The extracellular DNA lattice of bacterial biofilms is structurally related to Holliday junction recombination intermediates

Significance

Most chronic and recurrent bacterial infections are the result of biofilms. Extracellular DNA (eDNA) is a ubiquitous and pivotal structural component of biofilms that protects the resident bacteria from the host immune system and antimicrobial agents. It is of the highest priority to characterize the structure of the eDNA to understand the development of bacterial biofilm communities. Here, we employed the prototypic Holliday junction-specific (HJ) DNA-binding protein RuvA and demonstrated that eDNA within biofilms formed by 3 human pathogens, uropathogenic Escherichia coli, nontypeable Haemophilus influenzae, and Staphylococcus epidermidis was structurally related to HJ recombination intermediates. We further demonstrated that this HJ-like structure was critical to the structural and mechanical integrity of the bacterial biofilm matrix.

Abstract

Extracellular DNA (eDNA) is a critical component of the extracellular matrix of bacterial biofilms that protects the resident bacteria from environmental hazards, which includes imparting significantly greater resistance to antibiotics and host immune effectors. eDNA is organized into a lattice-like structure, stabilized by the DNABII family of proteins, known to have high affinity and specificity for Holliday junctions (HJs). Accordingly, we demonstrated that the branched eDNA structures present within the biofilms formed by NTHI in the middle ear of the chinchilla in an experimental otitis media model, and in sputum samples recovered from cystic fibrosis patients that contain multiple mixed bacterial species, possess an HJ-like configuration. Next, we showed that the prototypic Escherichia coli HJ-specific DNA-binding protein RuvA could be functionally exchanged for DNABII proteins in the stabilization of biofilms formed by 3 diverse human pathogens, uropathogenic E. coli, nontypeable Haemophilus influenzae, and Staphylococcus epidermidis. Importantly, while replacement of DNABII proteins within the NTHI biofilm matrix with RuvA was shown to retain similar mechanical properties when compared to the control NTHI biofilm structure, we also demonstrated that biofilm eDNA matrices stabilized by RuvA could be subsequently undermined upon addition of the HJ resolvase complex, RuvABC, which resulted in significant biofilm disruption. Collectively, our data suggested that nature has recapitulated a functional equivalent of the HJ recombination intermediate to maintain the structural integrity of bacterial biofilms.

Most bacteria in natural ecosystems prefer a biofilm lifestyle. Biofilm bacteria are encased within a self-produced extracellular matrix (extracellular polymeric substances, or EPS) comprised of extracellular DNA (eDNA), proteins, lipids, and exopolysaccharides (1). The biofilm EPS provides structural integrity; protects resident bacteria against physical, chemical, and environmental stresses that includes host effectors and antimicrobial therapies; and affects gene regulation and nutrient adsorption (reviewed in ref. 2). Hence, it is of utmost importance to characterize not only the EPS components but also their subsequent structure to gain insight into the development of bacterial biofilm communities and, consequently, for pathogenic biofilms, identify the means to undermine them.

eDNA is a key structural component of the EPS and therefore an attractive target for the control of bacterial biofilms. Although the importance of eDNA in the biofilm matrix has been established, the structure of the eDNA itself has not been well characterized. We have previously shown that eDNA in biofilms formed by nontypeable Haemophilus influenzae (NTHI) within a chinchilla middle ear (3) and by Pseudomonas aeruginosa in a murine lung model (4), as well as biofilms in pediatric sputum and otorrhea samples that were culture-positive for multiple mixed bacterial species (5–7), was present in a lattice structure. In addition, we have revealed that the DNABII family of proteins (integration host factor, IHF, and histone-like protein, HU) bind to and stabilize the eDNA lattice structure and are fundamental to the structural stability of bacterial biofilms (5, 7–14).

DNABII proteins that are localized at the vertices of the eDNA lattice within bacterial biofilms (5–8) have high affinity for branched DNA structures which include Holliday junction (HJ) DNA (15, 16). HJs are single-strand cross-over intermediates of homologous recombination and are common across both eukaryotes and prokaryotes (17). HJs appear as cross-like or cruciform structures with 4 double-stranded DNA (dsDNA) arms. In most eubacteria, the resolution of homologous recombination occurs through the association of HJ DNA with RuvA, RuvB, and RuvC, where RuvA binds to HJ DNA with high affinity in a structure-specific, but sequence-independent, manner (18). RuvA then recruits RuvB to the HJ, and the RuvAB complex drives translocation of the junction that expands the heteroduplex region in an adenosine 5′-triphosphate (ATP)-dependent fashion (19). Finally, the endonuclease RuvC binds the RuvAB complex, which results in cleavage of HJ DNA and resolution to yield 2 nicked duplexes (20). RusA, a resolvase of lambdoid phage origin, binds HJ in a sequence-independent manner and cleaves the phosphodiester bond 5′ of CC dinucleotides to resolve HJ into nicked duplexes (21).

