Peroxidase Activity and Involvement in the Oxidative Stress Response of Roseobacter denitrificans Truncated Hemoglobin

Roseobacter denitrificans is a member of the widespread marine Roseobacter genus. We report the first characterization of a truncated hemoglobin from R. denitrificans (Rd. trHb) that was purified in the heme-bound form from heterologous expression of the protein in Escherichia coli. Rd. trHb exhibits predominantly alpha-helical secondary structure and absorbs light at 412, 538 and 572 nm. The phylogenetic classification suggests that Rd. trHb falls into group II trHbs, whereas sequence alignments indicate that it shares certain important heme pocket residues with group I trHbs in addition to those of group II trHbs. The resonance Raman spectra indicate that the isolated Rd. trHb contains a ferric heme that is mostly 6-coordinate low-spin and that the heme of the ferrous form displays a mixture of 5- and 6-coordinate states. Two Fe-His stretching modes were detected, notably one at 248 cm-1, which has been reported in peroxidases and some flavohemoglobins that contain an Fe-His-Asp (or Glu) catalytic triad, but was never reported before in a trHb. We show that Rd. trHb exhibits a significant peroxidase activity with a (k cat/K m) value three orders of magnitude higher than that of bovine Hb and only one order lower than that of horseradish peroxidase. This enzymatic activity is pH-dependent with a pK a value ~6.8. Homology modeling suggests that residues known to be important for interactions with heme-bound ligands in group II trHbs from Mycobacterium tuberculosis and Bacillus subtilis are pointing toward to heme in Rd. trHb. Genomic organization and gene expression profiles imply possible functions for detoxification of reactive oxygen and nitrogen species in vivo. Altogether, Rd. trHb exhibits some distinctive features and appears equipped to help the bacterium to cope with reactive oxygen/nitrogen species and/or to operate redox biochemistry.


INTRODUCTION
Marine Roseobacters [1,2,3] are known to be abundant in the ocean ecosystem near the surface [4], and recent studies indicate that they are also responsible for producing reactive oxygen
Bacterial strains and culture conditions R. denitrificans OCh 114 cells were grown in Difco TM marina broth 2216 (Becton, Dickson and Company) [37] at 28°C with constant shaking at 180 rpm in 300 mL culture flasks with 100 mL of working volume. The cultures were grown to the stationary phase and harvested for extracting the genomic DNA using the protocol suggested in the QIAamp DNA Stool Mini Kit (Qiagen) with minor modifications.
Cloning, expression and purification of recombinant Rd. trHb The R. denitrificans glbO gene encoded a polypeptide of 152 amino acids, with a calculated molecular mass of 17.5 kDa. Polymerase chain reaction (PCR) was used to amplify the coding region of the glbO gene from the R. denitrificans genomic DNA with the method described previously [38]. DNA primers for gene amplification were: (forward primer, the NcoI site underlined) 5 0 -acctCCATGGcatgtgctttgcaaaagatgggcggc-3 0 and (reverse primer, the XhoI site underlined) 5 0 -ACCTCTCGAGTCAGGCCCGCCGCAGCGCACCGTGA-TCC-3 0 . The NcoI/ XhoI-digested fragment was inserted into NcoI/XhoI-cut pET-28a vector (Novagen/Merck) producing constructs for expression of Rd. trHb without N-or C-terminal His-tag. The ligation product was transformed to DH5α component cells (New England Biolabs) following the manufacturer's protocol. Plasmid DNAs from a number of clones were isolated and screened for the presence of the insert by restriction digestion and DNA sequencing. The construct with the correct sequence was transferred to BL21(DE3) cells (New England Biolabs) for protein expression.
E. coli BL21(DE3) cells containing the constructed plasmid were grown at 37°C in 1-L flasks containing 500 mL Luria-Bertani (LB) medium and 50 μg/mL kanamycin until the absorption at 600 nm reached 0.6-0.8, at which time, the temperature was reduced to 16°C and the cells were incubated overnight to maximize protein expression. IPTG was not added to induce protein synthesis because it was found to inhibit cell growth. The cells were harvested by centrifugation at 5000 x g for 20 min at 4°C, resuspended in a lysis buffer containing 1 mM EDTA, 2 mM NaCl, 1 mM β-mercaptoethanol and 0.1% Triton X-100 in 20 mM Tris-HCl at pH 7.8. The cells were sonicated for 30 min on ice and then centrifuged at 5000 x g for 20 min. The soluble proteins were collected for further purification on a 0.7 x 15 cm FLEX column packed with Q-Sepharose Fast-Flow (GE Healthcare Life Sci.) (130 mL) and equilibrated with buffer A (20 mM Tris-HCl pH 8.0) containing 50 μM EDTA at 4°C. After washing, the proteins were eluted with buffer A containing 150 mM NaCl. The proteins were concentrated by ultrafiltration (10 kDa MWCO) and then dialyzed against buffer A containing 50 mM NaCl overnight at 4°C in 8 kDa MWCO membranes (Spectra/Por). The protein mixture was then loaded on a Hi-load 16/60 Superdex 75 gel-filtration column (GE Healthcare Life Sci.) equilibrated with 50 mM NaCl in buffer A at 4°C. The purified Rd. trHb was estimated to be a monomer based on the elution volume from the gel-filtration column. A total of 0.2-0.25 mg protein was isolated from 1-liter growth medium. The protein sequence of the heterologously expressed Rd. trHb was confirmed by peptide-mapping via mass spectrometry [39]. The protein concentration was determined by the Bradford assay and by the A 280 using an extinction coefficient 15,470 M -1 cm -1 calculated from the deduced amino acid sequence of Rd. trHb. To obtain the ferrous deoxygenated state of Rd. trHb, the purified ferric protein was made anaerobic by flushing with N 2 and then reduced with 1.0% sodium dithionite.

