The ErpA/NfuA complex builds an oxidation-resistant Fe-S cluster delivery pathway

Fe-S cluster–containing proteins occur in most organisms, wherein they assist in myriad processes from metabolism to DNA repair via gene expression and bioenergetic processes. Here, we used both in vitro and in vivo methods to investigate the capacity of the four Fe-S carriers, NfuA, SufA, ErpA, and IscA, to fulfill their targeting role under oxidative stress. Likewise, Fe-S clusters exhibited varying half-lives, depending on the carriers they were bound to; an NfuA-bound Fe-S cluster was more stable (t½ = 100 min) than those bound to SufA (t½ = 55 min), ErpA (t½ = 54 min), or IscA (t½ = 45 min). Surprisingly, the presence of NfuA further enhanced stability of the ErpA-bound cluster to t½ = 90 min. Using genetic and plasmon surface resonance analyses, we showed that NfuA and ErpA interacted directly with client proteins, whereas IscA or SufA did not. Moreover, NfuA and ErpA interacted with one another. Given all of these observations, we propose an architecture of the Fe-S delivery network in which ErpA is the last factor that delivers cluster directly to most if not all client proteins. NfuA is proposed to assist ErpA under severely unfavorable conditions. A comparison with the strategy employed in yeast and eukaryotes is discussed.

Fe-S clusters rank among the oldest cofactors, and they are surmised to have been key factors for life origin and evolution (1,2). The first Fe-S proteins identified, by their electron paramagnetic signature, were those of mitochondrial respiratory complexes (3). In the following decades, biophysical and structural approaches revealed the great variety of Fe-S cluster proteins, whereas genomic analyses showed their widespread adoption in living organisms (4). Nowadays, in organisms, Fe-S cluster proteins participate in respiration, biosynthesis of cell building blocks and cofactors, central metabolism, gene regulation, tRNA modification, and DNA synthesis and repair (5).
In vitro studies revealed that, in favorable conditions, Fe-S clusters can form spontaneously on apoprotein clients (6,7). However, in the 1990s, studies initiated on nitrogenase maturation revealed that in vivo Fe-S cluster assembly requires multiprotein systems: the NIF system dedicated to nitrogenase maturation and two general systems, ISC and SUF, which maturate most, if not all, cellular Fe-S proteins (8 -10). Components of the ISC and SUF systems are conserved throughout eukaryotes and prokaryotes. In eukaryotes, the ISC and SUF Fe-S machineries are located in mitochondria and chloroplasts, respectively (11,12).
Fe-S cluster biogenesis proceeds in two steps, assembly and delivery. The former takes place on a scaffold protein, which allows sulfur and iron to meet and combine in an Fe-S cluster. Sulfur is produced via the catalytic degradation of L-cysteine by cysteine desulfurase, whereas the molecular source of iron remains elusive. The delivery relies on a series of Fe-S carriers that transport scaffold-bound clusters to the apo-clients. Numerous Fe-S carriers have been identified in both prokaryotes and eukaryotes. These include so-called A-type carriers (ATC), 3 named IscA, SufA, and ErpA in prokaryotes and ISA1 and ISA2 in eukaryotes (13)(14)(15)(16)(17). Other carriers include the P-loop NTPases, (Ind1 in mitochondria, ApbC in Salmonella) (18 -21), the monothiol glutaredoxins (Grx 5 in yeast and GrxD in E. coli) (22)(23)(24), and the highly conserved NFU-type proteins that have been shown to interact with target proteins (25)(26)(27)(28)(29)(30)(31)(32)(33).
Understanding how Fe-S clusters are delivered to the large number of functionally and structurally diverse Fe-S cluster proteins is a challenging task. Indeed, in most organisms, clients are actually a set of structurally and functionally diverse protein species, and it is difficult to think of a common maturation pathway shared by all of them. Moreover, Fe-S clusters are highly sensitive to environmental changes and, in particular, can be destabilized by reactive oxygen species (ROS), depending upon their solvent accessibility on the host protein (34 -36). Last, Fe-S clusters arise under different structure (i.e. 2Fe-2S, 3Fe-4S, or 4Fe-4S) or in more complex association with other metal and cofactors (1). How organisms keep on inserting clusters in such a great diversity of substrates and conditions and whether all Fe-S proteins get their clusters delivered by the same set of factors is a daunting issue. Multiple parameters can be foreseen as controlling the delivery process: (i) genetically controlled level of delivery factors available in a given condition, (ii) affinity between delivery factors and apo-targets, and (iii) intrinsic biochemical features of delivery factors. Our previous study had illustrated how genetic regulation orchestrated the choice of routes Fe-S clusters take to reach essential enzymes, IspG/H, and the two transcriptional factors, IscR and NsrR, throughout fluctuating conditions (16,37). The present study provides key findings on the two other parameters and provides us with an unprecedented view of Fe-S cluster trafficking and delivery.
We investigated the intrinsic capacity of IscA, SufA, ErpA, and NfuA to stabilize bound Fe-S cluster exposed to aerobic conditions. We proceeded by analyzing the contribution of each of them to the maturation of a series of Fe-S enzymes, including IspG and IspH, two essential Fe-S enzymes involved in the production of isoprenoids (38,39). This provided us for the first time with a direct ranking of the Fe-S cluster carriers according to their intrinsic capacity to resist ROS potential damages. Another key finding was the observation of a privileged partnership between ErpA and NfuA, providing the cell with a new type of "hybrid" carrier, whose Fe-S cluster exhibited the highest level of resistance to ROS. Altogether, this allowed us to propose a model of the Fe-S cluster delivery network that predicts ErpA to be in charge of the maturation of most if not all Fe-S enzymes. Its closer association with NfuA appears as a new strategy for the cell to meet with transiently occurring destabilizing conditions.

