Antivirulence Isoquinolone Mannosides: Optimization of the Biaryl Aglycone for FimH Lectin Binding Affinity and Efficacy in the Treatment of Chronic UTI

Abstract

Uropathogenic E. coli (UPEC) employ the mannose-binding adhesin FimH to colonize the bladder epithelium during urinary tract infection (UTI). Previously reported FimH antagonists exhibit good potency and efficacy, but low bioavailability and a short half-life in vivo. In a rational design strategy, we obtained an X-ray structure of lead mannosides and then designed mannosides with improved drug-like properties. We show that cyclizing the carboxamide onto the biphenyl B-ring aglycone of biphenyl mannosides into a fused heterocyclic ring, generates new biaryl mannosides such as isoquinolone 22 (2-methyl-4-(1-oxo-1,2-dihydroisoquinolin-7-yl)phenyl α-D-mannopyranoside) with enhanced potency and in vivo efficacy resulting from increased oral bioavailability. N-Substitution of the isoquinolone aglycone with various functionalities produced a new potent subseries of FimH antagonists. All analogues of the subseries have higher FimH binding affinity than unsubstituted lead 22, as determined by thermal shift differential scanning fluorimetry assay. Mannosides with pyridyl substitution on the isoquinolone group inhibit bacteria-mediated hemagglutination and prevent biofilm formation by UPEC with single-digit nanomolar potency, which is unprecedented for any FimH antagonists or any other antivirulence compounds reported to date.

Urinary tract infections (UTIs) represent the second most common infectious disease, imposing enormous socioeconomic and healthcare burdens worldwide. UTIs predominantly afflict women, who have a lifetime risk of 50% for developing an acute infection and 20% for experiencing multiple, recurrent infections.^[1] Antibiotic resistance^[2] among uropathogenic E. coli (UPEC), which account for the overwhelming majority of UTIs, is a rising problem. In addition, the ability of UPEC to establish quiescent reservoirs may lead to the observed recalcitrance to antibiotic therapy, as chronic bladder reservoirs may serve as seeds for recurrent UTI and contribute to the troubling nature of this disease.^[3] Furthermore, a history of UTI significantly predisposes one to recurrent UTI.^[4] Thus, antibiotic-sparing approaches, such as antivirulence strategies^[5] that prevent UPEC colonization, represent viable therapeutics to address this emerging threat. UPEC express long fibrillar appendages at their cell surface termed type 1 pili,^[6] which are tipped with FimH,^[7] to mediate attachment to superficial facet cells of the bladder epithelium. FimH is a two-domain adhesin^[8] comprised of an N-terminal lectin domain, which mediates specific recognition of mannosylated receptors on the bladder epithelium, and a C-terminal pilin domain, which anchors FimH to the tip of the pilus. Upon binding, FimH triggers host cell invasion and intracellular colonization by UPEC. A proportion of the UPEC population evades expulsion by escaping into the cytoplasm and replicating to form intracellular bacterial communities (IBCs), from which infection may propagate to neighboring cells. Formation of IBCs is part of a mechanism that allows UPEC to gain a foothold in the bladder and build up in numbers. Targeting FimH with small-molecule^[9] and multivalent carbohydrate mimetics^[10] to interfere with bacterial adhesion, invasion, and IBC and biofilm formation represents a promising therapeutic strategy for combatting recurrent and antibiotic-resistant UTIs.

