A mixed-valent Fe(II)Fe(III) species converts cysteine to an oxazolone/thioamide pair in methanobactin biosynthesis.

SignificanceMethanobactins (Mbns), copper-binding peptidic compounds produced by some bacteria, are candidate therapeutics for human diseases of copper overload. The paired oxazolone-thioamide bidentate ligands of methanobactins are generated from cysteine residues in a precursor peptide, MbnA, by the MbnBC enzyme complex. MbnBC activity depends on the presence of iron and oxygen, but the catalytically active form has not been identified. Here, we provide evidence that a dinuclear Fe(II)Fe(III) center in MbnB, which is the only representative of a >13,000-member protein family to be characterized, is responsible for this reaction. These findings expand the known roles of diiron enzymes in biology and set the stage for mechanistic understanding, and ultimately engineering, of the MbnBC biosynthetic complex.

The MbnBC enzyme complex is central to the biosynthesis of methanobactins (Mbns), copper-binding metallophores isolated from methane-oxidizing (methanotrophic) bacteria and currently under investigation as therapeutics for Wilson disease (1, 2). Mbns are ribosomally synthesized, posttranslationally modified peptide (RiPP) natural products generated from a precursor peptide, MbnA (3). Operons encoding MbnA along with the Mbn biosynthetic and transport machinery are found in a range of bacteria in addition to methanotrophs and are predicted to yield a diverse array of natural products (4). Mbns chelate copper with high affinity by using two nitrogen and two sulfur atoms derived from nitrogen-containing heterocycles and neighboring thioamide groups () (5–10). These paired oxazolone ring and thioamide groups, present in all Mbns characterized thus far, are generated from two conserved cysteine residues and the carbonyl groups of their preceding residues (Fig. 1) by two proteins, MbnB and MbnC, which form a heterodimeric complex denoted MbnBC (11). In a subset of Mbns, one nitrogen ligand is provided by a pyrazinedione ring proposed to result from further modification of an oxazolone ring originally produced by MbnBC () (7, 10). While additional modifications occur in the biosynthesis of some Mbns, MbnBC is universally present in the operons and is thus the linchpin of this pathway.

Fig. 1.

Reaction catalyzed by MbnBC and a model for the MbnB Fe-binding site. (A) A representative MbnA sequence from Ms. trichosporium OB3b contains a leader sequence (gray) and a core sequence (underlined). The cysteine residues that are modified are colored in orange. MbnBC performs a four-electron oxidation to generate an oxazolone ring and a thioamide group from each cysteine residue and neighboring carbonyl group. (B) The proposed metal-binding site in MbnB showing potential diiron and triiron clusters. Disruption of the amino acid ligands labeled in blue significantly diminished activity. Two other residues, D208 and H210 (labeled in purple), are proposed to bind a third Fe ion.

MbnB is the only biochemically characterized member of the DUF692 protein family, which includes >13,000 sequences (UniProt), while MbnC belongs to a completely uncharacterized protein family (11). Initial studies showed that MbnBC requires both Fe and O2 to convert cysteine into an oxazolone/thioamide pair (Fig. 1). According to top-down native mass spectrometry (nTDMS) measurements, MbnB ejected from the MbnBC complex contains three Fe ions. Mössbauer spectroscopic analysis further indicated that MbnBC contains a mixture of triferric [Fe(III)]3 and diferric [Fe(III)]2 clusters, and a model of MbnBC based on the unpublished crystal structure of a diiron center–containing homolog from Haemophilus somnus 129pt (Protein Data Bank [PDB] accession No. 3BWW) suggested possible ligand sets for three Fe ions (Fig. 1) (11). Variants in which predicted ligands His54, His90, Glu133, Asp163, His192, and Glu239 were individually replaced with alanine or serine (Fig. 1) exhibited diminished Fe content and significantly less cysteine modification activity, supporting a functional role for Fe coordinated by these residues (11).

Despite these advances, heterogeneity in the MbnBC Fe site, as evidenced by the observation of both dinuclear [Fe(III)]2 and trinuclear [Fe(III)]3 clusters by Mössbauer spectroscopy, has complicated identification of the catalytically active species. Moreover, modification of only the first, but not the second, modifiable cysteine in MbnA was reported previously (11), suggesting that the enzyme was not fully activated. The involvement of an Fe-containing enzyme in oxazolone and thioamide synthesis is unique, and establishing the nuclearity and redox state of the active species is critical to understanding the mechanism. Here, we provide evidence through electron paramagnetic resonance (EPR) and Mössbauer spectroscopies combined with enzymatic activity assays, crystal structures, and mass spectrometric analysis that the active species of MbnBC is a mixed-valent Fe(II)Fe(III) cluster coordinated by conserved residues in MbnB. These findings set the stage for detailed mechanistic studies and expand the diversity of reactions attributable to nonheme diiron enzymes in natural product biosynthesis.

Results and Discussion

MbnB Activity Derives from a Mixed-Valent Diiron Fe(II)Fe(III) Cluster.

To identify the Fe species that reacts with MbnA and O2, EPR, Mössbauer, and activity data were acquired for MbnBC samples in various Fe oxidation states. Methylosinus (Ms.) trichosporium OB3b MbnBC was heterologously expressed in Escherichia coli in the presence of 250 µM supplemental 57Fe and purified using established protocols (11). The resultant protein was first treated with ferricyanide to generate a fully oxidized (all ferric) sample (oxMbnBC). The EPR spectrum of this sample (Fig. 2, oxMbnBC + 0.0 equivalents (equiv) ascorbate [asc]) exhibits a strong signal at an effective g value (geff) = 4.3 and several weaker signals with geff values of ∼8.5, 6.9, 4.9, characteristic of “rhombic” S = 5/2 Fe(III). In addition, there is a weak signal at g < 2 attributable to the S = 1/2 ground state of an antiferromagnetically coupled Fe(II)Fe(III) cluster (Fig. 2); double integration of this signal reveals the presence of 9 μM Fe(II)Fe(III) clusters (∼1% of the total Fe in the sample). This spectrum of oxMbnBC is nearly the same as that obtained for aerobically purified MbnBC, but lacks features previously observed above 4,000 Gauss (11). The Mössbauer spectrum of oxMbnBC (Fig. 2) is similar to that published previously (11) () and exhibits the broad magnetically split features associated with an [Fe(III)]3 cluster (blue arrows) and/or adventitiously bound high-spin Fe(III) (blue arrows), in addition to a quadrupole doublet with parameters reported for that of the [Fe(III)]2 cluster (red spectrum, ∼20% of total intensity) (Fig. 2).

