FeMo maturation on NifEN SPECIAL FEATURE

Yilin Hu*, Mary C. Corbett†, Aaron W. Fay*, Jerome A. Webber*, Keith O. Hodgson†‡§, Britt Hedman‡§, and Markus W. Ribbe*§

*Department of Molecular Biology and Biochemistry, University of California, Irvine, CA 92697-3900; †Department of Chemistry, Stanford University, Stanford, CA 94305; and ‡Stanford Synchrotron Radiation Laboratory, Stanford Linear Accellerator Center, Stanford University, 2575 Sand Hill Road, MS 69, Menlo Park, CA 94025-7015

Edited by Douglas C. Rees, California Institute of Technology, Pasadena, CA, and approved May 10, 2006 (received for review March 31, 2006)

FeMo cofactor (FeMoco) is one of the most compli- Unlike the P cluster, FeMoco, which contains additional hetero- cated processes in metalloprotein biochemistry. Here we show that metal (Mo) and organic moiety (homocitrate), is first assembled on Mo and homocitrate are incorporated into the Fe͞S core of the a scaffold (17, 18) and then inserted into its destined FeMoco precursor while it is bound to NifEN and that the resulting location in MoFe protein (‘‘ex situ’’ assembly). Biosynthesis of fully complemented, FeMoco-like cluster is transformed into a FeMoco presumably starts with the production of the Fe͞Scoreby mature FeMoco upon transfer from NifEN to MoFe protein through NifB (encoded by nifB) (19, 20), which is then transferred to, and direct protein–protein interaction. Our findings not only clarify the further processed on, the ␣2␤2 tetrameric NifEN protein (17, 21). process of FeMoco maturation, but also provide useful insights into Sequence similarity between the respective subunit-encoding genes the other facets of chemistry. led to the proposal that NifE and NifN form a structurally homol- ogous complex to the MoFe protein (22, 23) and that, by analogy, biosynthesis ͉ nitrogenase NifEN also contains two types of metal cluster sites, one corre- sponding to the P-cluster site and the other to the FeMoco site of itrogenase is the key player in nature’s ingenious scheme to the MoFe protein (17, 21). The P-cluster analog was identified as Nconvert the inert atmospheric dinitrogen to the bioavailable a [4Fe-4S] cluster likely coordinated by conserved Cys residues at form of (for recent reviews, see refs. 1–8). The the NifE–NifN interface (21). The FeMoco analog had not been Mo-nitrogenase of is composed of the captured on NifEN until an efficient one-step purification proce- (Fe) protein and the –iron (MoFe) protein. The dure, which minimized the degradation of and the conse- homodimeric Fe protein has one nucleotide-binding site per quent loss of metal clusters, was applied to a His-tagged form of the subunit and a single [4Fe-4S] cluster bridged between the two NifEN protein (17). In a so-called FeMoco maturation assay ␣ ␤ subunits. The 2 2-tetrameric MoFe protein contains two comprising NifEN, molybdate, homocitrate, Fe protein, MgATP unique metal clusters per ␣␤-subunit: the [8Fe-7S] P-cluster (9), and ⌬nifB MoFe protein, the FeMoco analog on NifEN was proven ␣␤ which is located at the -interface and ligated to six protein to be a FeMoco precursor by its ability to activate FeMoco-deficient residues; and the [Mo-7Fe-9S-X-homocitrate] (the identity of X ⌬nifB MoFe protein. Extended x-ray absorption fine structure

is unknown but is considered to be C, O, or N; ref. 10) FeMo (EXAFS) analysis of the NifEN-bound precursor showed that it was BIOCHEMISTRY ␣ cofactor (FeMoco), which is situated within the -subunit and structurally similar to FeMoco except for the notable absence of Mo bound to only two protein residues and an exogenous homoci- (18). Therefore, FeMoco cannot be formed through condensation trate ligand. Both P-cluster and FeMoco are composed of of [Mo-3Fe-3S] and [4Fe-3S] partial cubanes; rather, the Fe͞S smaller substructures: the P-cluster comprises two [4Fe-4S] ␮ structure of FeMoco is formed first, possibly through condensation subclusters that share a 6-sulfide (9) and FeMoco consists of of smaller Fe͞S subclusters in a fashion similar to that proposed for [Mo-3Fe-3S] and [4Fe-3S] subcubanes that are bridged by three ␮ ␮ P-cluster assembly and, then, Mo and homocitrate are added to 2-sulfides and share a central 6-light atom (10). These metal complete the synthesis. Although instrumental in clarifying the clusters are essential for nitrogenase reaction, a process that trajectory of FeMoco biosynthesis, our original study left such involves ATP-dependent electron transfer from the [4Fe-4S] unanswered questions as: when Mo and homocitrate are inserted cluster of the Fe protein to the P-cluster of the MoFe protein and into the cluster; how FeMoco is transferred from NifEN to the finally to FeMoco where substrate reduction takes place, and MoFe protein; and what role the Fe protein and MgATP play in consequently become the major subjects in the vigorous studies FeMoco maturation. of nitrogenase catalysis (1–8). Meanwhile, there is an emerging The current study addresses these remaining questions by understanding of nitrogenase biosynthesis, in particular, P- following the final steps of FeMoco assembly in A. vinelandii cluster and FeMoco assembly in A. vinelandii, that is poised to ͞ further clarify the structure and function of these important using a similar biochemical spectroscopic strategy to that used previously (17, 18). Through this approach, we show that Mo and clusters while also serving as a paradigm for the field of complex ͞ metalloprotein biosynthesis (11, 12). homocitrate are incorporated into the Fe S core of the FeMoco The P cluster is a classical example of high-nuclearity clusters precursor while it is bound to NifEN and that the fully comple- containing only Fe and S, and it is likely assembled at its targeted mented cluster is subsequently transferred from NifEN to MoFe location (‘‘in situ’’ assembly). Using a biochemical͞spectroscopic protein through direct protein–protein interaction. The function approach, we identified the presence of two pairs of [4Fe-4S]-like clusters that likely represent P-cluster precursors in a FeMoco- Author contributions: Y.H., K.O.H., B.H., and M.W.R. designed research; Y.H., M.C.C., deficient MoFe protein purified from a nifH-deletion strain (13, A.W.F., J.A.W., and M.W.R. performed research; Y.H., M.C.C., K.O.H., B.H., and M.W.R. 14). This protein becomes catalytically active upon incubation analyzed data; and Y.H., M.C.C., K.O.H., B.H., and M.W.R. wrote the paper. with the deleted gene product, indicating that it might represent The authors declare no conflict of interest. a physiologically relevant intermediate during P-cluster assembly This article is a PNAS direct submission. (15). These results suggest that the P-cluster is formed through Abbreviations: FeMoco, FeMo cofactor; EXAFS, extended x-ray absorption fine structure; the fusion of its substructural units, a reaction mechanism that IDS, indigo disulfonate. is well known in synthetic inorganic chemistry (16) and partic- §To whom correspondence may be addressed. E-mail: [email protected], hogdson@ssrl. ularly appropriate considering the ‘‘modular’’ composition of the slac.stanford.edu, or [email protected]. P cluster. © 2006 by The National Academy of Sciences of the USA

