In vivo [Fe-S] cluster acquisition by IscR and NsrR, two stress regulators in Escherichia coli

Authors

  • Daniel Vinella,

    1. Laboratoire de Chimie Bactérienne, UMR 7283 (Aix-Marseille Université-CNRS), Institut de Microbiologie de la Méditerranée, Marseille, France
    Current affiliation:
    1. Institut Pasteur, Paris, France
    Search for more papers by this author
  • Laurent Loiseau,

    1. Laboratoire de Chimie Bactérienne, UMR 7283 (Aix-Marseille Université-CNRS), Institut de Microbiologie de la Méditerranée, Marseille, France
    Search for more papers by this author
  • Sandrine Ollagnier de Choudens,

    1. Laboratoire de Chimie et Biologie des Métaux, UMR 5249 (CEA-Université Grenoble I-CNRS), Grenoble Cedex, France
    Search for more papers by this author
  • Marc Fontecave,

    1. Laboratoire de Chimie et Biologie des Métaux, UMR 5249 (CEA-Université Grenoble I-CNRS), Grenoble Cedex, France
    2. Collège de France, Paris Cedex 05, France
    Search for more papers by this author
  • Frédéric Barras

    Corresponding author
    • Laboratoire de Chimie Bactérienne, UMR 7283 (Aix-Marseille Université-CNRS), Institut de Microbiologie de la Méditerranée, Marseille, France
    Search for more papers by this author

For correspondence. E-mail barras@imm.cnrs.fr; Tel. (+33) 4 91 16 45 79; Fax (+33) 4 91 71 89 14.

Summary

The multi-proteins Isc and Suf systems catalyse the biogenesis of [Fe-S] proteins. Here we investigate how NsrR and IscR, transcriptional regulators that sense NO and [Fe-S] homeostasis, acquire their [Fe-S] clusters under both normal and iron limitation conditions. Clusters directed at the apo-NsrR and apo-IscR proteins are built on either of the two scaffolds, IscU or SufB. However, differences arise in [Fe-S] delivery steps. In the case of NsrR, scaffolds deliver clusters to either one of the two ATCs, IscA and SufA, and, subsequently, to the ‘non-Isc non-Suf’ ATC, ErpA. Nevertheless, a high level of SufA can bypass the requirement for ErpA. In the case of IscR, several routes occur. One does not include assistance of any ATC. Others implicate ATCs IscA or ErpA, but, surprisingly, SufA was totally absent from any IscR maturation pathways. Both IscR and NsrR have the intrinsic capacity to sense iron limitation. However, NsrR appeared to be efficiently matured by Isc and Suf, thereby preventing NsrR to act as a physiologically relevant iron sensor. This work emphasizes that different maturation pathways arise as a function of the apo-target considered, possibly in relation with the type of cluster, [2Fe-2S] versus [4Fe-4S], it binds.

Introduction

[Fe-S] proteins are present in all living cells, where they participate in a wide range of physiological processes including respiration, photosynthesis, DNA repair, metabolism and regulation of gene expression (Kiley and Beinert, 2003; Fontecave, 2006; Py and Barras, 2010; Fleischhacker and Kiley, 2011; Py et al., 2011). The mechanism of [Fe-S] cluster formation and insertion into proteins involve several protein factors, most of them conserved throughout eukaryotes and prokaryotes (Kennedy and Dean, 1992; Zheng and Dean, 1994; Balk and Lobreaux, 2005; Barras et al., 2005; Johnson et al., 2005; Ayala-Castro et al., 2008; Fontecave and Ollagnier-de-Choudens, 2008; Lill and Muhlenhoff, 2008; Xu and Moller, 2008; Py and Barras, 2010). Briefly, the Isc system is thought to be the housekeeping biogenesis pathway, as its inactivation leads to a drastic reduction of the activity of all [Fe-S] enzymes tested under normal growth conditions (Schwartz et al., 2000; Takahashi and Tokumoto, 2002; Outten et al., 2004; Mettert et al., 2008; Trotter et al., 2009). In contrast, Suf is believed to be the back-up system for [Fe-S] cluster biosynthesis, which functions under stress conditions such as iron starvation or in an oxidative environment (Takahashi and Tokumoto, 2002; Outten et al., 2004; Ayala-Castro et al., 2008; Jang and Imlay, 2010).

The Isc and Suf systems include cysteine desulphurases, IscS and SufS, which provide sulphur to the scaffold proteins, IscU and SufB, respectively, onto which [Fe-S] clusters form transiently before being transferred to apo-proteins (Flint, 1996; Agar et al., 2000; Schwartz et al., 2000; Takahashi and Tokumoto, 2002; Loiseau et al., 2003; Tokumoto et al., 2004; Johnson et al., 2005; Layer et al., 2007; Trotter et al., 2009). The Suf and the Isc systems also include ATP-hydrolysing proteins, SufC, an ABC ATPase (Nachin et al., 2003; Layer et al., 2007), and HscBA, a DnaKJ-like co-chaperone system (Hoff et al., 2000; Vickery and Cupp-Vickery, 2007), respectively, which function in cluster transfer. Last, both systems include so-called A-type carriers (ATC), IscA and SufA, which play a predominant role in the delivery of already made [Fe-S] clusters to apo-proteins, although a role in iron donation has also been proposed (Krebs et al., 2001; Ding and Clark, 2004; Gupta et al., 2009; Vinella et al., 2009). However, it is fair to say that the more genome sequences analysed, the more diverse the organization of Suf-like and Isc-like systems appears. For instance, Gram-positive organisms contain a suf operon that comprises an iscU-like gene, encoding the so-called SufU scaffold, along with the sufBCD encoding scaffold, no sufE and a sufA-like elsewhere in the chromosome (Riboldi et al., 2009; Albrecht et al., 2010; Py et al., 2011). Besides the Isc and Suf factors, several others proteins have potential roles in [Fe-S] clusters biogenesis. These are (i) frataxin and ferritin proteins, which were proposed to provide iron (Patzer and Hantke, 1999; Nachin et al., 2003; Layer et al., 2006; Justino et al., 2007; Velayudhan et al., 2007), (ii) NfuA, GrxD and Mrp (ApbC), which have carrier or scaffold functions (Angelini et al., 2008; Boyd et al., 2009; Yeung et al., 2011; Py et al., 2012), (iii) ErpA, an ATC homologous to IscA and SufA, which is essential for Escherichia coli to grow under aerobiosis (Loiseau et al., 2007; Vinella et al., 2009), (iv) YtfE and YggX, which might be involved in [Fe-S] cluster repair (Overton et al., 2008; Thorgersen and Downs, 2009), and (v) YgfZ, the role of which remains to be ascertained (Waller et al., 2012).

