A Newly Identified Group of P-like (PL) Fimbria Genes from Extraintestinal Pathogenic Escherichia coli (ExPEC) Encode Distinct Adhesin Subunits and Mediate Adherence to Host Cells

Previously, we reported that ExPEC strain QT598, originally isolated from an infected turkey, contains a large ColV-type virulence plasmid, pEC598, that encodes a novel autotransporter protein, the serine-protease hemagglutinin autotransporter, Sha (27). The region adjacent to the sha gene on pEC598 contains a fimbrial gene cluster (Fig. 1). Due to the close protein identity and genetic organization with P fimbriae (see below), we have named these genes plf (for P-like fimbriae) and called these adhesins PL fimbriae. The plf gene cluster on the pEC598 plasmid is inserted beside a site-specific integrase gene located adjacent to the RepFIB region, which is one of the known replicons of IncF plasmids (Fig. 1). The RepFIB region and intM integrase genes are also present in most IncFII plasmids and are also commonly flanked by other predicted virulence genes, such as mig-14, hlyF, and ompT, in other APEC virulence plasmids, suggesting this conserved region may have led to acquisition of different genes through integration/recombination. Interestingly, this region of F and related plasmids is considered a “hot spot” for insertion of diverse virulence and antibiotic resistance genes, which have been termed cargo genes (Fig. 1) (28).

Fimbriae have been classed into specific groups by a variety of criteria, including comparison of usher, chaperone, and major fimbrial subunits (2, 29). Fimbria classification using the usher-encoding protein sequences has placed P fimbriae within the π fimbrial clade (2). Phylogenetic analysis using the usher proteins indicated that the PL fimbriae also belong to the π fimbrial clade and cluster with Pix, Sfp, and Pap fimbriae (Fig. 2A). Phylogenetic comparison based on the chaperone proteins also indicated PL fimbriae were most closely related to P, Pix, and Sfp fimbriae (see Fig. S1 in the supplemental material).

Girardeau et al. (29) also classified fimbriae based on amino acid sequence motifs within the major subunit proteins. The subfamily Ic (PapA-like) group, which included PapA variants, SmfA (Serratia marcescens), PmpA (Proteus mirabilis), and MrpA (Proteus mirabilis), also includes PixA, SfpA, and PlfA major subunit proteins that share a conserved sequence signature motif in segment S1 of the fimbrial subunits: GxG[KT]V[TS]FxG[TS]V[VI]DAP (Fig. 2B).

The PL fimbriae (plf) gene cluster contains 10 genes predicted to encode one regulatory protein and 9 structural/assembly proteins that share identity to equivalent proteins present in the pap gene cluster (Fig. 3). A predicted regulatory protein, PlfB, shares identity with members of the PapB regulatory family, which includes PapB (P fimbriae), PixB (Pix fimbriae), FocB/SfaB (F1C/S fimbriae), AfaA (Afa class III adhesin), Daa (F1845 fimbriae), and FanA/FanB (K99 pili) regulatory proteins (30). The highest identity was to PixB (57% identity, 76% similarity), followed by FanB (47% identity, 69% similarity) and PapB (45% identity, 67% similarity). No equivalent of the PapI regulatory gene was present. Some Plf proteins show higher identity to other pap-related fimbrial gene clusters, specifically from Pix fimbriae, identified in some E. coli urinary tract infection isolates (23, 24), and the plasmid-encoded Sfp fimbriae, present in sorbitol-fermenting diarrheagenic E. coli O157:H7 strains (25, 26) (Fig. 3). Despite demonstrating higher identity to certain proteins from these other fimbriae, only Plf demonstrates a complete set of structural/assembly protein genes equivalent to each in the pap gene cluster, as Pix and Sfp fimbriae both lack PapE or PapK protein paralogs, which are known to code for a fibrillum subunit and adaptor protein, respectively (Fig. 3). The gene products showing the greatest degree of diversity among these fimbrial protein paralogs were the G adhesins, which exhibited less than 30% identity. Taken together, the PL fimbrial system is highly similar to Pix and Sfp fimbrial systems, but shares a genetic organization more akin to that of P fimbriae, as it includes the PapK and PapE paralogous proteins PlfK and PlfE, which are predicted to be part of a thin fibrillum structure.

