Deciphering the genetic determinants for aerobic nicotinic acid degradation: The nic cluster from Pseudomonas putida KT2440

The nic Genes Are Responsible for NA Catabolism in P. putida KT2440.

The in silico analysis of the P. putida KT2440 genome revealed a gene cluster, named nicTPFEDCXRAB (Fig. 1), some of whose gene products showed significant similarities to proteins involved in the catabolism of N-heterocyclic aromatic compounds [supporting information (SI) Table S1]. When P. putida KT2440 was tested with different N-heterocycles (pyridoxal, pyridoxamine, pyridine, 2- and 3-hydroxypyridine, NA, isonicotinic acid, picolinic acid, quinoline, and isoquinoline) as sole carbon and energy sources, we observed growth only in NA (Fig. 1C), thus implicating the nic cluster in NA catabolism. To confirm this assumption, several P. putida KT2440 knockout mutants were constructed (Table S2). Disruption of the nicA, nicB, nicC, nicD, and nicX genes did not allow the corresponding P. putida KT2440 mutant strains to grow in NA as the sole carbon source (Fig. 1C). The nic cluster also contained two genes, nicP and nicT, encoding a potential porin and permease transport system, respectively, and a regulatory gene (nicR) (Fig. 1 and Table S1; see also SI Text).

Growth in NA could be restored when the P. putida knockout mutants were complemented with a broad-host-range plasmid (pNIC) harboring a 14-kb DNA cassette containing the complete nic cluster (Table S2). Interestingly, the pNIC plasmid also conferred the ability to use NA as sole carbon source to other bacteria, such as Pseudomonas sp. MT14 (Table S2), that are unable to degrade this aromatic compound (Fig. 1C). All of these results demonstrate that the nic cluster identified in P. putida KT2440 constitutes the first complete set of genes reported so far in any organism that is responsible for aerobic NA degradation.

The nicAB Genes Encode the NA Hydroxylase.

The observation that the disruption of the nicA and nicB genes abolished growth of the corresponding P. putida KT2440 mutants in NA but did not affect their growth in 6HNA (data not shown), as well as the fact that the nicA and nicB gene products show significant sequence similarity with the subunits of some members of the xanthine dehydrogenase family (Table S1), suggested that these two genes might encode the NA hydroxylase from P. putida KT2440 (Fig. 1). To confirm the role of the nicA and nicB genes, they were cloned and expressed in plasmid pNicAB (Table S2). When plasmid pNicAB was transferred to Pseudomonas strains that are not able to oxidize NA, such as Pseudomonas fluorescens R2f and Pseudomonas sp. DSM6412 (Table S2), the recombinant strains acquired the ability to efficiently convert NA into a stoichiometric amount of 6HNA (Fig. S1). NA hydroxylase assays using crude extracts from P. fluorescens R2f (pNicAB) revealed that the activity increased significantly when an external electron acceptor like phenazine methosulfate (PMS) was added and decreased below detection levels under anaerobic conditions (Fig. 2A). The hydroxylase activity was maximal at 30°C and pH 7.5, and it was highly specific for NA (Table S3). These data indicate that genes nicAB encode the two-component NA hydroxylase that converts NA into 6HNA, constituting the first aerobic NA hydroxylase whose primary structure is known.

