Regulation of dinitrogenase reductase ADP-ribosyltransferase and dinitrogenase reductase-activating glycohydrolase by a redox-dependent conformational change of nitrogenase Fe protein.

The nitrogenase-regulating enzymes dinitrogenase reductase ADP-ribosyltransferase (DRAT) and dinitrogenase reductase-activating glycohydrolase (DRAG), from Rhodospirillum rubrum, were shown to be sensitive to the redox status of the [Fe(4)S(4)](1+/2+) cluster of nitrogenase Fe protein from R. rubrum or Azotobacter vinelandii. DRAG had <2% activity with oxidized R. rubrum Fe protein relative to activity with reduced Fe protein. The activity of DRAG with oxygen-denatured Fe protein or a low molecular weight substrate, N(alpha)-dansyl-N(omega)-(1,N(6)-etheno-ADP-ribosyl)-arginine methyl ester, was independent of redox potential. The redox midpoint potential of DRAG activation of Fe protein was -430 mV versus standard hydrogen electrode, coinciding with the midpoint potential of the [Fe(4)S(4)] cluster from R. rubrum Fe protein. DRAT was found to have a specificity opposite that of DRAG, exhibiting low (<20%) activity with 87% reduced R. rubrum Fe protein relative to activity with fully oxidized Fe protein. A mutant of R. rubrum in which the rate of oxidation of Fe protein was substantially decreased had a markedly slower rate of ADP-ribosylation in vivo in response to 10 mM NH(4)Cl or darkness stimulus. It is concluded that the redox state of Fe protein plays a significant role in regulation of the activities of DRAT and DRAG in vivo.

Nitrogen fixation requires the transfer of single electrons through a series of redox-active metalloproteins. The oxidation states and reduction potentials of these metalloproteins have been the subjects of extensive research in the nitrogenase field (1). The nitrogenase enzyme itself is composed of two highly conserved iron-sulfur proteins, designated Fe protein 1 and MoFe protein. Fe protein is a 64-kDa homodimer, containing a single [Fe 4 S 4 ] cluster, bound between the subunits by symmetrical cysteinyl ligation (2,3). MoFe protein is a 230-kDa ␣ 2 ␤ 2 tetramer, containing two [Fe 8 S 7 ] clusters (P-clusters) located at the interface of each ␣␤ dimer and two molybdenum-iron-sulfur-homocitrate clusters (iron-molybdenum cofactor), the presumed sites of substrate reduction (4 -6). For any of the substrates of nitrogenase, including dinitrogen, acetylene, and protons (7,8), electron transfer is coupled with the hydrolysis of at least two MgATP molecules per electron (9).
Although the details of the nitrogenase catalysis model continue to be debated, there is general agreement on the basic sequence of the cycle, based on data obtained with both the Klebsiella pneumoniae system and the Azotobacter vinelandii system. Fe protein, containing the reduced (1ϩ) form of the [Fe 4 S 4 ] cluster, binds two MgATP molecules (10). The binding of this nucleotide shifts the redox potential of the [Fe 4 S 4 ] 2ϩ/1ϩ couple from about Ϫ300 mV to nearly Ϫ450 mV (11,12) and also elicits a specific conformational change in Fe protein (13), promoting its association with MoFe protein. An additional conformational change apparently occurs upon complex formation, shifting the Fe protein redox potential by an additional Ϫ200 mV (14) and allowing the transfer of a single electron from the Fe protein to a iron-molybdenum cofactor of MoFe protein, via a P-cluster (15,16). Coupled to electron transfer is the hydrolysis of the two MgATP molecules bound to Fe protein to MgADP and P i . ATP hydrolysis apparently serves to prevent the reverse flow of electrons, rather than to drive electron transfer (17), as single electron transfer without MgATP hydrolysis has been observed in a tight-binding complex of MoFe protein with a variant of Fe protein from an A. vinelandii mutant (18). Oxidized ([Fe 4 S 4 ] 2ϩ ) Fe protein, stabilized by the two MgADP molecules still bound (19), dissociates from oneelectron-reduced MoFe protein in what is generally thought to be the rate-limiting step of the cycle (20). To complete the cycle, Fe protein is reduced by a low potential electron donor (a ferredoxin or flavodoxin in vivo) and the MgADP molecules are exchanged for two MgATP molecules. The catalytic cycle is repeated until a sufficient number of electrons have been transferred to completely reduce the FeMoco-bound substrate (21). As can be seen in the catalysis description, the multifunctional Fe protein (which also has roles in MoFe protein metallocluster synthesis (22,23)) exists in several distinct redox and conformational states during a single enzyme turnover event.