Because HJ DNA is necessarily bent, it serves as an excellent substrate for the DNABII family that bind bent DNA with high affinity (15, 16). We therefore hypothesized that the structure of eDNA at the vertices was comprised of HJs. To test this hypothesis, we employed antibodies that are highly specific for HJ DNA as well as proteins that bind to and resolve HJ DNA. First, we demonstrated that the lattice structure found within the EPS of biofilms formed in vivo by NTHI (in the chinchilla middle ear during experimental otitis media) and in polymicrobial sputum samples recovered from cystic fibrosis (CF) patients was recognized by the these highly specific HJ-directed antibodies. Further, we took advantage of the proteins involved in the resolution of HJ DNA, RuvABC complex, and RusA to demonstrate that the HJ DNA-binding protein RuvA functionally complemented DNABII proteins within the EPS and stabilized biofilms formed by uropathogenic Escherichia coli (UPEC), NTHI, and Staphylococcus epidermidis in vitro. We further showed that NTHI biofilms stabilized by RuvA were biophysically indistinguishable from the control NTHI biofilms as measured by mechanical axial indentation. Finally, we also showed that the HJ resolvases RuvABC and RusA efficiently disrupted biofilms and directly targeted these HJ structures within the EPS of NTHI biofilms to inhibit the formation of the eDNA lattice structure. Collectively, our data suggested that eDNA lattice within bacterial biofilms is structurally related to HJ structures and is critical for the stability of the bacterial biofilm matrix.

Results

Bacterial Biofilms Formed by NTHI within the Middle Ear of the Chinchilla and Polymicrobial Sputum Samples Recovered from CF Patients Contained HJ-like DNA Structure.

We previously demonstrated that eDNA within the biofilm EPS formed by multiple single (3, 5, 8) and mixed bacterial species (6, 7) is organized into an interwoven web-like structure that is stabilized by DNABII proteins positioned at the vertices of each of the crossed strands of eDNA. Since these DNABII proteins have a high affinity for HJ DNA (K_d ∼ nanomolar) (15, 16), we hypothesized that these branched structures were related to HJ recombination intermediates. Toward this goal, we assessed whether the HJ-like structure was found within bacterial biofilms that had formed in vivo with a monoclonal antibody specific for cruciform DNA that exclusively binds to the elbow region of an HJ DNA structure (22). Immunohistochemistry analysis of middle ear sections from chinchilla infected with NTHI and sputum solids from CF patients that contained multiple mixed bacterial species revealed a complex lattice-like eDNA structure as indicated in green, with punctate labeling for cruciform DNA in white, at the majority of the crossed strands of the eDNA (Fig. 1). Given the specificity of the monoclonal antibody against cruciform DNA (SI Appendix, Fig. S1) (22), these data confirmed the presence of HJ DNA within the EPS of single and multispecies biofilms in vivo.

Fig. 1.

Labeling of dsDNA and cruciform DNA within the chinchilla middle ear infected with NTHI and within sputum collected from a CF patient. (A) Representative images of an optimal cutting temperature (OCT)-embedded section of the chinchilla middle ear infected with NTHI labeled for the presence of dsDNA (green) and cruciform DNA (white). (B) Representative images of an OCT-embedded section of sputum sample recovered from a CF patient. (Scale bars, 10 μm.) Note the complex lattice structure of eDNA and the punctate labeling of cruciform DNA at the vertices (yellow arrows) formed by the crossed strands of eDNA.

The Prototypic HJ DNA Binding Protein RuvA Compensated for the Removal of DNABII Proteins in Structural Stabilization of UPEC, NTHI, and S. epidermidis Biofilms.

To further confirm the presence of HJ-like structure within the EPS of bacterial biofilms, we employed 3 opportunistic pathogens, UPEC, NTHI, and S. epidermidis, all of which are known to persist in a biofilm lifestyle that is disrupted upon depletion of DNABII proteins. We depleted the DNABII proteins within the EPS of biofilms that were established in vitro by the addition of a hyperimmune polyclonal antibody directed against E. coli IHF (α-IHF, which recognizes both IHF and HU with variable avidities [SI Appendix, Fig. S1]) and simultaneously supplemented with purified recombinant E. coli RuvA, the prototypical HJ binding protein in bacteria, to determine if RuvA could functionally replace DNABII proteins to stabilize the biofilm EPS. Biofilms were then stained with LIVE/DEAD, visualized via confocal laser scanning microscopy (CLSM), and total biomass and average thickness were quantified by COMSTAT analysis (23). It was evident that α-IHF–mediated disruption of biofilms formed by each bacterial species, which included UPEC (Fig. 2 A and B), NTHI (Fig. 2C), and S. epidermidis (Fig. 2D), was prevented by the addition of RuvA. Addition of H-NS, a nonspecific DNA-binding protein, was unable to compensate for the loss of DNABII proteins within the biofilm matrix and thus served as a negative control (Fig. 2). This result was consistent with our previous findings that H-NS is not required for the structural integrity of biofilms formed by UPEC and NTHI (10, 11).

Fig. 2.

HJ-specific DNA-binding protein RuvA stabilized bacterial biofilm structure even when DNABII proteins were depleted. (A) Representative images of a UPEC biofilm. (B) Sixteen-hour UPEC and (C) 16-h NTHI biofilms were incubated with the indicated protein and/or antibody for 24 h. (D) Twenty-four-hour S. epidermidis biofilm was incubated with the indicated protein and/or antibody for 16 h. Biofilms were stained with LIVE/DEAD stain and visualized via CLSM. Images were analyzed by COMSTAT to calculate average thickness and biomass. Percent change in biomass compared to control was plotted. Bars represent the SEM. Statistical significance compared to control was assessed with unpaired t tests; *P < 0.05, **P < 0.01, ***P < 0.001; ns, not significant. Note that RuvA prevented α-IHF–mediated disruption of the biofilm structure of UPEC, NTHI, and S. epidermidis and thus confirmed the presence of HJ DNA within the biofilm matrix.

