Characterization of a mechanistic connection between persulfide transfer and ATP hydrolysis in the SufBC₂D scaffold of the Suf Fe-S cluster assembly pathway

Abstract

The Suf pathway is the most common pathway for bacterial iron-sulfur cluster assembly and uses the SufBC₂D complex as a scaffold for cluster formation. In most Gram-negative bacteria, the SufB subunit of SufBC₂D accepts a persulfide from the transpersulfurase, SufE, for incorporation into nascent clusters. There is no reported structure for the SufBC₂D-E complex and mechanistic details concerning the coordination of persulfide delivery with other SufBC₂D activities are unclear. Using the Suf pathway from Escherichia coli as a model system, we report that SufE acts as a noncompetitive inhibitor of SufBC₂D ATPase activity with a K_i value of 1.8 ± 0.2 μM. This value corresponds with a K_D value of 1.6 ± 0.2 μM for SufE binding to the SufBC₂D complex determined by fluorescence polarization. The rate of persulfide transfer from SufE to SufBC₂D is impaired in the presence of ATP suggesting the two reactions are mutually exclusive. An AlphaFold3 model of the SufBC₂D-E complex predicts electrostatic interactions between acidic residues on SufC and basic residues on the N-terminal helix of SufE. SufE variants at the K9 and R16 positions interfere with the ability of SufE to transfer persulfide to SufBC₂D and to inhibit SufBC₂D ATPase activity. In vivo complementation growth assays show that these SufE variants exhibit a slow-growth phenotype under iron starvation conditions, confirming the connection between SufE and SufC as important for optimal function in the Suf pathway. The mutual exclusivity of persulfide delivery from SufE and SufBC₂D ATPase activity suggests an ordered mechanism for cluster assembly.

Introduction

Iron-sulfur (Fe-S) clusters are essential cofactors in living systems, but the assembly blocks of these cofactors, iron and sulfur atoms, can be damaging in a cellular environment. To overcome this challenge, Nature has evolved highly choreographed assembly pathways using multiple protein participants to build Fe-S cluster in a regulated manner [1]. In bacteria and eukaryotes, these pathways utilize three common steps. Sulfur is mobilized from L-cysteine by cysteine desulfurase [2, 3] and transferred to a scaffold protein. The scaffold protein coordinates the acquisition of sulfur, iron, and reducing agents required to assemble the cluster. Following assembly, the nascent cluster will be trafficked by a carrier protein for insertion into apo-Fe-S proteins.

The scaffold proteins are central in each pathway as they form a hub that must regulate interactions with proteins that deliver the assembly materials and with proteins that traffic the assembled cluster. In the Isc pathway, found in the eukaryotic mitochondrion and some bacteria, the regulatory steps of cluster assembly on the IscU scaffold are well documented [4, 5]. These include stimulation of cysteine desulfurase perfsulfide transfer by accessory proteins, electron delivery from ferredoxin, identification of the site of cluster assembly, and a role for ATP hydrolysis by chaperone proteins to induce cluster transfer from the IscU scaffold.

In contrast to the understanding of cluster assembly on the IscU scaffold protein in the Isc system, many mechanistic details of the scaffold protein in the Suf pathway remain unclear. The Suf pathway is the most common pathway for cluster assembly in bacteria and in the chloroplast of photosynthetic eukaryotes but is missing from mammals. Better characterization of the Suf pathway could lead to its development as a novel antibacterial target. Work from several groups over the past decade has provided a description of the general architectures and overall functional roles for proteins in the Suf pathway [6, 7]. The SufS cysteine desulfurase mobilizes persulfide in a well-studied, PLP-dependent reaction resulting in persulfide formation on SufS C364. Full SufS activity also requires the SufE tranpersulfurase protein as a dedicated partner to remove the SufS C364 persulfide using C51 on SufE. The SufE protein then transfers the persulfide species to the SufBC₂D scaffold complex in preparation for cluster assembly. Following cluster assembly on SufBC₂D, the cluster is transferred to the SufA A-type carrier protein for downstream delivery to target proteins.

SufBC₂D is a heterotetramer composed of one subunit each of SufB and SufD, and two subunits of SufC (Figure 1) [8–10]. SufB and SufD form a dimeric structure with an unusual repeating β-strand helix and the two proteins share sequence and structural homology. The SufBD structure is thought to be the functional pairing within the complex, but both SufB₂ and SufD₂ dimers have been characterized when the proteins are overexpressed alone. The subunits of the SufBC₂D scaffold exhibit a number of functions related to cluster assembly. SufB is known to accept a L-cysteine derived persulfide from the SufS/SufE cysteine desulfurase/transpersulfurase system. A persulfide on C51 of SufE is transferred to C254 on SufB. [8, 11, 12] SufD is hypothesized to assist with iron acquisition, though the identity of the Fe donor in vivo is unknown [13]. In the functional tetramer, SufB and SufD are each attached to separate SufC monomers. SufC is an ABC-type ATPase, with hydrolysis of ATP hypothesized to play a role in iron acquisition [13]. Consistent with other ABC-type proteins, ATPase activity would require dimerization of the two SufC subunits, and transient dimerization has been observed in vitro using crosslinking approaches [10]. SufC dimerization may influence the conformation of the SufB/SufD subunits. Fe-S cluster formation is predicted to occur at the SufB/SufD interface with potential ligands from both proteins. Finally, it has been shown that when SufBC₂D is isolated under anaerobic conditions, the complex co-purifies with a molecule of FADH₂ [9]. The binding site and physiological role of the FADH₂ are currently unknown, but oxidation to FAD⁺ results in a drastic decrease in the binding affinity of the molecule for the complex causing the flavin to dissociate.

