Role of Capsule and O Antigen in the Virulence of Uropathogenic Escherichia coli

Urinary tract infection (UTI) is one of the most common bacterial infections in humans, with uropathogenic Escherichia coli (UPEC) the leading causative organism. UPEC has a number of virulence factors that enable it to overcome host defenses within the urinary tract and establish infection. The O antigen and the capsular polysaccharide are two such factors that provide a survival advantage to UPEC. Here we describe the application of the rpsL counter selection system to construct capsule (kpsD) and O antigen (waaL) mutants and complemented derivatives of three reference UPEC strains: CFT073 (O6:K2:H1), RS218 (O18:K1:H7) and 1177 (O1:K1:H7). We observed that while the O1, O6 and O18 antigens were required for survival in human serum, the role of the capsule was less clear and linked to O antigen type. In contrast, both the K1 and K2 capsular antigens provided a survival advantage to UPEC in whole blood. In the mouse urinary tract, mutation of the O6 antigen significantly attenuated CFT073 bladder colonization. Overall, this study contrasts the role of capsule and O antigen in three common UPEC serotypes using defined mutant and complemented strains. The combined mutagenesis-complementation strategy can be applied to study other virulence factors with complex functions both in vitro and in vivo.


Introduction
Urinary tract infections (UTIs) are one of the most common types of bacterial infection in humans. Approximately 150 million cases of UTI are estimated to occur across the globe annually [1]. UTI occurs frequently in women; it is estimated that 40-50% of adult healthy women experience at least one UTI episode in their lifetime [2]. Catheterized patients, the elderly and children are also at a higher risk for UTI [3][4][5]. If untreated, UTI may progress from the bladder to the kidneys and even cause life-threatening urosepsis. The estimated annual cost of treating UTI in the USA alone is greater than $1.6 billion, thus emphasizing the global economic burden these infections place on health care systems [2].
Uropathogenic Escherichia coli (UPEC) is the primary causative organism in more than 80% of all UTI cases [6]. UPEC isolates exhibit a high degree of genetic diversity, primarily due to the possession of specialized virulence genes located on mobile genetic elements called pathogenicity islands [7,8]. Although no single virulence factor is uniquely definitive of UPEC, their ability to cause symptomatic UTI is enhanced by adhesins (e.g. type 1, and P fimbriae), toxins (e.g. hemolysin), iron acquisition systems and cell surface polysaccharides [9][10][11]. The capsule and the O antigen represent two cell-surface associated polysaccharides that contribute to UPEC virulence [12][13][14]. These extracellular polysaccharides protect UPEC against phagocytosis by immune cells such as neutrophils and inhibit complement-mediated killing [15,16].
Bacterial lipopolysaccharide (LPS) plays a primary role in mediating UPEC resistance to the bactericidal activity of human serum [17]. LPS consists of the highly conserved lipid A-core and repeating O antigen subunits that differ greatly between strains based on the sugar residues and their linkage patterns within the repeating subunits [18]. More than 180 different serotypes have been reported for E. coli [18,19], with some O antigen types such as O1, O2, O4, O6, O7, O8, O16, O18, O25, and O75 being common among UPEC strains [20]. This variation is imparted by differences in the genes that encode for glycosyltransferase enzymes that modify O antigen structure. The lipid A-core and the O antigen subunits are assembled in separate pathways that come together for ligation at the inner membrane [21]. WaaL is a conserved O antigen ligase and the only known membrane enzyme implicated in the ligation of pre-assembled undecaprenyldiphosphate (Und-PP)-linked O antigen to lipid A-core oligosaccharide [22]. E. coli K-12 waaL deletion mutants are unable to cap the lipid A-core and are therefore disrupted in the expression of O antigen at the cell surface [23].
Like the O antigen, there is also significant variation in the E. coli capsule. More than 80 distinct types of E. coli capsule have been identified, with four major groups defined based on genetic assembly, biochemistry and physical attributes [24]. The group 2 capsular polysaccharides have mostly been associated with UPEC strains [12,25]. The role of the capsule in protecting UPEC from phagocytosis has been defined [26,27], however, its role in resistance to serum-mediated killing is unclear, with several conflicting reports in the literature [28][29][30][31]. Understanding the role of the capsule in UPEC serum resistance is further complicated by the use of multiple strains with different O antigen and capsule types. For example, a multi-strain study by Stawski et al. found that the contribution of capsule to serum resistance varied between strains and ranged from being important to having no protective role [32]. The genes involved in capsule production are located in the kpsFEDUCS operon [24,33]. KpsD is a 60 kDa outer membrane protein involved in the export of group 2 capsular polysaccharides across the outer membrane [34]. Mutation of kpsD results in the loss of surface expression of the capsule and accumulation of polysaccharide within the periplasmic space [35].
Given the importance of O antigen and capsule in UPEC virulence, we studied the function of these polysaccharides in three well-characterized UPEC strains (CFT073, RS218 and 1177) bearing different O and K antigens. We employed a recently described counter-selection strategy based on streptomycin-kanamycin resistance (mediated by the rpsL and neo genes, respectively) for the mutation of single genes and the subsequent repair of these mutations to generate complemented strains [36]. This enabled us to generate a defined set of wild-type (WT), mutant and complemented strains that were used to test the contributions of O antigen and capsule to UPEC virulence. Our study is the first to demonstrate the use of the rpsL-neo counter-selection system in UPEC, and shows that the O antigen is the major polysaccharide responsible for UPEC serum resistance and survival in the mouse urinary tract.

