Molecular tuning of farnesoid X receptor partial agonism

Abstract

The bile acid-sensing transcription factor farnesoid X receptor (FXR) regulates multiple metabolic processes. Modulation of FXR is desired to overcome several metabolic pathologies but pharmacological administration of full FXR agonists has been plagued by mechanism-based side effects. We have developed a modulator that partially activates FXR in vitro and in mice. Here we report the elucidation of the molecular mechanism that drives partial FXR activation by crystallography- and NMR-based structural biology. Natural and synthetic FXR agonists stabilize formation of an extended helix α11 and the α11-α12 loop upon binding. This strengthens a network of hydrogen bonds, repositions helix α12 and enables co-activator recruitment. Partial agonism in contrast is conferred by a kink in helix α11 that destabilizes the α11-α12 loop, a critical determinant for helix α12 orientation. Thereby, the synthetic partial agonist induces conformational states, capable of recruiting both co-repressors and co-activators leading to an equilibrium of co-activator and co-repressor binding.

Fig. 1

Structure and pharmacological profile of partial FXR agonist 1. a Constitution of 1 and endogenous agonist CDCA (2). b Dose-response curves of partial agonist 1, endogenous agonist 2 and ivermectin in a BSEP response element driven full-length FXR reporter gene assay. 1 partially activates FXR with an EC₅₀ value of 0.35 ± 0.06 µM and 21.1 ± 0.5% relative efficacy compared to the synthetic agonist GW4064 (3 µM). 1 also partially represses GW4064-induced FXR activation with an IC₅₀ value of 10.9 ± 0.2 µM to 26 ± 2% relative activation. CDCA (2) behaves as weakly potent (EC₅₀ = 4.0 ± 0.2 µM) agonist. Ivermectin has partial FXR agonistic activity (EC₅₀ = 0.060 ± 0.007 µM, 26 ± 1% rel. activation) and suppresses GW4064-induced FXR activation (IC₅₀ = 0.80 ± 0.16 µM, 22 ± 6% rel. act.). Results are mean ± SD, n = 3. c Partial agonistic activity of 1 was also observed in a hybrid Gal4-FXR reporter gene assay. d Profiling of 1 on NRs in hybrid reporter gene assays. Results are mean ± SEM, n = 3. e In vitro toxicity of 1 in HepG2 cells: 1 exhibits no acute toxicity in vitro up to 100 µM concentration. Results are mean ± SEM, n = 4. f In vitro metabolism analysis: 1 comprises good stability versus microsomal degradation with >60% of the parent compound remaining after 60 min. incubation. Results are mean ± SEM, n = 4. g Pharmacokinetic profile of 1 in C57BL/6j mice: With high oral bioavailability and a half-life of more than 2 h, 1 is suitable for in vivo studies. Results are mean ± SEM, n = 3. h Profiling of the effects of 1 on FXR regulated gene expression: Compared to endogenous FXR agonist CDCA, 1 partially induced FXR regulated genes in HepG2 cells (BSEP, SHP, CYP7A1, OSTα) and in HT-29 cells (IBABP, FGF19). Results are mean ± SEM, n = 3. i Hepatic mRNA levels of FXR-regulated genes upon treatment with 1 in mice: In mouse livers, 1 caused partial induction of SHP and partial repression of CYP7A1 compared to CDCA confirming its partial agonistic properties in vivo. Results are mean ± SEM, n = 3. Statistical significance was analysed by two-sided student’s t-test. ^#p < 0.1, *p < 0.05, **p < 0.01, ***p < 0.001 vs. 0.1% DMSO or as indicated. Source data are provided as a Source Data file

Fig. 2

Binding modes of physiological FXR agonist CDCA and partial agonist 1 to the FXR-LBD. a CDCA bound to FXR. The co-activator (NCoA) is highlighted in blue.  b Compound 1 bound to FXR. c, d Key residues involved in the interaction with CDCA and compound 1, respectively. e, f 2Fc-Fo omit maps (contour level 1.0 σ) for CDCA and compound 1 (chain B), respectively bound to the FXR-LBD. Chain B was considered for all interpretation due to preferable electron density for the ligand. Ligand density for chain A is poor due to twinning of the crystal. Additionally, the electron density of the loop region connecting α11 and α12 (V456-H459, AF-2 loop) in chain B is ambiguous or invisible, due to induced flexibility by binding of 1. g, h Compared to CDCA-bound FXR, binding of the partial agonist 1 causes an outward movement of W454 by 12 Å (2Fc-Fo omit maps, contour level 1.0 σ)

Fig. 3

FXR-LBD – ligand/activator (blue)/repressor (light-red) interactions. a (left) The natural FXR agonist CDCA (blue) stabilises α11-α12 (AF2-loop) and α12 (AF2-helix) while the α11 (purple) is partially formed. a (middle, right) Binding of the synthetic agonists fexaramine (yellow) and GW4064 (light blue) has a profound effect and extends the length of the α11. b Binding of ivermectin, destabilises the AF2-loop and in particular the α11 and the Ω-loop. The ivermectin containing FXR-LBD structure is complexed with NCoR-1 and thus represents an antagonistic conformation not fully reflecting ivermectin’s biological activity on FXR. c The partial agonist 1 by disturbing the AF2-loop influences the position of the AF2-helix and thereby affects position, orientation and binding of NCoA. In all structures a–c the α3-α4 loop is depicted in orange and the ligands are shown as sticks. d Overlay of the binding modes of different ligands. Key protein regions affected due to the ligand binding are highlighted with dashed-arcs

