaD− could be reverted using stationary-phase bacteria, which were likely partially lysed and therefore provided sufficient eDNA. Neisseria are known to be fragile bacteria that rapidly undergo cell lysis during stationary phase. This process is driven by factors other than MltA, MltB and AmpD, such as the outer membrane phospholipase A, which is activated in liquid cultures during stationary phase and which is absent from the ST-11 cc (Bos et al., 2005).
To test the hypothesis that the significant reduction in biofilm formation of mltA, mltB and ampD mutants of strain MC58siaD− was caused by reduced DNA release during early exponential growth, we assessed the DNA release into the culture medium of strain MC58siaD− and the respective mutants grown for 1–3 passages in liquid medium. The DNA release was determined after 1 h of growth using the PicoGreen dsDNA quantification reagent. As shown in Fig. 6A the mltA, mltB and ampD mutants released less dsDNA into the culture medium after three passages in liquid medium than the wild-type. These findings indicate that cells derived from agar plates were loaded with eDNA, which in the mutants is consecutively lost by serial passages.
We then asked whether the inhibitory effect of reducing the availability of eDNA on initial biofilm by growth in liquid culture could be reverted by exogenous DNA. To best mimic natural conditions we used crude DNA preparations obtained by autolysis. Initial attachment of wild-type and mutant strains in a static 24-well plate assay was tested, as shown for the ampD mutant (Fig. 6B). In the wild-type strain three passages in liquid medium led to a moderate reduction in initial attachment. The initial attachment of the wild-type after three passages could be almost completely prevented by DNase treatment, and could be reverted by addition of crude DNA, but not pure DNA (Fig. 6B). In all three mutants, as exemplified for the ampD mutant, the three passages in liquid medium led to a marked reduction of initial attachment, which could be stimulated by crude DNA, but not by pure DNA or DNase-treated or Proteinase K-treated crude DNA. We conclude that there is a likely coaction of DNA and proteinaceous constituents in initial attachment of meningococci. These results show that the reduction of DNA release by mutation of genes involved in cell wall remodeling and autolysis decreased initial biofilm formation. The addition of crude meningococcal DNA or the use of stationary-phase bacteria could restore initial biofilm formation.
OMPLA-mediated autolysis leads to mechanical stabilization of biofilm structures
Initial biofilm formation of strain MC58siaD− was dependent on the presence of eDNA, the release of which is, at least in part, dependent on MltA, MltB and AmpD. The pldA-encoded autolysin outer membrane phospholipase A has been reported to be active during stationary phase liquid in cultures, leading to DNA release, but inactive during exponential growth phase (Bos et al., 2005). OMPLA-mediated DNA release was not involved in initial biofilm formation because biofilm formation of a pldA knockout mutant of strain MC58siaD− was not reduced compared with the wild-type nor was less affected by DNase treatment (Fig. S1). We therefore asked for the impact of OMPLA-mediated DNA release for later stages of biofilm formation. For clarity, in this chapter, GFP-expressing variants of strains MC58siaD− and 2120siaD− are not termed ‘gfp+’. We selected strain 2120siaD− as a model because it does not express functional OMPLA owing to a point mutation in the pldA-coding sequence and therefore is non-autolytic (Fig. S2); this has recently been shown for other ST-11 strains (Bos et al., 2005). Furthermore, this strain exhibited very low proportions of dead cells in mature biofilms, which is in contrast to strain MC58siaD− (data not shown). A constitutively OMPLA-expressing variant of strain 2120siaD− was generated by the introduction of pldAMC58 into a neisserial expression vector. Approximately 90% of the OMPLA-expressing variant 2120siaD−/pldAMC58 were lysed within 48 h in a non-agitated liquid culture, while almost no cell lysis was seen for the OMPLA-negative strain 2120siaD− (Fig. S2).
