Bacterial strains used in this study are listed in Table 1 and Table S1. Neisserial strains were grown overnight on GC agar base medium (OXOID) supplemented with Oxoid™ Vitox at 37°C in a CO₂-enriched atmosphere provided by a candle jar. When appropriate, erythromycin (7 μg ml⁻¹), kanamycin (100 μg ml⁻¹), chloramphenicol (10 μg ml⁻¹), rifampicin (50 μg ml⁻¹) or gentamicin (60 μg ml⁻¹) was added to the medium. BB-1 and HB-1 are capsule-deficient derivatives of strains B16B6 (Frasch and Chapman, 1972) and H44/76 (Holten, 1979), respectively, which were isolated from patients with meningococcal disease in the United States and Norway, respectively. Strain α14 was isolated in Germany from a healthy carrier (Claus et al., 2005). A rifampicin-resistant derivative of HB-1ΔnalP was selected after plating the strain on GC plates supplemented with rifampicin.

For liquid cultures and biofilm experiments, bacteria grown on plates were resuspended in Tryptic Soy Broth (TSB) (Scharlau) or RPMI (Gibco) to an optical density at 550 nm (OD₅₅₀) of 0.1 and incubated in polystyrene cell culture flasks with shaking (120 rpm) until they reached the exponential growth phase (OD₅₅₀ of ~1.0). To induce the expression of genes from plasmids, 0.1 mM isopropyl-β-D-l-thiogalactopyranoside (IPTG) was added to the medium.

Escherichia coli strain DH5α was grown in lysogeny broth (LB) with shaking or on solid LB medium at 37°C. For plasmid selection, the following antibiotics were included in the medium: kanamycin (50 μg ml⁻¹), chloramphenicol (25 μg ml⁻¹) or gentamicin (20 μg ml⁻¹).

All plasmids and PCR primers used in this study are listed in Tables S1, S2, respectively. Regular PCR reactions were performed by using Dream Taq-DNA polymerase (Thermo Scientific), whilst PCR fragments generated for cloning were obtained using the High Fidelity polymerase (Roche Diagnostics GmbH) or Phusion DNA polymerase (Thermo Scientific). For purification of PCR products, the Wizard® SV Gel and PCR Clean-Up System (Promega) were used. Plasmids were isolated with the commercial E.Z.N.A.® Plasmid Mini Kit I (Omega Bio-Tek). PCR products and plasmids were both digested for 2 h with appropriate restriction enzymes (Thermo Scientific) for which cleavage sites were included in the primers, purified and ligated using T4 DNA ligase (5 U/μl) (Thermo Scientific).

For the preparation of pIN plasmids, plasmids pCRT_hrtA and phrtA_gm_rfp containing the hrtA region (Figure S1A) were used. The gfp gene together with an upstream region containing the lac promoter were amplified by PCR from plasmid mut3.1 and inserted into plasmid pCRT_hrtA via MluI and PpuMI digestion. Subsequently, a gentamicin-resistance cassette amplified by PCR from phrtA_gm_rfp was inserted via PpuMI, resulting in plasmid phrtA_gm_gfp. Fragments of different length (221–273 bp) located upstream of the opaB gene and containing the −35 and −10 boxes of the opaB promoter (opaBP) were amplified by PCR from Ng strain FA1090 using different primer pairs (Table S2) and used to generate three promoter variants: opaBP_M, opaBP_L, and opaBP_H (Figure S1C). The sequence of opaBP_M differs with respect to opaBP in four nucleotides creating an NheI site located between the ribosome-binding site (RBS) and the start codon (Figure S1C). This fragment and a gfp gene, amplified by PCR from plasmid phrtA_gm_gfp, were digested with NheI and ligated together. The resulting ligation product was purified and inserted into phrtA_gm_rfp via MluI and Van91I resulting in plasmid pIN_M (Figure S1B). The fragment amplified for opaBP_H was 18 nucleotides smaller than that for opaBP_M, replacing 52 nucleotides immediately upstream of the start codon by 34 nucleotides. This fragment was inserted into phrtA_gm_gfp and phrtA_gm_rfp upstream of the gene that encodes the fluorescent protein via MluI and SmaI. The resulting plasmids, pIN_H and pIN_H−RED, respectively, contain a different sequence between the -10 box and the start codon as compared with pIN_M (Figure S1C). The fragment amplified for opaBP_L was 3 nucleotides larger than opaBP, replacing 32 bp by 35 bp immediately upstream of the start codon, and this was cloned into phrtA_gm_gfp via MluI and SmaI. The resulting plasmid, pIN_L, contains a similar sequence between the -10 box and start codon as pIN_H but contains 21 additional bp in opaBP (Figure S1C). The correct insertion of the fragments was confirmed by PCR and subsequent sequencing of plasmids at the Macrogen sequencing service (Amsterdam).

