Structure-Based Drug Design of Quinoline CYP8B1 Inhibitors for the Potential Treatment of Metabolic Disease

Abstract

CYP8B1 has been identified as a potential target for the treatment of metabolic diseases, such as nonalcoholic fatty liver disease (NAFLD) and type 2 diabetes (T2D). CYP8B1 enzyme inhibitor 1 was discovered from a high-throughput screen of our compound collection; however, its poor metabolic stability prevented further in vivo evaluation. Leveraging structure-based drug design, metabolite identification and metabolic stability-guided DMPK optimization, and library synthesis to rapidly expand SAR, we identified compound 9 with substantially improved DMPK properties. Compound 9 showed robust in vivo CYP8B1 inhibition in mice and provided the foundation for further optimization of this class of CYP8B1 inhibitors.

Bile acids are amphiphilic molecules synthesized from cholesterol in the liver which function as physiological detergents that maintain cholesterol homeostasis. , Bile acid synthesis is a multi-step process consisting of two distinct pathways. The classical pathway (Figure ) accounts for the majority of bile acids produced. Hydrophobic cholic acid (CA), with a hydroxyl group at C12, is the major bile acid in the classical pathway. Cytochrome P450 (CYP), family 8, subfamily B, polypeptide 1 (CYP8B1) installs a hydroxyl group at C12 on 7α-hydroxy-4-cholesten-3-one (7-HCO), converting 7-HCO to 7α,12α-dihydroxy-4-cholesten-3-one (7,12-diHCO). Thus, CYP8B1 plays a critical role in 12α-hydroxy bile acids production, including CA. Inhibition of CYP8B1 reduces the amount of CA in the overall bile acid pool which causes an increase in the amount of alternative bile acids to make up for the deficiency. The alternate pathway in humans produces chenodeoxycholic acid (CDCA); whereas, in rodents the alternate pathway produces another major bile acid, muricholic acid (MCA).

1.

Formation of bile acids and the role of CYP8B1.

Mice with targeted disruption of CYP8B1 (CYP8B1^–/–) fail to produce CA. The fraction of the bile acid pool that would have been occupied by CA is predominantly replaced by MCA in CYP8B1^–/– mice. This leads to a reduction in both intestinal absorption and hepatic accumulation of cholesterol. , Singaraja and coworkers demonstrated therapeutic proof of concept for the treatment of NAFLD by targeting hepatic CYP8B1 in mice. Long-term knockdown (KD) of hepatic CYP8B1 by nonviral delivery of modified siRNA led to a profound remodeling of the BA pool, dramatically reducing lipid accumulation in the liver and inhibition of NAFLD.

It has been found that absence of CYP8B1 results in improved glucose tolerance, insulin sensitivity and β-cell function, mediated by absence of CA in CYP8B1^–/– mice. The absence of biliary CA results in reduced intestinal fat absorption and leads to increased free fatty acids reaching the ileal L-cells. Increased free fatty acids in the ileal L-cells cause the ileal L-cells to increase secretion of the incretin hormone glucagon like peptide-1 (GLP-1). GLP-1 in turn increases the biosynthesis and secretion of insulin from β-cells, leading to the improved glucose tolerance observed in the CYP8B1-/- mice.

Besides animal studies, Haeusler et al. examined the relationship between clamp-measured insulin sensitivity and plasma bile acid composition in a cohort of 200 healthy subjects and 35 T2D patients. They found higher 12α-hydroxy/non–12α-hydroxy bile acid ratios were associated with lower insulin sensitivity and higher plasma triglycerides. Furthermore, Zhong et al., reported haploinsufficiency of CYP8B1 in humans resulted in reduced 12α-hydroxy bile acids and increased muscle and whole-body insulin sensitivity.