NfuA is required for IspG/H maturation under oxidative stress
Earlier, we found that growth of the E. coli ⌬nfuA mutant on rich medium is severely affected in the presence of paraquat, a superoxide radical generator (25). Our previous analysis informed us that such a phenotype is probably due to an insufficient amount of isopentenylphosphate (IPP), due to a poor maturation of IspG and IspH, two Fe-S cluster-containing enzymes (17). Therefore, we tested whether the NfuA protein contributes to the Fe-S cluster delivery route toward maturation of IspG/H. For this, we introduced the eukaryotic IPP biosynthesis (Fe-S)-independent pathway, referred to as the mevalonate (MVA) pathway, in the ⌬nfuA mutant and found that ectopic expression of this pathway was sufficient to rescue viability of the nfuA mutant (Fig. 1). This result indicated that NfuA is required for Fe-S cluster delivery to IspG/H under oxidative stress.

NfuA interacts with ErpA
To know whether NfuA cooperates with the ATC, we tested the interaction of NfuA with the ATC using the bacterial twohybrid system. NfuA/ErpA interaction was observed as the BTH101 cells synthesizing the T18-NfuA and T25-ErpA hybrid proteins (pT18-NfuA and pT25-ErpA) exhibited ␤-gal activity ( Fig. 2A). In contrast, no interaction was indicated between NfuA and IscA or between NfuA and SufA using this assay ( Fig. 2A). ErpA/NfuA interaction was also tested by surface plasmon resonance (SPR) using apoproteins. Purified NfuA was immobilized onto a Biacore sensor CM5 chip, and ErpA was serially diluted and injected. The calculated dissociation constant (K d ) of the ErpA/NfuA interaction was determined to be 48.5 Ϯ 0.5 M (Fig. 2B). In contrast, no interaction was observed between NfuA and SufA or IscA (data not shown). Taken together, these analyses revealed that NfuA and ErpA physically interact with each other and suggest that they might partner to form a discrete path within the Fe-S cluster delivery network.