Fig. 2.

EPR and Mössbauer spectra of MbnBC. (A) EPR spectra of 57Fe MbnBC treated with different amounts of asc and dithionite (DT). (B) Mössbauer spectra of oxMbnBC with [Fe(III)]2 simulation (red) at 20% total Fe signal (Top), redMbnBC_asc (Middle), and difference spectrum (Bottom). Diferric simulation parameters were extracted from previous high-field Mössbauer work (11): ΔEQ = 0.80 mm/s, η = 0.75, and δ = 0.49 mm/s. The red vertical line highlights the high energy line associated with Fe(II) and/or [Fe(III)]3 species. All samples were prepared with 1 mM 57Fe MbnBC, and the equiv of reductants was calculated per MbnBC complex.

Treatment of oxMbnBC with asc (redMbnBC_asc) results in marked changes to the EPR spectrum. First, a rhombic EPR signal with = [1.94, 1.88, 1.76, average g value (gave) = ∼1.9] is observed and is well resolved at ∼10 to 30 K (Fig. 2). This signature is attributed to an antiferromagnetically coupled Fe(II)Fe(III) cluster with an S = 1/2 ground state, a unit that has been characterized extensively in both diiron enzymes and synthetic model complexes thereof (12–14). The intensity of the S = 1/2 EPR signal reached maximum upon the addition of 1 to 2 equiv of asc (Fig. 2), integrated to ∼80 μM spin, and persisted upon addition of 5 equiv of asc, indicating that asc cannot lead to efficient reduction of the cluster (). The ∼80 μM spin of the Fe(II)Fe(III) species accounted for 13% of the total iron in the sample, or 66% of the [Fe(III)]2 deduced from the Mössbauer spectrum of the fully oxidized sample (Fig. 2). Addition of 2 equiv of dithionite (redMbnBC_dt), however, resulted in a decrease of the intensity of the S = 1/2 signal, suggesting further reduction to an [Fe(II)]2 cluster (Fig. 2). In addition, the pronounced signal at geff = 4.3 was reduced approximately fivefold in intensity, and the low-field signals disappeared completely, suggesting that the majority of the S = 5/2 rhombic Fe(III) was reduced. A new broad signal extending from g = 4.5 to g = 6.7 appeared as well and could derive from an alteration in the zero-field splitting of residual high-spin Fe(III) or from a 2e− reduced [Fe(II)]2/Fe(III) cluster.

The Mössbauer spectrum of a sample of oxMbnBC treated with 1 equiv of asc (Fig. 2) exhibits broad, relatively featureless absorption, extending from approximately −7 mm/s to approximately +7 mm/s, revealing that the majority of the Fe exhibits magnetically split features. Most notably, the broad quadrupole doublet associated with the [Fe(III)]2 cluster is diminished in intensity (Fig. 2), consistent with the appearance of the EPR signal associated with the S = 1/2 Fe(II)Fe(III) cluster. This species is expected to exhibit magnetically split subspectra, but accounts for only ∼12% of the total intensity of the spectrum (80 μM spin × 2 Fe per dinuclear cluster/1.3 mM total Fe). Moreover, because the low-field EPR features are diminished significantly by asc treatment, the majority of the broad, magnetically split features must arise from another Fe species. One candidate for this species is a 1e−-reduced trinuclear cluster, Fe(II)/[Fe(III)]2, which would have an integer spin ground state that could give rise to magnetically split spectra (although the majority of integer-spin clusters exhibit quadrupole doublet features under these conditions). Alternatively, this species could be a 2e−-reduced trinuclear cluster, [Fe(II)]2/Fe(III), which has a half-integer spin ground state and is expected to exhibit magnetically split Mössbauer spectra, the presence of which could be inferred from the broad EPR signals at g = 5 to 6. We have not further analyzed these spectra due to their lack of resolution and the fact that the broad component is a superposition of multiple subspectra.

MbnBC samples in each of the three oxidation states, oxMbnBC (Fig. 3), redMbnBC_asc prepared with 2 equiv of asc (Fig. 3), and redMbnBC_dt (Fig. 3), were then tested for MbnA modification activity. For each assay, MbnBC prepared in the anaerobic chamber was mixed with an equal volume of MbnA in O2-saturated buffer using a stopped-flow apparatus, and MbnA modification was monitored by the appearance of an optical feature at 335 nm, arising from the oxazolone ring and the thioamide group (11). Although the oxidized protein has an overlapping optical feature at 340 nm, this feature forms slowly and is much weaker, suggesting that the absorbance is primarily attributable to product formation (11). A comparison of the kinetic traces indicates that redMbnBC_asc produces the modified MbnA product with the greatest initial velocity and yield at completion (Fig. 3).

Fig. 3.