In the nitrogen-fixing, photosynthetic, facultative anaerobe Rhodospirillum rubrum, nitrogenase can be rapidly, reversibly inactivated by mono-ADP-ribosylation of nitrogenase Fe protein (24). This modification event occurs in response to environmental conditions of fixed nitrogen sufficiency or of dark-ness, when the energy-expensive nitrogenase reaction is undesirable (24). The ADP-ribosylation of a specific arginine residue (Arg-101 in R. rubrum Fe protein) is catalyzed by dinitrogenase reductase ADP-ribosyltransferase (DRAT), using NAD ϩ as the ADP-ribose donor (25,26). The ADP-ribosylated Fe protein cannot reduce MoFe protein, although the [Fe 4 S 4 ] cluster is still accessible to small redox-active molecules. DRAT acts as a 30-kDa monomer with high specificity toward MgADP-bound Fe protein, possessing no measurable activity with other arginine residues or water as the ADP-ribose acceptor (27,28). Surprisingly, the Fe proteins from K. pneumoniae and A. vinelandii (which lack the dra operon) are better substrates for R. rubrum DRAT than the R. rubrum Fe protein itself. There are no measurable reverse or glycohydrolytic reactions catalyzed by DRAT. The removal of the ADP-ribose group is instead catalyzed by dinitrogenase reductase-activating glycohydrolase (DRAG), which restores fully active Fe protein with an intact Arg-101 side chain. DRAG is a 32-kDa monomeric binuclear manganese enzyme that is capable of cleaving the ␣-N-glycosidic bond of a number of analogs of ADP-ribosylarginine (29,30). However, only the MgATP-bound form (not the MgADP-bound or nucleotide-free forms) of ADPribosylated Fe protein is a substrate for DRAG (31). Although the exact modes of interaction of DRAT and DRAG with Fe protein are unknown, it is believed that each binds the same surface of Fe protein as does MoFe protein, as evidenced by the inhibition of cellular nitrogenase activity by overexpressed DRAT (32).
Although the means by which DRAT and DRAG are each regulated are not well understood, it is known that the activity of each enzyme is regulated in vivo (33,34). As the in vivo activation and inactivation rates are reflected by in vitro assay rates using purified components, it is believed that the regulatory signals involve either negative effectors or known assay components. As noted above, DRAT and DRAG have opposite specificities for MgADP-and MgATP-bound Fe protein. However, cellular fluctuations in ATP and ADP levels during inactivation/activation cycles are insufficient to account for the dramatic nitrogenase activity regulation (35). The cellular NAD ϩ concentration has also been suggested as a possible positive effector for DRAT (36,37). Similarly, studies proposing regulation of DRAG via membrane sequestration or activation have not yielded persuasive data (38,39). In this report, we demonstrate that the redox state of Fe protein controls the activity of DRAT and DRAG. The alteration of in vivo DRAT activity in a R. rubrum mutant in which Fe protein is trapped in the reduced state is shown. We also discuss the ramifications of these findings on the understanding of the ADP-ribosylation and nitrogenase catalytic models.
Protein Purification-Nitrogenase MoFe protein was purified from UR2 (wild-type R. rubrum) cells (broken by the French press method) as described (43). ADP-ribosylated Fe protein was purified from UR2 cells by a modification of the described method (44), in which a G-75 Sephadex gel filtration column (2.5 ϫ 70 cm; Amersham Pharmacia Biotech) was used instead of the preparative gel step. Unmodified (active) Fe protein was purified from UR212 (DRAT Ϫ ) cells with an identical protocol. DRAG was purified from UR276 (draG overexpresser) cells as described (39). DRAT was purified as described (44), from UR356 (draTGB overexpresser) cells broken by the French press method.
Oxidation of Fe Protein-Fe protein, containing 1.7 mM sodium dithionite (Na 2 S 2 O 4 ; Fluka) as isolated, was oxidized by indigo carmine (Sigma) bound to a Dowex chloride anion exchange column (10). In a VAC Atmospheres anaerobic glove box, an AG1-X8 Dowex chloride column (2 ml; Bio-Rad) was carefully poured on top of a small layer of G-25 Sephadex resin (0.2 ml; Amersham Pharmacia Biotech). Indigo carmine (1 ml of 20 mM) was bound to the AG1-X8 resin. The column was equilibrated with 20 volumes of anaerobic 50 mM MOPS, pH 7.8. Fe protein (1-5 mg; 3-10 mg/ml) was passed over the column, eluting in the equilibration buffer. The white color of the G-25 layer of the column allowed visualization of the brown Fe protein band eluting from the column. Indigo carmine-oxidized Fe protein was found to be Ͼ95% active.