IscR and NsrR are [Fe-S] transcriptional regulators, which have orthologues in a wide range of Gram-negative and Gram-positive bacteria (Keon et al., 1997; Rodionov et al., 2005; Fleischhacker and Kiley, 2011). IscR and NsrR regulate the expression of at least 40 and 60 genes respectively (Schwartz et al., 2001; Giel et al., 2006; Filenko et al., 2007; Nesbit et al., 2009; Partridge et al., 2009). NsrR and IscR are similar over the length of the proteins (Fig. S1). They have a helix–turn–helix motif close to the N-terminus and three cysteine residues that bind a [Fe-S] cluster close to the C-terminus (Tucker et al., 2010). IscR is the key regulator of [Fe-S] homeostasis, as the [2Fe-2S] holo-form of IscR represses the expression of the isc operon and the apo-form activates the transcription of the suf operon (Schwartz et al., 2001; Giel et al., 2006; Yeo et al., 2006). The balance between apo- and holo-IscR is predicted to be influenced by growth conditions such as oxidative and nitric oxide (NO) stresses or iron limitation. NsrR is dedicated to the detection of NO stress and the protection against reactive nitrogen species (Beaumont et al., 2004; Bodenmiller and Spiro, 2006; Rankin et al., 2008; Tucker et al., 2010). Despite that the Streptomyces coelicolor and Neisseria gonorrhoeae NsrR proteins were found to contain [2Fe-2S] clusters, most recent results with the Bacillus subtilis NsrR established that it carries a [4Fe-4S] cluster (Yukl et al., 2008). Regardless of the type of cluster, it is predicted to be destabilized by nitrosylation, which will yield a form unable to bind DNA (Rankin et al., 2008; Tucker et al., 2008; Isabella et al., 2009; Partridge et al., 2009; Kommineni et al., 2012).

In this work, we compared the [Fe-S] maturation requirement for IscR and NsrR. Use of lacZ transcriptional fusions allowed us to detect variations in the efficiency of maturation of these two stress regulators. This led us to define genetically the routes used by [Fe-S] clusters to reach the apo-forms of the two regulators, which were further established by biochemical [Fe-S] cluster transfer tests. A striking difference lies in the dependence of the two regulators towards ATCs: IscR maturation could arise in the total absence of ATCs whereas maturation of NsrR was strictly dependent on the presence of ErpA and IscA or SufA. Both IscR and NsrR regulators were actually found to sense iron limitation, presumably because under such conditions, clusters were not easily formed. A major observation was that in the case of NsrR, this iron-limitation sensing was masked by a very efficient cooperation between the Isc and Suf systems enabling NsrR to exist mostly under its repressing [Fe-S] holo-form. In contrast IscR, which appeared to be a poor substrate for either Isc or Suf occurred mostly under its apo-form under such conditions, hence transduced efficiently the iron limitation signal. Overall this study uncovers unsuspected difference in the way these two functionally and structurally related regulators acquire their clusters.

Results

Assessing the role of the Isc system in NsrR and IscR maturation

In its [Fe-S] bound form, NsrR acts as a transcriptional repressor of the hmpA gene expression (Bodenmiller and Spiro, 2006; Filenko et al., 2007). Hence, a decrease in the NsrR maturation efficiency will translate into an increased derepression of hmpA gene expression. A PhmpA::lacZ fusion was constructed and inserted into the E. coli chromosome at the λ attachment site, and β-galactosidase level from the PhmpA::lacZ fusion should reflect the level of matured NsrR in the cell. As expected, comparison between the wild-type strain and the nsrR mutant showed that NsrR caused a 160-fold repression of the expression of the PhmpA::lacZ fusion (Fig. 1A). A deletion of the suf operon had no effect on the repressing activity of NsrR (Fig. 1A). In contrast, repression of the PhmpA::lacZ fusion was decreased 15-fold in the iscU mutant, indicating that the Isc system participates in the maturation of NsrR (Fig. 1A). However, since some level of repression remained, this result indicated that part of the NsrR protein population remained under its holo-form in the iscU mutant. Assays with hscA, hscB, fdx and iscUA mutants yielded results similar to that reported with the iscU mutant (data not shown). A possibility was that the Suf system was responsible for the remaining amount of holo-NsrR since, in the iscU mutant, the expression of the suf operon is induced fourfold by apo-IscR (data not shown). Thus, we used an iscUA background and tested the effect of overexpressing the suf operon carried on a plasmid (pLAOS). This plasmid yielded an increased level of repression by NsrR (Fig. 1A), indicating that the Suf system can assist NsrR maturation if produced at a level sufficiently high.

Figure 1.

Maturation of NsrR and IscR by the Isc and Suf systems.

A. Strains are all derivatives of DV1301 (containing the PhmpA::lacZ fusion) and carried either mutations or plasmids, as indicated under the graph. DV1301 is referred to as wt. Values are expressed as repression folds. The levels of expression in the wild-type and in the nsrR mutant were 16 and 2260 units, respectively, and this latter was taken as the reference value. Strains were grown in LB medium supplemented with mevalonate (MVA), thiamine and nicotinic acid (LB* medium) at 37°C. Strains transformed with pBAD or pLAOS (sufABCDSE+) were grown in LB* supplemented with ampicillin 100 μg ml−1 and arabinose 0.2%. β-Galactosidase assays were carried out as described (Miller, 1972) using cultures grown in exponential phase (OD600 nm between 0.3 and 0.6) after 1/1000 to 1/2000 dilution of fresh overnight cultures. The erpA, sufA iscA and sufA iscA erpA mutants all carried the MVA cassette. Error bars represent the propagation of standard errors for three biological replicates.

B. Strains are all derivatives of DV1389 (containing the PiscR::lacZ fusion) and carried either mutations or plasmids, as indicated under the graph. DV1389 is referred to as wt. Values are expressed as repression folds. The level of expression found in the wild-type strain and in the iscR mutant were 170 and 1200 units, respectively, and the latter was taken as the reference value. Strains were grown in LB medium supplemented with mevalonate (MVA), thiamine and nicotinic acid (LB* medium) at 37°C. Strains transformed with pBAD or pLAOS (sufABCDSE+) were grown in LB* supplemented with ampicillin 100 μg ml−1 and arabinose 0.2%. β-Galactosidase assays were carried out as described above. Error bars represent the propagation of standard errors for three biological replicates.

The study of the maturation of the IscR regulator was undertaken using a similar strategy. In its [Fe-S] bound form, IscR acts as a transcriptional repressor of the iscRSUA operon. Hence, a decrease in the maturation of IscR will translate into an increased derepression of iscRSUA expression. Therefore, a PiscR::lacZ fusion was constructed, inserted into the chromosome and used as a reporter of IscR maturation. As expected, comparison between the wild-type strain and the iscR mutant showed that IscR caused a sixfold repression of IscR expression (Fig. 1B). Similar to NsrR, deletion of the suf operon had no effect on the repression level of the PiscR::lacZ fusion and mutation in iscU led to derepression of the fusion, indicating that most, if not all, IscR maturation depended on IscU (Fig. 1B). This was consistent with previous results (Schwartz et al., 2001). The Suf system was also found to enable IscR maturation if produced at high level. Hence, in the iscUA background, the pLAOS plasmid led to a restoration in PiscR::lacZ repression (Fig. 1B).