Construction of capsule and O antigen insertion mutants
A prerequisite for the rpsL-neo counter-selection system is that the bacterial strain should be streptomycin resistant (Strep R ) and that this resistance is the direct effect of a mutation within the rpsL gene. Streptomycin resistance is a recessive trait and in the presence of both WT and mutant alleles, the strain is sensitive to streptomycin (Strep S ) [37]. This trait can be exploited for the rapid mutation and repair of mutated genes, and a schematic outline of the strategy we employed is presented in Figure 1. Strep R derivatives of the UPEC strains CFT073 (O6:K2), 1177 (O1:K1) and RS218 (O18:K1) were generated spontaneously by sequential plating on increasing concentrations of streptomycin. These Strep R strains (CFT073 Strep , 1177 Strep , RS218 Strep ) were then transformed with pRedET, a temperature sensitive expression plasmid harboring the l-Red recombinase genes. Mutation of the waaL (O antigen) and kpsD (capsule) genes in the pRedETcontaining strains was performed by insertional inactivation following homologous recombination between the bacterial chromosome and the rpsL-neo cassette flanked by 50 bp homology regions and subsequent growth on kanamycin LB-agar. These mutants, which became sensitive to streptomycin due to the presence of a functional rpsL gene, were referred to as CFT073waaL, CFT073kpsD, 1177waaL, 1177kpsD, RS218waaL and RS218kpsD, respectively.

Generation of complemented strains
To repair the mutation and generate complemented strains, pRedET was re-introduced into the mutant strains and they were transformed with a 100 bp oligonucleotide that contained the same sequence as the WT strain in the regions flanking the point of insertion of the rpsL-neo cassette ( Figure 1). The re-constructed WT strains were selected based on resistance to streptomycin, as loss of a functional rpsL gene reverted them to a Strep R phenotype.
These strains were referred to as CFT073waaL C , CFT073kpsD C , 1177waaL C , 1177kpsD C , RS218waaL C and RS218kpsD C , respectively. In summary, we therefore generated a series of O antigen and capsule mutants, as well as repaired/complemented strains, in three UPEC strains with different O and K antigens. Hereon, all references to WT strains denote the Strep R derivatives of CFT073, E. coli 1177 and RS218, respectively.

Phenotypic characterization of capsule mutants and complements
The K1 phage assay was used to detect the presence of a K1 capsule on the surface of UPEC strains 1177 and RS218 ( Figure 2). In this assay, the 1177 and RS218 WT strains were killed upon encountering the phage due to its specific attachment to the K1 capsule and subsequent lytic replication life-cycle. In contrast, the kpsD mutants exhibited normal growth beyond the streak line, indicating a loss of K1 capsule expression and thus resistance to phage killing. A WT phenotype was fully restored in the complemented strains. In separate experiments, counter-current immunoelectrophoresis using K2 antiserum was used to detect K2 capsule on the surface of CFT073. Capsular extracts from CFT073, CFT073kpsD and CFT073kpsD C were electrophoresed against K2 antisera. A precipitin band was observed for CFT073 and CFT073kpsD C , while this was absent in CFT073kpsD. Taken together, these results confirm the K1 or K2 capsule phenotype of the three mutated and complemented UPEC strains.

