Insightful Improvement in the Design of Potent Uropathogenic E. coli FimH Antagonists

Selective antiadhesion antagonists of Uropathogenic Escherichia coli (UPEC) type-1 Fimbrial adhesin (FimH) are attractive alternatives for antibiotic therapies and prophylaxes against acute or recurrent urinary tract infections (UTIs) caused by UPECs. A rational small library of FimH antagonists based on previously described C-linked allyl α-D-mannopyranoside was synthesized using Heck cross-coupling reaction using a series of iodoaryl derivatives. This work reports two new members of FimH antagonist amongst the above family with sub nanomolar affinity. The resulting hydrophobic aglycones, including constrained alkene and aryl groups, were designed to provide additional favorable binding interactions with the so-called FimH “tyrosine gate”. The newly synthesized C-linked glycomimetic antagonists, having a hydrolytically stable anomeric linkage, exhibited improved binding when compared to previously published analogs, as demonstrated by affinity measurement through interactions by FimH lectin. The crystal structure of FimH co-crystallized with one of the nanomolar antagonists revealed the binding mode of this inhibitor into the active site of the tyrosine gate. In addition, selected mannopyranoside constructs neither affected bacterial growth or cell viability nor interfered with antibiotic activity. C-linked mannoside antagonists were effective in decreasing bacterial adhesion to human bladder epithelial cells (HTB-9). Therefore, these molecules constituted additional therapeutic candidates’ worth further development in the search for potent anti-adhesive drugs against infections caused by UPEC.


Introduction
One of the major pathogens responsible for urinary tract infections (UTIs) is uropathogenic Escherichia coli (E. coli) (UPEC), leading to a major burden in public health [1]. As part of the normal microbiota, E. coli exhibits diverse species with a wide spectrum of phenotypes that reside in the large intestine of humans and many animals [2]. Several virulence factors are responsible for the establishment of current infections [2,3]. While E. coli mainly live Scheme 1. Synthesis of a family of C-linked mannopyranosides using palladium-catalyzed Heck reaction.
Because C-linked mannopyranosides do not exhibit an anomeric effect in comparison to their O-linked analogs, there was a risk that the analogs had undergone conformational changes from the desired 4 C 1 chair to skew boats ( 0 S 2 or 0 S 2 ) [4]. The X-ray crystal structure of perbenzoylated compound 2 showed it to be in the desired chair conformation as previously observed with several related analogs ( Figure S4). Since the conformation of 2 in the solid-state could be different than that in solution, an in-depth NOESY analysis was also performed to further confirm its conformation ( Figure S3). The nOe between the geminal H1'a, H1'a' with the axial H3 and H5 unmistakably implied the 4 C 1 conformation. Analysis of the 1 H-NMR spectrum of C-linked mannosides (11-18, and 20) also showed values of ca. 1. 63 Hz for their 3 J 1,2 coupling constants of their vicinal equatorial/axial arrangements ( Figure S2).

X-ray and Molecular Dynamic Simulations
Unfortunately, no crystalline structure could be obtained from our However, co-crystallized FimH with the second-best inhibitor 11 was ob lution of 3 Å ( Figure S38, Table S1, PDB entry code 8BVD). The structure w the PHENIX [44] software to visualize the detailed binding interactions w between the FimH binding domain and the inhibitor.
Delightfully, the co-crystal structure of compound 11 and E. coli C [45,46] also revealed the expected 4 C1 chair conformation of the ligand 1 active site of the tyrosine gate. In the mannose-binding site of FimH, we in parallel, while the Tyr137 residue was in T-aromatic stacking with the of ligand 11. Four FimH lectin domains were observed in the asymmetric tal ( Figure S38) and the mannose-binding site of each FimH lectin domain ity with a symmetry-related neighbor in the crystal packing ( Figure 2) [34 ing sites of FimH molecules were held together by the two quinoline sub over in a parallel stacking (Figure 2), while their mannosides projected pocket of FimH. The distance between the axial O2 hydroxyl group of t end mannoside was 12

X-ray and Molecular Dynamic Simulations
Unfortunately, no crystalline structure could be obtained from our b However, co-crystallized FimH with the second-best inhibitor 11 was ob lution of 3 Å ( Figure S38, Table S1, PDB entry code 8BVD). The structure w the PHENIX [44] software to visualize the detailed binding interactions w between the FimH binding domain and the inhibitor.
Delightfully, the co-crystal structure of compound 11 and E. coli C [45,46] also revealed the expected 4 C1 chair conformation of the ligand 11 active site of the tyrosine gate. In the mannose-binding site of FimH, we in parallel, while the Tyr137 residue was in T-aromatic stacking with the of ligand 11. Four FimH lectin domains were observed in the asymmetric tal ( Figure S38) and the mannose-binding site of each FimH lectin domain ity with a symmetry-related neighbor in the crystal packing ( Figure 2) [34 ing sites of FimH molecules were held together by the two quinoline sub over in a parallel stacking (Figure 2), while their mannosides projected pocket of FimH. The distance between the axial O2 hydroxyl group of t end mannoside was 12 A° which exactly matched the distance previou divalently bound trimannoside to FimH (crystal structures with PDB en