Next, we used immunofluorescence to detect DNABII proteins and RuvA within bacterial biofilms and observed the depletion of DNABII proteins within the extracellular matrix of biofilms formed by UPEC upon treatment with α-IHF in the presence of RuvA (Fig. 3 A and B) and the concomitant incorporation of RuvA within the biofilm matrix (Fig. 3 C and D). While RuvA labeling was observed throughout the depth of the biofilms (see orthogonal projections in the bottom row of Fig. 3 C and D) treated with naive or α-IHF immunoglobulin G (IgG), RuvA was much more densely accumulated at the bottommost portions of those treated with α-IHF. We are further investigating the mechanism(s) of this spaciotemporal labeling pattern, as α-RuvA antibody was confirmed to neither adhere to the substratum nor to planktonic UPEC or NTHI cells (SI Appendix, Fig. S2). However, these results suggested that the observed distribution of RuvA was likely characteristic of UPEC biofilms. The relative abundance of IHF and RuvA within the biofilm EPS was determined by the ratio of the protein (α-IHF/α-RuvA–labeled) to total DNA (DAPI) and revealed a statistically significant decrease in DNABII proteins, which corresponded with a statistically significant increase in RuvA compared to the control (indicated by naive serum + RuvA; Fig. 3E). Given the skewed distribution of the fluorescence signal at the bottommost portion of the biofilm, we determined that even after digitally removing the 3-μm section from the bottom we still observed a statistically significant decrease in DNABII proteins within the remainder of the biofilm, an observation that corresponded with a correlating and statistically significant increase in RuvA compared to the control (SI Appendix, Fig. S3). These results suggested that the observed distribution of RuvA was characteristic of UPEC biofilms. The specificities of α-IHF and α-RuvA were determined by Western blot analysis and were confirmed to be highly specific for their target protein (SI Appendix, Fig. S1). We and others have shown that RuvA specifically binds to HJ DNA with high affinity (SI Appendix, Fig. S4A) (24). Given the high affinity and specificity of RuvA to HJ DNA, these results suggested that RuvA compensated for the loss of DNABII proteins and thus stabilized the eDNA structure by selectively binding to HJ DNA structures that were vacated by DNABII proteins as a result of DNABII protein depletion with α-IHF.

Fig. 3.

RuvA incorporated into the bacterial biofilm matrix when DNABII proteins were depleted. UPEC biofilms were formed for 16 h then incubated with naive IgG (1,000 nM) and RuvA (450 nM) (A and C) or α-IHF (1,000 nM) and RuvA (450 nM) (B and D) for 24 h. Immunofluorescence was performed on unfixed biofilms wherein the biofilms were incubated with either α-IHF antiserum (1:200 dilution) (A and B) or α-RuvA antiserum (1:200 dilution) (C and D) then incubated with goat anti-rabbit IgG conjugated to Alexa Fluor 594. eDNA was stained with DAPI (gray). Biofilms were visualized via CLSM. Images represent the top and side view of biofilms. (E) The relative abundance of DNABII proteins or RuvA was determined by the ratio of the respective protein (α-IHF/α-RuvA labeled) to total DNA (DAPI). Statistical significance was assessed with paired t tests; *P < 0.05. Note the depletion of DNABII proteins and the concomitant incorporation of RuvA within the UPEC biofilm matrix.

NTHI Biofilms Stabilized by DNABII Proteins and DNABII-Depleted Biofilms Stabilized by RuvA Exhibited Similar Mechanical Properties.

Rheological analysis was performed on NTHI biofilms to determine how DNABII depletion, and complementation of this depletion by addition of RuvA, induced any changes to the bulk biofilm mechanical properties. Axial mechanical indentation was performed on control (naive IgG) and DNABII-depleted biofilms (α-IHF) in the presence or absence of RuvA. Indentation has been commonly used to assess the impact of EPS components on biofilm mechanical stability (25, 26). An 8-mm geometry was lowered onto the biofilm, and the force required to compress the biofilm was determined. NTHI biofilms displayed a characteristic “J-shaped” stress–strain response (Fig. 4A), which indicated that as the biofilms were compressed they progressively became stiffer, which is typical of viscoelastic biological materials (27). It was evident from the differences in the stress–strain curves particularly at the lower strains that the different treatments influenced the stiffness of NTHI biofilms (Fig. 4A). To quantify these differences, the Young’s modulus (E) was calculated from the lower linear portion of the curve (Fig. 4 A, Inset) using Eq. 1. The Young’s modulus is a measurement of how stiff a material is, that is, how much a material deforms (measured as strain) in response to an applied normal force (i.e., force that is applied perpendicular to a material) (28). The Young’s modulus of DNABII-depleted biofilms was significantly reduced compared to control (naive IgG; no DNABII depletion) (Fig. 4B). This result suggested that the DNABII depletion, and subsequent disruption of the eDNA lattice network, resulted in NTHI biofilms that were mechanically less rigid than control biofilms. However, DNABII-depleted biofilms that had been complemented with RuvA exhibited a Young’s modulus similar to the control (Fig. 4B). These data suggested that complementation with RuvA mechanically compensated for the loss of DNABII proteins and restored biofilms to their normal stiffer phenotype.

Fig. 4.