Decoupled DRAG Assays-DRAG activity was determined by a modification of the nitrogenase-coupled DRAG assay (29), in which the removal of ADP-ribose was differentiated from the acetylene reduction reaction. The standard demodification step was conducted in an anaerobic, stoppered vial containing 5 mM ATP, 25 mM phosphocreatine, 50 g of creatine phosphokinase (Sigma), 25 mM MgCl 2 , 0.5 mM MnCl 2 , 10 mM Na 2 S 2 O 4 , 20 -40 g of ADP-ribosylated Fe protein, 350 ng of DRAG, and 50 mM MOPS, pH 7.8, in a 500-l total volume under nitrogen headspace. Na 2 S 2 O 4 or ATP were excluded from some assays. After a 10-min incubation at 30°C, 2 mM ADP-ribose (Sigma) was added to inhibit further DRAG activity. The acetylene reduction step was initiated by the addition of any components excluded from the demodification step, along with 75 g of MoFe protein and 10% acetylene in the headspace. After a 10-min incubation at 30°C, the reaction was stopped by the addition of 5% trichloroacetic acid. The headspace was analyzed for C 2 H 4 content by gas chromatography.
DRAG Assays by SDS-Polyacrylamide Gel Electrophoresis Analysis-Reactions were conducted in 50-l volumes in microcentrifuge tubes inside anaerobic, N 2 -filled vials. Inside the vials, but outside the reaction tubes, 0.5 ml of 100 mM Na 2 S 2 O 4 was present to act as an oxygen scavenger (26). Component concentrations were identical to those in the decoupled DRAG assays described above, with the following exceptions: creatine phosphokinase (2.5 g), ADP-ribosylated Fe protein (35 g), and DRAG (200 ng). The reactions were incubated 10 min at 30°C and then were stopped by the addition of 50 l of SDS sample buffer. A sample (20 l) of each stopped assay was loaded onto a 10% acrylamide (0.6% bis-acrylamide) SDS minigel. After electrophoresis, each gel was stained with Coomassie Blue R-250 stain (Sigma). The relative amounts of the Fe protein upper band (ADP-ribosylated) and lower band (unmodified) were quantitated by ImageQuant software (Molecular Dynamics). A standard correction was made in each assay for a contaminant of creatine phosphokinase, which comigrated with the upper band of Fe protein. Note that only one subunit of the Fe protein homodimer becomes ADP-ribosylated, and thus a 1:1 ratio of upper to lower band represents completely modified (inactive) Fe protein dimer.
Fluorometric Assay for DRAG Activity-A continuous assay for DRAG cleavage of the N-glycosidic bond of ⑀-ADPR-DAME has been described (45). Assays were performed in 5-mm (outer diameter) ϫ 3.5-cm sealed quartz tubes. Reaction mixtures contained 50 mM MOPS, pH 7.8, 1 mM MnCl 2 , 50 M ⑀-ADPR-DAME, and 150 ng of purified DRAG in a total volume of 250 l. Reactions were conducted anaerobically in the presence or absence of 0.5 mM Na 2 S 2 O 4 . The reactions were irradiated with 304 nm UV light, and the fluorescence emission at 405 nm was monitored. Reactions were initiated by the addition of DRAG. After 3.5 h, 1 g of phosphodiesterase (Sigma) and 0.5 mM MgCl 2 were added to completely cleave all ADP-ribosylarginine bonds. Finally, each reaction was exposed to air for 20 min, to react away dithionite, before obtaining the final fluorescence intensities.
DRAG Reaction Redox Dependence Determination-The extent of DRAG activity toward ADP-ribosylated Fe protein was determined in reactions poised at varied redox potentials. In an anaerobic glove box, DRAG decoupled assays (500 l total volume) were set up, as described above, containing 1 mM benzyl viologen (Sigma). High concentrations of indigo carmine-oxidized ADP-ribosylated Fe protein (100 -300 g/reaction) were required in these assays. Reactions were poised by the addition of Na 2 S 2 O 4 , and the redox potentials were measured by direct potentiometry, using a saturated Ag/AgCl 2 electrode (E m°Ј ϭ ϩ199 mV versus standard hydrogen electrode (SHE)) as a reference. Reactions were initiated by the addition of 150 ng of DRAG. After a 10-min incubation, the redox potentials were again measured, and the reactions were stopped by the addition of 2 mM ADP-ribose. Because benzyl viologen inhibits nitrogenase reactions at concentrations greater than 0.1 mM, portions of the stopped assays (50 l) were diluted 10-fold into acetylene reduction reactions. The acetylene reduction assays con-tained 5 mM ATP, 25 mM phosphocreatine, 50 g of creatine phosphokinase, 10 mM MgCl 2 , 2 mM ADP-ribose, 10 mM Na 2 S 2 O 4 , 75 g of MoFe protein, and 50 mM MOPS, pH 7.8, under a headspace of 10% acetylene in nitrogen. The reactions were incubated for 10 min at 30°C and then were stopped by the addition of 5% trichloroacetic acid. The headspace was analyzed for C 2 H 4 content by gas chromatography.
