1,1-Difluoroethyl-substituted triazolothienopyrimidines as inhibitors of a human urea transport protein (UT-B): New analogs and binding model

Abstract

The kidney urea transport protein UT-B is an attractive target for the development of small-molecule inhibitors with a novel diuretic (‘urearetic’) action. Previously, two compounds in the triazolothienopyrimidine scaffold (1a and 1c) were reported as UT-B inhibitors. Compound 1c incorporates a 1,1-difluoroethyl group, which affords improved microsomal stability when compared to the corresponding ethyl-substituted compound 1a. Here, a small focused library (4a–4f) was developed around lead inhibitor 1c to investigate the requirement of an amidine-linked thiophene in the inhibitor scaffold. Two compounds (4a and 4b) with nanomolar inhibitory potency (IC₅₀ ≈ 40 nM) were synthesized. Computational docking of lead structure 1c and 4a–4f into a homology model of the UT-B cytoplasmic surface suggested binding with the core heterocycle buried deep into the hydrophobic pore region of the protein.

Kidney urea transporters (UTs) are required for the formation of a concentrated urine.^1,2 Epithelial cells in some kidney tubules express UT-A proteins, which are encoded by the SLc14A2 gene, and endothelial cells in some microvessels express UT-B, which is encoded by the SLc14A1 gene.^3–5 Knockout mice lacking various UTs manifest impaired urinary concentrating ability, suggesting the application of UT inhibitors as ‘urearetics’ with a novel diuretic mechanism of action.^6–9

Early high-throughput screening using a human erythrocyte lysis assay identified phenylsulfoxyoxozole inhibitors of human UT-B with IC₅₀ ≈ 100 nM.¹⁰ Limitations of the original compounds included poor metabolic stability and poor activity against rodent UT-B. Further screening using a mouse erythrocyte lysis assay yielded triazolothienopyrimidine UT-B inhibitors (Fig. 1), with the most potent compound, 1a, having IC₅₀ ≈ 11 nM for inhibition of human UT-B and 25 nM for inhibition of mouse UT-B.¹¹ Administration of 1a to mice reduced urine osmolality and increased urine output. Compound 1a and related analogs inhibit UT-B by a competitive mechanism.

Figure 1.

Previously characterized trizaolothienopyrimidine inhibitors of UT-B.

Though 1a was effective in mice when administered in high concentration, it underwent rapid oxidation by hepatic enzymes, mostly at the benzylic position of the ethyl substituent. A collection of 273 commercially available analogs of 1a (exploring the R¹ position as well as variations at the amidine linkage) were evaluated, with no improved leads identified.¹² One interesting analog (1b) was synthesized containing a trifluoromethyl group at position R¹, which resisted microsomal degradation, but suffered significant loss of potency against UT-B. Reasoning that a two carbon atom substituent was required for potency, but that benzylic hydrogens are prone to radical abstraction, we synthesized the 1,1′-difluoroethyl inhibitor (1c), and found that the inhibitor resisted microsomal degradation with little effect on UT-B inhibition activity (IC₅₀ ≈ 14 nM) of the original lead 1a.¹² We chose to incorporate the amidine-linked thiophene-2-methyl group into 1c, which has generally been associated with more potent and stable inhibitors in this class. Herein, we prepared a small library of analogs of 1c, in which alternative amidine linked groups were evaluated for UT-B inhibitory potency and hepatic microsomal stability. Indeed, while the stability of 1c was enhanced greatly by the incorporation of the 1,1′-difluoroethyl substituent, a slow microsomal oxidation reaction could still be observed by LCMS, which we speculate might be modulated by variation of the amidine substituent.

The synthetic approach to generate this library of inhibitors, with differing amidine substituents tethered from the core heterocycle is summarized in Scheme 1. As was described previously, 4- bromo-(1,1-difluoroethyl)-benzene (2) was transformed to the corresponding triazolothienopyrimindine (3) by [2+3] cycloaddition with an azidothiophene ester.^12,11 Subsequently, lactam 3 was coupled with various primary amines, using PyBOP and DBU under microwave irradiation, to generate a library of amidine derivatives (4a–4f). While amidine formation using PyBOP and DBU has been described previously,^13,14 we observed that microwave irradiation was necessary to effect quantitative conversion in this class of molecules. Typically the compounds could be isolated with simple aqueous workup, and purified by trituration with cold methanol. All of the compounds were characterized by ¹H NMR and HRMS, and had a characteristic triplet (³J = ~18 Hz) due to coupling between the terminal methyl and neighboring difluoromethylene groups. Yields and spectroscopic characterization of these compounds are provided in Supplementary data.

Scheme 1.

Synthesis of amidine analogs of triazolothienopyrimidine lead structure 1c. Reagents and conditions: (a) described previously;¹² (b) 1° amines, PyBOP, CH₃CN, DBU, microwave irradiation, 100 °C, 30 min.

The library was designed to explore the effect of the amidine substituent on UT-B inhibition potency and on microsomal stability (Table 1). Compounds 1a and 1c were the previously characterized inhibitors.^11,12 The new inhibitors synthesized in this library had amidine substituents designed to explore: methyl substitution on the thiophene (4a), an isomeric thiophene connection point (4b), replacement of the thiophene heterocycle for furan (4c) or benzyl (4d), and the use of small hydrophobic alkyl ring systems such as cyclopropylmethyl (4e) or cyclobutylmethyl (4f).

Table 1.

