Cross-linking of Nitrogenase Components STRUCTURE AND ACTIVITY OF THE COVALENT COMPLEX*

The nitrogenase complex from Azotobacter vinelan- dii is composed of the MoFe protein (Avl), an az@z tetramer, and the Fe protein ( A V ~ ) , a yz dimer. Dur- ing turnover of the enzyme, electrons are transferred from Av2 to Avl in parallel with the hydrolysis of MgATP. Using the cross-linking reagent, l-ethyl-3-(3-dimethyIaminopropyl)carbodiimide, we have iden- tified some of the properties of the complex between the two components. The cross-linking reaction was highly specific yielding a single apparent M, = 97,000 protein. The amount of cross-linked product was essen- tially independent of whether MgATP or MgADP were in the reaction. Also, the amount was maximum at high ratios of Av2 to Avl. The M, = 97,000 protein was characterized by amino acid analysis and Edman degradation and was found to be consistent with a 1:l complex of an Av2 y subunit and an Avl @ subunit (the amino terminal serine sub- unit). The complex was no longer active in the nitrogenase reaction which supports, but does not prove, the requirement for dissociation of the complex after each electron transferred. Nitrogenase activity and cross-linking were inhibited in an identical way by NaC1, which suggests that electrostatic forces are crit- ical to the formation of the electron transfer complex. Biological dinitrogen reduction is catalyzed by the two protein complex, nitrogenase. The

The nitrogenase complex from Azotobacter vinelandii is composed of the MoFe protein (Avl), an az@z tetramer, and the Fe protein ( A V~) , a yz dimer. During turnover of the enzyme, electrons are transferred from Av2 to Avl in parallel with the hydrolysis of MgATP. Using the cross-linking reagent, l-ethyl-3-(3-dimethyIaminopropyl)carbodiimide, we have identified some of the properties of the complex between the two components. The cross-linking reaction was highly specific yielding a single apparent M, = 97,000 protein. The amount of cross-linked product was essentially independent of whether MgATP or MgADP were in the reaction. Also, the amount was maximum at high ratios of Av2 to Avl.
The M, = 97,000 protein was characterized by amino acid analysis and Edman degradation and was found to be consistent with a 1:l complex of an Av2 y subunit and an Avl @ subunit (the amino terminal serine subunit). The complex was no longer active in the nitrogenase reaction which supports, but does not prove, the requirement for dissociation of the complex after each electron transferred. Nitrogenase activity and cross-linking were inhibited in an identical way by NaC1, which suggests that electrostatic forces are critical to the formation of the electron transfer complex.
Biological dinitrogen reduction is catalyzed by the two protein complex, nitrogenase. The components of the complex are the MoFe protein, a a& tetramer of M, = 225,000, and the Fe protein, a y2 dimer of M, = 63,000 (3, 4). The MoFe protein contains the putative substrate reduction site which is a Fe6-7:Mo:Ss-9 cofactor. For substrate reduction, electrons are transferred from the Fe protein to the MoFe protein concomitant with the hydrolysis of ATP. An essential feature of the accepted model of nitrogenase turnover is the requirement for the Fe protein and the MoFe protein to dissociate after each electron transferred (5, 6). Because the substrates of nitrogenase require one or more pairs of electrons, the complex must undergo multiple cycles of association/dissociation for each mole of substrate reduced. A second prominent aspect of the reaction is the well documented structural change which occurs in the Fe protein upon binding MgATP and MgADP (3, 4). * This work was supported by Grants GM 34321 (to J. B. H.) and GM 31299 (to D. C. R.) from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore he hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact, $ Lynen Fellow of the Alexander-von-Humboldt Foundat,ion (Federal Republic of Germany).
A number of questions about the nature of the complex remain unanswered. For example, what is the symmetry of the complex, i.e. does the docking site on the MoFe protein involve both subunits? Can the complex reduce substrates without physically dissociating? Which amino acid residues are involved in the docking sites? What is the role of nucleotide binding in the docking process? Many of these questions can be evaluated by molecular model building once the crystallographic structures of the two components are known. On the other hand, model building will be greatly enhanced by having prior evidence for the docking sites. In addition some of these questions need to be considered for proteins under dynamic conditions. Towards these goals, we have used chemical cross-linking to study properties of the nitrogenase complex.

