Heme and nitric oxide binding by the transcriptional regulator DnrF from the marine bacterium Dinoroseobacter shibae increases napD promoter affinity

Under oxygen-limiting conditions, the marine bacterium Dinoroseobacter shibae DFL12T generates energy via denitrification, a respiratory process in which nitric oxide (NO) is an intermediate. Accumulation of NO may cause cytotoxic effects. The response to this nitrosative (NO-triggered) stress is controlled by the Crp/Fnr-type transcriptional regulator DnrF. We analyzed the response to NO and the mechanism of NO sensing by the DnrF regulator. Using reporter gene fusions and transcriptomics, here we report that DnrF selectively repressed nitrate reductase (nap) genes, preventing further NO formation. In addition, DnrF induced the expression of the NO reductase genes (norCB), which promote NO consumption. We used UV-visible and EPR spectroscopy to characterize heme binding to DnrF and subsequent NO coordination. DnrF detects NO via its bound heme cofactor. We found that the dimeric DnrF bound one molecule of heme per subunit. Purified recombinant apo-DnrF bound its target promoter sequences (napD, nosR2, norC, hemA, and dnrE) in electromobility shift assays, and we identified a specific palindromic DNA-binding site 5′-TTGATN4ATCAA-3′ in these target sequences via mutagenesis studies. Most importantly, successive addition of heme as well as heme and NO to purified recombinant apo-DnrF protein increased affinity of the holo-DnrF for its specific binding motif in the napD promoter. On the basis of these results, we propose a model for the DnrF-mediated NO stress response of this marine bacterium.

gen (N 2 ) (1). The intermediate nitric oxide is cytotoxic by accumulation. As a consequence, the transition process from oxygen respiration to denitrification is tightly controlled.
Many bacteria utilize so-called Crp (cAMP receptor protein)/Fnr (fumarate and nitrate reductase regulator) superfamily regulators for this purpose. This highly flexible group of transcription factors shares a low overall amino acid sequence identity (2). However, they all contain a C-terminal DNA-binding helix-turn-helix motif and an N-terminal sensory domain formed by multiple antiparallel ␤-strands that are connected by a long dimerization helix (2)(3)(4). According to the binding of the cofactor at the individual sensing region, a defined signal can be detected (2). Considering the localization of their binding sites, these proteins can act as activators or repressors of transcription. Binding sites upstream of or even overlapping with the Ϫ35 region of a standard bacterial promoter lead to transcriptional activation. Binding sites overlapping with or downstream of the Ϫ10 region are usually mediating gene repression (2,5).
One prominent example of Crp/Fnr regulators is the global transcription factor Fnr of Escherichia coli forming together with the homologous proteins Anr, FnrA, FnrL, and FnrP, a major subgroup of Crp/Fnr regulators. All of them sense oxygen via an oxygen-labile [4Fe-4S] cluster (5). This subgroup possesses four conserved cysteine residues at the N or C terminus for ligation of the oxygen and NO-sensitive iron-sulfur clusters (6 -11).
NsrR of Bacillus subtilis, another Fe-S cluster-containing regulator, plays a major role in NO detoxification (12). NsrR with intact Fe-S cluster represses hmp gene expression under anaerobic conditions (13). The hmp gene, encoding a flavohemoglobin, plays a central role in B. subtilis for the response to nitrosative stress (14,15). Accumulation of NO inactivates NsrR by destabilizing the Fe-S cluster and leads to the derepression of hmp transcription.
In Pseudomonas aeruginosa another Crp/Fnr-like regulator, Dnr (dissimilatory nitrate respiration regulator), was shown to create a fine-tuned hierarchical regulation of denitrification genes like nirS, norCB, and nosR (16 -19). In Pseudomonas stutzeri DnrD is involved in nitric oxide signaling and transcriptional control of denitrification genes (20).
The Dnr regulators are lacking the conserved cysteine residues of Fnr involved in iron-sulfur cluster ligation. But they still respond to nitric oxide and mediate the regulation of various denitrification genes (20). Members of this regulatory subfamily coordinate a heme cofactor (21)(22)(23). The crystal structure of the sensor domain of P. aeruginosa Dnr without bound heme was solved (24). Functional studies identified histidine residues 139 and 186 as axial ligands for heme binding (25). In response to the binding of the effector molecule NO, a conformational change of heme-bound Dnr was proposed, which results in efficient DNA binding (24,26).
Conflicting data exist about the DNA-binding capabilities of the protein. An initial EMSA study showed DNA binding solely for the P. aeruginosa Dnr with bound heme, whereas apo-Dnr failed to bind the nirS promoter sequence (24). Fluorescence anisotropy measurements of a Texas Red-labeled nor promoter fragment bound to P. aeruginosa holo-Dnr revealed DNA binding with a K D of 44 Ϯ 9 nM only in the presence of NO. Holo-Dnr without bound NO and the CO-bound form of the regulator failed to bind the promoter, even if the regulator is provided in a 50-fold excess over the DNA (21).
In P. aeruginosa the binding sites for Dnr and the Fnr analogue Anr can optically not be distinguished (27). Both regulators recognize the palindromic sequence TTGATN 4 ATCAA. However, clear-cut Anr, Dnr, and Anr ϩ Dnr (both regulators regulate the same gene via one binding site) regulons can be defined using transcriptomics (19). The molecular mechanism underlying promoter specificity and transcriptional activation by Dnr after binding to the promoter remains to be determined.
The marine bacterium Dinoroseobacter shibae DFL12 T (DSM 16493), as one representative of the highly abundant Roseobacter group of marine bacteria, possesses a fine-tuned hierarchical network for the adaptation to anaerobic conditions (28). One FnrL and three Dnr regulators (DnrD, DnrE, and DnrF) mediate the transition from aerobic to anaerobic conditions by activating denitrification genes in the presence of nitrate (28).
In D. shibae nitrate gets reduced to nitrite using a periplasmic dissimilatory nitrate reductase (nap-type), a periplasmic cd 1 -type nitrite reductase (NirS), a membrane-localized NO reductase (NorCB), and a periplasmic N 2 O reductase (NosZ) (29). All of these enzymes as well as corresponding maturation and regulatory proteins are encoded in a large denitrification gene cluster.
In a previous study, we determined the role of FnrL and the three Dnr regulators (DnrD, DnrE, and DnrF) for their response to anaerobiosis in the presence of nitrate (28). FnrL and DnrD were identified as key players for the regulation of the transition process.
Here, we investigated the regulatory function of DnrF. Because the toxic gas NO is produced during denitrification and released, the organism requires a fast and effective stress response. It must include detection of NO and a subsequent gene regulatory scenario that leads to a stop of NO production and enhances NO consumption. DnrF was identified as the NO stress regulator of D. shibae, where NO is detected via a bound heme. Repression of the nitrate reductase genes by DnrF prevented further NO production, although the parallel induction of the NO reductase genes ensures the consumption of the harmful N-oxide.

