comparative biochemical characterization of the iron-only nitrogenase and the molybdenum nitrogenase...

Eur. J. Biochem. 244, 789-800 (1997) 0 FEBS 1997

Comparative biochemical characterization of the iron-only nitrogenase and the molybdenum nitrogenase from Rhodobacter capsulatus Klaus SCHNEIDER, Ute GOLLAN, Melanie DROTTBOOM, Sabine SELSEMEIER-VOIGT and Achim MULLER

Lehrstuhl fur Anorganische Chemie I, Fakultat fur Chemie der Universitat Bielefeld, Germany

(Received 10 October 1996/3 January 1997) - EJB 96 1510/3

The component proteins of the iron-only nitrogenase were isolated from Rhodobacter capsulatus (AnifhTDK, AmodABCD strain) and purified in a one-day procedure that included only one column- chromatography step (DEAE-Sephacel). This procedure yielded component 1 (FeFe protein, RclFe), which was more than 95% pure, and an approximately 80% pure component 2 (Fe protein, R c ~ ~ ' ) . The highest specific activities, which were achieved at an R c ~ ~ V R C ~ ~ ~ molar ratio of 40: 1, were 260 (C,H, from C,H,), 350 (NH, formation), and 2400 (H, evolution) nmol product formed . min-' . mg protein-'. The purified FeFe protein contained 26 -t 4 Fe atoms ; it did not contain Mo, V, or any other heterometal atom.

The most significant catalytic property of the iron-only nitrogenase is its high H,-producing activity, which is much less inhibited by competitive substrates than the activity of the conventional molybdenum nitrogenase. Under optimal conditions for N, reduction, the activity ratios (mol N, reduced/mol H, produced) obtained were 1 : 1 (molybdenum nitrogenase) and 1 :7.5 (iron nitrogenase). The Rcl" protein has only a very low affinity for C,H,. The K,, value determined (12.5 kPa), was about ninefold higher than the K,, for RclM" (1.4 kPa). The proportion of ethane produced from acetylene (catalyzed by the iron nitrogenase), was strictly pH dependent. It corresponded to 5.5% of the amount of ethylene at pH 6.5 and was almost zero at pH values greater than 8.5.

In complementation experiments, component 1 proteins coupled very poorly with the 'wrong' component 2. RclF', if complemented with R c ~ ~ " , showed only 10-15% of the maximally possible activity. Cross-reaction experiments with isolated polyclonal antibodies revealed that Rcl'" and Rc lM" are immu- nologically not related.

The most active RclFe samples appeared to be EPR-silent in the Na,S,O,-reduced state. However, on partial oxidation with K,[Fe(CN),] or thionine several signals occurred. The most significant signal appears to be the one at g = 2.27 and 2.06 which deviates from all signals so far described for P clusters. It is a transient signal that appears and disappears reversibly in a redox potential region between - 100 mV and + 150 mV. Another novel EPR signal (g = 1.96, 1.92, 1.77) occurred on further reduction of RclF" by using turnover conditions in the presence of a substrate (N,, C,H,, H+).

Keywords: nitrogenase ; iron protein ; cofactor; EPR; Rhodobacter capsulatus.

Three genetically distinct types of nitrogenase systems (ng vnJ unf, have so far been proved to exist in nature. The most widespread and intensively characterized system is the classical Mo-containing nitrogenase (nif system) found in all diazotrophs [l, 21. During the last few years, two types of alternative, Mo- independent nitrogenases have been discovered and described. One is a vanadium-containing nitrogenase (vnf system) (for review see [3]) and the other enzyme lacks both Mo and V (anf system) and has been tentatively designated as Fe nitrogenase [4-6]. Although there is much circumstantial and genetic evi- dence for the occurrence of alternative nitrogenases in many aer- obic and anaerobic bacteria [7], only a few of these enzymes

Correspondence fo A. Muller, Lehrstuhl fur Anorganische Chemie I, Fakultat fur Chemie, Universitat Bielefeld, Postfach 100 131, D-33501 Bielefeld, Germany

Abbreviutions. nij nitrogen fixation; vnf, vanadium-dependent nitrogen fixation ; an5 alternative nitrogen fixation; Rc, Rhodobucter cupsulutus; RclM", MoFe protein from Rhodobacter cupsulatus; Rclr', FeFe protein; R c ~ ~ " , Mo nitrogenase Fe protein; R c ~ ~ ' , Fe nitrogenase Fe protein.

Enzyme. Nitrogenase (EC 1.18.6.2).

have been isolated and biochemically identified. These include the V nitrogenases from Azotobacter chroococcum [7] and Azo- tobacter vinelandii [8] and the Fe nitrogenases from A . vinelandii [6, 91, Rhodobacter capsulatus [5, 101 and Rhodospirillum rubrum [ 1 I].

All three nitrogenase systems consist of two dissociable component metalloproteins, component 1 (MoFe protein, VFe protein, FeFe protein) and component 2 (Fe protein). Whereas component 2, which is a dimer of two identical subunits bridged by a single [Fe,S,] cluster [12, 131, obviously has a practically identical structure in all nitrogenase systems [3], component 1 of the alternative nitrogenases differs from the conventional tet- rameric MoFe protein (a,p,, total molar mass =240 kDa) in that it contains, in addition to the a and p subunits, a small approximately 16kDa subunit of so far unknown function resulting in a hexameric structure

The component 1 protein of nitrogenases contains two types of unique metal clusters, the cofactor (FeMo cofactor, FeV cofactor, FeFe cofactor), also called M-cluster, which most likely represents the site of substrate reduction [16], and the P-cluster whose function is not yet definitely known, but whose primary

[3, 5 , 14, 151.

790 Schneider et a]. (ELK J. Biochern. 244)

role probably is to transfer electrons from the Fe protein to the cofactor [17]. In recent publications by Rees and co-workers [17-191 and Bolin et al. [20], structural models for both the FeMo cofactor (Fe,MoS&omocitrate) and the P-cluster (Fe8S,.J were proposed based on X-ray crystal structure analysis of the conventional nitrogenase MoFe protein from A. vinelundii and Clostridium pasteuriunum at up to 0.22 nm resolution.

The Mo- and V-independent nitrogenase from R. cupsulatus was identified by multielement analysis as heterometal-free iron- only nitrogenase 141. The component 1 of this enzyme was reported to contain 24 Fe atoms/protein molecule but no other metal atoms. Recent comparison of amino acid sequences of nitrogenase structural genes including the three systems of A. vinelundii, confirmed that the alternative Rhodobucter nitrogenase belongs to the group of Mo/V-free nitrogenases: it clearly exhibits the highest sequence similarity with the unfsys- tem of A. vinelandii (e.g. the dp-subunits show 73-75% identity with AnfDK, but only S1-54% identity with VnfDK and 26-30% with NitDK) 1151. Biochemical [lo] and genetic data 115 J suggest that a second non-conventional N,-fixing system, a V-containing nitrogenase, does not exist in R. cupsulntus.

The Rhodobacter Fe nitrogenase was shown to become strongly repressed in the presence of molybdate [lo, 211. A concentration of only 6 nM MOO:- in the growth medium causes a 50% repression of enzyme s This is in accordance with the regulation found in A. v where both alternative nitrogenase systems are repressed by MOO:- [22]. It is, however, contrary to the situation in R. rubrum where molybdenum appears not to have a regulatory function 1231. In this organism, the alternative nitrogenase is expressed, whenever an active Mo nitrogenase is lacking due to physiological or genetic inactivation.

The Mo effect (increase of the production of C,H, from C,H, after addition of MOO:- to a growing AnifHDK culture) [21] appears to be a specific feature of the iron-only nitrogenases. By EPR spectroscopic data and the use of R. cupsulatus mutants, which were defective in the FeMo cofactor synthesis, we were able to demonstrate that the Mo effect is based on the formation of a hybrid enzyme consisting of the Fe nitrogenase apoprotein and the FeMo cofactor of the conventional nitrogenase. Hybrid enzyme formation was shown to be mainly a result of a cluster exchange mechanism [21].

In a preceding publication [S], we reported on the rapid purification of the Fe nitrogenase component proteins (RclF', Rc2") and on some catalytic and EPR-spectroscopic properties of this type of nitrogenase system. A more comprehensive biochemical description of the Fe nitrogenase is presented in this work which (a) allows a well-founded distinction from the conventional Mo- containing system of the same organism, (b) yields important information on the specific influence of the Fe replacing the heterometal center (Mo), on the EPR-spectroscopic properties of the cofactor-containing Rcl" protein and on the reactivity of this enzyme with substrates, and (c) provides generally a better understanding of the catalytic particularities of the Fe nitrogenase and in this context, of the structure/function interrelation- ships. Part of this work was presented at the Seventh Interna- tional Conference on Bioinorganic Chemistry [24].

