|Home | About | Journals | Submit | Contact Us | Français|
Streptococcus sanguinis is a gram-positive, facultative anaerobe and a normal inhabitant of the human oral cavity. It is also one of the most common agents of infective endocarditis, a serious endovascular infection. To identify virulence factors for infective endocarditis, signature-tagged mutagenesis (STM) was applied to the SK36 strain of S. sanguinis, whose genome is being sequenced. STM allows the large-scale creation, in vivo screening, and recovery of a series of mutants with altered virulence. Screening of 800 mutants by STM identified 38 putative avirulent and 5 putative hypervirulent mutants. Subsequent molecular analysis of a subset of these mutants identified genes encoding undecaprenol kinase, homoserine kinase, anaerobic ribonucleotide reductase, adenylosuccinate lyase, and a hypothetical protein. Virulence reductions ranging from 2-to 150-fold were confirmed by competitive index assays. One putatively hypervirulent strain with a transposon insertion in an intergenic region was identified, though increased virulence was not confirmed in competitive index assays. All mutants grew comparably to SK36 in aerobic broth culture except for the homoserine kinase mutant. Growth of this mutant was restored by the addition of threonine to the medium. Mutants containing an insertion or in-frame deletion in the anaerobic ribonucleotide reductase gene failed to grow under strictly anaerobic conditions. The results suggest that housekeeping functions such as cell wall synthesis, amino acid and nucleic acid synthesis, and the ability to survive under anaerobic conditions are important virulence factors in S. sanguinis endocarditis.
Infective endocarditis is a life-threatening endovascular infection believed to occur when bacteria in the bloodstream adhere to damaged heart valves. Conditions that cause heart valve damage are risk factors for native valve endocarditis and include congenital heart disease, chronic rheumatic heart disease, and mitral valve prolapse with regurgitation (48). Fibrin and platelets are deposited at the site of endothelial cell trauma, forming a sterile vegetation where bacteria may adhere and colonize during bacteremia. Oral streptococci may enter the bloodstream through invasive dental procedures; therefore, antibiotic prophylaxis is recommended prior to such procedures for at-risk patients (48). It is also clear that some endocarditis cases caused by oral flora are not preceded by dental procedures (48, 59). Thus, other sources of bacteremia, which may include eating or other daily activities (48, 59), must be responsible for these cases. However, antibiotics cannot realistically be used to prevent such occurrences. Therefore, a vaccine would be a preferable prophylactic if one were available.
Several studies have examined putative streptococcal virulence factors for endocarditis to better understand the disease and to identify possible vaccine candidates. Some have implicated binding to laminin, fibrin, and intact extracellular matrix as important in causing endocarditis (48, 62). Platelet aggregation by Streptococcus sanguinis has also been shown to contribute to virulence in an animal model (26). Receptors for this interaction have been identified on both bacteria and platelets (6, 31, 55). Other studies have come to conflicting conclusions concerning the exact role of platelets in endocarditis (27, 60). This may be related to the finding that release by activated platelets of microbicidal proteins appears to aid in the clearance of streptococci from vegetations (15, 60). Therefore, successful pathogens may need to activate platelets to cause aggregation but also resist killing by platelet microbicidal proteins (15, 25). A family of related lipoproteins, including FimA from Streptococcus parasanguinis, has also been implicated in streptococcal adherence and virulence. A fimA mutant of S. parasanguinis shows decreased binding to fibrin monolayers in vitro and decreased infectivity in a rat model of endocarditis (9). A FimA ortholog, SloC, is also important for endocarditis causation in Streptococcus mutans (34, 53). Both FimA (52) and SloC (53) are members of ABC transport systems that take up manganese and iron, and this function has been shown to be required for endocarditis virulence in S. mutans (53). Exopolysaccharide production by S. mutans, which contributes to dental caries partly by contributing to adherence, has also been examined in relation to endocarditis. Mutants unable to produce exopolysaccharides adhere less well to fibrin, are less resistant to phagocytosis by human granulocytes in vitro, and show decreased virulence in a rat model of endocarditis (50). Thus, exopolysaccharides likely contribute to adherence and avoidance of immune clearance in S. mutans. However, the elimination of exopolysaccharide production in Streptococcus gordonii by mutation had no effect on the ability of this organism to cause endocarditis in the same model (66).
Among the viridans streptococci, S. sanguinis is most commonly isolated from infective endocarditis patients (17, 58). We therefore chose to examine this species to identify virulence factors that could serve as new targets for drugs or vaccines. The technique of signature-tagged mutagenesis (STM) was used to identify S. sanguinis virulence factors for endocarditis in an animal model. STM is a modified transposon mutagenesis system devised by Hensel and colleagues (24). It was originally developed for Salmonella species but has been successively applied to many other pathogens (13, 44). STM allows for the large-scale creation and simultaneous in vivo screening of random mutants for alterations in virulence.
For this study, STM was used to create and screen 800 mutants of S. sanguinis SK36 in a rabbit model of endocarditis. Two rounds of screening and additional experiments identified six mutants of interest, including strains with mutations in an intergenic region and in genes encoding undecaprenol kinase, homoserine kinase, anaerobic ribonucleotide reductase (RNR), adenylosuccinate lyase, and a hypothetical protein.
