The O2-independent pathway of ubiquinone biosynthesis is essential for denitrification in Pseudomonas aeruginosa

Many proteobacteria, such as Escherichia coli, contain two main types of quinones, benzoquinones represented by ubiquinone (UQ) and naphthoquinones such as menaquinone (MK) and dimethyl-menaquinone (DMK). MK and DMK function predominantly in anaerobic respiratory chains, whereas UQ is the major electron carrier used for reduction of dioxygen. However, this division of labor is probably not so stric. Indeed, a pathway that produces UQ under anaerobic conditions in an UbiU-, UbiV- and UbiT-dependent manner has been recently discovered in E. coli. Nevertheless, the physiological relevance of this pathway is not yet understood because of the presence of MK and DMK in this bacterium. In the present study, we established that UQ9 is the single quinone of P. aeruginosa and is required for growth under anaerobic respiration (denitrification). Moreover, we demonstrated that ORFs PA3911, PA3912 and PA3913, which are homologues to the ubiT, ubiV and ubiU E. coli genes in the anaerobic pathway, respectively, were essential for UQ9 biosynthesis and thus for denitrification in P. aeruginosa. We showed that UbiV from P. aeruginosa accommodates a [4Fe-4S] cluster, as is the case with its E. coli homolog. Moreover, we demonstrated that UQ copurifies with UbiU. Finally, UbiT, which possesses a sterol carrier domain 2, was able to bind lipidic molecules such as UQ. Since the denitrification metabolism of P. aeruginosa is believed to be important for pathogenicity in cystic fibrosis patients, our results highlight the O2-independent UQ biosynthesis pathway as a new possible target to develop innovative antibiotics.

The opportunistic pathogen Pseudomonas aeruginosa has a remarkable ability to grow under a variety of environmental conditions, such as soil and water as well as animal-, human-, and plant-host-associated environments. P. aeruginosa is responsible for numerous acute and chronic infections and poses a major health risk for patients with severe burns and cystic fibrosis (CF) or in severely immunocompromised states (1,2).
The utilization of various carbon sources and energy metabolism (respiration or fermentation) might contribute to the environmental adaptation of P. aeruginosa (3). Its main energy-producing system is respiration, which requires a protonmotive force used for ATP synthesis. The proton-motive force is produced by the transfer of electrons and protons from reduced donors to oxidized acceptors via the quinone pool. Whereas the dehydrogenases and reductases involved in respiratory metabolism have been well described and annotated in the genome of P. aeruginosa PAO1 (4,5), the composition of its quinone pool has not yet been fully established. Studies in the 1960s suggested ubiquinone 9 (UQ 9 ) is a major quinone of aerobically grown P. aeruginosa (6); therefore, UQ 9 is believed to be essential for aerobic respiration (7).
Proteobacteria contain two main types of quinones, benzoquinones and naphthoquinones, represented by UQ (or coenzyme Q) and menaquinone (MK)/demethylmenaquinone (DMK), respectively (8). Typically, MK and DMK function predominantly in anaerobic respiratory chains, whereas UQ is the major electron carrier used for the reduction of dioxygen by various cytochrome oxidases (8). Recent data indicated that the metabolic use of various quinone species according to environmental dioxygen availability is more complex than initially thought. Indeed, using E. coli as a model, we highlighted a pathway conserved across many bacterial species and able to produce UQ under anaerobic conditions (9). The classical UQ biosynthetic pathway requires O 2 for three hydroxylation steps (10). Obviously, the flavin-dependent monooxygenases UbiI, UbiF, and UbiH, which catalyze the O 2 -dependent hydroxylation steps, are not involved in the anaerobic pathway, and the accessory UbiK and UbiJ proteins are not implicated in the assembly and/or stability of the aerobic Ubi complex (11). Seven proteins (UbiA, UbiB, UbiC, UbiD, UbiE, UbiG, and UbiX) catalyzing the prenylation, decarboxylation, and methylation of the phenyl ring of the 4-hydroxybenzoate precursor are common to both pathways (9). In addition, the anaerobic pathway requires UbiT, UbiU, and UbiV proteins. UbiT is homologous to UbiJ, and UbiU and UbiV are expected to be involved in O 2independent hydroxylations (9). However, as explained above, the metabolic relevance of the O 2 -independent UQ pathway is not yet clearly understood.
In the absence of O 2 , P. aeruginosa is able to carry out anaerobic respiration with nitrate and nitrite as terminal electron acceptors of the respiratory chain. This process, called denitrification, allows the reduction of soluble nitrate (NO 3 2 ) and This article contains supporting information. ‡ These authors contributed equally to this work. * For correspondence: Ludovic Pelosi, ludovic.pelosi@univ-grenoble-alpes.
nitrite (NO 2 2 ) to gaseous nitrous oxide (N 2 O) or molecular nitrogen (N 2 ) (12). Because P. aeruginosa-infected mucus in CF airways is depleted of oxygen and enriched in nitrate and nitrite, the anaerobic metabolism of P. aeruginosa via the denitrification pathway is believed to be important for its pathogenicity (13). Four sequential reactions involving metalloenzymes are needed to reduce nitrate to N 2 , i.e. nitrate reductase, nitrite reductase, nitric oxide reductase, and nitrous oxide reductase. P. aeruginosa was considered a paradigm of the denitrification pathway, and all the reductases involved in this metabolism have been widely studied, as has the regulatory network controlling the denitrification genes (3,4,14). However, the anaerobic quinone pool of P. aeruginosa has not been characterized so far.
In the present study, we discovered that UQ 9 is essential for the growth of the P. aeruginosa PAO1 strain in denitrification medium. We identified in this bacterium the ORFs PA3911, PA3912, and PA3913 as homologs to E. coli ubiT, ubiV, and ubiU, respectively. Our results showed that these three genes, here called ubiT Pa , ubiV Pa , and ubiU Pa , are essential components of the O 2 -independent UQ 9 biosynthetic pathway of P. aeruginosa. We demonstrated that (i) UbiV Pa binds a [4Fe-4S] cluster and (ii) UbiU Pa and UbiT Pa copurify with UQ by recognizing the isoprenoid tail. Such a molecular pathway for UQ production was found only in proteobacteria (9), where it can exert an essential role under anaerobic conditions, as demonstrated here. Taken together, our results highlight that this pathway could be an interesting lead for the development of antibiotics targeting the denitrification metabolism.

