Identification of Veillonella parvula and Streptococcus gordonii adhesins mediating co-aggregation and its impact on physiology and mixed biofilm structure

ABSTRACT

The dental plaque is a polymicrobial community where biofilm formation and co-aggregation, the ability to bind to other bacteria, play a major role in the construction of an organized consortium. One of its prominent members is the anaerobic diderm Veillonella parvula, considered a bridging species, which growth depends on lactate produced by oral streptococci. Understanding how V. parvula co-aggregates and the impact of aggregation has long been hampered due to the lack of appropriate genetic tools. Here we studied co-aggregation of the naturally competent strain V. parvula SKV38 with various oral bacteria and its effect on cell physiology. We show that V. parvula requires different trimeric autotransporters of the type V secretion system to adhere to oral streptococci and actinomyces. In addition, we describe a novel adhesin of Streptococcus gordonii, VisA (SGO_2004), as the protein responsible for co-aggregation with V. parvula. Finally, we show that co-aggregation does not impact cell-cell communication, which is mainly driven by environmental sensing, but plays an important role in the architecture and species distribution within the biofilm.

IMPORTANCE

Our research explores the mechanisms of bacterial adhesion within the dental plaque, focusing on Veillonella parvula, a key player in the oral microbiome. Dependent on lactate from streptococci, V. parvula plays a crucial bridging role in the formation of dental biofilms by co-aggregating with other bacteria. Despite its importance, the understanding of the underlying mechanisms of co-aggregation remains limited. Our study shows that V. parvula uses different trimeric autotransporters to adhere to oral Streptococci and Actinomyces. We additionally identify a novel adhesin from S. gordonii, VisA (SGO_2004) facilitating this interaction. We found that although co-aggregation does not affect cell-cell communication, it is critical for biofilm structure and species distribution. This research opens up new avenues for exploring microbial interactions in dental health and diseases.

INTRODUCTION

Bacterial attachment to other bacteria is a key step in the formation of bacterial biofilm. This adhesion is termed auto-aggregation when the adhesion occurs with genetically identical bacteria and co-aggregation when different species or strains are involved. While auto-aggregation is known to enhance stress resistance, antibiotic tolerance, and virulence (1), the specific role of co-aggregation remains largely understudied (2), except in the contexts of dental plaque and certain aquatic environments (3–5).

The dental plaque is an important polymicrobial biofilm whose perturbation can lead to the development of caries and periodontitis (6, 7). The formation of the dental plaque is a stepwise process that begins with the adhesion to the teeth surface of early colonizers comprised of oral streptococci, including Streptococcus gordonii, S. oralis and S. mitis and Actinomyces spp. Then, bridging species such as Veillonella and Fusobacterium co-aggregate with the early colonizers forming an adhesion substrate for late biofilm commensal colonizers but also the opportunistic pathogens Porphyromonas gingivalis, Treponema denticola, and Tannerella forsythia (3). Co-aggregation is mostly driven by adhesins (8–13), few of which have been identified, including P. gingivalis major and minor fimbriae (14, 15), which interacts with S. gordonii SspB adhesin and GADPH, and the F. nucleatum autotransporters RadD and Fap2 (10, 12, 16). However, most of the molecular actors of oral biofilm co-aggregation mechanisms are currently unknown.

Veillonella are strict anaerobic diderm firmicutes and seven Veillonella species can be found in the dental plaque (17) where they rely on lactate produced by oral streptococci as a carbon source (18). Oral Veillonella species possess extensive aggregative properties contributing to their colonization of the oral environment (8) in which the physical proximity resulting from aggregation with their different partners likely facilitates their metabolic integration in the oral biofilm. For instance, V. parvula (formerly V. atypica) strain PK1910 induces the expression of the S. gordonii amylase amyB in a distance-dependent manner, possibly to increase lactic acid production, which it preferentially uses as a carbon source (19, 20). V. atypica was also shown to produce a catalase protecting F. nucleatum from reactive oxygen species produced by S. gordonii (21).

While Veillonella adhesive properties were first characterized more than 30 years ago (22, 23), the underlying molecular actors of co-aggregation and its physiological consequences remained elusive until recently. Indeed, it was recently shown that V. atypica OK5 possesses eight trimeric autotransporter adhesins (TAA) belonging to the type Vc secretion system family (24). One of them, Hag1, mediates adhesion to oral bacteria and buccal cells (25). On the other side, several oral Veillonella species, including V. atypica OK5, co-aggregate with S. gordonii in a Hsa adhesin-dependent manner (26). However, a more extensive mechanistic characterization of the Veillonella adhesin repertoire was hampered due to the lack of genetic tools described for this genus. V. parvula strain SKV38 is a recently described naturally competent isolate that is readily genetically engineered (27). We have recently shown that it possesses nine TAAs, named VtaA to -I, and 3 classical monomeric autotransporters, named VmaA to -C. Both VtaA and a gene cluster coding for 8 TAA adhesins were shown to be important for surface adhesion and biofilm formation (27). In addition, VtaA mediates the auto-aggregation of V. parvula (27).

Here, we investigated the capacity of V. parvula SKV38 to co-aggregate with common oral bacteria and studied the physiological impact of this co-aggregation. We found that, in addition to mediating auto-aggregation, VtaA is also involved in co-aggregation with S. oralis while two other adhesins encoded in an adhesin cluster, VtaE and VtaD, are involved in co-aggregation with S. gordonii and Actinomyces oris. We also identified a novel adhesin of S. gordonii, VisA (SGO_2004), as the possible direct or indirect interacting partner of V. parvula VtaE/VtaD. Analysis of the transcriptomic profiles of both bacteria in coculture with or without aggregation suggested a very limited impact of aggregation on gene expression. Furthermore, we showed that the absence of co-aggregation results in spatial segregation of the two species biofilms, suggesting that co-aggregation would be necessary to generate the architecture of a fully functional and healthy dental plaque biofilm (28, 29). In conclusion, this study contributes to provide a better mechanistic understanding of co-aggregation between oral bacteria, one of the key organization principles driving dental plaque formation.

