A host receptor enables type 1 pilus-mediated pathogenesis of Escherichia coli pyelonephritis.

Type 1 pili have long been considered the major virulence factor enabling colonization of the urinary bladder by uropathogenic Escherichia coli (UPEC). The molecular pathogenesis of pyelonephritis is less well characterized, due to previous limitations in preclinical modeling of kidney infection. Here, we demonstrate in a recently developed mouse model that beyond bladder infection, type 1 pili also are critical for establishment of ascending pyelonephritis. Bacterial mutants lacking the type 1 pilus adhesin (FimH) were unable to establish kidney infection in male C3H/HeN mice. We developed an in vitro model of FimH-dependent UPEC binding to renal collecting duct cells, and performed a CRISPR screen in these cells, identifying desmoglein-2 as a primary renal epithelial receptor for FimH. The mannosylated extracellular domain of human DSG2 bound directly to the lectin domain of FimH in vitro, and introduction of a mutation in the FimH mannose-binding pocket abolished binding to DSG2. In infected C3H/HeN mice, type 1-piliated UPEC and Dsg2 were co-localized within collecting ducts, and administration of mannoside FIM1033, a potent small-molecule inhibitor of FimH, significantly attenuated bacterial loads in pyelonephritis. Our results broaden the biological importance of FimH, specify the first renal FimH receptor, and indicate that FimH-targeted therapeutics will also have application in pyelonephritis.

Less is known about the host-pathogen interactions enabling establishment of upper-tract UTIs, including pyelonephritis and renal abscess. Another CUP pilus, the P pilus, is tipped with the PapG adhesin, which has been implicated in human pyelonephritis [2,[19][20][21]. However, data regarding the participation of P pili in murine pyelonephritis vary among prior studies, potentially due to disparities in host and bacterial strains used, and species specificity in the PapG glycolipid receptor [21,22]. Preclinical modeling of UTIs has been performed predominantly in female mice, and in a majority of murine backgrounds, upper-tract UTI resolves spontaneously and without significant sequelae [23,24]. Historically, female mice have been utilized in models of experimental UTI, typically initiated by catheter-directed inoculation of the mouse bladder; reliable catheter access to the male mouse bladder is technically challenging [25][26][27][28]. We previously developed an inoculation technique for initiating UTI in both male and female mice; when normal anatomic protections in males were thus bypassed, males evidenced more severe UTI, reflected throughout the course of infection by higher bladder and kidney bacterial loads, leukocyte infiltration, and inflammation [29]. Of note, these findings aligned with human epidemiologic data showing higher morbidity from complicated UTI in men, and higher UTI incidence in women with polycystic ovary syndrome (a common hyperandrogenic state) [30][31][32][33][34][35][36]. In C3H/HeN mice (a background inherently featuring vesicoureteral reflux, a primary risk factor for upper-tract UTI in humans), males and androgenexposed females developed severe pyelonephritis and renal abscesses [29,37]. Renal infection was nucleated by collections of UPEC occupying collecting ducts and more proximal segments of the nephron, which we termed kidney bacterial communities (KBCs) [37].
The consistent development of severe pyelonephritis in this model enables detailed investigation of bacterial and host factors involved in the establishment of ascending upper-tract UTI. We identified a previously unappreciated role in the kidney for the mannose-binding type 1 pilus adhesin FimH, previously recognized as a major urovirulence factor within the bladder. High-affinity mannosides, which neutralize FimH function and are in development for the treatment of cystitis [4,5,38,39], here significantly lowered bacterial burdens in the kidneys of mice with established pyelonephritis. To specify host factors required for upper-tract UTI, we performed a CRISPR/Cas9 screen in immortalized murine renal epithelial cells and identified desmoglein-2 (Dsg2) as a candidate receptor for FimH. The lectin domain of FimH (but not FimH with a point mutation that abolishes mannose binding) directly bound to the extracellular domain of human desmoglein-2 (DSG2). Finally, we observed co-localization of UPEC with Dsg2 in the collecting ducts of mice with ascending pyelonephritis. Our studies demonstrate that disruption of FimH binding to renal tubular epithelium represents a potential therapeutic intervention during pyelonephritis.