In preliminary studies we identified biphenyl mannosides 2 and 3 (Figure 1) as promising antagonists of FimH by X-ray-structure-based design.^[11] Compound 3 exhibits good inhibition of FimH function using UPEC strain UTI89 as determined by a hemagglutination assay (HAI=0.5 μM) and biofilm assay (IC₅₀=1.35 μM). This study revealed that meta substitution on the biphenyl B-ring with a hydrogen bond acceptor, such as the methyl amide of 3, was optimal. Further structure–activity relationship (SAR) studies from us and others^[12] produced mannosides such as 5 and 6, indicating that addition of a small substituent at the ortho position of the biphenyl A-ring leads to marked enhancements in potency. For example, ortho-methyl and chloro groups as with 5 and 6 were found to increase HAI titers (60 and 125 nM) relative to 3 (500 nM). The addition of another B-ring substituent also improved activity (compounds 7 and 8). This increase in potency was attributable to an increase in relative binding affinity for the FimH lectin domain as measured by differential scanning fluorimetry (DSF). In vivo investigation of mannoside 5 demonstrated that the compound is orally bioavailable and efficacious against acute,^[13] chronic,^[14] and antibiotic-resistant^[15] UTI in mice. However, despite this efficacy, the oral bioavailability of 5 is low, and compound exposure in the urine was significantly decreased by eight hours. Herein we report on newly optimized mannosides with significantly enhanced FimH potency, oral bioavailability, and efficacy in treating chronic UTIs.

Figure 1.

Progression of biphenyl α-D-mannoside FimH inhibitors.

To better understand the improved interactions of 5 and 7 with FimH, a high-resolution (1.67 Å) X-ray crystal structure of 7 (PDB ID: 5F2F) bound to the FimH lectin domain was obtained (Figure 2A). As noted in prior co-crystal structures,^[11a,16] the mannose ring is bound tightly in a deep hydrophilic pocket by an intricate network of hydrogen bonds between the C2, C3, C4, and C6 hydroxy groups of mannose and the main chain of F1 and side chains of N46, D54, Q133, N135, and D140. The O-linked aglycone projects toward the lid of the binding pocket and positions functional groups for favorable interactions. We discovered that the ortho-methyl substituent of the A-ring promotes hydrophobic interactions, as it packs snugly against the surface formed by I52, Y137, and N138. The B-ring phenyl forms favorable π–π stacking interactions with Y48 and the amide carbonyl of 7 indirectly interacts with the phenol of Y48 through a water-mediated hydrogen bond network, as evidenced by the ideal bent geometry and proximal distances (3–3.2 Å) of hydrogen bond donor–acceptor pairs. In addition, the secondary nitrogen atom of the two amides form electrostatic and potential H-bonding interactions with the neighboring R98 and Y137 residues, but these groups lie 4.1 and 4.6 Å away, respectively. However, the electron density of the second amide near Y137 reveals two different orientations of the amide carbonyl in which the carbonyl could be acting as an H-bond acceptor to the phenol. Comparison of mannoside 7 with 2 (Figure 2B) by structural overlay highlights a unambiguous difference in the orientation of the biphenyl group. Based on the structure of 7 presented here and a related analogue,^[12a] the improvement in activity of mannosides with ortho-substituted groups on the A-ring (5–8) relative to unsubstituted 2 and 3 can be explained by: 1) increased hydrophobic contact of the ortho substituent encompassed by the molecular surface formed by I52, Y137, and N138; 2) improved π–π stacking with Y48; and 3) decreased entropic cost of binding due to restriction in ring rotation. Altogether, this structure highlights the essential interactions that ortho-substituted mannosides form with the rim of the mannose-binding pocket, or the “tyrosine gate” of Y48 and Y137, responsible for increased potency. Despite the good activity of 5 and 7 from in vitro assays, in vivo pharmacokinetic (PK) studies show that both mannosides have low oral bioavailability and half-life. We hypothesize that they are metabolically unstable from hydrolysis of the methyl amide and/or hydrolysis of the O-glycosidic bond via the action of α-mannosidases. In an effort to improve the metabolic stability of these compounds, we examined the effects of replacing the amide with non-hydrolyzable bioisosteres.

Figure 2.

Structural basis of FimH–mannoside interactions. A) Crystal structure of compound 7 bound to FimH lectin domain (PDB ID: 5F2F). Unambiguous electron density from the F_o−F_c map (black mesh) and contoured at 1s reveals compound 7 (green sticks) bound in the mannose binding pocket of FimH (blue ribbon and surface). B) Comparison of mannoside binding poses. Overlay of compound 2 (yellow sticks; PDB ID: 3MCY) on 7 bound to the FimH lectin domain.