RESULTS

V. parvula uses specific adhesins to interact with S. oralis, S. gordonii, and A. oris

To identify potential ligands of V. parvula SKV38 adhesins, we used our model V. parvula SKV38 strain to perform co-aggregation assays with different bacterial members of the dental plaque. The assays were performed with independent cultures grown in an SK medium in which V. parvula auto-aggregation does not occur (Fig. S1). V. parvula SKV38 co-aggregated with several Streptococcus gordonii strains, Streptococcus oralis ATCC10557, and Actinomyces oris CIP102340. It did not, however, co-aggregate with Streptococcus mitis CIP 104996, Streptococcus parasanguinis CIP104372T, Fusobacterium nucleatum ATCC 25586, and Streptococcus mutans NG8, UA159, CBSm8, and CBSm38 and only very weakly with S. mutans UA140 (Fig. 1A; Fig. S2). We decided to further investigate the determinant of co-aggregation between V. parvula SKV38 and S. oralis ATCC 10557, S. gordonii DL1, and A. oris CIP102340. To identify which of the 12 V. parvula adhesins were involved in the co-aggregation with these different partners, we used previously constructed single deletion mutants of each of these adhesins (27) and performed co-aggregation assays by mixing independent cultures of each of the three tested oral bacterial strains and the 12 V. parvula adhesin mutants in aggregation buffer. Deletion of V. parvula trimeric autotransporter VtaA abolished co-aggregation with S. oralis, while deletion of the trimeric autotransporter VtaE abolished co-aggregation with S. gordonii and strongly reduced co-aggregation with A. oris (Fig. 1B through E; Fig. S3). A double mutant lacking both VtaA and VtaE showed reduced co-aggregation with A. oris compared to a ΔvtaE single mutant, suggesting that VtaA is a secondary adhesin involved in the co-aggregation with A. oris (Fig. 1G). Microscopy observation of V. parvula incubated with S. oralis, S. gordonii, and A. oris confirmed the observed co-aggregation phenotypes (Fig. S3). Moreover, the use of P_Tet-vtaA or P_Tet-vtaE constructs, in which the intact chromosomal vtaA and vtaE open reading frames and transcriptional terminators are placed under the control of an aTc inducible promoter, allowed us to recapitulate the aggregative phenotype in an aTc-dependent manner (Fig. 1C through E). Together with the phenotypes of the ∆vtaA and ∆vtaE mutants, this indicates that complementation of the phenotypes in the P_Tet-vtaA or P_Tet-vtaE constructs is likely not due to transcriptional readthrough. Both the P_Tet-vtaA and the P_Tet-vtaE strains partially co-aggregated with S. oralis and A. oris, even in the absence of aTc, suggesting a leakage of the used P_Tet promoter. While deletion of vtaE completely abolished co-aggregation with S. gordonii when mixed after independent growth, it only partially abrogated co-aggregation with S. gordonii when cocultured overnight (Fig. 1F), suggesting that another V. parvula adhesin could contribute to co-aggregation. Consistently, we identified VtaD as being this secondary adhesin since any residual co-aggregation between S. gordonii and V. parvula disappeared in the ∆vtaCDEF and ∆vtaDE mutants (Fig. 1F). vtaD is the gene located immediately upstream of vtaE and VtaD has a high similarity to VtaE (81%), which may explain why both corresponding proteins possess similar binding activities. However, vtaD encodes a shorter adhesin than VtaE (2,071 residues as opposed to 3,141 residues), mostly lacking part of the repetitive sequences found in VtaE stalk (Fig. S4 and S5). Interestingly, deletions of vtaC or vtaF in the ∆vtaE background increased the aggregative phenotype of V. parvula with S. gordonii (Fig. 1F) suggesting that these other adhesins may interfere with the VtaD-dependent co-aggregation process.

Fig 1.

VtaA and VtaE are the adhesins responsible for co-aggregation with S. oralis, S. gordonii, and A. oris. (A) Co-aggregation of independent cultures of both V. parvula SKV38 and various members of the dental plaque after 7 h, as measured by the % of the decrease in optical density between 0 and 7 h. SD and single points for 3–5 replicates are shown. See Fig. S2 for auto-aggregation of each strain. (B) Aggregation of V. parvula SKV38 WT and each single autotransporter mutant with S. oralis ATCC 10557, S. gordonii DL1, and A. oris CIP102340 after 7 h. SD and single points for three replicates are shown. The indicated P-values were calculated by comparing all conditions to the partner +Vp WT using a Brown-Forsythe and Welch ANOVA followed by Dunnett correction. (C–E and G) Co-aggregation curves of V. parvula WT, ∆vtaA, ∆vtaE, ∆vtaE∆vtaA, and P_Tet -vtaE or P_Tet -vtaA with 0, 100, or 250 ng/µL of aTc. Curves represent the mean and SD of 6–17 replicates. In panels A–E and G, 0% of initial OD corresponds to full auto or co-aggregation while 100% of initial OD corresponds to no auto or co-aggregation. (F) Representative pictures of co-aggregates after coculture between S. gordonii WT and V. parvula WT and different adhesin mutants; red arrow bars indicate the relative size of the aggregated fraction.

The Hag1 trimeric autotransporter is involved in the adhesion of V. atypica OK5 to human oral epithelial cells (25). Interestingly, the genes encoding VtaA and Hag1 are located at the same locus on the genome of V. parvula SKV38 and V. atypica OK5, respectively, with the difference that Hag1 is preceded by another trimeric adhesin. Comparison of this locus among different Veillonella revealed that this locus always contains adhesins, although the number of adhesins and their identity differs between strains, even within the same species (Fig. S6). This feature is reminiscent of V. parvula SKV38 (27) cluster of adhesins that is also present in a locus that consistently hosts diverse adhesins across Veillonella species.

Apart from its importance in dental plaque, V. parvula is also present throughout the gastrointestinal tract. We wondered whether some of its adhesins are involved in adhesion to oral or intestinal cells, rather than other bacteria. In contrast to the described strong interaction between V. atypica and host cells (25), we observed only a moderate adhesion of V. parvula SKV38 to TR146 oral and Caco-2 intestinal epithelial cells using microscopy (Fig. S7A through C). We then tested whether the major adhesins of V. parvula were involved in this interaction using a ∆vtaCDEF∆vtaA mutant. Deletion of the large adhesin group did not reduce adhesion to either cell type. Finally, we examined whether the other adhesins of V. parvula SKV38 could impact Caco-2 cell adhesion and showed that there were no significant differences in adhesion (Fig. S7D).

Identification of VisA (SGO_2004), a new S. gordonii adhesin mediating co-aggregation with V. parvula

To further characterize the molecular actors of co-aggregation, we focused on the pair V. parvula/S. gordonii and took advantage of a recently published collection of 27 S. gordonii DL1 surface proteins deletion mutants (30), corresponding to all 26 LPXTG cell wall anchor domain-containing proteins plus two mutants of the amylase-binding protein A (AbpA) and B (AbpB). We first investigated co-aggregation between wild-type V. parvula and all S. gordonii mutants and we identified two mutants, ∆padA (SGO_2005) and ∆SGO_2004, presenting either a reduced (∆padA) or total loss of co-aggregation (∆SGO_2004) with V. parvula (Fig. 2A and B).