To best identify potential fimbrial genes that are very closely related to the Plf system of strain QT598, alignment searches of the NCBI database were done against the predicted adhesin-encoding gene product PlfG using a cutoff of >90% amino acid identity. The search revealed 105 samples (104 E. coli strains and one Escherichia albertii strain) containing a pflG allele with high identity to PflG_QT598. Interestingly, among the sample sources, a majority were isolated from avian species as well as clinical isolates from urine or extraintestinal infections in humans (Table S1). However, some samples were also from a variety of livestock, healthy human fecal samples, exotic zoo animal fecal samples, and environmental sources. BLAST analyses against PlfG_QT598 also identified a series of proteins demonstrating from 44% to 77% identity to PlfG that were all associated with fimbrial gene clusters belonging to the Plf family, since these fimbrial gene clusters shared the same genetic organization as the plf gene cluster and had highly conserved identity (>94%) to the PlfD_QT598 gene products (data not shown).

The initial similarity search of the NCBI database was based on specifically identifying plf gene clusters with PlfG adhesin similar to that of strain QT598. In order to determine the general prevalence of PL fimbrial genes distributed among different isolates, a protein BLAST search specific to the predicted PlfK protein was used. We chose this predicted open reading frame for the search as it is not present in the other closely related π fimbria Pix or Sfp and since identity to the PapK protein ortholog is only 54% (Fig. 3). The search (conducted in November 2021) revealed 686 distinct isolates/strain entries containing plf sequences, most of which contained complete plf gene clusters. This included 171 entries from human infections (mainly urinary isolates or blood infections) and 66 human fecal source isolates. As well, many of the entries were associated with infections in dogs, samples from turkeys, chickens, and other avian species, and a variety of environmental sources. Details of the sources of entries containing plf sequences are presented in the supplemental material (Table S2). Taken together, sequence survey analyses indicate that the PL fimbria-encoding sequences are present in E. coli strains associated with extraintestinal infections as well as a variety of other animal and environmental sources.

Based on sequences in the database and identification of entries containing enough sequence data to span the length of the fimbrial gene clusters, a phylogenetic analysis of distinct protein entries for different PlfG adhesins was determined. In all, 21 protein entries sharing identity with PlfG were identified (Fig. 4). Analysis determined 5 distinct clades of the PlfG adhesins, including a group (class V) specific to some Cronobacter spp. The number of individual entries from the sequence database indicated that PlfG class I and PlfG class II families were predominant, whereas PlfG classes III to V were represented by only a few individual strains in the sequence database (Fig. 4). All of the plf gene clusters identified from E. coli strains regardless of G adhesin class were inserted adjacent to a site-specific integrase and RepFIB region (data not shown), suggesting that these fimbrial systems are likely to be plasmid encoded. Taken together, these results suggest that plf gene clusters are present in numerous E. coli strains and that the G adhesins of these fimbriae have diversified into distinct alleles.

Phylogenetic analysis based on the comparison of the PlfG adhesin sequences in the database demonstrated that two main classes of PlfG adhesins, class I and class II, are predominant in sampled genomes. Specific BLAST comparisons of the PlfG class II adhesin from E. coli strain QT598 with a representative gene encoding the class I adhesin from strain UMEA-3703-1, showed 45% identity and 65% similarity. This sequence divergence is similar to the difference between P fimbriae class I and class II G adhesins (46% identity, 64% similarity). As such, and since these two PlfG classes are the most common in the database, we cloned both of them for further investigation.