Taken together, these results indicated that both the Isc and the Suf systems have the biochemical properties to use both NsrR and IscR as substrates although under the growth conditions used Isc is the preferred pathway.

ATCs are essential for NsrR maturation

To determine the role of each ATC in maturation of NsrR, we constructed and analysed strains lacking individual or combinations of ErpA, SufA and IscA proteins. First, ErpA was found to be essential for NsrR maturation. Indeed an erpA mutation fully abolished the ability of NsrR to repress PhmpA::lacZ fusion expression, indicating NsrR is in its apo-form in this strain (Fig. 1A). Second, IscA and SufA were found to be functionally redundant for the maturation of NsrR. An iscA mutation caused a decrease in the repression of the PhmpA::lacZ fusion, indicating a large level of holo-NsrR remained (Fig. 1A). The lack of SufA had no effect (data not shown) but combining both iscA and sufA mutations yielded a complete loss in repression, similar to erpA (Fig. 1A). As expected, combining iscA, sufA and erpA mutations also eliminated NsrR maturation (Fig. 1A). Third, ErpA and IscA appear to be part of the same pathway for NsrR maturation since increasing erpA gene dosage in the iscA mutant, via the introduction of the pLAE-A plasmid, failed to restore NsrR-dependent repression of the PhmpA::lacZ fusion (Fig. 2A). Conversely, increased iscA gene dosage did not rescue NsrR repressing activity in an erpA background (Fig. 2B). These observations indicated that the absence of ErpA could not be compensated for by the presence of IscA, and, reciprocally, that IscA could not be replaced by ErpA for the maturation of NsrR. Fourth, additional evidence that IscA and SufA are functionally redundant was obtained from partial restoration of NsrR activity when sufA was overexpressed in an iscA mutant, via the pLAS-A plasmid (Fig. 2A). Surprisingly, an increase in sufA gene dosage appeared to have a suppressing effect, although modest, on the absence of ErpA, leaving open the possibility that SufA alone could fulfil the essential role of ErpA if produced at sufficiently high level (Fig. 2B). Finally, each of the three ATC was overexpressed into the iscA sufA erpA strain (Fig. 2C). The overexpression of IscA had no effect on the repressing capacities of NsrR. This was consistent with the view that IscA has to cooperate with ErpA for a [Fe-S] cluster to be transferred to apo-NsrR. In contrast, the over-production of SufA and, to a lesser extent, that of ErpA allowed a strong repressing activity of NsrR (Fig. 2C), indicating that, if present at high levels, the two ATCs SufA or ErpA could alone carry [Fe-S] cluster to apo-NsrR. Together, these results indicated that ErpA and IscA function in a common maturation pathway whereas SufA can compensate for the loss of either function.

Figure 2.

Assessing the contribution of each ATC to the maturation of NsrR. Strains are all derivatives of DV1301 (containing the PhmpA::lacZ fusion) and carried additional mutations in iscA (A), erpA (B) or sufA iscA erpA (C) genes. Strains were transformed with pBAD (vector), pLAS-A (sufA+), pLAI-A (iscA+) or pLAE-A (erpA+) plasmids, as indicated under the graphs. Values are expressed as repression folds. The level of expression in the nsrR/pBAD mutant was taken as the reference value. The dotted horizontal line represents the wild-type repression fold. Cultures were grown in LB* (see legend Fig. 1) supplemented with ampicillin 100 μg ml−1 and arabinose 0.2%. β-Galactosidase assays were carried out as described in Fig. 1. Error bars represent the propagation of standard errors for three biological replicates.

Lack of specificity in the transfer step occurring between scaffolds and ATCs

The above data suggested that the type of ATC was most important in determining the efficiency of maturation of NsrR. However, these data also raised the question of whether the efficiency of the overall process with a particular ATC was affected by the type of scaffold that partnered with the ATC. Thus, we tested whether IscA or SufA could also obtain, and transfer, cluster in the absence of IscU. For this purpose, the pLAI-A plasmid (iscA+) and the pLAS-A plasmid (sufA+) were introduced into the iscUA mutant (Fig. 3A). The same level of repression of the PhmpA::lacZ fusion than previously observed with iscU mutation alone (Fig. 1) was found (Fig. 3A). This indicated that a [Fe-S] cluster built on IscU will reach apo-NsrR whether it goes via IscA or via SufA. Then, we tested if a [Fe-S] cluster built on the SufB scaffold would go preferably to one of the ATCs. For this, a plasmid, referred to as pLAOHyb, wherein the chimeric operon iscAsufBCDSE was put under the control of an arabinose-inducible promoter, was introduced in a strain devoid of the Isc scaffold (iscUA). This allowed NsrR maturation, as shown by increased repression of the PhmpA::lacZ fusion expression (Fig. 3A). This indicated that in this case, [Fe-S] cluster went from SufBCD to IscA and eventually to apo-NsrR. Then, we tested if SufB could donate its cluster to ErpA. For this, the pLAOa plasmid that carries a sufA-less truncated version of the suf operon, was introduced into the erpA mutant. The pLAOa plasmid failed to restore NsrR maturation to wild-type level, showing that a SufBCD-made [Fe-S] cluster depends on ErpA to reach apo-NsrR (Fig. 3B). Next, we showed that a cluster made by SufB ought to go to either SufA or IscA as pLAOa plasmid also failed to restore NsrR maturation in an iscA sufA mutant (Fig. 3B). Last, multicopy plasmids carrying either iscU or iscUA genes were unable to promote maturation of NsrR in the erpA or in the iscA sufA erpA backgrounds, fully supporting the notion that ATCs are essential for IscU initiated maturation (Fig. S2). Altogether these results showed that there is no difference, which scaffold was initially used to build the [Fe-S] cluster, they must subsequently rely on ATC assistance to reach apo-NsrR.

Figure 3.

Analysis of the transfer of [Fe-S] cluster from the scaffolds to the ATCs in the maturation of NsrR. Strains are all derivatives of DV1301 (containing the PhmpA::lacZ fusion) and carried additional mutations in iscUA (A), sufA iscA (B, left) or erpA (B, right) genes. Strains were transformed with pBAD (vector), pLAOS (sufABCDSE+), pLAS-A (sufA+), pLAI-A (iscA+), pLAHyb (iscAsufBCDSE+) or pLAOa (sufBCDSE+) plasmids, as indicated under the graphs. Results shown are the mean of three independent experiments. Values are expressed as repression folds. The level of expression in the nsrR/pBAD mutant was taken as the reference value. Cultures were grown in LB* supplemented with ampicillin 100 μg ml−1 and arabinose 0.2%. β-Galactosidase assays were carried out as described in Fig. 1. Error bars represent the propagation of standard errors for three biological replicates.

