In Silico Study of Obeticholic Acid Analogs as FXR Agonists ^†

Abstract

Nuclear receptors are ligand-activated transcription factors that, in response to lipophilic hormones, vitamins, and dietary lipids, regulate numerous aspects of mammalian physiology. Bile acid receptors represent well-defined targets for the development of novel therapeutic approaches for metabolic and inflammatory diseases. The farnesoid X receptor (FXR) was identified as an orphan steroid receptor-like nuclear receptor, and its activation is crucial in many physiological functions of the liver. A vital function of FXR is to influence the amount of bile acids in hepatocytes by reducing bile acid synthesis, stimulating the bile salt export pump, and inhibiting enterohepatic circulation, thereby protecting hepatocytes from toxic bile acid accumulation. FXR activation induces distinctive changes in circulating cholesterol in animal models and humans. We present an evaluation of the interaction of various obeticholic acid analogs and other bile salts by studying their binding energies and receptor-ligand interactions using AutoDock 4.2.6 software. The results open the possibility of using new alternatives by deriving structures at position 3 of the steroid nucleus.

1. Introduction

The farnesoid X receptor (FXR), also known as NR1H4, is a prominent member of the nuclear receptor superfamily [1,2], found mainly in organs such as the liver and gastrointestinal tract. This receptor has been the subject of particular interest due to its pivotal role as a bile acid sensor in enterohepatic tissues. FXR positively regulates the cholesterol catabolism and exerts feedback inhibition on bile acid synthesis [3], underlining its importance in maintaining lipid homeostasis. Furthermore, this receptor is suggested to influence the regulation of plasma triglycerides, energy balance, and glucose homeostasis [4]. Given its multifunctionality, FXR is considered a promising potential target for the development of novel pharmacotherapies targeting metabolic diseases, including non-alcoholic fatty liver disease (NAFLD). In recent years, the synthesis and biological evaluation of steroidal FXR agonists have been extensively investigated [5,6,7,8,9,10,11,12,13]. Modifications were mainly based on the backbone and side chains of chenodeoxy cholic acid (CDCA). Of all the backbone derivatives, obeticholic acid with an ethyl group at the 6-position was found to be the most potent FXR agonist. The 6α-ethyl group may reside in a hydrophobic cavity within the ligand-binding domain of FXR, a key structural element for FXR potency [14].

2. Results and Discussion

In this communication we present a study of the interaction of a group of obeticholic acid derivatives with various functional groups on the crystal structure of the nuclear receptor FXR 1OSV obtained from the RSCB Protein Data Bank [15]. This structure presents obeticholic acid as a ligand and therefore seems ideal for comparing the behavior of the different OCA derivatives. For this purpose, the coordinates of the protein chain A of 1OSV were used and, using the Autodock 4.2.6 software [16] with MGLTools 1.5.7 [17], the different binding energies of compounds 1–42 were calculated (Figure 1), for which the binding energy values indicated in

Table 1. Binding energies of compounds 1–42.

Compound	-R1	-R2	kcal/mol
1	β-HOCH2CH2O-	-Et	−13.41
2	α-SO3NH2	-Et	−12.94
3	α-azido	-Et	−12.86
4	=NOH	-Et	−12.79
5	β-t-Bu	-Et	−12.78
6	α-OPO3PO3-	-Et	−12.67
7	β-N3	-Et	−12.58
8	α-OH (OCA)	-Et	−12.32
9	β-CN	-Et	−12.31
10	α-CN	-Et	−12.1
11	=NHNH2	-Et	−12.08
12	α-NH2	-Et	−12.07
13	α-SH	-Et	−11.99
14	β-SH	-Et	−11.96
15	α-t-Bu	-Et	−11.94
16	β-CH3	-Et	−11.89
17	β-OPO3PO3-	-Et	−11.86
18	β-NHCSNH2	-Et	−11.82
19	β-NHCONH2	-Et	−11.8
20	α-CH3	-Et	−11.79
21	β-imidazole	-Et	−11.64
22	β-OCH3	-Et	−11.64
23	-H	-Et	−11.53
24	β-OH	-Et	−11.44
25	α-OH (CDC)	-H	−11.30
26	α-OCH3	-Et	−11.4
27	α-HOCH2CH2O-	-Et	−11.22
28	β-NH2	-Et	−11.17
29	α-NHCSNH2	-Et	−11.16
30	α-NHCONH2	-Et	−11.16
31	α-PO3	-Et	−11.16
32	β-PO3	-Et	−10.95
33	α-OPO3	-Et	−10.91
34	α-imidazole	-Et	−10.81
35	β-NHC(NH)NH2	-Et	−10.78
36	β-OPO3	-Et	−10.5
37	β-CO2	-Et	−10.42
38	α-CO2	-Et	−9.61
39	α-NHC(NH)NH2	-Et	−9.52
40	α-SO3-	-Et	−8
41	β-SO3-	-Et	−7.96
42	β-SO3NH2	-Et	−7.61