MATERIALS AND METHODS

Bacterial strains. The organisms used were Rhodobucter capsulatus wild-type strain B 10s and two mutants generated from this strain : the kanamycin-resistant nifHDK deletion mutant [lo] and a kanamycin- and gentamycin-resistant double mutant with a nifHDK deletion and an additional deletion in the

modABCD region [2S]. The products of at least three of these genes (modABC) are involved in high-affinity molybdenum transport. ModB was similar to ChlJ, which has been described to be an integral membrane protein of the active MOO:- transport system in Escherichicl coli [26].

Growth medium and culture conditions. Because of the toxic/inhibiting effect of some salts and metal ions on the Fe nitrogenase activity, the composition of the growth medium was re-optimized. Under standard conditions the medium contained the following: 7.5 mM KH,PO,, 0.8 mM MgSO, . 7 H,O, 15 mM MnCI, . 4 H,O, 0.35 mM CaCl, . 2 H,O, 1.2 mM NaCl, 0.1 mM EDTA, 1 mM ferric citrate, 24 mM L-lactate and 5 mM serine. Under nitrogen-limiting growth conditions, the serine concentration used was 1.8 mM. The medium was additionally supplemented with thiamine (80 ng/ml), biotin (8 ng/ml), and kanamycin (25 pg/ml in the case of the An(fHDK mutant) or gentamycin (4 pg/ml in the case of the AnifHDK, AmodABCD mutant). The conventional trace element solution [ 101, which contained Zn, Cu, Ni, and Co chlorides, was omitted.

To remove traces of Mo impurities from the nutrient solution, stock solutions of the medium components were purified by the activated-carbon method [27]. To facilitate the application of this procedure for larger volumes, we introduced the following modifications. The nutrient stock solutions (0.1 M Fe citrate, 2.4 M L-lactic acid, 0.8 M KH,PO,) were not directly filtered through a layer of activated carbon, but were first mixed with the charcoal suspension (50 g/l H,O) in a ratio of 1 : 1, stirred overnight at room temperature, and finally suctioned through a filter of medium pore size. Since the procedure was performed with weakly acidic solutions [27], the pH values had to be adjusted afterwards with NaOH suprapur (Merck) to 6.8.

All other conditions for cultivating AnifHDK mutants were as described previously [lo], except that as standard incubation temperature 33°C instead of 30°C was applied. For cultivation of wild-type cells, nutrient solutions were not pretreated with activated carbon but supplemented with 10 pM molybdate. The medium also did not contain antibiotics. All other growth conditions were those used for the optimized expression of the Fe nitrogenase.

Preparation of cell-free extracts and purification of nitrogenase proteins. Anaerobic harvesting of the cells was carried out as described by Gollan et al. [21]. The cells, which were about 100-fold concentrated by centrifugation (corresponding to -50 mg protein/ml cell suspension), were disrupted exclusively by lysozyme treatment. This treatment was performed at pH 8.0 (in SO mM Tris) and optimized with respect to concentrations of lysozyme (from chicken egg) (2 mg/ml), polymyxin B sulphate (0.4 mg/ml), and EDTA (50 mM). The cell suspension was further supplemented with Na,S,O, (4 mM), deoxyribonuclease (0.2 mg/ml), superoxide dismutase (0.2 nig/ml), and phenyl- methylsulfonyl fluoride (2 mM), and stirred for 30 min under Ar at 30°C. All cell debris and membrane particles were removed by a 1 h anaerobic centrifugation at 1.SX 10' g.

Component 1 (FeFe protein, Rcl"") and component 2 (Fe protein, Rc2'') of the alternative nitrogenase were separated from each other and purified as single proteins by one DEAE- Sephacel chromatography step. The starting material was about 1.6 g protein from cells grown under N-limiting conditions with 1.8 mM serine. The cell-free extract resulting from lysozyme- treatment and subsequent high speed centrifugation, was loaded onto a DEAE-Sephacel column (internal diameter 2.6 cm), which contained about SO ml of the gel material and was previously equilibrated with 50 mM anaerobic Tris, pH 7.8, 4 mM Na2S,0,. This buffer was routinely used throughout the purification procedures if not stated otherwise. The DEAE column was stepwise eluted with five NaCl solutions of different concentra-

Schneider et al. ( E m J. Biochem. 244) 79 1

tions (each in Tris buffer). With about 60 ml each of 150 mM and 230mM NaCI, the bulk of contaminating proteins was extensively washed from the column. The Fe protein of the alternative nitrogenase was then eluted with 50 ml 260 mM NaCl. To achieve the most effective separation of the two nitrogenase components, an additional intermediate 290 mM NaCl fraction was also collected. This fraction contained low concentrations of both RclF" and Rc2'" and was discarded. RclF' was eluted with 30 ml 350 mM NaCI. Both protein components were first concentrated to about 8 ml by anaerobic ultrafiltration in a 50- ml chamber equipped with a PM30 membrane and desalted on Sephadex G-25 columns (internal diameter 5 cm; each contained approximately 70 ml gel). The resulting eluates were then concentrated directly in B15 chambers to final volumes of about 1 ml, which corresponds to total protein contents of 46 mg (Rcl'") and 29 mg ( R c ~ ~ ' ) .

The procedure for purification of the component proteins of the Mo nitrogenase (Rcl"", R c ~ ~ " ) isolated from wild-type cells, was identical with the procedure described for the Fe nitrogenase until the wash step with 230 mM NaCI. Since the MoFe protein and the Fe protein obviously bind to DEAE with almost the same strength, it was not possible to separate them by an- ionic exchange chromatography with a salt gradient. The two protein components were therefore eluted together and collected in one fraction. They started to move on the column at a concentration greater than 240 mM NaCl. However, to prevent a broad elution and thus an unnecessary dilution of the proteins, 300 mM NaCl was chosen as eluent. The resulting 25-ml fraction was concentrated to 2 ml and loaded onto a gel-filtration column (2.5 cmX80 cm) filled with about 500 ml Sephadex (3-150. The column was previously equilibrated with the standard buffer containing 30 mM NaCl to prevent ionic interactions. The protein was eluted at a flow rate of 36 ml/h, collected in 10-ml fractions, and checked for activity and purity. RclM0-containing fractions and R~2~"-containing fractions were pooled separately, concentrated about tenfold, and stored in liquid nitrogen.

During the purification, the columns were cooled to 10- 12°C by tap water flowing through the thermostat jacket of the columns. The bottles and tubes used for collecting the column fractions were kept in ice under a continuous stream of 0,-free argon during elution.

SDS electrophoresis. The purity of the protein components and the molecular masses of the subunits were analyzed by SDS/ PAGE according to the method of Laemmli [28]. The sharpest and most homogeneous bands were obtained, if the protein samples were mixed with 20 mM dithioerythritol before denatur- ation. As protein references a-lactalbumin (14.2 kDa), trypsin inhibitor (20.1 kDa), carbonic anhydrase (29 kDa), egg albumin (45 kDa), and bovine albumin (66 kDa) were used.

Assays of nitrogenase activity. Acetylene reduction. The whole-cell assay was carried out as described in [lo] except that shortened Hungate tubes (volume 6 ml) were used as reaction vessels and that, before the cell suspension was anaerobically transferred into the test tubes, a small amount of Ti(II1) citrate (0.25 mM) was added to the Ar-pregassed tubes to remove resid- ual traces of 0,. The reaction was finally started with 500 p1 C A .

The reaction mixture (0.5 ml) for the assay with the isolated proteins (Fe nitrogenase) contained the following: 100 mM Hepes, pH7.8, 5 m M ATP, 10mM MgC1,.7H,O, 6 m M Na,S,O,, 20 mM creatine phosphate, 0.2 mg creatine kinase and appropriate amounts of RclF' and Rc2". The amount of C,H, injected into the tubes was 500 pl as for the whole-cell assay. The reaction time was 15 min. All other experimental conditions were as previously described [lo]. The assay conditions applied for determination of the Mo nitrogenase activity were identical

except that the pH of the Hepes buffer was 7.4 and the concentration of MgCl, was 15 mM.

H2 evolution. The reaction mixture for measuring the H,- evolving activity was the same as for acetylene reduction except that different gas phases were used (N, alone, Ar alone, or vary- ing mixtures of C,H, and Ar). The H, formed was determined in a Hewlett Packard 5890A gas chromatograph fitted with a thermal conductivity detector. The GC column used (2 m in length, 3.2 mm diameter) contained Molecular Sieve 5A (60- 80 mesh) and was run at 100°C (oven temperature) and 250 kPa. The carrier gas was argon.

N2 reduction. N, reduction (ammonia formation) was determined by the indophenol method [29] as outlined by Eady et al. [30].

Protein determination. The protein content was determined by the Biuret method. The whole-cell procedure was carried out according to Schmidt et al. [31], the determination of extracts and isolated protein was according to Beisenherz et al. [32].

Metal analyses. Quantitative determinations of Mo and Fe in isolated nitrogenase proteins were carried out with an induc- tively coupled plasma mass spectrometer (VG Plasma Quad from VG Elemental, Winsford, UK) [27].

