Crystal Structure of Bacillus subtilis Cysteine Desulfurase SufS and Its Dynamic Interaction with Frataxin and Scaffold Protein SufU

The biosynthesis of iron sulfur (Fe-S) clusters in Bacillus subtilis is mediated by a SUF-type gene cluster, consisting of the cysteine desulfurase SufS, the scaffold protein SufU, and the putative chaperone complex SufB/SufC/SufD. Here, we present the high-resolution crystal structure of the SufS homodimer in its product-bound state (i.e., in complex with pyrodoxal-5ʹ-phosphate, alanine, Cys361-persulfide). By performing hydrogen/deuterium exchange (H/DX) experiments, we characterized the interaction of SufS with SufU and demonstrate that SufU induces an opening of the active site pocket of SufS. Recent data indicate that frataxin could be involved in Fe-S cluster biosynthesis by facilitating iron incorporation. H/DX experiments show that frataxin indeed interacts with the SufS/SufU complex at the active site. Our findings deepen the current understanding of Fe-S cluster biosynthesis, a complex yet essential process, in the model organism B. subtilis.


Introduction
Iron sulfur (Fe-S) clusters are amongst the most versatile enzyme cofactors in Nature, as they are involved in cellular respiration, carbohydrate metabolism, DNA repair and various other vital functions throughout all kingdoms of life [1][2][3][4]. The biosynthesis of Fe-S clusters must be tightly regulated because of the toxicity of free sulfur and iron. Therefore, the systems for Fe-S cluster biogenesis are mainly conserved from bacteria to human, although elaborate transport systems have diverged through evolution [5][6][7]. Three distinct systems have been described for prokaryotic cells: i.) the NIF system (nitrogen fixation, [8]), ii.) the ISC system (iron-sulfur cluster, [9]) and iii.) the SUF system (sulfur mobilization, [10]). While the NIF system is specific for nitrogenase maturation in azototrophic bacteria, the ISC and/or SUF systems function in housekeeping in most-if not all-bacteria. In some bacterial species (e.g., Escherichia coli), both systems are found, while others (e.g., B. subtilis) rely on a single type of Fe-S machinery Quest, Biorad) using a gradient of 5-100% HEPES buffer B (50 mM HEPES, pH 8, 300 mM NaCl, 250 mM imidazole) over 20 min. The elution fractions were concentrated and further purified by size exclusion chromatography (HiLoad 26/60 Superdex 200, GE Healthcare Life Sciences) in HEPES buffer C (50 mM HEPES, pH 8, 300 mM NaCl). Fractions containing the target protein were concentrated, flash frozen in liquid nitrogen and stored at -80°C in HEPES buffer C supplemented with 10% glycerol.
Crystallization, data collection, and structure determination Crystallization was performed by the sitting-drop method at 20°C in 0.6-μl drops consisting of equal parts protein and crystallization solutions. BsSufS crystallized at 20 mg/ml within one week in 0.1 M HEPES, pH 7.5, 50% (v/v) PEG 400. Prior to data collection, crystals were cryoprotected in a solution consisting of the well solution supplemented with 20% glycerol and then flash-frozen in liquid nitrogen. Data were collected under cryogenic conditions at the European Synchrotron Radiation Facility at Beamline ID29. Data were processed with XDS [34] and scaled with CCP4-implemented SCALA [35]. The structure was determined by molecular replacement with PHASER [36], manually built in COOT [37], and refined with PHENIX [38]. The SufS homolog from Brucella suis (PDB ID 4W91) was employed as a search model. Figures were prepared with Pymol (www.pymol.org). Coordinates and structure factors were deposited in the Protein Data Bank with the accession code 5J8Q.

Hydrogen/deuterium exchange experiments
H/DX mass spectrometric analysis of the samples was performed using an automated H/DX setup (Waters) including a two-arm robotic autosampler (LEAP Technologies), an ACQUITY UPLC M-Class System and HDX Manager (Waters). For the exchange reaction, BsFra, BsSufU, BsSufS, BsFra/BsSufU, BsFra/BsSufS, BsFra/BsSufU/BsSufS, and BsSufU/BsSufS (60 μM final concentration of each component) were individually prepared in H 2 O buffer (25 mM Tris-Cl, pH 7.5, 100 mM NaCl) [32] and pre-cooled to 1°C. For each LC-MS run, 7.5 μL of protein solution was pipetted into a fresh vial on an exchange plate at 25°C and diluted with 61.8 μL of either H 2 O buffer (t0 runs) or D 2 O buffer (exchange runs). After incubation for 15, 30, 60, or 600 s, 55 μL of the reaction solution was transferred to a fresh vial containing 55 μL of quenching solution (400 mM H 3 PO 4 /KH 2 PO 4 , pH 2.2, pre-dispensed and pre-cooled to 1°C for 10 min before the first run). After quenching, 95 μL of the resulting solution was immediately injected into the HDX Manager.