Results
UQ 9 is the major quinone of P. aeruginosa The quinone content of P. aeruginosa PAO1, grown under ambient air or anaerobic conditions (denitrification), was determined using electrochemical detection of lipid extracts separated by HPLC and compared with those obtained from E. coli. Whatever the conditions of growth, a major quinone species eluting at 11.5 min was present in the analyses of P. aeruginosa lipid extracts, with UQ 10 being used as the standard (Fig.  1, A and B). MS analysis showed a predominant ammonium adduct (M 1 NH 4 1 ) with an m/z ratio of 812.7 (Fig. 1C), together with minor adducts, such as Na 1 (817.6) and H 1 (795.7) (Fig. S1). These masses identify UQ 9 (monoisotopic mass, 794.6) as the major quinone produced by P. aeruginosa. Membranes of E. coli contain UQ 8 and naphthoquinones (DMK 8 and MK 8 ). The absence of detectable levels of naphthoquinones in P. aeruginosa lipid extracts, with or without oxygen (Fig. 1, A  and B), is in agreement with the absence of their biosynthetic pathways (MK or futalosine pathways) in P. aeruginosa genomes. It is also interesting that the UQ content of E. coli was higher under aerobic compared with anaerobic conditions (97 6 4 versus 42 6 6 pmol of UQ 8 per mg of cells), whereas we found the opposite for P. aeruginosa (95 6 4 versus 126 6 4 pmol of UQ 9 per mg of cells). Together, our results establish that UQ 9 is the major quinone of P. aeruginosa PAO1 and suggest that UQ 9 is used under denitrification conditions. Identification of ubi genes in the genome of P. aeruginosa PAO1 To identify the Ubi proteins of P. aeruginosa, IspB, UbiX, and UbiA to UbiK from E. coli MG1655 were first screened for homologs in the P. aeruginosa PAO1 protein sequence data set, available at RRID:SCR_006590, using the BLASTP software. As listed in Table S1, the analysis disclosed the presence of 11 homologous proteins (IspB, UbiA to UbiE, UbiG to UbiJ, and UbiX). As reported previously, the functional homolog of UbiF is a Coq7-like hydroxylase (15), and the corresponding PA0655 protein was shown to be essential for aerobic UQ 9 biosynthesis (16). Overall, we propose that the O 2 -dependent UQ biosynthetic pathways in P. aeruginosa and E. coli share a similar pattern (Fig. S2).
Under anaerobic conditions, E. coli still synthesizes UQ, and we recently identified three genes, which we called ubiT, ubiV, and ubiU, as essential for this process (9). Homologues of ubiT, ubiV, and ubiU were also identified in P. aeruginosa PAO1 and correspond to ORFs PA3911, PA3912, and PA3913, respectively (Table S1 and Fig. S2). These genes are called ubiT Pa , ubiV Pa , and ubiU Pa here. The three genes are predicted to form an operon, ubiUVT (RRID:SCR_006590). Interestingly, this operon is located downstream of the genes moeA1, moaB1, moaE, moaD, and moaC, involved in the biosynthesis of the molybdopterin cofactor (MoCo) (Fig. 2), which is essential for nitrate reductase activity (17). Next, we evaluated the metabolic relevance of the O 2 -independent UQ biosynthetic pathway in P. aeruginosa by studying mutants of ubiT Pa , ubiU Pa , and ubiV Pa genes.
Tn mutants of ubiV Pa and ubiU Pa present a growth defect for denitrification and an impaired UQ 9 content The physiological importance of the proteins UbiT Pa , UbiU Pa , and UbiV Pa was first investigated using transposon (Tn) mutants PW7609 (ubiT Pa ), PW7610 (ubiV Pa ), PW7611 (ubiV Pa ), PW7612 (ubiU Pa ), and PW7613 (ubiU Pa ) and the isogenic parental strain MPAO1 (WT strain from the Manoil collection) as a control (Table S2). Aerobic growth in LB medium was similar between the Tn mutants and the WT strain MPAO1 (Fig. S3A), and the Tn mutants presented a UQ 9 level comparable to that of the WT (Fig. S3B). Thus, UbiT Pa , UbiU Pa , and UbiV Pa are not involved in the O 2 -dependent UQ biosynthetic pathway of P. aeruginosa. In contrast, the growth of the ubiV Pa and ubiU Pa mutants was severely impaired under denitrification conditions (Fig. S3C), and their UQ 9 content was strongly lowered (Fig. S3B). These results suggest the overall requirement of ubiU Pa and ubiV Pa for denitrification in P. aeruginosa, supposedly via their involvement in O 2 -independent UQ biosynthesis. Surprisingly, the growth of the ubiT Pa Tn mutant PW7609 was not affected (Fig. S3C), and it showed around 40% UQ 9 of the WT level in anaerobic cultures (Fig. S3B). In this mutant, the Tn is inserted at the fifth base of the ubiT Pa gene, potentially leading to only partial inactivation of the gene (Table S2). We note that previous studies with E. coli ubi mutants showed that only 20% UQ was sufficient to maintain a WT growth phenotype (18,19).
Denitrification is dependent on ubiT Pa , ubiU Pa , and ubiV Pa genes via their involvement in O 2 -independent UQ biosynthesis As ubiT Pa , ubiV Pa , and ubiU Pa are localized next to each other in the genome of PAO1, the transposon inserted in the mutants previously studied might impact the expression of the neighboring genes. In addition, it is likely that the Tn mutant PW7609 is not properly disrupting the ubiT Pa gene. Thus, for each of the three genes, we constructed knockouts (KO) as well as complementation mutants in the parental strain PAO1. All deletion mutant strains (here called ubiTUV-KO) shared a growth defect under denitrification coupled to a strong decrease of UQ 9 content compared with the WT (Fig. 3, A and B), whereas UQ 9 content and growth were normal under aerobic conditions (Fig. 3, B and C). Under anaerobic conditions, ubiTUV-KO strains accumulated an early UQ biosynthetic intermediate corresponding to nonaprenylphenol (NPP) (Fig. S2 and Table S3). This result suggests that the O 2 -independent UQ biosynthetic pathway is blocked downstream of NPP in these three mutants. Interestingly, upon complementation, bacterial growth and UQ 9 levels were restored to those of the PAO1 strain used as a control (Fig. 4A).
To confirm that UQ was directly involved in the restoration of the anaerobic growth of the ubiTUV-KO strains, UQ 4 solubilized in methanol was added to the denitrification medium at 5 and 50 mM final concentrations. After 24 h of anaerobic incubation, numbers of CFU per ml (CFU/ml) of each KO strain were estimated and compared with the same strain cultivated without UQ 4 . The most significant results were obtained with 50 mM UQ 4 , which increased the number of CFU of ubiTUV-KO strains by 8-to 15-fold (Fig. 4B). We noted a substantial toxicity of methanol on the WT strain (Fig. 4B, Ct lane), suggesting that the positive effect of UQ 4 on the ubiTUV-KO strains is underestimated. Taken together, our results show unequivocally that UbiT Pa , UbiU Pa , or UbiV Pa is needed for denitrification via its involvement in UQ biosynthesis.
UbiT Pa , UbiU Pa , and UbiV Pa are needed for the process of denitrification We used soft-agar experiments to examine dioxygen and nitrate requirements of ubiTUV-KO strains with or without the WT allele. Soft agar was prepared anaerobically in LB medium containing KNO 3 at 100 mM final concentration and then exposed to ambient air. Oxygen diffuses through the agar to form a gradient, the highest concentration being at the top of the agar (Fig. 4C, lane 1). As shown in Fig. 4C, parental strain PAO1 and complemented ubiTUV-KO strains grew throughout the tube, because they were able to use aerobic respiration as well as denitrification. In contrast, the growth of the ubiTUV-KO strains harboring the empty vector was restricted to the oxygenated part of the medium, whereas the presence of the respective genes on the plasmids allowed growth in the anaerobic medium (Fig. 4C). The bubbles observed in the soft agar correspond to gas evolution of N 2 O and/or N 2 (12), suggesting a restoration of the denitrification process in the lower part of the tube. Taken together, these results point to the requirement for ubiT Pa , ubiU Pa , and ubiV Pa beyond the nitrate reduction step and support that UQ is probably essential for the entire denitrification process in P. aeruginosa.

