Role of the Escherichia coli ubiquinone-synthesizing UbiUVT pathway in adaptation to changing respiratory conditions

ABSTRACT Isoprenoid quinones are essential for cellular physiology. They act as electron and proton shuttles in respiratory chains and various biological processes. Escherichia coli and many α-, β-, and γ-proteobacteria possess two types of isoprenoid quinones: ubiquinone (UQ) is mainly used under aerobiosis, while demethylmenaquinones (DMK) are mostly used under anaerobiosis. Yet, we recently established the existence of an anaerobic O2-independent UQ biosynthesis pathway controlled by ubiT, ubiU, and ubiV genes. Here, we characterize the regulation of ubiTUV genes in E. coli. We show that the three genes are transcribed as two divergent operons that are both under the control of the O2-sensing Fnr transcriptional regulator. Phenotypic analyses using a menA mutant devoid of DMK revealed that UbiUV-dependent UQ synthesis is essential for nitrate respiration and uracil biosynthesis under anaerobiosis, while it contributes, though modestly, to bacterial multiplication in the mouse gut. Moreover, we showed by genetic study and 18O2 labeling that UbiUV contributes to the hydroxylation of ubiquinone precursors through a unique O2-independent process. Last, we report the crucial role of ubiT in allowing E. coli to shift efficiently from anaerobic to aerobic conditions. Overall, this study uncovers a new facet of the strategy used by E. coli to adjust its metabolism on changing O2 levels and respiratory conditions. This work links respiratory mechanisms to phenotypic adaptation, a major driver in the capacity of E. coli to multiply in gut microbiota and of facultative anaerobic pathogens to multiply in their host. IMPORTANCE Enterobacteria multiplication in the gastrointestinal tract is linked to microaerobic respiration and associated with various inflammatory bowel diseases. Our study focuses on the biosynthesis of ubiquinone, a key player in respiratory chains, under anaerobiosis. The importance of this study stems from the fact that UQ usage was for long considered to be restricted to aerobic conditions. Here we investigated the molecular mechanism allowing UQ synthesis in the absence of O2 and searched for the anaerobic processes that UQ is fueling in such conditions. We found that UQ biosynthesis involves anaerobic hydroxylases, that is, enzymes able to insert an O atom in the absence of O2. We also found that anaerobically synthesized UQ can be used for respiration on nitrate and the synthesis of pyrimidine. Our findings are likely to be applicable to most facultative anaerobes, which count many pathogens (Salmonella, Shigella, and Vibrio) and will help in unraveling microbiota dynamics.

(2). Isoprenoid quinones are composed of a quinone ring and a polyisoprenoid side chain whose length varies between organisms (for instance, UQ 8 in E. coli and UQ 9 in Pseudomonas aeruginosa). Many proteobacteria, such as E. coli, produce two main types of quinones: benzoquinones, represented by UQ, and naphthoquinones, such as MK and demethylmenaquinone (DMK). In respiratory chains, quinones transfer electrons from primary dehydrogenases to terminal reductases. For decades, E. coli aerobic and anaerobic respiratory chains were thought to rely on UQ and MK/DMK, respectively. Yet, we have recently discovered a new pathway for UQ biosynthesis under anaerobiosis, opening the way to a more complex and redundant model for bacterial respiratory metabolism (3).
Aerobic UQ biosynthesis pathway includes nine steps (4) (Fig. S1). It begins with the conversion of chorismate to 4-hydroxybenzoate (4HB) by the chorismate lyase UbiC. Then, the phenyl ring of the 4HB precursor undergoes condensation with a 40-carbonlong isoprenoid chain in a reaction catalyzed by the UbiA enzyme. Subsequently, a series of modifications on the 4HB ring by two methylases (UbiE and UbiG), a two-component decarboxylase (UbiD, UbiX), and three hydroxylases (UbiI, UbiH, and UbiF) generate the final UQ 8 product. The flavin adenine dinucleotide monooxygenases UbiI, UbiH, and UbiF use molecular O 2 for their hydroxylation reaction (5)(6)(7). An atypical kinase-like protein called UbiB is also involved in UQ 8 synthesis, but its exact role remains elusive (8). In addition, two non-enzymatic factors are required, UbiJ and UbiK, which may allow UbiIEFGH enzymes to assemble in a cytoplasmic 1 MDa complex, referred to as the Ubi metabolon (9). Also, UbiJ and UbiK bind lipids, which may help the hydrophobic UQ biosynthesis to proceed inside a hydrophilic environment.
Anaerobic UQ biosynthesis is formed by a subset of the enzymes of the aerobic pathway, namely UbiA, UbiB, UbiC, UbiD, UbiE, UbiG, and UbiX, that function with UbiT, UbiU, and UbiV proteins solely required under anaerobiosis (3) (Fig. S1). Like its homolog counterpart UbiJ, UbiT contains an SCP2 lipid-binding domain. Strikingly, UbiU and UbiV do not exhibit any sequence similarity or functional relatedness with the hydroxylases UbiI, UbiH, or UbiF. UbiU and UbiV each contain an iron-sulfur ([4Fe-4S]) cluster coordinated by four conserved cysteine residues embedded in the so-called protease U32 domain, and they form a soluble UbiUV complex (3). Interestingly, two other members of the U32 protein family, RlhA and TrhP, are involved in hydroxylation reactions. They introduce specific nucleotide modifications, respectively, in the 23S rRNA or in some tRNAs (10)(11)(12).
In this work, we aimed at identifying the conditions under which UbiUVT proteins are produced and the genetic regulatory mechanisms involved, and the physiological role of UbiUVT. We concluded that (i) thanks to Fnr control, UbiUV ensures the pro duction of UQ under a range of O 2 levels, from anaerobiosis to microaerobiosis, (ii) a dual anaerobic/aerobic regulation allows UbiT to secure a rapid shift from anaerobic UbiUV-dependent UQ synthesis to an aerobic UbiIHF-dependent UQ synthesis, and (iii) UbiUV-synthesized UQ can be used for nitrate respiration and anaerobic pyrimidine biosynthesis. We also showed that UbiUV acts as O 2 -independent hydroxylases paving the way for future studies toward the characterization of a new type of chemistry.

