Extraintestinal pathogenic Escherichia coli (E. coli) strains (ExPEC) are usually classified as facultative pathogens since they often colonize the gut without causing symptoms (Wiles et al., 2008; Köhler and Dobrindt, 2011). However, they harbor accessory traits that enable them to invade and colonize extraintestinal niches, eventually leading to symptomatic diseases (Vila et al., 2016; Sora et al., 2021). E. coli strains that have characteristic genetic markers that indicate the potential to cause extraintestinal disease, or that were directly isolated from infections outside the gastrointestinal system, are typically classified as ExPEC. ExPEC are usually divided into sepsis-causing (SEPEC), neonatal meningitis-associated (NMEC), and uropathogenic E. coli (UPEC), as well as strains causing systemic disease in animals such as avian pathogenic (APEC) or mammary pathogenic (MPEC) E. coli (Köhler and Dobrindt, 2011; Sora et al., 2021). Several virulence factors are described to promote extraintestinal pathogenicity in ExPEC, including adhesins, invasins, iron uptake systems, or toxins (Sora et al., 2021). Another bacterial trait that is involved in extraintestinal pathogenicity is quorum sensing (QS). Quorum sensing is a process in which bacteria sense and adapt to changing cell densities through signaling molecules called autoinducers (AI). Several AIs that are specific for certain bacterial strains or families have been described (Rutherford and Bassler, 2012). AI-2, on the other hand, serves as an interspecies signaling molecule that is produced and sensed by both Gram-positive and Gram-negative bacteria (Sun et al., 2004; Pereira et al., 2013). Even eukaryotic cells were proposed to participate in AI-2-mediated QS since it was found that Saccharomyces cerevisiae secretes molecules that mimic AI-2, as well as the intestinal epithelial cell (IEC) line Caco-2 in response to bacteria or a tight-junction disruption (Ismail et al., 2016; Valastyan et al., 2021). Moreover, the inflammatory interleukin IL-8 was found to be upregulated in response to external AI-2 in the IEC line HCT-8 (Zargar et al., 2015).

In bacteria, the precursor of AI-2, (S)-4,5-dihydroxy-2,3-pentanedione (DPD) (Xavier and Bassler, 2005), is produced as a byproduct of L-homocysteine synthesis by the synthase LuxS, a widely distributed and conserved enzyme that is part of the activated methyl cycle (Schauder et al., 2001; Vendeville et al., 2005). DPD is a hydrophilic molecule that is actively transported out of the bacterial cell after production (Khera et al., 2022). Outside of the cell, DPD spontaneously circularizes into either boron-containing (S)-2-methyl2,3,3,4-tetrahydroxytetrahydrofuran-borate or boron-free (R)-2-methyl-2,3,3,4-tetrahydroxytetrahydrofuran (R-THMF), resulting in two different types of active AI-2 (Chen et al., 2002; Miller et al., 2004). Depending on the environmental status, DPD can spontaneously convert between the two active forms and thus be sensed by mixed bacterial communities since both AI-2 forms are perceived by different receptors (Zhang et al., 2020). R-THMF, from now on referred to as AI-2, is an intracellular signaling molecule (Xavier and Bassler, 2005). Different receptors for AI-2 are known, but the best-studied is LsrB. LsrB was first described in Salmonella enterica serovar Typhimurium (S. Typhimurium) (Taga et al., 2001) and is part of a specialized transporter encoded in the lsr (LuxS-regulated) locus (Taga et al., 2003). The lsr locus, or homologs, have been found in various bacterial species, including E. coli (Pereira et al., 2009), and comprises two transcription units. The first transcription unit, lsrRK, encodes the kinase LsrK that phosphorylates the incorporated AI-2 and the repressor LsrR that represses transcription of both lsr operons (Xue et al., 2009), as well as of several other genes in E. coli (Li et al., 2007) and S. Typhimurium (Choi et al., 2012). The binding of phosphorylated AI-2 to LsrR abolishes the transcriptional repression, resulting in AI-2-dependent gene expression (Xavier and Bassler, 2005). The second transcription unit, lsrACDBFG, encodes structural proteins of the AI-2 transporter, i.e., LsrA, LsrC, LsrD, and LsrB as the receptor, as well as the isomerase LsrG that degrades the phosphorylated AI-2 and the thiolase LsrF that catalyzes the terminal step in processing phosphorylated AI-2 (Marques et al., 2014) ( Figure 1 ).

Interfering with AI-2-dependent QS, either by using QS inhibitors (Helmy et al., 2022) or gene mutations and deletions, led to altered behavior and lower virulence in ExPEC strains. For example, the deletion of luxS has led to a significant reduction of virulence in the APEC O78:K80:H9 strain χ7122 (Palaniyandi et al., 2013) and an attenuated virulence of the APEC strain DE17ΔaroA (Han et al., 2015). Also, the deletion of lsrACD in APEC94, which leads to an impaired AI-2 uptake and decreased bacterial motility, has led to lower bacterial loads and reduced virulence in a duckling infection model (Zuo et al., 2019). Additionally, it was shown that AI-2-dependent QS is associated with hydrogen peroxide (H₂O₂) resistance in the MPEC strain DCM5 (Wang et al., 2021) and antibiotic sensitivity in APEC strain APECX40 and the MPEC isolate DCM1 (Xue et al., 2016; Yu et al., 2020). Although AI-2-dependent QS has significant effects on APEC and MPEC, little is known about its effects on the virulence and fitness traits of UPEC, which are often genetically very similar to APEC and MPEC (Rodriguez-Siek et al., 2005; Tivendale et al., 2010). UPEC are the leading cause of urinary tract infections (UTI), one of the most prevalent bacterial infections worldwide, accounting for at least 75% of all complicated and uncomplicated cases (Flores-Mireles et al., 2015). Despite several host defense mechanisms, including the physicochemical composition of urine, the expression of antimicrobial peptides, or other innate immune response mechanisms (Abraham and Miao, 2015; Loubet et al., 2020), more than 50% of all women (Medina and Castillo-Pino, 2019) and about 20% of all men (Farrell et al., 2021) suffer at least from one symptomatic UTI episode during their lifetime. This indicates that UPEC strains are well adapted to the bladder environment and can combat the host’s defense strategies (Flores-Mireles et al., 2015; Mann et al., 2017).