Digestion was done online using an Enzymate BEH Pepsin Column (Waters) at 20°C with water/0.1% formic acid at a flow rate of 100 μL/min. Subsequently peptic peptides were trapped at 0.5°C using a C18 trap column. Separation of peptides was achieved at 0.5°C utilizing a 1 x 100 mm ACQUITY UPLC BEH C18 1.7 μm column (Waters) at a flow rate of 30 μL/min with the following gradient of solvents A (water/0.1% formic acid) and B (acetonitrile/0.1% formic acid): Linear increase from 5-35% B within 7 min, followed by a ramp to 85% B within 1 min and isocratic 85% B for additional 2 min. Finally, the column was washed with 95% B for 1 min and re-equilibrated with 5% B for 5 min.
During separation of peptides using the chromatographic column, the pepsin column was washed by injecting 3 × 80 μL 4% acetonitrile and 0.5 M guanidinium chloride. HDMSe was used for t0 peptide detection and HDMS for exchanged peptides. Lock mass spectra were measured every 45 s using Glu-fibrinopeptide B as a standard ([M+H] 2+ = 785.8427 m/z). t0 peptide identification was performed using ProteinLynx Global SERVER (Waters) with custom databases and the setting "no enzyme". Final assignment of deuterium incorporation was done with DynamX 3.0 (Waters).

Microscale thermophoresis
The determination of binding constants was done by microscale thermophoreis using a Monolith NT.115 instrument (NanoTemper). 100 μM BsFra was labeled at 8°C overnight using the Monolith NT.115 Protein Labeling Kit RED MALEIMID (cysteine-reactive; NanoTemper). Labeled BsFra was buffer-exchanged into a binding buffer (25 mM Tris-Cl, 100 mM NaCl, 10 mM 2-mercaptoethanol, pH 7.5) and the concentration was adjusted to 0.25 μM. A serial dilution series of BsSufS, BsSufU, and BsSufS/BsSufU was prepared in the same buffer in a range from 560 μM to 0.030 μM and then mixed 1:1 with labeled BsFra. The titration was transferred into NT.115 MST Premium Coated capillaries (NanoTemper). Measurements were performed at 25°C, 20% LED power and 20% MST power with a heating time of 30 s and cooling time of 5 s. The binding constant for each interaction was calculated from the average of three measurements using the NT Analysis software (NanoTemper).

Cysteine Desulfurase Activity Assay
The activity of BsSufS was measured by the amount of sulfide released during conversion of cysteine to alanine. Free sulfide was quantified using N,N-dimethyl-p-phenylenediamine sulfate (DMPD) and FeCl 3 as described previously [30,31,39]. We incubated 0.5 μM BsSufS with 10 μM BsSufU and 50 μM BsFra in 25 mM Tris-Cl, pH 7.5, 100 mM NaCl, 5 mM dithiothreitol (DTT) for 5 min at room temperature. The 200 μL reaction was started by addition of L-cysteine (2 mM) and quenched after 10 min by addition of 25 μL of DMPD (20 mM in 7.2 M HCl) and 25 μL FeCl 3 (30 mM in 1.2 M HCl). After 30 min of incubation in the dark, the absorbance was measured at 670 nm. All reactions were carried out in triplicate.

Results
Crystal structure of BsSufS in the product-bound state As no structure of B. subtilis SufS was available at the start of our study, we sought to fill this gap. BsSufS was thus crystallized and its structure was determined to 1.7 Å resolution by molecular replacement using the B. suis homolog (PDB ID 4W91) [41] as a search model ( Table 1). One molecule of SufS was found within the asymmetric unit (AU) and could be built to completeness. The crystal structure of BsSufS is highly similar to that of E. coli SufS (PBD ID 1I29) with a root mean square deviation (r.m.s.d.) of 1.28 Å over 407 Cα atoms (S1 Fig). The BsSufS monomer forms a tightly intertwined homodimer with another monomer across the crystallographic symmetry axis (Fig 1A). The interface and architecture of the BsSufS homodimer closely resemble those of E. coli SufS [42].
