Proc. Nati. Acad. Sci. USAVol. 89, pp. 6487-6491, July 1992Biochemistry

Plausible structure of the iron-molybdenum cofactor of nitrogenase(homocitrate/bicubane)

MARK S. MADDEN, ANDRZEJ M. KREZEL, RONDA M. ALLEN, PAUL W. LUDDEN*, AND VINOD K. SHAHDepartment of Biochemistry and Center for the Study of Nitrogen Fixation, College of Agricultural and Life Sciences, University of Wisconsin-Madison,Madison, WI 53706

Communicated by Robert Burris, March 20, 1992 (received for review February 6, 1992)

ABSTRACT A plausible structure of the iron-molybdenumcofactor of nitrogenase [reduced ferredoxin:dinitrogen oxidore-ductase (ATP-hydrolyzing), EC 1.18.6.1] is presented based onaltered substrate reduction properties of dinitrogenase contain-ing homocitrate analogs within the cofactor. Alterations on eachcarbon of the four-carbon homocitrate backbone were corre-lated with altered substrate reduction properties of dinitroge-nase containing these analogs. Altered substrate reduction prop-erties are the basis for a model in which homocitrate is orientedabout two cubane metal clusters.

Biological systems utilize nitrogenase (EC 1.18.6.1) to cata-lyze the ATP- and reductant-dependent reduction of N2 toammonium. Nitrogenase consists of two proteins: dinitroge-nase (MoFe protein, or component I) and dinitrogenasereductase (Fe protein, or component II) (1, 2). Each electrontransferred to dinitrogenase by dinitrogenase reductase iscoupled to the hydrolysis of two MgATP molecules (3).Electrons passed to dinitrogenase are channeled to a uniqueprosthetic group called the iron-molybdenum cofactor(FeMo-co) (4, 5), which is the ultimate site of substratereduction.FeMo-co has been the subject of numerous elemental

analyses and the latest data indicate a Mo/Fe/S/homocitrateratio of 1:7 ± 1:8 ± 1:1 (6, 7). FeMo-co has been subjectedto spectroscopic analyses, both in dinitrogenase and in iso-lated forms (8-16). Extended x-ray absorption fine structure(EXAFS) data indicate that FeMo-co has two or three ironatoms linked to molybdenum via bridging sulfur atoms andthat the remaining iron atoms are more distant (10-14).EXAFS data also suggest two or three light (0, N) ligands tomolybdenum. How FeMo-co is coordinated to apodinitroge-nase is not known (17).

Since the isolation of FeMo-co, significant contributionshave been made toward the synthesis ofFe-Mo-S complexesas possible analogs ofthe active site of nitrogenase. Althoughthe "linear" and "cubane" type of clusters reported earlier(18, 19) do not possess Fe/Mo/S ratios consistent with therange of ratios reported for FeMo-co, cubane clusters haveMo-S distances comparable to those observed by EXAFS inFeMo-co. Cubane clusters also contain Mo atoms in anoxidation state of 3+ to 4+, which is in agreement withexternal nuclear double resonance (ENDOR) studies (16). Onthe other hand, linear clusters contain Mo in a relatively highoxidation state (Mo5+, Mo6+) and shorter Mo-S distances.Recently, S-bridged Mo-bicubane structures with catecholligands have been synthesized as possible analogs ofFeMo-co (20-22), and the authors suggested the feasibility ofassembling MoFe3S4 and Fe4S4 units into S-bridged double-cubane clusters. The mixed-cubane clusters possess an Fe/Mo/S stoichiometry similar to that of FeMo-co, with at leastfour distinct iron sites compatible with ENDOR results of

FeMo-co (15). Doubly Mo-capped prismane clusters havebeen synthesized (23), but these clusters contain Fe/Moratios of 6:2. Mono-capped species were not isolated, eventhough these species are suggested to be in equilibrium withthe di-capped cluster in solution.