Hydroxylases of N-heterocyclic aromatic compounds incorporate oxygen derived from a water molecule into the product, and they typically contain a redox center, formed by a molybdenum ion coordinated to an organic cofactor (molybdopterin cytosine dinucleotide, MCD), two [2Fe-2S] clusters, and usually FAD (Fig. 2B), which transports electrons from the reducing substrate (N-heterocyclic compound) to the oxidizing substrate (the electron acceptor) (5, 10). The NA hydroxylase small subunit (NicA, 23.8 kDa) harbors the 41-C-X₄-C-G-X-C-X_n-C-59 and 100-C-G-X-C-X₃₁-C-X-C-137 conserved motifs likely involved in binding of the two [2Fe-2S] clusters that are located in the small subunit of other multicomponent xanthine dehydrogenases (5, 10) (Fig. 2B). However, analysis of the amino acid sequence of the NA hydroxylase large subunit (NicB, 128 kDa) revealed some unique features not reported yet in other hydroxylases of N-heterocyclic aromatic compounds. Thus, the N-terminal region of NicB shows a molecular architecture similar to that of the IorB subunit of isoquinoline oxidoreductase, containing the active site and the MCD binding sites in an arrangement (MPT2-MPT1-MPT3) that differs from that observed (MPT1-MPT2-MPT3) in most members of the xanthine dehydrogenase family (5, 10) (Fig. 2B). The existence of a MCD cofactor in NicB and the fact that Escherichia coli cannot synthesize this particular cofactor (11) explain the lack of NA hydroxylase activity in E. coli (pNicAB) cells (data not shown). On the other hand, the C-terminal region of NicB (residues 750-1187) contains three conserved cytochrome c (CytC) heme-binding motifs (817-CAVCH-823, 963-CTACH-969, and 1087-CLGCH-1093) (12) (Fig. 2B) that have not been described previously in any member of the xanthine dehydrogenase/oxidase family. The involvement of a CytC mediating the electron transfer from the intramolecular redox center to the final acceptor (oxygen) would explain the absence in NicB of the FAD-binding site usually present in enzymes of the xanthine-dehydrogenase family (5, 10) (Fig. 2B) and the observation that the basal NA hydroxylase activity could not be increased by adding external electron acceptors such as the organic cofactors NADP, NAD, FMN, and FAD or the inorganic ions Fe³⁺ and NO₃⁻ (data not shown).

When a truncated NicB subunit lacking the CytC domain was constructed and expressed from plasmid pNicABtr, the NicABtr truncated enzyme did not show NA hydroxylase activity (Fig. 2A). However, NicABtr acquired significant activity when PMS was added as a terminal electron acceptor (Fig. 2A). The NA hydroxylase assay could be also performed when the NicAB and NicABtr enzymes were loaded in a nondenaturing PAGE, and two bands of the expected molecular size for each heterodimer were observed (Fig. 2C). Nevertheless, whereas the band corresponding to the wild-type NicAB enzyme could be also detected by a heme-staining assay, the corresponding band of the NicABtr enzyme was not detected (Fig. 2C), which agrees with the predicted absence of heme groups in the truncated protein. The NA hydroxylase activity was partially restored (NicABtr) and increased (NicAB) when these enzymes were complemented in trans with the CytC domain of NicB cloned and expressed in plasmid pVLTCytC (Fig. 2A), which strongly suggests that this region behaves indeed as a functional CytC domain. On the other hand, it was observed that the NA hydroxylase activity was inhibited in the presence of KCN, but the addition of antimycin A had no effect (Fig. 2A). As expected, neither KCN nor antimycin A inhibited the NA hydroxylase activity when PMS was used as terminal electron acceptor (Fig. 2A). Because it is well known that the CytC oxidase is inhibited by KCN but not by antimycin A (12), it is likely that the CytC domain of NA hydroxylase may transfer the electrons from the redox center to a CytC oxidase that uses molecular oxygen as final electron acceptor.

In summary, all of these data show that the two-component NicAB protein constitutes the prototype of a previously uncharacterized class of hydroxylases within the xanthine dehydrogenase family whose modular architecture includes a CytC domain as component of the electron transport chain to molecular oxygen. Because the NA hydroxylase is used in industry for the formation of 6HNA, the key precursor of several insecticides (e.g., imidachloprid) and herbicides (e.g., imazapyr) (3), its cloning and successful heterologous expression reported here are of biotechnological interest, revealing also that P. fluorescens R2f and Pseudomonas sp. DSM6412 can be suitable hosts for the functional expression of other enzymes that require the unusual MCD cofactor for their activity.

The nicC Gene Encodes the 6HNA Monooxygenase.

The second enzymatic step in the maleamate pathway for the aerobic NA degradation is the oxidative decarboxylation of 6HNA to render 2,5DHP (6) (Fig. 1). This reaction is catalyzed by a 6HNA-3-monooxygenase (13), which shows 60% amino acid sequence identity with the product of the nicC gene (Table S1). The role of the nicC gene product as the 6HNA monooxygenase was in agreement with the observation that P. putida KT2440dnicC resting cells showed a complete conversion of NA to 6HNA (Fig. S1). On the other hand, when the nicC gene was cloned and expressed in plasmid pIZNicC (Tables S2 and S4), P. fluorescens R2f (pIZNicC) cell extracts revealed a 42-kDa protein that catalyzed the conversion of 6HNA to 2,5DHP when NADH and FAD were added to the assay (data not shown). These data indicate, therefore, that NicC is the 6HNA monooxygenase of the NA degradation pathway in P. putida KT2440.