Fig 2.

VisA (SGO_2004) is a novel adhesin interacting with V. parvula. (A) Co-aggregation of independent cultures of V. parvula SKV38 and S. gordonii DL1 WT and mutants for each LPXTG-containing protein and AbpA-B, as measured by the % of decrease in optical density between 0 and 7 h. SD and single points for 4–5 replicates are shown. The indicated P-values were calculated by comparing all conditions to the partner + Vp WT using a Brown-Forsythe and Welch ANOVA followed by Dunnett correction. Co-aggregation curves of S. gordonii WT, ∆visA, ∆padA (B) and P_Tet-visA or P_Tet-padA (C) with or without 250 ng aTc. Curves represent the mean and SD of 6–13 replicates. In panels A, B, and C, 0% of initial OD corresponds to full co-aggregation while 100% of initial OD corresponds to no co-aggregation. (D) Genetic organization of the SGO_2004/2005 locus. VWF_A: Von Willbrand factor A (IPR002035), Fim_isopep_form_D2: Fimbrial isopeptide formation D2 domain (IPR026466), G5 domain (IPR011098). (E) AlphaFold structural model of VisA without the signal peptide.

padA and SGO_2004 are part of an operon (Fig. 2D) and the observed loss of aggregation in the ∆padA mutant could be due to a polar effect on the downstream SGO_2004 gene (31). To test for this hypothesis, we inserted a P_Tet inducible promoter with the pVeg RBS (32) upstream of SGO_2004 while retaining or deleting the padA gene. In both cases, co-aggregation was fully recovered in the presence of aTc (Fig. 2C), demonstrating that SGO_2004 alone is the protein responsible for S. gordonii co-aggregation with V. parvula. SGO_2004 is a gene of previously unknown function coding for an 807 amino acid protein composed of a flexible chain of disordered/poorly predicted three short alpha helixes, 7 G5-domains, and an LPXTG domain (Fig. 2D and E). Homologs of this protein are found in other, sometimes distant, streptococci, next to a padA homolog (Fig. S8). Considering its newly identified role, we renamed this new aggregation-mediating adhesin VisA, for Veillonella Interacting Streptococcal protein A.

S. gordonii VisA directly interacts with V. parvula

To determine whether co-aggregation is mediated by a direct interaction of S. gordonii VisA with V. parvula surface, we purified the VisA region containing its 7 G5 domains (residues 138–698 with a C-terminal His-tag, see Fig. 2D) in E. coli and used the purified protein to assess potential direct interactions with V. parvula. When used at a concentration above 1 µg/ml, VisA_G5 was sufficient to induce aggregation of V. parvula on its own (Fig. 3A). Confirming our previous observations, a V. parvula ∆vtaE mutant retained a partial auto-aggregation phenotype, while a ∆vtaDE mutant did not, and ∆vtaCE and ∆vtaEF mutants displayed an intermediate phenotype (Fig. 3B). Moreover, immunofluorescence using an anti-His antibody detecting VisA_G5 incubated with V. parvula WT, ∆vtaE or ∆vtaDE showed that while VisA_G5 could be detected at the surface of V. parvula WT (Fig. 3C) and ∆vtaE (Fig. 3D), no signal could be seen for the ∆vtaDE mutant (Fig. 3E). Altogether, these results showed that VisA binds directly to V. parvula surface in a VtaE/VtaD-dependent manner.

Fig 3.

VisA binds directly to V. parvula in a VtaE- and VtaD-dependent manner. (A) Auto-aggregation curves of V. parvula SKV38 with various concentrations of VisA_G5. (B) Aggregation curve of V. parvula SKV38 or indicated adhesin mutants with 4 µg/mL of VisA_G5. For panels A and B, curves represent the mean and SD of three replicates and 0% of initial OD corresponds to full auto-aggregation while 100% of initial OD corresponds to no auto-aggregation. (C–E) Brightfield images and their corresponding immunofluorescence images targeting the His-tag of VisA_G5 after incubation of Vp WT, ∆vtaE, and ∆vtaDE with 10 µg/mL of VisA_G5 protein. The scale bar is 15 µm. The right image in panel C represents an enlargement of the WT + VisA_G5 immunofluorescence image (indicated by the white square), and the scale bar is 5 µm.

Co-aggregation in co-culture produces no significant alteration on the transcriptomic profiles of V. parvula and S. gordonii

While previous studies have compared the transcriptional responses of Veillonella and S. gordonii co-incubations to mono-incubation (19, 20, 33), they did not specifically evaluate the potential contribution of co-aggregation. Having identified the adhesins involved in V. parvula/S. gordonii co-aggregation, we set out to compare the transcriptional responses of these two strains in mono- and cocultures with and without co-aggregation or auto-aggregation. Here we used the rich medium BHIP (BHI +100 mM pyruvate) where V. parvula and S. gordonii wild types and mutants were co-cultured. In this medium, both V. parvula auto-aggregation and V. parvula/S. gordonii co-aggregation can be observed, and V. parvula can grow independently of the lactate produced by S. gordonii.

V. parvula transcriptional profiles of each condition were grouped mainly by the presence of S. gordonii and then by their strain type. In principal component analysis (PCA), calculated using normalized transcript counts, samples were strongly separated on the first principal component by their coculture status, thus indicating that the main determinant of the observed V. parvula response is the presence of its bacterial partner S. gordonii (Fig. 4A). The PCA on the second and third axis revealed a clustering by V. parvula mutant (Fig. S9), suggesting that the residual differences between conditions are associated with the nature of the V. parvula mutants. We could not identify specific signals due to co-aggregation in the V. parvula transcriptome analysis PCA analysis.

Fig 4.

Transcriptomic response of V. parvula to S. gordonii is mostly related to coculture. (A) PCA of all V. parvula samples (four biological replicates for 10 conditions). Colors and shapes represent the different conditions. The two circles separate monoculture samples from coculture samples. Green symbols indicate samples able to auto-aggregate in coculture, the blue shades samples unable to auto-aggregate. (B) Upset plot (a Venn diagram alternative) showing the number of differentially expressed genes (defined by an absolute log2fold change >1) shared for each condition compared to V. parvula WT monoculture. The green bar indicates the core response to coculture, the orange bar is the core answer to coculture without any aggregation, and the blue bar is the response to any aggregation in coculture. (C) Volcano plot of the coculture of V. parvula and S. gordonii WT compared to V. parvula in monoculture. Genes corresponding to identified key functions are differentially colored. (D) Upset plot for each condition compared to V. parvula ∆vtaA∆vtaDE and S. gordonii coculture, the blue bar shows the response to any aggregation in coculture.