X-ray and Molecular Dynamic Simulations
Unfortunately, no crystalline structure could be obtained from our b However, co-crystallized FimH with the second-best inhibitor 11 was ob lution of 3 Å ( Figure S38, Table S1, PDB entry code 8BVD). The structure w the PHENIX [44] software to visualize the detailed binding interactions w between the FimH binding domain and the inhibitor.
Delightfully, the co-crystal structure of compound 11 and E. coli C [45,46] also revealed the expected 4 C1 chair conformation of the ligand 11 active site of the tyrosine gate. In the mannose-binding site of FimH, we in parallel, while the Tyr137 residue was in T-aromatic stacking with the of ligand 11. Four FimH lectin domains were observed in the asymmetric tal ( Figure S38) and the mannose-binding site of each FimH lectin domain ity with a symmetry-related neighbor in the crystal packing ( Figure 2) [34 ing sites of FimH molecules were held together by the two quinoline subs over in a parallel stacking (Figure 2), while their mannosides projected pocket of FimH. The distance between the axial O2 hydroxyl group of th end mannoside was 12

X-ray and Molecular Dynamic Simulations
Unfortunately, no crystalline structure could be obtained from our However, co-crystallized FimH with the second-best inhibitor 11 was o lution of 3 Å ( Figure S38, Table S1, PDB entry code 8BVD). The structure the PHENIX [44] software to visualize the detailed binding interactions between the FimH binding domain and the inhibitor.
Delightfully, the co-crystal structure of compound 11 and E. coli [45,46] also revealed the expected 4 C1 chair conformation of the ligand active site of the tyrosine gate. In the mannose-binding site of FimH, w in parallel, while the Tyr137 residue was in T-aromatic stacking with th of ligand 11. Four FimH lectin domains were observed in the asymmetr tal ( Figure S38) and the mannose-binding site of each FimH lectin doma ity with a symmetry-related neighbor in the crystal packing ( Figure 2) [3 ing sites of FimH molecules were held together by the two quinoline su over in a parallel stacking (Figure 2), while their mannosides projected pocket of FimH. The distance between the axial O2 hydroxyl group of end mannoside was 12 A° which exactly matched the distance previo 19

X-ray and Molecular Dynamic Simulations
Unfortunately, no crystalline structure could be obtained from our b However, co-crystallized FimH with the second-best inhibitor 11 was ob lution of 3 Å ( Figure S38, Table S1, PDB entry code 8BVD). The structure w the PHENIX [44] software to visualize the detailed binding interactions w between the FimH binding domain and the inhibitor.
Delightfully, the co-crystal structure of compound 11 and E. coli C [45,46] also revealed the expected 4 C1 chair conformation of the ligand 11 active site of the tyrosine gate. In the mannose-binding site of FimH, we in parallel, while the Tyr137 residue was in T-aromatic stacking with the of ligand 11. Four FimH lectin domains were observed in the asymmetric tal ( Figure S38) and the mannose-binding site of each FimH lectin domain ity with a symmetry-related neighbor in the crystal packing ( Figure 2) [34 ing sites of FimH molecules were held together by the two quinoline subs over in a parallel stacking (Figure 2), while their mannosides projected pocket of FimH. The distance between the axial O2 hydroxyl group of th end mannoside was 12 A° which exactly matched the distance previou

X-ray and Molecular Dynamic Simulations
Unfortunately, no crystalline structure could be obtained from our b However, co-crystallized FimH with the second-best inhibitor 11 was ob lution of 3 Å ( Figure S38, Table S1, PDB entry code 8BVD). The structure w the PHENIX [44] software to visualize the detailed binding interactions w between the FimH binding domain and the inhibitor.
Delightfully, the co-crystal structure of compound 11 and E. coli C [45,46] also revealed the expected 4 C1 chair conformation of the ligand 11 active site of the tyrosine gate. In the mannose-binding site of FimH, we in parallel, while the Tyr137 residue was in T-aromatic stacking with the of ligand 11. Four FimH lectin domains were observed in the asymmetric tal ( Figure S38) and the mannose-binding site of each FimH lectin domain ity with a symmetry-related neighbor in the crystal packing ( Figure 2) [34 ing sites of FimH molecules were held together by the two quinoline subs over in a parallel stacking (Figure 2), while their mannosides projected pocket of FimH. The distance between the axial O2 hydroxyl group of th end mannoside was 12 A° which exactly matched the distance previou

