Uro-associated E. coli isolates (Total: 48) were curated between 2010 and 2015 from patients presenting with either cystitis, pyelonephritis, asymptomatic bacteriuria (ABU) or UTI-associated bacteraemia. Two reference E. coli strains, CFT073 (Welch et al., 2002) and NCTC10418 (Ali et al., 2017) were used as controls. The motility phenotypes of the isolates were assessed using semi-quantitative agar assays (Figure 1A) and 32/48 (67%) of the clinical isolates were identified as motile (Supplementary Table S1). In general, the swarms of isolates recovered from ABU patients measured between 0.8 and 5.4 cm, while cystitis strains exhibited an uneven motility distribution pattern with strains swarming less (n = 6) or greater (n = 4) than 5.4 cm, respectively (Figure 1B; ANOVA p < 0.005).

The impact of bacterial motility on the urothelial innate response was assessed in vitro using 24-h challenges of heat-killed isolates (1 × 10⁵ CFU/ml) and bladder RT4 cells stably transfected with a NF-κB luciferase reporter (Ali et al., 2017). Following these challenges 33/42 (78%) of the clinical isolates and the control strain CFT073 were associated with a 2–5 fold increase in NF-κB signaling activity compared to a PBS control (Figure 2A; p = 0.029). Seven motile isolates (Supplementary Table S1: 3406, 3,408, 3,411, 3,412, 3,414, 3,419, 3,424) equating to 30% of all motile strains and predominantly associated with cystitis, and the control strain NCTC10418 exhibited increases of >5-fold. Two of these isolates, 3,411 and 3,414, generated NF-κB signaling activities of >30 fold (Figure 2A) and swarm sizes of >7 cm (Supplementary Table S1).

UTIs are ascending infections and urothelial cells respond to potential uropathogens via flagellin detection, TLR5 signaling and the release of antimicrobial killing and proinflammatory agents (Ali et al., 2017). TLR5 proteins recognize a conserved motif found in the majority of flagellins (Andersen-Nissen et al., 2005) while antigenic variation amongst flagellins has defined 53 flagellin or H-serotypes in E. coli (Wang et al., 2003). Characterization of the clinical isolates used in these analyses identified seven H-serotypes, H1, H4, H5, H6, H18, H27 and H21 (Supplementary Table S1). To explore whether the signaling response reflected antigenic variation, flagellin was purified, in the form of flagellar filaments, from the control strains and 13 of the E. coli isolates associated with either ABU, cystitis or pyelonephritis infections (Supplementary Table S1), and the urothelial cell challenges repeated (Figures 2B,C). Compared to the PBS control, purified flagellin filaments from the 15 different preparations generated robust and comparable NF-κB signaling responses when used at concentrations of either 250 or 50 ng/ml (Figure 2B; ANOVA p = 0.13; Figure 2C; ANOVA p = 0.12). These measured responses did not reflect either the source of the isolate or H-serotype and argued against host variability, between isolates, being related to flagellin itself.

Using chemostat growth conditions the growth rate of E. coli strain RP437 was shown to correlate with flagellar abundance, suggesting that a change in growth fitness could impact host recognition (Sim et al., 2017). To explore whether the differences in host response linked to fitness, the maximum doubling times of the clinical isolates and the two control strains were explored. The maximum doubling time of all strains varied between 13.9 and 26.9 min, respectively (Supplementary Tables S1, S2). Density plot analysis, stratified for motility, identified a shift between these two sets of strains (Figure 3). The average maximum doubling time for the motile strains was 17.4 ± 3.0 min compared to 19.7 ± 2.5 min for the non-motile strains (Figure 3; ANOVA p = 0.011). However, the range of maximum doubling times within each grouping was similar (Motile: 13.9–26.9 min.; Non-Motile: 14.7–25.3 min.). These data argued against fitness being a primary driving factor in the variation in host recognition observed within the motile subset of the strain collection.