R. rubrum Fe Protein Midpoint Potential Determination-The midpoint potentials for modified and unmodified R. rubrum Fe protein were determined by dye-mediated, visible spectrum-monitored redox titration. Two solutions in matched glass cuvettes were prepared, containing 50 M Fe protein, 50 mM MOPS, pH 7.8, 3 mM MgATP, and 15 M methyl viologen (MV) (Sigma). The spectral differences between the cuvettes were recorded as 2-l aliquots of 2 mM Na 2 S 2 O 4 were added to one cuvette. The redox potential of the titrated solution was determined from the intensity of the absorption at 608 nm due to reduced MV. The fraction of Fe protein in the reduced ([Fe 4 S 4 ] 1ϩ ) state was determined from the change in the characteristic Fe protein absorbance at 420 nm (46,47).
DRAT Assays under Reducing or Oxidizing Conditions-An assay for DRAT activity by 32 P-NAD radiolabeling of Fe protein has been described (28). The standard reaction mix contains 50 mM MOPS, pH 7.8, 1 mM ADP, 5 mM MgCl 2 , 0.25 mM [␣-32 P]NAD (20 Ci/mol), 80 g of indigo carmine-oxidized, unmodified Fe protein, and DRAT in a volume of 50 l. To this mix were added 10 M MV and 4 g of carbon monoxide dehydrogenase (CODH) from R. rubrum. To produce reducing conditions, the headspace of the reaction vial was filled with CO, which was oxidized to CO 2 by CODH with a concomitant reduction of MV. Assay mixtures were incubated for 20 min at 30°C and then were stopped with 5% trichloroacetic acid. Precipitated protein was collected on nitrocellulose filters, and the amount of [␣-32 P]ADP-ribose incorporated into Fe protein was determined by a Packard Tri-Carb 2100TR liquid scintillation counter.
ADP-ribosylation Analysis in R. rubrum Whole Cells-R. rubrum strains UR2 (wild-type) and UR145 (nifD::kan (48)) were grown on supplemented malate-ammonium medium (40). Each strain was inoculated into malate-glutamate medium (33) at a 65-fold dilution. After illuminated, anaerobic growth for 2 days, for derepression of Fe protein, the cells were treated with darkness or 10 mM NH 4 Cl, as described (33). Proteins were extracted quickly by a trichloroacetic acid precipitation method (49,50). Fe protein subunits were separated by electrophoresis on a 10% acrylamide (0.6% bis-acrylamide) SDS minigel. Proteins from SDS-polyacrylamide gel electrophoresis were transferred onto nitrocellulose and then were immunoblotted with polyclonal antibody against A. vinelandii Fe protein and were visualized with ECL Western blotting reagents (Amersham Pharmacia Biotech). The protein bands on the x-ray film were quantitated by ImageQuant software.

Requirement for Reductant in DRAG Reactions with Fe Protein-
The activation of ADP-ribosylated Fe protein by DRAG was assayed with oxidized Fe protein or reduced Fe protein by the "decoupled DRAG assay" described under "Materials and Methods." In both cases, Fe protein was first oxidized as described under "Materials and Methods." To obtain reduced Fe protein, sodium dithionite was added to the incubation mixture. In the second phase of the decoupled DRAG assay, the extent of activation of Fe protein was determined from acetylene reduction assays conducted in the presence of MoFe protein, excess dithionite, and 10% acetylene in the headspace. In these decoupled assays for DRAG activity, with indigo carmineoxidized Fe protein from R. rubrum as the substrate, activity in the absence of Na 2 S 2 O 4 was found to be negligible in comparison to activity with the reductant added (Table I). This result was validated by demonstration of the efficacy of the assay system. In the decoupled assays, the addition of 2 mM ADPribose completely inhibited DRAG activity but only slightly inhibited the acetylene reduction reaction of Fe protein and MoFe protein (data not shown). The addition of this specific inhibitor stopped DRAG activation of Fe protein, effectively separating the DRAG-dependent activation of Fe protein from the Fe protein-dependent acetylene reduction phase of the assay and thus allowing analysis of the DRAG reaction in isolation. The decoupled assay system was demonstrated to be a linear assay for DRAG. The dependence of the DRAG reaction on the presence of MgATP, first reported by Saari et al. (31), was confirmed (Table I). Also, the dependence on reductant was shown to exist regardless of the divalent cation (Mn 2ϩ or Fe 2ϩ ) present in the assay (data not shown).
The stimulation of DRAG activity by reductant was correlated to a change in Fe protein subunit composition, indicative of the removal of ADP-ribose from Fe protein (51,52). In SDS-polyacrylamide gel electrophoresis, ADP-ribosylated subunits of Fe protein run anomalously slowly, allowing differentiation from unmodified subunits. The conversion of upper (ADP-ribosylated) to lower (unmodified) band was monitored by laser densitometry of Coomassie Blue-stained gels. These demodification assays, described under "Materials and Methods," were conducted under conditions similar to those of the decoupled assay but did not require the addition of acetylene reduction reaction components. In the presence of a complete reaction mix, including sodium dithionite, nearly half of the Fe protein was converted from the modified to the unmodified band in a 10 min assay (Fig. 1A). This conversion represents roughly 90% of the total possible subunit conversion, as ϳ50% of the N-glycosidic bonds liking ADP-ribose to Fe protein are in the ␤-configuration and thus are unreactive toward DRAG (52). When the system was oxidized by the addition of indigo carmine, little conversion to the lower band was observed. The absence of MgATP also prevented subunit conversion. Protein band densities were corrected for background in the absence of Fe protein (Fig. 1A, lane 1). The demodification assay was shown to be inhibited by 200 mM NaCl and 1 mM ADP-ribose by 59 and 61%, respectively, in close agreement to values reported by Saari et al. (31) for the DRAG radioassay.
