Structure and function of the hypochlorous acid–induced flavoprotein RclA from Escherichia coli

In response to microbial invasion, the animal immune system generates hypochlorous acid (HOCl) that kills microorganisms in the oxidative burst. HOCl toxicity is amplified in the phagosome through import of the copper cation (Cu2+). In Escherichia coli and Salmonella, the transcriptional regulator RclR senses HOCl stress and induces expression of the RclA, -B, and -C proteins involved in bacterial defenses against oxidative stress. However, the structures and biochemical roles of the Rcl proteins remain to be elucidated. In this study, we first examined the role of the flavoprotein disulfide reductase (FDR) RclA in the survival of Salmonella in macrophage phagosomes, finding that RclA promotes Salmonella survival in macrophage vacuoles containing sublethal HOCl levels. To clarify the molecular mechanism, we determined the crystal structure of RclA from E. coli at 2.9 Å resolution. This analysis revealed that the structure of homodimeric RclA is similar to those of typical FDRs, exhibiting two conserved cysteine residues near the flavin ring of the cofactor flavin adenine dinucleotide (FAD). Of note, we observed that Cu2+ accelerated RclA-mediated oxidation of NADH, leading to a lowering of oxygen levels in vitro. Compared with the RclA WT enzyme, substitution of the conserved cysteine residues lowered the specificity to Cu2+ or substantially increased the production of superoxide anion in the absence of Cu2+. We conclude that RclA-mediated lowering of oxygen levels could contribute to the inhibition of oxidative bursts in phagosomes. Our study sheds light on the molecular basis for how bacteria can survive HOCl stress in macrophages.

To kill invading bacteria, macrophages and neutrophils in the immune system employ myeloperoxidase to generate HOCl from H 2 O 2 , which is produced by the oxidative burst from molecular oxygen by NADPH oxidase (4). HOCl directly oxidizes amino acids containing amines and sulfurs and can cleave DNA once it is converted into a hydroxyl radical in the presence of iron or copper ions via Fenton(-like) reaction (5)(6)(7)(8).
HypT, RclR, and NemR are HOCl-responsive transcriptional factors in bacteria that increase bacterial survival under HOCl stress (9 -11). HypT, belonging to the LysR-type transcriptional regulators, employs a HOCl-sensitive methionine residue to repress genes related to the import of iron to the cytosol in Escherichia coli and Salmonella enterica serovar Typhimurium. HypT plays a critical role in the survival of Salmonella in macrophages (12). NemR of E. coli is a redox-regulated transcriptional repressor that utilizes the oxidation status of HOClsensitive cysteine residues. NemR expresses two enzymes, GloA and NemA, to detoxify methylglyoxal and reactive electrophiles (11).
RclR is a member of the AraC family of transcriptional regulators and senses HOCl with highly-conserved cysteine residues (9,13). RclA, RclB, and RclC proteins were identified as RclR target genes in both E. coli and Salmonella. In E. coli, RclR, RclA, RclB, and RclC contribute to the HOCl resistance of the bacterium. Deletion of any one of these genes decreases the survival of E. coli under HOCl stress (9). RclA putatively belongs to the flavoprotein disulfide reductase (FDR) family. RclB is a small uncharacterized periplasmic protein that has the signal sequence for secretion. RclC is a multipass inner membrane protein with homology to quinone-binding proteins. However, little is known about the biochemical functions of RclA, RclB, and RclC.
A recent study suggested that the FDR RclA is widely conserved among bacteria that colonize epithelial surfaces, and it contributes to host colonization in environments with HOClmediated oxidative stress (14). FDRs usually form homodimers and use noncovalently-bound flavin adenine dinucleotide (FAD) and a nonflavin redox center, catalyzing NAD(P)H-dependent reduction of a variety of substrates, such as sulfurcontaining peptides, metal ions, H 2 O 2 , and O 2 (15). Like other FAD-dependent reductases, the reducing power of the bound FAD is supplied from the cytosolic NAD(P)H pool.
FDRs commonly have four domains (FAD-binding domain, NAD(P)H binding domain, central domain, and the interface domain). However, they can be divided into three subgroups depending on the redox disulfide motif and the presence of the additional nonflavin redox centers at the extra N-or C-terminal motif (15). Very recently, Derke et al. (14) reported that RclA from E. coli exhibits Cu 2ϩ reductase activity, reducing Cu 2ϩ to Cu ϩ to cope with HOCl stress. Here, we determined the crystal structures of RclA from E. coli, containing FAD at its active site. Based on the structures and the subsequent biochemical analyses, we investigated the functions of RclA in terms of HOCl resistance in bacteria.