RuvA mechanically compensated for the depletion of DNABII proteins to structurally stabilize NTHI biofilms. (A) NTHI biofilms were formed for 16 h and incubated with either naive IgG (1,000 nM), α-IHF (1,000 nM), naive IgG and RuvA (450 nM), or α-IHF and RuvA (450 nM) for a further 24 h. Mechanical indentation analysis was depicted as stress-strain curves. (Inset) A closer view of the 0 to 40% strain (γ) portion of the curve. (B) Young’s modulus calculated from the lower linear portion of the curve, depicted in A, Inset. Data presented as mean ± SD; n = 4. Significance determined using a one-way ANOVA; *P < 0.05, ***P < 0.001; ns, not significant. Note that the Young’s modulus of DNABII-depleted biofilms stabilized by RuvA was comparable to control biofilms and thus confirmed that RuvA functionally and mechanically complemented for the depletion of the DNABII proteins within the NTHI biofilm EPS.

Bacterial Biofilm Matrix Stabilized by RuvA Was Disrupted upon Treatment with HJ- Specific Endonuclease Complex RuvABC.

Since RuvA readily and effectively replaced DNABII proteins to maintain the structural stability of biofilms formed by UPEC, NTHI, and S. epidermidis, we hypothesized that the biofilm matrix stabilized by RuvA is susceptible to disruption by the HJ-specific endonuclease complex RuvABC. To test this, established UPEC, NTHI, and S. epidermidis biofilms wherein the DNABII proteins had been experimentally replaced with RuvA (Figs. 2 and 3) were further incubated with RuvB and RuvC proteins at a concentration that has no effect on planktonic growth (SI Appendix, Fig. S5), so as to create the RuvABC complex followed by the addition of LIVE/DEAD stain, visualization with CLSM, and quantification with COMSTAT (23). Strikingly, as evident from Fig. 5, the addition of RuvABC complex to biofilms in which the EPS was stabilized by RuvA (indicated by α-IHF IgG + RuvABC) induced a significant reduction in biofilm biomass compared to control biofilms wherein the matrix was stabilized by DNABII proteins (indicated by naive IgG + RuvABC) in UPEC (Fig. 5A), NTHI (Fig. 5B), and S. epidermidis (Fig. 5C). Also, established UPEC and NTHI biofilms wherein the EPS was stabilized by DNABII proteins (no depletion) were only modestly disrupted by the addition of RuvABC (Fig. 5 A and B). With no depletion of DNABII proteins, RuvABC was ineffective at disruption of S. epidermidis biofilms (Fig. 5C). We have previously shown that DNABII proteins are limited in UPEC (11) (e.g., a situation wherein addition of exogenous DNABII proteins partitions bacteria from the planktonic to the biofilm state); however, they are not limited in NTHI (SI Appendix, Fig. S6). Nonetheless, exogenously added DNABII proteins do incorporate within their respective EPSs (10). These data suggested the presence of at least transiently free HJ DNA sites within the EPS of these biofilms, wherein RuvA could be incorporated. This outcome was confirmed by immunofluorescence, which revealed the incorporation of a modest amount of RuvA within the EPS of UPEC biofilms in the presence of naive serum, which does not deplete DNABII proteins (Fig. 3C), and was also in line with the modest disruption of UPEC and NTHI biofilms (Fig. 5 A and B) in the absence of depletion of DNABII proteins. DNABII proteins were not limited in S. epidermidis biofilms (SI Appendix, Fig. S6), which suggested the absence of free HJ within the biofilm EPS, and therefore was consistent with a lack of disruption of S. epidermidis biofilms by RuvABC (Fig. 5C). However, depletion of DNABII proteins with α-IHF allowed more HJ sites to be vacated within the biofilm EPS of UPEC, NTHI, and S. epidermidis, and as a result a significant amount of RuvA was incorporated within UPEC biofilm EPS (Fig. 3). Once RuvA was stably in place, the addition of RuvB and RuvC significantly disrupted UPEC, NTHI, and S. epidermidis biofilms (Fig. 5). These data implied that the observed significant disruption of biofilms by RuvABC was due to the incorporation of additional RuvA on the HJ DNA sites that were vacated by DNABII proteins as a result of depletion with α-IHF. In the absence of the endonuclease RuvC (indicated by naive IgG + RuvAB and α-IHF IgG + RuvAB), no significant disruption was observed in biofilms formed by UPEC, NTHI, and S. epidermidis. The RuvAB complex drives branch migration of the HJ in an ATP-dependent manner (19). This result suggested 3 possibilities: 1) that the HJs were immobile and could be in an antiparallel configuration, 2) that there was insufficient complementarity beyond the HJ, and/or 3) that there are other mitigating factors that impeded branch migration. Further, we confirmed the resolvase activity of the RuvABC complex on synthetic HJ DNA preincubated in the presence and absence of DNABII protein (SI Appendix, Fig. S4B). Collectively, these data indicated that the eDNA lattice structure within these biofilms contained HJ DNA structure that served a critical structural role in the stability of the bacterial biofilm EPS.

Fig. 5.

Disruption of bacterial biofilm structure by the HJ-specific endonuclease complex, RuvABC. (A) UPEC and (B) NTHI biofilms were established for 16 h then incubated with the indicated antibody (1,000 nM) and RuvA (450 nM) for 24 h (total 40 h). Biofilms were incubated with RuvB (1,130 nM) and RuvC (90 nM) in the final 16 h. (C) S. epidermidis biofilm was established for 24 h then incubated with the indicated protein and/or antibody for 16 h. Biofilms were stained with LIVE/DEAD stain and visualized via CLSM. Images were analyzed by COMSTAT to calculate biomass. Bars represent the SEM. Statistical significance compared to control was assessed with unpaired t tests; *P < 0.05, **P < 0.01, ***P < 0.001; ns, not significant. Note that upon replacement of DNABII proteins with RuvA within the biofilm matrix biofilms were susceptible to HJ-specific endonuclease RuvC that resulted in the statistically significant collapse of UPEC, NTHI, and S. epidermidis biofilm structure, consistent with the critical structural role of the HJ DNA in the stability of the bacterial biofilm extracellular matrix.

