GLRX3 Acts as a [2Fe–2S] Cluster Chaperone in the Cytosolic Iron–Sulfur Assembly Machinery Transferring [2Fe–2S] Clusters to NUBP1

Human cytosolic monothiol glutaredoxin-3 (GLRX3) is a protein essential for the maturation of cytosolic [4Fe–4S] proteins. We show here that dimeric cluster-bridged GLRX3 transfers its [2Fe–2S]2+ clusters to the human P-loop NTPase NUBP1, an essential early component of the cytosolic iron–sulfur assembly (CIA) machinery. Specifically, we observed that [2Fe–2S]2+ clusters are transferred from GLRX3 to monomeric apo NUBP1 and reductively coupled to form [4Fe–4S]2+ clusters on both N-terminal CX13CX2CX5C and C-terminal CPXC motifs of NUBP1 in the presence of glutathione that acts as a reductant. In this process, cluster binding to the C-terminal motif of NUBP1 promotes protein dimerization, while cluster binding to the N-terminal motif does not affect the quaternary structure of NUBP1. The cluster transfer/assembly process is not complete on both N- and C-terminal motifs and indeed requires a reductant stronger than GSH to increase its efficiency. We also showed that the [4Fe–4S]2+ cluster formed at the N-terminal motif of NUBP1 is tightly bound, while the [4Fe–4S]2+ cluster bound at the C-terminal motif is labile. Our findings provide the first evidence for GLRX3 acting as a [2Fe–2S] cluster chaperone in the early stage of the CIA machinery.


■ INTRODUCTION
The biogenesis of iron−sulfur (Fe−S) proteins is a highly conserved, multistep process, which involves dedicated machineries. 1 In eukaryotes, two distinct machineries are required for the maturation of mitochondrial, cytosolic, and nuclear Fe−S proteins. 2 In the current working model, a mitochondrial iron−sulfur cluster (ISC) assembly machinery de novo synthesizes a [2Fe−2S] cluster and then incorporates it into mitochondrial [2Fe−2S] and [4Fe−4S] target proteins. 3,4 Another machinery in the cytosol, named cytosolic iron−sulfur assembly (CIA) machinery, is in charge of the maturation of cytosolic and nuclear [4Fe−4S] proteins. 5 The CIA machinery has been mostly characterized in yeast, where it was proposed to begin with the assembly of a [4Fe−4S] cluster on a scaffold complex formed by two homologous P-loop nucleoside triphosphatases (NTPases) named Nbp35 and Cfd1. 6−8 It was also proposed that the latter process is assisted by a cytosolic electron transfer chain comprising the CIA components diflavin oxidoreductase Tah18 and the Fe−S binding protein Dre2. 9 In humans, this first phase of the CIA machinery involves the same set of proteins, which were proposed to perform the same function as their yeast homologues. Both in vivo and in vitro data indicated indeed that a [4Fe−4S] cluster assembly occurs on a scaffold complex formed by the human homologues of yeast Nbp35 and Cfd1, that is, NUBP1 and NUBP2, 10,11 and that this process is again assisted by an electron transfer chain composed by the human homologues of yeast Tah18 and Dre2, that is, NDOR1 and anamorsin. 9 NUBP1 and NUBP2 share a high degree of sequence identity with Nbp35 and Cfd1, respectively. While the two yeast proteins are known to form homodimeric and heterodimeric/-tetrameric complexes able to bind up to a maximum of three [4Fe−4S] clusters per dimer, 6,8,12,13 the cluster binding properties and the quaternary structure of NUBP1 and NUBP2 proteins are not so well characterized. We only know that they form a heterocomplex in vivo and that NUBP1 binds the [4Fe−4S] cluster(s), 10 but the biophysical data so far available do not permit one to decide how the cluster(s) are associated with NUBP1 and NUBP2. Both NUBP1 and NUBP2 have a CPXC motif in their C-terminal region as present in Nbp35 and Cfd1. This motif was found to be essential for the function of Cfd1 and of Nbp35 in the assembly of cytosolic [4Fe−4S] proteins. 12 It coordinates a labile [4Fe−4S] cluster bridging two protein molecules in homodimeric Nbp35, in homodimeric Cfd1, and in the Cfd1− Nbp35 heterocomplex. 8,12 Furthermore, NUBP1 and Nbp35 share a conserved N-terminal CX 13 CX 2 CX 5 C motif. 6,10,14 In Nbp35, this motif was found to be essential for the protein function and to tightly bind a [4Fe−4S] cluster. 12 How the assembly of the [4Fe−4S] clusters at both N-and C-terminal motifs of the two human NTPases occurs in the cytoplasm is still elusive. In particular, the source of iron and of the sulfide provided to the NUBP1−NUBP2 heterocomplex to assemble a [4Fe−4S] cluster is still a matter of debate. 2,15 Functional data clearly showed that, in yeast, Fe−S cluster assembly on Cfd1 and Nbp35 is independent of the proteins acting later in the CIA machinery. 8 A possible player of the CIA machinery responsible for the maturation of the [4Fe− 4S] clusters on Cfd1 and Nbp35 is the yeast cytosolic monothiol glutaredoxin Grx3, which is, indeed, a crucial component for the assembly of cytosolic Fe−S proteins. 16 In support of this model, in vivo data showed that Grx3 and its yeast paralog Grx4 bind a Fe−S cluster independent of the CIA machinery. 16 The human proteome contains only one monothiol glutaredoxin in the cytosol, GLRX3 (also commonly named PICOT), which is essential for the maturation of cytosolic [4Fe−4S] proteins. 17 GLRX3 has been shown to be involved in the CIA machinery by transferring its [2Fe −2S] clusters to the CIA component anamorsin, de facto acting as a [2Fe −2S] cluster chaperone in the cytosol. 18−20 Possibly, GLRX3 could transfer its [2Fe −2S] cargo to other targets, and not only to anamorsin. However, up to now, whether GLRX3 provides [2Fe−2S] clusters to NUBP1 and NUBP2 for assembling [4Fe−4S] clusters is still unknown.
