Fimbriae reprogram host gene expression – Divergent effects of P and type 1 fimbriae

Pathogens rely on a complex virulence gene repertoire to successfully attack their hosts. We were therefore surprised to find that a single fimbrial gene reconstitution can return the virulence-attenuated commensal strain Escherichia coli 83972 to virulence, defined by a disease phenotype in human hosts. E. coli 83972pap stably reprogrammed host gene expression, by activating an acute pyelonephritis-associated, IRF7-dependent gene network. The PapG protein was internalized by human kidney cells and served as a transcriptional agonist of IRF-7, IFN-β and MYC, suggesting direct involvement of the fimbrial adhesin in this process. IRF-7 was further identified as a potent upstream regulator (-log (p-value) = 61), consistent with the effects in inoculated patients. In contrast, E. coli 83972fim transiently attenuated overall gene expression in human hosts, enhancing the effects of E. coli 83972. The inhibition of RNA processing and ribosomal assembly indicated a homeostatic rather than a pathogenic end-point. In parallel, the expression of specific ion channels and neuropeptide gene networks was transiently enhanced, in a FimH-dependent manner. The studies were performed to establish protective asymptomatic bacteriuria in human hosts and the reconstituted E. coli 83972 variants were developed to improve bacterial fitness for the human urinary tract. Unexpectedly, P fimbriae were able to drive a disease response, suggesting that like oncogene addiction in cancer, pathogens may be addicted to single super-virulence factors.


Author summary
Urinary tract infections affect millions of individuals annually, and many patients suffer from recurring infections several times a year. Antibiotic resistance is increasing rapidly and new strategies are needed to treat even these common bacterial infections. One approach is to use the protective power of asymptomatic bacterial carriage, which has been shown to protect the host against symptomatic urinary tract infection. Instilling "nice" bacteria in the urinary bladder is therefore a promising alternative approach to a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

Introduction
Mucosal surfaces provide ideal living conditions for the normal flora but paradoxically, they also serve as attack sites for numerous bacterial pathogens that cause extensive morbidity and mortality. Understanding this dichotomy is critical for efforts to selectively target and remove pathogens without disturbing the commensal flora or its protective effects. The complex nature of disease predicts that virulence is multifaceted and that pathogens need multiple virulence factors to initiate tissue attack, disrupt immune homeostasis and create symptoms and pathology [1][2][3][4][5][6][7][8]. It is also well established that commensals fail to cause disease, due to a lack of critical virulence genes [9,10]. About 50% of asymptomatic bacteriuria (ABU) isolates have a smaller genome size than acute pyelonephritis strains due, in part, to inactivating virulence gene deletions or point mutations [11][12][13]. These strains continue to accumulate loss of function mutations in vivo, supporting the notion of a virtually irreversible reductive evolution process, where spontaneous recovery of a virulent phenotype is not likely to occur.
Surprisingly, these loss-of-function mutations also affect fimbrial subunits and adhesin genes [13], which are thought to be essential for bacterial persistence at mucosal sites [14][15][16][17][18][19][20][21]. Adhesive ligands arm bacteria with molecular tools to identify preferred tissue sites and attachment plays a decisive role in colonization and long-term adaptation. Certain fimbrial adhesins are ubiquitously expressed by commensals and pathogens alike, suggesting a homeostatic role. Others, in contrast, show a strong disease association in epidemiologic studies [14], suggesting more direct effects on pathobiology [22].
The urinary tract supports ABU; a commensal-like state [23], which has been shown to prevent super-infection with more virulent strains [24][25][26][27]. To reproduce this protective effect, we have established a protocol to create ABU, by inoculating patients with the ABU strain E. coli 83972 [28,29]. The therapeutic efficacy and safety of this procedure has been documented in placebocontrolled studies in patients with incomplete bladder voiding [30]. Genome sequencing of E. coli 83972 has revealed a general "loss of virulence" phenotype, which includes fimbrial genes [13,[31][32][33]. E. coli 83972 lacks functional P or type 1 fimbriae, due to attenuating point mutations in the papG adhesin gene and a large, inactivating deletion in the fim gene cluster [13]. Both fimbrial types have been proposed to enhance bacterial persistence in the urinary tract. equipped E. coli 83972 with functional adhesins, previously shown to enhance bacterial persistence in the murine or human urinary tract [34,35]. This approach also made it possible to address how fimbriae affect clinical outcome in inoculated human hosts.
E. coli 83972pap expressed functional P fimbriae as shown by P blood group specific agglutination of human erythrocytes (S1A

