Mechanism of Oxygen Sensing by the Bacterial Transcription Factor Fumarate-Nitrate Reduction (FNR)*

The facultative anaerobe Escherichia coli adopts different metabolic modes in response to the availability of oxygen. The global transcriptional regulator FNR (fumarate-nitrate reduction) monitors the availability of oxygen in the environment. Binding as a homodimer to palindromic sequences of DNA, FNR carries a sensory domain, remote from the DNA binding helix-turn-helix motif, which responds to oxygen. The sensing mechanism involves the transformation of a [4Fe-4S]2+ cluster into a [2Fe-2S] form in vitro on reaction with oxygen. Evidence is presented to show that this process proceeds by at least two steps, the first, an oxidative one, being the formation, on reaction with O2, of a [3Fe-4S]1+ cluster as an intermediate accompanied by the production of hydrogen peroxide. This is followed by a slower, non-redox, pseudo-first order step in which the [3Fe-4S]1+ form converts to a [2Fe-2S]2+ cluster. This must be accompanied by a substantial protein conformational change since the four cysteine ligands that bind the two forms of the FeS clusters have different spatial disposition. Hydrogen peroxide is also an oxidant of the [4Fe-4S]2+, causing a similar cluster transformation to a [2Fe-2S] form. Either the hydrogen peroxide formed on reaction with oxygen can be recycled by intracellular catalase or it can be used to oxidize further Fe-S clusters. In both cases, the efficacy of oxygen sensing by FNR will be increased.

Escherichia coli is a facultative anaerobe that adopts different metabolic modes in response to the availability of oxygen (1). A hierarchy of metabolism exists in which aerobic respiration is preferred to anaerobic respiration, which in turn is preferred to fermentation (1). In simple terms, the global transcriptional regulator FNR 1 (designated due to defects in fumarate-nitrate reduction of the corresponding mutants), along with other transcription factors, ArcBA, NarXL, NarPQ, and FhlA, maintains this hierarchy by monitoring the availability of oxygen in the environment (1). FNR is a member of a large family of transcription factors that modulate physiological changes in response to a variety of metabolic and environmental challenges (2). Members of the family are predicted to be structurally related to the catabolite gene activator protein, CAP (also known as the cAMP receptor protein). CAP and FNR proteins bind as homodimers to palindromic sequences of DNA, each monomer binding to one half-site (2). They consist of two functionally distinct domains, a DNA binding helix-turn-helix motif and an N-terminal region of antiparallel ␤-strands forming the sensory domain (3). The sensing regions are adapted to respond to different effectors (2). Thus, CAP reversibly binds cAMP to monitor glucose status (4), and the sensing domain of the CooA protein of Rhodospirillum rubrum possesses a b-type cytochrome that binds CO (5). The sensory region of FNR has four conserved cysteine residues (Cys-20, -23, -29, and -122) that are essential for in vivo activity and capable of ligating an oxygen-sensitive [4Fe-4S] iron-sulfur cluster (6 -11). A variety of studies have revealed that the active form of FNR contains one [4Fe-4S] 2ϩ cluster/protein monomer that is converted to a [2Fe-2S] 2ϩ cluster, together with other, less well defined iron species, following exposure to oxygen both in vitro and in vivo (10,(12)(13)(14)(15). The switch from a cubane [4Fe-4S] cluster, bound to the protein by four cysteine thiol ligands that sit at the vertices of a tetrahedron, to a planar [2Fe-2S] cluster, also thought to possess four cysteine ligands but lying in a plane, suggests that cluster transformation on exposure to oxygen will provide a large conformational change in the N-terminal region of FNR, presumably thereby initiating the switch of the protein from a DNA binding state to one incapable of binding DNA (8). Initial attempts to characterize the process of cluster conversion revealed that FNR requires at least 2.5 molecules O 2 per cluster for complete cluster conversion (8). Prolonged exposure to O 2 can lead to complete loss of cluster and the formation of apo-FNR (17).
The overall reaction of FNR with O 2 can be written as Reaction 1. The redox states of the individual Fe atoms in the cluster are indicated.