In this work, we investigated, through various spectroscopic techniques, the ability of human GLRX3 to act as a [2Fe −2S] cluster chaperone for NUBP1. We found that GLRX3  ■ EXPERIMENTAL SECTION Cloning, Overexpression, and Purification of wtNUBP1, NUBP1 38−320 , and NUBP1-C235A/C238A Mutant in Their Apo and Holo Forms. The gene coding for human NUBP1 (UniProtKB/ Swiss-Prot: P53384), inserted into the pUC-57 plasmid, was acquired from Sigma-Aldrich. The genes of wild-type NUBP1 (wtNUBP1, hereafter) and of a NUBP1 construct restricted to residues 38−320, containing only the C-terminal motif (NUBP1 38−320 , hereafter), were amplified by PCR and inserted into the pETDuet-1 expression vector using EcoRI and HindIII Fastdigest restriction enzymes (Thermo-Fisher Scientific). The NUBP1-C235A/C238A mutant was obtained through site-directed mutagenesis (Agilent QuikChange II sitedirected mutagenesis kit) performed on pETDuet-wtNUBP1 according to the producer's manual. His 6 -tagged wtNUBP1 and His 6 -tagged NUBP1-C235A/C238A were overexpressed in BL21-(DE3) and His 6 -tagged NUBP1 38−320 in BL21(DE3) Codon Plus RIPL competent Escherichia coli cells (Novagen). Cells were grown in Luria−Bertani containing 1 mM ampicillin at 37°C under vigorous shaking up to a cell OD 600 of 0.6. Protein expression was induced by adding 0.5 mM IPTG and 0.25 mM FeCl 3 . Cells were grown overnight at 21°C (wtNUBP1 and NUBP1-C235A/C238A) and 25°C (NUBP1 38−320 ). Cells were harvested by centrifugation at 7500g and resuspended in lysis buffer (40 mM sodium phosphate buffer pH 8.0, 400 mM NaCl, 5 mM imidazole, containing 0.01 mg/mL DNAase, 0.01 mg/mL lysozyme, 1 mM MgSO 4 , and 5 mM dithiothreitol (DTT)). Cell disruption was performed on ice by sonication, and the soluble extract was obtained by ultracentrifugation at 40 000g. The following purification steps were performed aerobically to obtain the apo protein, while in an anaerobic chamber (O 2 < 1 ppm) to isolate the protein in its holo form. The soluble fraction was loaded on a HisTrap FF column (GE Healthcare). The protein was eluted with 40 mM sodium phosphate buffer pH 8.0, 400 mM NaCl, and 400 mM imidazole, concentrated with Amicon Ultra-15 Centrifugal Filter Units with a MWCO of 10 kDa (Millipore), and the buffer was exchanged by a PD-10 desalting column in 40 mM sodium phosphate buffer pH 8.0, 400 mM NaCl, 5 mM imidazole. When required, cleavage of the His 6 tag was performed by TEV protease overnight at room temperature. The protein mixture was then loaded on a HisTrap FF column (GE Healthcare) to separate the digested protein from the His 6 tag.
The apo protein in the monomeric state was obtained by incubating overnight the aerobically purified protein at room temperature in 25 mM MOPS, 100 mM pyridine-2,6-dicarboxylic acid, at pH 7.0.
Chemical reconstitution was performed inside an anaerobic chamber (O 2 < 1 ppm), by incubating overnight at room temperature the monomeric apo protein in degassed 50 mM Tris, 100 mM NaCl, 5 mM DTT, at pH 8.0 with up to a 12-fold excess of FeCl 3 and Na 2 S. Excess of FeCl 3 and Na 2 S was anaerobically removed by passing the mixture on a PD-10 desalting column, and the holo protein was recovered.
