The β-latch structural element of the SufS cysteine desulfurase mediates active site accessibility and SufE transpersulfurase positioning

Under oxidative stress and iron starvation conditions, Escherichia coli uses the Suf pathway to assemble iron-sulfur clusters. The Suf pathway mobilizes sulfur via SufS, a type II cysteine desulfurase. SufS is a pyridoxal-5′-phosphate–dependent enzyme that uses cysteine to generate alanine and an active-site persulfide (C364-S-S-). The SufS persulfide is protected from external oxidants/reductants and requires the transpersulfurase, SufE, to accept the persulfide to complete the SufS catalytic cycle. Recent reports on SufS identified a conserved "β-latch” structural element that includes the α6 helix, a glycine-rich loop, a β-hairpin, and a cis-proline residue. To identify a functional role for the β-latch, we used site-directed mutagenesis to obtain the N99D and N99A SufS variants. N99 is a conserved residue that connects the α6 helix to the backbone of the glycine-rich loop via hydrogen bonds. Our x-ray crystal structures for N99A and N99D SufS show a distorted beta-hairpin and glycine-rich loop, respectively, along with changes in the dimer geometry. The structural disruption of the N99 variants allowed the external reductant TCEP to react with the active-site C364-persulfide intermediate to complete the SufS catalytic cycle in the absence of SufE. The substitutions also appear to disrupt formation of a high-affinity, close approach SufS–SufE complex as measured with fluorescence polarization. Collectively, these findings demonstrate that the β-latch does not affect the chemistry of persulfide formation but does protect it from undesired reductants. The data also indicate the β-latch plays an unexpected role in forming a close approach SufS–SufE complex to promote persulfide transfer.