UT-B inhibition activity^a and microsomal stability^b of synthesized compounds

Compd	R1	R2	IC50 (nM)	% Remainingb
1a			11	<5
1c			14	96
4a			38	22
4b			43	47
4c			353	14
4d			>800	79
4e			>800	75
4f			>800	24

^aDetermined by mouse erythrocyte lysis assay (n = 3). IC₅₀ determined by fit to a single-site saturation model.

^bPercentage of parent compound remaining after 30 min incubation with rat liver microsomes.

UT-B inhibition potency was determined by an erythrocyte lysis assay, as previously described¹⁰ in which lysis is measured in acetamide- loaded erythrocytes following dilution into isosmolar PBS (Supplementary data). Erythrocytes express the UT-B protein, which facilitates the transmembrane transport of urea, acetamide and other small polar solutes. UT-B inhibition increases hypotonic lysis, as measured by near-infrared light absorbance, because water influx is not opposed by acetamide efflux. Microsomal stability was determined by incubation of compounds in the presence of rat liver microsomes and NADPH, followed by organic extraction and LCMS analysis, as described¹² (Supplementary data).

In general, modification of the amidine substituent reduced inhibition potency and microsomal stability. Two of the more potent new inhibitors evaluated were 4a (containing 4-methylthiophene) and 4b (thiophene connected at the 3 position). Both of these inhibitors had modest potency, but suffered in terms of microsomal stability when compared to 1c. An analog containing furan (4c) instead of thiophene (1c) showed poor potency and stability, which was surprising due to the similarity between the thiophene and furan heterocycles. This is consistent with another observation, in which the furan analog of 1a was was found to have greatly reduced potency.¹² Replacement of the amidine-linked thiophene with non-heterocyclic rings such as benzyl (4d), cyclopropylmethyl (4e), and cyclobutylmethyl (4f) also greatly reduced potency. In the case of 4d and 4e, these modifications mildly improved microsomal stability compared to 1a.

Molecular modeling and docking was done to investigate the potential inhibitor binding site. While a high-resolution structural model of UT-B has not been reported, we constructed a homology model for UT-B using the SWISS MODEL utility, using the full human UT-B1 protein sequence, and have previously used this model to postulate a mode of binding of 1a.¹¹ Compounds 1c and 4a–4f were docked into the same homology model using FRED v2.2.5 (OpenEye Scientific).^15,16 Figure 2 shows a tentative model for binding of inhibitor 1c, zoomed in and out, with the key interactions shown. We offer several observations from the modeling:

Figure 2.

Model of the binding of 1c with the cytoplasmic channel region of human UT-B based on a homology model: (a) zoomed-in and (b) zoomed-out depictions; (c) potential interactions between 1c and UT-B cytoplasmic surface residues.

1. The computed mode of binding of inhibitor 1c is consistent with our previously reported model of 1a. A similar mode of binding was found for 4a and 4c–4f, with the orientation of the core heterocycle, aryl sulfone, and amidine substituents being consistent (Supplementary Fig. S1). In the postulated binding mode, the core heterocycle is buried deep into a hydrophobic pocket surrounded by Leu¹²¹, Leu³⁶⁵, Phe³⁰¹, Phe⁷¹, and other amino acids. There is a potential hydrogen bonding interaction between the Asn¹⁶⁵ sidechain and an exposed triazole nitrogen. The aryl sulfone is surrounded by Ser¹⁶⁹, Ser³⁶⁵, and Asp¹⁷², while the thiophene group resides near the Thr³⁶⁸ methyl group and Lys³⁶⁶. Overall, the inhibitor has a good complementary fit to the surface of UT-B, and aside from the one potential hydrogen bonding interaction, most likely is bound due to hydrophobic interactions with non-polar amino acid residues. Inhibitor 4b is an exception compared to the other compounds, in that the lowest energy conformation presented the thiophene deep into the pocket.

2. The docking software used has six independent scoring functions that evaluate protein–ligand interactions using different algorithms. Of the scoring functions available, five reliably distinguished the relatively active inhibitors (1a, 1c, 4a and 4b) from the relatively inactive inhibitors (4c–4f): Shapegauss, PLP, Chemgauss3, OEChemScore, and Screenscore. The sixth scoring function (ZapBind), which weights heavily toward electrostatic interactions, was not useful at discriminating active from inactive inhibitors in this series, which is expected given the limited electrostatic interactions in our postulated binding mode.

3. Deciphering the structural determinants of the active and inactive compounds is challenging. The thiophenemethyl amidine compounds (1a and 1c) appear to have the most complementary overlap, giving rise to the most potent compounds. Alkylation of the thiophene (4a) or simple use of an isomerically connected thiophene (4b) decreases potency, albeit not catastrophically. Most puzzling is the severe drop in potency in the furan-based inhibitor (4c), for which docking identified a similar mode of binding as the more potent thiophene-based inhibitor 1c. Nevertheless, with the FRED scoring functions named above (excluding ZapBind) all correctly ranked 1c as a better scoring ligand than 4c. The remaining less active inhibitors (4d–4f) were scored poorly by FRED, and apparently the amidine linked substituent had less complementary fit compared to the more active compounds.

In summary, a focused library of six analogs of UT-B inhibitor 1c was prepared to characterize the structural requirements for potency and microsomal stability. Two additional compounds with modest potency (4a and 4b) were discovered, but none were able to match 1c. Computational docking into a homology model of UT-B gave a tentative mode of binding for 1c and 4a–4f. While caution should be used in over-interpreting docking results from homology models, we were gratified to see that docking with multiple independent scoring functions effectively distinguished the active from inactive analogs, and that with exception of one compound (4b), the docked conformations were consistent.