X-ray and Molecular Dynamic Simulations
Unfortunately, no crystalline structure could be obtained from our best inhibitor 20. However, co-crystallized FimH with the second-best inhibitor 11 was obtained at a resolution of 3 Å ( Figure S38, Table S1, PDB entry code 8BVD). The structure was solved using the PHENIX [44] software to visualize the detailed binding interactions which took place between the FimH binding domain and the inhibitor.
Delightfully, the co-crystal structure of compound 11 and E. coli C43 (DE3) FimH [45,46] also revealed the expected 4 C 1 chair conformation of the ligand 11 bound into the active site of the tyrosine gate. In the mannose-binding site of FimH, we observed Tyr48 in parallel, while the Tyr137 residue was in T-aromatic stacking with the quinoline group of ligand 11. Four FimH lectin domains were observed in the asymmetric unit of the crystal ( Figure S38) and the mannose-binding site of each FimH lectin domain was in proximity with a symmetry-related neighbor in the crystal packing ( Figure 2) [34]. The two binding sites of FimH molecules were held together by the two quinoline substituents crossed over in a parallel stacking (Figure 2), while their mannosides projected into the binding pocket of FimH. The distance between the axial O2 hydroxyl group of the non-reducing end mannoside was 12 A • which exactly matched the distance previously observed in

Further Insights into the Design and Binding of Compound 20
Given that the best compound 20 failed to crystallize with the FimH lectin domain, we tried to explain its improved potency using molecular dynamics (MD) simulations. The best low energy score was obtained from its docking within the  [48][49][50][51].
According to previous work [41], the ortho-biphenyl substituted C-linked mannopyranoside with KD = 6.9 ± 5.7 nM had a higher affinity over the para-substituted analog (compound 18, Table 1) with KD = 17 ± 3.5 [41]. Previously published MD simulation with an ortho-substituted biphenyl derivative could nicely explain the origin of the better affinity observed with 20. Indeed, the MD simulation allowed us to postulate that the higher affinity might originate from a π-stacking between the first phenyl ring of 20 with Tyr48, while the second ortho pyridyl moiety can interact with the hydroxyl group of Tyr137 through a hydrogen bond (2.70 A°). Thus, the MD simulation was instrumental in the replacement of the second phenyl group with a heterocyclic pyridyl moiety to improve the solubility as well as the affinity of ligand 20.

Further Insights into the Design and Binding of Compound 20
Given that the best compound 20 failed to crystallize with the FimH lectin domain, we tried to explain its improved potency using molecular dynamics (MD) simulations. The best low energy score was obtained from its docking within the  [48][49][50][51].
According to previous work [41], the ortho-biphenyl substituted C-linked mannopyranoside with KD = 6.9 ± 5.7 nM had a higher affinity over the para-substituted analog (compound 18, Table 1) with KD = 17 ± 3.5 [41]. Previously published MD simulation with an ortho-substituted biphenyl derivative could nicely explain the origin of the better affinity observed with 20. Indeed, the MD simulation allowed us to postulate that the higher affinity might originate from a π-stacking between the first phenyl ring of 20 with Tyr48, while the second ortho pyridyl moiety can interact with the hydroxyl group of Tyr137 through a hydrogen bond (2.70 A°). Thus, the MD simulation was instrumental in the replacement of the second phenyl group with a heterocyclic pyridyl moiety to improve the solubility as well as the affinity of ligand 20. (Right) detailed molecular interactions of 11 with amino acids in the protein binding pocket of FimH: Phe1, Asp47, Gln133, Asp54, Asn135, Asp140 (PDB entry 8BVD). The carbon atom of the sugar ligand is grey and green for the FimH carbons.

Further Insights into the Design and Binding of Compound 20
Given that the best compound 20 failed to crystallize with the FimH lectin domain, we tried to explain its improved potency using molecular dynamics (MD) simulations. The best low energy score was obtained from its docking within the FimH half-open tyrosine gate (−90.70 kcal/mol) (PDB 4AUY) ( Figure 4). For comparison, the lowest FimH closed and open energy conformations were at −60.34 and −89.76 kcal/mol, respectively [48][49][50][51].
According to previous work [41], the ortho-biphenyl substituted C-linked mannopyranoside with K D = 6.9 ± 5.7 nM had a higher affinity over the para-substituted analog (compound 18, Table 1) with K D = 17 ± 3.5 [41]. Previously published MD simulation with an ortho-substituted biphenyl derivative could nicely explain the origin of the better affinity observed with 20. Indeed, the MD simulation allowed us to postulate that the higher affinity might originate from a π-stacking between the first phenyl ring of 20 with Tyr48, while the second ortho pyridyl moiety can interact with the hydroxyl group of Tyr137 through a hydrogen bond (2.70 A • ). Thus, the MD simulation was instrumental in the replacement of the second phenyl group with a heterocyclic pyridyl moiety to improve the solubility as well as the affinity of ligand 20.