While in vitro and clinical studies suggest TLR5 to be the key host receptor in detecting and protecting the bladder from uropathogenic E. coli, murine studies and genotype analyses also support a role for TLR4 (Samuelsson et al., 2004; Andersen-Nissen et al., 2007; Ragnarsdottir et al., 2007; Song et al., 2009; Smith et al., 2011; Ali et al., 2017). TLR4 recognizes lipopolysaccharide, the major outer membrane (OM) component of Gram-negative bacterial species such as E. coli (Tapping et al., 2000). Potential roles for TLR4 and LPS in the variable recognition of E. coli were further explored using outer membrane preparations from three motile isolates [3,398 (Pyl.-H1), 3,408 (cystitis-H1) and 3,412 (cystitis-H18)] (Supplementary Table S1), one non-motile isolate [3,416 (ABU (Mot⁻)] (Supplementary Table S2) and the two control strains NCTC10418 and CFT073 (Figure 4). These were used to challenge RT4 (TLR⁺) cells and RT4 cells where either TLR4 or TLR5 expression had been inhibited by siRNA knockdown (Supplementary Figure S1; Mowbray et al., 2018). The innate response was determined through quantification of the pro-inflammatory cytokine IL-8 (Figure 4A). Data indicated that the IL-8 responses to flagellin (50 ng/ml), whole bacterial and outer membrane (OM) sample challenges were significantly reduced following challenge of the TLR5^siRNA cells compared to wild-type (TLR⁺) cells (ANOVA p < 0.001). Analyses of the IL-8 responses following challenge of the TLR4 ^siRNA cells were more strain variable and supported minor roles for LPS, and TLR4 in the RT4 bladder cell innate response to an acute infection.

Effector IL-8 responses (< 2-fold) were observed in the TLR5^siRNA cells challenged with OM preparations. These were linked, potentially, to contamination of the preparations with flagellin, even though surface appendages were sheared by blending before OM isolation (see methods). To determine potential flagellar contamination residual levels of the flagellar hook protein FlgE (Paradis et al., 2017), were targeted. This approach addressed any serotype restrictions associated with using antibodies raised against specific flagellins. FlgE from Salmonella enterica serovar Typhimurium and E. coli exhibit 88% identity and 93% similarity (McClelland et al., 2001; Welch et al., 2002), which allowed an antibody raised against S. Typhimurium FlgE, α-FlgE_ST (Minamino and Namba, 2008), with cross reaction between the species (Figure 4B) to be used to identify potential contamination of the OM preparations. The α-FlgE_ST antibody was only able to detect FlgE in either whole cell lysates or flagellar preparations (Figures 4B,C) of motile isolates with no signal detected in the ABU (Mot⁻) samples. These data therefore argued that flagellin is the key factor in activating the bladder innate response to potential E. coli infections with LPS invoking a minor response.

The variability in host response, observed in Figure 2A, could not be explained by LPS alone so the roles of flagellar abundance were investigated. This was facilitated by exploiting a flagellar hook gene variant flgEA240C that allowed malemide-cysteine crosslinking of a fluorophore to assembled flagellar hooks. Visualization of the hooks by fluorescence microscopy (Sim et al., 2017), then allowed quantification of the hook numbers of cystitis (Schreiber et al., 2017), ABU (Lane et al., 2005) and PYL (Foxman, 2002) isolates (Figure 5). Individual flagella were visualized as distinct foci (Figure 5A) with the numbers of foci counted determining the number of flagella expressed by individual bacterial cells. Analysis of all clinical isolates was prevented by their antibiotic profiles restricting the use of pBADflgEA240C, which uses kanamycin as the selective agent. However, data relating to 6 cystitis and 11 ABU strains (Supplementary Table S1), were collated, which equated to 50% of the motile strains investigated in this study.

The control E. coli strain NCTC10418 averaged 2.4 foci or flagella per cell, while the clinical isolates and CFT073 averaged between 1 and 2 flagella per cell (ANOVA p = 0.148; Figure 5B) x-axis. Data also suggested that only 10 to 60% of the CFT073 and clinical isolate populations exhibited FlgEA240C foci compared to 80% of the NCTC10418 cell population (Figure 5B) y-axis. Comparison of these data to IκBα protein levels during flagellin challenge (Supplementary Figure S2) and urothelial IL-8 responses (Figure 5C; ANOVA p = 0.013) strongly suggested that uro-associated E. coli strains exploit population heterogeneity to preserve motility and avoid detection by TLR5. It is therefore predicted that variability in the host response, i.e., whether symptoms will or will not present, is linked to population heterogeneity and flagellar abundance of the invading uropathogen.