Clinical response and fimbrial expression in vivo
In this longitudinal study, five patients (P I to P V) were sequentially inoculated; first with the ABU strain E. coli 83972 and then with the fimbriated variants of this strain (Fig 1C). Each patient contributed a pre-inoculation sample as well as samples from five time points following each inoculation, resulting in 18 samples for patients undergoing three-and 12 samples for patients undergoing two inoculations. As a result of the study design, the response to inoculation was defined relative to the pre-inoculation sample in each patient and inoculation, and changes over time were evaluated intra-individually. Significant changes were evaluated intraindividually as well as between patient groups.
E. coli 83972 and the fimbriated strains established significant bacteriuria within 48 hours of inoculation and persisted for a period of at least 4 weeks or until the patients were treated to remove the strain (S2 Fig). Patients, who carried E. coli 83972 or E. coli 83972fim remained asymptomatic and P III and P IV carried E. coli 83972pap asymptomatically during the entire study period. Two patients, who carried E. coli 83972pap, developed symptoms, however ( Fig  1C). In P V, symptoms were recorded 17 days after inoculation (fever, general malaise and loin pain, S2A Fig). The patient recovered fully after antibiotic treatment, with a drop in Creactive protein levels from 245 μg/ml to 3.4 μg/ml after 7 days, normal kidney function on follow up and no evidence of focal tissue damage by intravenous excretory contrast tomography. P II reported a transient febrile reaction and local symptoms from the urinary tract on day 9 after E. coli 83972pap inoculation. Inoculations with E. coli 83972pap were therefore discontinued and the study outcome is evaluated here for P I-P V.
Bacterial fimA expression increased immediately after inoculation with E. coli 83972fim, followed by a rapid decline. PapA expression increased gradually, from 3 hours post inoculation with E. coli 83972pap until the time of symptoms in P V (Fig 1D, S3 Fig). This difference in activation kinetics may reflect the location of the fim and pap gene clusters in the bacterial chromosome, as the fim gene cluster is part of the core chromosome but pap resides in a chromosomal island that also includes other virulence genes, such as hlyA [13]. The urine IL-6 and IL-8 responses paralleled fimbrial expression, with an earlier peak for E. coli 83972fim and a later peak for E. coli 83972pap (Fig 1E).
Genome-wide microarray analysis was used to address how the patients respond to bacterial inoculation. RNA was harvested from peripheral blood leucocytes immediately before inoculation and at defined times post-inoculation and significantly regulated genes were identified relative to the pre-inoculation sample in each patient (cutoff FC � 2.0). Fimbriae-related effects on transcription were further defined by intra-individual analysis, comparing the response to the E. coli 83972 and E. coli 83972pap or E. coli 83972fim inoculations in each individual.

Rapid reprograming of host gene expression by E. coli 83972pap
E. coli 83972pap activated a rapid and sustained change in gene expression, which was detected after 3 hours (S4 Fig) and reached a maximum in P V during the symptomatic episode (1,020 regulated genes, S5 and S6A Figs). A peak response in P II was also observed in connection with the symptomatic episode (2 weeks, S6B Fig) but not in P III and P IV, who remained asymptomatic (S6C and S6D Fig). Intra-individual comparisons of E. coli 83972pap and E. coli 83972 inoculations detected little overlap, suggesting that E. coli 83972pap activates a novel, strain background independent repertoire of host genes (P V, S4 and S5B Figs). A similar discrepancy was observed between E. coli 83972pap in P V and E. coli 83972fim in P I (S4B Fig).
By Gene set enrichment analysis (GSEA) and canonical pathway analysis (Fig 2A, S7 Fig), type I interferon (IFN) signaling was identified as the top-scoring canonical pathway in P V at the onset of symptoms (Fig 2A and 2B). Genes in this pathway were activated in at least one sample from each patient inoculated with E. coli 83972pap (S8A Fig). Significantly regulated genes included STAT1, IFIT1, IFIT3, MX1 and PSMB8. The IFN pathway genes were not regulated in response to E. coli 83972 or E. coli 83972fim, except in P IV between 3 and 24 hours (Fig 2C, S8B Fig).
To address to what extent pap reconstitution in an ABU strain creates a disease-like response [7,8,[37][38][39], we compared the repertoire of regulated genes in P V at the time of symptoms to the in vitro response of human kidney cells to the genetically closely related The ABU strain E. coli 83972, does not express functional P or type 1 fimbriae, due to chromosomal PapG point mutations and a fimB-D deletion. In this study, the pap or fim gene clusters were reconstituted in E. coli 83972pap and E. coli 83972fim, respectively. To replace the defective papG gene with a functional copy, a papGX deletion mutant (E. coli 83972ΔpapGX) was generated using lambda red homologous recombination [59]. Briefly, the chloramphenicol acetyltransferase gene (cat) cassette of plasmid pKD3 was amplified with overhangs homologous to the 5´-and 3 -regions of the E. coli 83972 papGX gene fragment and cured upon transformation with plasmid pCP20 [60]. Meanwhile, the functional papGX genes from UPEC strain CFT073 was amplified with homologous overhangs to the tnpA and papF regions of E. coli 83972 and used for electroporation into E. coli 83972ΔpapGX cells. Chromosomal reconstitution of the functional papGX genes in the E. coli 83972 chromosome was achieved via homologous recombination [59]. B. E. coli 83972 carries an internal 4,253-bp fim deletion, comprising the fimEAIC genes and truncated fimB and fimD genes. To reconstitute the fim operon, truncated genes were replaced via lambda red-mediated recombination by a cat cassette [59], flanked by two FRT sites and removed by FLP recombinase-mediated recombination [60]. The resulting 83972Δfim strain was transformed with pCP20 and suicide vector pJZ1, carrying the entire fim operon from pPKL4 [36]. Chromosomal integration of the entire suicide vector including the functional fim operon in E. coli 83972fim resulted in a functional fim copy. C. The human therapeutic inoculation protocol, indicating individual patients (P I-P V) and strains (E. coli 83972, E. coli 83972fim or E. coli 83972pap). Five patients were inoculated with E. coli 83972, which established ABU for a period of at least two weeks [28,30]. After clearance of the strain by a short course of antibiotics, three patients were re-inoculated with E. coli 83972fim, followed after termination of bacterial carriage by a third inoculation with E. coli 83972pap (P II, P III and P IV). P I received E. coli 83972 followed by E. coli 83972fim but not E. coli 83972pap, and P V received E. coli 83972 followed by E. coli 83972pap but not E. coli 83972fim (S1 Table). D. Kinetics of papA and fimA expression after human inoculation. Bacterial RNA was isolated directly from urine of each patient at the indicated time points and papA or fimA expression was quantified by qRT-PCR. Changes in gene expression were defined relative to frr (ribosome-recycling factor) expression. Values for 0h correspond to the bacterial in vitro culture used for inoculation. E. Kinetics of the urine cytokine response to E. coli 83972fim (P I-P IV, left) P < 0.01 ( �� ) or E. coli 83972pap (P II-P V, right) P < 0.01 ( �� ). Data was normalized by subtraction of the pre-inoculation values in each patient (0h). Mean ± s.e.m. of 4 samples, 2-way ANOVA. https://doi.org/10.1371/journal.ppat.1007671.g001 Fimbriae reprogram host gene expression -Divergent effects of P and type 1 fimbriae UPEC strain CFT073 (both phylogroup B2 and same sequence type) [13,40]. A total of 115 genes were commonly regulated, including IRF7 and genes involved in interferon-and pattern recognition signaling (S10 Fig), suggesting that E. coli 83972pap activates similar facets of the innate immune response as a virulent strain. P fimbriae as IRF7 agonists IRF-7 controls inflammation and renal tissue damage in the murine acute pyelonephritis model, through a network of pathology-associated genes [8]. A potent IRF7 response was detected in P V at the time of symptoms and a more restricted response in P II, five days after the transient symptoms (Fig 3A and 3B). Furthermore, the IRF7 response was exclusive for the time of symptoms and was suppressed or not activated in the patients, who did not develop symptoms (Fig 3C-3E). Downstream of IRF-7, type I interferon genes and RIG-I pathway genes, cytokines and transcription factors were specifically activated during the symptomatic episode as well as cell surface receptors involved in innate immunity (Fig 3E and S9 Fig).
Consistent with these effects, E. coli 83972pap infection also stimulated a strong IRF-7 response in human kidney cells. Cytoplasmic and nuclear IRF-7 protein levels were increased in E. coli 83972pap-infected cells (Fig 4A and 4B) and internalization of the PapG adhesin [41] was detected, suggesting a P fimbriae-specific effect (Fig 4C). This was confirmed by exposing the cells to purified PapG II adhesin protein or the PapDG II protein complex (5 and 25 μg/ml, Fig 4C, S11A and S11B Fig). IRF-7 expression was activated (Fig 4D) as shown by an increase in IRF-7 protein-and IRF7, IFNB1, MYC and IFIT3 mRNA levels (Fig 4E), suggesting that P fimbriae may act as IRF7 agonists, in a PapG adhesin dependent manner.
Effects of PapG on the assembly of the IRF7 promoter complex were examined in an electrophoretic mobility shift assay (EMSA), using IRF7 promoter DNA as a probe (1563bp, -1308 to +255, Fig 4F and 4G). Band shifts were detected when the probe was mixed with protein extracts from uninfected cells (band 1, Fig 4G, S11C Fig) or E. coli 83972pap infected cells (band 2, Fig 4G). Anti-PapG antibodies created a further super-shift (bands 3 and 4) and bands 1 and 2 were strongly attenuated by anti-IRF3-, anti-IFNβ and/or anti-MYC antibodies, consistent with the presence of these proteins in the promoter complex (Fig 4G and 4H). IFNβ and MYC are known to regulate IRF7 expression by binding to the IRF7 promoter and IRF-3 forms heterodimers with IRF-7, after activation by phosphorylation [42,43].
The results suggest that the PapG adhesin affects the assembly of IRF7 promoter complexes in infected cells, together with IRF-3, IFN-β and/or MYC. Direct binding of purified PapG or PapDG to promoter DNA was not detected, however, suggesting that the other promoter constituents may be required for PapG to bind to the IRF7 promoter complex.