Unidirectional Fe-S cluster transfer from NfuA to ErpA
To test whether Fe-S cluster transfer occurs between ErpA and NfuA, an in vitro test was used. Holo-NfuA contains a 4Fe-4S cluster per dimer, and ErpA is able to bind a 2Fe-2S cluster per monomer (17,25,26,40). His-tagged holo-NfuA (1.2 Ϯ 0.1 iron and 1 Ϯ 0.2 sulfur/monomer; inset of Fig. 3A) was incubated, under anaerobic conditions, with one equivalent of untagged apo-ErpA, for 1 h. After separation by chromatography onto a Ni-NTA column, iron and sulfide contents of each protein were analyzed. ErpA contained 1 Ϯ 0.1 iron and 0.7 Ϯ 0.2 sulfur/monomer, and its visible spectrum was characteristic of a 2Fe-2S cluster protein with absorption bands at 420 and 320 nm. On the contrary, NfuA had lost its 420-nm Fe-S cluster absorption band and contained less than 0.1 iron and sulfur/monomer (Fig. 3A). In a separate experiment, holo-ErpA was incubated with apo-NfuA, and the two entities were submitted to the same analyses as described above after separation. No modification either in iron/sulfide content or in spectrum was observed before and after co-incubation (Fig.  3B). Together, these analyses showed that, in vitro, a unidirectional Fe-S cluster transfer occurs from NfuA to ErpA.

Stability of Fe-S cluster to oxidative damage varies with carrier identity
In vivo studies showed that ErpA and NfuA are required for Fe-S cluster transfer under aerobiosis and oxidative stress, respectively. Therefore, we tested the stability of Fe-S clusters bound to NfuA and to ErpA when exposed to O 2 . Before exposure to O 2 , reconstituted holo-NfuA contained 1.4 Ϯ 0.2 iron molecules per monomer and displayed a 4Fe-4S cluster-char- Figure 1. NfuA is required for IPP biosynthesis in E. coli during oxidative stress. WT (MG1655), ⌬nfuA, WT MVA ϩ , and ⌬nfuA MVA ϩ strains were grown overnight at 37°C in LB medium. Cultures were diluted in sterile PBS, and 5 l were directly spotted onto an LB medium plate and an LB medium plate supplemented with 100 M paraquat (PQ) and containing the MVA pathway inducer L-arabinose (0.2%) and the MVA substrate (1 mM). Growth was analyzed after overnight incubation at 37°C. Each spot represents a 10-fold serial dilution.

Biochemical properties of E. coli Fe-S cluster carriers
acteristic UV-visible spectrum in agreement with published data (40) (Fig. 4A). The degradation of the Fe-S cluster was monitored by the absorbance variation at 420 nm as a function of time when diluted into oxygenated buffer containing a controlled amount of O 2 . Under these conditions, the half-life of the NfuA Fe-S cluster was 100 Ϯ 10 min (Fig. 4A). Holo-ErpA exhibited a 2Fe-2S cluster-characteristic UV-visible spectrum (Fig. 4B). Upon controlled O 2 exposure, half-life of its cluster was 54 Ϯ 4 min (Fig. 4B). We also tested the stability of the Fe-S cluster of two other ATC, IscA and SufA, under the same experimental conditions and found half-lives of 45 Ϯ 3 and 55 Ϯ 7 min, respectively (Fig. 4C). All together, these results showed that Fe-S clusters bound to diverse carriers exhibit different capacities to resist oxidative damage with the following decreasing stability order: NfuA Ͼ ErpA ϭ SufA Ͼ IscA.

The NfuA/ErpA-bound cluster has an increased stability
Last, we tested whether stability of the ErpA bound Fe-S cluster was modified in the presence of NfuA. Both apo-NfuA and holo-ErpA proteins were mixed and exposed to O 2 , as above. Half-life of the Fe-S cluster bound to ErpA reached 90 Ϯ 6 min (i.e. an enhancement by 2-fold as compared with Fe-S bound to ErpA alone) (Fig. 5). Thus, our results indicated that the presence of NfuA led to increased stability of the ErpAbound Fe-S cluster toward O 2 .

NfuA and ErpA interact directly with the client proteins IspG and IspH
SPR was used to test the interaction between IspG/H and the Fe-S carriers. For the experiments performed with IspG, serially diluted IspG was injected into sensor chips coated with NfuA For the experiments performed with IspH, serially diluted NfuA or ErpA was injected into IspH-coated sensor chips. Although sensorgrams showed binding of NfuA and ErpA to IspH, the maximum response, was unattainable, preventing determination of a K d value (data not shown). We also used a bacterial two-hybrid approach to detect interactions between NfuA, ErpA, and IspH. We showed that the BTH101 cells synthesizing T18-IspH and T25-NfuA hybrid proteins exhibited ␤-gal activity, indicating an IspH/NfuA interaction (Fig.  7A). In addition, the two-hybrid system allowed us to show that the N-terminal domain of NfuA was sufficient to mediate the interaction with IspH (Fig. 7B). An interaction, albeit weak, was also observed between IspH and ErpA (Fig. 7A). In contrast, no interaction was observed between IspH and SufA or IscA (Fig. 7A). Collectively, these results indicated that NfuA and ErpA are able to interact directly with the client IspG and IspH proteins.