Under oxidative stress and iron starvation conditions, Escherichia coli uses the Suf pathway to assemble iron-sulfur clusters. The Suf pathway mobilizes sulfur via SufS, a type II cysteine desulfurase. SufS is a pyridoxal-5 0 -phosphate-dependent enzyme that uses cysteine to generate alanine and an active-site persulfide (C 364 -S-S -). The SufS persulfide is protected from external oxidants/reductants and requires the transpersulfurase, SufE, to accept the persulfide to complete the SufS catalytic cycle. Recent reports on SufS identified a conserved "β-latch" structural element that includes the α 6 helix, a glycine-rich loop, a β-hairpin, and a cis-proline residue. To identify a functional role for the β-latch, we used sitedirected mutagenesis to obtain the N99D and N99A SufS variants. N99 is a conserved residue that connects the α 6 helix to the backbone of the glycine-rich loop via hydrogen bonds. Our x-ray crystal structures for N99A and N99D SufS show a distorted beta-hairpin and glycine-rich loop, respectively, along with changes in the dimer geometry. The structural disruption of the N99 variants allowed the external reductant TCEP to react with the active-site C364-persulfide intermediate to complete the SufS catalytic cycle in the absence of SufE. The substitutions also appear to disrupt formation of a highaffinity, close approach SufS-SufE complex as measured with fluorescence polarization. Collectively, these findings demonstrate that the β-latch does not affect the chemistry of persulfide formation but does protect it from undesired reductants. The data also indicate the β-latch plays an unexpected role in forming a close approach SufS-SufE complex to promote persulfide transfer.
Fe-S clusters are ubiquitous enzyme cofactors involved in myriad biological pathways across all three domains of life. Owing to the reactive nature of the iron and sulfur building blocks, Fe-S clusters are assembled by highly choreographed multiprotein pathways. These pathways exhibit three conserved steps: (1) sulfur mobilization from cysteine, (2) assembly of the nascent Fe-S cluster on a scaffold complex, and (3) trafficking of the cluster to downstream apoproteins. Two major pathways exist for Fe-S cluster biogenesis, the ironsulfur cluster (Isc) pathway (1,2), which occurs in all three domains of life, and the sulfur formation (Suf) pathway (3) which occurs in plants and prokaryotes. Escherichia coli contains an example of both the Isc pathway, which is the primary source of Fe-S clusters under normal conditions, and the Suf pathway, which is activated under conditions that disrupt Fe-S metabolism, such as oxidative stress and iron starvation.
The first step in both E. coli pathways is the mobilization of sulfur from cysteine by a pyridoxal-5 0 -phosphate (PLP)dependent cysteine desulfurase (4). These enzymes catalyze the formation of alanine and a covalent persulfide intermediate on an active-site cysteine (-Cys-S-S -) on an S-transfer loop. The persulfide is then transferred to the next protein in the assembly pathway. In the E. coli Isc pathway, IscS acts as the cysteine desulfurase and transfers the persulfide directly to the IscU scaffold protein. This is thought to be accomplished by movement of a flexible S-transfer loop containing the activesite cysteine. In the E. coli Suf pathway, SufS is the cysteine desulfurase. The active-site cysteine of SufS (C364) is located on a short, inflexible S-transfer loop buried 10 Å below the surface of the active-site cleft. Owing to the difference in active-site architecture, SufS requires the accessory protein, SufE, to act as a transpersulfurase, which transfers the SufS-C364-S-Sto SufE-C51-S-S - (Fig. 1) (5). In vivo, SufE is regenerated by transferring the persulfide to SufB-C254 in the SufBC 2 D Fe-S cluster assembly complex. For in vitro activity, tris(2-carboxyethyl)-phosphine hydrochloride (TCEP) can reduce the SufE persulfide to generate SH 2 . It has been shown that the SufS persulfide is protected from external reductants (6), including TCEP and DTT, and the SufS/SufE system is functional in the presence of hydrogen peroxide (7). All of these characteristics are consistent with the biological function of the Suf system under conditions of oxidative stress.
IscS and SufS serve as examples of type I and type II cysteine desulfurases, respectively (4). Type I systems are defined by the long, flexible S-transfer loop containing the active-site cysteine and do not require a transpersulfurase partner. Type II enzymes are defined by a short, rigid active-site S-transfer loop and an insertion of amino acids that creates a β-hairpin motif. The β-hairpin reaches over and partially occludes the active site of the adjacent monomer ( Fig. 2A). Type II enzymes, like SufS, are faced with a mechanistic challenge of stabilizing and protecting the active-site persulfide on a short S-transfer loop in a desulfurase step and making it available to a flexible Stransfer loop on SufE in a transpersulfurase step. Based on combined results from site-directed mutagenesis, backbone amide hydrogen-deuterium exchange, x-ray crystallography, and bioinformatics experiments, it has been proposed that a βlatch structural element regulates activity in type II cysteine desulfurases (8)(9)(10). The β-latch contains four key components: the α 6 helix, a glycine-rich loop, a β-hairpin, and a cisproline (Fig. 2B). The α 6 helix contains three conserved residues. R92 and E96 (E. coli numbering) form cross-dimer electrostatic interactions with the α 6 helix of the adjacent monomer, and N99 forms hydrogen bonds to the backbone amide and carbonyl of G251 in the glycine-rich loop (Fig. 2, B and C). The glycine-rich loop (G251/G252/G253) immediately proceeds the β-hairpin (256-266) with both components forming part of the active site of the adjacent monomer. The backbone amide of S254, located directly between these two components, can form interactions with the active-site persulfide on C364 (11). A conserved cis-proline (P271) concludes the motif and is thought to acts as a rigid joint around which the β-hairpin pivots.
The β-latch structural element is dynamic, and motion of the components results in a 3-to 4-Å rotation between the SufS monomers (captured in structures of both WT and variant enzymes) with several consequences (Fig. 2D) (8). The modulation results in disruption of the cross-dimer electrostatic interactions on the α 6 helix and movement of the glycine-rich loop and β-hairpin away from the active site of the adjacent monomer, which may allow SufE closer access to the active site. There is also evidence that this modulation can shift the direction of the C364 persulfide from inward-facing to outward-facing (9). Importantly, backbone amide hydrogendeuterium exchange mass spectrometry data show that formation of the C364 persulfide results in changes to deuterium uptake in both the α 6 helix and glycine-rich loop, suggesting the β-latch is sensitive to the persulfide state of the enzyme (10). These structural changes are consistent with a role for the β-latch in the regulation of SufS. To investigate this hypothesis, site-directed mutagenesis was used to generate SufS variants at the conserved N99 residue. This residue physically links the α 6 helix and G251 of the glycine-rich loop through hydrogen bonds (Fig. 2B). Structural results confirm that the substitutions cause changes in the glycine-rich loop and β-hairpin regions and significant effects on the crossdimer interactions of the α 6 helix. This perturbation results in enzyme variants that no longer require SufE for turnover and can instead react with exogenous reductant. Additional results suggest that disruption of the β-latch also precludes formation of a close approach SufS-SufE complex. These results are consistent with N99, and by extension the β-latch structural element, playing a key regulatory role in SufS function. As the β-hairpin motif is located in a similar position as the binding site for type I cysteine desulfurase regulatory proteins (e.g., frataxin and CyaY), this site may be a conserved regulatory feature for cysteine desulfurases in general.