Optical spectral measurements
The UV-visible absorption spectra were recorded using a Shimadzu UV-1800 spectrophotometer. The CD spectra for Rd. trHb in a 1.5 mm path length quartz cuvette were recorded between 195 and 265 nm at 25°C using a Jasco J-810 CD spectrometer. All spectra were collected at 25°C.

Peroxidase activity assays
The peroxidase activity was assayed using H 2 O 2 as the oxidizing substrate and ABTS as the organic substrate. The reaction was first verified with HRP and monitored by the increase of absorbance at 414 nm (A 414 ). The Rd. trHb-dependent H 2 O 2 -reduction was performed with 0.5 mM H 2 O 2 , 0.5 mM ABTS and various concentrations of Rd. trHb in 20 mM Tris-HCl buffer at pH 8.0. The concentration of the H 2 O 2 stock solution was determined from the absorbance at 240 nm using an extinction coefficient = 43.5 M -1 cm -1 . The initial rates were determined in the initial linear phase of the progression curves. The concentration of oxidized ABTS was calculated from the increase at A 414 using a molar extinction coefficient of 36,800 M -1 cm -1 [40]. The steady-state kinetic parameters of HRP, Rd. trHb and bovine Hb were investigated using 13.5 nM HRP, 50 nM Rd. trHb and 77.5 nM bovine Hb and with various concentrations of H 2 O 2 and 1 mM ABTS. The initial rates versus H 2 O 2 concentration curves (6 to 7 H 2 O 2 /ABTS concentrations) were fitted to the Michaelis-Menten equation using the Kaleida-Graph software to obtain apparent k cat (k cat,app ) and apparent K m (K m,app ) values for the different enzymes and for reactions followed at different pHs. The pH dependence of the enzymatic activity was measured using the following buffers: MES (pH 6.0), phosphate (pH 6.1-6.8), MOPS (pH 7.0), Tris-HCl (pH 8.0), glycine (pH 9.0) and NaHCO 3 (pH 10.5). The pH dependences of k cat,app and K m,app were fitted to eq. 1: The pH dependence of k cat,app /K m,app was fitted to eq. 2: nance Raman spectra has been previously described [41]. The ferric, reduced and Fe(II)CO states were investigated with the 413.1 nm line from a krypton ion laser. The reduced state was also investigated with the 441.6 nm line from a He/Cd laser. The laser power on the samples was kept at less than 2 mW. The spectrometer was calibrated with the lines of indene in the low-and high-frequency regions. Cosmic ray lines were removed from the spectra by a routine of the Winspec software used for data acquisition (Roper Scientific, Princeton, NJ). Several spectra were acquired over a period of 30 minutes, averaged and analyzed using the Grams/AI software (ThermoGalactic).

Molecular modeling
The primary sequence of Rd. trHb (152 residues) was obtained from the UniProt Consortium (UniProt Consortium, 2012) (UniProt Q160B8). Three structures were identified as templates for homology modeling from a standard protein Blast (blastp) query using the Protein Data Bank (PDB) database on the NCBI/Blast web server (http://blast.ncbi.nlm.nih.gov/Blast.cgi). The three truncated hemoglobin structures identified are from Bacillus subtilis [42], Arabidopsis thaliana [43] and Agrobacterium tumefaciens [44], with UniProt identification numbers O31607, Q67XG0 and Q7CX73, respectively. A multiple sequence alignment of these sequences using the default parameters of Muscle v3.8.31 [45,46]  , using the CHARMM27 force field and the born implicit solvation model. Twenty-five backbone variations, each with twenty-five side chain positions, were explored. The best model was selected based on the GB/SI score in MOE and was validated with the MolProbity web server [48]. The heme cofactor from the B. subtilis structure was then added to the model, protonated and minimized with MOE.

Cultures prepared for gene expression assays
For investigating the gene expression profiles, R. denitrificans were grown in minimal medium supplied with 20 mM glucose as the sole carbon source and subjected to light/dark cycles (12 h:12 h) or with 20 mM acetate as the sole carbon source in darkness [37]. Reagents in 1-L minimal medium (pH 7.5) were: 3.2% (w/v) Instant Ocean sea salt, 0.  [7,37]. Phototrophic cultures were illuminated with low-intensity light (60 μmole/m 2 /s). Cultures in the mid-log growth phase were chosen for gene expression analyses.

