To investigate how PmrAB and PhoPQ respond in E. coli, we searched whether it was possible to induce polymyxin B resistance of E. coli W3110 strain (WT) by testing different growth conditions. Previous studies have examined the susceptibility of E. coli to polymyxin B by measuring survival of cells after 1 h of exposure to 2.5 μg/ml of this compound. In contrast, we here searched for conditions under which E. coli cell grew normally in a milieu containing 2.5 μg/ml of polymyxin B. Different stimuli were tested in the challenging milieu (i.e., containing polymyxin B) as well as in the pre-culture (referred as conditioning medium): two pH values (neutral, 7.5 and mildly acidic, 5.8), two concentrations of Mg²⁺ (10 μM, low and 1 mM, high) and the presence or absence of 300 μM FeSO₄. Fe²⁺ is quickly oxidized to Fe³⁺ serving as a specific signal for activation of the PmrA–PmrB system (Wösten et al., 2000). In minimal medium (N-min) at pH 7.5, both the WT and the PmrA-constitutive WD101 strain were susceptible to polymyxin B, irrespective of the concentration of Mg²⁺ (Figure 2A). At pH 5.8, WD101 became polymyxin B resistant irrespective of the Mg²⁺ concentration, but the WT remained sensitive (Figure 2B). On one hand, WD101 still required an external stimulus (e.g., mildly acidic pH) to express polymyxin B resistance. On the other hand, low pH combined with low Mg²⁺ did not elicit polymyxin B resistance of WT.

The addition of 300 μM Fe³⁺ at pH 7.5 and low or high Mg²⁺ did not elicit polymyxin B resistance of WT (data not shown). In contrast, WT became resistant at pH 5.8, with high Mg²⁺ and high Fe³⁺ irrespective of the conditioning medium, except for a combination of pH 7.5 and high Mg²⁺ (Figure 3A). These results indicate that the presence of stimuli in the conditioning medium was also determinant and that low pH or low Mg²⁺ are both appropriate to trigger resistance of WT. We showed that an additional 4-h induction at pH 5.8, high Mg²⁺ and high Fe³⁺ prior to antibiotic exposure induced resistance after a pre-culture at pH 7.5 and high Mg²⁺ (Figure 3A). When plated at pH 5.8, low Mg²⁺ and high Fe³⁺, WT displayed polymyxin B resistance, irrespective of conditioning medium (Figure 3B).

In conclusion, E. coli must be conditioned at least in a milieu of low Mg²⁺ or low pH to grow in presence of polymyxin B in a milieu of low pH and high Fe³⁺. Furthermore, E. coli cells grew in the presence of polymyxin B at low pH, low Mg²⁺ and high Fe³⁺, regardless of conditioning.

The requirement for multiple stimuli to elicit polymyxin B resistance suggested that both PmrAB and PhoPQ were involved. That was expected for PmrAB, while PhoPQ was thought to be only required in the absence of a direct activation of PmrB. We then assessed the role of PhoP and PmrD connector in polymyxin B resistance. The phoP and pmrD mutants were grown in N-min at pH 7.5 and low Mg²⁺, which provided a good conditioning of WT before exposure to polymyxin B at pH 5.8, high Mg²⁺ and high Fe³⁺ (Figure 3A). In contrast to WT, phoP and pmrD mutants were polymyxin B susceptible in these conditions (Figure 4A). These data showed that PhoPQ TCS contributes to polymyxin B resistance under these conditions, at least in part via PmrD. A 4 h-induction at pH 5.8, high Fe³⁺, and low or high Mg²⁺ before challenging, restored only partially the resistance of both strains (a decrease of 3 log units in viable counts with respect to WT) (Figure 4A). This implies that the direct activation of PmrAB by low pH and high Fe³⁺ during this 4 h-period was not sufficient in the absence of PmrD or PhoP, even though a partial phoP- and pmrD-independent resistance occurred.

When plated at pH 5.8, low Mg²⁺ and high Fe³⁺, the phoP and pmrD mutants still remained polymyxin B susceptible (Figure 4B). The phoP mutant displayed a partial resistance after 4 h-induction in the presence of high Fe³⁺, while pmrD mutant became fully resistant (Figure 4B). PhoP is required to obtain the full resistance in a PmrD-independent way; hence, the control of gene(s) of the PhoPQ regulon distinct from the PmrAB regulon is likely required. In contrast, the PmrD connector is dispensable as long as the cells are conditioned in low pH and high Fe³⁺ before exposure, which activates both TCS directly.