Both the findings in CYP8B1^–/– mice and in humans suggest that reduction of 12α-hydroxy bile acids could be used in the treatment of metabolic disease, including NAFLD, and T2D. Thus, we are interested in generating potent and selective small-molecule CYP8B1 inhibitors. To date, no small molecule CYP8B1 inhibitor has been reported. Liu and coworkers recently reported their efforts to inhibit CYP8B1 through pyridine containing substrate-mimics, but these efforts were ultimately unsuccessful at coordinating to the active site heme iron as designed. Here we report our hit to lead effort in discovering a potent and orally bioavailable quinoline CYP8B1 inhibitor, which showed robust reduction of 7,12-diHCO in a murine target engagement assay.

High-throughput-screening of our sample collection using a CYP8B1 liver microsome assay yielded a quinoline series of CYP8B1 inhibitors. As a representative example, compound 1 (Figure ), showed good potency on both human and mouse CYP8B1 enzyme (h8b1 IC₅₀ = 5.6 nM, m8b1 IC₅₀ = 43.4 nM). In general, this class of compounds had clean off-target profiles (e.g., hERG IC₅₀ > 10 μM, CYP3A4 IC₅₀> 40 μM). The key issue of this class was poor metabolic stability. For example, compound 1 has high scaled microsomal intrinsic clearance (Cl int). Mouse Cl int is 4288 mL/min/kg, and rhesus Cl int is >1271 mL/min/kg. This poor metabolic stability precluded further in vivo studies on this class of compounds.

2.

Structure and profile of CYP8B1 inhibitors. Cl int = scaled microsomal intrinsic clearance.

To gain a better understanding of the rapid metabolism and enable rational design, compound 1 was subjected to in vitro metabolite identification in mouse liver microsomes. LC-MS analysis of the metabolites showed a major peak of M + H⁺ = 441. Structure 2 (parent plus 4 oxygens) was proposed as the corresponding phase I metabolite, suggesting that many positions of the molecule were metabolically oxidized by CYP enzymes (Figure ). Metabolic site prediction by MetaSite suggested that the two methyl groups and the two piperidinyl carbons adjacent to the nitrogens were metabolically labile.

3.

Proposed metabolites using mouse liver microsomes. (*) Metabolic sites predicted by MetaSite.

A crystal structure of compound 1 bound to the CYP8B1 enzyme was obtained (Figure ). One of the key interactions was a hydrogen bond network where the quinoline nitrogen formed a hydrogen bond to a conserved water molecule, which in turn interacted with iron of the heme cofactor. The quinoline ring formed an edge-to-face interaction with Trp281 of the CYP8B1 binding site. The cyano group formed a dipole–dipole interaction with the Phe482 carbonyl group. The amide carbonyl group in compound 1 formed a hydrogen bond with Tyr102. The terminal piperidine occupied a hydrophobic pocket.

4.

Crystal structure of compound 1 with human CYP8B1 (PDB 11IH). Part of the enzyme has been removed for clarity.

Given that the methyl group at the 6-position of the quinoline is a potential metabolic site, the metabolically stable 6-Cl (6a, Table ) was designed to replace the 6-methyl. The general synthesis of compound 6 was similar to published literature procedures and started with compound 3 (Scheme ). Chloride displacement with corresponding amine 4 provided compound 5. Amide coupling of 5 with the requisite amine afforded compound 6. Although 6a maintained activity on CYP8B1, no improvement in the overall metabolic stability was made.

1. Representative SAR on the Core .

^and = not determined. *Cl int = Scaled microsomal intrinsic clearance.

1. General Synthesis of Analogs .

^a Reagents and conditions: (a) Et₃N, DMF, rt; (b) NHR₁R₂, HATU, HOAt, DIEA, DMF, rt.

Two general strategies are known to mitigate phase I oxidative metabolism mediated by CYPs. One is increasing the steric hindrance and constraint of the compound. Another is introducing polarity to hydrophobic compounds to reduce the overall ClogP. Both strategies could reduce binding to CYPs, thus, improving metabolic stability. We applied both strategies to the optimization of these compounds.