Biofilms were formed under static conditions in 24-wells plates as previously described (Arenas et al., 2013a) with modifications. Briefly, 5-h old cultures in TSB were adjusted to an OD₅₅₀ of 1, and 500-μl samples were seeded per well on a round glass cover slip. For mixed biofilms, 500-μl samples were mixed 1:1, unless mentioned otherwise, and the mixture was placed in the well. After 15 h of incubation, the medium was removed from each well, and the biofilm was washed twice with de-ionized water. Biofilms were chemically fixed with 0.1 mM PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, and 1.8 mM KH₂PO₄) containing 2% formaldehyde for 2 h for microscopy analysis. In some experiments, bacteria were harvested from cultures by centrifugation (4,500 g for 5 min) and resuspended in the supernatant recovered from the culture of another strain before initiation of biofilm formation. To determine DNase sensitivity of biofilm formation, cultures were treated with 100 mg ml⁻¹ of DNase I and biofilm formation was determined using crystal violet as described (Arenas et al., 2013a).

Fixed, 15-h old biofilms were used for microscopy. All microscopic observations and image acquisitions were performed using a Zeiss LSM 700 confocal laser scanning microscope (Carl Zeiss, Germany) equipped with detectors and filter sets for monitoring fluorescence. Images were obtained using a 20x/0.8 Plan-Apochromat, a 40x/1.30 Plan-Neofluar oil or a 63x/1.40 Plan-Apochromat oil objective. Phenotypes were considered when at least observed in three independent experiments performed in duplicate. For the analysis of the structural parameters of the biofilm (biomass, average thickness, roughness coefficient, and surface to volume ratio), image stacks at 0.4 μm z-intervals were acquired and analyzed with the program COMSTAT (Heydorn et al., 2000) in the image processing environment ImageJ (v1.48, NIH, http://imagej.nih.gov/ij/).

To determine the level of expression of fluorescent proteins expressed by each pIN variant, cell were fixed by adding formaldehyde (1% v/v) to an exponentially growing culture as described (Arenas et al., 2008) and formalin-fixed cells were visualized by fluorescence microscopy. An outline was drawn around 20 formalin-fixed cells, and the mean fluorescence was measured, along with several adjacent background readings. Then, the total corrected cellular fluorescence (TCCF) = integrated density – (area of selected cell × mean fluorescence of background readings), was calculated for each bacterium. Statistical analyses were performed considering TCCF values obtained from each construct.

To determine twitching motility in biofilms, time-lapse videos of fluorescent bacteria in 0.5- and 2-h old biofilms of strains following an eDNA-dependent and -independent strategy, respectively, were recorded. Sixty images were taken at 0.1-ms intervals. For the HB-1ΔpilE mutant, biofilms were stained with a LIVE/DEAD BacLight Bacterial Viability Kit (Molecular Probes).

