By mining the SK36 genome, 16 putative surface proteins with LPxTG motifs were identified. Of these proteins, we identified four putative cell-surface nucleotidases based upon the presence of an LPxTG motif and substrate specificities that suggest the ability to hydrolyze a nucleotide and water to a nucleoside and orthophosphate (Table S1). Putative Nt5e showed a substrate specificity that would most likely hydrolyze released platelet adenine nucleotides and yield adenosine. To test this possibility in strain SK36, allelic exchange mutants of nt5e and the other putative nucleotidases were each constructed using primers listed in Tables S3 (SK36) and compared for loss of hydrolysis of extracellular adenine nucleotides (Figure 1). Each strain was incubated with 10, 50, or 100 µM of ATP, ADP or AMP. Hydrolysis is reported as S. sanguinis Nt5e activity per cell. When strain SK36 was compared to deletion mutants of genes for the extracellular nuclease (nucH), cyclo-nucleotide phosphodiesterase (cnp) or DNA repair ATPase (rad3), only Δnt5e cells lost significant ability of whole cells to hydrolyze ATP (Figure 1A, P<0.05), ADP (Figure 1B, P<0.05) and AMP (Figure 1C, P<0.05) at each concentration examined. The ATPase, ADPase and AMPase activities of Δnt5e cells were also associated with a reduction in the lag time to platelet aggregation (Figure S1).

Platelets from some donors do not aggregate in response to strain SK36, reflecting donor specificity [21]. To determine whether the enzymatic activities were expressed by other platelet-aggregating strains, we also studied S. sanguinis 133-79. Whole cells of S. sanguinis 133-79 wild-type (wt) hydrolyzed the adenine nucleotides ATP (Figure 2A), ADP (Figure 2B) and AMP (Figure 2C), following Michaelis-Menten kinetics. The K_m and V_max for the different strains tested are summarized in Table 1. Using the Δnt5e strain, the ATPase activity was significantly decreased (P<0.01), whereas activity in this mutant was restored to wt levels by complementation (nt5e+) (Figure 2A). In the absence of nt5e, the ADPase activity was reduced, whereas activity in this mutant was restored to wt levels by complementation (nt5e+) (Figure 2B). Similarly, AMPase activity in S. sanguinis 133-79 was fully abrogated by deletion of nt5e (Figure 2C). When complemented, nt5e+ regained AMPase activity. Indeed, Δnt5e generated inorganic phosphate only with ATP as substrate, which may reflect activity of an ecto-ATPase on the cell surface of S. sanguinis as we reported [9]. When compared to NT5E, ecto-ATPase activity did not play a major role in the hydrolysis of extracellular adenine nucleotides. Collectively, these results suggest strongly that Nt5e on S. sanguinis 133-79 hydrolyzes the adenine nucleotides ATP, ADP and AMP.

We focused on the ability of S. sanguinis to produce adenosine, which could potentially affect the course of experimental endocarditis. Since our assays were performed with whole cells, we considered the possibility that total enzyme activity could include other phosphatases. A search of the S. sanguinis genome suggested the possible expression of two putative alkaline phosphatase (AP) ectoenzymes [13]. To better define the enzymes responsible for whole cell AMPase activity, the assay was run in the presence of α,β-methylene ADP (APCP), a known inhibitor of Nt5e [22] (Figures 2D and 2E). When expressed on a logarithmic scale, the APCP dose-inhibition curve was fitted to a sigmoidal, single-site model (Figure 2D), allowing determination of the half-maximal inhibitory concentration (IC₅₀ = 8.8±3.8 µM). In the presence of APCP (K_m = 66.7±14.1 µM; V_max = 0.9±0.1 nmoles Pi/min per 10⁶ cells), the whole cell AMPase activity showed an apparent increase in K_m (P<0.05), but no increase in the V_max (P = 0.3) compared to the no inhibitor control (K_m = 38.9±15.6 µM; V_max = 1.0±0.2 nmoles Pi/min per 10⁶ cells). Since the inhibition could be overcome at high concentrations of AMP substrate, APCP competitively inhibited AMPase activity (Figure 2E). Furthermore unlike Nt5e, AP is sensitive to an alkaline pH optimum [23]. In S. sanguinis 133-79, the AMPase activity did not significantly increase with increasing pH (Figure 2F). Therefore, the AMPase activity was not AP, but attributable to Nt5e.