The bacterial strains and plasmids used in this study are listed in Tables Tables11 and and2.2. S. sanguinis strain SK36 was obtained from Mogens Kilian (University of Aarhus, Denmark) (32). S. sanguinis strains were routinely grown in an atmosphere of 10% H2, 10% CO2, and 80% N2 at 37°C in brain heart infusion (BHI) broth (Difco Inc., Detroit, MI) or on tryptic soy broth (TSB; Difco Inc.) supplemented with 1.5% agar, unless otherwise indicated. Experiments performed in an anaerobic chamber (Forma Scientific, Marietta, Ohio) employed the same atmosphere with a palladium catalyst that maintained oxygen levels at less than 1 ppm. Aerobic growth experiments were performed in 200-μl volumes in microplates using a FLUOstar plate reader (BMG, Offenburg, Germany). Antibiotics for streptococcal selection were chloramphenicol (Cm) at 5 μg/ml and erythromycin (Em) at 10 μg/ml as needed. Todd-Hewitt (TH) broth (Difco Inc.) adjusted to pH 7.6 and containing horse serum (HS) was used for S. sanguinis transformation. TH broth was filter sterilized and aliquots stored at −20°C. Working stocks were kept at 4°C for up to a month. Heat-inactivated HS was stored at −20°C, thawed, and added to TH broth to 2.5% immediately before use. Escherichia coli DH10B (Invitrogen, Carlsbad, CA) used for cloning was routinely grown at 37°C with shaking in Terrific Broth (63). Cm (5 μg/ml), Em (300 μg/ml), ampicillin (Ap; 100 μg/ml), and kanamycin (Kan; 50 μg/ml) were added as needed for plasmid selection.
A pool of signature tags composed of a 40-bp variable central region flanked by 24- and 25-bp invariant ends was provided by David Holden (Hammersmith Hospital, London) (24) and modified to create BglII restriction enzyme sites at either end using primers PB1 and PB2 (Table (Table3).3). The plasmid pJFP1 and the signature tags were digested with BglII, ligated to each other, and transformed into E. coli. Transformants were selected by Cm resistance.
The MarC9 hyperactive form of the Himar1 transposase was purified from an E. coli strain carrying the transposase gene in a pET expression vector as described previously (36). A single purification yielded 17.6 mg of transposase. In vitro transposition reactions were performed with 0.5 μg of donor plasmid (pJFP1 or STM plasmids), 1 to 1.5 μg recipient DNA (SK36 DNA), 200 μg/ml bovine serum albumin, and 2 μg transposase in 20 μl 1× restriction endonuclease buffer 4 (New England Biolabs) at 30°C for 1 h (36, 37). Reactions were stopped by heat inactivation at 70°C for 10 min. Products were treated with T4 DNA polymerase and T4 DNA ligase to fill the gaps generated by transposition and to create intact double-stranded DNA. Each reaction was followed by heat inactivation at 70°C for 10 min. One-eighth of each reaction was used to transform S. sanguinis SK36 in the presence of competence stimulating peptide as described below.
The transformation method was based on previously published (21) and unpublished (R. Dwayne Lunsford, SmithKline Beecham Pharmaceuticals, Collegeville, Pennsylvania) methods. Briefly, cultures in 2 ml of TH-HS broth were incubated at 37°C overnight either in an anaerobe jar or with the tube's cap tightly closed. Ten or 50 μl of each culture, respectively, was transferred to 10 ml of TH-HS media similarly preincubated. Incubation was continued at 37°C for 3 to 4 h until the optical density at 660 nm (OD660) reached 0.06 to 0.08. Transforming DNA, 330 μl of cells, and 70 ng of S. sanguinis competence stimulating peptide (21) were added to 0.7-ml microfuge tubes, and incubation was continued for 1 h. Cells were plated on tryptic soy agar (TSA) with or without antibiotics and grown in an anaerobe jar at 37°C for 48 h.
Signature-tagged mutants were grown in individual wells of 48-well blocks in 1.25 ml BHI plus Cm in an anaerobe jar overnight at 37°C. Cells were mixed by vortexing, pooled, and diluted 10-fold into fresh, prewarmed BHI. Incubation was continued at 37°C for 3 h. Cells were harvested by centrifugation at 3,700 × g for 10 min at 4°C, washed, and resuspended in phosphate-buffered saline (PBS). Cells were diluted to an OD660 of approximately 0.8, equivalent to about 3 × 108 cells per 0.5 ml inoculum.
The previously described rat (49, 50) and rabbit (18) models of endocarditis were adapted for this study, employing male Sprague-Dawley rats weighing 200 to 250 g and specific-pathogen-free, male New Zealand White rabbits weighing 3 to 3.5 kg. The protocol received Institutional Animal Care and Use Committee approval (no. 9710-2082 and 0010-2865) and complied with all applicable federal guidelines and institutional policies. In brief, a catheter was inserted through the internal carotid artery of anesthetized animals, past the aortic valve to impose valve damage. The catheter was sutured and remained in the artery throughout the rest of the experiment. Two days later, streptococci were inoculated into the ear vein of catheterized rabbits or the tail vein of catheterized rats. One or two days later, rabbits were sacrificed by intravenous injection of euthasol (Virbac AH, Inc., Fort Worth, TX) and rats were sacrificed by CO2 inhalation. The heart was removed, and accurate catheter placement was assessed visually. The aortic valve and any apparent vegetations were removed and homogenized with PBS, and multiple dilutions were spread on TSA plates.
Pooled inoculum cells or cells from pooled vegetation homogenate plates containing 5,000 to 50,000 colonies were used for DNA extraction as described previously (53). Signature tags were amplified by PCR using PB1 and PB2 primers (Table (Table3)3) and 0.5 to 1 μg DNA. Cycling conditions were 95°C for 1 min; 25 cycles of 94°C for 30 s, 45°C for 60 s, and 72°C for 10 s; and a final 2-min extension at 72°C. Products were purified from 3% NuSieve agarose gels, and 1/10 of each product was used for labeling with a PCR digoxigenin probe synthesis kit (Roche, Mannheim, Germany) and primers PB11 and PB12. The cycling conditions were as described above except that the annealing temperature was 37°C and the number of cycles was 20. Digoxigenin-labeled amplicons were recovered from another 3% NuSieve agarose gel. The gel slice was melted by boiling in hybridization buffer and added directly to a hybridization bottle containing a dot blot membrane. Dot blots were created by transferring 1 μg of each STM plasmid onto a nylon membrane by using a vacuum manifold (Schleicher & Schuell Bioscience, Keene, NH). Signal detection was performed using Genius chemiluminescent kit reagents (Roche) and a Fluorchem Imager with FluorChem software, version 3.04A (Alpha Innotech Corporation, Alexandria, VA).