ATCs are dispensable for IscR maturation

The role of each ATC in IscR maturation was also investigated by both mutation and overexpression of iscA, sufA and erpA. IscA and ErpA were found to participate in IscR maturation, as shown by the two- to threefold reduction in IscR repression of PiscR::lacZ expression in both the iscA and the erpA mutants (Fig. 1B). Furthermore, like in the case of NsrR, IscA and ErpA were found to act within the same maturation route, as indicated by the lack of (i) additive effects of combining both iscA and erpA mutations (data not shown), (ii) suppressive effect of iscA mutant when overproducing ErpA and (iii) suppression of erpA when overproducing IscA (Fig. 4A and B). In contrast, unlike NsrR, SufA was found to have no role in IscR maturation. Indeed, overexpression of SufA in the iscA mutant or in the erpA mutant failed to increase IscR repression of the PiscR::lacZ fusion (Fig. 4A and B). Second, combining iscA and sufA mutations had no additive effect (Fig. 1B). These observations differed markedly from what had been found with NsrR, and suggested that SufA was a poor donor for cluster transfer to IscR. Most unexpectedly of all, IscR maturation could occur even in the absence of all three ATCs. Indeed, the iscA sufA erpA mutant exhibited a level of PiscR::lacZ repression that was only threefold reduced as compared with the wild-type (Fig. 1B). The ATC-free background (iscA sufA erpA) was used to test the effect of over-producing each of the ATC. None of the ATCs alone could even when over produced, allowed repression of the PiscR::lacZ fusion, i.e. maturation of IscR (Fig. 4C). Last, the pLAOa plasmid was introduced in an iscUA mutant wherein it led to a drastic increase in IscR-mediated repression of PiscR::lacZ expression showing that SufA is dispensable (Fig. 1B). In fact, the pLAOa plasmid allowed to enhance IscR-mediated repression in the absence of any ATC (in strain iscA sufA erpA) (Fig. 4C). These observations confirmed, first, that [Fe-S] cluster could be assembled on IscR independently of any ATC, i.e. IscA, SufA and ErpA and, second, that the Suf scaffold can transfer [Fe-S] clusters directly to IscR. In contrast, a multicopy plasmid carrying iscU or iscUA genes was unable to promote IscR maturation in the iscA sufA erpA background (Fig. S2). This indicated either that synthesizing extra-copies of IscU in addition to that produced from the chromosomal copy was not sufficient to augment maturation or that a carrier different from the ATCs could intervene between IscU and IscR.

Figure 4.

Assessing the contribution of each ATC in the maturation of IscR. Strains are all derivatives of DV1389 (containing the PiscR::lacZ fusion) and carried additional mutations in iscA (A), erpA (B) or sufA iscA erpA (C) genes. Strains were transformed with pBAD (vector), pLAS-A (sufA+), pLAI-A (iscA+), pLAE-A (erpA+) or pLAOa (sufBCDSE+) plasmids, as indicated under the graphs. Values are expressed as repression folds. The level of expression found in the iscR/pBAD mutant was taken as the reference value. The dotted horizontal line represents the wild-type repression fold. Cultures were grown in LB* supplemented with ampicillin 100 μg ml−1 and arabinose 0.2%. β-Galactosidase assays were carried out as described in Fig. 1. Error bars represent the propagation of standard errors for three biological replicates.

In vitro maturation of NsrR requires an ATC, ErpA or SufA

Our in vivo data indicated that neither IscU nor SufBC2D could function directly in maturation of NsrR. These hypotheses were tested in vitro. We incubated His-tagged apo-NsrR under anaerobic conditions with 1.5 equivalents of untagged holo-IscU (containing 1.5 iron and 1.6 sulphur/monomer), an amount sufficient to mature NsrR. Following Ni-NTA column chromatography, we found no [Fe-S] cluster transfer had occurred since IscU retained 1.3 Fe and 1.4 S per monomer and its UV-visible spectrum (Fig. 5A). Furthermore, the UV-visible spectrum of NsrR that was eluted from the Ni-NTA showed no absorption bands in the 350–600 nm range and contained less than < 0.1 Fe and S per monomer (Fig. 5A). The same result was obtained if HscA and HscB chaperones were added with IscU (data not shown). Also, no transfer was found using holo-SufBC2D complex as a [Fe-S] cluster source (data not shown). Our in vivo data also indicated that SufA and ErpA could function directly in maturation of NsrR. To test this hypothesis in vitro, we first incubated His-tagged apo-NsrR under anaerobic conditions with 2.5 equivalents of untagged holo-SufA (containing 1.1 Fe and 1.1 S/monomer). Following Ni-NTA column chromatography, we observed [Fe-S] cluster transfer had occurred since NsrR contained 1.4 Fe and 1.2 S/monomer and displayed light absorption at 420 nm whereas SufA had lost the majority of its [Fe-S] cluster (0.4 Fe/monomer) and was almost colourless (Fig. 5B). In a second experiment apo-NsrR was incubated anaerobically with 1.5 equivalents of holoErpA (containing 1.7 Fe and 1.6 S/monomer). After the affinity chromatography, we observed that [Fe-S] cluster transfer had occurred since NsrR contained 1.7 Fe and 1.6 S/monomer and displayed light absorption at 420 nm whereas ErpA had completely lost its [Fe-S] cluster (0.1 Fe/monomer) and was colourless (Fig. 5C). Finally, no transfer was observed using holo-IscA as a [Fe-S] cluster source (Fig. 5D). Results from these in vitro analyses fully support the notion that scaffold proteins transfer [Fe-S] cluster to NsrR via ErpA or SufA ATCs.

Figure 5.

Study of [Fe-S] cluster transfer to NsrR in vitro.

A. Apo-NsrR (20 nmoles), was mixed with holo-IscU (30 nmoles, 1.5 Fe and 1.6 S/monomer) in 0.1 M Tris-HCl pH 8, 50 mM KCl, 5 mM DTT. After separation on the Ni-NTA column, the UV-Vis spectra of the wash fraction containing IscU (solid line) and the eluate fraction containing NsrR (dashed line) were recorded.

B. For transfer from holo-SufA to apo-NsrR, apo-NsrR (25 nmoles) was mixed in Tris-HCl pH 8, KCl 50 mM, 5 mM DTT together with the holo-SufA (62 nmoles, 1.1 Fe and S/monomer) in such a way to provide sufficient amounts of Fe and S to NsrR. After separation on the Ni-NTA column, the UV-Vis spectra of the wash fraction containing SufA (solid line) and the eluate fraction containing NsrR (dashed line) were recorded.

C. For transfer from holo-ErpA to apo-NsrR, apo-NsrR (25 nmoles) was mixed in Tris-HCl pH 8, KCl 50 mM, 5 mM DTT together with the holo-ErpA (38 nmoles, 1.8 Fe and 1.7 S/monomer) in such a way to provide sufficient amounts of Fe and S to NsrR. After separation on the Ni-NTA column, the UV-Vis spectra of the wash fraction containing ErpA (solid line) and the eluate fraction containing NsrR (dashed line) were recorded.