The asymptomatic bacteriuria (ABU) E. coli isolate 83972 evolved from UPEC but has lost the ability to express many functional virulence factors (Zdziarski et al., 2008). This strain is well adapted to long-term bladder colonization without causing inflammation or other UTI-related symptoms (Hull et al., 1999). Additionally, E. coli 83972 has a better antioxidant defense than UPEC strains (Aubron et al., 2012) and can outcompete UPEC strains in human urine, amongst others, probably due to fast growth and nutritional adaptation (Ipe et al., 2016; George et al., 2024). Due to these characteristics, deliberate bladder colonization by E. coli 83972 has been suggested as a therapeutic approach to prevent recurring UTI by bacterial interference (Sundén et al., 2010; Köves et al., 2014; Wullt and Svanborg, 2016). Against the background of the increasing spread of multidrug-resistant bacteria, especially in the context of UTI, the use of antibiotics must be reduced (Cook and Wright, 2022). Bacterial interference by bladder colonization with ABU strain 83972 is an alternative way to prevent permanent colonization of the urinary bladder by symptomatic uropathogens (Wullt and Svanborg, 2016; Stork et al., 2018; Kenneally et al., 2022). Deeper insights into the importance of QS for the fitness and competitiveness of ABU strain 83972 may help to optimize the use of this bacterial strain for the deliberate therapeutic colonization of patients by uncovering underlying molecular mechanisms and thus improving the efficacy and safety of this form of treatment as an interesting alternative to antibiotic therapy. Interestingly, E. coli 83972 lacks the complete lsr locus but carries the luxS gene. We hypothesized that the absence of AI-2-dependent QS contributes to the strain’s fitness during bladder colonization. Therefore, we integrated the full-length lsr locus from the E. coli K-12 strain MG1655 into the chromosome of E. coli 83972 and analyzed the transcriptome of the lsr complemented strain to screen for differentially expressed genes that are affected by AI-2-dependent QS and may be relevant during bladder colonization. Additionally, we tested for relevant bacterial phenotypes that may increase this strain`s fitness in the urinary bladder.

Bacterial strains and culture conditions. The bacterial strains used in this study are listed in Supplementary Table S5 . The asymptomatic bacteriuria E. coli isolate 83972 has been obtained as a gift from C. Svanborg (Lund, Sweden). Bacterial cultures were either cultivated in lysogeny broth (LB) (10 g/L tryptone, 5 g/L yeast extract, and 5 g/L NaCl) or pooled human urine (four male and six female voluntary individuals, sterile filtered and mixed in a 1:1 male/female ratio (v/v)). When appropriate, antibiotics were added in the following concentrations: kanamycin (25 µg/mL), chloramphenicol (12.5 µg/mL), ampicillin (100 µg/mL), or zeocin (50 µg/mL). Bacterial strains were grown overnight at 37°C on LB agar plates (containing 1.5% agar (w/v)) with the appropriate antibiotics when needed. For overnight cultures, single colonies were picked and incubated in 2 mL of LB at 37°C and with orbital shaking at 180 rpm. E. coli DH5α was used as a host for plasmid construction. Genome manipulation was done by recombineering (Datsenko and Wanner, 2000). When indicated, cultures were supplemented with H₂O₂ (stabilized, Merck Millipore, Darmstadt, Germany) or synthetic DPD purchased from Rita Ventura’s research group at ITQB-UNL (Oeiras, Portugal) (https://itqb.unl.pt/research/chemistry/bioorganic-chemistry). Synthetic DPD was synthesized as published before (Ascenso et al., 2011), and the concentration was 16.8 mM. Both chemicals were subdiluted to working concentrations using sterile ddH₂O.

Cloning methods. DNA amplification for cloning and genetic manipulation was done using the Q5 High-Fidelity DNA Polymerase (New England Biolabs, Frankfurt/Main, Germany). Colony-PCR was done using the GoTaq Green Master Mix (Promega GmbH, Walldorf, Germany). After amplification, the correct size of amplicons was verified using agarose gel electrophoresis in 1x Tris-acetate-EDTA (TAE) buffer with 1-2% (w/v) agarose. The gels were run at 110–130 V for 30–60 min. Amplicons were purified using the NucleoSpin Gel and PCR Clean-up kit (Macherey-Nagel, Düren, Germany) or the Wizard SV Gel and PCR Clean-Up System kit (Promega, Walldorf, Germany). Plasmids were isolated using the NucleoSpin Plasmid Mini kit (Macherey-Nagel). Genomic DNA (gDNA) was isolated using the QIAamp DNA Mini kit (QIAGEN, Hilden, Germany). Purified PCR products were quantified using a spectrophotometer (NanoDrop 2000, Thermo Fisher Scientific, Schwerte, Germany). The restriction enzymes and T4 DNA ligase used for cloning were purchased from New England Biolabs (Frankfurt/Main, Germany).

Oligonucleotides and plasmids used. All oligonucleotides and plasmids used are listed in Supplementary Tables S6 , S7 .