Serological methods. Production and isolation of polyclonal antibodies against Rcl"", RC~"', and Rcl and immunodiffusion assays were performed as outlined in [33].

EPR measurements. EPR (X band) spectra were recorded on a Bruker ECS 106 spectrometer equipped with an ECS 041 MR Bruker microwave bridge and an Oxford Instruments ESR 900 helium flow cryostat.

RESULTS

Growth and in vivo nitrogenase activity of wild-type and AnifHDK cells of R. capsulatus. The absence or the almost complete absence of Mo in the growth medium is certainly the most fundamental prerequisite for the derepression of the Mo- independent Fe nitrogenase in R. cupsulatus [ 10, 21, 241. How- ever, there are several further growth factors that also significantly influence the expression of Fe nitrogenase activity. Some of these factors, such as the use of 1 mM Fe(II1) citrate or of serine as the nitrogen source, have already been described in preceding papers [5, lo]. Other unpublished improvements and modifications of the medium composition and growth conditions are described in detail in the Materials and Methods section. Under the currently optimized culture and derepression conditions, AnifHDK cells showed a remarkably good diazotrophic growth based on the Fe nitrogenase as sole nitrogen fixation system. They reached a doubling time of 4.5 h during the loga- rithmic growth phase and a final protein content of about 300 pg/ml (100-ml batch culture). Diazotrophic growth of wild- type cells in the presence of 10 pM MOO:- was 1.6-fold faster, the final protein content 1.3-fold higher, and the C,H,-reducing activity 15-fold (N,-fixing cells) and 8-fold (serine-grown cells) higher.

The occurrence of active and inactive forms of Rc2"'. In the first paper on the detection and preliminary characterization of an Fe- and V-independent nitrogenase in R. capsulatus [lo], we only measured very low activity with whole cells. This activity further decreased drastically during the preparation of ceI1-free extracts. Meanwhile, we have succeeded in achieving an up to 20-fold enhancement of the in vivo activity by successive improvements of enzyme derepression, stability, and activity-assay conditions. However, regardless of the extent of the whole-cell activity, extract preparations were still accompanied by a loss of

792 Schneider et al. (EUK J. Biochem. 244)

approximately 30 kDa) and the ADP-ribosylated, slightly larger form ( y ' ) [34]. In an inactive R c ~ ~ ' molecule, one of the subunits is ribosylated thus forming a yy' subunit complex. The electrophoretic pattern of bands, in correlation with the activity data, proved (a) that after growth with an excess of the N-source, most of the R c ~ ~ " molecules (about 90%) become inactivated during the course of the extract preparation, which involved several steps in the dark (cell disintegration, two centrifugations), and (b) that under N-limitation, this dark inactivation can be prevented and thus the active y,-dimeric form of Rc2" main- tained.

Fig. 1. Subunit composition of the Fe nitrogenase after growth with excess N source and after growth under N limitation. Extract preparation and the general procedure of DEAE-column chromatography were performed as outlined in the Materials and Methods section; special conditions for the elution of the whole enzyme complex comprising both component proteins, were as described in [ S ] . The figure shows Coomas- sie-brilliant-blue-stained SDS polyacrylamide gels after electrophoretic separation of the Fe-nitrogenase system into its subunits. Lane 1, subunits of R ~ l ~ ' / R c 2 ~ ' from cells grown with 5 inM serine; lane 2, subunits of RclF'/Rc2" from cells grown with 1.8 mM serine.

90-95 % of the original nitrogenase activity. Using the following two different approaches, we were able to conclu- sively confirm that this activity loss was due to a regulatory inactivation, i.e. a reversible ADP ribosylation of component 2 (Fe protein, R c ~ ~ ' ) of the Fe nitrogenase as described for the Fe protein of some Mo nitrogenases, e.g. of the conventional system of R. capsulatus (for review see [34]):

(a) Based on the knowledge that inactivation of Mo nitrogenases (catalyzed by the enzyme dinitrogenase reductase ADP- ribosyltransferase) in phototrophic bacteria can be induced by darkness, and reactivation (catalyzed by the enzyme dinitrogenase-reductase-activating glycohydrolase) can be induced by light, we subjected Anif lDK cells to a darMlight treatment. If a phototrophically grown culture was stored in the dark, the Fe nitrogenase lost about 80% of its activity within one hour, but a large portion of the activity reappeared only some minutes after the culture was placed into the light again (data not shown).

(b) When the nitrogenase protein components were isolated from cells grown under diazotrophic conditions or with a standard serine concentration ( 5 mM), the Rc2" subunits gave rise to a double band in SDS electrophoresis gels (Fig. 1, lane 1) designated as y and y'. When we cultivated the cells under N- limiting conditions with 1.8 mM serine, the R c ~ ~ ~ activity increased up to tenfold and in SDS gels only a single protein band occurred (Fig. 1, lane 2).

How can these results be explained on the basis of the inacti- vatiodactivation mechanism? The R c ~ ~ ' protein, like the Fe proteins of all nitrogenases, consists of two identical subunits ( y2 ) , but y can be present in two forms, the unmodified form ( y

One-column purification of the Fe nitrogenase component proteins. For studies on purification and subsequent characterization of the Fe nitrogenase system, we routinely used the Mo- tolerant AnifHnK, AmodABCD double mutant [25]. This mutant is defective in the Mo-transport system and allows us to neglect trace impurities of MOO:- in the medium. Cells were, if not otherwise stated, cultivated under conditions of N-limitation.

The procedure of enzyme isolation and purification has already been described in a preceding short paper [5]. In this work, the preparation steps are described in detail in the Materi- als and Methods section and are summarized in the scheme of Fig. 2. A comparison of the specific catalytic activities of whole cells, cell-free extracts, and isolated proteins of both the Fe nitrogenase and the Mo nitrogenase from R. cupsulatus is given in Table 1.

The scheme (Fig. 2) also documents the distribution of proteins among the cell fractions. In cells optimally derepressed for Fe nitrogenase, RclF" (component 1, FeFe protein) and Rc2"' (component 2, Fe protein) accounted for 10-15% of the total protein. After a gentle cell disintegration by lysozyme/polymyxin treatment and a centrifugation step (1.5X10' g), about 70% of the protein was recovered in the membrane-associated fraction. Since the Fe nitrogenase is a soluble enzyme system, the proportion of the nitrogenase proteins in the cytoplasmic fraction corresponded to 40- 50 % (according to densitrometric analyses of SDS gels). Therefore, by centrifugation alone, the nitrogenase proteins could be purified approximately threefold. This was, of course, an optimal prerequisite for the development of a rapid purification procedure. We actually succeeded in sepa- rating and purifying RclFe and Rc2"' in a one-day procedure that included only one column chromatography (DEAE-Sephacel). The column was developed with a NaCl step-gradient (in the range from 150-350 mM NaCl in 50 mM Tris, pH 7.8). The Fe protein (about 80% pure) was eluted first with 260 mM NaCl, the FeFe protein (greater than 95% pure) was eluted with 350 mM NaCl [ 5 ] . From a starting material of 1.6 g protein (re- ferred to the total content of the cells used), we obtained 46 mg purified FeFe protein and 29 mg purified Fe protein.

With the final protein preparations, the following maximal specific activities were determined: 260 (C,H, reduction), 175 (N, reduction), and 2400 (H, production) U/mg protein (Table I). Since these remarkably high reaction rates were obtained, if approximately saturating amounts of Rc2re were used, i.e. a 40- fold molar excess over Rcl", they represent the enzymatic activity of the FeFe protein. The Fe protein activity, determined by using an about twofold excess of RclFe in the assay, was dramat- ically lower corresponding to only 15-20% of the activities calculated for the FeFe protein.

The Fe nitrogenase could also be isolated as an RclF"- and R~2~"-containing pure enzyme complex, if the 260 mM NaCl elution step was omitted. From the intensity of the Coomassie- brilliant-blue-stained protein (subunit) bands in SDS gels (Fig. l), we estimated an R~2~'/Rcl' ' ' molar ratio of approximately 2: 1.

Schneider et al. (Eur: J. Biochern. 244) 193

Membrane-associated proteins (-70 %)

Centrifugation

PURIFIED

Chromatography I RclF: R c ~ ~ '

10-15 % Nitrogenase proteins

SS-W% Other proteins 40-50 % Nitrogenase

proteins 5040% Other proteins

Fig.2. Steps of isolation of the Fe nitrogenase component proteins. Experimental details of each step are described in the Materials and Methods section

Table 1. The specific activities of the two nitrogenase systems of R. cupsulutus, determined with whole cells, cell-free extracts, and purified protein components. The conditions for the activity assays were as described in the Materials and Methods section. The gas phases of the assays were 100% Ar (H, production), 100% N, (N, reduction) and 90% Ar/lO% C,H, (C,H, reduction). The cell-free extracts represent the 1.5X10' g supernatants. The activity of these extracts was determined without extra addition of complementary components. 1 U corresponds to 1 nmol substrate reduced/product formed min-'. n.d., not determined.

