Streptococcus sanguinis Class Ib Ribonucleotide Reductase

Background: Class Ib ribonucleotide reductase (RNR) is an essential enzyme for aerobic growth of S. sanguinis. Results: Its manganese form is 3.5-fold more active than the iron form when assayed with NrdH and thioredoxin reductase. Conclusion: This specific activity is the highest reported to date for the class Ib RNR. Significance: Our studies suggest why manganese is important in streptococcal pathogenesis. Streptococcus sanguinis is a causative agent of infective endocarditis. Deletion of SsaB, a manganese transporter, drastically reduces S. sanguinis virulence. Many pathogenic organisms require class Ib ribonucleotide reductase (RNR) to catalyze the conversion of nucleotides to deoxynucleotides under aerobic conditions, and recent studies demonstrate that this enzyme uses a dimanganese-tyrosyl radical (MnIII2-Y•) cofactor in vivo. The proteins required for S. sanguinis ribonucleotide reduction (NrdE and NrdF, α and β subunits of RNR; NrdH and TrxR, a glutaredoxin-like thioredoxin and a thioredoxin reductase; and NrdI, a flavodoxin essential for assembly of the RNR metallo-cofactor) have been identified and characterized. Apo-NrdF with FeII and O2 can self-assemble a diferric-tyrosyl radical (FeIII2-Y•) cofactor (1.2 Y•/β2) and with the help of NrdI can assemble a MnIII2-Y• cofactor (0.9 Y•/β2). The activity of RNR with its endogenous reductants, NrdH and TrxR, is 5,000 and 1,500 units/mg for the Mn- and Fe-NrdFs (Fe-loaded NrdF), respectively. X-ray structures of S. sanguinis NrdIox and MnII2-NrdF are reported and provide a possible rationale for the weak affinity (2.9 μm) between them. These streptococcal proteins form a structurally distinct subclass relative to other Ib proteins with unique features likely important in cluster assembly, including a long and negatively charged loop near the NrdI flavin and a bulky residue (Thr) at a constriction in the oxidant channel to the NrdI interface. These studies set the stage for identifying the active form of S. sanguinis class Ib RNR in an animal model for infective endocarditis and establishing whether the manganese requirement for pathogenesis is associated with RNR.

Streptococcus sanguinis is a causative agent of infective endocarditis. Deletion of SsaB, a manganese transporter, drastically reduces S. sanguinis virulence. Many pathogenic organisms require class Ib ribonucleotide reductase (RNR) to catalyze the conversion of nucleotides to deoxynucleotides under aerobic conditions, and recent studies demonstrate that this enzyme uses a dimanganese-tyrosyl radical (Mn III 2 -Y ⅐ ) cofactor in vivo. The proteins required for S. sanguinis ribonucleotide reduction (NrdE and NrdF, ␣ and ␤ subunits of RNR; NrdH and TrxR, a glutaredoxin-like thioredoxin and a thioredoxin reductase; and NrdI, a flavodoxin essential for assembly of the RNR metallocofactor) have been identified and characterized. Apo-NrdF with Fe II and O 2 can self-assemble a diferric-tyrosyl radical (Fe III 2 -Y ⅐ ) cofactor (1.2 Y ⅐ /␤ 2 ) and with the help of NrdI can assemble a Mn III 2 -Y ⅐ cofactor (0.9 Y ⅐ /␤ 2 ). The activity of RNR with its endogenous reductants, NrdH and TrxR, is 5,000 and 1,500 units/mg for the Mn-and Fe-NrdFs (Fe-loaded NrdF), respectively. X-ray structures of S. sanguinis NrdI ox and Mn II 2 -NrdF are reported and provide a possible rationale for the weak affinity (2.9 M) between them. These streptococcal proteins form a structurally distinct subclass relative to other Ib proteins with unique features likely important in cluster assembly, including a long and negatively charged loop near the NrdI flavin and a bulky residue (Thr) at a constriction in the oxidant channel to the NrdI interface. These studies set the stage for identifying the active form of S. sanguinis class Ib RNR in an animal model for infective endocarditis and establishing whether the manganese requirement for pathogenesis is associated with RNR.
Ribonucleotide reductases (RNRs) 5 catalyze the conversion of nucleotides to deoxynucleotides, providing the monomeric precursors required for DNA replication and repair in all organisms (1). All RNRs rely on a metallo-cofactor to oxidize a conserved cysteine in the active site of the enzyme into a thiyl radical, which then initiates reduction of nucleotides (2). In the case of the class I RNRs, two subunits are required, The ␤ subunit contains a dinuclear metallo-cofactor that oxidizes this cysteine in the ␣ subunit where nucleotide reduction occurs. The oxidation is reversible and occurs over a 35-Å distance (3). Recently, two new cofactors of the class I RNRs have been characterized, which has led to their subclassification (Ia, Ib, and Ic). The class Ia RNRs use a Fe III 2 -Y ⅐ cofactor in vitro and in vivo. In contrast, the class Ib RNRs are active with both Mn III 2 -Y ⅐ and Fe III 2 -Y ⅐ (4 -8) cofactors, although the class Ic RNRs require a Mn IV Fe III cofactor (9).
In most organisms, the genes for the class Ib RNRs are found in operons. The subunits for these RNRs, ␣ and ␤, are designated NrdE and NrdF, respectively. Two additional gene products are often found within this operon as follows: NrdI, a flavodoxin that we have recently shown behaves as a flavin oxidase (10), and NrdH, a glutaredoxin-like thioredoxin (11,12). NrdF has been known for some time to be able to self-assemble an active Fe III 2 -Y ⅐ cofactor from apo-NrdF, Fe II , and O 2 in a manner analogous to the corresponding ␤ (NrdB) in the class Ia RNRs (13,14). We have recently demonstrated with the Escherichia coli and Bacillus subtilis NrdFs (4,5), as have Sjöberg and co-workers (8) with Bacillus anthracis NrdF, that a Mn III 2 -Y ⅐ cofactor can be assembled from a Mn II 2 -NrdF, but only in the presence of the reduced form of NrdI (NrdI hq ) and O 2 . An x-ray structure of a complex of E. coli NrdI hq and Mn II 2 -NrdF, in conjunction with biochemical studies, recently demonstrated that the oxidant produced by NrdI hq is channeled directly to one of the manganese ions in the Mn II 2 -NrdF (15). Recent mechanistic studies on the B. subtilis Mn III 2 -Y ⅐ cofactor assembly further established that O 2 . is the oxidant required to convert Mn II 2 -NrdF to the Mn III 2 -Y ⅐ (16). In many, but not all, of the class Ib RNRs, NrdH can re-reduce the disulfide in ␣ produced concomitant with dNDP formation (Fig. 1) and thus is required for multiple turnovers (6,11,(17)(18)(19). NrdH itself can be re-reduced in vitro by TrxR and NADPH in E. coli, B. anthracis, Staphylococcus aureus, and Corynebacterium glutamicum (6,11,20,21).