D. For transfer from holo-IscA to apo-NsrR, apo-NsrR (25 nmoles) was mixed in Tris-HCl pH 8, KCl 50 mM, 5 mM DTT together with the holo-IscA (42 nmoles, 1.6 Fe and 1.5 S/monomer) in such a way to provide sufficient amounts of Fe and S to NsrR. After separation on the Ni-NTA column, the UV-Vis spectra of the wash fraction containing IscA (solid line) and the eluate fraction containing NsrR (dashed line) were recorded.

In vitro maturation of IscR can occur directly from a scaffold

Our genetic data indicate that either the IscU or the SufBC2D scaffolds could function directly in IscR maturation. To test this hypothesis in vitro, we incubated His-tagged apo-IscR with 1.3 equivalents of untagged holo-SufBC2D (containing 3 Fe and 4 S per complex). Following Ni-NTA chromatography, we found [Fe-S] cluster transfer had occurred since SufBC2D retained only 0.16 Fe and 0.18 S/complex and lost its characteristic UV-visible spectrum (Fig. 6A). In contrast, IscR, recovered from the chromatography step contained 0.97 Fe and 1.1 S/monomer and displayed a UV-visible spectrum characteristic of a [Fe-S] protein (Fig. 6A). The same result was obtained using holo-IscU as [Fe-S] source (Fig. 6B). This clearly indicated that in vitro, like in vivo, apo-IscR can get its cluster directly from either one of the two Isc or Suf scaffolds. A surprising observation in vivo was that SufA was dispensable for IscR maturation. To test this hypothesis in vitro, apo-IscR was incubated with holo-SufA (1.2 molar excess with regard to IscR and containing 1.1 Fe and 1.1 S/monomer). Upon separation by Ni-NTA chromatography, we observed no [Fe-S] cluster transfer since IscR contained 0.1 iron/monomer, and SufA retained 0.6 Fe and 0.7 S/monomer and still displayed a UV-visible spectrum characteristic of an [Fe-S] containing protein (Fig. 6C). This fully confirmed that IscR cannot obtain a cluster from SufA.

Figure 6.

Study of [Fe-S] cluster transfer to IscR in vitro.

A. The apo-IscR (25 nmoles), was mixed with holo-SufBCD (33 nmoles, 3 Fe and 4 S/complex) in 0.1 M Tris-HCl pH 8, 50 mM KCl, 5 mM DTT. After separation on the Ni-NTA column, the UV-Vis spectra of the wash fraction containing SufBCD (solid line) and the eluate fraction containing IscR (dashed line) were recorded.

B. Apo-IscR (25 nmoles) pre-reduced with DTT, was mixed with holo-IscU (50 nmoles, 1.4 Fe and 1.3 S/monomer) in 0.1 M Tris-HCl pH 8, 50 mM KCl. After separation on the Ni-NTA column, the UV-Vis spectra of the wash fraction containing IscU (solid line) and the eluate fraction containing NsrR (dashed line) were recorded.

C. For transfer from holo-SufA to apo-IscR, apo-IscR (25 nmoles) was mixed in Tris-HCl pH 8, KCl 50 mM, 5 mM DTT together with the holo-SufA (30 nmoles, 1.1 Fe and S/monomer). After separation on the Ni-NTA column, the UV-Vis spectra of the wash fraction containing SufA (solid line) and the eluate fraction containing IscR (dashed line) were recorded.

Analysis of NsrR maturation under iron starvation

Iron limitation is a common environmental stress for bacteria. Thus, understanding how cells carry out [Fe-S] cluster biogenesis under such conditions is of crucial importance. To address this question we tested the contribution of the Isc and Suf pathway components to the maturation of NsrR under iron-limiting conditions. In the presence of 2,2′-dipyridyl (DIP), an iron chelator, PhmpA::lacZ expression was partially derepressed (Fig. 7A and B), indicating that NsrR continues to obtain [Fe-S] clusters during iron limitation. Furthermore, NsrR [Fe-S] cluster synthesis during this period depends primarily on SufA, since the PhmpA::lacZ fusion was greatly derepressed in a sufA mutant indicating a major defect in NsrR maturation (Fig. 7A). This is in contrast to normal growth conditions where the absence of SufA had no detectable consequence (Fig. 1A). It may be explained by the loss of Fur-dependent repression of the suf operon and of the ryhB small RNA mediated decrease in iscSUA mRNA stability. In fact, the induction observed in the sufA mutant was suppressed by the ryhB mutation (Fig. 7A) indicating that when stability of the iscSUA mRNA is restored, the Isc system could function in the maturation of NsrR under iron-limiting conditions. Moreover, the pLAOS plasmid, which overexpresses the entire Suf system, restored repression by NsrR of PhmpA::lacZ in the presence of DIP but not the plasmid pLAOa lacking sufA (Fig. 7B and C). These results showed that the Suf system permits the maturation of NsrR under iron-limiting conditions, and confirmed the key role of SufA under these conditions. Nevertheless, these data clearly indicated that the regulatory wiring of the [Fe-S] biogenesis systems is designed to favour the Suf system for the NsrR protein to retain its cluster under iron limitation.

Figure 7.

Effect of iron limitation on NsrR maturation.

A. Overnight cultures of wild-type strain and its sufA, ryhB or sufA ryhB derivatives carrying the PhmpA::lacZ fusion were diluted 1/2000 in LB* and DIP (0.25 mM) was added when the OD600 of the cultures reached about 0.05–0.1. β-Galactosidase assays were then carried out at different time points as indicated. β-Galactosidase values recorded are indicated.

B. Overnight cultures of sufA PhmpA::lacZ strain carrying the plasmids pBAD (solid squares), pLAOS (sufABCDSE+, triangles) or pLAOa (sufBCDSE+, circles) were diluted in LB* supplemented with ampicillin 100 μg ml−1 and arabinose 0.2%. DIP (0.25 mM) was added when the culture reached OD600 about 0.01. β-Galactosidase assays were then carried out at different time points as indicated. β-Galactosidase values recorded are indicated.

C. Same as in (B) except that DIP was added when the culture OD600 was about 0.01. The experiment was run at least twice and results of one typical experiment are shown here. β-Galactosidase values recorded are indicated.

