Intestinally-targeted TGR5 agonists equipped with quaternary ammonium have an improved hypoglycemic effect and reduced gallbladder filling effect

TGR5 activation of enteroendocrine cells increases glucagon-like peptide 1 (GLP-1) release, which maintains glycemic homeostasis. However, TGR5 activation in the gallbladder and heart is associated with severe side effects. Therefore, intestinally-targeted TGR5 agonists were suggested as potential hypoglycemic agents with minimal side effects. However, until now no such compounds with robust glucose-lowering effects were reported, especially in diabetic animal models. Herein, we identify a TGR5 agonist, 26a, which was proven to be intestinally-targeted through pharmacokinetic studies. 26a was used as a tool drug to verify the intestinally-targeted strategy. 26a displayed a robust and long-lasting hypoglycemic effect in ob/ob mice (once a day dosing (QD) and 18-day treatment) owing to sustained stimulation of GLP-1 secretion, which suggested that robust hypoglycemic effect could be achieved with activation of TGR5 in intestine alone. However, the gallbladder filling effect of 26a was rather complicated. Although the gallbladder filling effect of 26a was decreased in mice after once a day dosing, this side effect was still not eliminated. To solve the problem above, several research strategies were raised for further optimization.


Part 1. Medicinal chemistry
Our initial goal was to find a new scaffold with higher activity compared with 2 in vitro.
Extensive explorations of the structure-activity relationship (SAR) of tetrahydroquinoxaline and phenoxy moieties have been done. However, the middle six-membered ring (pyridine of 2) was thought to be crucial for activity, and few modifications were carried out. In our on-going work, the middle pyridine of 2 was replaced with several five-membered rings. Among them, compound 9a (Table S1) containing a thiophene ring was the most active. In addition, 9a exhibited high logP and high liver microsomal clearance (1631/1390 mL/min/g for human/mice), which might decrease its systemic exposure according to studies on non-systemic drugs 1-4 . Therefore, a 3-phenoxythiophene-2-carboxamide scaffold was chosen for further SAR exploration.
On the basis of our initial SAR exploration, the tetrahydroquinoxaline moiety was critical for TGR5 activity and only 1-cyclopropyl-tetrahydroquinoxaline or 1-methyl-tetrahydroquinoxaline was tolerated. Substitutions at positon 4 or 5 of the thiophene ring were detrimental for activity (structure and data not shown). Hence, we focused on the phenoxy side chain and N4 positon of the tetrahydroquinoxaline. Human TGR5 (hTGR5) shares 83% amino acid sequence identity with mouse TGR5 (mTGR5) 5 , therefore, hTGR5 and mTGR5 were determined for most compounds.
SAR of the phenoxy side chain and N4 positon of the piperazine ring. Replacement of the pyridine ring of 2 with thiophene increased the activity of hTGR5 and mTGR5 (9a, Table S1). Removal of 2-Cl (9b) resulted in considerable reduction in potency, but replacement of 2-Cl of 9a with 2-Br (9c) retained activity. No significant change of activity was found after attachment of an additional 4-Br to 9a (9f). N-c-Pr derivatives displayed a 2-4-fold increase in activity compared with their N-Me derivatives (9a vs. 9d, 9b vs. 9e). According to SAR exploration of compound 1 by Roche, the CF 3 group is beneficial for hTGR5 activity 6 . Therefore, 2,5-di-Cl-phenyl of 9a and 9d was replaced with 4-CF 3 -Phenyl, but the resulting compound 9g and 9h showed decreased activity of hTGR5. Pleasingly, introduction of an additional 2-Br or 2-Cl substitution to 9g or 9h elicited a dramatic improvement in potency (9i-k), among which 9i exhibited comparable activity with 9a. In our previous work, introduction of a methyl group contributed to activity 7 . Therefore, the 2-Br group of 9i was replaced with 2-methyl, but the corresponding compound 10 displayed a slight decrease in activity. Taken together, 9a, 9c, 9f, and 9i were the most potent in our series. However, compounds containing Br demonstrated poor solubility (9c, 9f, 9i, 9j), which caused difficulty in syntheses and subsequent biological assay in vivo. Therefore, 9a was chosen for further modification. However, in a subsequent assay in Institute of Cancer Research (ICR) mice, 9a increased the gallbladder area by 210% at once a day oral dosing of 50 mg/kg (data not shown). This side effect was considered to be owing to its moderate membrane permeability (Papp = 0.55 × 10 -6 cm/s, Table S3), which might be decreased considerably by incorporation of quaternary ammonium. According to the SAR stated above, 2,5-dichloro substitution of the phenyl ring was critical for TGR5 activity, whereas the presence of 4-Br had little effect. Hence, we introduced the quaternary ammonium group to the 4-position of the phenyl ring, and a series of derivatives containing quaternary ammonium were synthesized (Table S2).