Identification of the metal cofactor (manganese and/or iron) in class Ib RNR in vivo under different growth conditions is an active area of investigation. Recently, Corynebacterium ammoniagenes, E. coli, and B. subtilis NrdFs isolated from their endogenous host organisms were all shown to contain a Mn III 2 -Y ⅐ cofactor (5,7,22,23). Human and most eukaryotic RNRs utilize class Ia RNRs exclusively, whereas pathogenic organisms, including Streptococcus sanguinis, Streptococcus pneumoniae, B. anthracis, S. aureus, and Corynebacterium diphtheriae depend on class Ib RNR for aerobic production of deoxynucleotides. Knowledge of how the biosynthetic pathways for the class Ia and class Ib RNRs differ could thus potentially lead to new targets for antimicrobial therapeutics (24).
S. sanguinis was chosen to understand the role of the manganese RNR in pathogenesis for several reasons. First, it has recently been shown in a screen for virulence factors for this organism that deletion of SsaB, a manganese transporter, severely reduces infectivity of this organism (25). Second, deletion of SodA (the manganese-dependent superoxide dismutase) does not adequately explain the reduced virulence of the ssaB mutant (26). Third, S. sanguinis contains a class Ib and a class III RNR, essential for aerobic and anaerobic growth, respectively (12). This organism can thus also serve as a model for many pathogenic organisms with class Ib RNR as the only aerobic source of deoxynucleotides (27). Fourth, based on phylogenetic analysis (28), S. sanguinis NrdF belongs to a subclass distinct from E. coli, B. subtilis, and C. ammoniagenes, and thus these studies may be informative as to whether all NrdFs utilize a Mn III 2 -Y ⅐ cofactor.
In this study, we report the cloning, expression, and isolation of NrdE, NrdF, NrdI paralogs (NrdI, FmnI, and FmnG), NrdH, and two putative thioredoxin reductases TrxR1 and TrxR2. We establish in vitro that an active cofactor of S. sanguinis class Ib RNR can be assembled with both manganese and iron and that only NrdI, not FmnI or FmnG, is essential in Mn III 2 -Y ⅐ cofactor formation in NrdF. The Mn III 2 -Y ⅐ -NrdF has activity of 6.1 s Ϫ1 , 3.5-fold higher than for the Fe III 2 -Y ⅐ -NrdF when assayed with NrdH and TrxR1. The Mn III 2 -Y ⅐ -NrdF turnover number is very similar to the E. coli class Ia RNR (Fe III 2 -Y ⅐ ) (29). Finally, an animal model for infective endocarditis is available. Data in the accompanying paper (30) show that in vivo only the manganese NrdF appears to be active. Because mammalian host organisms use a Fe III 2 -Y ⅐ for RNR activity, finding an inhibitor of NrdI/ NrdF interactions required for Mn III 2 -Y ⅐ cluster assembly could provide a new target for therapeutic intervention. H]CDP was obtained from ViTrax and diluted with CDP in buffer A to 6,000 -12,000 cpm/nmol before use. Radioactive samples were analyzed using a Beckman Coulter LS6500 scintillation counter. EPR spectra were acquired on a Bruker EMX X-band spectrometer at 77 K using a finger Dewar filled with liquid nitrogen and 706-PQ Wilmad EPR tubes or at room temperature (RT) using a Wilmad flat cell (150 l). UV-visible spectra of anaerobic samples were acquired on a Varian Cary 3 UV-visible spectrophotometer using anaerobic cuvettes (Starna) with Teflon silicon septa (Thermo Scientific) and anaerobic titrations used a 50-l gas tight syringe with a repeat dispenser (Hamilton). Fluorescence measurements were carried out using a Photon Technology International QM-4-SE spectrofluorometer and FELIX software. Manganese concentrations were measured using a PerkinElmer Life Sciences AAnalyst 600 atomic absorption spectrometer, and Fe II concentrations were measured by the ferrozine assay (31). DNA sequencing and MALDI-TOF mass spectrometry were performed at the Biopolymers Laboratory, Massachusetts Institute of Technology. All anaerobic procedures were carried out in a glove box (MBraun), and all proteins were purified at 4°C. For each protein molecular mass and the following extinction coefficients (⑀) were calculated using ExPASy: NrdF or ␤ 2 (⑀ 280 ϭ 133,620 M Ϫ1 cm Ϫ1 ); NrdE or ␣ 2 (⑀ 280 ϭ 161,140 M Ϫ1 cm Ϫ1 );   (32).
Expression and Purification of NrdI, FmnG, FmnI, Apo-NrdF, and NrdE-pET28a-nrdI was transformed into BL21(DE3) cells and expressed in LB in the presence of 50 g/ml kanamycin. The culture was grown at 37°C with shaking at 200 rpm to OD 600 ϭ 0.5, and then riboflavin was added to a final concentration of 10 mg/liter. NrdI culture was induced at A 600 ϭ 0.7-0.8 with isopropyl ␤-D-1-thiogalactopyranoside (Promega) to a final concentration of 0.4 mM and incubated at 30°C. In 4 -5 h, cells were harvested by centrifugation at 3,000 ϫ g for 10 min at 4°C, frozen in liquid nitrogen, and stored at Ϫ80°C. Typical yield was 1.8 g wet cell paste/liter of culture.
Cell pellet (18 g) was resuspended in 90 ml of buffer C, containing 1.7 mM FMN and two Complete protease inhibitor mixture tablets (Roche Applied Science). The cell suspension was passed twice through a French pressure cell at 14,000 p.s.i. and then centrifuged at 30,000 ϫ g for 20 min. Nucleic acids were precipitated by addition of streptomycin sulfate solution to a final concentration of 1.3% (w/v) with stirring for 15 min. After the solution was spun down at 30,000 ϫ g for 30 min, the supernatant was loaded onto Ni-NTA-agarose column (Qiagen, 1.5 ϫ 3.4 cm, 6 ml) pre-equilibrated with buffer C, and the column was washed with buffer C containing 100 mM NaCl and 20 mM imidazole, pH 7.6, until A 280 of the flow-through was Ͻ0.02. The protein was eluted with a 70 ϫ 70-ml linear gradient of 20 -250 mM imidazole in buffer C. NrdI-containing fractions, identified by A 280 and A 415 , were pooled, loaded onto a Q-Sepharose Fast Flow column (GE Healthcare, 2.5 ϫ 6.5 cm, 32 ml) pre-equilibrated with 100 mM NaCl in buffer C, and the column was washed with the same buffer. Flow-through fractions containing NrdI were pooled, and the protein was concentrated using an Amicon YM-10 centrifugal filter (Millipore), desalted on Sephadex G-25 column (Sigma, 1.5 ϫ 36.5 cm, 64 ml), and further concentrated to 1-1.5 mM. A typical yield was 3 mg/liter of culture, and the protein was Ͼ95% homogeneous by SDS-PAGE analysis. FmnI and FmnG were purified as described above for NrdI.
NrdF and NrdE were expressed in BL21(DE3) and purified using protocols previously optimized for E. coli class Ib RNR (10). NrdF and DNA bands partially overlap (judged by A 260 / A 280 and SDS-PAGE analysis); to remove DNA, the pooled NrdF fractions were chromatographed twice on a Q-Sepharose Fast Flow column.
Expression and Purification of NrdH, TrxR1, and TrxR2-pETSUMO-nrdH was transformed into BL21(DE3) cells and expressed in LB in the presence of 50 g/ml kanamycin. The culture was grown at 37°C with shaking at 200 rpm to OD 600 ϭ 0.6, and then the temperature was lowered to 30°C, and the culture was induced with isopropyl ␤-D-1-thiogalactopyranoside to a final concentration of 0.4 mM. In 4 h, cells were harvested by centrifugation at 3,000 ϫ g for 10 min at 4°C, frozen in liquid nitrogen, and stored at Ϫ80°C. Typical yield was 2.3 g of cell paste/liter of culture.