Effects of type 1 fimbriae on host gene expression resemble E. coli 83972
Type 1 fimbriae are ubiquitously expressed among gram-negative bacteria, suggesting a homeostatic role. This study provided a unique opportunity to identify such effects, in inoculated human hosts. We found no evidence that type 1 fimbriae create symptoms or pathology, when expressed in the background of E. coli 83972. Instead a rapid and profound inhibitory effect was identified in patients inoculated with E. coli 83972fim, by GSEA and analyzed using Fimbriae reprogram host gene expression -Divergent effects of P and type 1 fimbriae the gene ontology database (Fig 5A). Genes involved in RNA processing and post-transcriptional regulation were inhibited, suggesting effects on the post-transcriptional environment in infected host cells, including predicted inhibition of 5' RNA capping and 3' poly(A) tail elongation and translation as well as intron removal, exon splicing and ribosome biogenesis (Fig 5B,  S2 Table). Pro-inflammatory gene sets were not significantly regulated and by Ingenuity pathway analysis (IPA, S12 Fig), a limited number of weakly regulated pathways were identified, including Natural Killer (NK) cell signaling, which was inhibited (S12B-S12D Fig) and Integrin signaling, which was moderately activated (S12E-S12G Fig).
In addition, a number of gene sets were moderately activated by E. coli 83972fim in inoculated hosts (P I-P IV, 3 hours, Fig 5C), including ion channels and genes involved neuropeptide signaling, sensory perception of pain and other stimuli. Gene network analysis further revealed strong similarities between E. coli 83972 and E. coli 83972fim (S13 and S14 Figs). By aligning these responses, we detected a subset of overlapping genes that was inhibited by both strains, with effects on RNA processing and ribosome biogenesis (S13A-S13C Fig). Kinetic analysis suggested that the effects of E. coli 83972 were accelerated by E. coli 83972fim, suggesting that type 1 fimbriae act, in part, by enhancing the inhibitory effects of E. coli 83972 (S13B-S13D Fig). In contrast, the transcriptional response to E. coli 83972fim and E. coli 83972pap showed no similarity. A few genes were inversely regulated (n = 33); activated by E. coli 83972pap but inhibited by E. coli 83972fim (P I, S14 Fig).

