Molecular mechanism of the dual regulation of bacterial iron sulfur cluster biogenesis by CyaY and IscX

IscX (or YfhJ) is a protein of unknown function which takes part in the iron-sulfur cluster assembly machinery, a highly specialised and essential metabolic pathway. IscX binds to iron with low affinity and interacts with IscS, the desulfurase central to cluster assembly. Previous studies have suggested a competition between IscX and CyaY, the bacterial ortholog of frataxin, for the same binding surface of IscS. This competition could suggest a link between the two proteins with a functional significance. Using a hybrid approach, we show here that IscX is a modulator of the inhibitory properties of CyaY: by competing for the same site on IscS, the presence of IscX rescues the rates of enzymatic cluster formation which are inhibited by CyaY. The effect is stronger at low iron concentrations, whereas it becomes negligible at high iron concentrations. These results strongly suggest that iron-sulfur cluster assembly is an exquisite example of an enzymatic process which requires a double regulation under the control of iron as the effector.


Introduction
Iron and sulfur are elements essential for life thanks to their unique redox properties.
Yet, they are highly toxic. An efficient way to store them in cells in a nontoxic form is through formation of iron sulfur (Fe-S) clusters, labile prosthetic groups involved in several essential metabolic pathways (for a review (1,2)). Assembly of Fe-S clusters is carried out by highly conserved machines which, in prokaryotes, are encoded by the suf, nif and isc operons. Isc is the most general machine with highly conserved orthologs in eukaryotes.
In E. coli, the isc operon contains eight genes (i.e. iscR, iscS, IscU, iscA, fdx, hsca, hscb and iscX) (3). Among the corresponding gene products, the most important players are the cysteine desulfurase IscS (EC 2.8.1.7), which converts cysteine to alanine and IscS-bound persulfide (4), and IscU, a transient scaffold protein which forms a complex with IscS (5,6). The last component of the machine according to the order of genes in the operon is IscX (also known as YfhJ), a small acidic protein about which very little is known (7,8). IscX, the Cinderella of the isc operon, is not essential, in contrast to the other isc proteins (7). It is exclusively present in prokaryotes and in eukaryotes of the Apicomplexa, where it is highly conserved (9).
Based on its phylogenetic occurrence, IscX seems to depend on the presence of IscS whereas the reverse is not the case (9).
The structure of E. coli IscX consists of a classical helix-turn-helix fold often found in transcription regulators (9,10). In vitro studies have shown that IscX is able to bind IscS, thus suggesting a role for IscX as a molecular adaptor (9,11,12). IscX also binds to iron through a negatively charged surface, the same by which it recognises IscS (9). An exposed negatively charged iron binding surface which overlaps with the surface of interaction with IscS is also a feature present in another protein, CyaY (the ortholog in bacteria of the eukaryotic frataxin). This protein has attracted much attention because in humans it is associated with Friedreich's ataxia (13). In contrast to IscX, frataxins are proteins highly conserved from bacteria to high eukaryotes (14) and are essential in eukaryotes (15). In prokaryotes, CyaY is external to the isc operon but has extensively been implicated in Fe-S cluster assembly (16). We have in the past shown that CyaY is an IscS regulator, which dictates the enzymatic assembly of Fe-S clusters (17,18). Intriguingly, IscX and CyaY compete for the same site on IscS (11,12,18). A genetic interaction between CyaY and IscX was also demonstrated by a recent study which has validated a role of IscX as a new bona fide Fe-S cluster biogenesis factor (19). In some species, CyaY and IscX seem to replace each other.
These data raise the compelling question of whether there could be a functional link between CyaY and IscX, which could both elucidate the function of IscX and explain why these two proteins compete for the same binding site. Given the high conservation between prokaryotic and eukaryotic frataxins and desulfurases, answering this question could also inspire new studies on the regulation of eukaryotic frataxin. Using a complementary approach which makes use of enzymology, crosslinking and structural methods, we provide conclusive evidence indicating that IscX is a modulator of CyaY that switches off the inhibitory properties of CyaY as a function of the iron concentration. At low iron concentrations, IscS is under IscX control which has no inhibitory capacity. At high iron concentrations the system becomes controlled by CyaY. Based on our results, we suggest a general scheme that supports a role of CyaY as an iron sensor and provides the first testable indications concerning the function of IscX.

