F9 Fimbriae of Uropathogenic Escherichia coli Are Expressed at Low Temperature and Recognise Galβ1-3GlcNAc-Containing Glycans

Uropathogenic Escherichia coli (UPEC) is the leading causative agent of urinary tract infections (UTI) in the developed world. Among the major virulence factors of UPEC, surface expressed adhesins mediate attachment and tissue tropism. UPEC strains typically possess a range of adhesins, with type 1 fimbriae and P fimbriae of the chaperone-usher class the best characterised. We previously identified and characterised F9 as a new chaperone-usher fimbrial type that mediates biofilm formation. However, the regulation and specific role of F9 fimbriae remained to be determined in the context of wild-type clinical UPEC strains. In this study we have assessed the distribution and genetic context of the f9 operon among diverse E. coli lineages and pathotypes and demonstrated that f9 genes are significantly more conserved in a UPEC strain collection in comparison to the well-defined E. coli reference (ECOR) collection. In the prototypic UPEC strain CFT073, the global regulator protein H-NS was identified as a transcriptional repressor of f9 gene expression at 37°C through its ability to bind directly to the f9 promoter region. F9 fimbriae expression was demonstrated at 20°C, representing the first evidence of functional F9 fimbriae expression by wild-type E. coli. Finally, glycan array analysis demonstrated that F9 fimbriae recognise and bind to terminal Galβ1-3GlcNAc structures.


Introduction
Urinary tract infections (UTI) are among the most common infectious diseases of humans and a major cause of morbidity. In the USA, UTI account for approximately $1.6 billion in medical expenditures each year [1]. It is estimated that 40-50% of adult healthy women will experience at least one UTI episode in their lifetime. The recurrence rate of UTI is high and often the infections tend to become chronic with many subsequent episodes. UTIs usually start as cystitis but often evolve to encompass the kidneys and can ultimately result in dissemination into the bloodstream and/or renal failure. Catheter-associated UTIs are also very common and account for 40% of all nosocomial infections. Most patients with an indwelling urinary catheter for thirty days or more develop bacteriuria [2].
Uropathogenic Escherichia coli (UPEC) is the cause of the majority (.80%) of UTIs in humans. UPEC isolates contain numerous virulence factors, which allow for the successful colonisation of the urinary tract. Although no single virulence factor is uniquely definitive of UPEC, the ability to cause symptomatic UTI is enhanced by adhesins (e.g. type 1 and P fimbriae) and toxins (e.g. hemolysin) [3,4]. Adherence to the urinary tract epithelium is the first stage of UTI as it enables bacteria to resist the hydrodynamic forces of urine flow and establish infection. Among the best-described adhesins produced by UPEC are type 1, P, and F1C/S fimbriae of the chaperoneusher (CU) pathway [4].
The CU pathway is a highly conserved secretion system in Gram-negative bacteria that mediates the assembly of hair-like fimbrial polymers on the bacterial cell surface. CU fimbrial biogenesis requires a dedicated periplasmic chaperone and an outer membrane usher protein that functions as an assembly platform of the fimbrial organelle which is primarily composed of a helical array of 500 to 3,000 copies of major subunit protein [5,6]. The receptor-binding adhesin resides at the distal end of the fimbrial organelle and contains a C-terminal domain which connects the adhesin to the terminal major subunit protein sometimes aided by one or more minor subunits, and an Nterminal lectin domain which mediates binding to specific ligands [3]. The genes encoding the various components of CU fimbriae are typically organised in an operon and transcribed as a single polycistronic mRNA molecule [7].
Genomic analysis of the E. coli pan genome has revealed 38 distinct chaperone-usher fimbrial types based on genomic locus position and usher phylogeny [8]. Type 1 and P fimbriae are primary contributors to the colonisation of the urinary tract by UPEC and have been the focus of extensive study (for a review, refer to [9]). Type 1 fimbriae confer binding to a-D-mannosylated proteins such as uroplakins, which are abundant in the uroepithelial lining of the bladder [10]. P fimbriae contribute to UTI by binding to the a-Gal(1-4)b-Gal receptor epitope in the globoseries of glycolipids found in the kidney [11,12]. F1C/S fimbriae also contribute to UTI through their ability to bind to GalNAcb1-4Galb glycolipids and sialyl galactoside glycoproteins present on epithelial cells in the bladder and kidneys [13][14][15].
We previously characterised F9 fimbriae as a new CU fimbriae type in UPEC [16]. F9 fimbriae are part of the c1 fimbrial subclade and are closely related to type 1 and F1C/S fimbriae in genetic organization and structural composition [8,17]. Low levels of expression of the F9 major subunit have been detected in enterohemorrhagic E. coli (EHEC) strain O157:H7 EDL933 and in a UPEC CFT073fim foc null-mutant, however, to date there is no evidence of functional F9 fimbriae expression in any wild-type E. coli strain [16,18]. Cloning and expression of the f9 genes in a recombinant E. coli strain revealed F9 fimbriae mediate strong biofilm formation, however F9 expression did not confer hemagglutination or cellular adherence properties. In this study, we have examined the distribution and conservation of the f9 operon in E. coli, demonstrated the high frequency of f9 genes in extant E. coli strains, and evaluated the conservation of the F9 adhesin lectin domain. Additionally, we have demonstrated that the f9 fimbrial gene cluster is subjected to temperature-dependent repression by the global regulator H-NS. Repression was alleviated at lower temperatures, at which F9 fimbriae mediated significant biofilm formation on abiotic surfaces by wild-type E. coli. In order to study the ligand recognition properties of F9 fimbriae, we utilized a glycan array and identified Galb1-3GlcNAc and lacto-Ntetraose glycans as novel receptor targets for these fimbriae.