Construction of pMK2. The plasmid pMK2 was constructed by integrating the lsrA promoter-yfp fusion reporter cassette coupled to lsrR (lsrR:PlsrA:yfp fusion) from the previously described E. coli K-12 strain MG1655 (lsrA-G::yfp-cat) (Keizers et al., 2022) into the low-copy number plasmid pWKS30 (Wang and Kushner, 1991). For that, the lsrR:PlsrA:yfp fusion was amplified from the gDNA of E. coli K-12 strain MG1655 (lsrA-G::yfp-cat) using the oligonucleotides pKD3_lsr_seq_4 and LZP50 and introduced into the SmaI-digested pWKS30 by using DNA T4 ligase. The correct insertion of the PCR product into pWKS30 was confirmed by plasmid preparation and subsequent Sanger sequencing.

Construction of pLS1 and pLS2. The low-copy-number plasmids pLS1 and pLS2 were constructed by integrating the Pdps-cfp (cfp under the control of the stationary phase-dependent dps promoter) cassette or the Pdps-yfp (yfp under the control of the stationary phase-dependent dps promoter) cassette, respectively, into pWKS30. Both cassettes were amplified using the oligonucleotides bla-TEM-r and MC_185 and pPS1 (Pdps-cfp) or pPS2 (Pdps-yfp) (Schiller et al., 2021) as templates. Amplicons and pWKS30 were cut with KpnI and SmaI. All products were purified, and the digested Pdps-cfp cassette was introduced into digested pWKS30, resulting in pLS1. The digested Pdps-yfp cassette was introduced into digested pWKS30, resulting in pLS2. Ligation was done using DNA T4 ligase. Correct plasmids were confirmed by fluorescence microscopy.

Construction of E. coli strains 83972 ΔluxS and 83972 ΔybhC. For the construction of E. coli 83972 ΔluxS, the luxS gene was replaced by a zeocin resistance cassette (bleR) that was amplified using the oligonucleotides Rec_BleoR_fwd and Rec_BleoR_rev and pEM7/Zeo as a template. Replacement was confirmed by colony PCR using the oligonucleotides CFT073_ΔluxS_CP_fwd and CFT073_ΔluxS_CP_rev and Sanger sequencing of the purified amplicon. For the construction of E. coli strain 83972 ΔybhC, the ybhC gene was replaced by a chloramphenicol resistance cassette (cat) that was amplified using the oligonucleotides del_ybhC_fwd and del_ybhC_rev and pLP2 (Peng et al., 2022) as a template. Replacement was confirmed by colony PCR using the oligonucleotides wt_PybhC and LZP18 and Sanger sequencing of the purified amplicon.

Construction of E. coli strains 83972 attB::lsr and 83972 ΔluxS attB:: lsr. First, the plasmid pWKS30_LSR was generated. The chloramphenicol resistance cassette (cat) was introduced upstream of the lsr locus in the chromosome of E. coli strain MG1655 by recombineering (Datsenko and Wanner, 2000). The resistance cassette was amplified using the oligonucleotides CBL_fwd and CBL_rev and pMB54 (Berger et al., 2016) as a template. The correct chromosomal insertion was confirmed by colony PCR using the oligonucleotides LKRS_fwd and pMB54_lsrRK_CP1_rev. In the same way, the kanamycin resistance cassette (aph) was chromosomally inserted downstream of the lsr locus in the resulting E. coli strain MG1655 cat_lsr. The aph resistance cassette was amplified using the oligonucleotides RBL_fwd and RBL_rev and pKD4 as a template. The correct chromosomal insertion was confirmed by colony PCR using the oligonucleotides pKD3_lsr_seq_12 and LZP50. The full-length lsr locus, flanked by cat and aph, was amplified from gDNA of the resulting E. coli strain MG1655 cat_lsr_aph using the oligonucleotides LKRS_fwd and LZP50, and cloned into the SmaI-digested pWKS30 by using DNA T4 ligase. The relevant parts of the resulting plasmid pWKS30_LSR were afterward confirmed by Sanger sequencing. The complete lsr determinant was then inserted by recombineering (Datsenko and Wanner, 2000) next to the chromosomal attachment site of the bacteriophage λ (attB) in either E. coli 83972 or strain 83972 ΔluxS. For this, the lsr locus flanked by cat and aph (cat_lsr_aph) was amplified using the oligonucleotides pWKS30_attB_fwd and pWKS30_attB_rev and pWKS30_LSR as a template and introduced next to the λ attB site of the strains 83972 and 83972 ΔluxS, respectively. The correct chromosomal insertion was confirmed by colony PCR using the oligonucleotides CFT073_lsrRK_YFP_CP1_fwd and pKD3_lsr_seq_12 or CFT073_lsrRK_YFP_CP1_fwd and CFT073_lsrRK_YFP_CP2_rev and pKD3_lsr_seq_13. Afterward, gDNA was isolated, and the cat_lsr_aph region was amplified using the primer pair CFT073_lsrRK_YFP_CP1_fwd and CFT073_lsrRK_YFP_CP2_rev before the correctness of the DNA sequence was also verified by Sanger amplicon sequencing.

RNA isolation. Total RNA was isolated as previously described (Wallenstein et al., 2020) with the following differences: Overnight cultures of E. coli strains 83972 and 83972 attB::lsr were diluted to a final optical density OD600 = 0.02 in 200 mL LB. At indicated time points, 70 mL of culture (lag phase), 5 mL of culture (exponential phase), or 2 mL of culture (stationary phase) were added to the 0.2x volume of pre-chilled stop solution, and the bacterial pellets were directly lysed. Confirmation of complete gDNA removal was done by qPCR. RNA was extracted from three biological replicates for each strain.