Liquid cultures grown to an OD₅₅₀ of ~2.0 were centrifuged (4,500 g for 5 min), and the resulting cell pellet was resuspended in H₂O to an OD₅₅₀ of 10. The spent media were centrifuged at 16,000 g for 15 min to remove residual cells, and proteins were precipitated from the supernatant with 10% (w/v) trichloroacetic acid in H₂O. After centrifugation (16,000 g, 15 min), the resulting pellets were washed with ice-cold acetone, air-dried and resuspended in H₂O. For analysis of protein production in biofilms, 15-h old biofilms were disrupted by mechanical forces, and bacterial cells were peletted (4,500 g for 5 min) and resuspended in H₂O to a final OD₅₅₀ of 10. For analysis of secreted proteins in biofilms, biofilms and planktonic cells were mixed by pipetting and the suspensions were centrifuged at 16,000 g for 15 min to remove cells and debris, and proteins were precipitated from the supernatants as described above. All samples were resuspended in double-strength sample buffer for electrophoresis and boiled for 10 min.

For SDS-PAGE, the Mini-PROTEAN® Electrophoresis System (Bio-Rad) was used. After electrophoresis, gels were stained with Coomassie Brilliant blue. Western blot analysis was performed as described (Arenas et al., 2015a). Blots were developed with SuperSignal® West Pico Chemiluminescent Substrate (Thermo Scientific) for 1 min at room temperature, and the image was acquired on a light-sensitive film (Fuji Medical X-Ray Film) or in a bio-imaging system (BioRad). The antiserum directed against GFP was purchased from Sigma-Aldrich. The antisera directed against the translocator domain of IgA protease (Roussel-Jazédé et al., 2014), the translocator domain of NalP (Oomen et al., 2004), and the TPS domain of TpsA of system 1 (van Ulsen et al., 2008) were previously described. The antiserum directed against MafA_MGI−3 is from our laboratory collection, and the monoclonal antibody SM1 directed against PilE was generously provided by John Heckels (University of Southampton).

Bacteria with different antibiotic resistance markers were used to discriminate between two strains (indicated in Table S1). The biofilm was washed twice with Hanks' Balanced Salt Solution Gibco® (ThermoScientific). To disrupt the aggregates, biofilms were incubated for 15 min at room temperature with a solution containing 0.5 mg ml⁻¹ of DNase I (Sigma-Aldrich) and 0.02 mg ml⁻¹ of proteinase K (ThermoScientific) diluted in TSB. Complete disruption of aggregates was confirmed by microscopy. The bacterial suspensions were serially diluted and plated on GC agar containing appropriate antibiotics, and the colony-forming units (CFU) were quantified after 12 h of incubation.

For statistical comparisons, data from at least three independent experiments performed in duplicate were considered. To determine the structural parameters of the biofilm, at least five image stacks of each sample were obtained from representative experiments. Data were analyzed using an unpaired statistical t-test with GRAPH PAD v 6.0 (Graph Pad Software, Inc.).

To facilitate the visualization of bacteria by fluorescence microscopy and the discrimination between different strains or species in mixed biofilms, we designed new plasmids to integrate genes encoding fluorescent proteins into the chromosome. A high rate of transformation (hrtA) region (Claus et al., 1998) present in the chromosome of several Neisseria species with >90% of sequence identity (Figure S1A) was exploited for this purpose. In a plasmid containing this region, we introduced (i) the promoter of the opaB gene (opaBP), which encodes an abundant protein expressed in some Neisseria species and previously used for protein production (Ramsey et al., 2012), (ii) a gene encoding green or red fluorescent protein (GFP or RFP, respectively), and (iii) a gentamicin-resistance cassette (Figure S1B). This plasmid was designated pIN (plasmid for Integration in Neisseria). Three different opaBP variants were tested in the pIN backbone (Figure S1C), and they were called opaBP_H, opaBP_M, and opaBP_L. The pIN variants, which were called pIN_H, pIN_M, and pIN_L, respectively, were used to transform Nm strain HB-1 selecting for gentamicin-resistant recombinants. To test protein expression levels, the intensity of fluorescence emission was acquired for individual bacteria and analyzed with the image processing program ImageJ. The results revealed large differences in fluorescence intensity between the promoter variants in the order opaBP_H > opaBP_M > opaBP_L, in accordance with protein production levels detected on Western blots (Figure S1D). The fluorescence of pIN_H-derived transformants was visualized in all our imaging devices and allowed for the discrimination of fluorescent strains in mixed biofilms. Therefore, this plasmid was used to generate fluorescent bacteria. Indeed, we could generate fluorescent bacteria in strains of Nm, Nl, and Ng transformed with this plasmid (Figure S1E).