Molybdopterin cofactors are not involved in anaerobic UQ 9 biosynthesis
As mentioned previously, the ubiUVT operon is located downstream of the genes moeA1, moaB1, moaE, moaD, and moaC, involved in MoCo biosynthesis. Currently, MoCo-containing hydroxylases constitute the only family known to catalyze O 2 -independent hydroxylation reactions (20). Since three O 2 -independent hydroxylation reactions are needed to synthesize UQ anaerobically and UbiU and UbiV are suspected to be involved in these reactions (9), we reasoned that MoCo might  Ubiquinone is essential for denitrification in P. aeruginosa be involved in this process, providing a rationale for the colocalization of the ubi and moa/moe genes. To test this hypothesis, we evaluated the ability of Tn mutants PW7614, PW7615/ PW7616, PW7618/PW2470, PW7619/PW1920, and PW7621/ PW7622 (Table S2), corresponding to Tn insertions in the ORFs moeA1, moaB1 (two mutants), moaE (two mutants), moaD (two mutants), and moaC (two mutants), respectively, to synthesize UQ 9 without O 2 . However, MoCo is also essential for nitrate reductase activity and, thus, for denitrification (17). To overcome this problem, WT and Tn mutants were grown in LB medium using arginine as a fermentable energy source in rich medium. As expected, all the Tn mutants exhibited a growth defect in denitrification. However, anaerobic growth was rescued by the addition of arginine, as previously described (21); therefore, we were able to measure the UQ content of the cells under these conditions (Fig. S4). The UQ 9 content of the MoCo Tn mutants was comparable to that of the WT strain ( Fig. S4), suggesting that MoCo is not involved in anaerobic UQ 9 biosynthesis.