To identify potential coculture-specific response, we searched for genes upregulated or downregulated (log2Fold above 1 or below −1) in at least one condition compared to V. parvula WT monocultures. The resulting Upset plot (Fig. 4B) represents the common differentially regulated genes for different combinations of conditions. This plot shows that the core V. parvula coculture transcriptomic response in all conditions was composed of 68 genes (Fig. 4B green bar and Data S1). The most upregulated gene was FNLLGLLA_00352 (around 4.5 log2Fold increase compared to the monoculture), coding for an uncharacterized major facilitator superfamily-type (MFS) transporter, an inner membrane transporter of an unknown small molecule. We also found a strong upregulation of genes coding for enzymes of the histidine and arginine biosynthesis pathways (Fig. 4C). Interestingly, vtaB, encoding an uncharacterized trimeric autotransporter, and a gene cluster encoding a prophage was also induced, albeit at lower levels. Many genes associated with stress response were slightly upregulated (genes coding for the chaperones GroEL and GroES, their regulators CtsR and HcrA, ClpC and ClpE) (Data S1). Pyruvate metabolism appeared to be remodeled in coculture by the upregulation and downregulation of many pyruvate-associated genes (Fig. 4C; Data S1). Concerning lactate consumption, the malate/lactate antiporter mleN was slightly upregulated, while genes related to the L- and D-lactate dehydrogenases were downregulated (lutA-lutC, FNLLGLLA_01898, and fucO). Genes involved in iron or other metal uptake through the inner membrane were also both upregulated and downregulated.

We also compared specifically all coculture conditions compared to V. parvula ∆vtaADE with S. gordonii (Fig. 4D). Overall, only a few V. parvula genes involved in purine metabolism were upregulated specifically when aggregating in cocultures, either through co-aggregation or auto-aggregation (Fig. 4B and D, blue bar, supplementary data S1). By contrast, 22 genes were specifically differentially regulated in coculture in the absence of any type of aggregation among which genes involved in NADH regeneration through xanthine to urate conversion were slightly downregulated (Fig. 4B, orange bar, Data S1) suggesting a very slight effect of intra- and inter-bacterial proximity.

On the other hand, there were very few changes in S. gordonii transcriptome when cocultured with V. parvula. The only upregulated genes in all cocultures conditions (Fig. 5A and B, green bar, Data S2) are part of the Bfb PTS system (SGO_1575-82) already described as induced when co-aggregating with A. naeslundii (34). The only downregulated gene (SGO_1314) encoded a ZnuA-like metal-binding lipoprotein (Fig. 5C). No gene expression changes were found specifically associated with co-aggregation (Fig. 5D).

Fig 5.

Transcriptomic response of S. gordonii to V. parvula is limited to the upregulation of a PTS system and downregulation of a metal-binding lipoprotein. (A) PCA of all S. gordonii samples (four biological replicates for 10 conditions). Shades of green represent all S. gordonii ∆visA conditions, shades of blue cocultures of S. gordonii WT, and red the monoculture of S. gordonii WT. (B) Upset plot (Venn diagram alternative) showing the number of differentially expressed genes (defined by an absolute log2fold change >1) shared for each condition compared to S. gordonii WT or ∆visA monocultures (indicated by the ∆visA column). The green bar indicates the core response to coculture and the blue bar the core differences between S. gordonii WT and ∆visA. (C) Volcano plot of the coculture of V. parvula and S. gordonii WT compared to S. gordonii in monoculture. Genes of the PTS system are colored in red. (D) Upset plot for each coculture condition compared to S. gordonii + V. parvula ∆vtaA∆vtaDE coculture.

Altogether, these results indicate that (i) V. parvula transcriptional response to coculture is associated with changes in metabolism and stress, (ii) S. gordonii has a minimal transcriptional response, and (iii) aggregation has only a limited effect on both bacteria, without contribution of auto- or co-aggregation.

Co-aggregation strongly affects the structure of mixed V. parvula/S. gordonii biofilms

To assess the impact of co-aggregation on mixed biofilm formation, we imaged either mono-species or mixed biofilms of V. parvula and S. gordonii formed in tissue culture-treated 96-well plates for 24 h using confocal laser scanning microscopy (CLSM). To differentiate both bacteria, S. gordonii was stained using the monoderm-specific dye BacGO (35) while Syto61 was used to stain all bacteria (Fig. S10A and B). Comparison of co-aggregating mixed biofilms (Vp WT + Sg WT) with mixed biofilms without co-aggregation (Vp ∆vtaDE + Sg ∆visA, Vp ∆vtaDE + Sg WT and Vp WT + Sg ∆visA) showed that, in the absence of co-aggregation, the two partner bacteria were found in distinct patches (Fig. 6A; Fig. S11). This was confirmed by the measurement of roughness (capturing the variations of height over the biofilm) of the streptococcus biofilm in mixed biofilms (Fig. 6B; Fig. S10C and D). Co-aggregating biofilms presented a more homogeneous distribution of the two bacteria populations (Fig. 6A). Volume measurements were variable but suggested that co-aggregation results in a higher overall biofilm volume and an increased S. gordonii biofilm (Fig. S10E and F). Measures of total biofilm formation by crystal violet assay did not show an increase in biofilm formation when co-aggregating (Fig. S10G). Co-aggregation therefore seems to strongly impact on the organization of the two species in mixed biofilms which could profoundly modulate the behavior of these species in vivo.

Fig 6.

Confocal microscopy of mixed biofilms. (A) Representative section images of mixed biofilms, scale bar is 60 µm. Lower and side bands correspond to the orthogonal projections on the z-x and z-y axis, respectively. (B and C) Measured biofilm roughness parameter for the BacGO and the Syto61 dyes. Each point (9–12 per condition) represents the average roughness measurement of four images per well. The experiment was done in three biological independent replicates and three technical replicates. P-values, indicated by asterisks (*P < 0,05, ***P < 0,0005) were calculated using a Kruskal-Wallis test with Dunn’s correction for multiple testing. For all plots, Vp is V. parvula, ∆DE is V. parvula ∆vtaD-vtaE, Sg is S. gordonii, and ∆visA is S. gordonii ∆visA. The presence (or absence) of auto- and co-aggregation is indicated by the + (or -) symbols.

DISCUSSION

Interactions between bacteria and their environment, whether abiotic or biotic, play a key role in determining the nature and evolution of bacterial lifestyles and we previously characterized the V. parvula adhesins involved in its biofilm formation capacities. In this study, we investigated the molecular determinants at the origin of the co-aggregation mechanisms between V. parvula and different members of the dental plaque. We identified three V. parvula and one new S. gordonii adhesins involved in co-aggregation, and studied the impact of such co-aggregation on partner physiology and co-biofilm structure.