SAR of the linkers and substitution of quaternary ammonium derivatives.
At first, the role of linker a was explored. When the amide was attached directly to the phenyl ring (20), TGR5 activity was almost lost compared with 9a. Pleasingly, extension of linker a to a vinyl group (23) led to dramatic improvement in activity. Reduction of the double bond afforded the more active TGR5 agonist 26a, with an EC 50 of 4.1 and 0.71 nM for hTGR5 and mTGR5, respectively. However, further extension of linker a to allyl or propyl (30 and 33) resulted in a big decrease in activity. Then, replacement of N-methyl in the amide moiety with an ethyl group (26b) resulted in a slight decrease in mTGR5 activity, which suggested that substitution of the amide group with a large group was detrimental to activity.
Next, we turned our attention towards linker b and substitution of quaternary ammonium. Prolongation of linker b to a propyl group (26c) was unfavorable for mTGR5 activity. Cyclization of the amide N atom with quaternary ammonium gave compound S6 26d with slightly lower activity compared with 26a. Further placement of quaternary ammonium outside the ring yielded a compound (26e) with comparable activity for hTGR5, but its activity in mTGR5 was decreased slightly. Finally, we further introduced high-polarity group (quaternary ammonium or carboxylic acid group) to 26c and 26a, which resulted in a large decrease in TGR5 activity (26f, 27a-b).

S14
which could be hydrogenated to 31. Hydrolysis of 28 and 31 and further condensation with 39a yielded 30 and 33, respectively.

Figure S4. Synthesis of (E)-4-phenylbut-3-enamide and
Where k (min -1 ) is the slope of the line of the natural log of the percentage parent remaining versus the time of incubation; P is the microsomal protein concentration (mg/mL).
Drug transport from the apical side to the basolateral side (A-B) and from the basolateral side to the apical side was measured simultaneously under the same condition. Propranolol and atenolol were used as the hypertonic and hypotonic control, respectively. Digoxin was used as the positive control for Pgp-mediated drug efflux. In brief, the method was as follows. After washing the monolayer with HBSS three times, the compounds were diluted and added to the appropriate well (pH 6.8 for apical side and pH 7.4 for basolateral side). The plate was incubated at 37 °C for 95 min. Samples were collected from the donor side at 5 and 95 min, and from the receiver side at 35 and 95 min post-incubation. The concentration of samples was measured by LC-MS/MS. The P app was calculated from the following equation: Where V A is the volume of the acceptor well, SA is the surface area of the membrane, T is the total transport time, [drug] acceptor is the drug level at the acceptor side, and [drug]initial donor is the drug level at the donor side at T = 0.
Oral glucose tolerance test (OGTT) and gallbladder filling assay in ICR mice. 100 mg/kg of 26a or 0.25% CMC (control) was administered orally to overnight-fasted ICR mice (n = 7-8) 90 min prior to the oral glucose load (4 g/kg). Blood glucose levels were measured via blood drops obtained by clipping the tail of the mice using an Accu-Chek Advantage II Glucose Monitor (Roche, Indianapolis, IN, USA) before compound dosing and 0, 15, 30, 60, and 120 min after the glucose load. The area under the concentration−time curve from 0 to 120 min (AUC 0−120 min , Glu) of blood glucose after the glucose load was calculated by the trapezoidal rule.
After the OGTT experiment, the fasted mice were refed for 3 h, gallbladders were then removed, and the area was measured using vernier caliper. The relative gallbladder area was calculated from the length multiplied by the width of the gallbladder. The bile weight was measured using Analytical Balances.