To study the biofilm structure of various Nm and Nl strains, biofilms of fluorescent bacteria were formed on glass and visualized by confocal microscopy. All strains used were capsule deficient, as capsule has been reported to inhibit biofilm formation on abiotic surfaces (Yi et al., 2004; Lappann et al., 2006). Figure 1A shows the structures of 15-h old biofilms of various strains. Biofilms consisted of cell clusters, but the size, dispersion and number of the clusters differed largely between both species and between strains of the same species. Nm strains generally formed smaller clusters than did the Nl strains. Also, Nm strains of cc8 and cc11 formed much smaller and less compact clusters with, together, a larger coverage of the substratum than strains of cc32 and cc53 (Figure 1A). It is noteworthy that strains of cc8 and cc11 use an eDNA-independent strategy of biofilm formation in contrast to strains of other cc (Lappann et al., 2010). Both Nl strains formed biofilms that were sensitive to DNase I (data not shown). Thus, these results confirmed that aggregation is a common feature during neisserial biofilm formation.

AutA and type IV pili are known to be involved in bacterial aggregation during biofilm formation (Lappann et al., 2006; Arenas et al., 2015a). Here, we explored the contribution in this process of eDNA-binding proteins, which are cleaved from the cell surface by the protease NalP. Figure 1B shows the biofilm structure of mutants of Nm strains HB-1 and BB-1 lacking nalP. The nalP mutants formed bigger and more compact microcolonies than the corresponding wild types (Figure 1A), although the difference was less pronounced in BB-1, in accordance with the eDNA-independent strategy of biofilm formation of this strain. The stronger aggregation of HB-1ΔnalP is not due to increased piliation, as Western blotting assays showed a similar production of PilE, the major pilus subunit, in the wild type and the nalP mutant (Figure S2A). Since cc11 strain BB-1 produces a different type of pilin that is not recognized by the antibodies (Cehovin et al., 2010) we could not verify pilE expression in this strain and its nalP mutant. The autA gene is disrupted in HB-1 and BB-1 because of the presence of a premature stop codon (Arenas et al., 2015a) and can, therefore, not play a role in the differences between the NalP-producing and non-producing strains. Furthermore, aggregation was severely reduced when NalP was expressed in trans from plasmid pEN300 in the nalP mutants (Figure 1B) demonstrating that the increased aggregation of the nalP mutants is a direct effect of the lack of NalP synthesis. Microscopic examination of log-phase precultures grown under shaking conditions showed the presence of only few very small aggregates in the nalP mutant of HB-1 but not in the wild type (Figure S2B). The size of these aggregates does not match those observed in biofilms (Figure S2B). Probably, the abundance of eDNA at the surface of the nalP mutant cells facilitates aggregation and thereby microcolony formation. Such interactions may occur already in liquid cultures, but they are disrupted by physical forces during shaking. In conclusion, these results show that microcolony formation occurs during biofilm formation and that nalP expression influences this process, presumably by cleaving eDNA-binding proteins from the cell surface. Thus, these data expand the previously established role of NalP during the initiation of biofilm formation by demonstrating its effect on biofilm structuring.