Expression of the E. coli flagellar system is regulated by a transcriptional hierarchy controlled by the master transcriptional regulatory FlhD₄C₂ (Wang et al., 2006). FlhD₄C₂ levels are sensitive to a wide range of regulatory mechanisms that include transcription, translation and protein stability (Claret and Hughes, 2000; Soutourina and Bertin, 2003; Kitagawa et al., 2011; Takaya et al., 2012). To further examine the link between a population of uro-associated E. coli and flagellar gene expression, a high copy number plasmid encoding flhDC was transformed into the control strain NCTC10418 and isolate 3,408 (Cystitis; 25% Fla⁺; 1.66 Fla⁺/cell; 0.19 relative IL-8 production: Supplementary Table S1) and RT4 urothelial cell challenge experiments performed. IL-8 data, reflecting the host response, showed that increasing flhDC expression in 3408 resulted in a two-fold increase in IL-8 concentrations but no change for NCTC10418 (Figure 5D).

E. coli strains used in the study are described in Supplementary Tables S1, S2; Strains 3,398 to 3,710 were a kind donation from the Diagnostic Microbiology Unit at the Freeman hospital, Newcastle NHS Trust, Newcastle upon Tyne between 2010 and 2012. No ethics were necessary for the use of these strains as the researchers did not have access to clinical records and the only information provided by the unit was the type of UTI associated with each isolate. Strains 4,738–4,745 came from the clinical study of Drage et al. (2019) that was conducted under an ethically approved study protocol (ref: REC14-NE-0026).

Strains used during this study were propagated in or on Luria-Bertani (LB) medium using 1.5% agar for plates. Incubation, unless stated otherwise, was overnight at 37°C with liquid cultures aerated by orbital shaking at 160 rpm. All motility assays were performed by either direct inoculation using a toothpick or inoculating 3 μl of an overnight culture onto motility agar (1% Tryptone, 0.5% NaCl, 0.3% Agar) and incubating for 8 h at 30°C. Images of motility swarms were digitally captured, and the vertical and horizontal diameter measured to generate an average swarm distance using ImageJ. All swarm assays were performed with a minimum of three independent colonies. Transformation of the plasmids pflhDC or pBADflgEA240C were performed by electroporation as described previously (Schaeffer et al., 1987). Selection for plasmids was performed using either 100 μg/ml Ampicillin or 50 μg/ml Kanamycin. pflhDC was generated by cloning a PCR product using the primers flhD-42FBam [ggcggatccGGGTGCGGCTACGTCGCAC] and flhC + 616RBam [ggcggatccCGCTGCTGGAGTGTTTGTCC] into the high copy number vector pSE280 using standard cloning techniques.

The maximum doubling time was determined from growth curves of all strains grown in LB media at 37°C. Strains were grown overnight and diluted to a starting OD₆₀₀ of 0.02. Two hundred microliters were aliquoted into a 96-well plate leaving the outer wells empty. The plate was sealed with a BreatheEasy membrane and OD₆₀₀ measurements taken every 400 s for 10 h in a BMG Fluostar microplate reader, with orbital shaking between measurements. Data was processed in R Studio and coding is available on request. Background was subtracted, and all negative values were set at 0.02 defined as the calculated starting OD₆₀₀. The maximum slope for each strain was derived from log(OD₆₀₀) using a sliding sub-set of 10 time points (~60 min) across the timeline of the growth experiment. The maximum doubling time was derived from the calculated maximum slope before the mean value was determined. All growth experiments were conducted 3 times using independent colonies.

All flagellin and outer membrane (OM) preparations were based on 1 l cultures of strains grown to an OD₆₀₀ of 0.6–0.7. Cells were centrifuged at 3,890 g and cell pellets resuspended in cold 10 mM HEPES pH 7.4. For flagellin isolation, cell suspensions were sheared using an Ultra-Turrax blending stick for 2 min set at 13,500 rpm. The same protocol was used prior to OM isolation to reduce flagellin contamination. Blended supernatants were centrifuged at 100,000 g for 1 h at 4°C to collect sheared flagellar filaments. The pellets were washed by repeating this procedure three times. Pellets were resuspended in 10 mM HEPES pH 7.4 and centrifuged at 3890 g to improve the removal of cell debris between each ultra-centrifuge wash step. The washed flagellin pellets were resuspended in 500 μl 10 mM HEPES pH 7.4 and stored at −20°C.