were obtained.

Ligands were evaluated using SwissADME [15], to assess pharmacokinetics and small molecule drug-likeness, predictive models of physicochemical properties, pharmacokinetics, drug-likeness and compatibility with medicinal chemistry, including the BOILED-Egg diagram (an intuitive graphical classification model for gastrointestinal absorption and brain access) [16] (Figure 2).

In Figure 2, the white region is for a high probability of passive absorption by the gastrointestinal tract, the yellow region (yolk) is for a high probability of brain penetration, the blue dots indicate the prediction that they are actively effluent by P-gp (PGP+) and the red ones if it is predicted that they will not be effluxed from the central nervous system by P-glycoprotein P-gp (PGP−). Other results obtained are listed in

Table 2. SwissADME predictions for compounds 1–41. Ha: hydrogen bond acceptor, Hb: hydrogen bond donor, TPSA: topological polar surface area (Ų), LogP: consensus LogP_o/w, GIA: gastrointestinal absorption according to boiled egg diagram, BBBP: brain blood barrier permeation, PGP: P-glycoprotein substrate, CYP: cytochrome P450 inhibitor.

Cpd	Ha	Hb	TPSA	LOG	GIA	BBB	PGP	CYP1A2	CYP1A2	CYP2C19	CYP2C9	CYP2D6
1.27	5	2	89.82	3.89	High	No	No	No	No	No	No	Yes
2.42	6	2	128.9	3.14	Low	No	Yes	No	No	No	No	No
3.7	6	1	110.11	4.35	High	No	Yes	No	No	No	No	No
4	5	2	92.95	4.31	High	No	Yes	No	No	Yes	No	No
5.15	3	1	60.36	6.28	High	No	No	No	No	No	No	No
6.17	10	1	201.76	2.51	Low	No	No	No	No	No	No	Yes
8.24	4	2	80.59	3.90	High	No	Yes	No	No	No	No	No
9.10	4	1	84.15	4.64	High	No	Yes	No	No	Yes	No	No
11	4	2	98.74	4.09	High	No	Yes	No	No	No	No	No
12.28	3	2	88	2.44	High	No	Yes	No	No	No	No	No
13.14	3	1	99.16	4.63	High	No	No	No	Yes	No	No	No
16.20	3	1	60.36	5.03	High	Yes	No	No	No	No	No	No
18.29	3	3	130.5	3.88	Low	No	Yes	No	Yes	No	No	Yes
19.30	4	3	115.48	3.35	High	No	Yes	No	No	No	No	Yes
21.34	4	1	78.18	4.19	High	No	Yes	No	No	No	No	No
23	3	1	60.36	4.80	High	Yes	No	No	No	No	No	No
25	4	2	80.59	3.35	High	No	Yes	No	No	No	No	No
22.26	4	1	69.59	4.32	High	Yes	No	No	No	No	No	No
31.32	6	1	133.36	3.28	Low	No	Yes	No	Yes	No	No	No
33.36	7	1	142.59	3.25	Low	No	No	No	Yes	No	No	No
35.39	4	4	122.26	3.32	High	No	Yes	No	No	No	No	Yes
37.38	5	1	100.49	3.72	High	No	Yes	No	No	No	No	Yes
40.41	6	1	142.76	3.05	Low	No	Yes	No	Yes	No	No	No

