Discovery of New Genes Involved in Curli Production by a Uropathogenic Escherichia coli Strain from the Highly Virulent O45:K1:H7 Lineage

ABSTRACT Curli are bacterial surface-associated amyloid fibers that bind to the dye Congo red (CR) and facilitate uropathogenic Escherichia coli (UPEC) biofilm formation and protection against host innate defenses. Here we sequenced the genome of the curli-producing UPEC pyelonephritis strain MS7163 and showed it belongs to the highly virulent O45:K1:H7 neonatal meningitis-associated clone. MS7163 produced curli at human physiological temperature, and this correlated with biofilm growth, resistance of sessile cells to the human cationic peptide cathelicidin, and enhanced colonization of the mouse bladder. We devised a forward genetic screen using CR staining as a proxy for curli production and identified 41 genes that were required for optimal CR binding, of which 19 genes were essential for curli synthesis. Ten of these genes were novel or poorly characterized with respect to curli synthesis and included genes involved in purine de novo biosynthesis, a regulator that controls the Rcs phosphorelay system, and a novel repressor of curli production (referred to as rcpA). The involvement of these genes in curli production was confirmed by the construction of defined mutants and their complementation. The mutants did not express the curli major subunit CsgA and failed to produce curli based on CR binding. Mutation of purF (the first gene in the purine biosynthesis pathway) and rcpA also led to attenuated colonization of the mouse bladder. Overall, this work has provided new insight into the regulation of curli and the role of these amyloid fibers in UPEC biofilm formation and pathogenesis.

IMPORTANCE Uropathogenic Escherichia coli (UPEC) strains are the most common cause of urinary tract infection, a disease increasingly associated with escalating antibiotic resistance. UPEC strains possess multiple surface-associated factors that enable their colonization of the urinary tract, including fimbriae, curli, and autotransporters. Curli are extracellular amyloid fibers that enhance UPEC virulence and promote biofilm formation. Here we examined the function and regulation of curli in a UPEC pyelonephritis strain belonging to the highly virulent O45:K1:H7 neonatal meningitis-associated clone. Curli expression at human physiological temperature led (TraDIS) to identify genes involved in curli biogenesis during growth at 37°C. Our screen identified several novel genes, including genes involved in purine de novo biosynthesis and two new regulators, all of which were confirmed to affect curli production through the generation of specific mutants and complementation. Finally, the importance of these genes to UPEC virulence was demonstrated in a mouse UTI model.

RESULTS
MS7163 produces curli during growth at 37°C and binds Congo red. MS7163 is a curli-producing UPEC pyelonephritis strain resistant to LL-37 (24). In line with this phenotype, MS7163 bound strongly to CR following growth at both 37°C and 28°C on YESCA-CR agar (see Materials and Methods). Complete genome sequencing of MS7163 revealed it belongs to the E. coli B2 phylogenetic group and multilocus sequence type 95 (ST95). The MS7163 genome comprises a chromosome containing 5,074,586 nucleotides and two large plasmids (pMS7163A, 133,843 bp; pMS7163B, 84,078 bp) (see Fig. S1, Fig. S2, and Data Set S1 in the supplemental material). Whole-genome phylogenetic analysis of MS7163 and other completely sequenced B2 strains (listed in Data Set S2 in the supplemental material) revealed MS7163 clusters in a clade with the reference ST95 strains UTI89, a cystitis strain (48), and APEC-O1, an avian-pathogenic E. coli strain (49), and in a subclade with the neonatal meningitis E. coli (NMEC) strain S88 (50) (Fig. 1). Because curli production and CR binding by E. coli are frequently associated with the synthesis of cellulose, we examined the MS7163 genes involved in cellulose regulation and biosynthesis. MS7163 possessed a mutation in the essential cellulose production gene bcsA (with a G deletion at nucleotide 556 of bcsA causing its premature termination), leading us to predict that it is unable to produce cellulose. To demonstrate this, we mutated the csgA major curli subunit gene (to generate the mutant MS7163csgA) and examined the phenotypes of MS7163 and MS7163csgA on YESCA-CR agar. In contrast to wild-type (wt) MS7163, MS7163csgA did not bind CR, confirming that the production of curli (and not cellulose) is specifically associated with this phenotype (Fig. 2A).