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0602647103 PNAS ͉ November 14, 2006 ͉ vol. 103 ͉ no. 46 ͉ 17119–17124 Downloaded by guest on September 30, 2021 Table 1. Metal contents of NifEN؅ Metal

Protein Mo Fe NifEN Ͻ0.01 16.1 Ϯ 1.1 NifENcomplete 1.2 Ϯ 0.1 15.8 Ϯ 0.6 ⌬nifB NifEN Ͻ0.01 8.2 Ϯ 1.0 ⌬nifB NifENcomplete Ͻ0.01 7.7 Ϯ 0.1 NifENminus Mo͞homocitrate Ͻ0.01 14.9 Ϯ 0.1 NifENminus homocitrate 0.3 Ϯ 0.1 15.2 Ϯ 0.1 Fig. 1. Protein purification and visible region absorption spectroscopy. (A) NifENminus Mo Ͻ0.01 15.7 Ϯ 0.2 Coomassie blue-stained 10–20% gradient SDS͞PAGE of NifEN, ⌬nifB NifEN, NifENminus MgATP Ͻ0.01 15.6 Ϯ 0.2 complete ⌬ complete ␮ ␮ NifEN and nifB NifEN . Lane 1, 10 g protein standard; lane 2, 15 g NifENminus Fe protein Ͻ0.01 15.3 Ϯ 0.1 ␮ ⌬ ␮ purified NifEN; lane 3, 15 g of purified nifB NifEN; lane 4, 15 g of purified NifENapo Fe protein Ͻ0.01 14.8 Ϯ 0.2 NifENcomplete; lane 5, 15 ␮g of purified ⌬nifB NifENcomplete.(B) Visible region NifENA157S Fe protein 0.2 Ϯ 0.1 15.4 Ϯ 0.1 spectra of the same protein samples in A. Spectra of dithionite-reduced NifEN (1), M156C Fe protein Ϯ Ϯ ⌬nifB NifEN (2), NifENcomplete (3), and ⌬nifB NifENcomplete (4) are shown between NifEN 0.2 0.1 14.6 0.4 A157G Fe protein Ͻ Ϯ 350 and 550 nm. The samples were prepared at a concentration of 5 mg͞ml, as NifEN 0.01 16.1 1.0 described in Supporting Text, which is published as supporting information on NifENMgADP Ͻ0.01 16.4 Ϯ 1.1 the PNAS web site. NifENATP␥S Ͻ0.01 15.3 Ϯ 0.1 NifENAMPPNP Ͻ0.01 16.5 Ϯ 1.0

of Fe protein and MgATP in FeMoco maturation is described in Data are expressed as moles of metal per mole of protein. a companion paper (24).

