Carbohydrate scaffolds as glycosyltransferase inhibitors with in vivo antibacterial activity

The rapid rise of multi-drug-resistant bacteria is a global healthcare crisis, and new antibiotics are urgently required, especially those with modes of action that have low-resistance potential. One promising lead is the liposaccharide antibiotic moenomycin that inhibits bacterial glycosyltransferases, which are essential for peptidoglycan polymerization, while displaying a low rate of resistance. Unfortunately, the lipophilicity of moenomycin leads to unfavourable pharmacokinetic properties that render it unsuitable for systemic administration. In this study, we show that using moenomycin and other glycosyltransferase inhibitors as templates, we were able to synthesize compound libraries based on novel pyranose scaffold chemistry, with moenomycin-like activity, but with improved drug-like properties. The novel compounds exhibit in vitro inhibition comparable to moenomycin, with low toxicity and good efficacy in several in vivo models of infection. This approach based on non-planar carbohydrate scaffolds provides a new opportunity to develop new antibiotics with low propensity for resistance induction.


Virtual docking
The crystal structures with moenomycin A locates moenomycin in the donor binding site constituting the growing polysaccharide chain 3,8 , with the N-Ac group of ring F close to the catalytic site residue Glu100 (numbering by MGT from S. aureus). The virtual docking experiments using the 'best' receptor model 3HZSm (MGT from S. aureus), shows similar overall binding for both ACL20215 and ACL20964, but with different orientations of their benzimidazole groups. ACL20215 has its benzimidazole group close to site occupied by ring F of moenomycin, with the benzyl group along the donor site (occupied by ring E), pointing the trifluoro benzyl group towards the acceptor binding site. For ACL20964, the benzimidazole group is still close to site occupied by ring F of moenomycin, but the naphthyl group is orientated towards the acceptor site, while the trifluoro benzyl group towards the donor site. Even though the orientation of both inhibitor is switched around, their binding orientation is dominated by the three hydrophobic groups: trifluoro benzyl group, benzimidazole, and benzyl or naphthyl group.
One located at the donor site (similar to ring E), one oriented towards the membrane (similar to portion G of moenomycin) and one orientated towards the acceptor site. This last interaction site is not occupied by moenomycin itself.

Conformation of Inhibitors
The compounds with benzimidazole (ACL20215 and ACL20964) exist as atropisomer 13 , conformational isomers, due to the bulkiness of the benzimidazole group in vicinity to the carbohydrate ring, restricting the rotation of the benzimidazole linkage bond to the carbohydrate scaffold. An energy barrier of around 25 kcal mol -1 for this bond which -in relation to other known atropisomerscorresponds to an interconversion rate of a few hours and up to few days 14 . This corroborates our observations that while two isomers could be detected in the NMR experiment, they could not be separated using chromatography.

Conformational (rotamer) search of ACL20215
Both monosaccharide compounds showed distinct conformations in the 1 H-NMR, which is due to the benzimidazole group restricting the free rotation of the C 4 -N Benzimidazole bond. In order to calculate the rotation energy barrier of that bond, ACL20215 was built using Sybyl (Tripos, St.Louis), and the most predominant 4 C1 ring conformation for the monosaccharide scaffold. The structure was energy minimized using the MMFF96s force field before using the systematic torsion grid search, which

Hemolysis assay
Human male red blood cells were isolated from fresh whole blood sample and washed with phosphate buffered saline (PBS) 3 times. Red blood cells were diluted 1/200 with PBS for the haemolysis assay.
Each compound was assayed at 4 different concentrations in quadruplicate. Control items (100 μM tamoxifen and 1% triton X100) were also assayed in quadruplicate. Background was determined with 1% DMF. The assay was performed for 1 h at 37 °C. The samples were then centrifuged for 10 min at 4°C at 1000 × g, the supernatant was recovered and absorption was measured at 415 nm. Results are expressed as the % haemolysis compared to 1% triton X100 after background was subtracted, and fitting dose response curve with Prism (GraphPad) to extract HC50 values.

Bacterial Cytoplasmic Membrane Permeability Assay
The depolarization of the cytoplasmic membrane of S. aureus (ATCC25923) caused by the antibiotics was determined using the membrane potential-sensitive cyanine dye diSC35 and using the method described by Wu et al. 15 with some modification. Briefly, the overnight subcultures of bacteria were grown in LB broth at 37 °C with shaking to mid-logarithmic phase (OD600 = 0.5 to 0.6). Cells were collected by centrifugation at 5,000 × g for 10 min, washed once with the buffer (5 mM HEPES, pH 7. where Fm is the maximal fluorescence increase after the addition of a compound at a particular concentration; Fi is the initial fluorescence (the fluorescence level of cell suspension just after the addition of compound) and Fb is the fluorescence of blank. Citropin 1.1, which is known to disrupt membrane integrity of bacteria, was used as a positive control (100%) at a concentration of 32 µg mL -1 in this study [16][17][18] .

Inhibition of Lipid II and peptidoglycan biosynthesis
The cell-free particulate fraction of Bacillus megaterium KM (ATCC13632), capable of catalysing the polymerisation of PG from UPD-linked precursors was performed as described previously 19  The results with PBP1 from C. difficile showed some inhibition for ACL20215, while the experiment with ACL20964 displays no detection of either the substrate (NBD-lipid II) or the product of the glysyltransferase reaction (marked in Supplementary Figure 23 with 'Strange behaviour'). However the experiment with MGT A from S. aureus shows some inhibitory effect for both the compounds. These preliminary results were used to optimise the experiment to extend it to a dose response assay.