Reaction of 150 µM (A) oxMbnBC, (B) redMbnBC_asc, and (C) redMbnBC_dt with 300 µM MbnA in oxygenated buffer using stopped-flow spectroscopy over a period of 250 s. Arrows indicate the increase at 335 nm as a function of time with the starting spectrum colored red and the ending spectrum colored purple. Each inset shows the difference spectrum between the initial and final time points. The feature at 650 nm is proposed to derive from a charge transfer transition associated with the MbnA/oxidized MbnBC complex (11). (D) Difference absorbance traces at 335 nm attributed to the formation of an oxazolone ring and a thioamide group, obtained by subtracting the absorbance at each point from the initial time point. Inset shows the traces plotted between 0.01 and 10 s. AU, absorbance units.

Given that redMbnBC_asc gave the highest rate and extent of MbnA modification, 2 equiv MbnA was added to redMbnBC_asc under anaerobic conditions to generate the reactant complex, and the effects of substrate binding on different Fe species were examined by EPR and Mössbauer spectroscopies. In the EPR spectrum, the S = 1/2 gave ∼1.9 feature attributed to the antiferromagnetically coupled Fe(II)Fe(III) center was converted to a new sharper feature with g-values of 1.92, 1.89, and 1.82 (Fig. 4). The perturbation suggests that MbnA binding modifies the coordination sphere of the Fe(II)Fe(III) cluster. Indeed, similar perturbations upon addition of substrate have been observed for all three HD-domain mixed-valent diiron oxygenases and oxidases (MVDOs) that have been studied to date (15–17). The residual g = ∼5.8 signal also is perturbed upon substrate addition (). The Mössbauer spectrum of redMbnBC_asc in the presence of MbnA () is consistent with this assignment.

Fig. 4.

Effect of MbnA on the EPR spectrum of redMbnBC_asc. Samples were prepared with 1.2 mM redMbnBC_asc and 2 equiv of MbnA per MbnBC complex.

To further correlate activity with the prevalence of the Fe(II)Fe(III) cluster, catalytically impaired variants in which the proposed Fe ligands were replaced with alanine or serine (H54A, H90S, E133A, D163S, H192A, and E239A) (Fig. 1) (11) were isolated, treated with 2 equiv asc, and examined by EPR spectroscopy (). With the exception of D163S, these substitutions abolished the signals from the Fe(II)Fe(III) cluster, consistent with a first coordination sphere including H54, H90, E133, H192, and E239. The D163S substitution changed, but did not abolish, the Fe(II)Fe(III) EPR signal, consistent with this residue as a weakly interacting ligand.

Low Fe MbnBC Produces a Higher Yield of Modified MbnA.

The correlation between activity and the presence of an Fe(II)Fe(III) cluster combined with the observation of both [Fe(III)]2 and [Fe(III)]3 clusters by Mössbauer spectroscopy and nTDMS (11) suggested that our MbnBC sample preparation could be improved to maximize the amount of active Fe(II)Fe(III) species. We therefore tried to prepare an MbnBC sample with a homogeneous [Fe(III)]2 cluster. Initial attempts to remove Fe from MbnBC using ferrozine, which significantly diminishes activity (11), and to reload with either Fe(II) or Fe(III) were unsuccessful. As an alternative approach, MbnBC was heterologously expressed in E. coli with less Fe in the growth medium [160 µM compared to 250 µM used above and >250 µM used in prior work (11)]. Inductively coupled plasma mass spectrometry (ICP-MS) measurements indicated that the MbnBC isolated from these cultures (MbnBClowFe) contained 0.82 ± 0.02 equiv of Fe, whereas MbnBC isolated from cultures with 250 µM Fe (MbnBChighFe, used for above experiments) contained 1.05 ± 0.2 equiv Fe.

The enzymatic activities of MbnBClowFe and MbnBChighFe were then compared. The samples were first treated with 2 equiv of asc to maximize the presence of the Fe(II)Fe(III) species (redMbnBClowFe_asc, redMbnBChighFe_asc). Activity assays were then initiated by adding 1.3 equiv of the substrate MbnA to asc-treated MbnBC under ambient aerobic conditions to provide excess O2 to the reaction as a cosubstrate. Strikingly, the reaction with redMbnBClowFe_asc (Fig. 5) yielded significantly more product (∼1.8x) than the reaction with MbnBChighFe_asc (Fig. 5).

Fig. 5.

Activity analysis and Mössbauer spectra of MbnBC produced under low Fe conditions. (A) Activity assay of redMbnBClowFe_asc. (B) Activity assay of redMbnBChighFe_asc. The reactions were performed by addition of MbnA to a final concentration of 130 µM to a solution of asc-reduced 100 µM MbnBC under aerobic conditions at room temperature. The reaction was monitored every 15 s for a total assay time of 15 min. Arrows indicate the increase at 335 nm as a function of time, with the starting spectrum colored red and the ending spectrum colored purple. (C) Mössbauer spectra of oxMbnBClowFe (Top) with [Fe(III)]2 simulations (red) at 32% total Fe signal, redMbnBClowFe_asc (Middle), and difference spectrum (Bottom). Diferric simulation parameters were extracted from previous high-field Mössbauer work (11): ΔEQ = 0.80 mm/s, η = 0.75, and δ = 0.49 mm/s. The red vertical line highlights the high energy line associated with Fe(II) and/or [Fe(III)]3 species. AU, absorbance units.