Bacterial Biofilms Were Disrupted upon Treatment with Another HJ-Specific Resolvase, RusA.

To further validate the presence of HJ DNA structure within the bacterial biofilm EPS, biofilms formed by UPEC, NTHI, or S. epidermidis were incubated with varying concentrations of RusA, an HJ-specific endonuclease. Biofilms were then stained with LIVE/DEAD, visualized via CLSM, and quantified by COMSTAT analysis (23) to determine total biofilm biomass and average thickness. The addition of RusA at concentrations that have no effect on planktonic growth (SI Appendix, Fig. S5) destabilized the biofilm matrix and induced a significant dose-dependent reduction in UPEC (Fig. 6A), NTHI (Fig. 6B), and S. epidermidis (Fig. 6C) biofilm biomass compared to control. Although RusA bound with very high affinity to HJ and Y-DNA (SI Appendix, Fig. S7A), it only selectively cleaved HJ-DNA to nicked duplex DNA (SI Appendix, Fig. S7B). Also, RusA efficiently cleaved synthetic HJ prebound to HU (SI Appendix, Fig. S7C). Given the cleavage specificity of RusA for HJ DNA, these data further confirmed the presence of HJ DNA within the biofilm EPS and demonstrated that it was crucial for the structural integrity of bacterial biofilms.

Fig. 6.

Dose-dependent disruption of bacterial biofilms by the HJ-specific resolvase, RusA. Twenty-four-hour (A) UPEC, (B) NTHI, and (C) S. epidermidis were incubated with varied concentrations of RusA (1, 5, and 10 μg/mL for UPEC and NTHI; 10 and 20 μg/mL for S. epidermidis) for 16 h. Biofilms were stained with LIVE/DEAD stain and visualized via CLSM. Images were analyzed by COMSTAT to calculate biomass. Bars represent the SEM. Statistical significance compared to control was assessed with unpaired t tests; *P < 0.05, **P < 0.01, ****P < 0.0001; ns, not significant. Note that RusA disrupted UPEC, NTHI, and S. epidermidis biofilms in a statistically significant, dose-dependent manner.

RuvABC and RusA Targeted HJ DNA within the Biofilm Extracellular Matrix and Prevented the Formation of the eDNA Lattice-Like Network within an NTHI Biofilm.

Since treatment of biofilms formed by multiple bacteria with HJ-specific endonucleases disrupted biofilms, we reasoned that these endonucleases specifically targeted HJ DNA structure within the biofilm EPS to mediate biofilm disruption. To demonstrate this, immunofluorescence was used to visualize eDNA and evaluate the effect of RuvABC and RusA on the eDNA lattice structure of NTHI biofilms (used here as a representative model bacterial biofilm) formed in the absence or presence of RuvABC or RusA. Unfixed NTHI biofilms were then labeled with a monoclonal antibody against dsDNA to visualize the eDNA. While the eDNA was organized into a complex web-like structure in the absence of HJ-specific endonucleases (indicated by control in Fig. 7), the eDNA lattice structure was radically diminished with a few eDNA strands in the presence of RuvABC or RusA (Fig. 7). Upon addition of higher concentrations of RusA, either at initiation of biofilms or when added to established biofilms, a highly diminished lattice structure with fewer eDNA strands was observed (SI Appendix, Fig. S8). These results suggested that the remaining eDNA strands were either inaccessible to RusA or perhaps that other branched structures of eDNA were present within the biofilm matrix that could not be cleaved by RusA. In addition, biofilms were probed with a monoclonal antibody specific for cruciform DNA (SI Appendix, Fig. S1) (22) to directly visualize HJs within the EPS of biofilms formed by NTHI. In the absence of RusA, HJs were particularly visible in the lower, denser part of the biofilm as evidenced by the relative distribution of the yellow fluorescence within the biofilm matrix (Fig. 7E), whereas no fluorescence signal was detected when biofilms were incubated with naive IgG (Fig. 7D). The addition of RusA to biofilms at initiation of the biofilm significantly decreased the observed yellow fluorescence (Fig. 7F). Further, in the presence of RusA, the relative abundance of HJ DNA as determined by the ratio of the HJ DNA (α-cruciform–labeled) to the bacteria (FilmTracer) revealed a statistically significant decrease in the amount of HJ DNA within the biofilm matrix compared to the control (Fig. 7G). Next, we colocalized cruciform DNA and dsDNA within the EPS of established in vitro-formed biofilms in the absence (SI Appendix, Fig. S9A) and presence of RusA (SI Appendix, Fig. S9B) or RuvABC (SI Appendix, Fig. S9C) and observed a complex lattice-like eDNA structure in the control (as indicated in green), with punctate labeling for cruciform DNA (in white) at the majority of the crossed strands of eDNA. Further, in the presence of RusA or RuvABC, the eDNA lattice structure and the cruciform DNA were significantly reduced as compared to the control (SI Appendix, Fig. S9). Finally, we wanted to determine if the HJ structures exclusively colocalized with DNABII as our hypothesis suggests. Our hypothesis was supported by the fact that we were unable to colocalize the DNABII proteins and cruciform DNA within the NTHI biofilm EPS (SI Appendix, Fig. S10A) with specific antibodies when these were added simultaneously, a result that suggested that one antibody was blocking the other from also finding its target due to the shared physical location of their specific binding sites. Thereby, we used the alternative approach wherein we added the specific antibodies sequentially to determine if the failed ability to colocalize the DNABII proteins and the HJs at the same time was perhaps due to occlusion of each antibody to the same physical structure (i.e., DNABII-bound HJs). To demonstrate this likely occlusion, we sequentially labeled first the DNABII proteins then the cruciform DNA (and vice versa) and observed that the labeling of either DNABII proteins or cruciform DNA occluded the labeling of the other (SI Appendix, Fig. S10 B and C). Addition of H-NS, a nonspecific DNA-binding protein, had no effect on the labeling of cruciform DNA (SI Appendix, Fig. S10 D and E). These data provided additional support for our hypothesis that the DNABII proteins and cruciform DNA likely colocalize within the NTHI biofilm matrix. Collectively, these data further proved the presence and critical significance of HJ DNA to the stability of the bacterial biofilm EPS.