RclA contributes to the survival of Salmonella in macrophages
RclA is commonly found in Salmonella and E. coli (Fig. S1). Because HOCl is produced in macrophage phagosomes as a primary bactericidal agent, the function of RclA in resistance to HOCl suggests that RclA plays a role in the survival of the bacteria. Because Salmonella can proliferate in macrophages, we used S. enterica serovar Typhimurium to investigate the role of RclA in its proliferation in macrophages.
To confirm the role of RclA in the pathogenesis of Salmonella, we generated Salmonella mutant strains harboring the deletion of rclA or rclR. The WT and the RclR-or RclA-deletion mutant strains were precultured in LB broth and then incubated with macrophages. Deletion of rclA or rclR resulted in a 30% decrease in survival of Salmonella in macrophage after a 24-h incubation, indicating that RclA at least partly contributes to the survival of Salmonella in macrophages (Fig. 1). Taken together, our data indicate that RclA is involved in the pathogenesis of Salmonella by promoting survival in macrophage vacuoles containing sublethal doses of HOCl.

Structural determination and overall structure
Sequence comparison suggested that RclA belongs to the group I FDRs due to the presence of two essential motifs: CXXXXC for the redox-active disulfide as a nonflavin redox center, and the essential HXXXXE that interacts with and promotes the disulfide formation of the CXXXXC motif with its histidine residue ( Fig. 2A) (15). Like typical group I FDRs, RclA consists of the four domains and does not have the additional nonflavin redox center (Fig. 2B).
To gain mechanistic insights on RclA from its 3D structures, we expressed the full-length RclA proteins (441 amino acids) of E. coli and Salmonella and successfully determined the crystal structure of RclA from E. coli (EcRclA) at 2.9-Å resolution ( Fig.  2B and Table 1). The asymmetric unit contained four protomers consisting of two homodimers (Fig. S3). Each protomer contained a tightly-bound FAD molecule in its active site (Fig. 2B). Secondary structure elements of Fig. 2A were indicated in Fig.  S4.

Structural comparison to group I FDRs
Two identical active sites are at the subunit interfaces of the homodimer, which consists of chain A and chain B protomers (Fig. 3A). The overall arrangement of the CXXXXC motif in chain A and the HXXXXE motif in chain B protomer in the active site of RclA is similar to that of group I FDR family proteins ( Fig. 2A and Fig. S5). The two cysteines (Cys-43 and Cys-48) in the CXXXXC motif from chain A are in the reduced form, and the distance between the sulfur atoms of the two cysteine residues is ϳ4.4 Å in the crystal structure. The two cysteine residues are near the isoalloxazine ring of FAD from chain A, and Cys-48 is closer to FAD than to Cys-43 (the distance between the S␥ of Cys-48 and the C4a atom of flavin is 3.5 Å). The HXXXXE motif from chain B is involved in the interaction with the CXXXXC motif from chain A protomer and the bound FAD from chain A (Fig. 3A). His-426 and Glu-431 in the HXXXXE motif from chain B protomer are paired with a distance of 2.5 Å, and His-426 makes a polar interaction with Cys-43 from chain A protomer with a distance of 3.0 Å (Fig.  3A). Two lysine residues in chain A protomer (Lys-13 and Lys-16), which are unique to RclA proteins, were found near Cys-43 from chain A protomer at the active site. Because group I FDR proteins reduce substrates via the disulfide formation of the CXXXXC motif together with the HXXXXE motif, our findings suggest that RclA shares the enzymatic mechanism of group I FDR proteins.

Structural comparison to the group II FDR MerA
We searched for the closest protein to RclA in terms of structure using the DALI server (16). Mercuric reductase (MerA) (PDB code 4K7Z), a group II FDR, was ranked as the top entry (Z-score: 43.6). Like EcRclA reduces Cu 2ϩ to Cu ϩ , MerA converts Hg 2ϩ to Hg 0 (15). MerA contains the CXXXXC motif, similar to group I FDR proteins, and tyrosine replaces the histidine residue in the HXXXXE motif ( Fig. 2A). The structural superposition of EcRclA onto MerA revealed a high structural similarity between the matched regions (sequence identity: 30%; RMSD: 1.522 between 320 C␣ atoms; residues 5-433 of EcRclA, calculated with PyMOL (17)), except for the Cys-Cys motif for metal ion binding ( Fig. S6 and Fig. 3B). The C-terminal end loop containing the Cys-Cys motif was folded into the active site near the CXXXXC motif and the tyrosine residue in the MerA structure ( Figs. 2A and 3B). The Cys-Cys motif and the tyrosine residue interact with a mercuric ion at the active site (18).