To discern the differences between the MbnBChighFe and MbnBClowFe samples, MbnBClowFe was examined by EPR and Mössbauer spectroscopies. The Mössbauer spectrum of oxMbnBClowFe revealed an increased intensity of the quadrupole doublet associated with the [Fe(III)]2 cluster, 32% in oxMbnBClowFe versus 20% in oxMbnBChighFe (Fig. 5 and ). The remainder of the absorption comprised the broad, magnetically split features associated with the [Fe(III)]3 cluster and/or adventitiously bound high spin Fe(III). Upon reduction with asc, the Mössbauer spectrum showed a decrease in intensity of the quadrupole doublet from the [Fe(III)]2 state and an increased intensity of the magnetically split features, in addition to a new line at approximately +3 mm/s, which is assigned as the high energy line of high spin Fe(II) generated from reduction of adventitiously bound high spin Fe(III) (Fig. 5). The EPR spectra of redMbnBClowFe_asc are similar to those of MbnBChighFe_asc, exhibiting identical features at gave ∼1.9 and geff = 4.3, of which the latter signal likely derived from underloaded diiron cluster (). The signal at S = 1/2 with gave ∼1.9 integrates to ∼35 μM spin, which corresponds to ∼19% of total Fe and ∼60% reduction yield of [Fe(III)]2, which is consistent with the yield observed for MbnBChighFe. The activity profiles of MbnBClowFe in the three oxidation states () recapitulate those of MbnBChighFe (Fig. 3). The lesser quantity of [Fe(III)]3 cluster in the highest activity redMbnBClowFe_asc sample further supports the identification of the MbnBC active species as a mixed-valent Fe(II)Fe(III) cluster.

Crystal Structure of MbnBC Reveals Three Fe-Binding Sites.

To confirm the proposed ligand set for the Fe(II)Fe(III) cluster and assess the presence and importance of the third Fe ion detected by Mössbauer spectroscopy and nTDMS (11), we determined the crystal structure of MbnBC. Initial attempts to grow crystals of wild-type Ms. trichosporium OB3b MbnBC did not yield diffracting crystals. The crystal quality was improved significantly by removal of eight predicted disordered residues from the C terminus of MbnB and mutagenesis of predicted surface patches guided by the Surface Entropy Reduction prediction (SERp) server (MbnB E67A, E69A, K70A; MbnC E162A, E164A, K165A) (18). The structure of MbnBC expressed with 250 µM Fe was solved to 2.6 Å resolution by Fe single-wavelength anomalous dispersion (SAD), and additional structures of MbnBC expressed with 80 µM Fe, the H210S variant, and MbnBC crystallized anaerobically in the presence of the core peptide were determined by molecular replacement ().

As predicted, MbnB exhibits an eight-stranded triosephosphate isomerase (TIM)–barrel fold (Fig. 6) and closely resembles the H. somnus 129pt DUF692 protein (PDB accession No. 3BWW), with an rmsd of 3.8 Å for superposition of 224 Cα atoms. Three loop regions that were unmodeled in the H. somnus 129pt protein are well ordered in MbnB (). MbnC caps the MbnB TIM barrel, leaving a large open cavity above the iron site (Fig. 6), presumably for MbnA binding. There are numerous intersubunit interactions, with 2,113 Å2 buried surface area at the protein–protein interface (calculated using the PDBePISA server). MbnC consists of an N-terminal α-helical domain and two additional α helices packed against a four-stranded antiparallel β sheet (Fig. 6). The MbnC fold does not resemble any previously characterized protein. A search of the Dali server yielded a partial structural match to a member of Pfam family PF09836 from Neisseria gonorrhoeae (PDB accession No. 3DEE) (). Interestingly, this protein, proposed to play a role in transcriptional regulation, also neighbors a member of the DUF692 family (19). The MbnC surface housing the three surface entropy reduction substitutions E162A, E164A, and K165A formed an antiparallel β-strand interaction with a symmetry-related molecule, explaining why this crystallization strategy was successful ().

Fig. 6.

Crystal structure of Ms. trichosporium OB3b MbnBC. (A) Overall structure with MbnB shown in orange and MbnC show in green. The Fe ions in MbnB are shown as dark red spheres, and the ligands are shown as sticks. (B) The Fe center in MbnBC expressed in the presence of 250 µM Fe with partially occupied Fe3 site. The Fe anomalous electron density map (purple) is superimposed at 25σ. (C) The Fe center in MbnBC expressed in the presence of 250 µM Fe and crystallized anaerobically in the presence of the MbnA core peptide with all three Fe sites fully occupied. The Fe anomalous electron density map (purple) is superimposed at 20σ. Solvent molecules are shown as small red spheres.

Anomalous electron density maps calculated using data collected at the Fe absorption edge yielded two strong peaks (37σ), modeled as Fe ions (Fig. 6). The two Fe ions (Fe1 and Fe2) are six-coordinate, bridged by Glu133, along with two exogenous ligands modeled as a solvent molecule and a sulfate ion, analogous to the cacodylate ion in the H. somnus 129pt homolog structure and likely derived from the LiSO4 used for crystallization. Fe1 is further coordinated by the ε nitrogen atoms of His54 and His90 and a solvent molecule, while Fe2 is coordinated by δ nitrogen atom of His192 and by one carboxylate oxygen each from Asp163 and Glu239 (Fig. 6 and ). Given that activity (11) and the Fe(II)Fe(III) EPR signal () were disrupted upon replacement of these residues, these two Fe ions can be assigned as the active cofactor of MbnBC. A third peak (5σ) was also present in the anomalous electron density map, located between residues Asp208 and His210 at the proposed location of the third Fe ion (Fig. 1). The density was much weaker than that for Fe1 and Fe2 but was best modeled as an Fe ion at occupancies of 0.52 and 0.67 in the two molecules in the asymmetric unit. This Fe ion, designated Fe3, is coordinated by the ε nitrogen of His210 and the side chain oxygen of Asn166 (Fig. 6 and ), which was not anticipated as a ligand but is highly conserved. Proposed ligand Asp208 (Fig. 1) was disordered, likely due to partial Fe occupancy of this site. A second structure of MbnBC, determined to 2.3 Å resolution () using protein expressed in the presence of 80 μM rather than 250 µM Fe, showed even lower occupancy of the third Fe site (0.28 and 0.27) and no evidence for the bridging sulfate ion. Instead, Fe1 and Fe2 are each coordinated by an additional terminal solvent molecule ().