Results the formation of a FeMoco precursor and, accordingly, both ⌬nifB NifEN and ⌬nifB NifENcomplete are shown to have suffi- The FeMoco maturation assay, which was designed to test the cient Fe to form the permanent clusters alone: Ϸ8 mol Fe and capability of NifEN in FeMoco biosynthesis, comprises (i) no Mo per mol of protein (Table 1). Consistent with the metal NifEN, the source of FeMoco precursor; (ii) molybdate and analysis results, the visible region absorption spectra of these homocitrate, the constituents of FeMoco absent from the proteins reveal that the NifENcomplete spectrum (Fig. 1B,3)is precursor; (iii) Fe protein and MgATP, the factors assisting significantly more intense than that of NifEN (Fig. 1B,1)atthe the maturation of the precursor in an unknown fashion; and ⌬ same protein concentration, indicating that the precursor on (iv) FeMoco-deficient nifB MoFe protein, the ‘‘receptor’’ for NifENcomplete is modified from that on NifEN; whereas the fully converted FeMoco (17). Based on this assay, we devel- intensity of the ⌬nifB NifENcomplete spectrum (Fig. 1B,4)is oped a new strategy to test the extent of FeMoco maturation almost identical to that of the ⌬nifB NifEN (Fig. 1B, 2), on NifEN that involves (i) repurification of NifEN after indicating that the permanent clusters are unchanged. incubation with all of the components of the FeMoco matu- ⌬ EPR analyses provide additional, more detailed evidence that ration assay except nifB MoFe protein and (ii) subsequent the precursor on NifENcomplete is further processed (Fig. 2). In the analysis of re-purified NifEN (designated NifENcomplete). The analysis includes metal quantitation, EPR, visible region ab- sorption and x-ray spectroscopies, and activity assays that measure the capacity to reconstitute ⌬nifB MoFe protein. By systematically altering the composition of the incubation mix- ture, a suite of repurified NifEN proteins is produced, the analysis of which establishes the required components for FeMoco maturation on NifEN. ⌬nifB NifENcomplete, which is treated identically to NifENcomplete except for the replacement of NifEN by precursor-free ⌬nifB NifEN, serves as a negative control in this NifEN-specific FeMoco maturation assay. The complete set of repurified NifEN proteins are herein categor- ically designated NifENЈ and distinguished by different super- scripts (see Materials and Methods for the complete list of NifENЈ). As shown in Fig. 1A, NifENcomplete and ⌬nifB NifENcomplete, like their respective unprocessed counterparts NifEN and ⌬nifB NifEN, are composed of ␣ (Ϸ52 kDa) and ␤ (Ϸ49 kDa) subunits (17). The molecular masses of both proteins are Ϸ200 kDa based on their elution profiles on gel filtration Sephacryl S-200 HR column (data not shown), indicating that both proteins are ␣2␤2 tetramers. Metal analysis shows that unprocessed NifEN con- tains 16.1 Ϯ 1.1 mol Fe and no Mo per mol of protein (Table 1), allowing for the assignment of two permanent [4Fe-4S] clusters Fig. 2. EPR spectra of unprocessed (A) and processed (B) NifEN and ⌬nifB NifEN. (21) and one Mo-free precursor per protein molecule (17, 18). (A) EPR Spectra of dithionite-reduced NifEN (1), dithionite-reduced ⌬nifB NifEN ⌬ NifENcomplete, in contrast, contains 15.8 Ϯ 0.6 mol Fe and 1.2 Ϯ (2), IDS-oxidized NifEN (3), and IDS-oxidized nifB NifEN (4). (B) EPR Spectra of dithionite-reduced NifENcomplete (1), dithionite-reduced ⌬nifB NifENcomplete (2), 0.1 mol Mo per mol of protein (Table 1), which is consistent with IDS-oxidized NifENcomplete (3), and IDS-oxidized ⌬nifB NifENcomplete (4). All spectra the presence of one cluster having the same metal composition were measured at a protein concentration of 15 mg͞ml, as described in the as FeMoco in addition to the two permanent clusters per Supporting Text. The g values are indicated. (Inset) The spectrum of dithionite- molecule of NifEN. The absence of the nifB gene should prevent reduced NifENcomplete between 1,000 and 2,000 G at a magnification of 40-fold.

17120 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0602647103 Hu et al. Downloaded by guest on September 30, 2021 Table 2. Reconstitution of MoFe protein with NifEN؅

Activities* SPECIAL FEATURE

C2H4 formation H2 formation NH3 formation H2 formation Assay condition under C2H2͞Ar under Ar under N2 under N2

FeMoco maturation assay Complete 290 Ϯ 26 (100) 350 Ϯ 50 (100) 111 Ϯ 20 (100) 65 Ϯ 7 (100) MoFe protein reconstitution with NifENЈ† NifENcomplete 284 Ϯ 17 (98) 375 Ϯ 12 (107) 142 Ϯ 4 (127) 70 Ϯ 7 (108) ⌬nifB NifENcomplete 0 (0) 0 (0) 0 (0) 1 Ϯ 1(Ͻ1) NifENminus Mo͞homocitrate 0 (0) 0 (0) 0 (0) 0 (0) NifENminus homocitrate‡ 6 Ϯ 1 (2) 0 (0) 0 (0) 0 (0) NifENminus Mo 1 Ϯ 1(Ͻ1) 0 (0) 0 (0) 0 (0) NifENminus MgATP 0 (0) 0 (0) 0 (0) 0 (0) NifENminus Fe protein 0 (0) 0 (0) 0 (0) 0 (0) NifENapo Fe protein 0 (0) 0 (0) 0 (0) 0 (0) NifENA157S Fe protein§ 33 Ϯ 1 (11) 29 Ϯ 2 (8) 8 Ϯ 1 (7) 8 Ϯ 1 (12) NifENM156C Fe protein§ 23 Ϯ 1 (8) 31 Ϯ 1 (9) 5 Ϯ 1 (5) 8 Ϯ 1 (12) NifENA157G Fe protein 0 (0) 0 (0) 0 (0) 0 (0) NifENMgADP, NifENATP␥S or NifENAMPPNP 0 (0) 0 (0) 0 (0) 0 (0) Activity assays with NifENЈ alone¶ NifENcomplete or ⌬nifB NifENcomplete 0 (0) 0 (0) 0 (0) 0 (0)