We next examined the role of PhoPQ and PmrAB on deoxycholate susceptibility. We monitored the growth of E. coli strains in 2YT-broth containing 2.5 mg/ml of deoxycholate knowing that the MIC exceeds 100 mg/ml for WT. In contrast to WT and pmrD strains, the phoP and WD101 strains were susceptible to deoxycholate (Figure 5). In the same conditions, WT, phoP and pmrD strains were sensitive to polymyxin B, while WD101 was resistant (Figure 5). This finding demonstrates that PhoP is also essential in conferring innate resistance to deoxycholate. PhoPQ was reported to enhance lpxT expression by three fold under low Mg²⁺ conditions in E. coli (Hong et al., 2018). We then examined whether lpxT expression was also upregulated by PhoP when cells were grown on 2YT medium by monitoring the level of 3 × Flag-tagged LpxT. The amount of LpxT protein was similar in WT and pmrA strains, while it was reduced by two fold in the phoP mutant (Figure 6A) indicating that PhoP also significantly enhances lpxT expression in rich nutrient broth.

Since the WD101 strain displayed deoxycholate susceptibility, we then questioned whether the WT strain also exhibited deoxycholate susceptibility under PmrAB-inducing conditions. It was not possible to monitor the bacterial growth in the presence of deoxycholate in N-min at pH 5.8 (i.e., required for PmrAB induction) because deoxycholate precipitates at this pH. Therefore, we monitored the survival of the cells after exposure for 1 h to 10 mg/ml of deoxycholate, a dose relevant to the concentration of bile acids in the intestinal tract. The cells were grown in 2YT or in N-min at pH 5.8, low-Mg²⁺ and further conditioned in N-min pH 5.8, low-Mg²⁺ and high Fe³⁺ during 4 h. Then, the cells were challenged or not (control) with deoxycholate in PBS buffer and the surviving cells were numerated. The rate of survival to deoxycholate as compared to the control were 111 ± 25% when cells were grown in 2YT-broth and 39 ± 5% when cells undergo PmrAB-inducing conditions. These data demonstrated that LPS modifications or other event conferring polymyxin B resistance sensitized the cells to deoxycholate.

Contrary to CAMPs that are positively charged, bile acids are negatively charged. Thus, deoxycholate susceptibility could arise from a decrease of LPS negative charge under PmrAB-inducing conditions, either due to the addition of positively charged groups and/or the decrease of lipid A 1-PP species. We then inactivated pmrR in WD101 strain to assess whether the lack of LpxT inhibition restored deoxycholate resistance. The pmrR gene is divergently transcribed from the eptA-pmrAB operon and the 3′ end of pmrR overlaps with pmrB (Figures 7A,B). In order to maintain the expression of eptA-pmrAB operon, we only deleted the PmrA-box from the pmrR promoter (Figure 7B). The WD pmrR_prom mutant developed polymyxin B susceptibility in contrast to its parental strain, and the further deletion of lpxT restored polymyxin B resistance (Figure 8). These findings were consistent with the expected lipid A modifications following the relief of LpxT from PmrR inhibition in WD101 strain (i.e., phosphorylation instead of addition of pEtN) according to Herrera et al. (2010). These results thus confirmed that the polymyxin B susceptibility displayed by WD pmrR_prom is due to the lack of PmrR and not to a side effect on the expression of eptA-pmrAB. The functionality of eptA-pmrAB was also tested by monitoring the expression of arnT, which is under the control of PmrA. As expected, the arnT transcript was 50-fold higher in WD101 and WD pmrR_prom strains as compared to WT (Figure 7D). Moreover, an ectopic copy of pmrR restored polymyxin B resistance in WD pmrR_prom strain (data not shown). Contrary to its parental strain, the WD pmrR_prom strain displayed deoxycholate resistance (Figure 9A). Moreover, in trans expression of lpxT in WD101 also restored deoxycholate resistance, while conferring polymyxin B susceptibility (Figure 9B). We tested whether LpxT was required for deoxycholate resistance of WT; however, W3110 lpxT strain did not show any susceptibility up to 15 mg/ml of deoxycholate (Figure 9C). In conclusion, when we restored LpxT activity in WD101, we abolished polymyxin B resistance and, conversely, we restored deoxycholate resistance. Therefore, the deoxycholate susceptibility of WD101 is due to either a lack of LpxT-modification, the presence of EptA-modification, since both modifications occur at the same position of lipid A in a competitive manner (Herrera et al., 2010; Kato et al., 2012; Figure 1A), or both events.