Phylogenetic inference, sequence alignment and diversity estimations
The complete genomes of 42 E. coli strains representing the diversity of the species were investigated for presence of the f9 operon using the NCBI BLAST2.2.25+ program [19]. The f9 operon and its genomic context were aligned and visualised using Easyfig [20]. The evolutionary relationship of the 42 E. coli strains was predicted by Multi-Locus Sequence Typing (MLST) of the ,9 kb concatenated nucleotide sequences of 7 housekeeping genes (adk, fumC, gyrB, icd, mdh, purA, recA) as previously described [21]. Sequences were aligned in MEGA5 using the ClustalW DNA weight matrix under default settings. The Neighbour-Joining method of MEGA5 was used to infer the evolutionary history, with distances computed by the Jukes-Cantor method. The resulting phylogenetic tree was tested by a bootstrap test of 1000 replicates and visualised in iTOL [22] as a rooted phylogram. Diversity of the F9 adhesin protein was estimated with MEGA5. The mature adhesin (280 aa) and adhesin lectin domain (160 aa) sequences were aligned with ClustalW using the BLOSUM protein weight matrix under default settings; diversity was calculated using the Poisson model with a bootstrap confidence test for standard error measurements (1000 replicates).

Bacterial strains, plasmids and culture conditions
Strains and plasmids used in this study are listed in Table 1. E. coli CFT073 was isolated from the blood and urine from a woman with acute pyelonephritis [23]. E. coli strains were routinely cultured at 37uC on solid or in liquid lysogeny broth (LB) medium [24] or liquid M9 minimal medium (42 mM Na 2 HPO 4 , 22 mM KH 2 PO 4 , 9 mM NaCl, 18 mM NH 4 Cl, 1 mM MgSO 4 , 0.1 mM CaCl 2 and 0.2% (w/v) glucose). Where appropriate, media were supplemented with ampicillin (100 mg ml 21 ), kanamycin (100 mg ml 21 ) or chloramphenicol (25 mg ml 21 ). To induce expression of F9 fimbriae from plasmid pF9, culture media were supplemented with 0.2% (w/v) arabinose. Plasmid transformations into E. coli CFT073 were mediated by electroporation.

DNA manipulations and genetic techniques
Plasmid DNA was isolated using the QIAprep Spin Miniprep kit (Qiagen). Chromosomal DNA was purified using the GenomicPrep cell and tissue DNA isolation kit (GE Healthcare Life Sciences). PCR was performed using Taq DNA polymerase according to manufacturer's instructions (Roche). Restriction endonucleases were used according to the manufacturer's specifications (New England Biolabs). Oligonucleotide primers used in this study were purchased from Sigma-Aldrich and are listed in Table S1. For sequencing, PCR products were amplified using the BigDye Terminator v3.1 Cycle DNA Sequencing Kit according to manufacturer's specifications (AB SCIEX), and analysed subsequently by the Australian Equine Genome Research Centre.

Construction of plasmid pDW11
To generate a Green Fluorescent Protein (GFP) expressing plasmid compatible with pF9, the gfp gene from plasmid pKEN2 [25] was amplified with primers 2319 and 2320 containing 59 BamHI and SalI sites, respectively (Table S1). The PCR product was digested by BamHI and SalI and directionally cloned into the corresponding sites of cloning vector pACYC184. Plasmid transformed E. coli strains were screened for GFP expression by fluorescence microscopy.
Construction of CFT073 gene deletion mutants E. coli CFT073 gene deletion mutants were constructed using the l-Red mediated homologous recombination system as previously described [26]. Briefly, the FRT-flanked kanamycin resistance gene from pKD4 was amplified using primers containing 59 50 bp regions homologous to the start and end sequence of the gene(s) to be deleted. The resulting approximately 1.6 kb PCR products were introduced by electroporation into appropriate strains expressing l-Red recombinase from pKD46. Kanamycin resistant colonies were analysed by PCR and DNA sequencing to confirm deletion of the relevant gene. For the construction of E. coli CFT073 isogenic null-mutants, the kanamycin gene was removed using the pCP20 FLP-FRT site-specific recombination system, allowing for successive rounds of mutagenesis [26].