In spite of extensive efforts by many groups, no functionalanalog of FeMo-co capable of occupying the active site ofnitrogenase has been synthesized. FeMo-co contains 1 mol ofhomocitrate per Mo (7), however, no FeMo-co analog syn-thesized or suggested to date includes homocitrate.An in vitro system for the synthesis of FeMo-co has been

developed that requires molybdate, MgATP, dithionite, atleast the nijR, nifN, and niJE gene products (24), dinitroge-nase reductase (25-27), and homocitrate (7, 28). The in vitroFeMo-co synthesis system allows the incorporation of ho-mocitrate and its analogs (i.e., citrate, isocitrate, and ho-moisocitrate) to produce modified forms ofFeMo-co (29-31).The resulting aberrant forms of FeMo-co exhibit alteredsubstrate specificities and inhibitor susceptibilities and thesix-electron reduction of N2 by nitrogenase was dramaticallyaffected by substitution of any other organic acid for homo-citrate in the in vitro FeMo-co synthesis system. These datasuggest that none of the FeMo-co analogs tested wouldproduce a dinitrogenase capable of supporting significantdiazotrophic growth.

Imperial et al. (30) described the structural requirements ofan organic acid moiety of FeMo-co competent for substratereduction: (i) stereochemistry at the chiral C-2 carbon anal-ogous to that of the R isomer of homocitrate; (ii) a hydroxylgroup on the chiral carbon atom; (iii) a carboxyl group on thechiral carbon atom; (iv) a carboxyl group a to the chiralcenter; and (v) a four- to six-carbon chain containing twoterminal carboxyl groups.To test this model further, the two diastereomers of

fluorohomocitrate were utilized with a range of substrates(e.g., cyanide) and inhibitors (e.g., carbonyl sulfide) of ni-trogenase (31). Dinitrogenase activated with FeMo-co syn-thesized using threo-fluorohomocitrate reduces protons, cy-anide, and acetylene but cannot reduce N2. In addition,proton reduction is inhibited by carbon monoxide (CO), acharacteristic of dinitrogenase from NifV- mutants. Dinitro-genase activated with FeMo-co synthesized using erythro-fluorohomocitrate reduces protons, cyanide, acetylene, andN2. In this case, proton reduction is not inhibited by CO, acharacteristic of the wild-type enzyme.

In this paper we further investigate the properties ofdinitrogenase containing homocitrate analogs in which thechirality of the C-2 carbon has been altered [(R)-citroylfor-mate, and (S)-citroylformate] or the C-2 hydroxyl group hasbeen altered (3-carboxyglutamate), as well as alterations ofthe C-4 carbon R(R)-citroylformate]. The cumulative infor-mation on the substrate reduction properties and inhibitorsusceptibilities of dinitrogenase containing homocitrate ana-

Abbreviations: FeMo-co, iron-molybdenum cofactor; EXAFS, ex-tended x-ray absorption fine structure.*To whom reprint requests should be addressed.

6487

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

6488 Biochemistry: Madden et al.

logs provides insight into the functional groups on homoci-trate required for an active form of FeMo-co. This knowledgehas been rationalized toward the development of a plausiblestructural model of FeMo-co.

MATERIALS AND METHODSNitrogenase Assays. All assays (with each form of dinitro-

genase) were performed in the presence of excess dinitroge-nase reductase. Acetylene and proton reduction assays (24)with or without CO (29) have been described. N2 reductionwas assayed by the 15N2 fixation method (32) using 99% 15N2added after preincubation. Bacterial strains, preparation ofcrude extracts, and sources ofhomocitrate analogs have beendescribed (29-31). threo-Isocitrate and 3-carboxyglutamatewere obtained from Sigma. Chlorocitrate was obtained fromHoffmann-La Roche. (R,S)-Citroylformate was synthesizedby the method of Wiley and Kim (33). (R)-Citroylformate,(S)-citroylformate, and 3-hydroxyglutamate were gifts ofKiyofumi Maruyama (Gifu University, Gifu, Japan). threo-Fluorohomocitrate and erythro-fluorohomocitrate were syn-

thesized as described (31).FeMo-co Models. All available chemical and biochemical

evidence was converted into internal coordinates (bondlengths, bond angles, dihedral angles, and distances betweenatoms) (11, 34). Models compatible with these coordinateswere created and minimized with metric matrix distancegeometry algorithms of the DSPACE program (Hare Research,Woodinville, WA). Additional manual manipulations were

performed with the molecular modeling program INSIGHTII(Biosym Technologies, San Diego).