Phenotypic characterization of O antigen mutants and complemented strains
The waaL gene encodes an O antigen ligase and its mutation abolishes the attachment of the O antigen to the LPS core, resulting in a rough phenotype. In order to phenotypically characterize the waaL mutants and their complemented strains, LPS extracts were prepared from standardized late log phase cultures, separated by TSDS-PAGE and viewed by silver staining ( Figure 3). As expected, due to their different O types, the LPS banding patterns from WT CFT073, 1177 and RS218 varied. E. coli 1177 (O1) possessed an even distribution of short, medium and long O antigen side chains, while CFT073 (O6) and RS218 (O18) possessed predominantly short and long O antigen side chains. Disruption of the waaL gene resulted in complete loss of the bands corresponding to the O antigen subunits from all strains, with only the lipid A-core detected. In each of the complemented strains, the O antigen-banding profile was restored to WT.

O antigen is the major determinant of serum resistance
To compare the relative contribution of O antigen and capsule to UPEC serum resistance we tested our WT, mutant and complemented strains in a serum survival assay. In this assay, bacterial survival was measured as colony forming units (CFU) before and after incubation with pooled normal human sera. As expected, the WT strains CFT073, 1177 and RS218 were resistant to killing by human serum (Figure 4). In contrast, the CFT073waaL, 1177waaL and RS218waaL mutants were all highly sensitive to human serum and displayed a high serum sensitivity index (SSI; P,0.05, paired t test). This phenotype reverted back to WT (i.e. resistant) in each of the respective complemented strains. The contribution of the capsule to serum resistance was less clear. The CFT073kpsD mutant was most sensitive (SSI = 2, P = 0.046), while RS218kpsD displayed reduced sensitivity (SSI = 1, P = 0.027). Complementation of the kpsD mutation in CFT073kpsD and RS218kpsD resulted in reversion to a WT resistance phenotype. The 1177kpsD mutant, on the other hand, was completely resistant to serum killing. The E. coli K12 strain MG1655 was used as a control for the integrity of serum, and was completely killed in all assays. All the strains were resistant to heat-inactivated normal human sera (data not shown). Taken together, our data show that O antigen is the major determinant of serum resistance in the UPEC strains tested. Furthermore, given that 1177 and RS218 are K1 capsule type, the observed difference in their serum sensitivities indicate that the contribution of the capsule to serum resistance is closely linked to O antigen serotype.

UPEC survival in whole blood
The most well-characterized virulence-related role of capsular expression in UPEC relates to bacterial resistance to uptake and killing by phagocytes such as neutrophils and monocytes [15]. In view of this, and the data from the above assays that showed a less clear contribution of the capsule to serum resistance compared to O antigen, we performed a series of whole blood killing assays. These assays incorporate both of the cell types relevant to UPEC phagocytosis. In these experiments, the CFT073kpsD and 1177kpsD mutants displayed significantly enhanced sensitivity to killing by whole blood compared to their respective WT counterparts ( Figure 5; P,0.05, independent samples t test); in contrast, RS218 WT was so sensitive to killing in this assay that the data on the kpsD mutants could not be interpreted. Complementation of the kpsD mutation in CFT073kpsD and 1177kpsD restored survival to WT levels. Given that 1177 and RS218 both produce a K1 capsule, these data suggest that there may be differences in the composition of their K1 antigen, or that the role of the K1 capsule in resistance to killing by whole blood is strongly linked to O antigen type, as observed in serum killing assays.
The O6 antigen and K2 capsule are required for colonization of the mouse urinary tract Analysis of plasmid-based complemented strains in vivo is challenging as it is difficult to ensure consistent expression of the cloned gene or maintenance of the plasmid in the pathogen during infection. A major advantage of the counter selection strategy described here is that complementation is performed by repair of the mutated chromosomal gene, and thus the complemented strains can be tested in vivo. In order to demonstrate this, we analyzed the CFT073 WT, capsule and O antigen mutant and complemented strains in a mouse UTI model. Experiments were performed as mixed competition assays, whereby mice were inoculated with a 1:1 mix of CFT073:CFT073kpsD, CFT073kpsD: CFT073kpsD C , CFT073:CFT073waaL or CFT073waaL:CFT073-waaL C and the number of colonizing bacteria was calculated at 18 h post infection. In these assays, both the capsule and O antigen mutants were significantly outcompeted by the WT and complemented strains in colonization of the bladder ( Figure 6; P,0.05). No significant colonization of the kidneys was observed (data not shown); this is consistent with previous data from our laboratory using C57BL/6 mice [38][39][40][41]. Taken together, these data demonstrate that the O6 antigen has a major impact on the colonization of the urinary tract by CFT073, and that the K2 capsular antigen, while less important, also contributes to this phenotype. Figure 1. Schematic representation of the mutagenesis strategy utilizing rpsL counter selection. A 100 bp conserved region (shown as patterned region in WT sequence) was selected for the kpsD and waaL; Strep R strains harboring the pRedET plasmid were transformed with PCR product consisting of the rpsL-neo cassette flanked by 50 bp regions homologous to sequences on either side of the chosen point of insertion; double cross-over events during homologous recombination make the mutant strains Strep S Kan R due to integration of the rpsL-neo cassette; for complementation, mutant strains were transformed with 100 bp oligonucleotides bearing the same sequence as the WT across the region of insertion; complemented strains were again Strep R due to loss of the rpsL-neo cassette and restoration of the parental genotype; streptomycin resistant (Strep R ), streptomycin sensitive (Strep S ), kanamycin resistant (Kan R ). The methodology depicted in this figure is a modification of the rpsL counter-selection technique previously used for site-directed mutagenesis in bacterial artificial chromosomes [57][58][59]. doi:10.1371/journal.pone.0094786.g001