GLP-1 secretion in ob/ob mice. The ob/ob mice (female) were divided into 4 groups and treated with 0.25% CMC (control), 26a (100 mg/kg), DPP-4 inhibitor Linagliptin (3 mg/kg), 26a (100 mg/kg) plus Linagliptin (3 mg /kg), while the latter three groups were then divided into 3 subgroup of different time point (n = 8-9 each subgroup). All the animals were fasted for 6 hours before collecting blood samples at the indicated time points. At other times, animals were free accessed to water and food. 6 h, 12 h, 24 h later, blood samples were collected and placed in Eppendorf tubes containing the DPP-4 inhibitor valine pyrrolidide (Linco Research, DPP-4-010) with a final concentration of 1% blood samples and 25 mg/mL EDTA. Concentrations of active GLP-1[7-36 amide] in the serum were measured using ELISA kit from Linco Research. All time point shared the same control group collected after 12 h.
Pharmacokinetics assay of 26a in ob/ob mice. Compound 26a (100 mg/kg) was orally administered to 12 h-fasted ob/ob mice (n = 3 per time point). Blood and intestinal tissue samples were collected before dosing or 2 h, 4 h, 6 h, 10 h 16 h, 24 h after dosing. Blood samples were centrifuged at 11000 rpm for 5 min to isolate the plasma. 25 μL of the sample was mixed with 75 μL MeOH. After centrifugation at 11000 rpm for 5 mins, the supernatant was analyzed by LC-MS/MS. The tissues were ground and homogenated with 0.9% NaCl solution (1:4, w/v), and then the homogenates were centrifuged at 12,000 rpm for 10 min at 4 °C. The supernatant was collected, and 100 μL aliquot of supernatant was spiked with 300 μL methanol, and then treated as the same fashion as plasma sample. Density of intestinal tissue was taken as 1 g/ml.
Gallbladder filling assay in ob/ob mice. 100 mg/kg of 26a or 0.25% CMC (control) was administered orally to overnight-fasted ob/ob mice (male, n = 3-6). 1 h after dosing, mice were refed for 3 hours, and then gallbladders were removed, and the area was measured using vernier caliper. The relative gallbladder area was calculated from the length multiplied by the width of the gallbladder. The bile weight was measured using Analytical Balances.
While in the three days gallbladder filling assay of 26a, 2 groups of ob/ob mice were administered to 100 mg/kg of 26a or 0.25% CMC (control) for 3 days, while the following procedure were the same as the once a day dosing assay. Blood, bile and gallbladder tissue samples in this assay were collected for further drug level test.
Drug levels test in plasma, bile, and gallbladder tissue after 3 days oral administration to ob/ob mice. Plasma, bile, and gallbladder were taken from experiments performed in accordance with the guidelines for the use of experimental animals in the Shanghai Laboratory Animal Administration (Shanghai, China) after three days treatment of 26a to ob/ob mice. Blood samples were collected through cardiac puncture into heparinized tubes, and plasma was harvested by centrifugation. Bile and gallbladder tissue were collected from each mouse after blood withdrawal. To 10 mg of gallbladder was added 100 μL of MeOH, and the mixture was homogenized. A 15 μL aliquot of bile samples was diluted with 100 μL of MeOH and then mixed. Plasma, bile, and tissue homogenate were stored at −20 °C. These samples were protein precipitated with MeCN using clopidogrel as the internal standard, and the supernatant was injected onto a LC-MS/MS system for the quantification of 26a. Chromatographic separation was performed on a Waters Acquity UPLC BEH C18 (1.7 µm, 2.1 × 50 mm) at a flow rate of 0.5 mL min −1 using ternary mobile phase: A (H 2 O with 5 mM ammonium acetate and 0.5% TFA); B(MeCN: H 2 O = 95:5 with 0.1% TFA ); C (MeOH with 1% TFA). The MS/MS detection was carried out in MRM mode using a positive electrospray ionization interface.
Long-term study of 26a in ob/ob mice. Three groups of ob/ob mice (male) was orally administered to 26a (50 mg/kg), 26a (100 mg/kg) and 0.25% CMC (control) for 18 days. At day 0 (before dosing), 4, 8, 12 and 18, non-fasted blood glucose was measured before dosing. Mice were fasted for 6 h after dosing and then fasted blood glucose was measured. The serum level of HbA 1c was measured at day 0 (before dosing) and 18 by Adicon biochemical analyzer. The body weight of each mouse was recorded every day. The triglyceride and total cholesterol level was measured after the final dose.