Results
Substitution of N99 results in perturbations to the β-latch structure and loss of persulfide protection We hypothesized that substitutions of N99 that interfered with its hydrogen-bonding interactions with the glycine-rich loop would perturb the β-latch providing a tool to assay its contributions to enzyme function. We generated variants of SufS with either alanine or aspartate at position N99 and solved the crystal structures of the resulting proteins. Alteration of N99 results in significant, but localized, perturbations in components of the β-latch structural element. The most drastic effect is seen in the N99D SufS structure. In this structure, electron density is missing for residues 248 to 268, which includes the glycine-rich loop (G251-S254) and the βhairpin (I256-T266) (Fig. 3A). Additional perturbations are seen for residues R92 and E96 on the α 6 helix. In WT SufS, these two residues form a cross-dimer electrostatic interaction. The cross-dimer interaction between R92 and E96 is broken in N99D SufS, and the residue positions most closely resemble the open structure of WT SufS (Protein Data Bank [PDB] code 6uy5) with spontaneous shifts in the structure of the β-latch (Fig. 3B) (8). The structure of N99A SufS has continuous electron density for the β-latch; however, the glycine-rich loop and β-hairpin are shifted, exposing the SufS active site to solvent (Fig. 3C). In contrast to N99D, the N99A SufS structure does not show disruption of the cross-dimer interactions between R92 and E96. However, E250 is no longer able to participate in electrostatic interactions in the structure captured by crystallography.
With confirmation of structural perturbations to the β-latch caused by the N99 substitutions, we hypothesized that any protective function of the β-latch would also be disrupted. To Beta-latch disruption in SufS variants affects persulfide protection test this hypothesis, 30 μM WT or N99D SufS were mixed with 500 μM cysteine in the absence of SufE and the presence or absence of 2 mM TCEP, and the formation of alanine was followed using a fluorescence-based plate-reader assay (Fig. 4) (12). In the presence of 500 μM cysteine and 2 mM TCEP, N99D SufS catalyzes significant production of alanine over a 10-min period (filled diamonds). Under identical conditions, WT SufS is also capable of producing alanine (filled circles), but less that N99D SufS. In the absence of TCEP (open symbols), both WT and N99D SufS exhibit near identical alanine production kinetics to the WT enzyme in the presence of TCEP. Results for this experiment demonstrate that N99D SufS no longer inhibits the active-site persulfide from reacting with TCEP, consistent with the model that a key biological role of the β-latch is to shield the active-site persulfide from offpathway reactions.

Steady-state kinetics indicate the β-latch affects SufS-SufE interactions
Steady-state kinetics were followed by an HPLC-based fluorescence alanine detection assay to better quantitate the effect of β-latch perturbations on the catalytic function of SufS. The reactions reported in this section contain 2 mM TCEP, which is required for the regeneration of SufE in the WT system ( Fig. 1) (6). In our hands, the WT enzyme displays kinetic parameters similar to those previously reported (6,10) with an apparent k cat value of 22 ± 2 min -1 and apparent K m values of 0.8 ± 0.1 μM and 63 ± 4 μM for SufE and cysteine, respectively (Table S1). At SufE concentrations higher than 5 μM the activity of WT SufS decreased, and the overall kinetics are best fit by an equation including substrate inhibition (Fig. 5A). Similar SufE substrate inhibition kinetics for SufS have been previously reported, albeit at lower cysteine Figure 2. Components of the β-latch regulatory motif in SufS. A, the SufS dimer (PDB code 6mr2) is displayed with blue and tan monomers. The 5 0pyridoxal phosphate cofactor (PLP) and residue C364 mark the active site. A rectangle indicates the area highlighted in B. B, components of the β-latch are labeled as well as key residues described in the text. C, a sequence logo indicates the conservation of key β-latch residues. Residues of interest are colored magenta and E. coli numbering is used in the sequence logo. The frequency for each residue of interest is denoted below the label. D, superposition of the WT SufS dimer in two different crystal forms. The blue structure is PDB code 6mr2, and the tan structure is PDB code 6uy5. Differences in the β-latch are highlighted.
concentrations (7). The kinetics of SufE dependence for N99A and N99D SufS is complicated with both enzymes displaying a linear dependence on SufE concentrations up to 5 μM with inhibition occurring above 5 μM SufE (Fig. 5A). Owing to the linear dependence of the velocity at low [SufE], the Michaelis-Menten plots for the N99 variants are not well fit by an equation including substrate inhibition. The activity of the N99 variants in the absence of SufE is consistent with the TCEP-dependent activity described in the preceding section. The N99 substitutions do not appear to affect interactions with cysteine as N99A and N99D SufS exhibit Michaelis-Menten kinetics in response to varying cysteine concentrations in the absence of SufE with apparent K M values of 56 ± 10 μM and 62 ± 8 μM, respectively (Fig. 5B). These values are similar to those determined for WT SufS in the presence of SufE. The steady-state activity of N99D SufS was also measured under various concentrations of TCEP to address the mechanism of interaction ( Fig. S1). While increasing TCEP concentrations caused an increase in the rate of alanine formation, the data do not allow for discrimination between a second-order, collisional mechanism (linear dependence) and a mechanism that involves a saturable binding step prior to reduction (hyperbolic dependence).