Our structure of 7 shows an electrostatic or hydrogen bonding interaction of the meta-substituted amide carbonyl with R98, similar to the ester carbonyl of 2 (Figure 2B; PDB ID: 3MCY).^[11a] The importance of this interaction can be demonstrated by 1, in which the ester is located in the para position, resulting in a marked decrease in potency (Figure 1). To determine whether we could mimic the amide of 3 and retain this interaction with FimH, we designed analogues with non-hydrolyzable isoelectronic and isosteric groups (Figure 3). In our earlier studies, we evaluated a small set of mannosides containing simple amide isosteres in the B-ring meta position, such as nitro 9, nitrile 10, and sulfonamide 14. All isosteres retained similar potency, while analogues incapable of the same hydrogen bonding—phenol 11, fluoride 12, and methyl alcohol 13—all lost potency relative to that of 3. This data provided additional evidence that the salt bridge interaction with R98 provides the basis for increased FimH binding and inhibition of adhesion by the HA assay. In another set of compounds, we used nitrogen-containing heterocycles to mimic the spatial orientation and electronics of the amide carbonyl of compound 7. Isoquinoline 15 displays similar activity as observed with 3, while the isomeric isoquinoline analogue 17 had slightly improved activity, with HAI=0.25 μM. This discrepancy in activity is most likely explained by the different orientations of the nitrogen, predicted to interact with R98 which mimics the carbonyl oxygen in 3. Interestingly, the analogue with improved activity, 17, directs the carbonyl mimic in an ‘endo-like’ conformation pointing in the direction of mannose, in contrast to 15, which assumes an ‘exo-like’ orientation pointing away from the mannose ring. To maintain this interaction, we proposed that cyclizing or ‘fusing’ the amide of 3 onto the B-ring phenyl group as with isoquinolone 16 would maintain this hydrogen bonding interaction and also aid in protecting the amide from hydrolysis. Indeed, we found that 16 shows a fivefold improvement in HA potency (HAI=0.1 μM) over 3. This result suggests that the isoquinolone amide carbonyl is oriented in a way that mimics the more energetically favorable binding conformation to FimH retaining the interaction with R98. This rationale can also be used to explain the increased activity of 7 over 5. The addition of the second meta-substituted methyl amide in 7 proposes that the conformation where the amide is in proximity to the salt bridge (and Y48) is statistically higher than 5.

Figure 3.

Potency and SAR of meta-position amide bioisostere analogues of biphenyl mannoside 3.

Based on these exciting results, we synthesized new mannosides 22 and 17b, the matched pairs of 16 and 17, by introducing an ortho-methyl group on the A-ring of the biaryl aglycone to increase potency. Mannoside 22 was synthesized using two different synthetic routes, as shown in Scheme 1A and 1B. In route 1A, aryl boronate mannoside 18^[11b] was coupled to 7-bromo-1(2H)-isoquinolone via a palladium-mediated cross-coupling to give protected tetraacetate 21. In route 1B, the bromoisoquinolone was first converted into the boronate ester 19. Suzuki cross-coupling of 19 to bromophenyl mannoside 20^[11b] gave 21. In both routes, mannose acetate deprotection was performed with sodium methoxide to give 22 in good yields. Isoquinoline 17b was synthesized in a similar fashion from 18 and 5-bromoisoquinoline. When tested in the HA assay, both isoquinolone 22 and isoquinoline 17b had a significant fourfold increase in activity (Table 1) relative to 16 and 17, respectively.

Scheme 1.

Synthesis of A-ring o-methyl B-ring isoquinolone mannoside 22. Reagents and conditions: a) bis(pinacolato)diboron, cat. Pd(dppf)Cl₂, KOAc, DMSO; b) cat. Pd(PPh₃)₄, Cs₂CO₃, dioxane/H₂O, 80 °C; c) NaOMe/MeOH.