Data are expressed as nanomoles per minute per milligram of protein. Percentages are given in parentheses. *The lower detection limits were 0.01, 0.02, 0.001, and 0.02 nmol per min per mg of protein for C2H4 formation under C2H2͞Ar, H2 formation under Ar, NH3 formation under N2 and H2 formation under N2, respectively. †Except for ⌬nifB NifENcomplete, all NifENЈ proteins contain normal amounts of Mo-free precursor based on their activities in FeMoco maturation assays, which are not considerably lower compared with NifEN. ‡NifENminus homocitrate can be activated to Ϸ10% upon incubation with all components of FeMoco maturation assay except molybdate, indicating the accumulation of a lower amount of molybdenum on the NifEN-bound precursor in the absence of homocitrate. §A157S and M156C Fe proteins show 23% and 16% of MgATP hydrolysis activities, respectively, compared with the wild-type Fe protein (data not shown). ¶Assays were performed as described earlier (35) except that MoFe protein was replaced by NifENcomplete or ⌬nifB NifENcomplete.

dithionite-reduced state, NifEN exhibits an S ϭ 1͞2 signal in the g ϭ NifENcomplete alone does not show any substrate reducing 2 region, which has a rhombic line shape with a distinct feature activities, as expected (Table 2). Nevertheless, NifENcomplete can BIOCHEMISTRY between g values of 1.95 and 1.88 (Fig. 2A, 1). As established activate the FeMoco-deficient ⌬nifB MoFe protein to a maxi- previously, this signal arises from both the precursor and the mum activity of Ϸ300 nmol of C2H4 formation per mg of ⌬nifB permanent [4Fe-4S] clusters on NifEN (17). Compared to unproc- MoFe protein per min (Fig. 3). Upon activation by NifENcomplete, complete ϭ ͞ essed NifEN, NifEN shows an S 1 2 signal with a slightly MoFe protein shows not only C2H4-formation but also H2- rhombic line shape (Fig. 2B, 1) of significantly lower intensity. The formation and N2-fixation activities that are comparable with decreased signal intensity of NifENcomplete is not caused by the loss those resulting from the complete FeMoco maturation assay of clusters, as shown by the metal analyses (Table 1); rather, it (Table 2). Although activities in these maturation assays are reflects the disappearance of the signal arising from the precursor lower than those resulting from activation of ⌬nifB MoFe protein upon further processing. As a result, the signal of NifENcomplete,like with isolated FeMoco (13), the activities for all substrates are those of the ⌬nifB NifEN proteins, should arise solely from the proportionally reduced, which would not be expected if the permanent [4Fe-4S] clusters in the protein. Indeed, the S ϭ 1͞2 cofactor were structurally perturbed or homocitrate was incor- signal of NifENcomplete is identical to that of ⌬nifB NifEN (Fig. 2A, 2), which, as the spectrum of ⌬nifB NifENcomplete shows (Fig. 2B, 2), remains unchanged upon processing. Consistent with these results, the unique signal of NifEN in the indigo disulfonate (IDS)-oxidized state with a g value of 1.92 (Fig. 2A,3)isnot observed in NifENcomplete (Fig. 2B, 3), indicating the disappearance of this Mo-free precursor-associated signal (17) upon further processing. As expected, this unique signal is also missing from the spectra of ⌬nifB NifEN (Fig. 2A,4)and⌬nifB NifENcomplete (Fig. 2B, 4) due to the absence of precursors in these proteins (17). Additionally, in the dithionite-reduced state, NifENcomplete shows a very minor EPR signal that has a slightly rhombic line shape with g values of 4.44, 4.05 and 3.96 (Fig. 2B,1,Inset). This signal could arise from the processed precursor or some intermediates associ- ated with its assembly (24). It is important to note that despite the presence of stoichiometric Mo for FeMoco formation, the charac- teristic S ϭ 3͞2 signal associated with MoFe protein-bound complete FeMoco (1) is not observed in NifEN . These results suggest Fig. 3. MoFe protein reconstitution with NifENcomplete. Assays were per- complete that the processed cluster in NifEN has a different spin state formed as described in Materials and Methods except that the amounts of than both the unprocessed NifEN-bound precursor and the mature NifENcomplete were varied between 0 and 15 nmol. The data presented here are MoFe protein-bound FeMoco. the average of three independent experiments. The error bars are indicated.