Discussion
The polysaccharide capsule and O antigen are well-described UPEC virulence determinants that contribute to resistance against host defenses. Specific O and K types are associated with UPEC infection, often to different degrees depending on the serotype of the strain. In this study, we investigated the role of different O and K antigens in mediating resistance to serum and whole blood, and colonization of the urinary tract, using a novel set of WT, mutant and complemented strains.
Gene disruption in bacteria using l-Red mediated homologous recombination is a widely used and highly efficient technique [42]. In this method, the mutation of a gene or set of genes is attained by insertional inactivation using an antibiotic cassette that can subsequently be removed by FLP-mediated recombination to leave a characteristic scar sequence at the target site. Several modifications have been made to l-Red recombineering to improve recombination frequency that can otherwise be a major limitation in certain applications. These include using transformation amplimers with large flanking regions [43], the removal of endogenous nucleases from the host strain [44], and the use of three-way overlapping PCR methods that negate the need for multiple intermediary cloning steps [45,46]. Complementation of a specific mutation, a common procedure used to prove the function of a gene and fulfill Koch's postulates, is normally achieved in trans by introduction of the relevant gene into the cell on a plasmid. This approach, while effective, can be complicated by the need for antibiotic selection to maintain the plasmid [47,48], differences in copy number of the complemented gene due to its location on a plasmid [49] and differences in expression of the complemented gene [50]. These issues are often compounded when complementation is attempted in an animal infection model. Another approach for complementation involves the use of a transducing phage to repair the mutation on the chromosome [51]. This involves the replacement of up to 40 kb of the chromosomal region surrounding the target gene, and also requires the use of a strain-specific phage. A different allelic exchange strategy using suicide vectors can be employed when several rounds of modification are required, however problems such as the need for large flanking regions to facilitate recombination, loss of temperature sensitivity of the vector or drastically reduced recombination efficiency due to mismatched bases with target DNA sequences make this method less than ideal when dealing with non genome-sequenced clinical isolates [52]. Transposable elements also provide an alternative to l-Red  mediated homologous recombination. Transposition based strategies involving transposons such as Tn-5 and Tn-7, as well as their derivatives, have been used successfully to generate transposon mutant libraries [53,54] for site specific tagging [55] and for the generation of transcriptional/translational target gene fusions [56].
The rpsL counter selection system has been used previously for the modification of bacterial artificial chromosomes in viral and eukaryotic systems [57][58][59]. In E. coli, this system has been used in conjunction with loxP to generate markerless chromosomal deletions [60] and to introduce point mutations at specific sites on the chromosome [36]. Here, we have demonstrated its application in the construction of mutant and complemented UPEC strains. The method for complementation was highly efficient, and only required the use of a 100 base oligonucleotide that matched the target region. Since the complementation method stably generates the parental genotype without the need for a selectable marker, complemented strains can be used in animal infection experiments without antibiotic selection. This method is analogous to that previously described for sacB, another widely used counter selectable marker [61]. However, an advantage of rpsL counter selection is that unlike the sacB system, it can still be used in strains containing a native ColE1 replicon  Bacteria were incubated in fresh whole blood for 3-5 h and colony counts were used to compare survival phenotypes of mutants and complemented strains to the WT parents. Fitness of the CFT073 and 1177 capsule (kpsD) mutants was significantly impaired compared to the WT, and was restored completely in the respective capsule complemented (kpsD C ) strains. The survival data for the RS218 WT derivatives could not be analyzed due to the high sensitivity of WT strain to killing by whole blood. P values, as determined by the independent samples, t test were as follows: # 0.038, * 0.042. doi:10.1371/journal.pone.0094786.g005 [62,63]. A disadvantage of the rpsL counter selection system is that it requires the generation of a spontaneous streptomycin resistant derivative of the strain to be studied, and it cannot be applied directly to strains with an existing streptomycin resistance cassette [64,65]. Mutations in the native rpsL gene have also been shown to result in a fitness cost on E. coli growth, with the decline in bacterial fitness dependent on the type of rpsL mutation and the carbon source available in the medium [66,67]. Thus, this issue needs to be considered when comparing the virulence of knockout and complemented strains where one is Strep R and the other is Strep S . In this study, we used spontaneously generated Strep R derivatives of each parent strain (referred to as WT for simplicity), as well as mutants and repaired/complemented strains generated from these Strep R derivatives. Although we did observe a slight decrease in exponential growth rate for the Strep R strains, several of our mutants (which were Strep S ) still exhibited clear phenotypes with respect to serum resistance, whole blood killing and bladder colonization. Furthermore, our spontaneous CFT073 Strep R derivate (CFT0073 Strep ) colonized the C57BL/6 mouse bladder at a similar level to that we have reported previously for CFT073 [38,68].