Table 1.

In vitro biological and PK profile of initial and optimized mannosides.

Compd	HAI [μM]	Biofilm IC50 [μM]	DSF Tm [°C]	cLogD	Rat CL [mLmin−1kg−1][a]	Rat Vdss [Lkg−1][a]	Rat F [%][a]	Rat t1/2 [h][a]
2	1.0	0.94	73.4	1.17	ND	ND	ND	ND
3	0.50	1.35	73.1	0.28	ND	ND	ND	ND
5	0.062	0.16	75.5	0.93	98	1.3	1.4	2.8
7	0.016	0.12	75.8	1.35	ND	ND	ND	ND
17b	0.062	ND	ND	1.21	ND	ND	ND	ND
22	0.031	0.13	75.5	1.4	408	6.7	7.0	<3

^[a]Calculated from a 10 mgkg⁻¹ oral (p.o.) dose, 10% cyclodextrin formulation; ND: not determined.

Given the excellent potency displayed by 22, we wanted to more closely examine its biological behavior in both in vitro and in vivo assays. We first evaluated 5, 7, and 22 in a biofilm prevention assay followed by a thermal shift assay by differential scanning fluorimetry (DSF) to quantitatively calculate relative binding strength to the FimH lectin domain (Table 1). Early mannosides 2 and 3 prevent biofilm formation with IC₅₀ values of 0.94 and 1.35 μM, respectively, whereas 5 and 7 display ~10-fold higher potency (Table 1). We now report that 22 is equipotent to 5 and 7 in the biofilm and melting point shift assays. Upon the identification of 22, we next turned our attention to assessing the drug-like properties and pharmacokinetics of our compounds in order to prioritize them for in vivo studies. cLogD is often used as a good indicator of solubility. While mannosides are sugar-based, and are therefore predicted to have relatively good water solubility, our compounds have hydrophobic biaryl aglycones which encompass most of the molecular weight of the mannoside, causing the solubility to be variable. We calculated the LogD values of the five analogues in Table 1 and found that the change from a methyl ester 2 to amide 3 leads to a large drop in LogD from 1.17 to 0.28. Interestingly, introduction of the ortho-methyl group in amide 5 results in a large increase in LogD (0.93), almost recapitulating the solubility of ester 3. More surprising is that the addition of the second methyl amide in 7 further increases the LogD to 1.35, which is identical to that of cyclic amide/quinolone 22 (LogD 1.4). Therefore, we conclude that LogD is not well correlated to the improved activity of 7 and 22. Furthermore, evaluation in plasma and liver microsomes (Supporting Information) suggests that all compounds are metabolically stable in both mouse and human species.

Subsequently, we performed an in vivo pharmacokinetic (PK) study of the three lead mannosides in mice. In past studies, we have correlated compound levels in the urine at 6 h post-dosing as a good predictor of pharmacodynamic (PD) efficacy. Therefore, compounds were dosed via oral gavage (formulated in 10% cyclodextrin) at 50 mgkg⁻¹. Urine was collected at several time points (1, 3, and 6 h), and mannoside concentration was determined by mass spectrometry. As illustrated in Figure 4A, the initial urine concentration of 7 and 22 were elevated relative to 5 at 1 h post-doing, but level off and then display lower clearance for the remainder of the time course. A follow-up i.v./p.o. crossover study of 5 and 22 in rat (Table 1; Supporting Information), revealed that 22 (p.o. dose at 10 mgkg⁻¹) has higher bioavailability (F=7.0%) than 5, but also higher clearance (CL=408) and a larger volume of distribution (V_dss=6.7 Lkg⁻¹). Finally, we tested the lead mannosides in a murine model of chronic UTI, which was described previously.^[14] Briefly, introduction of bacteria into the urinary tract results in an acute infection, which in a subpopulation of mice can progress to a chronic infection as a result of an over-exuberant host response. Chronic infection is defined by severe pyuria and persistent bacteriuria. Elimination of bacteria in mice with chronic UTI requires antibiotic treatment. Shown in Figure 4B, all three compounds showed excellent efficacy relative to the control group, significantly dropping the bacterial burden (CFUs per bladder) of chronically infected mice by several log units. Excitingly, lead 22 displayed the best efficacy with a substantial 10-fold additional drop in CFUs per bladder relative to 5, 7, and isoquinoline analogue 17b (Supporting Information). The enhanced efficacy over 5 and 7 can be rationalized from the increased bioavailability of 22 and possibly the high volume of distribution (V_dss), indicating substantial tissue exposure, which is also supported by its rapid clearance (CL). The new lead mannoside 22 has elevated potency, better drug-like physical properties, bioavailability, and improved in vivo efficacy. To further improve the drug oral bioavailability and half-life, we synthesized a series of N-substituted mannosides based on 22, designed to explore structure–property relationships (SPR), specifically the effects of pK_a on half-life, clearance, tissue exposure, and oral bioavailability.