Cell pellet (9.2 g) was resuspended in 46 ml of buffer B containing one Complete protease inhibitor mixture tablet. The cell suspension was passed twice through a French pressure cell at 14,000 p.s.i. and then centrifuged at 20,000 ϫ g for 25 min. Nucleic acids were precipitated by the addition of a streptomycin sulfate solution to a final concentration of 0.5% (w/v) with stirring over 20 min. After the solution was spun down at 20,000 ϫ g for 30 min, the supernatant was loaded onto Ni-NTA-agarose column (1.5 ϫ 2.5 cm, 4.4 ml) pre-equilibrated with buffer B containing 30 mM imidazole, and the column was washed with the same buffer until A 280 of the flow-through was Ͻ0.02. The protein was eluted with 50 ml of 200 mM imidazole in buffer B, and 2-ml fractions were collected. Fractions containing SUMO-NrdH (21.6 kDa) were identified by SDS-PAGE (12% (w/v) acrylamide), pooled, and loaded onto a Q-Sepharose Fast Flow column (2.5 ϫ 5.5 cm, 27 ml) pre-equilibrated with 100 mM NaCl in buffer B. The column was then washed with the same buffer (300 ml) then SUMO-NrdH was eluted with 100 ϫ 100-ml linear gradient of 100 -500 mM NaCl in buffer B, and 3.6-ml fractions were collected. Fractions with high A 280 were analyzed by SDS-PAGE, pooled, and concentrated to 470 M using an Amicon YM-10 filter.
SUMO-NrdH (470 M) was divided into 0.6-ml aliquots, and each was incubated with SUMO protease (150 l, 50 M) overnight at 4°C. About 60% of the protein was cleaved under these optimized conditions. The digested SUMO-NrdH was loaded directly onto a Ni-NTA-agarose column (1.5 ϫ 2.5 cm, 4.4 ml) pre-equilibrated with 30 mM imidazole in buffer B. The column was washed with the same buffer, and 1-ml fractions were collected. Fractions containing NrdH (8.2 kDa), assessed by SDS-PAGE analysis (16% (w/v) Tricine gel, Invitrogen), were pooled and concentrated using an Amicon YM-3 filter. NrdH (2 ml) was then loaded onto Sephadex G-25 column (1.5 ϫ 33 cm, 58 ml) pre-equilibrated with buffer A containing 10 mM DTT and 1 mM EDTA, eluted with the same buffer, and concentrated to 1 mM using an Amicon YM-3 filter. The ratio A 280 /A 260 of homogeneous NrdH is 1.3. Successful removal of the tag was confirmed by MALDI-TOF MS.
To remove DTT from NrdH, required for the DTNB assay, NrdH (400 l, 1 mM) was loaded onto Sephadex G-25 column (1 ϫ 6.5 cm, 5 ml) pre-equilibrated with buffer A, and 0.5-ml fractions were collected. Fractions containing NrdH, as judged by A 280 /A 260 , were pooled and concentrated to 450 M as described above. pET28a-trxR1 (or pET28a-trxR2) was transformed into BL21(DE3) cells and trxR1 (trxR2) overexpressed in the presence of 50 g/ml kanamycin. The culture was grown at 37°C to OD 600 ϭ 0.5, and then riboflavin (10 mg/liter) was added, and the temperature was lowered to 30°C; 10 min later, the culture was induced with isopropyl ␤-D-1-thiogalactopyranoside to a final concentration of 0.4 mM. In 4 h, the cells were harvested, as above to give typical yields of 2.5 g wet cell paste/liter of culture.
Cell pellet (7.5 g) was resuspended in 38 ml of buffer A containing a Complete protease inhibitor tablet and 1.7 mM FAD. The cell suspension was passed twice through a French pressure cell at 14,000 p.s.i. and then centrifuged at 20,000 ϫ g for 25 min. Nucleic acids were precipitated by the addition of streptomycin sulfate solution to a final concentration of 1% (w/v). After the solution was spun down at 20,000 ϫ g for 30 min, the supernatant was loaded onto a Ni-NTA-agarose column (1 ϫ 4.2 cm, 3 ml), pre-equilibrated with buffer A with 30 mM imidazole, and the column was washed with the same buffer until A 280 was Ͻ0.02. The protein was eluted with 20 ml of 200 mM imidazole in buffer A. The yellow fractions were pooled, diluted 4-fold with buffer A, and loaded onto Q-Sepharose Fast Flow column (2.5 ϫ 5.5 cm, 27 ml), pre-equilibrated with buffer A and 100 mM NaCl, and the column was washed with the same buffer (ϳ150 ml). TrxR1 (TrxR2) was eluted with 100 ϫ 100-ml linear gradient of 100 -550 mM NaCl in buffer A and concentrated using an Amicon YM-30 filter; NaCl was removed by further dilution/concentration in buffer A.
Optimized Cofactor Assembly from Apo-NrdF Loaded with Fe II or Mn II -Buffers and proteins were degassed on a Schlenk line with at least five cycles of evacuation and refilling with argon. Concentrated NrdI ox (0.5-1 mM) and apo-NrdF (0.8 mM) were stable at 4°C for at least 5 days. To prepare Mn II 2 -NrdF, apo-NrdF (45 l, 800 M) was incubated with 6 eq of Mn II (2 mM solution, freshly prepared) in buffer A at 37°C for 15 min. A mixture of Mn II 2 -NrdF (60 M) and NrdI ox (120 M) in buffer A in a total volume of 600 l was titrated anaerobically with dithionite (ϳ3 mM, standardized using potassium ferricyanide (33)) until NrdI ox was completely reduced to NrdI hq (judged by the disappearance of the band at 580 nm associated with NrdI sq ). Cluster assembly was initiated by bubbling O 2 through the Mn II 2 -NrdF/NrdI hq solution for ϳ1 min. This protocol typically gave 0.6 Y ⅐ /␤ 2 . Increased yields of Y ⅐ /␤ 2 were obtained by recycling NrdI. After cluster assembly as described above, the mixture of Mn III 2 -Y ⅐ (0.6 Y ⅐ /␤ 2 ) and NrdI ox was again degassed on a Schlenck line (six cycles) and transferred inside the glove box to an anaerobic cuvette. The UV-visible spectrum was recorded to ensure that radical content remained intact, and then the mixture was titrated with dithionite (ϳ3 mM) to reduce NrdI ox to NrdI hq . O 2 was then bubbled through the sample for ϳ1 min. To remove free Mn II and NrdI, the mixture was incubated with EDTA (5 mM) at 4°C for ϳ30 min and loaded onto a Q-Sepharose Fast Flow column (1 ϫ 3 cm, 2.5 ml) pre-equilibrated with 200 mM NaCl in buffer A. The column was washed with 200 mM NaCl (buffer A, 25 ml), and Mn III 2 -Y ⅐ -NrdF was eluted with 500 mM NaCl (buffer A, 20 ml); 1-ml fractions were collected, and the protein was concentrated using an Amicon YM-30 filter. Typical radical content was 0.9 - For cluster assembly with iron, apo-NrdF (ϳ150 l, 900 M) was incubated anaerobically with 6 eq of Fe II at 37°C for 15 min and then diluted with buffer A to 60 M. Cluster assembly was initiated by addition of an equal volume of oxygenated buffer A. To remove excess iron, ferrozine (80-fold excess over NrdF) and sodium dithionite (40-fold excess) were added, and the mixture was incubated on ice for 5 min. The protein was then desalted on a Sephadex G-25 column (1 ϫ 11 cm, 8.5 ml) in buffer A. NrdF-containing fractions were pooled and concentrated. The amount of NrdF-bound Fe II was measured by the ferrozine assay (31).