Strain constructions
Most knockout strains were obtained by generalized Φ P1 transduction using donor strains from the Keio collection (13). For introducing the sequential peptide affinity (SPA) tag on the chromosome or for the generation of specific knockouts, PCR recombi nation with the lambdaRed system was used, using the oligonucleotides indicated in Table 1 (14,15). When necessary, the antibiotic resistance marker was removed using FLP recombinase expression from plasmid pCP20 as described previously (16). Cassette removal and plasmid loss were verified by antibiotic sensitivity and confirmed by PCR amplification. Point mutations were introduced on the chromosome using the pKO3 vector (17).
For mouse intestine colonization experiments, we used MP7 and MP13 strains, which derive from the commensal E. coli MP1 strain (22). MP7 and MP13 express, respectively, mCherry or green fluorescent protein (GFP) under the control of a tetracycline-inducible promoter. ∆menA and ∆ubiUV deletions were introduced in MP13 using generalized Φ P1 transduction.

Plasmid constructions
pUA66 and pUA-ubiUVp plasmids were obtained from the library of E. coli promoters fused to the GFP coding sequence (23). The ubiT transcriptional fusions were constructed using primers indicated in Table 3 and cloned in XhoI/BamHI sites of pUA66. Expression plasmids for ubiUV and fnr were constructed using primers indicated in Table 3 and cloned in EcoRI/SalI sites of the pBAD24 vector (24). Expression plasmids for ubiIHF and ubiM_Neisseria genes were constructed using primers indicated in Table 3 and cloned in EcoRI/XhoI sites of pTet vector. A region of 1,275 base pairs encompassing ubiU and ubiT promoters was cloned in pKO3 vector (17). Mutations were introduced in the pKO3-ubiTU vector, in the pBAD-ubiUV, and in the transcriptional fusions by PCR mutagenesis on a plasmid, using the oligonucleotides indicated (Table 2 and 3).

Media and growth conditions
Strains were grown in LB Miller (10 g/L of tryptone, 10 g/L of NaCl, and 5 g/L of yeast extract) or M9 medium (6 g/L Na 2 HPO 4 •7H 2 O, 3 g/L KH 2 PO 4 , 0.5 g/L NaCl, 1 g NH 4 Cl, 2 mM MgSO 4 , 1 mg/mL thiamine) supplemented with 0.2% glucose, 0.2% glycerol, or 50 mM succinate as the carbon source. For anaerobic cultures, media were degassed and incubated in anaerobic environment for at least 24 hours prior to use, if necessary supplemented with 25 mM KNO 3 as electron acceptor and uracil 25 μg/mL or casamino acids at 0.05%.
For microaerobic experiments, media and plates were pre-equilibrated and cells were cultured in a Whitley H35 hypoxic station with 95% N 2 , 5% CO 2 , and the desired O 2 concentration. Humidity and temperature were set up at 85% and 37°C, respectively. For anaerobic-aerobic shift experiments, all anaerobic steps were performed in a JACO MEX Campus anaerobic chamber under N 2 atmosphere at 1 ppm O 2 maximum. Cells were first incubated anaerobically in LB agar plates supplemented with 0.2% glucose overnight at 37°C. The next day, cells were cultured anaerobically in 3-mL LB supplemen ted with 25 mM NO 3 − for 24 hours at 37°C. Still under anaerobiosis, cells were collected by centrifugation, supernatant was discarded, and pellets were washed twice using 1 mL of M9 medium without carbon source and normalized at 0.1 OD units in M9 medium supplemented with 50 mM sodium succinate. At this point, cultures were moved out to atmospheric air, and growth was followed by triplicate at 37°C on 200 µL of culture in a 96-well plate using a TECAN infinite M200 plate reader. At 40 hours of culture, cells were diluted 1/20 in a new M9 50 mM sodium succinate medium and readings were resumed until 60 hours.

Aerobic and anaerobic cultures for quinone analysis
For aerobic cultures, 5 mL of LB medium, supplemented with ampicillin (100 µg/mL) and 0.05% arabinose when necessary to induce the expression from the pBAD vectors, was inoculated with 100 µL of overnight culture in glass tubes (15 cm long and 2 cm in diameter) and incubated at 37°C, 180 rpm overnight.
Anaerobic cultures were performed in Hungate tubes as previously described (3). Briefly, LB medium was supplemented with 100 mM KNO 3 as the final electron acceptor, 100 mg/L L-cysteine (adjusted to pH 6 with NaOH) to reduce residual molecular oxygen, and 2.5 mg/L reasazurin. This medium was distributed in Hungate tubes and deoxygen ated by high-purity argon bubbling for 40 minutes. The Hungate tubes were sealed and autoclaved. The resazurin was initially purple, it turned to pink after deoxygenation and become colorless after autoclave. The preculture was performed overnight at 37°C in Eppendorf tubes filled to the top with LB medium containing 100 mM KNO 3 . The Hungate tubes were then inoculated through the septum with disposable syringes and needles with 100 µL of precultures and incubated at 37°C without agitation. The resazurin remained colorless during culture indicating anaerobic conditions. For anaerobic to aerobic shift assay, MG1655 WT, ΔubiUV, and ΔubiT strains were grown anaerobically in Hungate tubes for ~4 hours. Then, 26 µL of chloramphenicol (200 µg/mL) was injected through the septum with a Hamilton syringe. After 20 minutes, the Hungate tubes were unsealed and 2 mL of cultures was taken for lipid extraction and quinone analysis. The rest of cultures was transferred to 250-mL Erlenmeyer flasks and placed at 37°C and 180 rpm for 2 hours. Two-milliliter aliquots of cultures were taken at 30 minutes and 120 minutes after the transition to ambient air for lipid extraction and quinone analysis.

SDS-PAGE and western blotting
Total cell extracts were prepared by resuspending cell pellets in Laemli buffer 1× at a concentration of 0.3 optical density at 600 nm (OD 600 nm ) units in 10 µL, and then heating for 10 minutes at 95°C. After the separation of 8 µL of total cell extracts on SDS-PAGE, electrotransfer onto nitrocellulose membranes was performed using Trans-Blot turbo transfer system from Bio-Rad. After blocking in phosphate-buffered saline (PBS) 1× + milk 5%, SPA-tagged proteins were detected with monoclonal anti-Flag M2 antibody purchased from Sigma. YbgF protein was used as an internal control and

Transcriptional fusions with GFP
We used several clones from the E. coli transcriptional fusions library (23) and we constructed the required additional transcriptional fusions (see above for plasmid construction and Table 2). ∆fnr E. coli strain was co-transformed with plasmids carrying the gfp transcriptional fusions and compatible pBAD24 or pBAD-fnr plasmids. Selection plates were incubated at 37°C for 16 hours. Six hundred microliters of LB medium supplemented with kanamycin and ampicillin, and with 0.02% arabinose for pBADdriven expression, were incubated (four biological replicates for each assay) and grown for 16 hours at 37°C in 96-well polypropylene plates of 2.2-mL wells in anaerobiosis. Cells were pelleted and resuspended in PBS supplemented with 30 µg/mL chloramphenicol and incubated at 4°C for 1 hour before fluorescent intensity measurement was per formed in a TECAN infinite M200 plate reader. One hundred fifty microliters of each well was transferred into a black Greiner 96-well plate for reading OD 600 nm and fluorescence (excitation: 485 nm; emission: 530 nm). The expression levels were calculated by dividing the intensity of fluorescence by OD 600 nm , after subtracting the values of a blank sample. These results are given in arbitrary units because the intensity of fluorescence is acquired with an automatic optimal gain and hence varies from one experiment to the other.