RESULTS AND DISCUSSIONIn preliminary experiments, various concentrations of ho-mocitrate analogs were tested for their ability to yield recon-

stituted dinitrogenase competent for acetylene reduction.The lowest concentration of organic acid that generatedmaximum acetylene reduction was chosen for the assayspresented in Table 1. Substrate reduction properties of dini-trogenase activated with some homocitrate analogs havebeen reported (30, 31). The experiments have been repeatedfor this report so that all results could be compared. Dini-trogenase capable of proton reduction can be formed in vitrowith many analogs ofhomocitrate incorporated into FeMo-co(29-31). However, the ability to reduce acetylene was moresensitive to modifications of the homocitrate structure, andonly homocitrate, erythro-fluorohomocitrate, or R-citroyl-formate was able to produce a dinitrogenase capable ofsignificant N2 reduction (Table 1).What follows is a discussion concerning alterations on each

carbon of the four-carbon homocitrate backbone and thecorresponding effects on the properties of dinitrogenasecontaining homocitrate analogs with these alterations. Amodel of FeMo-co based on properties of dinitrogenasescontaining homocitrate analogs is presented in Fig. 1. Thismodel does not include contributions from the protein hi-gands, which are of obvious importance. In this model,

Table 1. Substrate specificity of dinitrogenase activated in vitro by FeMo-co synthesized in thepresence of various organic acids

Proton reduction Acetylene reduction N2 reductionOrganic acid* Conc., mM activitytt activityt activity§

None - 0.7 (0) 0.5 0.1Homocitrate 0.08 28 (0) 27 6.8erythro-Fluorohomocitrate 0.16 21 (0) 16 1.9threo-Fluorohomocitrate 0.16 17 (59) 14 0.2Homoisocitrate 0.80 23 (0) 2.3 0.1Isocitrate 4.0 24 (0) 6.2 0.2threo-Isocitrate 8.0 8.1 (55) 2.4 ND(R)-Citroylformate 0.16 20 (0) 17 3.1(S)-Citroylformate 0.16 6.0 (54) 2.7 0.3(R,S)-Citroylformate 0.16 19 (0) 17 3.4(R)-Citramalate 2.0 15 (0) 11 0.5(S)-Citramalate 1.6 3.3 (70) 2.4 0.1D-Malate 8.0 8.6 (0) 4.3 0.3L-Malate 8.0 7.9 (0) 2.2 0.2Citrate 8.0 11 (59) 10 0.4Chlorocitrate 8.0 11 (55) 10 0.43-Hydroxyglutamate 8.0 12 (0) 6.8 0.23-Carboxyglutamate 8.0 4.0 (0) 0.54 ND

FeMo-co synthesis mixtures contained 0.2 ml of desalted UW45 extract (4 mg of protein) and 0.05ml of partially purified NifB protein from Klebsiella pneumoniae strain UN1217 (30, 31), along with 0.2ml of an ATP-regenerating mixture that contained 5 mM sodium dithionite and 0.05 mM sodiummolybdate and the indicated concentrations oforganic acids. Reaction mixtures were incubated at 300Cfor 35 min, after which 0.8 ml of additional ATP-regenerating mixture containing 5 mM sodiumdithionite was added, together with purified dinitrogenase reductase (24), and the activity was assayed.ND, not determined. Addition of excess purified FeMo-co to UW45 extracts used in these experimentsresulted in an activity of 44.9 nmol of ethylene formed per minute per assay.*Homocitrate lactone (Sigma) was converted to the free acid by adjusting the solution to pH 10 with4 M NaOH. The solution was stored overnight at 5°C and the pH was adjusted to 8 before use.Solutions of homocitrate analogs were prepared in dilute NaOH at a final pH of 8. Concentrations arethose present during FeMo-co synthesis.tExpressed as nmol ofhydrogen (proton reduction) or ethylene (acetylene reduction) formed per minuteper assay.tValues in parentheses show percent inhibition by CO. One hundred microliters ofCO gas was injectedinto an assay vial containing an 8-mil gas phase.§Expressed as nmol of 15N2 reduced minute per assay. N2 reduction assays were carried out for 1 hr.Atom % 15N excess in the samples ranged from 0.0030 to 0.1884 (32).

