Flavin Reductase Contributes to Pneumococcal Virulence by Protecting from Oxidative Stress and Mediating Adhesion and Elicits Protection Against Pneumococcal Challenge

Pneumococcal flavin reductase (FlaR) is known to be cell-wall associated and possess age dependent antigenicity in children. This study aimed at characterizing FlaR and elucidating its involvement in pneumococcal physiology and virulence. Bioinformatic analysis of FlaR sequence identified three-conserved cysteine residues, suggesting a transition metal-binding capacity. Recombinant FlaR (rFlaR) bound Fe2+ and exhibited FAD-dependent NADP-reductase activity, which increased in the presence of cysteine or excess Fe2+ and inhibited by divalent-chelating agents. flaR mutant was highly susceptible to H2O2 compared to its wild type (WT) and complemented strains, suggesting a role for FlaR in pneumococcal oxidative stress resistance. Additionally, flaR mutant demonstrated significantly decreased mice mortality following intraperitoneal infection. Interestingly, lack of FlaR did not affect the extent of phagocytosis by primary mouse peritoneal macrophages but reduced adhesion to A549 cells compared to the WT and complemented strains. Noteworthy are the findings that immunization with rFlaR elicited protection in mice against intraperitoneal lethal challenge and anti-FlaR antisera neutralized bacterial virulence. Taken together, FlaR’s roles in pneumococcal physiology and virulence, combined with its lack of significant homology to human proteins, point towards rFlaR as a vaccine candidate.


Results
FlaR bioinformatic analyses. To examine whether flavin reductase is ubiquitous among S. pneumoniae strains, the flavin reductase DNA sequence of the TIGR4 strain was compared to 29 completely sequenced genomes of S. pneumoniae. All 29 genomes were found to contain a highly similar locus to SP_RS 02775 [BLAST e-value < 10 −76 , >97% identity, >70% query coverage (>98% coverage in 27 of the genomes)].
A BLAST search carried out with WP_000580663.1 against all S. pneumoniae RefSeq proteins annotated as "flavin reductase" retrieved 63 proteins, which were subjected to multiple sequence alignment and a phylogenetic tree was constructed. The tree was used to select a representative set of 9 divergent proteins, one from each clade of the tree ( Supplementary Fig. S1). The position of a conserved flavin reductase domain is indicated, as well as three positions of conserved cysteines (equivalent to positions 66, 72 and 84 of WP_000580663.1). It is worth noting that cysteine was present in the three conserved positions in all 63 proteins, except 3 cases of substitution to tyrosine in the first or second position (data not shown).
A BLASTP search of WP_000580663.1 vs. RefSeq proteins excluding S. pneumoniae showed highly similar hits from other Streptococci species (e value < 10 −22 , 36-87% identify, query coverage >79%). Significant hits were also received from a wide range of non-Streptococci species (minimal e value was 10 −28 ), however none had a higher a percent identity than 39%. Using an e value cutoff of 0.001, no eukaryotic proteins were retrieved.
Recombinant FlaR protein production. rFlaR was overexpressed as HAT-tagged in Escherichia coli and purified under denaturing conditions. The theoretical molecular weight of the untagged protein was found to be 17.2 kDa. Separation of the HAT-tagged protein on reducing SDS-PAGE revealed a major band of 22 kDa and a minor band of 45 kDa (Fig. 1a). Non-reducing SDS-PAGE revealed the 45 kDa band to be the major band and 22 kDa to be the minor one (Fig. 1a). To identify the upper and lower bands were subjected to MALDI-TOF. Fourteen peptides, 9 derived from in-gel digestion of the lower band and 5 from the upper band, matched the database sequence of a hypothetical protein from S. pneumoniae R6 strain that is homologous to FlaR in the TIGR4 strain. The apparent molecular weight of the upper band was approximately double than that of the lower one, indicating that the upper band corresponds to a dimer of the lower band. Moreover, polyclonal antibodies, raised in rabbits against rFlaR, detected both bands by immunoblotting (Fig. 1b), none of which were detected by pre-immune sera.

FlaR cellular localization.
To validate the previous finding regarding the presence of FlaR on the bacterial surface 35 , we performed flow cytometry analysis on live R6 strain probed with a polyclonal mouse anti-rFlaR antibody. Indeed FlaR was found to be surface expressed (Fig. 1c). In addition, polyclonal rabbit antiserum recognized FlaR in the cytoplasm of WU2 WT strain in two forms: at ~17 kDa (untagged rFlaR monomer) and 34 kDa (untagged dimer; Fig. 1d). The 34 kDa band could also be observed in the bacterial cell-wall (CW) fraction SCIeNTIFIC REPORTS | (2018) 8:314 | DOI: 10.1038/s41598-017-18645-8 ( Fig. 1d) in accordance with our previous study 35 . A third, nonspecific band (~ 50kDa) was detected in the immunoblots of both the WU2 WT and WU2ΔflaR kan null mutant strains ( Supplementary Fig. S2). The cytoplasmic protein, malonyl-CoA:ACP transacylase (FabD), which is involved in lipid metabolism 36 , was used as control for successful separation between CW and cytoplasmic fractions. FabD could be visualized (33 kDa) in both the total bacterial protein extract and the cytoplasm, but not in the CW fraction. This was lately described in Mizrachi Nebenzahl et al. 2016 37 . Refolding and Solubilization of rFlaR. The rFlaR was insoluble under physiological conditions and its purification could be obtained under denaturing conditions in buffer supplemented with 8 M urea. Different refolding conditions were tested for their ability to renature the protein. Dialysis of the denatured protein against buffers at different pH levels revealed that rFlaR could be solubilized at pH ≥ 8.0 (Fig. 2a). The presence of cysteine-rich sequence suggests that the protein has a transitional metal-binding ability 38 . Thus, rFlaR was dialyzed against PBS (pH 7.3) supplemented with 2 mM salts of either transition metal or calcium and magnesium (2 mM of each). rFlaR became fully soluble in the presence of 2 mM FeSO 4 . Partial solubility could be obtained in the presence of 2 mM CoCl 2 or CuSO 4 (Fig. 2b).