Curli mediate biofilm formation and resistance of sessile cells to LL-37. To investigate the role of curli in virulence, we assessed the ability of MS7163 and MS7163csgA to form a biofilm in a microtiter plate assay employing two different  Table S2) was reconstructed using kSNP with a k-mer size of 21 and rooted using the E. coli phylogroup D strain UMN026 as an outgroup. Branches were colored according to their ST types (Achtman MLST scheme). MS7136 (boldface) clustered into the same group of E. coli ST95 and was closely related to NMEC strain S88.
Curli Biosynthesis in Uropathogenic Escherichia coli ® growth media. In LB, a condition under which curli are not produced, the ability to form a biofilm was similar for both strains (P ϭ 0.0994) (Fig. 2B). In contrast, growth in LB without salt, a condition known to induce curli production, resulted in stronger biofilm formation by MS7163 than MS7163csgA (P Ͻ 0.0001) (Fig. 2B). To specifically investigate how curli-dependent biofilm formation by MS7163 impacts its resistance to LL-37, we allowed MS7163 to form a biofilm by growth in LB broth without salt, separated planktonic (free-swimming) and sessile (biofilm) cells, and tested their susceptibility to LL-37. In these experiments, following a 2-h incubation in the presence of 64 g/ml of LL-37, most planktonic cells were killed (survival rate of 0.07%), while sessile cells exhibited significant resistance (survival rate of 30.4%; P Ͻ 0.0001) (Fig. 2C).
Curli enhance colonization of the mouse urinary tract. The role of curli in virulence was also assessed by examining the ability of MS7163 and MS7163csgA to survive in the mouse urinary tract in a competitive colonization experiment. Mice were coinoculated with MS7163 and MS7163csgA strains in a 1:1 ratio. In this mixed infection assay, MS7163 significantly outcompeted MS7163csgA in colonization of the bladder (P ϭ 0.0078) (Fig. 2D). Together with our biofilm and LL-37 resistance data, these results demonstrate an important role for curli in virulence.
Large-scale screening for curli-deficient mutants. We devised a screen based on CR binding in combination with transposon mutagenesis and TraDIS to define the Logarithmic values of competitive indices are plotted on the y axis; horizontal bars represent median values. A log 10 fitness index below 0 (shown by the dashed line) indicates that MS7163csgA was at a competitive disadvantage. Statistical significance was determined by the two-tailed Wilcoxon's signed rank test (*, P Ͻ 0.05). genetic basis of curli production by MS6173 at 37°C. MS7163 was subjected to transposon mutagenesis using a mini-Tn5 cassette carrying a chloramphenicol resistance gene, and mutants were screened on YESCA-CR agar supplemented with chloramphenicol. Under these conditions, mutants that did not make curli failed to bind CR and were white, mutants that made reduced levels of curli bound CR weakly and were light red, and mutants that were unaltered in their level of curli production bound CR and were red. In total,~132,000 mini-Tn5 mutants were screened, from which 71 mutants were identified that were "white" and unable to bind CR and 246 mutants were "light red" and exhibited reduced binding to CR. All of these mutants retained their CR binding phenotype over three successive rounds of subculture on YESCA-CR agar.