Study of IscR maturation under iron starvation

Maturation of IscR in the presence of DIP was also analysed. Derepression occurred within 10 min of adding DIP, as shown by the increase in PiscR::lacZ expression, suggesting that IscR maturation was more sensitive to Fe limitation than NsrR (Fig. 8A). The expression remained constant for about 15 min, presumably as a result of a new equilibrium between the amounts of holo- and apo-IscR (Fig. 8A). This equilibrium appeared to change again after 25 min where derepression resumed (Fig. 8A). In a separate experiment, cells at very low density were submitted to DIP for an increased period of time such as we could observe later time points (Fig. 8B). Expression of the fusion remained constant from 120 min to 190 min post DIP addition and then again increased (Fig. 8B), suggesting the equilibrium between apo- and holo-IscR continued to change during the adaptation to iron limitation, with the apo-form being eventually predominant. Overexpressing the whole Suf system, via the pLAOS plasmid, led to a drastic increase in repression in both set of conditions (Fig. 8A and B). This implied that the Suf system favoured formation of the holo-form of IscR, and that after 10 min the concentration of holo-IscR available was sufficient to repress PiscR::lacZ. This indicated that the Suf system can mature IscR under iron limitation, if Suf is overproduced. The overproduction of the Suf system derivative lacking the SufA protein, via the pLAOa plasmid, also led to steady repression (Fig. 8B), in full agreement with the hypothesis of dispensability of the SufA protein for IscR maturation. In contrast to what has been observed with NsrR, a ryhB mutation had no effect on its own (data not shown). This suggests that under iron limitation, the Isc system is not sufficiently efficient for IscR maturation. Overall these results indicated the Isc and Suf systems do not seem to be efficient enough for maintaining a high level of holo-IscR form and as a consequence in vivo, IscR can sense iron limitation.

Figure 8.

Effect of iron limitation on the IscR maturation.

A. Overnight cultures of sufA PiscR::lacZ strain carrying the plasmids pBAD (solid squares), pLAOS (sufABCDSE+, triangles) or pLAOa (sufBCDSE+, circles) were diluted in LB* supplemented with ampicillin 100 μg ml−1 and arabinose 0.2%. DIP (0.25 mM) was added when the culture reached OD600 about 0.05–0.1. β-Galactosidase assays were then carried out at different time points as indicated. The experiment was run at least twice and results of one typical experiment are shown here. β-Galactosidase values recorded are indicated.

B. Same as in (A), except that DIP was added when the cultures reached OD600 about 0.01.

Discussion

Escherichia coli is predicted to synthesize about 165 [Fe-S] cluster containing proteins (Fontecave, 2006; Py and Barras, 2010). Biochemical and structural analyses have provided insight into the mechanism allowing [Fe-S] clusters formation in vitro but it remains an open question whether the in vitro studies recapitulate the in vivo diversity and complexity needed for the large number of target proteins. The present study characterizes in detail how two stress-transcriptional regulators, IscR and NsrR, that exhibit 50% sequence similarity (Fig. S1) acquire their clusters. By defining maturation maps for NsrR and IscR, we found common features in the use of scaffolds but significant differences in the use of the ATCs. Our previous study on the maturation of IspG/H showed that these [4Fe-4S] clusters proteins can acquire clusters from different ‘routes’, providing us with the first evidence for functional redundancy (Vinella et al., 2009). Routes were defined by the identity of those biogenesis factors acting together, i.e. in the same pathway, to build and deliver a cluster to a given apo-protein. A limitation in this study was that maturation of IspG/H, which allows the biosynthesis of isopentenyl phosphate (IPP), is essential for E. coli survival. As a consequence, the alteration of a given [Fe-S] biogenesis factor, which would have yielded a decreased maturation of holo-IspG/H, yet keeping the level of IPP above a threshold needed for viability, would not have been detected as involved in IspG/H maturation. Therefore, in this study, we chose to analyse two [Fe-S] proteins, IscR and NsrR, the maturation of which could be monitored in a quantitative manner, regardless of the genetic background studied and/or the growth conditions used. Maturation of the [Fe-S] cluster homeostasis controlling regulator IscR was monitored by analysing the expression of the iscRSUA operon, which is repressed by the [Fe-S] cluster containing form of IscR. To date, this is the only regulator known to act at the iscR promoter level. Maturation of the NO-sensing transcriptional regulator NsrR was monitored, by studying the expression of the hmpA gene. Besides NsrR, several other regulators, i.e. Fur, Fnr (Poole et al., 1996; Cruz-Ramos et al., 2002; D'Autreaux et al., 2002; Yohannes et al., 2004; Flatley et al., 2005; Hernandez-Urzua et al., 2007) and possibly MetR and RpoS (Membrillo-Hernandez et al., 1997; 1998), have been proposed to act at the PhmpA promoter level. However, neither the MetE nor the RpoS binding sites were found to have an effect on the PhmpA::lacZ fusion activity in strains and conditions we used (data not shown). Moreover, we observed only a twofold induction of the PhmpA::lacZ fusion in the fnr fur and fnr fur rpoS mutants (data not shown), which we considered as negligible when compared with the induction ratios found in inactivating either nsrR or any of the isc genes studied here. Therefore, we considered the PhmpA::lacZ fusion activity as a reliable read-out of NsrR maturation level.

Maturation maps for both IscR and NsrR, as deduced from results of this study, are depicted in Fig. 9. A first conclusion is that [Fe-S] clusters can take different routes to reach each of the two regulators. This is a conclusion we reached previously in studying the maturation of IspG/H where Isc and Suf were predicted to intervene in this order with increasingly unfavourable conditions for [Fe-S] biogenesis (Vinella et al., 2009). A major difference between IscR and NsrR lies in their dependence on ATCs for acquiring their [Fe-S] clusters. Maturation of NsrR was found to be strictly dependent on the presence of a functional ErpA and at least one of the two other ATCs, SufA and IscA. In contrast, IscR maturation could occur in the total absence of ATCs, i.e. [Fe-S] clusters could go directly from scaffolds to apo-IscR. We do not yet understand the significance of the scaffold-only maturation route of IscR. It might relate to the central role of IscR in [Fe-S] cluster homeostasis. Hence, this ATC-independent pathway might allow IscR to sense the ability of the cells to make [Fe-S] cluster by directly assessing the presence of the [Fe-S] clusters at their source, i.e. on the scaffolds. However, we cannot at this point rule out the possibility that one of the other carriers, such as NfuA, Grx or Mrp, might take over the transfer of the [Fe-S] cluster from IscU to IscR. This will have to be tested in specifically designed backgrounds. Moreover, we found that the ATCs, IscA and ErpA, also contribute to IscR maturation. This is consistent with a previous study from the Outten's lab reporting on the effect of an iscA mutation on IscR-dependent regulation of biofilm formation (Wu and Outten, 2009). Last, quite unexpected was the difference in SufA utilization between IscR and NsrR. Both in vivo and in vitro studies demonstrated that SufA was unable to transfer [Fe-S] cluster to IscR whereas it was capable of doing so to NsrR.

Figure 9.

The in vivo maturation maps of NsrR and IscR. Arrows indicate Fe-S transfer paths. A solid arrow was put when all collected evidences pointed to the occurrence of Fe-S transfer. Occurrence of these transfer paths was deduced from either the inactivation or the overexpression of genes encoding the [Fe-S] biogenesis factors shown. Transfer paths deduced from the former are predicted to arise under normal growth conditions while those deduced from the second are predicted to arise under stress conditions. In a few cases, a broken arrow was drawn as to indicate that uncertainty remains about the physiological significance of the results.

