The twenty-one E. coli strains used in this study are described in Table 1; nine were pathogenic, three were commensal and nine ABU. The archetypal UPEC strain CFT073 was originally isolated from the blood and urine of a woman admitted to the University of Maryland Medical System 22 . Seven other UPEC prototypes, E. coli UTI 89, J96, UMN026, IAI39, IAI74, IH11128 and 536, were also studied 23 24 25 . E. coli ABU 83972, the archetypal ABU strain, is a clinical isolate capable of long-term bladder colonization 26 . Eight other ABU strains were also included 27 . E. coli K-12 MG 1655, a commensal, laboratory derived strain, was originally isolated from the faeces of a convalescent diphtheria patient in Palo Alto in 1922 28 . E. coli IAI1 is a commensal strain isolated from faeces of a young, healthy, military conscript in the 1980s in France 6 . E. coli ED1a, isolated in the 2000s from faeces of a healthy man in France, belongs to a human-specific, widespread, commensal clone that is increasing in frequency 29 . Finally, an enterohaemorrhagic E. coli (EHEC) O157:H7 Sakai strain isolated during an outbreak in Sakai City, Japan, in 1996 was studied 30 .

* TBARS values are expressed in micromoles per 10¹¹ cells. The data shown are means (mean ± SD) of three independent experiments in different batches of urine and LB broth. Differences between means were evaluated for statistical significance using the Tukey's HSD (Honestly Significant Difference) test.

**p values of < 0.05 were considered significant. compared to ABU83972 under the same conditions.

The study was performed in Luria Bertani (LB) broth, a nutritionally rich medium and in human urine. Three independent batches of urine from healthy volunteers were used. Midstream urine was collected over 24 hours from healthy male volunteers, with no history of UTI or antibiotic use in the last two months. The urine was pooled, centrifuged, filtered (filter size 0.22μm) and stored at −20°C. The growth experiments were assayed using 96 well plates and OD₆₀₀ was measured with a Tecan Infinite M200 plate reader. For the preparation of cell extracts, aliquots of 40 ml of pooled urine or LB broth in 50 ml vials were inoculated with the appropriate volume of the overnight culture to obtain 2 x 10⁵ bacteria/ml. Each experiment was performed in triplicate and repeated in 3 different batches of urine or LB broth. Cells were grown at 37°C under microaerobic conditions (1% O₂). Dissolved oxygen saturation was measured by luminescence with a measure probe (Hach Lange GmbH) in the different media during the exponential growth phase. The measure was repeated at least four times. Cultures were sampled in mid-exponential growth phase and 30 min after the beginning of stationary phase. Aliquots of 40 ml of culture were centrifuged at 4500 rpm at +4°C for 15 min. The bacteria were washed twice with 0.9% NaCl, pelleted and stored at −20°C until used. The cells resuspended in appropriate sonicating buffers (see below) were disrupted by sonication on ice for 3 min (30 s disrupt with 30 s rest) with an ultrasonic disrupter (Sonics & Materials Inc.).

The pellets were sonicated in phosphate buffer, pH 7.8. All assays, except catalase activity, were performed on a Roche Diagnostics/Hitachi 912. Catalase activity was determined using the Catalase Assay kit (Sigma). The Cu-SOD activity, which corresponds to the periplasmic SOD, was assayed using the SOD assay kit (Randox laboratories) based on the method of Mc Cord and Fridovich 31 . The cytosolic SOD activity, which is effected by the Mn- and the Fe-SODs, was calculated as the difference between the total SOD activity measured at pH 7.8 and the Cu-SOD activity measured at pH 10.2. The glutathione oxidoreductase was assayed by the method of Bleuter 32 . Oxidized glutathione (GSSG) was added and the disappearance of NADPH was monitored at a wavelength of 340 nm. The assay of glucose-6-phosphate-dehydrogenase (G6PDH) was based on Bleuter’s method 33 , where glucose-6-phosphate was added and the reduction of NADP to NADPH was monitored at a wavelength of 340 nm. The γ-glutamylcysteine synthetase (GshA) and the glutathione synthetase (GshB) were assayed as described previously 34 . Briefly, ADP generated by both enzymes in the presence of their substrates was determined using a coupled assay with pyruvate kinase, and lactate dehydrogenase. Oxidized and reduced glutathione concentrations were assayed by high-performance liquid chromatography (HPLC) equipped with a colorimetric detection system, using N-acetyl cysteine as an internal control 35 . Each experiment was performed in triplicate and repeated in 3 different batches of urine. The activities of the enzymes and the glutathione content in each sample were normalized with total proteins assayed by the method of Bradford 36 .

Lipid peroxidation was estimated as TBARS content. TBARS were measured by the standard procedure using the protocol according to the manufacturer’s instructions (OxiSelectTM TBARS Assay kit, Cells Biolabs). When cells were either in exponential growth phase or in stationary phase, OD₆₀₀ of the cultures and TBARS concentrations were determined. The pellets were sonicated in PBS buffer containing 1% Triton X-100 and 0.05% antioxidant butylated hydroxytoluene to prevent further oxidation of lipid. Each experiment was performed in duplicate and repeated in 3 different batches of human urine and LB broth.