Sample Specific activity

Fe nitrogenase Mo nitrogenase

C,H, reduction N, reduction H, production C,H, reduction N, reduction H2 production

U/mg protein

Whole cells Cell-free extracts Purified components

Rc2, Rcl = 2:l Rc2, Rcl = 40:1

15 n. d. 95 120 n. d. 1.50 12 n. d. 125 165 n. d. 205

42 n. d. 390 n. d. n. d. 590 260 17.5 2400 1200 235 1300

The FeFe protein is a hexamer consisting of three types of subunits (&& structure) [5, 151. The values of the individual molar masses, determined by SDS electrophoresis (Fig. I), devi- ated 3 - 13 % from the values obtained from nucleotide sequence analyses of the subunit structural genes [15]. They were 61 kDa (59.1 kDa, if calculated from the nucleotide sequence) for (I, 58 (50.7) kDa for p, and 15 (13.4) kDa for 6.

The purified FeFe protein contained, using the nucleotide- sequence-based molar mass of 246.4 kDa as a reference, 26 -C 4 Fe atoms, but did not contain Mo, V, or any other heterometal atom (compare Table 3), thus confirming the results of previous studies [4, 101.

Catalytic properties of the Fe nitrogenase. The composition of the assay mixture, which was required for the activity optimum of the iron-only nitrogenase, was, in principle, very similar to the assay compositions described for other nitrogenases (see Materials and Methods section). The most crucial factor, which apparently effects the Fe-nitrogenase activity more strongly than the activity of, for example, Mo nitrogenases, concerns the concentration of the electron donor sodium dithionite, which had to be kept as low as possible. If we required a relatively stable reaction (C,H, reduction, N, reduction), which must be at least linear for 15-20 min, the most suitable Na,S,O, concentration was 6 mM in 100 mM Hepes, pH 7.8 (standard conditions).

Titration of the RclF' protein with the R c ~ ~ ' protein, i.e. addition of increasing amounts of R c ~ ~ " to a fixed amount of RclFe, resulted in a rapid increase of the nitrogenase activity [with a similar dependency for all substrates (C,H,, N,, H') tested] up to a 20-fold molar excess of R c ~ ~ ' , followed by a saturation region with only insignificant activity increase and an apparent activity maximum at a 40-50-fold molar excess of

Rc2". Very similar dependencies on the component 2 concentration have been reported for MoFe proteins [30, 351 and the VFe protein of A. chroococcum [7]. With respect to the FeFe protein from R. cupsulutus, it has to be stressed that the optimal component 2/component 1 ratio was strictly dependent on the quality of the protein preparations, particularly on that of Rc2"", which appeared to be the much more labile component outside its cel- lular environment. The molar R ~ 2 ~ ' / R c l ~ ' ratio of 40-50: l was only optimal if samples of the most active and, as we assume, largely intact Rc2" protein (corresponding to a C,H, reduction activity of 55? 10 U/mg protein) were used. Less active R c ~ ~ ' samples required higher R ~ 2 ~ " / R c l Fc ratios for maximal Rcl"" activity. The activity of dark-inactivated Rc2" (from cells grown with 5 mM serine) was so low (1 -5 U/mg protein) that even a molar Rc2Fe/RclFe ratio of more than 150:l did not lead to a saturation of the Rcl"" activity.

Titration of the R c ~ ~ ' protein with the RclbC resulted, as described for other nitrogenases, in a rapid increase of nitrogenase activity reaching a maximum at a 1.5-2-fold molar excess of RclF'. A larger RclFe excess led to the inhibition of nitrogenase reactions. With R c ~ ~ ' samples of low activity (e.g. from dark- inactivated cells) even Rc2"/Rc1 ratios below 1 : 0.2 were in- hibitory.

Whereas Mo nitrogenases have been described to reduce acetylene with higher rates (1000-2000 U/mg protein) than the natural substrate N, (see [3] and data from Table l) , for alternative, Mo-independent nitrogenases, C,H, has been characterized to be only a poor substrate and thus an unsuitable indicator for catalytic activity. Particularly low specific activities (6- 60 U/ mg protein) were determined for iron-only nitrogenases of different origin (see activity comparison in [5]) . However, in all these cases, C,H, reduction was measured under standard condi-

794 Schneider et al. (Eur: J. Biochem. 244)

I , , , , , , , , , , ,\ J o g 6.0 6.5 7.0 7.5 8.0 8.5 9.0 K

pH value

Fig.3. pH dependence of C,H,, C,H, and H, production. The pH- dependent activities were carried out under standard assay conditions but with a three-component buffer system (75 mM Bistris, 38 mM Hepps. 38 mM Ches) which has a sufficient buffer capacity in the pH range 5.0-9.8 1371. A, C,H, production; 0, C2H, production; M, H, formation.

tions (with 10 kPa for C2H,) without consideration of the potentially low affinity of this substrate to the enzyme. We have therefore determined the c : H 2 for both the FeFe protein and the MoFe protein of the conventional system (RclM"). The K,,, for RclM" was 1.4 kPa and thus is in a partial pressure/concentration region found for other MoFe proteins [37]. The K, for RclF' was, i n contrast, about ninefold higher (12.5 kPa) proving that C,H, has a much lower affinity to this alternative enzyme system. This means that the use of 10 kPa C,H, yields a fully saturating substrate concentration for Rcl Ma, but leads to a clear underestimation of the C,H,-reduction activity of Rcl Fe. When we performed the assay under a C,H, atmosphere (100 kPa), the Rcl Fc activity increased about twofold (data not shown).

A general catalytic characteristic of alternative nitrogenase is the capability to reduce CLH, not only to ethylene but also partially to ethane. The proportion of this second product has been reported to account for 2-4% of the total amount of reduced acetylene. In the case of the Fe nitrogenase, the production of C,H, becomes enhanced up to 60% if a hybrid enzyme, i.e. the FeFe protein with the FeMo cofactor incorporated, is formed [6, 211. With the isolated FeFe protein, which is completely free of Mo, only 1.6% of C,H, was produced from C,H, under standard assay conditions. We have found, however, that the proportion of C,H, produced, is strictly pH dependent (Fig. 3). It was 5 . 5 % at pH 6.6, 1.6% at pH 7.8, and near zero at pH 9.0. The C,H, production catalyzed by the V nitrogenase (A. chroococcum) showed a different pH dependency : the proportion of C,H, produced from C,H, remained almost constant over a wide pHrange (6.5-7.5) and decreased only slowly above pH 7.5 [361.

100

5 80 - > > c .- ._ 4 60

6 40

T

m c ._

>

8" 20 \ .-., . ,

5 10 15

C,H, partial pressure @Pa)

3

Fig. 4. Inhibition of the H, evolution by acetylene. The Rc2/Rcl ratio applied was about 10:1 with both enzyme systems. The assay mixtures were incubated under Ar with partial pressures of C,H, varied as indicated in the figure. A, Fe nitrogenase, M, Mo nitrogenase.

The Fe nitrogenase exhibited an extraordinary high H,- evolving activity. With intact cells and also with the in vitro system, at Rc2/Rcl ratios as they exist in the cell (2: l), the Mo nitrogenase was only slightly superior (see the data of Table 2 and also [38], which contains a more detailed characterization of the itz vivo H, production by R. cupsulatus wild-type and mutant cells). When the isolated components at a 40-fold molar excess of component 2 were used, then the Fe nitrogenase showed higher H, production rates than the Mo nitrogenase. This H,-evolving capability was particularly drastic in the simul- taneous presence of a competing substrate. The ratio of the Fe- nitrogenase activity/Mo-nitrogenase activity increased with increasing proportion of C,H, in the gas phase (Table 2). In the presence of 5 % (by vol.) C,H, in Ar, the Fe nitrogenase produced approximately fourfold and in the presence of 20% (by vol.) C2H2, 36-fold larger amounts of hydrogen than the Mo enzyme. This has certainly to do with the extremely different extent of inhibition of the H,-evolving activity caused by C,H,. From the inhibition curves (Fig. 4), we can determine a 50% inhibition for the M o system at 0.7 kPa and for the iron-only system at 10.1 kPa. This weak inhibition of the H, production is consistent with the observed low affinity of the Fe nitrogenase towards C,H,, which suggests that in the case of this enzyme, C,H, does not compete very effectively with proton reduction. Although the inhibition by N, only played a comparatively mi- nor role (see Table 2), in 100% N, atmosphere, the rate of H, produced by the Fe nitrogenase was about sixfold higher than the rate determined for the Mo nitrogenase. Based on the data of Tables 1 and 2, the molar ratios of H, produced per N, reduced, were calculated to be 1:l (Mo nitrogenase) and 7.5:l (Fe nitrogenase).