RESULTS AND DISCUSSION
A variety of compounds have been screened as potential cross-linking agents for the nitrogenase components; only the water-soluble carbodiimide, EDC,' which cross-links amino and carboxyl groups has proven promising. As is shown in Fig. 1, the EDC cross-linking reaction is highly specific for the nitrogenase complex. A single new protein band (apparent Mr = 97,000) was observed upon gel electrophoresis if both nitrogenase components were present in the reaction. In contrast, no cross-linking between the constituent subunits of the individual nitrogenase components was observed. When creatine phosphokinase ( M , = 45,000) was present in the reaction as part of an ATP regenerating system, it was not cross-linked either to itself or to the nitrogenase components. Furthermore, formation of the M, = 97,000 product appears to require the active functional complex because only nonspecific, high molecular weight cross-linked material was observed with oxygen inactivated components (data not shown). The amount of cross-linking was independent of pH between 7 and 9 and of the EDC concentration from 1 to 50 mM. At higher concentrations of EDC (>lo0 mM), additional very ' Portions of this paper (including "Materials and Methods," Figs. 1-6, and Table 1) are presented in miniprint at, the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press.
Another line of evidence that the cross-linking reaction reflects the formation of an catalytically significant complex is the time dependence of the reaction. The results shown in Fig. 1 also indicate that the cross-linking is a rapid reaction with a tlh of 3 min for the protein concentrations and ratio used in the figure. The reaction had excess EDC and was not limited by this reagent. EDC added after the cessation of cross-linking (30 min) did not increase the yield of M, = 97,000 material.
One aspect of the nitrogenase model is that the component ratio as well as the protein concentration affect the amount of active complex and, thereby, the enzyme turnover. If the cross-linking reaction reflects the formation of the active complex, then the component ratio also might be expected to influence the rate of cross-linking. To investigate this further, the component ratio in the cross-linking reaction was varied from a 10-fold excess of Avl to a 10-fold excess of Av2. The results are shown in Fig. 2. Throughout this 100-fold range of component ratios, only a single cross-linked protein band (MI = 97,000) was observed. As the ratio of Av2:Avl was increased, the amount of M, = 97,000 material increased until the @ subunit of Avl was substantially depleted. Likewise, when the ratio of Av1:AvZ was increased (Fig. 2B), the amount of crosslinked material was limited by Av2 subunits. However, in contrast to the nearly complete loss of the Avl @ subunit, the Av2 subunits were decreased only 50-60%. Although we cannot exclude the possibility that 50% of our Av2 is inactivated by side reactions such that it can no longer react, it seems plausible that only one of the two subunits in each Av2 dimer is being cross-linked, whereas most of the Avl @ subunit is incorporated in the Mr = 97,000 material. This is also consistent with a specific interaction leading to cross-linking.
Perhaps the most surprising result of our studies was that the cross-linking reaction is essentially independent of the presence of nucleotides (see Fig. 3). Not only was the molecular weight and composition of the cross-linked material unrelated to the presence of nucleotides, but the amount of cross-linked material was similar or even somewhat less in the presence of either MgATP or MgADP. What makes this result so striking is that numerous other probes of the Fe protein structure uniformly show significant changes upon binding of nucleotides (3, 4). Under our experimental conditions of saturating nucleotide concentration, it is unlikely that the cross-linking reaction involves the nucleotide binding site or that nucleotide binding is prevented.
Because the EDC cross-linking reaction involves closely opposed carboxylic acids and amines, it is reasonable to assume that these groups may be part of ionic (salt) interactions which, in part, are responsible for the complex formation. Indeed, patches of acidic and basic residues have been postulated for "docking" sites in other electron transfer complexes and these residues can be cross-linked with EDC (e.g. . Studies by Watt and co-workers (15) and by Diets and Howard3 have shown that nitrogenase activity is substantially inhibited by NaCl and other salts. Based upon modeling of the salt inhibition of enzyme activity, Diets and Howard3 have concluded that salt prevents the formation of the Avl-Av2 complex required for enzyme turnover. Fig. 4 shows the results of salt effects on cross-linking of the nitrogenase complex. For comparison, the effect of salt on activity is also given. It is evident that both manifestations of the complex formation are equally inhibited and have a similar dependence on the salt concentration.
T. Diets and J. B. Howard, manuscript in preparation.
The effect of EDC and of the cross-linking reaction on the enzyme activity were investigated and the results are presented in Fig. 5. Just as the individual components were not cross-linked by incubation with EDC, they also were not inhibited. In contrast, the complex was rapidly inactivated by EDC. Furthermore, the inactivation occurred at a rate comparable to that of the cross-linking reaction, cf. results in Figs. 1 and 5 . In accordance with the cross-linking reaction, the rate of inactivation was dependent upon the component ratio but independent of the presence of nucleotides. Thus, the cross-linked material does not appear to have enzyme activity. Although we cannot exclude the possibility that cross-linking prevents ATP hydrolysis or electron transfer (14), our results support the previous hypothesis that the complex must dissociate after each electron transferred (5, 6).