NicX Is the First Member of a New Family of Extradiol Dioxygenases.

The third step in the aerobic NA degradation pathway is the one catalyzed by a 2,5DHP dioxygenase (Fig. 1) (6). Although the 2,5DHP dioxygenase activity had been characterized many years ago and ¹⁸O₂ incorporation experiments confirmed that the pyridine ring was opened by insertion of both atoms of ¹⁸O₂ (14, 15), the gene encoding the corresponding enzyme and the primary structure of the protein remained unknown. Although amino acid sequence analysis of the nic gene products did not reveal the presence of any putative ring-cleavage dioxygenase (Table S1), the P. putida KT2440dnicX mutant strain was shown to lack 2,5DHP dioxygenase activity (data not shown), pointing out that the nicX gene might encode such activity. To test this hypothesis, the nicX gene was overexpressed in E. coli BL21(DE3) cells containing plasmid pETNicX (Table S2). Cell extracts showed 2,5DHP dioxygenase activity (6 μmol·min⁻¹·mg⁻¹), revealing that the nicX gene product was the ring-cleavage enzyme in the NA degradation pathway from P. putida KT2440.

The NicX protein was purified and analyzed by SDS/PAGE, showing a band whose apparent molecular mass (39 kDa) (Fig. 3D) corresponded with that predicted for the nicX gene product. Analysis by inductively coupled plasma emission spectroscopy (ICP) of the metal content of purified NicX indicated the absence of any metal; however, after Fe²⁺-dependent reactivation of the enzyme (25 μM), 19.8 μM iron was detected, indicating that the ratio of moles of iron to moles of a monomer of NicX was near 1.0. Some biochemical properties of NicX (see SI Text), e.g., 1 mol of oxygen consumption per mole of 2,5DHP, and a strict substrate specificity for 2,5DHP with K_m and V_max values of 70 μM and 2.3 μM·min⁻¹·mg⁻¹, respectively, resemble those reported for the 2,5DHP dioxygenase from P. putida N9 (14).

The nature of the product generated by the 2,5DHP dioxygenase is controversial. Initial studies revealed that 2,5DHP ring cleavage was performed between C5 and C6, suggesting that N-formylmaleamic acid (NFM) was the actual ring-cleavage product, although it could not be detected (6). Further work suggested that the hydrolytic cleavage of NFM to form maleamate was also catalyzed by the 2,5DHP dioxygenase enzyme (15, 16). Therefore, to determine unequivocally the chemical structure of the product generated by the 2,5DHP dioxygenase, a combination of NMR spectroscopy and mass spectrometry was used to follow the course of the reaction catalyzed by purified NicX. The reaction of 2,5DHP with the dioxygenase enzyme revealed the complete transformation of this heterocyclic compound into a single opening product whose ¹H NMR and ¹³C NMR spectra are consistent with NFM as the ring-cleavage product (Fig. S2). On the other hand, negative mass spectrometry (electrospray) experiments revealed that the [M-H]⁻ peak at a m/z value of 110 due to 2,5DHP completely disappeared upon treatment with NicX, and a new major peak at a m/z value of 142 was observed (Fig. S2). The molecular ion at the m/z value of 142 corresponds to the ring-cleavage product of 2,5DHP by addition of two oxygen atoms generating the corresponding N-formyl derivative, i.e., NFM. Therefore, all of these data show that NicX catalyzes the dioxygenolytic ring cleavage of 2,5DHP between carbons 5 and 6 generating NFM as final product. In light of this finding, the previously suggested dioxygenase-mediated decarboxylation of NFM to form maleamate and formate (15, 16) can be now explained as a result of a deformylase activity (see below) that was contaminating the purified 2,5DHP dioxygenase enzyme.