Identification of the Redox-sensitive Species in DRAG Reactions-The studies above indicated that low redox potential was required for DRAG activity, but it was unknown whether low potential was required for reduction of DRAG, for reduction of Fe protein, or for the reaction chemistry. This question was answered by investigating reactions of DRAG with alternative substrates. The first such substrate was reduced Fe protein, from which excess Na 2 S 2 O 4 was removed by desalting on a gel filtration column (Table I). The reduced Fe protein was found to be activated to a similar extent as oxidized Fe protein when sodium dithionite was present (fully reducing Fe protein in each case). However, in the absence of additional reductant,  (46). Previously, the activity of DRAG with denatured Fe protein was shown to be similar to that with native Fe protein, but with no requirement for MgATP (31). All previously published data have corroborated the model of ADP-ribosylated, O 2 -denatured Fe protein as an effector-insensitive peptide. In demodification assays identical to those tested for native Fe protein, DRAG was shown to have full activity toward denatured Fe protein in the presence or absence of dithionite and in the presence or absence of MgATP (Fig. 1B). Reactions with denatured Fe protein were shown to be insensitive to NaCl but inhibited by ADP-ribose, verifying published data (31). It was noted that the presence of a fraction of Fe protein in the denatured form may have been responsible for the small activ-ities seen in reactions of DRAG with native Fe protein in the absence of MgATP or Na 2 S 2 O 4 (Fig. 1A, lanes 4 -6).
A third substrate of DRAG assayed for redox potential effect was ⑀-ADPR-DAME. With this fluorescent substrate, the DRAG reaction has a V max similar to that with ADP-ribosylated Fe protein (30). Cleavage of the ADP-ribosylarginine bond of ⑀-ADPR-DAME by DRAG was monitored by the intensity of fluorescence of the unquenched etheno-adenine group at 405 nm. The reaction curves in the presence or absence of 0.5 mM sodium dithionite are very similar (Fig. 2). After 3.5 h of reaction with DRAG, fluorescence equaled 41 and 36% (with and without dithionite, respectively) of the maximal fluorescence observed after treatment with phosphodiesterase. These values are consistent with the 40% value observed by Pope et al. (45) in the presence of dithionite. The difference in total fluorescence intensities was attributed to absorption of the excitation energy at 304 nm by dithionite ( max ϭ 310 nm). After oxidation by air, the final fluorescence intensities were identical (Fig. 2). Along with the denatured Fe protein data, the insensitivity of the fluorometric assay toward dithionite indicated that the requirement of DRAG for low redox potential occurs in response to a change of native Fe protein.
To verify the hypothesis that DRAG was sensitive to the redox state of the [Fe 4 S 4 ] cluster of Fe protein, the midpoint potential of Fe protein and the redox dependence of the activation reaction with DRAG were determined. The midpoint potential of the [Fe 4 S 4 ] cluster of Fe protein in the presence of 3 mM MgATP was determined by visible spectroscopy, using the A 608 of the redox buffer, methyl viologen (E m°Ј ϭ Ϫ446 mV versus SHE), as a standard. The resulting data predict a midpoint potential of Ϫ470 Ϯ 1 mV (versus SHE) and n ϭ 0.73 Ϯ 0.03. The Fe protein midpoint potential was unaffected by ADP-ribosylation (inactive Fe protein E m ϭ Ϫ471 Ϯ 1 mV, n ϭ 0.80 Ϯ 0.02). For the DRAG activation reactions, a variation of the decoupled DRAG assay was performed, in which the reactions with DRAG were poised at directly measured redox potentials in the presence of a redox dye (benzyl viologen, E m°Ј ϭ Ϫ325 mV versus SHE). After the DRAG reaction was stopped, a fraction of each reaction was injected into a standard nitrogenase assay. The activation of Fe protein by DRAG follows the redox state of the Fe 4 S 4 cluster of Fe protein (Fig. 3). Given the relative proximity of the two redox-dependent phenomena, it Fe protein was exposed to air for 30 min and then was made anaerobic by degassing and flushing with N 2 , before addition to reaction. Lanes 1 and 2 are controls as described for A. Lane 3 represents the reaction with DRAG in the presence of Na 2 S 2 O 4 and MgATP. In lanes 4 and 6, Na 2 S 2 O 4 was excluded. In lanes 5 and 6, reactions excluded MgATP, phosphocreatine, and creatine phosphokinase.