To show that Nt5e is a cell wall-associated enzyme, we characterized a cell surface protein fraction with the requisite enzyme activities. Recovery of the cell surface proteins fragments from S. sanguinis 133-79 was maximal after 7 minutes of TPCK-trypsin digestion (data not shown). The 7-minute tryptic digests (crude digest) of S. sanguinis 133-79 were chromatographed on a column of Sephadex G-100, pooled (Figure 3A), fractions resolved using SDS-PAGE (Figure 3B), and analyzed for platelet interactions (Table 2). Fraction 3 (G100-3) had the greatest ability to inhibit S. sanguinis-induced platelet aggregation, but had no effect on platelet-S. sanguinis adhesion (data not shown). After separation of Fraction 3 by two-dimensional SDS gel electrophoresis and analysis by mass spectrometry, two putative 5′-nucleotidase superfamily proteins were identified (Table 3) [13]. The crude digest and fraction G100-3 showed 5′-nucleotidase activities (Table 2). Compared with the crude digest, G100-3 had higher ATPase, ADPase and AMPase activities, which correlated with the inhibition of platelet aggregation (Table 2). In the plasma, both ADP removal and adenosine generation can inhibit platelet aggregation [24]. Taken together, 5′-nucleotidase is a protein cleaved from the surface of S. sanguinis 133-79, which appeared to metabolize adenine nucleotides to modulate platelet aggregation.

We next compared whole cells of S. sanguinis 133-79 wt, Δnt5e and nt5e+ strains for the ability to induce platelet aggregation. In response to either wt or nt5e+, platelets from a representative donor aggregated in approximately 9 minutes (Figure 3C). The aggregation response of Δnt5e showed a lag-time of approximately 5 minutes, which was significantly shorter than the wt and nt5e+ strains (Figure 3D). The magnitude of aggregation, however, was similar in response to all three strains (Figure 3C). The aggregation responses to S. sanguinis 133-79 wt, Δnt5e and nt5e+ were consistent for all three platelet donors tested (data not shown). Similar results were obtained using S. sanguinis SK36 (Figure S1). Therefore, S. sanguinis Nt5e prolonged the lag time to the onset of platelet aggregation.

Since the ability to induce aggregation is a function of adhesion of S. sanguinis to platelets [25], we next sought to determine whether deletion of NT5E affected aggregation because adhesion was affected. The percent adhesion to platelets was similar for wt (133-79: 60.1±0.9%; SK36: 60.1±1.9%), Δnt5e (133-79: 59.9±1.2%; SK36: 56.5±1.4%) and nt5e+ (133-79: 59.6±2.6%) strains. Hence, NT5E modulates platelet aggregation without affecting adhesion.

To investigate the contribution of Nt5e to S. sanguinis virulence in infective endocarditis, rabbits with experimental heart valve injury were infected by intravenous inoculation with 1×10⁹ CFU S. sanguinis 133-79 wt, Δnt5e, or nt5e+. Growth curves in TH broth were performed for all three strains to test for changes in cell growth or fitness. Although the nt5e deletion mutant showed a slight delay in entry to log-phase, all strains displayed similar growth rates in TH broth maintained in air with 5% CO₂ (not shown). The resulting vegetations ranged from non-apparent (Figure 4A) to macroscopic lesions (Figures 4B and 4C).