Recombinant UbiV Pa is an air-sensitive [4Fe-4S] cluster-containing protein
To gain insights into their biochemical properties, we produced and purified the three proteins in E. coli, with UbiV Pa being the most soluble. First, we showed that UbiV Pa , purified by size exclusion chromatography (SEC), behaved as a monomer (Fig. S5, A and B). Moreover, we noticed that the fraction containing the purified protein was slightly pink-colored with a UV-visible absorption spectrum characteristic of the presence of iron-sulfur species (22), with a band at 410 nm and broad and low-intensity shoulders between 450 and 600 nm (  (Table 1). As shown in Fig. S5C, the [Fe-S] cluster of UbiV Pa was sensitive to air.
Four strictly conserved cysteines (C39, C180, C193, and C197) arranged in a CX n CX 12 CX 3 C motif (where X represents any amino acid) are found in UbiV Pa (9). To test if these four cysteines are important for the chelation of the [4Fe-4S] cluster present in UbiV Pa , we generated two double mutants (C39AC180A and C193AC197A) and a triple mutant (C39AC193AC197A). All these mutants were colorless after purification under aerobic conditions and did not show any absorption band in the 350to 550-nm region of their UV-visible spectra (Fig. S5D), suggesting that they were impaired in their capacity to accommodate a [Fe-S] cluster. After reconstitution under anaerobic conditions, UbiV Pa C39AC193AC197A precipitated and its UV-visible spectrum could not be recorded. Although they also had a tendency to aggregate, 10% of the double mutants behaved as monomers, permitting us to perform some assays. Overall, their absorbance at 410 nm (Fig. 5B) and their iron  or the same vector carrying the corresponding WT allele (ubiT Pa , ubiU Pa , and ubiV Pa ). The parental strain PAO1 was used as a control (Ct), and the UQ 9 content of WT and ubiTUV-KO strains cultured anaerobically was assayed (n = 3). B, ubiTUV-KO strains were cultured in denitrification medium supplemented with methanol-solubilized UQ 4 at 5 or 50 mM final concentration. After 24 h of incubation, the numbers of CFU per ml (CFU/ml) of each KO strain were estimated and compared with those of the same strain grown without UQ 4 . As a control (P . 0.05 by unpaired Student's t test), the toxicity of methanol was tested on the parental strain grown in the same medium but supplemented with 4.5% (v/v) methanol (Ct), which is the final concentration of methanol corresponding to the adding of 50 mM UQ 4 . Data are representative of three independent experiments (**, P , 0.01; ***, P , 0.001; ****, P , 0.0001; by unpaired Student's t test compared with condition without addition of UQ 4 ; ns, not significant). C, As described for panel A, but in soft-agar tubes after overnight culture. All the strains studied were inoculated into anaerobic tubes and then exposed to ambient air to create an oxygen gradient. The controls correspond to soft-agar tubes supplemented with 2.5 mg/ml resazurin and then incubated with (1) or without (2) air. Oxic and anoxic parts of the agar are indicated. For all strains containing pSW196 vectors, denitrification medium was also supplemented with a 0.1% (w/v) final concentration of arabinose to induce the P BAD promoter. and sulfur contents (Table 1) were largely decreased compared with those of the WT protein, suggesting that the four conserved cysteines are good candidates as ligands of the [4Fe-4S] cluster present in UbiV Pa .

Recombinant UbiU Pa and UbiT Pa copurify with UQ 8 in E. coli
We have recently demonstrated that isoprenoid quinones were able to coelute with the Ubi proteins, such as UbiJ (11), and that UbiT exhibits a sterol carrier protein 2 (SCP2) domain, which is able to bind lipids (9). To that end, we performed lipid content analysis of the UbiT Pa , UbiU Pa , and UbiV Pa fractions purified from E. coli extracts. No isoprenoid quinones were detected coeluting with UbiV Pa (Table S4). In contrast, UQ 8 and DMQ 8 (2-octaprenyl-3-methyl-6-methoxy-1,4-benzoquinone) were shown to copurify with UbiU Pa . This protein was purified only in the presence of detergent, as it was insoluble without it. After a two-step purification protocol, including nickel-nitrilotriacetic acid (Ni-NTA) chromatography and SEC, the solubilized protein still had a tendency to form different oligomeric states, covering the fractions 14-44, as shown in Fig. 6, A and B. UQ 8 and DMQ 8 were mainly detected in the elution fractions 33-45 (Fig. 6A), corresponding to only a portion of the purified UbiU Pa (Fig. 6B). The highest contents, i.e. 488.17 pmol UQ 8 per mg of protein and 19,932 AU of DMQ 8 per mg of protein, were assayed in fractions 39 and 40, respectively (Table S4). This corresponds to a UQ 8 /protein ratio of 1.5%. Taken together, these results show that UQ 8 and DMQ 8 copurify with UbiU Pa depending on its oligomerization state.
Due to a lack of solubility, UbiT Pa was overproduced with E. coli thioredoxin (TrxA) as a gene fusion partner, as previously described (25). After the first step of the purification process (Ni-NTA chromatography), the 32-kDa TrxA-UbiT Pa fusion protein was digested with thrombin. To remove the TrxA Histagged protein, UbiT Pa then was purified, by Ni-NTA chromatography coupled to SEC, as a high oligomeric form (Fig. 6C). The purified UbiT Pa (pool of fractions 20-30) contained 9.75 6 4.57 pmol UQ 8 per mg of protein (Table S4), which corresponds to a UQ 8 /protein ratio of about 0.03%.

Recombinant UbiU Pa and UbiT Pa bind the isoprenoid tail of UQ
To further confirm the ability of UbiT Pa and UbiU Pa to bind UQ, a protein-lipid overlay assay was performed (Fig. 6D). We checked the possibility of these proteins recognizing UQ 10 , UQ 8 , solanesol, 3-methylcatechol, 1-palmitoyl-2-oleoyl-snglycero-3-phosphoethanolamine (POPE), and cholesterol. Solanesol is a noncyclic terpene alcohol that consists of nine isoprene units, as found in UQ 9 . 3-Methylcatechol was chosen to mimic the head group of UQ. POPE is the major lipid component of the inner membrane of E. coli (26). Finally, cholesterol was used as a sterol standard. Figure 6D shows that UbiT Pa and UbiU Pa did not interact with 3-methylcatechol, POPE, or cholesterol under our experimental conditions. In contrast, both proteins were able to recognize UQ 10 , UQ 8 , and solanesol. We established the ability of UbiT Pa to bind phosphatidic acid (PA), as previously demonstrated by Groenewold et al. (25) (Fig. S6). Together, we show that UbiT Pa and UbiU Pa are able to bind the isoprenoid tail of UQ, in agreement with their involvement in the O 2 -independent biosynthetic pathway of UQ.