The role of OMPLA-mediated DNA release in biofilms of strain 2120siaD− was then investigated. The GFP expressing variants were counterstained with SYTOX Orange. The biofilm of the OMPLA non-expressing variant harboured a very low proportion of dead cells and live cells covered with eDNA, whereas the biofilm of the OMPLA-expressing variant of strain 2120siaD− included high proportions of dead cells (Fig. 7A). The basal inner part of the microcolonies exhibited a distinct zone of predominantly dead cells. COMSTAT analyses of SYTOX Orange-stained biofilms revealed that the biofilm of the OMPLA-positive variant harboured five times more dead cell material than the biofilm of the OMPLA-negative variant (Fig. 7B).
We then asked whether the pronounced release of chromosomal DNA in mature biofilms of the OMPLA-expressing variant of strain 2120siaD− influences the mechanical stability of the biofilm. The biomass and surface coverage of 12-hour-old biofilms of strains 2120siaD− and 2120siaD−/pldA(MC58) were quantified with COMSTAT before and after the 50-fold increase of the flow for 10 min. Almost 80% of the biomass of the OMPLA negative variant was washed away by the high shear force, whereas the biomass of the OMPLA-positive variant was not affected (Fig. 8). In conclusion, eDNA released by OMPLA-mediated autolysis cross-linked the single cells within the microcolonies, thereby mechanically protecting the biofilm structure against shear stress.
DNase sensitivity was unequally distributed among the meningococcal population (Fig. 1A). Due to the functional relevance of OMPLA for mature biofilms, we also investigated the distribution of autolysis phenotypes among the same set of 51 strains. All ST-23 cc, ST-32 cc, ST-41/44 cc, ST-53 cc, ST-60 cc and ST-198 cc strains exhibited a high degree of autolysis, whereas only 1 of 6 ST-8 cc and 1 of 15 ST-11 cc strains displayed a 50% reduction in OD600 over 48 h (Fig. 1A; x-axis). There was a strong correlation between DNase sensitivity of initial biofilm formation and autolysis, and between DNase tolerance and diminished autolysis, among the tested strains (Fig. 1A). It should be stressed that autolysis and DNase sensitivity were not functionally linked, as pldA mutation and complementation in strains MC58 and 2120, respectively, did not alter DNase I sensitivity. Lack of autolysis and of DNase I sensitivity (Fig. 1A) was observed in strains of the ST-8 and ST-11 complex. Of interest, these lineages in the Bavarian meningococcal carriage collection were under-represented (Fig. 1B, data from Claus et al., 2005), although they contributed to a major share of invasive disease in Germany (Fig. 1C, data from Brehony et al., 2007).
Competition in biofilm formation between DNase sensitive and tolerant strains
Finally, competition in biofilm formation between DNase sensitive and insensitive strains was investigated. Using COMSTAT, biomasses of the DNase-sensitive biofilms of the ST-32 cc type strain MC58siaD− and the DNase-tolerant biofilms of the ST-11 type strain 2120siaD− were determined at 1, 6, 12 and 24 h after inoculation. After 1 h of biofilm formation the DNase-sensitive biofilm of strain MC58siaD− developed eightfold more biomass than the DNase-tolerant biofilm of strain 2120siaD− (Fig. 9A), pointing to more efficient initial biofilm formation by the DNase-sensitive strain. After 12 and 24 h, however, biofilms of both strains contained similar amounts of biomass (Fig. 9B).
To set up biofilm formation competition experiments, homogenous 1 : 1 mixtures of cells of CFP-labelled strain MC58siaD− and YFP-labelled strain 2120siaD−, which were controlled by microscopic enumeration, were used as inoculum in the continuous flow system. Twelve hours after inoculation biofilms almost completely consisted of strain MC58siaD− cells, whereas only few cells of strain 2120siaD− were visible (Fig. 9). The effect was independent of the fluorescent proteins used as suggested by colour swapping (Fig. 9). There was no growth inhibition during competitive growth in planktonic cultures (data not shown). The data suggest that the use of eDNA for initial attachment and as a stabilizer of late biofilm structures is beneficial for in vitro biofilm formation.