Mannosides Do Not Affect Bacterial Growth, Cell Viability, and Antibiotic Activities
To test whether the above mannoside derivatives could exert bactericidal and/or cytotoxic activities, bacterial growth and cell viability assays were undertaken in the presence of each mannoside at different concentrations. Natural D-mannose (D-Man) was included as control. As shown in Figure 5, bacteria grown in the presence of mannoside analogs showed similar growth rates compared to bacteria grown in LB (Luria-Bertani) medium that represents the positive control for bacterial replication. Moreover, no cytotoxic effect was recorded for urinary bladder cell line 5637 (HTB-9 cell) monolayers [48,52], as measured by the MTT test ( Figure 6). Since D-mannose can serve as nutrient replacement for E. coli when there is a shortage of D-glucose, we tested whether these new derivatives could be used as carbon sources as well [5]. To further this aim, bacteria were cultured in LB medium for 12 h and then diluted to 1.5 × 10 7 colony forming unit/mL (CFU/mL) in PBS buffer. Bacteria were incubated for an additional 24 h in the absence and in the presence of different mannosides and the number of viable bacteria was assessed by CFU/mL counting. Results show no statistically significant differences in the number of mannoside-treated bacteria compared to non-treated control (Figure 7). Conversely, the number of bacteria increased significantly in the presence of D-mannose, at both concentrations tested (500 μM and 83 mM) (Figure 7) [5]. The results indicated that the synthetic mannoside antagonists did not enter into the metabolic cycle of the bacterial cells to support their growth even in the absence of any other carbon sources. Differently, D-mannose as a natural sugar molecule is metabolized by bacteria, thereby maintaining bacterial replication [5]. Since carbohydrates can influence the activity of conventional antibiotics, the activity of different classes of antibiotics such as ampicillin (AMP 30 μg/mL), streptomycin (SM 50 μg/mL), and gentamycin (GM 50 μg/mL) in the presence of the synthetic mannoside analogs was evaluated. A broth-dilution test showed no differences in bacterial susceptibility, irrespectively in the presence of C-mannoside antagonists (Figure 8).

Mannosides Do Not Affect Bacterial Growth, Cell Viability, and Antibiotic Activities
To test whether the above mannoside derivatives could exert bactericidal and/or cytotoxic activities, bacterial growth and cell viability assays were undertaken in the presence of each mannoside at different concentrations. Natural D-mannose (D-Man) was included as control. As shown in Figure 5, bacteria grown in the presence of mannoside analogs showed similar growth rates compared to bacteria grown in LB (Luria-Bertani) medium that represents the positive control for bacterial replication. Moreover, no cytotoxic effect was recorded for urinary bladder cell line 5637 (HTB-9 cell) monolayers [48,52], as measured by the MTT test ( Figure 6). Since D-mannose can serve as nutrient replacement for E. coli when there is a shortage of D-glucose, we tested whether these new derivatives could be used as carbon sources as well [5]. To further this aim, bacteria were cultured in LB medium for 12 h and then diluted to 1.5 × 10 7 colony forming unit/mL (CFU/mL) in PBS buffer. Bacteria were incubated for an additional 24 h in the absence and in the presence of different mannosides and the number of viable bacteria was assessed by CFU/mL counting. Results show no statistically significant differences in the number of mannoside-treated bacteria compared to non-treated control (Figure 7). Conversely, the number of bacteria increased significantly in the presence of D-mannose, at both concentrations tested (500 µM and 83 mM) (Figure 7) [5]. The results indicated that the synthetic mannoside antagonists did not enter into the metabolic cycle of the bacterial cells to support their growth even in the absence of any other carbon sources. Differently, D-mannose as a natural sugar molecule is metabolized by bacteria, thereby maintaining bacterial replication [5]. Since carbohydrates can influence the activity of conventional antibiotics, the activity of different classes of antibiotics such as ampicillin (AMP 30 µg/mL), streptomycin (SM 50 µg/mL), and gentamycin (GM 50 µg/mL) in the presence of the synthetic mannoside analogs was evaluated. A broth-dilution test showed no differences in bacterial susceptibility, irrespectively in the presence of C-mannoside antagonists (Figure 8).
Altogether these results demonstrate the lack of toxicity of the synthetic mannoside derivatives toward both bacteria and eukaryotic cells. Moreover, these molecules are not used for bacterial metabolism and, unlike natural D-mannose, do not favor bacterial replication.