IscX has different effects on the kinetics of cluster formation as a function of concentration
We started by exploring the effect of IscX on the rates of enzymatic assembly of the Fe-S cluster on IscU using an assay in which the cluster forms, under strict anaerobic conditions, through IscS-mediated conversion of cysteine to alanine and persulfide and is reconstructed on IscU (17). We performed the experiment at increasing concentrations of IscX in the range 1-50 M using 1 M IscS, 50 M IscU, 250 M Cys, 2 mM DTT and 25 M Fe 2+ . We did not observe significant variations of the kinetics (Figure 1A,B) up to ca. 10 M, a range in which the initial rates in the presence or absence of IscX were superposable, revealing that IscX does not have inhibitory effects under these conditions. Higher concentrations of IscX (from 20 to 50 M, i.e. higher IscX/IscS ratios) led instead to inhibition of the reaction. Similar observations were made using circular dichroism (CD) (Figure 1C). The results at higher IscX/IscS ratios are in agreement with a previous study (12) in which the kinetics were performed at high concentrations of both Fe 2+ (125 M) and IscX (25 M), but the results at lower IscX/IscS ratios are new and surprising. Why does IscS have a different behaviour at low and high ratios? There are different hypotheses which could explain these observations. First, the effect we observed at high IscX/IscS ratios could be explained by loss of available iron which could be sequestered by IscX, since we have reported elsewhere that IscX can undergo ironpromoted aggregation (9,18). However, it was noted that aggregation occurs only when the assay is carried out at very low ionic strength. When salt is present, as was the case here, aggregation is not observed. Second, at low ratios the IscX occupancy on IscS could be too low to detect an effect because the complex has low affinity.
Third, there could be a secondary binding site for IscX on IscS which is occupied only when the primary site is fully occupied. Only the 2:1 IscX-IscS complex would behave as an inhibitor.

The presence of IscU does not influence the affinity of IscX for IscS
To test the complex occupancy, we reconsidered the dissociation constants. We had previously estimated by fluorescence labelling and calorimetry dissociation constants (Kds) of 12 M and 20 M for the binary IscX-IscS and CyaY-IscS complexes respectively (9,18). The IscX-IscS complex should thus be >50% populated under the conditions of our cluster formation assay. However, the presence of IscU could in principle modify these affinities as it is observed for binding to IscS when both IscU and CyaY are present. Also, although CyaY and IscX compete for the same site of IscS, the IscS dimer could have, at low IscX-IscS ratios, simultaneous occupancy of both CyaY and IscX. We used Biolayer Interferometry (BLI), a technique which can measure weak molecular interactions between several partners, to test this possibility.
When we immobilized IscS on the surface and titrated with IscX only, we obtained a Kd of 8±3 M for the IscX-IscS binary complex, in excellent agreement with the previous measurements (data not shown). When we immobilized CyaY on the surface, saturated with IscS and titrated with increasing quantities of IscX we obtained a Kd of 6±2 M for the IscX-IscS complex (Figure 2A). Thus, the presence of CyaY on IscS does not modify the affinity of IscX for the desulfurase, and so major allosteric effects or binding cooperativity between the two proteins are unlikely.
When we tested binding of IscX to the IscU-IscS complex (immobilizing IscS saturated with IscU), we obtained a Kd of 8±2 M, a value comparable to the one obtained in the absence of IscU ( Figure 2B). These data indicate that the presence of IscU bound to IscS does not affect the affinity for IscX and that the affinity of IscX for IscS is higher than that of CyaY. Thus, the effect observed at different IscX-IscS ratios cannot be ascribed to insufficient occupancy.