Y ⅐ Quantitation-All EPR spectra used for spin quantitation were acquired under nonsaturating conditions at 77 K (4). Spin quantitation was performed by double integration of the signal and comparison with a standard of E. coli Fe III 2 -Y ⅐ -NrdB. K d for the NrdI hq and Mn II 2 -NrdF Interaction-The procedure is a minor modification of that previously reported (16). Under anaerobic conditions at 23°C, NrdI hq (240 or 360 M, ϳ 40 l) in buffer C was added in 2-l portions using an air-tight Hamilton syringe into Mn II 2 -NrdF (1 or 3 M, 700 l) in the same buffer. To ensure that NrdI hq remains reduced throughout titration, buffer C also contained 100 M dithionite. After each injection, the cuvette was inverted several times, and the solution was allowed to equilibrate at RT in the dark for 1 min, and the spectrum was recorded. The excitation wavelength was 380 nm, and the emission was measured from 475 to 625 nm in 1 nm steps with a 1-s integration time. The excitation and emission bandwidth slit was 1.5 and 0.75 mm, respectively. No photobleaching was detected using these settings. Data were analyzed by the method of Eftink (34). The titration was performed four times using different concentrations of Mn II 2 -NrdF and NrdI hq , and the data were fit in IgorPro to obtain the stoichiometry of NrdI hq binding to NrdF (n) and the dissociation constant (K d ).
General Crystallographic Methods-All crystallographic datasets were collected at the Life Sciences Collaborative Access Team (LS-CAT) or General Medical Sciences and Cancer Institutes Collaborative Access Team (GM/CA-CAT) beamlines at the Advanced Photon Source and processed using the HKL2000 software package (35). Iterative rounds of refinement and model building were performed using Refmac5 (36) and Coot (37). Ramachandran plots were calculated with Molprobity (38) and figures were generated with the PyMOL Molecular Graphics System (Schrödinger, LLC). Internal channel calculations were performed with HOLLOW (39) using a 1.4 Å probe radius. Electrostatic surface potential calculations were carried out using the PyMOL APBS plugin (40). Electron density maps were calculated with FFT (41). Table 1 reports all data collection and refinement statistics.
X-ray Structure Determination of S. sanguinis Mn II 2 -NrdF-Crystals of S. sanguinis Mn II 2 -NrdF (25 mg/ml in 20 mM Hepes buffer, 5% (v/v) glycerol, pH 7.6) were generated using the hanging drop vapor diffusion method with 25% (w/v) PEG 3000, 250 mM magnesium formate, and 100 mM Hepes, pH 7.6, as the precipitating solution. Crystals appeared after 2 weeks of incubation at RT and were prepared for data collection by mounting on rayon loops and flash freezing in liquid nitrogen following cryoprotection by brief soaking in well solution containing 25% (v/v) glycerol. X-ray diffraction datasets were processed as described above with additional scaling performed using the UCLA MBI Diffraction Anisotropy Server (42). The structure was solved by molecular replacement using BALBES (43) with the Salmonella typhimurium Fe III 2 -NrdF structure (PDB accession code 1R2F) as the search model. Eight copies of NrdF, arranged into four ␤2 dimers, are present in the asymmetric unit. The quality of the electron density map varies widely between the eight monomers. The electron density is the least well defined for two of the monomers (chains A and G) and is of the highest quality for chains B and H. The latter subunits were used to draw conclusions about the structural features of the metal-binding site and oxidant channel. To aid in model building, tight noncrystallographic symmetry restraints were used in the initial phases of model refinement and released in the final rounds. The final model consists of residues 3-287 for chain A, residues 3-286 for chains B-F, residues 4 -286 for chains G and H, two Mn II ions per NrdF monomer, and 64 water molecules. Ramachandran plots show that 99.9% of residues are in allowed and generously allowed regions.
X-ray Structure Determination of S. sanguinis NrdI ox -Crystals of S. sanguinis NrdI ox (25 mg/ml in 20 mM Hepes buffer, 5% (v/v) glycerol, pH 7.6) were generated from a commercial screen (Qiagen) using the sitting drop vapor diffusion method with 30% (w/v) PEG 4000, 200 mM ammonium sulfate, and 100 mM sodium acetate, pH 5.6 as the precipitating solution. Crystals appeared after 1 week of incubation at RT and were prepared for data collection by addition of a well solution, in a 1:1 ratio, containing 50% (v/v) glycerol to the crystallization drop followed by mounting on rayon loops and flash freezing in liquid nitrogen. The structure was solved by molecular replacement using PHASER (44) with B. subtilis NrdI ox (PDB accession code 1RLJ) as the search model. Ramachandran plots indicate that 100% of residues are in allowed and generously allowed regions. The asymmetric unit contains two NrdI monomers and the final model consists of residues 2-66, 72-154 in chain A, residues 2-154 in chain B, two FMN molecules, two sulfate molecules, and 204 water molecules. Residues 67-71 in chain A are disordered and could not be modeled. Attempts to determine the structure of reduced forms of NrdI by soaking NrdI ox crystals in 10 -100 mM solutions of dithionite produced color changes in the crystals, but structures obtained from the resulting diffraction datasets did not exhibit any significant conformational changes in response to FMN reduction.
DTNB Assay for TrxR1/TrxR2-In a final volume of 290 l NADPH (300 M), variable amounts of DTT-free NrdH (0.06 -5 M), 100 mM Tris (pH 7.5 at 20°C), and 2 mM EDTA were mixed. DTNB was added to a final concentration of 1 mM, and the mixture was equilibrated to 25°C in a cuvette. The reaction was initiated by addition of TrxR1 (17.5 nM/dimer) and monitored by change in A 412 . The turnover number was calcu-lated as described previously (45). A similar experiment was carried out with TrxR2 (17.5 nM/dimer) and NrdH (5-30 M); no change in A 412 was observed.
Activity Assays-Three sets of conditions optimized to assay S. sanguinis RNR using DTT, NrdH/DTT, or NrdH/TrxR1/ NADPH as the reductant are described. 1) For DTT a typical activity assay contained in a final volume of 170 l the following: reconstituted NrdF (0.