RNAseq and data analysis. Strand-specific cDNA libraries were prepared from the isolated RNA and sequenced (Illumina NextSeq 500, 1 x 75 bp single reads) by vertis Biotechnology AG (Freising, Germany). The obtained raw reads were analyzed as follows: Quality control using FastQC and MultiQC (Bioinformatics; Ewels et al., 2016), adaptor trimming using Cutadapt (Martin, 2011) with subsequent quality control, alignment of reads using burrows-wheeler alignment (Li and Durbin, 2009) to the genome of E. coli 83972 attB::lsr with a subsequent strandedness check using RSeQC (Wang et al., 2012) and quality control, count reading using featureCounts (Liao et al., 2014) and a subsequent differentially expressed gene analysis using DESeq2 (Love et al., 2014) in R. Conversion of formats was done using AGAT (Dainat, 2022). The subsequent data analysis was done in R. The Venn diagram ( Figure 2A ) was made using venn.diagram (Chen, 2022). The heatmaps ( Figures 2B–D ) were made using pheatmap (Kolde, 2019). Differentially expressed genes were clustered using dist() and hclust(), and biological replicates were clustered using pheatmap (cluster_cols = T) with the complete linkage method for hierarchial clustering. The scatterplots ( Supplementary Figure S2 ) were computed using pairs.panels() from the package psych (Revelle, 2023).

Growth analysis and co-culture assays. For growth analysis, overnight cultures of the bacterial strains were diluted to a final optical density (OD600) of 0.01 in 1 mL of fresh medium with antibiotics when appropriate. 200 µL of freshly diluted cultures were added into one well of a 96-well plate (Thermo Fisher Scientific, Schwerte, Germany) in duplicates for each biological replicate. For the subsequent growth analysis, either an Infinite M NANO+ or an Infinite F200 microplate reader (both from TECAN, Männedorf, Switzerland) was used. The optical density was either measured at 595 nm (± 9 nm) (M NANO+) or 595 nm (± 10 nm) (F200). Signals were measured in ten-minute intervals. The resulting growth curves were analyzed using AMiGA (Analysis of Microbial Growth Assay (Midani et al., 2021)). For the lag phase length analysis, the data was not log-transformed. The co-culture assays were performed as previously described (Keizers et al., 2022). Briefly, overnight cultures of the strains to be analyzed were mixed in a 1:1 ratio with the E. coli K-12 reporter strain MG1655 ΔluxS attB::PlsrA-yfp to a final optical density (OD600) of 0.02. Growth analysis was performed as described above, and the fluorescence signal was either measured using an excitation wavelength of 514 nm (± 9 nm) and an emission wavelength of 550 nm (± 20 nm) (M NANO+) or using an excitation wavelength of 485 nm (± 20 nm) and an emission wavelength of 535 nm (± 25 nm) (F200).

Competition assays. For the competition assays, overnight cultures of E. coli strains 83972 (pLS2) and 83972 attB::lsr (pLS1) were mixed in a 1:1 ratio to a final optical density (OD600) of 0.02 in 25 mL of fresh medium supplemented with ampicillin. The cultures were grown at 37°C with orbital shaking at 180 rpm. After 3 h of growth, the cultures were subdiluted into 25 mL of fresh medium containing ampicillin with an appropriate dilution factor (1:200 for LB and 1:10 for pooled human urine). After subdilution, the cultures were grown for another 3 h. The subdilution procedure was repeated three times with a total assay time of 12 h. For the long-term competition, the cultures were grown at 37°C with orbital shaking at 180 rpm for 72 h. Directly after mixing and at the indicated time points, aliquots were taken from the cultures and microscopically analyzed using a Leica inverted microscope DMi8 with an attached camera at 400 x magnification. For each aliquot, ten microscopic pictures were taken with three channels each – Differential interference contrast (DIC), cyan fluorescent protein CFP (excitation 436 nm (± 20 nm); dichroic mirror 455 nm; emission 480 nm (± 40 nm) and fluorescein isothiocyanate (FITC) (excitation 480 nm (± 40 nm); dichroic mirror 505 nm; emission 527 nm (± 30 nm). Next, the number of bacteria per picture was quantified in the CFP channel and the FITC channel using an ImageJ script. The ratio of bacterial numbers in the FITC channel compared to the total bacterial numbers in both the FITC and CFP channels was calculated for each of the ten pictures, resulting in one mean ratio value with a standard deviation for each time point and biological replicate.

H₂O₂ resistance assay. For the H₂O₂ resistance assay, overnight cultures of the bacterial strains were diluted to a final optical density (OD600) of 0.01 in 1 mL of fresh medium. 190 µL of freshly diluted cultures were added into one well of a 96-well plate (Thermo Fisher Scientific) in duplicates for each biological replicate. After adding cultures, 10 µL of H₂O₂ diluted in ddH₂O was added into each well, resulting in the indicated final H₂O₂ concentrations. The growth was analyzed directly after the addition using an Infinite M NANO+ plate reader (TECAN) or an Infinite F200 (TECAN). The optical density was either measured at 595 nm (± 9 nm) (M NANO+) or 595 nm (± 10 nm) (F200). Signals were measured in ten-minute intervals. The resulting growth curves were analyzed using AMiGA (Analysis of Microbial Growth Assay (Midani et al., 2021)). For the lag phase length analysis, the data was not log-transformed.

Galleria mellonella larvae feeding assay. Galleria mellonella larvae were purchased from Fauna Topics Zoobedarf Zucht und Handels GmbH (Marbach am Neckar, Germany) and reared on an artificial diet (22% maize meal, 22% wheat germ, 11% dry yeast, 17.5% beeswax, 11% honey, and 11% glycerin) as previously described (Mukherjee et al., 2020). To evaluate the fitness of the E. coli wild type strain 83972 and its lsr-complemented mutant in the larval digestive tract, overnight cultures (OD 600 = 1, grown in LB) of both strains were mixed in a 1:1 ratio and force-fed to sixth-instar larvae (weighing approximately 250–300 mg) (Lange et al., 2019). A 10-μL aliquot of this bacterial suspension was gently introduced into the larval mouth using 1-mL disposable syringes fitted with 0.4 x 20-mm needles mounted on a microapplicator. Control larvae received an equivalent volume (10 μL) of sterile LB. Following force-feeding, larvae were incubated at 37°C for 24 hours without food. All larvae survived the incubation period and were subsequently flash-frozen in liquid nitrogen, ground into a fine powder, and homogenized in LB. The homogenates were then plated onto LB agar, and colony-forming units (CFUs) were counted following a 24-hour incubation at 37°C. The survival of the wild type and lsr-complemented mutant strains was analyzed by comparing their initial inoculation ratio to the ratio recovered from larvae after 24 hours of incubation.