Type 1 pili are required for both bladder and kidney infection in male C3H/HeN mice
Type 1 pili have long been implicated in pathogenesis of UPEC cystitis, but only recently have new experimental models enabled molecular interrogation of their role in pyelonephritis. We infected male C3H/HeN mice with either wild-type UTI89 (a prototypic UPEC strain), or UTI89ΔfimH, an isogenic mutant lacking the type 1 pilus adhesin [16,40]. Two weeks post infection (wpi), UTI89ΔfimH was sharply attenuated compared to wild-type in bladder and kidney bacterial loads (p<0.0001; Fig 1A) and in incidence of visible renal abscess (1/16 vs 13/ 13; p<0.0001). We confirmed that UTI89ΔfimH reached the kidney normally following bladder inoculation, as kidney bacterial loads 24 hours post infection (hpi) were equivalent to wild-type (S1 Fig). Of note, an analogous defect was confirmed using the urosepsis UPEC isolate CFT073, whose isogenic CFT073ΔfimH mutant failed to colonize the kidney 2 wpi (S2 Fig). Using immunofluorescence microscopy, we demonstrated type 1 pili expression by UPEC in KBCs located within renal abscesses 2 wpi (Fig 1B). As UTI89 carries ten full or partial CUP operons [41], we similarly tested deletion mutants in each of the annotated CUP pili. No other CUP pilus mutants exhibited a defect in bladder or kidney infection in male C3H mice 2 wpi, after inoculation either alone (S3 The pilin and lectin domains of FimH together exist in an equilibrium between tense and relaxed states with differential mannose binding affinity. Conversion between these states is governed in part by a catch-bond mechanism [9,[42][43][44]. In the relaxed state, the FimH lectin domain samples conformational ensembles, allowing it to act as a molecular tether, with the mannose binding pocket in a conformation able to bind mannose tightly; in the tense state, the pocket is open and thus binds mannose only weakly [8]. In the bladder, the ability of FimH to transition between these states enables optimal epithelial binding by UPEC [8,45,46]. We infected male C3H/HeN mice with wild-type UTI89 or with UTI89ΔfimH complemented with FimH A27V/V163A , a variant that predominantly adopts the relaxed conformation (high-affinity mannose-binding state). Paradoxically, UPEC expressing this variant are attenuated during cystitis in female mice [8,45,46]. Here, we found similarly that UTI89 FimH A27V/V163A also failed to establish pyelonephritis, with significantly lower bacterial loads in kidneys (p = 0.001) as well as bladder (p<0.0001) compared to wild-type UTI89 (S5 Fig). Further, the type 1 pilus rod, formed from FimA subunits, normally can unwind from a tightly coiled helix to a more linearized rod; this action is hypothesized to help UPEC withstand the shear force of urine flow [47][48][49]. A variant in the type 1 pilus helical rod, FimA A22R , requires less force to adopt

PLOS PATHOGENS
Type 1 pili in UPEC pyelonephritis the unwound form, but UPEC expressing this variant are attenuated in the female mouse bladder [50]. We found that UTI89 expressing FimA A22R was also attenuated in the kidneys (p = 0.0038) and bladders (p<0.0001) of male C3H/HeN mice 2 wpi (S5 Fig). Thus, the conformational dynamics of type 1 pili, critical in establishing bladder infection, are similarly important in ascending pyelonephritis. Collectively, our data demonstrate that beyond their well-described role in establishing bladder infection, UPEC type 1 pili are also important in initiating ascending pyelonephritis and renal abscess.

Murine collecting duct cells display type 1 pili-dependent UPEC binding
The collecting duct is the first nephron segment encountered by ascending UPEC. Consistent with prior findings [51,52], ascending UPEC were located within collecting ducts in C3H/ HeN mice 5 dpi (Fig 2A). Therefore, we chose mouse intramedullary collecting duct (IMCD-3) cells to establish an in vitro model of type 1 pili binding to kidney epithelium. In standard bacterial binding assays, UTI89 ΔfimH was significantly attenuated compared to wild-type UTI89 (Fig 2B), recapitulating what we observed in in vivo infection. Following IMCD-3 cell infection and anti-E. coli antibody staining, we confirmed by flow cytometry the in vitro binding defect of ΔfimH (Fig 2C and 2D). This defect was complemented by chromosomal re-