Hu et al. PNAS ͉ November 14, 2006 ͉ vol. 103 ͉ no. 46 ͉ 17121 Downloaded by guest on September 30, 2021 Fig. 4. Fe K-edge EXAFS (A) and Fourier transforms (B) of (top to bot- tom), the NifENcomplete precursor (blue) with the NifEN precursor (red); ⌬nifB NifENcomplete (blue) with ⌬nifB NifEN (red); and NifENcomplete (blue) with NifEN (red). EXAFS data for the NifENcomplete precursor were obtained by subtraction of ⌬nifB NifENcomplete EXAFS from NifENcomplete EXAFS in an 8͞7:15͞7 ratio. NifEN precursor data were obtained by a 1:2 subtraction of ⌬nifB NifEN EXAFS Fig. 5. Mo K-edge XAS and EXAFS. (A and B) Mo K-edge x-ray absorption complete from NifEN EXAFS (18). spectra (A) and smoothed second derivatives (B) of NifEN (blue), MoFe protein (red), and molybdate in 33% glycerol (dashed black). (C and D)Mo K-edge EXAFS (C) and Fourier transforms (D) of data (dotted black) and fits for NifENcomplete (blue) and MoFe protein (red). A four-component fit is shown for rectly bound (25). Thus, these activity results indicate that the complete complete NifEN comprising Mo-O at 2.00 and 2.17 Å, Mo-S at 2.37 Å, and Mo-Fe precursor on NifEN has all components necessary to at 2.70 Å. The MoFe protein fit comprises Mo-O at 2.22 Å, Mo-S at 2.37 Å, and become a mature FeMoco upon delivery from NifEN to the Mo-Fe at 2.69 Å. Complete fit results are given in Table 5. FeMoco binding site in MoFe protein. Furthermore, these data suggest that the transfer of the precursor from NifEN to MoFe protein can occur through direct protein–protein interaction, precursor are only subtly changed from those of the NifEN excluding the absolute requirement for a specific cluster carrier precursor (Fig. 4), consistent with a lack of major structural in this process. differences between these two species. Importantly, the signa- Formation of the processed, activatable precursor on ture feature of a FeMoco-like cluster, the intense peak at Ϸ3.5 Å NifENcomplete requires molybdate, homocitrate, Fe protein and in the Fourier transform (18) is still present in the NifENcomplete MgATP, because NifENЈ proteins prepared in the absence of precursor (Fig. 4). However, EXAFS fitting results indicate that one or more of these factors, such as NifENminus Mo/homocitrate, there is a change in the components that comprises this peak NifENminus homocitrate, NifENminus Mo, NifENminus Fe protein and from two Fe–Fe scatterers, at 3.68 and 3.80 Å, in the unprocessed NifENminus MgATP, are unable to activate the ⌬nifB MoFe protein precursor to one at 3.70 Å after processing (Table 4, which is (Table 2). Fe protein-mediated MgATP hydrolysis is also im- published as supporting information on the PNAS web site). plicated in this process, based on the observations that (i) MoFe protein-bound FeMoco is similarly fit by only one peak NifENMgADP, NifENATP␥S, and NifENAMPPNP, prepared by re- (Table 4) (18). Compared to FeMoco, the NifENcomplete precur- placing MgATP with MgADP or nonhydrolysable ATP analogs sor has slightly longer scattering distances on average, whereas (Table 3, which is published as supporting information on the it has slightly shorter differences than those found in the PNAS web site), cannot reconstitute the ⌬nifB MoFe protein unprocessed NifEN precursor (Table 4). These results confirm (Table 2) and (ii) NifENA157S Fe protein, NifENM156C Fe protein and that NifENcomplete still contains a FeMoco-like cluster, and NifENA157G Fe protein, prepared by substituting wild-type Fe pro- furthermore, suggest that the Fe͞S core of the processed pre- tein for variants defective in MgATP hydrolysis (Table 3), show cursor in NifENcomplete is tightened relative to that of NifEN, greatly diminished or no capacity to activate the ⌬nifB MoFe although it is still not as compact as FeMoco in its dithionite- protein (Table 2). In addition, the requirement of electron reduced state. Given the differences in the spin states of these transfer for cluster conversion is suggested by the inability of clusters, it is expected that they would have slightly different NifENapo Fe protein, prepared by replacing wild-type Fe protein bond lengths. Similar differences are observed for the different with cluster-deficient apo Fe protein (Table 3), to reconstitute states of FeMoco in MoFe protein (26). Because the contribu- the ⌬nifB MoFe protein (Table 2). The metal contents of the tion from Mo scattering is small relative to the other components NifENЈ proteins indicate considerably less or no Mo (and in the average Fe environment, Fe EXAFS fits cannot provide homocitrate; ref. 24) incorporation into the precursors in cases proof of Mo incorporation into the cluster. This proof comes where one or more of the MoFe protein maturation components from direct analysis of the Mo environment in NifENcomplete are absent or disrupted (Table 1) and correlate well with their through Mo K-edge XAS. The Mo edge spectrum of NifENcomplete respective capacities to activate the ⌬nifB MoFe protein (Ta- is highly similar to that of MoFe protein (Fig. 5) and consistent with ble 2). Mo in an octahedral environment of mixed O and S ligation like EXAFS analysis provides a structural probe of the pro- that found in FeMoco (27). The small, Ϸ0.6 eV, difference between cessed precursor on NifENcomplete. The Fe K-edge EXAFS of the edge positions of these proteins (Fig. 5) could be due to subtle NifENcomplete are slightly shifted in frequency relative to those differences in the Mo ligand environment, which are expected given of NifEN in the high-k region (Fig. 4); whereas the EXAFS of the absence of the FeMoco His ligand on NifE (23), or to ⌬nifB NifENcomplete are largely unchanged from those of ⌬nifB differences in the cluster oxidation state. Of particular significance NifEN (Fig. 4), indicating that it is the precursor, not the is the lack of any preedge features in NifENcomplete (Fig. 5), which permanent clusters, that is altered. Using an established method provides evidence that there is no free molybdate associated with (18), the EXAFS for the processed precursor were obtained by this protein after repurification (molybdate has a characteristic subtraction of ⌬nifB NifENcomplete from NifENcomplete in an sharp preedge transition at Ϸ20,008 eV, Fig. 5) (28). The Mo 8͞7:15͞7 ratio. The Fe EXAFS features of the NifENcomplete K-edge EXAFS of NifENcomplete clearly indicate that the majority