RNA extraction and quantitative real-time polymerase chain reaction (QRT-PCR)
RNA was isolated from cell pellets using the TRIzol reagent (Invitrogen) according to the manufacturer's protocol, and DNase was used to remove DNA. The absence of contamination of the RNA samples by genomic DNA was verified by PCR and agarose gel electrophoresis. QRT-PCR was carried out to profile the gene expression levels under different growth conditions. cDNA was synthesized from 1 μg of RNA and 100 μM random 9-mer DNA using the Superscript III reverse transcriptase (Invitrogen). QRT-PCRs were performed with the Mx-3000P qPCR systems (Agilent Technologies, Inc.). The primers for QRT-PCRs (shown in Table 1) were designed with the program Primer3 (http://www.ncbi.nlm.nih.gov/tools/primer-blast/) and analyzed with Oligo-Analyzer 3.0 (Integrated DNA Technologies). The Power SYBR green master mix (Qiagen) was used to amplify DNA with the following cycles: an initial denaturation step (15 min at 95°C), followed by 40 amplification cycles (15 s at 95°C, 30 s at 60°C and 45 s at 72°C) and then 1 dissociation cycle (15 s at 95°C, 1 min at 60°C and 15 s at 95°C). The threshold cycle (C T ) was calculated as the cycle number at which ΔRn crossed the baseline. 16S rRNA was used as the internal control gene transcript. The following parameters were calculated: ΔC T that corresponds to C T (target gene)-C T (16S rRNA), ΔΔC T that corresponds to absolute value of ΔC T in two different culture conditions and the relative expression level that corresponds to 2 (absolute value of ΔΔCT) [49]. Three biological replicates, with three technical replicates for each biological sample, were analyzed and the mean value was reported. The amplified DNA fragments were verified by agarose gel electrophoresis.

Phylogenetic tree
A phylogenetic tree was constructed based on the primary sequences of 46 trHbs with the Neighbor-Joining method [50] using the phylogenetic software MEGA6 [51]. The percentage of replicate trees for which the associated taxa clustered together in the bootstrap test (1000 replicates) are shown next to the branches. The tree is drawn to scale, with branch lengths in the same units as those of the evolutionary distances used to infer the phylogenetic tree. All positions containing gaps and missing data were eliminated. The final data set contained 88 positions.

RESULTS AND DISCUSSION
Optical spectra of the recombinant Rd. trHb The glbO gene encoding trHb from R. denitrificans (Rd. trHb) was sub-cloned and the recombinant protein was heterologously expressed in E. coli and purified using the protocol reported for a group II trHb from Mycobacterium tuberculosis (Mt. trHbO) [52] with minor modifications. The recombinant Rd. trHb appears to be red (not shown), implying that the purified protein contained heme. Fig. 2 shows the UV-visible absorption and circular dichroism (CD) spectra of recombinant Rd. trHb and of bovine Hb (i.e. a "full-length" Hb). The UV-visible absorption spectrum of Rd. trHb ( Fig. 2A) is typical for a heme-bound protein with a characteristic Soret band (412 nm). The heme appears to be six-coordinate and lowspin heme with Q bands at 538 and 572 nm. We have also isolated a trHb from R. denitrificans cultures. This native protein displayed the same optical absorption spectrum (data not shown). In contrast, ferric bovine Hb exhibits a mixture of high-spin and low-spin states with a Soret band at 406 nm and Q bands at 497, 537, 573 and 630 nm. The absorption spectrum of Rd. trHb in the deoxygenated ferrous state shows the spectral features of a hexacoordinate heme, with two Q bands at 532 and 560 nm [53,54] (Fig. 2B). The CD spectra ( Fig. 2C) show that Rd. trHb, like bovine Hb, largely contains alpha-helical secondary structure elements, in agreement with the 2-on-2 alpha-helical sandwich structure reported for trHbs [29,30,33].

Sequence alignment and phylogenetic analysis
The multiple protein sequence alignment (Fig. 3) shows that Rd. trHb has the conversed His at F8 (His being the proximal ligand to the heme in globins) and Trp at G8 (important for O 2 binding and stabilization in Mt. trHbO (vide infra) [55], lacks Tyr at CD1 (with Phe instead) but has Gln at E11 and Tyr at B10 that are the important amino acids of group I Mt. trHb (Mt. trHbN) [30]. The percentages of identity and similarity were 16% and 27% between Rd. trHb and Mt. trHbN, respectively. Although the phylogenic analyses suggest that Rd. trHb falls in group II trHbs, which include Mt. trHbO, it has an interesting mix of potential heme pocket distal amino acids distinct from those of Mt. trHbO but similar to those of the group II trHb of Bacillus subtilis (Bs. trHbO) [56]. The phylogenetic tree built from 46 well characterized and putative trHbs from different organisms for illustrating evolutionary relationships among trHbs is shown in Fig. 4. TrHbs from cyanobacteria and green algae are identified as group I trHbs. TrHbs from R. denitrificans and several members of the Roseobacters genus are identified as group II trHbs and thus fall into the same group of Mt. trHbO [57] and Bs. trHbO [42] along with trHbs of higher plants.