The high rates of Nm colonization (Sim et al., 2000) suggest an intense traffic of strains within the nasopharynx. Consequently, strains could interact with each other to compete or to benefit during colonization. To study how different strains affect each other during biofilm formation, we analyzed pairwise combinations of three Nm strains in biofilm experiments. We selected strains HB-1, BB-1, and α14 because different traits relevant for biofilm formation. HB-1 and BB-1 were chosen as representatives of strains following an eDNA-dependent and -independent strategy of biofilm formation, respectively (Arenas et al., 2013a). Strain α14 was chosen because it produces AutA, which causes autoaggregation and thereby affects biofilm architecture (Arenas et al., 2015a). HB-1 and BB-1 do not produce AutA (Arenas et al., 2015a). Other relevant characteristics of these strains are listed in Table 1.

First, several relevant properties of these strains were further analyzed. HB-1 and BB-1 clearly showed twitching motility in biofilms (Videos 1, 2, respectively, in Supplementary Material). Strain α14, however, showed no twitching motility, similar as a pilE mutant of HB-1 (Videos 3, 4, respectively, in Supplementary Material). Interestingly, Western blot analysis revealed a considerable difference in the electrophoretic mobility of the PilE proteins of HB-1 and α14 (Figure S2A), even though the sequences of these proteins are identical according to the available genome sequences (Schoen et al., 2008; Piet et al., 2011). Additionally, the genome sequence of α14 showed that nalP is out of phase; Western blotting confirmed that NalP is not synthesized in this strain (Figure S2A). All relevant properties of the three strains are listed in Table 1.

Each strain was grown independently in TSB, and, after adjusting them to the same OD, they were mixed 1:1 for biofilm formation. First, green- and red-fluorescent variants of strain HB-1 were combined (Figure 2A). Both variants of the strain formed separate clusters with hardly any intermixing, although these separate clusters extensively interacted (Figure 2A). A similar result was observed when green- and red-fluorescent variants of α14 were mixed (Figure 2A). Presumably, these separate clusters result from clonal outgrowth. In contrast, small intermixed clusters were abundantly found when green- and red-fluorescent variants of BB-1 were combined (Figure 2A), suggesting that the less aggregative nature of this strain allows for intermixing.

Next, strains of different genetic backgrounds were mixed. Interestingly, strain BB-1, which forms only very small clusters dispersed over the substratum in single-strain biofilms (Figure 1A), formed large clusters, which coincided with those of α14, in dual-strain biofilms (Figure 2B), indicating that these strains can form intermixed microcolonies. In contrast, HB-1 and α14 formed separate but interacting clusters (Figure 2C), whilst strains BB-1 and HB-1 attached to the substratum without much apparent association between the clusters formed (Figure 2D). Lack of intermixing without much association between clusters was also observed in most cases when Nm strains were combined with Nl strains (Figure S3). Only associations of clusters of α14 with those of the Nl strains were observed.

Next, we studied the possible role of AutA in the formation of intermixed clusters of α14 and BB-1. Indeed, intermixing was drastically reduced when an autA mutant of α14 was used, and separate but interacting clusters of the two strains were observed (Figure 2B). Unfortunately, the introduction of plasmid pFPAutA in the autA mutant of α14 for complementation studies failed for unknown reasons (Arenas et al., 2015a). To further investigate the possible role of AutA in the formation of intermixed clusters with BB-1, the AutA-encoding plasmid was introduced in another genetic background, i.e., HB-1. The synthesis of AutA in HB-1, which was confirmed on Western blots (Figure S4A), increased the interaction of its clusters with those of BB-1, but it did not lead to intermixed clusters (Figure 2D). Thus, additional traits of α14, such as the absence of NalP or of twitching motility (Table 1), may contribute to its ability to form intermixed clusters with BB-1.