The arbitrary primed PCR (AP-PCR) technique used was provided by Glen Tamura (University of Washington, Seattle, Washington) and modified during the course of the study (16). The final procedure employed 10 ng of chromosomal template DNA, 0.5 μM arbitrary primer, and 0.2 μM transposon primer for first-round reactions. Cycling conditions were as follows: 95°C for 5 min, followed by six cycles at 95°C for 30 s, 30°C for 30 s, and 72°C for 1.5 min; 30 cycles at 95°C for 30 s, 45°C for 30 s, and 72°C for 2 min; and 72°C for an additional 4 min. PCR products were purified using a QIAquick PCR purification kit (QIAGEN, Valencia, CA). One-eighth of each product was used as template for second-round reactions that employed the Arb2 primer (Table (Table3)3) and a nested transposon primer at concentrations of 0.2 μM each. Cycling conditions were as follows: 95°C for 1 min, followed by 30 cycles at 95°C for 30 s, 52°C for 30 s, and 72°C for 2 min, and then 72°C for an additional 4 min. Purified second-round products were used for DNA sequence analysis performed by the Virginia Commonwealth University Nucleic Acids Research Facility with the Arb2 and nested transposon primers.
Overnight cultures of SK36 and mutant strains were grown in BHI broth at 37°C. Seven hundred microliters each of the SK36 strain and a mutant culture were coinoculated into a tube containing 12.6 ml prewarmed BHI. Incubation was continued for 3 h. Cells were washed and suspended in PBS to an OD660 of 0.8. Five hundred microliters was inoculated into the peripheral ear veins of rabbits. Dilutions of the inoculum were plated using a two-layer plating technique based on a previously published procedure (42). Cells were mixed with 12.5 ml of molten 1.0% low-melting-point (LMP) agarose in TSB at 37°C. The cell suspension was poured into a petri dish, allowed to solidify at room temperature, and incubated for 2 h at 37°C. Molten TSA (12.5 ml at 50°C) with or without 10 μg/ml Cm was then overlaid onto each plate and allowed to solidify. Plates were returned to 37°C and incubated aerobically for 2 days, colonies were counted, and the ratio of mutant to SK36 CFU determined. Catheterization, inoculation, and necropsy were performed as described above. Dilutions of vegetation homogenates were incorporated into layer plates as described above for enumeration. The competitive index (CI) was determined as the mutant/SK36 ratio of the homogenate divided by the mutant/SK36 ratio of the inoculum. Ratios were determined from colony numbers obtained from two to three plates. Since colony numbers are often strongly skewed, the log-transformed values were used. Repeated measures mixed-model analysis of all the colony counts was used to determine CI values, 95% confidence intervals, and whether CI values were significantly different from 1, with α = 0.05. All strains were tested in three to four animals each.
Gene splicing by overlap extension (Gene SOEing ) was used to construct an internal, in-frame deletion of the nrdD gene, fusing the gene's first 12 and last 14 codons. Oligonucleotide pairs ARTR-SalI-24495-F and ARTR-23327-R and ARTR-21212-F and ARTR-SalI-20031-R were used in the first PCR (Table (Table3).3). The products of the first reaction served as the template for amplification by primers ARTR-SalI-24495-F and ARTR-SalI-20031-R in a second reaction. The ARTR-24495/20031 product was then cloned into suicide plasmid pVA2606 (Table (Table1)1) via SalI restriction sites, creating pJFP27. E. coli DH10B was electrotransformed with pJFP27 and plated on Lennox Luria-Bertani agar containing Kan and 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal). Plasmid DNA was prepared using a Quantum Prep Plasmid Mini-Prep kit (Bio-Rad) and screened by PCR with the ARTR-SalI-24495-F and ARTR-SalI-20031-R primers. DNA sequence analysis confirmed proper insertion and in-frame deletion of nrdD. S. sanguinis SK36 was transformed with 500 ng SspI-linearized pJFP27 as described above and plated on TSA without antibiotics. Colonies were inoculated into overnight broth cultures for rapid PCR analysis. Briefly, 1 ml of each overnight culture was centrifuged (18,000 × g for 2 min) and the pellet was washed in 1 ml 10 mM Tris-Cl, pH 8.0, and suspended in 100 μl 10 mM Tris-Cl. Tubes were then boiled for 5 min, cooled, and centrifuged again, and 100 μl supernatant was transferred to a new tube. Two microliters was used as template for PCR screening reactions employing ARTR-SalI-24495-F and ARTR-SalI-20031-R. Of the 20 colonies screened, 2 produced a band of the expected size for the deleted gene. One of these was selected for further study and named JFP27.
Two nrdD constructs were designed for complementation studies. Primers specific for the two constructs, ARTR-EcoI-F1, ARTR-EcoRI-F2, and ARTR-EcoRI-R (Table (Table3),3), were used in PCRs with S. sanguinis SK36 genomic DNA as template and with Supermix (Invitrogen). PCR products were purified using Quantum Prep Freeze 'N Squeeze DNA gel extraction spin columns (Bio-Rad) and cloned into the pVA838 (41) shuttle vector via EcoRI restriction sites, creating plasmids pJFP29 and pJFP30 (Table (Table1).1). E. coli DH10B was electrotransformed with pJFP29 and pJFP30, and transformants were selected on Luria-Bertani agar containing Em. Plasmid DNA was extracted using the Quantum Prep Plasmid Mini-Prep kit (Bio-Rad) and then screened by EcoRI digestion. Sequencing with pVA838-specific primers pVA838-2939-F and pVA838-5331-R (Table (Table3)3) confirmed correct construct insertion at splicing junctions. The plasmids were introduced into S. sanguinis by transformation with 250 ng of DNA. Transformants were selected by plating on TSA containing Em. PCR with pVA838-specific primers was used to confirm the presence of appropriate plasmids.