Molecular modeling
The homology model of Rd. trHb was built using the Bs. trHbO crystal structure, except for a 4-residue insertion in the D-helix that was modeled based on the A. thaliana trHb and A. tumefaciens trHb crystal structures (see Materials and Methods for details). The distal site of the homology model of Rd. trHb is presented in Fig. 5A and the sequence alignment used for generating the model is shown in Fig. 5B. Fig. 5A shows that the highly conserved residues Tyr22(B10), His38(CD region), Phe39(CD1), Gln51(E11) and Trp91(G8) are found in the distal site pointing toward to the heme, and His78(F8) is the proximal His coordinating to the Fe. In Mt. trHbN, Mt. trHbO and Bs. trHbO, the residues involved in polar interactions with heme-bound ligands were described as a ligand-inclusive network of hydrogen bond interactions [42,56,58]. The amino acids involved in ligand stabilization are Tyr(CD1) and Trp(G8) for Mt. trHbO [58] and Tyr(B10) as well as Gln(E11) for Mt. trHbN. The homology model indicates that as for Bs. trHbO, Rd. trHb possesses a mixture of the conserved residues of group I and group II trHb of M. tuberculosis that are involved in the stabilization of the ligands. The only significant difference being His38, next to position CD1, that is present in trHbs from Roseobacters but absent in other trHbs. His38 in the CD region has not yet been reported and its role in ligand stabilization remains to be determined.

Resonance Raman spectra
To gain information about the active site of Rd. trHb, resonance Raman spectra were obtained. The high-frequency region of the resonance Raman spectra of heme proteins comprises several in-plane vibrational modes of the porphyrin that are sensitive to the oxidation, coordination and spin-state of the heme-iron [59,60]. In the low-frequency region, in-plane and out-ofplane vibrational modes of the porphyrin macrocycle contribute to the spectra, in addition to the modes associated with the axial ligands of the heme-iron and the heme substituents (i.e. vinyl and propionate groups) [61]. Fig. 6 shows the high-frequency region of resonance Raman spectrum of Rd. trHb. The ν 4 line at 1373 cm -1 is consistent with the heme being in the oxidized state. Two sets of ν 3 and ν 2 lines indicated the presence of a 6-coordinate high-spin heme (ν 3 at 1481 cm -1 and ν 2 at 1562 cm -1 ) and a 6-coordinate low-spin heme (ν 3 at 1506 cm -1 , ν 2 at 1584 cm -1 and ν 10 at~1643 cm -1 ) ( Table 2). The C = C stretching mode of the vinyl groups was detected at~1631 cm -1 . The ν 3 and ν 2 frequencies of both the high-spin and low-spin states are similar to those of ferric Mt. trHbO [52]. In the latter, the 6-coordinate high-spin signal comes from water and the 6-coordinate lowspin signal from a hydroxide ion bound to the heme. However, in contrast to Mt. trHbO, which shows a clear transition of low-spin to high-spin (or high-spin to low-spin) as the pH decreases (or increases) [55], only small spectral changes were detected in our pH titration (S2 Fig.). The Soret band was red-shifted from 412 nm to 416 nm as the pH was increased from 6 to 12 whereas the changes in the Q bands were rather small. Rd. trHb stayed mostly low-spin even at pH 6, which is clearly different from Mt. trHbO. It seems highly unlikely that a hydroxide ion would stay bound heme at pH 6. Rather, an amino acid is most likely the axial ligand at pH 6-7. As the pH increased, this amino acid may remain ligated or a transition to the hydroxide ion bound complex may take place. In either case, the optical spectra may only display small or no  differences. Based on a structural model of Rd. trHb, potential amino acids that could coordinate the ferric heme-iron are Tyr22(B10) and Gln51(E11). Tyr(B10) was identified as the sixth ligand to the heme of ferric trHb of Chlamydomonas eugametos, a group I trHb [62]. To our knowledge, there is no precedent for Gln coordination to heme in trHbs.
We also investigated the high-frequency and low-frequency regions of Raman spectra of the ferrous state. Fig. 6 shows that the heme of the ferrous form contains some 5-coordinate highspin heme with a strong ν 3 line at 1471 cm -1 , along with some 6-coordinate low-spin heme with ν 3 at 1494 cm -1 . The low-spin signal may come from an amino acid coordinating the reduced heme, presumably the same amino acid that coordinates the heme in the ferric state. It is interesting to note that for most trHbs characterized so far, the heme was found to be in the 5coordinate state [63]. Rd. trHb with a mixture of 5-and 6-coordinate states appears more similar to the trHbs of photosynthetic organisms; namely Synechocystis trHb, which contains a 6-coordinate heme having a histidine as the sixth axial ligand [25], and Chlamydomonas trHb, which contains mostly a 5-coordinate heme at pH 7.5 that becomes 6-coordinate at higher pH [64].