C-Mannoside Antagonists Are Effective in Decreasing Bacterial Adhesion to Human Bladder Epithelial Cells
To evaluate the efficacy of the synthetic mannoside antagonists to inhibit the ability of CFT073 strain to adhere to epithelial cells, an in vitro adhesion assay was performed. For this purpose, equal amounts of strain CFT073 strain were inoculated in PBS supplemented with different concentrations of each mannoside at final concentrations of 100, 500 µM, and 1 mM and incubated for 3h under static conditions. Bacterial inoculation was used to infect  Figure 9). No significant inhibition of bacterial adhesion to bladder cells was obtained using mannoside inhibitors at 100 µM. Conversely, a significant reduction (more than 1 log) in the number of adherent bacteria was observed by increasing the concentration of the mannosides to 500 µM in comparison to non-treated bacteria ( Figure 10). The same extent of reduction in bacterial adhesion to bladder cells was obtained by increasing mannosides concentration to 1 mM, thereby showing a dose-dependent effect. Interestingly, among the mannosides studied, the quinoline analog 11, having an IC 50 of 3.17 nM, was the most efficient at reducing FimH-mediated bacterial adherence (Figures 9 and 10). On the other hand, compound 23 (MeMan) was the less efficient; however, it achieved the same extent of bacterial adhesion inhibition of natural D-mannose but at a 164-fold lower concentration. These results nicely confirmed previous observations from the control reference compounds 21, 23, and D-mannose [46]. To appraise qualitatively the reduction of bacterial adhesion, parallel infected cells were fixed, and Giemsa stained. As shown in Figure 10, no macroscopic differences in the shape, integrity, adhesiveness, cytoplasmic vacuolization, proliferation, or cytotoxic effects were observed in HTB-9 cell monolayers incubated with the synthesized mannosides, in line with the results of the MTT assay [5]. Overall, these results revealed that these synthetic inhibitors finely mimic the interaction between FimH and its natural receptor, thereby significantly decreasing the adhesion of strain CFT073 to bladder cells.      Altogether these results demonstrate the lack of toxicity of the synthetic mannoside derivatives toward both bacteria and eukaryotic cells. Moreover, these molecules are not used for bacterial metabolism and, unlike natural D-mannose, do not favor bacterial replication.

C-Mannoside Antagonists Are Effective in Decreasing Bacterial Adhesion to Human Bladder Epithelial Cells
To evaluate the efficacy of the synthetic mannoside antagonists to inhibit the ability of CFT073 strain to adhere to epithelial cells, an in vitro adhesion assay was performed.  Altogether these results demonstrate the lack of toxicity of the synthetic mannoside derivatives toward both bacteria and eukaryotic cells. Moreover, these molecules are not used for bacterial metabolism and, unlike natural D-mannose, do not favor bacterial replication.

C-Mannoside Antagonists Are Effective in Decreasing Bacterial Adhesion to Human Bladder Epithelial Cells
To evaluate the efficacy of the synthetic mannoside antagonists to inhibit the ability of CFT073 strain to adhere to epithelial cells, an in vitro adhesion assay was performed. shown in Figure 10, no macroscopic differences in the shape, integrity, adhesiveness, cytoplasmic vacuolization, proliferation, or cytotoxic effects were observed in HTB-9 cell monolayers incubated with the synthesized mannosides, in line with the results of the MTT assay [5]. Overall, these results revealed that these synthetic inhibitors finely mimic the interaction between FimH and its natural receptor, thereby significantly decreasing the adhesion of strain CFT073 to bladder cells. Figure 9. Mannoside antagonists efficiently inhibit bacterial adhesion to bladder cells. Bacteria were pre-incubated or not (non-treated) with mannose molecules at a final concentration of 500 μM. One mL of inoculation was used to infect HTB-9 cells at a MOI of 10. The total number of adherent bacteria was determined after 2.5 h of incubation and expressed as a percentage of CFU/mL (%) relative to the non-treated bacteria considered as 100%. Data represent the means ± SDs of three independent experiments performed in triplicate. p values were evaluated by one-way ANOVA; * p < 0.05, ** p < 0.01, *** p < 0.001. Figure 9. Mannoside antagonists efficiently inhibit bacterial adhesion to bladder cells. Bacteria were pre-incubated or not (non-treated) with mannose molecules at a final concentration of 500 µM. One mL of inoculation was used to infect HTB-9 cells at a MOI of 10. The total number of adherent bacteria was determined after 2.5 h of incubation and expressed as a percentage of CFU/mL (%) relative to the non-treated bacteria considered as 100%. Data represent the means ± SDs of three independent experiments performed in triplicate. p values were evaluated by one-way ANOVA; * p < 0.05, ** p < 0.01, *** p < 0.001.  Giemsa-stained after 2.5 h cell incubation. Representative images of three independent experiments are shown. Scale bar, 10 µm. Images were recorded with the 40× objective using a Leica DM5000B microscope and processed using the Leica Application Suite 2.7.0.R1 software (Leica). Arrows show the adherent bacteria.