Defining NO stress in D. shibae
During the process of denitrification, several dangerous N-oxides, including NO, are generated as intermediates. From previous experiments we knew that the nirS gene, encoding the enzyme nitrite reductase, responsible for NO generation, is controlled by NO. To determine the role of sublethal NO concentrations as signal for gene expression in D. shibae, we grew the wild-type strain DFL12 T harboring a nirS-lacZ reporter gene fusion under aerobic conditions until an optical density at 578 nm of 0.5 was reached. Subsequently, the cells were shifted to anaerobic conditions. Depletion of oxygen was determined using a PreSense Fibox 3 trace assay in combination with a PSt3-type oxygen sensor (detection limit 15 ppb dissolved oxygen, 0 -100% oxygen). When all oxygen was consumed by the culture, we added NO-saturated water to final NO concentrations of 50, 1000, 10,000, and 70,000 nM and monitored growth and ␤-galactosidase activities from the nirS-lacZ fusion after 15, 30, and 60 min (Fig. 1). These highly different NO concentrations were used because in the literature quite diverse NO concentrations were employed to study NO-dependent transcriptional changes. In P. stutzeri 50 nM NO caused an inducing effect on nirS gene expression. In contrast, in P. aeruginosa 10,000 nM were used to induce norC-lacZ reporter gene expression (20,30), and in Gram-positive organisms like B. subtilis survival of 50,000 nM NO can be observed (31).
After shifting the cells to gas tide serum flasks, oxygen consumption occurred within 15 min. Because of the loss of an electron acceptor, growth stopped after the anaerobic shift indicating a major metabolic crisis for D. shibae (32). After addition of 50, 1000, and 10,000 nM NO, a slight decrease in optical density was recorded, but after that the cell density remained stable. However, 70,000 nM NO leads to a significant drop in cell density indicating cell death and cell lysis (Fig. 1A). For ␤-galactosidase assays, samples were taken 0, 15, 30, and 60 min after the addition of indicated NO concentrations (Fig. 1B 1B). This NO-independent increase of nirS-lacZ expression is due to an anaerobic induction mediated by FnrL (28). Addition of 50 nM NO led to ␤-galactosidase activities of 16 Miller units after 15 min, 93 Miller units after 30 min, and 119 Miller units after 60 min. The presence of 50 nM NO increased the nirS-lacZ reporter gene expression by a factor of 1.5. Addition of 1000 or 10,000 nM resulted in ␤-galactosidase activities comparable with the untreated anaerobic culture. The NO concentration of 70,000 nM led to a loss in activity presumably caused by the cell death generated by the toxic effect of NO. Thus, 50 nM was identified as the NO concentration with best gene-inducing effect without severe toxic effects at the same time.

NO-dependent gene expression in D. shibae
The anaerobic NO-dependent global gene expression was investigated by DNA arrays comparing transcripts from D. shibae wild-type cells grown anaerobically in the presence and NO stress regulator DnrF of Dinoroseobacter shibae absence of 50 nM NO. For this purpose, shift experiments as outlined above were performed using a final concentration of 50 nM NO. RNA samples were collected 30 min after the anaerobic shift and 30 min after the addition of NO. A differential gene expression was assumed for genes with log2 fold changes higher or equal to 0.8. Overall, five genes were found induced and 10 repressed by the presence of NO (supplemental Table  S1). The denitrification gene cluster appeared to be the major target for NO-dependent gene expression. Genes significantly up-regulated after NO treatment were nirH and nirG as part of the nir operon encoding proteins for the maturation of the nitrite reductase, and norC encoding the small subunit of the nitric-oxide reductase. Moreover, the dnrE gene encoding the transcription factor DnrE was induced (supplemental Table  S1).