Four days after inoculation with S. sanguinis, bacterial load recovered from the vegetations ranged from 10³ to 10⁹ per rabbit; the vegetative masses correlated with the recovered bacterial CFUs (Figure 4D; R² = 0.66, n = 31). Eight of 31 infected rabbits had no visible vegetations, which might due to insignificant injury caused by catheter placement. Total recovered bacterial CFUs indicated the relative ability of each strain to colonize and proliferate on the heart valve.

As expected, 11 out of 12 rabbits infected with wt formed aortic vegetations with a mean mass of 18.6 mg and mean recovered bacterial load of 1.0×10⁸ CFU. Similarly, after infection with nt5e+, vegetations formed in 7 of 8 rabbits with a mean mass of 12.5 mg and mean bacterial load of 0.2×10⁸ CFU on TH plates, and 0.3×10⁶ CFU on TH plates with appropriate antibiotics. The CFUs recovered with and without antibiotics were not significantly different. In contrast, 6 of 11 rabbits injected with Δnt5e showed no vegetations. Two other rabbits died before euthanasia and were excluded from statistical analysis. In rabbits challenged with Δnt5e, the mean weight of vegetations was 4.0 mg and the mean bacterial load was 0.7×10⁶ CFU on non-selective and 0.1×10⁶ CFU on selective plates, which were the lowest infectious loads among these three groups. Taken together, Δnt5e-infected rabbits developed significantly smaller aortic vegetations (P<0.01) with significantly lower bacterial CFUs (P = 0.01) than rabbits challenged with the wt strain. Notably, the Δnt5e group showed an approximately 100-fold reduction in mean CFUs as compared with the wt group (Figure 4E). After rescuing the nt5e gene, nt5e+ and wt infected rabbits showed comparable aortic vegetations (P = 0.32) and bacterial CFUs (P = 0.55).

S. sanguinis strains were routinely grown in Todd Hewitt broth (TH broth, Difco; Sparks, MD) or on TH agar plates at 37°C in 5% CO₂. E. coli cells were grown aerobically at 37°C in Luria-Bertani broth (LB broth, Bacto; Sparks, MD). When required, antibiotics were added to the medium at the indicated concentrations: erythromycin (Em), 10 µg ml⁻¹ (S. sanguinis); and kanamycin (Km), 50 µg ml⁻¹ (E. coli) or 400 µg ml⁻¹ (S. sanguinis).

Standard recombinant DNA techniques were employed as described [43]. Plasmids (listed in Table S2) were purified from E. coli cells using the QIAquick Spin Miniprep purification Kit (Qiagen Inc., Valencia, CA). Oligonucleotides for strain SK36 (Table S3) and 133-79 (Table S4) were synthesized for deletion (A1 and A2) and complementation (ACom) using integrated DNA Technologies.

Chromosomal DNA was prepared from mutanolysin-treated streptococcal cells using the Qiagen 100/G Genomic Tip system [45]. PCR products were purified using the High Pure PCR Product Purification Kit (Roche Ltd., Indianapolis, IN). DNA restriction and modification enzymes were used according to the manufacturer’s directions (New England Biolabs Inc., Ipswich, MA).

The target genes of S. sanguinis (wt) were inactivated by allelic exchange with the erythromycin-resistance determinant, ermAM. Briefly, ermAM was amplified from plasmid pVA891 [46] and cloned into pPCR-Amp SK (+) (Stratagene Corp., La Jolla, CA). Two DNA fragments constituting the flanking sequences of the target genes were then amplified and fused with the ermAM genes sequentially [47]. The fused construct was then PCR-amplified, purified and transformed into wt. For transformation, overnight cultures were grown in TH broth. The next day, cells were inoculated into fresh aliquots of the same medium (1∶40), containing 10% heat-inactivated horse serum (Sigma-Aldrich, St. Louis, MO), and the fused DNA construct was added. Competence stimulating peptide (CSP, a gift from Dr. Jens Kreth, University of Oklahoma) was added to a final concentration of 200 ng ml⁻¹. Incubation continued for 5 h at 37°C and cells were plated on Em selective TH plates, generating the deletion mutant. The mutation was confirmed by PCR amplification and sequencing.