Lipid extraction and quinone analysis
Cultures of 2, 5, or 10 mL were cooled on ice for at least 30 minutes before centrifu gation at 3,200× g at 4°C for 10 minutes. Cell pellets were washed in 1-mL ice-cold PBS and transferred to pre-weighted 1.5-mL Eppendorf tubes. After centrifugation at 12,000× g at 4°C for 1 minute, the supernatant was discarded, the cell wet weight was determined, and pellets were stored at −20°C until lipid extraction, if necessary. Quinone extraction from cell pellets was performed as previously described (6). The dried lipid extracts were resuspended in 100 µL ethanol, and a volume corresponding to 1 mg of cell wet weight was analyzed by high performance liquid chromatography (HPLC) electrochemical detection-mass spectrometry (ECD-MS) with a BetaBasic-18 column at a flow rate of 1 mL/minute with a mobile phase composed of 50% methanol, 40% ethanol, and 10% of a mix (90% isopropanol, 10% ammonium acetate [1 M], and 0.1% formic acid). 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 880-881, 10-17 minutes. MS spectra were recorded between m/z 600 and 900 with a scan time of 0.3 seconds. 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 obtained with electrochemical detection or MS detection were quantified with a standard curve of UQ 10 as previously described (6).

O 2 labeling
MG1655 wild type (wt) and ΔubiIΔubiHΔubiF containing, respectively, the pBAD24 empty vector or pBAD-ubiUV were grown overnight at 37°C in LB medium supplemented with ampicillin (100 µg/mL) and 0.05% arabinose. These precultures were used to inoculate 20 mL of the same fresh medium at an OD 600 of 0.05 in Erlenmeyer flasks of 250 mL. The cultures were grown at 37°C, 180 rpm, until an OD 600 of 0.4-0.5 was reached. An aliquot was taken for lipid extraction and quinone analysis (0 minute of 18 O 2 ), and 13 mL of each culture was transferred to an Hungate tube. Five milliliter of labeled molecular oxygen ( 18 O 2 ) was injected through the septum with disposable syringes and needles, and the incubation was continued at 37°C, 180 rpm for 2 hours. Then 5 mL of each sample was taken for quinone analysis (120 minutes of 18 O 2 ).

Mouse intestine colonization experiments
Four-week-old female BALB/cByJ were purchased from Charles River Laboratories (Saint-Germain-Nuelles) and were acclimatized in a controlled animal facility under specific pathogen-free conditions for 2 weeks prior to the beginning of the coloniza tion assay. Mice were randomly assigned to groups of three or five per cage, and ear punching was used to identify each mouse in a given cage.
The colonization experiments were adapted and performed as previously described (25,26). Mice were given drinking water containing streptomycin sulfate and glucose (both 5 g/L) for 72 hours to remove existing resident anaerobic facultative microflora.
For the clearance of streptomycin, freshwater devoid of antibiotics and glucose was then given to mice for 48 hours before the inoculation of E. coli strains and for the rest of the experiment. To start the competition experiment, the mice were orally inoculated with 200 µL of a mixture in a 1:1 ratio of the two competing strains at ~20,000 cells/mL in PBS. Mice from each cage were orally inoculated with the same solution of bacteria. An aliquot of inoculum was plated on LB agar containing 15 µg/mL tetracycline to compute the input value.
The relative abundance of both competing strains was then monitored for several days post-inoculation in fecal samples. Fecal samples were collected from each mouse in pre-weighed 1.5-mL Eppendorf tubes containing the equivalent of 100-µL glass beads (diameter 0.25-0.5 mm) and 80-µL PBS, and the feces weight was determined. A volume of PBS was then added to each tubeto obtain a final concentration of 0.15 g of feces per 1-mL PBS. The feces were homogenized by vortexing for 2 minutes, serially diluted by 10-fold steps up to a 10 5 -fold dilution, and aliquots of 70 µL were plated on LB agar medium containing 15 µg/mL tetracycline. The plates were incubated overnight at 37°C and were transferred at 4°C for at least 2 hours the following day, before imaging under blue light which revealed the fluorescent markers carried by each colony. The red and green colonies corresponding, respectively, to MP7 and MP13 strains were counted by an adapted version of ImageJ. Then, the CFU was computed per gram of feces for each strain and a competitive index (CI) was calculated as a ratio of (MP13 mutant CFU/MP7 wt CFU)/(input MP13 mutant CFU/input MP7 wt CFU), where the input CFU was determined from the inoculum for which an aliquot was plated on the day of gavage. The limit of detection in fecal plate counts was 10 2 CFU/g feces. At all time points, the wt strain was detectable on the fecal plates. The absence of CFU count and CI for 1 day in one mouse corresponds to the absence of feces for that day. Significance of CI was calculated by GraphPad Prism using one sample t-test compared to one.