Adhesion strategies in Veillonella

We showed that the previously identified V. parvula VtaA adhesin interacts with S. oralis and A. oris while VtaE is responsible for co-aggregation with A. oris and S. gordonii, in which the highly similar, but truncated VtaD has a secondary contribution (Fig. 7).

Fig 7.

The multiple roles of V. parvula adhesins. Model of the interactions mediated by the different V. parvula adhesins. V. parvula auto-aggregates through its VtaA autotransporter, which also mediates co-aggregation with S. oralis. VtaA and VtaE TAAs mediate co-aggregation with A. oris. VtaE and VtaD mediate co-aggregation with S. gordonii by direct or indirect binding to its VisA surface protein.

Contrary to what has been described for V. atypica, where a single adhesin, Hag1, is responsible for all aggregative phenotypes (25), the different adhesive functions in V. parvula are located on different proteins. A comparison of the predicted structures of the Hag1 with VtaE, VtaD, and VtaA shows that Hag1 head section is much more complex than the other adhesins, which could explain its pleiotropic role (Fig. S4B). In addition, Hag1 is almost twice the size in residues (7,187 residues) compared to Hag2 (3,838 residues), the second longest adhesin in V. atypica OK5. In V. parvula SKV38, all major adhesins, including VtaA (3,041 residues), VtaE (3,142 residues), VtaC (2,811 residues) and VtaF (3,193 residues) are of similar size. One hypothesis is that Hag1, because of its long size, could mask other adhesins at the cell surface thus explaining the concentration of activities on the only surface-accessible adhesin. By contrast, in V. parvula SKV38, activities are distributed across multiple co-expressed surface adhesins. VtaD and VtaE head domains are very similar, but VtaE is estimated to be around 100 nm longer than VtaD (Fig. S4A) and we observed that more discrete aggregative phenotypes are associated with VtaD compared to VtaE, which could be due to masking of VtaD by VtaE. In addition, the fact that the double mutants ∆vtaE∆vtaF and ∆vtaE∆vtaC aggregate faster with the purified S. gordonii VisA_G5 than the simple ∆vtaE mutant is also in favor of the hypothesis that a shorter VtaD adhesin is partially masked by the longer VtaC and VtaF adhesins. This masking interference between adhesins has been commonly observed as a possible regulatory mechanism of the surface structures (15, 16). Therefore, selection pressure on adhesion could either apply toward ensuring that the main adhesins do not mask each other by remaining of similar size, still allowing some potential interference relief of shorter adhesins (V. parvula case) or toward accumulating all functions on the tallest adhesin (V. atypica case).

Here we identified VtaA as an adhesin-promoting co-aggregation with S. oralis, whereas we previously showed that VtaA promotes auto-aggregation in BHI (36). This auto-aggregation does not happen after growth in SK medium, which was used to grow V. parvula for co-aggregation assays. This switch from an auto-aggregative to a co-aggregative behavior depends on environmental conditions. This could be an efficient means to rapidly adapt to abrupt changes in environment without affecting the quantity of a single adhesin at the cell surface.

Different Veillonella species occupy different niches within the oral microbiome. V. parvula is strongly associated with dental plaque while V. atypica and V. dispar are found on soft surfaces. Veillonella HMT 780 has a strong specialization for keratinized gingiva (17). It would be interesting to know if differential colonization sites stem from different co-aggregation capacities. This site specialization has been associated with certain genes (e.g., thiamine biosynthesis genes) but no difference in the number of adhesins between sites could be seen (17). However, older studies have shown that Veillonella isolates from different origins within the mouth presented site-specific co-aggregation capacities (8). Revisiting the concept of strain-specific co-aggregation with a modern genetic approach leveraging genome sequencing and genetic manipulation could help us decipher whether Veillonella adhesins specificity to different bacteria is related to the site specificity.

We found that V. parvula binds weakly to human epithelial cells. This differs from what has been described in V. atypica (25). Different species of Veillonella might show different adhesion capacities to host cells, maybe linked to their isolation niche or their adhesin repertoire. However, it could also be due to experimental differences between the two studies. Here, we performed a series of quantitative experiments using cancerous epithelial cell lines, whereas Zhou et al. reported a single observation using fresh human buccal cells collected by swabbing the buccal mucosa. Cancer cells harbor specific surface glycans (37) that may not be recognized by V. parvula. Importantly, Zhou et al. did not perform washes of the buccal cells prior to the addition of V. atypica which may complicate the interpretation of V. atypica adhesion images as native oral bacteria may have been taken up with the fresh buccal cells, so the reported V. atypica adhesion must be taken with caution.

VisA, a novel adhesin of S. gordonii

Like V. parvula, S. gordonii DL1 seems to use different adhesins to bind to different partners. For instance, it binds to certain Veillonella species, including V. atypica OK5, through Hsa, a sialic-acid-binding protein also involved in platelet activation (26, 38, 39). Here, we showed that Hsa is not involved in S. gordonii co-aggregation with V. parvula SKV38 (Fig. 2A) and we have identified a second and new adhesin, VisA (SGO_2004), responsible for this interaction. The use of purified VisA G5 domains demonstrated that they are the portion of VisA recognized by V. parvula. G5 domains are structural folds that are part of the stalk of monoderm surface proteins and are often found associated with an active site (40). For instance, SasG from Staphylococcus aureus or Aap from Staphylococcus epidermidis promote auto-aggregation through interaction between the G5-E domain repeats forming their B domain (the E-domain, another structural domain, is absent in VisA) and have been described to undergo a zinc-mediated dimerization (41). While VisA does not seem to induce auto-aggregation of S. gordonii, the purified protein migrated exclusively at a size corresponding to a dimer in denaturing western blot (something also observed for trimeric autotransporters), suggesting that it also possesses the ability to dimerize even without the E-linker domain (Fig. S12). While we have formal proof that VisA directly interacts with V. parvula, we are still missing formal evidence that this interaction is mediated through direct interaction with VtaE/VtaD (Fig. 7), a hypothesis that we favor. Biochemical studies including biosensor methodologies using purified VtaE/VtaD and VisA would need to be used to clarify the nature of the interactions between VisA and VtaE/VtaD.