FIG. 2.
Reaction of DRAG with ⑀-ADPR-DAME. Fluorescence intensity of etheno-adenine moiety released from the DAME group by DRAG was monitored in the absence (trace 1) or presence (trace 2) of 0.5 mM Na 2 S 2 O 4 . Treatment with phosphodiesterase released all ⑀-adenine. After O 2 treatment, total fluorescence in each reaction cuvette was determined.
seems likely that the change of Fe protein recognized by DRAG is, indeed, the change in redox state of the [Fe 4 S 4 ] cluster. Additionally, A. vinelandii Fe protein that was 100% ADPribosylated by extensive reaction with DRAT was tested for reaction with DRAG by the DRAG activation protocol. The midpoint potential for this reaction was ϳϪ440 mV, which is in close agreement with the published value of Ϫ430 mV for the midpoint potential of nucleotide-bound A. vinelandii Fe protein (11,12).
DRAT Reactions under Oxidizing or Reducing Conditions-Previous studies on the reaction of DRAT with unmodified Fe protein have noted the deleterious effect of Na 2 S 2 O 4 (26), which apparently reduces NAD ϩ to NADH; NADH does not serve as an ADP-ribose donor in the DRAT-catalyzed modification of Fe protein. Therefore, another reductant had to be found for examination of DRAT reactions under reducing conditions. A coupled assay was developed in which CO, the ultimate source of reducing power, is oxidized to CO 2 by CODH, in the process reducing the redox dye MV. Thus MV concentrations could be kept low to prevent inhibition of DRAT, while maintaining a constant source of MV as a low potential reductant. Under the reaction conditions, the redox states of NAD ϩ or NADH are unaffected by the presence or absence of the reductant CO (Fig.  4A). Oxidized Fe protein is unaffected by the presence of oxidized CODH and MV (Fig. 4B) but is fully reduced when CO is added to the system. The ratio of A 430 for Fe protein (oxidizing conditions versus reducing conditions) is 1.7, which is identical to the ratio (1.7) reported by Ashby and Thorneley (10) for A. vinelandii Fe protein (87% reduced versus fully oxidized).
Using the CO/CODH/MV system as a source of reductant, the activity of DRAT with oxidized or reduced Fe protein from R. rubrum or A. vinelandii was examined by 32 P-ADP-ribosylation assay (Table II) or reversible inactivation of Fe protein (Table III). In all cases, the DRAT reaction with oxidized Fe protein was superior to the reaction with reduced substrate. In the reaction of DRAT with reduced R. rubrum Fe protein (Table  II), significant ADP-ribosylation activity (ϳ15% of maximum) was observed. It is possible that some label was deposited on CODH, as labeling is diminished in the absence of CODH. Also, there was no reversible inactivation of R. rubrum Fe protein under reducing conditions (Table III), suggesting that DRAT activity with reduced Fe protein is minimal. Under oxidizing conditions, the DRAT reaction characteristics reflected published data for the activity of DRAT (27,28), with similar extent of reaction under the given reaction conditions and a K m for NAD ϩ of about 2 mM (data not shown). Reactions with A. vinelandii Fe protein achieved much more labeling and inactivation than with R. rubrum Fe protein, as noted previously FIG. 3. Redox potential curves for Fe protein Fe 4 S 4 cluster reduction and activation by DRAG. DRAG activities were determined at the indicated redox potentials by decoupled DRAG assays. The DRAG activation data were determined with 1 mM benzyl viologen (q) as the redox buffer, in addition to standard reaction components. Redox potentials were measured by direct potentiometry. DRAG activities were normalized to average activities of standard decoupled assays using freshly oxidized Fe protein. The fraction of Fe protein reduced (E) at poised potentials was determined from the change in the characteristic Fe protein absorbance at 420 nm. The redox potential of the titrated solution was determined from the intensity of the absorption at 608 nm due to reduced MV. The fraction of Fe protein reduced at each potential was normalized to the predicted maximal value. The midpoint potential of Fe protein (DRAG activation data excluded) was fitted with least-square curves using Prism software, with both curve slope and midpoint potential as parameters.   c In reactions lacking CODH, MV, or CO, Fe protein was not reduced as determined by UV-visible spectroscopy. (27). Although some DRAT activity was seen in reducing conditions, reactions in oxidizing conditions resulted in greater labeling (Table II) and 100% inactivation of Fe protein (Table  III).