To demonstrate that the plf-containing clones produced fimbriae, the plasmids encoding plf genes were transformed into the afimbriated E. coli K-12 strain ORN172. Transmission electron microscopy (TEM) demonstrated that both plf_QT598- and plf_UMEA-3703-1-containing plasmids produced peritrichous fimbrial filaments at the surface of cells of strain ORN172 (Fig. 5; Fig. S2). In contrast, ORN172 containing the empty vector did not produce any fimbriae, as expected.

P fimbrial adhesins are known to be mannose-resistant hemagglutinins, and they demonstrate lectin activity specific to Gal(α1-4)Gal-containing glycolipids present on the surface of erythrocytes and other host cells. To compare the hemagglutination activity of P fimbria reference clones with that of clones producing PL fimbriae, we tested hemagglutination activity of fimbriae expressing clones in the nonfimbriated E. coli strain ORN172 for a variety of erythrocytes from different species (Fig. 6). The reference clone encoding P fimbriae with the PapG class I adhesin from E. coli J96/pPap5 demonstrated strong hemagglutinin activities with human, pig, dog, and rabbit erythrocytes. The reference clone encoding P fimbriae containing the Pap class II adhesin from E. coli IA2/pDC5 strongly agglutinated pig and human erythrocytes and, to a lesser extent, sheep and chicken erythrocytes. The reference clone encoding Prs fimbriae with the PapG class III (PrsG) adhesin from uropathogenic E. coli J96/pJFK102 agglutinated dog, pig, and sheep erythrocytes. The clone encoding PL fimbriae containing a Plf class I adhesin from E. coli UMEA-3703-1 agglutinated a broad range of erythrocytes from all species tested, except dog blood, although hemagglutinin titers were higher for human and sheep blood. Interestingly, the clone encoding the Plf class II adhesin from E. coli QT598 only strongly agglutinated human and turkey blood (Fig. 6). Taken together, these results indicate that, as with the P fimbrial classes of adhesins, PL fimbriae are hemagglutinins and that the PlfG class I and class II adhesins demonstrate distinct hemagglutination activities compared to the P fimbrial adhesins.

To determine whether lectin-based hemagglutination by PL adhesins was similar to those of Pap and related fimbriae, we tested whether globoseries glycolipids known to be recognized by P fimbriae (Gb3, Gb4, or Gb5) were able to inhibit PL fimbria-mediated hemagglutination. For the hemagglutination tests, we used a macroagglutination assay using human and turkey erythrocytes, since these were both recognized by the PlfG class I and class II fimbriae. In this assay, the PlfG class I and II adhesins strongly agglutinated human and turkey erythrocytes, and addition of either Gb3, Gb4, or Gb5 did not inhibit or reduce this agglutination. For the test controls, none of the PapG adhesins agglutinated turkey erythrocytes, but all strongly agglutinated human erythrocytes. For PapG class I, Gb3 and Gb4 reduced this agglutination, but not Gb5. For PapG class II, only Gb4 was able to inhibit agglutination. For PapG class III (Prs), Gb4 and Gb5, but not Gb3, were able to inhibit agglutination (Table 1). Taken together, these results indicate that the PL fimbriae belonging to either class I or class II recognize receptors on both human and turkey erythrocytes, but these receptors are distinct from those recognized by P fimbrial adhesin classes.