Regulon of the NO stress regulator DnrF
To determine the role of the DnrF regulator in mediating NO-dependent regulation, its regulon was defined by comparing the transcriptome of the wild-type strain with the corresponding mutant strain DS004 (⌬dnrF). A large NO-dependent regulon was found for DnrF with 365 induced and 199 repressed genes compared with wild type, indicating a major regulatory function of DnrF (supplemental Table S2). The comparison of the NO-mediated response of the wild-type strain and the DnrF regulon using a heat map analysis revealed significant overlap in the regulation of the denitrification gene cluster (Fig. 2). The napDAGHBC operon, encoding the dissimilatory periplasmic nitrate reductase, was found to be repressed by DnrF, which was indicated by the increased transcription in the ⌬dnrF mutant strain compared with the wild type (Fig.  2). This finding is in accordance with previous observations (28). This repression is mediated via the palindromic sequence 5Ј-TTGATN 4 ATCAA-3Ј, which is located 96.5 bp upstream of the napD translational start site (Fig. 5B). Furthermore, a DnrFdependent activation was found for the nirSECFDGHJN operon, indicated by a reduced transcription in the ⌬dnrF mutant strain compared with the wild type (Fig. 2). Interestingly, only nirG and nirH exhibited a significant alteration in gene expression but lacked an assignable regulator-binding site. Moreover, a significant activation was found for the norC/B genes encoding nitric-oxide reductase. A corresponding potential DnrF regulator-binding site was found 67.5 bp upstream of the translational start site (Fig. 2). However, no significant alteration in transcription of the nosRZDFYLX operon dependent on DnrF was found after NO treatment. Overall, the observed regulatory scenario with an NO-dependent repression of the NO-producing system (nitrate reductase Nap) and an induction of the NO-consuming enzyme (NO reductase NorCB) finally identifies DnrF as NO stress regulator of D. shibae.
Because several of the observed DnrF-regulated genes might also be regulated by the DnrD regulator, the corresponding mutant strain (DS002 (⌬dnrD)) was analyzed under NO-inducing conditions, and genes of overlapping regulons were defined. The NO-dependent DnrD regulon consists of 17 genes and is given in supplemental Table S1. In total, 22 genes belong to the overlapping regulon of DnrD and DnrF (supplemental Table  S1). Thus, a specific DnrF regulon was determined, which is composed of a total of 80 genes. Because not all genes exhibit a potential DnrF-binding site, a significant portion of indirect regulatory scenarios can be deduced from the DnrF-dependent repression of various transcriptional regulator genes like zntR, an ATP-dependent transcriptional regulator (Dshi_1881) and an XRE family transcriptional regulator (Dshi_2488), which exhibited a potential DnrF-binding site in their promoter sequences. However, most potential DnrF-binding sites were mainly found in promoters of genes of unknown function (Dshi_0391, Dshi_0625, Dshi_2048, Dshi_2148, and Dshi_3542). Furthermore, a striking influence of DnrF on numerous genes encoding transport systems was observed. A direct activation of an S-adenosylmethionine uptake transporter (Dshi_2148), a sulfite exporter tauE/safE (Dshi_0205), sugar transporter systems (Dshi_0488/Dshi_1808), a periplasmic dicarboxylate shibae DFL12 T wild-type strain carrying the nirS-lacZ reporter gene fusion was grown under aerobic conditions until the mid-exponential growth phase (gray) and then shifted to anaerobiosis. NO-saturated water for final concentrations of 0 nM (black), 50 nM (blue), 1000 nM (green), 10,000 nM (red), and 70,000 nM (violet) was added to the anaerobic cultures. A, optical density at 578 nm of the differentially treated cultures was measured over time. B, nirS-lacZ-derived ␤-galactosidase activity was measured for samples taken 0, 15, 30, and 60 min after NO addition. Results given in Miller units represent the average of at least three independent experiments performed in triplicate with a standard error of less than 10%.

NO stress regulator DnrF of Dinoroseobacter shibae
transporter (dctQ/dctM), and cation/H ϩ antiporter (mnhC/ mnhG) were found. However, significant repression by DnrF was found for the genes of two ABC transporters (znuC/ Dshi_1421/Dshi_2434). The DnrF-binding site in the promoter of Dshi_1421 encoding a potential ribose ABC transporter indicated direct regulation. An impact on the general stress response was previously shown for periplasmic transporter (dctQ/dctM) and cation/H ϩ antiporter (33). Only a few genes of the general stress response, including the exonuclease (urvB), a cytochrome c peroxidase gene with potential DnrF-binding site (Dshi_2749), and the cell division factor (mraZ) were found controlled by DnrF. A significant activation was found for a potential quorum-sensing gene (Dshi_3013) and the plasmidencoded replication protein C (repC). A significant role of quorum-sensing genes and plasmid-encoded genes in the general stress response was previously shown (33,34). Alternative metabolic strategies like arginine fermentation via ornithine carbamoyltransferase (arcB), acetate utilization via acetate-CoA ligase (acsA), or control of the central metabolism via the malic enzyme (maeB) were found to be regulated by DnrF. This makes sense in the light of the metabolic problems occurring after the utilization of residual oxygen amounts (32). Nevertheless, regulation by DnrF was observed for numerous other genes like hisD, encoding a histinol dehydrogenase. Moreover, a direct regulation via a potential DnrF-binding site in the promoter was found for a gene encoding a potential hemolysin (Dshi_2763) and the gene encoding a cobalamin synthesis protein P47K (Dshi_2592). The advantage in regulating these genes under nitrosative stress conditions remains unclear.