Proc. Natl. Acad. Sci. USA 89 (1992)

Proc. Natl. Acad. Sci. USA 89 (1992) 6489

FIG. 1. Plausible model of FeMo-co based on substrate reductionproperties of dinitrogenase containing homocitrate analogs.

homocitrate links two cubane clusters, as opposed to sulfidelinking the two cubanes as in the structure suggested byCoucouvanis and coworkers (20, 22).The C-1 Carbon: Substrate Specificities of Dinitrogenase

Activated by FeMo-co Synthesized with Fluorohomocitrate.Dinitrogenase activated with FeMo-co synthesized by usingeither fluorinated diastereomer of homocitrate reduces pro-tons 60-74% and acetylene 51-59o as effectively as homo-citrate-containing dinitrogenase (Table 1). On the other hand,N2 reduction by dinitrogenases activated with FeMo-cosynthesized by using the two fluorinated diastereomers isdramatically different. Dinitrogenase activated withFeMo-co containing threo-fluorohomocitrate is ineffective inN2 reduction. In contrast, dinitrogenase activated withFeMo-co containing erythro-fluorohomocitrate is capable ofN2 reduction 27% that of dinitrogenase activated with ho-mocitrate-derived FeMo-co. The size of the fluorine atom isvery similar to that of a hydrogen atom attached to carbon;therefore, steric effects from the fluorine atom of fluoro-homocitrate are probably not causing the dramatic differ-ences in N2 reduction. However, the three electron pairs ofthe fluorine atom may be contributing toward altered sub-strate specificities of the dinitrogenases activated withFeMo-co containing fluorinated analogs of homocitrate.

Fig. 1 shows how substrate reduction properties of dini-trogenase might be affected by erythro vs threo placement offluorine atoms on the C-1 carbon of homocitrate. Fluorineplacement at the threo position may interfere with the pro-posed site for substrate (i.e., N2) binding and reduction dueto the close proximity of the three electron pairs of thefluorine atom. Placement of a fluorine atom at the erythro sitepositions it away from the proposed substrate binding site,and thus it would not interfere as strongly as the threofluorine atom. It is also possible that a hydrogen atom at thethreo position of the C-1 carbon of homocitrate might beinvolved in proton donation and therefore placement of afluorine atom on the C-1 carbon might affect proton donation.The C-2 Carbon: Substrate Specificities of Dinitogenase

Activated by FeMo-co Synthesized with Citroylformate, Cit-ramalate, or Malate. The R or S isomers of citroylformate (ahomocitrate analog in which the CA methylene carbon ofhomocitrate is replaced with a carbonyl carbon) also were

incorporated into dinitrogenase. Table 1 shows that dinitro-genase containing R-citroylformate and S-citroylformate willreduce N2 45% and 3.5% as effectively as homocitrate-containing dinitrogenase, respectively. In addition, dinitro-genase containing R-citroylformate or S-citroylformate willreduce acetylene 63% and 10% as well as homocitrate-containing dinitrogenase, respectively. Proton reduction ac-tivity of dinitrogenase containing S-citroylformate is inhib-

ited by CO, whereas no CO inhibition of proton reduction isobserved for dinitrogenase containing R-citroylformate.

Dinitrogenase activated with the racemic mixture of cit-roylformate reduces N2, acetylene, and protons similarly tothe R-citroylformate-containing dinitrogenase, and protonreduction activity is not inhibited by CO. If the FeMo-cosynthesis system is unable to discriminate between the twoisomers, then CO inhibition of proton reduction by dinitro-genase activated by FeMo-co synthesized with the racemicmixture should be -27%. Since similar properties are ob-served when dinitrogenase is activated with FeMo-co syn-thesized with either R-citroylformate or the racemic mixture,it seems that R-citroylformate is preferentially incorporatedinto dinitrogenase. The R isomer of homocitrate seems to berequired for the synthesis of active FeMo-co. Only the Risomer of homocitrate is synthesized in biological systems(28). Imperial et al. (30) have shown that the substratereduction properties of dinitrogenase containing R-citra-malate or D-malate are markedly different from those ofdinitrogenases containing the corresponding S-citramalate orL-malate isomers. Dinitrogenase containing R-citramalate orS-citramalate reduces acetylene 40%o and 7% as effectively ashomocitrate-containing dinitrogenase, respectively (Table 1).