For soft-agar studies (61), LMP agarose was dissolved in TSB by autoclaving and aliquoted in 10-ml volumes into glass Hungate tubes. Em and 5 μg/ml of the oxygen dye indicator, resazurin sodium salt (Acros) (5), were added to selected tubes. All tubes were then placed in the 37°C incubator of a Forma Scientific anaerobe chamber for 48 h. Tubes were then inoculated with 100 μl of overnight streptococcal cultures, mixed by inverting, and incubated at ambient temperature in the anaerobe chamber for 30 min to allow the agar to solidify. Tubes were then removed from the anaerobe chamber and incubated aerobically with caps loosened at 37°C for 24 h. The depth of growth was measured in each tube and documented using a digital camera. Values were compared using analysis of variance with the Tukey-Kramer multiple-comparisons test with α = 0.05.
DNA sequences were viewed, aligned, and edited using SeqMan II software (DNASTAR Inc, Madison, WI). Sequences were searched against the unfinished S. sanguinis genome (www.sanguis.mic.vcu.edu/) using the stand-alone version of BLASTN available from National Center for Biotechnology Information (1). GenBank searches were performed using BLASTP. Similarity of paired amino acid sequences was determined by Needleman-Wunsch global alignment using the default parameters of the European Molecular Biology Open Software Suite (EMBOSS) program Needle (57). Oligonucleotide primers were designed using Oligo (Molecular Biology Insights, Cascade, CO) and FastPCR (www.biocenter.helsinki.fi/bi/bare-1_html/fastpcr.htm) software. Multiple sequence comparisons were made using the Genetics Computer Group program Pileup (Wisconsin Package, version 10.3; Accelrys Inc., San Diego, CA) and ClustalW (12). Transcriptional terminators were predicted using the Genetics Computer Group program Terminator (7). DNA sequences of genes identified in the study are available as supplementary material.
STM mutants were generated by in vitro transposition using purified Himar1 transposase and a mariner-based minitransposon, magellan2. In this system, the purified transposase is sufficient to cause transposition of the minitransposon into the target DNA included in the reaction, although subsequent treatment with polymerase and DNA ligase is required to fill and seal the gaps created by transposition (37). The target for transposition was the genomic DNA of S. sanguinis strain SK36. Transposition products were then introduced into cells of the same strain by transformation. The SK36 strain was used for these studies because it was selected for genomic sequencing (Macrina et al., manuscript in preparation) based on a number of properties including adhesion to saliva-coated hydroxyapatite, platelet aggregation, and virulence for endocarditis in a rat model (www.sanguis.mic.vcu.edu/). SK36 was also sensitive to Em, Cm, and Kan.
The pEMCat vector containing the magellan2 minitransposon (22) was modified by removing the bla gene to prevent the possible introduction of β-lactamase resistance into streptococci. The resulting plasmid was designated pJFP1. Pooled signature tags were ligated into the minitransposon of pJFP1 as described by Hava and Camilli (22). Plasmids were screened to select 48 that produced strong signals in blots washed under high stringency conditions. These tagged plasmids were combined into pools for further screening to eliminate cross-hybridizing tags. One plasmid was eliminated. Finally, 40 of the 47 remaining plasmids that contained uniquely hybridizing signature tags, referred to as STM plasmids, were selected. Individual in vitro transposition reactions were performed with each STM plasmid and S. sanguinis genomic DNA. A portion of each product was transformed into SK36. Transformants were arrayed in microtiter plates to create sets of 40 uniquely tagged mutants and cryopreserved. In total, 20 sets of 40 mutants each were prepared and tested.
The first set of 40 mutants was inoculated into 18 catheterized rats, which served as a model for endocarditis (49). Inocula of 1010, 109, and 108 CFU were used. Seven rats were successfully infected: three each from the 1010 and 108 inocula and one from the 109 inoculum. This infection rate is typical for viridans streptococci in this model (9, 34, 53). Dot blot analysis of the inoculum pool and the bacteria recovered from the infected vegetations (output pool) was performed essentially as described previously (22). The output pool of bacteria was recovered from rats 2 days postinfection by spreading heart valve homogenates on agar plates. Replicate dot blots containing 41 STM plasmids were prepared on nylon membranes. Each blot was hybridized with labeled signature tags derived from the pooled inoculum or from the bacteria recovered from each infected vegetation sample. The 41st tagged plasmid served as a negative control since this plasmid was not used for mutant creation. The number of mutant strains that infected each animal ranged from 2 to 27, with a median value of 9. The results from three rats are indicated in Fig. Fig.1A.1A. As shown, strain recovery was inconsistent. This suggested that a colonization bottleneck had occurred (44) so that failure to recover most strains from a vegetation sample was probably the result of chance rather than reduced virulence. Therefore, additional trials were performed using the rabbit endocarditis model. Three inoculum levels were used: 2 × 108, 2 × 109, and 1 × 1010. The largest inoculum sample proved lethal within 24 h, but there was no apparent difference in the results from the other two inocula. All three inocula produced infections in all animals, and dot blotting revealed consistent results from each animal. An inoculum size of 2 ×108 to 4 ×108 was used for the remaining STM trials. In addition, animals were sacrificed 1 day postinoculation rather than 2.