Identification of the number and types of subunits that compose the cross-linked complex has not been as simple as might have been expected for the reasons outlined above. Amino acid analysis (see Table 1) of the isolated complex was compared with theoretical values calculated using various ratios for the known Avl and Av2 subunit compositions. The best fit of the data was for a 1:1 complex of an Avl p subunit and an Av2 y subunit. Because the amino acid compositions for the subunits are so similar (1, 16), unambiquous assignment of the ratio could not be made with confidence by this method.
Amino-terminal sequence analysis by Edman degradation gave a clear identity of the subunits in the M, = 97,000 material. The results are shown in Fig. 6. At each cycle two phenylthiohydantoin-derivatives were observed which corresponded to the Av2 y and Avl @ subunits. Equally importantly, no residues for the Avl a-subunit were found. The calculated initial yields (based on M, = 89,000) were 88 pmol (15%) for the @ subunit and 712 pmol (29%) for the Av2 y subunit, with repetitive yields of 93.6 and 91.2%, respectively. The ratio of the initial yields would suggest a cross-linked complex composed of 0.5 @-subunit and 1.0 y-subunit. However, the low percent initial yield for both amino termini, which is often observed for material isolated using denaturing buffers or material from cross-linking reactions, leaves in doubt how reliable the ratio is when determined by Edman degradation.
Another approach to characterizing the composition of the complex is the analysis of the cross-linking reaction itself. Although molecular weights determined by denaturing gel electrophoresis may give anomalous results for branched (cross-linked) proteins, the molecular weight of 97,000 seems to best fit a ratio of 1:l (Mr = 89,000). Furthermore, at saturating ratios of Av2, most of the Avl @ subunit was converted to the cross-linked complex whereas only -50% of the Av2 subunits were cross-linked at saturating Avl. Since the Av2 subunits do not dissociate or exchange between native molecules,4 the 50% cross-linking of the Av2 subunits is consistent with a 1:1 complex.
Thus, taking all of the results together, we favor the M , = 97,000 cross-linked protein as being a 1:l complex of the Av2 y and Avl @ subunits. It should be emphasized that there is no evidence for the presence of the Avl a subunit in the isolated cross-linked material. Identification of the specific amino acid residues involved in the cross-linking may help to determine the ratio. This is presently under way.
Our results provide several important conclusions about the interaction between the components of the nitrogenase complex, which may be related to the turnover of the enzyme during catalysis. First, at least part of the docking site for Fe D. Ikeda and J. B. Howard, unpublished observations. protein on the MoFe protein is in the (3 subunit. Of course, we cannot exclude a docking site between the two MoFe protein subunits where only the (3 subunit has the chemical cross-linking site. Second, the docking site has some element of asymmetry in that only one of the two identical subunits of the Fe protein is cross-linked. Furthermore, because there is nearly complete incorporation of the (3 subunit, we conclude that the 2-fold symmetry of the MoFe protein (17) provides two separate binding sites and that each binding site interacts with a Fe protein dimer.
Third, a primary component of the docking recognition and binding energy comes from ionic interactions, at least some of which are between amino and carboxyl functions. The lack of cross-linking between subunits of the individual components, i.e. Avl and Av2, suggests that ionic interactions may be more important in component docking than in subunit stabilization.
Fourth, the cross-linked complex appears to be inactive which supports the earlier hypothesis that the components must dissociate after each electron transferred.
Fifth, the individual components were neither inactivated nor cross-linked, whereas the complex was both rapidly crosslinked and inactivated. This is evidence for a specific interaction in the docking process.
Last, the docking of the two components is independent of the conformational changes induced by nucleotides bound to the Fe protein. This is quite significant, because the general belief is that these conformational changes are large if not global. However, whatever the scope of the structural change, it does not alter the region involved in binding to the MoFe protein. Since the docking site and the Fe:S cluster in the Fe protein might be expected to be contiguous for efficient electron transfer, an alternate model for the effect of ATP binding can be suggested. Namely, nucleotide binding could cause reorientation of the subunits with respect to each other without large intrasubunit conformational changes. In addition, if the nucleotide-dependent conformational change is a gating mechanism for electron transfer, then the component docking and electron transfer steps must be uncoupled to some degree.