Figure 1. Structure and functions of the SufBC₂D scaffold complex from E. coli.

The ribbon structure for each subunit (SufB, blue; SufD, pink, SufC, yellow) is labeled with proposed functional roles as described in the main text. The structure is based on PDB entry 5awf.

Despite our understanding of the stand-alone SufBC₂D activities, an understanding of molecular mechanisms regulating the choreography of events in Fe-S cluster assembly on the scaffold remain unclear. Here, we have investigated the interaction between persulfide delivery from SufE and the ATPase activity of the SufBC₂D scaffold. Steady-state kinetic assays show SufE acts as a noncompetitive inhibitor of SufC ATPase activity within SufBC₂D with a K_i value of ~2 μM. Conversely, addition of ATP diminishes the ability of SufE to deliver persulfide to SufBC₂D suggesting the two activities are mutually exclusive. An AlphaFold3 model of the SufBC₂D-E complex predicts electrostatic interactions between SufC and the N-terminal helix of SufE. Substitution of SufE residues K9 or R16 with glutamate disrupts SufE inhibition of ATP hydrolysis and persulfide delivery to SufB. In vivo complementation assays with the same SufE variants result in slow-growth phenotypes under iron starvation conditions suggesting the SufE-SufC interaction is biologically relevant. This characterized functional connectivity provides a potential mechanism for the regulation of persulfide delivery in the Suf pathway.

Methods

Protein expression and purification.

A pDuet plasmid harboring the sufB(his₆)CD genes from E. coli was transformed into electrocompetent BL21(DE3)Δsuf E. coli cells. Cells were plated on LB-agar plates containing 100 mg/L ampicillin and 50 mg/L kanamycin and incubated at 37 °C overnight. A single colony was used to inoculate an overnight culture of LB media at 37 °C with 100 mg/L ampicillin and 50 mg/L kanamycin. The next day, a 10 mL aliquot of the overnight culture was used to inoculate 1 L of sterile LB media containing 100 mg/L ampicillin and 50 mg/L kanamycin). Cells were incubated at 37 °C with agitation until OD600 reached 0.5–0.6. At that time, 0.5 mM isopropyl-β-D-thiogalactoside (IPTG) was added to the to induce the protein expression, and cells were grown for 4 more hours. Following the growth, cells were harvested via centrifugation at 8000 ×g for 10 minutes at 4 °C and stored in −80 °C until purification. To purify SufB(his₆)C₂D, cell pellets were resuspended into Buffer A (50 mM Tris, pH 7.5, 500 mM NaCl, 20 mM imidazole, 10 mM β-mercaptoethanol (BME) with 1 mM phenylmethylsufonyl fluoride (PMSF), 1 mM MgCl₂, and 10 μg/mL DNase I). The resuspended cells were lysed via sonication, and insoluble debris was removed by centrifugation. The clarified lysate was loaded onto a 5 mL HisTrap Ni column (Cytivia), and His₆-SufBCD was eluted with a linear gradient from Buffer A to Buffer B (50 mM Tris pH 7.5, 500 mM NaCl, 500 mM imidazole, 10 mM BME). The fractions containing His₆-SufBCD were pooled, concentrated, and further purified with a Superdex 200 column. His₆-SufBCD was eluted with 25 mM 3-(N-morpholino)-propanesulfonic acid (MOPS) pH 7.5, 150 mM NaCl, 10 mM BME. Post SDS-PAGE analysis, the selected fractions were pooled, concentrated, and supplemented with 10% glycerol, flash frozen in liquid nitrogen, and stored in −80°C until use. His₆-SufBCD concentration was determined by UV-Vis, with an ε₂₈₀ = 130,000 M⁻¹ cm⁻¹. Results in this work were obtained from the His₆-SufBCD protein complex, hence referred to as SufBC₂D for simplicity.

Wildtype SufE was expressed and purified as described previously [14]. SufE variants (K9A/E, R12A/E, R16A/E, and an N-terminal truncation of the first 17 residues) were obtained via gene synthesis from Twist Biosciences. A pET-21c plasmid containing the desired SufE gene sequence was transformed into BL21(DE3)Δsuf E.coli cells, and the expression was performed as described above.

Carbamidomethylation of SufE (SufE-alk).

Wildtype SufE (400 μM) in 25 mM MOPS pH 8.0, 150 mM NaCl, was reduced with 5 mM dithiothreitol (DTT) on ice for 30 minutes. Post desalting with a PD-10 column, the reduced SufE was alkylated with 5 mM of iodoacetamide in the dark for 1 h, 25 °C. A subsequent PD-10 desalting step followed, and the alkylated protein was concentrated and analyzed through peptide mass fingerprinting with alkylated-cysteine modifications enabled. Peptides containing C51 were found in their alkylated form.

ATP hydrolysis assay.

