All animal experimental procedures were conducted according to the guidelines of Experimental Animal Management Measures of Jiangsu Province and were approved by the Laboratory Animal Monitoring Committee of Jiangsu Province (China).

Strains and plasmids used in this study are listed in Supplementary Table S1. Aerobic growth was achieved by shaking in air at 160 rpm, and anaerobic growth by incubating in a BugBox chamber (Ruskinn, UK) filled with a gas mixture (N₂, 85%; CO₂, 5%; H₂, 10%). For genetic manipulations, all E. coli strains were grown routinely in lysogenic broth (LB) medium (OXOID). For growth and gene expression studies, bacteria were generally grown aerobically or anaerobically in M9 minimal salts with certain carbon sources indicated, supplemented with 2 mM MgSO₄, 0.1 mM CaCl₂, and 1 μg ml^-1 vitamin B1 (Cai et al., 2013). Glucose (0.4% m/v) or glycerol (0.4% v/v) (Sinopharm, China) was added as energy substrates, as indicated. When used, electron acceptor trimethylamine N-oxide (TMAO) and dicarboxylates were present at 40 mM. Selective antibiotics and IPTG (Isopropyl β-D-1-thiogalactopyranoside) were added when necessary at the following concentrations: ampicillin (Amp), 100 μg ml^-1; kanamycin (Kan), 50 μg ml^-1;chloramphenicol (Chl), 25 μg ml^-1; IPTG, 1 mM. All reagents were purchased from Sigma-Aldrich unless otherwise noted.

For growth studies, fresh colonies of wild-type (WT) CFT073 and its derivative mutant strains were picked and inoculated in 3 ml LB medium. After OD₆₀₀ of bacterial culture reached about 1.0, bacteria were washed once with phosphate-buffered saline (PBS) and normalized, followed by being subcultured 1:100 into M9 medium. OD₆₀₀ value at different time points was measured and recorded using spectrophotometer (Eppendorf, basic model).

Polymerase chain reaction (PCR), DNA ligation, electroporation and DNA gel electrophoresis were performed according to Sambrook and Russell (2001) unless otherwise indicated. All oligonucleotide primers were purchased from BGI Technology Solutions Co., Ltd. (BGI, Guangzhou, China) and are listed in Supplementary Table S2. All restriction and DNA-modifying enzymes were purchased from New England Biolabs and were used based on the suppliers’ recommendations. Recombinant plasmids, PCR products, and restriction fragments were purified using TaKaRa PCR purification kit or gel extraction kit (TaKaRa, a Clontech company) as recommended by the supplier. DNA sequencing was performed at Shanghai Sunny Biotech Co., Ltd. DNA and amino acid sequence analyses were performed using Clone Manager software (Scientific & Educational Software, Morrisville, NC, USA) and the search engine¹ was used to find homologous sequences. ClustalW program was used to perform multiple sequences alignments. Phylogenetic trees were made using MEGA software (Tamura et al., 2011; Hall, 2013).

Deletion mutants were constructed using the lambda red recombinase system described by Datsenko and Wanner (2000). For complementation studies, the coding sequences of genes plus their promoter regions (400 bp upstream of start codons) were amplified from the CFT073 genome and independently cloned into pGEN-MCS (Lane et al., 2007) using EcoRI and SalI restriction sites. Plasmid pCJ112 for plasmid-borne LacZ fusion studies was constructed by replacing the R6K origin in pVIK112 [this plasmid was created for making chromosomal transcriptional fusions of lacZ reporter gene (Kalogeraki and Winans, 1997)] with p15A origin from pBAD30 (Guzman et al., 1995) plasmid using EcoRI and BamHI restriction sites (See Supplementary Figure S1 for detail). The resulting plasmid was tested to show the functionality and an undetectable basal level without inserting any promoters upstream of promoterless lacZ. P_bla constitutively expressed promoter was cloned from pGEN-MCS plasmid by amplifying 400 bp upstream of bla gene as previous described (Sperandio et al., 1999). All constructs were verified by Sanger Sequencing (Shanghai Sunny Biotech Co., Ltd).