RESULTS
Identification of the Genes for Cluster Assembly and Activity of S. sanguinis RNR-The genome of S. sanguinis SK36 has been sequenced, and nrdH-nrdE-nrdK-nrdF were annotated in a single operon (48). A search for nrdI using B. subtilis nrdI as the query sequence revealed three candidate genes: SSA_2263 (nrdI), SSA_1668 (fmnG), and SSA_1683 (fmnI) (Fig. 2A). A general screen for genes essential under aerobic conditions identified only SSA_2263 (49). Thus, SSA_2263 was tentatively annotated as NrdI. Further analysis of this nrdI sequence, specifically the spacing between the ribosomal binding site and start codon, and ClustalW sequence alignments of characterized NrdIs suggested that Met 5 is the actual start site and that the annotated start site is incorrect (Fig. 2B and supplemental  Fig. S1). SSA_1668 and SSA_1683 were shown to bind FMN, and their genes were named fmnG and fmnI, respectively. Finally, to identify candidates for TrxR, B. anthracis thioredoxin reductase (BA5387) was chosen for a BLAST search, and two candidate genes, SSA_1865 or trxR1 (TrxR1) and SSA_0813 or trxR2 (TrxR2), were identified. Studies in the accompanying paper (30) reveal that only SSA_1865 is essential under aerobic growth conditions.
Expression and Purification of Apo-NrdF, NrdE, NrdI, FmnI, FmnG, NrdH, and TrxRs-Genes encoding NrdE, NrdF, NrdI, FmnI, FmnG, NrdH, TrxR1, and TrxR2 were amplified by PCR using S. sanguinis SK36 genomic DNA as a template. The gene for each protein was cloned into pET vectors for expression in BL21(DE3), and the sequences were verified. Apo-NrdF was expressed without a tag in pET24a. NrdE, NrdI, FmnI, FmnG, TrxR1, and TrxR2 all were expressed in pET28a with an N-terminal His 6 tag and a 10-amino acid linker (MGSSHHHH-HHSSGLVPRGSH). To obtain high yields of soluble protein, NrdH was cloned into pETSUMO and expressed as a fusion with a His 6 -SUMO tag, and subsequent to protein purification, the His 6 -SUMO tag was removed using SUMO protease. All of the proteins were purified to Ͼ95% homogeneity by conventional methods (Ni-NTA and ion exchange chromatography).
Characterization of NrdI-Phylogenetic analysis suggests that S. sanguinis NrdI is likely distinct from previously characterized NrdIs (5,10,16). Thus, NrdI (full-length and truncated by four amino acids, Fig. 2B) along with FmnI and FmnG were expressed and purified. As discussed subsequently, neither FmnI nor FmnG support Mn III 2 -Y ⅐ assembly in NrdF, whereas both the full-length and truncated NrdIs support Mn III 2 -Y ⅐ assembly in NrdF to the same extent. Given our re-annotation of NrdI, the truncated NrdI was used in subsequent experiments.
Assembly of an Active NrdF Cofactor with Iron and Manganese-For Fe III 2 -Y ⅐ , apo-NrdF was incubated anaerobically with Fe II for 15 min at 37°C followed by O 2 addition. The spectrum of the resulting product revealed a sharp feature at 408 nm and broad bands at 325 and 370 nm corresponding to the Y ⅐ (1.2 Y ⅐ /␤ 2 ) and the diferric cluster, respectively. The EPR spectrum of Fe III 2 -Y ⅐ is similar to those previously reported ( Fig. 3B, red) for E. coli, C. ammoniagenes, and S. typhimurium (4,14,52). Typically in cluster assembly studies, Mn II 2 -NrdF is generated as above and incubated anaerobically with NrdI hq , without Mn II removal, followed by exposure to O 2 . The UV-visible spectrum of the resulting NrdF (Fig. 3D, line 2) reveals a typical Y ⅐ feature at 408 nm and absorption features at 550 and 330 nm associated with the Mn III 2 cluster. Quantitative analysis typically gives 0.5-0.6 Y ⅐ /␤ 2 and 3.1-3.4 Mn III /␤ 2 . Many variables (pH, temperature, and ratio of manganese/NrdF) were examined in an effort to increase the amount of Y ⅐ /␤ 2 ; all conditions resulted in similar yields. A higher yield of Y ⅐ /␤ 2 , however, was achieved by recycling the NrdI. The NrdI ox -NrdF mixture resulting from the first effort to assemble cluster (Fig. 3D, line 2) was made anaerobic and titrated with dithionite to reduce NrdI ox into NrdI hq . The sample was then re-exposed to O 2 (Fig.  3D, line 3). This recycling process typically gives 0.9 Y ⅐ /␤ 2 and 3.7 Mn III /␤ 2 . The EPR spectrum of the Mn III 2 -Y ⅐ at 77 K is a broad signal with a line width of 150 G and is shown in Fig. 3B  (black line). Furthermore, as in the case of the other Mn III 2 -Y ⅐ -NrdFs (4,5,22), the spectrum dramatically sharpens at 4 K (data not shown). Attempts to assemble Mn III 2 -Y ⅐ cluster using FmnI hq or FmnG hq and conditions optimized for NrdI were unsuccessful.
The increase in active Mn III 2 -Y ⅐ -NrdF cluster by NrdI recycling likely mimics the assembly process in vivo where a flavodoxin reductase would recycle catalytic amounts of NrdI (5,23  NrdF. Thus, titration studies were carried out as described previously (16), and the analysis reveals a K d of 2.9 Ϯ 1 M with 1.5 Ϯ 0.4 NrdI hq per NrdF. This K d value is higher than those we previously reported for the NrdI-NrdF interactions in E. coli (K d Ͻ 0.05 M) and B. subtilis (K d ϭ 0.6 M) (16). Given the K d value and the concentrations of Mn II 2 -NrdF (60 M) and NrdI hq (120 M) used to study cluster assembly described above, Ͼ95% of Mn II 2 -NrdF is complexed. To further explore the similarities/differences among NrdIs from various organisms, an additional experiment was carried out. Previous studies in the E. coli system showed that reduction of NrdI ox in the presence of apo-NrdF (1 eq) led to formation of an anionic FMN semiquinone (ϳ30%). In a similar experiment with B. subtilis NrdF, no anionic form was detected (16). We evaluated the FMN form that accumulated upon reduction of S. sanguinis NrdI in the presence of apo-NrdF, and no anionic semiquinone flavin was observed.
Crystallographic Analysis of the S. sanguinis Mn II 2 -NrdF-Because the streptococcal NrdFs are in a phylogenetic group that has not been previously characterized, we determined the structure Mn II 2 -NrdF (2.65 Å resolution, Table 1) to compare with the corresponding structures of the E. coli and B. subtilis NrdFs. As shown in Fig. 4, the Mn II 2 -binding site strongly resembles that of the E. coli Mn II 2 -NrdF structure (15). Each Mn II ion is six-coordinate with three bridging carboxylates, including the unusual bridging mode for Glu 157 (S. sanguinis numbering), two His ligands, and two coordinated water molecules. The side chain of Asp 66 hydrogen bonds to Tyr 104 , the site of Y ⅐ formation in the active metallo-cofactor, and also resembles the linkage between the corresponding Tyr residue and the metal-binding site in E. coli NrdF. Although the moderate resolution prevents conclusive determination of the orientation of Glu 191 , the residue is modeled in a -1 , 2 binding mode, as observed in E. coli Mn II 2 -NrdF, and the refined model is consistent with this assignment. The structural similarity to E. coli Mn II 2 -NrdF is consistent with similarities observed in the EPR spectra of these proteins.