To complement the deletion mutant, a DNA fragment encompassing the entire target gene and upstream promoter sequence was amplified by PCR from wt and cloned into the E. coli-streptococcal shuttle vector pDL276 [48], generating plasmid pDL276-gene. The construct was amplified in E. coli, purified and used to transform the deletion mutant to obtain the complemented strain using the method described above.

Streptococcal cells in stationary phase were harvested from overnight cultures. Harvested cells were washed twice with 30 mM Tris•HCl buffer (pH 7.4) containing 0.25 mM ethylenediaminetetraacetic acid (EDTA) and 30 mM sodium chloride (NaCl), followed by washing with 50 mM Tris•HCl (pH 7.4) containing 130 mM sodium chloride (NaCl) and 5 mM magnesium dichloride (MgCl₂), and resuspended to 2×10⁹ cells per ml. For phosphate hydrolase activity, streptococcal cells were washed twice with 50 mM Tris•HCl (pH 8.0) buffer containing 150 mM NaCl, 5 mM CaCl₂, and 5 mM MgCl₂. The bacterial suspension (0.5 ml) was mixed with AMP at final concentrations up to 100 µM in a 2 ml microcentrifuge tube, or with up to 400 µl ADP or ATP, and incubated at 37°C for 30 minutes. After incubation, cells were centrifuged at 10,000×g for 5 minutes, and 50 µl of the supernatant was transferred into 96-well plates (Corning, N.Y.). Similarly, the crude tryptic digest or fractions from gel filtration (final concentration 10 µg/ml) were incubated with 50 µM AMP, ADP or ATP at 37°C for 15 minutes. The reactions were stopped with an equal volume of HCl (final concentration 0.1N). The final solution (50 µl) was then transferred into 96-well plates. The enzymatic activity was measured as the amount of inorganic phosphate (Pi) released into the supernatants using the QuantiChrom Phosphate Assay Kit DIPI-500 (Bioassay systems, Hayward, CA). The results were expressed as nM of Pi produced/min per 10⁶ cells or nM of Pi produced/min per ng of protein. The K_m (Michaelis constant) and V_max (maximum velocity) for AMPase activity of intact S. sanguinis cells were calculated from substrate concentration curves using nonlinear regression of four replicates for each concentration point. To minimize the likelihood that Pi was generated by other enzymes, some streptococcal cells were pretreated with 1 µM to 1 mM tetramisole (Sigma-Aldrich, St. Louis, MO), an inhibitor of alkaline phosphatase, or 1 µM to 1 mM adenosine 5′-[α,β-methylene] diphosphate (APCP) (Sigma-Aldrich, St. Louis, MO), an inhibitor of mammalian Nt5e, before incubating with AMP.

Minimal tryptic digests of S. sanguinis 133-79 were prepared as described [49], which leaves the cell wall intact. The crude digests representing minimally digested surface proteins were concentrated and desalted using an ultrafiltration column (10 kDa cutoff; Millipore, Billerica, MA) into 2 mL deionized water (dH₂O). The salt-free protein fragment concentrates were then chromatographed on a column (1.25×95 cm) of Sephadex G-100 (GE Healthcare, Pittsburgh, PA) at a flow rate of 0.3 ml/min in PBS. The fraction with the greatest ability to inhibit S. sanguinis-induced platelet-rich plasma (PRP) aggregation (G100-3) was then analyzed using two-dimensional SDS gel electrophoresis. Gels were stained with silver stain and spots were excised for mass spectrometry analysis (Center for Mass Spectrometry and Proteomics, University of Minnesota).