Detection of the lsr determinant in E. coli genomes. To screen for the presence and conservation of the lsr determinant in E. coli, all published E. coli genomes that were available at the time point of the analysis were downloaded from NCBI Assembly (Kitts et al., 2016) (time point of download: 15.03.2023) in FASTA format (Status: latest RefSeq). In total, 32,594 genomes were downloaded. Next, the phylogroup was determined using ClermonTyping (Beghain et al., 2018) for each genome. The analysis was continued with 32,404 genomes belonging to the phylogroups A, B1, B2, C, D, E, F, and G. Next, each genome was analyzed using the basic local alignment search tool (BLAST+ (Camacho et al., 2009)) to screen for the nucleotide sequence (blastn) of the full-length lsr locus (lsrRK-lsrACDBFG), the luxS gene and the tam gene (reference sequences from the E. coli K-12 strain MG1655). Output possibilities were “complete” (blastn hit length equal query length), “incomplete” (blastn hit length inequal query length), “no match” (no blastn hit) or “multiple matches” (multiple blastn hits). The output “< 100 bp” is a summary of a blastn hit length of below 100 bp and the “no match” hits. Please note that our analysis only focuses on gene length, whereas the functionality of the encoded gene product was not part of the analysis.

Statistical analysis. Statistical analysis was performed using GraphPad Prism v8.0.2 (San Diego, USA).

AI-2-dependent QS was shown to play an essential role in APEC and MPEC strains for virulence, colonization, or resistance (Palaniyandi et al., 2013; Han et al., 2015; Xue et al., 2016; Zuo et al., 2019; Yu et al., 2020; Wang et al., 2021; Helmy et al., 2022). However, except for knowledge about the role of QS-induced biofilm formation in catheter-associated UTIs (Laganenka and Sourjik, 2018; Henly et al., 2021; Lila et al., 2023), little is known about the function of AI-2-dependent QS in UTIs. E. coli 83972 is naturally lsr-negative but luxS-positive and causes asymptomatic bacteriuria (Medina and Castillo-Pino, 2019). We analyzed this strain’s ability to produce, export, import, and phosphorylate AI-2. We found that E. coli 83972 produced AI-2, which could be sensed by another E. coli strain during co-culture ( Figure 2B ). The AI-2 production was LuxS-dependent, as judged by the inability to induce AI-2-dependent YFP expression in our reporter strain when luxS was deleted in E. coli 83972 ( Figure 2B ). However, we observed no AI-2-dependent YFP expression when E. coli 83972 ΔluxS (pMK2) was challenged with extracellular DPD ( Figures 2D, E ). In comparison, the addition of low concentrations of extracellular DPD to E. coli 83972 ΔluxS (pMK1) has led to a significantly higher YFP expression compared to the water control ( Figure 2F ). This result was expected since E. coli 83972 does not encode LsrK, which is so far the only described kinase that phosphorylates incorporated AI-2 (Pereira et al., 2012). Since AI-2 is initially taken up in an Lsr-independent way by a phosphoenolpyruvate phosphotransferase system, the lack of an active AI-2 kinase would not lead to an AI-2 sequestration inside the cell because unphosphorylated AI-2 is transported back into the extracellular space (Marques et al., 2011; Pereira et al., 2012; Trappetti et al., 2017). When pMK1, a plasmid encoding LsrK, was introduced into E. coli 83972, externally added AI-2 was incorporated, phosphorylated, and sequestered inside the cell, while the introduction of pMK2, lacking lsrK, did not lead to AI-2 phosphorylation. These results let us conclude that E. coli 83972 produces AI-2, which can be sensed by other bacteria but cannot process AI-2 itself. This is not unusual, as other bacteria are also known to be unable to sense AI-2 and express luxS only because of its role in the activated methyl cycle (Rezzonico and Duffy, 2008). One example is the uropathogenic Proteus mirabilis strain BB2000, which does not possess a luxP or lsrB homolog but does express luxS (Rezzonico and Duffy, 2008). Furthermore, the deletion of luxS did not lead to altered colonization or pathogenicity of this strain during UTI in a mouse model (Schneider et al., 2002), suggesting that AI-2-dependent QS does not play a role during UTI in this strain. To reintroduce the ability to use AI-2 as a signaling molecule, we integrated the lsr locus into the chromosomal attB site of E. coli 83972 ( Figure 2A ). Indeed, when E. coli 83972 ΔluxS attB::lsr (pMK2) was challenged with extracellular DPD, we observed a significantly higher YFP expression as compared to the water control ( Figures 2D, E ). This result suggests that lsrK is functionally expressed in E. coli 83972 attB::lsr. Additionally, E. coli 83972 attB::lsr could export AI-2, which could be sensed by another E. coli strain in coculture ( Figure 2C ).