A. The maturation map of NsrR. Cluster transfer (C.T.) between ErpA and NsrR was deduced from the lack of maturation of NsrR in the erpA mutant and from the fact that the overproduction of erpA restored NsrR maturation in the triple iscA erpA sufA mutant. C.T. between IscU and SufA was deduced from the lack of complementation of the iscUA mutant by the pLAS-A (sufA+) plasmid, whereas this later suppressed the iscA mutation. C.T. between SufA and NsrR was deduced from the complementation of the triple sufA iscA erpA mutant by the pLAS-A (sufA+) plasmid. C.T. between SufA and ErpA, and between IscA and ErpA were deduced from the functional redundancy of the iscA and sufA genes. C.T. between SufB and IscA was deduced from the following observations: (i) complementation of the iscUA mutant by the pLAHyb (iscAsufBCDSE+) plasmid, (ii) lack of complementation of iscUA by the pLAI-A (iscA+) plasmid and (iii) lack of complementation of the iscA sufA mutant by the pLAOa (sufBCDSE+) plasmid.

B. The IscR maturation map. C.T. between SufB and IscR is deduced from the complementation of the iscA sufA erpA mutant by the pLAOa (sufBCDSE+) plasmid. C.T. between IscU and IscR is uncertain: the fact that there is no maturation of IscR in the iscU mutant demonstrates the involvement of IscU. However, whereas IscR maturation occurs in the iscA sufA erpA mutant, which could be interpreted as reflecting a direct transfer from IscU to IscR, multicopy plasmid carrying iscU or iscUA (Fig. S2) failed to suppress the inducing effect of the iscA sufA erpA genotype. C.T. between IscA and IscR or between ErpA and IscR remain defined more precisely as the results point to the requirement of both IscA and ErpA, i.e. no multicopy suppression of iscA by erpA or of erpA by iscA. Therefore, IscA and ErpA are depicted as acting together.

Studying maturation of IscR and NsrR under iron-limiting conditions illustrated how E. coli makes use of both the Isc and the Suf systems in stress conditions. In the wild-type strain, NsrR maturation appeared unaffected by iron limitation. In fact, NsrR maturation was sensitive to it as indicated by the fact that hmpA gene expression was partially derepressed in the wild-type strain in the presence of DIP as this had already been reported (Poole et al., 1996; Bodenmiller and Spiro, 2006). However, the expression level reached was much lower than in a sufA mutant. This indicated that NsrR remained mostly under its repressing holo-form, thanks to the action of the Isc system for 30 min after adding DIP, and subsequently that of the Suf system. Interestingly, during iron limitation, the shift from the Isc to the Suf system appeared to involve the small non-coding RNA RyhB, as its inactivation acted as a suppressor of a sufA mutation. Presumably, inactivating RyhB allowed synthesis of Isc to continue and compensate for the lack of sufA (Desnoyers et al., 2009). The situation with IscR was almost opposite to that described for NsrR as the apo-IscR/holo-IscR ratio kept changing throughout the period cells were exposed to DIP and eventually apo-IscR was the predominant form. This suggested that IscR is not a good substrate for either one of the two systems under such stressful conditions. As a matter of fact, that apo-IscR is maturated quite inefficiently by the Isc system is part of the underlying autoregulatory principle controlling synthesis of the Isc system (Schwartz et al., 2001; Giel et al., 2006; Mettert et al., 2008). Such a homeostatic loop is proposed to allow E. coli to carry out maturation of all cellular proteins before maturing IscR, hence ensuring sufficient expression of the Isc system. Similarly, the hypothesis that IscR is a poor substrate of the Suf system also makes physiological sense. Since apo-IscR activates the Suf system, maturating IscR under iron limitation would lower expression of the suf operon and the [Fe-S] clusters biogenesis capacity of E. coli would be lessened. Future studies will test whether the poor maturation of IscR under such conditions is due to the inability of SufA to mature IscR.

NsrR and IscR maturation maps differ markedly from that proposed for the [4Fe-4S] Fnr transcriptional regulator, whose maturation was found to be dependent on the Isc system under both aerobiosis and anaerobiosis, and on the Suf system under anaerobiosis only (Mettert et al., 2008). These different maturation maps illustrate the differences in the interactions between targets and the [Fe-S] biosynthesis systems. This is consistent with the view of Kiley and collaborators who argued that the reason Fnr is not a substrate for Suf under aerobiosis reflects its poorer affinity as compared with that of other apo-targets (Mettert et al., 2008). Here we show that the reason NsrR does not relay the ‘iron limitation’ signal is because both Isc and Suf systems are highly efficient in keeping the level of the holo-NsrR form high even under iron limitation. Hence, we propose that, at the cellular level, [Fe-S] biosynthesis systems exert a decisive influence on which signal a cluster containing regulator will be responsive to. Considering that Fnr, IscR and NsrR control the expression of hundred of genes, this unsuspected integrated level of regulation could have a broad implication on genomic expression.

The two membrane bound hydrogenases, the Nar nitrate reductase, the Fdh formate dehydrogenases, the aconitase B, the thiamin biosynthesis pathway belonging ThiC, and the endonuclease III, all of these [4Fe-4S] proteins were found to require ATCs for their maturation (Mettert et al., 2008; Tan et al., 2009; Pinske and Sawers, 2012a,b). In contrast, under aerobic conditions, maturation of the redox transcription factor SoxR, ferredoxin, and the siderophore-iron reductase FhuF, which are all [2Fe-2S] proteins, was proposed to be independent of at least IscA and SufA (Tan et al., 2009). Hence, an attractive model is that in vivo scaffolds can transfer directly clusters to [2Fe-2S] apo-targets and that ATCs are required for maturation of [4Fe-4S] proteins only.

Experimental procedures

Media and growth conditions

The rich medium used in this work was LB broth (Miller, 1972). Glucose (0.2%), arabinose (0.2%), casaminoacids (0.2%), amino acids (0.005%), thiamine (50 μg ml−1), nicotinic acid (12.5 μg ml−1) and mevalonate (1 mM) were added when required. Solid media contained 1.5% agar. Antibiotics were used at the following concentrations: chloramphenicol (Cm) 25 μg ml−1, kanamycin (Km) 25 μg ml−1, tetracycline (Tc) 25 μg ml−1, spectinomycin (Spc) 50 μg ml−1 and ampicillin (Amp) 50 μg ml−1.