BsSufS assumes a type I fold of the aminotransferase class V family, consisting of an overall α/β fold that is highly similar to that of E. coli SufS, IscS, NifS, and CsdA [41,[43][44][45][46]. The enzyme consists of two domains: A large, N-terminal domain (residues 1-294) and a smaller, C-terminal domain (residues 295-406). In the N-terminal domain, a 7-stranded, parallel β-sheet is sandwiched between several α-helices to form a tightly packed core. This domain harbors the active site pocket, where the cofactor PLP is bound to Lys224 as an aldimine. Furthermore, residues 253-265 (herein referred to as the 'β-hook') form a hairpin-like structure that latches on to the other monomer. The small, C-terminal domain consists of a 4-stranded, parallel β-sheet and four α-helices containing the flexible 'Cys361-loop' with the nucleophile Cys361 ( Fig 1B). The dimer interface covers much of the active site pocket and protects PLP and Cys361. We found BsSufS in the product-bound state, with alanine bound near Cys361-persulfide, suggesting that one reaction cycle had occurred.

BsSufS forms a homodimer in solution
We performed time-resolved H/DX experiments in order to understand the dynamics of BsSufS in solution. We identified 176 peptides between 5 and 20 amino acids in length, resulting in 98.6% sequence coverage and a redundancy of 4.4 (Fig 2). BsSufS was incubated in D 2 O buffer for 15, 30, 60, and 600 s, and the exchange of backbone amide hydrogen to deuterium was measured for each peptide. In good agreement with the expectations from our crystallographic analysis, the solvent-exposed N-and C-termini of BsSufS exchanged readily, whereas most of the core structure was strongly protected (compare to Fig 1). Strong protection was observed for BsSufS regions involved in formation of the homodimer interface, validating the existence of the enzyme as a homodimer in solution [47] ( Fig 2B).
BsSufU binds to the BsSufS homodimer Next, we sought to investigate the interaction interface and conformational dynamics of SufS and SufU. Therefore, SufS was incubated with SufU in deuterated buffer. Upon completion of the H/DX reaction and peptic digest, we observed that several peptides of SufS showed a change in deuterium uptake when compared to their counterparts from H/DX with SufS alone (see above). Most notably, the C-terminal α-helix of SufS showed a high degree of protection in the presence of SufU (Fig 3). This observation agrees with recent findings that the C-terminus of E. coli IscS is required for interaction with IscU [48,49]. In addition, residues 50-63 of SufS (herein referred to as the 'α-hinge') were protected against deuterium incorporation, suggesting the presence of a second SufU interaction site at the SufS homodimer interface. We then utilized our H/DX approach to analyze changes in SufU upon binding SufS. We identified 87 peptides, which covered 93% of the sequence with a redundancy of 6.0 (Fig 4).
The NMR solution structure of apo-SufU (PDB ID 2AZH) reveals a compact, α/β fold. Our H/ DX analysis supports this structure in solution, as medium to low H/D exchange was observed for the core of SufU, whereas high H/D exchange was observed for an extended loop region between the N-terminal helix α1 and strand β1 (the 'α/β linker') as expected for a solvent- exposed, unstructured region. In the presence of SufS, the 'α/β-linker' of SufU is strongly protected from deuterium uptake, which indicates that SufS interacts with SufU at this location ( Fig 4C). Residues 117-130, which form a loop carrying Cys128 (the 'Cys128-loop'), are also protected from deuterium uptake, suggesting a second SufS binding site. Therefore, we conclude that the 'α/β-linker' and 'Cys128-loop' of SufU bind to the C-terminus and 'α-hinge' of SufS, respectively, bringing the active site of SufS and SufU into close proximity. A similar scenario has been proposed for the interaction of IscU and IscS in E. coli [49].

Binding of BsSufU to BsSufS induces conformational changes in both proteins
Next, we analyzed the H/DX data for changes in the dynamic behavior of SufS in the presence of SufU. Significantly higher H/D exchange rates were observed for the active site pocket of SufS when SufU was present, particularly at the 'Cys361-loop' and 'β-hook' of SufS (Fig 3). These observations strongly suggest that the SufS homodimer opens in the presence of SufU, allowing the 'Cys361-loop' to move freely. This is in agreement with findings in E. coli, where the 'Cys361-loop' of IscS undergoes a major, 14-Å movement during transfer of sulfur to IscU [48,49].