The ATP hydrolysis rates were measured with a coupled enzyme assay as described previously [15]. Briefly, the solution contained 50 mM Tris, pH 7.5, 10 mM KCl, 2 mM MgCl₂, 1 mM phosphoenolpyruvate (PEP), 0.16 mM NADH, 12 U lactate dehydrogenase, 6 U pyruvate dehydrogenase, and varying concentration of ATP. The reactions were initiated with addition of His₆-SufBCD at 37 °C. NADH concentrations were monitored at 340 nm for 15 minutes, and the rate of NADH oxidation was determined using its extinction coefficient of 6,220 M⁻¹ cm⁻¹. All inhibition assays were performed in the same standard mixture as the ATP hydrolysis measurements described above, with varying concentrations of SufE or SufE-alk (0–4 μM).

NDA-alanine detection assay.

Cysteine desulfurase activity of SufS was measured by quantifying the amount of L-alanine produced using naphthalene-2,3-dicarboxaldehyde (NDA) as a labelling agent, which forms a fluorescent adduct with L-alanine [16]. The assay mixture contained 50 mM MOPS pH 8.0, 150 mM NaCl, 2 mM TCEP, 0.2 μM SufS (based on PLP concentration), 500 μM L-cysteine, and varying amount of SufE (0.25 – 3 μM) in 500 μL. The react was initiated by the addition of SufS and 50 μL aliquots were quenched at various time points using 5 μL of 10% trichloroacetic acid (TCA). Next, 200 μL of NDA-labelling mix containing 10 mM borate pH 9.0, 2 mM KCN, and 0.2 mM NDA was added to the quenched mix and the samples were incubated in dark for 20 min. Following incubation, 100 μL of each labeled reaction was transferred to a dark 96-well plate. Lastly, fluorescence (excitation 390 nm, emission 440 nm) of the alanine-NDA adduct was measured using BioTek Synergy2 multiwell plate reader. Alanine standards were also made under same conditions to construct the standard curve. Linear progress curves for alanine production were determined from at least four time points.

SufSE activation assay.

To determine if SufE variants can activate SufS to the same extent as WT SufE when SufBC₂D is included in the reaction, a methylene blue assay was performed to quantify the amount of sulfide formed in the whole system [16]. The assay contained 50 mM MOPS pH 7.5, 150 mM NaCl, 0.25 μM SufS, 0.25 μM SufBC₂D, 2 mM cysteine, varying amounts of SufE (0 – 6 μΜ) and 2 mM DTT. DTT was used to reduce the persulfide formed on SufBC₂D to allow reaction turnover. 1300 μL reaction was initiated by adding cysteine, and 200 μL aliquots were quenched at various time points with 20 μL of 25 mM N,N-dimethyl-p-phenylenediamine (DMPD) in 7.2 M HCl followed by addition of 20 μL 30 mM FeCl₃ in 1.2 M HCl. The labeled mix was incubated for 30 minutes at room temperature and plated on clear 96-well plate. Absorbance at 670 nm was measured on BioTek synergy2 multi-well plate reader. Sodium sulfide standard was made under similar conditions and used to calculate the amount of sulfide generated in the enzyme reaction. Linear progress curves for sulfide production were determined from at least four time points.

Fluorescence anisotropy SufE binding assay.

The fluorescently labeled SufE C51A/E107C was prepared as described previously [14]. A SufBC₂D and SufE binding titration was conducted in a buffer containing 50 mM MOPS, 150 mM NaCl, and 0.1 mg/mL bovine serum albumin. Varying concentrations of SufBC₂D (0.0002–45.5 μM) was incubated with 0.1 μM labeled SufE C51A/E107C at 37 °C in the dark for 30 minutes. After incubation, the fluorescence polarization (480 nm excitation, 520 nm emission) was measured using a BioTek Synergy2 multiwell plate reader.

AlphaFold3 model generation.

AI-generated models of the SufBC₂D, SufBC₂D-E, and the (Mg-ATP)₂-SufBC₂D-E complexes were generated using the AlphaFold3 webserver (alphafoldserver.com) and sequences for the SufBCD-E from E. coli (Uniprot IDs: SufB, P77522; SufC, P77499; SufD, P77689; and SufE, P76194).[17] Five 3D models were generated for each complex, and the highest ranked structure was selected for analysis.

Circular Dichroism Spectroscopy.

CD spectra of all SufE variants were obtained in the far-UV region (190–280 nm) at room temperature using a spectropolarimeter (J-1500 CD Spectrometer). 0.2 mg/mL final protein concentration was used to obtain the CD spectra. Protein sample was made to final concentration by diluting with 50 mM phosphate buffer, pH 8. The blank solution was also prepared by mixing same volume of protein storage buffer (25 mM MOPS pH 7.5, 150 mM NaCl and phosphate buffer). Each spectrum is an average of five scans and baseline corrected using a 50 mM phosphate buffer, pH 8.

Genetic complementation tests.

Single colonies of E. coli wild-type strain MG1655 and the ΔsufE::cm^R strain containing various pET21a plasmid derivatives were used to inoculate Lennox broth (LB) with 100 μg/mL ampicillin (and 30 μg/mL chloramphenicol for ΔsufE::cm^R). After 16 hours growth with shaking at 37°C, the overnight cultures were normalized to an OD₆₀₀ = 2.0 using fresh LB. Serial dilutions of each stock culture were prepared and 5 μl of each dilution was spotted on LB plates containing 0.4% glucose, 100 μg/mL ampicillin, and with or without 250 μM 2,2’-bipyridyl (Sigma). Plates were monitored for 24 – 48 hours at 30°C, and growth was recorded by photography.

