Quantification of Extracellular DNA Network Abundance and Architecture within Streptococcus gordonii Biofilms Reveals Modulatory Factors

ABSTRACT

Extracellular DNA (eDNA) is an important component of biofilm matrix that serves to maintain biofilm structural integrity, promotes genetic exchange within the biofilm, and provides protection against antimicrobial compounds. Advances in microscopy techniques have provided evidence of the cobweb- or lattice-like structures of eDNA within biofilms from a range of environmental niches. However, methods to reliably assess the abundance and architecture of eDNA remain lacking. This study aimed to address this gap by development of a novel, high-throughput image acquisition and analysis platform for assessment of eDNA networks in situ within biofilms. Utilizing Streptococcus gordonii as the model, the capacity for this imaging system to reliably detect eDNA networks and monitor changes in abundance and architecture (e.g., strand length and branch number) was verified. Evidence was provided of a synergy between glucans and eDNA matrices, while it was revealed that surface-bound nuclease SsnA could modify these eDNA structures under conditions permissive for enzymatic activity. Moreover, cross talk between the competence and hexaheptapeptide permease systems was shown to regulate eDNA release by S. gordonii. This novel imaging system can be applied across the wider field of biofilm research, with potential to significantly advance interrogation of the mechanisms by which the eDNA network architecture develops, how it can influence biofilm properties, and how it may be targeted for therapeutic benefit.

IMPORTANCE Extracellular DNA (eDNA) is critical for maintaining the structural integrity of many microbial biofilms, making it an attractive target for the management of biofilms. However, our knowledge and targeting of eDNA are currently hindered by a lack of tools for the quantitative assessment of eDNA networks within biofilms. Here, we demonstrate use of a novel image acquisition and analysis platform with the capacity to reliably monitor the abundance and architecture of eDNA networks. Application of this tool to Streptococcus gordonii biofilms has provided new insights into how eDNA networks are stabilized within the biofilm and the pathways that can regulate eDNA release. This highlights how exploitation of this novel imaging system across the wider field of biofilm research has potential to significantly advance interrogation of the mechanisms by which the eDNA network architecture develops, how it can influence biofilm properties, and how it may be targeted for therapeutic benefit.

INTRODUCTION

Biofilm development is characterized by the production and release of extracellular polymeric substance (EPS) to form a matrix that encases the microbial community. EPS accounts for >90% of biofilm dry weight and comprises a rich tapestry of components including extracellular DNA (eDNA), which has been found as a common component of biofilms across a range of environments (1–3). Diverse roles have been ascribed to eDNA, including maintenance of biofilm structural integrity, facilitating initial adhesion to surfaces, acting as a reservoir for genetic exchange, providing protection against antimicrobial compounds, and serving as a nutrient source (4). As a consequence, eDNA is often considered an attractive target for the management of biofilms, which account for up to 80% of all nosocomial infections in humans (5).

Within the biofilm, eDNA is proposed to conform to an “electrostatic net” model where, under low-pH conditions, negatively charged eDNA forms electrostatic interactions with positively charged DNA-binding proteins within EPS, acting as a net that interconnects cells (6, 7). Advances in techniques for the visualization of fluorescently stained eDNA networks have provided insights into their structural composition (8–10). Specifically, eDNA has been shown to form Holliday junction-like (9) and G-quadruplex (8) structures, stabilized by DNA-binding proteins (11–14), that ultimately form cobweb- or lattice-like networks across the biofilm (9, 10). However, understanding of the mechanisms by which eDNA is released, how this is regulated, and the spatiotemporal dynamics of eDNA network formation remains limited. This is, in part, driven by a lack of tools with the capacity to reliably detect and quantify the abundance and architecture of eDNA networks within biofilms.

One ecological niche in which eDNA is recognized as a prominent component of biofilms and a promising therapeutic target is within the oral cavity and, specifically, dental plaque. Streptococcus gordonii is a pioneer colonizer and ubiquitous constituent of dental plaque biofilms, where it can influence the accretion of the dental plaque community on salivary pellicle (3, 15). DNA extraction techniques that enable the quantification of soluble eDNA have confirmed the capacity for S. gordonii to produce eDNA during biofilm formation (16, 17). From such studies, S. gordonii eDNA is hypothesized to be of chromosomal origin, and its release has been shown to be hydrogen peroxide (H₂O₂) dependent (16, 18). However, further insights into the parameters that may affect S. gordonii eDNA networks and their overall architecture are lacking.

Here, we demonstrate use of a novel, high-throughput image acquisition and analysis platform to reliably quantify the abundance and architecture of eDNA networks in situ within early S. gordonii biofilms. By exploiting this technology, these studies provide evidence of glucan stabilization of the eDNA matrix, reveal that a surface-bound nuclease can modulate the eDNA networks, and identify cross talk between the competence and hexaheptapeptide permease (Hpp) systems in regulating eDNA release. The high-level detail of eDNA network analysis that this imaging system provides has potential to significantly advance current understanding of biofilm development and manipulation across the spectrum of biofilm research.

RESULTS

Evaluation of eDNA production in early-stage S. gordonii biofilms.

Pilot studies had indicated the capacity for S. gordonii to produce an eDNA network during biofilm formation, alike in architecture to the yarn-like eDNA structures produced by Enterococcus faecalis biofilms (10). S. gordonii biofilms were therefore selected as the model to verify the capacity for our image analysis approach to reproducibly quantify eDNA networks in situ within biofilms. Before such studies could be undertaken, however, it was necessary to establish the optimal stage during biofilm development at which S. gordonii produces eDNA. Previous reports had indicated that S. gordonii releases eDNA during early biofilm formation (16, 17), but detailed, time-dependent changes in eDNA release were lacking. Phenol-chloroform-isoamyl DNA extraction was combined with crystal violet staining to systematically evaluate changes in soluble eDNA and biomass quantities over time. A time-dependent increase in eDNA concentration was seen that peaked at 5 h and then began to decline, while biomass levels continued to increase beyond 5 h (Fig. 1). This indicated that eDNA levels did not simply correlate with S. gordonii cell numbers. As it represented the peak for eDNA concentration, a 5-h time point was selected to further evaluate eDNA within S. gordonii biofilms.