Overnight LB cultures of CFT073 and its derivative strains containing the gene of interest-lacZ fusions were washed with PBS once and then were diluted 1:100 in LB or M9 medium with the carbon sources indicated. Empty vector pCJ112 and pCJ112 containing P_bla-lacZ were used under all conditions as controls. Cultures were grown at 37°C to log phase or stationary phase. Samples were diluted 1:1 in Z buffer and assayed for β-galactosidase activity using ortho-Nitrophenyl-β-galactoside (ONPG) as a substrate as described previously (Miller, 1972).

The transcriptional start site (TSS) site of c5038 was identified by using ExactSTART 5′- and 3′ -RACE Kit (Epicentre) with some modifications. Briefly, bacteria were grown in M9 media supplemented with glycerol and KG to log phase. RNA was stabilized by RNAprotect Bacterial Reagent (QIAGEN) and extracted using an RNeasy Mini Kit (QIAGEN) with a 1-h in-tube DNase digestion (QIAGEN) to remove possible DNA contamination according to the manufacturer’s instructions. Complete removal of DNA contamination was confirmed by using reverse transcription PCR. Three biological replicates of each sample were prepared. The concentration of RNA was determined using a Spectrophotometer (Eppendorf, basic model). The Steps A and B were omitted as they were designed for treating Eukaryotic RNA. About 5 μg of RNA were used for the Step C in the manual: 5’-RACE Acceptor Oligo Ligation step. Control reactions were included according to the manual. The primers which yielded reliable PCR products were listed in Supplementary Table S2.

To study the binding of proteins of interest to DNA probes, electrophoretic mobility shift assays (EMSAs) were performed using the commercialized EMSA kit (Invitrogen, Carlsbad, CA, USA) (Shen et al., 2011; Cai et al., 2013). pET28a prokaryotic expression system (Novagen) was used to over-produce recombinant proteins. His₆-ArcA (Jiang et al., 2015), His₆-(FNR-FNR)2 (Shan et al., 2012) and His₆-RpoN fusion proteins were individually purified to homogeneity using Ni-NTA Spin Columns (QIAGEN) and dialyzed against the binding buffer. Protein concentrations were measured using BCA protein assay kit (Pierce). DNA probes (230 bp upstream of the start codon of kgtP or c5038) were PCR amplified using specific primers and gel-purified. The negative control probe was amplified from the coding region of Chloramphenicol resistance gene-cat. The concentrations of DNA fragments were determined by Biospectrometer (Eppendorf). EMSAs were performed by adding increasing amounts of purified fusion proteins (0240 nM) to DNA probes (10 nM) in binding buffer [10 mM Tris (pH 7.5), 1 mM EDTA, 1 mM dithiothreitol, 50 mM KCl, 50 mM MgCl₂, 1 μg ml^-1 bovine serum albumin (NEB)) and incubating for 30 min at room temperature (His₆-ArcA and His₆-RpoN), or 20 min at 37°C (His₆-(FNR-FNR)2]. Reaction mixtures were then mixed with the loading dye (bromophenol blue, 0.25%; xylene cyanol FF, 0.25%; Ficoll 400, 15%), followed by electrophoresis on a 6% polyacrylamide gel in 0.5 × TBE buffer (44.5 mM Tris, 44.5 mM boric acid, 1 mM EDTA, pH 8.0) at 200V for 45 min in an ice-bath. The gel was stained in 0.5 × TBE buffer containing 1 × SYBR Gold nucleic acid staining solution (Life technology, USA) for 15 min. Gels were then visualized and photographed using GelDoc-it Imager (UVP).

Mouse infection studies were performed according to the methods of Johnson et al. (2005). Female Balb/C mice (about 6 to 7 weeks of age) were anesthetized by ketamine/xylazine/Midazolam and inoculated via transurethral catheterization with a 50 μl (∼10⁹ CFU) challenge inoculum per mouse. Overnight statically grown LB cultures of CFT073 and the mutant strains were pelleted and resuspended in sterile PBS, mixed in equal number, and adjusted to make challenge inocula. To determine the initial CFU/ml, dilutions of each inoculum were plated onto LB plates with and without chloramphenicol (kgtP::Chl^R mutant) or kanamycin (c5038::Kan^R mutant). After 48 h, the mice were sacrificed and the bladder and kidneys were aseptically removed and homogenized in tubes containing PBS using a Bioprep-24 bullet homogenizer (Allsheng, Hangzhou in China). Dilutions (10 μl in each droplet) of the homogenized tissue were then spotted onto quadruplicate LB plates with and without antibiotics to determine the bacterial load. The numbers of colonies on selective plates were subtracted from those on LB plates to obtain the number of WT bacteria. A group of 10 or 6 (for complementation studies) mice for each dual-strain challenge were used to determine alterations in fitness, and assays were performed at least twice. Competitive index is defined as the ratio of mutant to the WT isolated divided by the ratio of mutant to the WT in the inoculums. the single-gene deletion mutants