General Information
Nuclear magnetic resonance (NMR) spectra were recorded on Varian Geminin 300 MHz or Innova 600 MHz spectrometer. Chemical shifts are reported in parts per million (ppm) (δ) relative to CDCl 3 (δ 7.27 and 77.23 ppm for 1 H-and 13 C-NMR, respectively) or relative to the signal of CD 3 OD (δ. 3.31 and 4.8 ppm for 1 H-and 49.7 ppm for 13 C-NMR, respectively). Where necessary, DEPT, APT, and two-dimensional 1H-1H COSY and HSQC experiments were performed for complete signal assignments. Coupling constants (J) are reported in Hertz (Hz). Melting points were measured on a Fisher Jones apparatus and are uncorrected. Purification of some compounds were done by semi-preparative HPLC (Agilent, Santa Clara, CA, USA). All eluents contained 0.1% formic acid and flow rate was set to 5 mL/min. Solvents were dried by distillation from drying agents as follows: DMF (Barium oxide), CH 2 Cl 2 (P 2 O 5 ), Et 3 N and pyridine (CaH 2 ), MeOH was stored over 4A molecular sieves. Some aryl iodides were synthesized according to literature procedures [53][54][55]: 2-iodophenyl acetate, 4-iodophenyl acetate, 4-iodobenzyl acetate and methyl 4-iodobenzoate. Compound 9 and 18 were prepared as previously published [30].

General procedure for Heck Coupling of Protected C-Mannopyranosides 2-10
To a solution of mannopyranosides 1 [40] in degassed anhydrous DMF, were added the various iodoaryl derivatives (2 equiv.), 10% palladium(II) acetate, tetrabutylammonium bromide (1 equiv.), and sodium bicarbonate (3 equiv.). The reaction mixture was heated at 85 • C under N 2 . The course of the reactions was followed by TLC. The solution was evaporated under reduced pressure and the residue was purified by flash column chromatography on silica gel (from 0 to 425% AcOEt-Hexanes).

Expression and Purification of FimH
The FimH lectin (Phe1-Thr158) was produced and purified as described previously [46] by expression from the pET-24a vector in E. coli C43(DE3) [45] in MinA medium complemented with the 20 amino acids, the vitamins biotin and thiamin, glucose and MgCl 2 . Soluble FimH lectin secreted in the periplasm was extracted by applying 30% sucrose and 2.5 mM EDTA, in 20 mM HEPES at pH 7.4, onto the washed bacterial pellet followed by a 30-fold dilution in the same buffer to cause the desired osmotic shock. A 30' centrifugation at 13,000× g was carried out to eliminate the cellular debris present in the pellet and isolate the soluble proteins in the supernatant. The supernatant was acidified using HCl to pH = 3.9 and centrifuged again before cation exchange chromatography onto an HiTrap Sulfopropyl Fast Flow column (SPFF, Cytiva). The SPFF column was washed in 20 mM formic acid (pH = 3.9) and eluted using a salt gradient. The fractions containing FimH lectin eluted between 150-250 mM NaCl and were immediately neutralized upon elution by adding a drop of 1 M HEPES at pH 7.4. Finally, pure protein fractions were pooled and dialyzed against 20 mM HEPES at pH 7.4 containing 150 mM NaCl.

Co-Crystallization of Antagonist 11 with FimH
Purified FimH lectin was concentrated to 17.19 mg·mL −1 and 1 mM of ligand 11 was added for co-crystallization at 20 • C using the sitting-drop vapor-diffusion method. A single crystal was obtained in a condition from the JCSG crystallization screen (Molecular Dimensions), containing 3.0 M NaCl and 0.1 M BIS-TRIS at pH = 5.5 ( Figure S5). Cryoprotection prior to flash-freezing in liquid nitrogen was performed by dragging the crystal through a drop containing 3.5 M NaCl, 50 mM BIS-TRIS at pH 5.5 and 30% glycerol. The crystal diffracted to 3 Å resolution at the PX1 beamline of the French Soleil synchrotron. Molecular replacement using PDB entry 2VCO (same as above) [46] of its oligomannose-3 ligands, led to the placement of four FimH lectin protomers in the unit cell of the hexagonal crystal. Iterative rounds of refinements were performed using PHENIX and the model was adjusted manually using Coot [56]. The crystal structure was run through PDB_REDO [57], for further optimization, and validated using Molprobity [58] and Staraniso [59].