FIG 1.

Changes in S. gordonii biofilm biomass and eDNA over time. WT S. gordonii biofilms were grown at 37°C in YPTG on saliva-coated 24-well plates for up to 24 h, and levels of biomass were determined by crystal violet staining (line), or eDNA was assessed using the phenol-chloroform-isoamyl DNA extraction method (columns). Data are presented as mean values ± standard deviation. *, P < 0.05; **, P < 0.01, or ****, P < 0.0001, compared to 1-h value as determined by one-way ANOVA followed by post hoc Tukey test (n = 3).

Quantification of eDNA networks within S. gordonii biofilms.

The quantification of DNA by phenol-chloroform-isoamyl alcohol extraction has been used widely to quantify levels of eDNA within biofilms (19–21). However, this approach only indicates the concentration of soluble eDNA and can provide no information on the structural complexity of eDNA within the biofilm architecture. To address this limitation, a novel mass image acquisition and high-throughput image analysis system was devised to both visualize and quantify eDNA networks in situ within biofilms. Immunolabeling of double-stranded eDNA combined with TO-PRO-3 staining of S. gordonii cells revealed web- or constellation-like networks of eDNA extending across the S. gordonii biofilm (Fig. 2). Our image analysis software could then be exploited to detect and quantify these eDNA structures. Due to differences in pixel density between the background of the image and eDNA matrices, our software was able to detect and subsequently highlight eDNA structures. Comparison of automated eDNA detection with manual detection rate confirmed a relatively high level of accuracy, with a nondetection rate of <8% (Fig. 3). Nondetected fragments were <5 μm and predicted to comprise colloidal particles and debris. To optimize accuracy, a minimum detection threshold of 5 μm was therefore set for subsequent analyses.

FIG 2.

Interwoven networks of eDNA in S. gordonii biofilm. WT S. gordonii biofilms were grown at 37°C in YPTG on saliva-coated 24-well plates for 5 h. Networks of eDNA (red) and S. gordonii biofilm cells (green) were fluorescently labeled and visualized by widefield microscopy. Representative images are shown. Scale bar, 50 μm.

FIG 3.

Optimization of automated eDNA detection. WT S. gordonii biofilms were grown at 37°C in YPTG on saliva-coated 24-well plates for 5 h. Networks of eDNA were then immunolabeled and visualized by widefield microscopy (A). Image analysis software was used to detect and quantify eDNA strands, and the reliability of this system was assessed (B). Different colors denote complete detection (green), fragmented detection (orange), undetected eDNA (magenta), non-eDNA (cyan), and background particles (white/not highlighted). Representative images are shown. Scale bars, 50 μm.

The software was designed to highlight eDNA structures with various colors to indicate different points of origin of each eDNA structure (Fig. 4). Information regarding their quantity and architecture could then be output. Specifically, data could be obtained regarding the total percentage of each field of view comprising eDNA, total eDNA stranding (micrometers per square millimeter, total length of eDNA strands per square millimeter), average branch length (micrometers), and average number of eDNA branches diverging from a single point (number of junctions/number of branches per junction). To test this analysis capability, while verifying the accuracy and sensitivity of this imaging approach in detecting eDNA, studies were repeated in the presence of an increasing concentration (10 to 25 μg/mL) of DNase I. As was anticipated, a significant reduction in eDNA levels was seen for both DNase I concentrations tested (Fig. 5A). This was reflected in the quantification, as percent field of view comprising eDNA networks (Fig. 5B) and total eDNA stranding (Fig. 5C) were significantly diminished following DNase I application. Variations in eDNA architecture could also be measured. The average number of eDNA branches reduced with increasing DNase I concentration (Fig. 5F), likely correlated with the general reduction in eDNA levels, but no significant effect on average branch length (Fig. 5D) or eDNA junction structure (Fig. 5E and G) was seen. DNase I had no significant impact on overall biofilm biomass levels (see Fig. S1 in the supplemental material). As a comparison with the time-dependent differences in eDNA shown in Fig. 1, the imaging tool was also applied to analysis of 5-h versus 24-h biofilms (Fig. S2). No significant differences in eDNA architecture were observed, but eDNA levels were significantly diminished at 24 h compared to 5 h, correlating with the soluble eDNA data. To explore if the differences in eDNA levels were linked to time-dependent changes in nuclease activity, fluorescence-based DNase activity assays were performed on cell-bound and secreted biofilm fractions (Fig. S3). However, DNase activity was significantly lower at 24 h than at 5 h. Together, these data provided confidence that the imaging system could accurately detect eDNA networks within biofilms and provide information relating to both quantities of eDNA and the overall architecture of the eDNA networks. These data also implied that DNase I could drive the removal and/or release of eDNA, thus reducing bulk quantity, but did not significantly impact its fundamental organization.

FIG 4.

Visualization of eDNA in S. gordonii biofilms at 5 h. WT S. gordonii biofilms were grown at 37°C in YPTG on saliva-coated 24-well plates for 5 h. Networks of eDNA were then immunolabeled and visualized by widefield microscopy (A). Image analysis software was used to detect and quantify eDNA strands, as shown in panel C. Panels B and D correspond to higher-resolution images of the section indicated by the red box in panels A and C, respectively. Representative images are shown. Scale bars, 50 μm.

FIG 5.

eDNA detection and quantification following DNase I treatment. WT S. gordonii biofilms were grown at 37°C in YPTG with or without 10 to 25 μg/mL DNase I on saliva-coated 24-well plates for 5 h. Networks of eDNA were then immunolabeled and visualized by widefield microscopy (A, i to iii), and image software was used to detect eDNA networks (A, iv to vi). Quantifiable differences in the percentage of field of view comprising eDNA (B), total eDNA stranding per square millimeter (C), average eDNA branch length (D), average maximum eDNA branch length (E), average number of branches per field of view (F), and average number of junctions per eDNA structure (G) were then assessed using Excel. Data are presented as mean ± standard deviation. **, P < 0.01, and *, P < 0.05, relative to untreated (UT) control, as determined via one-way ANOVA followed by Tukey test (n = 3). Scale bars, 50 μm.