Discussion
UQ acts as a membrane-embedded electron and proton shuttle and is a key molecule in the respiratory metabolism of proteobacteria. The biosynthesis of UQ under aerobic conditions has been widely studied and includes a series of enzymatic reactions in which a benzene ring undergoes a series of modifications involving a prenylation, a decarboxylation, three  methylations, and three hydroxylations (27). Our chemical analysis identified UQ 9 as the major quinone in the membranes of aerobically grown P. aeruginosa cells, which is in agreement with the literature (7,16). We found in P. aeruginosa homologues of the genes known to be involved in UQ biosynthesis in E. coli, except ubiF (Table S1). Indeed, as already published, P. aeruginosa exhibits a yeast COQ7 protein homolog, which catalyzes the same reaction as UbiF from E. coli (15,16). Therefore, both bacteria share a similar UQ biosynthetic pathway involving three hydroxylases, UbiI, UbiH, and UbiF, or COQ7, using O 2 as a cosubstrate (Fig. S2). As no other isoprenoid quinone was detected in lipid extracts of P. aeruginosa, we suppose that UQ 9 is essential for aerobic growth of this bacterium.
In the absence of oxygen, P. aeruginosa can grow by dissimilatory nitrate respiration by using nitrate or nitrite as alternative terminal electron acceptors of the respiratory chain. This metabolic process, known as denitrification, has been widely studied (3). However, the component acting to transfer electrons from primary dehydrogenases to nitrate or nitrite reductases was not clearly identified to date (3). In the present study, we identified UQ 9 as the major quinone synthesized under anaerobic conditions. By LC-MS analysis, we also identified two related redox molecules, UQ 8 and DMQ 9 , present in small amounts (see Fig. 1B, the peaks around 8 and 9.5 min) and nonaprenylphenol (NPP; see Fig. S2). NPP and DMQ 9 are UQ 9 biosynthetic intermediates (Fig. S2), and UQ 8 was already detected in Pseudomonas lipid extracts under aerobic conditions (16). Taken together, these results suggest that P. aeruginosa possesses an O 2 -independent UQ biosynthetic pathway, which produces the major quinone species observed under anaerobic conditions. We recently identified such a pathway in E. coli (9). Here, we characterized three genes, ubiT Pa (PA3911), ubiU Pa (PA3913), and ubiV Pa (PA3912), as essential for the anaerobic UQ 9 biosynthesis in P. aeruginosa and dispensable for the aerobic one. These genes are homologs to the ubiT, ubiU, and ubiV genes previously identified in E. coli, which grows normally under anaerobic conditions in a UQ-independent manner because of the presence of naphthoquinones (9,28,29). In contrast, we demonstrated that these three genes were essential to anaerobic denitrification metabolism of P. aeruginosa, which is in agreement with the presence of a single quinone corresponding to UQ. In line with our results, ubiV Pa (PA3912) and ubiU Pa (PA3913) are expressed in response to anaerobic conditions (30), and the abundance of UbiU Pa protein was increased during anaerobic growth (31). Moreover, using random transposon mutagenesis, ubiV Pa (PA3912) and ubiU Pa (PA3913) were already reported as essential for anaerobic growth of P. aeruginosa on nitrate and nitrite as alternative terminal electron acceptors of the respiratory chain (21).
Although its contribution is still poorly understood for within-host growth, anaerobic respiration of P. aeruginosa is likely to be significant for promoting virulence mechanisms in chronic lung infections (13). Indeed, the infected endobronchial mucus of CF patients is subject to severe hypoxia or even anoxia (32). A likely hypothesis is that accelerated O 2 consumption in the biofilm results from activated polymorphonuclear leukocytes that produce superoxide (33) and nitric oxide (34). Indeed, high levels of nitrate and nitrite have been measured in sputum from CF patients (35). From all these observations, and as UQ is an essential component of the denitrification metabolism in P. aeruginosa, we propose that UbiT Pa , UbiV Pa , and UbiU Pa contribute to the CF lung infection in patients (work in progress in our laboratory). This hypothesis is supported by a recent quantitative proteomics approach revealing the increased abundance of the three proteins in anaerobic biofilms grown under conditions of the cystic fibrosis lung (25). Moreover, as deduced from a high-throughput sequencing of Tn libraries from P. aeruginosa strain PA14, it appears that the ubiT gene was found to be essential for this bacterium to colonize the murine gastrointestinal tract (36), which suggests that O 2 -independent UQ biosynthesis is essential for bacterial virulence. This hypothesis is also supported by the essential contribution of UbiU and UbiV homologs to Yersinia ruckeri virulence (37).
As already suggested, homologs of UbiU and UbiV would belong to a new family of O 2 -independent hydroxylases (9). To date, only the MoCo-containing hydroxylases using waterderived oxygen are known to catalyze hydroxylations under anaerobic conditions (20). In our study, we have demonstrated that MoCo is not essential to the anaerobic UQ pathway, strengthening the hypothesis that hydroxylation reactions performed in a UbiU-and UbiV-dependent manner do not involve MoCo.
To better understand their functions, we decided to overproduce in E. coli, purify, and biochemically characterize UbiT Pa , UbiU Pa , and UbiV Pa . Our results showed that recombinant UbiV Pa is an air-sensitive Fe-S-containing protein, as UbiV from E. coli (9), and we demonstrated that cysteines 39, 180, 193, and 197 were ligands to the [4Fe-4S] cluster found in UbiV Pa . These results confirm the conservation of a four-cysteine pattern coordinating an Fe-S cluster across homologs of UbiV. This pattern is also found in RlhA and TrhP (38,39), two proteins that belong to the same protease U32 family as UbiU and UbiV, and are also involved in O 2 -independent hydroxylation reactions in E. coli (38,39). However, the function of the iron-sulfur centers in the hydroxylation mechanism remains to be understood.
As a member of the U32 protease family, UbiU Pa also presents four conserved cysteines (C169, C176, C193, and C232). Unfortunately, we failed to reconstitute an Fe-S cluster and instead obtained protein precipitation. Indeed, UbiU Pa is an unstable protein. Unlike UbiU from E. coli, which forms a stable heterodimer UbiU-UbiV complex (9), we were not able to solubilize UbiU Pa by coproducing it with its potential part-ner, UbiV Pa . The fact that we produced P. aeruginosa proteins in E. coli could explain this difference in behavior. Nevertheless, a UbiU-UbiV complex in P. aeruginosa remains a reasonable hypothesis that needs further investigation. We were able to purify UbiU Pa alone with significant quantities of UQ 8 and DMQ 8 , whereas purified UbiV Pa contained no quinones. This result was confirmed by a protein-lipid overlay assay, which showed that the isoprenoid tail of UQ was the structural determinant for the recognition by UbiU Pa . From these results, we propose that UbiU would bind UQ and reaction intermediates of the anaerobic UQ pathway.
Homologs of UbiT and UbiJ contain an SCP2 domain (9,40), involved in protein-lipid interactions, and UbiJ from E. coli copurified with UQ 8 (11). Moreover, a previous study showed that PA3911 (UbiT Pa ) was able to bind specifically PA, the central hub of phospholipid metabolism (25). Here, we showed that UbiT Pa binds to UQ 8 and shares with UbiU Pa the recognition of the isoprenoid tail of UQ. Taken together, these results support the hypothesis that UbiT is the counterpart of UbiJ under anaerobic conditions. We propose that UbiT binds UQ intermediates and stabilizes a putative anaerobic Ubi complex that has yet to be demonstrated.