Differences between means of at least 3 to 9 experiments were evaluated for statistical significance using the Tukey's HSD (Honestly Significant Difference) test. Non-parametric data were analysed using a Mann–Whitney U-test. P values of < 0.05 were considered significant. Data are presented as mean ± standard deviation or as box-plots based on medians and quartiles.

The growth capacity of twenty-one E. coli strains (8 UPEC, 1 EHEC, 9 ABU, 3 commensal strains) was studied (Figure 2). As expected, growth in pooled human urine was significantly less than in LB medium, for all strains and supplementation of urine with casaminoacids improved growth (data not shown). Unlike LB broth, urine limits cell growth. Moreover, in LB broth as in urine, it was found that all strains produced similar growth curves. Only both strains ABU 83972 and IAI1 grew slightly faster than four ABU strains (57, 64, 27 and 5) during the exponential phase in urine (p < 0.0001). Surprisingly, the growth capacity of ABU in the urine is not better than that of UPEC and commensal strains.

The content of TBARS, corresponding to the accumulation of membrane lipid peroxidation products was measured during exponential growth in both culture media, pooled human urine and LB broth (Table 1). The levels of damage products accumulated have been used to assess oxidative stress induced by intracellular ROS 16 37 . In all cases, p values were versus ABU 83972 strain. No significant difference was observed in TBARS content of twenty-one strains grown in LB broth while differences occurred during growth in urine. Similar amounts of TBARS were produced by ABU 83872 and fourteen other strains. These amounts were significantly higher than those produced by five other E. coli strains (Sakai, UTI89, MG1655 and ABU 38 and 62). IAI1 with a p value at 0.075 was at an intermediate position. These data show that during exponential growth in urine, the intracellular ROS level differs between strains. Furthermore, the ROS level is not linked to the phylogenetic groups. Strains belonging to different phylogenetic groups were found in both groups with high and low TBARS content, respectively. In the same way, no clear delineation between virulence categories was apparent (Table 1). Interestingly, it also appeared that strains which grew slightly faster or slower in urine e.g. ABU strains 83972 and 64 respectively, exhibited similar levels of TBARS.

Changes in ROS levels produced in the exponential and stationary growth in both pooled human urine and LB broth were studied using a representative panel of strains [three UPEC strains (CFT073, UTI89, 536), all belonging to the phylogenetic B2 group, three commensal strains (ED1a, IAI1, MG1655) belonging to various phylogenetic groups, the ABU 83972 from phylogenetic group B2 and Sakai from phylogenetic group E] (Table 2). Due to the sampling procedure, data obtained were subject to a new analysis of variance. The statistical analysis performed on a limited number of strains showed results quite similar to the first analysis. Similar amounts of TBARS were produced by ABU 83972 and CFT073 during exponential growth in urine. These amounts were significantly higher than those produced by the four strains IAI1, Sakai, UTI89 and MG1655. ED1a and 536 with a p value at 0.070 and 0.048 respectively were now at an intermediate position. No significant changes were observed in the stationary phase of growth. As a consequence, similar amounts of TBARS were produced during the two phases of growth except for ABU 83972 in urine. In strain ABU 83972, the level of TBARS was higher in the exponential phase and decreased significantly in the stationary phase showing the ability of strain ABU 83972 to control the endogenous oxidative stress during growth in urine. In contrast, all isolates grown in LB medium exhibited similar levels of ROS regardless of the growth phase.

* TBARS values are expressed in micromoles per 10¹¹ cells. The data shown are means (mean ± SD) of three independent experiments in different batches of urine and LB broth. Differences between means were evaluated for statistical significance using the Tukey's HSD (Honestly Significant Difference) test.

** p values of < 0.05 were considered significant. compared to ABU83972 under the same conditions.

The behavior of the commensal strains and UPEC in urine was also compared. As expected, no significant difference was observed in the amount of TBARS produced (data not shown). E. coli is a diverse species, both in terms of gene content and sequence divergence 24 38 , so we then analysed strains from the phylogenetic B2 group only, which includes both commensal and pathogenic strains. No difference was observed between the UPEC and the ED1a intestinal commensal strains (p = 0.968). However, clear differences were demonstrated in urine between the three UPEC strains selected (5.19 ± 1.31) and the ABU strain 83972 (7.26 ± 1.03) with a p value = 0.009.

The non-enzymatic and enzymatic components involved in antioxidant defense systems (Figure 1b) were studied during growth in pooled human urine in a subset of four B2 UPEC and ABU strains selected from the previous panel (CFT073, UTI89, 536 and ABU 83972) (Additional file 1: Table S1 and Additional file 2: Table S2). To increase the statistical power of our analysis, antioxidant defense mechanisms of the three UPEC were compared with those of ABU 83972. The results are presented Figure 3. We also compared antioxidant defense systems between ABU 83972 and CFT073 alone. Similar results to those obtained for ABU 83972 and the three UPEC were obtained, however, the p values were less significant (between 0.03 and 0.15) (data not shown).