Three UPEC strains with different O and K antigens were used to demonstrate the application of the rpsL counter selection system. For UPEC, the potential influence of waaL deletion and disruption of expression of O antigen on virulence is undefined except for a recent study that showed waaL was required for efficient colonization of mouse bladder [14]. Disruption of the waaL gene has been shown to result in uncapped lipid A-core due to the absence of a functional O antigen ligase [23,69]. Our waaL mutants had a similar lipid A-core region, while the characteristic O antigen was absent (and restored in complemented strains). The role of O antigen in mediating serum resistance in multiple bacterial pathogens has been extensively reported in the literature [32,43,70,71]. Our results demonstrate that the O1, O6 and O18 antigens are important for survival in serum. This is consistent with previous studies on the O18 antigen, which has been extensively studied in the context of neonatal meningitis strains [72][73][74]. Our results for the O1 and O6 antigens, while expected, represent to the best of our knowledge the first demonstration of this phenotype in well-defined mutant and complemented strains.
Mutation of the kpsD gene resulted in mutants lacking any surface exposed K antigen, and this was restored by complementation. In our experiments, the K2 capsule contributed to the resistance of CFT073 against the bactericidal activity of human serum. A study by Buckles et al. has shown the importance of the K2 capsule in UTI pathogenesis and its role in preventing complement mediated killing [31]. Notably, the serum sensitivity index for the K2 mutant in our study was significantly lower than its corresponding O antigen mutant (P,0.01). However, both of the strains were resistant to heat-inactivated normal human sera indicating a role for complement attack that depends at least in part, on the presence of O antigen. There are conflicting reports in the literature about the importance of K1 capsule in effecting bacterial survival in serum. A study by Leying et al. has suggested that K1 expression is a prerequisite for serum resistance [28]. However, the mutants used were not defined and complementation was not described. Likewise, Kim et al. reported that the K1 capsule plays a more important role than the O18 antigen in establishing bacteremia in a neonatal rat meningitis model [29]. In contrast, an earlier report concluded that in a similar animal model, the K1 capsule, while important for establishing meningitis, had no effect on bacterial survival in serum [75]. Several other reports have corroborated these findings [30,76,77]. Both the RS218 and 1177 strains used in our study produce a K1 capsule. We observed that while the absence of the capsule renders RS218 slightly susceptible to the bactericidal activity of serum, it had no significant effect on 1177. We also analyzed UPEC survival in whole blood to elucidate the role of capsule in protecting against whole blood killing. Survival of the CFT073 and 1177 capsule mutants was significantly impaired, which is similar to the role of K5 capsule antigen that mediates resistance to phagocytosis in UPEC [15]. The RS218 WT strain exhibited a poor survival rate and its corresponding mutants could not be analyzed by this assay. Thus, the importance of the K1 capsule in mediating serum resistance and survival in whole blood is variable, and possibly linked to the O antigen type and other factors such as hemolysin Figure 6. Fitness of CFT073 capsule (kpsD) and O antigen (waaL) mutants during in vivo competition with CFT073 WT and corresponding complemented (kpsD C , waaL C ) strains. C57B/L6 mice were infected transurethally with a 1:1 mixture of CFT073kpsD and CFT073 WT or CFT073kpsD C ; CFT073waaL and CFT073 WT or CFT073waaL C . (A) Each marker represents LOG 10 total CFU recovered from each mouse per 0.1 g of bladder tissue. Lines connect data points for the same mouse, and horizontal bars represent median values. (B) Each marker represents the LOG 10 competitive index calculated for each individual mouse; competitive indices are the ratio of the mutant 0.1 g of bladder tissue to that of WT or complemented strain. Dashed lines represent hypothetical competitive index of 1 (LOG 10 1 = 0), which indicates no difference in fitness between the two strains. Horizontal bars represent group medians, and each competition group had 6 to 8 mice. All mutants were significantly outcompeted by CFT073 WT or their respective complemented strains for bladder colonization (P,0.05). The O6 antigen was significantly more important than the K2 capsular antigen for bladder colonization (P,0.05). doi:10.1371/journal.pone.0094786.g006 [78], outer membrane proteins (OmpA, OmpC) [79,80] and plasmid-encoded virulence factors (TraT, Iss) [81,82].
A key advantage of the rpsL counter selection strategy is that mutants can be complemented and studied in vivo. To demonstrate this, the ability of the CFT073 capsule and O antigen mutant strains to cause disease in a mouse UTI model was analyzed. We observed that both the K2 and O6 mutants were outcompeted by the WT and the complemented strains in bladder colonization during mixed infection challenges. However, based on competitive index the O6 antigen was significantly more important for disease pathogenesis than the K2 capsule. These results confirm the importance of the O6 antigen and, to some extent, the K2 capsular antigen, as colonization factors of the mouse urinary tract. The results are also consistent with a recent report demonstrating that restoration of the O-antigen producing capacity of the E. coli K12 strain MG1655 was sufficient for it to kill Caenorhabditis elegans at rates comparable to other pathogenic bacteria [83]. The expression of O antigen and capsule on the bacterial surface can also affect the function of other virulence factors [84,85]. Differences in the role of other UPEC K capsular antigens during infection have also been described previously. For example, the K15 antigen is important for virulence but not for serum resistance in a neonatal mouse UTI model [86], while a protective role for O4 but not for the K54 antigen has been described during colonization of the mouse urinary tract [87].
The O antigen and capsule are well known virulence factors of UPEC, however the interplay between common O and K antigens remains to be assessed using defined mutant and complemented strains. Here we have described a strategy to make single gene mutants and their complements that can be analyzed in vitro and in vivo. With the aid of mutant and complemented strain sets we have shown that the surface expression of the O1, O6 and O18 antigens and to a lesser extent, the K2 antigen, are important for bacterial survival in human serum. O1, O6, K1 and K2 antigens were all found to provide a survival advantage in the presence of whole blood. The O6 antigen was shown to be more important than the K2 antigen for colonization of the mouse bladder. The strategy described herein can be applied to study other virulence factors with complex functions and thus avoid the use of plasmid-based systems for complementation.