Considering that NalP inactivation increased aggregation in single-strain biofilms (Figure 1B), we asked whether the presence or absence of NalP may also affect the interaction between different strains in mixed biofilms. Combinations of strains HB-1 and BB-1 with their respective nalP mutant derivatives generated biofilms consisting of separate clusters of the individual strains, which, however, extensively interacted with those of the other strain as evidenced by their co-localization on the substratum (Figure 2D). The ectopic expression of nalP from pEN300 in both nalP mutants drastically reduced these interactions (Figure 2D). Furthermore, the combination of BB-1 with a derivative of HB-1 lacking, besides nalP, also the nhbA and iga genes showed only small clusters of the HB-1 derivative that did not interact with those of BB-1 (Figure 2D), confirming that the increased aggregation observed for the nalP mutant is mediated by the increased cell-surface exposure of intact NHBA and αP in this strain. Interestingly, inactivation of nalP in strain BB-1 interfered with the formation of intermixed microcolonies with α14 and led to the formation of separate clusters of the individual strains, which barely interacted (Figure 2B). Probably, in this case, the increased interaction between the BB-1 cells due to the inactivation of NalP prevents intermixing with the α14 cells. In conclusion, Nm strains can form microcolonies constituted of mixed lineages or separate clusters of individual strains that variably interact depending on the strains studied and on the synthesis of AutA and/or NalP.

To gain more insight into the implication of AutA and NalP synthesis on biofilm architecture, the individual contribution of each strain in mixed biofilms was analyzed with COMSTAT software and statistically compared with results of single-strain biofilms. Single-strain biofilms of HB-1, BB-1, and α14 differed considerably; the biomass, that reflects the amount of both live and dead bacteria in the biofilms, was in the order α14 > HB-1 > BB-1 (Figure 3A). Also other biofilm parameters, such as thickness, surface/volume ratio, and roughness, differed (Figure S5). To test bacterial viability, the biofilms were disrupted and the numbers of CFU were determined on selective GC plates. Viability was determined at 15, 24, and 48 h after initiation of biofilm formation and found to decrease drastically for HB-1 and α14 at 48 h (data not shown). To better compare the differences in viability between wild types and mutants of the three strains, results are shown for 24-h old biofilms (Figure 3B), although similar results were observed for 15-h old biofilms. In single-strain biofilms, the numbers of CFU of strains HB-1 and BB-1 did not differ but they were significantly different to those of α14 (Figures 3B).

Single-strain biofilms of the nalP mutants of BB-1 and HB-1 revealed higher biomasses compared to the corresponding wild types, but the numbers of CFU were similar or drastically decreased, respectively (Figures 3A,B). We did not find differences in viability between the wild types and the nalP mutants in liquid cultures (data not shown). The discrepancy between biomass and numbers of CFU indicates the presence of more dead cells in the biofilms of the nalP mutants and could be explained by the more compact clusters formed in the absence of NalP (Figure 1), which could result in lower exposure of the biomass to the nutrients. In summary, the pronounced aggregation observed in nalP mutants (Figure 1B) correlates with increased biomass and altered biofilm architecture but reduced bacterial viability.

In dual-strain biofilms, the biofilm biomass and structure of HB-1 remained almost unaltered when in consortium with BB-1 or BB-1ΔnalP (Figure 3A and Figure S5 for additional biofilm details). In contrast, the phenotypes of the HB-1ΔnalP mutant, such as increased biomass and reduced viability in biofilms, appeared to be complemented when this strain was grown with BB-1, but barely with BB-1ΔnalP (Figures 3A,B and Figure S5). This complementation was not caused by the cleavage of the αP from the cell surface of HB-1ΔnalP by NalP secreted by BB-1 as revealed by Western blotting (Figure S4B). Consistently, the phenotype of the HB-1ΔnalP mutant was not complemented when biofilms were formed in culture supernatants from BB-1. Apparently, the interaction of BB-1 with clusters of HB-1ΔnalP (Figure 2D) had an impact on the architecture of HB-1ΔnalP biofilms (Figure 3A and Figure S5) and on the survival of the HB-1ΔnalP bacteria within these biofilms (Figure 3B).