The observation that global gene expression of UPEC strains in vivo during human UTI (hUTI) or murine UTI (mUTI) was more similar to their in vitro gene expression in LB than to their in vitro gene expression in pooled human urine (Frick-Cheng et al., 2020) allows us to assess the importance of lsr-dependent gene expression for colonization of the urinary bladder using our own in vitro RNA-seq data in LB. Our transcriptome analysis revealed that the transcriptomes of the E. coli strains 83972 and 83972 attB::lsr were not drastically different during the three growth phases ( Supplementary Figure S2 ). Only 60–72 genes were differentially expressed in E. coli strains 83972 and 83972 attB::lsr with a log2-fold change of higher or lower 1 ( Figure 3 ). By comparing the results of Frick-Cheng and colleagues (Frick-Cheng et al., 2020) to our data, we observed that some genes or operons that were differentially expressed in vivo during hUTI/mUTI relative to in vitro growth in LB, also seem to be affected by AI-2-dependent QS. For example, several genes from the flagellar machinery (flgCFGLM and fliS) were downregulated during mUTI/hUTI as compared to growth in LB (Frick-Cheng et al., 2020). More than 50% of all downregulated genes in E. coli 83972 attB::lsr in the lag phase were linked to the flagellar machinery ( Table 2 ). The link between AI-2-dependent QS and flagellar gene expression was also seen in the EHEC strain 86-24, where the deletion of luxS led to a downregulation of flagellar genes (Sperandio et al., 2001), which is contradictory to our results where AI-2-dependent QS seems to repress flagellar gene expression. In UPEC strain CFT073, flagella expression was not crucial for bladder colonization efficiency, whereas the loss of flagellar genes was a disadvantage during bladder colonization (Lane et al., 2005). During UTI, the flagella induce a Toll-like receptor-dependent immune response and, thus, inflammation (Subashchandrabose and Mobley, 2015; Acharya et al., 2019). However, since the flhDC genes coding for the master regulator of the flagellar biosynthesis cascade (Subashchandrabose and Mobley, 2015), and the highly antigenic flagellin, encoded by fliC (Acharya et al., 2019), were not differentially expressed in E. coli 83972 attB::lsr ( Table 2 , Supplementary Table S1 ), and since E. coli 83972 expresses only little flagella (Hancock et al., 2008), it appears unlikely that the further downregulation of already weakly expressed flagellar genes is negatively impacting this strain’s bladder colonization ability.

One of the main characteristics of E. coli 83972 is its fast growth in urine and, thus, the outgrowth of competing UPEC strains in direct competition (Roos et al., 2006; Ipe et al., 2016). One factor that might contribute to this higher competitiveness is metabolic adaptation (Ipe et al., 2016). We found several genes associated with metabolism to be differentially expressed in all three growth phases in E. coli 83972 attB::lsr ( Table 1 ). AI-2-dependent QS was already linked to sugar metabolism (Ha et al., 2018), catabolite repression (Wang et al., 2005) and carbon regulation (Mitra et al., 2016). The strong upregulation of genes involved in the valine, leucine, and isoleucine biosynthesis (ilv operon) was interesting since ilvA and ilvC were found to be essential factors during growth in urine (Vejborg et al., 2012), probably because the overall concentrations of the three amino acids are relatively low as compared to other amino acids in urine (Bouatra et al., 2013) ( Table 1 , Supplementary Table S1 ). In E. coli strain CFT073, valine overproduction is considered a metabolic adaptation during biofilm formation (Valle et al., 2008). We also found several genes related to biofilm formation and adhesion upregulated in the transition to the stationary phase upon restoration of the lsr operon ( Supplementary Table S1 , Table 2 ). Thus, we checked for biofilm formation according to the protocol of (Laganenka et al., 2016). in E. coli strains 83972 and 83972 attB::lsr, but observed no significant differences ( Supplementary Figure S7 ). Thus, the observed lsr-dependent upregulation of the ilv genes had no marked effect on biofilm formation of E. coli strain 83972 attB::lsr.

We wondered whether the differential regulation of metabolic genes in E. coli 83972 attB::lsr might affect growth. Indeed, we observed a growth retardation of this strain compared to the lsr-negative wild type 83972, independent of growth medium or oxygen availability ( Figure 4 , Supplementary Figure S3 ). Consistent with the findings of Sperandio and colleagues made in EHEC strain 86-24 (Sperandio et al., 2001), we observed that the deletion of luxS resulted in a significantly longer lag phase of E. coli strain 83972 ( Supplementary Figure S8 ). Interestingly, the restoration of the lsr determinant in E. coli 83972 significantly extended the lag phase as well as the doubling time of E. coli 83972 ( Supplementary Figure S8 ). As we did not observe an additive effect of luxS deletion and lsr restoration on the length of the lag phase in E. coli 83972, and since luxS was not differentially expressed in E. coli 83972 attB::lsr in all three growth phases, the longer lag phase of the strains 83972 ΔluxS and 83972 ΔluxS attB::lsr as compared to the wild type might thus be due to the loss of LuxS as metabolic enzyme (Schauder et al., 2001; Vendeville et al., 2005; Zdziarski et al., 2008) and not due to AI-2 secretion, as E. coli 83972 and E. coli 83972 ΔluxS are not able to sense AI-2. Taken together, these findings showed that the presence of the full-length lsr locus alone was not sufficient to cause the observed slowdown in growth but that it most likely resulted from an AI-2-dependent deregulation of specific gene expression. Because we performed the complementation as a single copy-insertion of the lsr determinant into a neutral site of the chromosome, overexpression effects and artificial phenotypes due to incompatibilities in gene regulation or metabolic stress caused by overexpression should be excluded. To validate our assumption, additional E. coli isolates will have to be tested for their lsr-dependent growth characteristics. In this context, fitness costs from the two cat and aph resistance cassettes used for complementation could be considered as well as non-QS-related functions of the lsr determinant.