Data analysis.

Initial velocity kinetic data for the NDA-alanine assay and the methylene blue persulfide assay were fit to the Michaelis-Menten equation to determine steady-state kinetic parameters. Initial velocity inhibition data was globally fit to equation 1 in GraFit (Erithacus Software). Equation 1 describes non-competitive inhibition, where [I] is the concentration of the inhibitor added to the reaction, and $K_i$ is the inhibition constant. Fluorescence polarization binding data was fit with equation 2 to determine a $K_D$ value for the interaction. 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. Experimental data to measure IC50 values for partial inhibition were fit to equation 3 where relative activity is expressed as the ratio of $v_i$, the velocity in the presence of a given SufE concentration, and $v_0$, the velocity in the absence of SufE, β is the fractional activity remaining at saturating SufE, and ${\mathrm{IC}}_{50}$ is the SufE concentration that results in 50% fractional inhibition between $v_0$ and β [18].

$$v=\frac{Vmax\left[S\right]}{Km\left(1+\frac{\left[I\right]}{K_i}\right)+\left[S\right]\left(1+\frac{\left[I\right]}{K_i}\right)}$$ (1)

$$Flpolarization=A_0+DA\times \left(\frac{\left(\left[SufE\right]+\left[SufS\right]+K_D\right)-\sqrt{{\left(\left[SufE\right]+\left[SufS\right]+K_D\right)}^2-(4\times [SufE]\times \left[SufS\right])}}{2\times [SufE]}\right)$$ (2)

$$\frac{v_i}{v_o}=\frac{{(v}_0\times {IC}_{50})+{(v}_0\times \beta \times \left[I\right])}{{IC}_{50}+\left[I\right]}$$ (3)

Results

SufBC₂D ATPase activity is inhibited by SufE

In an attempt to unravel the mechanistic steps for SufBC₂D cluster assembly, we began investigating how the various ligands/substrates for SufBC₂D would affect their non-cognate reactions, starting with the effect of SufE on the SufBC₂D ATPase activity. Initial velocity kinetics for ADP formation from ATP were determined under a range of SufE concentrations (Figure 2A). Increasing concentrations of SufE inhibited the reaction, and the data could be globally fit to a noncompetitive inhibition mechanism (equation 1) with a $K_{\mathrm{i}}$ value of 1.8 ± 0.2 μM. Previously, it was shown that SufE binding to its cysteine desulfurase partner SufS can be increased ~10-fold by chemically alkylating the C51 residue (SufE-C51_alk) [19]. This modification is thought to mobilize the C51 loop of SufE out of its binding pocket on the SufE surface, causing SufE to adopt a high-affinity binding conformation that mimics its sulfur-accepting conformation. We tested if the C51_alk-SufE was a more potent inhibitor of SufBC₂D ATPase activity (Figure 2B). Despite the effect on SufS-SufE interactions, the “activated” versions of SufE exhibited a $K_{\mathrm{i}}$ value of 2.5 ± 0.3 μM, similar to the value determined with the unmodified wildtype SufE enzyme.

Figure 2. Inhibition of ATPase activity by SufE.

Initial velocity plots for SufBC₂D ATPase activity are shown with titrations of (A) WT SufE and (B) C51-alkylated SufE. The SufE concentrations are listed on the right side of each plot. Solid lines are from a global fit of the data to equation 1. Kinetic parameters determined from the fits are listed in Table S1. Assay conditions are described in Methods and Materials.

To validate the $K_{\mathrm{i}}$ value determined from the inhibition plot was due to a physical interaction, SufE binding to SufBC₂D was measured by a fluorescence polarization assay. This assay utilizes the C51A/E107C variant of SufE (SufE_DM) which can be labeled by a BODIPY-FL maleimide dye at the engineered E107C position [14]. Titration of the BODIPY-labeled SufE_DM with SufBC₂D resulted in increases in polarization that could be fit with the quadratic binding equation (Equation 2). This gave a $K_{\mathrm{D}}$ value of 1.6 ± 0.2 μM for formation of the SufBC₂D-E complex (Figure S1), consistent with the $K_{\mathrm{i}}$ value of 1.8 μM determined in the steady-state inhibition assay above and with the previously reported $K_{\mathrm{D}}$ value of 2.8 ± 0.5 μM for SufE-SufBC₂D interactions measured by surface plasmon resonance [12].

Mg²⁺-ATP impairs SufE-catalyzed persulfide transfer to SufBC₂D.