Interestingly, the locus encompassing genes encoding VisA-like proteins, PadA, and a thioredoxin reductase is conserved in distant pathogenic streptococci (Fig. S8). PadA (Platelet adherence protein A), in interaction with Hsa, is known to bind to platelets triggering their activation (39). While in laboratory conditions, VisA (formerly known as SGO_2004) does not play a role in platelet interaction (31), its conservation could suggest otherwise in vivo. S. oralis ATCC 10557 also possesses homologs of PadA (HRJ33_07090) and VisA (HRJ33_07095). However, V. parvula adhesins responsible for co-aggregation with the S. gordonii and S. oralis species are not the same, which strongly indicates that S. oralis likely uses a protein different from VisA to co-aggregate with V. parvula. The S. oralis VisA homolog possesses only five G5 domains while S. gordonii VisA has seven domains. The protein could be too short in S. oralis and masked by other surface components or not expressed. This could explain why VisA does not contribute to S. oralis co-aggregation with V. parvula.

Taken together, these results further illustrate the versatility in the use of various adhesins to co-aggregate both streptococci and Veillonella species.

What drives the response to coculture in oral bacteria?

Although limited, modifications of gene expression during the coculture of V. parvula and S. gordonii were observed. For S. gordonii, the main answer to coculture with V. parvula was the upregulation of a PTS system encoded by the bfb operon (SGO_1575–1582). This system was found upregulated in S. gordonii when co-aggregating with Actinomyces oris (34) and one gene of the operon downregulated when co-aggregating with Fusobacterium nucleatum (42). The bfb operon is associated with biofilm formation as the deletion of several genes led to a decrease in adhesion and biofilm formation while the operon promoter was 25% more active in biofilms (43). An increase in arginine concentration could be at the origin of the induction of this S. gordonii PTS system. Indeed, arginine is known to be important for S. gordonii biofilm formation and arginine restrictions result in strong downregulation of the bfb operon in monoculture (44). Co-aggregation of S. gordonii with A. oris resulted in the downregulation of arginine biosynthesis and the upregulation of the bfb operon through the uptake of A. oris-produced arginine. One of the upregulated pathways in Veillonella when cocultured with S. gordonii is arginine biosynthesis. Therefore, one can hypothesize that V. parvula would favor S. gordonii biofilm formation by producing arginine. We have, however, not detected any decrease in the arginine biosynthesis pathway in S. gordonii or changes in the expression of arginine-dependent regulators argC, argR, or argC.

Globally, coculture did not result in major changes in gene expression in our experiments performed in anaerobic conditions using a rich and buffered medium without metabolic dependency. Auto-aggregation and co-aggregation themselves had a negligible impact on the observed responses by both bacteria. The induction of the alpha-amylase amyB gene expression in S. gordonii caused by an unknown diffusible signal produced by V. parvula (19, 20) was not observed in our experiments. This may be due to our specific conditions that did not allow the production of the signal by V. parvula. Other examples of oral bacteria responding weakly to co-aggregation are F. nucleatum interacting with S. gordonii (42) and S. mutans interacting with V. parvula (45, 46). These results suggest that oral bacteria do not actually sense attachment to other bacteria but rather changes in nutrient availability and environmental conditions such as pH or oxidative stress. Auto-aggregation and biofilm lifestyle are known to induce large metabolic changes in common aerobic bacteria, inducing genes involved in stress response and anaerobic metabolism in E. coli (47) which seem mostly driven by oxygen gradients, as shown in aggregates of P. aeruginosa (48).

While anaerobic conditions could explain the limited response caused during coculture and interactions between V. parvula and S. gordonii, exposure to oxygen could strongly impact the response to co-aggregation of anaerobic bacteria. Indeed, in another study looking at S. gordonii and V. parvula co-transcriptomes, Mutha et al. reported broad changes in Veillonella including a predominant response to oxidative stress with 39 out of 272 regulated genes associated with it while S. gordonii samples presented high inter-variability (33). No common gene regulation could be detected between our results and their results, possibly due to different experimental settings, as they looked at response from short (30 min) aerobic co-aggregation in saliva while we looked at transcriptional responses after 6 h of anaerobic coculture in BHI pyruvate-rich medium. The aerobic conditions used during this short co-aggregation period could explain the strong V. parvula response to oxidative stress exacerbated by S. gordonii. These different results obtained depending on the level of oxygen also pinpoint how modification of environmental conditions could strongly impact the physiological response that can be induced by aggregation. One can anticipate that modification of the growth medium composition, for example, the presence of saliva, cariogenic sugars, etc., could strongly impact the structure and activity of the oral flora, and the cooperation/competition between bacterial oral members, and their physiological response to co-aggregation.

Could the proximity within the biofilm enhance synergistic or antagonistic interactions?

We hypothesized that co-aggregation could influence the localization of the two bacteria within the biofilm. Indeed, co-aggregation was necessary to promote the colocalization of the two bacteria. Proximity within the biofilm would be essential to Veillonella parvula as it can favor the uptake of lactate by bringing it closer to the producer streptococci. It could also favor signal transduction as demonstrated for the distance-dependent induction of S. gordonii amyB.

Without co-aggregation, both bacteria were distant from each other in the biofilm. This could be explained by a passive clonal development but also by an active prevention of biofilm colonization by non-aggregating partners. This could have a strong effect in vivo by limiting the entry of non-co-aggregating members (including S. mutans) into the dental plaque biofilm while permitting the presence of cooperative partners in close vicinity. A similar mechanism has been demonstrated in Vibrio cholerae, where deletion of rbmA, the gene encoding RbmA, a matrix protein involved in mother-daughter cell cohesion, resulted in higher penetration by invaders as cells were less tightly packed in the biofilm (49). In addition, mixed biofilms between RbmA producers and deficient strains resulted in patchy structures reminiscent of our observation.

Mixed biofilms have often been described to increase stress resistance compared to single-species biofilms. For instance, synergistic biofilm formation by four marine bacteria promoted protection against invasion by the pathogen Pseudoalteromonas tunicata and increased resistance to hydrogen peroxide and tetracycline compared to monospecies biofilms (50). The resistance in a three-species biofilm was due to the protective capacity of one of the resident members (51). We hypothesize that, while mixed biofilms are already more stress-resistant, co-aggregation between members could further increase stress resistance.

In conclusion, we have shown that V. parvula uses specific sets of multiple trimeric autotransporters to specifically interact with other members of the oral dental plaque. While these adhesive capacities are not necessary for intercellular communication under conditions tested, they reduce the distance between members of the biofilm. The co-aggregation phenomena are likely to contribute to the highly organized process of dental plaque formation by modulating the successive addition of interacting bacterial species.