The unusual configuration of the carboxylate side chains in the E. coli Mn II 2 -NrdF structure opens a solvent-lined channel near Mn2 to the surface of protein (Fig. 5A). The channel is further accommodated by small hydrophobic residues near   (Fig. 5B). The overall cavity shape and volume are nearly identical to the E. coli NrdF channel, and electron density for water molecules is also observed (Fig. 5, C and D) Another interesting conserved feature in the S. sanguinis and E. coli Mn II 2 -NrdF structures involves a constriction in the channel immediately above Mn2. In the E. coli system, the constriction is formed by the side chain of Ser 159 and the backbone carbonyl of the bridging ligand Glu 192 , whereas in the S. sanguinis system it involves Thr 158 and Glu 191 (Fig. 5, A and B, supplemental Fig. S2). Because oxidant passage through this constriction may be a slow step in metallo-cofactor activation, substitution to a more sterically bulky residue could translate into a slower rate of reaction with the oxidant in the S. sanguinis system.
Crystal Structure of S. sanguinis NrdI ox -NrdIs have been structurally characterized thus far from several species of Bacillus (50, 51) and E. coli (in complex with Mn II 2 -NrdF) (15). As noted above, streptococcal NrdIs represent a third phyloge- netic group for which no structural information exists. Sequence alignments (supplemental Fig. S3) reveal that S. sanguinis NrdI has three distinct insertions. The 1.88 Å resolution crystal structure supports this prediction showing an extension of helix ␣1 (residues 14 -30), insertion of a new helix between the ␤-strand of ␤2 and ␤3 (␣1*, residues 40 -47), and extension of the loop between ␤3 and ␣2 near the FMN cofactor (Fig. 6A). This extended loop is 11 amino acids long (residues 65-76), making it significantly larger than the corresponding loops in E. coli NrdI (eight residues) and the Bacillus NrdIs (three residues).
The S. sanguinis NrdI ox structure contains two copies of the protein in the asymmetric unit and thus provides two independent views of this extended loop's conformation. In one monomer, it is fully ordered and extends directly away from the FMN (Fig. 6B). In the second monomer, it is partially disordered (residues 67-71 are not modeled) and folds in toward the si face of the flavin (Fig. 6C). The fully ordered view of the loop (residues 65-76) shows that it adopts a complex secondary structure composed of an extension of the ␤-sheet secondary structure from which the loop emanates followed by a series of successive turns. Participation of the loop in an extended ␤-sheet hydrogen-bonding pattern prevents the 70s loop backbone from interacting directly with the FMN cofactor in the oxidized state. This feature is distinct from what is observed in the E. coli NrdI ox structure in which the corresponding loop is flexible and glycine-rich, allowing for hydrogen bonding interactions with the N5 of the FMN via a backbone amide NH (15). The short loop (XFG) in the Bacillus NrdIs is similar to S. sanguinis NrdI ox in that neither can form this interaction. The Bacillus and E. coli NrdIs have also been structurally character-ized in the sq-or hq-reduced states (15,50,51). All of these structures reveal a backbone amide flip that positions a carbonyl group near the N5 position to interact with the now protonated form of the cofactor.
To understand whether a similar conformational change occurs upon reduction of S. sanguinis NrdI, diffraction datasets were collected on dithionite-soaked NrdI ox crystals. This treatment reduced the FMN to the sq or hq state, based on observation of color change in the crystals, but the resulting electron density maps did not reveal any significant conformational changes in the 70s loop backbone. The protonated FMN cofactor (FMNH ⅐ or FMNH Ϫ ) interacts instead with a water molecule (not observed in the NrdI ox structure). The observed outcome in the dithionite-soaked S. sanguinis NrdI crystals could be the result of crystal packing interactions that preclude conformational changes in the loop or it could indicate that the 70s loop backbone cannot, in any oxidation state, bind directly to the FMN N5 position due to the ␤-sheet hydrogen bonding pattern in the loop near the FMN (Fig. 6B).
The S. sanguinis NrdI 70s loop is also unusual in that it contains many more bulky and negatively charged residues than the corresponding regions in other NrdIs. These residues could affect the electrostatic properties of the flavin and its accessibility to O 2 . However, evaluation of the electrostatic environment in NrdI near the FMN isoalloxazine ring shows that the interior of the FMN pocket remains positively charged, similar to what is observed in E. coli and the Bacillus NrdIs (Fig. 7). The negatively charged residues in the S. sanguinis NrdI 70s loop instead generate a strong negative patch on the exterior of the protein. We propose that this patch may be involved in interaction with a flavin reductase required to recycle NrdI ox in vivo. Physiological Reductant for NrdE-The reduction of NDPs to dNDPs is accompanied by oxidation of two active site cysteines in NrdE to a disulfide ( Fig. 1) (2, 55). A number of artificial and endogenous systems are capable of mediating this re-reduction step, including dithiothreitol (DTT),Trx/TrxR/NADPH, Grx/ GSH/GR/NADPH, and NrdH/TrxR/NADPH, where Trx is thioredoxin, Grx is glutaredoxin, and GR is glutaredoxin reductase (6,11,20,21,56). The observation that nrdH is colocalized in the same operon with nrdE and nrdF in S. sanguinis and many other organisms makes this protein the most reasonable candidate for the endogenous reductant. Our initial attempts to isolate untagged NrdH were unsuccessful due to its low solubility. To overcome this problem, NrdH was fused to a His 6 -SUMO tag and then isolated by nickel-affinity chromatography. Subsequent removal of the tag with SUMO protease gave soluble NrdH, which has been used with TrxR to assay RNR. Two candidate genes (trxR1 and trxR2) for thioredoxin reductases were identified, and the corresponding proteins, TrxR1 and TrxR2, were overexpressed and purified to homogeneity by nickel-affinity chromatography. To ensure that the isolated TrxRs were fully loaded with cofactor, FAD was added to crude cell lysates prior to purification (6,57). This protocol gave homogeneous FAD-bound TrxR1 and TrxR2 with the ratio A 272 /A 455 of 6.0 and 6.4, respectively. Complete cofactor loading was further confirmed by anion exchange chromatography.
The turnover number for each TrxR was measured using NADPH, NrdH, and the DTNB assay (45). The K m value of TrxR1 for NrdH was 0.09 M and k cat ϭ 3.5 s Ϫ1 giving k cat /K m of 4.03 ϫ 10 7 M Ϫ1 s Ϫ1 . TrxR2, however, showed no activity under the same conditions. Thus, the results suggest that TrxR1 is the reductant for NrdH in vivo.
Activity of Fe III 2 -Y ⅐ and Mn III 2 -Y ⅐ NrdFs Using DTT, NrdH/ DTT, and NrdH/TrxR1/NADPH-The activity of the class Ib RNRs has been predominantly reported for the Fe-loaded NrdFs (14,21,58,59), as the role of NrdI in generation of active Mn-loaded RNRs was not elucidated until 2010. Although a number of recent studies have reported activities for Mn III 2 -Y ⅐ NrdFs, in most cases the proteins contained substoichiometric manganese loading and Y ⅐ and/or the endogenous reductant was not used (4 -6, 8). As a starting point to identify the metallo-cofactor required for S. sanguinis class Ib in cultures and in an animal model for S. sanguinis-mediated endocarditis, we have measured the activity of pure reconstituted Mn III 2 -Y ⅐ and Fe III 2 -Y ⅐ with DTT, DTT/NrdH, and the endogenous reductants NrdH and TrxR1 (Fig. 1). Some representative results of optimization of the concentrations of NrdE, NrdF, and variable reductants are shown in Fig. 8, A-C.