Identification of curli mutants using TraDIS. The above mutants were examined by TraDIS to enable en masse identification of the insertion sites that led to altered CR binding. Analysis of the pooled 246 light red colonies generated 1,004,719 sequence reads that mapped to 105 unique mini-Tn5 insertion sites on the MS7163 chromosome. These insertion sites were further localized to 22 genes and 6 intergenic regions, and included genes involved in LPS biogenesis (waaGHILPQ, wzxE, and rfaH), regulation (envZ and rpoS), metabolism (pgm, galE, pgi, and fbp), sodium transport (nhaA), and septum formation (nlpD and amiB) (see Table S1 in the supplemental material). Similarly, analysis of the pooled 71 white colonies generated 238,182 sequence reads that mapped to 67 unique mini-Tn5 insertion sites. These insertion sites were further localized to 19 genes and 3 intergenic regions (Table 1). Among the 19 genes, nine (47.4%) were well characterized with respect to curli production; these included three genes encoding components of the curli assembly machinery (csgA, csgB, and csgG) and six genes associated with curli regulation (mlrA, ihfA, ihfB, dksA, rne, and ompR). The remaining ten genes were novel or poorly characterized with respect to curli synthesis, and thus we focused the remainder of our investigation on these genes. Disruption of genes involved in purine de novo biosynthesis prevents curli production. Six of the 19 genes identified in our TraDIS analysis (31.6%) form part of the purine de novo biosynthesis pathway: i.e., purF, purD, purL, purM, purK, and purH ( Table 1). The purine de novo biosynthesis pathway comprises 10 genes for the conversion of 5-phospho-␣-D-ribose 1-diphosphate (PRPP) to ribosylhypoxanthine monophosphate (IMP), the precursor for AMP and GMP (Fig. 3A). To validate this finding, we examined the involvement of selected purine biosynthesis genes in curli production using defined knockout mutants. Four pur mutants were generated by -Red-mediated homologous recombination: MS7163purF, MS7163purD, MS7163purM, and MS7163purK. Compared to wt MS7163, all four pur mutants were unable to bind CR following growth on YESCA-CR agar and failed to express the CsgA major subunit protein based on Western blot analysis of whole-cell lysates employing a CsgA-specific antibody ( Fig. 3B and C). Furthermore, complementation of the corresponding gene in each pur mutant restored curli production ( Fig. 3B and C). Finally, supplementation with IMP, the last product of the purine de novo biosynthesis pathway, restored curli production in MS7163purM and MS7163purK (Fig. 3D). Taken together, these results demonstrate that purine de novo biosynthesis plays a critical role in curli production.
Disruption of yrfF prevents curli production. TraDIS analysis identified seven independent insertions within the yrfF gene and two insertions immediately upstream of the yrfF coding region, strongly suggesting the involvement of YrfF in curli production. YrfF, also known as IgaA in Salmonella enterica, is a negative regulator of the Rcs phosphorelay system, which controls genes involved in colanic acid production (51,52). Mutation of yrfF causes overexpression of the Rcs system, leading to increased colanic acid synthesis (53). To confirm the role of yrfF in curli production, we mutated the yrfF  4A). In addition, complementation of MS7163yrfF with a plasmid containing the yrfF gene (pYrfF) led to loss of mucoidy and restoration of curli production, demonstrated by CR binding and anti-CsgA Western blot analysis ( Fig. 4A and B). We also examined the transcription of colanic acid regulatory and biosynthesis genes (rcsCDB-A-F and wcaH) and selected curli genes (csgD and csgA) in wt MS7163 compared to MS7163yrfF and the complemented strain MS7163yrfF(pYrfF). Consistent with our phenotypic data, the transcription of colanic acid genes was increased and curli genes decreased in MS7163yrfF, while complementation restored the transcription of these genes to approximately wt level (Fig. 4C). Taken together, these results demonstrate that yrfF plays an opposite regulatory role in curli and colanic acid production, respectively.
Overexpression of the novel regulator RcpA abrogates curli production by repressing the transcription of csgD and csgA. Our TraDIS analysis also identified an enrichment of three independent mini-Tn5 insertions in the region upstream of a coding sequence (CDS) identified by the locus tag MS7163_03831 (at bp Ϫ23, Ϫ25, and Ϫ26 relative to the MS7163_03831 predicted ATG start codon) (Fig. 5A). In all three cases, the mini-Tn5 cassette was inserted with the chloramphenicol resistance gene in the same orientation as the MS7163_03831 CDS, referred to henceforth as a repressor of curli production, or rcpA. We hypothesized that the mini-Tn5 insertions led to enhanced transcription of the rcpA gene via readthrough from the promoter of the chloramphenicol resistance gene within the mini-Tn5, leading to increased expression of RcpA and abolishment of curli production. We tested this by examining the mini-Tn5 mutant with an insertion at position Ϫ25 relative to the predicted rcpA start codon (referred to as strain Tn5-03831). Indeed, quantitative reverse transcription-PCR (qRT-PCR) analysis revealed that rcpA transcript levels in Tn5-03831 were increased~3.2-fold compared to that in wt MS7163 (Fig. 5B). In contrast, transcript levels of csgD and csgA Western blot analysis of CsgA performed using whole-cell lysates prepared from MS7163, MS7163yrfF, and MS7163yrfF(pYrfF). Bacteria were grown on YESCA agar for 24 h to induce curli production and treated with formic acid to dissolve polymerized CsgA. Anti-OmpA antibody was used as a loading control. (C) Impact of yrfF on the transcription of genes involved in curli biosynthesis, the Rcs phosphorelay system, and colanic acid production. Shown are the relative fold changes in transcript level of curli genes (csgD and csgA), Rcs phosphorelay system genes (rcsCDB-A-F), and the colanic acid biosynthesis gene wcaH in MS7163yrfF and the complemented MS7163yrfF(pYrfF) strain compared to wt MS7163 as determined by qRT-PCR. Total mRNA was extracted from bacteria grown on YESCA agar at 37°C for 24 h. Results are displayed as the mean log 2 FC with standard deviation from three independent replicates.