FlaR is a Fe 2+ -binding protein. To further verify that rFlaR can directly bind Fe 2+ , solubilized rFlaR was incubated with ferrozine in the presence of free Fe 2+ ions. Ferrozine is well known as an effective chelator of Fe 2+ forming a complex that peaks at 562 nm 39 . During the experiment the absorbance was significantly lower (a) rFlaR was purified using Ni-NTA beads under denaturing conditions and resolved by SDS-PAGE using sample buffer with or without β-mercaptoethanol (β-ME). (b) rFlaR was separated using SDS-PAGE. The protein was transferred to a nitrocellulose membrane and probed with rabbit antiserum raised against rFlaR (lane 1) or preimmune rabbit serum (lane 2). (c) Expression of FlaR on the surface of R6 was detected by flow cytometry using mouse anti-rFlaR antibodies. (d) Cytoplasmic (cyto) and cell wall (CW) fractions prepared from WU2 were subjected to electrophoresis and immunoblotted with rabbit anti-rFlaR antiserum. Of note, this gel was cropped from Supplementary Fig. S2.
rFlaR has NADP reductase activity. Flavin-binding domains are found in proteins with oxidoreductase activity. Nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) is used as an electron donor in the enzymatic reactions of flavin reductases [38][39][40] . To test whether rFlaR is involved in flavin-dependent redox reactions, we performed a flavin reductase activity assay as previously described 40,41 . Oxidation of NADPH, measured at A 340  nm, could not be observed in the presence of iron-refolded rFlaR and flavin adenine dinucleotide (FAD) (data not shown). Significant increase in the reduction of NADP mediated by rFlaR was observed only in the presence of all the reaction components (NADP and FAD) over time (Fig. 4a, Spearman correlation, r = 1, p = 0.0014). No significant NADP reduction could be observed in the absence of either one of the components (Fig. 4a, p = 0.0001). After the first hour of the reaction in the presence Fe 2+ and rFlaR, the absorbance change at 340 nm was 0.09, equivalent to 14.5 µM/hour of generated NADPH (calculated by using ε 6.22 mM −1 cm −1 ) 42,43 . Hence, rFlaR specific activity is 40.3 µM min −1 mg −1 protein. It should be noted that iron-refolded rFlaR exhibited significantly higher NADP-reductase activity compared to rFlaR refolded at pH 8.0 in the absence of Fe 2+ ions. BSA was used as a negative control (Fig. 4b, p = 0.0001). The NADP reductase activity of rFlaR refolded at pH 8.0 was still significantly higher in comparison to the negative control (Fig. 4b, p = 0.02). rFlaR NADP reductase activity is Fe 2+ dependent. To evaluate the importance of Fe 2+ for rFlaR enzymatic activity, different divalent ion chelators were introduced into the reaction mixture. EDTA (10 nM), Nitrilotriacetic acid (NTA; 10 nM) and citrate (10 nM) all significantly inhibited rFlaR NADP reductase activity ( Fig. 5a; p = 0.0001). To rule out the possibility that Fe 2+ is oxidized by H 2 O 2 , NADP reductase activity was measured in the presence of the H 2 O 2 scavenger, salicylate. Indeed, salicylate did not inhibit NADPH production by rFlaR (Fig. 5b), indicating that H 2 O 2 does not affect FlaR activity directly.
Fe 2+ is the electron donor in the rFlaR-NADP reductase activity. The next step was to determine the electron donor in rFlaR reductase activity. When cysteine was used as a reducing agent, the rFlaR FAD-dependent NADP reduction was enhanced in a concentration-dependent manner (Fig. 6a, Spearman correlation, r = 1, p = 0.0083). To detect which component in the reaction mixture was reduced cysteine was applied to solutions containing either ammonium iron sulfate, or FAD or NADP. Ammonium iron sulfate incubated for 2 hours at room temperature was found to be significantly oxidized (Fig. 6b, p = 0.0001). Addition of cysteine reduced the oxidized Fe 3+ to Fe 2+ , as detected by the ferrozine stain ( Fig. 6b; p = 0.0001). Both NADP and FAD were tested in their oxidized forms and no reduction in NADP and FAD could be observed in the presence of cysteine. These findings suggest that the stimulation of rFlaR activity occurs through the recycling of Fe 3+ to Fe 2+ by cysteine. To verify this point, excess Fe 2+ ions were supplemented into the reaction mixture containing rFlaR, FAD and NADP. Indeed, rFlaR NADP reductase activity was stimulated in the presence of Fe 2+ in a concentration-dependent manner (Fig. 6c, Pearson correlation, r = 0.9947, p < 0.0001).