With SufE acting as an inhibitor for SufBC₂D ATPase activity, this suggests that persulfide delivery may be conversely regulated with ATP hydrolysis during cluster assembly. Therefore, we tested if Mg²⁺-ATP might reduce persulfide transfer from SufE C51 to SufB C254. It has previously been shown that SufBC₂D enhances the cysteine desulfurase activity of the SufS-SufE system by acting as a preferred persulfide acceptor of the C51 persulfide on SufE, increasing the rate of SufE recycling relative to inefficient SufE persulfide reduction by the non-physiological reductant DTT (Figure 3A) [8, 20]. Reduced sulfide can be detected via a colorimetric methylene blue assay to quantitate persulfide transfer through the Suf pathway. A typical persulfide transfer reaction between SufS, SufE, and SufBC₂D is shown in Figure 3B. In this assay, SufS and SufBC₂D are kept at a 1:1 stoichiometry and the concentration of SufE is varied. The data are consistent with SufBC₂D activation of SufS-SufE-dependent persulfide production (Figure 3B, red squares vs. black circles) [8, 20]. Using a C51A SufE variant (Figure 3B, green triangles) results in minimal sulfide production because DTT alone is not an efficient reductant for the C364-persulfide on SufS, resulting in trapping of SufS in the intermediate persulfide form and blocking the transfer pathway. Finally, use of a C254A SufB variant (Figure 3B, blue diamonds) results in similar kinetics to SufS and SufE alone, confirming that the increase in persulfide production in the presence of SufBC₂D is directly linked to persulfide transfer from SufE C51 to SufB C254 [11]. Next, we tested if the inclusion of 500 μM Mg²⁺-ATP alters SufBC₂D enhancement of SufS-SufE activity. The results showed a modest, but reproducible, 20% decrease in the maximal rate of desulfurase activity (Figure 3C, black circles vs. blue diamonds). The addition of Mg²⁺-ATP has no effect on the reaction with SufS-SufE alone (Figure 3C, green triangles), showing that the initial persulfide transfer between SufS-SufE is not affected.

Figure 3. SufBCD stimulation of SufS cysteine desulfurase activity.

(A) Overall reaction scheme for persulfide transfer in the assay. The persulfide on SufS is resistant to reduction by DTT. Sulfide is shown with a blue oval to indicate the methylene blue detection assay. (B) Control reactions for SufBC₂D-dependent reaction stimulation. (C) Inhibition data for Mg-ATP-containing reactions. The error bars represent the standard error for the averaged data from triplicate measurements. The solid lines are fits to the Michaelis-Menten equation. Kinetic parameters from the fits are reported in Table S2. All reaction conditions are described in Materials and Methods.

While both activities (ATPase and persulfide transfer) can be reciprocally affected, the SufE-induced inhibition of the ATPase activity is more substantial. This may be due to the fact that SufE binding to SufBC₂D can physically restrict the dimerization of SufC required for ATPase activity. In contrast, the rate determining step for the ATPase reaction is suggested to be ADP release occurring from the un-dimerized SufC conformation [21]. This would lower the steady-state population for the SufC-dimer form of SufBC₂D, which in theory is the source of inhibition in the persulfide transfer assay.

AlphaFold3 models of SufBC₂D-E complex identifies evolutionarily conserved interactions between SufC and SufE.

In the absence of a high-resolution experimental structure, AlphaFold3 (AF3) was used to predict the structures for various forms of the E. coli SufBC₂D-SufE complex (Figure 4). As a control, AF3 was used to predict the SufBC₂D complex for comparison with the reported crystal structure (PDB id: 5awf) (Figure 4). Superposition of the AF3 results show the generated structures are in good agreement with the published structure (Figure S2) [10]. The experimental structure lacks density for SufB residues 1–34 and 80–156. The AF3 models do not make strong predictions for residues 1–34; however, residues 80–156 (magenta region in Figure 4) are now modeled with several elements of secondary structure. These regions may be mechanistically important as they are predicted to form a cap over the SufB C254 residue, which is known to accept the persulfide from C51 on SufE [11].

Figure 4. AlphaFold3 models for the various SufBC₂D complexes.

Amino acid sequences for each protein were submitted to the AlphaFold3 webserver. The highest ranking models for SufBC₂D (0.77 ipTM, 0.8 pTM), SufBC₂D-E (0.78 ipTM, 0.79 pTM), and SufBC₂D-E-2MgATP (0.76 ipTM, 0.79 pTM) are shown. Models are colored by subunit with key cysteine residues shown as spheres with the thiol sulfur in yellow. SufB residues 79–157 are highlighted in magenta as these residues are not visible in the 5awf structure. The red arrows indicate the distance between C51 SufE and C254 SufB where persulfide transfer takes place. Mg-ATP is shown in yellow within SufC protomers. Models colored by AlphaFold3 confidence score and predicted alignment error plots are shown in Figure S3.

To predict possible SufE binding sites, AF3 was used to model the SufBC₂D-SufE complex (Figure 4). All five models generated by the AlphaFold3 server placed SufE at the SufB/SufC interface. This quaternary arrangement agrees with previous experimental results showing SufB and SufC are required for interactions with SufE as determined by IMAC-affinity pull-down assays [12]. Interactions between SufB and SufE require a predicted conformational change in SufB residues 95–162 (magenta in Figure 4). The predicted conformational change would allow SufE access to the C254 residue on SufB for persulfide transfer. Persulfide transfer has been shown to occur between C51 on SufE and C254 on SufB, and the AF3 models are consistent with this experimental result as the two cysteine residues are located within 13 Å of each other in the AF3 model. Extension of the C51 mobile loop from SufE during persulfide transfer would allow an even closer approach. The AF3 model of the SufB-SufE interface identifies a number of residues with potential for mediating persulfide transfer that are subject to ongoing investigations. Interactions between SufE and SufC are found between positively charged residues on the N-terminal helix of SufE and negatively charged residues on SufC (discussed in more detail below).