For outer membrane isolation cell suspensions were lysed using a Constant Systems cell disruptor at 23kPSI. Lysed cell suspensions were centrifuged at 12,000 g at 4°C for 40 min, the supernatant layered onto a sucrose gradient and centrifuged at 56,000 g for 36 h at 4°C. Outer membrane fractions were resuspended and washed once in 10 mM HEPES pH7.4. The washed outer membrane fraction was collected by centrifuging at 134,000 g for 6 h at 4°C and the resulting pellet resuspended in 500 μl 10mmM HEPES. The quality of preparations was assessed using standard SDS polyacrylamide gel electrophoresis.

Cell lysate normalized to OD₆₀₀ (Aldridge et al., 2006) and outer membrane samples (0.5 μg) were separated using a NuPAGE™ 4 to 12%, Bis-Tris, 1.0–1.5 mm, Mini Protein Gel. Proteins were transferred onto a nitrocellulose membrane (GE Healthcare) using the Bio-Rad Trans-Blot Turbo Transfer system (at 1.3 A, 25 V for 7 min). Membranes were blocked overnight at 4°C in 5% skimmed milk powder; 0.1% Tween 20 PBS solution (Aldridge et al., 2006). Following washing the membranes were incubated for 1 h with 1:10,000 dilution of α-FlgE_ST antibody (Minamino and Namba, 2008), washed and further incubated 1 h with 1:10,000 dilution of horseradish peroxidase (HRP)-conjugated Goat α-Rabbit antibody (Southern Biotech). After washing blots were exposed to Clarity Max Western ECL substrate (Bio-rad, 1:1 ratio) for 5 min and data captured using an iBright™ Imaging System (Invitrogen).

Growth conditions for the bladder RT4 cell line has been previously described (Ali et al., 2017). All challenges were performed using 24 well plates seeded with 500,000 cells in 500 μL, and the cells grown until 80–90% confluent. Bladder RT4 cells were challenged in triplicate with either isolated outer membrane preparations (100 ng/ml protein content), flagellin (0–250 ng/mL) or heat-killed whole cells at 37°C and 5% CO₂ (Ali et al., 2017). Heat-killed whole cells of each bacterial strain were prepared as previously described (Ali et al., 2017). Strains were grown to late log phase (OD600 = 0.6–0.8), cells harvested by centrifugation and resuspended in PBS prior to heating for 15 min at 100°C. No further centrifugation steps were conducted prior to the challenges allowing for intact and heat depolymerized flagellin to remain in the suspension. Prior to heating an aliquot was taken to numerate the colony forming units to calculate the necessary volume required to challenge with 1 × 10⁵ CFU/ml. Challenges were stopped after 24 h, the extracellular media collected and stored at −20°C. Interleukin 8 concentrations (pg/ml) were assayed using an eBioscience IL-8 ELISA kit following the manufacturer’s instructions. Measurement of RT4 NF-κB reporter activity was as previously described (Ali et al., 2017). For IκBα immunoblots, challenged RT4 cells were lysed in RIPA Buffer collected, quantified using a Micro BCA protein assay kit (Thermo) and stored at −20°C before 10 μg was used for immunoblots with an α-IκBα antibody (New England Biolabs).

TLR4 and TLR5 knockdown experiments using RT4 cells and siRNA were performed as described previously (Mowbray et al., 2018). siRNAs used were as follows: s14196 (TLR4) and s14199 (TLR5) and AM4611 (negative siRNA #1) as a control.

Expression of flgEA240C was analyzed following bacterial growth at 37°C in LB media containing 0.1% arabinose with shaking until an OD600 of 0.6 to 0.7. Staining of the cells was performed using malemide conjugated AlexaFluor 568 (Sim et al., 2017). Bacterial cell suspensions were immobilized on a 1% agarose padded microscope slide and samples analyzed in triplicate at 100x objective using a Nikon Eclipse Ti inverted microscope capturing both phase contrast (100 ms exposure) and red channel images (1,000 ms exposure) at five different fields of view. Five randomly chosen fields were analyzed manually using the ImageJ cell counter plugin generating data where n = 200–300 cells. Foci per cell was averaged across the five fields of view to enumerate the level of cell flagellation, as well as the distribution of flagella over the population.

Data and statistical analysis were performed using R Studio and MS Excel including the use of ANOVA and regression analysis where appropriate. Images for figure panels were processed and cropped using ImageJ and imported into Adobe illustrator for formatting. All figures were collated in Adobe illustrator to achieve the correct resolution.