Fig. 4

¹H/¹⁵N-HSQC spectra of fully labeled FXR-LBD. Left panels show superposition of apo and apo+ligand, right panels show superposition of apo and apo + ligand + co-activator peptide. Blue—apo; red—with ligand; cyan—with ligand and co-activator peptide (NCoA). Concentrations of protein (200 µM), peptide (500 µM) and ligand (500 µM) were constant for all experiments to ensure saturation. All spectra were processed and compared at the same S/N. A representative signal is shown in each panel to visualise equal S/N. In unliganded state, the ¹H/¹⁵N-HSQC reveals only a fraction of all signals indicating conformational flexibility of part of the FXR-LBD. Upon addition of an FXR agonist (CDCA, GW4064 or fexaramine), additional but still not all signals appear suggesting partial stabilisation of the protein. Addition of co-activator peptide causes marked further stabilisation as observed by appearance of almost all ¹H/¹⁵N-HSQC signals. Addition of the FXR partial agonists ivermectin or 1 also induced stabilisation of the FXR-LBD but in contrast to the FXR agonists, addition of co-activator peptide had no further effect. Addition of the FXR antagonist guggulsterone had only minor/no effect on the HSQC indicating much less stabilisation of the FXR-LBD

Fig. 5

NMR studies on the FXR activation mechanism. a Amino acid sequence of co-repressor (NCoR) and the co-activator (NCoA) with ¹⁵N/¹³C-isotpically enriched amino acids highlighted with larger font. A stick model of leucine with its Hβ highlighted. The inset shows the Hβ-region of ¹H, ¹³C-HSQC spectra of the labelled peptides. The chemical shifts for the NCoA and NCoR are not overlapping and are easily distinguishable. b Schematic representation of the typical behaviour of an LBD in response to ligand-dependent activation and c resulting changes in the NMR signal (Hβ of leucine) of the co-repressor (NCoR) at different points of interaction. Disappearance of NMR signals upon binding to FXR, partial reappearing (release) upon addition of agonist and further reappearance upon addition of NCoA indicating complete release of NCoR. d Monitoring of the Hβ-signal of leucine in NCoR in response to addition of FXR, followed by addition of ligand (e–i) and the co-activator peptide: The Hβ signals of the co-repressor (d) are severely broadened upon addition of FXR, indicating binding. Upon addition of an agonists (e–g), an increase in signal intensity is observed indicating that agonist binding induces partial release of NCoR (e–g, left spectra). Further, upon addition of the co-activator peptide, the intensity of the NCoR signals further increases and the signals of NCoA disappear (blue circled region), indicating the complete release of the co-repressor and binding of NCoA to the FXR-agonist complex (e–g, right spectrum). The bar graphs adjacent to the spectra represent the relative populations of the released NCoR/NCoA peptides upon binding of the ligand. Populations of each reporter signal were determined from their intensities as P[(CDCA^NCoR) = 100∙I^NCoR/(I^NCoR + I^NCoR)] and vice versa. The spectra obtained in the presence of partial agonists (h, i) show marginal increase in the intensity of the repressor signals and minor line broadening for the signals of the activator, indicating that both co-repressor and the co-activator bind. This is also reflected by the relative populations of the released NCoR/NCoA peptides. Antagonist guggulsterone, in contrast, in this setting induces full binding of NCoR peptide with no recruitment of NCoA peptide (see Supplementary Fig. 7)

Fig. 6

NMR study on the interaction of α12 with the FXR-LBD core domain. a Model depicting transient interaction of the AF2 helix with the core domain. b Superimposition of ¹H, ¹⁵N-HSQC of fully 13C/15N-labeled FXR-LBD (blue) and an AF2 truncated FXR-LBD mutant (red) reveals significant shifts and disappearance of signals indicating that α12 interacts transiently with the core domain of the FXR-LBD even in the absence of a ligand/agonist

Fig. 7

Proposed model for partial agonism. Regions in the FXR-LBD structures with major changes induced upon binding of ligands (agonist, antagonist and partial agonist) are highlighted with arcs (red/dashed for antagonist characteristics, blue/bold for agonist characteristics). a In the apo FXR, the α11 adopts a loop structure and the α3-α4 region predominantly exists as a loop. b Co-repressor (NCoR) binds to the apo FXR. c Antagonist binding does not induce significant structural changes, however, destabilises the α11-α12 (AF2-helix). The α3-α4 loop adopts a loop structure and the Ω-loop (α2’-α3 loop region) is destabilised. d Binding of the natural agonist CDCA, induces partial formation of α11 (purple) and also the α3-α4 loop adopts a 3₁₀—helix conformation. e Binding of the synthetic agonist fexaramine, induces extended α11 formation (purple) and the α3-α4 loop adopts a 3₁₀— helix conformation. f Co-activator (NCoA) binds to the CDCA bound FXR-LBD. g Co-activator (NCoA) binds to the FXR-LBD bound to fexaramine. h Binding of the partial agonist 1, induces α11 formation (purple), destabilises the α11-α12 (AF2) and the Ω-loop region. The α3-α4 loop adopts a loop conformation as observed for the antagonist binding. i, j Partial agonist 1 bound FXR is competent of binding either the co-activator (NCoA) or the co-repressor (NCoR)