FlaR contributes S. pneumoniae resistance under oxidative stress conditions. To determine the importance of FlaR to S. pneumoniae resistance to oxidative stress, flaR null mutant (WU2ΔflaR Erm ) and complemented strains (WU2ΔflaR flaR/Erm/Kan plasmid and WU2ΔflaR flaR/Erm/Kan chromosome) were constructed as FlaR contributes to S. pneumoniae virulence. To test whether FlaR contributes to S. pneumoniae virulence, BALB/c mice were inoculated intraperitoneally (IP) with WU2 WT or WU2ΔflaR Kan strains. Increased survival was observed in mice inoculated with WU2ΔflaR Kan compared to inoculation with the WT WU2, though it did not reach statistical significance (Fig. 8a). In an attempt to increase signal-to-noise responses, IP inoculation was performed using CBA/CaHN-Btkxid/J mice (CBA/N xid ). This mouse strain carries a mutation in the Bruton kinase (btk) gene that renders it highly susceptible to S. pneumoniae infection 44,45 . Significantly increased survival rates were observed in CBA/N xid mice inoculated IP with WU2ΔflaR Kan compared to mice inoculated with WU2 WT (Fig. 8b, p = 0.0236).

Mechanisms underlying FlaR contribution to S. pneumoniae virulence. The reduced virulence in
the peritoneum in the absence of FlaR may result from two different mechanisms: increased sensitivity to oxidative stress elicited by immune cells or reduced adhesion and spread in the host. To test the first possibility, the extent of phagocytosis of WU2 WT, WU2ΔflaR Erm and WU2ΔflaR flaR/Erm/Kan complemented strains was evaluated. Interestingly, no significant differences in the phagocytosis by primary mice macrophages were observed for all of these strains (Fig. 9a, p = 0.1190). These data suggest that the reduced pneumococcal virulence in the peritoneum in the absence of FlaR is not due to increased sensitivity to oxidative stress produced by the recruited immune cells. This is in accordance with data published recently, showing that oxidation is not the major killing mechanism of S. pneumoniae by phagocytic cells 46 .
The surface expression of FlaR, presented in the previous 35 and the current studies ( Fig. 1), encouraged testing the second mechanism of FlaR involvement in virulence i.e adhesion. As demonstrated in Fig. 9b, WU2ΔflaR Erm mutant adhesion to lung-derived epithelial (A549) cells was significantly reduced compared to the WU2 WT (~76%; Fig. 9, p = 0.0001), WU2ΔflaR flaR/Erm/Kan plasmid (~73%, p = 0.0001) and WU2ΔflaR flaR/Erm/Kan chromosome (~64%, p = 0.0001). Complementation in WU2ΔflaR flaR/Erm/Kan plasmid restored bacterial adhesion to A549 cells to about 88%, not significantly differing from the WU2 WT. The adhesion of WU2ΔflaR flaR/Erm/Kan chromosome strain was restored only up to 65%, remaining significantly lower compared to the WU2 WT (Fig. 9b, p = 0.0011). These data suggest that FlaR is involved in bacterial adhesion to the host cells in addition to its enzymatic function. Similarly, reduced adhesion to the glioblastoma cell line U251 47 and the motor neuron cell Vaccine potential of rFlaR. The finding that FlaR contributes to S. pneumoniae virulence, combined with the lack of significant similarity to any human protein, directed us to test the vaccine potential of rFlaR. Immunization of BALB/c mice with rFlaR elicited the production anti-rFlaR antibodies. An increase in antibody levels could be observed following the secondary immunization however this increase reached significance only after the tertiary immunization. The antisera titer following the third immunization reached ~1:10000 (Supplementary Fig S4, In addition, significant protection against IP lethal S. pneumoniae challenge was observed following immunization with rFlaR in comparison to adjuvant only immunization ( Fig. 10a 15 µg rFlaR + alum, p = 0.0388; Fig. 10b 25 µg rFlaR + alum, p = 0.0001; Fig. 10c 35 µg rFlaR + CFA/IFA, p = 0.0053). To further demonstrate rFlaR ability to elicit protective immune response we have performed a modified method of passive immunization. Bacteria were incubated ex-vivo with either pre-immune or anti-rFlaR antiserum. Mice were inoculated IP with bacteria pre-treated with the anti-rFlaR antiserum survived significantly longer than mice inoculated with bacteria pre-treated with pre-immune serum (Fig. 10d, p < 0.0001). The oxidative state of Fe 2+ was assessed with ferrozine staining and monitored at A 562 nm. The oxidative state of FAD was monitored at A 450 nm and that of NADP monitored at A 340 nm. Fe 2+ , FAD and NADPH in aqueous solutions were allowed to oxidize for 2 hrs, and then cysteine was added for additional 2 hrs. Oxidation was observed only in Fe 2+ solution which was significantly reduced upon cysteine addition. The results presented are the mean of three different experiments (one way ANOVA with Dunnett's post hoc multiple comparisons test, n = 3 in each group, n = 3 in each group, ****p = 0.0001). (c) Increasing concentrations of Fe 2+ significantly enhanced rFlaR NADP-reductase activity (Pearson correlation, r = 0.9947, ****p < 0.0001).