Since the PlfG class I and II adhesin sequences from E. coli strains QT598 and UMEA-3703-1, respectively, are quite distinct from each other and as the plf gene clusters share close genetic organization with pap gene clusters, we generated chimeric fimbrial gene clusters encoding different G adhesin alleles. These chimeric clones were based on the plf_QT598 gene cluster by generating a clone lacking the plfG gene (pIJ598) and then cloning the plfG_UMEA-3703-1 or papG alleles from each of the three PapG adhesin classes (Fig. S3A and B). Electron microscopy demonstrated that each of the five chimeric clones introduced to nonfimbriated E. coli ORN172 produced fimbriae on the surface of the cells (Fig. 5A). The PlfA subunit protein was also detected from cell surface extracts, as shown in immunoblots, although the level of protein present was decreased when the papG recombinant alleles were expressed compared to the plfG_QT598- and plfG_UMEA-3703-1-expressing clones (Fig. 5B). We also verified the hemagglutination capacity of hybrid fimbrial systems with some types of blood. The PL hybrid fimbriae with Pap class I or Pap class II adhesin could agglutinate human and pig blood, but were no longer able to agglutinate turkey blood. The PL fimbria hybrid with Pap class III Prs adhesin showed agglutination of sheep and human blood, but was also unable to agglutinate turkey erythrocytes (Table S3). This indicates that replacement of the PlfG allele with a Pap adhesin resulted in production of hybrid fimbriae with an altered binding specificity. It was also possible to replace the PapG adhesin in reference clone pDC1 (31) with the PlfG class II adhesin, resulting in production of fimbriae, as shown by electron microscopy (Fig. S3). However, this hybrid fimbrial system was unable to agglutinate either human or turkey erythrocytes. These results indicate that the PlfG adhesin can complement loss of the PapG adhesin for biogenesis of fimbriae, but suggest that an altered conformation of the adhesin in the hybrid system results in loss of hemagglutination capacity. As such, the Pap and PL fimbrial systems, although they share some characteristics, are not fully interchangeable with regard to adhesin function and specificity.

The adherence of bacteria to host epithelial cells such as bladder and kidney cells is an important step in colonization of the urinary tract. To investigate whether PL fimbriae can mediate adherence to host cells, we used clones expressing PlfG class I, PlfG class II, or PapG class I, II, or III fimbrial adhesins.

All the clones encoding P or PL fimbrial adhesins (reference and chimeric clones) demonstrated increased adherence to the bladder 5637 and kidney HEK-293 epithelial cell lines compared to the strain containing the empty vector (Fig. 7). The chimeric clones containing the plf gene cluster with hybrid pap or plfG adhesin-encoding genes also promoted adherence to epithelial cells. However, the clone containing a plf gene cluster, lacking the plfG or papG gene showed no appreciable adherence compared to the empty vector-containing clone (Fig. 7). These results demonstrate that PL fimbriae producing distinct types of G adhesins can mediate adherence to urinary tract epithelial cells and suggest that these fimbriae may potentially play a role during host colonization.

Since fimbriae can contribute to biofilm production, we tested for biofilm formation in polystyrene microtiter plates at different temperatures (25, 37, and 42°C) (Fig. 8). The clone expressing PL fimbriae with the PlfG class II adhesin showed a high level of biofilm production at all tested temperatures, even above that of a positive-control biofilm-forming Serratia liquefaciens reference strain. The clone expressing PL fimbriae with the PlfG class I adhesin also produced biofilm at 25 and 37°C at moderate levels compared to the clone producing the PlfG class II adhesin. However, biofilm production was very low at 42°C. The Pap class I- and III-producing reference clones were also able to form significantly more biofilm at 25, 37, and 42°C than the negative control. The Pap class II-expressing reference clone produced biofilm at higher temperatures (37 and 42°C), but biofilm levels were reduced at 25°C.

The chimeric clones that expressed different Pap or Plf adhesins fused to the plf_QT598 gene cluster were all able to produce appreciable levels of biofilm at both 25 and 37°C, although biofilm was much reduced at 42°C. The clone expressing the plf_QT598 gene cluster lacking a papG or plfG adhesin-encoding gene (ΔplfG clone) as well as clone containing the empty vector were not able to produce biofilm at all the tested temperatures (Fig. 8). Taken together, these results indicate that PL and P fimbriae expressing different types of G adhesins can mediate biofilm production in E. coli K-12 and that the PlfG class II adhesin in particular can contribute to strong biofilm formation.