Type 1 fimbriae activate ion channel expression
In depth analysis of the regulated gene sets revealed that potassium channels, ion anti-porters, voltage gated cation channels and substrate specific ion channels were up-regulated by E. coli 83972fim (Fig 6A). By kinetic analysis of consecutive samples from P I-P IV, we detected an increase in Ca 2+ and K + channel expression after 3 hours, and K + channel expression was sustained, with a maximum after 48 hours, especially in P IV (Fig 6B). In addition, a moderate increase in Na + and Clchannel expression was recorded. These gene sets were not unique for E. coli 83972fim but were regulated also by E. coli 83972 after 24 hours, further suggesting that type 1 fimbriae enhance the effects of E. coli 83972 (S15 Fig).
The activation of ion channel expression was confirmed in vitro in human bladder epithelial cells, representing the site of infection in human hosts. E. coli 83972fim and E. coli 83972 infection generated an increase in K + channel protein levels, compared to uninfected control cells (TWIK, TRAAK and KCNJ11), with E. coli 83972fim showing the most pronounced effects (Fig 6C). In addition, we detected an increase in cation channel protein levels (TRPC1, TRPV6), exclusively in E. coli 83972fim infected cells (Fig 6D and 6E). The soluble receptor analogue α-D-methyl-mannopyranoside (α-D-man., 2.5%) effectively blocked the TRPC1 and TRPV6 response, suggesting that the effects are type 1 fimbriae specific (Fig 6F).
This hypothesis was confirmed by treating human bladder epithelial cells with purified FimCH protein complexes (1.25-5 μg/ml), [44][45][46][47]. A dose-dependent increase in TRPC1 expression was detected (Fig 6G and 6H) and the FimCH complex was shown to activate rapid Ca 2+ , K + and Zn 2+ fluxes, which were inhibited by α-D-methyl-mannopyranoside ( Fig  6I), suggesting that type 1 fimbriae stimulate ion fluxes, in an adhesin-dependent manner. As ion fluxes regulate a variety of cellular responses, we suggest that these findings identify a A. An IRF-7-centric gene network was activated in P V, at the time of symptoms (n = 103). B. A more moderate IRF-7 response was detected in P II, 5 days after transient symptoms (n = 45, 2 weeks). C, D. Patients II and IV, who did not experience symptoms, also did not show evidence of IRF-7 activation. Instead, IRF7-dependent genes were inhibited in P IV (n = 39, 2 weeks). E. Heatmap of disease associated, IRF-7-driven genes [8] 2 weeks after E. coli 83972pap inoculation (P II-P V). Genes in the network were activated in P V, at the time of symptoms and a partial response was seen in P II, 5 days after transient symptoms. Red = FC � 2.0 and blue = FC � -2.0. https://doi.org/10.1371/journal.ppat.1007671.g003 Fimbriae reprogram host gene expression -Divergent effects of P and type 1 fimbriae general mechanism by which type 1 fimbriae may affect tissue homeostasis at different mucosal sites.

Nerve cell related responses to type 1 fimbriae
Further analysis of patients inoculated with E. coli 83972fim identified 22 significantly activated gene sets (nominal p-value � 0.01), involved in neuronal sensing, neurotransmitter receptor activity and nervous system development (Fig 7A and 7B). Genes within these categories were regulated in all inoculated individuals but the time and amplitude of the maximum response varied between the patients (Fig 7B).
The response was reproduced in human bladder epithelial cells, where NK1R and SP protein levels were increased in vitro after E. coli 83972fim infection [48] (Fig 7C). Furthermore, the purified FimCH complex stimulated NK1R and SP expression in human bladder epithelial cells (Fig 7D), suggesting that type 1 fimbriae contribute to the activation of neuropeptides and neuropeptide receptors in an adhesin-dependent manner.
The results identify pronounced early effects of E. coli 83972fim on the host environment, with inhibition of RNA processing and translation and activation of ion channel-and neuropeptide responses. E. coli 83972fim enhanced both the inhibitory and activating effects of E. coli 83972, in a FimH adhesin-dependent manner, but did not significantly alter the profile of expressed genes, compared to E. coli 83972. In contrast, E. coli 83972pap activated host gene expression and changed the gene expression repertoire.