Crystal structure of RclA Cu 2؉ and Hg 2؉ accelerate the oxidation of NADH in RclA
To elucidate the mechanism of RclA, we performed further biochemical analyses. We observed the basal level of the NADH oxidase activity of RclA in the absence of any additional material (control in Fig. 4A and 0 mM Cu 2ϩ in Fig. 4B), which was found in many FDR enzymes (19 -21). Previous studies of other FDRs suggested that the basal activity of RclA resulted from oxidation by the molecular oxygen in the buffer (19,(21)(22)(23).
Next, we tested whether RclA exhibits the MerA-like activity in which NADH oxidase activity is increased by adding metal ion (24,25). We measured the NADH oxidase activity of RclA in the presence of various metal ions, such as Hg 2ϩ , Cu 2ϩ , Fe 3ϩ , Mn 2ϩ , Mg 2ϩ , and Zn 2ϩ (Fig. 4A). Notably, Hg 2ϩ and Cu 2ϩ increased oxidation of NADH by ϳ3-fold compared with the basal activity without the metal ions, and Ag ϩ resulted in an ϳ2-fold increase. However, Fe 2ϩ , Mg 2ϩ , Zn 2ϩ , and Mn 2ϩ did not affect the NADH oxidase activity. Similar results were observed in a preprint (14). We further determined that Cu 2ϩ accelerates the NADH oxidation rate in a dose-dependent manner until reaching 0.3 mM CuCl 2 (Fig. 4, B and C). The data suggest that at higher concentrations of CuCl 2 , the dose-dependent effect would continue. A, sequence alignment of FDR proteins with the secondary structure elements. GR, GSH-disulfide reductase from Bartonella henselae (group I FDR; GenBank TM accession no. WP_011180567); Lpd, dihydrolipoyl dehydrogenase from Mycobacterium multispecies (group I FDR; GenBank TM accession no. WP_003402301.1); MerA, mercuric reductase from Pseudomonas aeruginosa (group II FDR; Gen-Bank TM accession no. WP_023980594.1). GR, Lpd, and MerA were chosen based on the highest score of DALI server structural analyses. The sequence alignment was performed using the T-Coffee server (35) and the ESPript server (36). For clarity, the regions containing the signature motifs are displayed. The full sequence alignment is presented in Fig. S2. The cysteine residues in the CXXXXC motifs and histidine and glutamate residues in the HXXXXE motif are indicated by To identify other substrates, we tested the NADH oxidation activity with oxidized GSH (GSSG), lipoic acid, cystine, and HOCl, which could be putative substrates of RclA. No increase was observed in the NADH oxidation rate by RclA with these substances (Fig. 4D). We thus focused on Cu 2ϩ in the following analyses, because Hg 2ϩ and Ag ϩ are not relevant in host immune responses, and Cu 2ϩ is employed in phagosomes of host innate immune systems.

O 2 is required for NADH oxidation in the presence of Cu 2؉
Although Derke et al. (14) reported the reduction of Cu 2ϩ with the reducing power of NADH, the redox stoichiometry was not fully accounted for. Only 25-50% of the reducing power from NADH was used in the Cu 2ϩ reduction to produce Cu ϩ (Fig. 4B in Ref. 14). Given that molecular oxygen dissolved in the reaction buffer is the oxidant candidate, we investigated molecular oxygen as the second oxidant in the RclA-mediated reaction.
Because a significant amount of O 2 is dissolved in ambient conditions, we investigated the NADH oxidase activity of RclA under oxygen-depleted conditions by thoroughly purging the buffer with nitrogen gas (Fig. 5A). The NADH oxidation rate of EcRclA was compared in the presence and absence of Cu 2ϩ . The depletion of dissolved oxygen almost completely shut down the NADH oxidation by RclA regardless of the presence of Cu 2ϩ . Stoichiometric analyses of this reaction accounted for 50% of the NADH oxidation in the presence of Cu 2ϩ . Thus, our findings suggest that the presence of Cu 2ϩ and molecular oxygen are interlinked in the oxidation of NADH by RclA, and O 2 is required for Cu 2ϩ -mediated NADH oxidation by RclA.