Discussion
A putative flavin reductase (FlaR) was recently annotated in S. pneumoniae TIGR4 strain in the NCBI database. We have previously shown that there is an age-dependent enhancement of the antibody response to this protein 35 . Therefore, we initially named it pneumococcal surface immunogenic protein B (PsipB), but in order to conform with the NCBI database we currently refer to this protein as FlaR. FlaR was found to be surface expressed, although it is devoid of a signal peptide or any other known export sequences 35 . Several additional S. pneumoniae proteins such as α-enolase 49,50 , glyceraldehyde 3-phosphate dehydrogenase, fructose-bisphosphate aldolase 35 , glutamyl tRNA synthetase 51 , NOX 52 and PtsA 39 were also shown to be cell surface localized albeit lack of known export signals 53,54 . The current study further confirmed the cell surface localization of FlaR, using immunoblotting and flow cytometry.
FlaR was found to be highly conserved in different S. pneumoniae strains. In addition, highly conserved cysteines at positions 66, 72 and 84 of WP_000580663.1 were found in all S. pneumoniae sequenced flavin reductases. Multiple cysteines in a protein may enable the formation of multimers. Indeed, resolving rFlaR protein on SDS-PAGE demonstrated the existence of both dimeric and monomeric forms of FlaR. Equilibrium between the dimeric and monomeric forms may be important for FlaR enzymatic activity, as hydrogenases are known to often function as homo-dimers 55 or multi-enzyme complexes 56 . Moreover, multiple cysteines in a protein sequence may indicate a transitional metal ion binding capacity 57 . Metal ions can interact with metal-binding proteins during their folding and create a local structure that initiates and directs the protein-folding process and is considered critical for its biological activity 58,59 . In accordance, rFlaR could be refolded/solubilized at physiological pH only  in the presence of Fe 2+ and to lesser extent with Co 2+ and Cu 2+ . Noteworthy is the finding that rFlaR, refolded at pH 8 without Fe 2+ , bound Fe 2+ as detected by the ferrozine stain.
We validated the putative flavin reductase activity of rFlaR and found that NADP and FAD are necessary components for its enzymatic activity, as described for NADPH reductases 60 . In contrast, the flavin reductase activity of ferredoxin-NADP + reductase from Pseudomonas putida was FAD-independent 42 . The activity of NADPH-cytochrome p 450 reductase is generally accompanied by measurable alterations in FAD and FMA oxidative state 60 . Although rFlaR catalyzed NADP reduction in a FAD-dependent manner, we failed to detect the change in FAD oxidative state.
Refolding of rFlaR in the presence of Fe 2+ enhanced its enzymatic activity in comparison to the rFlaR refolded at pH 8, possibly as a result of increased initial binding of Fe 2+ . The ability of divalent ion chelators to inhibit rFlaR activity reinforce Fe 2+ necessity for NADP reduction by FlaR.
The midpoint potential of Fe 2+ in solution, as a complex with EDTA, is +0.1 V while the midpoint potential of NADP is much lower (−0.32 V), rendering reduction of NADP by Fe 2+ improbable. However, the midpoint potential of Fe 2+ bound to 4Fe-4S cluster of ferredoxin from Butyribacterium methylotrophicum was found to be −0.410 V 61 and −0.472 V in 2Fe-2S cluster of oxidoreductase from Paracoccus denitrificans 62 . Additionally, it was shown that substitution of amino acid residues around an Fe 2+ -sulfur cluster, by site-directed mutagenesis, affects the midpoint potential of the Fe 2+ in the cluster 63 . The finding that under physiological conditions rFlaR remains soluble only in the presence of Fe 2+ , impedes determination of the midpoint potential of Fe 2+ bound to FlaR. Yet, the demonstration of Fe 2+ dependent rFlaR NADP reductase activity indicates that the FlaR-bound Fe 2+ midpoint potential is probably lower than that of NADP.
Mucosal surfaces lack free Fe 2+ due to chelation by host Fe 2+ -binding proteins 64 . S. pneumoniae has several mechanisms enabling it to cope with iron deficiency, among them hijacking iron from hemin, hemoglobin or lactoferrin 33,65 , or direct iron binding by ABC iron transporters Pit1B and Pit2A 31 . A double mutant of pit1B and pit2A demonstrated growth deficiency in a cation-depleted medium, which was restored by the addition of Fe 2+ 31 . In addition, the 12 mer Dpr/Dps that bind, oxidize and store ~500 Fe 2+ atoms in its hollow cavity has been identified in streptococci 27,66 . The current study revealed that FlaR belongs to the Fe 2+ binding proteins. rFlaR activity was found to be enhanced in the presence of increasing concentration of the reducing agent, cysteine. Our demonstration that cysteine reduced Fe 3+ but not FAD or NADP, implies that the increase in NADPH in the presence of cysteine stemmed from Fe 3+ to Fe 2+ recycling, which stimulated rFlaR activity. In line is the finding that rFlaR NADP reductase activity was enhanced in a Fe 2+ concentration-dependent manner.