Both ErpA and NfuA are required for the maturation of aconitase B and the respiratory complexes I and II
Our previous study revealed a role of NfuA in the maturation of AcnB and the respiratory complex I (Fig. S1) (40). We then asked whether ErpA participates in the maturation of these Fe-S cluster-containing proteins.
First, we assayed the AcnB activity in the conditional mutant in which the expression of the erpA gene was under arabinose induction and glucose repression referred to as LL401 (ara p ::erpA), in which the ⌬acnA mutation has been introduced (BP721) (17). ErpA-depleted cells were obtained after 3.5-4 h of growth in glucose-supplemented medium (Fig. 8). In the ErpA-depleted cells, aconitase activity was decreased by 80% when compared with the ErpA-replete cells grown in the pres-ence of arabinose, whereas AcnB was detected at an identical level in both ErpA-replete and ErpA-depleted cells (Fig. 8). The NfuA protein level was slightly increased in ErpA-depleted cells (Fig. S2). SPR was used to investigate the interaction between ErpA and AcnB. Serially diluted ErpA (0 -270 M) was injected into sensor chips coated with AcnB (Fig. S3A). The K d value for the ErpA/ AcnB interaction was 50 Ϯ 5 M. We also used SPR experiments to quantify the previously shown NfuA-AcnB interaction (40). Serially diluted NfuA (0 -160 M) injected into sensor chips coated with AcnB yielded a K d value of 29 Ϯ 3 M (Fig. S3B).
We then tested whether ErpA was also required for the maturation of complex I. ErpA-depleted cells of the LL401 strain exhibited a drastic decrease (70%) for complex I activity when compared with ErpA-replete cells grown in the presence of arabinose (Fig. 8).
Last, we tested the contribution of ErpA and NfuA to the maturation of the respiratory complex II (Sdh). When grown in glucose, the Sdh activity of the LL401 strain was found to be drastically reduced (85%) (Fig. 8). In the nfuA mutant, the Sdh activity was decreased by 35% (Fig. 9). Collectively, all of these data indicate that maturation of the respiratory complexes I and II is with the assistance of both ErpA and NfuA under aerobic conditions and that maturation of AcnB requires ErpA under aerobic conditions and NfuA under oxidative stress conditions.

ErpA interacts with IscA and SufA
Using the Biacore experiment, we tested whether ErpA can physically interact with the ATC of the ISC and SUF machineries, IscA and SufA, respectively. ErpA was immobilized onto a Biacore sensor CM5 chip, and IscA and SufA were serially diluted and injected. A dose-response curve was obtained, where an increase in response units was observed with increasing concentrations of IscA (0 -200 M) and SufA (0 -600 M), indicating that ErpA interacted with IscA and SufA (Fig. 10). The sensorgrams indicated that IscA and SufA were released in a short time without the necessity of chip regeneration. The calculated dissociation

Biochemical properties of E. coli Fe-S cluster carriers
constant (K d ) of the ErpA/IscA and ErpA/SufA interactions was determined to be 40 Ϯ 3 and 90 Ϯ 2 M, respectively. These results indicated that ErpA is able to interact with IscA and SufA.

Multicopy suppression of the nfuA growth defect by erpA
The results above showed that NfuA and ErpA contribute to the maturation of the same set of enzymes. Therefore, we tested whether NfuA and ErpA are functionally redundant in vivo. For this, we made use of a multicopy-based suppression approach. A pBAD derivative plasmid expressing erpA (pBAD-erpA) suppressed paraquat sensitivity of the nfuA mutant (Fig. 11). In contrast, increased nfuA gene dosage failed to suppress nonviability caused by erpA mutation under aerobiosis (Fig. 12). Last, pBAD-iscA and pBAD-sufA, plasmids expressing iscA and sufA, respectively, failed to suppress ⌬nfuA mutant sensitivity to paraquat (Fig. 11). Altogether, these data showed that ErpA overproduction overcomes the need of NfuA for IspG/H maturation but that the converse is not true.