Extensive attempts at cocrystallization with full-length MbnA were not successful, but a 2.2 Å resolution structure was determined from crystals grown anaerobically in the presence of 10 equiv MbnA core peptide. Interestingly, anomalous electron density maps for this structure showed full occupancy of all three Fe sites (30 to 35σ) (Fig. 6). In this structure, Asp208 adopts a new position, coordinating directly to Fe3, which is also coordinated by Asn166 and His210 (Fig. 6 and ). Given that MbnBC used for this structure was also expressed in the presence of 250 µM Fe, it is not clear why full occupancy of Fe3 was achieved. No density corresponding to the core peptide was apparent, and additional structures determined from crystals grown anaerobically in the absence of core peptide gave identical structures with all three Fe sites fully occupied.

In the absence of a cocrystal structure with MbnA, we used Alphafold2 (20) as accessed through ColabFold (21) to generate a model of the MbnABC complex (). In this model structure, modifiable Cys21 of MbnA is positioned adjacent to Fe1, supporting the assignment of an Fe(II)Fe(III) cluster comprising Fe1 and Fe2 as the active cofactor. Strikingly, residues 6 to 11 of the MbnA leader peptide form an extended β sheet with four-stranded antiparallel β sheet of MbnC, interacting with MbnC residues 158 to 162 (), a finding recently confirmed by the crystal structures of Rugamonas rubra ATCC 43154 and Vibrio caribbenthicus BAA-2122 MbnBCs in complex with their cognate MbnAs (22). This interaction resembles a common RiPP precursor peptide recognition element involving the formation of an extended β sheet between a winged helix-turn-helix motif of a biosynthetic enzyme and its substrate (23, 24). This interface also explains why cocrystallization of MbnBC with MbnA was not successful; MbnC residues 158 to 162 interact closely with a symmetry-related molecule in the crystal lattice ().

MS2 Analysis Shows That Both MbnA Modifiable Cysteine Residues Are Converted to Oxazolone/Thioamide Groups.

Having identified the active species and optimized the conditions for MbnBC activity, we analyzed the final product of the reaction between redMbnBClowFe_asc and MbnA by liquid chromatography–tandem mass spectrometry (MS/MS). In the MS1 chromatogram, two related species were observed with a mass difference corresponding to two protons (2.02 Da), suggesting the presence of disulfide bond in one of the species. The triply charged ions of species A presented a monoisotopic peak at 1,065.154 m/z, which corresponds to a mass of 3,192.42 Da. The m/z of the monoisotopic peak for species B was 1,065.826, deconvoluting to a mass of 3,194.46 Da (Fig. 7 and ). These parental masses from MS1 are consistent with the mass of MbnA in which two cysteine residues were modified to an oxazolone ring and a thioamide group along with the binding of one Cu (Fig. 7).

Fig. 7.

MS analysis of the product of the optimized MbnBC reaction with MbnA. The reaction mixture contains species A with a disulfide bond and species B in which the disulfide bond is reduced. Cys21 and Cys27 (orange) have each been modified to an oxazolone ring and a thioamide group. Cys24 and Cys29 (gray) form the disulfide bond in species A. All species A y ions colored in purple bind Cu. The y- and b-type ions mentioned in the text are labeled.

To localize specific modifications, both species were targeted for MS2 fragmentation experiments. In species A (disulfide bond intact), fragmentation localized the loss of 4H to Cys21, consistent with its modification to an oxazolone/thioamide group (Fig. 7). Because there was no further fragmentation between Cys24 and Cys29, the second cysteine modification could not be localized in species A, possibly due to the presence of the disulfide bond. Notably, all y-type fragment ions bound Cu (Fig. 7, fragments shown in purple), and some of the smaller fragments, including y7, y8, and y9, showed an isotopic pattern consistent with the presence of 63Cu and 65Cu (), supporting the identification of the Cu-bound form of the product. Extra Cu was not added to the sample, suggesting that modified MbnA binds trace Cu from the environment.

In MS2 fragmentation experiments of species B (reduced disulfide bond), the Cu ion was not present (presumably lost during the MS2 experiment), and more fragment ions were detected, including b25, y2, and y3. While b25 suggests that Cys24 was not modified, y2 and y3, containing Cys29, also showed no modification thereof, suggesting that the second modification, corresponding to a loss of 5H, was localized on Cys27 (Fig. 7). The loss of an extra proton compared to Cys21 (−4H) is likely due to deprotonation of the −SH group upon Cu binding. Modification at Cys27 is consistent with the structure of Ms. trichosporium OB3b Mbn and the presence of conserved modification motifs (4, 6, 25). Thus, the MS2 analysis shows that the optimized reaction conditions resulted in modification of both cysteine residues in the MbnA precursor peptide and that cleavage of the leader peptide was not required for Cys27 modification, as proposed previously (11). Although MbnBC is sufficient to modify both cysteine residues, prior work showed that Cys21 is required for any modification to occur (11). Exactly how MbnBC recognizes MbnA, selects Cys21 for initial modification, and proceeds to Cys27 remains unclear.

Role of the Proposed Third Fe Site.

The combined data indicate that the active cofactor is a mixed-valent Fe(II)Fe(III) species, leaving the role of the third Fe binding site detected by Mössbauer spectroscopy, nTDMS of MbnBChighFe (11), and crystallography unclear. To probe the functional significance of this site, two of its ligands, D208 and H210 (Fig. 1), were individually replaced with serine, and the resultant D208S and H210S variants were expressed in the presence of 250 µM Fe. According to ICP-MS measurements, the D208S and H210S variants contained 0.92 ± 0.084 and 1.17 ± 0.04 equiv Fe, respectively, slightly less than or comparable to the Fe content of MbnBChighFe (also expressed with 250 µM Fe). The 2.7 Å resolution crystal structure of the H210S variant, obtained using the surface entropy reduction variant expressed at 250 µM Fe, showed the same coordination environment for Fe1 and Fe2 as the structure obtained from MbnBC expressed in the presence of 250 µM Fe (Fig. 6), including a bridging sulfate ion, but no density whatsoever for Fe3 (Fig. 8 and ).