Each STM mutant pool was inoculated into two or three rabbits. Figure Figure1B1B shows the results of analyzing in rabbits the same mutant pool that was analyzed in rats in Fig. Fig.1A.1A. If a mutant exhibited a hybridization signal comparable to that of the negative control in all output blots but not in the inoculum blot, the strain was identified as a putative avirulent mutant. If a mutant exhibited a strong signal in all output blots but a weak inoculum blot signal, the strain was identified as a putative hypervirulent strain. Strains that produced weak signals in both blots (indicated by double circles in Fig. Fig.1)1) could not be assessed for virulence. In contrast to our results with rats, most experiments resembled that shown in Fig. Fig.1B1B in that output pools obtained from different rabbits receiving the same inoculum were comparable.
Primary screening was performed with 800 STM mutants, identifying 53 strains with apparent decreased virulence and 7 with apparent increased virulence. Of these 60 strains, 49 were repooled into three secondary inocula for retesting. STM mutants that appeared to have normal virulence in the initial screen were used to supplement the secondary inocula so that 40 strains would again be tested in each experiment. The secondary screening eliminated 6 mutants, leaving 38 putative avirulent and 5 hypervirulent mutants.
Sequences flanking transposon insertions were identified by AP-PCR (16). These sequences were compared to the unfinished SK36 genomic sequence (www.sanguis.mic.vcu.edu/). Analysis of the first four mutants examined by AP-PCR, those corresponding to the circles in Fig. Fig.1B,1B, produced surprising results. These mutants appeared to have vector DNA inserted along with the transposon. Furthermore, chromosomal DNA sequences immediately flanking the insertion were apparently not contiguous prior to mutagenesis (51). Therefore, restriction-digested DNA from remaining mutants was examined by Southern blotting with the pJFP1 plasmid as a probe to identify mutants containing vector sequences. Of the 32 mutants so examined, 12 were eliminated by this screen. Ten additional strains were eliminated because sequencing data showed the presence of the pJPF1 vector sequence adjacent to the minitransposon or because the right and left flanking sequences were noncontiguous. Four other strains either did not generate bands in AP-PCR or were not analyzed. Finally, six mutants with the expected structures, 3-24, 4-4, 4-29, 6-26, 9-10, and 16-29, were collected for further analysis (Table (Table2).2). In all mutants except 3-24, a single gene was disrupted by the transposon insertion. In the 3-24 mutant, the transposon was located in a 182-bp intergenic region. In the 4-4 mutant, there was a 48-bp duplication at the site of transposon insertion whereas the other five mutants contained a TA dinucleotide duplication (data not shown). While TA duplication was observed most often upon insertion of the parent transposon of magellan2, longer duplications have been observed (37).
Production of a strong signal in inoculum blots by five of these six mutants suggested that they likely grew to comparable numbers in the inoculum and were thus well represented there. To examine this possibility more closely, the six mutant strains and SK36 cells were grown in BHI broth in a 96-well plate. Mutant 4-29 grew more slowly than SK36, whereas the other five mutants grew indistinguishably from SK36 and each other (Fig. (Fig.2A),2A), suggesting that any growth defects in these five mutants occurred in the animals but not in the broth culture prior to inoculation.
The final six mutants were examined for altered virulence using a CI assay in the rabbit model. SK36 and each mutant strain were coinoculated into catheterized rabbits. The CI was calculated as the ratio of mutant to SK36 in the cells recovered from the infected vegetation samples divided by the ratio of mutant to SK36 in the inoculum. Ratios were determined by plating dilutions of cells with and without Cm. Because we found that Cmr strains were recovered with reduced and variable efficiency when spread on the surfaces of plates containing Cm, we used a two-layer plating technique in which cells were embedded in LMP agarose and incubated for 2 h at 37°C prior to adding a top layer with or without Cm. With this technique, Cmr strains were recovered with equal efficiency on plates with and without Cm. The CI results are shown in Table Table2.2. A CI of >1 indicates increased virulence whereas a CI of <1 indicates reduced virulence.
Strain 3-24 was identified as a putatively hypervirulent mutant. This was the only mutant in which the transposon insertion did not interrupt a gene. Instead, the insertion was located between genes with similarity to a putative lactose repressor from Clostridium acetobutylicum (and other bacteria) and a conserved hypothetical protein from Streptococcus agalactiae, with characteristics of an isoprenylcysteine carboxyl methyltransferase (Table (Table2).2). The CI assays suggested that 3-24 was moderately more virulent than SK36, though the difference was not significant (Table (Table22).
Two strains, 9-10 and 16-29, were hypothesized to be defective in nucleotide synthesis. The purB gene, encoding adenylosuccinate lyase and mutated in strain 9-10, is involved in de novo purine synthesis (23). The CI value of 9-10 indicated a 150-fold virulence reduction (Table (Table2).2). In the 16-29 mutant, the disrupted gene is homologous to a conserved hypothetical protein from Desulfitobacterium hafniense and other bacteria (Table (Table2).2). This protein shows homology to the fic gene in E. coli, a mutant of which was not able to grow without ρ-aminobenzoate or folate. This suggests that the fic gene product may be involved in the synthesis of p-aminobenzoic acid (PABA) or folic acid (35). These compounds are required for purine and pyrimidine synthesis in bacteria. The twofold virulence reduction demonstrated by this mutant in CI assays was not statistically significant.
Mutant 4-4 was found to contain a disrupted gene encoding an ortholog of BacA from Streptococcus pneumoniae and other bacteria (Table (Table2).2). This gene's name was derived from the finding that the E. coli ortholog increased bacitracin resistance when expressed from a multiple-copy plasmid (10). Mutation of the bacA gene in S. pneumoniae and Staphylococcus aureus confers bacitracin sensitivity (11). Mutant 4-4 was found to exhibit sensitivity to bacitracin similar to that of Streptococcus pyogenes in liquid MIC assays and in disk diffusion assays. S. pyogenes is a prototypical bacitracin-sensitive species. Neither SK36 nor another unrelated STM mutant (6-26) exhibited sensitivity (data not shown).