Tight regulation of the free Fe 2+ pool can be especially important for pneumococci, which produce high levels of H 2 O 2 but lack catalase, and hence are at high risk of damage by ROS formed via the Fenton reaction 13 . We hypothesized that FlaR is involved in preventing the accumulation of toxic levels of free Fe 2+ that may react with The explanation may be that following flaR trans-complementation, multi-copies of flaR are expressed, whereas following cis-complementation (into the same chromosome) a single copy of flaR is expressed, under the control of its native promoter. This may stem from two concomitantly occurring mechanisms: i) FlaR's function in binding and oxidizing Fe 2+ to protect the bacterium from the toxic effects of the Fenton reaction; ii) FlaR's NADP reductase activity provides an additional source of NAD(P)H beside the Embden-Meyerhof-Parnas and the pentose phosphate pathways. Increased NAD(P)H levels are essential under high H 2 O 2 levels for NADH oxidase activity 67 . The action of NADH oxidase makes pyruvate available for SpxB oxidation, resulting in the production of two additional ATP molecules per one glucose molecule, beyond that available from conventional glycolysis 13,68 . Moreover, the electron transport to NADPH by FlaR enzymatic activity may increase the reduction of cysteine containing enzymes 69 to maintain redox balance thus enhancing bacterial growth and virulence 70 .
In contrast to our finding that the presence of FlaR in the cells confer resistance to oxidative stress, Pericone et al. 13 claimed that iron chelators had no effect on the survival of WT or a mutant lacking pyruvate oxidase (ΔspxB) following a challenge with 20 mM H 2 O 2 . This led the authors to suggest that the Fenton reaction does not play a major role in the killing of S. pneumoniae by H 2 O 2 . The mechanism by which FlaR contributes to bacterial survival under oxidative stress conditions should be further elucidated. Of note is the finding that Fe 2+ dependent-rFlaR activity was not affected by the H 2 O 2 scavenger, salicylate, suggesting that Fe 2+ oxidation is not directly executed by H 2 O 2 .
The contribution of FlaR to pneumococcal virulence was tested in a mouse model of sepsis. The reduced virulence of S. pneumoniae in the peritoneum, in the absence of FlaR, could be explained by increased sensitivity to phagocytosis and oxidative stress elicited by immune cells. The phagocytic role of macrophages in S. pneumoniae engulfment and killing has been previously demonstrated for bronchoalveolar lavage derived macrophages 71 , peritoneal macrophages 72 and neutrophils 46,73 . Interestingly, we found that FlaR absence did not significantly affect the extent of phagocytosis by primary mice peritoneal macrophages. In line, in vitro studies demonstrated that the predominant killing mechanism of pneumococci in the phagosome of neutrophils is executed by the proteolytic activity of two serine proteases rather than oxidation 46,73 . The involvement of the serine proteases in pneumococcal killing was further confirmed in vivo demonstrating higher bacterial load in the spleen of mice lacking the serine proteases in a systemic infection model 46 . These authors also emphasized that in the case of systemic infection with S. pneumoniae, most of the bacteria can be found in the spleen and to a lesser extent in other organs 46 .
The peritoneum is characterized by low oxygen pressure 74 . Thus, the role of FlaR in protection against oxidative stress could not explain the reduced virulence of WU2ΔflaR kan in this niche suggesting an additional role for FlaR in virulence. FlaR was found to be surface expressed using immunoblotting of cell-wall proteins and flow cytometry. Pre-incubation of S. pneumoniae with anti-FlaR antisera, which interacts only with the surface exposed FlaR, reduced bacterial virulence, reinforcing the surface localization of FlaR. Many surface proteins were found to mediate bacterial adhesion to the host 37 . Hence, we hypothesized that FlaR may be involved in bacterial adhesion to host cells. Indeed, WU2ΔflaR Erm strain was found to adhere to A549 cells significantly less than its parental WT and the two complemented strains. We have also used glioblastoma cells U251 47 and motor neuron cells NSC34 48 to establish FlaR contribution to S. pneumoniae adhesion and similar results were obtained. The ability of anti-rFlaR antibodies to reduce bacterial virulence also implies a direct involvement of the surface expressed FlaR in virulence and diminishes the possibility that downstream consequences of the mutation, such as upregulation of other adhesins, are responsible for the bacterial virulence.
The following findings highlight FlaR as a potential candidate vaccine: 1) possesses age dependent antigenicity; 2) is conserved among S. pneumoniae strains; 3) lacks significant homology to human proteins; 4) protects the bacterium against oxidative stress and 5) possibly mediates S. pneumoniae adhesion to the host. Indeed, FlaR elicited significant protection against IP challenge. In addition, sera obtained from immunized rabbits were able to neutralize S. pneumoniae virulence in the peritoneum, probably by enhancing opsonophagocytosis and/or interfering with bacterial adhesion and spread. Hence, rFlaR can be considered as a future vaccine candidate. It should be noted that proteins, such as PspA, CbpA, PavA, PavB and PhtD, which are involved in different aspects of bacterial virulence including adhesion, are known to be immunogenic and to elicit a protective immune response in experimental infection models [75][76][77][78][79] . Furthermore, PhtD has recently been shown to elicit an immune response in phase I/II clinical trials 80,81 .