Figure 4.

Mannoside PK/PD. A) In vivo mouse PK of 5, 7, and 22 following oral delivery at 50 mgkg⁻¹. Urine mannoside concentrations over time indicate equivalent concentrations of 7 and 22. Data are the mean ± SEM from at least three animals. B) In vivo mouse PD model of UTI89 chronically infected mice dosed with 5, 7, and 22 at 50 mgkg⁻¹ p.o. (10% cyclodextrin formulation). Compound 22 shows significantly 10-fold enhanced efficacy relative to 5 and 7 at 6 h post-dosing. Differences were tested for significance within the designated treatments using Mann–Whitney U-test (****p<0.0001, *p<0.05).

To access this library based on 22, we used a synthetically feasible route for regioselectively substituting the isoquinolone nitrogen versus the oxygen. Thus, substitution reactions with isoquinolone 21 were performed with various alkyl bromides as shown in Scheme 2. Subsequent acetyl deprotection with NaOMe/MeOH was completed to obtain target mannosides 23a–d in respectable yields. Acetamide ester 23d was used as a key intermediate to access further diversity in substitution of the isoquinolone through installation of a terminal amide. To this end, saponification of 23d to carboxylic acid 24, followed by standard peptide coupling to various amines with HATU, produced isoquinolone mannosides 25a–d in excellent yields. These compounds were designed to combine the conformational restrictions of the isoquinolone ring and the increased basicity of the ester- and amine-bearing substituents, in an attempt to improve both affinity and PK. To verify that the rigidity afforded by the isoquinolone moiety in compounds 23–25 was responsible for the increase in binding observed, we synthesized compounds 28a–b to maintain similar interactions, while affording greater flexibility in the FimH binding pocket. We used a similar route involving a Suzuki cross-coupling of 4-methoxycarbonylphenylboronic acid to bromophenyl mannoside 20, followed by saponification of compound 26 to 27, and subsequent HATU-mediated amine coupling (Scheme 3). Subsequent to synthesis, the focused set of substituted quinolones were tested in the HA, biofilm, and DSF assays described above. Shown in Figure 5, almost all analogues of 22 had improved activity in the HA and biofilm assays at a level correlating well with the relative increased binding strength by DSF. The pyridyl acetamides 25b and 25c were identified as the most potent analogues, with nearly identical activity (HAI= 1 nM, biofilm IC₅₀=20 nM, and DSF T_m=76.3°C). While the two acyclic amide derivatives, 28a and 28b, display potency (31 and 62 nM) in the HA assay, their matched quinolone pairs, 23b and 23c, have better HAI titers of 4 and 8 nM. As elucidated by DSF melting point data, this notable eightfold enhancement in activity is derived largely from increased FimH binding as a result of the conformational restriction of the amide in quinolones 23b and 23c. Notably, the intermediate acid 24, used to synthesize 25b and 25c, showed a lower DSF melting temperature (T_m=74.8°C) and biofilm inhibition with an IC₅₀ value of 120 nM, whereas the HA inhibition assay showed an improvement, with HAI=16 nM. Interstingly, when evaluated for oral PK in rats (Supporting Information), we discovered that the two analogues 23b and 25d showed only minor evidence of bioavailability and when tested in the animal model of chronic UTI, analogues 23a and 23c showed no significant effect in decreasing the bacterial burden in mice when dosed orally (50 mgkg⁻¹). On the other hand, 28b displayed similar activity to that of 5 and 7 in the animal model (Supporting Information). The basis for the lack of oral bioavailability or efficacy in mice or rats remains unclear. All substituents of 22 have a basic, high pK_a and create a tertiary amide on the mannoside, so it is possible that either change prompts alternate metabolism in the stomach, gut, and/or plasma, resulting in the lack of oral exposure. Nonetheless, this new class of quinolone mannosides has the highest FimH affinity and potency of any antagonists reported to date (e.g., 25b HAI=1 nM). They are also capable of preventing bacterial biofilm formation of UPEC at activity levels well below that of any previously reported small molecule (e.g., 25c biofilm IC₅₀=18 nM).