The low-frequency region was investigated using an excitation wavelength of 442 nm to enhance the signal from the 5-coordinate state and to identify the Fe-His stretching mode. In contrast to most reported Hbs, which contain one Fe-His stretching mode (ν Fe-His ) in the 200-220 cm -1 region, Rd. trHb exhibits two Fe-His stretching modes at 228 and 248 cm -1 , respectively (Fig. 7). We considered the possibility that the line at 248 cm -1 originated from the ν 9  mode of the porphyrin, but this mode would be expected to be present in the ferric (Fig. 7) and Fe(II)-CO ( Fig. 8) spectra, and these are featureless in this region. We thus conclude that the 248 cm -1 line more likely originates from a Fe-His mode that is known to be specifically enhanced from the 5-coordinate state of heme proteins and is probed here with an excitation wavelength at 442 nm. The Fe-His stretching mode at 248 cm -1 appears as a shoulder to the 228 cm -1 line but the fit of the spectrum in the 200-280 cm -1 region revealed that the area under the curve of the 248 cm -1 line was similar to that of the 228 cm -1 line, indicating that both Fe-His stretching modes are almost equally populated (Inset in Fig. 7). The frequency of the Fe-His stretching mode at 228 cm -1 is similar to that of Mt. trHbO (226 cm -1 ) and is accordingly assigned to a neutral histidine forming a favorable interaction with the heme-iron [52,63]. In contrast, the much higher frequency of the stretching mode at 248 cm -1 is similar to that of peroxidases where the proximal histidine is partially deprotonated by a strong hydrogen bond with a nearby aspartate/glutamate residue (i.e. an Fe-His-Asp/Glu triad) (  peroxidases [69], which is highly relevant to the peroxidase activity we report for Rd. trHb (vide infra) and/or NO detoxification activities discussed further below. From the sequence alignment (Fig. 3), we note that Rd. trHb has an Asp (Asp129) at the position H23. However, the homology model of Rd. trHb (Fig. 5A) cannot be used to evaluate if this Asp may be in a position to form a strong hydrogen bond with the proximal His at F8 since it does not extend further than the position H19. Nevertheless, this Asp is notably conserved in trHbs of  Roseobacters (Fig. 3), suggesting a functional role in these organisms. We also note that the Cterminus of Rd. trHb is more than 20 amino acids longer than trHbs of known structures and that this sequence is rich in Asp/Glu residues. This region may be another source for the amino acid involved in a Fe-His78-Asp/Glu catalytic triad. Altogether, the resonance Raman spectrum in the low-frequency region confirms iron coordination to a histidine in Rd. trHb and suggests a possible catalytic triad. The simple scenario is that the proximal histidine is His78 at F8, i.e. the same histidine as Mt. trHbO and Bs. trHbO, as suggested by the structural model (Fig. 5A). The low-frequency regions also displays lines at 350 and 676 cm -1 that are assigned to the ν 7 and ν 8 modes of the porphyrin ring [70], a line at 375 cm -1 that is assigned to a bending mode of a heme propionate, δ(C β C c C d ), and a line at 409 cm -1 that is assigned to a bending mode of a vinyl group, δ(C β C a C b ) [71] ( Table 2 and Fig. 7). The protein environment surrounding the propionate and vinyl groups did not change significantly upon heme reduction as the frequencies of the δ(C β C c C d ) and δ(C β C a C b ) modes remained nearly the same in the reduced state compared to the ferric state. Fig. 6 shows that the Fe(II)CO complex of Rd. trHb is 6-coordinate and low-spin with ν 3 at 1496 cm -1 and ν 2 at 1582 cm -1 . The relatively high-frequency of the oxidation state marker ν 4 at 1373 cm -1 is typical of Fe(II)CO complexes of heme proteins and is indicative of πback-bonding from the reduced iron to CO. Even at relatively low power (less than 2 mW), a small amount of photo-dissociation of CO occurred as revealed by the detection of the ν 4 (~1356 cm -1 ) and ν 3 (1471 cm -1 ) lines of a 5-coordinate reduced state (Fe II ). These band increased in intensity at higher laser power (13 mW) and decreased at lower intensity (0.5 mW) indicating that this process was reversible (data not shown). The frequency of the ν 4 line of the photo-dissociated reduced state (1356 cm -1 ) differed somewhat from that of the equilibrium reduced state (1362 cm -1 ) suggesting that only the 5-coordinate state may be present if the sixth ligand to the 6-coordinate state observed in the equilibrium sample had not formed yet.