PLOS PATHOGENS
Type 1 pili in UPEC pyelonephritis integration of wild-type fimH, but not by integration of the fimH Q133K mutant, which lacks mannose binding activity [10] (S6 Fig). The binding defect was also replicated in CFT073 ΔfimH (S6 Fig). Finally, UPEC binding to collecting duct cells was significantly inhibited by methyl α-D-mannopyranoside and to an even greater extent by mannoside FIM1033 (S7 Fig). Thus, the in vivo requirement for type 1 pilus function is reflected in in vitro binding of UPEC to IMCD-3 cells.
A CRISPR screen identifies candidate type 1 pilus receptor desmoglein-2 Although type 1 pili are well known as critical virulence factors in the bladder [6][7][8][9][10][11], they have not been implicated in ascending pyelonephritis, and a cognate receptor has not been identified. To screen for host genes participating in type 1 pili-dependent binding of UPEC to collecting duct epithelium, we performed a genome-wide CRISPR-Cas9 screen using the Brie sgRNA library in IMCD-3 cells (Fig 3A), providing 4× nominal coverage of each gene within the mouse genome [53]. The pooled library of edited cells were inoculated with UPEC, fixed and stained, and subsequently flow-sorted for the population of cells with the lowest fluorescence (Fig 3A). Genomic DNA was extracted from sorted cells, subjected to Illumina sequencing, and analyzed using the probability mass function of a hypergeometric distribution to identify candidate genes statistically associated with bacterial binding (S1 Table) [53,54]. Among these candidates (Fig 3B), we focused on genes encoding proteins that would localize to the cell surface and could be available to interact with bacterial pili [55][56][57]. The screen revealed deficient UPEC binding to cells edited with guides targeting desmoglein-2 (Dsg2; Fig  3B), which encodes a mannosylated cell junctional protein displayed on kidney tubular To screen for host genes responsible for type 1 pili-dependent binding to this cell line, we transduced the Brie library of mouse guide RNAs into IMCD-3 cells bearing Cas9, providing 4× nominal coverage of each gene within the mouse genome. Cells were then bound (MOI 150) by UTI89 and sorted by fluorescent labeling, isolating cells unbound by bacteria. Genomic DNA was extracted and sequenced, identifying candidate genes that may be required for UPEC binding to IMCD-3 cells. B) Volcano plot results from sorted and sequenced cells. Colored dots represent genes having an average log 2 fold change >0.5 and a-log 10 (p-value) >2. 5
Finally, we hypothesized that if the Dsg2-FimH interaction was important for pyelonephritis, mannosides could be employed therapeutically in our mouse model. Male C3H/HeN mice were inoculated with UTI89, and early KBCs were allowed to form over 5 days [37,61]; mice were then treated with mannoside FIM1033 (formerly termed 29R [5]) (Fig 6A). Compared with mock-treated mice, bladder and kidney bacterial loads were significantly reduced by mannoside treatment for 24 h (kidney p = 0.0434, bladder p = 0.0255; Fig 6C) and even more so with treatment for 48 h (kidney p = 0.0042, bladder p = 0.0003; Fig 6D). These in vivo data support a model in which desmoglein-2 on collecting duct epithelium serves as a receptor for UPEC FimH during pyelonephritis in vivo.

Discussion
In this study, we detail the importance of type 1 pili in host-pathogen interaction during ascending pyelonephritis and identify desmoglein-2 as the first candidate receptor for FimH on renal tubular epithelium. Our findings in male C3H mice confirm that type 1 pili are essential for cystitis (as in female mice [6][7][8][9][10][11]), but more importantly provide both in vitro and in vivo evidence of the function of FimH in bacterial infection of the kidney. Furthermore, our ability to mitigate kidney infection with mannosides suggests that type 1 pilus-directed therapeutics currently under development for recurrent cystitis [2,4,5,38,39] may also be useful in pyelonephritis.