Based on the inhibition results reported above, we hypothesized that the interaction between SufE and SufC would be affected by Mg²⁺-ATP-induced SufC dimerization. To test this hypothesis, a second set of AF3 models were created for the SufBC₂D-SufE complex in the presence of two equivalents of Mg²⁺-ATP. In this series of models, the two SufC monomers move asymmetrically with the SufC_B monomer approaching the SufC_D monomer to form a canonical ATP-bound ABC ATPase dimeric structure. As a result of the SufC_B movement, SufE is dragged away from SufB such that the interface is significantly decreased, and the two cysteine residues are now ~30 Å apart.

Substitutions to the N-terminal helix of SufE affect functional interactions with SufBC₂D in vitro.

Analysis of the SufE-SufC interfaces predicted by AF3 identified two conserved electrostatic interactions between positively charged residues on SufE (K9 and R16) and negatively charged residues on SufC (Figures 5A,B), with an additional residue (SufE R12) making interactions in some of the AF3 models (Figure S4). These residues are well conserved in Enterobacterales but are not globally conserved in the SufE-like InterPro family (IPR003808) (Figure S5). Site-directed mutagenesis was used to substitute the three positively charged SufE residues with alanine or negatively charged glutamate to generate K9A/E, R12A/E, and R16A/E SufE variants. To induce a more drastic change, an N-terminal truncated SufE was also created by removing the first 17 residues. Each of the SufE variants was expressed and purified using methods similar to the wildtype enzyme. Results from SDS-PAGE and circular dichroism spectroscopy show that all variants were purified to homogeneity and remained folded (Figure S6). As an additional control, the SufE variants were tested for their ability to support SufS cysteine desulfurase activity (Figure S7). Other than the N-terminal truncation, all SufE variants supported SufS activity at levels comparable to or better than the WT SufE protein. The N-terminal truncated SufE was unable to stimulate SufS activity (data not shown).

Figure 5. AF3 models and kinetic impacts of mutations at the SufE-SufC interface.

(A) Overview of AF3 model for SufBC₂D-SufE interactions. The red boxed region is lighted in panel B. (B) Structural details for SufE (green) and SufC (yellow) interactions with labels residues that form a predicted electrostatic interaction. (C-E) IC₅₀ assays for inhibition of the SufBC₂D ATPase activity with N-terminal variants of SufE. Solid lines are from fits of the data to equation 2. Kinetic parameters derived from the fits are listed in Table S3. (F-H) Effect of N-terminal substitutions on persulfide transfer assays using various Suf components as noted in the legend. Solid lines are from fits of the data to the Michaelis-Menten equation. Kinetic parameters derived from the fits are shown in Table S2. Experimental conditions for panels C-H are described in Methods section.

Each SufE variant was then screened for their ability to inhibit SufBC₂D ATPase activity in the presence of 500 μM Mg²⁺-ATP (Figure 5C–5E). The inhibition curves were fit to equation 3, which describes the IC₅₀ value for partial inhibitors. As shown in Figure 2, WT SufE inhibits SufBC₂D ATPase activity with an IC₅₀ value of ~ 1 μM. Substitutions at K9 and R16 both diminish SufE inhibition of SufBC₂D ATPase activity causing changes in both IC₅₀ values and extent of inhibition (β). In both cases, the negatively-charged glutamate substitutions result in a much larger defect in inhibition relative to that seen with the neutral, alanine substituted SufE variants (Figure 5C–5E). Substitutions at R12 had very minor effects on SufE inhibition of ATPase activity, in agreement with AF3 predictions that R12 may not interact strongly with SufC in all conformations. These results suggest that K9 and R16 make specific interactions with SufBC₂D as predicted by AF3.

Further characterization showed that the N-terminal SufE substitutions also diminished the ability of SufBC₂D to activate persulfide transfer (Figure 5F–5H). Titration of K9A, R12A, and R12E SufE exhibited some activation, but none reached the extent of WT SufE. The effects of the K9E, R16A, and R16E substitutions were more severe with these variants appearing to be insensitive to the presence of SufBC₂D (Figure 5F–5H). In sum, these results validate the structural interactions between SufC and SufE that were predicted by the AF3 SufBC₂D-SufE structures and provide further support that the SufE-SufC interaction plays a role in stabilizing SufE binding for persulfide transfer to SufB.

Genetic complementation assays link severe in vitro defects to in vivo functionality of the Suf pathway.

A single substitution in the N-terminal helix of SufE is sufficient to disrupt the SufBC₂D-SufE function in vitro; however, it is unclear if this functional defect is sufficient to perturb cluster assembly by the Suf pathway in vivo. Therefore, we tested if the variant SufE genes could genetically complement a ΔsufE mutant strain of E. coli in vivo. The ΔsufE mutant is sensitive to stress that perturbs Isc-dependent Fe-S cluster biogenesis, such as iron starvation stress caused by the addition of 2,2’-bipyridyl (BiPy), a cell-permeable ferrous iron chelator (Figure 6) [22]. Addition of a plasmid carrying wild-type SufE was able to rescue most of the ΔsufE sensitivity to iron starvation. In contrast, a SufE C51A mutant protein, which lacks the active site Cys residue needed to accept persulfide from SufS, could not rescue growth under iron starvation conditions (Figure 6). The SufE mutant that lacks the first 16 residues of the N-terminal region (N-trunc) was also not able to rescue growth, similar to the inactive SufE C51A mutant and empty vector control and consistent with the failure of this mutant to enhance SufS activity (data not shown). The SufE K9E and K16E mutants were able to provide a partial rescue, consistent with the biochemical data suggesting they have reduced but not abolished interactions with SufBC₂D. All strains grew equally well on LB plates without BiPy (data not shown).