Production of [2Fe−2S] 2 -GLRX3 2 -GS 4 . GLRX3 was expressed, purified, and chemically reconstituted following previously reported procedures. 18 Protein, Iron, and Acid-Labile Sulfide Quantification. Protein quantification was carried out with the Bradford protein assay, using BSA as a standard. Nonheme iron and acid-labile sulfide content was determined as described previously. 21 Biochemical and Spectroscopic UV−Vis, CD, EPR, and NMR Methods. The quaternary structure of the proteins was analyzed through analytical gel filtration on a Superdex 200 10/300 Increase column (GE Healthcare). Column was calibrated with a gel filtration marker calibration kit, 6500−66 000 Da (Sigma-Aldrich), to obtain the apparent molecular masses of the detected species. Samples in degassed 50 mM phosphate buffer pH 7.0, 200 mM NaCl, 5 mM DTT (plus 5 mM GSH for [2Fe−2S] 2 -GLRX3 2 -GS 4 ), were loaded on the pre-equilibrated column. Elution profiles were recorded at 280 nm with a flow rate of 0.65 mL/min. UV−visible (UV−vis) spectra were anaerobically acquired on a Cary 50 Eclipse spectrophotometer in degassed 50 mM phosphate buffer pH 7.0 (plus 5 mM GSH for [2Fe−2S] 2 -GLRX3 2 -GS 4 ).
Circular dichroism (CD) spectra were acquired on a JASCO J-810 spectropolarimeter in 20 mM phosphate buffer pH 7.0.
CW EPR spectra were recorded before and after the anaerobic reduction of the cluster(s) by addition of up to 5 mM sodium dithionite. Protein concentration was in the range 0.5−0.7 mM, in degassed 50 mM Tris buffer pH 8.0, 100 mM NaCl, 5 mM DTT, and 10% glycerol. EPR spectra were acquired at 10 and 45 K, using a Bruker Elexsys 580 spectrometer working at a microwave frequency of ca. 9.36 GHz, equipped with a SHQ cavity and a continuous flow He cryostat (ESR900, Oxford instruments) for temperature control. Acquisition parameters were as follows: microwave frequency, 9.36 GHz; microwave power, 1 mW at 10 K and 0.12 mW at 45 K; modulation frequency, 100 kHz; modulation amplitude, 10 G; acquisition time constant, 163.84 ms; number of points 1024; number of scans 4; field range 2000−4000 G.
Paramagnetic 1D 1 H NMR experiments were performed on a Bruker Avance spectrometer operating at 400 MHz 1 H Larmor frequency and equipped with a 1 H dedicated 5 mm probe. Water signal was suppressed via fast repetition experiments and water selective irradiation. 22 Experiments were typically performed using an acquisition time of 50 ms and an overall recycle delay of 80 ms. Sample concentration was in the range 0.5−0.7 mM, in degassed 50 mM phosphate buffer pH 7.0. Squared cosine and exponential multiplications were applied prior to Fourier transformation. Manual baseline correction was performed using polynomial functions.
Cluster Transfer from [2Fe−2S] 2 -GLRX3 2 -GS 4 to His 6 -Tagged Apo wtNUBP1, His 6 -Tagged Apo NUBP1 38−320 , and cluster, that is, three per dimeric wtNUBP1, one per monomeric NUBP1-C235A/C238A, and one per dimeric NUBP1 38−320 . Separation of the His 6 -tagged NUBP1 species from untagged GLRX3 after reaction was performed in anaerobic conditions by loading the reaction mixtures on a His GraviTrap column pre-equilibrated with 40 mM sodium phosphate buffer pH 8.0, 400 mM NaCl, and 5 mM imidazole. The His 6 -tagged NUBP1 species was eluted with 40 mM sodium phosphate buffer pH 8.0, 400 mM NaCl, and 400 mM imidazole. After concentration, the buffer was exchanged by a PD-10 desalting column in the appropriate degassed buffer required to perform analytical gel filtration, iron and acid-labile sulfide quantification, and to acquire UV−vis and paramagnetic 1D 1 H NMR spectra.
Cluster Transfer from [2Fe−2S] 2 -GLRX3 2 -GS 4 to His 6 -Tagged Apo wtNUBP1 in the Presence of Different GSH Concentrations. His 6 -tagged monomeric apo wtNUBP1 was incubated under anaerobic conditions with 1.5 equiv of [2Fe−2S] 2 -GLRX3 2 -GS 4 , in the presence of increasing amounts of GSH (0, 1, 5, and 10 mM), for 1 h at room temperature in 40 mM sodium phosphate buffer pH 8.0, 400 mM NaCl, and 5 mM imidazole. Separation of His 6 -tagged wtNUBP1 from untagged GLRX3 after reaction was performed in anaerobic conditions as described above. After concentration, the buffer was exchanged by a PD-10 desalting column in degassed 50 mM phosphate buffer pH 7.0. and UV−vis spectra were then acquired.