Discussion
Interest in Fe-S cluster biology keeps expanding as one realizes how much these cofactors are central to multiple issues, from basic knowledge to molecular medicine, antibiotic resistance, and biotechnological applications (41,42). In most organisms, several dozen structurally and functionally diverse client protein species need to acquire an Fe-S cluster for func-

Biochemical properties of E. coli Fe-S cluster carriers
tioning. The diversity of clients, the diversity of the clusters, and the fluctuating environmental conditions under which Fe-S cluster trafficking takes place make the question of the delivery a most challenging issue. Here we show that synthesizing different carriers can afford different ROS resistance environments for transported Fe-S clusters, hence delivery circuits better adapted to sustain oxidative stress. We show that enhanced resistance to ROS can also be achieved by interaction between Fe-S carriers. We propose a model in which, right after the assembly step, the Fe-S trafficking network diversifies in multiple branches, which eventually converge toward ErpA. This unprecedented view of Fe-S cluster trafficking and delivery is discussed compared with the eukaryotic situation.
The present study allowed us to quantitate ROS resistance capacity of a cluster bound to NfuA, ErpA, SufA, and IscA carriers. Likewise, NfuA-and SufA-carried clusters showed the highest resistance to ROS. This is fully consistent with the genetically based view of SufA and NfuA as stress-responding factors and predicted to be better adapted to deliver clusters under oxidative stress conditions (16,25,26). NfuA results from the fusion between an NFU domain, in the C-terminal region, and, in the N-terminal region, a degenerated ATC domain lacking the Fe-S cluster-liganding cysteine residues (25,26,40). NfuA forms a dimer and binds a 4Fe-4S cluster in its NFU domain, probably at the dimer interface (25,26,40). Then it is possible that the cluster stability of NfuA is related to the four cysteinyl ligations and/or shielding by the dimer formation.
An unsuspected association between NfuA and ErpA was observed both by SPR and two-hybrid-based methods. Interestingly, the presence of NfuA led to increased stability of the ErpA-bound Fe-S cluster toward O 2 . An attractive hypothesis is that the ErpA/NfuA association resulted in an enhanced stability of the ErpA-bound cluster as compared with when bound to ErpA alone. Next, biochemical and structural studies will aim to test this hypothesis and investigate the molecular basis for the enhanced ROS resistance procured by NfuA. In particular, we will investigate whether apo-forms of NfuA stabilize the ErpAbound cluster via its Cys-reactive residues and whether the A-type Fe-S carriers, IscA and SufA, might also exert the same effect.
Our previous phylogenetic studies permitted us to classify ATC into two different families. ErpA and IscA/SufA were classified into the ATC-I and -II families, respectively (16). Members of the ATC-I family were predicted to partnership with the apo-targets, whereas the ATC-II members were thought to be connected to scaffolds (16). The present bacterial two-hybrid and SPR-based investigations fully support the phylogenetically based functional prediction. Direct interactions were observed between apo-targets (IspG/H and AcnB) and ErpA, but not IscA and/or SufA (data not shown). Taken together with our previous genetic analysis, this observation confirms the view that scaffold-bound Fe-S clusters are transferred to ATC-II (IscA or SufA), which transfers them to ErpA, which delivers them to the apo-targets. In mitochondria, once assembled by the core biogenesis machinery, 2Fe-2S clusters reach a heteromeric platform, the ISCA1-ISCA2-Iba57 complex, which converts them into 4Fe-4S before targeting them to a so-called dedicated factor, which in turn allows maturation of cellular Fe-S proteins (15,(43)(44)(45). Mammalian ISCA1/2 proteins contain 2Fe-2S clusters. Heterodimer ISCA1-ISCA2 can assemble a 4Fe-4S cluster in the presence of GRX5, implying that the heterodimeric complex is the functional unit for 4Fe-4S cluster formation before their transfer to specific targets (45). ISCA1 is an ATC-II, like IscA and SufA, whereas ISCA2 is an ATC-I, like ErpA. In essence, therefore, the complex ISCA1-ISCA2 resembles the partnership IscA/ErpA or SufA/ErpA discussed above except that no stable bacterial ATC-II-ATC-I complex has ever been isolated. Note, however, that in mice, despite forming a complex, ISCA1 and ISCA2 are able to carry out separate tasks, presumably depending upon the substrates and the conditions (46).