Statistical analysis. All data were expressed as the mean ± SEM or mean ± SD. For the experiment including multiple time points, one-way ANOVA statistical analysis was used; otherwise, unpaired Student's t test was used. P < 0.05 was considered to be statistically significant.

Part 5. Experimental procedure (chemistry)
Synthetic materials and methods. All reagents were purchased from commercial suppliers and used without further purification. Microwave reactions were performed in a Biotage Initiator. Column chromatography was carried out on silica gel (200-300 mesh) or with pre-packed silica cartridges (4-40 g) from Bonna-Agela Technologies Inc. (Tianjin, China) and eluted with a CombiFlash @ Rf 200 from Teledyne Isco. Prep-HPLC separation was carried out in Unimicro Easysep-1010 series LC( UV 254 nM, 25 °C, flow rate = 10 mL min −1 ) with the column of Agilent Prep-C18 10 μm, 21.2 × 250 mm, while the mobile phase was 30-90% MeCN/H 2 O (containing 0.1%TFA) in 1h. Melting point (mp) of target compounds was measured by SGWX-4 melting point apparatus (Shanghai Precision and Scientific Instrument Corporation, Shanghai, China). 1 H NMR and 13 C NMR spectra were recorded on a Varian-Mercury Plus-300 or a Bruker Avance III 400 or a Bruker Avance III 500 NMR spectrometer using tetramethylsilane or solvent signals as an internal reference. IR spectra were recorded on IS5 FT-IR (Thermo). High-resolution mass spectra (ESI) were obtained on a Q-TOF or Thermo Orbitrap Elite. The purity of tested compounds was determined by HPLC (Agilent LC1260, Agilent ChemStation, ZORBAX SB-C18, 5 μm, 4.6 × 150 mm, UV 254 nM, 30 °C, flow rate = 1.0 mL min −1 ). All of the assayed compounds possess >95% purity. General procedure for preparation of compound 6a-c. To a solution of 5 (1.20 g, 7.59 mmol) in dry DMF (25 mL) was added K 2 CO 3 (1.15 g，8.32 mmol) and the corresponding substituted 2-fluoronitrobenzene (7.59 mmol). The reaction mixture was stirred at room temperature overnight and then poured into water (50 mL) and extracted with ethyl acetate for three times. Organic layers were combined, washed with saturated brine for three times, dried over anhydrous magnesium sulfate and evaporated under vacuum. The residue was purified by flash column chromatography to yield the desired compound.   General procedure for preparation of compound 7a-g. To a solution of the 2-nitro benzene intermediate 6a-c (1.06 mmol) in THF (50 mL) and H 2 O (50 mL) was added iron power (0.47 g, 8.4 mmol) and NH 4 Cl (113 mg, 2.11 mmol). The resulting mixture was stirred at 65 °C overnight. Then the reaction mixture was cooled to room temperature, filtered and the filtrate evaporated in vacuum, diluted with water and extracted with ethyl acetate for three times. Organic layers were combined, washed with saturated brine for three times, dried over anhydrous magnesium sulfate and evaporated under vacuum to yield the crude product which was subjected to the next step without further purification.
A solution of the crude product obtained above in MeCN or DMF was added CuCl 2 (2.12 mmol) and t BuONO (2.12 mmol), or CuBr 2 (2.12 mmol) and t BuONO (2.12 mmol), or t BuONO (2.12 mmol). The resulting mixture was stirred at room temperature for 0.5 h and then heated at 60 °C or 80 °C overnight. Then the reaction mixture was cooled to room temperature and evaporated in vacuum, diluted with water and extracted with ethyl acetate for three times. Organic layers were combined, washed with saturated brine for three times, dried over anhydrous magnesium sulfate and evaporated under vacuum. The residue was purified by flash column chromatography to yield the desired compound. General procedure for preparation of compound 8a-g. To a solution of 7a-g (0.66 mmol) in 1,4-dioxane (10 mL) and H 2 O (10 mL) was added NaOH (1.32 mmol). The resulting mixture was stirred at room temperature overnight. The reaction solution was then concentrated under vacuum, diluted with water, acidified with HCl (4N). The solid was filtered and dried at 45 °C overnight to afford the desired acid 8a-g.   General procedure for preparation of compound 9a-k. To a solution of 8a-g (3 mmol) in DCM was added DMF (3 drops) and (COCl) 2 (9 mmol). The resulting mixture was refluxed for 3 h. After the reaction was cooled to room temperature, the solvent and (COCl) 2 was removed under vacuum. Then the residue was dissolved in DCM before adding Et 3 N (9 mmol) and 1-cyclopropyl-1,2,3,4-tetrahydroquinoxaline (3 mmol) or 1-methyl-1,2,3,4-tetrahydroquinoxaline (3 mmol). The resulting mixture was stirred at room temperature overnight, diluted with water, extracted with DCM. The organic layer was combined, washed with saturated brine for three times, dried over anhydrous magnesium sulfate, and evaporated under vacuum. The residue was purified by flash column chromatography to provide 9a-k as target compounds.