Fe−S Cluster Binding Properties and Quaternary
Structure of Human NUBP1. The characterization of the Fe−S cluster binding properties and quaternary structure of NUBP1 is an essential prerequisite to investigate in detail Fe− S cluster transfer from GLRX3 to NUBP1. For this purpose, in addition to wild-type NUBP1 (wtNUBP1, hereafter), we produced a construct where the first 37 N-terminal residues were deleted (NUBP1 38−320 , hereafter), which thus contains only the C-terminal CPXC motif, and a mutant containing only the N-terminal CX 13 CX 2 CX 5 C motif, by mutating the two cysteines of the CPXC motif, that is, C235 and C238, into alanines (NUBP1-C235A/C238A, hereafter). The N-terminal deletion of NUBP1 as well as C235A/C238A mutations did not affect protein folding, as shown by circular dichroism and 1D 1 H NMR spectroscopy ( Figure S1). UV−vis spectra of anaerobically purified wtNUBP1 and NUBP1-C235A/C238A exhibit a broad absorption band at ∼410 nm, which is characteristic of a [4Fe−4S] 2+ cluster 23 ( Figure 1A and B). The samples are EPR silent, as expected for the S = 0 ground state of an oxidized [4Fe−4S] 2+ cluster. 24 Chemical reduction with sodium dithionite of both samples gave rise to similar intense rhombic EPR signals at 10 K, with g values of 2.05, 1.92, and 1.86, which broadened beyond detection at 45 K ( Figure 1D and E), in agreement with the presence of [4Fe− 4S] + clusters. These results compare quite well with what was previously observed for the yeast NUBP1 homologue Nbp35. 8 Anaerobically purified wtNUBP1 and NUBP1-C235A/ C238A showed differences in their quaternary structure, as determined by analytical gel filtration: wtNUBP1 is dimeric, whereas NUBP1-C235A/C238A is monomeric (Figure 2A and B). Anaerobically purified wtNUBP1 and NUBP1-C235A/ C238A contained 0.9 and 0.3 [4Fe−4S] clusters per dimer and monomer, respectively, according to protein, acid-labile sulfide, and iron analysis (Table 1). These data indicate the presence of substoichiometric amounts of cluster(s) in anaerobically purified wtNUBP1 and NUBP1-C235A/C238A (0.9 vs 3 clusters and 0.3 vs 1 cluster, respectively).
Paramagnetic 1D 1 H NMR spectra of these samples showed four intense hyperfine shifted signals in the 18−11 ppm spectral region ( Figure 3, traces a and c), whose chemical shift values and line widths are typical of βCH 2 of cysteines bound to a [4Fe−4S] 2+ cluster. 25,26 The anti-Curie temperature dependence of these signals further confirms the presence of an oxidized [4Fe−4S] 2+ cluster ( Figure S2). Moreover, the observation that the paramagnetic 1D 1 H NMR spectrum of anaerobically purified wtNUBP1 is well superimposable to that of anaerobically purified NUBP1-C235A/C238A (Figure 3), which can only bind a [4Fe−4S] 2+ cluster at the N-terminal motif, indicates that the [4Fe−4S] 2+ clusters in the anaerobically purified wtNUBP1 are bound to the N-terminal motif, and thus the occupancy of the C-terminal site by [4Fe−4S] 2+ clusters is very low.
Contrary to wtNUBP1 and NUBP1-C235A/C238A, anaerobically purified NUBP1 38−320 was essentially colorless, contained less than 5% of [4Fe−4S] cluster (Table 1), and showed no signals in the UV−vis spectrum ( Figure 1C) as well as in the paramagnetic 1D 1 H NMR and EPR spectra, the latter even after dithionite reduction (data not shown), indicating that it does not bind Fe/S clusters, as was previously observed in a truncated form of Nbp35 lacking the N-terminal 52 amino acids. 8 Anaerobically purified NUBP1 38−320 is homodimeric, as determined by analytical gel filtration ( Figure  2C). Incubation of anaerobically purified NUBP1 38−320 as well as of anaerobically and aerobically purified wtNUBP1, which are both totally in a dimeric state, although they contain from zero to substoichiometric percentages of bound Fe/S clusters,   Journal of the American Chemical Society pubs.acs.org/JACS Article with pyridine-2,6-dicarboxylic acid (PDC), a strong divalent transition metal chelator, led to their monomerization ( Figure  2A and C). This indicates that their dimerization is promoted, not only by the binding of a bridged Fe/S cluster, but also by the binding of an adventitious metal ion(s), shared by C235 and/or C238 from each subunit of the dimer. Consistent with this behavior, the treatment of both anaerobically and aerobically purified NUBP1-C235A/C238A with PDC does not affect the quaternary structure that is always monomeric ( Figure 2B). Upon chemical reconstitution of monomerized apo wtNUBP1, monomeric apo NUBP1-C235A/C238A, and monomerized apo NUBP1 38−320 species in the presence of a 12-fold excess of FeCl 3 and Na 2 S, under reductive, anaerobic conditions, the broad absorption band at ∼410 nm, typical of oxidized [4Fe−4S] 2+ protein-bound clusters, appeared in the UV−vis spectra of NUBP1 38−320 , wtNUBP1, and NUBP1-C235A/C238A, having, in the latter two, higher intensity with respect to that in the UV−vis spectra of the anaerobically purified proteins (Figure 1). After chemical reconstitution, the quantification of iron and acid-labile sulfide indicated that ∼1. The latter were acquired, in the presence of 5 mM DTT and in anaerobic conditions, as a series of 1 h spectra, over a period of roughly 12 h. Over this period of time, the intensity of the paramagnetic signals of [4Fe−4S] NUBP1 38−320 decreased, while that of the signals of [4Fe− 4S] NUBP1 C235A/C238A did not, clearly indicating cluster loss from the former mutant. To further corroborate the latter proposal, a freshly prepared sample of chemically reconstituted NUBP1 38−320 was anaerobically sealed in a quartz cuvette and in the presence of 5 mM DTT, to completely avoid cluster oxidation over time, and UV−vis spectra were recorded for 24 h. It resulted that the band at ∼410 nm characteristic of a [4Fe−4S] 2+ cluster decreases in intensity ( Figure S3), indicating that the [4Fe−4S] cluster bound to the C-terminal motif is kinetically labile. The same behavior was not observed for the chemically reconstituted NUBP1-C235A/C238A, which, indeed, does not lose the cluster over time. A freshly prepared chemically reconstituted NUBP1 38−320 was anaerobically reduced by addition of sodium dithionite and rapidly frozen. The EPR spectrum, acquired at 10 K, showed an axial signal with g values of 2.05 and 1.92 that broadened beyond detection at 45 K ( Figure 1F), in agreement with the presence of a [4Fe−4S] + cluster bound to the C-terminal motif. This spectrum is similar to that previously reported for a construct of Nbp35 lacking the N-terminal motif. 8 Analytical gel filtration of the species obtained chemically reconstituting the monomeric proteins showed that dimeriza-tion occurred for wtNUBP1 and NUBP1 38−320 , but not for NUBP1-C235A/C238A (Figure 2). Specifically, wtNUBP1 and NUBP1 38−320 were eluted in two fractions: the one eluting at smaller elution volume (at ∼15 mL), that is, the dimeric species, has a dark brown color and thus contains a dimeric [4Fe−4S] 2+ NUBP1 species, while the fraction eluting at ∼16 mL, that is, the monomeric species, is colorless and thus contains the apo protein. This result indicates that cluster binding to the C-terminal motif induces protein dimerization by bridging two C-terminal motifs, as was previously observed in yeast Nbp35. 12 Moreover, the presence of the monomeric apo species in the gel filtration profiles of the chemically reconstituted monomerized apo proteins (Figure 2) is in agreement with the partial cluster loading resulting from the iron and acid-labile sulfide quantification data (Table 1). Taking into account the molar fractions of holo wtNUBP1 versus the apo form in the chemically reconstituted sample (i.e., 0.6 and 0.4, respectively, estimated from the gel filtration profile, Figure 2A) and the cluster content ( clusters, while the C-terminal cluster binding site is occupied at very low extent, that is, ∼20% occupied. This model is supported by the paramagnetic 1D 1 H NMR spectra of freshly prepared chemically reconstituted NUBP1 38−320 and wtNUBP1. The first showed very weak hyperfine shifted signals at 12 and 10 ppm in the spectral region typical of βCH 2 of cysteines bound to a [4Fe−4S] 2+ cluster (Figure 3, trace d), and their very low signal intensity agrees with a very low [4Fe− 4S] cluster occupancy at the C-terminal site, as a consequence of the [4Fe−4S] 2+ cluster binding lability at this motif. The second showed very intense hyperfine shifted signals of cysteines bound to the N-terminal cluster, and a very weak hyperfine shifted signal at 10 ppm diagnostic of a C-terminal bound cluster, which is indicative of a low [4Fe−4S] cluster occupancy at the C-terminal site ( Figure 3, trace b). This signal was not observed in anaerobically purified wtNUBP1 ( Figure 3, trace a), which confirms that the C-terminal site was not occupied by [4Fe−4S] clusters in the latter sample.