To address this issue further, we inactivated eptA in WD101 strain. The expression of eptA-pmrAB operon relies on two promoters: a PmrA-dependent promoter located upstream of eptA and a constitutive promoter located at the 3′ end of eptA, which overlaps with the 5′ end of pmrA (Figures 7A,C; Soncini and Groisman, 1996; Lee et al., 2004). Therefore, to maintain pmrAB expression, we generated the WD eptA_part strain, which held 120-nt and 420-nt at the 5′ and 3′ ends of eptA open reading frame, respectively. The level of anrT transcript was similar in WD eptA_part and WD101 strains, i.e., 50-fold more than WT, attesting pmrAB functionality (Figure 7D). As was expected for a lack of pEtN decoration, which was previously demonstrated with a similar mutant (Herrera et al., 2010), WD eptA_part displayed polymyxin B susceptibility (Figure 10A). We tested functional complementation with plasmids carrying an ectopic copy of eptA under the control of its own promoter (pET21d:eptA), a trc promoter (pTrcH60:eptA) or a T7 promoter (pET2160:eptA). The plasmids pET21d:eptA and pTrcH60:eptA restored polymyxin B resistance in WD eptA_part strain (Figure 10B). Collectively, these data supported that WD eptA_part strain remained PmrA-constitutive, while EptA activity was lost.

Interestingly, WD eptA_part strain displayed similar deoxycholate susceptibility as WD101 (Figure 10A), supporting that the addition of pEtN per se does not confer deoxycholate susceptibility and, thus the lack of lipid A 1-PP, due to the inhibition of LpxT by PmrR, is directly responsible for bile acid susceptibility. We generated the WD eptA_part pmrR_prom strain, which exhibited deoxycholate resistance (Figure 10A). The resistance was also restored in WD eptA_part upon in trans expression of lpxT (Figure 10A). The plasmid pTrcH60:eptA restored deoxycholate susceptibility in WD eptA_part pmrR_prom likely due to the fact that overexpressed EptA outcompetes LpxT to modify lipid A at position 1. Meanwhile, it did not restore polymyxin B resistance (Figure 10A), suggesting that not enough pEtN decorations have been completed with respect to the CAMP. The WD eptA_part lpxT strain displayed sensitivity to both polymyxin B and deoxycholate as WD eptA_part (Figure 10A), demonstrating that according to polymyxin B, it is the lack of pEtN that conferred susceptibility and not the LpxT-dependent modification that may occur in the absence of EptA.

We further addressed the effect of other LPS modifications according to polymyxin B and deoxycholate by individually deleting arnT, eptB, eptC, pmrG, and pagP in WD101 strain. In accordance with Herrera et al. (2010), WD arnT mutant displayed increased polymyxin B susceptibility similar to WD eptA_part and WD pmrR_prom strains and all other mutants displayed polymyxin B resistance (Figure 11). These data indicated that the charges displayed at the level of lipid A play a determinant role in CAMPs resistance, while the charges displayed at the core level have no effect at least taken individually. All these mutants displayed deoxycholate susceptibility as WD101 and in contrast to WD pmrR_prom (Figure 11). Notably, WD eptA_part arnT also displayed deoxycholate susceptibility, supporting that only the lack of diphosphate group at position 1, due to the inhibition of LpxT by PmrR, is responsible for deoxycholate susceptibility. Unexpectedly, the inactivation of lpxT in WD arnT mutant restored polymyxin B resistance (Figure 11). This observation indicated that LpxT, whose enzymatic activity is inhibited by PmrR in WD101 background (Touzé et al., 2008), somehow hindered polymyxin B resistance in WD arnT strain.