Curli Biosynthesis in Uropathogenic Escherichia coli ® in Tn5-03831 decreased~70-and~4,000-fold, respectively. The abolishment of curli production by Tn5-03831 was also confirmed by its inability to bind CR and anti-CsgA Western blot analysis ( Fig. 5C and D).
RcpA is a small 282-bp CDS that encodes a putative 93-amino-acid protein of unknown function. To demonstrate that increased expression of RcpA in Tn5-03831 abolished curli production, and rule out the possibility of polar effects or secondary mutations, we generated two types of RcpA-overexpressing strains: (i) PcL-rcpA, which contained a strong constitutive promoter (PcL) inserted immediately upstream of the rcpA CDS on the MS7163 chromosome and (ii) MS7163(pRcpA), which contained the rcpA gene cloned behind the lac promoter in plasmid pSU2718 (Fig. 5A). The transcript level of rcpA was strongly increased in PcL-rcpA and MS7163(pRcpA) compared to wt MS7163 [log 2 fold changes (FCs) of 5.5 and 11.4 in PcL-rcpA and MS7163(pRcpA), respectively] (Fig. 5B). In contrast, the transcription of csgA and csgD was significantly decreased in both strains [csgA, log 2 FCs of Ϫ13.9 and Ϫ7.3 in PcL-rcpA and MS7163(pRcpA), respectively; csgD, log 2 FCs of Ϫ8.5 and Ϫ6.6 in PcL-rcpA and MS7163(pRcpA), respectively]. Both PcL-rcpA and MS7163(pRcpA) did not produce curli based on CR binding and anti-CsgA Western blot analysis (Fig. 5C and D).

Repression of curli production by RcpA is not strain specific.
To extend our findings on the role of RcpA in curli production, we examined the impact of overexpression of RcpA in UPEC strain UTI89, which produces curli during growth at 28°C on YESCA-CR agar (12,20). Transformation of UTI89 with plasmid pRcpA led to reduced binding to CR (Fig. 6A) and abolishment of CsgA production compared to wild-type UTI89 following growth on YESCA agar (Fig. 6B). Taken together, these results demonstrate that overexpression of RcpA causes repression of curli production, and this function is conserved in different UPEC strains.
Mutation of purF and overexpression of RcpA attenuate colonization of the mouse urinary tract. Given the impact of mutations in the purine biosynthesis pathway and overexpression of RcpA on curli production, we tested the ability of MS7163purF and PcL-rcpA to survive in the mouse urinary tract using a competitive infection assay. Female C57BL/6 mice were infected with a 1:1 ratio of MS7163 to MS7163purF or MS7163 to PcL-rcpA cultured statically in LB without salt, respectively, and colonization was assessed at 1 day postinfection. In these experiments, both the MS7163purF and PcL-rcpA mutants were significantly outcompeted by wt MS7163 in the bladder and urine of infected mice (Fig. 7). No significant colonization of the kidneys was observed. Taken together, these data demonstrate that disruption of the purine de novo biosynthesis pathway and overexpression of RcpA, both of which abrogate curli production, result in attenuated colonization of the mouse urinary tract.

DISCUSSION
Recent literature has established an increasing appreciation for the role of curli in UPEC virulence (20,21,23,47,54). These amyloid fibers are produced by many UPEC strains at 37°C, mediate strong biofilm formation, confer resistance to the soluble cationic peptide LL-37, and enhance the production of proinflammatory mediators. Here, we characterized the role of curli in the virulent UPEC strain MS7163 and employed a forward genetic screen to identify new genes involved in the production of curli.