To further characterize the interaction of BsFra with its partners we applied H/DX experiments. Analysis of BsFra alone yielded 61 unique peptides with 98% sequence coverage and a redundancy of 4.7 (Fig 6). Human, S. cerevisiae, and E. coli frataxin (PDB ID 3S4M, 2GA5, and 2EFF, respectively) assume an α/β fold in which two α-helices are stacked against a 6-stranded β-sheet, with the N-terminal α-helix harboring several acidic residues (the 'acidic ridge') [21]. BsFra (PDB ID 2OC6) has an additional short, N-terminal helix (herein referred to as 'α1-helix') followed by the helix carrying the 'acidic ridge' (compare S2 Fig). The 'α1-helix' shows medium to high deuterium uptake in BsFra, suggesting that in solution it can move freely. Furthermore, a helix connecting β4 to β5 (the 'EDDI-helix') is present in BsFra but not homologous frataxin structures [19][20][21]. A high rate of H/D exchange was observed for this 'EDDI-helix', especially for residues E75, D76, D77, and I78.
We repeated the H/DX experiments of BsFra in the presence of BsSufU (S3 Fig), BsSufS (S4  Fig), and the BsSufS/BsSufU complex (Fig 7). These measurements were compared with those of BsFra alone for differences in deuterium uptake. In the presence of BsSufU, we observed a decrease of H/D exchange in β-strands 1 and 2 of BsFra (residues 37-46, herein referred to as the 'KWNloop'), which harbor residues K40, W41, and N42 ( Fig 7C). As we found that the same residues were protected in the reaction of BsFra and BsSufU, we conclude that BsFra binds to BsSufU at the 'KWN-loop' area. The involvement of the 'KWN-loop' in the interaction with BsSufU is in agreement with previous studies in which mutations of in this region disrupted the interaction [52][53][54][55].
In the presence of both BsSufU and BsSufS, BsFra was additionally to the KWN loop protected from H/D uptake at the 'acidic ridge' (residues 9-25) (Fig 7 and S4 Fig), suggesting an interaction with BsSufS. This is in accordance with previously published work on E. coli IscS/CyaY [51,56]. Furthermore, BsSufS induces an increased rate of deuterium uptake at the N-terminal 'α1-helix' of BsFra (Fig 7). We propose that, upon binding of BsFra to BsSufS, the 'α1-helix' undergoes a structural rearrangement. Interestingly, BsFra was protected from deuterium uptake at the C-terminus and 'EDDI helix' (residues 74-85) compared to the protein alone. However, this was found in the BsFra/BsSufS and BsFra/BsSufU reactions (Fig 7, S3 Fig and S4 Fig), suggesting that the either protein can stabilize these highly dynamic areas (compare Fig 6). Taken together, the results indicate that BsFra binds to BsSufS and BsSufU as well as to the BsSufS/BsSufU complex. A similar binding of frataxin to a cysteine desulfurase/scaffold protein complex was described in E. coli [51,56].
Binding of BsFra to BsSufU/BsSufS induces conformational changes in all three proteins Next, we tested whether the binding of BsFra would alter the interaction of BsSufS with BsSufU. We analyzed the H/DX data of the BsFra/BsSufU/BsSufS interaction and compared it to that of the BsSufS/BsSufU complex and proteins alone (see above). The scaffold protein BsSufU showed decreased deuterium uptake at the C-terminal helix ('α5-helix'), indicating binding to BsFra (Fig 8 and S3 Fig). The cysteine desulfurase BsSufS showed protection from H/D exchange at the 'Cys361-loop' and the 'β-hook' when BsFra was present (Fig 8), also indicating a binding event. Several positively charged residues are located in this region (e.g., R356, H359, H360, and K367) and may bind the negatively charged residues of the 'acidic ridge' of BsFra (see above). In contrast to the reaction of BsSufS/BsSufU with BsFra, no increase in deuterium uptake was observed at the active site pocket of BsSufS when BsFra was present (compare Fig 8  to Fig 3). The BsSufU binding site on BsSufS (see above) was still observable, supporting the idea of simultaneous binding.