Fig. 1 shows FeMo-co containing R-homocitrate, whichpositions the C-2 hydroxyl group away from the metal-boundN2 molecule. However, analogous coordination using S-ho-mocitrate would place the hydroxyl group close to thesubstrate binding site. Such close proximity of the hydroxylgroup could explain the dramatic differences between dini-trogenases activated with R and S analogs of homocitrate.The C-3 Carbon: Substrate Specificities of Dinitrogenase

Activated by FeMo-co Synthesized with Chlorocitrate. Sub-strate reduction properties of dinitrogenase containing chlo-rocitrate (Table 1) are identical to those of the citrate-containing enzyme. Citrate is a prochiral molecule, whereaschlorocitrate has two chiral centers; therefore, four arrange-ments of the chlorocitrate molecule within the FeMo-costructure are possible. The chlorine atom might be positionedon the C-1 carbon at either the erythro or threo position;however, dinitrogenase activated with FeMo-co containingthreo-chlorocitrate should show weaker substrate reductionproperties than citrate-containing dinitrogenase. Placementof a bulky chlorine atom at the threo position might interferewith substrate binding and reduction as previously discussed.Therefore, the chlorine atom would more likely be positionedon the C-1 carbon at the erythro position, where it should notinterfere with substrate binding and/or reduction. However,a preferable position for the chlorine atom is at the C-3carbon. Note that the C-3 carbon places its substituents wellaway from the substrate reduction site; therefore, this wouldbe the preferred arrangement for chlorocitrate when coordi-nating the metal clusters. Structural requirements of homo-citrate analogs for incorporation into FeMo-co are the leaststringent in the tail region (i.e., alterations on C-3 and C-4)(30).The C-4 Carbon: Substrate Specificities of Dinitrogenase

Activated by FeMo-co Synthesized with Citroylformate andCitrate. Homocitrate and R-citroylformate differ only withrespect to substituents placed on the C-4 carbon. Homoci-trate has a methylene (-CH2-) moiety whereas citroylformatehas a carbonyl (-CO-) moiety. Substrate reduction propertiesof dinitrogenase containing R-citroylformate or homocitrateare similar (Table 1). Both dinitrogenases have high protonreduction activities that do not suffer from CO inhibition. Inaddition, dinitrogenases containing R-citroylformate will re-duce N2 and acetylene 45% and 63% as effectively as ho-mocitrate-containing dinitrogenase, respectively.

Dinitrogenase containing erythro-fluorohomocitrate exhib-its substrate reduction properties similar to those of R-cit-roylformate-containing dinitrogenase. The arrangement of

Biochemistry: Madden et al.

6490 Biochemistry: Madden et al.

homocitrate around the metal clusters of FeMo-co shown inFig. 1 places the erythro fluorine atom of erythro-fluorohomocitrate and the carbonyl oxygen atom of R-cit-roylformate in the same region, when these analogs aresubstituted for homocitrate. The staggered conformation ofthe tail region of homocitrate allows the substituents of C-4to be in similar proximity to that of the erythro atom on C-1.

Citrate lacks the C-4 carbon of homocitrate. Therefore,when citrate is arranged around the two cubanes theFeMo-co molecule becomes distorted. Citrate may rearrangearound the clusters to place the C-1 carbonyl oxygen atom(and the electron pairs) in close proximity to the substratebinding site. Similarly, this model predicts that FeMo-cocontaining threo-fluorohomocitrate places the fluorine atom(and its electron pairs) in close proximity to the proposedsubstrate binding site. In addition, proton reduction bydinitrogenase containing either citrate or threo-fluorohomo-citrate is inhibited 59% by CO.

Role for the Hydroxyl Group of Homocitrate or Its Analogs.The hydroxyl group is required for FeMo-co synthesis (29,30). Both homocitrate and citrate contain a C-2 hydroxylgroup, whereas homoisocitrate and isocitrate contain thehydroxyl group on the C-1 carbon. Dramatic differences insubstrate reduction properties are observed between dinitro-genases containing homocitrate or homoisocitrate (as well ascitrate or isocitrate). This may be due to an altered orienta-tion of FeMo-co when attached to apodinitrogenase, sincethe hydroxyl group may be required to anchor the cofactor tothe protein.