Biochemical properties of E. coli Fe-S cluster carriers
like ErpA, NfuA appears to have all features required to interact directly with targets and could be positioned at the ultimate step within the Fe-S cluster delivery process. If direct transfer between NfuA and target proteins can occur in vivo, its contribution appears very modest as compared with the contribution of ErpA. Hence, lack or depletion of ErpA caused a stronger phenotype than the lack of NfuA, as shown here by a drop in complex I and II activities by over 70 and 85%, whereas an nfuA mutant exhibited a Ͻ2-fold reduction in complex I and complex II activities. Similarly, aconitase B activity was down by Ͼ80% in ErpA-depleted cells, whereas it was only slightly altered under stress conditions in cells lacking NfuA. Also, erpA was able to act as a multicopy suppressor of nfuA, whereas the reverse was not true. We thus favor the hypothesis that in E. coli, NfuA acts conjointly with ErpA, by providing it an Fe-S cluster, rather than being an ultimate Fe-S donor for target proteins (Fig. 13).
The yeast and human Nfu1 and the E. coli NfuA proteins share a conserved NFU-type C-terminal domain that binds a 4Fe-4S cluster. Interestingly, as NfuA, the yeast and human Nfu1 were found associated with the ISA proteins (33,46). In yeast, defects exhibited by nfu1 mutation were of much lesser extent than those caused by mutations in the ISA complex

Biochemical properties of E. coli Fe-S cluster carriers
(ISCA homologs in yeast) (13,14,30,47). In fact, defects were mainly apparent under oxidative stress conditions (33,47). Hence, the parallel with the E. coli situation is striking. Moreover, overexpression of Isa2, which we classified as an ATC-I member, was able to suppress a defect of the NFU1 mutant, whereas overexpression of Isa1, an ATC-II member failed to do so (33). This is highly reminiscent of what we observed here with E. coli, wherein erpA, an ATC-I, but not iscA or sufA, both ATC-II members, was able to act as a multicopy suppressor of nfuA. This argues that in both prokaryotes and eukaryotes, NFU-containing proteins are called upon in stress conditions at the delivery step but that their contribution can be bypassed, given that the level of the ATC-II member (ISCA2 or ErpA) is sufficiently high and that the proposed scenario of NFU proteins as a Fe-S refuel for ATC-I can also be envisioned at the late stage of the mitochondrial Fe-S delivery step.
In conclusion, independently of the organisms considered and of the type of Fe-S client proteins to be matured, a similar overall strategy has been retained throughout evolution. Despite the multitude of apoprotein clients waiting for their clusters, cells evolved a handful of carriers, which they either synthesize at appropriate levels under a set of conditions or combine in different higher-order organization, from homodimers to heterotetramers, each possibility providing a better solution to adapt to the diversity of client proteins and growth conditions. This study pinpoints how diversifying carriers and combining different carriers constitute two strategies that the cell has evolved to yield multiple delivery pathways, in particular a robust ROS-resistant delivery network.

Strains and growth conditions
Strains used in this study are listed in Table 1. The ⌬acnA::kan KEIO mutation was introduced by P1 transduction (48). E. coli strains were grown in Luria-Bertani (LB) rich medium at 37°C. Solid medium contained 1.5% agar. Ampicillin was used at 50 g/ml. Arabinose (0.2%) and mevalonate (1 mM) were added when required.

Paraquat sensitivity test
The paraquat sensitivity test was performed on overnight cultures that were diluted in sterile PBS and directly spotted onto LB plates containing 100 M paraquat (25). The plates were incubated overnight at 37°C before growth was scored.