Comparative analysis of upstream regulators in patients inoculated with E. coli 83972pap or E. coli 83972fim
Last, to provide a molecular context to these divergent effects, we identified upstream regulators of the responses to E. coli 83972pap or E. coli 83972fim, respectively. This analysis predicts key transcriptional regulators of the response, in this case to E. coli 83972pap or E. coli 83972fim (Fig 8). We selected the time of maximal response of each patient and fimbrial type and included all regulated genes in the respective data sets. IRF-7 was identified as a potent upstream regulator of the response to E. coli 83972pap, consistent with the effects in patients and animal models of acute pyelonephritis (-log (p-value) = 61, Fig 8A and 8B). Other identified transcriptional nodes included IRF-3, which has been shown to balance the IRF-7 response by forming heterodimeric complexes and STAT1, regulating the expression of interferon stimulated genes. E. coli 83972, in contrast, was predicted to inhibit IRF-7 and E. coli 83972fim had no predicted effect.
A weak, more pleiotropic pattern was detected for E. coli 83972fim, suggesting, that host gene expression is inhibited more broadly, by mechanisms unrelated to specific transcription factors. MYC was identified as a transcriptional regulator in P I, after 3 and 24 hours and was Increase in IRF-7 protein levels after infection of human kidney cells with E. coli 83972pap but not E. coli 83972fim (10 5 cfu/ml, 4 hours). Western blot analysis. B. Increase in nuclear and total IRF-7 staining, quantified by confocal imaging. Mean + s.e.m. of two experiments, 50 cells/experiment. Two-tailed unpaired t-test compared to PBS. Scale bars = 20 μm. C. PapG internalization after infection or stimulation of cells with purified PapG protein (5-25 μg/ml), quantified by confocal imaging, using polyclonal anti-PapGII antibodies. Scale bar = 20 μm. D. The purified PapG adhesin or the PapDG protein complex stimulated an IRF-7 response, in treated cells (5-25 μg/ml). Scale bars = 50 μm. E. Increase in IRF7, IFNB1, MYC and IFIT3 mRNA levels quantified by qRT-PCR. Cells were infected with E. coli 83972pap or stimulated with the PapG adhesin or the PapDG protein complex (5 or 25 ug/ml). Mean + s.e.m. of two experiments, multiple unpaired t-test compared to PBS. F. IRF7 gene and promoter map. IRF7 promoter DNA (1563bp, -1308 to +255) was used as a probe, in an electrophoretic mobility shift assay (EMSA). G. Extracts from uninfected or E. coli 83972pap infected cells were mixed with the indicated IRF7 promoter fragment. By polyacrylamide gel electrophoresis, one band shift was detected in uninfected cells (band 1) and a second band in E. coli 83972pap infected cells (band 2). Specificity for PapG was supported by two super-shifted bands (bands 3 and 4), in the presence of anti-PapG antibody. Bands 3 and 4 were inhibited by using anti-IRF3 antibody. All bands were attenuated by combining anti-IFNβ with anti-IRF-3 or anti-MYC antibodies. H. Model of IRF7 activation by PapG, including IRF-3, IFNβ and MYC. P < 0.05 ( � ), P < 0.01 ( �� ), P < 0.001 ( ��� ). https://doi.org/10.1371/journal.ppat.1007671.g004 Fimbriae reprogram host gene expression -Divergent effects of P and type 1 fimbriae predicted to be inhibited, consistent with the overall inhibition of gene expression by E. coli 83972fim (-log (p-value) = 6.4 and 8, respectively, Fig 8C-8E). MYC was also predicted to be inhibited in P I by E. coli 83972 after 24 hours (-log (p-value) = 4) but was not regulated by E. coli 83972pap.

Discussion
Bacterial pathogens have evolved sophisticated molecular strategies to colonize the appropriate host niche and adherence is an essential first step to enhance their virulence [39]. Like pathogens, commensals have evolved adhesive surface ligands to enhance their fitness, but the outcome is very different, suggesting that the quality of the adhesive interactions may distinguish commensals from pathogens (Fig 9). Here, we address this question by comparing P fimbriae, which are expressed by uropathogenic E. coli strains [49] to type 1 fimbriae, which are expressed among Gram-negative bacteria, with no apparent disease association. By reconstituting the pap or fim gene clusters in the non-virulent E. coli strain 83972 and inoculating human hosts with the fimbriated variants of this strain, we have had the unique opportunity to study fimbrial function and define molecular effects in human hosts.
We made the unexpected observation that the acquisition of functional P fimbriae made the ABU strain virulent in two susceptible hosts. This is explained mechanistically by bacterial reprogramming of host gene expression, including the activation of IRF-7; a transcription factor that defines tissue pathology in the murine pyelonephritis model [8]. The PapG adhesin is identified as a transcriptional IRF7 agonist, in the context of IRF-3, IFN-β and MYC. We contrast this effect against type 1 fimbriae, which transiently inhibited genes involved in RNA processing and activated the expression of ion-channels and neuro-transmitters, with no evidence of symptoms in the host. Rather than reprogramming host gene expression, type 1 fimbriae broadly enhanced the inhibitory effects of the non-fimbriated wild type strain, suggesting a more homeostatic function. This does not exclude, however, a virulence-enhancing effect of the fimbriae, when expressed in the background of a fully virulent strain [50,51]. The findings illustrate the remarkably divergent effects of fimbriae in the infected host.
The rationale for this study was clinical, as protective effects of E. coli 83972 inoculation have been documented, in placebo-controlled studies [29]. Fimbriae were introduced in an attempt to increase the fitness of E. coli 83972 for the urinary tract and extend the use of human inoculation therapy. There was no indication from earlier human studies that P fimbriae alone would cause a disease-like response in inoculated hosts [49,52] and unlike fully virulent strains, E. coli 83972pap did not activate a disease response in the murine UTI model. UPEC-associated virulence genes are attenuated in E. coli 83972 and even after prolonged carriage further attenuation of virulence has been shown to occur, suggesting that the strains evolve towards commensalism [13]. It is important to emphasize that E. coli 83972pap is sensitive to antibiotics and that antibiotic therapy resulted in rapid resolution of symptoms and infection, without sequels.
These observations suggest, for the first time, that type 1 fimbriae may have potent inhibitory effects on the post-transcriptional machinery of the host, as E. coli 83972fim inhibited genes involved in RNA processing and translation. In addition we observed a rapid activation Volcano plot of gene sets regulated by E. coli 83972fim after 3 hours compared to pre-inoculation samples in P I-P IV (Gene Ontology). Identified gene sets are plotted as the -log (pvalue) against their Normalized Enrichment Score (NES) and functionally annotated (see S2 Table). Most inhibited gene sets were involved in RNA processing and translation (purple). Activated genes were mainly involved in ion channel (green)-and neuropeptide (red) regulation as well as immune signaling. B. Top five gene sets identified at the time of maximum response to E. coli 83972fim (3 hours). C. Top 15 activated gene sets at the time of maximum response to E. coli 83972fim (3 hours). https://doi.org/10.1371/journal.ppat.1007671.g005 Fimbriae reprogram host gene expression -Divergent effects of P and type 1 fimbriae of K + channels and solute carriers, as well as neuropeptides and their receptors, providing novel mechanistic insights into potential homeostatic effects and mechanism to broadly regulate cellular functions. Consistent with previous studies, NK cell function and integrin signaling was moderately affected by E. coli 83972fim (Fig 9B). A lack of distinct upstream transcriptional regulators suggested an entirely different level of control of the host response compared to P fimbriae, mainly executed at the post-transcriptional level. While type 1 fimbriae have been shown to increase mucosal inflammation in the murine UTI model and promote the formation of intracellular communities, type 1 fimbriae alone did not act as virulence factors in this study, when expressed in the background of a non-virulent strain.
The dual role of P fimbriae as bacterial sensors and transcriptional regulators is fascinating and challenges the dogma that virulence must rely on a complex set of virulence genes in every case. The difference between P and type 1 fimbriae further suggests that the repertoire of regulated host genes may distinguish disease-generating adhesins like P fimbriae from adhesive ligands that are involved in more homeostatic tissue functions. Based on these findings, we propose that bacteria may suffer from "virulence gene addiction", in analogy with the "oncogene addiction" of cancer cells [53,54]. While pathogens generally rely on multiple genes to survive in the host [55] this study suggests that a single, potent virulence determinant may be sufficient to enhance or attenuate virulence. It follows that a loss of P fimbria would represent a first step towards virulence attenuation and adaptation to long-term persistence in the urinary tract. This is supported by a high frequency of inactivating papG mutations in ABU isolates [13]. The findings raise the question whether therapeutic efforts should be focused on "super-virulence" gene attenuation rather than on functions that help the normal flora to maintain homeostasis in the host.