Statistical tests were performed using GraphPad Prism (version 5.0 for Windows, San Diego, CA, USA). For animal experiments data, Wilcoxon signed rank test was used to determine significance of competitive indices with a hypothetical median of 1 (number_WT = number_mutant). For other data, unpaired Student t-test was used to estimate differences between samples. The threshold for statistical significance was a P-value < 0.05.

Gene c5038 is encoded on a metabolic island implicated in the anaerobic utilization of KG in UPEC (Cai et al., 2013). Bioinformatic analysis revealed that C5038 belongs to the Divalent Anion:Na⁺ Symporter (DASS) Family (Saier et al., 2006). Characterized DASS members in bacteria include succinate transporter VcINDY, citrate transporter CitT, L-tartrate/succinate transporter TtdT, C₄-dicarboxylates uptake system DccT, YbhI, and SdcS (Markovich, 2012). C5038 on average shares 29% homology with these proteins. Sequence alignment of these DASS members (Supplementary Figure S2A) revealed that C5038 contains a SAT (serine–alanine–threonine) signature, which is similar to the relatively conserved SNT (serine–asparagine–threonine) motif involved in substrates-binding (Mancusso et al., 2012). The kgtP locus in K-12 is 96.4% identical to kgtP in UPEC str. CFT073 at DNA level, with 47 differences between them. At the amino acid level, only three differences exist between them, all of which are non-charged amino acids. KgtP shares no homology with C5038 at both DNA and amino acid levels. The promoter regions of kgtP in K-12 and CFT073 have an 88% sequence identity to each other (Supplementary Figure S2B).

To determine the roles of C5038 and KgtP in growth of CFT073 in various media, we compared the growth of the WT, the single-gene deletion mutants (Δc5038 and ΔkgtP, respectively), and the double-gene deletion mutant (Δc5038ΔkgtP) in different media under aerobic conditions. We observed no significant difference in growth kinetics among these strains in LB or M9 medium supplemented with glucose (Gluc) or glycerol (Glyc) as the sole carbon source (data not shown).

When cultured in M9 using KG as the sole carbon source under aerobic conditions, the Δc5038 mutant showed similar growth kinetics as WT, indicating that c5038 was not important for growth under these conditions (Figure 1A). In contrast, deletion of kgtP abolished CFT073’s growth (P < 0.0001) and kgtP mutant displayed no detectable growth within 40 h. To rule out the possible masking effect of kgtP on c5038’s role, we compared the growth of two mutants, ΔkgtP mutant and Δc5038ΔkgtP mutant. The results showed no significant difference in growth between ΔkgtP mutant and Δc5038ΔkgtP mutant (Figure 1A), further supporting that c5038 played a negligible role in KG utilization in aerobiosis. Additionally, the introduction of a complementation plasmid (pkgtP) carrying kgtP controlled by its native promoter restored the growth of kgtP mutant (Figure 1B). These results suggest that kgtP is required whereas c5038 is dispensable for growth on KG under aerobic conditions.

Under anaerobic conditions, deletion of kgtP did not affect CFT073’s growth, whereas loss of c5038 dramatically reduced its growth in M9 with KG as the sole carbon source, as compared to WT (Figure 1C). Transformation of the c5038 mutant with a complementation plasmid (pc5038) carrying c5038 controlled by its native promoter significantly improved c5038 mutant’s growth (Figure 1D). To rule out the possible masking effect of c5038 on kgtP’s role under these conditions, we compared the growth of Δc5038 mutant and Δc5038ΔkgtP mutant. We observed a difference in growth between Δc5038 mutant and Δc5038ΔkgtP mutant at 60 h time point and thereafter (P < 0.05) (Figure 1C). These data indicate that under anaerobic conditions, c5038 played a major role and kgtP a minor role in growth on KG. Of note, in both aerobiosis and anaerobiosis, the complementation strains apparently had much shorter lag phases and higher growth rates than the WT strain carrying the empty vector. This was likely due to the fact that the complementation plasmid vector is a multi-copy plasmid (∼15 copies per cell), which can cause overexpression of kgtP or c5038.