A perturbed β-latch does not affect formation of initial cysteine-PLP intermediates
To determine if the β-latch plays a role in the reaction of cysteine with PLP, the reaction catalyzed by N99 variants of SufS was followed by UV-visible spectroscopy. Cysteine was mixed with 30 μM SufS in the presence of 2 mM TCEP to monitor formation of cysteine-PLP intermediates along the reaction pathway in a photodiode array spectrophotometer. The as-purified enzymes exhibit an initial peak at 420 nm, representing the internal aldimine species. As previously reported, addition of cysteine to WT SufS results in the rapid (<5 s) reduction of the 420 nm peak by 50% (13). Concomitant with loss of the 420 nm peak, a new peak at 343 nm is formed (Fig. 6A). The enzyme requires 2 to 3 h for the 343 nm peak to disappear and the full 420 nm peak to be restored. Previously, the two prominent SufS absorbance peaks after addition of cysteine were hypothesized to result from formation of cys-ketoenamine (420 nm) and cys-enolimine (343 nm) resonance structures of a stalled external cysaldimine SufS species (13). With the demonstration that WT SufS completes multiple turnovers in the absence of SufE (albeit very slowly), these species are better characterized as the predominant intermediates under steady-state conditions Beta-latch disruption in SufS variants affects persulfide protection (i.e., intermediates preceding slow steps in the reaction) in the absence of SufE. When cysteine is added to N99D SufS, a rapid decrease of 50% is seen at 420 nm with a corresponding increase at 340 nm, similar to the results with the WT enzyme (Fig. 6B). However, the species do not persist, and the initial 420 nm absorbance state is regained after only 10 to 15 min rather than the 2-h time frame for the wildtype enzyme. N99A SufS exhibits results similar to N99D SufS (Fig. 6C). The rapid shift back to the 420-nm species is consistent with the kinetic data in Figure 4 showing N99D rapidly converts cysteine to alanine in the presence of TCEP. When combined with the steady-state kinetics results, it appears that substitution of N99 has not substantially affected the interactions of SufS with cysteine or the reactivity of the PLP cofactor. Instead, the kinetic defects resulting from a perturbed β-latch are localized to interactions with SufE.

β-Latch perturbations affect the interactions of SufS with SufE
The altered steady-state kinetics measured when varying SufE suggested that the β-latch might play a role in SufS binding to SufE. As a first step in answering this question, we constructed a RosettaDock model of SufS bound to SufE since no crystal structure of the complex exists. The best scoring model from this computation had an orientation of SufS and SufE similar to the crystal structure of the cysteine desulfurase CsdA bound to the transpersulfurase CsdE reported in PDB code 4lw4 (Figs. 7A and S2) (14). In both cases the transpersulfurase (SufE or CsdE) interacts extensively with an αhelix in the cysteine desulfurase (termed α 16 ) that sits above the active site and points toward solvent (Fig. 7B). The activesite cysteines in the cysteine desulfurase and transpersulfurase point toward each other but are too far apart in space for sulfur transfer consistent with the identification of additional,  Beta-latch disruption in SufS variants affects persulfide protection possibly transient, conformational changes in SufS and SufE necessary for the chemistry (15,16). The computational model likely represents an on-pathway, initial encounter complex that is populated just before the conformational change that facilitates chemistry. However, the computational model is useful because it suggests specific residues in the α 16 helix of SufS (N353, F350, and D346) whose interactions with SufE drive binding.
The model supports the rational design of SufS mutants with disrupted binding to SufE, a useful tool in developing and validating SufS-SufE binding assays. To this end, a SufS/SufE binding assay based on fluorescence polarization was developed such that a fluorescently labeled SufE should show an increase in fluorescence polarization upon formation of the larger SufS-SufE complex. Using the Rosetta SufS-SufE model as a guide, E107 was selected as a position in SufE that could be covalently labeled with BODIPY-maleimide dye without disrupting SufE function. Thus, E107 and the persulfide accepting C51 were substituted to create a C51A/E107C variant of SufE to serve as a reagent in the fluorescence polarization binding assay. A titration of WT SufS against labeled C51A/E107C SufE results in an increase in fluorescence polarization ( Fig. 8A and Table 1). The data can be fit to Equation 3, which describes a single-affinity binding site to determine a K D value of 5.2 ± 1.2 μM, consistent with binding affinities measured by other methods (16). D346R SufS, a substitution predicted to abrogate SufS-SufE binding based on a Rosetta-derived model of the protein complex, was developed to test the assay (Fig. 8A, yellow triangles). Consistent with the prediction above, titration of the D346R SufS shows limited increases in fluorescence polarization under conditions where the wildtype proteins show substantial interactions and suggests a K D value greater than 40 μM, the highest concentration of SufS used in the assay. Figure 8A also shows the fluorescence polarization results for a titration of N99A and N99D SufS against labeled SufE. The N99 variant enzymes are able to interact with SufE with low micromolar K D values similar to the value determined for