In the intestinal tract, where the bacterial composition is diverse and the bacterial load is high, a negative effect of AI-2-dependent QS on growth rate does not necessarily lead to a reduction in fitness as AI-2-dependent QS significantly impacts microbiota composition, which in turn affects the fitness of individual bacterial species (Christiaen et al., 2014; Thompson et al., 2015; Deng et al., 2022). Nevertheless, our feeding experiments of G. mellonella larvae with equal ratios of E. coli 83972 wild type and the lsr-complemented strain impressively showed that the restoration of the lsr operon significantly reduced the colonization ability and competitiveness of E. coli 83972 attB::lsr in the digestive tract compared to the lsr-negative parent strain. We are fully aware that the human intestinal tract and that of G. mellonella larvae differ substantially, not only in their microbiome composition. We applied G. mellonella larvae as a 3R-compliant model to investigate the ability of E. coli 83972 to assert itself in a niche densely populated by a complex microbiota. We view this experiment as a proxy for intestinal colonization and understand that a more relevant colonization model for the human digestive tract must be used in further studies to clarify the impact of AI-2-dependent QS on the intestinal colonization capacity of phylogroup B2 strains. It has recently been described in a mouse model that AI-2 production increases chemotaxis and thus the metabolic trait-dependent fitness of E. coli strain Z1331 in the intestine. AI-2-dependent chemotaxis can also promote the coexistence of different E. coli strains in the intestine through niche segregation (Laganenka et al., 2023). Related to hardly flagellated E. coli strain 83972, we interpret our results as indicating that the importance of AI-2 in bacterial fitness in the digestive tract can be strain-dependent and influenced by individual metabolic and phenotypic characteristics. In the human bladder, however, where bacterial numbers are typically not as high as in the intestinal tract (Perez-Carrasco et al., 2021), regulating growth by monitoring cell density may be a disadvantage for urine isolates, because of rapid changes of the population size in the bladder. The urine void drastically reduces the overall bacterial count and volume of the growth compartment, urging non-voided bacteria to regrow fast to maintain bladder colonization, especially when competing with a second bacterial species (Ipe et al., 2016). QS is known to slow down growth and metabolism as soon as a high cell density is reached (Sperandio et al., 2001; An et al., 2014). When the bacteria encounter a rapid reduction in cell density, this switch leads to a subsequent adaptation of gene expression in quorum sensing strains (Pereira et al., 2013), which likely differs from that in bacteria that cannot sense the bacterial cell density. In our co-cultivation experiment, where the lsr-positive strain 83972 attB::lsr (pLS1) and the lsr-negative strain 83972 (pLS2) were in direct competition, the lsr-positive variant capable of QS was outcompeted by the isogenic lsr-negative wild type, which is unable to sense bacterial cell density but displays a shorter lag phase and doubling time ( Figure 4 ). The observed slowdown in growth and prolonged doubling time of the lsr-positive variant could underlie the displacement of this mutant by the QS-insensitive wild type due to QS-dependent adaptations in gene expression. Especially in the case of frequent changes in population density, here mimicked by serial passage after 3 h of growth ( Figure 5A ), compared to the long-term competition experiment ( Figure 5B ), the displacement by the lsr-negative strain is particularly evident. This property benefits E. coli 83972 in bacterial interference in the urinary bladder because it can compensate for bottlenecks due to rapid reductions in population size as a result of repeated voiding. In summary, our results suggest that the differential expression of metabolic genes by AI-2-dependent QS may result in different growth characteristics and reduced competitiveness of E. coli 83972 attB::lsr. The direct evidence, however, is still missing, and the precise regulatory mechanisms remain to be elucidated.

The nrdHIEF operon was upregulated during mUTI/hUTI as compared to growth in LB (Frick-Cheng et al., 2020), and we found nrdHIEF upregulated in E. coli 83972 attB::lsr as compared to E. coli 83972 in the lag phase ( Table 3 ). Though the operon encodes a ribonucleotide reductase that is required for dNTP synthesis and is thus associated with DNA synthesis, nrdHIEF expression was found to be stimulated upon oxidative stress, particularly in mutants that miss major antioxidant defenses (Monje-Casas et al., 2001). It is thought that E. coli overexpresses several reductases and electron donors, including nrdHIEF, to cope with oxidative stress (Monje-Casas et al., 2001; Abdelwahed et al., 2022). Interestingly, the gene ybhC, which leads to a significantly lower minimal inhibitory concentration against H₂O₂ when deleted in the E. coli strain BW25113 (Chen et al., 2021), was the most strongly downregulated gene during the lag and exponential growth phase and was also strongly downregulated during the transition to the stationary phase in E. coli 83972 attB::lsr ( Supplementary Table S1 ). We wondered whether this strong downregulation might contribute to a higher sensitivity to H₂O₂. Therefore, we challenged E. coli 83972, its isogenic ybhC deletion mutant 83972 ΔybhC, and E. coli 83972 attB::lsr with H₂O₂ concentrations that can be physiologically measured in human urine (Varma and Devamanoharan, 1990). Indeed, we observed that E. coli 83972 was more resistant to H₂O₂ than the strains 83972 ΔybhC and 83972 attB::lsr ( Figure 6 ). We observed that the ybhC deletion did not affect the length of the lag phase as compared to E. coli 83972 (unpaired t-test; p = 0.19), whereas E. coli 83972 attB::lsr had a significantly longer lag phase than both, E. coli 83972 and E. coli 83972 ΔybhC ( Figure 6 ). Thus, the overall longer lag phase of the lsr-positive variant of strain 83972 does not explain its higher sensitivity to H₂O₂, while the strong downregulation of ybhC in this strain might be one factor responsible for this phenotype. Notably, the increased H₂O₂ sensitivity of E. coli 83972 attB::lsr is in agreement with a previous report of increased H₂O₂ sensitivity in the MPEC strain DCM5 expressing a functional LsrR (Wang et al., 2021). However, in this MPEC strain, the higher sensitivity was explained by LsrR-mediated repression of H₂O₂ scavenging enzyme-encoding genes, including ahpCF. As opposed to that, we found that ahpCF was slightly upregulated in strain 83972 attB::lsr in the lag phase ( Supplementary Table S1 ). Therefore, the decreased expression of ahpCF is unlikely to explain the lsr-dependent sensitivity of E. coli 83972 to H₂O₂. E. coli 83972 has an increased level of endogenous reactive oxygen species while growing in urine compared to other ABU and UPEC strains, but correspondingly, it also has a more active antioxidant defense system (Aubron et al., 2012). We found several differentially expressed genes in E. coli 83972 attB::lsr that have been associated with oxidative stress in other studies ( Supplementary Table S4 ) or other stress-related responses ( Table 4 ). Thus, the overall H₂O₂ detoxification or stress response seems to be imbalanced in the presence of functional AI-2-dependent QS in E. coli 83972 attB::lsr, resulting in a higher sensitivity against oxidative stress. During ABU, however, increased sensitivity to H₂O₂ is a disadvantage since the infiltration of activated neutrophils into the bladder represents one host defense mechanism to kill bacteria with reactive oxygen species during UTI (Demirel et al., 2020). Especially in a competition situation in the bladder, a significantly prolonged lag phase and doubling time of E. coli 83972 attB::lsr in addition to the increased H₂O₂ sensitivity could constitute a clear colonization disadvantage relative to the lsr-negative wild type strain ( Supplementary Figure S5 ).