Study participants
Five patients with recurrent lower UTI and incomplete bladder emptying were included. The patients had experienced a minimum of four symptomatic episodes/year prior to enrolment and conventional treatment, including clean intermittent catheterization (CIC), had been tried but failed (S1 Table). The patients had anatomically normal urinary tracts as defined by cystoscopy and CT scanning. Renal function tests were normal. All patients had incomplete voiding (residual urine between 50-300 ml; if > 100 ml treated with CIC), and had recurrent UTI. Inoculations were performed during a four-year period (October 2007-June 2011). Time between inoculations ranged between 4-13 months (median 6 months).

Human epithelial cells
The A498 human kidney carcinoma cell line from a female (A498, American Type Culture Collection #HTB-44) and the 5637 human bladder grade II carcinoma cells (5637, ATCC# HTB-9) are established models to study UTI pathogenesis [37]. Cells were cultured in RPMI-1640 supplemented with 1 mM sodium pyruvate, 1 mM non-essential amino acids, and 10% heat-inactivated fetal bovine serum (FBS) (PAA) at 37˚C, 90% humidity and 5% CO 2 . For experiments, epithelial cells were cultured the previous day in six-well plates (4-6x10 5 cells/ well for Western blots and RNA extraction), or eight-well chamber slides (4-6x10 4 cells/well for confocal imaging), (Thermo Fisher Scientific). Cells were washed and exposed to bacteria in fresh, serum-free supplemented RPMI. Cells were infected with appropriately diluted bacteria in PBS and incubated for 4 hours at 37˚C with 5% CO 2 . To investigate FimH specificity, cells were treated with 2.5% α-D-methyl-mannopyranoside for 30 minutes prior to bacterial infection.

Human therapeutic inoculation
The protocol for therapeutic bladder inoculation with E. coli 83972 has been described [30,52]. In the present study, the protocol was modified to include only one inoculation, and this was enough for the patients to establish bacteriuria. Prior to inoculation, patients were treated with antibiotics to sterilize their urine. E. coli 83972 wild type or the fimbriated derivatives were cultured overnight (16 h) in lysogeny broth (LB), cells were harvested by centrifugation (10 min, 4,000 rpm) and re-suspended in PBS to a concentration of 10 5 cfu/ml. Patients were inoculated with 30 ml of the solution through a catheter, which was then removed. Each patient was closely monitored. Blood and urine samples were collected prior to inoculation, three, 24 and 48 hours and at one, two, four and seven weeks after inoculation. Patients had access to a direct telephone number to the study physician at all times and were prescribed antibiotics to be used immediately in case of symptoms and upon instruction by the physician. After seven weeks the patients received antibiotic treatment, which eradicated bacteriuria in all cases. in each patient, as exemplified in P I after 3 hours. NES = normalized enrichment score. B. Significant NES data (P < 0.01) for gene-sets activated at different time points in patients inoculated with E. coli 83972fim (P I-P IV). Regulated categories included genes involved in neurotransmission, nervous system development and taste receptors. C. Increased levels of SP (red) and its receptor NK1R (green) in bladder epithelial cells infected with E. coli 83972fim or E. coli 83972 (10 5 cfu/ml, 4 hours). D. The purified FimCH protein complex (1.25-5 ug/mL) stimulated SP (red) and NK1R (green) responses, as shown by confocal imaging. Mean + s.e.m. of three experiments, 50 cells/experiment. Two-tailed unpaired t-test compared to PBS. Scale bars = 20 μm. P < 0.05 ( � ), P < 0.01 ( �� ). https://doi.org/10.1371/journal.ppat.1007671.g007 Fimbriae reprogram host gene expression -Divergent effects of P and type 1 fimbriae Fimbriae reprogram host gene expression -Divergent effects of P and type 1 fimbriae