DnrF senses NO via bound heme
The result of the transcriptome analyses pointed toward a role of DnrF in mediating the observed NO-dependent gene regulation in D. shibae. For thorough biochemical characterization of the DnrF protein, the transcription factor was heterologously produced as a His-tagged fusion protein in E. coli and chromatographically purified to apparent homogeneity under strict anaerobic conditions. The ability of DnrF to bind heme was determined spectroscopically by titration of heme to the DnrF apoprotein under anaerobic conditions. A stoichiometric binding of approximately 1 mol of heme/mol of subunit of DnrF was observed (Fig. 3A). A mostly dimeric state of the purified DnrF protein was determined by gel-permeation chromatography. DnrF has a calculated molecular mass of 26,260 Da. Apo-DnrF, holo-DnrF, and holo-DnrF treated with NO eluted when analyzed under anaerobic conditions as two peaks with relative molecular masses of ϳ52,000 Ϯ 3000 and 106,000 Ϯ 5000 Da (Fig. 3B). In all analyzed cases the major oligomeric form observed was dimeric. The higher-ordered form of DnrF detected corresponded to a tetramer, most likely as dimer of dimers. The detected amounts of this form were decreasing from apo-via holoto holo-DnrF with NO. Higher-ordered complexes were also observed during the electrophoretic mobility shift assays (EMSA) experiments (Fig. 4). Perhaps these higher ordered complexes represent a DNA-bound storage form of inactive DnrF.
UV-visible absorption spectroscopic analysis of the purified apo-DnrF protein failed to detect absorption indicative of a bound heme (Fig. 3, B and C). Free heme has an UV-visible absorption spectrum with maxima at 365 and 385 nm. When heme was added to apo-DnrF in a stoichiometric range, the resulting holo-DnrF with bound heme in its ferric form showed a Soret peak at 417 nm ( Fig. 3C). Reduction of the bound heme to its ferrous form by the addition of dithionite changed the corresponding absorption maxima of the Soret, ␣, and ␤ peak to 427, 530, and 560 nm, respectively (Fig. 3, B and C). However, shoulders at 360 and 392 nm indicated residual amounts of unbound heme in the reaction (Fig. 3C). Because the heme molecule is not covalently bound to the DnrF protein, an equilibrium reaction between the bound and unbound state was assumed. These recorded spectra are similar to those of heme proteins with histidine residues as axial ligands, like the heme regulator HrtR of Lactococcus lactis (35). In Dnr of P. aeruginosa, an essential function in heme coordination was assumed for His-187 (25). However, an amino acid sequence alignment of DnrF from D. shibae with Dnr of P. aeruginosa and DnrD from P. stutzeri failed to detect a conserved histidine residue, which may be involved in heme coordination. Even the previously experimentally determined histidine residues of P. aeruginosa Dnr His-167 and His-187 were not conserved in D. shibae Dnr (25,36). The addition of 1% NO in gaseous form to holo-DnrF altered the absorption maximum of the Soret

NO stress regulator DnrF of Dinoroseobacter shibae
peak from 427 to 393 nm, and the solution became red (Fig. 3C) (24,25). Furthermore, a blur of the Q bands was observed due to vibrational excitations. To investigate the NO-dependent alteration in the absorption spectra of DnrF, electron paramagnetic resonance (EPR) measurements were performed. Thirty M heme-reconstituted DnrF were analyzed in its ferric, ferrous ion, and NO-bound state at 4 -5 K in an X-band EPR spectrometer (Fig. 3D). Data plotting for heme-reconstituted DnrF versus pure heme revealed comparable spectra showing an axial EPR resonance with a g value of 6.0. The ferric heme displayed an EPR signal that is typical and consistent with a five-coordinated high-spin Fe 3ϩ system S ϭ 5/2. However, a slight shift in the heme-containing DnrF spectrum was noticeable. Thus, a weak coordination of the Fe 3ϩ to an s-donor or other secondary forces in a polar protein pocket can be assumed. The reduction of the Fe 3ϩ with 2 mM DTT resulted in a loss of signal, due to the EPR silent behavior of Fe 2ϩ . The exposure of ferrous heme con-taining DnrF to 1% NO gas mixture resulted in a broad signal with an isotropic g value of g ϭ 2.05. The resonance is consistent with a five-coordinated, low-spin Fe 2ϩ nitrosyl complex {FeNO} 7 (Fig. 3D) (37).

Functional identification of the DnrF-binding site
The transcriptome analyses comparing D. shibae wild type and the corresponding ⌬dnrF mutant strain also led to the identification of several genes showing a clear-cut DnrF-dependent regulation (Fig. 2). Five of them, namely hemA3, nosR2, norC, dnrE, and napD, were selected to study their DnrF-binding properties to the corresponding promoters using EMSA. As a reference, the nosR1 promoter was chosen. Initial experiments were performed under aerobic conditions. Doublestranded promoter fragments of 75 bp in length containing the putative DnrF-binding sequences with nucleotide sequence identity to the palindromic sequence 5Ј-TTGATN4ATCAA-3Ј