Surface plasmon resonance analysis
The binding studies were performed using a Biacore TM T200 instrument and CM5 sensor chips. Proteins were covalently immobilized onto the chip using a standard primary aminecoupling procedure. The same amine-coupling procedure, but without any protein, was performed on the reference channel as a blank to subtract any nonspecific binding signal. The analytes, dissolved in running Tris buffer (50 mM Tris, 150 mM NaCl, pH  7,4) or HBS buffer (10 mM Hepes, 150 mM NaCl, 3 mM EDTA, 0.005% Tween, pH 7.4), were serially diluted to various concentrations with the running buffer. According to their kinetics of association and dissociation, they were injected for 50 -125 s during the association phase at a constant flow rate of 10 l/min at 25°C. The dissociations were subsequently followed for 75-250 s at the same flow rate. For each cycle, the sensor surface was regenerated with 10 mM Gly-HCl buffer (pH 2.0) when the data collection was finished. To calculate the rate and affinity constants, the results from the sensograms were fit globally with BIAcore 3000 analysis software (BIAevaluation version 4.1) that proposes 1:1 Langmuir binding modes.

Bacterial two-hybrid technique
We used the adenylate cyclase-based two-hybrid technique (51). DNA inserts encoding the proteins of interest were obtained by PCR and were cloned into pUT18C and pKT25 plasmids. After co-electroporation of the BTH101 strain with the two plasmids expressing the hybrid proteins, plates were incubated at 30°C for 2 days. Three milliliters of LB medium supplemented with antibiotics and isopropyl 1-thio-␤-D-galactopyranoside at the recommended concentration (51) were inoculated and incubated overnight at 30°C. ␤-Galactosidase activity was determined in Miller units. Strains were grown overnight in LB supplemented with arabinose. Fresh LB medium with 0.2% glucose (Glu) or arabinose (Ara) was inoculated. Samples were taken and analyzed when cells reached late-exponential phase (3.5-4 h of growth). A, aconitase activity was assayed using the arap::erpA ⌬acnA strain (BP721). B, immunoblot analysis was performed on glucose-or arabinose-grown cells of the BP721 strain, using antibodies raised against AcnB (top) and ErpA (bottom); identical amounts of total proteins were loaded. Extract of the ⌬acnA ⌬acnB double mutant (BP444) was used as a negative control for immunoblotting using antibodies raised against AcnB. The molecular mass (kDa) of the proteins used in the molecular weight size marker (MW) is indicated. D-NADH oxidase activity (C) and succinate dehydrogenase (D) activities were assayed from the LL401 strain (arap::erpA) strains. The experiments were run in triplicate, and the S.E. values are shown (error bars).

Biochemical properties of E. coli Fe-S cluster carriers In vitro Fe-S reconstitution
NfuA and the three ATC were reconstituted as described (25,49). The purified proteins were obtained primarily in the apoform and were reconstituted anaerobically in the glove box at 18°C as follows. 0.08 mM protein was mixed with 5 mM DTT, 10 M IscS, 1 mM L-cysteine, and 0.2 mM ferrous iron in 50 mM Tris, pH 8, 50 mM KCl. After a 90-min incubation, samples were passed over a desalting column and concentrated when needed.

Cluster transfer
All of the following procedures were performed anaerobically in the glove box at 18°C.
Fe-S cluster transfer from holo-ErpA to apo-NfuA-Apo-NfuA His (75 M) was mixed for 1 h with holo-ErpA (reconstituted ErpA, as described above) in buffer A (0.1 M Tris-HCl, pH 8, 50 mM KCl, 5 mM DTT) in a molar ratio to give sufficient amounts of iron and sulfur per NfuA monomer. After incubation, DTT was removed from the solution on a Nap-10 column equilibrated with buffer B (0.1 M Tris-HCl, pH 8, 0.1 M NaCl) before separation of the proteins on a Ni-NTA affinity column (1 ml) equilibrated with the same buffer. Fe-S donor proteins, which do not contain a polyhistidine tag, were recovered in the flow-through and wash fractions, whereas NfuA His was eluted with buffer B containing 0.2 M imidazole and desalted over a Nap-10 column to remove imidazole. Each fraction (wash and elution fractions) was analyzed for its iron content, and a UVvisible spectrum was recorded.
Fe-S cluster transfer from holo-NfuA to apo-ErpA-Apo-ErpA was first pre-reduced with 5 mM DTT for 10 min. DTT was removed onto a MicroBio-Spin 6 chromatography column. Reduced ErpA (150 M) and holo-NfuA His were mixed together in buffer B in a molar ratio allowing for the provision of one iron and one sulfur atom per ErpA monomer. After a 1-h incubation, proteins were separated on the Ni-NTA affinity column and analyzed as described above.