Type 1 pili in UPEC pyelonephritis
Compared to bacterial cystitis, significantly less is known about the pathogenesis of UPEC pyelonephritis, primarily due to the fact that female mice of most backgrounds resolve kidney infection spontaneously [23,24]. Work in humans has identified genetic factors that confer susceptibility to pyelonephritis and renal scarring, including polymorphisms reducing IRF3 or CXCR1 (encoding IL-8 receptor) expression, in certain UTI-prone kindreds [62][63][64][65]. P pili have been considered the major adhesin in kidney infection [2,19,21]. This work has been complicated by allelic variation in the P pilus adhesin PapG (UTI89 encodes the PapGIII allelic variant but does not express P pili under laboratory conditions, while the common pyelonephritis strain CFT073 expresses the PapGII variant [21]). These three alleles exhibit differing affinity for various glycolipid receptors that are differentially expressed in the kidneys of humans and model animals-including mice, where further differences exist among backgrounds [2,66]. Beyond our in vivo results, we found that P pili were also unnecessary for UPEC binding to IMCD-3 cells (S6 Fig). Finally, while P pili are enriched in the genomes of pyelonephritis-associated UPEC isolates for children with acute pyelonephritis, a significant proportion of UPEC isolates from women with acute and recurrent UTI lack P pili, suggesting a role for other adhesin(s) during ascending infection [21,[67][68][69][70].
Recent works have employed new mouse models (including C3H/HeOuJ mice, male C3H/ HeN mice, and androgenized female mice) which now enable detailed study of the pathogenesis of kidney infection and abscess formation [29,37,51]. In our model of ascending UTI, type 1 pili were shown to be essential for maintenance of renal infection, through both genetic mutation of the pilus and pharmacological inhibition with mannosides. Other work has hinted at a role for type 1 pili and mannose targets in kidney infection [71][72][73][74][75]. For instance, an early study in C3H/HeN female mice noted significant attenuation in both bladder and kidneys upon deletion of fimH in E. coli strain 1177 that was rescued upon complementation [75]. Further, signaling through C5a receptor 1 (C5aR1), which regulates inflammatory cell recruitment in UPEC infection, may also enhance presentation of mannosylated glycoproteins by primary renal tubular epithelial cells [72,73]. Meanwhile, experiments in which tubular infection was initiated via microinjection of UPEC directly into rat nephrons posited a role for type 1 pili in interbacterial interactions and biofilm formation [71]. While our work indicates that FimH mediates renal epithelial binding (in agreement with prior staining of kidney sections using purified adhesins [21,76]), the present work does not exclude a role for type 1 pili in interbacterial interactions and/or biofilm formation within the kidney. Inhibition of any of these interactions by administration of mannosides might confer therapeutic benefit in pyelonephritis.
As the C3H mouse background used here features vesicoureteral reflux, it is conceivable that ongoing bladder infection (enhanced by type 1 pili expression) may replenish the kidney niche. The converse is also likely true, as suggested by the data in S1 Fig, where the lack of attenuation of ΔfimH in the kidneys 24 hpi appears to obscure the bladder pathogenesis defect that would be expected from prior work. Thus, the infected kidneys presumably can continually re-seed the bladder niche. The absence of a kidney phenotype with ΔfimH 24 hpi may also indicate that an alternative host-pathogen interaction is responsible for initial UPEC binding. The timing of FimH action during UPEC kidney colonization is therefore an exciting avenue for future study.