4-Methoxybenzyl 3-(2,5-dichloro-4-(methoxycarbonyl)phenoxy)thiophene-2-carboxylate (16).
To a solution of 15 (1.9 g, 4.0 mmol) in THF (80 mL) and H 2 O (40 mL) was added iron power (1.8 g, 32 mmol) and NH 4 Cl (0.43 g, 8.0 mmol). The resulting mixture was stirred at 60 °C overnight. Then the reaction mixture was cooled to room temperature, filtered, evaporated in vacuum, diluted with water and extracted with ethyl acetate for three times. Organic layers were combined, washed with saturated brine for three times, dried over anhydrous magnesium sulfate and evaporated under vacuum to yield the crude product which was subjected to the next step without further purification.
To a solution of the crude product obtained above in MeCN (50 mL) was added CuCl 2 (0.80 g, 6.0 mmol) and t BuONO (0.95 mL, 8.0 mmol), The resulting mixture was stirred at room temperature for 0.5 h and then heated at 60 °C overnight. Then the reaction mixture was cooled to room temperature, evaporated in vacuum, diluted with water and extracted with ethyl acetate for three times. Organic layers were combined, washed with saturated brine for three times, dried over anhydrous magnesium sulfate and evaporated under vacuum.

Methyl 3-chloro-4-fluoro-5-nitrobenzoate (36).
To a mixture of 35 (4.6 g, 24 mmol) and H 2 SO 4 (1.7 mL) was added the solution of H 2 SO 4 (2.2 mL) and HNO 3 (2.2 mL) at 0 °C. The resulting mixture was stirred at room temperature overnight. Then the reaction mixture was poured into water, extracted with ethyl acetate, washed with aqueous NaHCO 3 , H 2 O and brine, dried over anhydrous magnesium sulfate, and evaporated under vacuum. The residue was purified by flash column chromatography (petroleum ether/ethyl acetate=50:1) to provide 2.5 g (43%) of 36 as a white solid. 1   General procedure for preparation of compound 38a-c or 43. To a solution of 37a-c or 42 (20 mmol) in THF (80 mL) and H 2 O (80 mL) was added NaHCO 3 (2.52g, 30 mmol) followed by Boc 2 O (5.13 mL, 30 mmol). The resulting mixture was stirred at room temperature overnight. Then the reaction mixture was evaporated under vacuum, diluted with water, extracted with DCM for three times. The organic layer was combined, dried over anhydrous magnesium sulfate and evaporated under vacuum. The residue was purified by flash column chromatography (DCM/MeOH=10:1) to provide 38a-c or 43.      (N,N,N-trimethylpropan-1-aminium) dichloro hydrochloride (47). To a solution of 46 (0.19 g, 0.66 mmol) in MeCN (10 mL) was added CH 3 I (162μl, 2.60 mmol). The resulting mixture was stirred at room temperature overnight. Then the reaction mixture was evaporated under vacuum and dissolved in DCM to which 4N HCl/EA (2 mL) was added. The resulting mixture was stirred at room temperature overnight. Then the reaction mixture was evaporated under vacuum, dissolved in DCM/MeCN and cooled to -10 °C. The solid formed was filtered, washed with DCM, and dried under vacuum to afford 47 in yield of 65%. 1 H NMR (400 MHz, Methanol-d 4 ):