In summary, this study attributes FlaR multiple functions in S. pneumoniae physiology and virulence: i) Fe 2+ binding and oxidizing to prevent the toxic effect of the Fenton reaction; ii) reducing NADP to provides an additional source of NAD(P)H, resulting in maintenance of redox balance and ATP production; and iii) mediating bacterial adhesion to the host. Importantly, we found that rFlaR can be considered for future vaccine development. Further studies should concentrate on detailed understanding of FlaR's role in pneumococcal biology and the nature of the immune protection elicited by FlaR. Ex-vivo neutralization was performed as follows: a lethal dose (~10 2 CFU) of WU2 WT strain was incubated at 37 °C for 1 hour with 1:10 diluted rabbit pre-immune serum or anti-rFlaR serum and subsequently inoculated IP to BALB/c mice (n = 10 in each group). Survival was monitored daily.

Inoculation and immunization of mice and ethics statements. This study was carried out in
For all animal experiments, mice were humanely euthanized by CO 2 asphyxiation, as recommended by the American Veterinary Medical Association (AVMA) guidelines for euthanasia 2013, if they became moribund or showed evidence of distress. The following criteria were considered sufficient evidence of distress to warrant such intervention in order to minimize pain and suffering to animals: severe weight loss (20% body weight); reluctance or inability to move freely; appearance of bristle fur; social disengagement; refusal or inability to eat or drink. No analgesic treatment was provided as such treatment may alter the immune response and may independently affect the outcome of the experiments 82 . Cloning, Expression and Purification of rFlaR, rFabD, rATCase and antibodies production. The flaR gene was amplified from S. pneumoniae strain WU2 genomic DNA by PCR using the primers FlaR-F and FlaR-R (Supplementary Table S1, flaR pHAT expression) designed according to the sequence in the R6 strain (spr0489). The amplified and BamHI-SacI (Takara Bio Inc, Shiga, Japan)-digested DNA-fragments were cloned into the pHAT expression vector (BD Biosciences Clontech, Palo Alto, CA) and transformed into DH5α E. coli cells. The vector was purified using the Qiagen High Speed Plasmid Maxi Kit (Qiagen GMBH, Hilden, Germany) and transformed into E. coli BL21. Sequencing was performed to rule out any mutation. rFlaR expression was induced by incubation for 5 h with 1 mM of IPTG at 37 °C. The protein was purified under denaturing conditions using Ni-NTA agarose beads (Qiagen GMBH). This preparation was dialyzed against PBS pH 7.3, containing 2 mM FeSO 4 for 48 h with three changes of buffer (named rFlaR Fe 2+ ). Isolation of the protein was confirmed by immunoblotting with anti-HAT (BD Biosciences Clontech) and thereafter with anti-rFlaR antibodies (described below) and trypsin in-gel digestion of the protein to derive peptides for MALDI-TOF mass spectrometry sequencing (Bruker Reflex-IV mass spectrometer Bruker-Daltonik, Bremen, Germany). Alignment to existing databases was performed using both the Mascot software package (Matrix Science Ltd., UK, http://www.matrixscience.com) and the "Profound" Program provided by the Rockefeller University.
Untagged protein was produced by amplifying the gene using the S. pneumoniae TIGR4 strain (ATCC, Rockville, MD, USA) DNA as template, using specific primers (Supplementary Table S1, flaR pET30a (+)). Digestion of pET30a(+) (Novagen, Madison, WI, USA) with NdeI and Bpu1102I restriction enzymes, prior to ligation removed all tags from the vector. Then, flaR was cloned into the vector. Protein expression was induced with IPTG in fermented E. coli BL21 followed by inclusion bodies isolation. The inclusion bodies were centrifuged and resuspended in double distilled water and frozen in aliquots. The inclusion bodies were refolded by dialysis for 1 hour against a buffer adjusted to pH 11.3, containing 4.5 M urea, 40 mM Tris base and 1 mM cysteine, followed by the addition of 0.67 mM arginine and continued dialysis overnight. The resulting solution was concentrated ~4.4 fold. This preparation was applied to a preparative Superdex 200 column and elution was performed with buffer adjusted to pH 10, containing 4.5 M urea, 40 mM Tris base and 1 mM cysteine. The tubes containing the protein were pooled and dialyzed against NaHCO 3 at pH 11 followed by lyophilization. The lyophilized protein was suspended in PBS pH 8 (named FlaR pH 8).
For antibodies production, three-month-old New Zealand white rabbits and 7-week old BALB/c mice were immunized with Ni-NTA-purified rFlaR formulated with CFA in the first immunization and IFA in the two booster immunization with two-week intervals between immunizations. The rabbits were bled following the third immunization.
FabD was cloned and expressed as a His-tagged soluble protein and antibodies were produced as previously described 36 .