Monitoring of clinical parameters
To examine the effects of fimbriae on the establishment of bacteriuria and on the host response, intra-individual comparisons were performed. Three patients were first inoculated with E. coli 83972, subsequently with E. coli 83972fim and finally with E. coli 83972pap (S1 Table). Following inoculation, the establishment of bacteriuria was followed with repeated urine cultures and the host response was monitored by urine neutrophil counts quantified in uncentrifuged urine using a hemocytometer chamber. Interleukin-6 (IL-6) and IL-8 concentrations were quantified by Immulite (Siemens) in urine and blood. Urine samples obtained before inoculation and at each subsequent sampling point were diluted in PBS and semi-quantitatively cultured on TSA plates overnight (37˚C). Prior to the inoculation the urine was sterile and neutrophil numbers, IL-6 and IL-8 concentrations were below reference values for infection.

Fimbrial gene expression in vivo by qRT-PCR
Bacteria were harvested from urine samples immediately after delivery of urine samples by the patients, briefly centrifuged and resuspended immediately in RNAprotect Bacteria (Qiagen). Total RNA was extracted using the RNeasy mini kit (Qiagen) and reversely transcribed (Super-Script III, Invitrogen) in a two-step process with random hexamer primer. Prior to qPCR, the optimal annealing temperature and primer efficiency were determined. The fimA and papA transcripts were amplified using primers listed in S3 Table. Gene expression was quantified relative to frr (ribosome-recycling factor). For details, see Supplementary methods.

Fimbrial function
The expression of P or type 1 fimbriae was quantified by hemagglutination. Briefly, erythrocytes were harvested from heparinized human A 1 P 1 blood, resuspended in PBS or 2.5% α-Dmethyl-mannopyranoside in PBS and mixed with bacteria on microscopy slides. Agglutination was recorded as +++, ++, + or -. Bacterial adherence to the A498 kidney epithelial cell line was assessed as previously described [9] and evaluated by differential interference contrast (DIC) microscopy (Carl Zeiss). For details, see Supplementary methods.

Whole genome transcriptomic analysis
RNA was extracted from 1 ml of heparinized peripheral whole blood collected from the participants before inoculation and at seven time points after inoculation (3, 24 and 48 hours, and Innate immune recognition of P-fimbriated E. coli. The PapG adhesin binds Galα1-4Galβ-oligosaccharide motifs in glycosphingolipid receptors. Release of ceramide, the membrane anchor of the glycolipids, activates TLR4 signaling [37,61,62], and phosphorylation of the TLR4 adaptor proteins TRAM (TIR domain-containing adapter molecule 2 or TICAM2) and TRIF (TIR domain-containing adapter molecule 1 or TICAM1) activates downstream signaling, involving the phosphorylation of mitogen-activated protein (MAP) kinases, phospholipase C, p38, activating JNK (c-Jun N-terminal kinases), CREB (cyclic AMP response element-binding) and FOS-JUN (AP1), leading to IRF3-and IRF-7-dependent and AP-1 dependent transcription of cytokine-and chemokine genes, as well as type I interferons (IFN) including IFN-β [7]. Activation results in inflammatory cell recruitment and symptoms depend on the genetic make up of the host. In this study, PapG is defined as an agonist of IRF-7, in hosts and cells inoculated with E. coli 83972pap. B. Type 1 fimbriae recognize several mannosylated host cell glycoconjugate receptors. The FimH adhesin binds to uroplakins [63], to integrins through N-oligosaccharides [64], to the Tamm-Horsfall protein (or uromodulin) [65] and to immunoglobulins [66] as well as CD48 on mucosal mast cells [67]. Downstream, type 1 fimbriae have been proposed to stimulate the innate immune response, trigger apoptosis, and stimulate mast cell degranulation as well as promoting actin rearrangement in bladder epithelial cells [2,68,69]. In this study, FimH is defined as a broad, mostly inhibitory regulator of RNA processing and translation and inducer of ion channel-and solute carrier expression as well as neurokinin ligand-receptor networks. C. ABU strains like E. coli 83972 have developed successful adaptation strategies, which include the deletion or inactivation of virulence genes, leading to a reduction in genome size [33]. In addition, we have shown that ABU strains actively create a calm, non-reactive environment in the host by inhibiting gene expression. An intriguing mechanism is the suppression of by RNA polymerase II (Pol II) phosphorylation, which results in a protected phenotype [23]. one, two, four and seven weeks). After purification with the QIAamp RNA Blood Mini Kit (Qiagen), 100 ng of RNA was amplified using GeneChip 3´IVT Express Kit, after fragmentation and labeling, aRNA was hybridized onto Human Genome U219 arrays (all Affymetrix) for 16 hours at 45˚C, either by Aros Applied Biotechnology or in-house using the GeneAtlas system (Affymetrix). Transcriptomic data was normalized using Robust Multi Average (RMA) implemented in the Partek Express software. Fold change was calculated by comparing each sample to the pre-inoculation samples in each individual. Genes with absolute fold change >2.0 were considered differentially expressed. Heat-maps were constructed using the Gitools software. Differentially expressed genes and regulated pathways were analyzed using the Gene Set Enrichment Analysis (GSEA, Broad Institute) and the Ingenuity Pathway Analysis (IPA, Qiagen Bioinformatics) softwares. Fimbriae-specific effects on transcription were distinguished by comparing the response to E. coli 83972 inoculation at each time point and in each patient.