We propose that BsFra binds via the 'acidic ridge' to the positively charged 'Cys361-loop' and 'β-hook' of BsSufS. BsFra may act as a clamp at the BsSufS homodimer interface, locking the 'Cys361-loop' and 'β-hook' and thus preventing the opening of the BsSufS active site pocket. Furthermore, BsFra binds via its 'KWN-loop' to the 'β-sheet surface' of BsSufU. A similar binding site was suggested for E. coli IscS/CyaY [51,56].
To test whether the binding of BsFra to the BsSufS/BsSufU complex alters the activity of BsSufS we conducted cysteine desulfurase activity assays in vitro. We found that BsSufU greatly enhances the activity of BsSufS, in agreement with previous reports [30,33]. The introduction of BsFra, however, does not appear to affect the activation of BsSufS by BsSufU (Fig 9). This is in agreement with the E. coli IscS/CyaY interaction, where conversion of cysteine to alanine by  IscS is not affected by CyaY [56]. We next analyzed whether frataxin has an effect on the formation of Fe-S clusters in vitro. In an assay described by Albrecht and colleagues [30], we incubated BsSufS and BsSufU in the presence of ferrous iron and cysteine and monitored the formation of Fe-S clusters by UV-Vis spectroscopy. When BsFra was present in the reaction mixture the initial rate of formation was unaffected, although we observed a very minor increase in the yield of Fe-S clusters formed on BsSufU. Such a subtle change could indicate that either BsFra does not efficiently donate iron to BsSufU under assay conditions or that its presence is redundant under the supplied concentration of iron. Additionally, other, unidentified components may be required for efficient transfer.

Discussion
Bacillus subtilis has five paralogous cysteine desulfurase genes, namely nifS [57], yrvO [58], nifZ [59] and ycbU [60]. However, only the product of the sufS gene is involved in biosynthesis of Fe-S clusters. Until now, no structural information was available on the B. subtilis cysteine desulfurase. We were able to determine the crystal structure of the BsSufS dimer to 1.7 Å resolution, revealing high structural homology to its E. coli counterpart. To better understand the conformational dynamics of SufS and its interaction with SufU and Fra, we conducted H/DX experiments with proteins from the Gram-positive model organism B. subtilis. Many insights into the SUF system were previously obtained from studies on E. coli, in which the SUF system only acts as a backup under stress conditions [7]. The ISC system mediates the major "housekeeping" functions for Fe-S cluster biosynthesis in E. coli [9]. In B. subtilis, we find similarities to both E. coli ISC and SUF systems [30,31,33]. We applied hydrogen/deuterium exchange experiments on BsSufS in order to investigate structural dynamics upon its interaction with the putative iron delivery protein BsFra and the scaffold protein BsSufU. H/DX detects the exchange of hydrogen on the backbone amides with deuterium in the solvent, where the exchange rate of highly dynamic or surface-exposed areas is rapid compared to residues that are buried in the protein core or otherwise protected. Investigation of BsSufS confirmed the homodimer interface in solution and suggested that the enzyme is tightly packed. Nevertheless, the strength of H/DX lies in its ability to detect changes in residues following a binding event. We showed that the C-terminal region and 'α-hinge' of BsSufS interact with the long 'α/β-linker' and 'Cys128-loop' of BsSufU (Fig 10). This interaction is accompanied by an opening of the BsSufS dimer interface and rearrangement of the 'Cys361-loop'. BsSufU, on the other hand, showed H/D exchange in the loop carrying Cys41 and Asp43. BsSufU fits on the BsSufS interaction site, which brings the 'Cys361-loop' of BsSufS and the 'Cys41-loop' of BsSufU into close proximity for persulfide transfer.
The E. coli SufS/SufE interaction was previously characterized by H/DX [61] and the crystal structure of the homologous E. coli CsdA/CsdE enzyme complex was solved recently [41]. In both cases, SufE/CsdE binds near the active site of SufS/CsdA to accept the sulfur as persulfide, which is then passed on to the scaffold protein complex SufBC 2 D of the E. coli SUF system. E. coli SufB has been shown to act as a scaffold and carries an FAD binding site that is possibly important for the reduction of Fe 3+ to Fe 2+ in the generation of Fe-S clusters [18]. In B. subtilis, the roles of SufB, SufC, and SufD are still unclear. BsSufB, though conserved, lacks the Fe-S cluster and FAD binding sites, suggesting an important role which differs from that of its E. coli homolog. We hypothesize that the BsSufBCD enzyme complex is involved in later steps of Fe-S cluster maturation, similar to the E. coli HscA/HscB proteins of the ISC system [62][63][64]. Taken together, our data suggest that binding of BsSufU to BsSufS activates the latter for transfer of sulfur from persulfide to the active site of BsSufU.