The transposon insertion in mutant 4-29 was in an ortholog of the S. pneumoniae thrB gene, encoding homoserine kinase (Table (Table2).2). This enzyme is required for the synthesis of threonine (64), and its inactivation has been shown to result in threonine auxotrophy in many bacteria and yeast (67). To confirm the gene's function in S. sanguinis, growth studies were performed in BHI supplemented with threonine at 10, 20, 30, 50, and 75 mM. All five concentrations improved the growth of mutant 4-29, with the 75 mM supplementation showing the least effect (Fig. (Fig.2B).2B). In contrast, the effect of supplemental threonine on the growth of the SK36 cells was minimal (Fig. (Fig.2C).2C). Moreover, the addition of threonine to 50 or 75 mM inhibited SK36 growth. The results thus suggest that the thrB gene of S. sanguinis was properly identified.
In the 6-26 mutant, the minitransposon was inserted in a gene with similarity to nrdD of S. pneumoniae and other bacteria (Table (Table2),2), encoding anaerobic RNR (Fig. (Fig.3).3). Anaerobic, or class III, RNRs become permanently inactivated by oxygen (30). Therefore, it was hypothesized that the 6-26 strain would survive only in aerobic conditions. The growth study showed that 6-26 cells grew indistinguishably from SK36 under aerobic conditions (Fig. (Fig.2A).2A). Initial studies failed to demonstrate a growth defect using our standard procedure for producing an anaerobic atmosphere.
To test for the possibility that the 6-26 mutant might have acquired an extragenic suppressor mutation to allow anaerobic growth, as well as to eliminate the possible polar effects of transposon insertion, we created an in-frame deletion of the nrdD gene in SK36. DNA encoding 705 amino acids was removed, resulting in the fusion of the first 12 codons with the last 14, creating strain JFP27 (Fig. (Fig.3).3). This mutant also grew well on plates in anaerobe jars. A soft-agar oxygen gradient assay (61) employing an anaerobe chamber was therefore used to compare the oxygen requirements of 6-26, JFP27, and SK36. As indicated in Fig. Fig.4,4, parental strain SK36 grew throughout each tube (typically 7.0 cm in depth), indicating growth in both oxidized and reduced agar. In contrast, 6-26 and JFP27 growth was restricted to oxidized agar, indicated by the pink hue of the oxidized resazurin (Fig. (Fig.44 and data not shown). This suggests sensitivity of 6-26 and JFP27 to anaerobic conditions due to the loss of NrdD.
To confirm this, we attempted to complement nrdD mutant strains with the nrdD gene expressed from its native promoter. Plasmids pJFP29 and pJFP30 contained nrdD preceded by 337 or 36 nucleotides, respectively, cloned into the shuttle vector pVA838 (Fig. (Fig.3).3). Sequence analysis predicted a possible terminator 149 bp upstream of nrdD (Fig. (Fig.3).3). Plasmid pJFP29 was designed to include the possible terminator and the postulated subsequent native promoter. To exclude the possibility of the nrdD gene being expressed from a plasmid-borne promoter, pJFP30 was created, in which the nrdD gene was preceded by 36 bp—enough to contain a ribosome binding site but not a complete promoter (Fig. (Fig.33).
S. sanguinis SK36, 6-26, and JFP27 were each transformed with pVA838, pJFP29, and pJFP30, and evidence of NrdD complementation was evaluated in soft-agar studies. SK36 growth was unaffected by transformation with pVA838, pJFP29, or pJFP30 (Fig. (Fig.4).4). The nrdD mutant strain JFP27 containing pJFP29 exhibited growth that was indistinguishable from that of parent strain SK36. In contrast, pJFP30 did not complement the growth of JFP27, suggesting that nrdD in pJFP29 was expressed from its native promoter. Typical dense growth in reduced agar was not restored in strain 6-26 by either pJFP29 or pJFP30. The soft-agar growth of 6-26 (pJFP29) and 6-26 (pJFP30) was distinctive, however, in that isolated colonies were observed growing in reduced agar (Fig. (Fig.4A).4A). This phenotype was not exhibited by mutant 6-26 (pVA838). (Smaller spots apparent in some of the other tubes are bubbles in the agar rather than colonies. These were easily distinguished by the lenticular disk morphology of the colonies.)
PCR analysis with chromosome-specific primers (ARTR-SalI-24495-F and ARTR-SalI-20031-R) was used to assess the genomic nrdD region in selected 6-26 (pJFP29) and 6-26 (pJFP30) colonies isolated from reduced agar. Analysis of the PCR products by gel electrophoresis indicated that the transposon had been lost from the genomic nrdD in the 6-26 (pJFP30) cells (data not shown). In contrast, the size of the 6-26 (pJFP29) product indicated that the transposon was still present in the nrdD gene in these cells. When overnight cultures derived from the characterized colonies were regrown in soft agar as described above, dense growth was observed for both 6-26 (pJFP29) and 6-26 (pJFP30) throughout the tubes, indicating that the capacity for anaerobic growth had been restored in these cells (data not shown).
Despite advances in treatment, infective endocarditis mortality remains high, indicating a need for better therapeutic or prophylactic agents. The STM technique was adapted for S. sanguinis in the hope of identifying virulence factors for infective endocarditis that could serve as targets for these agents. Initially, 800 STM mutants were screened through two rounds of infection and dot blot analysis procedures using a rabbit model for endocarditis. After the primary and secondary screening, 37 avirulent and 5 hypervirulent strains were identified. Thirty-seven avirulent strains represent 4.6% of the total mutants screened in this study. In previous STM studies employing S. pneumoniae in a mouse pneumonia model, the percentage of avirulent strains ranged from 7 to 21% (22, 38, 56), although other analyses have typically produced smaller percentages (44). Coulter et al. (14) performed primary screens of S. aureus STM mutants in sepsis, abscess, and wound models and then examined six selected mutants for virulence in an endocarditis model. We are unaware of any study other than ours to perform a primary screen in an endocarditis model.