In Salmonella, the deoxycholate susceptibility of the PmrA-constitutive strain was found to rely on the expression of wzz_ST, whose product controls the modality of O-antigen synthesis favoring the abundance of long chains (16 to 35 subunits) (May and Groisman, 2013; Islam and Lam, 2014). In contrast, very long O-antigen (>100 subunits) or the addition of pEtN and L-Ara4N had no effect on deoxycholate susceptibility (van Velkinburgh and Gunn, 1999; May and Groisman, 2013). The LPS of E. coli W3110 lacks the O-antigen chain due to a disruption of wbbL gene. We then examined whether restoring the O-antigen biosynthesis in E. coli and varying the modality of its synthesis could modulate deoxycholate susceptibility in a way comparable to Salmonella. SDS-PAGE analysis of LPS extracted from WT and WD101 carrying an ectopic copy of wbbL showed a restored synthesis of the O-antigen moiety (Figure 12A). As judged from the SDS-PAGE, the latter polymer mostly exhibited about 25 subunits and could therefore be qualified as long. The expression of wbbL and thus the abundance of these long O-antigens did not modify the phenotypes of WD101 and WT strains that remained susceptible and resistant to deoxycholate, respectively (Figure 12B). We assessed whether wzz_EC expression was upregulated in WD101 as compared to WT by monitoring wzz_EC transcript; however, wzz_EC was expressed at similar levels in both strains (Figure 12C). Thus, contrary to Salmonella, wzz_EC is not under the control of PmrA. This observation was correlated with the existence of similar O-antigens pattern (i.e., long O-antigens) in both WD101 and WT. Moreover, the inactivation of wzz_EC in WD101 led to the production of a majority of shorter O-antigens, further demonstrating the functionality of Wzz_EC (Figure 12A). The WD wzz_EC strain, which exhibited drastically less long O-antigen, remained deoxycholate susceptible (Figure 12B). In contrast to Salmonella, wzz_fepE was not upregulated upon PmrA activation as no wzz_fepE transcript was observed in WD101 or WT, which correlates with similar O-antigen patterns in both strains (Figure 12A). To investigate whether the presence of O-antigen changes the requirement for lipid A-1PP to resist deoxycholate in PmrA-constitutive background, the phenotype of a series of WD101 mutants carrying wbbL was examined. All mutants remained deoxycholate susceptible, except WD pmrR_prom (Figure 12D), demonstrating that in E. coli, the presence and the modality of the O-antigen synthesis have no major effect with respect to deoxycholate contrary to Salmonella.

In P. aeruginosa, the lpxT gene transcript was identified in a screen for mRNA whose translation depends on temperature. The translation is blocked by a secondary structure that disassembles when temperature rises from 25 to 37°C (Delvillani et al., 2014). Although the 5’ UTR regions of lpxT from P. aeruginosa and E. coli do not present any similarity, we tested whether lpxT is also under the control of temperature in E. coli. We monitored the level of 3 × Flag-tagged LpxT in WT background after 1 h growth in 2YT-medium at 25, 30, 37, and 42°C. LpxT protein was present at a low level at 25°C and significantly increased with temperature rising: 2.9 and 3.6-fold increases were observed at 37 and 42°C, respectively, compared to 25°C (Figure 6B).

Such a temperature-dependent profile suggested a role of LpxT within the host. We then investigated a potential role in colonization of the mammalian gut. The lpxT deletion was introduced in the mouse-adapted E. coli WT strain MG1655 and the kinetics of colonization of WT and lpxT strains was followed in the feces from day 1 to day 7 post-oral challenge. WT was recovered in a significantly higher number as compared to lpxT strain (p < 0.05) from feces at day 1 and this gap was no longer visible at day 2 (Figure 13A). Of note, the maximal bacterial load in the feces was observed at day 1 for WT and this number decreased throughout the experiment (from 2 × 10⁹ cfu/g down to 1 × 10⁸ cfu/g). We performed competition experiments by inoculating WT and lpxT strains in an equal mixture. The WT was recovered from feces in much higher numbers at day 1 (about 25-fold in excess), also at days 2 and 4 but to a lesser extent (Figure 13B). We then monitored bacterial loads at different parts of the gut in the competitive assay after 1 day post-inoculation. Panel C shows that it was especially in the small intestine (duodenum, jejunum and ileum) that we observed at day 1 the biggest and equivalent disadvantage of the lpxT mutant (about 1.5–2 log; 50 to 100 times less), while only a 1 log of difference (10 times less) (p < 0.0001) was observed in the colon (Figure 13C). All this suggested that LpxT rather brought an advantage in the initial stages of colonization where E. coli must pass through areas with high bile acid concentrations.

Bacterial strains and plasmids are listed in Table 1. Primers are listed in the Supplementary Material. Bacteria were grown at 37°C in 2YT broth or N-min minimal medium (pH 7.5 or 5.8) containing 0.1% casamino acids and 38 mM glycerol (Chamnongpol et al., 2002), and the indicated concentrations of MgCl₂ and FeSO₄. When required, the medium was supplemented with ampicillin, kanamycin or chloramphenicol at 100, 50, and 25 μg/ml, respectively. The different enzymes used for molecular biology techniques were from New England Biolabs, and DNA purification kits were from Macherey-Nagel. Primer synthesis and DNA sequencing were performed by Eurofins-MWG. All other materials were reagent grade and obtained from commercial sources.