To investigate the potential role of the PL fimbriae in virulence in the UTI model, 6-week-old female CBA/J mice were infected with wild-type strains QT598 and UMEA-3703-1 or with mutant Δplf strains lacking the genes encoding PL fimbriae. In the mouse infection model, loss of PL fimbriae did not have a significant effect on colonization of the bladder or kidneys by strain QT598 (Fig. S4). Strain UMEA-3703-1 and its Δplf mutant also showed no significant differences in colonization. However, UMEA-3703-1 was only able to colonize at lower levels (10² to 10³ CFU/g) compared to strain QT598 (10⁵ to 10⁶ CFU/g) (Fig. S4). Interestingly, the expression level of plf_QT598 was upregulated by more than 40-fold in the bladder of infected mice compared to expression following growth in vitro in LB medium (see Fig. 9 below). This suggests that the expression of these fimbriae is favored by environmental cues during infection in the urinary tract.

We also conducted a competitive infection model using the plf mutant and a wild-type lac-negative derivative of strain QT598. In this infection model, there was no significant difference in colonization of the mouse bladders (Fig. 10A). In contrast, the bacterial numbers of the plf mutant were significantly lower than those of the plf-expressing strain (P = 0.0084) and showed a median decrease of more than 16-fold (Fig. 10B) and an overall competitive index of −0.65 (Fig. 10C) in the kidneys. Taken together, these results indicate that the plf fimbrial gene expression is increased in the mouse urinary tract and that in competitive infection, loss of the PL fimbriae decreases competitive fitness in the mouse kidneys.

The bacterial strains and plasmids used in this study are listed in Table 2. ExPEC strain QT598 (a passaged derivative of strain MT156 [40]) is an O1:K1 sequence type ST1385 strain originally isolated from a turkey suffering from colibacillosis in France (27). UPEC strain CFT073 was isolated from the blood and urine of a woman suffering from urinary tract infection (41), and UMEA-3703-1 (NCBI Biosample no. SAMN01885978) was isolated from the urine of a human with bacteremia. E. coli K-12 laboratory strains DH5α and ORN172 (type 1 fimbria fim-negative strain) were used for cloning and protein expression. Reference clones expressing fimbria genes encoding different PapG adhesin classes were used as controls, including plasmids pPap5 (papG_J96 class I) (42, 43), pDC5 (papG_IA2 class II) (31), and pJFK102 (prsG_J96 class III) (44, 45).

Bacteria were routinely grown at 37°C on solid or liquid Luria-Bertani (LB) medium (Alpha Bioscience, Baltimore, MD). When required, antibiotics were added to a final concentration of 100 μg/mL of ampicillin, 30 μg/mL of chloramphenicol, or 50 μg/mL of kanamycin.

Identification and comparison of sequences in the databases were achieved by accessing data on completed genomes and BioProjects publicly available in the NCBI database (www.ncbi.nlm.nih.gov). Analyses included BLAST against both nucleotide and protein entries. Figures presenting the organization and comparison of genes and gene clusters were generated from the nucleotide accession numbers and entries using SnapGene (version 5.2.1) (www.snapgene.com). For comparison of the protein sequences, entries were obtained from either the NCBI or the Universal Protein Resource (UniProt) (www.uniprot.org) website. Phylogenetic analyses of protein sequences were done with the platform at Phylogeny.fr (http://www.phylogeny.fr) using the default (“one click”) parameters (46). Analyses consisted of multiple-sequence alignment with MUSCLE (47), alignment curation with GBlocks (48), maximum likelihood phylogeny analysis using PhyML 3.0 (49), and TreeDyn for generation and editing of trees (www.treedyn.org). Specific parameters are described at the Phylogeny.fr web site platform.