Protein immunoblotting
Rabbit immune serum against an F9 fimbriae over-expressing E. coli strain was previously generated [16]. Sera were absorbed against cell lysates of f9 negative stains and f9 null-mutants. For western blot analysis, bacterial EDTA heat-induced outer membrane vesicles (OMVs) were generated to enrich for the outer membrane-associated protein fraction, using a previously described method [27] with several modifications. Briefly, 50 ml LB or M9 medium was inoculated with 100 ml pre-culture (grown in the same medium) and incubated for 18 h at 37uC, 28uC or 20uC 250 rpm. Cells were harvested at 10,0006g for 10 min at 4uC and washed in 25 ml 4uC PBS. The bacterial pellet was resuspended in 1 ml EDTA buffer (0.05 M Na 2 HPO 4 , 0.15 M NaCl, 0.01 M ethylenediaminetetraacetic acid (EDTA), pH 7.4) and incubated 30 min at 56uC, statically. Cells were centrifuged at 10,0006g for 10 min at 4uC and the supernatant was filtered using a 0.22 mm PVDF low protein binding filter (Millipore). Trichloroacetic acid (TCA) was added to a final concentration of 20% (w/ v) to precipitate proteins overnight at 4uC. Protein suspensions were separated according to electrophoretic mobility using SDS-PAGE and transferred to a polyvinylidene difluoride (PVDF) membrane, which was subsequently incubated in 1:500 rabbit polyclonal absorbed anti-F9 primary sera, followed by 1:10,000 goat anti-rabbit immunoglobulin G-alkaline phosphatase-conjugated secondary antibody (Sigma-Aldrich). Signal development was performed using the substrate 5-bromo-4-chloro-3-indolylphosphate-nitroblue tetrazolium (BCIP/NBT; Sigma-Aldrich).

Rapid amplification of cDNA ends (RACE)
The transcription start site of the f9 operon was determined using the 59 RACE System v2.0 (Invitrogen) [28]. Experiments were performed according to manufacturer's specifications except for the modifications listed below. Three gene specific primers were used for this assay: 4235, 4236 and 4237 (Table S1). To verify that the first nucleotide of the mRNA was a guanine, the cDNA was dA-tailed with a dATP substitution. PCR amplification of dA-tailed cDNA was performed using the (dT)17-adaptor primer 4296 (Table S1). Amplified cDNA was sequenced by the Australian Equine Genome Research Centre.

Electrophoretic mobility shift assay
Gel shift assays were performed as previously described [29]. Briefly, a DNA mixture containing an equimolar ratio of the 251 bp PCR amplified f9 promoter region and TaqI-SspI digested pBR322 was incubated with native purified H-NS protein in 30 ml H-NS binding buffer (40 mM HEPES pH 8, 60 mM potassium glutamate, 8 mM magnesium aspartate, 5 mM dithiothreitol, 10% glycerol, 0.1% octylphenoxypolyethoxyethanol, 0.1 mg/ml bovine serum albumin) for 15 min at room temperature. DNA fragments and DNA-protein complexes were resolved by gel electrophoresis (0.56 Tris/Borate/EDTA buffer, 3% agarose MS gel, ran at 50 V, 4uC), stained with ethidium bromide and visualised by ultraviolet illumination.

Microtitre plate biofilm formation assay
Bacterial biofilm formation was assessed on sterile non-coated 96-well polyvinyl chloride (PVC) microtitre plates (BD Falcon) as previously described [30]. Briefly, cells were cultured with aeration at various temperatures for 24 hours in 150 ml M9 medium containing 0.2% (w/v) glucose. After incubation, cells were washed, stained with 0.1% crystal violet for 30 min at 4uC, and washed three additional times. Bound bacterial cells were quantified by adding ethanol-acetone (80:20 v/v) and measurement of the dissolved crystal violet at an optical density of 595 nm.