Effect of a nifD Mutation on ADP-Ribosylation of Fe Protein from R. rubrum-The nifD gene encodes the ␤-subunit of the MoFe protein. R. rubrum cells lacking nifD, grown as described under "Materials and Methods," are incapable of the reduction of nitrogen (48). The Fe protein in an R. rubrum mutant (UR145) containing an insertion cassette in the nifD gene was ADP-ribosylated more slowly in response to darkness or ammonia than was Fe protein in wild-type cells (Fig. 5). In these experiments, the cells were derepressed for nitrogenase expression on a medium with malate as the major carbon source and glutamate as the nitrogen source. Under these growth conditions, either removal from light or addition of 10 mM NH 4 Cl causes the loss of nitrogenase activity due to inactivation of Fe protein by DRAT-dependent ADP-ribosylation. Lacking the MoFe protein, the UR145 mutant lacks the primary electron acceptor for Fe protein, trapping the Fe protein in the reduced state. DISCUSSION We have demonstrated the sensitivity of the DRAT and DRAG regulatory enzymes to the redox state of the substrate, Fe protein. In previous studies of these enzymes, treatment of the regulatory and substrate proteins precluded investigation of any redox effect. In the case of DRAG, the enzyme was thought to be oxygen-labile until recently (41), and the truly oxygen-labile substrate, Fe protein, was generally handled and stored in the presence of Na 2 S 2 O 4 as an oxygen scavenger. Thus, DRAG reactions were always performed in the presence of the strong reductant, sodium dithionite, creating the conditions required for activity. In reactions with DRAT, it was recognized that the reaction mix had to be treated with excess NAD ϩ to react away all dithionite. However, the ability of dithionite to reduce NAD ϩ masked the separate effect of activity inhibition under reducing conditions.
In these studies, a pair of activity assays for DRAG was developed, in which the activation of ADP-ribosylated Fe protein by DRAG could be studied in isolation. In previous studies of DRAG, activity assays were normally coupled directly with nitrogenase activity assays (29,53,54), which prevented investigation of the DRAG reaction requirements in isolation. A direct assay for DRAG activity has also been reported (31), in which the release of tritiated ADP-ribose from Fe protein is measured. The decoupled DRAG assay, described under "Materials and Methods," was shown to effectively separate the activation of Fe protein from acetylene reduction by activated Fe protein in conjunction with MoFe protein. The demodification assay also allows the study of DRAG activity in isolation. Both assays were shown to produce results similar to those of the radioassay but offer the advantages of speed and ease of implementation.
The salient discovery of this study is the demonstration that DRAG and DRAT activities are sensitive to the redox state of Fe protein in vitro. DRAG was clearly shown to have no activity with oxidized Fe protein. Some DRAG activity was observed with isolated, reduced Fe protein. This activity was limited by the fact that reduced Fe protein, in the absence of excess reductant, becomes oxidized during handling. It is not clear whether this oxidation occurs due to O 2 leaking into the reaction vials or due to Fe protein oxidation in scant dithionite solutions, noted by Watt et al. (12). As DRAG is insensitive to redox poise in reactions with other ADP-ribosylarginine substrates, it is concluded that DRAG senses the change in Fe protein in the redox status of the [Fe 4 S 4 ] 1ϩ/2ϩ cluster. The midpoint potential of DRAG activation was found to coincide with the Fe protein midpoint. The midpoint potential of R. rubrum Fe protein in the presence of MgATP was determined to be Ϫ470 mV, which is comparable to the values reported for Fe proteins from A. vinelandii (11,12) and K. pneumoniae (55). DRAT was found to have a specificity for Fe protein opposite that of DRAG, acting only upon the oxidized form of Fe protein. The CO/CODH/MV reducing system was shown to reduce oxidized Fe protein effectively, while leaving NAD ϩ unaffected. The small amount of activity seen for R.  rubrum Fe protein under reducing conditions may be due either to nonspecific deposition of 32 P label on CODH present in the assay, to a small (Ͻ10%) fraction of Fe protein remaining oxidized, or to incomplete inhibition of DRAT under reducing conditions. We propose that DRAT recognizes the same redox switch as DRAG, that is, the [Fe 4 S 4 ] 1ϩ/2ϩ switch in Fe protein.
The alteration of nitrogenase regulation in the nifD mutant demonstrates the physiological relevance of these in vitro studies. In R. rubrum, there are multiple low redox potential electron donors to Fe protein. 2 However, MoFe protein is the primary electron acceptor for Fe protein. Therefore, in the UR145 mutant, oxidation of Fe protein is significantly inhibited, presumably trapping the Fe protein in the reduced state. This result is predicted by the data collected for in vitro DRAT and DRAG reactions, which demonstrate less efficient ADP-ribosylation and more efficient removal of ADP-ribose when Fe protein is predominantly reduced. Note that the response curves to darkness and ammonia treatment resemble those of wild-type except in the magnitude of response. In response to the strong negative stimulus of darkness treatment, transient DRAT activity is observed, followed by a leveling of nitrogenase activity. In response to high levels of ammonia, the inactivating response, although weaker than that of darkness, is maintained for a much longer period of time (over 1 h). These results are consistent with a model in which the signal transduction pathways for regulation of DRAG and DRAT activity are unaltered in UR145, but the fraction of Fe protein available for ADPribosylation is reduced. It is not clear whether the slower rate of modification of Fe protein in the nifD mutant is a result of decreased rate of ADP-ribosylation, increased rate of removal of ADP-ribose, or a combination of both. Future analysis of the turnover of label on Fe protein will address this question.