In conclusion, wtNUBP1 binds three [4Fe−4S] 2+ clusters, two, as stable, at each N-terminal CX 13 CX 2 CX 5 C motif, and one at the C-terminal CPXC motif, which is labile and essential for promoting protein dimerization. This behavior is the same as that observed for the yeast homologue of NUBP1, that is, Nbp35. 8 [ The cluster transfer/assembly reaction is relatively fast, as it occurs within 1 h after the mixing of the two proteins. Acidlabile sulfide and iron analysis on NUBP1-C235A/C238A, isolated after incubation with 1 equiv of [2Fe−2S] 2 -GLRX3 2 -GS 4 , showed that NUBP1-C235A/C238A contains 0.5 [4Fe− 4S] clusters per monomer (Table 2), consistent with a [4Fe− 4S] cluster assembly efficiency of ∼60% with respect to chemical reconstitution. The absorbance peaks in the UV−vis spectra of [2Fe−2S] 2 -GLRX3 2 -GS 4 , once incubated with apo NUBP1-C235A/C238A and then isolated, were significantly reduced with respect to the starting material ( Figure 4C). However, their intensities were not completely bleached at any protein ratio, being higher as they went from substoichiometric ratios to the 1:1 ratio ( Figure 4C). This result indicates no complete cluster transfer/assembly reaction. An excess of     Figure S4). At the same time, the intensity of the UV−vis signals of GLRX3, isolated after incubation with NUBP1 C235A/C238A in the presence of 5 mM DTT and separation on Ni-NTA column, is lower than that observed in the absence of DTT ( Figure S4). Overall, these data indicate that, upon DTT addition, more [2Fe −2S] clusters are released from GLRX3 and concomitantly more [4Fe−4S] clusters assembled on NUBP1 C235A/C238A. This effect can be rationalized considering the higher reduction potential of DTT (E pH=8 = −366 mV) 33 than that of GSH (E pH=8 = −299 mV), 34   protein dimerization ( Figure 5C). The cluster transfer/ assembly reaction is relatively fast as it occurs within 1 h after the mixing of the two proteins. Acid-labile sulfide and iron analysis of NUBP1 38−320 , isolated from the 1:0.5 apo NUBP1 38−320 :[2Fe−2S] 2 -GLRX3 2 -GS 4 mixture, showed that NUBP1 38−320 contains ∼0.5 [2Fe−2S] clusters per dimer (Table 2), indicating an incomplete [2Fe−2S] cluster transfer reaction at the [2Fe−2S] 2 -GLRX3 2 -GS 4 stoichiometric amount required to fully saturate NUBP1 38−320 with a [4Fe− 4S] cluster. The observed partial cluster transfer could be possibly due to similar [2Fe−2S] cluster binding affinities between the C-terminal motif of NUBP1 38−320 and the two CGFS motifs of GLRX3. Consistent with the observed partial cluster transfer, the absorbance peaks of the [2Fe−2S] 2+ clusters of [2Fe−2S] 2 -GLRX3 2 -GS 4 , once incubated with NUBP1 38−320 and then separated, were significantly reduced in the UV−vis spectra with respect to the starting material, but their intensities were not completely bleached in the UV−vis spectra at all protein ratios and increased in intensity when moving from substoichiometric ratios to the 1:0.5 ratio ( Figure  5D). Acid-labile sulfide and iron analysis on [2Fe−2S] 2 -  (Table 2). Thus, ∼30% of [2Fe−2S] clusters per dimer were released from [2Fe−2S] 2 -GLRX3 2 -GS 4 at the 1:0.5 ratio, corresponding to ∼0.5 [2Fe−2S] clusters, which is the amount experimentally determined on NUBP1 38−320 isolated from the mixture ( Table  2). This indicates that [2Fe−2S] clusters are not lost in solution during the cluster transfer reaction, but are transferred to NUBP1 38−320 . In conclusion, [2Fe−2S] 2 -GLRX3 2 -GS 4 is able to transfer its [2Fe−2S] 2+ clusters to the C-terminal motif by inducing NUBP1 dimerization, but no formation of a [4Fe−4S] 2+ cluster is observed, even when 5 mM DTT was added to the mixture ( Figure S4). Different mixtures of wtNUBP1 and [2Fe−2S] 2 -GLRX3 2 -GS 4 (0.50, 1.0, and 1.5 equiv of [2Fe−2S] 2 -GLRX3 2 -GS 4 per monomerized wtNUBP1, with 1.5 equiv being the stoichiometric amount required to form three [4Fe−4S] clusters on dimeric wtNUBP1) were then analyzed following the same approach as that used for the two mutants, and similar conclusions were drawn. Indeed, the UV−vis and paramagnetic 1D 1 H NMR spectra indicated a stepwise and relatively fast assembly of a [4Fe−4S] 2+ cluster bound to the N-terminal motif of wtNUBP1 ( Figure 6A and B). Analytical gel filtration showed that cluster transfer induces protein dimerization ( Figure 6C), indicating that Fe/S cluster binding also occurs at the C-terminal motif of wtNUBP1, that is, the motif exclusively found to promote NUBP1 dimerization. UV−vis spectra collected on [2Fe−2S] 2 -GLRX3 2 -GS 4 , once incubated and separated from wtNUBP1 ( Figure 6D), as well as acid-labile sulfide and iron analysis on wtNUBP1 and [2Fe− 2S] 2 -GLRX3 2 -GS 4 indicated an incomplete cluster transfer/ assembly reaction; that is, ∼0.6 [4Fe−4S] clusters are bound