Effects of carbon source on eDNA networks.

Having confirmed the capacity for the imaging system to reliably detect and analyze eDNA networks, the next step was to exploit this approach to gain an improved understanding of eDNA within S. gordonii biofilms. For this work, a series of parameters were selected that had previously been implicated in modulating eDNA. The first of these was the effect of sugars. Prior studies had identified carbon catabolite-dependent modulation of eDNA release in S. gordonii and Streptococcus sanguinis biofilms (22–24), and sucrose has been shown to promote eDNA-dependent Streptococcus mutans biofilm formation, in which glucans were proposed to stabilize the eDNA matrices (25, 26). To validate whether the same trend could be observed within S. gordonii biofilms, our imaging system was exploited to examine the differential effects of glucose and sucrose on S. gordonii total eDNA stranding levels (Fig. 6A). Levels of eDNA for biofilms cultured in sucrose were 69% higher than those observed for glucose-cultured biofilms (Fig. 6B), while biomass levels differed by only 13% (Fig. 6G). No difference in eDNA architecture was observed (Fig. 6C to F). Furthermore, levels of eDNA in glucose-grown biofilms were unaffected by dextranase, although there was a 25% reduction in biomass (Fig. 6B and G). In contrast, a 76% decrease in eDNA levels was observed for the sucrose-grown biofilms following dextranase application, alongside a 25% reduction in biomass (Fig. 6B and G). Dextranase had no impact on eDNA branch length (Fig. 6C and D), but reductions were seen in numbers of junctions/branches per eDNA structure for both glucose- and sucrose-grown biofilms (Fig. 6E and F). These data supported a role for glucans in eDNA networks within S. gordonii biofilms.

FIG 6.

Glucans enhance eDNA levels within sucrose-grown biofilms. WT S. gordonii biofilms were grown on saliva-coated 24-well plates for 5 h at 37°C in YPT with or without 0.2% glucose or sucrose in the absence (UT) or presence of 10 μg/mL dextranase (Dex). eDNA stranding (A and B), eDNA branching (C and D), number of junctions/branches per eDNA structure (E and F), and levels of biomass (G) were then determined by microscopy (A to F) or crystal violet staining (G). Data are presented as mean ± standard deviation. *, P < 0.05, and **, P < 0.01, as determined via one-way ANOVA followed by Tukey test (n = 3).

To further verify an association between eDNA levels and glucans, an S. gordonii ΔgtfG mutant was tested. Glucosyltransferase G (GtfG) is the only glucosyltransferase expressed by S. gordonii, is located extracellularly and is responsible for the generation of glucans during S. gordonii biofilm formation (27). GtfG hydrolyzes dietary sucrose, synthesizing glucose moieties into glucan polymers with α-1,6 and α-1,3 linkages (28, 29). In the presence of glucose, loss of GtfG reduced levels of eDNA by 53%, but this effect was much more pronounced in the presence of sucrose, with a reduction of 84% (Fig. 7A and B). No effect was seen on biomass levels or eDNA branch length upon loss of GtfG under either condition (Fig. 7C, D, and G), but as for dextranase, loss of GtfG resulted in reductions in numbers of junctions/branches per eDNA structure for sucrose-grown biofilms (Fig. 7E and F). Finally, to enable glucans to be visualized alongside eDNA, dextran conjugated to Alexa Fluor 647 was applied to the biofilms over the 5-h period. The fluorescently labeled dextran acts as an acceptor that is incorporated into newly formed glucans by Gtfs. As was expected, sucrose-cultured biofilms exhibited a significantly higher fluorescence output than their glucose-cultured counterparts, confirming a greater abundance of glucans (Fig. 8). Taken together, these data suggest a potential synergy between eDNA and glucans during S. gordonii biofilm formation, in which the glucans may serve to promote the structural stability of eDNA matrices.

FIG 7.

Inability to synthesize glucans impairs eDNA levels within biofilms. WT and ΔgtfG S. gordonii biofilms were grown for 5 h at 37°C in YPT with or without 0.2% glucose or sucrose. eDNA stranding was assessed via widefield microscopy (A) and quantified (B to F). Levels of biomass (G) were determined by crystal violet staining. Data are presented as mean ± standard deviation. *, P < 0.05; **, P < 0.01; and ***, P < 0.001, as determined via one-way ANOVA followed by Tukey test (n = 3). Representative images are shown. Scale bars, 50 μm.

FIG 8.

Sucrose elevates glucan levels within S. gordonii biofilms. WT S. gordonii biofilms were grown on saliva-coated 24-well plates for 5 h at 37°C in YPT with or without 0.2% glucose or sucrose. Relative levels of glucans were measured by inclusion of Alexa Fluor 647-conjugated dextran during biofilm development and subsequent quantification of fluorescence levels. Data are presented as mean ± standard deviation. *, P < 0.05, as determined via Student’s t test (n = 3). RFU, relative fluorescence units.

Effects of DNase SsnA on eDNA networks.