[2Fe(III)2Fe(II)-4S] 2ϩ ϩ xO 2 3 y[2Fe(III)-2S] 2ϩ ϩ other products REACTION 1 Neither the product(s) of O 2 reduction nor the overall stoichiometry of the reaction is known. The chemical state of iron and sulfur released is also unknown. An EPR signal characteristic of a [3Fe(III)-4S] 1ϩ cluster (S ϭ 1 ⁄2) has occasionally been observed in FNR samples following brief exposure to O 2 but has never accounted for more than about 5% of the original [4Fe-4S] 2ϩ cluster in the wild type protein (18).
Here we investigate the transformation of the [4Fe-4S] 2ϩ cluster of FNR into the [2Fe-2S] form in vitro by the effect of oxygen. Evidence is presented to show that the process proceeds by at least two steps, the first, the oxidative one, being the formation on reaction with O 2 of a [3Fe-4S] 1ϩ cluster as an intermediate accompanied by the production of hydrogen peroxide. This is followed by a slower, non-redox, pseudo-first order, step with the [3Fe-4S] 1ϩ form converting to a [2Fe-2S] 2ϩ cluster presumably accompanied by a substantial protein conformational switch.
Protein purification and handling was carried out under strictly anaerobic conditions in a Faircrest anaerobic cabinet, typically operating at Ͻ2.0 ppm O 2 by volume, and all buffers were sparged with oxygen-free nitrogen gas for a minimum of 2 h. Bacterial extract (10 ml) was applied to buffer without protease inhibitor equilibrated GSH-agarose affinity columns (2 ml). Bovine thrombin (Sigma), incubated at ambient temperature (ϳ25°C) for 16 h, released apo-FNR, which was collected and stored at 4°C until required. Protein concentration was determined using the Bio-Rad protein reagent with bovine serum albumin as the standard. A correction factor of 0.83 was applied to FNR samples according to Ref. 18. Purity of the isolated FNR was checked by SDS-PAGE.
Spectroscopy and Oxidation of FNR-A Hitachi U3200 spectrophotometer, scanning at 120 nm min Ϫ1 , or a Jasco J-810 spectrophotometer, scanning at 200 nm min Ϫ1 , were used to measure absorbance or CD spectra of FNR samples in sealed anaerobic cuvettes (1 cm). Time dependence for a 1-ml sample of FNR, containing catalytic amounts of catalase (370 units ml Ϫ1 ), was determined after the sample was injected with an aliquot (4 l) of a 15.4 mM H 2 O 2 stock solution, equivalent to 30.7 M O 2 , and the decrease in A 420 was monitored continuously using a Hewlett Packard 8453 spectrophotometer fitted with a thermostatic cell holder set to 25°C.
The optical spectra of oxidized holo-FNR a 2-ml sample of FNR, containing catalase (176 units ml Ϫ1 ), was injected with 10 l of a 30.7 mM H 2 O 2 stock solution, giving a final concentration of 72.7 M O 2 , and then incubated for 15 min at an ambient temperature prior to measurement. Spin intensities of paramagnetic samples were estimated by integration of EPR spectra using 1 mM Cu(II), 10 mM EDTA as the standard. An X-band Bruker EMX EPR spectrometer was equipped with a TE-102 microwave cavity and an ESR-900 helium flow cryostat (Oxford Instruments). The microwave frequency was monitored using a Marconi Instruments microwave counter, model 2330.
Determination of Dissolved Oxygen-Oxygen concentrations were determined in all buffer solutions at 15, 20, and 25°C by chemical analysis using the method of Winkler (22). Standardization of Hydrogen Peroxide Solution-A stock solution of ϳ20 mM H 2 O 2 was prepared from a commercially available 30% (v/v) H 2 O 2 solution, by dilution with purification buffer, and maintained on ice. Oxygen-free nitrogen gas was slowly bubbled through an aliquot (25 ml) of the stock solution, which was then treated with an aliquot (100 ml) of 1 M H 2 SO 4 . The addition of 10 ml of 10% (w/v) NaI in 1 M H 2 SO 4 followed by five drops of a 3% (w/v) Mo 7 O 24 ⅐4H 2 O catalyst solution liberated iodine, which was immediately titrated with a 0.1 M solution of Na 2 S 2 O 3 ⅐5H 2 O, using 1% (w/v) starch solution as indicator (23). Stock solutions (21.6 Ϯ 2.16 mM H 2 O 2 ). were made freshly, calibrated, maintained at Յ4°C, and used on the same day.