The K d value for NrdE-NrdF interactions in the class Ib RNRs still remains largely unknown. However, recent studies on B. subtilis class Ib RNR demonstrated that the active form is a 1:1 complex of subunits (5,47). Thus, our initial studies focused on establishing the ratio of NrdE to NrdF for maximum activity using DTT and NrdH/DTT as reductant. First, Mn III 2 -Y ⅐ -NrdF (0.9 Y ⅐ /␤ 2 ) at 0.2 M was assayed with increasing concentrations of NrdE (0.4 -4 M) and DTT in excess. Under these conditions, maximum activity is observed at 5 eq of NrdE (Fig. 8A, inset). When the assay was carried out with Mn III 2 -Y ⅐ -NrdF (0.07 M) using NrdH as the reductant, the SA of Mn III 2 -Y ⅐ was 20-fold higher relative to DTT and only 1 to 2 eq of NrdE were required for maximum activity (Fig. 8A). Differences in the NrdE/NrdF ratios required for maximal activity with DTT and NrdH could in part reflect the higher efficiency of NrdH in recycling NrdE, which results in higher concentrations of reduced NrdE available for each turnover. In almost all subsequent assays, a 1:2 ratio of NrdF to NrdE was used.
The effect of NrdH on RNR activity was also examined by maintaining a 1:1 ratio of NrdF/NrdE while increasing the concentration of reduced NrdH. The apparent K m value for NrdH is 0.4 M (Fig. 8B). Studies of the DTT requirement for NrdH cycling revealed that levels as low as 1 mM were sufficient for optimal activity. Thus with 0.07 M Mn III 2 -Y ⅐ , 10 M NrdH, and 20 mM DTT, a specific activity of 5,700 units/mg was observed. Although DTT is routinely used as a reductant to assay RNR directly and, as just noted, can recycle oxidized NrdH, the endogenous reductant for NrdH is TrxR1. Thus, DTT was replaced with 0.5 M TrxR1 and 1 mM NADPH and a specific activity of 5,000 units/mg was measured ( Table 2). The observed turnover numbers with NrdH/DTT and NrdH/ TrxR1/NADPH are very similar and considerably higher than the specific activity of Mn III 2 -Y ⅐ -NrdF reported for other organisms with their endogenous reductants (6,8). One additional variable, the concentration of 1:1 ratio NrdE and NrdF, was examined. The results, shown in Fig. 8C, reveal a V max of 6,000 units/mg and a K m of 6.4 nM.
The SA of Fe III 2 -Y ⅐ -NrdF (1.2 Y ⅐ /␤ 2 ) was also measured using the conditions optimized for the Mn III 2 -Y ⅐ and gave an activity of 170 units/mg with DTT, 1,500 units/mg with NrdH/DTT, and 1,500 units/mg with NrdH/TrxR1/NADPH. A comparison of the Mn-and Fe-loaded NrdF activities indicates that the former has a turnover number that is 3.5-fold higher than the Fe-loaded enzyme, establishing that Mn-loaded NrdF is a better catalyst.
In the assays described in Fig. 8, CDP was used as a substrate and dATP (100 M) as an allosteric effector. Our recent studies of B. subtilis class Ib RNR showed that dATP inhibits RNR activity at concentrations Ͼ5 M (47). This result was surprising because class Ib RNRs do not have an ATP cone or activity domain, the typical binding site for dATP that facilitates class Ia RNR inhibition (52,60,61). To determine whether the S. sanguinis class Ib RNR is inhibited by dATP, the activity of Mn III 2 - Y ⅐ -NrdF was measured in the presence of increasing dATP concentrations (Fig. 8D). The apparent K m is 2.4 M, and no inhibition was observed even at 1 mM dATP.

DISCUSSION
Since our discovery that NrdI plays an essential role in Mn III 2 -Y ⅐ formation, many reports have appeared describing the activity and properties of manganese-loaded class Ib RNRs (4 -8, 22). Although the identification of the endogenous reductant has been reported in some of the studies, in most cases the activity of the class Ib RNR remained very low even with the endogenous reductant (6,8). We have chosen to examine the importance of manganeseversus iron-loaded cofactor in S. sanguinis class Ib RNR because this organism causes infective endocarditis, and deletion of a manganese transporter results in loss of virulence (25). In addition, its NrdI and NrdF proteins belong to a group phylogenetically distinct from the E. coli, C. ammoniagenes, and B. subtilis class Ib enzymes, the only reported manganese-RNRs at the time we began this work.
Studies of the class Ib RNR have been hampered by the poor efficiency of Mn III 2 -Y ⅐ cluster assembly and consequently low catalytic activity of the enzyme. Our recent work on the B. sub-tilis class Ib RNR revealed that chromatographic removal of apo-and substoichiometrically metallated NrdF and identification and use of the endogenous reductant (Trx/TrxR) increased the activity of Mn III 2 -Y ⅐ -NrdF to 1,475 units/mg, 18 times that measured for RNR isolated directly from B. subtilis. Furthermore, the activity of the Mn III 2 -Y ⅐ -NrdF was 12 times that observed for Fe III 2 -Y ⅐ (125 units/mg) (47). In a similar study of the B. anthracis class Ib RNR, using the endogenous reductant Trx/DTT, the activity of the Mn III 2 -Y ⅐ was 10 times higher than the Fe III 2 -Y ⅐ . However, the overall activity of Mn III 2 -Y ⅐ -NrdF was only 65 units/mg, possibly associated with the inefficiency of manganese cluster assembly (6). With S. sanguinis RNR, the activity of 6,000 units/mg is very high, but the activity difference between the manganese-and iron-loaded NrdFs is only 3.5-fold using the endogenous reductant, NrdH/TrxR1/NADPH. This result is distinct from all other systems studied to date and raises an important question of which metal is used as a cofactor in vivo and how the organism's environment might influence the enzyme's metallation state.
Class Ib RNRs are found in actinobacteria, firmicutes, and ␣and ␥-proteobacteria, and recent studies have used bioinformatic analysis of the genomes of interest to identify candidate genes for the endogenous reductant(s) (6,47). In some organisms, nrdH is found in an operon with nrdE and nrdF, whereas in the Bacillus genus, the nrdH is in a distant location (12). Recent studies of B. anthracis class Ib RNR identified several candidates for its endogenous reductant system as follows: two Trxs, three putative NrdHs, and three TrxRs. Biochemical analysis of the reductant's efficiency (k cat /K m ) in recycling NrdE, accompanied by Western blot analysis of concentrations of the  most interesting reductants, led to the conclusion that both Trx1 and NrdH could function in this capacity but that the former is the most likely candidate in vivo (6). Importantly, the reported V max for the B. anthracis Mn III 2 -Y ⅐ (45 to 65 units/mg) was 100-fold lower than our S. sanguinis turnover number and independent of the reductants Trx1/DTT, Trx1/TrxR1, NrdH/ DTT, and NrdH/TrxR1. Finally, Trx1 had no stimulatory effect at all on the turnover number of Fe III 2 -Y ⅐ , and therefore it was concluded that the manganese-loaded NrdF was the likely form of the class Ib RNR in vivo. Our studies of the B. subtilis class Ib RNR also support this hypothesis by showing that thioredoxins TrxA and YosR (NrdH-like) give a 10-and 5-fold stimulation of Mn III 2 -Y ⅐ activity, respectively, relative to DTT, but they stimulate Fe III 2 -Y ⅐ activity only 2-fold. In contrast, S. sanguinis NrdH significantly increased activity of both Mn III 2 -Y ⅐ and Fe III 2 -Y ⅐ 20-fold and 9-fold, respectively, relative to DTT. Recently, we have measured an K m of 25 nM for the B. subtilis class Ib NrdE-NrdF (␣␤) subunit interactions, which is 4 -10fold lower than the interactions in the E. coli class Ia RNR (K d of 0.06 to 0.2 M (62)). These results suggested that class Ib RNR can be assayed using a 1:1 ratio of subunits (47), instead of with an excess of one subunit over the other to ensure complete ␣␤ complex formation (29). Our studies of S. sanguinis NrdF-NrdE, conducted using a 1:1 ratio of subunits in the concentration range from 0.001 to 0.1 M (Fig. 8C), gave a K m of 6.4 nM, similar to the B. subtilis class Ib RNR. The NrdE-NrdF interaction presents an opportunity to crystallize the active ␣␤ complex, and this work is in progress.