Biochemical function of UbiUV in vivo
To get further insight into the UbiUVT system in vivo, we tested whether the overproduc tion of UbiU and UbiV could substitute for the three oxygen-dependent hydroxylases UbiI, UbiH, or UbiF. Thus, we cloned the ubiUV operon in the pBAD24 vector downstream the arabinose-inducible pBAD promoter (pES154 plasmid). In parallel, we also cloned ubiUV upstream the SPA tag encoding sequence to assess the quantities of proteins produced. The pBAD-ubiUV-SPA plasmid produces a level of UbiV protein approximately 30-fold higher than that produced by a chromosomal copy of ubiV-SPA under anaerobio sis (Fig. S2). After the transformation of mutant strains, selection, and precultures with LB medium in absence of O 2 , growth on M9 succinate was tested as it strictly depends on an aerobic UQ-dependent respiratory chain (Fig. 1). In the presence of an inducer, the pES154 plasmid was able to suppress the growth phenotype of the ∆ubiF, ∆ubiH, ∆ubiIK, and ∆ubiIHF mutants (Fig. 1A). Note that as a control, we used the Neisseria meningitidis ubiM gene that we previously showed to substitute for the growth phenotype of a ∆ubiIHF mutant (19). Also, in M9 succinate, the ∆ubiI mutation alone has no growth phenotype and needs to be combined with ∆ubiK mutation for a defect to be observed (27). To test the importance of the UbiU-bound [Fe-S] cluster, a complementation test was carried out in the same conditions, using a pBAD derivative carrying the ubiU(C176A) allele that produces an UbiU variant lacking its [Fe-S] cluster (3). Accordingly, the suppression of ∆ubiH, ∆ubiF, ∆ubiIK, and ∆ubiIHF was no longer observed (Fig. 1A). In addition, the pES154 plasmid was unable to suppress the growth phenotype of ∆ubiA, ∆ubiD, ∆ubiE, or ∆ubiG strains (data not shown) and was also unable to suppress the growth phenotype of ∆ubiH∆ubiA or ∆ubiH∆ubiD mutants (Fig. 1B), showing that UbiUV intervene specifically at the hydroxylation steps and otherwise depend on all the other components of the aerobic UQ biosynthesis pathway to do so. These results indicate that in the presence of O 2 , expression of UbiUV can substitute for the O 2 -dependent UbiIHF hydroxylases and that integrity of the UbiU [Fe-S] cluster is required.
Remarkably, the expression of the pES154 plasmid was also able to suppress growth defects of the ∆ubiJ mutant ( Fig. 2A). UbiJ is an auxiliary factor important for organizing the aerobic Ubi metabolon. We reasoned that suppression was made possible thanks to the presence of the chromosomally encoded UbiT that shares sequence similarity with UbiJ. To test this, we repeated the complementation test in two new strains, ∆ubiH∆ubiJ and ∆ubiH∆ubiT. The pES154 plasmid still complemented the growth defects of the ∆ubiH∆ubiJ mutant, but it was unable to complement the ∆ubiH∆ubiT mutant (Fig. 2B).

Research Article mBio
Similarly, pES154 was found to suppress the growth defect phenotype of a ∆ubiF∆ubiJ mutant but not a ∆ubiF∆ubiT mutant (Fig. 2C). These results showed that in the presence of O 2 , increased dosage of ubiUV genes suppresses the lack of O 2 -dependent hydroxyla ses UbiF and UbiH in an UbiT-dependent/UbiJ-independent manner. To confirm that phenotypic suppression was due to UQ 8 synthesis, we quantified the UQ 8 content by HPLC analysis coupled to electrochemical detection (ECD) for all strains described above (Fig. 3A). Results showed that mutant strains lacking UbiI-UbiK, UbiH, and/or UbiF were severely deficient in UQ. The pES154 plasmid-enabled ∆ubiH, ∆ubiF, or ∆ubiIH strains to synthesize 30%-50% of the UQ level of the wt strain (Fig. 3A, first panel). The levels of UQ obtained in the ∆ubiIH∆ubiF and ∆ubiIK mutant strains with the pES154 plasmid were much lower. We stress that the UQ levels cannot be directly correlated with the phenotypic analysis ( Fig. 1 and 2) since culture media were different (LB versus M9 succinate) to allow the recovery of enough biological material for the HPLC-ECD analyses. Importantly, the pBAD-ubiU(C176A)V plasmid was unable to promote UQ synthesis in ∆ubiH (Fig. 3A, second panel). Last, UQ 8 content assay confirmed that UbiT, but not UbiJ, was necessary for UbiUV to synthesize UQ in aerobic conditions (Fig. 3A, third panel).
The results above showed that UbiUV hydroxylate UQ precursors, when expressed under aerobic conditions. This result raised the possibility that under such conditions, O 2 might be used as a co-substrate of the hydroxylation reactions, as is the case for UbiI, UbiH, and UbiF in wt cells (5). To test this hypothesis, we exposed cells to 18 O 2 and monitored the labeling of UQ by HPLC-ECD-MS. Two hours after 18 O 2 addition, the level of UQ 8 increased in both strains (Fig. 3B). Before adding 18 O 2 , the mass spectra of UQ synthesized by wt or ΔubiIHΔubiF cells containing pES154 displayed H + and NH 4 + adducts with m/z ratio characteristic of unlabeled UQ (Fig. 3C and D). As expected, 2 hours after adding 18 O 2 , most of the UQ 8 pool in wt cells contained three 18 O 2 atoms (Fig. 3E), in agreement with O 2 being the co-substrate of the aerobic hydroxylation steps (5). In contrast, we detected only unlabeled UQ 8 in the ΔubiIHΔubiF strain expressing UbiUV (Fig. 3F), demonstrating that UbiUV utilizes another oxygen donor than O 2 , even when operating under aerobic conditions. Altogether, both phenotypic and UQ 8 quantification results allowed us to conclude that UbiU and UbiV, when produced at sufficiently high level, function in the canonical "aerobic" UQ 8 biosynthesis pathway by catalyzing [Fe-S]-dependent hydroxylation of the benzene ring in an O 2 -independent reaction. Remarkably, UbiT is necessary for such aerobic UbiUV-mediated synthesis to occur and cannot be substituted by UbiJ.

The ISC [Fe-S] biogenesis machinery is required for anaerobic UQ biosynthe sis
The UbiU and UbiV proteins each contain a [4Fe-4S] cluster, which is essential for the synthesis of UQ 8 in anaerobic conditions (3). Assembly of [4Fe-4S] clusters requires complex biosynthetic machineries, ISC and SUF (28). Therefore, the UQ 8 levels were monitored in Δisc and Δsuf mutants grown in anaerobic conditions (Fig. S3). UQ 8 content in Δisc mutants was strongly impaired (around 15% of the wt), while it was much less affected in ∆suf mutants (60%-80% of the wt). This indicated that the ISC system contributes to anaerobic UQ 8 biosynthesis likely through the maturation of [4Fe-4S] clusters in UbiU and UbiV. An alternative explanation would be that isopentenyl phos phate (IPP), which is the precursor of UQ 8 and whose synthesis depends on [4Fe-4S] containing IspG and IspH proteins, is not efficiently synthesized in the Δisc mutants. However, DMK 8 and MK 8 levels, which also rely on IspG/IspH-synthesized IPP, remained mostly unaltered in the Δisc mutants. It is likely that in this case, the SUF system takes over in a more efficient way as it does for maturating UbiU and UbiV proteins. Collec tively, these results showed that the ISC system and to some minor extent the SUF system are necessary for anaerobic UQ 8 biosynthesis.