In vivo efficacy and PK studies
The

In vivo Maximum Tolerated Dose
Compound dissolved in 0.05% Tween 80/0.9% NaCl were each administered IP at doses of 20, 60 and For the corresponding IV studies compounds at 4 mg kg -1 were dissolved in 10% DMSO/0.1 M HCl/1% Tween 80/5% glucose, and compounds were each administered intravenously at 10 min and/or 180 min after S. aureus (Smith) challenge.

In vivo Mastitis mouse model
For the infection of the mice S. aureus Newbould 305 (ATCC 29740) was used , which was originally isolated from a clinical mastitis case 21 . Bacterial preparation was previously described 22 . Briefly, an overnight brain heart infusion (BHI; Oxoid Limited) culture of the bacteria was diluted in sterile phosphate-buffered saline (PBS; Gibco) and quantified by flow cytometry (BD biosciences).
Subsequently, 1 ml of a 1000-fold PBS-diluted bacterial suspension was added to a TRUcount tube (BD biosciences), which contained a known number of fluorescent beads. The number of bacteria was then calculated using the following equation: The actual colony forming units (CFU) of the inocula was confirmed by overnight culture of a serial logarithmic dilution on tryptic soy agar (TSA; Oxoid Limited) plates.

Determination of the MIC of compounds for the experimental bovine mastitis isolate S. aureus
Newbould 305 was performed using the Mueller-Hinton agar dilution assay according to CLSI guidelines 23 . Plates were incubated at 35 °C (+/-2 °C) for 16-20 h in an aerobic atmosphere. In addition the MIC of a random selection of recent bovine mastitis field isolates (10 isolates for each antibiotic) was determined in order to ensure that the sensitivity of the experimental isolate concurred with the sensitivity of a naïve bacterial population encountered in the field.
The procedure for mouse mammary gland infection has been recently described 22  After sacrifice of the mice by cervical dislocation at 14 h post-treatment, mammary glands (two per mouse) were harvested, weighed and homogenized on ice in sterile PBS using a TissueRuptor (QIAGEN Benelux BV). The mammary glands, which are structurally separate, were considered as individual samples. Bacterial CFU counts were obtained after quantification of serial logarithmic dilutions of mammary gland homogenates on TSA plates and transformation of the raw CFU counts into base 10 logarithm values. The detection limit (DL) was 1.7 log10 CFU g -1 gland weight.
The mean intramammary CFU g -1 gland of infected glands from mice treated with different doses of the antimicrobial compound (test group) are compared with the mean CFU g -1 gland of infected glands from mice injected with excipient only (negative control group). To evaluate the dose-response relationship for the antimicrobial the ED2logCFU, ED4logCFU, PD50 and PD100 values were calculated. The ED2logCFU and ED4logCFU are the concentrations of the antimicrobial needed to reduce the mean CFU g -1 gland from the control group by 2log10 and 4log10, respectively. The PD50 and PD100 are the doses of equiv.). The reaction was stirred at room temperature for two days, diluted with DCM (25 ml), washed with water (2 x 25 ml), dried (MgSO4) and the solvent was removed in vacuo to give disaccharide 4.

Synthesis of ACL19378.
To a solution of disaccharide 4 in dry DMF (69 mol, 0.5 ml, 0.138 M) in the glove box was added a solution of 4-(trifluoromethoxy)phenyl isocyanate (17 mg, 83 mol, 1.2 equiv.) in dry DMF (0.2 ml, 0.41 M). The reaction was hand shaken and left to sit in the glove box overnight.
The solvent was removed in vacuo. To the resulting disaccharide (49 mol) was added a solution of TFA (5%), triethylsilane (15%) in dry DCM (4 ml). The reaction was stirred at room temperature for 2 h and the solvent was removed in vacuo. The resulting disaccharide was then dissolved in a solution of saturated ammonia in methanol (3 ml). The reaction mixture was heated at 55 °C for 2 h and the solvent was removed in vacuo to give ACL19378. The crude product was purified by LCMS on standard C18 column using a water/acetonitrile gradient.

Analysis data for ACL19378 (NMR and MS)
MS spectra of ACL19378 were obtained using an AB Sciex TripleTOF® 5600 (ABSCIEX, Canada) equipped with a DuoSpray Ion Source. ACL19378 was infused into the MS system using a Harvard Syringe Pump at 10 µL min -1 . Data was collected in (+)-ESI-mode from m/z 50 to 1000 and product ions from m/z 50 to 1000 in high resolution TOF mode (collision energy (CE) 20, ionspray voltage 5300V, declustering potential (DP) 100V, curtain gas flow 25, nebuliser gas 1 (GS1) 12, GS2 to 5, interface heater at 150 o C and the turbo heater to 450 o C). NMR data were collected in DMSO-d6 using a Bruker Avance 600 MHz spectrometer (Bruker, Germany) at the temperatures noted with spectra calibrated to residual solvent signals (δH 2.49 and δC 39.0). .