The ability to induce experimental endocarditis of biofilm-deficient mutants of Streptococcus gordonii was studied in an isogenic background. Strains were inactivated in either comD, fruK or pbp2b genes, which are involved in biofilm formation. These strains were clearly impaired (>75% reduction) in biofilm production in vitro. However, this did not result in a decreased severity of infection in vivo.
The contribution of microbial surface adhesins such as fibronectin- and fibrinogen-binding proteins, glucosyltransferases and other components recognizing adhesive matrix molecules to the establishment of infective endocarditis (IE) has been demonstrated [reviewed in (Moreillon et al., 2002)]. In addition, other as yet undetermined factors might be implicated as well (Lebeau et al., 1993; Cree et al., 1995; Hogevik et al., 1998). In particular, even though biofilm production is considered significant in a wide array of diseases (Costerton et al., 1999; Donlan & Costerton, 2002; Parsek & Singh, 2003; Hall-Stoodley et al., 2004), its contribution to the establishment of streptococcal IE has not been assessed in an isogenic background.
Only a few studies investigated the possible implication of biofilm formation in IE. One showed that endocarditis isolates of Enterococcus faecalis produced significantly more biofilm than nonendocarditis isolates (Mohamed et al., 2004). Yet, two studies showed that biofilm production was neither more frequently present nor higher in endocarditis-related E. faecalis when compared to bacteriemic, nonendocarditis E. faecalis isolates (Baldassarri et al., 2004; Di Rosa et al., 2006). Finally, another study comparing viridans streptococci isolated from endocarditis patients with those isolated from nonendocarditis sepsis (in neutropenic patients) did not reveal significantly greater biofilm production in endocarditis strains. Nevertheless, it suggested that infection due to biofilm-producing strains may need longer periods or higher doses of antibiotics (Presterl et al., 2005). Thus, the importance of biofilm production in the pathogenesis of IE remains unclear.
Viridans streptococci are a frequent cause of IE (Moreillon & Que, 2004). In order to assess whether biofilm formation is relevant in the pathogenesis of this infection, we investigated the virulence of Streptococcus gordonii Challis (Pozzi et al., 1990) mutants carrying deletions in biofilm-related genes in the rat model of experimental endocarditis.
Materials and methods
DNA manipulations and transformation
The preparation of S. gordonii genomic DNA was carried out according to a published method (Pozzi et al., 1990). Restriction endonuclease digests, DNA ligations and PCR amplifications were performed using standard techniques (Sambrook et al., 1989). The preparation and transformation of S. gordonii competent cells was done as previously described (Pozzi et al., 1990).
The following oligonucleotides were used in this study: erm-K7-DAM104 (5′-CCGGGCCCAAAATTTGTTTGATTTGTATCTTAAAATTTTGTATAATAGG-3′), erm-K7-DAM105 (5′-GGGGATCCAAAAACTGCCGACTGTAAAAAGTACAGTCGGCAGCGACTCATAGAATTATTTC-3′), erm-PA (5′-GGCGCGCCCCGGGCCCAAAATTTGTTTGAT-3′), erm-PB (5′-GGCCGGCCAGTCGGCAGCGACTCATAGAAT-3′), comD_L5 (5′-TTCTATATTTGGAAAATGTCCCTGT-3′), comD_L3 (5′-GGCGCGCCAACAAACGATAAAAACAAATTCAA-3′), comD_R5 (5′-GGCCGGCCAAAGGAGAAAATCGAGGATTAGG-3′), comD_R3 (5′-TTCGAATTTCGTTTCCTCTAAAA-3′), pbp2b_L5 (5′-GGTCTGATGAAGGCAGGAAA-3′), pbp2b_L3 (5′-GGCGCGCCATCAGCAATCTGGCGTTTTT-3′), pbp2b_R5 (5′-GGCCGGCCCCGACAGAATCCCAAGGTTA-3′) and pbp2b_R3 (5′- CCGAGTAACGCTGTGGCTA-3′).
Construction of biofilm-deficient strains of S. gordonii
Three genes which had previously been shown to be involved in biofilm formation and operating in separate, unrelated systems were selected. First, comD, encoding an histidine kinase involved in quorum sensing (Loo et al., 2000). Secondly, fruK which encodes a fructose-1-phosphate kinase (Loo et al., 2003). Thirdly, pbp2b encoding the penicillin binding protein 2B (Loo et al., 2000).
All three genes were inactivated using the so-called PCR ligation mutagenesis technique (Lau et al., 2002) in order to avoid the emergence of revertants in vivo (Giammarinaro & Paton, 2002; Loo et al., 2003).
The inactivation of both comD and pbp2b was performed as follows. First, an erythromycin resistance cassette was PCR-amplified from plasmid pJDC9 (Chen & Morrison, 1988) using primers erm-K7-DAM104 and erm-K7-DAM105 (Claverys et al., 1995). Next, a second PCR was performed using this cassette as a template with primers erm-PA and erm-PB in order to introduce a 5′AscI and a 3′FseI restriction site (Lau et al., 2002). This final resistance cassette was double digested with these two enzymes.