Dinitrogenase containing 3-hydroxyglutamate has signifi-cant acetylene and proton reduction activities, with protonreduction not inhibited by CO. However, dinitrogenase con-taining 3-carboxyglutamate (in this case the hydroxyl groupis replaced with a carboxyl group) exhibits poor substratereduction activity (Table 1). Therefore, the C-2 hydroxylgroup of homocitrate analogs cannot be replaced with acarboxyl group to form a functional FeMo-co. In this model,the best fit to the DSPACE molecular modeling parameters isobtained when the hydroxyl group is positioned out of theligation sphere of molybdenum or iron. This suggests anotherrole, such as binding to the protein, for the hydroxyl group.Models Containing a Five-Membered Ring Ligating Molyb-

denum. The seven-membered ring model of FeMo-co (Fig. 1)shows the C-1 and C-2 carboxyl groups of homocitratecoordinated to the molybdenum atom of one cluster while theC-4 carboxyl group is coordinated to one Fe of the othercluster. Seven-membered ring structures of tricarboxylates(e.g., citrate and isocitrate) coordinated to an octahedral (Fe)atom have been reported (34). However, a five-memberedring can be constructed that coordinates the C-2 hydroxyl andcarboxyl groups of homocitrate to the molybdenum center(not shown). This five-membered ring structure coordinatesthe C-1 carboxyl group to the 4Fe-4S cluster and can positionthe threo hydrogen close to the substrate binding site, or itcan arrange to interfere with the carbonyl oxygen of thefive-membered ring. Evidence favoring the five-memberedring over the seven-membered ring includes the following: (i)five-membered rings are entropically more favorable thanseven-membered rings; (ii) other organic acid/metal clustermodels exist showing five-membered rings containing car-boxyl-metal-hydroxyl coordination (36); and (iii) isolatedFeMo-co contains a net negative charge, which may be dueto a free carboxyl group. However, this five-membered ringmodel is unable to explain how alterations in the tail (i.e., C-3and C-4) region of homocitrate affect the substrate reductionproperties of dinitrogenase containing the correspondinghomocitrate analogs. In addition, this model cannot explainthe effects of C-1 hydroxyl acids when incorporated intoFeMo-co. Therefore, Fig. 1 seems the best representation of

FeMo-co based on substrate reduction properties ofFeMo-co containing homocitrate analogs.A model with a five-membered ring structure, but coordi-

nating the C-4 carboxyl group to the 4Fe-4S cluster, was alsoconstructed (not shown). Although this model does notexplain differences in substrate reduction data for substitu-ents placed on the C-1 carbon (erythro- vs. threo-fluorohomocitrate), it may explain how alterations in the tailregion of homocitrate affect substrate reduction properties.Energy differences between this model and the five-membered ring model discussed above are quite small, andthis indicates that substitution of citrate for homocitrate inthis model would have minor effects. Therefore, the dramaticdifferences in N2 reduction observed between homocitrate-containing dinitrogenase and citrate-containing dinitrogenasedo not make this model plausible. The dramatic differencesin substrate reduction properties of dinitrogenase containingR vs. S analogs of homocitrate also may rule out the possi-bility ofa five-membered homocitrate-metal ring in FeMo-co.A Pentlandite Model of FeMo-co. Pentlandite models of

FeMo-co (22, 37) possess a reasonable Fe/Mo/S stoichiom-etry and are in fair agreement with much of the spectroscopicdata acquired on FeMo-co. An attractive feature of the model(not shown) is a five-membered Mo ring that includes the C-2carboxyl and hydroxyl groups of homocitrate. However,negative features of the pentlandite model include (i) anine-membered ring containing iron and homocitrate, (ii) nobinuclear (Mo-substrate-Fe) interaction as previously pro-posed (35), and (iii) no easy explanation for the dramaticdifferences between dinitrogenase containing homocitrateanalogs possessing a C-1 fluorine (or hydroxyl group) at theerythro vs. threo position. According to this model, FeMo-cocontaining citrate would have an eight-membered citrate-Fering, whereas homocitrate would have a nine-membered ring.Even though the eight-membered citrate-Fe ring is entropi-cally favorable, substrate reduction activities of dinitroge-nase containing this analog are substantially lower than thedinitrogenase containing homocitrate. However, it is possi-ble that shortening the tail region of homocitrate by onecarbon may move the C-1 carboxyl group away from thesubstrate binding site in dinitrogenase containing citrate.The pentlandite model does not explain differences in