Fig. 7.

HJ-specific resolvases targeted HJ DNA within the NTHI biofilm extracellular matrix to disrupt the lattice-like eDNA network. NTHI biofilm growth was initiated in the absence (A) or presence (B) of RusA or (C) RuvABC for 16 h. Unfixed biofilms were incubated with α-dsDNA monoclonal antibody then incubated with goat anti-mouse IgG conjugated to Alexa Fluor 488. (Scale bars, 10 μm.) NTHI biofilm growth was initiated in the absence (D and E) or presence of RusA (10 μg/mL) (F) for 16 h. Unfixed biofilms were incubated with α-cruciform DNA monoclonal antibody then incubated with goat anti-mouse IgG conjugated to Alexa Fluor 488 (yellow). NTHI were stained with FilmTracer FM 4-64 (gray). Biofilms were visualized via CLSM. (G) The relative intensity of cruciform DNA was determined by the ratio of cruciform DNA (yellow) to NTHI (gray). Bars represent the SEM. Statistical significance compared to control was assessed with paired t tests; *P < 0.05. Note the complex web-like structure of eDNA in the control and the loss of this eDNA structure in the presence of RusA or RuvABC. Also, note the distribution of cruciform DNA throughout the biofilm matrix, particularly visible in the lower, denser part of the biofilm within an NTHI biofilm (E) and the loss of cruciform DNA in the presence of RusA (F).

Discussion

Our overarching hypothesis is that in a multispecies biofilm with coaggregating partners, the DNABII proteins in conjunction with eDNA assemble a common nucleoprotein complex that creates an inclusive EPS infrastructure within the means of all eubacteria which is permissive for bacteria to enter into a community biofilm architecture. Multiple human pathogens, which include NTHI, UPEC, Neisseria gonorrhoeae, P. aeruginosa, S. epidermidis, Staphylococcus aureus, Streptocoocus pneumoniae, Enterococcus faecalis, Helicobacter pylori, and Campylobacter jejuni incorporate eDNA into their biofilms (reviewed in ref. 29). Bacteria not only release their own DNA in a multitude of ways but also secrete toxins that induce lysis of host cells by apoptosis and necrosis wherein the released host DNA now facilitates biofilm development (30). Neutrophil extracellular traps, a host defense mechanism wherein neutrophils release nuclear DNA associated with histones and cytoplasmic granules to combat pathogens, also serves as a source of eDNA. In particular, P. aeruginosa exhibits an enhanced biofilm formation in the presence of neutrophils (31), an outcome that suggests that the eDNA within bacterial biofilms is likely comprised of both host-derived and bacterial-derived DNA. eDNA was first shown to be critical for biofilm formation of P. aeruginosa (32) and several other studies revealed its structural role within the biofilm EPS of gram-positive and gram-negative bacteria in natural, industrial, and medical ecosystems (32–35). Now, it is known that eDNA is a common component in bacterial biofilms; however, the structural configuration of eDNA within biofilms has not been well characterized.

A filamentous network of eDNA has been previously described for Reinheimera sp. F8 and Pseudomonas sp. FW1 isolated from freshwater streams (33). A similar structural organization of eDNA is also evident in NTHI, Myxococcus xanthus, E. faecalis, and Streptococcus mutans (3, 36–38) biofilms. We have previously shown that eDNA in single-species biofilms formed by NTHI and P. aeruginosa in a chinchilla experimental otitis media and murine lung infection model, respectively, in pediatric sputum samples that were culture-positive for Burkholderia cenocepacia, P. aeruginosa, and Staphylococci, and in pediatric otorrhea samples that were culture-positive for Haemophilus influenzae, methicillin-resistant S. aureus, S. pneumoniae, Moraxella catarrhalis, and P. aeruginosa is arranged into an interwoven lattice structure that is stabilized by the DNABII family of proteins (4–8).