F9 Immunogold labelling and electron microscopy
Cells for immunogold labelling and transmission electron microscopy (TEM) were prepared from liquid cultures grown overnight at 20uC in M9 minimal medium. A glow-discharged carbon-coated Formvar copper grid was placed on a drop of the bacterial suspension for 1 min to allow the cells to adsorb. Grids were washed twice on drops of water (1 min), and blocked for 30 min in blocking buffer (PBS containing 0.2% BSA, 0.2% fish skin gelatin, 20 mM glycine). Samples were exposed to 1:25 anti-F9 rabbit immune serum for 30 min and washed four times in blocking buffer (5 min) before incubation with Protein A-gold conjugate (10 nm diameter, diluted 1:60 in blocking buffer) for 30 min and four washes (5 min) in PBS. Cells were fixed with 4% paraformaldehyde in PBS (5 min) and grids were washed four times (2 min) in sterile ultrapure water before examination under a JEOL 1010 TEM operated at 80 kV. Images were captured using an analySIS Megaview III digital camera.

Glycan array analysis
Glycan array slides and whole-cell binding assays were essentially performed as previously described [31,32]. Glycan arrays comprised 120 unique carbohydrates (Table S2) printed on super epoxy slides (Arrayit). All array experiments consisted of a minimum of three independent biological repeats. E. coli strains MS428 (pF9, pDW11) and MS428 (pBAD, pDW11) were cultured overnight at 37uC with gentle agitation in M9 minimal medium containing ampicillin (100 mg ml 21 ), chloramphenicol (25 mg ml 21 ) and 0.2% arabinose. Post-incubation, cells were diluted to an OD of 0.6, representing approximately 1610 7 CFU ml 21 . A volume of 125 mL of cells was hybridised to a pre-blocked (0.1% BSA in PBS with 2 mM MgCl 2 and CaCl 2 , 5 mins) glycan array for 20-30 mins at room temperature in the dark. Glycan array slides were placed in a 50 mL tube and washed in filter sterilized buffer 1 (PBS with 2 mM MgCl 2 and CaCl 2 ) for 5 mins, buffer 2 (buffer 1 with 0.01% tween-20) for 2 mins, and finally rinsed in fresh buffer 1. Cells were fixed for 10 mins in PBS with 10% formaldehyde and dried by centrifugation for 5 mins at 2006g. The array slide was scanned using a ProScan Array microarray scanner (Perkin Elmer) using a 488 nm argon laser. Images were attained and analysed using the ScanArray Express software package (Perkin Elmer). To determine minimum binding concentration, glycans were printed in serial dilutions from 5 mM to 5 fM on an array slide. For competition assays, cells were pretreated with the glycan of interest for 15 mins prior to hybridisation on the slide.

Statistical analyses
The frequency of intact f9 operons in different E. coli strain collections and between different E. coli phylogenetic groups was compared using Fisher's exact test with a two-tailed P value. Biofilm formation was compared between f9 encoding strains and their isogenic f9 null mutant using a two-tailed t test. P values , 0.05 were considered significant. For glycan array analysis, binding was classified as RFU (relative fluorescence units) above average background (defined as background mean plus 3 standard deviations) and was tested for statistical significance using a twotailed t test with a P value ,0.001.

Results
Genetic organisation and distribution of f9 fimbrial operons in E. coli In order to investigate the distribution and genetic location of the f9 operon in E. coli, we examined the genome sequence of 42 diverse E. coli strains available on the NCBI database, including representatives of all E. coli lineages and various pathotypes ( Table 2). The E. coli species exhibits extensive genetic substructure and can be divided into 5 major monophyletic clades (phylogroup A, B1, B2, D and E) [21]. To evaluate the conservation and evolutionary history of F9 fimbriae among E. coli phylogroups, a phylogenetic tree based on multi-locus sequence typing (MLST) of 7 concatenated housekeeping genes (,9 kb) was constructed and combined with f9 genomic context alignments ( Figure 1). The f9 operon consists of six structural genes, encoding, from 59 to 39: the major subunit, chaperone, usher, two minor subunits and an adhesin. f9 operons containing deletions, truncations and/or insertion elements were considered disrupted. Comparative genomic analysis revealed that all strains possessed at least part of the F9 encoding DNA sequences. In 60% (25/42) of E. coli strains the f9 operon appeared intact ( Figure 1 and Table 2). Among E. coli phylogenetic groups, the f9 operon was conserved in the majority of B1 and E strains, and to a lesser extent in B2 and D strains. The f9 operon was disrupted in all phylogroup A strains. In a pathotype context, the intact f9 operon was highly prevalent in intestinal pathogenic E. coli, including adherent-invasive E. coli (AIEC; 3/3), enteroaggregative E. coli (EAEC; 2/2), enteropathogenic E. coli (EPEC; 3/3) and enterohemorrhagic E. coli (EHEC; 7/8), but not in enterotoxigenic E. coli (ETEC; 0/2) ( Table 2). The f9 encoding genes were not detected in genome sequences from other bacterial genera available in the NCBI database (except for Shigella, a subgenus of Escherichia).
The f9 operon is flanked by the highly conserved ydeP gene involved in acid resistance (downstream) and the relatively wellconserved hipBA cell persistence gene cluster (upstream). The immediate 59 region is variable, and contains a range of different insertions and/or deletions, including three hypothetical transcriptional regulators (Figure 1). Currently, there is no evidence linking these putative regulators to the transcriptional control of f9 genes.