Our first STM mutant trial employed the rat endocarditis model. As in previous studies (9, 34, 53), over half of the rats inoculated did not develop infections. The rats that were infected yielded 105 to 108 CFU (data not shown). This suggested a possible strict dichotomy such that rats developed either no infection or heavy infections. The STM results shown in Fig. Fig.1A,1A, however, suggest that most infections were initiated by fewer than 40 bacteria. Surprisingly, there was no correlation between either the number of bacteria or the number of different strains recovered from a given rat and the inoculum size within the range tested. In contrast to the rat model, all 91 of the rabbits employed in this study for STM and CI assays became infected, with a mean recovery of 1.2 × 108 CFU per animal. Nevertheless, seven of these rabbits were infected with only 105 to 106 CFU. Output blots from six of the seven rabbits were missing many spots, representing STM mutants, that were present in the output blots from other rabbits infected with the same inocula. This suggests that a colonization bottleneck occasionally occurred in the rabbit model as well. Moreover, the mutants that did not infect the rabbits in the first inoculum also did not infect any of the rats (Fig. (Fig.11 and data not shown). This suggests that the rabbit model is more permissive for infection than the rat but that the virulence factors required for the infection of both species are similar.
Unexpectedly, the process of mutant creation was more problematic than mutant screening. It is not clear why most of the avirulent mutants we identified had undergone complex recombination events since this was not observed when the same transposition system was used for STM analysis of S. pneumoniae (22; A. Camilli, personal communication). Two mutants that did not produce signals in any of the inoculum or output blots (indicated by the double circles in Fig. Fig.1)1) were found to have had their signature tags deleted (data not shown). Other mutants, such as those circled in Fig. Fig.1,1, were found to have undergone recombination events that were sufficiently complex to preclude their characterization. We have not yet clarified the mechanism of recombination that led to these mutants; however, subsequent experiments in which individual ~2.5-kb amplicons of SK36 genomic DNA were used as targets for transposition produced aberrant mutants at frequencies of 0 to 100% in side-by-side reactions. This suggests either that the sequence of certain regions causes more aberrant mutants to be formed or that there is selection for aberrant mutants in some genes, perhaps because a conventional mutation would be lethal. The latter possibility is consistent with our findings that both aberrant mutants in which this was examined possessed wild-type copies of the genes surrounding the transposon in addition to the interrupted copies (51).
Despite the high rate of aberrant mutant creation, several strains that had simple transposon insertions were recovered. Strain 3-24 was identified as potentially hypervirulent, since it showed a weak signal in an inoculum blot paired with strong signals in output blots. Four additional uncharacterized mutants also appeared to have increased virulence. This is perhaps not surprising. Competition for the survival of the viridans streptococci occurs in the oral cavity, and there is likely no evolutionary pressure to maintain virulence for endocarditis per se. Moreover, in S. aureus, disruption of at least two genes has been shown to cause increased virulence for endocarditis (3, 47).
The 4-4 mutant with a disrupted bacA gene exhibited a 55-fold reduction in virulence. Bacitracin acts by binding to undecaprenyl diphosphate, which is the lipid carrier for peptidoglycan and teichoic acid synthesis (2). This binding prevents dephosphorylation to undecaprenol monophosphate, which is required for recycling of the lipid carrier for successive rounds of cell wall synthesis (2). The bacA gene product, undecaprenol kinase, is a membrane-bound isoprenol kinase that can phosphorylate cytoplasmic undecaprenol monophosphate, which overcomes the inhibition caused by bacitracin (10). In S. pneumoniae, a bacA mutant showed markedly reduced virulence in a mouse respiratory tract infection model, though the reason for virulence reduction was not determined (11). The gene was identified as a virulence factor in a later STM analysis of the same organism (38). In another close relative of S. sanguinis, S. gordonii, a bacA mutant was shown to be defective in biofilm formation (40). This study modeled the oral biofilm of dental plaque, but our findings suggest the possibility that bacA is also required for formation of the biofilm that occurs in infected vegetations (54).
The thrB gene encoding homoserine kinase was disrupted in the 4-29 mutant. This enzyme catalyzes the first committed step for synthesis of threonine and isoleucine (4). Both are essential amino acids for humans who do not possess a homoserine kinase. Therefore, this enzyme has been proposed as a new antibiotic target (4). The thrB gene was identified as a virulence factor in two independent STM analyses of S. aureus (14, 45), while thrC, also involved in threonine synthesis, was identified in S. pneumoniae (38). Purine biosynthesis, which was disrupted by an insertion in the purB gene in mutant 9-10, has also been commonly identified in STM studies. The purB gene was identified in S. pneumoniae (38) and Streptococcus iniae (46) STM studies, while purL was identified in three S. pneumoniae (22, 38, 56) and two S. aureus (14, 45) STM studies.
Mutant 6-26 has a transposon insertion in the gene encoding a class III, or anaerobic, RNR—another enzyme that is not present in humans. RNRs are essential for de novo deoxyribonucleotide synthesis (30). Analysis of the unfinished genome suggests that S. sanguinis possesses RNRs of classes Ib and III, which is also true for related bacteria such as S. aureus and S. pyogenes (43). Class I RNRs require O2 for their activity, whereas class III enzymes are permanently inactivated by O2 (30). Masalha et al. (43) examined both enzymes in S. aureus and found that anaerobic RNR mutants grew more slowly than wild-type S. aureus cells under anaerobic conditions, though growth was not completely eliminated. It was suggested that growth in these mutants was due to trace amounts of oxygen in the medium, since addition of the class I RNR inhibitor, hydroxyurea, further reduced growth in these cultures. We also found it difficult to eliminate the growth of S. sanguinis RNR mutants under standard anaerobic conditions, suggesting that the class Ib RNR requires only trace amounts of oxygen to support growth. RNR genes have not been identified in STM studies employing streptococci or staphylococci, though a recent STM analysis of Proteus mirabilis identified an nrdD ortholog as important for virulence in a urinary tract infection model (8).