In the present study we provide evidence that eDNA contributes to meningococcal biofilm formation and is present in the extracellular matrix of meningococcal biofilms. The inhibitory effect of DNase I on biofilm formation could be reverted by adding chromosomal DNA. Some meningococcal lineages did not use eDNA for biofilm formation. However, eDNA-dependent biofilm founders more rapidly produced biomass and out-competed eDNA-independent biofilm founders. A model is proposed for meningococcal biofilm formation, which suggests two discrete functions for eDNA: first, as a mediator of the initial attachment; and second, as a structural component of the biofilm matrix ensuring mechanical stability of the overall biofilm architecture. We furthermore relate the in vitro model to the population biology of meningococci, for which segregation in settlers and spreaders is proposed.
Most of the previous studies addressing the role of eDNA in biofilm formation by the model organisms S. aureus, S. epidermidis and P. aeruginosa (Whitchurch et al., 2002; Qin et al., 2007; Rice et al., 2007) have been restricted to the investigation of only one strain or a few clinical isolates. Owing to the high genetic diversity of meningococci, we included 51 pathogenic and non-pathogenic strains to address the importance of eDNA in meningococcal biofilm formation. DNase sensitivity of biofilm formation did not correlate with the serogroup or the virulence properties of the parental strains; however, it was uniformly present for strains of ST-23 cc, ST-32 cc, ST-41/44 cc, ST-53 cc and ST-198 cc, and uniformly absent for strains belonging of the related ST-8 cc and ST-11 cc. This finding highlights the similarity of latter lineages, which are known to share MLST alleles at three of seven loci, serogroup and serosubtype profiles (Wang et al., 1993), and several chromosomal markers such as restriction modification systems (Claus et al., 2000; 2001). Furthermore, ST-11 cc and ST-8 cc strains, as shown in this study, were the only lineages containing autolysis-negative strains.
To the best of our knowledge, this is the first report that chromosomal DNA added to eDNA-free cells stimulates initial attachment. This is in contrast to the results of studies in S. epidermidis in which the addition of homologous or heterologous chromosomal DNA did not significantly affect biofilm formation of cells not being treated with DNase I beforehand (Qin et al., 2007). Of note, in the present study any further supply of DNA to cells already saturated with eDNA did not further increase initial attachment. The effect of DNase treatment could not be fully compensated by the use of highly purified chromosomal DNA, but rather by the use of crude DNA. The factors or substances other than DNA that promote initial attachment remain to be identified.
Several attempts have been made to visualize eDNA in bacterial biofilms. Extracellular DNA was shown to be present in a matrix of P. aeruginosa PAO1 biofilms (Matsukawa and Greenberg, 2004) and to affect the structure of the microcolonies (Allesen-holm et al., 2006). We now show that DNase-sensitive N. meningitidis biofilms contain eDNA, whereas DNase-tolerant biofilms are almost totally lacking eDNA. In contrast to the spatial structuring of eDNA in P. aeruginosa biofilms, eDNA in early meningococcal biofilms was not confined to certain areas of the microcolonies, but was localized on the surface of vital cells that were randomly distributed in early microcolonies.
How is eDNA released during N. meningitidis biofilm formation? Active excretion of chromosomal DNA by type IV secretion system (T4SS) has been described for N. gonorrhoeae (Hamilton et al., 2005). The gonococcal genetic island (GGI) encodes the DNA secretion system (Dillard and Seifert, 2001). By genomic screens using DNA microarrays, the GGI, and remnants thereof, were only rarely found in N. meningitidis. Only strains of serogroups H and W-135 were positive, but no functional proof of the T4SS was provided (Snyder et al., 2005).