Isotope substitution was used to identify the modes of heme-bound CO in the low-frequency region of the resonance Raman spectra of the Fe(II)CO complex of Rd. trHb (Fig. 8). The Fe-CO stretching mode (ν Fe-CO ) was identified at 509 cm -1 (496 cm -1 with 13 C 18 O). The frequency of the ν Fe-CO mode remained the same (509 cm -1 ) for spectra recorded at 0.5 mW (not shown) and 2 mW laser power (Fig. 8). The frequency of the Fe-CO stretching mode of Rd. trHb is lower than that of wild-type Mt. trHbO (525 cm -1 ) [52], where CO interacts with the side-chain of two amino acids (Trp(G8) and Tyr(CD1)), and is more similar to its Trp(G8)Phe mutant (with ν Fe-CO at 514 cm -1 ) and Tyr(CD1)Phe mutant (with ν Fe-CO at 515 cm -1 ), where CO is interacting with a single amino acid residue [52,55] (Table 3). The Fe-CO stretching frequency of Rd. trHb is also notably much lower than those of the two conformers observed for Bs. trHbO (520 and 545 cm -1 ), although both proteins share common polar groups in the heme pocket (Tyr(B10), Thr(E7), Gln(E11) and Trp(G8)) [42,56] (Table 3). Trp(G8) was shown to be the a key player for ligand stabilization in trHbO of B. subtilis along with Tyr(B10) [56]. When CO is not experiencing any interaction with a nearby group, the ν Fe -CO frequency is found near 495 cm -1 , such as with the Trp(G8)Phe/Tyr(CD1)Phe double mutant of Mt. trHbO (497 cm -1 ) [55] and one conformer of the Trp(G8)Leu mutant of Bs. trHbO (489 cm -1 ) ( Table 3) [56]. So the relatively high-frequency of ν Fe-CO of Rd. trHb indicates that an amino acid residue interacts with the heme-bound CO but that this interaction is weaker than in Mt. trHbO and Bs. trHbO. Contrary to the latter, only one Fe-CO stretching mode was observed for Rd. trHb providing no evidence that more than one conformer is present. The multiple sequence alignment and the structural model (Fig. 5) indicate that a number of polar residues are present in the heme pocket, such as Trp91 at G8, Tyr22 at B10, Gln51 at E11 and His38 next to CD1, which is conserved in putative trHbs of Roseobacters (Fig. 3). These will be examined for their roles in ligand binding in our future studies. The C-O stretching mode could not be detected because the sample was highly fluorescent in the 1500-2200 cm -1 region (not shown).
Gene expression of Rd. trHb, peroxidase reaction and implication of potential function(s) The genomic sequence of Roseobacters revealed that the glbO gene (encoding trHb) is either adjacent or a part of a gene cluster encoding proteins responsible for nitrogen/ammonium transfer and transport, such as glutamate synthase, glutamine synthetase, (glutamine) amidotransferases, and ammonium/amino acid transporters and permeases (Fig. 9). Glutamate is a key metabolite connecting the central carbon and nitrogen metabolism and thus is involved in reducing or inducing oxidative stress [72,73]. Acquisition of glutamate via the glutamate transporter was suggested to be important for Francisella tularensis for protection against oxidative stress [74]. Further, glutamine synthetase is also known to be sensitive to oxidative stress [75]. Thus the genomic environment suggests that Rd. trHb may be involved in nitrogen metabolism and oxidative stress response in Roseobacters. In addition to binding molecular oxygen, CO and nitric oxide (NO), trHbs are known to undergo allosteric regulation like Hbs in Metazoan [76] and also function as biological catalysts (enzymes) [17]. In terms of potential function(s) of Rd. trHb, one of the Campylobacter jejuni trHbs was suggested to function as a peroxidase [77] and Mt. trHbO, as well as its homologous trHb protein from Thermobifida fusca (Tf. trHbO), have been shown to function as peroxidases that reduce hydrogen peroxide while oxidizing an organic substrate [78,79]. Tyr at position CD1 of Mt. trHbO is involved in radical propagation in the protein following exposure to hydrogen peroxide and two surface Tyr residues (Y55 and Y115) have been shown to form radicals, which led to the formation of protein dimers [78]. In contrast, Rd. trHb has Phe at CD1, like in all full-length Hbs (i.e. those with a 3-on-3 alpha-helical fold) and many trHbs, and the two surface Tyr residues of Mt. HbO are not conserved in Rd. trHb, which has Phe and Thr at those sites. Thus any peroxidase activity by Rd. trHb would neither propagate a radical through Tyr at CD1 nor involve the surface Tyr residues identified in Mt. trHb. However, Rd. trHb could still oxidize substrates by radical propagation through other residue(s) or from the compound I intermediate (ferryl species with an associated radical) as in most peroxidases.
To examine the possible catalytic activity of Rd. trHb, we assayed the peroxidase activity using H 2 O 2 and ABTS as substrates. The reaction was monitored by the increase of absorbance at 414 nm (A 414 ) caused by the oxidation of ABTS. No increase of A 414 was detected without the addition of Rd. trHb. Fig. 10A shows the increase of absorbance at 414 nm with different concentrations of Rd. trHb (0 to 250 nM), and Fig. 10B illustrates the linear relationship of the initial velocities (μM ABTS/s) versus the concentration of Rd. trHb (0 to 250 nM). The straight line indicates that with ABTS and H 2 O 2 both at 0.5 mM concentration, Rd. trHb did not show apparent enzymatic inactivation.