Beta-latch disruption in SufS variants affects persulfide protection
the WT SufS-SufE interaction (Table 1). This result suggests that the N99 substitutions have not disrupted the initialencounter complex between the two proteins. However, previous studies have shown that a C51-alkylated SufE binds to SufS with K D value of 0.5 μM (10). This 10-fold increase in affinity is hypothesized to occur because the alkylated SufE represents a better mimic of the functional interaction that results in persulfide transfer. In an attempt to mimic this functional interaction, 500 μM cysteine was added to the titration conditions. WT SufS will process cysteine slowly and will build up the steady-state intermediates as shown in Figure  6A, which could be trapped by the inactive BODPIY-C51A/ E107C SufE. In the absence of TCEP, the N99 variants should behave in a similar manner. In the presence of 500 μM cysteine the WT SufS-SufE interaction is dramatically improved to a value of 0.6 ± 0.2 μM, consistent with the value determined with C51-alkylated SufE (Fig. 8B and Table 1) (10). This suggests that SufE is able to probe the higher-affinity "close approach" functional interaction between SufS and SufE when cysteine is present. Under identical conditions, the two N99 SufS variants show only modest improvement in their K D values for SufE, and the values remain similar to those determined in the absence of cysteine. These results suggest that, when the β-latch of SufS is perturbed, SufE can still bind in an initial-encounter complex but cannot form the close approach conformation, the structure of the complex required to promote efficient persulfide transfer. This finding is consistent with the altered steady-state kinetics parameters described above.

Discussion
Cysteine desulfurases are ubiquitous PLP-dependent enzymes responsible for the mobilization of sulfur from cysteine for use in the biosynthesis of multiple biological compounds (4). The type II cysteine desulfurases are defined by a short, rigid S-transfer loop and a β-latch structural element that forms part of the active site on the adjacent monomer. Owing to the short S-transfer loop, type II desulfurases require accessory transpersulfurase proteins to transfer the covalent persulfide to downstream protein partners. Although the β-latch appears to be ubiquitous in type II enzymes, there is not a clear understanding of the functional role of this evolutionarily conserved feature. To answer this question, site-directed mutagenesis was used to generate variants of SufS from E. coli targeted to a conserved asparagine residue (N99), which forms a key interaction in the β-latch.
The β-latch provides a protective mechanism for the E. coli SufS persulfide As described above, a key feature of E. coli SufS (and the E. coli Suf pathway in general) is its ability to protect reactive persulfide intermediates under harsh cellular conditions (6,7). The most striking result from this work is the loss of that protection in the N99 SufS variants as evidenced by the robust alanine production using TCEP as a persulfide reductant. These kinetic results are readily explained when the structures of the N99 variants are examined. Disruption of the N99 interactions with the β-latch results in changes to all facets of the motif. One can hypothesize that the changes most relevant to the protective function of the β-latch are the disruption of the interaction between the N99 position and the backbone of the glycine-rich loop, which makes up a wall of the active-site opening. In the N99D SufS structure, the loss of density for the glycine-rich loop and the β-hairpin suggests increased solvent access to the active-site persulfide (Fig. 3A). In the N99A SufS structure density for the glycine-rich loop and the β-hairpin is seen, suggesting they are less dynamic and are still capable of physically blocking the active site from solvent. However, the glycine-rich loop is in an altered conformation compared with the WT SufS structure (Fig. 3B). This altered conformation shifts the position of the S254 amide away from the active site by over 8 Å. The backbone amide of S254 has been shown to interact with the terminal sulfur of the C364 persulfide in the outward-facing position (11). Loss of this interaction could affect the stability of the persulfide intermediate to nucleophilic attack.
The β-latch is also responsible for productive SufS-SufE interactions in E. coli While the protective nature of the E. coli SufS-SufE system is critical for its biological function, the β-latch is conserved in all type II cysteine desulfurases (8). It is not clear if the protective mechanism is exhibited in systems where it is not specifically required (i.e., systems that do not operate under conditions of oxidative stress). This raises the possibility of a more general function of the β-latch in type II cysteine desulfurases. As described above, type II cysteine desulfurases require an accessory transpersulfurase protein to transfer the covalent persulfide intermediate. The best models for the SufS-SufE interaction come from costructures of the homologues CsdA-CsdE in E. coli (17,18) and the SufS-SufU costructure from Bacillus subtilis (19). Both structures show key interactions between the desulfurase and the transpersulfurase occurring in the α 16 helix region of the desulfurase (residues 344-354 in E. coli SufS). In the current costructures the distance between the persulfide donors and acceptors is too far to support efficient persulfide transfer. This suggests that additional conformational changes are required to create a "functional" interaction.
Based on homologous structures and the Rosetta model, the β-hairpin motif lies 10 Å across from a face of the SufE protein, including the region directly upstream of the Beta-latch disruption in SufS variants affects persulfide protection persulfide accepting cysteine. It is possible that rotation around the SufS α 16 helix-SufE interaction axis could close the gap between the two proteins to form a fully functional interaction capable of protected and efficient persulfide transfer. The data presented here and in previous reports support this two-step interaction model (Fig. 9). The D346R SufS variant exhibits severely diminished interactions with SufE under standard assay conditions, consistent with the α 16 helix region mediating initial SufS-SufE interactions. In the absence of cysteine, WT and the N99 SufS variants bind to SufE with low micromolar K D values representative of the initial interaction and likely represented by homologous desulfurase-transpersulfurase costructures. The addition of cysteine to WT SufS allows for formation of the functional SufS-SufE complex and a 10-fold decrease in the K D value. However, with the structural disruptions in the β-hairpin region caused by the N99 substitutions, the SufS variants are not able to form this functional complex and exhibit minor improvements in their K D values.
Comparison of the type II β-latch-mediated SufS-SufE interaction with type I systems identifies a regulatory hotspot As described above, cysteine desulfurases can be described as type I and type II systems (Fig. 10A). The two types contain similar active-site architecture (Fig. 10, B and C) but are differentiated by a relative sequence insertion resulting in a flexible S-transfer loop for the persulfide-containing cysteine in type I enzymes compared with a buried, inflexible S-transfer loop in type II enzymes. The flexible Stransfer loop in the type I enzymes allows for direct transfer of the persulfide to acceptors, while the type II systems require an accessory transpersulfurase with a flexible Stransfer loop containing the persulfide acceptor. Both systems share the common challenge of creating a complex capable of transferring the persulfide intermediate between the donor and acceptor cysteines with one cysteine residing on a flexible loop. In the type II systems, we propose that the β-latch is responsible for mediating interactions by guiding the flexible persulfide acceptor loop on SufE toward the donor cysteine on SufS to promote a functional interaction (Fig. 10C). In type I cysteine desulfurases, the challenge is guiding the flexible S-transfer donor loop to specific S-transfer acceptors. In NFS1, the type I mitochondrial cysteine desulfurase, this challenge is overcome by the accessory protein frataxin. Frataxin is required for optimal function of NFS1 and has been shown to kinetically promote persulfide transfer (20) and structurally guide the flexible S-transfer loop for functional interactions with the acceptor protein ISCU (Fig. 10B) (21). The type I bacterial cysteine desulfurase, IscS, is inhibited by interactions with CyaY, a frataxin homologue, and requires IscX to disrupt the IscS-CyaY interaction (22). A previously reported mutagenesis screen of IscS identified several residues essential for binding of CyaY (23). A comparison of IscS and SufS shows these residues occupy the same space as the βlatch (Fig. 10D). Surprisingly, the mode of activation or inhibition by the accessory protein is encoded by the desulfurase structure as frataxin and CyaY both activate NFS1 and both inhibit IscS (24).
These commonalities suggest that the type I and type II cysteine desulfurases utilize a similar structural feature adjacent to the active site to coordinate persulfide transfer between donor and acceptor S-transfer loops. Type I enzymes, which are capable of interacting with multiple acceptor proteins, can be regulated by forming a ternary complex with various acceptor and accessory proteins to promote specific conformations of the S-transfer donor loop. Type II cysteine desulfurases have dedicated persulfide acceptors and utilize the native β-hairpin, rather than exogenous accessory proteins, to promote S-transfer donoracceptor loop interactions.