MS7163 was isolated in Slovakia from a patient with severe pyelonephritis (24). Whole-genome sequencing of MS7163 revealed it is most closely related to the neonatal meningitis-causing E. coli (NMEC) strain S88. E. coli S88 was isolated from the cerebrospinal fluid of a neonate and belongs to the highly virulent O45:K1:H7 clone, which accounts for one-third of all neonatal meningitis cases in France (50). Based on in silico analysis, MS7163 also possesses this serotype, and both strains share a large plasmid (pMS7163A and pECOS88, respectively) that harbors multiple virulence genes, including three iron uptake systems. Our analysis of the MS7163 genome revealed it contains a mutation in the essential cellulose synthase gene bcsA that results in the inability to produce cellulose. Intriguingly, S88 also contains exactly the same mutation, Curli Biosynthesis in Uropathogenic Escherichia coli ® a G deletion at nucleotide 556 of bcsA that causes its premature termination. Thus, we speculate that the production of curli, together with the inability to produce cellulose, might represent a previously unrecognized virulence feature of the O45:K1:H7 lineage. Given that the potent immune response stimulated by curli through the activation of interleukin-6 (IL-6) and IL-8 is amplified in the absence of cellulose (21,23,31), this phenotype is consistent with our hypothesis. Indeed, a link between the curli-positive, cellulose-negative phenotype and enhanced disease severity has also been demonstrated in the highly virulent E. coli O104:H4 outbreak strain that infected nearly 4,000 people and caused 54 deaths in Germany in 2011 (43).
Growth of MS7163 in LB without salt at 37°C promoted strong biofilm formation, which was abrogated in non-curli-inducing medium (LB broth) and by mutation of the csgA gene. We also demonstrated that sessile MS7163 cells were highly resistant to LL-37 compared to their free-swimming planktonic counterpart cells, thus providing direct experimental evidence that curli-mediated biofilm formation significantly enhances LL-37 resistance. LL-37 is a soluble antimicrobial peptide that forms an integral component of the host innate defense system of the urinary tract (23). Thus, curli production would be expected to enhance colonization of the urinary tract, and we demonstrated this in a mixed competitive infection model, where wt MS7163 outcompeted MS7163csgA at 24 h postinfection of the mouse bladder. These results were also consistent with previous data demonstrating a role for curli in the early stages (6 h postinfection) of bladder colonization in experimental mice (20).
The application of transposon mutagenesis together with TraDIS represents a powerful combinatorial approach to dissect genetic pathways linked to specific phenotypes (55)(56)(57). In this study, we screened~132,000 mini-Tn5 mutants by using CR binding following growth on YESCA-CR agar as a proxy for curli production. Based on the size of the MS7163 genome and our previous analysis of unique insertion sites in  other TraDIS libraries (55,57), we estimate this represents a coverage of approximately one insertion every 80 to 100 bp. Overall, two types of mutants were identified: those that were unable to bind CR (white colonies) and those that displayed reduced binding to CR (light red colonies). Using TraDIS, we then mapped the mini-Tn5 insertion sites of these mutants en masse, leading to the identification of 19 genes that were absolutely required for curli production and a further 22 genes that impacted curli production. Many of the genes identified in our screen overlapped with a recent large-scale screen of the E. coli K-12 Keio library (58). In addition to known structural and regulatory genes, these included genes involved in purine biosynthesis, lipopolysaccharide (LPS) biogenesis, stress, and stationary-phase regulation, metabolism, sodium transport, and septum formation. Despite this, we note there were also differences in the total list of genes identified in both studies (see Table S2 in the supplemental material), suggesting strain-specific variation, as well as differences in growth and temperature conditions, may account for divergence in the complex pathway of curli production.