to dimeric wtNUBP1 (Table 2) consistent with a [4Fe−4S] cluster assembly efficiency of ∼50% with respect to chemically reconstituted wtNUBP1. [4Fe−4S] 2+ cluster assembly is more efficient in the presence of 5 mM DTT, a condition that gives   2+ cluster is assembled on the C-terminal motif of wtNUBP1, but, because it is not stably bound to it, it is lost upon protein isolation. The formation of a [4Fe−4S] 2+ cluster on the C-terminal motif is confirmed by the paramagnetic 1D 1 H NMR spectrum of wtNUBP1, isolated from the wtNUBP1 and [2Fe−2S] 2 -GLRX3 2 -GS 4 final mixture. Indeed, a paramagnetic signal having the same chemical shift, that is, 10 ppm ( Figure 6B), and intensity as that present in the paramagnetic 1D 1 H NMR spectrum of the chemically reconstituted [4Fe−4S] 2+ NUBP1 38−320 (Figure 3, trace d), which exclusively contains a [4Fe−4S] cluster bound at the C-terminal motif, is detected. Moreover, the intensity of this signal is much lower than that of the signals due to the N-terminal [4Fe−4S] bound cluster ( Figure 6B) To investigate the role of GSH as the reductant promoting the [4Fe−4S] 2+ cluster assembly on wtNUBP1, we performed the reaction between [2Fe−2S] 2 -GLRX3 2 -GS 4 and monomerized apo wtNUBP1 at the 1.5:1 stoichiometric ratio using different GSH concentrations (0, 1, 5, and 10 mM). The UV− vis spectrum, collected on wtNUBP1 isolated from the reaction mixture at 0 mM GSH, showed bands at 320, 420, 510, and 580 nm ( Figure S6), which are typical of a [2Fe−2S] 2+ cluster. By addition of up to 10 mM GSH, changes in the UV−vis spectra of wtNUBP1 isolated from the mixture were observed: the bands at 320, 510, and 580 nm decreased in intensity, and a broad unresolved band at 410 nm increased in intensity ( Figure S6). These changes indicated that (i) [2Fe−2S] 2+ clusters are transferred from GLRX3 to wtNUBP1 forming [2Fe−2S] 2+ -bound wtNUBP1 species in the absence of GSH; and (ii) the latter species is then transformed into [4Fe−4S] 2+bound wtNUBP1 species upon the addition of increasing amounts of the GSH reductant.

■ DISCUSSION
In this work, we showed that [2Fe−2S] 2 -GLRX3 2 -GS 4 is able to transfer its [2Fe−2S] 2+ cargo to NUBP1 in its monomeric apo form (Figure 7). Although a dimeric apo state of NUBP1 was purified from E. coli cells, we showed that this species was induced by the presence of adventitious metal ion(s) bridging two NUBP1 molecules. Thus, it is likely that the monomeric apo NUBP1 is a potential physiologically relevant species accepting [2Fe−2S] clusters from GLRX3. Our data showed that cluster transfer from [2Fe−2S] 2 -GLRX3 2 -GS 4 to monomeric apo NUBP1 occurs on both N-terminal and C-terminal motifs, with GLRX3 thus acting as a [2Fe−2S] cluster chaperone for both cluster binding sites. The proposed chaperone function is in agreement with the functional data available on yeast Nbp35. Indeed, it was observed that depletion of Nbp35 resulted in an increase of iron level in Grx3/4, 16 as, when the [2Fe−2S] cluster from Grx3/4 cannot be transferred anymore to Nbp35, iron accumulates on the Grx3/4 proteins as a [2Fe−2S] cluster. A similar effect has been also observed in the ISC assembly machinery. Depletion of the ISC components Ssq1, Jac1, and Grx5, which are involved in the transfer of the [2Fe−2S] cluster from Isu1 to target proteins, leads to an accumulation of iron on Isu1, likely in the form of a [2Fe−2S] cluster. 37 The role of GLRX3 in chaperoning [2Fe−2S] clusters to NUBP1 is also supported by other in vivo findings in yeast, showing that Nbp35 was dispensable for iron incorporation into Cfd1, and vice versa, 8,9 suggesting that [4Fe−4S] cluster assembly on Cfd1 or Nbp35 does not depend on the other P-loop NTPase partner, but depends on Grx3/4 only. 16 The kinetic lability observed for the C-terminal [4Fe−4S] 2+ cluster with respect to the N-terminal [4Fe-4S] 2+ cluster suggests a distinct functional role of the two types of clusters bound to NUBP1. As NUBP1 is involved in the formation and distribution of [4Fe−4S] clusters, 10,11 we could speculate that the labile [4Fe−4S] cluster assembled at the C-terminal site is transferred to apo target proteins, this step being facilitated by several other assembly proteins of the CIA machinery. 5 On the contrary, the N-terminal [4Fe−4S] 2+ cluster might play a structural role or an electron transfer role (see later). Different roles for the two types of [4Fe−4S] clusters bound to NUBP1/Nbp35 were also supported by experimental data in yeast cells. Indeed, when yeast cells were labeled with 55 Fe followed by exposure to a cell-permeant chelator, ∼50% of the 55 Fe bound to Nbp35 (yeast homologous of NUBP1) was rapidly lost, whereas a nearly equal portion of 55 Fe was more stably associated with Nbp35 for as long as 3 h in the presence of the chelator. 