Conclusions
The presence of a conserved β-hairpin structure is a defining characteristic of type II cysteine desulfursases. Recent kinetic, structural, and bioinformatic results proposed the βhairpin is part of a larger, connected set of features dubbed the β-latch structural element. Here we have shown that the conserved N99 residue acts as a linchpin in the β-latch. Substitution of this residue in SufS from E. coli resulted in a Figure 9. Model for SufS-SufE interactions. Initial SufS (blue and tan) and SufE (green) interactions involve formation of an "initial interaction" complex centered on binding of SufE to the α 16 helix on SufS. Upon formation of the SufS persulfide and rearrangement of the beta-latch, a "close approach" conformation is formed with new interactions between SufE and the beta-hairpin motif.
kinetically impaired enzyme. More in-depth characterization showed the N99 enzyme variants lost the native protection of the covalent persulfide intermediate to external reductants and could no longer form a high-affinity, functional SufS-SufE complex. Structures of the two SufS variants showed these functional defects could be explained by perturbations to the β-latch. Comparison of the location of the β-latch with regulatory accessory proteins of type I cysteine desulfurases suggests a conserved architectural feature in cysteine desulfurases plays an important regulatory role in selecting specific conformations of the S-transfer donor and acceptor loops in all cysteine desulfurases.

Generation of the SufS and SufE variants
Mutagenesis of the SufS (Uniprot ID:P77444) or SufE (Uniprot ID:P76194) gene in vector plasmid pET-21(a) was accomplished using Quikchange Lightening Mutagenesis Kit (Agilent) for SufS and inverse PCR for SufE. Forward and reverse primers with desired mutations were introduced to obtain the N99A, N99D, and D346R SufS variants as well as the C51A/E107C SufE variant (Table S2). The resulting plasmids were transformed into XL-10 Gold E. coli cells, and the purified plasmid was sent for DNA sequencing (Eurofin Genomics) to ensure the desired mutation.