NO stress regulator DnrF of Dinoroseobacter shibae
were labeled with digoxigenin and incubated with increasing amounts of purified His-tagged apo-DnrF. This palindromic sequence was considered to be specifically recognized by Dnr regulators (38). The equilibrium dissociation constants (K D ) of apo-DnrF/promoter fragment were estimated by non-linear regression according to the Hill equation: y ϭ [DnrF]/(K D ϩ [DnrF]). By using the nosR1 promoter fragment only a faint retarded DNA-DnrF complex was observed at DnrF concentration above 64 nM (Fig. 4A). Because of the poor binding of DnrF to the promoter DNA the determination of a specific dissociation constant was not possible. This weak interaction of DnrF with the nosR1 promoter fragment may be due to the improper palindromic sequence 5Ј-TTGATGTCCATGGG-3Ј lacking the conserved half-site motif at the 3Ј end (conserved bases are given in bold). In contrast, binding of apo-DnrF to the hemA3 promoter fragment already occurred at concentrations higher than 8 nM (Fig. 4B). A dissociation constant (K D ) of 44.18 nM was calculated. Here, a higher degree of conservation of the potential DnrF-binding sequence was observed (5Ј-TT-GACGTTGGTTAA-3Ј). An even better binding affinity of DnrF was found for the nosR2 promoter that was localized directly upstream of the nor operon (Fig. 4C). A K D of 14.7 nM was determined. Moreover, a second retarded complex became visible at concentrations higher than 64 nM (Fig. 4C). The second complex found in the EMSA may represent a higher-ordered complex, possibly consisting of a dimer of dimers (Fig.  3B). Next, analysis of the affinity of DnrF for the norC promoter yielded a K D of 11.95 (Fig. 4D). Similar to the results of the nosR2 promoter analysis, a higher-ordered complex was observed at higher DnrF concentrations of 32 nM. Binding studies using the dnrE promoter sequence revealed a DnrF/ DNA interaction already at concentrations of 2 nM and a higher ordered complex above concentrations of 32 nM DnrF (Fig. 4E). Quantitative analyses revealed a dissociation constant of DnrF for the dnrE promoter of 10.68 nM. Binding studies of DnrF to the napD promoter sequences resulted in the highest affinity of DnrF for a promoter sequence. Efficient DNA binding occurred already at concentrations of 2 nM DnrF with a calculated dissociation constant K D of 9.8 nM. Again, the second higher-ordered complex occurred at a DnrF concentration of 32 nM. The high affinity of DnrF to the napD promoter fragment led to a complete shift of the employed DNA fragment in the EMSA, and at DnrF concentrations higher than 128 nM free DNA was no longer detectable (Fig. 4F).
To determine the role of the potential binding site 5Ј-TT-GATCTCGATCAA-3Ј centered at position Ϫ96.5 with respect to the translational start site of napD, the binding sequence was mutated to 5Ј-GCGATCTCGATCGC-3Ј in the context of the 75-bp napD promoter sequence and was used in EMSA studies. DnrF formed a stable complex with the wild-type napD promoter sequence, whereas no DNA binding of DnrF was observed for the mutated napD promoter sequence (Fig. 5A,  lanes 2 and 3). In addition, competition experiments were performed by adding increasing amounts of unlabeled wild-type and mutated napD promoter DNA to the binding assay. Addition of the wild-type napD in a 10-, 50-, and 100-fold molar excess completely abolished DNA binding to the labeled napD promoter fragment. The shifted complex disappeared already with a 50-fold molar excess (Fig. 5A, lanes 4 -6). In contrast, even the highest amount of the mutated napD promoter DNA

NO stress regulator DnrF of Dinoroseobacter shibae
used as a competitor for the binding of DnrF to the labeled napD promoter fragment failed to compete efficiently (Fig. 5A,  lanes 7-9). These experiments clearly demonstrate that DnrF is regulating napD expression via the proposed DnrF-binding site 5Ј-TTGATCTCGATCAA-3Ј at position Ϫ96.5 with respect to the translational start site.

Regulation of napD gene expression by DnrF in vivo
To demonstrate regulation of target gene expression by DnrF in vivo, a napD-lacZ reporter gene fusion was tested in D. shibae. For this purpose, a 165-bp DNA fragment spanning napD promoter sequences from Ϫ145 to ϩ20 with respect to the translational start site was cloned upstream of the lacZ reporter gene resulting in the plasmid pBBRnapD-lacZ (Fig. 5B). The plasmid carrying the napD-lacZ reporter gene fusion was transformed into the D. shibae wild-type and the dnrF mutant strain (DS004 (⌬dnrF)). The resulting D. shibae strains were grown under aerobic and anaerobic denitrifying conditions, and ␤-galactosidase activities were determined from samples taken in the mid-exponential growth phase. Under aerobic conditions, only very low ␤-galactosidase activities close to the detection limit were measured for the wild-type and the dnrF mutant strain (Fig. 5C). Under anaerobic conditions, a 10-fold increase of reporter gene expression up to 139 Miller units was detected for the wild-type strain. However, in the dnrF mutant strain a 32-fold increased expression up to 2690 Miller units was determined (Fig. 5C). These result clearly demonstrated a DnrF-dependent repression of the napD-lacZ reporter gene under denitrifying growth conditions.

Heme binding and NO coordination increase DNA-binding affinity of DnrF
The regulator protein DnrF in its apo-form is already able to bind efficiently to its palindromic target sequence (5Ј-TTGATN 4 ATCAA-3Ј) within the napD promoter. To investigate the consequences of heme and subsequent NO binding to the regulator, DNA-binding affinities of apo-DnrF and holo-DnrF with bound heme in the reduced state and in the presence of NO were determined using EMSA experiments. The anaerobically purified regulator DnrF was incubated with the napD fragment as described above. With increasing amounts of apo-DnrF, two shifted complexes were identified (Fig. 6A). The amounts of shifted complexes were quantified, and binding affinity was deduced by calculating the equilibrium dissociation constant (K D ). For apo-DnrF, a K D value of 8 Ϯ 1.5 nM protein was determined (Fig. 6, A and D). Addition of heme in the ferrous state resulted in a slightly increased binding affinity to a K D value of 4.7 Ϯ 1.3 nM (Fig. 6, B and D). However, addition of NO in gaseous form further increased the binding affinity to a K D value of 1.6 Ϯ 0.4 nM (Fig. 6, C and D). Thus, NO sensing by heme-bound holo-DnrF resulted in a 5-fold increased binding affinity of the regulator to the DNA.

Discussion
Under anaerobic conditions, the utilization of nitrate as a terminal electron acceptor is essential for the survival and growth of D. shibae (32). Corresponding reduction processes lead to the production of the toxic N-oxide NO. Thus, a stringent NO-triggered stress response is needed. Recent investigations of the D. shibae adaptation strategies to anaerobic nitrate-reducing