Stoichiometry of Oxygen Reaction with FNR-
The optical spectrum of FNR in the absence of O 2 displays absorbance maxima at 320 and 405 nm, with values of ⑀ 320 17,443 M Ϫ1 cm Ϫ1 and ⑀ 405 13,559 M Ϫ1 cm Ϫ1 , respectively, together with the broad shoulder at 420 nm, giving the sample a characteristic straw brown color (Fig. 1A)  Taking the initial slope and the asymptote at high oxygen levels, a binding stoichiometry of 0.58 Ϯ 0.04 O 2 /[4Fe-4S] cluster is obtained. A complete binding curve cannot validly be fitted since the reaction turns out to be complex (see "Intermediates in the Oxidation of FNR by Oxygen"). However, the initial slope should be a reliable indicator of the stoichiometry at low oxygen levels.
The reaction of O 2 with FNR was also monitored using CD spectroscopy (Fig. 3A). Iron-sulfur clusters gain optical activity from the fold of a polypeptide chain. In the absence of O 2 , the FNR CD spectrum displayed weak bands in the region 280 -800 nm with three positive features at max 330, 380, and 420 nm. The ⌬⑀ values were of the same order of magnitude as those of other proteins containing [4Fe-4S] cluster types (25,26 and ϳ84% complete at a ratio of 1.0, in agreement with the observations made by optical spectroscopy (see above). The CD spectra of FNR, the first to be reported, provide a useful means of monitoring the status of the FNR iron-sulfur cluster.
Product of Oxygen Reduction by FNR-The maximum number of electrons available from complete oxidation of one [4Fe-4S] 2ϩ cluster to two Fe(III) ions is two, suggesting that hydrogen peroxide may be a product. To test this, an anaerobic sample of FNR was titrated with O 2 in the presence of Amplex Red (27) and horseradish peroxidase (HRP). Peroxide reacts with Amplex Red and HRP to generate the highly fluorescent dye Resorufin (27). The addition of aliquots of O 2 caused a progressive increase in fluorescence intensity at 587 nm (   has also been followed by CD spectroscopy (Fig. 3B) Intermediates in the Oxidation of FNR by Oxygen-The time course of the reaction of FNR with O 2 (generated by the reaction of catalase with H 2 O 2 ) was monitored at 420 nm, revealing a 50% decrease in 1 min, 82% decrease after 5 min, and 98% decrease after 10 min (Fig. 6) (18,28). During the course of an O 2 titration, samples of FNR were taken within ϳ1 min after O 2 addition, placed in an EPR tube, and frozen rapidly to 77 K. The EPR spectra of these samples, measured at 10 K, display a signal centered at g Ϸ 2.014, characteristic of a [3Fe-4S] 1ϩ cluster, which decreased rapidly in intensity on increasing the temperature until it was lost at ϳ30 K (Fig. 7A). Integration of the signal showed an initial EPR signal intensity, which declined steeply during the course of the titration (Fig. 7B) 2ϩ cluster upon exposure to O 2 has been observed in proteins that generate a radical species by reductive cleavage of S-adenosyl methionine such as biotin and lipoate synthase (29). This appears to be a very large family with over 600 members (30 We have demonstrated the generation of H 2 O 2 following reaction of FNR with O 2 with the amount of peroxide produced being dependent on the amount of O 2 added. However, quantification of H 2 O 2 yielded only ϳ45% of that expected for oxidation of one [4Fe-4S] 2ϩ cluster. This suggests either that the reaction with O 2 produces products in addition to H 2 O 2 , or more likely, that the H 2 O 2 generated can oxidize unreacted [4Fe-4S] 2ϩ clusters. Evidence to support this latter possibility has been obtained. This could account for the observation that more than one iron-sulfur cluster can be disassembled by one O 2 (see below). If the overall oxidation process of the FNR by O 2 is written as Reaction 2, then the stoichiometry of oxygen to [4Fe-4S] is not as determined.   REACTION 4 The first in a two-step process is oxidative, involving a twoelectron oxidation of the [4Fe-4S] 2ϩ cluster, leading to a [3Fe(III)-4S] 1ϩ cluster and the loss of one Fe 3ϩ . No inorganic sulfur would be lost at this stage. The product of this reaction is H 2 O 2 (Reaction 3). This reaction is too fast to follow accurately by conventional spectrophotometry, although at the shortest time intervals, a fast phase was observed (Fig. 6). The second slower, rate-determining step (Reaction 4) occurs, in which sulfide ions and further Fe 3ϩ are lost. These two steps lead to the overall reaction scheme given by Reaction 2, in accord with the kinetic observations provided that k 1 Ͼ Ͼ k 2 .