Strains of S. sanguinis were tested for the ability to induce platelet aggregation using fresh PRP obtained from a single donor as described previously [24]. A single donor was used to eliminate variability in platelet response between donors [21] and the procedures were reviewed and approved by the IRB of the University of Minnesota. Each bacterial strain (50 µl suspension containing 4×10⁹ cells/ml) was incubated with 450 µl of PRP (4×10⁸ cells/ml). PRP aggregation was performed at 37°C with controlled stirring in a recording aggregometer (model 660, Chronolog Corp., Havertown, PA), and the lag time or delay to onset (minutes) was measured.

All procedures were performed as described previously [24]. In brief, platelets from outdated PRP (Memorial Blood Center, St. Paul, MN) were washed with PBS and fixed with 10% formalin. Washed platelets and washed streptococcal cells were incubated together or alone (controls) in microwells; the small clusters of adhering platelets and bacteria were separated from non-interacting particles by centrifugation. The sedimentation of adhering mixtures relative to controls was quantitated by the following formula: percent adhesion = 100×{1–[mixture A_{620 nm}/(bacterium A_{620 nm}+washed-platelet A_{620 nm})/2]}. Based on previous studies of the variability of the method, only adhesion scores of ≥20% were considered positive.

Injury-induced experimental endocarditis was initiated by placement of a catheter into the left side of the heart in healthy, adult New Zealand White rabbits (2 to 3 kg; obtained from Bakkom Rabbitry, Red Wing, MN), essentially as described previously [4], [34]. The procedures were reviewed and approved by the IACUC of the University of Minnesota. The catheter was retained in place for 2 hours to abrade the heart valve and then removed. After closure of the neck incision, viable S. sanguinis was injected intravenously via the marginal ear vein. A total of 33 rabbits were inoculated with 1×10⁹ of S. sanguinis 133-79 wt (n = 12 rabbits), nt5e deletion mutant (Δnt5e) (n = 13) and complemented strain (nt5e+) (n = 8). After four days, the animals were euthanized, hearts were removed, aortic valves were excised and vegetations were weighed. To determine infecting bacterial colony forming units (CFUs), vegetations from each rabbit were homogenized separately in 1 ml of TH broth and plated onto TH agar and CFUs were enumerated.

To determine whether antibiotic resistance markers were lost during the four-day infection in vivo without antibiotics, samples from vegetations were also enumerated on replica plates with appropriate antibiotics. For Δnt5e, the CFUs on TH plates with or without antibiotics were not significantly different (P = 0.25), indicating that resistance markers were retained. In contrast, the nt5e+ group tended to show reduced growth on antibiotic-containing medium (P = 0.15), suggesting loss of the complementation plasmid. Less than 1% of the bacteria enumerated from the nt5e+ group still retained the complemented plasmid.

When vegetations were not visualized, all of the aortic valve leaflets were scraped and cultured to enumerate the bacteria colonizing the valves. Since all bacteria recovered from the aortic valve potentially contributed to the infection and vegetation formation, CFUs on TH plates without antibiotics were used for statistical analysis. All experiments were conducted under the established guidelines of the University of Minnesota Institutional Animal Care and Use Committee.

We have partially sequenced the genome of S. sanguinis 133-79. The putative nt5e gene of S. sanguinis 133-79 and S. sanguinis SK36 [13] shared 95% sequence identity. The annotated genome of S. sanguinis SK36 (NC_009009) is available at http://www.ncbi.nlm.nih.gov. The GenBank accession number for the S. sanguinis 133-79 nt5e gene is BankIt1529021 Seq1 JQ920433.

Descriptive statistics, including the means and standard errors, were calculated. Total CFUs were converted to log₁₀ values prior to statistical analysis. Statistical analysis of data was performed using the Student’s t-test, one-way analysis of variance (ANOVA), non-linear regression or 4-parameter logistic regression (4-PL) with GraphPad Prism 5 (GraphPad Software, La Jolla, CA). An α = 0.05 was considered to be statistically significant.