The present study has a significant impact of the model for regulation of nitrogenase in R. rubrum. It is clear that the model for Fe protein regulation by ADP-ribosylation must address the availability and state of the Fe protein for reactions of the regulatory enzymes. We propose such a general model for the regulation of Fe protein activity by ADP-ribosylation in R. rubrum (Fig. 6). Acceptance of this model prompts an examination of previous results in this system. Johnson et al. (56) have demonstrated that a significant portion of A. vinelandii Fe protein remains oxidized when reductant is limiting in an in vitro assay system. If reducing power is also limiting in vivo, it is likely that a mixture of reduced and oxidized Fe protein is present at any given time. Analysis of previous work (32) suggests that DRAT, even when inactive, may be associated with oxidized, MgADP-bound Fe protein. The ability of reduced or oxidized Fe protein to form such an inhibiting complex with DRAT, without the ADP-ribosylation reaction occurring, may be studied in vitro by protein-protein cross-linking, after the method of Grunwald and Ludden (44). The inability of overexpressed DRAG to inhibit active Fe protein (32) may be due to sequestration of MgATP-bound Fe protein in complexes with MoFe protein. ADP-ribosylated, inactive Fe protein may, on the other hand, be available for DRAG under conditions conducive to nitrogenase and DRAG activity. Although the possibility of change in cellular ATP/ADP ratios has been ruled out as a regulatory signal to DRAT and DRAG (35), it is possible that local changes in redox potential, along with variations in ATP/ ADP ratios and NAD ϩ concentrations, may give rise to the dramatic regulation observed for DRAT and DRAG in vivo.
This study has also produced a result that is important to the understanding of the general nitrogenase catalytic cycle in all nitrogen-fixing organisms. Here, we clearly demonstrate that a significant conformational change must occur in Fe protein in response to the change in the redox state of the [Fe 4 S 4 ] cluster, independent of the nucleotide-induced conformational change (55). Circular dichroism studies of Fe protein, first reported by Stephens et al. (13), have long been used to demonstrate a change in the environment of the [Fe 4 S 4 ] cluster upon reduction or oxidation. However, these studies do not demonstrate a significant change in the overall protein structure. The results here suggest that Fe protein exists in different conformations in the oxidized and reduced forms in the presence of excess MgATP. DRAG is able to cleave the N-glycosidic bond that links ADP-ribose to Arg-101 only when the Fe protein is in the reduced form with MgATP bound. Thus, either DRAG cannot bind to the [Fe protein] ox ⅐MgATP complex due to a different conformation of the Fe protein or the N-glycosidic bond is inaccessible to DRAG in a [Fe protein] ox ⅐MgATP⅐DRAG complex. Either possibility demands that the conformation of [Fe protein] ox ⅐MgATP be significantly different from [Fe protein] red ⅐MgATP. The conformational change in Fe protein concurrent with the change in the Fe 4 S 4 redox state is independent of the nucleotide-dependent conformational change, as the redox switch was observed in 5 mM MgATP, saturating for either oxidized or reduced Fe protein (12,57). The redox-dependent conformational change may also explain the difference in the affinity of [Fe protein] ox and [Fe protein] red for MgADP and MgATP (12), as well as the redox-dependent difference in accessibility of the [Fe 4 S 4 ] cluster to ␣,␣Ј-dipyridyl (58).
The differences observed between R. rubrum and A. vinelandii Fe proteins in these studies provide some insight into the importance of subtle differences between these two very similar enzymes. Although the sequence identity between these two proteins is quite high, on the "upper" surface, the site of interaction with MoFe and presumably with DRAG and DRAT, 2 S. Nordlund, personal communication. there are only three differences in solvent-exposed residues. From the crystal structure of A. vinelandii Fe protein, it can be seen that Thr-66, Asn-173, and Ser-174 are exposed on the upper surface. These residues are replaced by Ser-66, His-173, and Thr-174 in R. rubrum Fe protein. Whether it is the actual residue changes that are critical, or rather subtle variations in the relative positions of upper surface residues, the A. vinelandii Fe protein is a significantly better substrate for DRAT than the natural substrate, due in part to a lower apparent K m for NAD ϩ (27). Here we have shown that DRAT still senses the redox switch in A. vinelandii Fe protein but that the inactivity toward reduced Fe protein is less strict than with R. rubrum Fe protein. Further studies are required to determine the amino acid differences in A. vinelandii and R. rubrum Fe proteins that cause the different reactivities toward DRAT. Interestingly, DRAG, which is generally less selective of its substrates than is DRAT, retains absolute specificity for the reduced form of A. vinelandii Fe protein. Thus, the significant conformational change occurring upon change of the redox state of the [Fe 4 S 4 ] cluster is likely to be important in all nitrogenase Fe proteins.