NO stress regulator DnrF of Dinoroseobacter shibae
conditions revealed a fine-tuned transcriptional regulatory network composed of four different Crp/Fnr-type regulators. Besides one oxygen-sensing FnrL homologue, a set of three potentially NO-sensing Dnr regulators were identified. Initial transcriptome analysis of D. shibae grown under anaerobic, denitrifying conditions identified DnrD as a major denitrification regulator. Here, we addressed the function of DnrF. We essentially were answering the following question. How does D. shibae deal with NO as an intermediate of denitrification? B. subtilis possesses the NO stress regulator NsrR, sensing NO via a bound [4Fe-4S] controlling a large detoxification machinery to avoid NO accumulation (14,15) and that enabled B. subtilis to survive amounts of 50 M NO (31). However, for D. shibae a sublethal NO concentration of 10 M was determined. This is in accordance with other members of the Alphaproteobacteria, like Rhodobacter sphaeroides (39) and Paracoccus denitrificans (40,41). Both organisms possess Dnr homologues called NnrR and Nnr, respectively. An essential role of those regulators under nitrosative stress conditions was demonstrated (39,40,42). However, their detailed biochemical function remains to be determined. With respect to the D. shibae DnrF regulon, the major findings were the repression of the genes for the nitrate reductase (nap) to prevent further NO 2 Ϫ and subsequent NO formation together with the induction of the NO-consuming NO reductase (nor). Production of NO reductase is also a strategy of several pathogenic bacteria to withstand NO produced by macrophages of the human immune system (43). Interestingly, only one of the two genes encoding potential flavohemoglobins, Dshi_1666, is under the control of DnrD and DnrF under anaerobic conditions. Biochemical analyses of D. shibae DnrF using UV-visible and EPR spectroscopy revealed a non-covalent coordination of heme and the binding of NO by forming a low-spin Fe 2ϩ nitrosyl complex {FeNO} 7 . The distal and proximal ligands of the heme iron of DnrF are unknown. Amino acid sequence alignments of Dnrs from various Pseudomonas species revealed a conserved stretch of 26 amino acids, of which 15 residues were assumed to form an apolar pocket to accommodate the heme molecule (24). This amino acid stretch is missing in D. shibae DnrF. Moreover, conserved histidine residue were functionally identified to be responsible for heme coordination in P. aeruginosa Dnr (25). All Dnr regulators of D. shibae lack the corresponding conserved histidine residues.
The palindromic sequence 5Ј-TTGATN 4 ATCAA-3Ј upstream of the napD gene was shown to be recognized specifically already by apo-DnrF. The reconstitution of apo-DnrF with heme leads to a significant increase in binding affinity of DnrF. Additional coordination of NO to holo-DnrF increased the binding affinity. This is in contrast to findings of P. aeruginosa Dnr, where apo-Dnr failed to bind the nirS promoter sequence (24). Fluorescence anisotropy measurements of holo-Dnr revealed DNA binding with a K D of 44 Ϯ 9 nM only in the presence of NO. Free holo-Dnr and the CO-bound form of the regulator failed to bind the promoter fragment (21). In contrast, DnrF is binding already in its apo-form, which enables D. shibae to have a fast and precise stress response to NO (Fig. 7). Generation of holo-DnrF via heme binding to apo-Dnr increased DNA affinity. Heme binding might induce already the structural changes to Dnr responsible for the observed increase in DNA-binding affinity. In the presence of NO, the DNA affinity of holo-DnrF was found further increased. This resulted in either strong gene repression, like in the case of the nap

NO stress regulator DnrF of Dinoroseobacter shibae
operon, or gene activation as observed for the other tested DnrFdependent promoters. The regulatory activity of NO-bound holo-DnrF is finally dependent on the localization of the corresponding binding site in the target promoter as observed for regulators of the Crp/Fnr class (2,5). In the case of the nosR2 promoter, the binding site is located at position Ϫ41.5 with respect to the transcriptional start site, a typical site for transcriptional activation (44). Transcriptional repression of the nap promoter might be due to blocking of the corresponding promoter at the Ϫ10 region. Furthermore, DnrF also differentially modulates the transcriptional transactivation as observed for the nor operon (Fig. 2) most likely via differential interaction with the RNA polymerase upon NO. As summarized in Fig. 7, a fast, DnrF-controlled and solely denitrification enzyme triggered response stop NO production and leads to the consumption of residual NO. Interestingly, potential flavohemoglobin genes for NO detoxification were not induced. These observations describe a novel bacterial strategy for the detoxification of the hazardous N-oxide NO.

Bacterial strains, media, and growth conditions
Cultures of the type strain D. shibae DFL12 T , ⌬dnrD mutant strain DS002, and ⌬dnrF mutant strain DS004 (Table 1) were grown in marine-bouillon (MB, Roth, Karlsruhe, Germany) at 30°C in flasks shaking at 200 rpm in the dark or on MB plates solidified with 1.5% agar. For D. shibae mutant strains, the DS002 and DS004 medium was supplemented with 80 g/ml gentamycin. E. coli strains were routinely grown in lysogenic broth (LB) supplemented with the appropriate antibiotics and amino acids at 37°C and shaking at 200 rpm (Table 1) (45). To investigate the growth behavior under aerobic and anaerobic conditions, the D. shibae strains were grown in artificial seawater medium (SWM) 2 (46) supplemented with 16.9 mM succinate in flasks shaking at 200 rpm. For nitrosative stress experiments, 25 mM pyruvate was added, and incubation was performed in serum flasks sealed with rubber stoppers and shaking at 100 rpm (26). NO-saturated water solution was prepared according to Moore et al. (31).

Growth curve and shift experiments
For growth experiments, a pre-culture of the appropriate D. shibae strain was inoculated in SWM supplied with 16.9 mM succinate and grown overnight at 30°C and 200 rpm in the dark. Next, 125 ml of the main culture was inoculated to an OD 578 nm of 0.05 in SWM supplied with 16.9 mM succinate in a 1-liter baffled flask. After reaching an OD 578 nm of ϳ0.5, the cultures were shifted to anaerobic conditions. For anaerobic cultivation, the rubber stopper sealed serum flask was used and 25 mM pyruvate was added. Oxygen tension was measured every 5 min using a PreSense Fibox 3 LCD trace version 7 and an oxygen sensor type PSt3 with an accuracy of Ϯ0.15% (Pre-Sense, Regensburg, Germany). After 30 min of anaerobic cultivation, NO saturated water was added, under strict anaerobic conditions, for final amounts of 50, 1000, 10,000, and 70,000 nM NO (31). Samples were taken for RNA preparation after 30 min of anaerobic starvation and 30 min after injection of appropriate amounts of NO-saturated water solution.