Application of exogenous nuclease enzymes has been shown to disrupt eDNA networks within biofilms (30, 31). Nuclease enzymes are also expressed by several bacterial species, but little is known about their capacity to modulate biofilm eDNA. Previous studies had identified the nuclease activity of S. gordonii, and we have characterized this enzyme as streptococcal surface nuclease A (SsnA) (32). We therefore generated a ΔssnA mutant strain and utilized our imaging system to determine if SsnA can influence eDNA network formation during S. gordonii biofilm development. Levels of eDNA for ΔssnA biofilms were 2.3-fold greater than those observed for S. gordonii wild-type (WT) biofilms (Fig. 9A), while total biomass levels (Fig. 9B) and eDNA architecture (Fig. S4) were comparable. This suggested that SsnA can influence S. gordonii eDNA levels and may have the capacity to manipulate or disperse eDNA networks as the biofilm develops. These studies were then extended to monitor the effects of SsnA in glucose or sucrose environments, since nuclease activity can be regulated by carbon catabolite availability (33, 34). Addition of glucose resulted in eDNA levels that were comparable to those of the ΔssnA mutant in the absence of sugars, and as before, higher levels of eDNA were seen in the presence of sucrose. However, for both sugars, these effects were independent of SsnA, as no significant differences were seen for eDNA or biomass between the ΔssnA mutant and S. gordonii WT (Fig. 9A and B). One potential explanation for this effect was that utilization of the sugars via glycolysis and concomitant production of lactic acid reduced the local pH to below the activity threshold for SsnA. To explore this, the pH of the culture medium with or without sugar supplementation following biofilm formation was measured. For both glucose and sucrose, it was confirmed that pH levels fell below pH 7.0, which would have significantly reduced SsnA activity (Table 1).

FIG 9.

SsnA can modulate eDNA levels but is affected by carbon source. S. gordonii WT and ΔssnA biofilms were grown on saliva-coated 24-well plates for 5 h at 37°C in YPT in the absence (UT) or presence of 0.2% glucose or sucrose. eDNA stranding (A) and levels of biomass (B) were determined by microscopy or crystal violet staining, respectively. S. gordonii WT and ΔssnA biofilms were also grown with or without 5 μg/mL SsnA, and eDNA stranding (C) and biomass (D) were determined as described above. Data are presented as mean ± standard deviation (n = 3). *, P < 0.05, and ****, P < 0.0001, as determined via one-way ANOVA followed by Tukey test.

TABLE 1.

Medium pH of 5-h biofilms

Treatment	Strain	pHa
Untreated	WT	7.11 ± 0.03
	WT + SsnA	7.18 ± 0.06
	ΔssnA	7.16 ± 0.11
	ΔssnA + SsnA	7.15 ± 0.11
0.2% glucose	WT	6.85 ± 0.06
	ΔssnA	6.84 ± 0.10
0.2% sucrose	WT	6.26 ± 0.09
	ΔssnA	6.66 ± 0.11

^aData are presented as mean ± standard deviation, n = 8.

As a final assessment of the capacity for SsnA to modulate eDNA networks, recombinant SsnA (rSsnA) was applied to biofilms formed by S. gordonii WT and ΔssnA strains (Fig. 9C and D). As before, no significant differences were seen in total biomass levels between the two strains (Fig. 9D). In contrast, while exogenous SsnA had no impact on the levels of eDNA for WT S. gordonii biofilms, the enhanced eDNA stranding levels seen for ΔssnA biofilms were reduced to WT levels following application of rSsnA (Fig. 9C). Exogenous SsnA had no impact on overall eDNA architecture (Fig. S5) These data provide further evidence of a role for SsnA in manipulating the eDNA networks of S. gordonii biofilms under conditions permissive for enzymatic activity.

Modulation of eDNA networks via competence and Hpp systems.

A number of studies have implicated the competence (comCDE) system in regulating eDNA release by S. gordonii (16–18, 35). We therefore used our imaging system to verify the modulatory effects of the competence system on eDNA networks within S. gordonii biofilms. A panel of knockout mutants defective in different stages of the competence pathway were utilized for these studies: ΔcomC (cannot express competence-stimulating peptide [CSP]), ΔcomDE (expresses but cannot detect CSP), ΔcomCDE (cannot express or detect CSP), and ΔcomR1/R2 (cannot upregulate competence genes in response to CSP). Biomass levels were comparable for biofilms formed by all the strains tested (Fig. 10B). In contrast, relative to WT, biofilms formed by ΔcomC, ΔcomCDE, and ΔcomR1/R2 strains exhibited significantly lower levels of eDNA, with reductions of 56%, 83%, and 68%, respectively (Fig. 10A). This confirmed the proposed role of competence genes in mediating S. gordonii eDNA release and of CSP as the signal to induce these effects. Unexpectedly, however, eDNA levels for ΔcomDE biofilms were comparable to those of WT, despite the absence of the cognate two-component signal system (ComDE) to detect the CSP signal (Fig. 10A). This suggested that S. gordonii may be utilizing an alternative mechanism to detect CSP, and this hypothesis was further supported by complementation studies using exogenous CSP (Fig. 10C and D). As was anticipated, application of exogenous CSP to ΔcomC biofilms restored eDNA networks to WT levels (Fig. 10C). No effect was seen for the already higher eDNA levels of WT and ΔcomDE biofilms. However, a significant (6-fold) increase was also seen in eDNA following application of exogenous CSP to ΔcomCDE biofilms, despite lacking the ComDE CSP detection apparatus (Fig. 10C). Importantly, this response was specific to CSP, as no such effect was seen following application of a scrambled CSP as control (data not shown). Assessment of eDNA architecture revealed some variation in branch length or eDNA junction composition for the competence mutant biofilms relative to WT (Fig. S6), but exogenous CSP had no significant effects (Fig. S7).

FIG 10.

CSP modulates eDNA stranding in S. gordonii biofilms, even in the absence of ComDE apparatus. WT S. gordonii and various comCDE operon mutant strains were grown at 37°C in YPTG on saliva-coated 24-well plates in the absence (white bars) or presence (gray bars) of CSP for 5 h. Levels of eDNA stranding (A and C) and biomass (B and D) were then determined by microscopy or crystal violet staining, respectively. Data are presented as mean ± standard deviation (n = 3; ΔcomDE, n = 2). *, P < 0.05; **, P < 0.01; ***, P < 0.001; or ****, P < 0.0001, as determined by one-way ANOVA followed by Tukey test.