Table 2. H, production by the isolated Fe- and Mo-nitrogenase systems under various gas atmospheres. The component ratio Rc2/Rcl was 40: 1 in all cases.

Gas atmosphere Fe nitrogenase (Rcl") Mo nitrogenase (RclM") Activity ratio (RclYRclM")

H, production inhibition H, production inhibition ~

U/mg protein o/o U/mg protein %

1.8:1 Ar 2400 Ar/l% (by vol.) C,H, 2250 6 585 65 3.8: 1 Ar/S% (by vol.) C,HL 1800 25 170 87 10.5: 1 Ar/20% (by vol.) C2H, 910 62 25 98 36 : I N2 1300 46 210 75 6.2: 1

- 1300 -

Schneider et al. (Eul: J. Biochm. 244) 795

Table 3. Comparison of the Fe- and Mo-nitrogenase component proteins. The data on molar masses are based on amino acid sequence analyses [IS]. For component complementation, the ratio was 1 :40 in all cases.

Property studied Result

Metal content (mollmol) MoFe protein FeFe protein

MoFe protein FeFe protein Fe protein (Rc2"") Fe protein (Rc2")

a subunits (RclM", Rcl'") y subunits ( R c ~ ~ " , Rc2")

Rcl"" to Rcl" R c ~ ~ " to R c ~ ~ '

(70 activity) Rcl '"IRc2" Rc 1 F'/R~2M" Rcl " " l R ~ 2 ~ " Rc1""IRc2'"

Subunit structurelmolar masses (kDa)

Amino acid sequence identity (70)

Serological relationship

Component complementation

30 (24-36) Fe", 2 Mo, no V 26?4 Fe, no Mo, no V

u*P* (59, 57)

y 2 (31) a 2 ~ 2 d z (59, 51, 13.5)

y 2 (30)

26 62

no relationship partial identity

100

100 10-15

1 - 2

" The value of 30 mol Felmol protein represents the iron content determined from the crystal structure. The values given in brackets are based on metal analyses of the most active, soluble preparations of MoFe proteins [SO].

Fig.5. Comparison of the component proteins of the Fe- and Mo- nitrogenase by Ouchterlony immunodiffusion. (A) The center well of the agarose gel contained 0.3 mg anti-Rcl'" IgG and 0.35 mg anti-RclM" TgG; wells 1, 3, and 5 each contained 20 pg RclF' protein and well 4 contained 36 pg Rcl'" protein. (B) The center well contained 0.55 mg antiLR~2~' IgG; well 1 contained 14 pg R c ~ ~ ' , well 5 28 pg Rc2'", well 2 contained 17 pg R c ~ ~ " , well 4 34 pg R c ~ ~ " .

The pH optimum of the H,-evolving activity (Fe nitrogenase) was at 7.0-7.2 (Fig. 3). At pH 7.1, we obtained an about 20% higher H,-production rate than at pH 7.6, the optimum determined for the C,H, reduction, if assayed in a Bistris/ Hepps/Ches combined buffer system (see legend of Fig. 3). The profiles of the two pH-dependency curves were evidently similar only differing by a pH shift of 0.5-0.6. With 100 mM Hepes, the buffer routinely used in activity assays, the pH optimum region (N, and C,H, reduction) was broader and showed only a weakly pronounced activity maximum around pH 7.8 (data not shown).

Complementation experiments and immunological comparison of nitrogenase proteins. To find out to what extent RclM" and RclF", and R c ~ ~ " and R c ~ ~ " , are functionally interchange- able, heterologous component protein mixing (cross-complementation) experiments were performed. The results demon-

strate that component 1 of both enzyme systems coupled very badly with the 'wrong' component 2. Rcl", when complemented with R c ~ ~ " , showed only an activity (C,H, reduction) of maximally 10-15% if compared with the activity of the natural R C I " / R C ~ ~ " enzyme complex (Table 3). Vice versa, when Rcl'" was complemented with R c ~ ~ " , the activity was even lower (1 - 2 %), regardless of which component ratio had been applied.

We also isolated polyclonal, monospecific antibodies (produced in rabbits) against the component proteins and performed an immunological comparison. Antibodies against Rc 1" (anti- Rcl" IgG) did not cross-react with the RclM" protein, and, in agreement with this result, antibodies against Rcl M'l (anti-Rcl M''

IgG) did not cross-react with the RclFe protein. When anti-RclF' IgG and anti-Rcl'" IgC were both placed in the middle well of an immunodiffusion gel, the respective antigens into two neigh- boring outer wells, then two precipitation lines occurred which crossed each other forming two spurs (Fig. 5A, lower bands). This pattern of bands represents the non-identical reaction type and confirms that a serological relationship between RclF' and Rcl"" does not exist. For comparison, into two other neighbor- ing wells, two identical protein samples (Rcl Fr) were placed. This resulted in two precipitation bands that completely fused into one line without spur formation indicating their identity (Fig. 5 A, upper bands).

The two Fe proteins (Rc~"", R c ~ ~ " ) were proved to be sero- logically related. This was indicated by the partial-identity reaction type, i.e. the partial fusion of precipitation lines with the formation of a single spur (Fig. 5 B).

EPR properties of the FeFe protein. The Na,S,O,-reduced iron proteins ( R c ~ ~ " , Rc2'") of both nitrogenases did not show any unusual EPR characteristics. Their spectra were very similar [5] and were also similar to those described for the Fe proteins of other Mo- and of V-nitrogenases [8]. We therefore focused our interest on the EPR properties of the FeFe protein.

The MoFe protein of R. c f f ~ s u ~ f f ~ u ~ ~ showed, in the semi- reduced state (preparation in the presence of 4 mM dithionite), the typical S = '/, EPR signal at g = 4.3, 3.7, and 2.01 (Fig. 6, spectrum A). Under the same conditions (4 K, 20 mW), the FeFe protein (we used the so far most active and almost homogeneous sample in this study) was EPR-silent. In the g = 1.9-2.1 region, only a very weak and poorly resolved resonance was detectable (Fig. 6, spectrum B). This signal, which became more prominent at 16 K, did not correlate with the catalytic activity and corresponded to less than 0.05 spinsiRcl'" molecule [5]. An S = '1, signal was definitely absent no matter what the conditions were. The EPR resonance peak at g = 5.44, described in an earlier paper [4]: also did not correlate with the catalytic activity. In- stead, it became weaker as the specific Rcl Fe activity increased, and was absent or of negligibly low intensity at activities greater than 150 U/mg protein (C,H, reduction). We therefore conclude that the intact and fully active FeFe protein is EPR-silent in the dithionite-reduced state and that the previously observed g =

5.44 signal was due to a preparation artifact possibly resulting from an altered FeFe cofactor or of a cofactor break-down product.

Instead, we have detected some novel and significant EPR signals upon partial oxidation of RclF' as well as under turnover conditions. A sample that contained 0.5 mM dithionite and was stepwise oxidized with K,[Fe(CN),] (0.25 -2 mM), yielded the following types of signals (Fig. 6, spectrum C): (a) A (sugges- tive of an S = 9/2 ground state) weak signal at g = 6.50 and a resonance peak at g = 2.03. Both features have not been studied in detail and cannot be assigned at present. (b) A g = 4.3 signal which may only in part represent non-functional ferric iron. This signal is not typically isotropic but rather shows rhombic sym-

796 Schneider et al. (Eur: J. Biochem. 244)

4.29

2.01 + v A

2.05 1.93

2.27 4.30

100 2 00 300 400 Magnetic field (mT)

Fig. 6. Basic EPR spectra of the MoFe protein and the FeFe protein. Spectrum A, MoFe protein (10.5 mg protein) reduced with 4 mM Na,S,O,; spectrum B, FeFe protein (45 mg protein) reduced with 4 mM Na,S,O,; spectrum C, FeFe protein (4 mg protein) reduced with 0.5 mM Na,S,O, and subsequently reoxidized with 0.5 mM K,[Fe(CN),]. Spectra A and B were recorded at 4 K, a microwave power of 20 mW, and a microwave frequency of 9.44 GHz; spectrum C was recorded at 16 K, 100 mW and 9.44 GHz.

2.00 2.03 1.98

A 227 2.06) I 1 (( 1.96

1

0.2 mW

W I I I I I

280 320 360

B

i

, I I I I

280 320 360 Magnetic field (mT)

Fig. 7. Temperature and microwave-power dependence of the EPR signals derived from partially oxidized FeFe protein. Sample conditions (redox state, protein content) were the same as for spectrum 2 of Fig. 8. Microwave-power-dependence measurements were performed at 16 K (A), and temperature-dependence measurements at 100 mW (B). The applied microwave frequency was 9.44 GHz in all cases.