17122 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0602647103 Hu et al. Downloaded by guest on September 30, 2021 of Mo in this protein is in an Fe͞S cluster. The strong scattering formation, and deliver Mo to a FeMoco assembly site through contributions at high-k and the ratio of Fourier transform peaks

protein-protein interactions with NifEN (11, 12). In a related SPECIAL FEATURE (Fig. 5) are consistent with Mo in a FeMoco (29, 30) or FeMoco- study we establish that the Fe protein functions as a specific like environment (31). Accordingly, the NifENcomplete EXAFS are Mo͞homocitrate insertase in FeMoco biosynthesis (24). fit with Mo-S and Mo-Fe scattering components at similar distances The combined Mo and Fe EXAFS results indicate that the to those in FeMoco (Table 5, which is published as supporting fully complemented cluster species, formed upon incorporation information on the PNAS web site). The total magnitude of the of Mo and homocitrate into the Mo-free precursor, is structur- complete NifEN Fourier transform is greatly diminished relative to ally very similar to mature FeMoco. The bond distances in this that of the MoFe protein and the long-range features are absent, processed form of the precursor are consistent with a compac- indicating a significant amount of disorder in the Mo environment tion of the Fe͞S core of the precursor upon processing and are of NifENcomplete (Fig. 5). This disorder could be due to asymmetric similar to, although not the same as, those of FeMoco. EXAFS coordination as compared to that in FeMoco, or weak Mo binding studies clearly indicate that Mo is associated with the Fe͞S core, in the Fe-S cluster leading to a mixture of partially incorporated states, contamination from Mo bound elsewhere in the protein, or but also indicate that Mo binding in this cluster is likely a combination of these effects. EXAFS fitting results indicate asymmetrical or relatively weak compared to that in FeMoco. that NifENcomplete is best fit with a reduced number of Mo-S and These EXAFS results are consistent with EPR results showing Mo-Fe scatterers compared to the MoFe protein and with either that the processed cluster is in a different spin state than both 3 statically disordered Mo-O͞N scatterers at an average distance FeMoco and the unprocessed precursor. The final transforma- of 2.13 Å, or two types of Mo-O͞N scatterers at 2.00 Å and 2.17 Å tion of this cluster probably involves another change in cluster (Table 5). Note that the Mo-Fe scattering observed for oxidation state, leading to further tightening of the core Fe͞S NifENcomplete must be associated with the Fe in the precursor structure and stronger Mo binding. Additionally, the cluster will and not the permanent clusters, as there is no change to the be changed, especially at the Mo site, upon release from NifEN spectroscopic features of ⌬nifB NifEN, which represents the and ligation to the MoFe protein. This argument is based on the permanent clusters upon Mo incubation. In combination, fact that residues that either provide a covalent ligand or tightly the Fe and Mo K-edge XAS and EXAFS results indicate that pack FeMoco within the MoFe protein are not duplicated in the the NifENcomplete precursor consists of a FeMoco-like struc- corresponding region within the NifEN complex. For example, ture with slightly elongated bond lengths and an asymmetri- ␣-His442, which coordinates the Mo atom and anchors FeMoco cally or loosely bound terminal Mo atom that is also coordi- to MoFe protein, is substituted by Asn at the corresponding ͞ nated by a mixture of O N-containing ligands. position in NifEN. It is likely that the differences in protein environments not Discussion only account for some of the minor differences between the FeMoco biosynthesis is one of the most complicated processes in clusters on NifEN and MoFe protein, but also create an ‘‘affinity metalloprotein biochemistry, requiring a multitude of compo- gradient’’ between NifEN and MoFe protein for cluster binding nent proteins and encompassing a variety of synthetic strategies. that enable the negatively charged cluster to ‘‘escape’’ from By employing an experimental approach of analyzing only purified proteins, the roles of key components and the timing of NifEN (cluster donor), travel down the positively charged BIOCHEMISTRY key events in this process are unambiguously determined. Fur- ‘‘FeMoco insertion funnel’’ of the MoFe protein and eventually thermore, the combination of spectroscopic characterization ‘‘lock’’ into the binding site within the MoFe protein (cluster with biochemical analyses enables a more detailed picture of the receptor). This ‘‘locking’’ mechanism is likely directed by the specific FeMoco assembly mechanism. Previously, we identified change in Mo coordination to a stronger protein ligand upon a Mo-free, NifEN-bound FeMoco precursor with a core struc- MoFe protein binding, and potentially, also encompasses a ture similar to FeMoco as an early intermediate in this process change in cluster oxidation state. Our results indicate that no (18). In this work, we show that Mo and homocitrate are specific carrier proteins or factors are absolutely required to incorporated into the precursor while it is bound to NifEN, ‘‘escort’’ the cluster from NifEN to MoFe protein in vitro. resulting in the formation of a Mo͞homocitrate-containing, Therefore, it is possible that the respective cluster binding sites FeMoco-like cluster. We also show that this cluster is trans- within NifEN and the MoFe proteins are brought into close formed into a fully matured FeMoco and inserted into its target vicinity by interactions between the two proteins thereby facil- location in MoFe protein without the aid of a carrier or itating the ‘‘diffusion’’ of the cluster between the two binding chaperone. sites without additional assistance. The proposed diffusion re- The results of the FeMoco maturation assay clearly indicate action in which MoFe protein interacts with its FeMoco donor that Mo, homocitrate, [4Fe-4S]-containing Fe protein and ATP may have important implications for the structure of MoFe hydrolysis are all required for FeMoco maturation on NifEN. protein in different states and its mechanism for protecting its The impact of homocitrate on Mo binding to NifEN (Tables 1 air-sensitive clusters. Furthermore, the observed plasticity of Mo and 2) indicates that homocitrate and Mo are possibly preasso- in a FeMoco-like structure, and the changes observed with ciated and concurrently incorporated into the precursor. It has different spin states may provide insights into the structural been speculated that homocitrate, the organic moiety of FeMoco changes that take place in FeMoco during catalysis. Thus, the responsible for the overall negative charge of the cluster, may results presented herein not only significantly clarify the process assist FeMoco insertion by ‘‘steering’’ it down a positively charged ‘‘FeMoco insertion funnel’’ into its final location in the of FeMoco maturation; they may also prove useful for under- MoFe protein (32). It is interesting that some of the NifD standing other facets of nitrogenase chemistry. residues (␣ subunit of MoFe protein) providing the ‘‘FeMoco The required components for FeMoco maturation have been insertion funnel’’ are also positively charged in the correspond- identified, but other components are likely required to facilitate ing NifE residues (␣ subunit of NifEN), indicating the possible this process. For example, if the FeMoco assembly proceeds presence of an analogous, positively charged ‘‘Mo͞homocitrate through an [8Fe-9S] cluster, as has been proposed (18), then insertion funnel’’ in the NifEN complex (11). Mobilization of Mo there will be a need for a component to remove and sequester (and, apparently, homocitrate) for FeMoco biosynthesis has toxic Fe upon Mo incorporation into the cluster. Further re- been proposed to be carried out by proteins that can sequester search is needed to continue to fine-tune the understanding of Mo, place Mo in the correct oxidation state for FeMoco the final step of FeMoco biosynthesis.