Prevalence of f9 genes in E. coli
Based on the genomic analysis described above, primers were designed in conserved regions of the f9 gene cluster to screen for the major subunit, usher and adhesin genes in two large E. coli strain collections. These included fifty-one UPEC isolates collected from patients presenting with urosepsis at the Princess Alexandra Hospital (Brisbane, Australia) as well as seventy-two strains of the well-defined and diverse ECOR reference collection. In the UPEC collection, 80% (41/51) of strains screened positive by PCR for all three f9 genes (Figure 2A). A further 16% (8/51) screened positive for at least one f9 gene, while 4% (2/51) of strains screened negative for all genes. In the ECOR collection, 61% (44/72) of strains screened positive for all three f9 genes, 29% (21/72) screened positive for at least one f9 gene and 11% (8/72) did not yield any positive PCR results. F9 operon prevalence (as judged by screening positive for the major subunit, usher and adhesin genes)  [85] was significantly higher (P,0.05) in UPEC isolates compared to the ECOR collection ( Figure 2A). PCR data from the two E. coli collections were merged to evaluate f9 prevalence relative to strain phylogenetic group ( Figure 2B). Consistent with the genomesequenced strains, the frequency of intact f9 operons in phylogroup A strains was significantly lower in comparison to strains belonging to other phylogroups (P,0.05). The F9 operon was detected in 100% of phylogroup B1 strains and the majority (.70%) of strains from phylogroup B2, D or E.

Transcription of the f9 operon is repressed by H-NS
We previously detected very weak expression of F9 fimbriae in UPEC strain CFT073fim foc cultured at 37uC [16]. However, to date there have been no reports of functional F9 fimbriae expression in wild-type UPEC. To investigate the genetic basis of f9 gene regulation in UPEC, we employed a previously described panel of CFT073 isogenic single gene deletion mutants lacking a selection of defined/putative regulatory genes (c0421  [33]. Expression of the F9 fimbrial major subunit protein was assessed by western blot analysis of CFT073 wild-type and mutant strains employing an F9 specific antiserum. In this experiment, strong expression of the F9 major subunit protein was only detected in the CFT073hns mutant strain, but not in any of the other regulator mutants following growth in LB broth at 37uC (Fig. 3A). To confirm these results, the hns mutant strain (referred to as CFT073hns) was complemented with the H-NS expressing plasmid pH-NS. No detectable F9 major subunit protein was detected in CFT073hns(pH-NS) ( Figure 3B). Additionally, the strong F9 major subunit signal was absent in a CFT073f9 hns double mutant. Interestingly, a faint band similar in size to the F9 major subunit was observed in CFT073f9 hns, suggesting some non-specific cross reactivity of the F9 antiserum with a similar sized protein. Since H-NS negatively controls the expression of various distinct fimbrial operons, this observation could be the result of alleviation of repression of an F9 related fimbrial type [34]. We addressed this by constructing a mutant deleted for gene clusters encoding type 1, F1C, P1 and P2 fimbriae (referred to as CFT073D4), and a CFT073D4 strain deleted for the f9 genes (CFT073D4 f9). Indeed, mutation of the hns gene in CFT073D4 and CFT073D4 f9 resulted in the complete loss of this cross-reacting band ( Figure 3B). Combined, these results demonstrate that H-NS represses the expression of F9 fimbriae in CFT073.

H-NS binds to the promoter region of the f9 operon
In order to determine whether H-NS influences f9 gene transcription by directly binding to the promoter region, the f9 promoter was characterised using 59-RACE and investigated for H-NS interaction by electrophoretic mobility shift assays. The f9 transcription start site was identified as a guanine residue, 251 nucleotides upstream of the f9 major subunit gene start codon ( Figure 4A). The transcription start site was preceded by a strong 210 promoter consensus sequence (CATAAT) and a moderate 2 35 promoter consensus sequence (TAGTCG) with an 18 bp spacer region. In silico analysis of the promoter region discerned a ribosomal binding site (RBS) directly upstream of the translation initiation site, and identified six putative H-NS binding motifs at positions 2111, 2103, +8, +14, +57 and +89 ( Figure 4A) [35]. To investigate f9 promoter/H-NS interactions, the 251 bp promoter region was amplified by PCR and mixed with TaqI-SspI-digested pBR322 DNA (containing the H-NS recognised bla promoter). The DNA mixture was incubated with increasing concentrations of purified H-NS protein and analysed by mobility shift electrophoresis. The f9 promoter region and the positive control bla-promoter fragment were equally impeded in gel migration in the presence of increasing concentrations of H-NS ( Figure 4B). In contrast, the mobility of pBR322 fragments lacking the blapromoter sequence was not altered in the presence of H-NS. These results demonstrate that H-NS binds to the f9 promoter region.