As mentioned above, the amino acid sequence of NicX does not show any significant similarity to that of other ring-cleavage dioxygenases described so far. Instead, NicX shows amino acid sequence similarity to members of the MQ family of proteins such as the aminopeptidases AmpS and AmpT from Staphylococcus aureus (17) and Thermus thermophilus (18), respectively. Since the three-dimensional structures of AmpS and AmpT have been recently determined (17, 18), homology modeling for NicX was accomplished. According to this model, the protein is composed of two globular domains (Fig. 3A). The N-terminal domain (residues 1–123 and 177–191) is predicted to have a Rossmann fold with a central six-stranded mixed β-sheet (β₂-β₁-β₃-β₄-β₉-β₈) flanked by α-helices (α₁, α₂, α₃, α₄, α₅, and α₆) on both sides. The C-terminal domain is organized into two antiparallel-β-sheets (β₂₁-β₂₀-β₁₉-β₅-β₆-β₁₁-β₁₄-β₁₅ and β₇-β₁₀-β₁₆-β₁₇-β₁₈-β₁₂-β₁₃) that form a pseudo β-barrel capped by the α₇ and α₈ helices (Fig. 3A). The active site of AmpS contains two high-affinity Co(II) binding sites, i.e., Co₁ constituted by residues E253 and H348 and Co₂ formed by residues E319, H381, and D383 (17). According to the structural model of NicX, residues corresponding to the first Co(II) binding site are F194 and L300, indicating that the Co₁ site is lost in this protein. However, residues H265, H318, and D320 located on the surface of one of the β-sheets at the C-terminal domain may constitute the iron coordination sphere in NicX replacing the equivalent Co₂ site of AmpS (Fig. 3B). Despite the phylogenetic diversity of extradiol ring-cleavage dioxygenases, the iron (II) binding motif constituted by two histidines and one carboxylate group is a common feature of this family of enzymes (19). In this sense, the modeled active site of NicX fits perfectly the geometry of the canonical 2-His-1-carboxylate metal binding motif found in classical extradiol ring-cleavage dioxygenases (19). Crucial side-chain atoms coordinating the metal are in the same spatial position despite coming from different local architectures (Fig. 3C). It is worth noting that although 2,5DHP is not a canonical peptidic substrate like those recognized by the aminopeptidases AmpS and AmpT, it may acquire a peptidic-like structure with an amide bond (5-hydroxy-2-pyridone) as a result of its keto-enolic tautomerization in the active site of the ring-cleavage enzyme. This fact might explain the similarity found between the active site of NicX and that of such aminopeptidases. In summary, NicX is proposed as the founding member of a new family of extradiol ring-cleavage dioxygenases that evolved from the MQ family of metalloproteases with partial replacement of active site residues to obtain new facial catalytic triad.

The nicD Gene Encodes a N-Formylmaleamate Deformylase.

We have shown above that the 2,5DHP dioxygenase generates NFM as final product. To check whether nicD was responsible for the required deformylase activity, this gene was overexpressed from plasmid pETNicD (Table S2). SDS/PAGE analysis of crude extracts from E. coli BL21(DE3) cells harboring plasmid pETNicD showed the presence of an intense band corresponding to a protein of 29 kDa (data not shown), in good agreement with that predicted for the product of the nicD gene (29.1 kDa). Interestingly, these extracts were able to transform NFM to maleamic and formic acids as shown by HPLC and ¹H NMR analyses (Fig. S3). These data reveal that deformylation of NFM to generate formic and maleamic acid is catalyzed by the NicD enzyme in a reaction that was not outlined in the proposed NA degradation pathway (Fig. 1).

The primary structure of NicD shows significant similarity to that of members of the α/β-hydrolase-fold superfamily of enzymes (Table S1 and Fig. S4). A homology model was proposed for the structure of NicD based on the known three-dimensional structure of the lactamase from Aureobacterium sp. (20) (Fig. S5). The α/β-hydrolases share a conserved catalytic triad formed by a nucleophile residue (usually Ser), an acidic residue, and a histidine (21). The sequence 98-G-H-S-M-G-104 in NicD perfectly matches the consensus motif for the nucleophile residue (Fig. S4). The conserved histidine in α/β-hydrolases is located in a loop after the β₈ strand, and it corresponds to H245 in NicD (Figs. S4 and S5). The acidic residue in α/β-hydrolases is generally placed in a loop that follows the β₇ strand; however, a multiple sequence alignment revealed that the corresponding position is occupied by residue G217 in NicD (Fig. S4), being the two closest acidic residues, E221 and D224, in an α-helix located outside the predicted active site pocket (Fig. S5). However, a detailed analysis of the predicted 3D structure for NicD revealed that another acid residue, D125, located after the β₆ chain faces the putative active site (Fig. S5). Interestingly, this position is equivalent to that found for the acidic residue in some α/β-hydrolases (21) (Fig. S4).