We focused our experiments on genes that were absolutely required for curli production, all of which led to a "white" colony phenotype on YESCA-CR agar. Of the 19 genes identified in this category, three were integral to the curli operon (csgA, csgB, and csgG), and disruption of these genes has previously been shown to abolish curli production in E. coli (25)(26)(27)(58)(59)(60). We did not identify the curli gene csgC; however, this is consistent with other reports that show its mutation does not affect CR binding (26,29,58,(61)(62)(63). In addition, we did not find insertions in the curli chaperone genes csgE and csgF in our screen. Mutation of these genes has previously been linked to reduced CR binding, resulting in a light red colony phenotype following growth at Յ28°C on YESCA-CR agar (26,29,58,(61)(62)(63). It is possible that mutation of these genes in MS7163 led to a subtle change in CR binding phenotype that was missed in our screen or even that there were no insertions in either gene in our library. Interestingly, our screen also failed to identify the csgD regulator gene. We generated a targeted csgD mutant in MS7163 using -Red-mediated homologous recombination and confirmed it does not produce curli (data not shown). Closer examination of csgD revealed it contains a lower GC content (41.8%) than the entire MS7163 genome (50.7%), which could explain its absence from our screen as the mini-Tn5 transposon preferentially inserts into GC-rich regions (64). This has been observed in two previous TraDIS studies, where an increased mini-Tn5 transposon insertion frequency was detected in high-GC versus low-GC regions (57,65). Consistent with current knowledge of curli gene control (35,36,42), we also identified mlrA (which encodes a master regulator of CsgD), ihfA and ihfB (which together encode IHF), and ompR (which encodes a regulator of envelope stress) as positive regulators of curli production, demonstrating their regulatory function is not subjected to temperature control.
In UPEC, purine de novo biosynthesis is required for IBC growth in the mouse bladder (66). Our TraDIS screen identified six genes within the purine biosynthesis operon that were required for curli production, and we confirmed the role of four of these genes (purF, purD, purM, and purK) by the construction of defined mutants and complementation. We hypothesize that disruption of de novo purine biosynthesis results in a decrease in the level of cyclic-di-GMP in the cell. Since cyclic-di-GMP activates the transcription of the csgD curli regulator gene (67), this would lead to abolishment of curli production in a pur-deficient background. Genes involved in purine biosynthesis were also recently shown to be required for curli production by E. coli K-12 (58), and mutation of purF attenuates the ability of UPEC UTI89 to colonize the bladder of C3H/HeN mice (66). Our study demonstrated attenuated bladder colonization by MS7163purF in C57BL/6 mice in a mixed competitive colonization experiment. The log 10 competitive index (CI) of wt MS7163 versus MS7163purF (Ϫ1.45) was significantly lower than the log 10 CI of wt MS7163 versus MS7163csgA (Ϫ0.96; P ϭ 0.04), suggesting that the attenuated colonization of MS7163purF may be due to the combined effect of reduced fitness via disruption of the purine de novo biosynthesis pathway and the inability to produce curli.
The yrfF gene was also identified in our screen, and we showed its mutation abolishes curli production. YrfF, known as IgaA in Salmonella enterica serovar Typhimurium, is a negative regulator of the Rcs phosphorelay system, a complex signal transduction pathway that controls the production of colanic acid (51,52,68). The Rcs phosphorelay system consists of the histidine sensor kinase RcsC, the phosphoryl group transporter RcsD, and the response regulator RcsB (69). The outer membrane lipoprotein RcsF is also an integral activator of this system and functions by interacting with RcsC and YrfF, both of which are located in the cytoplasmic membrane (51,52,70). The yrfF gene has been described as essential in several studies (51,52,55,68,71), as its mutation results in overactivation of RcsCDB and uncontrolled colonic acid biosynthesis (53,69,72). This lethal phenotype can be overcome by secondary suppressor mutations in rcsB, rcsC, and rcsD, all of which affect the Rcs system (73). Here, we provide direct evidence for the role of YrfF in curli production through our screen, which identified seven independent insertions in yrfF that abrogated CR binding, as well as the construction of a defined mutant (MS7163yrfF) and its complementation. We also demonstrated an increase in the transcription of genes in the Rcs system, together with a corresponding decrease in curli gene transcription, in MS7163yrfF. This result is consistent with studies in E. coli K-12 that showed the Rcs system negatively controls transcription of the csg genes (41). Although yrfF was not essential in MS7163, we were unable to test the colonization of MS7163yrfF in the mouse bladder, as static culture to enrich for the production of type 1 fimbriae led to spontaneous loss of its mucoid and noncurliated phenotype (see Table S3 in the supplemental material).