Ethical approval
This study was carried out in strict accordance with the recommendations in the Animal Care and Protection Act

Bacterial strains and growth conditions
The UPEC strains used in this study are listed in Table 1. Wildtype clinical isolates and their derivatives were routinely cultured at 28uC or 37uC on solid or in liquid Luria-Bertani (LB) medium [88] supplemented with the appropriate antibiotics: streptomycin (Strep, 100 mg/ml), kanamycin (Kan, 50 mg/ml), tetracycline (Tet, 5 mg/ml). Streptomycin-resistant derivatives of UPEC strains CFT073, 1177 and RS218 were generated by sequential growth on increasing concentrations of streptomycin (10 mg/ml, 25 mg/ ml, 50 mg/ml and 100 mg/ml).

DNA manipulations and genetic techniques
Oligonucleotides used in this study were sourced from Sigma-Aldrich (Australia). Plasmid pRedET and the rpsL-neo template were obtained from the Counter selection BAC modification kit (GeneBridges). Isolation of plasmid DNA was carried out using the QIAprep Spin Miniprep Kit (Qiagen). All polymerase chain reactions were performed using Taq DNA polymerase (New England Biolabs) under standard conditions except for reactions requiring high fidelity, in which case Phusion DNA polymerase (Thermo Scientific) was used. DNA sequencing for mutant and complement validation was performed using the Big Dye v3.1 kit (Applied Biosystems) at the Australian Equine Genetics Research Centre, University of Queensland.