Polyclonal antibody pre-absorption
Polyclonal rabbit anti-PapGII antibody was made from native PapGII truncated protein at Sigma Biogenysis using standard protocol (Rabbit #127). An overnight culture of E. coli 83972 complemented with the plasmid pDD3 containing all pap genes from UPEC J96 except papG [31] was resuspended in PBS. 2 ml of the bacterial cells was lysed using an ultrasound sonicator (30 min at 4˚C). After centrifugation, the pellet was resuspended in PBS and mixed with 1:100 of anti-PapG serum. The lysate-antibody mix was incubated for 2 h at room temperature and centrifuged. The resulting supernatant was used for experiments.

Electrophoretic mobility shift assay
IRF7 promoter fragment [8] was amplified from human genomic DNA (for primers see S3 Table) and used as probe. Each reaction contained 3-5 μL of DNA probe and 2-5 μg of cell extract from E. coli 83972pap infected or uninfected A498 cells in binding buffer. For the band shift/competition assay, 1-2 μg of anti-IRF-3, anti-MYC, anti-IFNβ, anti-PapGII or IgG 2A control were used. Binding reactions were incubated at 15˚C for 30 min and loaded onto a 6% nondenaturing, nonreducing polyacrylamide gel. Alternatively, samples were loaded on a 2% agarose gel. Gels were imaged using the Bio-Rad ChemiDoc System. For details, see Supplementary methods.

Cellular ion flux assays
Intracellular calcium was measured by Fluo4 NW (Molecular Probes), intracellular potassium was measured by FluxOR (Molecular Probes) according to manufacturer's instructions in human bladder epithelial cells grown in 96-well plates (60,000 cells/well) after exposure to FimCH (5 μg). Extracellular Zn 2+ was measured by FluoZin-3 (Thermo Fischer Scientific) by addition of 1 μg/ml of indicator salt for 60 minutes prior to FimCH treatment. Fluorescent intensity was measured by Infinite F200 (Tecan) microplate reader at 20 seconds intervals for indicated times.

Statistical analysis
Data was examined in Prism version 6.02 (GraphPad). Normalized cytokine concentrations were compared using two-way ANOVA and Sidak's multiple comparisons tests. Changes in pathway gene expression (fold change) after inoculation were compared intra-individually using paired t-test (two tailed p values). Staining quantifications was analyzed using unpaired t-test (two tailed p-values) and qRT-PCR data using multiple unpaired two-tailed Student's ttest for homoscedastic variances. Results are presented as mean + s.e.m. and are representative Fimbriae reprogram host gene expression -Divergent effects of P and type 1 fimbriae of at least two independent experiments. Significance was accepted at P < 0.05 ( � ), P < 0.01 ( �� ) or P < 0.001 ( ��� ). The analysis was not blinded to condition.

Ethics statement
The study was approved by the Human Ethics Committee of the Medical Faculty, Lund University, Sweden (Dnr 298/2006; 463/2010) and informed consent forms were signed by all patients.
Supporting information S1 Fig. Phenotyping of E. coli 83972fim and E. coli 83972pap Table). PBLs and urine samples were collected prior to inoculation and after 3, 24, 48 hours, 1, 2 and 4 weeks. Bacterial numbers (cfu/ml), PMNs (x 10 4 /ml), IL-8 (ng/l), IL-6 (ng/l) were quantified in urine at each sampling point. Gene expression in CFT073-infected human kidney epithelial cells (A498, microarray data GEO: GSE43790) was compared to the symptomatic episode in PV. A. Heatmap of significantly regulated genes in the two data sets. 115 genes were commonly regulated between PV at the time of symptoms and in vitro CFT073 (FC<2.0). In addition, 882 genes were specifically regulated in response to E. coli 83972pap and 768 genes were regulated in response to CFT073. B. The shared genes between the two data sets were used to construct the gene network shown connecting 89 genes. Gene network analysis revealed genes in the IRF7-dependent network, immune response and cytokine genes as well as type I interferon pathway genes. The most strongly regulated genes included IRF7, transcriptional regulators (IRF1, RELB, JUN, ATF3), cytokines/chemokines (IL6, CXCL8, TNF, IL1B, CCL5, CXCL1, IL24 and IL15), acute phase response mediators (CRP), interferon-induced genes (IFIH1, IFIT5, IFI44L, ISG20, IFIT2, IFIT3) and cell migration and adhesion genes. The results suggest that the transcriptional reprogramming in E. coli 83972pap creates a transcriptional response that resembles the one in epithelial cells infected with fully virulent strains.  E. coli 83972fim. A. Rapid inhibition of host gene expression after inoculation with E. coli 83972fim in P I after 3 hours (61% of regulated genes). B. Heat map comparing the 3 hours response to E. coli 83972fim in P I to E. coli 83972 from 3h to 1w. Inhibited genes were shared, with more rapid kinetics for E. coli 83972fim than E. coli 83972. In addition, a set of genes was specific for E. coli 83972fim. C. Biological processes regulated by genes in the E. coli 83972fim "mega network". Top regulated functions included RNA processing, RNA translation and immune related functions. D. Inhibitory profile especially in P I and P IV, which lasted for at least 48 hours but was lost thereafter. Red = FC � 2.0 and blue = FC � -2.0. (TIF) S14 Fig. Host gene expression response to E. coli 83972 or E. coli 83972pap inoculations. A. Rapid inhibition of gene expression, after inoculation with E. coli 83972 (P I, 3 hours). A "mega-network" was generated by merging the five top-scoring expression networks detected by IPA. B. Heatmap illustrating the differential effect of E. coli 83972fim and E. coli 83972pap on host gene expression. The regulation of genes in the P I, 3 hours, E. coli 83972fim network is shown (see Fig 6A). Red