Anaerobic and microaerobic UQ biosynthesis
Genome-scale studies have predicted that ubiUV genes are under the control of the anaerobic Fnr transcriptional activator (29,30). In contrast, ubiT did not appear as a potential Fnr target. This prompted us to investigate the effect of anaerobiosis (0% O 2 ), microaerobiosis (0.1% O 2 ), and aerobiosis (21% O 2 ) on the level of UbiU, UbiV, and UbiT proteins. To follow the quantity of UbiTUV proteins in physiological conditions, we constructed a series of recombinant strains producing the UbiT, UbiU, or UbiV proteins with a C-terminal SPA tag (14) encoded from a gene fusion at their chromosomal loci. We examined protein production by western blot assay using an anti-flag antibody and assessed loading with a polyclonal antibody against YbgF (CpoB). All three UbiTUV-SPA tagged proteins were present in strains grown in anaerobiosis (Fig. 4A) and microaero biosis (Fig. 4B). In aerobiosis, the production of UbiU and UbiV was no longer observed, whereas a significant level of UbiT was still visible. The contribution of Fnr to anaerobio sis-or microaerobiosis-mediated activation of ubiU and ubiV genes was confirmed as no cognate UbiU or UbiV-associated band was observed in a ∆fnr mutant (Fig. 4A and B). Interestingly, UbiT level was also reduced in the ∆fnr mutant in −O 2 . Last, to validate the physiological significance of the Fnr regulatory circuit depicted above, we quantified the amount of UQ 8 produced in wt and ∆fnr strains, during aerobiosis and anaerobiosis (Fig.  4C). In comparison to the UQ content found in the wt strain in aerobiosis, the level in anaerobiosis was reduced by half. Importantly, we observed that almost no UQ was

Genetic control of ubiUVT gene expression
Previous genome Chip-seq analysis reported binding of Fnr within the ubiT-ubiUV intergenic region. Additionally, in a whole-genome sequence search study, one transcrip tion start site has been described upstream of the ubiUV operon (ubiUV p ) and two sites described upstream of ubiT (ubiT p1 , ubiT p2 ) (31) (Fig. 5A and B). On inspection of that Research Article mBio region, we were able to identify two potential Fnr-binding sites fitting well with the described Fnr-binding consensus. The F1 site, reading TTGATTTAAGGCAG is located 36 nucleotides (nt) upstream the ubiUV p transcription start site (Fig. 5A). The F2 site reading TTGATTTATACCGC locates 33 nt upstream the proximal +1 transcription starting site ubiT p2 and 19 nt downstream the distal ubiT p1 (Fig. 5A and B).
To detail the molecular mechanism of regulation and to dissect the promoter organization of the intergenic region between ubiUV and ubiT, we used transcriptional fusions with GFP (23). We used four different transcriptional fusions encompassing ubiUV p , ubiT p1 , ubiT p2 , and a construction ubiT p1p2 containing the two promoters of ubiT (Fig. 5B). We compared the expression of these transcriptional fusions in anaerobiosis, in a ∆fnr mutant complemented or not with a pBAD-fnr plasmid. The ubiUV p and ubiT p2 promoters were strongly activated in the presence of pBAD-fnr, whereas the ubiT p1 promoter was not ( Fig. 5C and D). This suggested that the ubiUV p promoter was activated by Fnr binding to the F1 site and that the ubiT p2 promoter was activated by Fnr binding to the F2 site. When we introduced mutations in the F1-binding site (five mutated nucleotides; mutF1; Fig. 5A), the activation of the expression from the ubiUVp transcrip tional fusion was severely reduced (Fig. 5C). Mutations of the F2 site (mut∆F2 complete deletion or mutF2 with five mutated nucleotides; Fig. 5A) also affected the expression of the ubiT p1p2 transcriptional fusion, but a basal level of expression was maintained, probably due to the expression from the distal ubiT p1 promoter (Fig. 5E).
Next, we introduced the same mutations in the F1 and F2 Fnr-binding sites at the locus in the ubiU-ubiT intergenic region in the chromosome of the strains producing UbiV-SPA or UbiT-SPA tagged proteins. Mutation within the F1 site upstream ubiU completely prevented the production of UbiV in the absence of O 2 (Fig. 5F). Mutation  (Fig. 5F; Fig. S4). Notably, the mutation in the F1-binding site did not Research Article mBio affect the expression of ubiT and conversely, the mutation of the F2-binding site did not affect the expression of ubiV. Altogether, these results showed that Fnr activates ubiUV transcription under anaerobiosis, while ubiT expression can be triggered from two promoters, one aerobi cally active (P1) and the other anaerobically active (P2) under Fnr control.