In the present study, we further established an interaction between BsSufS/BsSufU and the B. subtilis frataxin homolog BsFra. Previous biochemical characterization showed that BsFra can bind ferric and ferrous iron, and plays a role in Fe-S cluster biosynthesis and transfer to a target protein [28,29]. The frataxin family has been extensively studied in E. coli, S. cerevisiae, and H. sapiens. Generally, this family consists of a conserved α/β sandwich fold and harbors several acidic residues at the N-terminal α-helix and the first β-sheet (S2 Fig). In vitro enzyme assays measuring the rate at which Fe-S clusters form on the scaffold protein in the presence of a cysteine desulfurase and frataxin highlights a discrepancy: While frataxin inhibits cluster formation in the E. coli system [51,56], it enhances the rate of cluster formation in the yeast system [65]. It was shown that the cysteine desulfurase determines whether frataxin acts as an inhibitor or activator [66]. It was further shown that a single point mutation on the scaffold protein renders E. coli as frataxin-dependent and S. cerevisiae as frataxin-independent [67][68][69]. We conducted analogous assays with BsFra and found that, in contrast to E. coli and S. cerevisiae homologs, it did not appear to alter the initial rate of cluster formation, but resulted in a very minor increase in cluster yield. The cluster formed on BsSufU is highly labile and most likely degrades to an uncharacterized species upon isolation of holo-BsSufU [30,50]. That we were unable to isolate holo-BsSufU may suggest that our current model for Fe-S biosynthesis includes only the minimal number of participants.
We performed H/DX experiments to analyze the interaction of BsSufS and BsSufU with BsFra. We observed binding of BsFra to BsSufU and BsSufS, which did not change in the presence of BsSufU/BsSufS and therefore indicates formation of a BsFra/BsSufU/BsSufS complex in solution. The binding epitope of BsFra associates with the 'Cys361-loop' and 'β-hook' motifs of BsSufS (Fig 10). The interaction points between BsFra and BsSufU are the 'KWN-loop' and 'βsheet surface', respectively. The short 'α1-helix' of BsFra appears to rearrange upon binding BsSufS, forming an extended α1-α2 helix hybrid, which then fits in the groove of BsSufS/ BsSufU. This interaction brings the putative iron binding site of BsFra into close proximity to the active site of BsSufU, and it is reasonable to propose that iron incorporation is mediated by the 'acidic ridge' of BsFra. Previous SAXS measurements of E. coli IscS/IscU/CyaY revealed a similar binding mode [51], and characterization of the interaction of frataxin with the ferrochelatase established that BsFra binds with its acidic residues in a similar fashion to the iron acceptor BsHemH [32]. In B. subtilis, it is well established that frataxin is involved in incorporating iron into the nascent Fe-S cluster in vivo [33], but whether frataxin also serves as an intracellular iron carrier remains elusive.
We determined that BsSufU binds BsSufS and further showed how BsFra binds to the BsSufU/BsSufS complex as well as the individual proteins. In accordance with what is known about E. coli IscS/IscU/CyaY [51,56], the interaction of BsSufU/BsSufS is tight compared to that of BsFra with the complex. We assume that the interaction of BsFra to the complex is transient and that BsFra competes with additional participants of the Fe-S biogenesis pathway, as is known for E. coli IscS/Fdx [17]. The results presented here represent a single snapshot in a highly dynamic assembly process whose parts have not been fully identified.
The biogenesis of Fe-S clusters is a multistep process consisting of sulfur abstraction from cysteine, persulfide transfer, iron delivery and incorporation, and reductive generation of the Fe-S cluster, followed by transfer onto a target protein. Further studies will be necessary to identify any additional participants and determine how the biosynthetic steps are organized. In particular, the identity of the electron donor for Fe-S cluster biogenesis in vivo is unknown, and the role of SufB/SufC/SufD is yet to be determined.
The authors declare no financial conflict of interest.