RNA preparation and DNA microarray
RNA preparation following DNA array analysis was performed as described elsewhere (28,32). Generated data have been deposited in the NCBI Gene Expression Omnibus (47) and are accessible by Geo Series accession number GSE95560.

Promoter-lacZ reporter gene fusions
A 165-bp napD DNA fragment corresponding to promoter sequences from position Ϫ145 to ϩ20 with respect to the translational start was PCR-amplified using primers EH689 and EH670 (Table 2) together with the LIC tail sequences and cloned by ligation-independent cloning into pBBRLIC-lacZ resulting in plasmid pBBRnapD-lacZ as described previously (28).

Recombinant DnrF production
The dnrF gene (Dshi_3270) was PCR-amplified from D. shibae genomic DNA using primers oPT163 and oPT164 containing NdeI and BamHI restriction sites, respectively ( Table 2). The resulting PCR product was digested using NdeI and BamHI and ligated into the equally treated vector pET14b (Novagen, Darmstadt, Germany). The resulting vector pET14DnrF was 2 The abbreviations used are: SWM, seawater medium; DIG, digoxigenin.

NO stress regulator DnrF of Dinoroseobacter shibae
used for production of the N-terminal His-tagged DnrF fusion protein. For heterologous production of DnrF, the E. coli BL21 (DE3) pLysS strain was used. For the production of the Histagged DnrF protein, pET14DnrF carrying the E. coli cells was grown from a starting OD 578 nm of 0.05 in 500 ml of LB medium containing 100 g/ml ampicillin in a 1-liter flask. Incubation was carried out at 37°C with shaking at 200 rpm. After reaching an OD 578 nm of 0.5-0.6, production of His-tagged DnrF protein was induced by adding 50 M isopropyl ␤-D-thiogalactopyranoside, and the cultures were transferred to 17°C with shaking at 100 rpm for 16 h. Cells were sedimented by centrifugation for 15 min at 4000 ϫ g at 4°C. The resulting cell pellet was resuspended in 10 ml of binding buffer (100 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 20 mM imidazole). A French press (19,200 p.s.i.) was used for cell disruption, and the soluble protein fraction was obtained by ultracentrifugation at 160,000 ϫ g for 65 min at 4°C. The supernatant was loaded onto a 1-ml nickelpacked chelating Sepharose Fast Flow column (GE Healthcare, Solingen, Germany). The column was washed twice with 10 ml of washing buffer (100 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 30 mM imidazole). The bound proteins were eluted by adding 10 ml of elution buffer (100 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 200 mM imidazole). The elution fraction was dialyzed against 1 liter of 100 mM Tris-HCl, pH 7.5, 150 mM NaCl buffer for 16 h at 4°C in darkness, using a dialysis tube membra-cel TM (Carl Roth GmbH ϩ Co. KG, Karlsruhe, Germany). The purified protein was stored at Ϫ20°C. Protein fractions were analyzed by SDS-PAGE (48,49). Protein concentrations were determined using the Beer-Lambert law as well as the Bradford reagent (Sigma) according to the manufacturer's instructions. BSA and ovalbumin were used as reference proteins. UV-visible spectra were recorded on a Jasco V-650 spectrophotometer (Jasco Inc., Easton, MD) using a high precision cell with light path of 10 mm (Hellma Analytics, Müllheim, Germany), and all steps were performed under strict anaerobic conditions. In initial experiments, the His tag was cleaved off and removed. However, in UV-visible spectroscopy, the heme-binding gel filtration, and EMSA analyses, we did not observe any difference in behavior between the His-tagged and the tag-free DnrF. Consequently, all experiments were performed using DnrF-His in accordance with Karnaukhova et al. (50).

Heme titration to apo-DnrF and binding of NO
To investigate heme binding in vitro, 10 M His-tagged DnrF was used. For determination of heme-binding properties, heme solution was freshly prepared. For this purpose 10 mg of heme granulate were incubated in 1 ml of 100 mM NaOH solution for 30 min. The mixture was dissolved by addition of 1 ml of 1 M  Table 2 Oligonucleotides used in this study a Dnr-binding sites are underlined; base exchanges are given in bold letters; restriction sites are given in italic letters

NO stress regulator DnrF of Dinoroseobacter shibae
Tris-HCl, pH 7.6 solution. Residual solid matter was sedimented by centrifugation (10 min at 13,000 ϫ g). The supernatant was filtered using a 0.2-m filter tip. Solution was measured at OD 385 nm , and molar concentration was determined using the Beer-Lambert law. Heme titration was performed under reducing conditions by using 16 nM purified apo-DnrF and stepwise addition of 2 M heme up to a final concentration of 28 M. Molar ratios were determined (Fig. 3A). Reduced derivative was obtained by addition of 2 mM sodium dithionite. The NO-bound derivative was achieved by injecting 0.5 ml of 1% gaseous NO. NO dilution occurred within a N 2 -saturated atmosphere. All steps were performed under strictly anaerobic conditions in a Coy anaerobic chamber (Coy, Grass Lake, MI).