Protein expression and purification
SufS and SufE enzymes were expressed and purified as discussed (10). For recombinant protein expression, sequenced pET21 plasmids carrying the desired gene sequence were transformed into electrocompetent BL21(DE3)Δsuf E. coli cells. Cells were grown overnight in Luria Broth with 100 μg/ ml ampicillin and 50 μg/ml kanamycin at 37 C. To initiate 1-L cultures, the overnight growth was diluted to 100-fold in LB and allowed to incubate at 37 C until the cells reached an A 600 of 0.4 to 0.6. Protein expression was then induced using 500 μM isopropyl-β-D-thiogalactoside. Cells were grown for 3 h post induction, harvested via centrifugation, and cell pellets were stored at −80 C. To begin purification, cell pellets were thawed and resuspended in 25 mM Tris pH 7.5, 100 mM NaCl, 5 mM dithiothreitol (DTT), 1 mM phenylmethylsufonyl fluoride, 1 mM DNase, and 1 mM MgCl 2 . The resuspended cells were lysed via sonication, and cellular debris was pelleted via centrifugation. The clarified lysate containing SufS was subjected to purification using three different columns: Q-XL, Phenyl HP, and Superdex 200 (Cytivia). The Q-XL column used a linear gradient from 25 mM Tris-HCl, pH 7.5, 10 mM β-mercaptoethanol (BME) to 25 mM Tris-HCl, pH 7.5, 10 mM BME, 1 M NaCl. The Phenyl HP column used a linear gradient from 25 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 M (NH 4 ) 2 SO 4 , 10 mM BME to 25 mM Tris-HCl, pH 7.5, 10mM BME. The Superdex 200 column was run with 25 mM 3-(Nmorpholino)propanesulfonic acid (MOPS), pH 7.5, 150 mM NaCl, 10 mM BME. Lysate containing SufE was purified using Q-XL and Superdex 200 columns. Fractions from all the columns containing the protein of interest were identified using SDS-PAGE. Selected fractions were pooled, concentrated, and stored in gel filtration buffer supplemented with 10% glycerol at −80 C for further use. SDS-PAGE was used to assess the purity of the final sample (Fig. S3). SufS concentration was determined based on quantification of PLP at 388 nm in 0.1 M NaOH (ε 388 = 6600 M −1 cm −1 ) (25). SufE concentration was determined with a calculated extinction coefficient of ε 280 = 20,970 M −1 cm −1 .

Alanine-detection assay
Cysteine desulfurase activity was measured by monitoring L-alanine production in a naphthalene 2,3-dicarboxaldehyde (NDA) labeled reaction (12). The assay mixture contained the following components: 0.25 to 1 μM SufS (based on PLP absorbance), 100 mM MOPS, pH 8.0, 150 mM NaCl, 2 mM TCEP, and varying amounts of SufE and cysteine. For the HLPC-based assay, a 700-μl reaction was initiated by addition of SufS and, at various time points, 100-μl aliquots were quenched using 20 μl of 10% trichloroacetic acid. This was followed by adding 1 ml of NDA labeling mix (100 mM borate, pH 9.0, 2 mM KCN, and 0.2 mM NDA) and incubating the reaction mixtures in the dark for 60 min. After the incubation, 200 μl of the reaction mixtures was transferred to vials in the HPLC autosampler. The autosampler was set to inject 10 μl of each sample onto a Zorbax C18 column (Agilent) at a flow rate of 0.4 ml/min with an isocratic gradient of 50% 10 mM ammonium acetate with 50% methanol (pH 6.0) for 11 min. The fluorescent adduct was measured with a Shimadzu HPLC coupled with a florescence detector set to 390 nm excitation and 440 nm emission. Alanine standards were also prepared under the same conditions to construct a standard curve. The alanine-NDA peak was integrated, and the peak area was converted into nanomoles of alanine using the standard curve. For the plate-reader assay, reactions were run using similar conditions to the HPLC-based assays. Reaction aliquots of 50 μl were quenched with 5 μl of 10% trichloroacetic acid and mixed with 200 μl of the NDA labeling mix. After a 20-min incubation with the NDA labeling mix in the dark, the fluorescence (390 nm excitation, 440 nm emission) for each time point was measured using a Biotek Synergy2 multiwell plate reader.