To assess the proposed catalytic triad of NicD, the residues S101, D125, H245, and E221 were substituted through site-directed mutagenesis by Ala residues. The NicD mutant proteins were overexpressed as the wild-type protein, and their specific activities were compared. Whereas substitutions S101A, D125A, and H245A lead to a complete loss of the deformylase activity, the NicD mutant protein harboring the substitution E221A retained ≈70% of the wild-type activity (Fig. S5). Thus, these data suggest that residues S101, D125, and H245 are essential for the enzyme activity, constituting the catalytic triad of the NicD deformylase.

Although some members of the α/β-hydrolase superfamily are able to break a C–N linkage in a peptidic bond, e.g., serine-carboxypeptidase and proline iminopeptidase, no deformylase activity had been reported so far within this superfamily (21). The NicD protein, therefore, expands further the wide functional diversity associated to the α/β-hydrolase common fold.

The nicF and nicE Gene Products Catalyze the Last Two Steps of the Maleamate Pathway.

The primary structure of the nicF gene product is similar to that of members of the isochorismatase superfamily such as the N-carbamoylsarcosine amidohydrolase from Arthrobacter sp. and the pyrazinamidase (nicotinamidase) from Pyrococcus horikoshii (22) (Table S1), two enzymes that catalyze the substitution of the amine group in an amide bond by a hydroxyl group. The catalytic mechanism of these proteins involves a cysteine residue as nucleophile, an aspartic acid, and a lysine (22). Because these three residues were conserved in the primary structure of NicF (D31, K121, and C154), it was tempting to speculate that this enzyme is the maleamate amidase that transforms maleamic acid into maleic acid plus ammonia in the NA degradation pathway (Fig. 1), thus explaining the growth of P. putida KT2440 when cultivated in NA as sole nitrogen source. To confirm the predicted role for NicF, the nicF gene was overexpressed from plasmid pETNicF (Table S2). Crude extracts of E. coli BL21 (DE3) (pETNicF) cells showed maleamate amidohydrolase activity (19.3 μmol·min⁻¹·mg⁻¹), indicating that the nicF gene encodes the maleamate amidohydrolase of the NA degradation pathway in P. putida KT2440. Although another amidohydrolase has been described in the anaerobic NA degradation pathway (4), this enamidase recognizes a different substrate (1,4,5,6-tetrahydro-6-oxonicotinate), and it does not show any amino acid sequence similarity to NicF, indicating that these two enzymes belong to different protein families.

The maleamate pathway ends with the isomerization of maleic acid to fumaric acid, a Krebs cycle intermediate (6) (Fig. 1). Because the nicE gene product shows a significant amino acid sequence similarity with a previously characterized maleate cis/trans-isomerase (Table S1) and contains the two conserved Cys residues (C104 and C222 in NicE) involved in the catalytic activity (23), the nicE gene was assigned as that encoding the last step of the aerobic NA degradation pathway in P. putida KT2440.

Distribution of the nic Genes in Other Bacteria and Evolutionary Considerations.

Within the Pseudomonas genus the nic genes are present only in strains of P. putida, e.g., KT2440, F1, GB-1, and W619, where they show high sequence identity (>95%) and the same organization. Although a few P. fluorescens strains (7, 24) have been reported to use NA, it cannot be ruled out that these strains are not properly classified and/or that they harbor a mobile genetic element, such as the NIC plasmid from a P. putida strain (25), that confers the ability to degrade NA. Outside the Pseudomonas genus, orthologs of the nic genes were found only in the genomes of β-proteobacteria from the Burkholderiales family (Fig. 1B). In most β-Proteobacteria the nicB2 and nicB1 genes encode proteins that correspond to N-terminal (first 740 residues) and C-terminal (last 480 residues) regions of the nicB gene product from Pseudomonas, respectively (Fig. 1B). This observation suggests that the large subunit (NicB) of the NA hydroxylase from Pseudomonas has a modular architecture of two domains that have evolved as two different proteins in β-Proteobacteria: (i) the N-terminal domain (NicB2 protein) would be involved in binding to the substrate and MCD cofactor, and (ii) the C-terminal CytC domain (NicB1 protein) would participate in the electron transfer to the CytC oxidase. Nevertheless, the nicAB genes from Burkholderia xenovorans constitute an exception because they retain the same molecular architecture as the corresponding nicAB genes from P. putida and they are located not within the nic cluster (Fig. 1B) but in an additional cluster predicted to be involved in an alternative NA degradation pathway via 1,4,5,6,-tetrahydro-6-oxonicotinate that is also found in some α-Proteobacteria (4). Most nic clusters harbor a gene (nicR) that encodes a MarR-type transcriptional regulator and a putative NA transport system of the MFS superfamily (nicT) or an ABC transporter (nicT1-T4) (Fig. 1B).