We here provide evidence that autolysis mediated by the putative autolysins MltA/MltB and the phospholipase OMPLA is the mechanism of DNA release required for N. meningitidis biofilm initiation and stabilization respectively. MltA is involved in autolysis and cell separation in E. coli (Jennings et al., 2002; Adu-bobie et al., 2004). As mutation of mltA, which led to decreased DNA release, suppressed eDNA-dependent biofilm formation of strain MC58siaD−, but not eDNA-independent biofilm formation of strain 2120siaD−, decreased DNA release of the mltA mutant is considered to be the major reason why biofilm formation of strain MC58siaD− was reduced in the mltA mutant. Furthermore, we have clearly shown that the mutant forms biofilms per se because biofilm initiation by bacteria harvested directly from plates was successful. We suggest that mutant strains coming from agar plates carry surface-bound DNA liberated, e.g. by OMPLA-mediated autolysis in stationary phase cultures; however, in contrast to bacteria expressing MltA, MltB or AmpD they are impaired in releasing DNA in consecutive liquid growth cycles. We believe that DNA is constitutively released during all stages of the growth cycle. Low levels of DNA release have been demonstrated for many bacterial species (reviewed in Lorenz and Wackernagel, 1994), and have been shown to be sufficient for ensuring biofilm formation of S. aureus (Rice et al., 2007) and P. aeruginosa (Allesen-holm et al., 2006).
How might lytic transglycosylase-mediated DNA release be controlled in N. meningitidis? In S. aureus the murein hydrolase regulator CidA, which shows homologies to phage holins, controls DNA release in S. aureus biofilm formation (Rice et al., 2007), presumably by the formation of a multimeric pore in the cytoplasmic membrane giving murein hydrolases access to their target structures (Bayles, 2007). CidA orthologues, i.e. CidA and CidB (NMB2003 and NMB2004), are conserved in N. meningitidis (Bayles, 2007), and show 24% and 29% amino acid sequence identity respectively. In preliminary studies we have replaced the cidAB gene cluster in strain MC58 by a kanamycin resistance cassette to elucidate the impact of the staphylococcal homologues on biofilm initiation. Interestingly, the double mutant completely lacked the propensity to form biofilms despite of normal growth in liquid medium. In analogy to S. aureus, this finding suggests that meningococcal cidAB directly or indirectly contributes to DNA release. However, the cidAB genes in meningococci have not been explored at all, and future research is needed to unravel the precise mode of action and interaction with murein hydrolases, and to dissect the function of each protein.
Neisseria meningitidis can be carried asymptomatically as microcolonies in tonsillar tissue (Sim et al., 2000). For maintenance in the host population, successful between-carrier transmission is needed, which requires the release of bacteria from microcolonies. Duration of carriage and transmission frequencies presumably are out-balanced to maintain successful lineages in the population. The formation of DNase-sensitive or DNase-tolerant biofilms might reflect two different adaptation strategies, which are schematically depicted in Fig. 10. We hypothesize that the ‘settler’ strategy is characterized by robust biofilm formation, which allows a long-term colonization of the host. By contrast, the meningococcal ‘spreader’ strategy implies instable biofilm formation, short-term colonization of the host, and outright release of bacteria from the nasopharyngeal microcolonies providing the bacteria with the opportunity to sweep through the host population by aerosol transmission. In line with this hypothesis, we could demonstrate that the ST-32 cc type strain (settler) out-competed the ST-11 cc type strain (spreader) in mixed biofilm experiments. Increased transmission rates might compensate for the low point prevalence of spreaders and help to maintain the lineage in the host population as proposed recently (Yazdankhah et al., 2004). Longitudinal carriage studies revealed that asymptomatic carriage of particular clones is a long-term relationship, with 90% of carriers hosting the same clone for at least 5–6 months (Caugant et al., 2007). ST-11 (spreader) was scarcely found and carriage was lost in less than 1 month (Caugant et al., 2007). Frequent encounter of new hosts furthermore might explain the fairly high incidence of invasive disease caused by these lineages (Moxon and Jansen, 2005).
It will be interesting to determine, whether the dynamics of geographical migration of distinct lineages correlate to the spreader/settler model. Unfortunately, longitudinal data on geographical movement of lineages are not available for carriage. It is nevertheless noteworthy that ST-41/44 cc (a settler) causes prolonged and geographically confined epidemics such as exemplified in New Zealand commencing in the 1990s (Baker et al., 2001). In contrast, a specific clone within the ST-11 cc (a spreader), which was first observed in 1986 in Canada, rapidly spread out to many countries on a worldwide scale (Jelfs et al., 2000).