Given that kgtP and c5038 contributed differentially to growth on KG under aerobic and anaerobic conditions, we sought to unravel their transcriptional levels under these conditions. The TSS of c5038 was obtained using 5′-RACE PCR (Figure 2B). The –10 and –35 sites of kgtP promoter were predicted by Virtual Footprint program (Munch et al., 2005) and Bprom program (Solovyev and Salamov, 2011), which presented consistent output with high reliability. The putative ribosome binding sites were also indicated. The predicted TSS of kgtP in CFT073 was highly similar to that in K-12 identified by 5′-RACE and deep sequencing (Cho et al., 2009). The promoter regions of kgtP and c5038 were then cloned to drive promoterless lacZ gene on a plasmid, respectively, and the resulting plasmid constructs were individually introduced into LMP10 strain (CFT073 ΔlacZYA). In parallel, a control plasmid carrying constitutively expressed promoter P_bla-lacZ fusion was also transformed into LMP10 strain. β-galactosidase activities were measured to indicate expression levels. As shown in Figure 2C, c5038 expression was not detectable in aerobiosis, but was induced in anaerobiosis. The expression of kgtP was moderate under aerobic conditions, but was greatly repressed under anaerobic conditions, with a 40-fold reduction. It is also evident that anaerobic expression level of c5038 was significantly higher than aerobic expression level of kgtP (about 10-fold difference). In the control strain, P_bla-lacZ expression did not exhibit difference between under aerobic and anaerobic conditions. These results clearly showed that under anaerobic conditions, c5038 was highly expressed whereas kgtP was severely repressed, thereby at least partially explaining the phenotypic difference in growth.

Fumarate and nitrate reduction and ArcA are two master regulators mediating bacterial adaptation to environmental oxygen availability (Constantinidou et al., 2006; Park et al., 2013; Jiang et al., 2015). To probe the regulatory roles of arcA and fnr in expression of c5038 and kgtP, we examined the effects of single deletion and double deletion of fnr and arcA on expression levels of c5038 and kgtP under aerobic and anaerobic conditions. Under aerobic conditions, deletion of fnr did not affect kgtP expression while loss of arcA upregulated kgtP expression about fivefold. Under anaerobic conditions, deletion of either fnr or arcA caused upregulation of kgtP expression (4-fold and 700-fold increase, respectively). Deletion of both fnr and arcA led to a greater increase in kgtP expression than individual deletions of fnr and arcA. These results indicate that both fnr and arcA negatively modulate kgtP expression. Further, in LMP10 strain, anaerobiosis repressed kgtP expression about 40-fold, as compared to aerobic expression, whereas in LMP10Δfnr and LMP10ΔarcA mutants, anaerobiosis repressed kgtP expression 10-fold and 4-fold, respectively (Figure 3A). These results clearly demonstrate that in either fnr or arcA mutant, anaerobic repression of kgtP was partially relieved compared with that in their parental strain LMP10, suggesting both fnr and arcA are involved in the anaerobic repression, with ArcA being a major player.