The crystal structure suggested that we could further optimize van der Waals interactions around the central piperidine ring to improve the potency. Therefore, we increased the steric hindrance and constraint around the piperidine (analogs 6b–6h, Table ) to improve metabolic stability. Methyl analog 6b only showed modest CYP8B1 inhibition. An interesting SAR detail was that the smaller fluorine in analog 6c maintained good CYP8B1 potency, with little impact on metabolic stability. Looking at polar substitutions to reduce ClogP, hydroxy analog 6d (ClogP = 3.1) and the amino analog 6e (ClogP = 2.6) lost some potency, although 6e improved the metabolic stability. For bicyclic analogs, 6f was tolerated but did not improve metabolic stability, while 6g was much less potent at CYP8B1. After exploring symmetric cores, we turned our efforts to incorporating asymmetric cores. Interestingly, a single methyl on the core in 6h significantly boosted CYP8B1 potency, although it did not improve metabolic stability.

In parallel to the central core optimization, we studied the impact of the substitutions on the terminal piperidine (structure 7, Table ) on CYP8B1 potency and metabolic stability. The X-ray crystal structure suggested that substitution could be potentially tolerated on the 3 and 4 positions of piperidine. In order to enable rapid SAR exploration, parallel library synthesis was employed to prepare compounds with diverse sizes and polarities on 3 and 4 positions. Representative compounds 6i–6o are shown in Table . For analogs of the 4 position, compared with methyl analog 6a, CF₃ analog 6i and spiro cyclopropyl analog 6j maintained excellent CYP8B1 potency, while still exhibiting poor metabolic stability. On the other hand, the polar cyano analog 6k (ClogP = 0.77) showed modest potency with improved metabolic stability both in mouse and rhesus microsomal stability assays. This suggested that reducing the ClogP is a viable strategy to improve metabolic stability. Larger substituents such as a pyridyl group at the 4-position (6l) were not well-tolerated. Substitution at the 3-position was well-tolerated, since it reaches toward a solvent-exposed channel. For example, CF₃ analog 6m, pyridyl analog 6n, and tetrazole analog 6o all showed decent CYP8B1 potency, with the most polar analog 6o (ClogP = 1.7) having the best metabolic stability among the three analogs.

2. Selected SAR on the Amide .

^and = not determined. *Cl int = Scaled microsomal intrinsic clearance.

Given that both 3 and 4 positions of the piperidine were tolerated for CYP8B1 potency, and inspired by the strategy used for the discovery of DPP-4 inhibitor Januvia, we incorporated the 3,4-heterocyle fused piperazine/piperidine (structure 8, Table ) to optimize the ligand-enzyme interaction and reduce ClogP to improve metabolic stability. Of the polar heterocycles we studied, triazole analog 6q (Table ) showed better human CYP8B1 potency than imidazole analog 6p. We were delighted to find out that ethyl substitution (6r) showed good CYP8B1 potency and, more importantly, significantly improved metabolic stability compared with 6a. This further validated that reducing ClogP (0.73 vs 3.0, 6r vs 6a) was a viable strategy to improve metabolic stability. Replacing the ethyl with CF₃ (6s) further improved metabolic stability with a modest loss of CYP8B1 activity. Adding an additional methyl substituent to this compound reduced both potency and metabolic stability (6t). Compared to other heterocycles we surveyed, 6u had acceptable CYP8B1 potency, though reduced metabolic stability.

3. SAR of 3,4-Heterocycle Fused Piperazine/Piperidine .

^and = not determined. *Cl int = Scaled microsomal intrinsic clearance.

With metabolically stable 6s in hand, we sought to hybridize with the methyl piperdine moiety present in 6h to improve CYP8B1 potency. Indeed, the “magic” methyl group − in compound 9 significantly increased both human and mouse CYP8B1 potency (>50-fold, Figure ), as well as provided acceptable metabolic stability. An X-ray crystal structure of compound 9 bound to CYP8B1 showed that compound 9 maintained all the key interactions as seen in compound 1 (Figure ). Additionally, a new hydrogen bond interaction was observed between the nitrogen of the triazole and Thr355. The methyl on the central piperidine ring was in close contact with Trp281 and Phe482, which helped to explain the potency improvement.