Expression of F9 fimbriae in UPEC CFT073 is temperaturedependent
The global regulator H-NS modulates the expression of a large subset of genes in response to external stimuli such as temperature [36,37]. To evaluate whether temperature had an effect on the expression of F9 fimbriae, CFT073 and the isogenic f9 null mutant were cultured at various temperatures and examined by western blot analysis employing an F9 specific antiserum. No protein bands were detected when CFT073 was cultured at 37uC or 28uC, but at 20uC an 18.3 kDa band corresponding to the mature F9 major subunit was observed ( Figure 5). This band was not detected in the CFT073f9 null-mutant at all temperatures examined, confirming the identity of the band as the F9 major subunit protein ( Figure 5). To strengthen these findings we also examined F9 fimbriae expression on the cell surface by immunogold electron microscopy. We detected F9 fimbriae on the surface of CFT073D4 but not CFT073D4 f9 following culture at 20uC (Fig. 5B and 5C). These data represent the first evidence of F9 fimbrial expression by UPEC, and based on the temperature expression profile suggest a role for F9 fimbriae outside the mammalian host.

F9 fimbriae mediate biofilm formation in UPEC strain CFT073
We previously demonstrated that F9 fimbriae mediate a strong biofilm on abiotic surfaces using a plasmid-based system in a recombinant E. coli strain [16]. To determine whether F9 fimbriae expressed by wild-type UPEC are involved in biofilm formation, we investigated CFT073 and CFT073f9 for biofilm formation at 20uC using a microtitre plate assay. Consistent with our F9 fimbrial expression findings using western blot analysis and immunogold-TEM, the expression of F9 fimbriae at 20uC by CFT073 correlated with significant biofilm formation compared to CFT073f9 under these growth conditions (P,0.001; Figure 6). Complementation of CFT073f9 with the F9 fimbrial expression plasmid pF9 restored the strong biofilm phenotype. These data confirm that F9 fimbriae promote significant biofilm growth on abiotic surfaces by wild-type CFT073 at 20uC.
The F9 fimbrial adhesin is highly conserved and displays receptor specificity to Galb1-3GlcNAc terminating glycans The predicted F9 adhesin is encoded by the last gene in the f9 operon and contains a characteristic two-domain structure comprising a C-terminal fimbrial integration domain and an Nterminal receptor-binding region. Comparison of the amino acid sequence of the full-length F9 adhesin among the 25 E. coli strains that contained an intact f9 operon revealed a high degree of conservation, with a mean diversity of 0.01360.004 amino acid The above analysis demonstrates that the F9 adhesin sequence is highly conserved, and indicates that the F9 adhesin from CFT073 can be used to define the overall receptor-binding characteristics of F9 fimbriae. We therefore employed a glycan array in combination with a F9 over-expressing E. coli strain labelled with GFP (MS428[pF9, pDW11]) to evaluate the binding specificity of F9 fimbriae to different carbohydrates. In this assay, F9 fimbriae mediated specific binding to Galb1-3GlcNAc terminating structures, including lacto-N-tetraose (Galb1-3GlcNAcb1-3Galb1-4Glc), globotriose (Gala1-4galb1-4Glc) and the globotriose terminal disaccharide (GalNAcb1-3Gal) (P,0.05). The presence of fucose in Galb1-3GlcNAc glycans eliminated or reduced affinity by at least 100-fold (data not shown). Of the glycans that were bound by F9 fimbriae, lacto-N-tetraose displayed the highest affinity. A glycan competition analysis using 50 mM free lacto-N-tetraose resulted in no observable F9 fimbriaemediated binding to any of the glycans on the array. Taken together, these data provide the first evidence for Galb1-3GlcNAc glycans as specific receptors for F9 fimbriae, and identify lacto-Ntetraose as a high affinity glycan.