IscX has two different binding sites on IscS
BLI proved inconclusive towards the presence of a secondary binding site. This could either be because we did not explore sufficiently high molar ratios, or because BLI depends on there being a significant change in the distance between the sensor's internal reference layer and the solvent interface. For formation of some complexes this can be rather small and thus undetectable. We thus used electrospray ionisation (ESI) mass spectrometry (MS) under non-denaturing conditions, a technique which, if optimised, provides direct information on all complexes present in a solution. The m/z spectrum (4100 -6300 m/z) of IscS displayed well-resolved charge states ( Figure   S1A). The deconvoluted spectrum of IscS revealed a major peak at 91,035 Da consistent with the presence of dimeric IscS (predicted mass 91,037 Da) ( Figure S1B and Table 1). Occasionally, shoulder peaks at +32 Da intervals on the high mass side of the peak were observed; these are likely due to the presence of one or more sulfane sulfur atoms, as previously observed for other proteins (20). An unknown adduct at +304 Da relative to the main IscS peak was also observed ( Figure S1B This evidence conclusively supports the presence of two binding sites for IscX on IscS.

Mapping the two binding sites by cross-linking
We next used cross-linking experiments to map the IscX binding sites on IscS, attempting to transform the non-covalent interaction in a stable covalent bond. In this assay, a cross-linking agent was added to link covalently proteins in close spatial proximity (21). We used bis[sulfosuccinimidyl]suberate (BS3), a cross-linking agent that reacts with primary amino groups up to 11.5 Å apart, to mixtures of IscS and IscX at different molar ratios. The reaction produced two covalently attached protein complexes with apparent molecular weights of 52 kDa and 60 kDa, tentatively identified on PAGE as 1:1 and a 2:1 IscX-IscS complexes respectively (Figure 4).

Mass spectrometry (MS) confirmed the presence of only these two proteins in both
bands. Identification of the cross-link nature was achieved by subjecting to trypsin digestion the two species, as well as the isolated IscS. The obtained peptide mixtures were then analysed by MALDI/MS (Table S2 of

Structural characterization of the secondary binding site of IscX on IscS
Since it is difficult to assess the significance of these binding sites without a working structural model, we used small-angle X-ray scattering (SAXS) to grasp the shape of  To construct a low resolution model of the IscX-IscS complex, rigid body modelling was used taking into account the possibility of 1:1 and 2:1 complexes.
Multiple runs of SASREFMX starting from random initial configurations with applied P2 symmetry yielded models with qualitatively good fits to the data at a 1:1 molar ratio with a χ 2 of 1.7. The interaction sites of IscS and IscX determined by NMR and cross-linking experiments were used as restraints (9,12). The presence of unbound (free) components were taken into account by using free IscS dimers and IscX monomers as independent components in addition to the SASREFMX models.
Their relative volume fractions were refined with the program OLIGOMER ( Table   S4). The volume fraction of free IscS decreases at a large excess of IscX consistent with an almost full occupancy of IscX. While it is difficult to distinguish between the fitting for 1:1 and 2:1 complexes, fits assuming both types of complexes are marginally better than those obtained assuming the presence of 2:1 complexes only.
We estimated ca. 60-70% of 2:1 IscX-IscS complexes. The structure of the 1:1 complex is in excellent agreement with a previous model obtained by the same hybrid technique (12), IscX sits in the catalytic pocket of IscS contributed by both protomers of the IscS dimer (Figure 5B,C). This is the site where also CyaY binds (12,18) ( Figure 5D) thus explaining the direct competition between the two molecules.
Experimental validation of this binding site has been supported by mutation studies of IscS (11). The second site is close but shifted towards the IscU-binding site.
The SAXS model suggests a logical explanation for the observed effects. We have previously demonstrated by molecular dynamics that binding of CyaY in the primary site restricts the motions of the catalytic loop which is thought to transfer persulfide, bound to Cys237, from the catalytic site to IscU (23). As compared to CyaY, IscX is smaller and would not be sufficient to block the movement of the catalytic loop as long as only the primary site of IscX is occupied. In this position, IscX simply prevents CyaY from binding, having higher affinity, but does not act as an inhibitor. However, binding of another IscX molecule with its N-terminus close to K101, as observed by cross-linking, can strongly increase the steric hindrance and result in inhibition of iron sulfur cluster formation resulting in an effect similar to that observed for CyaY.