MATERIALS AND METHODS

Growth conditions

Bacterial strains are listed in Table S1. Streptococcus spp. and A. oris were grown in brain heart infusion (BHI) medium (Bacto brain heart infusion; Difco). V. parvula was grown in BHI supplemented with 0.6% sodium DL-lactate (BHIL) or SK medium [10 g/L tryptone (Difco), 10 g/L yeast extract (Difco), 0.4 g/L disodium phosphate, 2 g/L sodium chloride, and 10  mL/L 60% (wt/vol) sodium DL-lactate; described in Knapp et al. (52)], in which it does not auto-aggregate (Fig. S1). F. nucleatum was grown in BHI medium supplemented with 5 µg/mL hemin and 1 µg/mL vitamin K3. For all experiments using oral bacteria, they were incubated at 37°C under anaerobic and static conditions in anaerobic bags (GENbag anaero; bioMérieux no. 45534) or in a C400M Ruskinn anaerobic-microaerophilic station. Escherichia coli was grown in lysogeny broth (LB) (Corning) medium under aerobic conditions at 37°C (agitation at 150 rpm). Antibiotics were added to the cultures as required for mutant production and plasmid stability: 20 mg/L chloramphenicol (Cm), 200 mg/L erythromycin (Ery), 300 mg/L kanamycin (Kan), or 2.5 mg/L tetracycline (Tet) for V. parvula cultures; 5 mg/L Ery for S. gordonii cultures; and 25 mg/L Cm or 100 mg/L ampicillin (Amp) for E. coli cultures. All chemicals were purchased from Sigma-Aldrich unless stated otherwise.

Veillonella parvula natural transformation

Cells grown overnight on the plate were scraped and resuspended in 1 mL SK medium adjusted to an optical density at 600 nm (OD₆₀₀) of 0.4 to 0.8, and 15 µL was spotted on SK agar petri dishes. On each drop, 1–5 μL (75–200  ng) linear double-stranded DNA PCR product was added. The plates were then incubated anaerobically for 24–48 h. The biomass was resuspended in 500 µL SK medium, plated on SK agar supplemented with the corresponding antibiotic, and incubated for another 48 h. Colonies were streaked on fresh selective plates, and the correct integration of the construct was confirmed by PCR and sequencing.

Veillonella parvula mutagenesis and complementation

V. parvula site-directed mutagenesis was performed as described by Knapp et al. (52) and Béchon et al. (27) Briefly, upstream and downstream homology regions of the target sequence and the V. atypica kanamycin (aphA3 derived from the pTCV-erm (53) plasmid under the V. parvula PK1910 gyrA promoter) or tetracycline resistance cassette were PCR amplified with overlapping primers using Phusion Flash high-fidelity PCR master mix (Thermo Scientific, F548). PCR products were used as templates in a second PCR round using only the external primers, resulting in a linear dsDNA with the antibiotic resistance cassette flanked by the upstream and downstream sequences. vtaE chromosomal complementation was done by inserting in the promoter region the previously described Veillonella P_Tet promoter (27) associated with an erythromycin resistance cassette. Primers used in this study are listed in Table S2 in the supplemental material.

Streptococcus gordonii natural transformation

25 µL of an O/N culture, 100 µL of heat-inactivated horse serum (Sigma), 900 µL of THY Broth, 2 µL of competence-specific peptide (1 mg/mL, DLRGVPNPWGWIFGR, synthesized by GenScript), and 1–5 µL of linear double-stranded DNA PCR product were mixed in a microcentrifuge tube, incubated anaerobically for 5–8 h at 37°C, and plated on selective agar medium for 1–3 days. Colonies were streaked on fresh selective plates, and the correct integration of the construct was confirmed by PCR and sequencing.

Streptococcus gordonii complementation

To create a markerless mutant of SGO_2004 with a P_Tet promoter, we took advantage of the described IDFC2 cassette (54), containing an erythromycin resistance and a mutant pheS gene encoding the A314G missense mutation providing sensitivity to p-chlorophenylalanine (4 CP). Briefly, the IFDC2 cassette and homology regions before and after the promoter of SGO_2004 were amplified from an S. gordonii strain containing IDFC2. PCR products were used as templates in a second PCR round using only the external primers, which generated a linear dsDNA with the IFDC2 cassette flanked by the upstream and downstream sequences. Streptococcus gordonii DL1 WT was transformed with this construct and selected for insertion of the cassette with erythromycin.

For a second time, the IDFC2 cassette was replaced by the P_Tet promoter of pRPF185 plasmid fused with the pVeg RBS (55) by creating a construct with similar homologies regions than for the IFCD2 cassette or using an homology region upstream of padA to create the ∆padA,pTet-SGO_2004 mutant. After the transformation of S. gordonii IDFC2-DL1 with either construct, counter selection was done on BHI + 4 CP plates, and selected mutants were verified by Sanger sequencing and for sensibility to erythromycin.

Auto- and co-aggregation assays

These assays were performed using independent cultures. Overnight cultures were centrifuged for 5 min, 5,000 g and resuspended in aggregation buffer (25) (1 mM Tris-HCl buffer, pH 8.0, 0.1 mM CaCl₂, 0.1 mM MgCl₂, 150 mM NaCl) to a final OD₆₀₀ of 1. For auto-aggregation, 800 µL of each bacterial suspension was added to a microspectrophotometer cuvette (Fisherbrand). For co-aggregation, 400 µL of bacterial suspension was mixed with 400 µL of the other bacterial suspension. For both auto- and co-aggregation, the bacteria were left to sediment on the bench in the presence of oxygen, so no growth should occur. The OD₆₀₀ was measured every hour in a single point of the cuvette using a SmartSpec spectrophotometer (Bio-Rad). OD₆₀₀ was then normalized to the initial OD₆₀₀ by the formula ${\displaystyle 100\times \frac{\mathrm{O}\mathrm{D}{600}_{\mathrm{T}\mathrm{\_}\mathrm{X}}}{\mathrm{O}\mathrm{D}{600}_{\mathrm{T}\mathrm{\_}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{a}\mathrm{l}}}}$.

Purification of SGO_2004 (VisA) G5 domains

The portion of SGO_2004 coding for G5 domains (residues 138–698) was amplified from S. gordonii and the pET22b-HIS vector was linearized by PCR. The PCR products were then purified and annealed by Gibson reaction. The plasmid was dialyzed and transformed into electrocompetent E. coli DH5-alpha. After verification of the construct by sequencing, the plasmid was purified and transformed in E. coli BL21(DE3)-pDIA17. After growth to OD₆₀₀ 0.4, cells were induced with 0.1 mM IPTG and grown for 3 h at 37°C before harvesting. The cell pellet was frozen O/N, then resuspended in Buffer A (30 mM Tris-HCl pH 7.5, 300 mM NaCl, 30 mM Imidazole), and lysed by sonication. Debris was pelleted by ultracentrifugation (50,000 g, 30 min) and supernatant run through a HisTrap 5 mL column on an AKTA Explorer (GE) against a gradient of imidazole (30–300 mM). After resuspension in Laemmli buffer with 2.5% betamercaptoethanol, the purified protein was assessed for purity by SDS-Page followed by SafeStain SimplyBlue (ThermoFisher) staining and western blot against the HIS-tag (Fig. S12) and dialyzed twice against 30 mM Tris-HCl pH 7.5, 300 mM NaCl using a SnakeSkin 3500 Da (ThermoFisher).