Cu 2؉ promotes O 2 consumption without producing superoxide anion
To determine whether O 2 is the additional oxidant during the RclA reaction, we measured the amount of oxygen dissolved in the reaction buffer with an oxygen probe during the oxidation of NADH by RclA (Fig. 5B). The specific O 2 consumption rate in the presence of 200 M Cu 2ϩ was about four times faster than that in the absence of Cu 2ϩ (Fig. 5B), as observed in the NADH oxidation rates in the presence and absence of Cu 2ϩ (Fig. 4C). These results indicate that Cu 2ϩ catalyzes the reduction of oxygen during the oxidation of NADH. About 50 M oxygen was consumed by the time 200 M NADH was entirely used in the reaction (Figs. 4B and 5B), indicating that four NADH molecules were required for one oxygen molecule in this reaction of RclA.
Many FDR enzymes produce superoxide anion as the basal or side activity by a reaction between the bound FADH 2 and molecular oxygen (22). We thus measured the amount of superoxide anion produced in the RclA-mediated reaction without Cu 2ϩ using cytochrome c, which is rapidly reduced by superoxide anion with an increase in light absorbance at 550 nm. Like other FDR enzymes, EcRclA without Cu 2ϩ produced superoxide anion (Fig. 5C). We performed similar experiments in the presence of Cu 2ϩ . Production of superoxide anion was not observed in the reaction, although more oxygen was consumed that in the absence of Cu 2ϩ (Fig. 5C). These findings indicate that Cu 2ϩ promotes the reaction between oxygen and FADH 2 at the active site of RclA while suppressing the production of superoxide anion.

Roles of the two cysteine residues in the CXXXXC motif
To investigate the roles of the two conserved cysteine residues (Cys-43 and Cys-48) in the CXXXXC motif and the additional cysteine residue (Cys-396) at the C terminus, we produced EcRclA C43S, C48S, and C396S, and we then tested their activities in the presence or absence of Cu 2ϩ . The mutation at Cys-396 did not affect the tested RclA properties (data not shown), which is consistent with the structural analyses. The cysteine residue was buried and surrounded by two methionine Crystal structure of RclA residues in the hydrophobic region in the crystal structure (Fig.  S7). Because the cysteine residue is far from the active site, it is unlikely that Cys-396 is involved in the catalysis. However, analyses of the point mutation at Cys-43 or Cys-48 in the CXXXXC motif yielded interesting results (Fig. 6A). When Cys-48 was changed to a serine, the NADH oxidase activity was highly increased regardless of the presence of Cu 2ϩ , compared with those of WT enzyme (Fig. 6, A and B, and Fig.  S8). These observations indicate that the Cu 2ϩ -dependent NADH oxidase activity of RclA is controlled by Cys-48. In terms of production of superoxide anion, C48S showed the increased basal superoxide production rate in the absence of Cu 2ϩ , indicating that Cys-48 may contribute to the suppression of the basal superoxide production by RclA (Fig. 6B).
The RclA C43S showed a dramatic increase in NADH oxidase activity in the presence of Cu 2ϩ compared with the WT (Fig. 6A). However, the mutation at Cys-43 increased the NADH oxidase activity in the presence of other metal ions as well. Ag ϩ accelerated the oxidation rate of NADH as much as Cu 2ϩ , and Zn 2ϩ and Hg 2ϩ also further increased the NADH oxidase activity compared with that of WT enzyme (Fig. 6C). Thus our findings suggest that Cys-43 is involved in the metal specificity and further in the binding of Cu 2ϩ at the active site.

Multiple Cu 2؉ -binding sites
To analyze how Cu 2ϩ affects the activity of RclA at the molecular level, we determined crystal structures of EcRclA in complex with Cu 2ϩ in the presence of 5 mM CuCl 2 . To prevent the formation of the disulfide bond in the CXXXXC motif by Cu 2ϩ , we also determined the complex structure of the EcRclA C43S mutant (see below). We found that RclA is a histidinerich protein containing 16 histidine residues (15 surface-exposed histidine residues and His-426 at the active site). Strong Cu 2ϩ densities were observed near all of the histidine residues except for His-426 in both complex structures, indicating that Cu 2ϩ is bound to almost all of the surface-exposed histidine residues of RclA (Fig. 7A). However, no apparent structural changes were observed in the complex structures despite the multiple Cu 2ϩ binding at the histidine residues (Fig. S9). It is noteworthy that surface-exposed histidine residues are abundant in RclA. Because almost all of the histidine residues of RclA can bind to Cu 2ϩ (Fig. 7A), RclA could lower the amount  Crystal structure of RclA of free Cu 2ϩ available to participate in the Fenton reaction with HOCl to produce highly-toxic hydroxyl radicals.