substrate reduction properties of dinitrogenase containing(R)- vs. (S)-citroylformate. In a five-membered ring confor-mation (not shown), (S)-citroylformate would orient in aninverted manner when compared with FeMo-co containing(R)-citroylformate or (R)-homocitrate. Since both isomerscontain four-carbon backbones with two terminal carboxylgroups, both normal and inverted orientations can effectivelycoordinate one carboxyl group to an Fe atom, with theremaining carboxyl group available for effective proton do-nation. Therefore, proton donation via the C-4 carboxyl ofthe S analog (instead of the C-1 carboxyl of the R analog)should be unaffected. In fact, the more flexible tail regionshould conform more easily to place the carboxyl group nearthe substrate binding site with the S isomer. In addition, theS isomer would give an eight-membered Fe ring, which ismore desirable than the nine-membered ring associated withR isomers. However, poor substrate reduction is observedwith (S)-citroylformate compared with (R)-citroylformate.Therefore, the dramatic differences in substrate reductionobserved between dinitrogenase containing (R)- vs. (S)-citroylformate do not make this model highly plausible.

CONCLUSIONThe bicubane model presented here is based on (i) metricmatrix distance geometry algorithms of the DSPACE program(Hare Research), (ii) spectroscopic (e.g., EXAFS) data ac-quired on FeMo-co (11, 12), and (iii) the substrate reduction

Proc. Natl. Acad. Sci. USA 89 (1992)

Proc. Natl. Acad. Sci. USA 89 (1992) 6491

properties of dinitrogenases containing homocitrate analogs.Although other possible models presented did not fit theconstraints imposed by the altered substrate reduction prop-erties of dinitrogenases containing homocitrate analogs, thedata presented here should stimulate future modeling andsynthesis of the active site of nitrogenase. The fact that somany organic acids can be incorporated to make FeMo-coanalogs should provide bioinorganic chemists with manyavenues ofcofactor synthesis and insight into the mechanismof N2 reduction.

Special thanks go to Robert Burris for his valuable help with 15N2experiments and to Mark Gallop, Rip Lee, Larry Dahl, HelmutBeinert, Marie Claire Kennedy, Scott Ensign, Charles Casey, RobertBurris, Gary Roberts, and Timothy Paustian for their helpful dis-cussions. We thank Jeff Allen for technical assistance. This researchwas supported by the College ofAgricultural and Life Sciences at theUniversity of Wisconsin-Madison, by United States Department ofAgriculture Grant 8801471 to P.W.L. and V.K.S., and by NationalInstitutes of Health Grant GM35332 to P.W.L.

1. Bulen, W. A. & LeComte, J. R. (1966) Proc. Natl. Acad. Sci.USA 56, 979-986.

2. Hageman, R. V. & Bums, R. H. (1978) Proc. Natl. Acad. Sci.USA 75, 2699-2702.

3. Ljones, T. & Burris, R. H. (1978) Biochemistry 17, 1866-1872.4. Shah, V. K. & Brill, W. J. (1977) Proc. Natl. Acad. Sci. USA

74, 3249-3253.5. Hawkes, T. R., McLean, P. A. & Smith, B. E. (1984) Biochem.

J. 217, 317-321.6. Nelson, M. J., Levy, M. A. & Orme-Johnson, W. H. (1983)

Proc. Nat!. Acad. Sci. USA 80, 147-150.7. Hoover, T. R., Imperial, J., Ludden, P. W. & Shah, V. K.

(1989) Biochemistry 28, 2768-2771.8. Rawlings, J., Shah, V. K., Chisnell, J. R., Brill, W. J., Zim-

merman, R., Munck, E. & Orme-Johnson, W. H. (1978) J. Biol.Chem. 253, 1001-1004.

9. Huynh, B. H., Munck, E. & Orme-Johnson, W. H. (1979)Biochim. Biophys. Acta 527, 192-196.

10. Cramer, S. P., Gillum, W. D., Hodgson, K. O., Mortenson,L. E., Stiefel, E. I., Chisnell, J. R., Brill, W. J. & Shah, V. K.(1978) J. Am. Chem. Soc. 100, 3814-3819.

11. Conradson, S. D., Burgess, B. K., Newton, W. E., Hodgson,K. O., McDonald, J. W., Rubinson, J. F., Gheller, S. F.,Mortenson, L. E., Adams, M. W. W., Mascharak, P. K.,Armstrong, W. A. & Holm, R. H. (1985) J. Am. Chem. Soc.107, 7935-7940.

12. Conradson, S. D., Burgess, B. K., Newton, W. E., Morten-son, L. E. & Hodgson, K. 0. (1987) J. Am. Chem. Soc. 109,7507-7515.