Protein analysis
Protein concentrations were measured by the method of Bradford using BSA as a standard. Iron and sulfide quantifications were carried out as described previously (53,54). UV-visible spectra were recorded on a Cary Bio (Varian) spectrophotometer.

Fe-S cluster stability toward oxygen
To control the amount of oxygen that would react with the reconstituted protein Fe-S cluster, the experiment was performed using a cuvette containing 100 l of a solution (0.1 M Tris, 0.1 M KCl, pH 8) equilibrated with air outside the glove box. At 20°C and at 760 mm Hg, this solution contains 284 M oxygen (55). The cuvette was sealed with a septum and intro- Figure 13. Working model of the E. coli Fe-S delivery step. Arrows represent the routes (i.e. the carriers used, IscA, SufA, ErpA, and NfuA) to deliver the Fe-S clusters (green/red triangles) from the scaffold to the target proteins. The thickness of the arrows increases with the level of ROS met by the bacterium. Table 1 Strains and plasmids used in this study duced into the glove box. A controlled amount of the reconstituted proteins was introduced into the cuvette using a Hamilton syringe to give a ratio of 1:20 for amount of iron/O 2 . The kinetics of degradation of the Fe-S cluster was followed by UVvisible spectroscopy.

Enzymatic activity
Aconitase activity-Strains were grown in LB at 37°C to an A 600 of 0.8. The cells were harvested; washed with 50 ml of 50 mM Tris-HCl, pH 7.5, buffer; frozen in liquid nitrogen; and stored at Ϫ80°C. The cell pellets were transferred to a Coy anaerobic chamber (90% N 2 , 10% H 2 ) and resuspended in anaerobic 50 mM Tris-HCl, pH 7.5 (0.4% of the culture volume). Cells were lysed using a French press and centrifuged at 16,000 ϫ g for 5 min. Supernatant was immediately frozen in liquid nitrogen. In the Coy anaerobic chamber, cell extracts containing 100 mg of protein were added to 50 mM Tris-HCl, pH 7.5, 0.6 mM MnCl 2 , 30 mM citrate, 0.2 mM NADP ϩ , 1.7 units of isocitrate dehydrogenase in a 1-ml final volume. Aconitase activity was assayed by following the formation of NADPH in the coupled assay as an increase in absorbance at 340 nm (56).
NADH dehydrogenase activity-NADH dehydrogenase activity was adapted from a method described previously (57). Briefly, cells were harvested by centrifugation; resuspended in 50 mM phosphate buffer, pH 7.5; lysed using a French press; and frozen immediately in liquid nitrogen. NADH activity was assayed on the thawed samples by immediately adding D-NADH (200 mM) as substrate and by following A 340 .
Succinate dehydrogenase activity-Cells were harvested by centrifugation; resuspended in 50 mM phosphate buffer, pH 7.5; and lysed using a French press. Following centrifugation (11,000 r.p.m. for 15 min at 4°C), the supernatant was submitted to ultracentrifugation (45,000 r.p.m. for 2 h at 4°C). SDH activity was assayed on the pellet fraction resuspended in 50 mM phosphate buffer, pH 7.5. Samples were preincubated for 30 min at 30°C in 4 mM succinate, 1 mM KCN, 50 mM phosphate buffer, pH 7.5. The assay was performed by adding dichlorophenolindophenol (100 mM) and phenazine ethosulfate (1 mM) as substrate and by following A 600 .

Immunoblotting
Equal quantities of proteins were separated on SDS-polyacrylamide gels and transferred onto nitrocellulose membranes. The membrane filter was incubated with appropriate antibodies (anti-AcnB, anti-ErpA) diluted 1:2000. Immunoblots were developed by using horseradish peroxidaseconjugated goat anti-rabbit antibody, followed by chemiluminescence detection.