Fig. 8.

MbnA modification activity and structure of third Fe site variants. (A) The Fe center in the H210S variant expressed in the presence of 250 µM Fe. The Fe anomalous electron density map (purple) is superimposed at 10σ. Solvent molecules are shown as small red spheres. For activity assays with (B) D208S and (C) H210S, 40 µM MbnBC variant was reacted with 200 µM of MbnA under aerobic conditions. The insets show a difference spectrum between the final and initial time points. AU, absorbance units.

Activity assays were then performed as described for redMbnBC_asc. Both variants were able to modify MbnA. While D208S yielded about half as much product as wild-type MbnBClowFe (Fig. 8), the H210S variant was significantly less active (Fig. 8). Nevertheless, these results indicate that a third Fe ion is not essential for activity. To assess the presence of the active Fe(II)Fe(III) species, both variants were reduced with 2 equiv asc and examined by EPR spectroscopy. Each variant exhibited the gave = ∼1.9 feature attributed to the S = 1/2 Fe(II)Fe(III) species (). While the D208S spectrum was identical to that of wild-type MbnBC in terms of g-values, the spectrum of H210S at 10 K exhibited unresolved g-values, possibly reflecting the presence of multiple cofactor conformers. It is possible that this substitution disrupted the active site geometry, as evidenced by the EPR spectrum, leading to reduced MbnA modification activity. This perturbation is even more evident in the EPR spectrum of the H210S MbnBC–MbnA complex, which is very different from that of wild-type MbnBC in the presence of MbnA. This substitution likely impairs the ability to properly bind the substrate. These combined findings further support the assignment of a diiron Fe(II)Fe(III) active species.

Mechanistic Implications.

The identification of the MbnBC active species as a diiron, rather than triiron, cluster adds to an ever-expanding group of enzymes that use nonheme diiron cofactors to activate dioxygen and oxidize substrates (26, 27). There are three major families of nonheme diiron enzymes: i) the ferritin-like oxidases and oxygenases (FDOs), which include soluble methane monooxygenase (28), ribonucleotide reductase (29), aldehyde deformylating oxygenase (30), and stearoyl-acyl carrier protein desaturase (31); ii) the recently discovered family of heme-oxygenase–like diiron enzymes (HDOs); and iii) the MVDOs. All FDOs and HDOs studied to date activate O2 with the [Fe(II)]2 forms of their cofactors; a two-electron oxidative addition forms a peroxo-[Fe(III)]2 intermediate, which can either directly transform the substrate or convert to high-valent, substrate-oxidizing intermediates upon O-O cleavage. FDOs and HDOs typically carry out 2e− oxidations of their substrates and require two additional electrons, which are usually provided by a reduced nicotinamide, to convert the [Fe(III)]2 “product” of substrate oxidation back to the O2-reactive [Fe(II)]2 form.

By contrast, the MVDOs, which include the enzymes myo-inositol dioxygenase (MIOX), PhnZ, and TmpB (15–17), activate O2 via a one-electron oxidative addition at the Fe(II) of their Fe(II)Fe(III) cofactors, while the Fe(III) site coordinates the substrate. The resultant superoxo-[Fe(III)]2 intermediate can initiate substrate oxidation by abstraction of hydrogen from a partially activated aliphatic carbon. Importantly, this strategy allows MVDOs to carry out four-electron oxidations of their substrates and not to require a cosubstrate. The four-electron oxidation of a cysteine to an oxazolone ring and thioamide amide group by MbnBC is yet another reaction for this family. Given evidence that the Fe(II)Fe(III) cofactor of MbnBC can interact with its substrate MbnA in the absence of dioxygen, it may be that the reaction is initiated similarly to the MIOX and PhnZ reactions () (27). While the Fe(III) ion binds to the cysteine thiol, O2 binds to the Fe(II), forming an iron-superoxo intermediate. Then an oxidative cyclization analogous to that in the isopenicillin N synthase mechanism (32) could occur by cleaving the Cβ–H bond in the cysteine residue to form a β-lactam ring. After C–S desaturation, presumably by a ferryl intermediate, the β-lactam ring would be opened by an enzyme nucleophile followed by formation of a five-membered oxazolone ring () (27). However, we cannot exclude the possibility that O2 addition occurs before substrate binding. While these mechanistic steps are speculative, the identification of an Fe(II)Fe(III) cluster as the starting point for the catalytic cycle sets the stage for future spectroscopic and kinetic investigations. Moreover, MbnBC is unique in that it performs the reaction twice on the same substrate, modifying two specific cysteine residues. How MbnBC distinguishes the modifiable cysteines from other cysteines in MbnA and why disruption of Cys21 precludes modification of Cys27 (11) remain open questions.

Materials and Methods

Expression and Purification of MbnBC Proteins.

A construct of Ms. trichosporium OB3b MbnBC that includes the mbnC gene with an N-terminal His6 tag and a tobacco etch virus cleavage site and the mbnB gene with a C-terminal Strep-II tag was used (11). The D208S and H210S MbnBC variants were generated by site-directed mutagenesis using a QuikChange Lightning kit (Agilent) (). All three proteins were expressed and purified as reported previously (11) except that the concentration of Fe in the autoinduction growth medium was varied. MbnBClowFe was expressed with 160 µM ferrous ammonium sulfate, while all other proteins including MbnBChighFe, D208S, and H210S were expressed with 250 µM ferrous ammonium sulfate (MilliporeSigma). 57Fe-labeled MbnBC was also prepared as described previously (11) except that the concentration of 57Fe (Isoflex) added to the growth medium was 160 µM for MbnBClowFe. A buffer containing 25 mM 3-(N-morpholino)propanesulfonic acid (MOPS), pH 7.2, 250 mM NaCl, and 10% glycerol was used for all experiments unless noted.