Using a CRISPR-Cas9 screen, we identified desmoglein-2 as a receptor for UPEC during ascending infection. DSG2 is a member of the cadherin family of Ca 2+ -binding proteins, involved in intercellular junctions via the desmosome [77,78]. Its identified roles in mammalian disease states are limited; of note, cardiomyocyte-specific conditional knockout of Dsg2 in mice phenocopies human arrhythmogenic cardiomyopathy, which is correlated with loss-offunction DSG2 mutations [79,80]. Interestingly, other members of the desmoglein family also act as receptors for unrelated microbes [81]. Dsg2 has been specifically implicated as a receptor for group B adenovirus serotypes 3, 7, 11, and 14 [82,83]; of note, serotype 11 is most PLOS PATHOGENS Type 1 pili in UPEC pyelonephritis commonly associated with hemorrhagic cystitis in renal and other transplant recipients [84]. Along the length of the nephron, DSG2 expression is highest within the collecting duct and decreases as one ascends to more proximal segments of the tubule [59]. As a family, desmogleins (and other cadherins) exhibit a unique form of mannosylation in which α-D-mannose (the binding target of type 1 pili) is present as a novel O-linked glycosylation modification [58]. Prior studies have demonstrated FimH binding to N-linked glycans containing a terminal mannose [7,10,85,86]. While N-linked glycans are present on each EC domain of DSG2 [58], our data suggest that FimH interaction with native DSG2 is mediated through EC4 and/or EC5 (Fig 4). Of particular interest is the possibility that the cadherin-specific O-linked α-Dmannose on EC4 may represent the preferred binding target. While FimH canonically binds mannose, existing studies have not aimed to interrogate whether the spatial arrangement of nearby amino acid residues may also influence binding affinity. Future work, including determination of the structural basis of DSG2-FimH binding, will further address this question.
One limitation of this study is that we were unable to obtain a complete Dsg2 knockout cell line, indicating that complete absence of Dsg2 may confer a significant defect in in vitro growth of IMCD-3 cells. It is becoming better appreciated that CRISPR-generated knockout cell clones display phenotypic plasticity, and residual low-level protein expression often persists [87]. Therefore, it is difficult to precisely quantify what proportion of UPEC binding is to Dsg2 as opposed to other mannosylated cell-surface receptors. It is possible that Dsg2 knockdown leads to compensatory production of other proteins; that Dsg2 knock-down reveals the participation of more minor binding partners for type 1 pili; or that the amount of residual Dsg2 in the knock-down clones provides sufficient receptors for UPEC to bind a modest proportion of cells. The present data do not distinguish among these possibilities.
Our results indicate that Dsg2 may not be the sole receptor for UPEC within the kidney, but rather the most important receptor among others on renal tubular epithelium. The apical surfaces of bladder epithelial facet cells are coated with a comparatively restricted set of proteins, largely four uroplakins [88]. Among these, uroplakin Ia bearing N-linked oligomannose is thought to be the primary receptor important for UPEC binding of intact bladder epithelium [6][7][8]10,11,89]. In contrast, the renal tubular epithelium bears a wider variety of surface proteins, a number of which might be mannosylated. Of note, CRISPR knock-down of the glycosyltransferase Dad1 on IMCD-3 cells also conferred a UPEC binding defect (Fig 3B), suggesting that this enzyme may play a role in mannosylation of Dsg2 and/or other potential receptors for UPEC type 1 pili. Additionally, multiple innate immune genes were statistically enriched in the CRISPR screen (Fig 3B), hinting that innate response pathways may alter receptor expression and thereby influence UPEC binding; this too represents an avenue for future study.
The present work highlights the importance of type 1 pili and the mannosylated epithelial receptor desmoglein-2 in UPEC colonization of renal tubules. As the type 1 pilus adhesin FimH is already targetable by small-molecule inhibitors and vaccines in patients with cystitis [3,68,90], these therapeutics may prove useful in prevention or treatment of upper-tract UTI as well. Finally, desmoglein-2 is expressed widely across many epithelia [55,57,91,92], suggesting that it may serve as a FimH receptor in other UTI-relevant niches, perhaps as a secondary receptor in the urinary bladder or in the gut.