ATCase is known to be involved in pyrimidine synthesis. ATCase is a cytoplasmic protein that was found to be surface exposed 37 . ATCase (SP_RS06260) was cloned and expressed as a His-tag soluble protein. Briefly, ATCase was amplified from the R6 strains using specific primers (Supplementary Table 1, ACTaase gene for pET32a+) and ligated into pET32a+ (Novagen, Madison, WI, USA) following digestion with EcoR1 and Xho1 restriction enzymes. The plasmid was transformed into DH5α E. coli cells. The vector was purified and transformed into E. coli BL21. Sequencing was performed to rule out any mutation. Protein expression was induced by 0.1 mM IPTG for 3 h. Following bacterial lysis the protein was purified using NI-NTA column under native condition.
Flow cytometry of S. pneumonia. Flow cytometry was performed as previously described 51 . Briefly, R6 bacteria were incubated with mouse anti-rFlaR antibodies or control mouse serum, washed, and stained with Alexa Fluor 647 ® -conjugated goat-anti-mouse-IgG (Jackson ImmunoResearch, West Grove, PA). Flow cytometry was performed using a FACSCalibur flow cytometer (Becton Dickinson, Mountain View, CA), and data were acquired and analyzed using BD CellQuest TM 3.3 software. Ability of Cysteine to Reduce Iron. Freshly prepared solution of 20 µM ammonium iron sulfate in PBS was incubated for 2 h at 25 °C to allow Fe 2+ oxidation prior to the addition of 5 mM cysteine. The extent of Fe 2+ oxidation in PBS solution was tested with ferrozine stain, and absorbance monitored at A 562 nm. The same manipulations were performed for 200 µM NADP and 50 µM FAD solutions in PBS, and absorbance was measured at A 340 nm and at A 450 nm, respectively. Bioinformatics Analyses. All completely sequenced genomes of S. pneumoniae were downloaded from NCBI's Genome database (29 genomes as of October 2 nd , 2016). The flavin reductase amino acid sequence of the TIGR4 strain (WP_000580663.1) was used as a query in tblastn and blastp searches against each of these genomes and their predicted proteins. WP_000580663.1 sequence was then used as a query in a blastp search against all RefSeq proteins annotated as S. pneumoniae and "flavin reductase". The 63 retrieved protein hits were subjected to multiple sequence alignment using Tcoffee with default parameters. A representative set of 9 divergent proteins was selected from the multiple sequence alignment by inspecting a Neighbor Joining tree and choosing a representative protein from each clade. Multiple sequence alignment of these proteins was performed using Tcoffee and visualized using the Jalview software. Finally, the entire RefSeq protein database except for S. pneumonia proteins was searched by blastp with WP_000580663.1 as a query. Only Genbank records containing the word "Flavin" and having blastp e value < 0.001 were retrieved.

Preparation of WU2ΔflaR Mutant Bacteria (kanamycin or erythromycin resistance).
To create null mutants of the flaR gene, the kanamycin (Kan) or erythromycin (Erm) resistance cassette was inserted into the coding sequence of flaR via homologous recombination. The transformation procedure was performed as previously described 85 with minor modifications. The upstream and downstream flanking regions of flaR were amplified by PCR from DNA of WU2 strain using the primer combinations upwing-F/upwing-R (Supplementary  Table S1, Upwing WU2ΔflaR), for the upstream region, and downwing-F/downwing-R, for the downstream region (Supplementary Table S1 Table S1, Kan AB cassette). The Erm cassette was amplified by PCR from the genome of WU2Δnox Erm52 using the primer combinations Erm AM-F/Erm AM-R (Supplementary Table S1, Erm AM cassette). The PCR products were digested with the corresponding restriction nucleases, as specified in Supplementary Table S1, purified, ligated and transformed into WU2 in the presence of competence stimulating factor, CSP1 and CaCl 2 . Transformants were selected on THY solidified with 1.5% agar, containing kanamycin (80 µg/ml) or erythromycin (125 µg/ml). Verification of flaR deletion was done by PCR using specific primers (Supplementary Table S1, Kan AB cassette for Kan resistance and Erm AM cassette for Erm resistance). It is worth mentioning that phenotypically, flaR mutation did not affect the size and mucosity of the colonies.
Of note, the kanamycin resistant mutant strain, WU2ΔflaRKan, was used in the mice inoculation studies. The erythromycin resistant mutant strain, WU2ΔflaRErm, was used in the susceptibility to oxidative stress experiment and the adhesion to A549 cells assay.
Cis-complementation of WU2ΔflaR Erm . To confirm that mutation of flaR in WU2ΔflaR Erm introduced no polar effects, WU2ΔflaR Erm was complemented with an intact copy of the gene using pCEP, which is a nonreplicative plasmid that allows controlled gene expression under its native promoter, following ectopic integration into the chromosome 87,88 . Briefly, flaR was amplified from WU2 with flaR-F and flaR-R primers (Supplementary Table S1, flaR pBAV-K1-plasmid and pCEP complementation), which introduce NcoI and PstI sites. Then the amplicons were ligated into NcoI and PstI digested vector. An aliquot of ligation mixture was transformed into Stellar ™ competent cells (Clonetech, Saint-Germain-en-Laye, France) as described by the manufacturer. The transformants were selected for kanamycin resistance (500 μg/ml). Successful ligation was determined by colony PCR using Mal-F and pCEP-R primers (Supplemental Table S1, flaR pCEP verification), whose recognition sites are localized immediately up and downstream of the cloning site, respectively. These primers amplified 263 bp product in empty vector, while they produce a product of 913 bp in the recombinant clones (additional 650 bp represents the cloned fragment containing flaR). The recombinant plasmid was purified using a commercial kit (Qiagen), and an aliquot was transformed into WU2ΔflaR Erm as described previously 89 to produce the complemented strain WU2ΔflaR flaR/Erm/Kan chromosome. The transformants were selected on blood agar plates supplemented with erythromycin (125 μg/ml) and kanamycin (500 μg/ml). Integration of flaR into the genome was confirmed by PCR (Supplementary Table S1, Kan resistance in pCEP verification).