Cu 2؉ binding at the active site
Strong electron density maps near the cysteine residues at the active site also appeared in Cu 2ϩ -soaked crystals of the WT RclA protein. The maps were suggested as heavy-atom sites by a strong signal in the difference anomalous map. Thus, it is very likely that the strong electron density maps represent the Cu 2ϩ binding (Fig. 7B). We noted that the Cu 2ϩ density maps were between the sulfur atoms of Cys-43 and Cys-48 in the CXXXXC motif. Moreover, the densities were not found in the C43S mutant structure, indicating the importance of Cys-43 for the binding of Cu 2ϩ (Fig. 7C). Combined with the biochemical results with the C43S mutant RclA, the binding between the two cysteine residues is essential in the selectivity to Cu 2ϩ over the other metal ions.
Regarding Cys-48, the Cys-48 residue is located between the Cu 2ϩ and the FAD molecule. This structure suggests that Cys-48 prevents the direct contact between Cu 2ϩ and FAD, and it allows the electron flows from FAD to Cu 2ϩ via Cys-48. This structural arrangement of Cu 2ϩ -Cys-48 -FAD is well-supported by our biochemical results that Cys-48 contributed to the prevention of basal production of superoxide anion from oxygen.
Two lysine residues (Lys-13 and Lys-16), which are unique to RclA proteins, were located at the entrance to the active site. To examine the role of the lysine residues, we produced the mutant protein K13A/K16A and compared the K m and V max values for Cu 2ϩ with those of the WT protein. The mutant protein showed both K m and V max values decreased by ϳ1.3-fold with a similar V max /K m value (Fig. 7D). These biochemical results show that the two lysine residues provide the positivelycharged environment to limit the prolonged staying of metal ions, promoting the turnover of the reaction cycles. Alterna-

Discussion
RclA from E. coli was reported to increase resistance to HOCl treatment (9). This study using Salmonella strains extends the role of RclA in bacterial survival in macrophage phagosomes. To understand the roles of RclA at the molecular level, we elucidated high-resolution structures of RclA from E. coli exhibiting the typical dimeric features of FDR family proteins. The structure of EcRclA showed intermingled properties of groups I and II FDRs. Although the overall structure of EcRclA was the most similar to the group II FDR MerA, the key motifs in the active site were more closely related to group I FDRs. While this manuscript was in preparation, Derke et al. (14) reported that EcRclA is a Cu 2ϩ reductase that converts Cu 2ϩ into Cu ϩ using the reducing power of NADH to cope with HOCl stress in Drosophila. Consistently, we found that Cu 2ϩ boosted the NADH oxidation activity in a dose-dependent manner, and Cu 2ϩ was bound to the active site. Furthermore, we discovered the O 2 consumption activity of RclA in the presence of Cu 2ϩ , with a limited production of superoxide anion.
We further found that O 2 reduction is required for the Cu 2ϩ reduction activity of EcRclA and vice versa, indicating that the NADH oxidation is coupled with the reduction of O 2 dissolved in the buffer. Furthermore, neither the reduced portions of Cu 2ϩ nor O 2 was able to account fully for the released electrons from NADH. These findings indicate that the reduction of Cu 2ϩ and O 2 should be simultaneously considered a reaction cycle. Because O 2 is the starting substance to produce HOCl via oxidative burst in immune cells, the reduced level of O 2 in the phagosomes would lower the level of HOCl. To decrease the level of HOCl would be an important role of RclA for bacterial survival in immune cells.
Despite the structural and chemical similarities to MerA mediating a two-electron transfer to a metal ion, a previous study (14) presented evidence for a one-electron transfer to Cu 2ϩ producing Cu ϩ . However, it seems odd that Cu ϩ is more active than Cu 2ϩ in Fenton-like reactions to produce hydroxyl radicals from HOCl. We believe that more sophisticated reaction networks with other proteins would be involved.
FDRs are involved in many different chemical reactions using the bound FAD and various electron acceptors, such as disulfide, O 2 , or metal ions. In the group I FDR GSH reductase, electrons from NADPH are transferred to the disulfide of the oxidized GSH using the CXXXXC motif as the key intermediate. However, a small portion of electrons flows into oxygen to generate superoxide anion or H 2 O 2 as a side reaction of FDRs (21). Furthermore, the reduction of GSH is strongly inhibited by Cu 2ϩ (26), indicating that Cu 2ϩ is bound at the active site, and thus it alters the electron flow in the reaction mechanism as a potential regulator. In the cases of MerA and RclA, the electrons flow into Hg 2ϩ and Cu 2ϩ , respectively, as the final acceptor, which is the primary reaction of these enzymes. Thus, our findings suggest that binding of metal ions at the active site alters the pathway of electrons from FDRs toward the bound metal ions.
Our structural and mutational studies provide hints on the reaction mechanism of RclA (Fig. 8A). This study observed the Cu 2ϩ binding in the presumably-oxidized FAD-bound structure, presenting the FAD-Cys-48 -Cu 2ϩ -Cys-43 network as the critical electron flow route. The reaction would be triggered when the electrons are provided to FAD by NADH in the presence of O 2 . The electrons released from FADH 2 are first transferred to Cys-48 and then transferred to the bound Cu 2ϩ with the Cys-43. This Cys-48 -mediated electron transfer to Cu 2ϩ is consistent with the result that the Cu 2ϩ dependence in NADH oxidation disappeared with increased levels of superoxide anion production in the C48S mutant. We thus propose that