13. Antonio, M. R., Teo, B. K., Orme-Johnson, W. H., Nelson,M. J., Groh, S. E., Lindahl, P. A., Kauzlarich, S. M. &Averill, B. A. (1982) J. Am. Chem. Soc. 104, 4703-4705.

14. Arber, J. M., Flood, A. C., Garner, C. D., Alasnain, S. S. &Smith, B. E. (1986) J. Phys. Colloq. 47, C8-1159.

15. Hoffman, B. M., Venters, R. A., Roberts, J. E., Nelson, M. &Orme-Johnson, W. H. (1982) J. Am. Chem. Soc. 104, 4711-4712.

16. Venters, R., Nelson, M. J., McLean, P. A., True, A. E., Levy,M. A., Hoffman, B. M. & Orme-Johnson, W. H. (1986) J. Am.Chem. Soc. 108, 3487-3498.

17. Dean, D. R., Scott, D. J. & Newton, W. E. (1990) in NitrogenFixation: Achievements and Objectives, eds. Gresshoff, P. M.,Stacey, G. & Newton, W. E. (Chapman & Hall, New York),pp. 95-102.

18. Coucouvanis, D. (1981) Acc. Chem. Res. 14, 201-209.19. Christou, G., Mascharak, P. K., Armstrong, W. H., Pa-

paefthymiou, G. C., Frankel, R. B. & Holm, R. H. (1982) J.Am. Chem. Soc. 104, 2820-2831.

20. Challen, P. R., Koo, S., Kim, C. G., Dunham, W. R. &Coucouvanis, D. (1990) J. Am. Chem. Soc. 112, 8606-8607.

21. Coucouvanis, D., Challen, P. R., Koo, S., Davis, W. M.,Butler, W. & Dunham, W. R. (1989) Inorg. Chem. 28, 4183-4186.

22. Coucouvanis, D. (1991) Acc. Chem. Res. 24, 1-8.23. Coucouvanis, D. & Kanatzidis, M. G. (1985) J. Am. Chem.

Soc. 107, 5005-5006.24. Shah, V. K., Imperial, J., Ugalde, R. A., Ludden, P. W. &

Brill, W. J. (1986) Proc. Nat!. Acad. Sci. USA 83, 1636-1640.25. Robinson, A. C., Dean, D. R. & Burgess, B. K. (1987) J. Biol.

Chem. 262, 14327-14332.26. Imperial, J., Shah, V. K., Hoover, T. R. & Ludden, P. W.

(1988) in Nitrogen Fixation: Hundred Years After, eds. Bothe,H., de Bruijn, F. J. & Newton, W. E. (Fischer, New York), p.128.

27. Shah, V. K., Hoover, T. R., Imperial, J., Paustian, T. D.,Roberts, G. P. & Ludden, P. W. (1988) in Nitrogen Fixation:Hundred Years After, eds. Bothe, H., de Bruijn, F. J. &Newton, W. E. (Fischer, New York), p. 115.

28. Hoover, T. R., Robertson, A. D., Cerny, R. L., Hayes, R. N.,Imperial, J., Shah, V. K. & Ludden, P. W. (1987) Nature(London) 329, 855-857.

29. Hoover, T. R., Imperial, J., Liang, J., Ludden, P. W. & Shah,V. K. (1988) Biochemistry 27, 3647-3652.

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

31. Madden, M. S., Kindon, N. D., Ludden, P. W. & Shah, V. K.(1990) Proc. Nat!. Acad. Sci. USA 87, 6517-6521.

32. Bums, R. H. & Wilson, P. W. (1957) Methods Enzymol. 4,355-366.

33. Wiley, R. H. & Kim, K. (1973) J. Org. Chem. 38, 3582-3585.34. Glusker, J. P. (1980) Acc. Chem. Res. 13, 345-351.35. Hardy, R. W. F., Burns, R. C. & Parshall, G. W. (1973) in

Inorganic Biochemistry, ed. Eichhorn, G. L. (Elsevier,Amsterdam), pp. 745-793.

36. Carrell, C. J., Carrell, H. L., Erlebacher, J. & Glusker, J. P.(1988) J. Am. Chem. Soc. 110, 8651-8656.

37. Christou, G., Hagen, K. S. & Holm, R. H. (1982) J. Am. Chem.Soc. 104, 1744-1747.

Biochemistry: Madden et al.