In parallel, a sequence overlapping the 5′ portion of either comD or pbp2b was amplified from S. gordonii genomic DNA using primer couples comD_L5/comD_L3 or pbp2b_L5/pbp2b_L3, respectively. A second sequence overlapping the 3′ portion of either comD or pbp2b was amplified using primer couples comD_R5/comD_R3 or pbp2b_R5/pbp2b_R3, respectively. The amplified 5′ and the 3′ overlapping portions were digested using AscI and FseI, respectively, and separately ligated to the resistance cassette. Each ligation product was amplified using either the primer couples comD_L5/comD_R3 or pbp2b_L5/pbp2b_R3, in order to generate the whole chimera constructs. The inactivation of fruK was achieved as previously described using a spectinomycin resistance cassette (Loo et al., 2003). Finally, either of the three constructs was transformed into competent wild-type S. gordonii (SG). Recombinants were selected on antibiotic-containing agar plates, and the correct inactivation was verified by PCR (data not shown).
The three SG mutants were named SGΔfruK, SGΔcomD and SGΔpbp2b, respectively.
Solid phase adhesion assay
Binding to immobilized fibronectin was quantified using a previously described method (Styriak et al., 1999). Briefly, 96-well flat-bottomed microtiter plates (MaxiSorp, Nunc, Roskilde, Denmark) were coated overnight at 4°C with two-fold decreasing concentrations (from 125 to 0.5 µg mL−1) of purified human fibronectin (Sigma, Buchs, Switzerland). The plates were washed with phosphate buffered saline (PBS) and wells incubated with 2 mg mL−1 bovine serum albumin in PBS for 1 h to block unoccupied sites. After washing with PBS, 50 µL of exponential bacterial cultures (corresponding to about 5 × 108 CFU) were added to each well and incubated at 37°C for 2 h. The plates were then washed with PBS, and adherent bacterial stained with 0.5% (w/v) crystal violet for 45 min. Wells were rinsed with PBS, and crystal violet solubilized in 150 µL of ethanol-0.1 M sodium citrate (1 : 1, v/v) (pH 4.3). Absorbances were read with a Multiskan RC plate reader (Thermolabsystems, Helsinki, Finland) at 570 nm (A570 nm) and strains classified as strongly adherent (A570 nm>0.3), weakly adherent (0.1≤A570 nm,≤0.3) or nonadherent (A570 nm<0.1).
Adhesion to human platelet-fibrin clots in vitro
Adherence to human platelet-fibrin clots was assessed as previously described (Moreillon et al., 1995). Platelet-fibrin clots were produced by pouring 0.5 mL of plasma into 30-mm-diameter cell culture plates (Corning Costar, Corning, NY) in the presence of 100 µL of a 500-U mL−1 National Institutes of Health bovine thrombin solution and 100 µL of a 0.2 mM CaCl2 solution. The clots were then dehydrated overnight at 37°C and kept at 4°C before being used. To determine bacterial adherence, 2 mL of saline containing about 105 CFU mL−1 was added to the wells, and the plates were agitated for 3 min at 120 r.p.m. on a rotating incubator. The fluid was gently decanted, and the clots were washed twice for 5 min with 2 mL of PBS to remove nonadherent bacteria. The clots were then overlaid with 3 mL of nutrient agar, which was allowed to solidify before incubation at 30°C. The number of adherent bacteria giving rise to colonies was determined after 24 h of incubation and expressed as a multiple of adherent organisms relative to the original inoculum.
Biofilm was quantified using an adaptation of a published method (O'Toole et al., 1999). Briefly, exponential growing cultures in CDEN (Garcia-Bustos et al., 1987) supplemented with 0.015% of yeast extract were diluted 1 : 500 in fresh medium and 100 µL aliquots were dispensed into round-bottomed 96-well PVC plates (Falcon 353911, Becton Dickinson Labware) previously coated with 16 µg mL−1 of fibronectin. Because adhesion to fibronectin was the same for all strains (see results), coating the wells with fibronectin was carried out in the biofilm assay to reduce nonspecific adhesion to PVC and to diminish background. The concentration of 16 µg mL−1 of fibronectin was chosen because it was the lowest that allowed maximum binding for all strains in the solid phase assay. The plates were incubated for 48 h in microaerobic conditions. Wells were subsequently stained for 15 min with 50 µL of 0.5% (w/v) crystal violet solution and carefully rinsed with water three times. The crystal violet that stained biofilm cells was solubilized in 150 µL of ethanol-0.1 M sodium citrate (1 : 1, v/v, pH 4.3). Biofilm formation was quantified by measuring the A570 nm. Biofilm produced by SG was assigned a value of 100, and tested strains quantified relatively to it. Experiments were repeated independently at least three times in 10 wells for each strain.