WD eptA_part, WD pmrR_prom and BW lpxT-Flag were generated by using the Datsenko and Wanner method (Datsenko and Wanner, 2000). For WD eptA_part and WD pmrR_prom, the primers were designed to amplify the Cm^R resistance cassette from the template plasmid pKD3, flanked by 50 bp from the targeted chromosomal region for homologous recombination (Supplementary Table 1 and Figures 7B,C). To generate the BW lpxT-Flag strain, the primers (Supplementary Table 1) were designed to amplify the Kan^R resistance cassette and the 3 × Flag tag-coding sequence from the template plasmid pSUB11 (Uzzau et al., 2001), flanked by 50 bp from the targeted chromosomal region. The PCR products were transformed in the WD101 or BW25113 recipient strains harboring the pKD46 plasmid for λ red recombinase expression. After selection of the mutated strains under the appropriate selective pressure, the thermosensitive pKD46 plasmid was cured. For WD eptA_part and WD pmrR_prom, the antibiotic resistance cassette was finally removed from the chromosome by using the pCP20 plasmid for expression of the FLP recombinase, yielding an 85-nt FRT scar in place of the deleted region (Datsenko and Wanner, 2000).

WT lpxT, WT pmrD, WT phoP, WD arnT, WD eptB, WD eptC, WD pmrG, WD pagP, WD wzz_EC, WD wzz_fepE, WD ΔpmrR_prom lpxT, WD eptA_part lpxT, WD arnT lpxT, WT lpxT-Flag, WT phoP lpxT-flag, and WT pmrA lpxT-Flag were generated by P1-mediated transduction by using the following strains as donors: DMEG3 (El Ghachi et al., 2005), BW25113-based Keio collection (Baba et al., 2006), or BW lpxT-Flag. The antibiotic resistance cassette was always removed by FLP recombination except for the lpxT-Flag strains. The MG lpxT strain was generated by P1 transduction using DMEG3 strain as donor and MG1655 strain as recipient strain. All the mutant strains were systematically controlled by PCR using appropriate primers (Supplementary Table 2).

The plasmids used for the expression of an ectopic copy of lpxT or eptA were generated by PCR amplification of the corresponding ORFs with primers listed in Supplementary Table 3, followed by their insertion in the appropriate expression vectors (Table 1). The primers and vectors which were used are specified in Table 1. For the construction of the plasmid for wbbL expression, the gene was amplified by overlap extension PCR in two steps. The two parts of the gene, flanking the inserted IS5 element, were PCR amplified from BW25113 chromosomal DNA with primers wbbL-NcoI and wbbL-mid1 (5′-end of the gene) and wbbL-HindIII and wbbL-mid2 (3′-end of the gene). The resulting PCR products were then used as templates in a second PCR for the amplification of the reconstituted gene, which was then inserted in the pET21d vector. All plasmids were checked by sequencing.

The bacteria were grown overnight at 37°C in 2YT medium or N-min medium at pH 7.5 or 5.8, in the presence of MgCl₂ (10 μM or 1 mM). When additional conditioning was performed as specified, the overnight culture was washed twice in fresh N-min medium pH 5.8 or 7.5 and the cells were diluted 1:50 in N-min at pH 7.5 or 5.8, 10 μM or 1 mM MgCl₂ and 300 μM FeSO₄ and subsequently incubated at 37°C under agitation for 4 h. The culture (overnight culture or conditioned cells) was used to prepare cellular suspensions at 10³ to 10⁸ CFU/ml in sterile water. The number of CFU of the culture was determined according to the OD_600nm (1 unit corresponding to 3 × 10⁸ CFU/ml). 5-μl aliquots of these serial dilutions were deposited on 2YT or N-min agar plates containing or not polymyxin B and deoxycholate at the indicated concentrations, which were incubated at 37°C for 24 to 48 h.

Alternatively, the bacteria were grown overnight at 37°C in 2YT medium and a suspension at 10⁸ CFU/ml was prepared in 5 ml of sterile water. A 2YT agar plate was flooded with this suspension for 1 min to allow the bacteria to sediment before removing the excess of water. 5-μl drops of polymyxin B solutions at various concentrations were added at the surface of the plates, which were then incubated at 37°C for 16 h.