The DNABII family of proteins condense DNA upon binding and in doing so play a critical role in intracellular bacterial nucleoid structure and function (39). Members of the DNABII protein family exhibit high affinity toward prebent secondary structures of DNA that includes HJ DNA (15, 16). The DNABII family of proteins are also found within the extracellular matrix of various single and multispecies biofilms and serve as lynchpin proteins in stabilization of the lattice-like structure of the eDNA (4–8). The universal conservation of DNABII in eubacteria and the presence of eDNA in bacterial biofilms, combined with the observation of the DNABII protein-stabilized lattice-like arrangement of eDNA in the biofilms formed by multiple bacterial species and under various conditions, indicated the likely universality of this organization of eDNA in bacterial biofilms. Given the preference of DNABII proteins for branched DNA structures that include HJ DNA, and the positioning of DNABII proteins at each of the vertices of the crossed strands of the eDNA, we hypothesized that the eDNA lattice in bacterial biofilms was structurally related to HJ DNA, and that other HJ DNA-binding proteins would provide similar structural integrity. Accordingly, RuvA, the prototypic HJ DNA-binding protein, stabilized the bacterial biofilm structure upon the depletion of the DNABII proteins and thus functionally replaced DNABII proteins within the EPS of bacterial biofilms. Since RuvA exclusively binds to HJ DNA (24), these data strongly implied that HJ DNA was a significant component within the EPS of these biofilms.

Herein, we also analyzed the mechanical properties of NTHI biofilms using axial indentation. All analyzed NTHI biofilms displayed a J-shaped stress–strain response, which has been observed for S. mutans (40, 41) and P. aeruginosa biofilms (26), as well as in mixed biofilms of P. aeruginosa, Pseudomonas fluorescens, Klebsiella pneumoniae, and Stenotrophomonas maltophilia (42) under shear and compression. It therefore appears that this classic J-shape response to applied forces is a common property of bacterial biofilms. Furthermore, the Young’s modulus of the NTHI biofilms determined here is greater than that previously determined for both S. mutans biofilms (20 to 40 kPa; refs. 40 and 41) and for mixed biofilms (0.04 kPa; ref. 42). However, these values are on the same order as those calculated for wild-type P. aeruginosa biofilms (>100 kPa; ref. 26).

How does an HJ configuration figure into the eDNA-dependent EPS of biofilms? HJ, a universal intermediate formed during repair and homologous recombination events, consists of a branched structure with 4 double-helical arms that extend from the center. HJ adopts 2 configurations dependent on the local concentration of cations and interaction with HJ DNA-binding proteins: An open-X form wherein the 4 double-helical arms are extended in a square planar geometry is the preferred configuration at low ionic strength and a stacked-X form wherein the arms coaxially pair and stack into a more compact structure is favored at high ionic strength (reviewed in ref. 43). The DNA strands in the stacked-X configuration align either parallel or antiparallel to each other. While the stacked-X configuration with parallel strands can migrate along the DNA strands (“branch migration”), the junction with antiparallel DNA strands is topologically incapable of branch migration (44). DNABII proteins, IHF, and HU recognize and bind to stacked-X HJ. While HU locks the HJ in stacked-X configuration, IHF induces the junction to adopt the open-X configuration (45–47). RuvA, on the other hand, binds and stabilizes the open-X configuration (48). Although the binding preference of each of these HJ DNA-binding proteins are different, the fact that RuvA complemented the loss of DNABII proteins implied that the eDNA lattice within bacterial biofilms was composed of a structure sufficiently similar to a bona fide HJ that complementation was possible.

The second HJ binding protein we utilized was the resolvase RusA that efficiently cleaved synthetic HJ prebound to HU and also significantly disrupted biofilms formed by UPEC, NTHI, and S. epidermidis in a dose-dependent manner. RusA binds and stabilizes stacked-X HJ (49). In this configuration, the faces of the junction are distinct such that one side of the HJ exhibits minor groove characteristics and the other side exhibits the characteristics of the major groove (50). Hence, the 4 angles in the junction are sterochemically distinct. Several junction-specific endonucleases that include T4 endonuclease VII, yeast endonuclease X2, and Calf thymus junction-specific endonuclease bind to the 120° angle of the HJ DNA on the minor groove side and cleave HJ DNA. DNABII protein HU failed to inhibit T4 endonuclease VII activity and therefore it was proposed that HU bound to the 60° angle of the HJ DNA (15). In our study, RusA was effective at disruption of biofilm EPS wherein the eDNA was stabilized by DNABII proteins. This result suggested that RusA bound to a site on the HJ DNA that was distinct from the site bound by DNABII proteins to mediate cleavage of HJ DNA.

While our data suggested that HJ DNA was present within single- and mixed-species biofilms in vitro and in vivo, there is the potential for variability in the number of HJ DNA sites and endogenous steady-state levels of DNABII that stabilize these sites in each biofilm, which likely contributes to the differences in the efficiency of biofilm disruption by the HJ resolvases. Also, the possibility of variable proportions of the respective HJ topologies, which likely depends on the microbial species from which they are derived, could not be excluded. Since the DNABII family of proteins bind to a variety of other DNA structures which include dsDNA, dsDNA with nicks, gaps and overhangs, single-strand fork, double-strand fork, and 3-way junction with nicks (16), the presence of these specific structures within the EPS of bacterial biofilms cannot be excluded and remains to be investigated.

Finally, while cleavage/removal of these HJ structures was coincident with biofilm disruption, it is unclear why other nucleases fail to likewise disrupt extant biofilms (32). In accordance with our model, DNABII proteins bound to HJs and stabilized the eDNA lattice. Disruption of extant biofilms either by sequestration of the DNABII proteins or competition for the HJs by HJ resolvases demonstrated the importance of these structures. However, independent of these HJ structures, the remaining eDNA enters into a nuclease-resistant state as various DNases prevent bacterial biofilm formation but fail to affect mature biofilms (5, 8, 32). Future work will explore the nature of this nuclease-recalcitrant state and the capacity of the resident bacteria to create a formidable eDNA-dependent extracellular matrix.