Physiological role of UbiUVT at different O 2 levels
We have previously reported that UbiU, UbiV, and UbiT are essential for the anaero bic synthesis of UQ in E. coli when grown in LB, glycerol/DMSO, or lactate/NO 3 − (3). However, the contribution to E. coli physiology of UQ synthesized by UbiUVT in anaerobic conditions was not investigated in detail. We made use of a set of mutants altered in aerobic (ubiH) or anaerobic (ubiUV, ubiT) UQ 8 synthesis, as well as mutants altered in DMK/MK biosynthesis (menA) to assess the contribution of each type of quinone for growth in a wide range of O 2 level, 21% (aerobic), 0.1% (microaerobic), and 0% O 2 (anaerobic), and with varying carbon sources (e.g., glycerol or glucose) and electron terminal acceptors (e.g., O 2 or NO 3 − ). In the presence of glycerol and NO 3 − under aerobic conditions (Fig. 6, upper left panel), ΔubiUV and ΔubiT strains showed no growth phenotype. In such conditions, while NO 3 is present, O 2 is used for respiration. This contrasted with the ΔubiH mutant, which was severely affected. Combining ΔubiH and ΔubiUV bore no aggravating effect. In contrast, combining both ∆ubiH and ∆menA had an aggravating effect, indicating that in addition to UQ, DMK and/or MK can support E. coli growth even in aerobiosis, as previously suggested (32). In microaerobic conditions (Fig. 6, upper center panel), no phenotype was observed for ΔubiUV or ΔubiT strains. In contrast, the ΔmenA ΔubiH strain still exhibited a clear defect, suggesting that ubiUV and ubiT do not bear a prominent role in NO 3 − -dependent respiratory metabolism under microaerobic conditions, despite being expressed in microaerobiosis (see above). This notion was also supported by the fact that at 0.1% O 2 , the ΔubiH and ΔubiHΔubiUV strains did not show any phe notype. At 0.1% O 2 , UQ-dependent metabolism through cytochrome bd or bo oxida ses would remain inconsequential, and cells presumably rely on DMK/MK-dependent metabolism for anaerobic respiration (33). Last, in anaerobic conditions, with NO 3 − used for respiration, ΔubiUV, ΔubiT, and ΔmenA strains showed wt-like growth phenotype (Fig.  6, upper right panel). However, combining ΔmenA and ΔubiUV mutations or ΔmenA and ΔubiT mutations drastically hampered NO 3 − respiratory capacities. In fact, the growth of these mutants on M9 glycerol NO 3 − was barely better than a Δfnr strain (Fig. 6), which was used as a control since it was shown that such strain is unable to respire nitrate but can still use glucose anaerobically (34). These results indicated that anaerobically UbiUVT-synthesized UQ and MK are fully interchangeable electron carriers during NO 3 − respiration under full anaerobiosis (35). Furthermore, we could exclude that the aerobic UQ biosynthetic pathway could contribute to growth in such conditions as the ΔubiH and ΔmenA ΔubiH mutants exhibited no growth phenotype.
In the presence of glucose as a carbon source and under aerobiosis, ΔubiUV and ΔubiT mutants exhibited wt-like growth capacity (Fig. 6, left middle panel). The ∆ubiH mutant showed some slower growth, but a most spectacular negative additive effect was observed on combining ∆ubiH and ΔmenA mutations. It likely points out a role for DMK/MK in aerobic electron transport (36). In anaerobiosis, neither ∆ubiH nor ∆menA, alone or in combination, showed defect in the presence of glucose as a carbon source (Fig. 6, middle right panel). In contrast, ∆menA ∆ubiUV or ∆menA ∆ubiT mutants exhibited additive growth defects (Fig. 6, middle right panel). This indicated that the UbiUVTbiosynthesized UQ was crucial for growth in glucose fermentative conditions, in the absence of MK. A possibility was that this negative effect reflected auxotrophy for uracil, whose synthesis depends on electron transfer from PyrD dihydrooratate dehydrogenase to fumarate reductase (FrdABCD) via quinones in anaerobiosis (37). As a matter of fact, adding uracil to the medium had a rescuing effect (Fig. 6, lower right panel), supporting the notion that uracil deficiency was responsible for the growth defect observed in the ∆ubiUV ∆menA mutant in anaerobiosis. This was an important observation as early studies had proposed that the PyrD/FrdABCD electron transfer chain relied mostly on MK/DMK and marginally, if at all, on UQ (37). Our observation clearly shows that were grown aerobically at 37°C in LB medium or LB glucose 0.2% (for ∆menA ∆ubiH), washed, and resupended in M9 medium without carbon source to OD 600 of 1. Serial dilutions were spotted in agarose M9 medium plates supplemented with carbon source (glycerol or glucose), KNO 3 , or uracil and incubated at 37°C at the indicated O 2 concentration until growth was observed. Experiments were performed in triplicates and confirmed with at least four independent biological replicates.

Research Article mBio
anaerobically synthesized UQ can also allow the functioning of PyrD. Incidentally, we noticed that the addition of uracil did not rescue the growth defect of the ∆menA ∆ubiH mutant in aerobiosis, but we have no explanation for this observation.

Contribution of the O 2 -independent UQ biosynthesis pathway to mouse intestine colonization
Since enterobacteria evolve mostly in anaerobic conditions in their natural habitat, we evaluated the physiological importance of the O 2 -independent UQ biosynthesis pathway in the mouse intestine. To do so, we performed competition experiments between two isogenic strains, MP7 and MP13, which respectively express mCherry and GFP in the presence of tetracycline (22). We deleted ubiUV in the MP13 background and confirmed, as expected, that this strain was deficient for UQ 8 when grown anaerobically (Fig. S5A). MK was previously shown to be important for the efficient colonization of the mouse intestine by E. coli (38). Thus, we also constructed a ∆menA mutant in the MP13 background. We checked that the deletion of ∆menA abrogated the synthesis of DMK and MK ( Fig. S5B and C). The fitness of the ∆ubiUV and ∆menA mutants was tested in competition experiments with the MP7 wt strain. We monitored the abundance of each strain in the feces of mice up to 10 days after co-inoculation by oral gavage (Fig. 7A). In both experiments, the total CFU count reached ~10 8 /g of feces 24-hour post-inoculation ( Fig. 7B and C; Fig. S6A and B) and then gradually decreased to ~10 5 , showing efficient colonization of the MP7 strain. The abundance of the ubiUV mutant was slightly decreased compared to wt ( Fig. 7B; Fig. S6A), which translated into an average CI <1 ( Fig. 7D; Fig. S6C) at days 1, 2, 4, and 10. We noticed, however, a rather high inter-individual variability (Fig. S6C). In contrast, the ∆menA mutant was markedly less abundant than the wt ( Fig. 7C; Fig. S6B) and was even undetectable at day 10. CI <1 were observed for every mouse at every sampling ( Fig. 7E; Fig. S6D), and the values obtained were much lower than in the case of the ∆ubiUV mutant. Collectively, these data confirm that DMK/MK is the most important quinone for the physiology of E. coli in the mouse intestine (38). However, they also reveal a contribution, albeit minor, of the O 2 -independent UbiUV-mediated UQ biosynthesis pathway.