Experimental procedures
Bacterial strains and growth conditions P. aeruginosa and E. coli strains used in this study are listed in Table S2. We obtained the collection of transposon (Tn) mutants in the P. aeruginosa MPAO1 strain from the Manoil Laboratory, Department of Genome Science, University of Washington (41,42). The Tn insertion sites of the mutant strains were verified by sequencing (GATC Biotech, Konstanz, Germany) using PCR primers recommended by the library curators. P. aeruginosa strains were aerobically maintained at 37°C on lysogeny broth (LB) agar plates. For quinone assay, aerobic cultures (5 ml) were performed in LB medium at 37°C with rotary shaking at 200 rpm. Anaerobic growth of P. aeruginosa was performed in 12-ml Hungate tubes containing LB medium supplemented with KNO 3 as an electron acceptor (100 mM final concentration) (43), here called denitrification medium, and deoxygenated for 30 min by argon bubbling (O 2 at ,0.1 ppm) prior to autoclaving. In some experiments, LB medium was supplemented with arginine at a final concentration of 40 mM instead of KNO 3 (44). Hungate tubes were inoculated through the septum with 100 ml of overnight cultures taken with disposable syringe and needles from closed Eppendorf tubes filled to the top. Cultures in Hungate tubes were used for measuring the quinone contents. For aerobic growth studies, aerobic overnight cultures were used to inoculate a 96-well plate to obtain a starting optical density at 600 nm (OD 600 ) of 0.05 and further incubated with shaking at 37°C. Changes in OD 600 were monitored every 10 min for 12 h using the Infinite 200 PRO microplate reader (Tecan, Lyon, France). For anaerobic growth curve studies, overnight cultures in 50-ml closed tubes of non-degassed denitrification medium were used to inoculate 400-ml bottles to obtain a starting OD 600 of 0.05. The bacteria then were grown anaerobically by sparging argon (O 2 at ,0.1 ppm), and bacterial cultures were monitored spectrophotometrically (OD 600 ) at 30-min intervals for 9 h. E. coli MG1655 and DH5a were grown on LB agar or in LB liquid. When required, the medium was supplemented with ampicillin at 100 mg/ml for E. coli, carbenicillin at 250 mg/ml for P. aeruginosa, tetracycline at 60 mg/ml for E. coli and 100 mg/ml for P. aeruginosa, or gentamicin at 200 mg/ml for P. aeruginosa (Table  S2). When necessary, UQ 4 or 0.1% (w/v) arabinose, final concentration, was added to the medium to enhance bacterial anaerobic growth or to induce P BAD expression of pSW196derived plasmids, respectively. Pseudomonas isolation agar medium (from DB) containing irgasan (25 mg/liter) was used for triparental mating to counterselect E. coli. For CFU counting, bacteria were suspended in PBS and cell suspensions were serially diluted in PBS. For each sample, 100 ml of at least four different dilutions were plated on LB plates and incubated for 24 h at 37°C, and CFU were counted using a Scan 100 Interscience.