ICP-MS Analysis.

ICP-MS experiments were performed at the Quantitative Bio-element Imaging Center at Northwestern University. The samples were digested in concentrated nitric acid (Thermo Fisher Scientific) by incubating at 65 °C for 1 h and were then diluted with Milli-Q water to a final solution of 3% nitric acid. An element standard containing Fe (Inorganic Ventures) was used to prepare a 100-ng/mL standard solution in 3% nitric acid (vol/vol). Online dilution was carried out by the prepFAST system and was used to generate calibration curves consisting of 100, 50, 10, 2, 1, and 0.5 ng/g of Fe. Samples were measured using a Thermo iCAPQ ICP-MS instrument equipped with a CETAC ASX260 autosampler (Thermo Fisher Scientific). The isotopes selected for analysis were 56Fe, 57Fe, 89Y, and 115In (89Y and 115In were used as internal standards for data interpolation and machine stability). The data were analyzed using QTEGRA software. All reported values were measured with three to five independent protein samples.

Preparation of 57Fe-Labeled MbnBChighFe and MbnBClowFe for Activity Assays, EPR Spectroscopy, and Mössbauer Spectroscopy.

Purified 57Fe-labeled MbnBC was either oxidized or reduced for further experiments. For oxMbnBC, samples were incubated with 1.5 equiv of potassium ferricyanide (MilliporeSigma) for 20 min at 4 °C followed by removal of excess potassium ferricyanide using a Superdex 75 size exclusion column. All reduced MbnBC samples were prepared in either a Coy anaerobic chamber or an anoxic chamber (Labmaster, MBraun) using oxMbnBC samples. The reduction titration was performed by adding 0.5, 1, and 2 equiv of sodium asc (Acros Organics) or 2 equiv sodium dithionite (MilliporeSigma) followed by incubation at 4 °C for 20 min. Samples of oxMbnBC, redMbnBC_asc (2 equiv asc), and redMbnBC_dt (2 equiv of dithionite) were used for activity assays and characterization by Mössbauer spectroscopy. Samples of MbnA-bound redMbnBChighFe_asc for EPR and Mössbauer spectroscopic analyses were prepared in the anaerobic chamber, by incubation with 2 equiv of MbnA for 20 min at 4 °C.

Activity Assays.

Stopped-flow absorption spectroscopy was performed using an Applied Photophysics Ltd. (Leatherhead) SX20 stopped-flow spectrophotometer at 5 °C in an anoxic chamber (Labmaster, MBraun). The instrument was set for single-mixing with an optical path length of 1 cm and a photodiode-array detector. An anaerobic solution of MbnBC (300 µM) in purification buffer (25 mM MOPS, pH 7.2, 250 mM NaCl, and 10% glycerol) was rapidly mixed with an equal volume of O2-saturated buffer containing MbnA (600 µM), and 1,000 points of absorption spectra were acquired in a logarithmic time scale.

All other activity assays were performed using an Agilent Cary 3500 Compact Peltier UV-vis spectrophotometer. For the comparison between MbnBClowFe and MbnBChighFe samples, 100 µM MbnBC (150 µL) was reduced with 2 equiv sodium asc (Acros Organics) in a Coy anaerobic chamber and transferred into a quartz cuvette. The reaction was initiated by adding 1 µL of 20 mM MbnA (Bio-Synthesis Inc.) and exposure to air. Spectra were collected every 15 s for 15 min. Activity assays with 57Fe-labeled MbnBClowFe to compare different oxidation states and with the D208S and H210S variants were initiated by mixing 80 µM MbnBC in 120 µL of deoxygenated buffer with 400 µM MbnA in 120 µL aerobic buffer (aerobic buffer was used to provide O2 for the reaction). The reaction was then monitored for 1 h, collecting spectra with an interval of 15 s.

EPR Spectroscopy.

Samples of 57Fe MbnBChighFe used for reductive titrations and investigation of MbnA binding were prepared at a protein concentration of 1 mM. Samples were transferred to custom quartz X-band EPR tubes (Quartz Scientific, Inc.), and continuous wave X-band EPR spectra were collected on a Magnettech MS5000X spectrometer equipped with an Oxford Instruments ESR-900 continuous flow cryostat and an Oxford Instruments ITC-300 temperature controller. Spin quantification was performed using a spin standard of equine skeletal muscle myoglobin (Mb) (Sigma) coordinated by azide. The low-spin Mb-N3 standard was quantified using the visible absorption of the Soret band of a control Mb sample without azide (ε = 188,000 M−1 cm−1 at 409 nm) (33).

For 57Fe-labeled MbnBClowFe, 0.45 mM protein was used to prepare samples of oxMbnBClowFe, redMbnBClowFe_asc, and redMbnBClowFe_dt. The oxMbnBClowFe sample (100 µL) was transferred into a custom quartz Q-band EPR tube and frozen in liquid nitrogen under aerobic conditions, while the two reduced samples were prepared anaerobically in a Coy chamber, transferred to Q-band EPR tubes, and frozen and stored in liquid nitrogen until analysis. X-band EPR measurements were performed on a Bruker ESP-300 spectrometer with a liquid helium flow Oxford Instruments ESR-900 cryostat. All samples were measured using 200 µW of microwave power, 6 G modulation, 10 G/s scan rate, and 320-ms time constant. Spectra were acquired at 3.9 K to optimize the signals from rhombic high spin Fe(III) and at 10 to 12 K to optimize the signals from Fe(II)Fe(III) antiferromagnetically coupled species and subsequently scaled to relative intensities to facilitate comparison. Spectra were also subtracted against a spectrum of buffer in order to correct the baseline and to remove the signal of a copper contaminant associated with buffer reagents. Samples of 56Fe MbnBChighFe prepared using the same procedures were also utilized to test the effects of different amounts of asc and of MbnA.