Figure 6. Genetic complementation of the ΔsufE strain with SufE point mutants.

E. coli strains carrying the pET21a empty vector or various pET21a-sufE plasmids were grown as described in the Methods section. Serial dilutions of liquid cultures were spotted on Lennox broth (LB) plates with or without 250 μM 2,2’-bipyridyl (BiPy). Growth on plates was monitored for 24 – 48 hours at 30°C. The image shown is after 40 hours growth on BiPy plates.

Discussion

The SufBC₂D complex is the most common scaffold for Fe-S cluster assembly in bacteria [23]. It must coordinate multiple steps in cluster assembly including persulfide delivery/reduction, iron acquisition, nucleotide hydrolysis, and cluster trafficking. Despite the centrality of the SufBC₂D scaffold to bacterial Fe-S cluster assembly, there are limited mechanistic insights into the coordination or timing of the assembly steps. A first step to characterizing the overall mechanism of assembly is to identify specific reactivities of the SufBC₂D complex that are affected by the presence of substrates/ligands for other reactions. Here, we have used the Suf system from E. coli as a model and focused on the connection between persulfide delivery by SufE and the SufBC₂D ATPase activity.

SufE is a potent inhibitor of SufBC₂D ATPase activity, which could prevent futile hydrolysis of ATP.

The clearest impact of this work is the identification of SufE as an inhibitor for the ATPase activity of the SufBC₂D complex. Agreement between the independently determined K_i and K_D values of 1.8 μM and 1.6 μM, respectively, provides strong support that the inhibition is directly due to binding of SufE to the SufBC₂D complex. It was previously reported that SufB alone was not sufficient to bind SufE and that the presence of SufC was also required [12], foreshadowing a SufE-SufC interaction. The most logical interpretation of these results is that formation of the SufBC₂D-SufE complex physically prevents dimerization of the two SufC subunits, which is required for ATP hydrolysis, through specific SufE-SufC interactions. SufC alone is monomeric with poor basal ATPase activity. The ATPase activity of SufC is enhanced upon formation of the SufBC₂D complex due to the conformational change of an active site loop in SufC[10], although SufC dimerization would still be required for full ATP binding and hydrolysis. Therefore, the presence of SufE may reduce SufC dimerization to help prevent uncoordinated hydrolysis of ATP by the activated SufC subunits.

Insight into a mechanism for SufE exchange between SufS and SufBC₂D

The use of a transpersulfurase partner between the desulfurase and scaffold proteins in type II systems, rather than direct transfer of persulfide to the scaffold in ISC systems, raises several mechanistic challenges for the pathway. In the E. coli Suf pathway, SufE must interact with both SufS and SufBC₂D to facilitate persulfide trafficking. SufE has been shown to bind to SufS in a two-step process with a K_D value of ~5 μM in the absence of a SufS-persulfide intermediate that increases to ~0.5 μM when the SufS-persulfide is present [14]. A conformational change in a SufS active site loop linked to persulfide formation is hypothesized to generate the tight-binding interaction, which could revert back the moderate-binding interaction after persulfide transfer [24, 25]. The ~2 μM binding constant for formation of the SufE-SufBC₂D complex lies in between these two values and would allow SufBC₂D to compete for the SufE-persulfide form.

The SufE exchange mechanism also presents structural issues as there must be some overlap between the region that SufE uses to interact with SufS and SufBC₂D as the SufE C51 residue serves as a common locus in the reaction. There would likely need to be unique regions of SufE that would only interact with SufS or SufBC₂D to coordinate persulfide transfer between them. Indeed, a superposition of SufE positions in the SufBC₂D-E AF3 model and the SufS-SufE structure does show the C51-face of SufE may have mutually exclusive interactions with SufS and SufB along two faces of the SufE fold (Figure S8). However, the interactions between the N-terminal helix of SufE and SufC provide a unique SufE face that is not involved in SufS interactions which may help drive persulfide delivery specificity.

A potential mechanistic link for persulfide delivery and ATP hydrolysis.

The disruption of inhibition parameters and slow-growth phenotypes for the N-terminal SufE variants suggest persulfide transfer and ATP hydrolysis are not fully independent activities. In vitro, ATP is not required for delivery of persulfide from SufS-SufE to SufBC₂D nor is it required for Fe-S cluster assembly on SufBC₂D. However, it is possible that in vitro reconstitution of Fe-S clusters on SufBC₂D is at least partially driven by the ability of non-physiological reductants (such as DTT) to reduce the C254 persulfide intermediate on SufB to generate free SH₂. In that case, release of sulfide in solution locally near SufBC₂D may facilitate in vitro assembly of clusters by SufBC₂D via simple sulfide diffusion to the cluster binding site within SufBC₂D (Figure 7, pathway A). Such a process would be similar to non-catalyzed Fe-S cluster reconstitution on SufBC₂D using Na₂S and FeCl₃ as simple sulfur and iron donors.

Figure 7. Potential model for persulfide delivery regulation by ATP.