Construction of capsule and O antigen insertion mutants
Primers were designed using Vector NTI Advance 10 software (Invitrogen). Due to the unavailability of the RS218 and E. coli 1177 genome sequence, sequences of kpsD and waaL genes from UPEC strains CFT073, APEC01, 536 and UTI89 were aligned to identify 100 bp regions of complete sequence conservation. These sequences were incorporated as the 50 bp homology regions at the 59 end of the forward and reverse knockout primers. The 39 end of the primers contained the 24 base sequences specified in the Counter selection BAC modification kit (GeneBridges), which were used to amplify the rpsL-neo cassette. The primer sequences used are as follows: kpsD forward (F): TCCAACGTCAACGTC-TACGCCTCCTTATTACAGGCGCAGCCAGTAAAAGTGG-CCTGGTGATGATGGCGGGATCG; kpsD reverse (R): ACGT-CACACCGCCATACAGACCAGGATTACGCACAAATCCG-GTCACGTACTCAGAAGAACTCGTCAAGAAGGCG; waaL F: CGCGTGCAGCCATATTAGTGTTTCCATTCTTTGCG-TTACTATTAATCGTAGGCCTGGTGATGATGGCGGGA-TCG; waaL R: ATAAAACAATATAACTTATAATTAATTCG-CTTATTAATATAACTATCCATTCAGAAGAACTCGTCAA-GAAGGCG. Screening primers were designed at least 70 bp upstream and downstream of the homology regions, respectively. Their sequences are as follows: kpsD screen F: GATGGTGCGT-TACAGGTTGATC; kpsD screen R: CCGCGCTTGACCA-CAATATC; waaL screen F: CAACAGCAGGCTACATCATCC; waaL screen R: CTTCATACATTGCCAGACGG. Streptomycin resistant UPEC strains CFT073, 1177 and RS218 were transformed with pRedET by electroporation and selected in the presence of tetracycline. These strains were then grown at 28uC to an optical density at 600nm (OD 600 nm ) of 0.3, induced with L-arabinose (0.3% w/v) for approximately 1 h to allow expression of the recombinase genes, and then electroporated with the rpsLneo cassette bearing the 50 bp homology regions. Mutants that were streptomycin sensitive (Strep S ) and kanamycin resistant (Kan R ) were identified and insertion of the rpsL-neo cassette within the kpsD and waaL genes, respectively, was confirmed by PCR and sequencing.

Complementation of capsule and O antigen insertion mutants
Plasmid pRedET was re-introduced into the CFT073, 1177, and RS218 waaL and kpsD mutants, respectively. To repair the waaL and kpsD mutations, these strains were prepared as described above and transformed with 2 mM single-stranded 100 base oligonucleotides comprising a sequence identical to the WT strain in the mutated region (kpsD:TCCAACGTCAACGTCTACG-CCTCCTTATTACAGGCGCAGCCAGTAAAAGTGTACG-TGACCGGATTTGTGCGTAATCCTGGTCTGTATGGC-GGTGTGACGT; waaL:CGCGTGCAGCCATATTAGTGTT-TCCATTCTTTGCGTTACTATTAATCGTAATGGATAGT-TATATTAATAAGCGAATTAATTATAAGTTATATTGTTT-TAT). Repair of the specific waaL or kpsD mutation in Strep R colonies was confirmed by PCR screening and sequencing. K1 phage assay K1 phage solution was sourced from the Statens Serum Institut (Denmark). The cross-brush method recommended by the manufacturer was used to determine the presence of K1 antigen [89]. Briefly, a vertical line of phage suspension was streaked across an agar plate and allowed to dry. Horizontal lines of overnight E. coli broth cultures were then streaked from left to right across the phage suspension and plates were incubated at 37uC for 12 h. Inhibited growth or complete killing of bacteria after crossing the phage suspension was deemed a positive reaction. A negative reaction, in the absence of K1, was characterized by normal bacterial growth across the entire bacterial streak. CFT073 WT strain with K2 capsule was used as a negative control for the assay.
Counter current immunoelectrophoresis E. coli diagnostic K2 antisera (Statens Serum Institut, Denmark) was used to detect the presence of the K2 antigen. Capsule extracts were made according to manufacturer guidelines. Briefly, the bacterial cells from an agar plate were suspended in 1 ml phosphate buffered saline (PBS) (pH 7.4) and incubated at 60uC for 20 min. The cell suspension was then centrifuged for 30 min at 14,000 rpm and the supernatant transferred to a fresh tube and heated for 1 h at 95uC. Extracts were then diluted in PBS (1:50) for electrophoresis. A 1.5% agarose gel was applied to a glass plate with 2 rows of wells and a distance of 5 mm between each row. 15 ml of K2 antisera and diluted capsule extracts were loaded at the anode and cathode ends, respectively. Counter-current immunoelectrophoresis was performed for 70 min at 80 V in 1X TAE buffer. A precipitin band between the wells demonstrated the presence of a K2 capsule.
LPS isolation and silver staining LPS was extracted as described previously by West et al. [90]. Tricine-SDS PAGE (TSDS PAGE) was used to obtain a high resolution of LPS. Gels were cast with a 16.5% resolving phase containing 4 M urea, overlaid with a 4% stacking phase. A twobuffer system was used for electrophoresis as previously described [91]. Electrophoresis was performed in the Mini-PROTEAN gel apparatus (BioRad) at 50 V for 30 min and then at 150 V for 70 min. A modified silver staining method originally described by Tsai and Frasch [92] was used to view LPS bands. Gels were transferred to a glass tray and fixed overnight (40% ethanol, 5% glacial acetic acid) at room temperature. The fixing solution was replaced with an oxidizing solution (40% ethanol, 5% glacial acetic acid, 0.7% periodic acid) and incubated for 7 min. Gels were washed repeatedly in copious amounts of milliQ water and then stained (20 mM sodium hydroxide, 0.4% ammonium hydroxide, 0.7% silver nitrate) for 10 min. The staining solution was removed and gels were incubated in the dark in a developing solution (0.02% formaldehyde, 0.26 mM citric acid) until distinct bands were visible. Gels were then washed in water to stop the reaction.