Plasmids and genetic manipulations
The plasmids and the primers used in this study are listed in Tables S2 and S4. All the plasmids produced in this work were checked using DNA sequencing (GATC Biotech, Konstanz, Germany). To generate P. aeruginosa deletion mutants, overlapping upstream and downstream flanking regions of ubiT Pa , ubiU Pa , and UbiV Pa genes were obtained by PCR amplification using the PAO1 genome as the template and the oligonucleotides described in Table S5. The resulting fragments were then cloned into SmaI-cut pEXG2 plasmid by sequence-and ligation-independent cloning (45). To complement the mutants, the ubiT Pa , ubiU Pa , and ubiV Pa fragments were generated by PCR amplification using the oligonucleotide pairs ubiT-PA-F/ ubiT-PA-R, ubiU-PA-F/ubiU-PA-R, and ubiV-PA-F/ubiV-PA-R, respectively, and PAO1 genome as the template (Table S5). The fragments were EcoRI-SacI digested and inserted into the P BAD -harboring pSW196 plasmid, yielding the pubiT Pa , pubiU Pa , and pubiV Pa plasmids, respectively (Table S2). The pEXG2-and pSW196-derived vectors were transferred into the P. aeruginosa PAO1 strain by triparental mating using pRK2013 as a helper plasmid (46). For allelic exchange using the pEXG2 plasmids, cointegration events were selected on Pseudomonas isolation agar plates containing gentamicin. Single colonies then were cultured on NaCl-free LB agar plates containing 10% (w/v) sucrose to select for the loss of the plasmid, and the resulting sucrose-resistant colonies were checked for mutant genotypes by PCR. To overproduce C-terminally His-tagged UbiV Pa , the ubiV Pa gene was cloned into the pET22b(1) vector. The ubiV Pa insert was obtained by PCR amplification using the oligonucleotide pair pET22-UbiV-F and pET22-UbiV-R and the ubiV Pa ORF as a template (Table  S5). NdeI-XhoI-digested amplicon was ligated to NdeI-XhoIdigested pET22b(1) vector to obtain pET22-UbiV Pa (Table  S2). Variants of UbiV Pa were obtained using the Q5 sitedirected mutagenesis kit (New England Biolabs) (UbiV Pa C193AC197A and UbiV Pa C39AC193AC197A) and the Quik-Change II XL site-directed mutagenesis kit (Agilent) (UbiV Pa C39AC180A) according to the manufacturer's specifications, using pET22b-UbiV Pa as the template (Table S2 and S5). The ubiU Pa gene was cloned into pET-22b(1) by following the same protocol as that for the ubiV Pa gene. The ubiT Pa gene was synthesized by Eurofins with E. coli codon optimization. The synthetic gene then was cloned into the EcoRI/NotI sites of vector pET32a(1) (Novagen), resulting in plasmid pET32-TrxA-UbiT Pa (Table S2).

Soft-agar study to evaluate the O 2 dependence of growth
Soft-agar studies were performed in denitrification medium supplemented with agar at a 0.7% (w/v) final concentration. After argon bubbling (O 2 at ,0.1 ppm) for 30 min, the suspension (13 ml) was autoclaved in Hungate tubes. They were then placed in a 40°C incubator and inoculated through the septum with 100 ml of overnight cultures taken with disposable syringe and needles from Eppendorf tubes filled to the top, mixed by inverting, and incubated at room temperature for 30 min to allow the agar to solidify. The tubes then were incubated under aerobic conditions with caps loosened at 37°C for 24 h. A control experiment was performed with resazurin (0.25 mg/ml final concentration), used as an indicator of medium oxygenation. When required, the medium was supplemented with antibiotics.