Pathway A represents a scaffold model where DTT reduces the SufB persulfide to generate sulfide locally. This sulfide could then be used with exogenous iron to assemble a cluster on the SufBC₂D scaffold while bypassing any in vivo mechanistic requirements. Pathway B represents a possible catalytic model where ATP and FADH₂ are required for sulfide transfer to the cluster assembly site within the SufB cavity.

In contrast, in vivo SufBC₂D is thought to traffic the C254 persulfide along a tunnel within the complex to a cluster binding site where the persulfide is reduced to sulfide using a physiological reductant, likely FADH₂, which binds SufBC₂D and copurifies with it (Figure 7, pathway B). Furthermore, in vivo SufBC₂D must likely acquire iron from a donor partner since the Suf pathway is operational under conditions when labile iron is scarce (i.e. iron starvation and oxidative stress). Previous results have suggested that SufC ATPase activity is required for iron acquisition from an unknown donor in vivo [13]. Point mutations in SufC that abolish ATPase activity (but do not alter the SufBC₂D complex) reduce the iron content in as-purified SufBC₂D to a greater extent than the sulfide content. That previous finding is consistent with our results here where SufE binding (and presumably persulfide delivery) does not require ATP and is in fact antagonistic with SufC ATP hydrolysis.

In light of the antagonistic relationship between SufE persulfide delivery and ATP hydrolysis, it may be reasonable to suggest these two activities are carefully coordinated during the in vivo SufBC₂D-catalyzed cluster assembly mechanism. We propose that SufE persulfide delivery and ATP binding are mutually exclusive, and both may be required prior to iron acquisition from the native donor (Figure 7, pathway B). The measurement of similar rates for SufS/E cysteine desulfurase and SufBC₂D ATPase activities of ~10–20 min⁻¹ provides additional circumstantial evidence of functional linkage that would prevent a buildup of intermediates in the assembly process.

Our results here are consistent with recent biochemical analysis of the Archaeal SmsBC complex published while this manuscript was under review.[26] SmsB and SmsC from Archaea have been proposed to represent the ancestral precursors to the modern bacterial SufB and SufC proteins, respectively [27]. The Sms system is phylogenetically distinct from Suf and lacks homologues to SufA, SufS, or SufE. It exists primarily in obligate anaerobes that live in H₂S-rich environments. The recent SmsBC studies revealed that SmsC ATP binding is mutually exclusive with binding of an intact [4Fe-4S] cluster [26]. The presence of non-hydrolyzable ATP analogues prevents cluster formation on SmsBC and, conversely, the presence of an intact Fe-S cluster on SmsBC blocks ATP analogue binding. Finally, the authors did not observe any alteration of the bound Fe-S cluster upon subsequent addition of ATP to the SmsBC complex. These independent results support that Suf and Sms scaffold complexes utilize conformational gating to ensure that inappropriate reaction intermediates or protein-protein interactions do not occur at the wrong step in cluster assembly. Moving forward, a careful comparison of Suf and Sms systems will be informative as the two pathways diverge in several important ways. Most notably, Sms appears to use H₂S or a derivative as the sulfur source rather than a dedicated transpersulfurase system like SufSE, and the SmsC protein has a C-terminal extension that allows it to coordinate the [4Fe-4S] cluster, while SufC does not.

While the work presented here provides evidence for reciprocal regulation of SufE persulfide delivery and SufC ATPase activity, many questions remain unanswered. It is possible that FADH₂, which copurifies with SufBC₂D, serves as the reductant for converting persulfide to sulfide in the SufBC₂D catalyzed reaction, but this has yet to be examined experimentally. Furthermore, while we hypothesize that ATP binding and/or hydrolysis drives conformational changes needed for iron acquisition, we do not yet know the in vivo iron donor(s) for SufBC₂D, and we cannot fully reconstitute the physiologically relevant pathway in vitro. Relying on low-molecular weight iron salts for the reconstitution is problematic since it is unlikely such compounds are used for Fe-S cluster biogenesis under iron starvation or oxidative stress conditions. It is also unclear how the delivery of multiple persulfides is regulated and how they are stored prior to reduction and cluster formation. SufBC₂D can bind a mixture of cluster types ranging from [2Fe-2S] to [4Fe-4S], so multiple rounds of sulfide and iron acquisition are needed. Elucidation of the coordinated mechanism of persulfide transfer and reduction, ATP hydrolysis, and iron acquisition will require a multiprong approach that includes identification of an in vivo iron donor, biochemical and structural characterization of all proteins involved, and further advanced kinetic analysis carried out under physiological conditions.

Conclusions

The lack of mechanistic details for Fe-S cluster assembly in the bacterial Suf pathway stands in contrast to the well-developed Isc pathway. Here, we have shown that SufE binding to the SufBC₂D complex results in inhibition of ATP hydrolysis by the SufC subunits. As SufE is responsible for persulfide delivery, this result suggests that persulfide delivery and ATP hydrolysis are mutually exclusive steps in the cluster assembly process. Disruption of the SufE-SufC interface, as predicted by AlphaFold3, results in defects in ATPase inhibition parameters and N-terminal SufE variants result in slow-growth phenotypes under iron-stress complementation assays. The in vitro and in vivo results support an ordered mechanism for cluster assembly where persulfide delivery from SufE may limit ATPase-dependent iron acquisition by SufBC₂D.

Data Availability Statement

The data underlying this article will be shared on reasonable request to the corresponding author.