We have shown previously that the promoter of the chloramphenicol resistance gene in our mini-Tn5 transposon can drive the transcription of a downstream gene if the insertion position is favorable (57,74). This led to the identification of the rcpA gene as a novel repressor of curli production. We confirmed this function by overexpression of RcpA using two independent systems: one based on insertion of a constitutive PcL promoter upstream of rcpA to drive strong transcription of the gene (PcL-rcpA) and a second based on plasmid-mediated overexpression [MS7163(pRcpA)]. In both cases, overexpression of RcpA led to significantly decreased transcription of curli genes (csgD and csgA), abolishment of CsgA expression, and the inability of the modified strains to bind CR. The PcL-rcpA strain was also attenuated in its capacity to colonize the mouse bladder in a mixed infection assay.
The rcpA gene encodes a protein 93 amino acids in length. Small proteins, notably those less than 100 amino acids, are frequently ignored due to difficulties in annotation and biochemical detection (75). However, examples of such proteins that localize to the inner membrane and regulate membrane-associated two-component sensor-kinase systems have been described (75). One of the best-characterized small proteins in E. coli is MgrB, a hydrophobic inner membrane protein that functions as a negative-feedback regulator of the PhoP/PhoQ system (76). Analysis of the translated RcpA protein sequence did not identify any known conserved domains or a putative signal sequence. However, using TMHMM2.0 (77), the RcpA protein is predicted to contain two transmembrane helices (amino acids 7 to 26 and 31 to 50) and a cytoplasmic C-terminal domain (amino acids 51 to 93). These properties, in addition to its high hydrophobic amino acid ratio, suggest RcpA may localize to the inner membrane like MgrB. Our analyses showed repression of curli synthesis by overexpression of RcpA was not strain specific and also occurred in UTI89. While the precise mechanism by which RcpA represses curli synthesis remains to be elucidated, we demonstrated its overexpression led to decreased curli gene transcript levels. The rcpA gene is conserved in E. coli, Shigella, and Achromobacter, suggesting its function as a negative regulator of curli synthesis may be broadly conserved.
In summary, this study describes the complete genome sequence of UPEC MS7163, a curli-producing pyelonephritis strain. MS7163 belongs to the highly virulent O45: K1:H7 clone, and we provide the first description of curli production at 37°C by a strain from this lineage. We also describe the identification of several new genes involved in the production of curli and demonstrate their role in UPEC pathogenesis in experimen-tal mice. Further work is now required to understand the precise molecular mechanisms by which these genes impact curli synthesis and the development of UTI.

MATERIALS AND METHODS
Ethics statement. Approval for mouse infection studies was obtained from the University of Queensland Animal Ethics Committee (SCMB/242/16/NHMRC). Experiments were carried out in strict accordance with the recommendations in the Animal Care and Protection Act (Queensland, 2002) and the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes (8th ed., 2013).
Strains, growth conditions, and genome sequencing. Bacterial strains and plasmids are listed in Data Set S2. Strains were grown at 37°C on solid or liquid lysogeny broth (LB) medium. To induce curli production, strains were cultured on YESCA agar (10 g/liter Casamino Acids, 1 g/liter yeast extract, 20 g/liter agar) or LB without salt supplemented with 50 g/ml CR and 1 g/ml Coomassie brilliant blue as indicated. When required, chloramphenicol (30 g/ml), gentamicin (20 g/ml), or IMP (0.5 mM) was added to the medium. Methods for MS7163 genome sequencing and analysis are described in Text S1 in the supplemental material.
Transposon mutagenesis and TraDIS. Generation of the transposon mutant library was performed essentially as previously described (55). MS7163 mini-Tn5 mutants were plated directly onto YESCA-CR agar supplemented with chloramphenicol and incubated at 37°C for 24 h. Under these growth conditions, CR binding was used as an indicator for curli production: curliated colonies were red, mutants that produced less curli were light red, and mutants that did not produce curli were white. Mutants with an altered CR binding phenotype (i.e., white or light red colonies) were restreaked onto YESCA-CR agar over three successive rounds of subculture to confirm their phenotype and then stored individually in 20% glycerol at Ϫ80°C. Multiplex TraDIS analysis of pooled mutants was performed using the Illumina MiSeq platform as previously described (65). Tn5-specific reads were identified by FASTX-Toolkit (v.0.0.13) and mapped to the MS7163 complete genome using MAQ (v0.7.1) (78).