We next tried to understand the mechanisms by which ArcA and FNR repress the expression of kgtP. Sequence analyses of promoters repressed by ArcA or FNR suggest that ArcA and FNR inhibit gene expression through competitive binding to –10 or/and –35 region, thereby blocking occupancy of RNA polymerases (Myers et al., 2013; Park et al., 2013). Using a recently developed binding consensus sequence of ArcA ([GT][TA][TA][AG][AC][AT][AT][AT][AT][AT], each pair of letters in each bracket indicate two most frequent bases in that position, with the first being the most frequent) (Park et al., 2013), we were able to identify one perfect match (GTTGCATAAT) that overlaps the putative –35 region and two imperfect matches close to the perfect one (Figure 2A). Similarly, using a recently developed binding consensus sequence of FNR ([TA][TC][GA][AC][TC]nnnn[AG][TA][CT][AG][AC]) (Myers et al., 2013), we were able to identify a perfect match (ATGATatgcAACAA) that lies between –10 and –35 region. To test whether ArcA and FNR directly binds to the promoter region of kgtP, EMS assays were carried out. DNA fragments with sizes around 230 bp containing the predicted sites were prepared. Purified His₆-tagged ArcA and FNR variant (FNRD154A) proteins (Shan et al., 2012) were obtained by nickel affinity chromatography. As shown in Figure 3B, the DNA probe of kgtP promoter region can be shifted by ArcA protein, but the negative control probe cannot be shifted. With the same amounts of protein, the presence of acetyl phosphate appears to not affect binding of ArcA to DNA under our testing conditions. Also, FNR recombinant protein can shift the DNA probe of kgtP, but not the control probe (Figure 3C). These results indicate that ArcA and FNR can bind to the promoter region of kgtP directly. Altogether, these data suggest that ArcA and FNR likely repress kgtP expression through binding to promoter region and occluding the occupancy of RNAP.

For c5038, deletion of either fnr or arcA or both resulted in a significant reduction in c5038-lacZ expression in comparison to that in the LMP10 strain (P < 0.01), suggesting both fnr and arcA positively modulate c5038 expression. We next used EMSA to test if ArcA directly associates with DNA probe of c5038 promoter region. The results show that ArcA protein cannot shift the DNA probes (Supplementary Figure S3), raising the possibility that ArcA may not regulate c5038 expression directly. Taken together, these data indicate that FNR and ArcA stimulated c5038 but repressed kgtP expression.

In contrast to c5038, what stimuli induce kgtP’s expression is still unclear. To test if the expression of kgtP is influenced by the addition of KG, we measured and compared the expression levels of kgtP during bacterial growth in M9 (Glyc), M9 (Glyc and KG), and M9 (KG), respectively. Surprisingly, kgtP expression in all three media is comparable (Figure 4A), indicating that kgtP expression does not respond to the presence of KG. We next tested and compared the expression of kgtP in LB rich medium, M9 (Gluc), and M9 (Glyc). The results showed that the order of kgtP expression levels is as follows: M9 (Glyc) > M9 (Gluc) > LB.

The cAMP-CRP complex is a global regulatory element that induces a number of genes involved in the utilization of carbons in the absence of glucose (Botsford and Harman, 1992; Kolb et al., 1993), including glycerol catabolic regulon (Iuchi et al., 1990). We therefore examined whether CRP contributes to the regulation of kgtP expression. As shown in Figure 4B, kgtP expression in the strain lacking crp was downregulated, as compared to the parental strain LMP10. This result indicates that CRP positively regulates kgtP expression. Collectively, these data suggest that kgtP might be responsive to the carbon state of the cell, and be used for carbon scavenging.

KguR involved in direct activation of c5038 was predicted to belong to the category of bacterial enhancer-binding proteins (bEBPs) which interact with σ⁵⁴ factor (Bush and Dixon, 2012), raising the possibility that c5038 is a new member of the σ⁵⁴ regulon. The “A,” –30 site relative to the start codon was identified to be the TSS of c5038 (Figure 2B). Further analysis of the promoter region of c5038 did reveal a motif that is highly similar to the σ⁵⁴ binding consensus “–24 to –12” region (Figure 2B), suggesting that c5038 transcription might depend on sigma54. To assess the contribution of rpoN (the gene coding for sigma54) in c5038 expression, we compared the expression levels of c5038 in LMP10 with that in LMP10ΔrpoN. As shown in Figure 5A, deletion of rpoN abolished the KG induction of c5038 expression, as compared to parental strain LMP10. To ensure that no unwanted mutation was responsible for such phenotype, a complementation study was performed by introducing a copy of plasmid-borne rpoN controlled by its putative native promoter. The result showed that with rpoN reintroduced, the expression level of c5038 was rescued.