Mössbauer Spectroscopy.

Samples of 57Fe-labeled oxMbnBC, 57Fe-labeled redMbnBC_asc, and 57Fe-labeled redMbnBC_asc with MbnA (300 µL) were prepared at protein concentrations of 1 mM (MbnBChighFe) or 0.9 mM (MbnBClowFe). Mössbauer spectra were recorded on a spectrometer from SEE Co. equipped with a Janis SVT-400 variable-temperature cryostat. The reported isomer shift is given relative to the centroid of the spectrum of α-iron metal at room temperature. Simulations were carried out using the program WMOSS (SEE Co.).

Crystallization and Structure Determination.

The sequence of wild-type MbnBC from Ms. trichosporium OB3b was analyzed by the XtalPred (34) and SERp (18) servers to predict disordered loops and identify high entropy surface patches. Based on the results, eight residues at the MbnB C terminus were deleted, and residues comprising two surface patches, MbnB E67, E69, and K70 and MbnC E162, E164, and K165, were identified for mutation to alanine. This construct, with an N-terminal His6 and small ubiquitin-like modifier (SUMO) tag, was synthesized and inserted into the pCDFDuet-1 vector (GenScript). Proteins were expressed and purified as described above. The H210S MbnBC variant for crystallization was generated using a QuikChange Lightning kit (Agilent). After the initial purification using Ni-loaded 5-mL HiTrap Chelating column as described previously (11), protein was incubated with SUMO protease overnight at 4 °C with nutation. Cleaved proteins were then purified by loading the entire reaction onto the Ni-loaded 5-mL HiTrap Chelating column followed by washing with two column volumes of 25 mM MOPS, pH 7.2, 250 mM NaCl, 10% glycerol, and 10 mM imidazole and collection of the flow-through fraction. Further purification was performed using a Superdex 75 size exclusion column.

Before crystal trays were set up, all protein samples were exchanged into 25 mM MOPS, pH 7.2, and 250 mM NaCl. Initial crystals were obtained via sitting drop vapor diffusion at 10 °C overnight using 10 mg/mL protein in 2-µL drops at a 1:1 ratio with a precipitant containing 0.2 M lithium sulfate, 0.1 M Bis-Tris, pH 6.5, and 25% (wt/vol) PEG3350 (B11 in MCSGII, Anatrace). All other crystals were grown in similar conditions containing 0.2 M lithium sulfate (Alfa Aesar), 0.1 M Bis-Tris propane, pH 7.0 (Hampton Research), and 20 to 26% (wt/vol) PEG3350 (Hampton Research). Crystals were soaked in a cryoprotectant solution containing the crystallization solution with 10% ethylene glycol for 1 to 3 min and then frozen in liquid nitrogen. Data were collected at 66 K at a wavelength of 1.722 Å at the Life Sciences Collaborative Access Team (LS-CAT) beamline 21 ID-D at the Advanced Photon Source at Argonne National Laboratory. The crystals belong to space group C2221 with two MbnBC molecules in the asymmetric unit. The data were processed using XDS (35), and the initial structure was solved by Fe SAD using autoSHARP (36). All additional structures were solved by molecular replacement using Phenix (37). Model building and refinement were performed using Coot (38) and Phenix (37). Data collection and refinement statistics are summarized in .

The interaction with MbnA was modeled using the newly accessible advanced version of ColabFold (https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/beta/AlphaFold2_advanced.ipynb) with the following parameters: MbnABC oligomeric ratio of 1:1:1; msa_method jackhmmer; msa_format fas; pair_mode unpaired+paired; pair_cov 50; pair_qid 20; rank_by pTMscore; use_turbo unchecked; num_models 5; use_ptm checked; num_ensemble 1; max_recycles 3; tol 0; is_training unchecked; num_samples 1; subsample_msa checked (21). The first ranked prediction model is presented. All crystal structure figures were prepared with UCSF ChimeraX (39).

Tandem MS.

A reaction mixture of redMbnBClowFe_asc with MbnA was analyzed by tandem MS. Prior to analysis, the sample was reduced by the addition of 2 mM dithiothreitol (MilliporeSigma) and incubated at 40 °C for 30 min. Ultra–high performance liquid chromatography (UHPLC)–high-resolution MS/MS analysis was performed using an Agilent 1290 Infinity II UHPLC System placed in line with a Thermo Q-Exactive mass spectrometer. Analytes were separated on a Kinetex C18 reversed-phase column [50 × 2.1 mm i.d., 1.3-µm particle size; (Phenomenex)] using a flow rate of 0.7 mL/min. The solvent system consisted of solvent A (0.1% formic acid, 5% acetonitrile [ACN] in H2O) and solvent B (0.1% formic acid in ACN). Metabolite separation was achieved following a 10-min gradient as follows: 0 to 45% B (0 to 8.4 min), 45 to 100% B (8.4 to 8.8 min), 100% B (8.8 to 9.0 min), and 100 to 0% B (9.0 to 10 min). Electrospray ionization was achieved using a heated electrospray ionization source with capillary temperature at 280 °C, sheath gas at 60 units, aux gas at 15 units, and spray voltage at 3.5 kV. In positive polarity mode, MS1 data were collected at a resolution of 35,000 in a scan range of 200 to 2,000 m/z. Precursors were fragmented in the HCD cell, and MS/MS scans were collected at a resolution of 17,500 using targeted parallel reaction monitoring analysis with an isolation window of 2.0 m/z and a range of normalized collision energies between 20 and 25 units. Spectra were processed using Xcalibur Qual Browser (Thermo Scientific), deconvoluted using the Xtract algorithm, and analyzed with ProSight Lite (Proteinaceous, Inc). Fragment ions were matched with an error tolerance of 15 ppm.