Reaction of FNR with Hydrogen Peroxide-The major product of oxygen reduction by FNR, hydrogen peroxide, will itself oxidize the [4Fe-4S] 2ϩ cluster in FNR. The CD spectrum suggests the product of the reaction with H 2 O 2 to be a [2Fe-2S] 2ϩ cluster, although further spectroscopic characterization is needed. The reaction scheme, Reaction 5, is also a two-electron process but with water as the product. REACTION 5 A cooperative interaction between a pair of iron sulfur clusters in the dimeric form of FNR would lead to the four-electron reduction of oxygen to water. If the H 2 O 2 produced from oxidation of one cluster by O 2 were subsequently to react with the partner [4Fe-4S] cluster in a co-operative reaction, this would result in the reduction of one O 2 molecule to two H 2 O molecules by a pair of [4Fe-4S] clusters of dimeric FNR (Reaction 6)  (38) using a function that fitted the amino acid sequence of FNR to that of the CooA monomer followed by energy minimization and further fitting to minimize the root mean square (between 1 and 2 Å) (16, 34,36). The 2Fe cluster was fitted using the [2Fe-2S] cluster of the oxidized form of ferredoxin from Chlorella fusca (1AWD.pdb) (39). cysteine residues to form a [2Fe-2S] cluster. This results in an inability to bind DNA and the switching on of transcription. By tuning the redox potential of the FNR [4Fe-4S] 2ϩ/3ϩ cluster, the range of oxygen levels sensed could be altered to fit the purpose. Note that two sequential one-electron oxidations of the cluster [4Fe-4S] 2ϩ would not occur at the same potential, a higher oxidizing potential being required to remove the second electron. However, by loss of an Fe(III) ion from the [4Fe-4S] 3ϩ state to form [3Fe-4S] 1ϩ , a concerted two-electron loss can take place at a single potential.
Although the crystal structure of FNR has not been determined, it is possible to model its structure and response to cluster interconversion now that structures are available for the analogues of FNR, CAP, and CooA, the CO-sensing protein of R. rubrum. The structure of CooA shows that, in the CO-off form, the DNA binding domain of the protein has swung almost through 180 degrees, lengthening the long helix that forms the dimer interface (5). We have generated structures using CAP as a template for the DNA binding state via the Swiss model and the structure of CooA as a template for the form that fails to bind DNA. Fig. 8 shows our results. In the DNA binding state, the DNA recognition helices lie exposed at the top of the structure, and this domain makes contact via a two-turn helix with the ␤-sheet domain that binds the [4Fe-4S] cluster at its lower extremity. The conformational change from a [4Fe-4S] cluster to the [2Fe-2S] form, induced by oxygen, breaks the interface between the upper and lower domains, and the long central helix, which lies at the dimer interface, lengthens ,causing the DNA binding and recognition helices to be swung toward the middle of the structure. The Fe-S cluster binding sites have been modeled on the assumption that both the 4Fe and the 2Fe clusters bind the four conserved cysteine residues. The modeling predicts considerable rearrangement around the clusters, as dictated by the differing steric demands of each cluster.
Finally, we note that the [4Fe-4S] 2ϩ cluster is not chemically specific toward oxidants. This work has shown that oxygen and hydrogen peroxide are both oxidants that lead to a [2Fe-2S] cluster, whereas our earlier studies of the reaction of FNR with nitric oxide (35) revealed nitrosylation of the [4Fe-4S] cluster, apparently yielding several iron-nitrosyls and loss of iron and sulfur (35). All three processes are triggers of transcription. This is somewhat surprising but may imply that effective oxidants other than oxygen are maintained at low intracellular levels in order that the oxygen switch prevails.