Schneider et al. (Eur: J. Biochem. 244) 797

Fig. 8. EPR signals of thionine-oxidized RclM" and RclFe. The protein samples (RclM": 9 mg/ml; RclF': 11 mg/ml), prepared in the presence of 4 mM Na,S20,, were reoxidized with 10 mM thionine. To guarantee anaerobic conditions, solid thionine was preferably used. A few grains (0.5 mg) were put into the EPR tube and the tube was flushed with 02- free argon for 15 min. The protein sample (200 pl) was then transferred into the EPR tube and the thionine directly dissolved in the protein solution. After 2 min incubation at room temperature, the sample was frozen in liquid N,. The spectra were recorded at 16 K, 100 mW, and 9.44 GHz.

metry. Further, the signal intensity increased on increasing the K,[Fe(CN),] concentration up to 0.75 mM, but it decreased again above 0.75 mM. (c) A very narrow rhombic signal at g = 2.0, 1.98, and 1.96 which was easy to saturate (Fig. 7A) and was strongest at 23 K. At this temperature, it was not disturbed by other, overlapping signals (Fig. 7 B). As the intensity maximum was not reached during titration with K,[Fe(CN),], and also because this signal was only partially reversible by rereduc- tion with Na,S,O,, it appeared to represent an oxidatively darn- aged rather than an intact cluster. (d) The most important EPR signal is certainly that at g = 2.27 and 2.06. Because of the wideness of this signal, the possibility of a third negative peak must be considered (in the g-region of 1.8- 1.9), which however might be too broad to be detectable. This g = 2.27 signal is particularly interesting because of the following characteristics : (a) It is only detectable in a narrow temperature region rapidly disappearing at less than 10 K and above 16 K (Fig. 7B). (b) It is a transient signal, which means it disappears on further oxidation as well as on reduction of Rcl"', thus going through two redox steps. (c) It is reversible and is therefore unlikely to be an artifact, and (d) it deviates from all signals so far described for P clusters [39, 401. The g =2.27 signal was strongest at a concentration of 0.5 mM K3[Fe(CN)J. The maximum intensity of this signal was roughly determined to occur at -20 mV (versus a normal hydrogen electrode).

After reoxidation of the 4 mM Na,S,O,-reduced FeFe protein with 10 mM thionine (2-min incubation), the g = 2.27 signal was the only clearly recognizable EPR signal in the high- field region (Fig. 8, lower spectrum). Other potentially present resonance peaks near g = 2 were overlaid by the strong thionine radical signal. The g = 2.27 signal was definitely absent in the case of the MoFe protein after the same thionine treatment. In-

1.96 +

1.92 1.77

I I

I I I

3 40 360 380 Magnet ic f ie ld (mT)

Fig.9. EPR signal of the FeFe protein obtained under steady-state turnover conditions with N, as substrate. The sample was anaerobically prepared (under N,) in a 6-ml assay tube. It contained Rcl'" (25 mg/ml), R c ~ ~ " (3.5 mglml), and all other constituents (Hepes buffer, ATP, creatine phosphate, creatine kinase, MgCI, . 7 H,O, Na,S,O,) required for the catalytic reaction (concentrations as for the routine activity assays, see Materials and Methods section). The reaction was started with ATP and the mixture incubated for 5 min at 30°C. The whole sample (0.21 ml) was then anaerobically withdrawn from the assay tube, passed through a syringe into the EPR tube, and immediately frozen in isopentane cooled by liquid nitrogen. The spectrum was recorded at 16 K, 100 mW, and 9.44 GHz.

stead, a distinct signal at g = 2.06, 1.96, and 1.83 was detectable (Fig. 8, upper spectrum). From the position of the g values, this signal is strongly reminiscent of the signal interpreted by Titts- worth and Hales [40] as S = 1/2 signal arising from one- electron-oxidized P-clusters (AvIMo protein). This signal was, however, absent in the case of the FeFe protein. We are thus left with the exciting question of whether the RclF' signal at g = 2.27 is an atypical P-cluster signal characterized by a pronounced broadening and a dramatic g-value shift, or whether this signal is actually due to a paramagnetic state of the FeFe cofactor.

According to the Mo nitrogenase system, i t has been postulated that during enzyme turnover the semi-reduced and EPR- detectable FeMo cofactor (S = '/? signal) is further reduced by one electron to an EPR-silent state [42 I. In steady-state experiments with the nitrogenase from K. pneumoniae, i.e. EPR measurements of samples in which, in analogy to the activity assay, both component proteins (Kpl, Kp2) and all other additions nec- essary for the catalytic reaction (substrates, ATP, MgC1,) are simultaneously present, a 60-90% decrease in the intensity of the FeMo cofactor EPR-signal was observed [42].

Provided it is true that the FeFe cofactor of the iron-only nitrogenase in the Na,S,O,-reduced state contains an equal number of ferric and ferrous iron-centers resulting in an EPR- silent, most likely S = 0 ground state (51, then it should be expected that not only on partial oxidation but also on further reduction under steady-state turnover conditions, the FeFe cofactor becomes converted into an EPR-detectable state. In fact, when

798 Schneider et al. (Eur: J. Binchern. 244)

turnover conditions were applied, a new EPR signal at g = 1.96, 1.92, and 1.77 appeared (Fig. 9). This signal, which was most pronounced at 16 K (100 mW), was qualitatively identical regardless of the substrate used (N?, C,H,, H+). It was, however, strongest in the presence of molecular nitrogen. This turnover signal may thus represent the fully reduced state of the FeFe protein.

DISCUSSION

In several preceding papers, we reported the detection [lo], identification [4], purification 151, and partial biochemical characterization of the iron-only nitrogenase from R. cupsulntus. This characterization included data on subunit structure and some basic catalytical and EPR-spectroscopic properties [5], on the formation of a FeFe protein/FeMo cofactor hybrid enzyme [21], and on the in vivo production of molecular hydrogen [38]. A more comprehensive description and discussion of the properties of this enzyme system including a detailed comparison with the Mo-containing nitrogenase, has so far been lacking and is presented in this article.

An apparent iron-only nitrogenase or ANF system, respectively, has also been isolated from two other bacteria, i.e. A. vinelaiidii [6, 91 and R. rubrum [ 1 1 1 . In the case of the Azotobuc- ter enzyme (nitrogenase 3) , the investigator is confronted, i n particular, with two problems: either the isolated Avl'" protein is Mo-free but of very low activity ([9], Pau, R., personal com- munication), or it has, when isolated from a Mo-tolerant AnijHDK, AvnfHDGK mutant, better activity (e.g. 110 nmol NH: formed . min-' . mg protein-'), but contains about 1 mol Mo/mol of enzyme and exhibits the typical features of a hybrid enzyme (FeMo cofactor EPR-signal, increased production of C,H, from C2HZ) 161. These observations led to the widespread assumption that the natural cofactor, the FeFe cofactor, gets lost during extract and enzyme preparation and that this type of nitrogenase is, at least in vitro, only functional if the FeMo cofactor is incorporated into the active site. The ANFl protein (Krl'") of R. rubrum also showed only low activity (maximally 40 nmol C2H, reduced . min-' . mg protein-'). It should, however, be noted that the enzyme components were isolated from cells grown in the presence of 1 inM tungstate. In spite of these unusual conditions for enzyme expression 1231, Rrl"" was determined not to contain significant amounts of W.

These examples illustrate that the previous lack of a detailed characterization of iron-only nitrogenases is due to the fact that in the past highly active and stable protein samples were not available. Studies with only low active proteins are not reliable, because one never knows whether the results obtained really represent the intact enzyme. The great problem is that the potential activity maximum of a newly discovered enzyme, as that of the Fe nitrogenase, is not known and, therefore, the activity which reliably represents the intact enzyme is not known either. Our main intention was therefore to find procedures and experimental conditions that enable us to purify the Fe nitrogenase component proteins furnished with the highest possible catalytic activity and thus to disprove the hypothesis that the iron-only nitrogenase system cannot be isolated in a functional form.

Already in earlier studies the Fe nitrogenase turned out to be much more unstable than the conventional Mo nitrogenase. This conceined not only O2 sensitivity [38], but also lability towards dilution, freezinghhawing [ 101, and excess concentrations of salts and metal ions [ S ] . The only possibility to obtain an enzyme system with good activity i n the purified state despite this instability was to develop a very rapid purification procedure that makes time-consuming, protein-diluting chromatography

steps (e.g. gel filtration) and freezingkhawing cycles unnecessary. Based on two essential prerequisites, (a) the elaboration of optimal conditions for enzyme expression (which corresponds to a pre-enrichment i n the cell) and (b) a gentle lysozyme cell disintegration method, we succeeded in establishing a rapid one- day procedure resulting in specific activities that were on average ten times higher than the activities reported earlier for the same enzyme but also for the analogous nitrogenase 3 system from A. vinelandii (compare [9] and Table 1 in [5]). The functional intactness of the FeFe protein and the presence of a catalytically active iron-only cofactor (FeFe cofactor) was also indicated by the formation of only low amounts of C,H,, the accompanying product of C,H, in the C,H, reduction assay, by the definite absence of Mo and of an EPR signal attributable to the FeMo cofactor, and by the fact that RclFe contained distinctly more Fe atoms (26 -C 4) than required for P-clusters.