Affinity Evaluation of Mannosides through FimH LEctPROFILE Kit
FimH LEctPROFILE kit assays from GLYcoDiag (Orléans, France) were performed according to GlycoDiag's protocol already described [60][61][62]. Briefly, the interaction profiles of each compound were determined through a competitive inhibition assay based on the inhibition by the compounds of the interaction between FimH lectin coated onto the microplate surface and a biotinylated neoglycoprotein NeoM (Man-BSA) as a tracer. A mix of biotinylated Man-BSA (fixed concentration) and the corresponding compounds (range of concentrations) prepared in PBS supplemented with 1 mM CaCl 2 and 0.5 mM MgCl 2 was deposited in each well (50 µL each) in triplicate and incubated for two hours at room temperature. After washing with PBS buffer, the conjugate streptavidin-DTAF (dichlorotriazinylamino fluorescein) was added (50 µL) and incubated 30 min more. The plate was washed again with PBS. Finally, 100 µL of PBS was added for the readout of fluorescent plate performed with a fluorescence reader (Pherastar microplate reader, BMG labtech, λex = 485 nm, λem = 530 nm). The signal intensity is inversely correlated with the capacity of the compound to be recognized by the lectin and expressed as inhibition percentage with comparison with the corresponding tracer alone. Data analysis was performed with GraphPadPrism software (version 5.03 for windows, San Diego, CA, USA). 50% inhibitory concentration (IC 50 ) was determined according to a standard dose-response/inhibition fitting model with the following equation: y = 100 / (1 + [inhibitor]/IC 50 ) and expressed in nanomolar units.

Bacterial Strains and Cell Line
The well characterized UPEC strain CFT073 (ATCC 700928) was used as the uropathotype in this study. Strain CFT073 was grown at 37 • C in LB or seeded onto MacConkey agar plates. The presence of the fimH gene was confirmed by PCR using the primers fimH-F 5 -TGCAGAACGGATAAGCCGTGG-3 and fimH-R 5 -GCAGTCACCTGCCCTCCGGTA-3 and E. coli fimH-proficient and -deficient strains served as positive and negative controls (E1P and I2P strains), respectively [5,63,64]. The human bladder epithelial cell line 5637, (ATCC HTB-9) (ATCC-LGC, Milan, Italy) was routinely cultured in T25 flasks at 37 • C in a humidified atmosphere with 5% CO 2 using Roswell Park Memorial Institute (RPMI) 1640 medium supplemented with 10% FBS (both Gibco, Milan, Italy).

Effect of Mannosides on Bacterial Growth and Metabolism
Mannosides dissolved in dimethyl sulfoxide DMSO or water at a final concentration of 5 mM were prepared. To test the mannosides' toxicity, strain CFT073 was grown on LB medium supplemented with each molecule reported in Table 1, at a final concentration of 50, 100 and 500 µM. Natural D-mannose (D-Man) at equal concentration to the mannosides was used as control. LB supplemented with DMSO was included as growth control. Bacterial cultures were then incubated in a 96-well plate at 37 • C over a period of 10 h with a 30 min temperature equilibration period before data acquisition started. Readings of culture turbidity (OD 600 ) were determined using a plate reader (POLARstar Omega BMG Labtech plate reader, Germany). To evaluate whether mannosides could be used as carbon sources, an initial inoculum of strain CFT073 of~1.5 × 10 7 CFU/mL was incubated in PBS supplemented with each molecule at a final concentration of 500 µM for 24 h at 37 • C. Following incubation, the growth of strain CFT073 was determined by colony forming unit (CFU/mL) counting by spot-plating serial dilutions.

Cell Viability and Toxicity Assay
To evaluate whether synthesized mannosides can affect eukaryotic cells viability, HTB-9 cells were seeded onto 24-well plates at 5 × 10 5 cells per mL and incubated in RPMI supplemented with 10% FBS in the presence of each mannoside molecule at a final concentration of 50, 100 and 500 µM at 37 • C in a 5% CO 2 atmosphere for 24 h. At this point, cell viability was determined by the MTT assay. The medium was replaced by fresh RPMI supplemented with 10% FBS and 1 mg/mL MTT, and the cells were further incubated for 1 h. Viable cells, with active metabolism able to metabolize yellow tetrazole (MTT) into purple formazan crystals, were quantified by measuring the absorbance at 570 nm of formazan crystals formed and solubilized with isopropanol.

Antibiotic-Mannoside Interference Assay
To assess any interference with antibiotic activities, strain CFT073 (approximately 10 5 -10 6 CFU/well) was inoculated into a 96-well microplate supplemented with LB containing ampicillin (AMP 30 µg/mL), streptomycin (SM 50 µg/mL) and gentamycin (GM 50 µg/mL) with or without the addition of each mannoside antagonist (500 µM concentration). LB supplemented with DMSO was included as growth control in this experiment. The microplate was incubated at 37 • C and bacterial growth kinetics were monitored by measuring the OD 600 over a period of 16 h.