Crystal structure of RclA
Cys-48 prevents the direct single electron transfer from FADH 2 to O 2 by absorbing electrons from FADH 2 , leading to suppression of superoxide anion production, and transfers the received electrons to Cu 2ϩ .
The Cu 2ϩ -bound structure, together with the results of the C43S mutation, demonstrated that Cys-43 is directly involved in the binding of Cu 2ϩ . This C-terminal cysteine residue (Cys-43) in the CXXXXC motif would affect the specificity of various metal ions, as the C43S mutant showed different NADH oxidation rates with metal ions compared with the WT. Because the sulfur atom of Cys-43 and the imidazole ring of His-426 are within a distance to make a hydrogen bond, the His-426 -Glu-431 pair in the HXXXXE motif would deprotonate Cys-43, promoting binding with the Cu 2ϩ . Furthermore, the His-Glu pair could also transfer protons to proton or electron acceptors during the catalysis. To reduce oxygen, the electron transfer should accompany the proton transfer. Thus, this proton transfer mediated by this HXXXXE motif would also be important for the catalysis. Because of the complexity of the reactions, we believe that multiple rounds might be required to complete the reaction. Combined with the previous result (14), we propose an overall reaction mechanism, which meets the stoichiometry (Fig. 8B). However, further studies are required to elucidate the complete reactions.
In conclusion, we used a high-resolution structure of RclA to investigate its biochemical roles and functions in the HOCl resistance of bacteria. Although further study is required to elucidate the molecular mechanism, our study provides a molecular basis for how bacteria can cope with our immune system, which may help in the creation of strategies to control pathogenic bacteria.

Plasmid construction and protein production
To overexpress the full-length (amino acids 1-441) RclA WT, the gene encoding the protein was amplified and cloned into the pProEX-HTa expression vector (Invitrogen). The resulting plasmid encodes the hexahistidine tag and the TEV protease cleavage site at the N terminus of the protein. For each RclA mutant protein (C43S, C48S, and K13A/K16A), site-directed mutagenesis was performed. The plasmids encoding RclA WT protein were transformed into the E. coli C43 (DE3) strain, and each plasmid encoding the RclA mutant was transformed into E. coli BL21 (DE3). The transformed cells were cultured in 1.5 liters of Luria Broth medium containing 100 g/ml ampicillin at 37°C until an OD 600 of 0.7 was measured, and the protein expression was induced with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside at 30°C for 6 h. After harvesting, cells were resuspended in lysis buffer containing 20 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 2 mM 2-mercaptoethanol. The suspensions were disrupted using a continuous-type French press (Constant Systems Ltd., United Kingdom) at 23,000 p.s.i. After centrifugation at 19,000 ϫ g for 30 min at 4°C, the supernatant was loaded onto a nickel-nitrilotriacetic acid column (GE Healthcare). After washing the column with ϳ300 ml of lysis buffer supplemented with 20 mM imidazole, the RclA WT protein was eluted with lysis buffer supplemented with 250 mM imidazole. The eluate was treated with TEV protease to cleave the hexahistidine tag. Cleaved proteins were diluted 4-fold with 20 mM Tris-HCl (pH 8.0) buffer containing 2 mM 2-mercaptoethanol and loaded onto a HiTrap Q column (GE Healthcare). A linear gradient of increasing NaCl concentrations was applied to the column. The fractions containing the protein were pooled, concentrated, and subjected to size-exclusion chromatography using a HiLoad 26/600 Superdex 200 pg column (GE Healthcare) equilibrated with 20 mM Tris-HCl (pH 8.0) buffer containing 150 mM NaCl and 2-mercaptoethanol. The purified protein was concentrated to 13 mg/ml and stored at Ϫ80°C until used for crystallization. Protein concentrations were determined based on FAD content using a previously determined ⑀ 456 of 11.3 mM Ϫ1 cm Ϫ1 .