It is highly significant to note that if we determined the activity of the two component proteins individually by reciprocal titration, the activity of the non-ribosylated R c ~ ~ " component was at best 20% of the activity obtained with RclF'. From these results, we conclude the following: (a) The low component 2 activity leads to a significant underestimation of the individual FeFe protein activity and limits the catalytic potential of the iron-only nitrogenase in general. (b) It is not the modified structure of the cofactor, i.e. the absence of a heterometal atom, but rather the weakly active component 2 proteins that might princi- pally be responsible for the low reaction rates measured for Fe nitrogenases. Based on the fact that, in the case of the Rhodo- hacter enzyme, the inactivation by ADP-ribosylation can be overcome by isolating the enzyme from cells grown under N limitation (Fig. 2), relatively low Rc2'" activity may be related to the following two basic causes:

(a) Since we have determined that the stated instability of the Fe nitrogenase system is mainly due to KC^"', we assume that a considerable portion of R c ~ ~ " becomes irreversibly inactivated i n the course of the enzyme isolation despite the rapidity of the involved procedures. Because of the relatively good activity yield (40-60%) after the DEAE-column chromatography, we assume that the greater part of Rc2"" is inactivated during the steps of extract preparation. From the first impression, the specific activity determined for cell-free extracts seemed to be satisfactory because it reached high values similar to those calculated for whole cells (compare Table 2). It has, however, to be considered that the extracts used in the assays, represent the centrifugation supernatants already separated from more than two-thirds of the protein and thus a 3-4-fold higher extract activity was rather to be expected. However, extract and whole- cell activities are very difficult to compare, because in the in vitro assay the natural electron donor (probably a ferredoxin 1431) is not used but a chemical reductant (Na,S,O,) is employed which is known to weaken the interaction between component 1 and 2 [44] and may have an additional inhibitinghnactivating influence on Rc2"' itself [lo]. From these arguments, it follows that the proportion of Rc2"' that actually became inactivated upon the isolation procedure is very hard to estimate.

(b) The second cause of the relatively low Rc2"' activity could simply be that low activity is a specific characteristic of the Kc2'-' component. Even the highest specific activities determined (50-60 nmol C,H, reduced . min-' . mg protein <) can still be categorized as low activity, but nevertheless they inay represent the native, functionally intact and, at least under in vitro conditions, maximally active protein. This means that Rc2" has a drastically lower activity than RclF', and also lower activity than component 2 proteins from Mo and V nitrogenases. Either the structural requirements (protein conformation, loca- tion of the Fe,S, cluster, position/orientation of the amino acids

Schneider et al. ( E M J. Biochem. 244) 799

involved in component-protein interaction) are unsuitable for a better activity or, as already suggested, the Rc2" protein is specifically inhibited by dithionite and/or another, yet unknown factor of the in vitro system. Detailed kinetic studies are planned to help to answer this question.

In spite of the relatively low Fe protein activity and the presumably not fully exhausted catalytic capacity of the FeFe protein, in the presence of a large R c ~ ~ ' excess, surprisingly high specific activities were achieved. These were, in the case of N2 and CZHZ reduction (Table l ) , in the same order of magnitude as those reported for the activities of V nitrogenases. With respect to H L production, the reaction rates determined for the Fe nitrogenase were distinctly higher than those published for V nitrogenases and even higher than the rates measured for the Rho- dobacter Mo nitrogenase in this study (Table 2). This superior H,-evolving capability was particularly pronounced in the simul- taneous presence of C,H, or N,. In the case of the Fe nitrogenase, the reduction of these substrates obviously does not compete very efficiently with proton reduction, thus resulting in H,-evolution inhibition effects that are substantially weaker than those analyzed for other nitrogenase systems. In this context, it is interesting to compare the electron allocation in the two Rhodobacter nitrogenases under optimal conditions for N, reduction. The activity values taken from Tables I and 2 yielded molar ratios (N? reduced/H, formed) of approximately 1 : 1 (Mo nitrogenase) and 1 :7.5 (Fe nitrogenase). This means that the Mo nitrogenase allocates, as one would expect and as also found for other conventional systems, 75 % of the total electron flux to N, reduction, the remainder to proton reduction. In the case of the Fe nitrogenase, electrons are distributed the other way round. This enzyme allocates only a smaller proportion of electrons (25-30%) to N, reduction, whereas the bulk of electrons are utilized for the reduction of H'. In this respect, the V nitrogenases apparently take a middle position, as they have been postulated to allocate 50% of the total electron flux to both substrates [7] .

It was a striking observation that under in vivo conditions the FeFe cofactor of the iron-only nitrogenase can be replaced by the conventional FeMo cofactor forming a hybrid enzyme [21]. The fact that such a cluster exchange may occur, implies (a) that the FeMo cofactor and the FeFe cofactor must be at least topologically comparable and (b) that the immediate protein vi- cinity of the cofactors, their binding pocket, must be similar in both the MoFe protein and the FeFe protein. These suggestions were confirmed and extended by recent genetic analyses [15] that revealed several even identical structural details of the proteins as well as of the cofactors: (a) the two conserved amino acid residues in the (x subunit of the MoFe protein (cysteine, histidine), which are the only protein ligands of the FeMo cofactor, are also present in the a subunit of the FeFe protein; (b) at least two genes (ni f l nip) are required for the biosynthesis of both cofactors thus proving that homocitrate and the NifB cofactor (an FeS-cluster fragment formed by the nip gene product [45, 461) are also essential constituents of the iron-only cofactor.

Based on this structural conformity, it appears that the only fundamental structural difference between the two cofactors is the presence of a diverse heterometal atom or the absence of such a heterometal atom in the Fe nitrogenase. Of course, this heterometal atom may have, even if it is not directly involved in the catalytic reaction [47], a significant influence on the ge- ometry, the electronic/magnetic structure, and the stability of the cofactor and therefore, in a more indirect manner, also on the catalytic activity. Several of the described specific catalytical characteristics of the Fe nitrogenase (lower N,-reducing activity than the Mo enzyme, extraordinary high H,-evolving activity, weak affinity to C2H2, partial and pronounced pH-dependent re-

duction of C,H, to C,H,) might be due to the influence of the replacement of the heterometal center by an iron center. In a recent theoretical study on the electronic structure (molecular orbital calculations) of idealized structural models of the different types of cofactors, it has been postulated that N2, if bound to the FeFe cofactor, becomes more difficult to be reduced than if it were bound to the FeMo cofactor or the FeV cofactor due to a substantially larger highest-occupied molecular orbitalAowest- unoccupied molecular orbital (HOMO-LUMO) energy separation 1471. This consequently implicates that N2 reduction, catalyzed by the Fe nitrogenase, either follows an alternative mechanism or the ratio of reaction products (NH,, H,) greatly differ from those of other nitrogenase systems. This latter possibility is indicated by the results described and discussed in this work.

Another point should not be neglected, namely that, with respect to some of the deviating properties (e.g. the height of reaction rates and K,,, values), it is hard to differentiate between those that actually arise from influence of the additional Fe center and those that arise from the protein influence, i.e. from dif- ferences in amino acid sequence, protein conformation, pres- encehbsence and strength of non-covalent interactions of the cofactor with its environment etc. Several examples for the strong influence of the protein on cofactor and activity properties have been more comprehensively discussed in 1481. In this context, we wish to point out that in spite of the mentioned structural similarities/identities close to the catalytic center of Rcl and RclFe, the overall similarity of the two functionally analogous proteins is small. This is documented by the results of three approaches using different methods (compare Table 3) : (a) The amino acid sequence of the 0: subunits in which the cofactors are located only shows 26% identity 1251. (b) If the Rcl components were each cross-complemented with the wrong component 2 (RclF' with R c ~ ~ " and vice versa), the catalytic activities decreased to 10-15% and 1 -2%, respectively. (c) Se- rologically RclM" and Rcl" are not related (see Fig. 5) .

The most striking difference between these two enzymes, which actually appears to be due to the substitution of Fe for M o in the cofactor, concerns the EPR-spectroscopic properties. No EPR signal has been described to date that could be unequiv- ocally attributed to the FeFe cofactor. Two signals ( S = 3/2, S = 1/2) reported earlier in the literature 141, were both found not to represent the intact enzyme. The most active RclF' sample was EPR-silent in the dithionite-reduced state. This is i n line with the suggestion that, based on theoretical considerations, the seven Fe centers in the FeMo cofactor consist formally of four Fe(I1) and three Fe (111) [47]. Assuming an EPR-silent, integer spin state for the FeFe cofactor, presumably resulting from anti- ferromagnetic coupling, the additional Fe center replacing the Mo atom should be an Fe(II1) one leading to an equal number of ferric and ferrous Fe centers within the cofactor [5, 471. It therefore meets the expectations that the FeFe protein can be converted into EPR-detectable states by oxidation (g-values : 2.27, 2.06; potential region: -100 mV to +150 mV) and also by the use of steady-state turnover conditions (at g = 1.96, 1.92, 1.77). These are signals that have so far not been described for any of the existing types of nitrogenase. From signal shape and g-value position they significantly deviate from P-cluster signals [39, 401 and other types of S = signals in the g = 1.8-2.1 region observed with Mo- and V-nitrogenases [3, 401. The detected signals may actually arise from the protein-bound FeFe cofactor. However, a definite assignment is not yet possible and must await future studies (a) on the isolated FeFe cofactor and (b) on the cofactor-less FeFe apoprotein from a n$Y mutant.