To investigate the steady-state kinetics of the peroxidase activity catalyzed by Rd. trHb, we estimated the initial velocities of ABTS oxidation by Rd. trHb at various concentrations of [H 2 O 2 ] and compared the results with those of the well characterized HRP (Fig. 10C). When 50 nM Rd. trHb or 13.5 nM HRP were employed, Rd. trHb displayed a similar K m,app (apparent K m ) and approximately one order of magnitude lower k cat,app (apparent k cat ) and k cat,app /K m,app values compared to HRP ( Table 4). The peroxidase activity of Rd. trHb (50 nM) was also much higher than that of bovine Hb (77.5 nM) (Fig. 10D and Table 4), which is a heme protein with a very low but measurable peroxidase activity [80]. Unlike Rd. trHb, the trHbO from M. tuberculosis and T. fusca did not show saturation kinetics with respect to the concentration of H 2 O 2 when ABTS was the organic substrate [78,79]. To compare Rd. trHb with these trHbOs, we used the slope of the initial rates versus concentration of H 2 O 2 with data points up to 0.25 mM H 2 O 2 . This slope indicates the apparent first-order rate dependence of the reaction with respect of H 2 O 2 as described for Mt. and Tf. trHbO [78,79]. This analysis of the Rd. trHb data gives a very similar value (1.5 x 10 3 M -1 s -1 ) (Inset in Fig. 10C) compared to those determined for Mt. trHbO (1.4 x 10 3 M -1 s -1 ) and Tf. trHbO (1.9 x 10 3 M -1 s -1 ). This slope is also a measure of the apparent k cat /K m when the concentration of the variable substrate is small relative to K m which reduces the Michaelis-Menten equation to V = V max /K m or k = k cat /K m when the initial rates are divided by the concentration of enzyme. Such an analysis gives a k cat /K m ratio of 3.0 x 10 4 M -1 s -1 which is consistent with the k cat /K m ratio calculated from the apparent k cat and K m values (Table 4). Thus, our results show overall that Rd. trHb has a very significant peroxidase activity similar to that of Mt. and Tf. trHbO although it differs from the latter in showing saturation kinetics with respect to H 2 O 2 when ABTS is the organic substrate.
To enhance our understanding on the peroxidase reaction catalyzed by Rd. trHb, we performed steady-state kinetic measurements with H 2 O 2 /ABTS in buffers of different pHs (pH 6 to 10.5) ( Table 5). The peroxidase reaction catalyzed by Rd. trHb showed similar catalytic efficiencies in the pH range investigated, with k cat,app /K m,app values within a 5-fold range   11A). The k cat,app /K m,app ratios were slightly greater at lower pH than at higher pH (Fig. 11A). Nevertheless, compared with the reactions catalyzed at neutral to mild-alkaline pH (pH 7 and 8), at pH 6, both the K m,app and the turnover rate were smaller ( Fig. 11B and 11C). The smaller K m,app value suggests that H 2 O 2 apparently binds more tightly to the heme at low pH despite being harder to deprotonate. It should be cautioned that the steady-state kinetic parameters reported in this paper were measured from the oxidation of ABTS, so possible pH-dependent effect of ABTS binding to the active site of Rd. trHb cannot be ruled out. A slower catalytic turnover may be correlated with a lower fraction of the proximal histidine expected to be deprotonated at low pH (i.e. having an imidazolate-character). Data fitted with equations 1 and 2 described in Materials and Methods indicate that all pH profiles yield one pK a value: 6.5 ± 0.2 (k cat,app ), 7.0 ± 0.3 (K m,app ) and 6.8 ± 0.2 (k cat,app /K m,app ) ( Table 5). Thus, a pK a of~6.8, which is close to the pK a of the side chain of histidine free form (~6.0), is observed in k cat,app , K m,app and k cat,app /K m,app . If the pK a value of~6.8 represents the pK a of the proximal histidine of Rd. trHb, it could explain that a significant fraction of the protein displayed a proximal histidine with imidazolate character as detected from the resonance Raman spectrum collected in a buffer at pH 8.0 (Fig. 7). Although Rd. trHb displays clear differences of its active site compared to Mt. and Tf. trHbO, it shares with the latter a significant peroxidase activity (see above), and thus the ability to oxidize an organic substrate using hydrogen peroxide [78]. Meanwhile, bi-functional catalase/peroxidase (encoded by the katG gene), which has been detected in several Roseobacters [81,82,83], may be related to oxidation of secondary metabolites and production of biofilms [84,85,86,87,88], as well as glutathione peroxidase (encoded by the gpo gene), which is known to detoxify ROS in aerobic organisms [8,37]. Herein, we examined the possible physiological role of Rd. trHb in vivo using gene expression experiments (Fig. 12A). R. denitrificans was grown in minimal medium containing acetate with or without H 2 O 2 (1 mM) and the transcript levels of glbO (RD1_4240), katG (RD1_2195) and gpo (RD1_0599) were examined during the exponential growth phase. Significant up-regulation of the gpo and katG genes, along with moderate increment of glbO, were observed in the cultures exposed to H 2 O 2 , compared with cultures without H 2 O 2 , implying potential detoxification of reactive oxygen via Rd. trHb in vivo.