5.

Structure and profile of compound 9.

6.

Crystal structure of compound 9 with human CYP8B1 (PDB 11IR). Part of the enzyme has been removed for clarity.

We were also particularly interested in the inhibitory activity of compound 9 against other CYPs and were pleased to discover that compound 9 was quite selective over all the CYPs we screened, showing >3000-fold selectivity (Table ) over all isoforms tested, likely due to the water-mediated binding mode as seen in Figure , rather than direct coordination. Compound 9 also exhibited a clean profile in cardiac ion channels and excellent permeability (P_app) in mouse MDR1a cells (Table ).

4. Profile of Compound 9 .

CYP IC50 (μM)
3A4	2D6	2C9	2C8	8A1	7A1	51	46	17
>50	>50	29	>50	1.4	22	>25	>25	>30

ion channels IC50 (μM)			Papp (10–6 cm/s)
Na 1.5	hERG	Ca 1.2	mouse MDR1a
>30	22	28	42

Compound 9 was then evaluated in pharmacokinetic studies in mice and rats (Table ) where it showed relatively high in vivo clearance in mice and modest clearance in rat. Regarding bioavailability, 9 achieved 43% F in mice and 28% F in rats. Despite moderate oral bioavailability, given that compound 9 has reasonable exposure in mice with a high unbound fraction (9.2% unbound) in mouse plasma and picomolar potency in mouse (IC₅₀ = 0.5 nM), the observed plasma exposures were expected to provide an unbound concentration sufficient to engage CYP8B1 in vivo, supporting evaluation in a mouse target engagement assay. Extensive work was done to identify the key metabolites. Chemical modification to block metabolism leads to an improved oral bioavailability. This work will be published in the near future.

5. PK Profile of Compound 9 .

PK	Clp (mL/min/kg)	t 1/2 (h)	F (%)	f u%
mouse C57BL/6 (2 mpk IV, 10 mpk PO)	65	0.4	43	9.2
Rat Wistar Han (2 mpk IV, 10 mpk PO)	27	0.8	28	9.9

In a mouse target engagement study, the level of 7,12 diHCO, the CYP8B1 enzymatic product was measured to determine the level of inhibition of CYP8B1. Mice were orally administered (PO) either vehicle (10% Tween-80) or compound 9 (0.3 mpk, 1 mpk, 3 mpk, and 10 mpk, Figure ). Compound 9 and 7, 12 diHCO concentrations were measured after 6 h. Despite low exposures (0.001 μM @ 0.3 mpk, 0.003 μM @ 1 mpk, 0.02 μM @ 3 mpk, 0.01 μM @ 10 mpk), compound 9 showed dose-dependent reduction of 7, 12-diHCO when compared to vehicle. The non-dose proportionality of plasma exposure is likely attributable to the variability of pharmacokinetics (PK) of compound 9. This compound exhibits high clearance and a short half-life (T _1/2 = 0.4 h) in mice leading to low exposures.

7.

Compound 9 showed a dose-dependent inhibition of CYP8B1 in mouse. ** p < 0.01 versus vehicle group via ANOVA and Dunnett’s post-test.

In summary, a novel class of quinoline CYP8B1 inhibitors was discovered from an HTS. Structure-based drug design, metabolic stability, and a metabolite identification study guided DMPK optimization, which led to the improvement of potency and metabolic stability. Lead compound 9 demonstrated robust CYP8B1 inhibition in mice. Further optimization of this class of compounds will be discussed in future papers.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.6c00387.

• Syntheses and characterization data for compound 9, assay protocols, and computational modeling (PDF)

Research was funded by Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA.

The authors declare no competing financial interest.