Because plants are the principal source of NA in the environment, it was not surprising to find nic genes in soil bacteria that are able to colonize the plant rhizosphere, such as P. putida, Burkholderia, or Ralstonia strains. On the contrary, it was unexpected to find conserved nic clusters in pathogens such as Bordetella spp. that are able to colonize mammalian respiratory tract. The presence of the nic genes in the genome of several Bordetella spp. provides some clues to the ability of these strains to metabolize NA and to control the NA levels that regulate their infectiveness (26).

Concluding Remarks.

The complete set of genes involved in aerobic NA degradation via the maleamate pathway has been studied at the molecular level and successfully expressed in heterologous hosts. Some unique proteins, such as the two-component NA hydroxylase and the 2,5DHP dioxygenase, constitute the founding members of previously undescribed classes of enzymes. These results provide also a framework for developing new biotechnological processes that use N-heterocyclic aromatic compounds, such as 6HNA and 2,5DHP, as building blocks, and for detoxification of nicotine in tobacco wastes (3, 27). The nic genes from P. putida are also useful to explore the existence of homologous gene clusters in saprophytic bacteria and some animal pathogens where they might stimulate studies on their role in virulence.

Bacterial Strains, Plasmids, Growth Conditions, and Molecular Biology.

The bacterial strains and plasmids used in this study are listed in Table S2. The construction of recombinant plasmids and P. putida KT2440 mutants harboring disrupted nic genes is detailed in SI Text. Bacteria were grown in LB medium (28) at 37°C (E. coli) or 30°C (Pseudomonas). Pseudomonas strains were also grown at 30°C in M63 or MC minimal media (29) containing citrate (0.2%), NA (5 mM), or 6HNA (5 mM). Standard molecular biology techniques were performed as previously described (28).

Enzymatic Assays.

NA hydroxylase and 6HNA 3-monooxygenase were assayed in resting cells and cell extracts as detailed in SI Text. The 2,5DHP dioxygenase (NicX) was assayed spectrophotometrically by measuring the disappearance of 200 μM 2,5DHP at 320 nm (ε = 5,200 M⁻¹·cm⁻¹) as previously described (14). Because purified NicX is inactive, the protein (5 μg) was reactivated by incubation with 10 mM DTT and 0.5 mM FeSO₄ for 15 min at 25°C. Kinetic parameters and the substrate specificity were calculated by measuring the substrate-dependent oxygen consumption rate of the reaction mixture at 25°C using a DW1 Hansa-Tech Oxygen Electrode. NFM deformylase activity (NicD) was measured by coupling deformylase activity to a commercial formate dehydrogenase enzyme from Candida boidinii (Sigma) and monitoring NADH formation as indicated by the manufacturer. The two products of the NicD-catalyzed reaction, i.e., formate and maleate, were identified by HPLC and RMN analyses. Maleamate amidohydrolase (NicF) was assayed by coupling NH₄⁺ formation to a commercial glutamate dehydrogenase (Sigma) and by measuring NADPH oxidation at 340 nm (ε = 6,220 M⁻¹·cm⁻¹), in the presence of α-ketoglutarate.

Other Techniques.

The overproduction and purification of NicX were carried out as indicated in SI Text. Proteins were analyzed by SDS/PAGE and Coomassie-stained as described previously (28). The protein concentration in cell extracts was determined by the method of Bradford by using BSA as the standard (29). NA hydroxylase activity staining and heme staining in nondenaturing PAGE were performed with nitroblue tetrazolium and dimethoxybenzidine dihydrocholoride, respectively, as previously described (7, 30). Fe²⁺ was determined by ICP using a ICP-OES Optima 2000DV equipment (PerkinElmer). The different techniques used for metabolites detection are detailed in SI Text. Three-dimensional modeling of NicX and NicD was performed as indicated in SI Text, and figures were generated with PyMOL (DeLano Scientific).