Ethics statement
All animal protocols complied with relevant ethical regulations and received prior approval from the Institutional Animal Care and Use Committee at Washington University (approval number 20180159).

Bacterial strains and growth
For mouse infections and for binding studies with IMCD-3 cells, bacteria were grown statically at 37˚C for 18 h in Luria-Bertani (LB) broth. For the CRISPR screen, type 1 pili were induced by static growth at 37˚C for 18 h, then 1:100 subculture and static growth for an additional 18 h. Bacterial strains were all previously published, with mutations generated using the λ Red Recombinase system [93] (see S2 Table). Bacteria were pelleted at 7,500 × g at 4˚C for 10 min, then resuspended to an OD 600 of 1.0 (~4 × 10 8 colony-forming units [CFU]/mL) in sterile phosphatebuffered saline (PBS) for mouse infections, or in DMEM F-12 (Invitrogen 1132-0033) supplemented with 10% fetal bovine serum (FBS; VWR 97068-085) for tissue culture experiments.

Mouse infections
Male C3H/HeN mice (Envigo) aged 8-9 weeks were infected as previously described [29]. Briefly, mice were anesthetized with inhaled 3% isoflurane, and the lower abdomen was shaved and sterilized with 2% chlorhexidine solution. A 3-mm midline abdominal incision was made through the skin and peritoneum, exposing the bladder. The bladder was aseptically emptied before 50 μl of inoculum (1-2 × 10 7 CFU in PBS) was injected into the bladder lumen via 30-gauge needle over 10 s. The bladder was allowed to expand for an additional 10 s before the needle was removed. The peritoneum and skin incisions were closed with simple, interrupted sutures. At the time of surgery, mice were given sustained-release buprenorphine (1 mg/kg SQ) for analgesia. At the indicated time points, mice were euthanized by CO 2 asphyxiation. Bladders and kidney pairs were sterilely removed and homogenized in 1 mL or 0.8 mL PBS, respectively. Homogenates were serially diluted and plated on LB agar for CFU enumeration.

Immunofluorescence microscopy
Infected bladders and kidneys were bisected and fixed in 10% neutral buffered formalin for 24 h. Fixed tissues were embedded in paraffin, sectioned, and stained. Unstained slides were deparaffinized in xylenes, rehydrated in isopropanol, boiled in 10 mM sodium citrate, and blocked in 1% bovine serum albumin (BSA), 0.3% Triton-X 100 in PBS for 1-2 h at room temperature (RT). Tissue culture cells were fixed in methanol at -20˚C for 10 min and then blocked in 3% BSA for 1 h. Slides were incubated with primary antibodies for 4 h or overnight at RT and then (after washing in PBS) with secondary antibodies for 1 h at RT. Slides were then mounted with ProLong Gold antifade reagent (Invitrogen) and images acquired on an Olympus FV1200 confocal microscope. Primary antibodies utilized were: rabbit anti-type 1 pili (1:500 dilution), rabbit anti-E. coli

PLOS PATHOGENS
Type 1 pili in UPEC pyelonephritis streptomycin (Gibco 15140-122) until the day before an experiment, when medium was changed to the above without antibiotics. Cells were maintained at 37˚C in a humidified atmosphere with 5% CO 2 .

In vitro binding assays
The day before infection, IMCD-3 cells were seeded in 24-well plates. Medium was removed and cells were washed with PBS supplemented with Mg 2+ and Ca 2+ (PBS-MgCa; Sigma-Aldrich D8662). After inoculation with UPEC at a multiplicity of infection (MOI) of 20, plates were centrifuged at 400 × g for 3 min, then returned to the incubator. To enumerate bound CFU, after 45 min, wells were washed 5 times with PBS-MgCa, then lysed with 0.1% Triton X-100 (Sigma T9284). To enumerate internalized CFU, after 60 min, cells were washed with PBS-MgCa, treated with medium containing 50 μg/mL gentamicin (Thermo Fisher 15750060), and incubated at 37˚C for 90 min, then lysed with 0.1% Triton X-100. Lysates were serially diluted and plated to LB agar. Binding efficiencies were calculated in comparison to input wells, in which cells were inoculated with UPEC, incubated for 45 min, then lysed by addition of Triton X-100, and the well contents plated to LB agar for enumeration of input CFU.

Flow cytometry
The day before infection, IMCD-3 cells were seeded in 6-well plates. Cells were infected with UPEC (at MOI 20-150) for 45 min, then washed as described above; cells were liberated with 0.05% trypsin-0.02% EDTA (Gibco), pelleted, and fixed in 4% paraformaldehyde (Electron Microscopy Sciences) for 30-60 min at RT. Cells were then blocked in 3% BSA for 1 h at RT. Cells were stained with primary rabbit anti-E. coli antibody (E3500-06C; US Biological) or isotype control (Rabbit serum; Sigma-Aldrich R9133) and secondary AlexaFluor 488-conjugated donkey anti-rabbit IgG (Invitrogen A21206) for 1 h each. Samples were stored in 3% BSA until analyzed on a Becton Dickinson (BD) LSR II Fortessa flow cytometer. Gating strategy is shown in S10 Fig.
For screening, 5 × 10 6 IMCD-3 cells were seeded into eighty 15-cm dishes the day before infection. Cells were infected with UTI89 at MOI 150 and incubated at 37˚C for 45-75 min,

PLOS PATHOGENS
Type 1 pili in UPEC pyelonephritis then fixed and stained for flow cytometry as described above. Samples were placed at 4˚C overnight on a tube roller; the following day, the 5% least FITC-positive cells were collected on a Sony iCyt Synergy BSC sorter. Tubes containing these low-FITC sorted cells were centrifuged, resuspended in 250 μl PBS, and stored at -20˚C. A total of 6 × 10 8 mock-treated cells were harvested, separated into aliquots of 8 × 10 7 cells, resuspended in 2 mL PBS, and stored at -20˚C until genomic DNA preparation. DNA was extracted using QIAamp DNA Blood Maxi kit (Qiagen 51192) for mock samples and QIAamp DNA FFPE Tissue kit (Qiagen 56404) for experimental samples.