Hu et al. PNAS ͉ November 14, 2006 ͉ vol. 103 ͉ no. 46 ͉ 17123 Downloaded by guest on September 30, 2021 Materials and Methods with the same assay conditions as i except that one or two of the Unless otherwise noted, all chemicals and reagents were ob- components required for FeMoco maturation were omitted. (iv) tained from Fisher, Aldrich, or Sigma. Cell growth, protein NifENapo Fe protein, NifENA157 Fe protein, NifENM156C Fe protein,or purification, preparation of apo Fe protein, FeMoco maturation NifENA157G Fe protein. These were control assays with the same assay, metal analysis, visible region absorption, EPR and x-ray conditions as i except that wild-type Fe protein was replaced by 120 spectroscopies were performed as described (13, 17, 18, 24, 33, mg of apo, A157S, M156C, or A157G Fe proteins, respectively. (v) MgADP ATP␥S AMPPNP 34). See Supporting Text, which is published as supporting NifEN , NifEN or NifEN . These were control information on the PNAS web site, for more information on assays with the same conditions as i except that ATP was replaced Ј ␥ these procedures. by 2.4 mM ADP, adenosine 5 -O-(3-thiotriphosphate) (ATP S) or 5Ј-adenylylimidodiphosphate (AMPPNP). Creatine phosphate and FeMoco Maturation on NifEN. To monitor FeMoco maturation creatine phosphokinase were omitted when the function of ADP while it is bound to NifEN, FeMoco precursor-bound NifEN was evaluated. Table 3 summarizes the designations of repurified Ј was subjected to FeMoco maturation assays (17) with modified NifEN and the assay conditions used to obtain these proteins. conditions, repurified, and subsequently examined in MoFe protein reconstitution assays (below). Such repurified NifEN MoFe Protein Reconstitution with NifEN؅. To test whether the Ј proteins are categorically designated NifENЈ, with different NifEN proteins (above) contain mature FeMoco, the following ⌬ superscripts indicating different assay compositions. The fol- assay was designed to reconstitute the nifB MoFe protein. Such ⅐ lowing NifENЈ proteins were prepared. (i) NifENcomplete. This an assay contained, in a 0.8 ml total volume, 25 mM Tris HCl (pH ⌬ assay contained, in a 50 ml total volume, 25 mM Tris⅐HCl (pH 8.0), 20 mM Na2S2O4, 0.5 mg of purified nifB MoFe protein 8.0), 2 mM Na S O , 100 mg of FeMoco precursor containing from A. vinelandii strain DJ1143 (32). FeMoco insertion was 2 2 4 Ј NifEN, 120 mg of Fe protein, 0.4 mM homocitrate, 0.4 mM initiated with the addition of 2 mg of isolated NifEN to the mixture above. The reaction mixture was incubated and stopped Na2MO4, 2.4 mM ATP, 4.8 mM MgCl2, 30 mM creatine phosphate, and 24 units͞ml of creatine phosphokinase. This as described for the FeMoco maturation assay (17). The enzy- mixture was stirred for1hat30°C and then NifEN was matic activities were subsequently determined as described (35). repurified as described (17). Note that the ⌬nifB MoFe protein was omitted from the assay to allow for accumulation of This work was supported by National Institutes of Health Grants complete ii ⌬nifB complete GM-67626 (to M.W.R.) and RR-01209 (to K.O.H.). The Stanford processed FeMoco on NifEN .( ) NifEN . Synchrotron Radiation Laboratory (SSRL) is supported by the Depart- This was a control assay with the same conditions as i except ment of Energy (DOE) Office of Basic Energy Sciences, and the SSRL that NifEN was replaced by precursor-free ⌬nifB NifEN (17). X-Ray Absorption facilities are supported by DOE Office of Biological (iii) NifENminus Mo/homocitrate, NifENminus homocitrate, NifENminus Mo, and Environmental Research and National Institutes of Health National NifENminus MgATP,orNifENminus Fe protein. These were control assays Center for Research Resources Biotechnology Training Program.