Understanding the competition between IscX and CyaY
We then explored more closely the conditions under which IscX competes with CyaY (11,12). We first repeated the enzymatic assays keeping the conditions unchanged (1 IscX has no effect on IscS activity as demonstrated above, the experiment results in a progressive rescuing of CyaY inhibition (Figure S7), in agreement with a competition between the two proteins for the same binding site on IscS. The effect of IscX becomes noticeable at substoichiometric concentrations as compared to CyaY, as expected from the respective dissociation constants. However, even under the maximal condition of rescuing activity (10 M), IscX is able to reach only ca. 70% of the rate compared to a control experiment performed in the absence of CyaY where there is no inhibition. This reflects the fact that, at relatively high IscX/IscS ratios, the secondary binding site, which also leads to inhibition, starts being populated probably favoured by a conformational change or an electrostatic rearrangement. Above a 10:1 IscX/IscS ratio, increase of the IscX concentration results in progressively marked inhibition ( Figure 6A). When we titrated instead the system with CyaY (0-20 M), keeping IscX fixed at 5 M, we started observing a decrease of the intensities at 1:1 IscX/CyaY molar ratios and a marked inhibition for higher CyaY ratios (Figure 6B).
We then put our results within the context of the protein concentrations observed in E.
coli. Several independent studies have reported these under different growth conditions (24)(25)(26)(27). Under non-stress conditions, both CyaY and IscX are usually, but not always, present at substoichiometric ratios as compared to IscS, with IscX in excess over CyaY (25). Different conditions can however remarkably change the molar ratios. We thus repeated the enzymatic assay using concentrations comparable to the non-stress conditions to verify that IscX controls the reaction under these conditions (4.1 M IscS, 250 M Cys, 2 mM DTT, 1.0 M IscX, 0.7 M CyaY and 25 M Fe 2+ ). We used IscU in a large excess (50 M) because it is the reporter. We preferred to use this protein rather than another reporter because the IscU binding site on IscS does not interfere with CyaY or IscX (9,18). At these ratios we do not observe inhibition indicating that the reaction is under the control of IscX ( Figure   6C). Taken together, these results fully confirm the competing role of IscX and CyaY. because we had already observed that CyaY inhibition is enhanced by iron (17) ( Figures 7A,B). As a control, we compared the variation of the initial rates as a This implies that, at lower iron concentrations, IscX has higher affinity for IscS than does CyaY, impairing the ability of CyaY to inhibit cluster formation. The effect is reversed at higher iron concentrations where CyaY must acquire a higher affinity.
To dissect whether iron dependence is controlled by CyaY, IscX or both proteins together, we fixed the concentration of IscX to 1 M and repeated the experiments varying the concentrations of Fe 2+ from 10 to 100 M in the presence and absence of IscX in the absence of CyaY. We observed that the initial rates increased at increasing Fe 2+ concentrations but were independent of the presence of IscX ( Figure 7C). Thus, the interaction of IscX to IscS is Fe 2+ independent in this range of concentrations. This is at variance with CyaY whose interaction with IscS is clearly iron mediated, as previously demonstrated (17): the same experiment carried out with CyaY showed a marked difference in the absence and presence of CyaY which increases with iron.
We thus must conclude that IscX acts as a modulator of CyaY inhibition, resulting from the ability of CyaY to sense iron and bind to IscS more tightly at high iron concentrations.