Cloning of the plf gene clusters and plfG and papG genes encoding the different classes of adhesins was performed by PCR amplification using specific primers (Table 3) and Q5 high-fidelity DNA polymerase (New England Biolabs [NEB]). The A-Tailing kit (NEB) was then used to add additional deoxyadenosine (A) to the 3′ end of the PCR products. The insert possessing the additional A at the 3′ end was ligated to the linearized vector with additional deoxythymidine (T) residues using T4 DNA ligase (NEB). The plf gene cluster from strain QT598 (plf_QT598) was amplified using primers CMD1847_F and CMD1900_R and cloned into vector pUCm-T (Bio Basic, Markham, ON, Canada), generating plasmid pIJ507. This plasmid contains the full plf gene cluster with a PlfG class II adhesin. The plf gene cluster from strain UMEA-3703-1 (plf_UMEA) was amplified using primers CMD2119_F and CMD2120_R and cloned into vector pBC sk+, generating plasmid pIJ523. This plasmid contains the full plf_UMEA gene cluster with a PlfG class I adhesin. To generate chimeric gene clusters comprised of plf_QT598 with different types of G adhesin-encoding genes, pIJ507 was used as a template. plfG_QT598 was deleted using an inverse-PCR method with primers CMD2168_F and CMD2169_R, which introduced PmeI sites and amplified a linear fragment lacking the plfG_QT598 gene. The linearized product was then treated with DpnI endonuclease (NEB) to cleave any methylated template sequence. The linear fragment was either ligated using T4 DNA ligase (NEB) to generate pIJ598, which contains plfBAHCDJKEF (plf_QT598 ΔplfG), or used as a template to generate chimeric fimbrial gene clusters containing different G adhesin genes using the T4 DNA ligase (NEB). PCR fragments containing G adhesin genes were obtained using primer pairs CMD2171_F and CMD2172_R (for papG class I from strain J96), CMD2174_F and CMD2175_R (for papG class II from strain CFT073), CMD2177_F and CMD2178_R (for prsG [papG class III] from strain J96), and CMD2180_F and CMD2181_R (for plfG_UMEA-3703-1::plfGI from strain UMEA-3703-1). Cloning experiments to generate recombinant plasmids or subclones were first achieved using E. coli strain DH5α. The plasmids were extracted using a Miniprep kit according to the manufacturer’s recommendations (Bio Basic, Inc.) and then transformed into E. coli fim-negative strain ORN172 for phenotypic testing. Strains containing the following reference plasmids that contain full P fimbrial gene clusters were used as reference controls: pPap5 (encoding P fimbriae PapG class I from E. coli J96) (42, 43), pDC5 (encoding P fimbria PapG class II from strain IA2) (31), and pJFK102 (encoding Prs fimbria PrsG [PapG class III] from E. coli J96) (44, 45).

A plf knockout mutant of APEC strain QT598 was obtained by the lambda red recombinase method (50). First, plasmid pKD46 expressing lambda red recombinase was transformed into QT598 by electroporation, and then the kanamycin resistance cassette was amplified from plasmid pKD4 by PCR with primers CMD2112_F and CMD2113_R and transformed into QT598 carrying plasmid pKD46 by electroporation. Mutants were selected at 37°C, and then the loss of genes was confirmed by PCR using screening primers CMD1849_F and CMD1900_R to obtain the QT598 Δplf strain (QT4420). The same method was used to create the Δplf deletion mutation in UMEA-3703-1 using primers CMD2112_F and CMD2114_R to generate strain QT4598 (UMEA-3703-1 Δplf). The deletion was confirmed using primers CMD2115_F and CMD2116_R.