Immunofluorescence of surface-bound VisA_G5

V. parvula was grown overnight in SK and washed two times in PBS. VisA_G5 was preincubated 1 h in the dark at 0.1 mg/mL with 1/10 of an anti-His Tag monoclonal antibody coupled with Alexa Fluor 488 (MA1-135-A488, Invitrogen). 50 µL of bacteria at OD₆₀₀ 1 was incubated for 2 h with 5 µL of the fluorescent VisA_G5 and subsequently mounted on a slide. Cells were imaged using a Zeiss Axioplan 2 microscope equipped with an Axiocam 503 mono camera (Carl Zeiss, Germany). Epifluorescence images were acquired using the ZEN lite software (Carl Zeiss, Germany) and processed using Fiji (ImageJ).

RNA extraction

The RNAseq experiments were performed on co-culture of V. parvula and S. gordonii. 600 µL of anaerobic media BHIP (BHI + 100 mM sodium pyruvate) in a 1.5 mL tube was inoculated with each of the bacteria at OD₆₀₀ 0.05 and incubated for 6 h anaerobically. The resulting co-cultures were mixed with 1.2 mL of RNAprotect Bacteria reagent (QIAGEN), vortexed, and incubated at RT for 5 min, before centrifugation (10,000 rpm, 4°C) for 5 min. The supernatant was removed and pellets were kept at −80°C before RNA extraction. For lysis, pellets were washed with 700 µL of PBS and resuspended in 200 µL of lysis buffer (15 mg/mL Lysozyme, 100 µL/mL Proteinase K) before incubation for 3 h at 37°C with constant shaking (750 rpm). Each sample was then added to a matrix B lysis tube with 800 µL of TRIzol and lysed using a FastPrep (two times S. mutans preregistered protocol). 800 µL of 100% ethanol was added and samples were centrifuged to pellet debris (8,000 g, 2 min). Lysate was transferred to a column from the kit Direct-zol RNA Miniprep plus (Zymol) and the rest of the extraction was done following the provider’s manual.

RNA sequencing

Libraries were prepared using Illumina Stranded Total RNA Prep from 440 ng of RNA. RiboZero kit Microbiome kit (Illumina) was used to eliminate ribosomal RNA. The subsequent steps were as follows: RNA fragmentation, cDNA synthesis (incorporating uracils into the second strand), adapter ligation, indexing by PCR with 17 cycles (amplifying only the first strand), purification of unbound adaptors and primers on AMP beads (Beckman Coulter). The resulting stranded libraries comprised fragments from 200 to 1,000 bp with peaks lying between 390 and 470 bp as visualized on a 5300 Fragment Analyzer (Agilent Technologies). No low-molecular peaks corresponding to unbound adaptors and primer dimers were observed. Libraries were pooled and sequenced on a NovaSeq X 10 B flow cell (Illumina) producing 1,200 million 150 × 150 bp pair-end reads. As a result, each sample was represented by 18–55 million reads.

Ribofinder was used to verify the efficiency of ribodepletion: only around 5% of reads mapped to ribosomal RNA. Taxonomy analysis using the Kraken module confirmed the presence of S. gordonii and V. parvula RNA according to the co-infection design. In coinfection samples, reads from the two species were present in more or less equal proportions. The RNA-seq analysis was performed with Sequana (56). In particular, we used the RNA-seq pipeline [v0.19.2 (https://github.com/sequana/sequana_rnaseq)] built on top of Snakemake v7.32.4 (57). Reads were trimmed from adapters and low-quality bases using fastp software v0.22.0 (58), then mapped to the reference genome using Bowtie2 v2.4.6 (59). Genomes and annotations were downloaded from the NCBI website using Veillonella parvula SKV38 (GenBank LR778174.1) and S. gordonii DL1 (GenBank CP000725.1) genome references. FeatureCounts 2.0.1 (60) was used to produce the count matrix, assigning reads to features using the annotation aforementioned. Statistical analysis on the normalized count matrix was performed to identify differentially regulated genes. Differential expression testing was conducted using DESeq2 library 1.34.0 (61) scripts, and HTML reporting was made with the Sequana RNA-seq pipeline. Parameters of the statistical analysis included the significance (Benjamini-Hochberg adjusted P-values, false discovery rate FDR < 0.05) and the effect size (fold-change) for each comparison.

Confocal laser scanning microscopy

Biofilms were formed in a 96 PhenoPlate 96-well, black, optically clear flat-bottom, tissue-culture-treated plate adapted to microscopic observation (PhenoPlate, PerkinElmer) by inoculating 150 µL of anaerobic media BHIP (BHI + 100 mM sodium pyruvate) with overnight culture of each species at OD₆₀₀ 0.05 for each of them. After 1 h of adhesion, media was replaced to remove planktonic bacteria and incubated for 24 h. Biofilm was stained by the addition of 50 µL of BHIP media containing both the BacGO (1 µM final concentration) and the Syto61 dyes (5 µM final concentration). Three images set at defined positions within each well were acquired on an Opera Phenix Plus High Content Screening System running with Harmony software v.5.1 (Revvity, formerly known as PerkinElmer), using the following modalities: 20× water/NA 1.0, Z-stack, 40 planes, 2 µm step between planes, for the Syto61 dye: λ_exc: 640 nm/emission filter 650–760 nm), for the bacGO dye: λ_exc: 561 nm/emission filter 571–596 nm. The resulting images were analyzed using BiofilmQ 1.0.1 (62).Images were first denoised by convolution (dxy = 5, dz = 3) and top hat filter (dxy = 25), then segmented into two classes using an OTSU thresholding method with a sensitivity of 0.15 for the Syto61 channel and 0.25 for the BacGO channel. Images were then declumped in 10-pixel wide cubes and surface properties (range 30 pixels) and global biofilm properties were calculated (supplementary data S3). Illustrative images were generated with Imaris 9.0.

DATA AVAILABILITY

Supplementary data are available at https://github.com/ldorison/Coaggregation_streptococcus_Veillonella-.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/mbio.02171-24.

Supplemental material. mbio.02171-24-s0001.pdf.

Tables S1 and S2; description and link for Data S1 to S3; Figures S1 to S12; supplemental methods.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.