In summary, this report underlines the importance of eDNA for biofilm formation. A dual role of eDNA for early and late biofilms was demonstrated. Early DNA release from a small fraction of cells was mediated by cell wall-modifying enzymes and was crucial for initial biofilm formation, whereas late DNA release of a larger fraction of cells by a cell membrane-targeting autolysin was beneficial for mechanical stability of the biofilm structure. Meningococci are interesting model organisms for research because they include lineages that substantially differ in their mechanisms of biofilm formation. The in vitro findings reported here could be associated with maintenance strategies in the host population.
Bacterial strains and culture conditions
Meningococcal strains are summarized in Table S1. They were obtained from a variety of sources including own strain collections (Caugant et al., 1987; Dunn et al., 1995; Vogel et al., 1998; Claus et al., 2002; 2005; Kurzai et al., 2005; Lappann et al., 2006; Weber et al., 2006; Elias and Vogel, 2007). Few strains were kindly provided by Dominique Caugant, Oslo. Standard cultivation was performed with GC agar and was supplemented with PolyViteX (bioMerieux). When appropriate, erythromycin (7 µg ml−1), kanamycin (100 µg ml−1), chloramphenicol (7 µg ml−1) and spectinomycin (100 µg ml−1) were added to the medium. For liquid cultures and biofilm experiments a modification of the neisserial minimal medium (NDM) (Archibald and DeVoe, 1978) was used and supplemented with PolyViteX and 5 mM NaHCO3 as described recently (Lappann et al., 2006). Most experiments were performed with derivatives expressing a fluorescent protein from the plasmid pEG2-Ery (Lappann et al., 2006), which also contains an erythromycin-resistance gene. The pEG2-Ery derivatives proved to be stable for several days of biofilm culture without the addition of antibiotics throughout the study. All experiments were performed at 37°C. A CO2-enriched atmosphere was used for GC agar cultures only.
Recombinant DNA techniques
Restriction enzymes and DNA-modifying enzymes were obtained from New England Biolabs. N. meningitidis chromosomal DNA was purified using the Qiagen Genomic tips system according to the manufacturer's instructions. Southern blot hybridizations were performed with digoxigenin-labelled probes (Roche). Recombinant plasmids were isolated using the QIAprep Spin miniprep kit (Qiagen). Primers were purchased from Sigma-Aldrich. PCR analyses were performed using the AmpliTaq DNA polymerase (Applied Biosystems).
Inactivation of ampD, mltA and mltB
For cloning of ampD, a 2003 bp DNA fragment was amplified from strain MC58 with the primers ML63 and ML64 (Table S2) and ligated into pBluescript II SK(+) (Stratagene), resulting in plasmid pML1 (Table S3). Using plasmid pML1 as a template, an inverse PCR was performed with primers ML65 and ML66. The resulting PCR product was restricted with NsiI and ligated with the PstI fragment of the pUC4K vector (GE Healthcare) that contained a kanamycin resistance cassette, resulting in plasmid pML2. MltA was cloned by amplification of a 2608 bp DNA fragment from strain MC58 with primers ML85 and ML86, followed by ligation into pBluescript II SK(+) vector, resulting in plasmid pML3. Inverse PCR was performed with primers ML51 and ML52, using plasmid pML3 as a template. The resulting PCR product was digested with NsiI and ligated with the PstI fragment of pUC4K, to give plasmid pML4. To clone mltB, a 2307 bp DNA fragment from strain MC58 was amplified with primers ML60 and ML45 and cloned into the pBluescript II SK(+) vector, resulting in plasmid pML5. Using plasmid pML5 as a template, inverse PCR was performed with primers ML42 and ML44. The PCR fragment was digested with EcoRI and ligated with the EcoRI fragment of pUC4K, resulting in plasmid pML6. The unencapsulated strains expressing GFP, namely strains 3349 and 3379, were transformed with plasmids pML2, pML4 or pML6 to generate ampD, mltA or mltB knockouts, respectively, which were confirmed by PCR and Southern blot hybridization.