Oxidative stress assay. Oxidative stress assays were performed as previously described 87 with minor modifications. Briefly, WU2 WT, WU2ΔflaR Erm and WU2ΔflaR flaR/Erm/Kan plasmid and chromosome strains were grown under anaerobic conditions in 20 ml of THY medium to an OD 620 = 0.3. Five ml of culture were harvested by centrifugation at 4000 rpm for 20 min and resuspended in 5 ml of PBS. Even volumes of the bacteria were dispensed, mixed with equal volumes of either H 2 O 2 (final concentration 20 mM) or PBS and incubated at 37 °C for 30 min. Following the incubation serial dilutions were plated onto blood agar plates for bacterial enumeration. The percent survival was calculated by dividing the CFU of cultures after exposure to H 2 O 2 by the CFU in the control culture without H 2 O 2 of the respective strain. The results presented are the average of three independent experiments performed in triplicates.
Primary macrophage phagocytosis assay. Mice were injected intraperitoneally with 3% thioglycolate broth. Three days later mice were euthanized, using CO 2 , and their peritoneal cells were harvested by flushing the peritoneum with 10 ml of PBS. Cells were centrifuged and resuspended in complete (c) RPMI 1640 medium supplemented with 10% FCS and 2% L-glutamine. 2 × 10 5 cells/well were plated onto 96 well culture plates. The macrophages were allowed to adhere for 9 hours and the supernatant was removed. 2.5 × 10 4 of midlog harvested bacteria in RPMI 1640 were added to the well for 40 minutes incubation at 37 °C. Following the incubation, aliquots were taken from the supernatant and plated onto blood agar plates for enumeration. The number of residual live bacteria CFU was subtracted from the initial bacterial CFU resulting in the number of phagocytosed bacteria. The data presented is the summary of 2 biological independent experiments each performed in triplicates.
S. pneumoniae adhesion to A549 cells assay. A549 cells (lung adenocarcinoma cells; ATCC, Rockville, MD, USA) retain morphological, biochemical and immunological characteristics resembling type II lung epithelial cells [90][91][92] and have been widely used as a model to study S. pneumoniae interaction with human cells 37,51 . A549 cells were cultured on 96-well plates (2.5 × 10 5 cells/well) in DMEM without antibiotics. Following overnight incubation each well contained ~10 5 A549 cells. WU2 WT, WU2ΔflaR Erm and WU2ΔflaR flaR/Erm/Kan strains (~10 6 CFU; multiplicity of infection (MOI) 10:1) were added to the cells for 1 hr incubation at 37 °C. Bacteria were enumerated after the 1 h incubation and no bacterial death under the experimental conditions could be observed in WT, mutant and complemented strains (data not shown). Excess bacteria were removed, and cells detached with 0.25% trypsin-EDTA and plated onto blood agar plates for enumeration. This experiments were performed in triplicate and repeated three times. rFlaR elicits antibody response in immunized mice. Mice were immunized with 25 µg rFlaR refolded in the presence of 2 mM Fe 2+ using CFA in primary immunization and IFA in the following booster immunizations. Microtiter plates (F96 Maxisorp Nunc (ThermoFisher) were coated with 1 µg/ml solution of rFlaR in bicarbonate buffer pH 9.6. Plates were blocked with 2% BSA solution in PBS. Sera obtained from mice immunized with rFlaR following each immunization were diluted with PBS and used as the primary antibodies. Washes were performed with PBS supplemented with 0.05% tween. Secondary antibody used was HRP conjugated anti-mouse IgG (Jackson laboratories). Detection was performed using tetramethylbenzidine (TMB; DakoCytomation) as substrate. Reaction was stopped using 1 M H 2 SO 4 . Absorbance was determined at 450 nm using ELISA reader multimode detector DTX 880 (BECKMAN COULTER). Serum obtained from mice immunized with rPtsA 37 (PtsA IV) was used as a negative control.
Statistical Analysis. The Shapiro-Wilk test was used and post-hoc statistical power was calculated, respectively, to verify that small sample sized data sets assumed normal distribution and are of the sufficient size to justify the use of one-way ANOVA for parametric data, followed by the Dunnett's test for multiple comparisons. One-tailed Student's t-test with Welch's correction was used for bacterial load comparisons of two groups. Data was reported as the mean ± SEM, unless stated otherwise. Pearson and Spearman correlations were used to assess the significance of change. Survival of S. pneumoniae-inoculated mice was determined using Log-rank (Mantel-Cox) test. Differences were considered significant at p < 0.05. All statistical analyses were performed with the software package in GraphPad Prism version 7 (La Jolla, CA, USA).