Sequencing and bioinformatics
Illumina sequencing was performed as described previously [53]. Briefly, genomic DNA was aliquoted into multiple wells of a 96-well plate (up to 10 μg of DNA in 50 μL total volume). Samples were sequenced on an Illumina HiSeq 2000. Barcodes in the P7 primer were deconvoluted, and the sgRNA sequence was mapped to a reference file of sgRNAs in the Brie library. To normalize for different numbers of reads per condition, read counts per sgRNA were normalized to 10 7 total reads per sample; this normalized value was then log 2 transformed. We used the hypergeometric distribution method to rank sgRNAs and calculate gene p-values using the probability mass function of a hypergeometric distribution (https://portals. broadinstitute.org/gpp/public/analysis-tools/crispr-gene-scoring-help). We considered candidate genes those having an average log fold change >0.5 and a false discovery rate >2.5 [53,54]. For input of the hypergeometric distribution ranking, we subtracted the log 2 normalized read values of the uninfected unsorted IMCD-3 Brie library from the log 2 normalized read values of the 5% lowest UPEC-bound sorted cells. We used R Studio to visualize the results of the hypergeometric distribution analysis.

Generation of Dsg2 knock-down clones (C4 and F11)
Dsg2 knock-down clones C4 and F11 were among hundreds generated by the Genome Engineering and iPSC Center at Washington University School of Medicine. IMCD-3 cells were nucleofected with Cas9 and a Dsg2-specific sgRNA (5' GGAACTACGCATCAAAGTTCTGG 3'). Single-cell clones were isolated by FACS and expanded in 96-well plates. Cells were harvested, and genomic DNA was amplified and subjected to targeted deep sequencing of ã 400-bp amplicon flanking the gRNA target. Clones were screened for frameshifts by sequencing the target region with Illumina MiSeq at~1500× coverage. We obtained no clones that completely lacked Dsg2 expression. Frameshift mutations identified in knock-down clones C4 and F11 are detailed in S3 Table. Quantitative PCR RNA was isolated from tissue culture cells using the Qiagen RNeasy kit. qPCR was performed using the Applied Biosystems TaqMan RNA-to-Ct 1-Step Kit (ThermoFisher 4392938) and the probes listed in S4 Table (Integrated DNA Technologies).

Protein production and biolayer interferometry (BLI)
Methods for purification of full-length FimH and the lectin domain (FimH LD ; amino acids 1-160) have been described previously [10]. Briefly, untagged FimH LD was purified from E. coli periplasmic extracts using ion-exchange and size-exclusion chromatography, dialyzed into sterile PBS, and stored at 4˚C until use.

Mannoside treatment
Male C3H/HeN mice were infected as described above. Beginning 5 dpi, mice were given mannoside FIM1033 (gift of Fimbrion Therapeutics), 8 mg/kg in sterile PBS with 4% DMSO, by intraperitoneal injection every 8 h for 24 or 48 h. Mock-treated mice were injected with 4% DMSO in sterile PBS. Mice were sacrificed 6 h post last treatment dose.

Statistical analysis
Statistical analysis was performed using Prism 8 (GraphPad Software). Differences were analyzed with the unpaired, two-tailed, nonparametric Mann-Whitney U test. P values <0.05 were deemed significant. Cultured murine collecting duct cells were infected with the indicated UPEC strains, and cells were stained with anti-E. coli antibody and analyzed by flow cytometry; median fluorescence intensities are shown. Mutation of fimH in UTI89 abrogated binding, while deletion of the P pilus usher (papC) had no effect. Chromosomal re-integration of wild-type fimH, but not fimH Q133K , restored binding in the UTI89 fimH mutant. Deletion of fimH in CFT073 similarly abrogated binding by this urosepsis strain. n = 9-10 wells per condition (aggregate of three triplicate experiments).