Fluorescence anisotropy SufE binding assay
The pET21 plasmid carrying the C51A/E107C SufE gene was transformed into BL21(DE3) E. coli cells and expressed and purified as described for SufE above with a change in purification buffer pH from 7.5 to 8.0. Following purification and concentration, the protein was dialyzed into 50 mM MOPS (pH 8) and 150 mM NaCl to remove DTT. Fluorescently labeled samples of C51A/E107C SufE were created by incubating 50 μM protein with 500 μM BODIPY FL maleimide (Invitrogen) for 4 h at 25 C. After the incubation period, unreacted dye was removed using a PD-10 column. Labeling efficiency was determined by comparing the ratio of label concentration (ε 505 = 80,000 M −1 cm −1 ) with protein concentration (ε 280 = 20,970 M −1 cm −1 ) and was determined to be between 60 and 80% using this method. SufS-SufE binding titrations were carried out in black 96-well plates in 50 mM MOPS (pH 8), 150 mM NaCl, and 0.1 mg/ml bovine serum albumin. SufS (0.2-40 μM) was mixed with 100 nM labeled C51A/E107C SufE and allowed to equilibrate at room temperature for 30 min. After equilibration, the fluorescence polarization (480 nm excitation, 520 nm emission) was measured using a BioTek Synergy2 multiwell plate reader.

Crystallography methods
Crystallization and structure solution of SufS site-directed mutants was performed essentially as reported (9). Briefly, 8 Beta-latch disruption in SufS variants affects persulfide protection to 9 mg/ml protein was mixed 2:1 (vol/vol) with crystallization solution, 4 to 4.5 M NaCl, and 100 mM MES pH 6.5, and subjected to sitting drop vapor diffusion at 20 C. Crystals appeared in 1 to 7 days. Crystals were cryoprotected by soaking for several minutes in a solution consisting of mother liquor with 50% glycerol. Crystals were plunge-frozen in liquid nitrogen and x-ray data collection proceeded at 100 K with 1.54-Å x-rays generated by a Rigaku Synergy DW equipped with a HyPix6000-HE detector. Data reduction was performed with XDS (SufS N99D) or with Rigaku CrysAlis Pro software (SufS N99A). The structure given by PDB code 6mr2 (with heteroatoms removed) provided phases for structure solution. Model building and refinement were performed iteratively using Coot and Phenix. X-ray data statistics and model quality data are reported in Table S3. Final coordinates were submitted to the PDB as PDB codes 7ruj (N99A SufS) and 7rw3 (N99D SufS).

Rosetta model construction
A computational model of SufE bound to SufS was constructed using RosettaDock in the local docking mode in Rosetta version 3.11. The SufS coordinates given by PDB code 6mr2 and the SufE coordinates given by PDB code 1mzg were prepared in the following manner. The SufS homodimer sits on a crystallographic symmetry axis in the crystal form reported in 6mr2. Therefore, the symmetry mate of SufS was generated in Pymol from the 6mr2 coordinates to produce the SufS homodimer with chains labeled A and B. Both the SufS coordinates and the SufE coordinates were cleaned (stripped of heteroatoms), renumbered, and repacked in Rosetta. SufE was positioned approximately 10 Å from SufS with its active site, marked by residue C51, facing toward the SufS active site in monomer A, marked by the SufS residue C364. The known, enzymatic mechanism of persulfide transfer from SufS to SufE dictates the two active-site cysteines must be physically adjacent to each other, which informed the starting position for SufE in the calculation. In addition, homologous structures such as SufS-SufU (PDB code 5xt6) and CsdA-CsdE (PDB code 4lw4) confirm the active sites of SufS and SufE must be physically adjacent to each other for persulfide transfer to occur. Rosetta was launched with command: rosetta/3.11/main/source/bin/docking_protocol.default. linuxgccrelease @flag_local_docking using the flag_local_docking file: -in:file:s 1mzg_6mr2_repack.pdb -unboundrot 1mzg_6mr2_repack.pdb -nstruct 1000 -partners AB_C -dock_pert 3 8 -ex1 -ex2aro -out:path:all Output_files -out:suffix _local_dock A plot of the interface score (I_sc) versus the interface rms difference (I_rms) among the decoys was generated, which identified Model 0014 as the lowest energy structure, existing at the nadir of a cluster of related structures in the plot (Fig. S3). Model 0014 is publicly available at https://dunklelab. ua.edu/structures.

Data analysis
Nonlinear regression fits of the data to various equations were performed in Kaleidagraph 5.0 (Synergy Software). Initial velocities at various substrate concentrations were determined using a minimum of three time points. Initial velocity data were then fit to the Michaelis-Menten equation (Equation 1) or an equation describing substrate inhibition (Equation 2) to determine the kinetic parameters for each enzyme. Here, v is the initial velocity, E t is the total enzyme concentration, k cat is the maximal turnover number, S is the substrate concentration, K M is the Michaelis constant, and K i is the inhibition constant.
Fluorescence polarization data used to measure proteinprotein interactions were fit to Equation 3 where A o is the polarization in the absence of the ligand, ΔA is the total change in polarization, and K D is the dissociation constant.

Data availability
All kinetic and biophysical data are contained within the article and supporting information. Structural data have been deposited in the PDB under accession codes 7ruj for N99A SufS and 7rw3 for N99D SufS.
Supporting information-This article contains supporting information.
responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.