The bacteria were grown at 37°C overnight in 2YT medium or N-min medium at pH 5.8, 10 μM MgCl₂. The respective cultures were diluted 1:50 in 2YT medium or in N-min at pH 5.8, 10 μM MgCl₂ and 300 μM FeSO₄ and subsequently incubated at 37°C with agitation for 4 h. Two suspensions at 10⁵ CFU/ml were prepared in PBS buffer to be challenged or not with deoxycholate at 10 mg/ml final concentration at 37°C with agitation for 1 h. The suspensions were diluted in PBS buffer and spread on 2YT agar plates for the enumeration of survivals, which was done after overnight incubation of the plates at 37°C.

The LPS were prepared according to the protocol already described (Crawford et al., 2012). They were analyzed by 15% SDS-polyacrylamide gel electrophoresis and visualized by silver staining as described previously (Tsai and Frasch, 1982).

Total RNA were extracted from bacteria grown to the middle of exponential phase (OD_600nm = 0.5) using RNeasy Protect bacteria Mini Kit system (Qiagen) according to the manufacturer’s instructions. cDNA synthesis was performed from 1 μg of total RNA with random hexanucleotides as primers using the Superscript IV First Strand Synthesis system for RT-PCR (Invitrogen). The quantitative PCR reactions were then carried out with the appropriate primers (Supplementary Table 4) using DyNAmo ColorFlash SYBR Green qPCR kit (Thermo Scientific) and they were run in a StepOnePlus Real-Time PCR system (Applied Biosystems). The data were analyzed with StepOne software v2.3 using ΔΔCt method and normalized using the housekeeping genes rrsA, gyrA and ffh as reference genes.

Bacterial strains expressing the 3 × Flag-tagged LpxT (WT lpxT-Flag, WT pmrA lpxT-Flag, and WT phoP lpxT-Flag) were grown overnight in 2YT at 37°C, diluted 100 fold in fresh 2YT medium and then incubated at 37°C until the OD_600nm reached 0.7. To test the expression of Flag-tagged LpxT at different temperatures, the WT lpxT-Flag strain was grown overnight in 2YT at 25°C and diluted in four flasks containing fresh 2YT medium at an initial OD_600nm = 0.2, which were then incubated at 25, 30, 37, or 42°C for 1 h under agitation. 1 ml of each culture was then used for total protein extraction. The cells were centrifuged and the pellet was resuspended in SDS buffer (50 mM Tris–HCl pH 6.8, 100 mM β-mercaptoethanol, 2% SDS, 0.1% bromophenol blue and 10% glycerol) and boiled for 15 min. Similar amounts of proteins, according to the OD_600nm of the culture, were analyzed by 15% SDS polyacrylamide gel electrophoresis and transferred to a polyvinylidene difluoride membrane. The membrane was incubated first with monoclonal antibody Anti-Flag M2 (Sigma) and then with antibody Anti-IgG from mouse (Fc Specific) coupled to a peroxidase (Sigma). The blots were developed with ECL system (Bio-Rad) according to the manufacturer’s protocol. The images were collected with ImageQuant^TM LAS 500 (GE Healthcare-Life Sciences) and the proteins were quantified by Image Studio Lite by measuring the density of the corresponding bands.

OF1 female mice purchased from Charles River Laboratories and aged 5 weeks were given bottles of water with 5 g/L of streptomycin (treatment started 2 days prior to gavage) and infected by gavage with feeding needles with MG1655 strains (2 × 10⁸ bacteria per mouse) resistant to streptomycin. For the competitive assay, mice were infected by MG1655 and MG lpxT strains in equal proportions (2 × 10⁸ total bacteria per mouse). Colonization rates were determined by enumeration of CFU per gram of feces. The samples were diluted and spread onto 2YT agar plates with streptomycin (5 μg/ml) supplemented with 20 μg/ml of chloramphenicol to enumerate MG lpxT cells and/or without chloramphenicol to enumerate total E. coli cells (i.e., WT and lpxT mutant). Colonization rates were also determined by enumeration of CFU per gram of different parts of the gastrointestinal tract. Mice were euthanized with CO₂, parts of the gut were ground and homogenized in peptone broth and MG1655 and MG lpxT cells were enumerated. The results of two independent colonization experiments (seven mice by cage) were pooled and a one tailed Mann-Whitney test was used to determine statistical significance of observed differences (GraphPad Prism v5.0 GraphPad Software, CA).