Targeted gene mutation and complementation. Specific mutants were generated using -Redmediated homologous recombination as previously described (55,79). Where necessary, the chloramphenicol resistance cassette was removed using plasmid pCP20 (80). Strain PcL-rcpA was generated using a previously described methodology (81). Plasmids for complementation were generated by PCR amplification and cloning of genes from MS7163 into plasmid pSU2718 and confirmed by sequencing. All primers are listed in Data Set S2.
Transcription analysis and Western blotting. To assess transcript levels of genes involved in curli production, strains were grown on YESCA agar for 24 h at 37°C, harvested directly, resuspended in KPi buffer (28.9 mM KH 2 PO 4 and 21.1 mM K 2 HPO 4 , pH 7.2) and standardized to an optical density at 600 nm (OD 600 ) of 1. A volume of 500 l of this bacterial suspension was added to 1 ml of RNAprotect bacterial reagent (Qiagen) to stabilize total RNA. RNA extraction, conversion to cDNA, and quantitative reverse transcription-PCR (qRT-PCR) were performed as previously described (82). All primers are listed in Data Set S2. The relative transcript level of target genes from each mutant was compared to wt MS7163 using gapA as a control. Relative transcript levels were calculated by the threshold cycle (2 ϪΔΔCT ) method (83) and are presented on a log 2 scale (log 2 FC) (84). Western blotting for the major curli subunit CsgA was performed following growth in YESCA agar for 24 h at 37°C as described previously (85).
Biofilm analyses. Biofilm assays were performed using polyvinyl chloride microplate plates (Corning) and growth in LB or LB without salt with subsequent crystal violet staining as previously described (85). To compare the sensitivity of sessile (biofilm) cells versus planktonic cells to LL-37, MS7163 biofilms were established following growth in LB without salt at 37°C for 24 h. Planktonic cells were removed, and the biofilm was washed 3 times with phosphate-buffered saline (PBS). One set of biofilm cells were incubated in LB for 2 h at 37°C, while another set were incubated in LB containing 64 g/ml of LL-37 for 2 h at 37°C. After incubation, the biofilm was washed with 1ϫ PBS, cells were detached from the biofilm by repeated pipetting and resuspended in 1ϫ PBS, and the total CFU were enumerated by plating on LB agar. In parallel, planktonic cells were centrifuged, resuspended in either LB or LB containing 64 g/ml of LL-37, and incubated for 2 h at 37°C, and the total CFU were enumerated by plating on LB agar. The difference in survival rates of planktonic and biofilm cells following incubation with LL-37 was analyzed by t test.
Mouse model of UTI. The C57BL/6 mouse model of ascending UTI was employed as previously described (16,86). Strains were enriched for expression of type 1 fimbriae by three successive rounds of static growth in LB without salt for 48 h followed by one round of static growth for 24 h for inoculum preparation and did not display any difference in type 1 fimbria production as assessed by yeast cell agglutination and a fim switch orientation PCR (see Fig. S3 in the supplemental material). Curli expression was confirmed by dot blot analysis with CsgA antibody. Infections were performed as competitive assays; the inoculum contained 1:1 strain mixture of wt MS7163 versus MS7163csgA, MS7163purF, or PcL-rcpA. Bacterial loads corresponding to each strain in the urine, bladder, and kidney at 24 hours postinfection (hpi) were enumerated by plating onto LB agar and LB agar supplemented with chloramphenicol. Competitive indices were determined as the ratio of each respective mutant versus wt MS7163 to the ratio of the two strains in the inoculum. Statistical analyses were performed using the two-tailed Wilcoxon matched-pair signed-rank test (Prism7; GraphPad).
Accession number(s). The sequences for the MS7163 chromosome and plasmids pMS7163A and pMS7163B have been deposited in the NCBI GenBank database under GenBank accession no. CP026853, CP026854, and CP026855, respectively. The raw PacBio sequence reads have been deposited in the sequence Read Archive (SRA) under accession no. SRR6727962. The TraDIS reads have been deposited at the SRA under accession no. SRR6706193, SRR6706194, SRR6706195, and SRR6706196.

ACKNOWLEDGMENTS
We thank Barbara Arnts for technical assistance with the mouse UTI model.