Role of UbiT within the anaerobiosis-aerobiosis shift
Phenotypic analysis above revealed that anaerobically UbiUVT-synthesized UQ 8 was contributing to growth via glucose fermentation or NO 3 − respiration. In both conditions, anaerobic UbiUVT-synthesized UQ 8 was functionally redundant with anaerobically synthesized DMK/MK. Because UQ 8 is crucial under aerobiosis, we wondered whether anaerobically synthesized UQ 8 might prepare the cells to adapt to an aerobic environ ment, that is, before the aerobic UbiIHF-dependent synthesis takes over. Thus, we investigated the role of UbiUVT-synthesized UQ 8 in the anaerobiosis-aerobiosis transi tion.
First, we used ΔmenAΔubiH and ΔmenAΔubiUV strains that only produce UQ 8 under anaerobiosis and aerobiosis, respectively. Strains were grown in LB supplemented with NO 3 − under anaerobic conditions for 24 hours, then switched to aerobic conditions with succinate as a carbon source, that is, in conditions wherein growth strictly relies on UQ 8 (35). The wt strain showed differential efficiency in shifting from anaerobiosis to aerobio sis as compared with the ∆menA and ΔmenAΔubiUV strains. Indeed, by taking the end of the lag period at the time point at which growth resumes an upward trajectory, lag periods were 2 hours for the wt and 7 hours for the ∆menA and the ∆menA∆ubiUV mutant strains (Fig. 8A). This suggested that UbiUVT-synthesized UQ8 does not bear a significant influence on the shift between anaerobiosis and aerobiosis. Eventually, all three strains showed the same growth rate in the exponential phase and reached the same final OD 600 value, suggesting that the UbiIHF-synthesized UQ 8 was activated and provided UQ 8 in extended aerobic conditions. To confirm this hypothesis, we reinoculated these cells into the same medium (Fig. 8A, refresh), and as expected we observed that lag periods were the same for all three strains, indicating that they had accumulated the same level of UQ 8 since the beginning of the growth.
Surprisingly, the ∆menA∆ubiH mutant-a strain defective for the aerobic UQ 8 synthesis pathway-was able to grow after the transition to aerobic conditions, with the same lag period as the ∆menA and ∆menA∆ubiUV mutant strains, that is, 7 hours. This showed that UbiUVT-synthesized UQ 8 in the ∆menA∆ubiH allowed a shift from anaero biosis to aerobiosis. Yet, the ∆menA∆ubiH strain showed a slower and shorter exponential of feces for that day. ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 by one sample t test. Changes in total CFU counts and CI throughout the experiment in each mouse are shown in Fig. S6.

Research Article mBio
phase and a lower final OD 600 value as compared with the wt, ∆menA, and ∆menA∆ubiUV strains. Then, as expected, the ∆menAΔubiH mutant failed to resume growth on reinocu lation in a fresh medium, indicating that the level of anaerobically UbiUV-synthesized UQ 8 failed to sustain protracted aerobic growth (Fig. 8A). The fact that anaerobically synthesized UQ 8 has a positive, yet minor effect on the anaerobic-aerobic transition, somehow contradicted the first conclusion reached when studying the ∆menA and ∆menA∆ubiUV mutant strains (see above). A possible explanation is that in the ∆menA∆ubiUV strain, newly synthesis of UQ 8 by UbiIHF might be quick enough to compensate for the lack of UbiUV-synthesized UQ 8 . UQ 8 content was subsequently measured over a shorter time period during the transition from anaerobic to aerobic conditions in a separate experiment (Fig. 8B). For Research Article mBio this, cultures in LB of ∆ubiUV or ∆ubiT mutants were subjected or not to chloramphenicol (Clp) treatment prior to the shift, and samples were taken at 0 minute, 30 minutes, and 120 minutes for UQ quantification. UQ 8 level increased with time in both the wt and the ∆ubiUV mutant, but in the 30-120 minute period, it stopped increasing in the presence of translation inhibitor Clp. The likeliest explanation is that UQ 8 biosynthesis is driven by UbiUV before the shift and later de novo synthesized by UbiIHF in aerobic conditions. This suggested that the three hydroxylases UbiI, H, and F were already present under anaerobiosis, in a standby state, waiting for O 2 to allow hydroxylation. Importantly, this was confirmed as levels of UbiI, H, and F proteins were found to be similar in both aerobic and anaerobic conditions (Fig. S7). Also, this is consistent with the hypothesis of a very quick synthesis of UbiIHF-synthesized UQ 8 (see above). Second, the role of the accessory factor, UbiT, was investigated using the ∆menA ∆ubiT mutant. As described before, the ∆menA∆ubiT strain was grown first in LB with NO 3 − under anaerobiosis, subsequently shifted in succinate minimal medium, and growth was monitored. A most unexpected and spectacular effect was observed as a lag period with this strain in these conditions was approximately 20 hours whereas that of the wt was approximately 2 hours (Fig. 8A). The ∆menA∆ubiT strain finally reached a final OD 600 value similar to WT, ∆menA, ∆menA∆ubiUV strains at 40 hours and also resumed growth on re-inoculation at 40 hours (Fig. 8A). This highlighted a crucial role of UbiT in the anaerobic-aerobic transition phase. This result was strengthened by direct quantification of UQ 8 synthesized with time after shifting cultures from anaerobiosis to aerobiosis (Fig. 8B). The ∆ubiT mutant exhibited a two-fold reduction in UQ 8 as compared with the ∆ubiUV mutant after the shift. When Clp was added, the difference was much smaller. This confirmed that UbiT is necessary at the onset of aerobic UQ 8 biosynthesis, presumably via the UbiIHF complex.

The yhbS gene is not involved in UQ 8 -based metabolism
The yhbS gene predicted to encode an acetyltransferase lies downstream the ubiT gene (Fig. S8A). It was recently proposed to intervene in small noncodingRNA (sncRNA)-medi ated expression control (39). Using RT-PCR, we showed that yhbS and ubiT genes share a single transcription unit (Fig. S8B). Using YhbS-SPA tag protein, we observed that YhbS protein synthesis takes place both under aerobiosis and anaerobiosis. The level of YhbS-SPA protein appears slightly higher in −O 2 , and this induction seems to be lost in the ∆fnr mutant, as expected if yhbS and ubiT genes are co-expressed and co-regulated by Fnr (Fig. S8C). The ∆yhbS mutant shows no defect in NO 3 − respiratory capacity, and no aggravating effect was observed on combining ∆yhbS and ∆menA mutations (Fig. S8D). Last, we carried out shift experiments, from −O 2 to +O 2 , as described above for ubiT and failed to identify any defect in the ∆yhbS mutant (not shown). Altogether with previous assays failing to reveal a defect in UQ 8 levels in anaerobiosis in the ∆yhbS mutant (3), these results allowed us to rule out a role of YhbS in UQ 8 synthesis.