The interaction of S. sanguinis NrdF with its NrdI is also of interest because they belong to a third phylogenetic subgroup that remained uncharacterized until our studies (24,28,51). This subclassification based on bioinformatics is supported by the measured K d of 2.9 M for S. sanguinis NrdI-NrdF interactions, distinct from tighter interactions measured for E. coli (Ͻ0.05 M) and B. subtilis (0.6 M) systems (16). Although our crystallization efforts have not yielded a structure of the S. sanguinis NrdF-NrdI complex, crystallization of the individual proteins has been successful. As suggested by sequence alignments (supplemental Fig. S2), the coordination environment of the S. sanguinis Mn II 2 -NrdF is much more similar to that of E. coli NrdF than the B. subtilis protein. The three distinguishing features between the E. coli and B. subtilis NrdFs are the H-bonding interactions between the Tyr residue to be oxidized in the active cofactor and the aspartate residue coordinated to Mn1, the presence of a H 2 O molecule coordinated to Mn1, and the unusual bridging coordination of Glu 158 to Mn1 and Mn2 (15,53). Using all three criteria, S. sanguinis Mn II 2 -NrdF is similar to E. coli Mn II 2 -NrdF (Fig. 4). Another distinguishing feature between E. coli and B. subtilis NrdFs is the presence of a water-lined channel linking Mn2 to the FMN in NrdI containing a constriction created by Ser 159 (Fig. 5A). The channel has been proposed to provide a pathway for the oxidant from the flavin in NrdI to the metal-binding site in NrdF. The crystal structure of S. sanguinis Mn II 2 -NrdF reveals the presence of a similar channel with a Thr 158 counterpart to Ser 159 at the constriction (Fig. 5B). The substitution for a bulkier side chain at this site may provide a more stringent selectivity filter for the oxidant. This property may be particularly critical in strepto-cocci because these organisms accumulate high concentrations of H 2 O 2 (63,64), and the observed constriction may restrict NrdF Mn II 2 site access by this molecule. In addition, the presence of this constriction might explain why Mn II 2 -NrdF cannot be activated directly with H 2 O 2 itself (4,14).
The x-ray structure of S. sanguinis NrdI confirms predictions about the unique features of the streptococcal class Ib proteins. The longer sequence of S. sanguinis NrdI translates into an extended helix ␣1, an additional helix ␣1*, and an extended 70s loop when compared with structures of E. coli and B. subtilis NrdIs (Fig. 6A). These extensions could be involved in interaction with another protein, such as a NrdI reductase. A crystal structure of E. coli flavodoxin reductase (FlxR) (65) was used to construct a docking model with S. sanguinis NrdI (models generated with ClusPro (66)). As shown in Fig. 9A, a model was produced that places the long and bulky 70s loop and inserted helix ␣1* of NrdI in close contact with FlxR near its FAD cofactor. A similar model for interactions between FlxR and flavodoxin suggested that flavodoxin binds in a bowl-shaped pocket close to FAD with positively charged residues of FlxR positioned to interact with conserved negatively charged residues in the flavodoxin (65). A negative electrostatic surface potential near the FMN is a defining characteristic of flavodoxins, but NrdIs instead use positively charged residues to facilitate reduction of FMNH ⅐ to FMNH Ϫ (67,68). Based on analysis of the S. sanguinis NrdI electrostatic surface potential near its FMN cofactor, the negatively charged 70s loop does not seem to influence the positive electrostatic environment around the FMN and instead is localized at the outer surface of the protein (Fig. 7). Thus, we hypothesize that the negatively charged 70s loop may play a role in facilitating interaction between the positively charged pocket of FlxR (Fig. 9B) and the positively charged patch around FMN in NrdI.
Using the E. coli crystal structure of the NrdF-NrdI complex and sequence alignments (supplemental Fig. S2 and S3) with representative class Ib enzymes from the other subgroups, residues involved in forming the NrdF-NrdI interface were predicted for the B. subtilis and S. sanguinis systems. Interestingly, the NrdF portion of the protein-protein interface is more conserved than the NrdI portion. Therefore, differences in the K d values measured for E. coli, B. subtilis, and S. sanguinis may be due to variations in the NrdI surface (supplemental Table S3) rather than differences in the NrdF component. Moreover, electrostatic analysis of S. sanguinis NrdF revealed that the FIGURE 9. Model of an interaction between E. coli FlxR (PDB accession code 1FDR) and S. sanguinis NrdI. A, docking model shows that the 70s loop and ␣1* of NrdI may interact with FlxR. FlxR is in green; NrdI is in cyan with the 70s loop and ␣1* in blue, and the FMN (red) and FAD (magenta) cofactors are shown as sticks. The model was generated using ClusPro (66). B, electrostatic surface potential diagram of FlxR near the modeled interface with NrdI is shown contoured at Ϫ15 k B T (red) and ϩ15 kBT (blue). The FAD cofactor is shown in stick format. region surrounding the predicted NrdI-binding site is negative, which, in combination with the negatively charged 70s loop, could account for the particularly low affinity between these proteins.
An active cluster, Mn III 2 -Y ⅐ -NrdF and Fe III 2 -Y ⅐ -NrdF, can be assembled with 1 Y ⅐ /␤ 2 , and both forms exhibit relatively high activity with the physiological reductant NrdH/TrxR. However, the 3.5-fold difference in activities relative to the 10-and 12-fold differences between the Mn III 2 -Y ⅐ and Fe III 2 -Y ⅐ observed with the B. anthracis and B. subtilis NrdFs, respectively, suggests that this organism might be able to stay active in vivo with either cofactor, with loading dependent on the growth environment. The accompanying paper by Rhodes et al. (30) demonstrates in a rabbit model for infective endocarditis that a strain of S. sanguinis in which nrdI has been deleted does not colonize heart valves, unlike the WT-strain. These studies thus provide the first evidence that a class Ib NrdF requires manganese under conditions in which the organism is pathogenic. Considering that streptococci and other pathogens, including enterococci, staphylococci, and Bacillus sp., contain class Ib RNR as their only aerobic RNR, prevention of Mn III 2 -Y ⅐ formation in the class Ib RNR may be an attractive target for new antimicrobials.