. These include the number of acceptor and donor atoms in hydrogen bonds in each ligand molecule, the topological polar surface area (TSPA), the mean LogP, the gastrointestinal absorption (GIA), the permeation across the blood–brain barrier (BBB), the consensus LogP by various methods, transport by P-glycoprotein (PGP), Knowledge about whether the compounds are substrate or non-substrate of the permeability glycoprotein is key to assess the active efflux through biological membranes, for example, from the gastrointestinal wall to the lumen or from the brain. An important function of P-gp is to protect the central nervous system (CNS) from xenobiotics [15] and the prediction about its ability to inhibit cytochromes P-450: 1 A2, 2C19, 2C9, 2D6, and 3 A4; the inhibition of these isoenzymes is undoubtedly one of the main causes of pharmacokinetic interactions between drugs, which cause toxic adverse effects or other unwanted adverse effects due to the lower clearance and accumulation of the drug or its metabolites [15].

For each of the ligands, the complex that the Autodock program indicated with a more favorable binding energy—that is, more negative—was chosen. As a reference, the 1OSV crystal structure was analyzed, and the ligand-receptor interactions were extracted using the Discovery Studio Visualizer v25.1 program [17] to identify the residues that interact with the ligand (Figure 3).

The noncovalent interaction (NCI) method, also known as the reduced density gradient (RDG) method, is very popular for studying weak interactions [18]. These interactions are crucial for determining how a ligand, such as a drug or a substrate in a chemical reaction, binds to a target molecule, such as a protein or a catalyst. Calculating the reduced density gradient (RDG) using the promolecular approach is a reasonably reliable alternative for working with macromolecules, where the binding mode of proteins and ligands is studied; it avoids the high time and resource costs of ab initio calculations. The promolecular density is constructed by simply superimposing the electron densities of the free atoms and can therefore be easily evaluated. RDG (Figure 4) and NCI surfaces (Figure 5 and Figure 6) were calculated with software Multiwfn 3.8 [19,20].

The energy binding calculated for obeticholic acid is −12.32 kcal/mol. Complexes 1–7 present more favorable binding energies, the interaction with near residues to 3-C substituent with TYR358 remains in 1, 2, 4, HIS444 in 3, 6, 7 and HIS444, MET447, and TRP451 for 5.

In complex 9, the β-cyano group in 3C has no interaction with residues but the epimer 10 with the α-cyano group interacts with HIS444. In complex 11, hydrazone interacts with TYR358, as does complex 12 which is under zwitterionic form, but this also shows an unfavorable positive-positive interaction with HIS444.

Thiol complexes 13 and 14 both show interaction by hydrogen bonding with TYR358, but 14 shows additional pi-sulfur interactions with HIS444.

Chenodeoxycholic acid (CDC, 25) lacks the ethyl substituent at 7-C present in obeticholic acid (OCA, 8). Interactions with nearest residues are similar; only the ethyl substituent interactions are absent as expected (Figure 7 and Figure 8).

In Figure 9b it can be observed that there are no interactions between 7-C and residues in the peptide. When complex 23 is calculated, the interactions with MET362, PHE363, and TYR366 present in 8 are maintained even with the absence of hydroxyl group in 3-C (Figure 10).

Compounds 2, 18, 33, 31, 32, 36, 40, 41, and 42 are out of scale in the BOILED-Egg diagram, so in principle they would be discarded despite the favorable binding energy of 2. If we consider the binding energy of chenodeoxycholic acid as a lower limit, the compounds that would penetrate the BBB would be 16 (β-CH3), 20 (α-CH3), 22 (β-OCH3), 23 (-H), and 26 (α-OCH3), which could be interesting options. The rest of the compounds present acceptable gastrointestinal absorption and PGP and CYP inhibition.

3. Conclusions

After having studied 41 variants for the substitution at carbon 3 of the skeleton of obeticholic acid (8) and chenodeoxycholic acid (25), it has been found that compounds 1–7 have better binding energies than 8. Of these 7 compounds, only 2 is not expected to present a good gastrointestinal absorption, so that compounds with the following functional groups at carbon 3: β-HOCH2CH2O-, α-SO3NH2,α-N3, =NOH,β-tBu,α-OPO3PO3-,β N3 would be candidates with potential biological activity for in vitro tests.