Oxygen levels in vegetations have not to our knowledge been measured. Coulter et al. (14) determined that five of five unidentified anaerobic S. aureus mutants were attenuated for virulence in the rabbit endocarditis model, suggesting that the interior of the vegetation becomes anaerobic despite the oxygenated blood flowing past. The moderate but significant reduction of virulence seen for the 6-26 mutant could suggest either a spatial or temporal gradient of oxygen within the vegetation or a partial ability to overcome the loss of anaerobic RNR activity. Some Bacillus species that apparently possess only class I RNRs can use external DNA as a source of deoxyribonucleotides under anaerobic conditions, thereby overcoming the requirement for anaerobic ribonucleotide reduction (19). Given that S. sanguinis is also naturally competent and that the biofilm of the vegetation may represent a site of increased natural competence (39), DNA uptake may have aided the survival of the RNR mutants in the CI assays.
Our studies also suggest a model for the transcriptional organization of the nrdD locus in S. sanguinis. Examination of the incomplete genomic sequence of S. sanguinis suggests that the nrdD gene is preceded upstream by a putative membrane protein gene and downstream by two acetyltransferase genes and nrdG, encoding the beta subunit of the RNR holoenzyme (30) (Fig. (Fig.3).3). The failure of pJFP29 to complement the nrdD mutation in strain 6-26 suggests that the transposon mutation was polar on the expression of downstream genes. Complementation of the in-frame nrdD deletion mutant JFP27 by pJFP29 but not pJFP30 suggests that the nrdD promoter is contained within pJFP29. These conclusions suggest an explanation for the colonies observed in anaerobic agar in some strains (Fig. (Fig.4).4). It appears that in 6-26 (pJFP30), occasional recombination between the chromosomal and plasmid-borne nrdD genes caused restoration of the genomic nrdD gene and transcription of the remainder of the operon, resulting in colony formation in reduced agar. This likely was not observed in JFP27 (pJFP30) because of the limited sequence shared by the genomic and plasmid-borne loci (72 and 260 bp at the 5′ and 3′ ends of nrdD, respectively) (Fig. (Fig.3).3). The retention of the transposon in the genomic nrdD gene in colonies of mutant 6-26 (pJFP29) suggests that when nrdD is expressed from the pJFP29 plasmid, restoration of growth results from an uncharacterized mutation that restores expression of the chromosomal genes downstream from nrdD. Further studies will be needed to confirm these hypotheses. The regions preceding nrdD from S. sanguinis SK36, two strains each of S. pyogenes and S. pneumoniae, S. mutans, and S. agalactiae were compared (see Fig. 2 in the supplemental material). With the exception of SK36, all the species shared potential −35, −16, and −10 promoter sequence elements (65) and other potential promoter elements in the 59 bp upstream of nrdD. A homologous promoter region was not found in SK36, suggesting that a unique promoter may regulate nrdD expression in S. sanguinis.
These findings indicate that housekeeping functions such as cell wall synthesis, amino acid and nucleic acid synthesis, and the ability of the bacteria to survive in anaerobic conditions are important virulence factors for S. sanguinis endocarditis. Given that most of the enzymes involved in these processes are not found in humans, this study suggests several possible new drug targets. The finding that the bacA, thrB, and purB genes identified here have also been identified by STM in models ranging from pneumonia in mice (22, 38, 56) to systemic infections in zebra fish (46) suggests that they have a general role in virulence in gram-positive cocci.
There are two groups of genes not identified in this screen that we might have expected to find based on previous studies. The first is composed of the adhesins mentioned in the introduction as having been shown or suspected to be important for endocarditis. One possible explanation for this is the limitation inherent in STM, that certain mutants may be complemented by the other strains in the inoculum, thus masking their virulence defects (24). Indeed, few adhesins have been identified in previous streptococcal STM screens and this may be the reason (22, 29, 38, 46, 56). The second group is composed of those genes identified as induced in S. gordonii in the rabbit model of endocarditis, employing in vivo expression technology (33). This could be explained by the observed tendency for this technique to identify genes different from those identified by STM (13). The failure to find either class of genes in our screen could also simply result from the analysis of too few mutants. We recently created a mutation in the ssaB gene (20), which is an ortholog of the fimA and sloC genes required for endocarditis virulence in other viridans streptococci (9, 53). This mutant appeared avirulent by STM and produced a value of 0.00016 in a competitive index assay (Das et al., unpublished data). This suggests that other S. sanguinis virulence genes remain to be identified. It also suggests the feasibility of using STM to test strains with targeted mutations in suspected virulence factors to determine their role in endocarditis virulence. We are currently pursuing this approach.
We thank A. D. Baughn and M. H. Malamy of Tufts University and J. P. Lewis and F. L. Macrina of Virginia Commonwealth University for helpful conversations regarding anaerobic methodology, M. J. Climo for assistance initiating the rabbit endocarditis model, Taisei Kanamoto for assistance with CI experiments, Sarah Maziarz for the initial characterization of mutant 4-4, Al Best for statistical analyses, and Nicai Zollar for technical assistance.
This work was supported by grants 9807751U from the American Heart Association and R01AI47841 and K02AI054908 from the National Institutes of Health to T.K. and by a scholarship from the United States Air Force Institute of Technology to J.C.N.
Editor: J. N. Weiser
†Supplemental material for this article may be found at http://iai.asm.org/.