Another regulatory system that has been associated with competence in S. gordonii is the hexaheptapeptide permease (Hpp) system (36). The Hpp system is an oligopeptide permease system comprising four constituents: HppA, HppB, HppG, and HppH. HppA has been implicated in substrate-specific binding and, along with HppH, transports peptides comprising 5 to 7 amino acid (aa) residues across the cell envelope and into S. gordonii cells. To ascertain whether the Hpp system may have a capacity to detect CSP in the absence of ComDE and so facilitate CSP modulation of eDNA networks, knockout mutants lacking HppA or HppH, individually or in combination with ΔcomCDE, were generated and tested. Slight variations were seen in biofilm biomass levels across the strains, but the addition of exogenous CSP had no significant effects (Fig. 11A and C). In contrast, biofilms formed by mutants lacking HppA or HppH were reduced in eDNA levels relative to WT, and these were restored upon application of exogenous CSP (Fig. 11A and B). For biofilms formed by strains lacking ComCDE in addition to HppA or HppH, levels of eDNA were significantly lower than those for WT biofilms, but addition of exogenous CSP had no effect (Fig. 11A and B). It was also noted that the mutant lacking HppH formed biofilms that exhibited diminished numbers of branches/junctions per eDNA structure relative to WT (Fig. S8), suggesting that HppH (but not HppA) may contribute to eDNA architecture. Taken together, these data support the hypothesis that the Hpp system can engage CSP and that via CSP detection, both the ComCDE and Hpp systems can modulate eDNA networks within S. gordonii biofilms.

FIG 11.

Hpp system responds to CSP to modulate eDNA. WT S. gordonii or hpp (with or without comCDE) system mutants were grown at 37°C in YPTG on saliva-coated 24-well plates with or without 10 μg/mL CSP for 5 h. Levels of eDNA stranding (A and B) and biomass (C) were then determined by microscopy or crystal violet staining, respectively. Panel A indicates representative images of eDNA stranding. Data are presented as mean ± standard deviation. *, P < 0.05, or **, P < 0.01, as determined by two-way ANOVA followed by Tukey test; n = 3/4. Scale bars, 50 μm.

DISCUSSION

Advances in fluorescence microscopy techniques have provided novel insights into the architecture of eDNA networks, showing them to form “web-” or “lattice-like” structures across the biofilm (9, 10). However, studies requiring the quantification of eDNA have had to rely on the analysis of soluble eDNA, which is disconnected from this complex eDNA architecture. To address this gap, this study presents use of a high-throughput image analysis tool that enables the visualization and quantification of eDNA networks in situ within biofilms. Furthermore, alongside quantification of eDNA abundance, this imaging platform provides the ability to interrogate the detail of eDNA networks with regard to, for example, eDNA branch length and number (Table S1). Such high-level analysis of eDNA architecture has not previously been possible.

To validate the capacity of this imaging system to both reliably detect eDNA and exhibit sufficient sensitivity to detect differences in abundance, the effects of DNase I and sugars were examined. As predicted, DNase I reduced total eDNA levels in a dose-dependent manner. Nonetheless, some eDNA structures clearly remained following DNase I application. These may represent Z-form eDNA, which accumulates as biofilms mature and is recalcitrant to treatment with DNases (37). Additionally, as the biofilm develops, eDNA matrices can be stabilized by DNA-binding proteins, which in turn may limit access to eDNA structures by DNase enzymes (9, 12). In contrast to the effects of DNase I, the presence of sucrose promoted eDNA production relative to glucose. This correlates with the established role of H₂O₂ in regulating eDNA release by S. gordonii (35, 38). H₂O₂ production is governed by SpxB, which in turn is under the control of a carbon catabolite regulator, CcpA (38). Moreover, these effects on eDNA directly correlated with glucan production. The number of junctions within eDNA networks of sucrose-grown biofilms was significantly diminished in the presence of dextranase or the absence of GtfG. It is possible, therefore, that glucans may stabilize eDNA matrices at points where eDNA branches, serving a role similar to that of DNA-binding proteins (9). This correlates with studies using S. mutans, for which eDNA has also been shown to increase in a glucan-dependent manner within biofilms (25, 39, 40). GtfB expressed by S. mutans acts synergistically with eDNA to promote bacterial adherence to surfaces (41). With several Streptococcus species known to express Gtfs (42), glucan-mediated support of eDNA matrices may represent a common mechanism during biofilm development under conditions permissive for glucan production.

This study also demonstrated that surface-associated nuclease SsnA of S. gordonii could modulate eDNA levels. SsnA has homology to streptococcal wall-anchored nuclease (SWAN) of S. sanguinis, which has been shown to degrade neutrophil extracellular traps (43), but this is the first evidence of a surface-expressed nuclease influencing eDNA levels within biofilms. In the absence of SsnA, S. gordonii biofilms exhibited a greater abundance of eDNA networks, suggesting that SsnA may act directly on the eDNA strands to release or reorganize the networks. However, the impact of SsnA was significantly affected by conditions within the local environment. SsnA is primarily active in the pH range 7 to 10 (data not shown) and thus was rendered largely inactive in the presence of fermentable carbohydrate due to the resultant acidification of the environment from glycolysis. Nuclease enzyme expression has been observed from an array of oral biofilm commensals (32). As such, going forward, it will be interesting to determine the contribution that surface-bound nucleases make to organization of the eDNA matrices found within polymicrobial biofilms of the oral cavity and at other sites and the implications of variations in eDNA architecture for overall biofilm properties.