Crystallization of RclA
RclA proteins (WT, C43S) were crystallized at 16°C using the hanging-drop vapor-diffusion method after mixing 1 l of protein solution and 1 l of well solution. The RclA WT protein was first crystallized in a well solution containing 0.2 M ammonium sulfate, 0.1 M BisTris-HCl (pH 5.8), 25% (w/v) PEG 3350, and 2 mM TCEP. Subsequently, the crystals were disrupted for microseeding and diluted with the well solution containing 0.2 M ammonium sulfate, 0.1 M BisTris-HCl (pH 5.8), 13% (w/v) PEG 3350, and 2 mM TCEP. Microseeding was performed by mixing 1 l of protein and 1 l of microseeds diluted with the well solution. The RclA WT crystals were briefly incubated in the well solution supplemented with 20% (v/v) glycerol as a cryoprotectant, flash-frozen in liquid ethane at Ϫ88.5°C, and subsequently transferred into liquid nitrogen to collect the dataset. For the Cu 2ϩ -soaked RclA WT structure, the crystals were incubated in a reservoir solution supplemented with 5 mM CuCl 2 for 15 h. The crystals were incubated in the well solution supplemented with 30% (v/v) glycerol and 5 mM CuCl 2 for 6 min, and flash-frozen in liquid nitrogen to collect the dataset. RclA C43S was crystallized in a solution containing 0.05 mM ammonium sulfate, 0.05 mM BisTris-HCl (pH 5.8), and 30% pentaerythritol ethoxylate (15/4_EO/OH). Microseeding was performed using disrupted RclA C43S crystals and was diluted Crystal structure of RclA with 0.05 mM ammonium sulfate, 0.05 mM BisTris-HCl (pH 5.8), 25% pentaerythritol ethoxylate (15/4_EO/OH). For Cu 2ϩsoaked RclA C43S structure, the RclA C43S crystals were incubated in the well solution supplemented with 5 mM CuCl 2 for 15 h. After incubation of the crystals in reservoir solution supplemented with 30% (v/v) glycerol and 5 mM CuCl 2 for 30 s, crystals were flash-frozen in a nitrogen stream at Ϫ173°C.

Data collection and structural determination
All diffraction datasets were collected at the Pohang Accelerator Laboratory Beamline 5C (27) and were processed using the HKL-2000 package (28,29). The structure of the RclA WT was determined using the molecular replacement method with MOLREP in the CCP4 package (29), using a model built by the Galaxy web server (30, 31) as a search model. The initial model of the RclA WT was used to solve other structures of RclA (RclA WT-Cu 2ϩ and RclA C43S-Cu 2ϩ ) by the molecular replacement method. All of the structures were built using COOT and refined using PHENIX refine software (32,33). The detailed statistics are shown in Table 1.

Enzyme kinetic assays
All assays were performed using UV-visible spectrophotometer, GeneQuant 1300 (GE Healthcare) or Multiscan Go (Thermo Fisher Scientific). The RclA activity was assayed by adding 2 M protein (as measured based on flavin content) in 10 mM MES buffer (pH 7.0) containing 100 mM NaCl, 0.2 mM NADH, and 0.2 mM putative substrate. Initial rates of NADH oxidation by RclA with various metal ions were monitored by the loss of absorbance at 340 nm due to NADH oxidation (⌬⑀ 340 ϭ 6.2 mM Ϫ1 cm Ϫ1 ); the loss of absorbance was monitored for 10 min, and the oxidation profile was expressed in a time-dependent manner as the mean Ϯ S.D. of three independent experiments. Control assays lacking the substrate or enzyme were routinely included as references. NADH oxidation by RclA mutants C43S and C48S were recorded as described previously for the RclA WT assay.
For the examination of enzyme dependence on Cu 2ϩ concentration, NADH oxidation was measured at 340 nm with various concentrations of CuCl 2 . Before the assay, the RclA WT protein was desalted with 10 mM MES buffer (pH 7.0) supplemented with 100 mM NaCl to remove any remaining 2-mercaptoethanol from the RclA size-exclusion chromatography buffer. The activity assays were initiated by adding 2 M desalted RclA WT protein to 10 mM MES buffer (pH 7.0) containing 100 mM NaCl, 0.2 mM NADH, and 0, 0.025, 0.05, 0.075, 0.1, 0.2, or 0.3 mM CuCl 2 . Initial rates were monitored by the loss of absorbance at 340 nm due to NADH oxidation (⌬⑀ 340 ϭ 6.2 mM Ϫ1 cm Ϫ1 ) and expressed as the mean Ϯ S.D. of four independent experiments.
For the determination of K m and V max values of RclA WT and RclA K13A/K16A, NADH oxidation was measured at 340 nm with various concentrations of CuCl 2 . Before the assay, each RclA protein was desalted with 10 mM MES buffer (pH 7.0) supplemented with 100 mM NaCl. The activity assays were initiated by adding 2 M desalted RclA WT or RclA K13A/K16A protein to 10 mM MES buffer (pH 7.0) containing 100 mM NaCl, 0.2 mM NADH, and 0, 0.6, 0.15, 0.3, 0.6, 1.5, 3, or 6 mM CuCl 2 . Initial rates were monitored by the loss of absorbance at 340 nm due to NADH oxidation (⌬⑀ 340 ϭ 6.2 mM Ϫ1 cm Ϫ1 ) and expressed as the mean Ϯ S.D. of three independent experiments. The data were fit with a Michaelis-Menten equation, and the K m and V max values were determined using an enzyme kinetics tool of the SigmaPlot Version 14.0 (Systat Software, San Jose, CA).