Bacterial strains, phages and plasmids

All the strains used in this work are E. coli K-12 derivatives; the principal strains are described in Table S1. All the donor strains for kanR gene deletions, kindly provided by P. Moreau (LCB, Marseille) were from the Keio collection (Baba et al., 2006). The ΔiscA::cat, ΔiscUA::cat, ΔiscS::cat, Δsuf::cat and ΔerpA::cat mutations were constructed as described (Datsenko and Wanner, 2000; Trotter et al., 2009; Vinella et al., 2009). The donor of the MVA+ KanR cassette was strain DV1093 (Vinella et al., 2009). The so-called MVA cassette was used as described in previous papers (Loiseau et al., 2007; Vinella et al., 2009). Briefly, the MVA cassette contains isoprenoid synthesizing eukaryotic genes, which are under the control of the arabinose-inducible promoter (Campos et al., 2001). All mutations were introduced into strains by P1 vir transduction (Miller, 1972), selecting for the appropriate antibiotic resistance. The antibiotic resistance cassettes were eliminated when needed using plasmid pCP20 as described (Cherepanov and Wackernagel, 1995). The construction of all the conditional-lethal mevalonate-dependent strains was carried out in anaerobiosis with mevalonate in the plates to limit the occurrence of suppressor mutations (Vinella et al., 2009).

To construct the transcriptional PhmpA::lacZ fusion, PCR amplification of chromosomal DNA from the MG1655 strain was carried out using primers hmpUP and hmpDO (Table S2). The PCR product was digested by EcoRI and BamHI and cloned into pRS415 (Simons et al., 1987). Note that we failed to observe differences in the behaviour of the fusion above described and a fusion constructed with hmpUP1 (absence of the upstream MetR binding site) except a lower activity in all the experiments (data not shown). The fusions were introduced into the bacterial chromosome of strain DV206, a MG1655 ΔlacZ derivative (Vinella et al., 2000) using the RS45 phage and we verified that strain DV1301 was mono-lysogen as described (Powell et al., 1994). β-Galactosidase assays were carried out as described (Miller, 1972). erpA, sufA iscA and erpA sufA iscA mutants all carried the MVA cassette which also contains a lacZ gene. Thus, we substracted the β-galactosidase measured in the corresponding derivatives of a ΔlacZ MVA strain (with no transcriptional fusion) to obtain values shown in all figured. Assays were performed at least three independent times. Standard errors for data expressed as ‘Fold Repression’ were calculated using a propagation of standard error formula (Ku, 1966).

To construct the pLAOa plasmid, we first constructed the pLUS-A, a pUC derivative containing the sufA gene. PCR product obtained by using MG1655 chromosomal DNA with primers EcoRI–sufA/XhoI-sufA (Table S2) was digested with EcoRI/XhoI and inserted into pUC18 restricted by EcoRI/SalI. We next subcloned the HindIII/HindIII insert of pGSO164 (Outten et al., 2003) containing the 3′ end of sufA and the rest of the suf operon into pLUS-A deleted of the HindIII/HindIII insert containing the 3′ end of sufA, giving the pX plasmid. In parallel, we cloned the wild-type sufB gene into EcoRI/SalI digested vector pBAD24 by ligation of a PCR product from MG1655 using primers 5SufB/3SufB (Table S2) and digested by EcoRI/XhoI. The resulting p(sufB) plasmid was then digested by BstXI (cutting inside sufB) and HindIII and ligated with a BstXI/HindIII insert containing part of sufB and the rest of the suf operon obtained by digestion of the pX plasmid. To construct the pLAOHyb plasmid, we cloned the HindIII–HindIII fragment containing the sufBCDSE genes obtained from the pX plasmid into the pLAI-A plasmid (Vinella et al., 2009) opened with HindIII. Note that the resulting pLAOHyb plasmid carries a mutation causing a H106G change in IscA that is neutral.

Protein purification

In order to overproduce IscR and NsrR in E. coli, plasmids pET::iscR and pET::nsrR were constructed as follows. The iscR and nsrR coding regions were amplified by PCR from E. coli MG1655 genomic DNA using the primers NdeI-iscR/XhoI-iscR and NdeI-nsrR/XhoI-nsrR respectively. The resulting PCR products were digested with NdeI and XhoI, and the released iscR and nsrR fragments were inserted into the pET22b+ overexpression vector (Novagen) digested with the same enzymes. E. coli strains BL21(DE3)/pETnsrR and BL21(DE3)/pETiscR were grown in LB medium containing 100 μg ml−1 ampicillin at 37°C. Protein expression was induced for 3h30 by the addition of 0.5 mM IPTG to an exponentially growing culture. The bacterial pellets were resuspended in buffer A (50 mM Tris-HCl pH 8, 0.1 M NaCl, 1 mM PMSF) and sonicated before ultracentrifugation (90 min, 45000 r.p.m., 4°C). Supernatants were treated with streptomycin sulphate (2%) at 4°C for 45 min and, after centrifugation (30 min, 10000 r.p.m., 4°C), soluble proteins were loaded onto an Ni-NTA column equilibrated with 50 mM Tris-HCl pH 8, 0.1 M NaCl. After an extensive washing with the same buffer, NsrR and IscR were eluted with 50 mM Tris-HCl pH 8, 0.1 M NaCl containing 400 mM imidazole. Pure proteins were immediately desalted onto NAP-25 columns.

Apo-SufA, IscU and SufBC2D were obtained according to the procedure previously described (Agar et al., 2000; Layer et al., 2007; Sendra et al., 2007). Holo-forms of these proteins were obtained as already described (Agar et al., 2000; Sendra et al., 2007; Wollers et al., 2010). Protein concentrations were measured by the method of Bradford using bovine serum albumin as a standard which, in the case of SufBC2D, overestimates the concentration by a factor of 1.12. Iron and sulphide were, respectively, assayed by the methods of Fish (Fischer, 1967) and Beinert (Beinert, 1983).

In vitro [Fe-S] cluster transfer

All the following procedures were performed anaerobically in the glove box at 18°C. Holo-protein (SufBC2D, IscU, SufA, ErpA or IscA) were mixed in buffer 0.1 M Tris-HCl, 50 mM NaCl, 1 mM DTT pH:8 to an apo-target containing an His-Tag (NsrR or IscR) in a molar excess varying from 1.2 to 2.5 depending the experiments. After 1 h incubation, DTT was removed from the solution on a Nap-10 column equilibrated with buffer 0.1 M Tris-HCl, 50 mM NaCl before separation of the proteins on a Ni-NTA affinity column (1 ml) equilibrated with buffer 0.1 M Tris-HCl, 50 mM NaCl. [Fe-S] donor proteins (SufBC2D, IscU, SufA, ErpA or IscA) were recovered in the flow-through and wash fractions, as they did not contain a polyhistidine-tag. NsrR and IscR were eluted with buffer 0.1 M Tris-HCl, 50 mM NaCl containing 0.2 M imidazole. The flow-through and elution fractions were analysed both by UV-visible absorption spectroscopy and for their metal content.

Acknowledgements

We thank all members of the Barras group for fruitful discussions. We also thank Dr P. Moreau (LCB, Marseille) for handling the KEIO collection. We are indebted to Dr Erin L. Mettert and Prof. Tricia Kiley (U. Wisconsin, Madison) for sharing unpublished data and careful editing of the manuscript. This work was supported by grants from the CNRS, the ANR (Blanc SPV05511), the Institut Universitaire de France (IUF), the CEA, Aix-Marseille Université and the Université Joseph Fourier at Grenoble.

Ancillary