Lipid extractions and quinone analysis
Cultures (5 ml under ambient air and 13 ml under anaerobic conditions) were cooled down on ice before centrifugation at 3200 3 g at 4°C for 10 min. Cell pellets were washed in 1 ml ice-cold PBS and transferred to preweighted 1.5-ml Eppendorf tubes. After centrifugation at 12,000 3 g at 4°C for 1 min and elimination of supernatant, the cell wet weight was determined (;5-30 mg) and pellets were stored at 220°C. Quinone extraction from cell pellets was performed as previously described (18). Lipid extracts corresponding to 1 mg of cell wet weight were analyzed by HPLC electrochemical detection-MS (ECD-MS) with a BetaBasic-18 column at a flow rate of 1 ml/min with mobile phases composed of methanol, ethanol, and a mix of 90% isopropanol, 10% ammonium acetate (1 M), and 0.1% TFA with mobile phase 1 (50% methanol, 40% ethanol, and 10% mix). When necessary, MS detection was performed on an MSQ spectrometer (Thermo Scientific) with electrospray ionization in positive mode (probe temperature, 400°C; cone voltage, 80 V). Single-ion monitoring detected the following compounds: UQ 8  ), m/z 724-725, 8-13 min, scan time of 0.2 s. MS spectra were recorded between m/z 600 and 900 with a scan time of 0.3 s. ECD and MS peak areas were corrected for sample loss during extraction on the basis of the recovery of the UQ 10 internal standard and then were normalized to cell wet weight. The peaks of UQ 8 and UQ 9 obtained with electrochemical detection were quantified with a standard curve of UQ 10 as previously described (18).
Overproduction and purification of UbiV, UbiU, and UbiT from P. aeruginosa in E. coli WT and UbiV Pa variants were expressed and purified as previously described for E. coli proteins (9). Briefly, the pET-22b (1) plasmid, encoding WT or UbiV Pa variants, were cotransformed with pGro7 plasmid (Takara Bio Inc.) into E. coli BL21 (DE3) DubiUV competent cells grown at 37°C in LB medium, which was supplemented with ampicillin (50 mg/ml), kanamycin (50 mg/ml), and chloramphenicol (12.5 mg/ml). At an OD 600 of 1.2, D-arabinose was added to the cultures at a final concentration of 2 mg/ml. At an OD 600 of 1.8, cultures were cooled down on ice for 20 min, and IPTG was added at a final concentration of 0.1 mM. Cells then were allowed to grow further at 16°C overnight. WT UbiV Pa and the different variants were purified by Ni-NTA chromatography followed by SEC in buffer A (50 mM Tris-HCl, 25 mM NaCl, 15% [v/v] glycerol, pH 8.5) containing 1 mM DTT. The purified proteins were concentrated to 30-40 mg/ml using Amicon concentrators (30-kDa cutoff; Millipore).
The overproduction of WT UbiU Pa was performed in E. coli BL21(DE3) DubiUV cells by following the same protocol as that for UbiV Pa , except that UbiU Pa overexpression was induced with 0.05 mM of IPTG, and the cell pellets were resuspended in buffer B (50 mM Tris-HCl, 500 mM NaCl, 15% [v/v] glycerol, pH 8.5) containing 0.2% (w/v) N-lauroylsarcosine sodium salt. After cell disruption by sonication, the clarified cell-free extracts were loaded onto a His-Trap FF crude column (GE Healthcare) pre-equilibrated with buffer B containing 0.1% (w/v) N-lauroylsarcosine sodium salt. The column was washed with 10 column volumes of buffer C (50 mM Tris-HCl, 500 mM NaCl, 15% [v/v] glycerol, 10 mM imidazole, pH 8.5) containing 6 mM N,N-dimethyldodecylamine N-oxide (LDAO) to remove nonspecifically bound E. coli proteins and then eluted with a linear gradient of 10 column volumes of buffer C containing 500 mM imidazole and 6 mM LDAO. Fractions containing UbiU Pa were pooled and then loaded on a HiLoad 16/600 Superdex 200 pg (GE Healthcare) pre-equilibrated in buffer D (50 mM Tris-HCl, 150 mM NaCl, 15% [v/v] glycerol, pH 8.5) containing 3 mM LDAO. The purified proteins were concentrated using Amicon concentrators (100-kDa cutoff; Millipore), aliquoted, frozen in liquid nitrogen, and stored at 280°C. For protein-lipid overlay, fractions 34-43 were pooled.
Overproduction of UbiT Pa fused with the thioredoxin (TrxA-UbiT Pa ) in E. coli BL21(DE3) DubiUV cells was performed by following the same protocol as that for UbiU Pa , except that overexpression of the chimeric gene was induced at an OD 600 of 0.5 and the cell pellet was resuspended in buffer B containing 5% (w/v) sodium cholate. TrxA-UbiT Pa was first purified by following the same protocol as that for UbiU Pa by Ni-NTA chromatography, except that the HisTrap FF crude column was preequilibrated with buffer B containing 0.5% (w/v) sodium cholate and then eluted with buffer C containing 500 mM imidazole and 0.5% (w/v) sodium cholate. Fractions containing TrxA-UbiT Pa were pooled and detergent was removed using a Hiprep 26/10 desalting column (GE Healthcare) preequilibrated with buffer D. The fusion protein was digested with thrombin (10 units/mg of TrxA-UbiT Pa) at room temperature and then loaded on a HiLoad 16/600 Superdex 200 pg (GE Healthcare) coupled with a HisTrap FF crude column (GE Healthcare) preequilibrated with buffer D. The purified proteins were concentrated using Amicon concentrators (100-kDa cutoff; Millipore), aliquoted, frozen in liquid nitrogen, and stored at 280°C.

[Fe-S] cluster reconstitution
The [Fe-S] cluster(s) of holo-UbiV Pa and holo-variants was reconstituted as previously described (9). Briefly, a solution containing 100 mM of metalloproteins was treated with 5 mM DTT for 15 min at 20°C and then incubated for 1 h with a 5fold molar excess of both ferrous ammonium sulfate and L-cysteine. The reaction was initiated by the addition of a catalytic amount of the E. coli cysteine desulfurase CsdA (1-2% molar equivalent) and monitored by UV-visible absorption spectroscopy. After 1 h of incubation, the holo-proteins were then loaded onto a Superdex 75 Increase 10/300 GL column (GE Healthcare) preequilibrated with buffer A. The fractions containing the holo-proteins were pooled and concentrated to 20-30 mg/ml on a Vivaspin concentrator (30-kDa cutoff).

Protein-lipid overlay
To assess the lipid-binding properties of UbiT Pa and UbiU Pa , a protein-lipid overlay was performed as previously described (47). Briefly, 2 ml of 20 mM lipids in dichloromethane was spotted onto PVDF membrane and allowed to dry at room temperature for 1 h. The membranes were blocked in 3% (w/v) fatty acid-free BSA in TBST (50 mM Tris-HCl, 150 mM NaCl, and 0.1% [v/v] Tween-20, pH 7.5) for 1 h. The membranes were then incubated overnight at 4°C with gentle stirring in the same solution containing 0.2 mg/ml of the indicated proteins. After washing six times for 30 min in TBST buffer, the membranes were incubated for 1 h with a 1/1000 dilution of anti-polyHis mAb (Sigma) and then for 1 h with a 1/10,000 dilution of anti-mouse-horseradish peroxidase conjugate (Thermo Fisher Scientific). His-tagged proteins bound to the membrane by virtue of its interaction with lipid were detected by enhanced chemiluminescence using Clarity Max Western ECL substrate (Bio-Rad).

Quantification methods
Protein concentrations were determined using the method of Bradford (Bio-Rad), with BSA as the standard. The iron and acid-labile sulfide were determined according to the method of Fish (48) and Beinert (49), respectively, before and after [4Fe-4S] cluster reconstitution.

UV-visible spectroscopy
UV-visible spectra were recorded in 1-cm-optic-path quartz cuvettes under aerobic conditions on a Cary 100 UV-visible spectrophotometer (Agilent) and under anaerobic conditions in a glove box on an XL-100 Uvikon spectrophotometer equipped with optical fibers.

Data availability
All data are contained within the manuscript and supporting material.