Gel-permeation chromatography
The native molecular mass of DnrF was determined on an Äkta purifier system equipped with a Superdex TM 200 increase 10/300 GL column (GE Healthcare, Solingen, Germany). The column was equilibrated with 100 mM Tris-HCl, pH 7.5, 150 mM NaCl buffer at a flow rate of 0.5 ml per min. For calibration, gel filtration markers kit (M r of 12,400 (cytochrome c), 29,000 (carbonic anhydrase), 66,000 (albumin), 150,000 (alcohol dehydrogenase), and 200,000 (Amylase)) (Sigma) was used (supplemental Fig. S1) according to the manufacturer's instructions. An amount of 200 l of DnrF at a concentration of 160 M was injected. Elution was monitored at 280 nm for DnrF protein, at 417 nm for bound ferric heme (apo-DnrF), 427 nm for bound ferrous heme (holo-DnrF), and 398 nm for bound NO/heme in the case of NO/heme/DnrF.

EMSA
The extent of DNA binding of apo-DnrF, heme-reconstituted DnrF, and NO-bound DnrF was assessed using EMSA. Strand-specific promoter fragments with a central localization of the potential DnrF-binding site were synthesized (Metabion, Planegg, Germany). For analyses of the nosR1 (Dshi_0686) promoter, oligonucleotides nosR1 and nosR1-rev were used, spanning the 5Ј-region Ϫ32 to ϩ43 (Table 2). For analyses of the hemA3 (Dshi_3190) promoter, oligonucleotides hemA and hemA-rev were used, spanning the promoter region Ϫ101 to Ϫ26 (Table 2). For analyses of the dnrE promoter, oligonucleotides dnrE and dnrE-rev were used, spanning the promoter region Ϫ116 to Ϫ41 (Table 2). For analyses of the nosR2 promoter, oligonucleotides nosR2 and nosR2-rev were used, spanning the promoter region Ϫ78 to Ϫ3 (Table 2). For analyses of the norC promoter, oligonucleotides norC and norC-rev were used, spanning the promoter region Ϫ111 to Ϫ36 with respect to the translational start (Table 2). For analyses of the napD promoter, a fragment spanning from Ϫ135 to Ϫ60 resulting in oligonucleotides napD and napD-rev was used (Table 2). In addition to the native napD promoter fragment, a mutated napD promoter fragment was synthesized, exhibiting base exchanges at position Ϫ106/Ϫ105 from TT to GC and at position Ϫ91/Ϫ90 from AA to GC, respectively. For the resulting napDmu promoter fragment, the oligonucleotides napDmu and napDmu-rev were used (Table 2). For the annealing reaction, 10 pmol of the complementary oligonucleotides were heated for 5 min at 95°C in 10 mM Tris-HCl, pH 8, 1 mM EDTA, and 100 mM NaCl and then stepwise cooled down (5°C steps, every 10 min) to room temperature. The resulting doublestranded DNA fragments were DIG-labeled using the DIG gel shift kit, 2nd generation, following the manufacturer's instructions (Roche Applied Science, Basel, Switzerland). Unincorporated nucleotides were removed using the nucleotide removal kit from Qiagen (Hilden, Germany). Indicated amounts of DnrF (2,4,8,16,32,64,128, and 256 nM) were incubated with 4 nM DIG-labeled DNA fragments in 20 l of binding buffer (10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 5 mM DTT, 5% glycerol, 10 mM NaCl, 1 mM MgCl 2 , 0.1 mg/ml BSA). For competition experiments, the appropriate unlabeled DNA was added in amounts of 40 nM (10-fold excess), 200 nM (50-fold excess), and 400 nM (100-fold excess). Incubation was carried out for 20 min at room temperature. Subsequently, the mixture was loaded onto an 8% polyacrylamide gel and electrophoresed for 55 min in 1ϫ Tris borate/EDTA buffer at 100 V. Then, the gel was blotted for 2 h onto a positively charged nylon membrane, followed by immunodetection. For this purpose, the membrane was washed in 100 mM malic acid, 150 mM NaCl, 175 mM NaOH, and 0.3% Tween 20 and incubated for 40 min in 1ϫ blocking solution (Roche Applied Science). Immunodetection was carried out by the addition of 0.375 units of anti-digoxigenin-AP Fab fragments (Roche Applied Science) and incubated for 30 min. After two washing steps, the pH of the membrane was increased by adding detection buffer (100 mM Tris-HCl, pH 9.5, and 100 mM NaCl). The luminescence reaction was initiated by coating the membrane with 0.25 mM CDP-Star and incubated at 37°C for 15 min in darkness. Luminescence was measured using a high resolution camera (Photometrics, Cool SNAP HQ 2 ) in darkness. Quantitative analyses were made by using Gelscan Version 6.0 (BioSciTec, Frankfurt am Main, Germany).

Electron paramagnetic resonance spectroscopy
X-Band EPR spectra were recorded with a Bruker EMX spectrometer equipped with an Oxford ESR900 gas flow cryostat. Helium was used as cooling gas. The spectra were measured at the stated temperature and frequency. The samples are prepared under nitrogen atmosphere in a glove box and filled in thin wall quartz EPR sample tubes (inner diameter 4 mm) 707-SQ-250 M manufactured by Wilmad-LabGlass (Vineland, NJ).
Author contributions-M. E. provided substantial contributions to the conception and design of the work, data collection, data analysis and interpretation, drafting the article, critical revision of the article, and final approval of the version to be published. P. S. contributed to the conception and design of the work, data collection, data analysis, and interpretation. M. B. contributed to the conception and design of the work, data collection, data analysis, and interpretation. S. L. contributed to conception and design of the work, data collection, data analysis, and interpretation. E. H. provided substantial contributions to the conception and design of the work, data analysis and interpretation, drafting and critical revision to the article, and final approval of the version to be published. D. J. provided substantial contributions to the conception and design of the work, data analysis and interpretation, drafting and critical revision to the article, and final approval of the version to be published.