Inactivation and complementation of pldA
For cloning of pldA, a 2171 bp DNA fragment was amplified from strain MC58 with primers TA3 and TA4 and ligated into pBluescript II SK(+) vector, to give plasmid pTA1. Using primers TA1-R and TA2-R, inverse PCR was performer using plasmid pTA1. The PCR product was digested with NsiI and ligated with the PstI DNA fragment of pUC4K containing the kanamycin resistance cassette, to generate plasmid pTA2. The unencapsulated meningococcal strains expressing GFP, namely strains 3349 and 3379, were transformed with plasmid pTA2 to create a pldA knockout. For complementation of the pldA knockout strains the pldA gene was amplified with primers TA5 and TA6, and was then ligated into the SpeI and EcoRI restriction sites of pAP2-1 (Lappann et al., 2006). The resulting plasmid was designated pTA3. The pldA knockout mutants of strains 3349 and 3379 were transformed with pTA3. All mutants were validated by PCR and Southern blot hybridization analyses. Furthermore, pldA inactivation and the complementation were phenotypically confirmed by an autolysis assay, as described previously (Bos et al., 2005).
Inactivation of meningococcal capsule expression
The inactivation of the serogroup B, C and Y capsule polymerases was performed as described previously (Ram et al., 2003; Kurzai et al., 2005). To generate a capsule-deficient serogroup 29E mutant, the first gene in region A of the capsule gene cluster was inactivated by insertion of a kanamycin resistance cassette. In brief, a DNA fragment of positions 288–2674 (GenBank accession number AJ576117, submitted by H.C., to be published separately) of strain α707 was cloned into the pCR-Scrip Amp SK(+) vector (Stratagene). Subsequently, the kanamycin resistance cassette of pUC4K was inserted into the SpeI site within open reading frame 1, resulting in plasmid pNB10. The capsule knockout was confirmed by slide-agglutination with a serogroup-specific antibody (Bio-Rad).
Flow cell biofilm experiments
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 using a culture room at 37°C. The flow system was assembled and prepared as described previously (Christensen et al., 1999). A 24 mm glass coverslip (Knittel Gläser) was used as substratum for biofilm growth. Sterilization of the flow system was performed as described previously (Lappann et al., 2006). Inocula were prepared as follows: meningococcal cultures grown on GC agar plates for a maximum of 8 h were resuspended in modified NDM, diluted to a density of 1 × 108 cells ml−1, and incubated at 37°C and 200 r.p.m. for 5–6 h. Samples (200 µl) of these suspensions of 1 × 108 cells ml−1 were injected into each channel of the flow cell. The flow cell was turned upside down for 1 h without flow to allow adherence to the glass surface. After 1 h, the flow was resumed. During growth of biofilms, medium was pumped through the flow cells at a constant rate of 0.2 mm s−1 using a Watson Marlow 205S peristaltic pump. To expose flow cell biofilms to high shear forces, the standard medium flow velocity of the pump of 0.2 mm s−1 was increased to 10 mm s−1 for 10 min. For seeding wild-type and mltA, mltB and ampD mutant biofilms, samples of passages 1 and 3 after 1 h of growth were used.
For determination of the initial attachment of cells under different conditions in the flow cell system, cells were allowed to adhere for only 30 min before resuming the flow. When necessary, DNase I or chromosomal meningococcal DNA was added to the inoculum.
Microscopy and image analysis
All microscopic observations and image acquisitions were performed using a Zeiss LSM510 confocal laser scanning microscope (Carl Zeiss), equipped with detectors and filter sets for monitoring GFP and SYTOX Orange. Images were obtained using a 40×/1.3 Plan-Neofluar oil objective lens and a 60×/1.3 water objective lens. Simulated three-dimensional images and sections through the biofilm were produced using the IMARIS software package (Bitplan). Structural parameters of biofilms were calculated with the COMSTAT computer program (Heydorn et al., 2000). Strains were cultivated in three discrete channels. From each channel, six image stacks were generated randomly along the centre of the channel.