Fimbriae were extracted using the heat extraction method as described previously by (51), with some modifications. Briefly, overnight cultures were incubated at 56°C for 1 h and harvested by centrifugation at 4,000 rpm for 15 min. Supernatants were incubated in 10% trichloroacetic acid (TCA) to precipitate proteins. Proteins were then concentrated by centrifugation at 12,000 rpm for 20 min, washed twice with Tris-EDTA (0.05 M) at pH 12 and 8.5, and resuspended in 0.1 mL of Tris-EDTA (0.05 M) at pH 8.5. Western blotting was performed as previously described (52), with some modifications. Proteins were separated using 15% polyacrylamide gel, transferred onto a nitrocellulose membrane (Bio-Rad Laboratories, CA, USA), and blocked with 15 mL of blocking buffer TBST (0.15 M NaCl, 0.025 M Tris, 0,05% Tween, 3% skim milk) for 1 h at 4°C. Fimbrial major subunit protein was detected using rabbit polyclonal antibodies provided by New England Peptide (1:1,000) against a peptide corresponding to part of the PlfA major fimbrial subunit (Ac-CAHLAADGISVKKD-amide) for 45 min at room temperature. The membrane was then washed 3 times with wash buffer (0.15 M NaCl, 0.025 M Tris, 0,05% Tween) and incubated with an anti-rabbit-conjugated secondary antibody (1:20,000) for 45 min at room temperature. After four washes with TBST, proteins were detected using SuperSignal West Pico chemiluminescent substrate (Pierce) according to the manufacturer’s instructions.

Bacteria for electron microscopy were grown overnight at 37°C. Cultures were then adsorbed onto a glow-discharged Formvar-coated copper grid for 1 min and stained with 1% phosphotungstic acid. The excess of liquid was removed with a filter paper. Samples were then dried and observed under a Hitachi H700 transmission electron microscope.

Hemagglutination was performed in 96-well round-bottom plates as described in reference 53. Briefly, different types of blood were tested for this assay, human (A and O), horse, bovine, sheep, pig, rabbit, chicken, turkey, and dog red blood cells (RBCs) were suspended in phosphate-buffered saline (PBS) at a final concentration of 3% and added to 96-well plates. Clones expressing different classes of PapG were grown in LB broth at 37°C and centrifuged at 3,000 × g for 15 min, and pellets were suspended in PBS (pH 7.4) and adjusted to an optical density at 600 nm (OD₆₀₀) of 0.6, centrifuged, and then concentrated 100-fold in PBS (pH 7.4). The agglutinating titer was determined as the most diluted well with agglutination after 30 min of incubation on ice.

To investigate potential inhibition of hemagglutination by carbohydrates known to be recognized by P fimbriae, a macroagglutination assay was performed. Overnight bacterial cultures were centrifuged and concentrated 10× in PBS. Thirty microliters of 6% human O blood group or turkey erythrocyte suspensions were deposited into each well of a 10-well glass plate. Two microliters of the different inhibitors was added to the erythrocytes. Finally, 30 μL of the different control and test strains was mixed using toothpicks. Samples were left on ice for 30 min followed by gentle agitation, then hemagglutination levels were noted. The carbohydrates used for inhibition tests were ceramide trihexosides (CTH) (globotriaosylceramides [Gb3]) and globotetrahexosylceramide-globosides (Gb4) obtained from Matreya (State College, PA, USA), and Forssman antigen pentaose (Gb5) from Biosynth Carbosynth (San Diego, CA, USA). Stock solutions of the carbohydrates were prepared in dimethyl sulfoxide (DMSO). Final concentrations used for hemagglutination assays were as follows: Gb3, 32 μg/mL or 1.8 nmol per well; Gb4, 161 mg/mL or 7.5 nmol per well; and Gb5, 161 μg/mL or 11 nmol per well. Samples were tested in duplicate and gave similar results.

Biofilm formation in 96-well microtiter plates was examined as previously described (54). Fimbrial clones were grown statically in LB at 25, 30, 37, and 42°C for 48 h. After 48 h of incubation, the liquid was discarded, and plates were washed and stained with 0.1% crystal violet (Sigma) for 15 min. Ethanol-acetone solution (80:20) was used to dissolve biofilm, and the optical density was measured at 595 nm to determine the production of biofilm.