Sigma54 protein was shown to be capable of binding to target promoter regions independent of core RNA polymerase (Shingler, 2011). To evaluate if σ⁵⁴ protein directly associates with c5038 promoter region, EMSA was used to study the binding of σ⁵⁴ proteins to DNA molecules. σ⁵⁴ was over-produced in E. coli as His₆-RpoN fusion protein, which was then purified to homogeneity against nickel column. glnA promoter region, which was previously shown to be directly associated with RpoN (Huo et al., 2006), was included as a positive control; on the other hand, xylA promoter region, a member of σ⁷⁰ regulon (Song and Park, 1997; Desai and Rao, 2010), was included as a negative control. As we expected, His₆-RpoN fusion protein can shift P_glnA and P_c5038 DNA probes, but not the P_xylA (Figure 5B; Supplementary Figure S4) probe. Altogether, these data indicate that c5038 is a new member in the RpoN regulon.

Because RpoN is required for c5038 expression which is important for growth of CFT073 on KG, we then examined the growth kinetics of the WT and ΔrpoN mutant in M9 (KG) containing abundant nitrogen source NH₄Cl under anaerobic conditions. The results demonstrate that ΔrpoN mutants had a longer lag phase and grew significantly slower than their parental WT strain (P < 0.001) (Figure 5C). These data suggest that RpoN contributes to anaerobic utilization of KG, possibly through regulating c5038 expression.

In contrast to c5038, deletion of rpoN had no effect on kgtP expression (Figure 5A). We have shown earlier that kgtP expression remains high in M9 (Glyc), which is considered to be a carbon- and amino acid-limited environment. RpoS, also known as sigma38 factor, is a stress-responsive global regulator, which is induced under stress conditions such as carbon, phosphorus, nitrogen, or amino acid scarcity (Notley and Ferenci, 1996). Thus, rpoS was deleted from LMP10 strain, and its effects on kgtP expression were determined. Unexpectedly, loss of RpoS increased kgtP expression by about twofold, irrespective of growth phase (Figure 5D). Therefore, RpoS is a negative modulator of kgtP expression.

Now that c5038 and kgtP were demonstrated to play differential roles in vitro, we next tried to investigate if loss of c5038 or kgtP affects UPEC fitness in vivo. Mice were challenged transurethrally with a 1:1 mixture of WT and a mutant, followed by recovering bacteria from tissues 48 h post inoculation and calculating competitive indices. Note that when inoculated in LB medium with 1:1 ratio, any two strains contained in the inoculums were recovered equally in numbers during exponential growth phase, indicating that they were equally fit in vitro. Figure 6A showed that the c5038 mutant was significantly outcompeted by the WT in both bladder and kidneys (P-value equals 0.0005 and 0.0458, respectively). In vivo complementation experiments were performed to confirm that the mutation is not polar and the disadvantage is not caused by a secondary mutation (Figure 6B). The stable low-copy pGEN plasmid was used as it was shown to be well maintained in CFT073 up to 48 h even in the absence of antibiotic pressure. The Δc5038 mutant carrying empty vector displayed the expected competitive disadvantage in bladder and kidneys when co-challenging mice with the WT containing pGEN (P-value equals 0.0161 and 0.0342, respectively) whereas Δc5038 harboring a copy of c5038 on pGEN plasmid was equally fit as WT (pGEN). Therefore, reintroduction the c5038 gene into the c5038 mutant can potentially complement the negative effect on colonization caused by mutation of c5038.

In contrast to the c5038 mutant, the kgtP mutant colonized murine urinary tracts as well as the WT (Figure 6A, CI was not significantly different from the hypothetical value 1). To rule out the possible masking effect of c5038 on kgtP, we tested the effects of mutating kgtP in Δc5038 genetic background on fitness in vivo. Deletion of kgtP in Δc5038 background had no impact on colonization of mice by CFT073, further corroborating that kgtP is not important for colonization of mice by UPEC. Taken together, these results suggest that c5038 but not kgtP is important for UPEC fitness in vivo.

In light of distinct roles of C5038 and KgtP in KG utilization and in vivo fitness of CFT073, their phylogenetic trees were constructed in order to understand their evolutionary routes (Escherichia spp. strains were excluded). Comparing the two phylogenetic trees (Supplementary Figures S5A,B), it is evident that the two transporters have different patterns: KgtP is clustered with many E. coli close relatives whereas C5038 is clustered with only two of them, with others being Pseudomonads and Burkholderiales. These data strongly indicate that C5038 and KgtP have distinct evolutionary trajectories, supporting our suggestion that c5038 was acquired through horizontal gene transfer.