DISCUSSION
UQ is an essential component of electron transfer chains and of respiratory metabolism. For decades, the dogma has been that UQ was exclusively used for aerobic respiratory metabolism, whereas DMK/MK was used for electron transfer in anaerobic respiratory chains. Following our recent discovery that UQ is also synthesized under anaerobiosis, which contradicted the above assumption (3), the present study identified two versatile anaerobic physiological processes that rely on the anaerobic UQ biosynthesis pathway, namely NO 3 − respiration and uracil biosynthesis. Moreover, we provide clear evidence that UbiUV catalyzes hydroxylation steps independently from O 2 . Last, UbiT was found to play a key role in both anaerobiosis and aerobiosis conditions, allowing a smooth transition between the two conditions. Overall, this analysis uncovers a new facet of the strategy used by E. coli to adapt to changes in O 2 levels and respiratory conditions. This is of particular interest in the context of gut microbiota studies, as changes in O 2 level and in respiratory electron acceptors are key factors that the host uses to select the type of flora present through the different sections of the intestine (40).
UbiUV-mediated UQ synthesis takes place under anaerobiosis. Here we showed that this is made possible by Fnr-mediated activation of expression of the ubiUV operon that takes place from microaerobiosis (0.1% O 2 ) to anaerobiosis. In contrast, expression of the ubiT gene is more versatile with two promoters, one under Fnr control, allowing UbiT synthesis under microaerobiosis and anaerobiosis, simultaneously with UbiUV, and the second constitutive one, insuring expression in aerobiosis. This genetic regulation is consistent with the presence of UbiT proteins under both aerobic and anaerobic conditions. Such a versatile expression meets with other evidence we collected, which together pave the way to an important role of UbiT in the anaerobiosis to aerobiosis transition: (i) UbiT is required for insuring continuous UQ synthesis on shifting from anaerobiosis to aerobiosis, (ii) ubiT was found to compensate for the lack of ubiJ in conditions where high dosage of ubiUV genes suppressed absence of ubiIHF under aerobiosis, and (iii) UbiIHF enzymes are present in anaerobiosis but not active as one would expect for O 2 -dependent hydroxylases. This indicates that the O 2 -dependent pathway is in a standby mode in anaerobic conditions, waiting only for the presence of O 2 to activate the O 2 -dependent hydroxylases and produce UQ, as proposed previously (41). This is also consistent with the fact that UbiUV synthesis is strictly controlled at the transcriptional level, whereas expression of ubiIHF is constitutive. Altogether, this leads us to propose that UbiT and UbiJ are required for the formation of two related but distinct metabolons, respectively, an anaerobic one containing UbiUV and an aerobic one containing UbiIHF. Besides, both UbiJ and UbiT are likely to bind UQ biosynthetic intermediates via their SCP2 domain, thereby providing the substrates to UbiUV and UbiIHF (9,42).
UbiUV catalyzes hydroxylation of the benzene ring in the absence of O 2 . Moreover, our results show that they can substitute for aerobic hydroxylases UbiIHF in the presence of O 2 , but that they still catalyze the hydroxylation without relying on O 2 in this condition. This raises the question of the source of the O atom under anaerobiosis. Previous analysis on RhlA, a member of the U32 protein family to which UbiU and V belong, indicated that prephenate, an intermediate within the aromatic amino acid biosynthesis pathway, could act as an O donor (11). Our ongoing studies aim at investigating such a possibility in the case of anaerobic UQ biosynthesis. [Fe-S] clusters seem to play a role in the process since isc mutants devoid of anaerobic [Fe-S] biogene sis machinery and UbiU variant lacking [Fe-S] cluster fail to produce UQ. The simplest hypothesis is that [Fe-S] clusters are transferring electrons from the O source to a terminal reductase, both to be identified.
UbiUVT-synthesized UQ has a significant contribution to growth in anaerobiosis and microaerobiosis (0.1% O 2 ). Indeed, we found that UbiUVT-synthesized UQ are key for NO 3 − respiration in the absence of DMK, in agreement with early biochemical work on formate-nitrate reductase (37) and with our previous study reporting that P. aeruginosadenitrifying activity depends on UbiUVT-synthesized UQ (42). Moreover, we observed that the anaerobically synthesized UQ greatly contributes to uracil synthesis. This was unexpected as uracil synthesis was reported to depend mainly on the oxidation of (S)-dihydroorotate to orotate with fumarate as a hydrogen acceptor and DMK/MK as an electron carrier (37). Our present physiological studies demonstrate that the anaerobi cally produced UQ can fully compensate for the DMK/MK loss, likely through an as yet unknown reductase since UQ is too electropositive to be a frdABCD substrate (43). Last, UQ could be used as an electron sink to other catabolic processes taking place in both aerobiosis and anaerobiosis such as heme biosynthesis, wherein the HemG enzyme utilizes UQ or MK for the conversion of protoporphyrinogen IX into protoporphyrin IX (44).
The contribution of anaerobically synthesized UQ for E. coli multiplication in the gut appeared as marginal. This implies that either absence of UV-synthesized UQ was masked by MK/DMK synthesis or anaerobic UQ-dependent processes such as NO 3 respiration or uracil biosynthesis is dispensable. Clearly, the first possibility is the likeliest given the paramount importance of anaerobic respiration for E. coli multiplication in the gut (45,46), as nicely confirmed by the drastically altered multiplication of MK/DMK-deficient cells (Fig. 7). This is of particular interest as the presence and nature of respiratory electron acceptors were proposed to be drivers of bacterial community composition in the different regions of the intestine (40). Likewise, the relatively high O 2 level in the duodenum, of NO 3 − in ilium, and hypoxia in the cecum were proposed to be causal of the different flora hosted in these regions in a healthy host. Strategies used by E. coli to live in such different respiratory and fermentative conditions are therefore key aspects of its adaptation to the host. In this context, it is important to understand the mechanism underlying the switch from O 2 -rich to NO 3 − -rich and/or hypoxic compartments, and the present study highlights the added value of having overlapping systems permitting a smooth shift from anaerobic NO 3 − to aerobic respiration.

ACKNOWLEDGMENTS
We thank Marc Fontecave and Murielle Lombard from College de France, and the members of the SAMe unit at Pasteur for discussion and help. We thank Mark Goulian (University of Pennsylvania, USA) for providing the MP7 and MP13 E. coli strains and Laurent Loiseau for providing the UbiUVT SPA-tagged strains. We gratefully acknowledge the help of TrEE team members with the mouse intestine colonization experiments, Françoise Blanquet, Dalil Hannani, Clément Caffaratti, and Amélie Amblard. We are also grateful to Arnold Fertin for developing the ImageJ plugin used for the automatic counting of red and green colonies. This project was supported by Institut Pasteur and CNRS and by grants from the ANR (ANR-10-LABX-62-IBEID and ANR-19-CE44-0014O2-TABOO).