Cells of the human bladder 5637 (ATCC HTB-9) and kidney HEK-293 (ATCC CRL-1573) epithelial cell lines were grown to confluence in 24-well plates in RPMI 1640 or Eagle’s minimal essential medium (EMEM) (Wisent Bio Products, St-Bruno, Canada) supplemented with 10% fetal bovine serum (FBS) at 37°C in 5% CO₂. Fimbrial clones expressing different classes of PapG or PlfG adhesins were grown in LB medium at 37°C, cultures were then centrifuged, resuspended in RPMI 1640 or EMEM with 10% FBS, and added to cells at a multiplicity of infection (MOI) of 10 for 2 h, as described previously (55). After 2 h, cells were washed three times with PBS, lysed with 1% Triton X-100, diluted, and plated onto LB agar plates supplemented with selective antibiotics.

To determine the potential role of PL fimbriae in virulence, wild-type strains QT598 and UMEA-3703-1 as well as the QT598 Δplf (QT4420) and UMEA-3703-1 Δplf (QT4598) mutants were tested in 6-week-old CBA/J female mice using an ascending UTI model adapted from reference 38. A total of 5 mice in each group were infected with 10⁹ CFU/mL of bacteria. After 48 h, the infected mice were euthanized, and bladders and kidneys were harvested aseptically for the bacterial count on MacConkey’s agar plates. To study the expression of plf in vivo, bladder samples after necropsy were homogenized with TRIzol LS reagent (Thermo Fisher Scientific) for RNA extractions.

A competitive infection model was also used as described in reference 27. Briefly, a murine ascending UTI model with 9 mice in each group was used for coinfection, in which a virulent ΔlacZYA derivative of QT598 (QT4567) was coinfected with the Δplf strain (QT4420). Twenty-five microliters (10⁹ CFU) of a mixed culture containing equal amounts of each strain were inoculated through a catheter in 6-week-old CBA/J female mice. Mice were euthanized after 48 h, and bladders and kidneys were harvested aseptically, homogenized, diluted, and plated on MacConkey agar plates.

We compared the expression of PL fimbriae by comparing RNA levels of the plfA gene under different conditions: LB medium, minimal M63 medium, and during infection in bladders of mice. For in vitro analysis, total RNAs from bacterial samples were extracted according to the manufacturer’s protocol EZ-10 spin column total RNA Miniprep kit (BioBasic). For in vivo analysis, bladder samples were homogenized with TRIzol LS reagent (Thermo Fisher Scientific), incubated with chloroform followed by centrifugation and incubation in ethanol (95 to 100%) to separate the aqueous phase that contains RNA. Then, RNA samples were extracted using a Direct-zol RNA Miniprep kit (Zymo Research, Irvine, CA, USA) according to the manufacturer’s recommendations. All RNA samples were treated with Turbo DNase (Ambion), to eliminate any DNA contamination. The Iscript reverse transcription supermix (Bio-Rad Life Science, Mississauga, ON, Canada) was used to synthesize cDNAs from samples according to the manufacturer’s protocol. Primers were specific to the plfA gene and the RNA polymerase sigma factor rpoD (housekeeping control). Quantitative reverse transcription-PCR (qRT-PCR) was performed in the Corbett Rotorgene (Thermo Fisher) instrument using 50 ng of cDNA, 100 nM each primer, and 10 μL of SsoFast Evagreen supermix (Bio-Rad). Data were analyzed using the threshold cycle (2^−ΔΔCT) method (56).

All data were analyzed with the Graph Pad Prism 6 software (GraphPad Software, San Diego, CA, USA). A Mann-Whitney test was used for mouse infection experiments to determine statistical significance. Analysis of variance (ANOVA) was used to compare the means of samples. Differences between groups were considered significant for P values of <0.05.

Protocols for mouse urinary tract infection were approved by the animal ethics evaluation committee (Comité Institutionel de Protection des Animaux; CIPA no. 1608-02) of the INRS-Centre Armand-Frappier Santé Biotechnologie.