Preliminary Mossbauer studies on the dithionite-reduced FeFe protein revealed spectra similar to those published for the VFe protein (Avl") [50] (Krockel, M., Krahn, E., Schneider, K.,

800 Schneider et al. ( E m J. Biochem. 244)

Trautwein, A. X. and Miiller, A., unpublished results). Moss- bauer measurements on "Fe-enriched RclFe samples, which were subject to redox treatments analogously applied in the EPR study, are currently under way. From the data w e hope to gain additional information for a conclusive differentiation of the cofactor and the P cluster in this enzyme.

This work was financially supported by the Deutsche Forschungs- gemeinschaft. We wish to thank Prof. D. Gatteschi and Dr W. R. Hagen for helpful discussions, Dr A. Radunz for the cooperation in the immunological work and Prof. A. X. Trautwein and M. Krockel for measuring the Mossbauer spectra.

REFERENCES 1. Stiefel, E. I., Coucouvanis, D. & Newton, W. E. (1993) Molybdenum

enzynzes, cofactors und tnodel systems, American Chemical Soci- ety, Washington, D. C.

2. Dean, D. R., Bolin, J. T. & Zheng, L. (1993) J. Bacteriol. 175, 6737-6744.

3. Eady, R. R. (1991) Adv. lnorg. Chem. 36, 77-102. 4. Muller, A,, Schneider, K., Knuttel, K. & Hagen, W. R. (1992) FEBS

5 . Schneider, K., Gollan, U., Selsemeier-Voigt, S., Plass, W. & Muller,

6 . Pau, R. N., Eldridge, M. E., Lowe, D. J., Mitchenall, L. A. & Eady,

7. Eady, R. R., Robson, R. L., Richardson, T. H., Miller, R. W. &

8. Hales, B. J., Case, E. E., Morningstar, J . E., Dzeda, M. F. & Mau-

9. Chisnell, J. R., Premakumar, R. & Bishop, P. E. (1988) J. Bacteriol.

10. Schneider, K., Muller, A., Schranim, U. & Klipp, W. (1991) Eur: J.

11. Davis, R., Lehmann, L., Petrovich, R., Shah, V. K., Roberts, G. P. &

12. Meyer, J., Gaillard, J. & Moulis, J.-M. (1988) Biochemistry 27, 6150-6156.

13. Georgiadis, M. M., Komiya, H., Chakrabarti, P., Woo, D., Kornuc, J. J. & Rees, D. C. (1992) Science 257, 1653-1659

14. Robson, R. L., Woodley, P. R., Pau, R. N. & Eady, R. R. (1989) EMBO J. 8, 1217- 1224.

15. Schuddekopf, K., Hennecke, S., Liese, U., Kutsche, M. & Klipp, W. (1993) Mol. Microhiol. 8, 673-684.

16. Imperial, J., Hoover, T. R., Madden, M. S., Ludden, P. W. & Shah, V. K. (1989) Biochemistn 28, 7796-7799.

17. Kim, J. & Rees, D. C. (1992) Science 257, 1677-1682. 18. Kim, J. & Rees, D. C. (1992) Nature 360, 553-560. 19. Chan, M. K., Kim, J. & Rees, D. C. (1993) Science 260, 792-794. 20. Bolin, J. T., Campobasso. N., Muchmore, S. W., Morgan, T. V. &

Mortenson, L. E., (1993) in Molybdenum enzymes, cqfactors and model systems (Stiefel, E. I., Coucouvanis, D. & Newton, W. E., eds) pp. 186- 195, American Chemical Society, Washington, DC.

Lett. 303, 36-40.

A. (1 994) Naturwi.wenschc$en 81, 405 -408.

R. R. (1993) Biochem. 1. 293, 101-107.

Hawkins, M. (1987) Biochem. J. 244, 197-207.

terer. L. A. (1986) Biockemistry 25, 7251-7255.

170, 27-33.

Biochem. 195, 653-661.

Ludden, P. W. (1996) J. Bacteriol. 178, 1445- 1450.

21. Gollan, U., Schneider, K., Muller, A,, Schuddekopf, K. & Klipp, W.

22. Luque, F. & Pau, R. N. (1991) Mol. & Gen. Genet. 227, 481-487. 23. Lehmann, L. J . & Roberts, G. P. (1991) J. Bacteriol. 173, 5705-

24. Muller, A,, Schneider, K., Gollan, U., Krahn, E. & Drottboom, M.

25. Wang, G., AngermukT, S. & Klipp, W. (1993) 1. Bucteriol. 175,

26. Johann, S. & Hinton, S. M. (1987) J . Bucteriol. 169, 1911-1916. 27. Schneider, K., Miiller, A,, Johannes, K.-U., Diemann, E. & Kott-

mann, J. (1991) Anal. Biochem. 193, 292-298. 28. Laemmli, U. K. (1970) Nature 227, 680-685. 29. Chaykin, S. (1969) And. Biochem. 31, 375-382. 30. Eady. R. R., Smith, B. E., Cook, K. A. & Postgate, J. R. (1972)

Biochem. J. 128, 655-675. 31. Schmidt, K., Liaanen-Jensen, S. & Schlegel, H. G. (1963) Arch.

Microbiol. 46, 117 - 126. 32. Beisenherz, G., Bolze, H. J., Biicher, T., Czok, R., Garbade, K. H.,

Meyer-Arendt, E. & Pfleiderer, G. (1953) Z. Nuturjorsch. Sect. B Chem. Sci. 8, 555-577.

33. Hornhardt, S., Schneider, K., Friedrich, B., Vogt, B. & Schlegel, H. G. (1990) EUK J. Biochem. 189, 529-537.

34. Ludden, P. W. & Roberts, G. P. (1989) Cum Top. Cell. Regul. 30,

35. Hallenbeck, P. C., Meyer, C. M. & Vignais, P. M. (1982) J. Bacte-

36. Dilworth, M. J., Eady, R. R. & Eldrige, M. E. (1988) Biochem. J.

37. Pham, D. N. & Burgess, B. K. (1993) Biochemistry 32, 13725-

38. Krahn, E., Schneider, K. & Muller. A. (1996) Appl. Microbiol. Bio-

39. Tittsworth, R. C. & Hales, B. J. (1993) J. Am. Chem. Sue. 115, 9763 -9767.

40. Pierek, A. J . , Wassink, H., Haaker, H. & Hagen, W. R. (1993) Eur: 1. Biochem. 212, 51-61.

41. Burgess, B. K. (1993) in Molybdenum enzymes, cofactors and model systems (Stiefel, E. I., Coucouvanis, D. & Newton, W. E., eds) pp. 144- 169, American Chemical Society, Washington, DC.

42. Smith, B. E., Lowe, D. J. & Bray, R. C. (1973) Biochern. J. 135,

43. Hallenbeck, P. C. (1991) Biochim. Biophys. Acta 1057, 97-101. 44. Thornley, R. N. F., Eady, R. R. & Yates, M. G. (1975) Biockim.

45. Shah, V. K., Allen, J. R., Spangler, N. J. & Ludden, P. W. (1994) 1.

46. Miiller, A. & Krahn, E. (1995) Angew. Ckem. Int. Ed. Engl. 34,

47. Plass, W. (1994) J . Mol. Struct. (Theochem.) 315, 53-62. 48. Schneider, K., Miiller, A,, Krahn, E., Hagen, W. R., Wassink, H. &

Kniittel, K. (1995) Eur: J . Biochem. 230, 666-675. 49. Lowe, D. J.. Thorneley, R. N. F. & Smith, B. E. (1985) in Metallo-

proteins (Harrison, P., ed.) vol. 1, pp. 207-249, Macmillan, Lon- don.

50. Ravi, N., Moore, V., Lloyd, S. G., Hales, B. J. & Huynh, B. H. (1994) J. Biol. Chem. 269, 20920-20924.

(1993) ELK J. Biochem. 215, 25-35.

5711.

(1995) J. Inorg. Biochem. 59, 551.

3031-3042.

26-56.

rial. 149, 708-717.

249, 745-151.

13731.

technol. 46, 285-290.

331 -341.

Bioplzy.7. Acta 403, 269-284.

Bid. Chem. 269, 1154-1158.

1071-1078.

comparative biochemical characterization of the iron-only nitrogenase and the molybdenum nitrogenase...

Documents