Bitmap images (960 × 980 pixels each) of gfp labelled meningococci stained with SYTOX Orange were analysed with the package ‘pixmap’, version 0.4–10., created by R. Bivand, F. Leisch and M. Maechler for the program ‘R’, version 2.10. (R Foundation for Statistical Computing, Vienna, Austria). Pixels were mapped to four exclusive categories according to values in red and green channels, which ranged from 0 to 1: (i) cells covered with extracellular DNA (eDNA) (red > 0.3 and green > 0.3), (ii) dead cells (red > 0.3 and green ≤ 0.3), (iii) cells not covered with eDNA (red ≤ 0.3 and green > 0.3), and (iv) background (red ≤ 0.3 and green ≤ 0.3). The ratio of a/(a + b), reflecting proportionate area of cells covered with eDNA, was calculated for six images per strain. Differences between clonal complexes were analysed with the Mann–Whitney test; differences between strains within a clonal complex were assessed using analysis of variance (anova).
Live/dead staining and eDNA staining of the biofilms
Bacterial viability and amounts of eDNA within biofilms were determined using SYTOX Orange (Invitrogen). A 1 mM stock solution of SYTOX Orange in demethyl sulphoxide was diluted to a concentration of 5 µM in NDM and injected into the flow cell. Live cells, dead cells and eDNA-covered cells were visualized by CLSM.
Measurement of eDNA in planktonic cultures
Extracellular DNA levels in planktonic cultures were determined using the PicoGreen double stranded DNA (dsDNA) quantification reagent (Invitrogen) according to the manufacturer's instructions. In brief, planktonic culture samples were centrifuged at 13 000 r.p.m. for 3 min. The supernatants were put in new eppendorf tubes and were incubated with 1 volume of working solution. Finally, the sample was excited at 485 nm and the emission was measured at 535 nm using a Tecan GENios fluorescence microplate reader.
Static biofilm formation
Meningococcal biofilms under static conditions were initiated in 24-well plates, as described previously (Lappann et al., 2006). Meningococcal cells from a 12 h culture were adjusted to an OD600 of 0.4 in modified NDM with and without DNase I (Sigma-Aldrich). One millilitre of the resulting suspensions was seeded per well. After 1 h of incubation at 37°C, each well was washed with 1 ml de-ionized water and the adherent cells were determined with crystal violet as described previously (Lappann et al., 2006).
To evaluate the effects of cell passage in liquid medium on initial attachment, cells from a 16-hour-old agar plate were set to an OD600 of 0.1 in liquid medium and incubated at 37°C and 200 r.p.m. for 3 h. Then, the OD600 of the culture (passage 1) was determined and a new liquid culture was set to an OD600 of 0.1 with cells of passage 1 (passage 2). After another 3 h of growth and subsequent determination of OD600, cells from passage 2 were used to set a new liquid culture to OD600 of 0.1 (passage 3). Samples of passages 1 and 3 after 1 h of growth were used for measuring initial attachment in 24-well plates, as described above. When required, DNase I, crude DNA or highly purified meningococcal DNA was added to the samples of passage 3.
DNase I treatment of biofilms and addition of exogenous DNA to biofilms
A final concentration of 100 µg ml−1 DNase I (46 Kunitz units ml−1) was used in all static experiments. Owing to a constant supply of DNase I in the medium by the peristaltic flow, a lower concentration of 1 µg ml−1 DNase I was used for all flow cell biofilm experiments. DNase I was added at different at time points. When necessary, chromosomal meningococcal DNA was added to the pre-cultures to a final DNA concentration of 0.1 µg ml−1.
Extraction of crude meningococcal DNA
To obtain crude meningococcal DNA, cells of strain MC58 were adjusted to OD600 of 5.0 in NDM and incubated without shaking at room temperature for 48 h. Then the cells were gently resuspended and the cultures were centrifuged for 5 min at 13 000 r.p.m. The resulting supernatant was supplemented with NaCl to a final concentration of 0.25 M. Crude DNA was precipitated by addition of two volumes of ethanol and subsequent centrifugation. The precipitant was dissolved in fourfold diluted Tris/EDTA buffer. The DNA concentration was determined by ultraviolet spectrophotometry.
This study was supported by the Deutsche Forschungsgemeinschaft (Collaborative Research Centre 479, University of Würzburg, UV). We thank the Boehringer Ingelheim Fonds for travel support to M.L.