38 We also showed that two [2Fe−2S] 2+ clusters donated by GLRX3 were assembled to form a [4Fe−4S] 2+ cluster at the Nterminal motif of NUBP1 and that its formation is independent of the presence of the C-terminal cluster motif (Figure 7). Generation of a [4Fe−4S] 2+ cluster requires two electrons for reductively coupling the two [2Fe−2S] 2+ clusters donated by GLRX3. In our in vitro experiments, we showed that the [2Fe−2S] 2+ cluster coupling reaction is driven by the presence of GSH (Figure 7). We showed, indeed, that GSH was strictly required to assemble a [4Fe−4S] 2+ cluster at the N-terminal motif, as otherwise a [2Fe−2S] 2+ NUBP1 species was obtained in its absence. The cluster transfer/assembly process is, however, not complete in the presence of GSH and is more efficient when a reducing agent stronger than GSH, that is, DTT, although not physiological, was also present. In vivo the two required electrons could be supplied to the N-terminal site by the electron transfer chain formed by the NDOR1 and anamorsin CIA components (Figure7). 39,40 Indeed, it has been demonstrated that the yeast homologues of NDOR1 and anamorsin, that is, Tah18 and Dre2, respectively, are specifically required for the insertion of the [4Fe−4S] cluster into the N-terminal motif of Nbp35, and human NDOR1anamorsin can functionally replace yeast Tah18-Dre2. 9 In support of the presence of the electron transfer chain assembling a [4Fe−4S] 2+ cluster on the N-terminal motif of NUBP1, it was shown that a specific interaction between the anamorsin and NUBP1 homologues occurs in plants. 41,42 In addition to the NDOR1-anamorsin electron transfer chain driving [4Fe−4S] assembly on the N-terminal motif of NUBP1, our data suggested that GSH, which is largely abundant in the cytosol, and determines a low half-cell reduction potential ranging from −200 to −240 mV, 34 might promote the cluster coupling reaction in vivo. The observed incomplete [4Fe−4S] cluster assembly driven by GSH reductant might be determined by its higher redox potential with respect to that provided by NDOR1-anamorsin electron transfer chain. Indeed, a redox potential of −415 mV, much lower than that of GSH, was required to reduce the [2Fe−2S] cluster of Dre2, 9 which is the terminus of the Tah18−Dre2 electron transfer chain and thus the redox center expected to provide the electrons to reductively couple two [2Fe−2S] 2+ clusters into a [4Fe−4S] 2+ cluster on the N-terminal site.
A GSH-driven assembly of a [4Fe−4S] 2+ cluster via a reductive [2Fe−2S] cluster coupling process also occurs at the C-terminal motif of NUBP1 (Figure 7), although the lability of the cluster binding to the C-terminal motif determines a low [4Fe−4S] cluster occupancy at that site. In the latter reaction, cluster binding to the C-terminal motif of NUBP1 promotes protein dimerization, at variance with cluster binding to the Nterminal motif, which, indeed, does not affect the quaternary structure of NUBP1 (Figure 7). However, at variance with what is observed in the formation of the [4Fe−4S] cluster at the N-terminal motif, we found that [2Fe−2S] 2+ clusters were not converted in a [4Fe−4S] 2+ cluster at the C-terminal motif once the N-terminal motif was absent. This suggests a cooperativity effect between the N-and the C-terminal motifs in the formation of a [4Fe−4S] cluster at the C-terminal motif, but not in the formation of a [4Fe−4S] cluster at the Nterminal motif. It is possible that the presence of a [4Fe−4S] cluster bound at the N-terminal site is required to drive [4Fe− 4S] cluster formation at the C-terminal site, but not vice versa. A possible mechanism might require that the [4Fe−4S] cluster bound at the N-terminal site would receive the electrons from the NDOR1-anamorsin electron transfer chain, and then transfer them to the C-terminal site, thus assembling a [4Fe− 4S] cluster on it. In such a case, the [4Fe−4S] 2+ N-terminal cluster is therefore playing an electron transfer role instead of a structural role. In yeast, it was proposed that the Tah18−Dre2 electron transfer chain is not required for the insertion of the [4Fe−4S] cluster into the C-terminal motif. 9 This interpretation was based on the fact that depletion of Tah18 and Dre2 does not abolish iron incorporation on the C-terminal motif of Cfd1, but even increases iron incorporation on Cfd1. However, we have now observed that [2Fe−2S] 2 -GLRX3 2 -GS 4 can transfer a [2Fe −2S] cluster to the C-terminal motif of NUPB1, which stably binds it without releasing it in solution. Thus, iron accumulation on Cfd1 upon depletion of Tah18 and Dre2 9 can be now interpreted as a consequence of the lack of the electron transfer chain required for assembling a [4Fe−4S] cluster on the C-terminal motif. According to our data, Grx3/4 proteins would transfer the [2Fe −2S]