1. Burgess BK, Lowe DJ (1996) Chem Rev 96:2983–3011. 21. Goodwin PJ, Agar JN, Roll JT, Roberts GP, Johnson MK, Dean DR (1998) 2. Howard JB, Rees DC (1996) Chem Rev 96:2965–2982. Biochemistry 37:10420–10428. 3. Smith BE (1999) Adv Inorg Chem 47:159–218. 22. Roberts GP, Brill WJ (1980) J Bacteriol 144:210–216. 4. Rees DC, Tezcan FA, Haynes CA, Walton MY, Andrade S, Einsle O, Howard 23. Brigle KE, Weiss MC, Newton WE, Dean DR (1987) J Bacteriol 169:1547– JB (2005) Philos Trans R Soc A 363:971–984. 1553. 5. Christiansen J, Dean DR, Seefeldt LC (2001) Annu Rev Plant Physiol Plant Mol 24. Hu Y, Corbett MC, Fay AW, Webber JA, Hodgson KO, Hedman B, Ribbe MW Biol 52:269–295. (2006) Proc Natl Acad Sci USA 17125–17130. 6. Igarashi RY, Seefeldt LC (2003) Crit Rev Biochem Mol Biol 38:351–384. 25. Madden MS, Kindon ND, Ludden PW, Shah VK (1990) Proc Natl Acad Sci 7. Seefeldt LC, Dance IG, Dean DR (2004) Biochemistry 43:1401–1409. USA 87:6517–6521. 8. Peters JW, Szilagyi RK (2006) Curr Opin Chem Biol 10:1–8. 26. Christiansen J, Tittsworth RC, Hales BJ, Cramer SP (1995) J Am Chem Soc 9. Peters JW, Stowell MHB, Soltis SM, Finnegan MG, Johnson MK, Rees DC 117:10017–10024. (1997) Biochemistry 36:1181–1187. 27. Conradson SD, Burgess BK, Newton WE, Hodgson KO, McDonald JW, 10. Einsle O, Tezcan FA, Andrade SLA, Schmid B, Yoshida M, Howard JB, Rees Rubinson JF, Gheller SF, Mortenson LE, Adams MWW, et al. (1985) JAm DC (2002) Science 297:1696–1700. Chem Soc 107:7935–7940. 11. Dos Santos PC, Dean DR, Hu Y, Ribbe MW (2004) Chem Rev 104:1159–1173. 28. Kutzler FW, Scott RA, Berg JM, Hodgson KO, Doniach S, Cramer SP, Chang 12. Rubio LM, Ludden PW (2005) J Bacteriol 187:405–414. CH (1981) J Am Chem Soc 103:6083–6088. 13. Ribbe MW, Hu Y, Guo M, Schmid B, Burgess BK (2002) J Biol Chem 29. Conradson SD, Burgess BK, Newton WE, Mortenson LE, Hodgson KO (1987) 277:23469–23476. 14. Corbett MC, Hu Y, Naderi F, Ribbe MW, Hedman B, Hodgson KO (2004) J Am Chem Soc 109:7507–7515. J Biol Chem 279:28276–28282. 30. Liu HI, Filipponi A, Gavini N, Burgess BK, Hedman B, Di Cicco A, Natoli CR, 15. Hu Y, Corbett MC, Fay AW, Webber JA, Hedman B, Hodgson KO, Ribbe MW Hodgson KO (1994) J Am Chem Soc 116:2418–2423. (2005) Proc Natl Acad Sci USA 102:13825–13830. 31. Nordlander E, Lee SC, Cen W, Wu ZY, Natoli CR, Di Cicco A, Filipponi 16. Lee SC, Holm RH (2004) Chem Rev 104:1135–1158. A, Hedman B, Hodgson KO, Holm RH (1993) J Am Chem Soc 115:5549– 17. Hu Y, Fay AW, Ribbe MW (2005) Proc Natl Acad Sci USA 102:3236–3241. 5558. 18. Corbett MC, Hu Y, Fay AW, Ribbe MW, Hedman B, Hodgson KO (2006) Proc 32. Schmid B, Ribbe MW, Einsle O, Yoshida M, Thomas LM, Dean DR, Rees DC, Natl Acad Sci USA 103:1238–1243. Burgess BK (2002) Science 296:352–356. 19. Robinson AC, Burgess BK, Dean DR (1986) J Bacteriol 166:180–186. 33. Clark LJ, Axley JH (1955) Anal Biochem 27:2000–2003. 20. Allen RM, Chatterjee R, Ludden PW, Shah VK (1995) J Biol Chem 270:26890– 34. Van de Bogart M, Beinert H (1967) Anal Biochem 20:325–334. 26896. 35. Ribbe MW, Burgess BK (2001) Proc Natl Acad Sci USA 98:5521–5525.

17124 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0602647103 Hu et al. Downloaded by guest on September 30, 2021