Discussion
Since its discovery (28), iron-sulfur cluster biogenesis has rapidly become an important topic of investigation both because it is a molecular machine at the very basis of life and because of its crucial role in an increasing number of human diseases (29). The study of this topic in the bacterial system is also specifically important given the important role that iron-sulfur cluster in bacterial infections (30). Much progress has been made in understanding this metabolic pathway but several questions have so far remained unanswered. Why does IscS bind different proteins using the same surface and how is recognition regulated? What is the role of IscX in the isc machine? What is the relationship between IscX and CyaY? We have now tentative answers to these questions.
Previous studies carried out at high IscX concentrations (>10 M) led to the conclusion that IscX, like CyaY, is an inhibitor of cluster formation (12). We could reproduce these results but demonstrated that they strongly depend on the relative and absolute concentrations used in vitro. At high IscX-IscS ratios (i.e. >10:1), the inhibitory effect of IscX is the consequence of a secondary binding site on IscS which becomes populated only once the primary site is saturated. At low IscX-IscS ratios, IscX has no effect on cluster formation but efficiently competes with CyaY and modulates its strong inhibitory power also at substoichiometric ratios. This is an important result which changes our perspective on IscX: this protein is not just "another frataxin-like protein" but the modulator of the inhibitory function of CyaY in bacteria; it is what silences CyaY. Our data are fully consistent with and explain in molecular terms a previous report that conclusively demonstrated a genetic interaction between CyaY and IscX and an additive positive effect on cluster maturation upon deletion of both genes (19).
We characterized the 2:1 IscX-IscS complex to obtain, albeit at low resolution, clues about the mechanism by which CyaY and IscX act as inhibitors. The two proteins compete for the same binding site on IscS but IscX is appreciably smaller than CyaY. In the complex with only the primary site occupied, IscX can thus allow free movement of the catalytic loop, which transports the persulfide from the active site to IscU (23). Occupancy of the secondary binding site would interfere instead with the loop movement by steric hindrance, thus producing an inhibitory effect. We also demonstrated that the effect of IscX on cluster formation depends on the iron concentration and that the iron sensor is CyaY and not IscX, whose effect on IscS is iron independent.
We can thus suggest a model based on these results which fully explains the role of both proteins (Figure 8 present. If this balance is lost, there will be too many "highly reactive" Fe-S clusters not being delivered which will be degraded. From an evolutionary perspective, our results explain the presence of IscX in prokaryotes and not in most eukaryotes: bacterial IscS is a fully active enzyme, whereas the eukaryotic ortholog Nfs1 is inactive in the absence of its activators frataxin and the eukaryotic specific Isd11 and ACP (31,32). As demonstrated in a recent paper, this behaviour can be explained by a different quaternary assembly of the desulfurase complex with Isd11 mediated by the acyl carrier protein in this structure (31,32). These results indicate a different regulation of Fe-S cluster biogenesis in eukaryotes compared to prokaryotes. We propose the mechanistic basis of the bacterial regulation based on our results. We can also speculate that, as a regulator of the inhibitory function of CyaY in bacteria, IscX could have disappeared in the passage from prokaryotes to eukaryotes where the desulfurase was deactivated making the double repression unnecessary. This hypothesis would fully explain the difference in the function of frataxin between prokaryotes and eukaryotes, and suggests that IscX is the evolutionary "missing link".

Protein production
All proteins used are from E. coli. Their sequences were subcloned in a pET 24d vector modified as fusion proteins with His-tagged glutathione-S-transferase (GST).
The constructs were expressed in BL21(DE3). Bacteria expressing IscU were grown in Luria Broth enriched medium containing 8.3 MgZnSO4 (33) to stabilize its fold.
The proteins were purified as previously described (9,34) (36). For further details on data treatment see Suppl. Mat.