It has been recognized for some time that the competence (comABCDE) operon can regulate the release of eDNA by S. gordonii and S. mutans (4). During the competence pathway in S. gordonii, pre-CSP (encoded by comC) is a 50-aa polypeptide that is cleaved by ComA to produce the mature 19-aa CSP. Mature CSP is transported out of the cell by the ComAB ABC binding cassette transporter and detected by the two-component system (TCS) ComDE. ComD autophosphorylates upon detection of CSP and phosphorylates its intracellular response regulator, ComE. ComE subsequently modulates expression of the competence-specific alternative σ factor, ComR, which regulates transcription of the competence genes, including murein hydrolase LytF, enabling the bacterial cell to take up DNA from the environment (35, 44). Specific to eDNA release, it has been proposed that detection of CSP induces upregulation of AtlS that, in turn, upregulates expression of SpxB. This results in an increase in the intracellular concentration of H₂O₂, with the resultant oxidative stress ultimately inducing LytF expression and eDNA release (4). The data presented in this study support the role of CSP in eDNA release. Specifically, our image analysis system revealed that the abundance of eDNA within S. gordonii biofilms was significantly diminished in the absence of CSP. Unexpectedly, however, it was also revealed that detection of CSP was not dependent on ComDE. Rather, the data imply that the Hpp system can serve as an alternative system for CSP detection and subsequent induction of downstream gene regulation. Cells lacking ComDE but with an intact Hpp system could respond to exogenous CSP, with a concomitant increase in eDNA abundance. Importantly, this was a specific effect, as no such elevation in eDNA levels was seen using a scrambled CSP. Production of eDNA could not be rescued by the application of exogenous CSP to cells lacking both the ComDE and HppA/H detection apparatus, indicating that the cross talk does not extend beyond these two systems.

As Hpp has been described as a hexaheptapeptide permease system (36), it is yet to be understood how the 19-aa CSP can be detected. It is possible that some form of extracellular interaction causes signal transduction, without requiring full CSP entry to the cell. For example, for bacteria such as Lactococcus lactis, the Opp family proteins have been shown to detect peptides between 4 and 35 aa in length, as the whole peptide does not enter the recognition site of OppA (homologous to HppA in S. gordonii) (36, 45). Alternatively, CSP may be cleaved to a shorter length peptide prior to translocation into the cell via the Hpp system. In S. mutans, the 17-aa peptide ComS is processed at a double tryptophan motif (WW), releasing a 7-aa SigX-inducing peptide (XIP) that is imported into the cell via an Opp system (46). As the mature S. gordonii CSP also possesses a WW motif, it is possible that this peptide may be processed in a similar way for recognition via the Hpp system. Exploring such possibilities will be the focus of future studies.

In summary, by exploiting our high-throughput image analysis tool, this study has provided a more detailed understanding of the factors that can modulate eDNA networks within S. gordonii biofilms. Evidence is provided of the capacity for glucans to stabilize eDNA matrices, while surface-bound nuclease SsnA has been shown to modify these structures under conditions permissive for enzymatic activity. Furthermore, while the role of CSP in inducing eDNA release is confirmed, a more complex regulatory mechanism has been revealed, with cross talk with the Hpp system evident. Extending beyond S. gordonii, a critical feature of this imaging system is its capacity to discriminate between eDNA strands, allowing a detailed quantification of the eDNA architecture in situ within biofilms that has not before been possible. In the studies presented here, it was possible to determine changes in eDNA branch length and junction composition, and while current understanding is not yet sufficient to fully appreciate the biological implications of these modifications, there is clear potential for the high-level interrogation provided by this tool to help advance understanding of biofilm matrices. Going forward, it will be important to develop techniques that overcome or minimize current limitations relating to the need to detect Z-forms of eDNA, alongside B-form eDNA (37), and potential antibody accessibility issues (e.g., due to DNA-binding proteins or matrix accumulation), particularly for mature biofilms. Nonetheless, incorporation of this tool across the field of biofilm research offers the capacity to undertake a detailed assessment of how eDNA networks develop, how these networks contribute to the properties of the biofilm, and how this can be modulated. Such opportunities should significantly advance attempts to disrupt eDNA matrices within biofilms for therapeutic benefit, including oral biofilms.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

Bacterial strains utilized in this study are listed in Table 2. S. gordonii wild-type and isogenic mutants were routinely cultured in brain heart infusion broth (Lab Neogen) supplemented with 0.5% (wt/vol) yeast extract (BD; BHY) under stationary conditions for 16 h in a candle jar at 37°C. As needed, broth cultures were supplemented with 100 μg/mL spectinomycin (Sp), 1.5 to 5 μg/mL erythromycin (Ery), or 250 μg/mL kanamycin (Kan). A defined medium (YPT) was used for eDNA secretion studies comprising 20 mM NaH₂PO₄ (pH 7), 1× yeast nitrogen base (Difco), and 0.1% (wt/vol) Bacto tryptone with or without supplementation with 0.2% (wt/vol) glucose or sucrose (47).

TABLE 2.

Strains used in this study

Identifier	Strain	Relevant characteristic(s)	Reference
UB1507	DL1 (Challis)	Parental strain	57
UB653	ΔgtfG	gtfG::aad9	58
UB2660	ΔcomC	comC::aad9	44
UB2661	ΔcomDE	comCD::aad9	This study
UB2347	ΔcomCDE	comCDE::aad9	44
UB2975	ΔcomR1R2	comR1::aad9 comR2::ermAM	This study
UB2953	ΔhppA	hppA::ermAM	This study
UB2958	ΔhppA ΔcomCDE	hppA::ermAM comCDE::aad9	This study
UB3097	ΔhppH	hppH::aphA3	This study
UB3098	ΔhppH ΔcomCDE	hppH::aphA3 comCDE::aad9	This study
UB2886	ΔssnA	ssnA::aad9	Rostami et al. (unpublished)

Mutagenesis of S. gordonii.

Streptococcus gordonii DL1 (Challis) is predicted to express a 779-amino-acid protein with 76% homology to SWAN, a nuclease in Streptococcus sanguinis capable of modulating the eDNA of neutrophil extracellular traps (43). The gene encoding this protein, designated streptococcal surface nuclease A (ssnA), was deleted by allelic exchange. In brief, flanking regions of ssnA were amplified by PCR using primer pairs SsnA.F1/R1 and SsnA.F2/R2 (Table 3), while the aad9 spectinomycin resistance cassette was amplified from plasmid pFW5 using aad9_SsnAF/R (Table 3) (48). Amplicons were joined by overlapping PCR using primers SsnA.F1/R2 (Table 3), yielding a final amplicon of 1,936 bp. This was transformed into wild-type S. gordonii, and successful mutagenesis was confirmed by sequencing.