The biomass and viability of BB-1 in biofilms increased significantly when it was grown in a consortium with HB-1 or HB-1ΔnalP (Figures 3A,B). This shows that BB-1 benefits when mixed with HB-1, but BB-1 benefited less in consortium with HB-1ΔnalP than with HB-1. The biomass and viability of BB-1ΔnalP were barely affected when in consortium with HB-1 or HB-1ΔnalP (Figures 3A,B).

Biofilms of strain α14 had a lower biomass than those of its autA mutant derivative (Figure 3A). However, the number of CFU of the wild type in biofilms appeared higher (Figure 3B). In consortium with BB-1, the biomass and number of CFU of α14 and its ΔautA mutant derivative were similar or decreased (Figures 3A,B). The biofilm biomass and CFU of BB-1 were increased in consortium with α14 or α14ΔautA (Figure 3A) and also other biofilm parameters were altered (Figure S5). Together with the results of BB-1 grown in consortium with HB-1 or its nalP mutant derivative, these data indicate that biofilm formation of BB-1 increases when it is grown in consortium with other strains, and the degree of the increase is influenced by the synthesis of NalP or AutA in the partner strain.

Although Nm strains synthesize a variety of polymorphic toxins against congeners (Arenas et al., 2013b, 2015b; Jamet et al., 2015), the results in Figure 3B do not provide evidence of inter-bacterial competition in dual-strain biofilms. The expression of toxins in the biofilms was therefore evaluated in Western blotting assays. TpsA1 was detected in considerably higher amounts in the medium of disrupted biofilms than in liquid cultures grown under shaking conditions (Figure S4C). Large differences in the amounts of TpsA1 between the strains were detected in the order BB-1 > HB-1 > α14 (Figure S4C). As we have no suitable antisera available directed against MafB proteins, we had recourse to an antiserum directed against the MafA encoded by the Maf Genomic Island (MGI) 3, where the corresponding gene is located in an operon with a mafB gene (Jamet et al., 2015). This antiserum detected the MafA_MGI−3 protein in biofilms of BB-1 and α14 (Figure S4D), indicating that also MafB_MGI−3 is expressed in biofilms of these strains. MafA_MGI−3 was not detected in HB-1, where the gene is disrupted by a premature stop codon. Notably, MafA_MGI−3 was produced in higher abundance in biofilms than in liquid cultures (Figure S4D). In conclusion, the apparent absence of competition in the biofilm assays (Figure 3B) cannot be explained by the lack of expression of TpsA or MafB proteins.

Changing the ratio of HB-1 and BB-1 in dual-strain biofilms to 1:0.2 or 0.2:1, did not result in the suppression of the growth of either strain in the consortium (data not shown). Considering that competition between strains is particularly important under nutrient-limiting conditions, we also studied biofilm formation in RPMI medium, a synthetic medium that contains a low concentration of nutrient metals (Stork et al., 2010). In single-strain biofilms, the autoaggregative character of HB-1 and α14 already observed in TSB medium (Figure 1A) was retained (Figure 4A), while the biofilm biomass was reduced ~3-fold (compare Figure 3A and Figure 4B). Biofilms of BB-1 were dispersed on the substratum and showed less compact clusters as compared with those formed in TSB medium. Remarkably, the biofilm biomass of BB-1 in the nutrient-limited RPMI medium was increased ~4-fold relative to that in TSB (compare Figure 3A and Figure 4B). The three strains revealed a similar biomass in single-strain biofilms (Figure 4B), but the number of CFU in 24-h old biofilms varied by orders of magnitude in order α14 > HB-1 > BB-1 (Figure 4B). In dual strain biofilms, BB-1 formed compact clusters that interacted with those of α14 or HB-1 (Figure 4C). The biofilm biomass and CFU of BB-1 increased when in combination with HB-1 and, to a lesser extent, with α14 (Figure 4B). The number of CFU of HB-1 increased significantly in consortium with BB-1, whilst the CFU of α14 remained unaltered. In conclusion, under low-nutrient conditions, the biofilm architecture of the strains and the interactions with other strains may deviate from those in TSB, but we did not find evidence for competition.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.