Methods

Bacteria Strains.

NTHI strain 86-028NP isolated from the nasopharynx of a child with chronic otitis media at Nationwide Children’s Hospital was used in this study. This strain has been sequenced (51) and well characterized (52). UPEC strain UTI89 was isolated from a patient with cystitis (53). S. epidermidis strain 1618 was originally isolated from a child with serous otitis media in 1987 and has been maintained at low passage number in liquid nitrogen since its isolation.

Visualization of eDNA and Cruciform DNA within Biofilms Formed In Vivo.

Middle-ear sections from chinchilla infected with NTHI strain 86-028NP were prepared as previously described (8). Animal work was performed in accordance with federal, state and institutional guidelines and under protocol #01304AR approved by the Nationwide Children’s Hospital Institutional Animal Care and Use Committee. The facilities within the Animal Resource Core at the Abigail Wexner Research Institute are fully accredited by the American Association for Accreditation of Laboratory Animal Care. Laboratory animals are maintained in accordance with the applicable portions of the Animal Welfare Act and the guidelines prescribed in the Department of Health and Human Services publication, Guide for the Care and Use of Laboratory Animals (54). Sputum samples were collected after receipt of written informed consent and under a protocol (IRB11-00790) approved by Nationwide Children’s Hospital Institutional Review Board. Samples were then deidentified and sectioned as described in ref. 7. Sections were air-dried for 15 min at room temperature and fixed in cold acetone for 10 min. Sections were then equilibrated in wash buffer that contained 0.05 M Tris⋅HCl, pH 7.4, 0.15 M NaCl, and 0.05% Tween 20 at room temperature for 5 min in a humidified chamber. Image-iT FX signal enhancer (Molecular Probes) was added to the sections and incubated at room temperature for 30 min. The sections were then washed 3 times with wash buffer. The sections were incubated with SuperBlock (Thermo Fisher Scientific) at room temperature for 10 min. Zenon Alexa Fluor 488 mouse IgG_2a labeling kit (Thermo Fisher Scientific) was used to label the monoclonal antibody against dsDNA as per the manufacturer’s instructions. Sections were then incubated with 1.5 μg of monoclonal antibody against dsDNA conjugated to Alexa Fluor 488 and 1.5 μg of monoclonal antibody against cruciform DNA at room temperature for 1 h. The sections were incubated with naive IgG as a negative control. The sections were fixed with 4% formaldehyde at room temperature for 10 min. The sections were then rinsed 3 times in wash buffer and incubated with goat anti-mouse IgG1 conjugated to Alexa Fluor 594 (Molecular Probes) for 30 min at room temperature. The sections were cover-slipped with ProLong Gold antifade mountant (Molecular Probes). Sections were imaged with a ×63 objective on a Zeiss 800 laser scanning confocal microscope (Zeiss).

Mechanical Indentation of NTHI Biofilms.

Mechanical indentation was performed using a TA Instruments Discovery Hybrid Rheometer-2 (HR-2) with the Peltier plate connected to a heat exchanger (TA Instruments). The rheometer was fitted with 8-mm-sandblasted Smart Swap parallel plate geometry. Rheology measurements were performed at 25 °C. TRIOS v4 (TA Instruments) software was used for data collection. Biofilms formed by NTHI strain 86-028NP were established in 35-mm FluoroDishes (World Precision Instruments) for 16 h as described in SI Appendix, Supplementary Methods, Stabilization of Bacterial Biofilm Structure by RuvA and Disruption by RuvABC Complex. After 16 h of incubation at 37 °C, 5% CO₂, the medium was replaced with fresh medium that contained one of the following: naive IgG (1,000 nM), α-IHF IgG (1,000 nM), naive IgG + RuvA (450 nM), or α-IHF + RuvA (450 nM). After an additional 8-h incubation period, the medium was replaced again as described above and the biofilms were incubated for an additional 16 h. Prior to rheological analysis, biofilms were washed twice with sterile phosphate-buffered saline (PBS) and the dishes were filled with 3 mL PBS. Dishes were transferred to the Peltier plate, and mechanical indentation was performed using an approach rate of 1 μm/s, with a termination step set to 8N. For data interpretation, the force–displacement curves were converted to stress–strain curves. Force (F) was converted to normal stress (σ) by dividing by the area of the geometry (σ = F/πr²). Displacement was converted to strain (γ) by dividing the resultant change in thickness by the original thickness (γ = ΔL/L). The Young’s modulus (E) was calculated using the force–displacement relationship previously described (28):

$$E=\frac{slope\cdot \left(1-v^2\right)}{2r},$$ [1]

where the slope is of the force–displacement curve (newtons per meter), r is the radius of the geometry (r = 0.004 m), and v is the assumed Poisson’s ratio of a biofilm (v = 0.5) (40). The slope of the lower, linear portion of the force–displacement curve was measured, which corresponded to 0 to 40% strain. Two biological replicates were analyzed, with duplicate biofilms analyzed per biological replicate and 2 technical replicates per biofilm.

Statistical Evaluation.

Statistical significance was assessed by unpaired or paired t test (GraphPad Prism version 6.0) and denoted as *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001.

Detailed materials and methods can be found in SI Appendix.

Data Availability.

Raw data files are available from the corresponding author upon fair request.