TABLE 3.

Primers used in this study

Mutant generated	Primer name	Primer sequence	Function
ΔssnA	SsnA.F1	TTTTATCAGAAATTGATTG	Amplify 484-bp amplicon upstream of ssnA
	SsnA.R1	AAAGTTCTCCTTTTCCTA
	SsnA.F2	CCTAGAGTAAGCTCTAAACA	Amplify 674-bp amplicon downstream of ssnA
	SsnA.R2	TGTCAAAGCTACCAGTAC
	aad9_SsnAF	AGGAGAACTTTATGAATACATACGAACAAATTAATA	Amplify 782-bp aad9 cassette from pFW5 with overlaps for ssnA flanking regions
	aad9_SsnAR	GCTTACTCTCTAGGTTATAATTTTTTTAATCTGTTATTTAA
ΔcomDE	ComCD.F1	CGACTCAGTCGTTTTACGAAAG	Amplify 448-bp amplicon upstream of comDE
	ComDE.R1	GGAGATTGAAATGATATTTACAATGGATCCGACAAAG
	ComDE.F1	TTACAATGGATCCGACAAAGCGAGATAAACTGG	Amplify 619-bp amplicon downstream of comDE
	ComCDE.R2	CTACTTCGCGGATATTGGC
	ComDE_aad9F	GGAGATATTTTTTTGAATACATACGAACAAATT	Amplify 1,100-bp aad9 cassette from pFW5 with overlaps for comDE flanking regions
	ComDE_aad9R	GTTAGAGGATTTTAATATTAAAAAAATTAGACAATAAAT
ΔcomR1	ComR1.F1	GATATTCCAGGATCCTGCTG	Amplify 586-bp amplicon upstream of comR1
	ComR1.R1	TATGTATTCATTGACTAGTCCTTTCTTTTTG
	ComR1.F2	AAAAAATTATAAAAAGAAGGGAGAGGCAATC	Amplify 1,075-bp amplicon downstream of comR1
	ComR1.R2	CCTCAGCGTCAGTTACAGAC
	aad9.comR1F	GACTAGTCAATGAATACATACGAACAAATTAATA	Amplify 770-bp aad9 cassette from pFW5 with overlaps for comR1 flanking regions
	add9.comR1R	CCTTCTTTTTATAATTTTTTTAATCTGTTATTTAA
ΔcomR2	ComR2.F1	TCCAGGTGCATATAATCCAC	Amplify 840-bp amplicon upstream of comR2
	ComR2.R1	ATTTTTGTTCATTGACTAGTCCTTTCTTTTTG
	ComR2.F2	GGAGGAAATAAAAAGAAGGGAGAGGCAATC	Amplify 1,075-bp amplicon upstream of comR2
	ComR2.R2	CCTCAGCGTCAGTTACAGAC
	ermAM.comR2F	ACTAGTCAATGAACAAAAATATAAAATATTCTCAAAAC	Amplify 755-bp ermAM cassette from pVA838 with overlaps for comR2 flanking regions
	ermAM.comR2R	CCCTTCTTTTTATTTCCTCCCGTTAAATAATAG
ΔhppA	HppA.F1	CAACAATCCAGACCAATACTC	Amplify 953-bp amplicon upstream of hppA
	HppA.R1	GAAATGGAGAATATACGATGAACAAAAA
	HppA.F2	CGGGAGGAAATAACCAATCATTAGAACTTTC	Amplify 932-bp amplicon downstream of hppA
	HppA.R2	CCATCCATGCTTGTTAGC
	ermAM.hppAF	AATATACGATGAACAAAAATATAAAATATTCTC	Amplify 753-bp ermAM cassette from pVA838 with overlaps for hppA flanking regions
	ermAM.hppAR	TGATTGGTTATTTCCTCCCGTTAAATA
ΔhppH	HppH.F1	CCCGATTCACTTAGATCTTC	Amplify 901-bp amplicon upstream of hppH
	HppH.R1	CATTTTAGCCATGAAATACTCCTTTCAAAATA
	HppH.F2	ATTGTTTTAGCAATTACCCTAACGAGGAGG	Amplify 906-bp amplicon upstream of hppH
	HppH.R2	GATACTTGTCGGGTCAGTAGC
	aphA3.hppHF	AGTATTTCATGGCTAAAATGAGAATATCACC	Amplify 813-bp aphA3 cassette from pDL276 with overlaps for hppH flanking regions
	aphA3.hppHR	AGGGTAATTGCTAAAACAATTCATCCAGTAAAATA

A similar allelic exchange approach was used to generate a ΔcomDE mutant using primer pairs ComCD.F1/ComDE.R1 and ComDE.F1/ComCDE.R2 to amplify the upstream (884-bp) and downstream (619-bp) flanking regions, respectively, and primers ComDE_aad9F/R to amplify aad9 from pFW5 (Table 3). Likewise, a ΔcomR1/R2 mutant was generated using primer pairs ComR1.F1/R1 with ComR1.F2/R2 or ComR2.F1/R1 with ComR2.F2/R2 to amplify the flanking regions of comR1 or comR2, respectively (Table 3). These were joined to the aad9 cassette from pFW5 (48) or the ermAM erythromycin resistance cassette from plasmid pVA838 (49) using primers aad9.comR1F/R or ermAM.comR2.F/R, respectively (Table 3). The hppA gene was inactivated by allelic exchange with ermAM using primers hppA.F1/R1, hppA.F2/R2, and ermAM.hppAF/R (Table 3). The hppH gene was inactivated by allelic exchange with the aphA3 kanamycin resistance cassette from plasmid pDL276 (50) using primers hppH.F1/R1, hppH.F2/R2, and aphA3.hppHF/R (Table 3). Final amplicons were transformed into wild-type S. gordonii. Those for hppA and hppH were additionally transformed into S. gordonii ΔcomCDE (44).

Preparation of saliva.