Bacterial Adhesion Assay
HTB-9 cells were routinely seeded in cell culture plates and maintained 2-4 days at 37 • C in a humidified atmosphere containing 5% CO 2 . For the adhesion assay, cells were seeded in 35 mm tissue culture plates at a density of 1 × 10 5 cells/well and incubated at 37 • C for 48 h to reach confluency. CFT073 was grown in LB under mild shaking conditions overnight and resuspended in phosphate buffer (PBS) to an inoculum of~10 6 CFU/mL (normalized according to OD 600 ). Each mannoside molecule was added to CFT073 inoculant at the final concentrations of 100, 500 µM and 1 mM and incubated for 3 h in static conditions. One ml of these bacterial/mannosides mixtures was used to infect HTB-9 cell monolayers at a multiplicity of infection (MOI) of 10; monolayers were centrifuged (10' at 2000× g) and incubated at 37 • C with 5% CO 2 for 2.5 h. The CFT073 strain incubated without any mannoside molecules was used as control. Monolayers were extensively washed (seven times) with PBS and lysed with 0.1% Triton X-100 in PBS. Cell lysates were serially diluted and spot-plated onto LB agar plates for CFU/mL counting. Parallel infected cells were Giemsa stained for qualitative assessment of bacterial adhesion, as previously described [5,65]. Images were recorded with a Leica DM5000B microscope equipped with DFX340/DFX300 camera and processed using the Leica Application Suite 2.7.0.R1 software (Leica).

Molecular Dynamics Simulations
The complex of compound 20 and the FimH lectin domain (a.a. 1-158) with the best score using induced fit was used as the starting configuration for the molecular dynamics (MD) simulations. The complex was solvated and the structural waters were added using the same structural information as in the docking (PDB code: 4AUY) [66] and the ionic concentration was set to 0. 15 M NaCl. In accordance with propKa [67] the standard protonation state at pH 7 was used for all protonatable groups of FimH. The generated molecular system comprised about 45,000 atoms including around 15,000 water molecules. The CHARMM36 force field with CMAP corrections was used to describe protein, water, and ion atoms [68][69][70]. Missing force field parameters for compound 20 were initially generated with CGenFF [68] with standard parameters and afterwards adapted. The integrity of the compound was verified in a 50-ns long MD simulation of the compound alone in water using the adapted force field.
Two independent simulations of the so generated system were performed. In each of them a three-step equilibration was applied: first, a 2.5 ns long equilibration of the water and ions molecules, second a 2.5 ns long equilibration in which only the protein backbone was fixed, and third unrestrained simulations was carried out for 2.5 ns. This was followed by a 30-ns long production run.
All MD calculations were performed in the isothermal-isobaric ensemble at 300 K with the program NAMD2.9 [48]. Long-range electrostatic interactions were calculated using the particle-mesh Ewald method [49]. A smoothing function was applied to truncate shortrange electrostatic interactions. The Verlet-I/r-RESPA multiple time-step propagator [51] was used to integrate the equation of motions using a time step of 2 and 4 fs for short-and long-range forces, respectively. All bonds involving hydrogen atoms were constrained using the Rattle algorithm [50].

Conclusions
In this study, the design, synthesis, and function of a small library of C-mannose inhibitors containing heteroaryl moieties were reported. Their relative binding affinity was measured using a competitive inhibition assay against the binding of FimH with mannosylated BSA conjugate. Among them, the best results were obtained with compounds 20 and 11, respectively. Although compound 20 had a higher inhibitory potency (IC 50 0.82 ± 0.4) against isolated FimH lectin binding domain, ligand 11 (IC 50 3.17 ± 2.3) showed better potency in inhibiting bacterial adhesion to bladder HTB-9 cell monolayers without adverse side effects. The crystal structure of the FimH lectin-binding domain co-crystallized with inhibitor 11 was obtained at a resolution of 3 Å. It also confirmed the expected 4 C 1 chair conformation of antagonist 11 bound into the active site of the tyrosine gate. Interestingly, the inter-molecular π-π stacking of two quinolines residues of 11 triggered the interlacing of two FimH lectins, providing a "bidentate" complex. Furthermore, in the mannosebinding site between FimH and ligand 11, Tyr48 was shown to be p-stacked in parallel to the quinoline moiety, while its Tyr137 appeared to form T-aromatic stacking. On the other hand, molecular dynamics (MD) simulations were done for the best antagonist 20, which unfortunately failed to provide co-crystals with FimH. The results indicated that the lowest potential energy was obtained with the FimH in its half-open conformation. The docking of compound 20 to this conformer helped raising the hypothesis that its high affinity may originate from a p-stacking of the first phenyl ring with Tyr48, as well as from the interaction of the ortho-pyridyl moiety with Tyr137 through a potential hydrogen bond.
In addition, the synthetic C-linked mannopyranoside inhibitors discussed herein neither affected bacterial growth or cell viability, nor interfered with antibiotic activity. The latter aspect is particularly important because antibiotics still represent the standard treatment for UTIs. However, literature data evidenced the increase of the number of cleared infections when antibiotics were administrated in combination with D-mannose [71]. Moreover, the preventive use of D-mannose showed a reduced number of UTIs in patients suffering from rUTIs. Hence, the reported mannoside derivatives, and in particular molecules 11 and 20, represent good candidates to be analyzed in clinical trials to definitively accelerate the inclusion of mannoside-based FimH inhibitors in the clinical guidelines for the treatment of UTIs.