Serum killing assay
Overnight bacterial cultures were washed in PBS and then standardized to a final OD 600 = 0.8 (approximately 4610 8 CFU/ ml). A volume of 50 ml of each standardized culture was mixed with an equal volume of PBS and serially diluted for colony counts on agar to enumerate CFU prior to serum treatment (t = 0 min). Fresh blood was collected in EDTA-free vacutainer tubes on the day of the assay from two healthy human donors who gave informed consent. The same donors were used for all assays in order to reduce serum variability. Blood was allowed to clot for 15 min and the tubes were centrifuged at 1500 g for 10 min. A volume of 50 ml of each standardized culture (approximately 3610 7 CFU) was mixed with an equal volume of pooled human sera, mixed and incubated for 90 min at 37uC. All experiments were performed in triplicate. Colony counts were used to enumerate bacterial cells post treatment (t = 90 min). E. coli MG1655 was used as a serum sensitive control strain. All the strains were tested similarly in a parallel assay using heatinactivated normal human sera (Millipore). Strains showing at least a log 10 difference (i.e. 10-fold) in CFU/ml were deemed to have altered sensitivity to serum as compared to WT. Serum sensitivity indices (SSI) were calculated according to these differences. Paired samples t tests were performed on CFU/ml values at the two time points (0 min and 90 min) to determine statistical significance.

Whole blood killing assay
Approximately 10 4 colony forming units (CFU) of UPEC prepared from a late exponential phase culture in LB were added to 1.5 ml of whole blood collected from a healthy human volunteer using plasma preparation tubes (PPT) containing 9.0 mg K2EDTA (BD Vacutainer, NJ, USA). The infected cultures were incubated rolling at 37uC and after 3-5 h diluted blood samples were plated in triplicate onto agar to enumerate the surviving bacteria. The survival rate was calculated as a percentage of the starting inoculum ([CFU (at a given time point)/CFU (at the start)]6100) and P values were determined using the independent samples t test.

Mouse model of UTI
The mouse model of UTI described by Roos et al. (2006) was used in this study [93]. Female C57BL/6 mice (8-10 weeks) were purchased from the University of Queensland Animal Facility and housed in sterile cages with ad libitum access to sterile water. All strains used in this experiment were enriched for type 1 fimbriae expression after 3 passages under static conditions. Bacterial cultures were prepared by overnight static growth in LB broth; all strains were strongly positive for type 1 fimbriae expression as determined by yeast cell agglutination. For competitive (mixed infection) assays, inocula contained 1:1 strain mixtures of CFT073kpsD and CFT073 WT (group 1) or CFT073kpsD C (complement; group 2); CFT073waaL and CFT073 WT (group 3) or CFT073waaL C (group 4). An inoculum of 20-40 ml, containing 5610 8 CFU of bacteria in PBS was injected directly into the bladder of each mouse. Urine was collected from mice 18 h after challenge following which they were euthanized by cervical dislocation. Bladders were then excised aseptically, weighed, and homogenized in PBS. Urine and bladder homogenates were serially diluted in PBS and plated onto streptomycin and kanamycin-containing LB agar for colony counts. The two strains in each group were differentiated by their resistance to streptomycin (WT and complements) or kanamycin (kpsD and waaL mutants). Significance was determined using the two-tailed Wilcoxon matched pairs test on CFU/ml urine, or CFU/0.1 g bladder tissue using Prism6 (GraphPad). P values ,0.05 were considered significant.