Unstimulated whole saliva was collected on ice and pooled from a minimum of 5 healthy adult donors who provided written consent (approved by the National Research Ethics Committee Central Oxford C; 08/H606/87). Pooled saliva was treated with 2.5 mM dithiothreitol (DTT), incubated on ice for 10 min, and centrifuged at 10,000 × g for 10 min to sediment mucins and bacteria. The supernatant was transferred to sterile plasticware, diluted to 10% with distilled water (dH₂O), and sterilized through a 0.45-μm filter. Saliva-coated plates were assessed for DNase activity, and the levels were found to be negligible (see Fig. S9 in the supplemental material).

Biofilm formation.

Black, clear-bottom 24-well plates (Sensoplate; Greiner Bio-one) were incubated with 10% saliva (500 μL) for 16 h at 4°C. Overnight broth cultures of S. gordonii were harvested (5,000 × g, 7 min) and resuspended to an optical density at 600 nm (OD₆₀₀) of 0.25 in YPT-glucose (YPTG; equivalent to approximately 2.5 × 10⁶ CFU/mL). Saliva was aspirated from the plates, wells were inoculated with 500 μL bacterial suspension, and plates were incubated in a humid environment at 37°C under gentle agitation (50 rpm) for up to 24 h. Following incubation, nonadherent cells were aspirated, and the biofilms were washed twice with YPT and either fixed with 4% (wt/vol) paraformaldehyde (PFA) for 16 h at 4°C for microscopy or resuspended in phosphate-buffered saline (PBS) for alternative applications. For some studies, bacterial suspensions were treated with dextranase (10 μg/mL; Sigma-Aldrich), DNase I (10 to 25 μg/mL; Sigma-Aldrich), or competence-stimulating peptide (CSP; DVRSNKIRLWWENIFFNKK; 10 μg/mL; GenicBio) following inoculation of the plates. To measure glucan levels within the biofilm, Alexa Fluor 647-conjugated dextran (1 μM; ThermoFisher Scientific) was applied alongside the bacterial suspension. Following incubation (5 h), wells were washed twice with YPT and fluorescence levels (excitation [ex]/emission [em]: 650/668, respectively) were measured with a plate reader (Infinite F200 Pro; Tecan). For assessment of biomass, biofilms were stained with 0.5% (wt/vol) crystal violet and washed with PBS to remove excess stain, and then biomass was quantified by release of stain using 10% (vol/vol) glacial acetic acid and measurement of A₅₉₅.

Soluble eDNA extraction and quantification.

Biofilms from quadruplicate wells were collected into PBS, and the soluble fraction was recovered following centrifugation. Fractions were treated at 37°C for 1 h with proteinase K (5 μg/mL; Sigma-Aldrich), and then the eDNA was extracted using phenol-chloroform-isoamyl alcohol (25:24:1). The aqueous phase was collected, mixed with 3 M sodium acetate and isopropanol, and incubated for 1 h at 20°C. DNA was precipitated and resuspended in dH₂O. DNA concentration and quality were then assessed by measurement of A₂₆₀/A₂₈₀.

High-throughput eDNA image capture and analysis.

Following PFA fixation of biofilms, 2% (wt/vol) bovine serum albumin (Sigma-Aldrich), mouse anti-double-stranded DNA (anti-dsDNA) antibody (ab27156; Abcam; 1:1,000), and Alexa Fluor 594-conjugated secondary antibody (1:1,000) were applied sequentially for 45 min each. When required, S. gordonii cells were additionally stained with TO-PRO-3 (1:1,000 dilution; ThermoFisher Scientific) for 15 min. A ×20 magnification lens (HC PL APO 20×/0.75 CS2) on a widefield Leica DMi600 microscope (Leica) coupled to a Photometrics Prime 95B complementary metal oxide semiconductor (cMOS) camera (1,200 × 1,200, 11-μm pixels, 8 bit; Photometrics) was employed to capture eDNA images using Leica acquisition software (LASX; Leica). eDNA structures were visualized using a cube consisting of a 560/40-nm excitation filter, 595-nm long-pass (LP) dichroic filter, and a 645/76-nm emission filter at an exposure time of up to 100 ms. Positions within each well were defined automatically using a custom-made MATLAB (MathWorks) program which generated xyz positions to be used within the ‘mark and find’ function of LASX, facilitating the acquisition of at least 6 images per well. Each image covered an area of 660 by 660 μm, and all images were taken in the same position in each well. All images were acquired as 10-μm Z-stacks (13 slices by 0.8-μm steps) to ensure that images of eDNA at the optimum focus level were taken. Glucans within biofilm were visualized in a similar manner using a cube consisting of a 620/60-nm excitation filter, 660-nm LP dichroic filter, and 700/38-nm emission filter.

Quantification of eDNA networks was performed using the Wolfson Bioimaging Facility modular image analysis Fiji plugin, MIA (51–53). Initially, eDNA was segmented from fluorescence images using two-dimensional (2D) ridge detection (54, 55). Small gaps between proximal eDNA ends were then bridged, subject to user-defined alignment filters (end-end distance and maximum angular difference). Finally, length and branching metrics for the eDNA structures were obtained using the Analyze Skeleton plugin (56). Structural composition and abundance of eDNA were then assessed using Excel software (Microsoft).

Statistical analyses.

Data were processed utilizing Excel software (Microsoft), and analyses were performed using Prism (GraphPad Software, CA, USA). All experiments were performed at least in triplicate, unless otherwise stated, and data were analyzed using Student’s t test (when comparing two groups) or general linear model (GLM) followed by one-way analysis of variance (ANOVA) and Tukey test (when comparing three or more groups).

Data availability.

All experimental data associated with this work are openly available at the University of Bristol data repository, data.bris, at https://doi.org/10.5523/bris.3v9229rx5z79w1zxnwphwzh4mf. The modular image analysis macro is available at https://doi.org/10.5281/zenodo.3401275. The 2D ridge detection macro is available at https://doi.org/10.5281/zenodo.845874.