AI-2-dependent QS in ExPEC has pleiotropic effects since it contributes to virulence, colonization, and antibiotic resistance but also H₂O₂ sensitivity, metabolism, and growth defects, as can also be seen in our study (Xue et al., 2009; Kathayat et al., 2021; Wang et al., 2021). Epidemiological studies have shown that ExPEC frequently belong to the phylogroup B2 (Picard et al., 1999; Micenková et al., 2016). However, in the collection of E. coli genomes we examined, genomes of phylogroup A and B1 isolates were predominant ( Supplementary Table S2 ). Intriguingly, it was already suggested that E. coli strains of the phylogroup B2 might have lost the lsr locus during evolution (Quan and Bentley, 2012; Brito et al., 2013). However, all B2 strains analyzed by Brito and colleagues encoded a functional LuxS synthase and thus should be capable of producing AI-2 (Brito et al., 2013). Following Brito and colleagues, we found that the vast majority of E. coli belonging to the phylogroup B2 lack a full-length lsr locus ( Figure 8 , Supplementary Figure S6A ), while the presence of luxS seems to be remarkably conserved ( Supplementary Figure S6B ). The high prevalence of luxS in E. coli and also over a long time in bacteria (Santiago-Rodriguez et al., 2014) underlines its crucial role as a metabolic enzyme in the activated methyl cycle (Vendeville et al., 2005). The tam gene, which is located downstream of the lsrACDBFG operon, was also widely distributed (> 88%) in the different phylogroups ( Supplementary Figure S6C ). Since E. coli 83972 encodes Tam and its role in AI-2-dependent QS remains unclear, we omitted tam from the complemented lsr locus. Indeed, we observed that tam was upregulated during the transition to the stationary phase (data not shown) but was not differentially expressed in E. coli 83972 attB::lsr as compared to E. coli 83972. Evolutionary studies suggest that the lsr locus was passed on via horizontal gene transfer (Rezzonico et al., 2012), and lateral gene transfer events were also described (Pereira et al., 2009). Thus, it is interesting that the operon was not reintroduced into the phylogroup B2 (Manges and Johnson, 2015). These observations and our data suggest that the loss of AI-2-dependent QS could be an evolutionary advantage for E. coli 83972, which efficiently colonizes the urinary bladder as an extraintestinal body niche. Whether the absence of the lsr determinant in almost all phylogroup B2 isolates investigated so far may be correlated, at least in part, with generally increased fitness properties in extraintestinal niches or superior colonization properties in densely populated niches such as the intestinal tract will require more detailed future analyses of the effect of lsr complementation in a diverse collection of phylogroup B2 strains.

It is imperative that we explore non-antibiotic solutions to prevent UTI, with bacterial interference being a promising avenue for treating symptomatic episodes of uncomplicated cystitis. Our study provides new information on gene regulation in E. coli 83972. The more we understand which bacterial characteristics improve fitness and colonization properties in the urinary bladder, the better we could rationally improve the engineering of E. coli 83972 or predict outcomes in therapeutic colonization. The impact of AI-2-mediated QS for pathogenicity or fitness of ExPEC is not uniform and is probably dependent on the individual strain background. Whereas interference with AI-2-dependent QS resulted in downregulation of ExPEC virulence traits and attenuation in some isolates, increased fitness due to enhanced resistance to oxidative stress was observed in others. It is tempting to speculate that the absence of the lsr operon in the majority of phylogroup B2 isolates may potentially suggest that the loss of AI-2-dependent QS could be associated with a fitness advantage in extraintestinal niches or their reservoir, i.e., the intestinal tract. However, this must be further analyzed in detail in the future. In this study, we investigated the impact of AI-2-mediated QS on fitness traits of asymptomatic bacteriuria E. coli isolate 83972, a strain that has already been used for therapeutic bladder colonization, thereby interfering with bladder infections caused by uropathogens. We found that the reintroduction of AI-2-dependent QS in E. coli 83972, which is optimally adapted to the growth conditions in the urinary bladder, has led to a phenotype that is disadvantageous for efficient and long-term bladder colonization, especially in a competitive situation. Restoration of the lsr determinant in E. coli 83972 resulted in growth retardation, loss of competitiveness and increased susceptibility to oxidative stress, all characteristics relevant to this strain’s colonization ability in the bladder. Thus, our findings indicate that the absence of AI-2-dependent QS in E. coli 83972 may be an advantage during the colonization of the urinary bladder. Further studies on the benefits of the lack of the lsr determinant in a broad spectrum of ExPEC isolates with different phylogenetic backgrounds and genome content are needed to gain deeper insights into the general importance of AI-2-dependent QS for E. coli fitness and pathogenicity in extraintestinal niches and or the colonization of the intestinal tract.