Discussion
Bacterial adhesins mediate attachment to host tissues and abiotic surfaces and provide the first step in colonisation and biofilm formation. Despite the large repertoire of CU fimbriae encoded by UPEC [8], there are only a few well-studied examples of fimbriae that are directly associated with pathogenesis or mediate tissue tropism. Many UPEC fimbriae are cryptic in nature and have not been thoroughly characterised. We previously described F9 fimbriae in UPEC as a functional CU fimbrial type promoting formation of E. coli biofilms [16] and have recently demonstrated that they are closely related to the type 1 and F1C/S fimbriae [38], which are both involved in colonisation of the human urinary tract [3]. In this study, the distribution and conservation of F9 fimbriae in diverse E. coli lineages was investigated and evaluated in an evolutionary and pathotype associated context. Evolutionary diversity analysis of the F9 adhesin sequence revealed a high conservation of the receptor recognising lectin domain. Furthermore, H-NS was identified as a temperature dependent negative regulator of F9 expression by binding directly to the f9 promoter region. F9 fimbriae were expressed by CFT073 at 20uC and mediated significant biofilm formation at this temperature. This is the first report of functional F9 expression in wild-type E. coli, and provides the first evidence that F9 fimbriae specifically recognise Galb1-3GlcNAc and lacto-N-tetraose glycans.
E. coli population genetics have identified five major monophyletic clades (phylogroups A, B1, B2, D and E) [21]. Despite the high frequency of f9 DNA sequences in the E. coli species, the conservation of the f9 operon between E. coli phylogenetic groups varied significantly. A genomic comparison of the f9 operon from 42 E. coli genomes showed that intact f9 operons were particularly prevalent in phylogenetic group B1 and E, and to a lesser degree in phylogroups B2 and D. In strains from phylogenetic group A, all f9 operons were disrupted. Variation was also observed among E. coli pathotypes, with the f9 fimbrial genes particularly conserved in intestinal pathogenic isolates representing AIEC, EAEC, EPEC and EHEC, but not ETEC, suggesting a potential role in the pathogenic lifestyle of these bacteria. Indeed, signature-tagged mutagenesis screens using EHEC strains of serotype O157:H7 and O26:Hhave previously identified insertion mutants in the f9 gene cluster that were significantly impaired for intestinal colonisation in young calves [39,40]. The f9 operon was moderately conserved in UPEC genomes. A PCR screen of the 51 isolates in our UPEC collection suggested that the f9 operon is intact in 80% of the strains, significantly higher than the 61% prevalence of intact f9 operons in the 72 strains of the diverse and well defined ECOR reference collection. In a phylogenetic context, the results from the f9 gene prevalence screen of the two collections were consistent with the genomic data. F9 encoding sequences were not found in any other species (except for Shigella, a subgenus of Escherichia), indicating this fimbrial operon is unique to E. coli. The ubiquity of f9 genes in extant E. coli strains suggests that the f9 operon is ancient and was present in the E. coli common ancestor. H-NS is a histone-like DNA-binding protein that shows affinity for A-T rich and bent nucleation sites on DNA [41]. In this study, several lines of evidence demonstrated a role for H-NS in the regulation of F9 fimbrial expression. In a CFT073 hns mutant background, F9 expression was de-repressed, and this effect could be reversed through the introduction of a plasmid containing the hns gene. In addition, H-NS bound to a 251 bp DNA fragment containing the mapped f9 promoter region and a positive control bla-promoter fragment with equal affinity. H-NS has been shown to repress multiple other virulence-associated genes in UPEC, including genes encoding alpha-hemolysin, iron uptake systems, fimbriae and autotransporter proteins [33,[41][42][43]. In E. coli K-12, several cryptic chaperone-usher fimbrial genes are also repressed by H-NS [34]. The data presented here is the first direct demonstration that H-NS represses F9 fimbriae, and is consistent with a role for H-NS in the regulation of multiple UPEC virulence factors.
F9 fimbriae expression by UPEC CFT073 also displayed a temperature-dependent profile. At 20uC, we detected expression of the F9 major fimbrial subunit protein by western blot and F9  fimbriae structural organelles on the cell surface by immunogold electron microscopy. This F9 expression profile correlated with a strong biofilm phenotype for CFT073 grown under these conditions. Previous studies in EHEC O157:H7 using a chromosomally integrated lacZ-f9 promoter fusion have also suggested F9 temperature-dependent regulation, with stronger activity of the f9 promoter observed at 28uC versus 37uC [18]. In addition, it has been shown that the expression of the F9 major subunit is increased at 28uC compared to 37uC, however expression levels were too low to detect F9 fimbriae by immunofluorescence [18]. While F9 fimbrial expression in a recombinant E. coli K-12 strain promoted enhanced binding to bovine rectal epithelial cells, there were no significant differences in colonization of the terminal rectum of cattle by a wild-type and F9 mutant strain, suggesting that F9 fimbriae are not responsible for EHEC O157 rectal tropism in cattle [18]. Combined, these data suggest that F9 fimbriae contribute to the E. coli lifestyle outside the mammalian host, potentially involving colonisation of epidermal surfaces and persistence in the environment through biofilm formation.