Rat model of experimental endocarditis
Sterile aortic vegetations were produced in female Wistar rats as previously described (Heraief et al., 1982). In a first set of experiments, groups of animals were inoculated intravenously with 5–7 × 105 CFU of exponential-phase streptococci. This was the minimum inoculum producing ≥90% of endocarditis with the SG strain (Entenza et al., 1999), thus permitting a clear differentiation with putatively less-infective strains. Rats were sacrificed early (16 h) after inoculation and colony counts in the vegetations were determined. In a second set of experiments, rats were challenged with a higher inoculum (106 CFU), thus ensuring that all the isolates would colonize the damaged valves at the onset of infection. Animals were subsequently sacrificed either 3 or 5 days after inoculation and colony counts in the vegetations were determined. The frequency of infection was compared by the Fisher's exact test. The median bacterial titers were compared by the nonparametric Mann–Whitney unpaired test. A value of P<0.05 was considered significant.
Results and discussion
Both parent and mutant strains showed a similar weak adherence to fibronectin (0.1≤A570 nm≤0.2) in the solid phase assay. Accordingly, the mean adherence (±SD of 12 individual determinations) to platelet-fibrin clots, expressed as a function of the original inoculum, was similar for all strains, i.e. 0.036±0.01 for SG as compared with 0.039–0.049±0.01 for the transformants (P>0.05 by anova with Bonferroni's correction). Thus, adherence to components of the vegetations was not affected.
On the contrary, in accordance with previously published results (Loo et al., 2000, 2003), we found that all mutagenized strains exhibited an important reduction in biofilm formation (Fig. 1). Of note, none of the mutants were affected in their growth rates in liquid cultures.
In vivo, early sacrifice experiments (16 h after inoculation with 5–7 × 105 CFU) showed that all the strains had comparable abilities to colonize damaged valves (Fig. 2a). Moreover, they had similar bacterial densities in aortic vegetations (P>0.05). Hence, biofilm-deficient mutants were not impaired in their ability to initiate endocarditis. To investigate the contribution of biofilm to disease progression, we subsequently infected rats with a higher inoculum (106 CFU) and sacrificed them 3 days postinoculation. As depicted in Fig. 2b, no statistical difference in CFU counts in vegetations could be observed between the strains. However, rats inoculated with SGΔcomD exhibited a more distributed pattern, some being infected with as few as 3 log10 CFU g−1 of vegetation. Therefore, we wanted to assess the progression of infection by prolonging the delay between inoculation and sacrifice. Five days postinoculation (Fig. 2c), vegetations of rats infected with the SG or either of the three biofilm-deficient strains had similar bacterial load. Hence, even though SGΔcomD exhibited scattered bacterial counts after 3 days, the results at 5 days clearly show that later progression of the infection was not hindered. Therefore, an altered biofilm formation affected neither early valve colonization nor later evolution of infection.
In summary, S. gordonii mutants deficient in biofilm formation in vitro were not impaired in producing endocarditis in vivo. The fact that both the parent and mutant strains exhibited similar levels of adhesion to fibronectin and platelet-fibrin clots may explain comparable levels of virulence. Thus, initial S. gordonii adherence to cardiac vegetations (Stutzmann Meier et al., 2001) appears to be more important than biofilm formation for the establishment of endocarditis. Yet, it cannot be excluded that, in vivo, other pathways of biofilm production may complement the deficiency observed in vitro.
Pioneer studies in a rabbit model have shown that bacteria quickly adhere to the nonbacterial thrombotic endocarditis and that within hours the microorganisms begin to form microcolonies in the vegetation mesh. Moreover, they indicated that the metabolic activity of colonies was more important on the surface of the vegetation than inside, where bacteria were in a resting state (Durack & Beeson, 1972; Durack, 1975). Likewise, the influence of pretreatment duration of infection on the efficacy of antibiotic treatment may be related to the lower metabolic activity of bacteria growing in the vegetation (Cremieux et al., 1993; Entenza et al., 1994). Furthermore, Mghir (1994) confirmed the presence of biofilm-compatible structures in the vegetations of rabbits infected with streptococci. The biofilm-like structure was related to dextran production and treatment of the animals with dextranase decreased the infection severity and facilitated antibiotic-induced bacterial eradication.
Thus, over the long-term course of endocarditis, the ability to form a biofilm may confer a competitive advantage or protect against antibiotic treatment (Costerton et al., 1999). However, the present results indicate that biofilm formation by streptococci was not a prerequisite for successful valve infection, and that care must be taken when linking biofilm formation in vitro with in vivo virulence.
This work was supported by grants 3235-62′698 (MD/PhD program, to A.B.), 3200-65371.01 and 3200-65371/2 from the Swiss National Fund for Scientific Research. We thank Marlyse Giddey and Jacques Vouillamoz for outstanding technical assistance. Millán Ortiz is acknowledged for the construction of the SGΔcomD mutant.