Different globins are known to possess a NO dioxygenase activity that converts NO to harmless nitrate from the reaction of the oxygenated state of these protein with NO [57,89,90,91]. M. tuberculosis was proposed to use Mt. trHbN to detoxify NO produced by macrophages [92], whereas Mt. trHbO was shown to have a 1,200-fold reduced NO dioxygenase activity compared with Mt. trHbN making it unlikely that it is involved in such a reaction  [58]. It was noted that proteins that possess a NO dioxygenase activity share the property of having more than one residue interacting with the heme bound ligand, such as carbon monoxide. For these proteins, a dynamic conversion of open and closed state was suggested from the detection of two Fe-CO stretching modes of the Fe(II)-CO complex [68]. The single Fe-CO   [93]. For this protein, it was proposed that the reaction of the ferric protein with H 2 O 2 would lead to the formation of a ferryl species that would then react with NO to produce nitrite. Such a reaction bypasses the need of a reductase to reduce the heme and to allow oxygen binding as occurs in flavohemoglobins as they carry out the NO dioxygenase reaction. The relatively efficient reaction with peroxide catalyzed by Rd. trHb discussed above may be relevant to such a NO detoxification reaction. An enzyme with a NO detoxification activity, either through a NO dioxygenase reaction or through a reaction involving a ferryl heme and NO, would be helpful during aerobic denitrification in R. denitrificans. While aerobic denitrification is no longer recognized as an uncommon process in nature [94], R. denitrificans has not yet been shown to grow under strictly denitrification condition [10] and thus, it seems unlikely that it uses nitrate denitrification as an energy source. Since R. denitrificans cannot grow aerobically without the supply of nitrate (data not shown), the biological and physiological roles of its aerobic denitrification activity, in addition to consuming excess reducing equivalents to minimize the production of ROS [95], remains to be understood. Here, we used QRT-PCR to monitor the expression level of gene transcripts from cultures grown in the presence of glucose in day-night growth cycles. Fig. 12B shows that glbO and genes involved in denitrification, biomass production and carbon assimilation are stimulated by light, whereas genes involved in carbohydrate metabolism and the TCA cycle are repressed. These observations are in agreement with the fact that Roseobacters perform photosynthesis during phototrophic growth and obtain energy from the oxidation of organic carbons via aerobic respiration during chemotrophic growth [96]. Moreover, up-regulation of the NO reductase gene indicated from our data, along with the known higher cellular levels of the nitrite reductase (cytochrome cd 1 complex) [11] and activation of the aerobic denitrification process by light [9], suggest that more NO molecules are generated during phototrophic growth.
NO produced by aerobic denitrification may function as a signal transducer to inhibit the cytochrome c oxidase activity and thus could down-regulate aerobic respiration, as reported in mitochondria [97]. NO is also known to inactivate aconitase of the TCA cycle and 6-phosphogluconate dehydratase of the Entner-Doudoroff (ED) pathway [98]. The latter is the pathway by which R. denitrificans breaks down sugars to produce pyruvate molecules and synthesizes ATP, NADH and NADPH to preserve energy [37]. Alternatively, active carbon assimilation and biomass production [99], which consume reducing equivalents (NADPH) produced during aerobic photosynthesis, would be stimulated to minimize the production of ROS. NADPH is known to be one of the major source for the production of ROS [100] and Roseobacters have been suggested recently to be the major source of ROS in ocean [5]. ROS are not desirable during aerobic photosynthesis. They also contribute to the production of RNS and toxic peroxynitrite via reaction with NO. It is possible that R. denitrificans produces more trHb to convert excess NO and O 2 to nitrate/nitrite during phototrophic growth. A similar mechanism was also suggested recently to link the activity of THB1, which is a trHb from Chlamydomonas reinhardtii that exhibits NO dioxygenase activity, to nitrogen assimilation in this organism [101]. This hypothesis will be verified in future studies.

CONCLUSIONS
Hemoglobins are known to bind various gas molecules and some of them also possess enzymatic activities. Analysis using bioinformatic tools suggest that putative globin-like proteins belonging to the truncated globin family are present in Roseobacters. In this paper, we characterized the recombinant trHb from R. denitrificans using functional, spectral and modeling studies. The sequence alignment suggests that Rd. trHb belongs to group II trHbs and contains distal amino acids known to participate in ligand stabilization in this group of heme proteins. Resonance Raman spectra reveal a heme pocket structure consistent with the very significant peroxidase activity measured by steady-state kinetics with Rd. trHb. Together, these properties suggest a role for Rd. trHb in mechanisms of defense against reactive oxygen and/or nitrogen species or in redox biochemical reactions that remain to be explored further. Gene expression analyses also suggest that Rd. trHb contributes the detoxification of reactive oxygen and nitrogen species in vivo. Current genomic sequence data indicate that trHbs genes are widely distributed in the ecologically important marine Roseobacters. Some of the distinctive properties reported in this paper set the stage for further in-depth analyses of this interesting enzyme.