Two other types of E. coli adhesins that are expressed strongly at 20uC have also been described, namely Mat (or ECP) fimbriae and curli fibres [44,45]. Mat fimbriae mediate biofilm formation by neonatal meningitis E. coli and UPEC at low temperature [46,47] The expression of Mat fimbriae has also been observed more generally in E. coli following cultivation in DMEM, suggesting that temperature-mediated regulation is linked to specific growth conditions [44]. Curli are also strongly expressed at 20uC and associated with biofilm formation [48], however to the best of our knowledge CFT073 has not been shown to produce curli. In our experiments, although the reduction in biofilm formation at 20uC between wild-type CFT073 and the CFT073f9 mutant was significant, CFT073f9 still formed a reasonable biofilm ( Figure 6). This suggests that CFT073 produces other biofilm formation mechanisms under these conditions, which may include Mat fimbriae. It remains to be determined whether Mat fimbriae are produced by CFT073 at 20uC under the conditions used in our experiments, whether F9 and Mat fimbriae can be co-expressed at 20uC, and if there are additional layers of regulatory control in E. coli strains that have the capacity to express both of these fimbriae.
The sequence of the F9 adhesin lectin domain was shown to be highly conserved in E. coli strains from different phylogenetic lineages. In order to examine the receptor binding specificity of F9 fimbriae, a glycan array containing 120 structures was utilized. The glycans on the array represented host cell surface glycoconjugates including terminal galactose, mannose, fucosylated and sialylated structures and glycosaminoglycans [31,32]. These glycans mimic those found on mucosal surfaces, the extracellular matrix, blood antigens and cells of the immune system. The analysis revealed F9 fimbriae bind to Galb1-3GlcNAc containing glycans, with lacto-N-tetraose identified as a high affinity glycan. Epithelial cells of the human urinary tract and kidney are rich in the globoseries glycolipids [49], whereas lacto-N-tetraose is a common oligosaccharide found in human milk [50]. In addition, lacto-N-tetraose is the oligosaccharide moiety of the lactotetraosylceramide glycosphingolipid receptor present in human gastric epithelium, which is recognised and bound to by Helicobacter pylori [51]. Given that many H-NS repressed genes encode virulence factors associated with human infection, it is possible that F9 fimbriae expression in the human host could also contribute to colonisation. In this respect, we were unable to demonstrate binding of a recombinant E. coli strain over-expressing F9 fimbriae to human exfoliated urothelial cells, human T24 bladder epithelial cells, human A498 kidney epithelial cells, human Caco-2 intestinal epithelial cells, or human type A red blood cells (data not shown). Thus, the Galb1-3GlcNAc glycan-containing target cells bound to by F9 fimbriae in the mammalian host remain to be identified.
In conclusion, we have shown that the f9 fimbriae genes are common to many different E. coli lineages and pathotypes and are regulated by H-NS and temperature. F9 fimbriae bind with high affinity to Galb1-3GlcNAc glycans, including lacto-N-tetraose. Finally, UPEC CFT073 expresses F9 fimbriae at 20uC which (A) Western blot analysis of wild-type CFT073 and isogenic f9 nullmutants cultured at various temperatures. The F9 specific antiserum reacts with the F9 mature major subunit protein (,18.3 kDa) in overexpressing strain CFT073f9 (pF9). No expression is observed in wild-type CFT073 when cultured at 37uC or 28uC. F9 expression is observed in CFT073 at 20uC, and lost again in isogenic null-mutant CFT073f9, illustrating the temperature dependent regulation of F9 fimbriae in UPEC. The mature F9 major subunit (MS) is indicated by an arrow. The 22 kDa higher molecular weight cross-reacting band detected from CFT073 following growth at 20uC is consistent with the size of the unprocessed F9 major subunit protein. TEM micrograph of CFT073D4 (B) and CFT073D4 f9 (C) labelled with immunogold anti-F9 serum after growth at 20uC. Scale bars (500 nm). doi:10.1371/journal.pone.0093177.g005 correlates with strong biofilm formation on abiotic surfaces. Further characterisation of F9 fimbriae is now required to identify its potential role in the colonisation of specific biotic surfaces.   Figure 6. F9 fimbriae mediate biofilm formation in UPEC CFT073 at 206C. PVC microtitre plate biofilm formation assay of CFT073 and isogenic f9 null mutants cultured in M9 medium at 20uC. The optical density (OD) at 595 nm (mean 6 SD) is an indication of biofilm formation. CFT073 formed a significantly stronger biofilm in comparison to the isogenic f9 null mutant. The strong biofilm phenotype was restored by complementation of the f9 deletion mutant with the F9 expression vector pF9 but not with the empty vector pBAD. * Statistically significant (P, 0.001). doi:10.1371/journal.pone.0093177.g006