NADH oxidation assay under low O 2 conditions
The dissolved oxygen (DO) was removed by nitrogen purging. The buffer with 10 mM MES (pH 7.0), 100 mM NaCl, 0.2 mM NADH, and with or without 0.1 mM CuCl 2 was contained in a 1-ml cuvette and sealed with parafilm. The solution was purged with N 2 for 5 min. After nitrogen purging, 2 M RclA was added. Initial rates of NADH oxidation by RclA were monitored by the loss of absorbance at 340 nm due to NADH oxidation (⌬⑀ 340 ϭ 6.2 mM Ϫ1 cm Ϫ1 ), monitored using a GeneQuant 1300 UV-visible spectrophotometer (GE Healthcare), and expressed in time-dependent manner as the mean Ϯ S.D. of three independent experiments. A control assay in which the solution was not purged with N 2 was included for comparison.

Detection of DO in water
DO concentrations were measured using the Gravity Analog Dissolved Oxygen Sensor (SKU:SEN0237; DFRobot) with an Arduino Nano board and the Arduino IDE program (34). The programming code was obtained from the DFRobot website. The reaction volume was 5 ml, containing 10 mM MES (pH 7.0) buffer with 100 mM NaCl, 0.2 mM NADH, and 1 M RclA, with or without 0.2 mM CuCl 2 . The solution was stirred gently with a 1-cm magnetic stir bar to distribute DO evenly throughout the entire solution. After preincubation for 5 min with stirring for equilibrium, the RclA protein was added and the DO concentration was recorded for 10 min. The specific oxidation rate was measured within the initial 1 min and expressed as the mean Ϯ S.D. of three independent experiments.

Measurement of O 2 . formation by cytochrome c assay
Superoxide anion (O 2 . ) formation was determined by monitoring the reduction of cytochrome c, which was isolated from the equine heart (Sigma). The reaction was performed with 200 l of buffer containing 10 mM MES buffer (pH 7.0), 100 mM NaCl, 200 M NADH, 100 M CuCl 2 , 2 M RclA, and 50 M cytochrome c. The increase in the absorbance at 550 nm due to cytochrome c reduction was recorded at 37°C for 10 min using an UV-visible spectrophotometer (Multiscan Go, Thermo Fisher Scientific). The amount of O 2 . released was calculated using an extinction coefficient of 21 mM Ϫ1 cm Ϫ1 and expressed as the mean Ϯ S.D. of three independent experiments.

Intracellular survival assay
RAW264.7 macrophage-like cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% heatinactivated fetal bovine serum, penicillin (50 units/ml), and streptomycin (50 units/ml). The cells were seeded at 2 ϫ 10 5 cells/well in a 24-well tissue culture plate and incubated for 24 h at 37°C with 5% CO 2 . A monolayer of cells was infected with bacteria at a multiplicity of infection of 10. After 30 min of Crystal structure of RclA incubation at 37°C under 5% CO 2 , nonengulfed bacteria were removed by washing with PBS and incubated for 1 h with medium containing 100 g/ml gentamycin to kill extracellular bacteria. Subsequently, the medium was exchanged with fresh medium containing 10 g/ml gentamycin, and cells were cultured for the indicated times (2 or 24 h). For counting bacterial cells replicated within RAW264.7, the macrophage cells were lysed with 1% Triton X-100.