Scheme 2.

Synthesis of isoquinolone mannoside analogues with varied pK_a substituents. Reagents and conditions: a) Br-R, NaH, DMF, 0°C; b) NaOMe/MeOH (0.02M), RT; c) NaOH (0.2M), MeOH, RT; d) HATU and DMF added at 0°C, DIEA and 1°/2°-amine-R′ added, brought to RT; e) NaOMe/MeOH (0.02M), RT.

Scheme 3.

Synthesis of acyclic mannoside amides as matched pairs to pyridyl isoquinolones 23b–c and 25 b–c. Reagents and conditions: a) cat. Pd(PPh₃)₄, Cs₂CO₃, dioxane/H₂O, 80°C; b) NaOMe/MeOH (0.02 M), RT; c) NaOH (0.2 M), MeOH, RT; d) HATU and DMF added at 0°C, DIEA and 1°-amine added and brought to RT.

Figure 5.

Summary plot of mannoside structure–activity data as measured by hemagglutination inhibition (HAI) assay (EC_>90), biofilm prevention assay (IC₅₀), and DSF thermal shift assays (T_m). Good correlation of mannoside activity in the HAI and biofilm prevention assays are observed when plotted against the relative binding strengths derived from the DSF melting point assay.

The development of small-molecule antagonists of the type 1 pilus adhesin FimH for the clinical treatment and prevention of UTI holds promise as an antibiotic-sparing strategy. Our current pursuit in identifying a compound for clinical development has involved a number of lead optimization studies on improving interactions with FimH, as well as understanding structure–property relationships from these modifications in vivo. In this study we pursued a structure-based approach in which we obtained a high-resolution X-ray structure of 7 bound to FimH for identifying key interactions aimed at improved potency. Mannosides containing simple isosteric or isoelectronic replacements of the biaryl B-ring amide^[11–12] have been previously studied and have similar potency and properties to those having the amide. In contrast, the conformationally restricted cyclic amide isoquinolone and isoquinoline isosteres reported herein produce improved mannosides with significantly elevated potency, better drug-like properties, and in vivo efficacy. Consequently, quinolone mannoside 22 has substantially improved bioavailability and a 10-fold increased efficacy in the treatment of chronic UTI in mice. Furthermore, substitution of 22 has led to a new class of the most potent isoquinolone-based mannosides ever reported. In summary, we have discovered new lead biaryl mannoside 22, which is a promising preclinical candidate drug for the treatment and prevention of urinary tract infections.

Abbreviations

UTI

urinary tract infection

UPEC

uropathogenic E. coli

SAR

structure–activity relationship

HA

hemagglutination

HAI

HA inhibition titer

IBCs

intracellular bacterial communities

PD

pharmacodynamics

PK

pharmacokinetics