Discovery of Potent Selective Nonzinc Binding Autotaxin Inhibitor BIO-32546

Abstract

Autotaxin (ATX) is a lysophospholipase D that is the main enzyme responsible for generating LPA in body fluids. Although ATX was isolated from a conditioned medium of melanoma cells, later it was discovered to play a critical role in vascular and neuronal development. ATX has also been implicated in primary brain tumor, fibrosis, and rheumatoid arthritis, as well as neurological diseases such as multiple sclerosis, Alzheimer’s disease, and neuropathic pain. As ATX and LPA levels are increased upon neuronal injury, a selective ATX inhibitor could provide a new approach to treat neuropathic pain. Herein we describe the discovery of a novel series of nonzinc binding reversible ATX inhibitors, particularly a potent, selective, orally bioavailable, brain-penetrable tool compound BIO-32546, as well as its synthesis, X-ray cocrystal structure, pharmacokinetics, and in vivo efficacy.

Autotaxin (ATX or ENPP2) is an extracellular enzyme which is one of the seven members of the ecto-nucleotide pyrophosphatase/phosphodiesterase family with structurally related catalytic domains.¹ Originally isolated from a conditioned medium of melanoma cells in 1992,² ATX was discovered to play a major role in the development of vascular³ and nervous⁴ systems and malignancy.⁵ Its enzymatic activity remained to be an enigma until it was shown to be identical to glycoprotein lysophospholipase D (lysoPLD) in 2002.⁶ ATX is the main enzyme responsible for generating lysophosphatidic acid (LPA) in body fluids by catalyzing the hydrolysis of lysophosphatidylcholine (LPC) to LPA and choline.⁷

LPA is a bioactive phospholipid with diverse functions in almost every mammalian cell line.⁸ LPA signals through six distinct G-protein-coupled receptors (i.e., LPA1–6), leading to various cellular responses including the stimulation of cell migration, proliferation, and survival. The ATX–LPA signaling pathway is implicated in a variety of human pathologies including angiogenesis,⁹ autoimmune diseases,¹⁰ cancer,¹¹ fibrotic diseases,¹² inflammation,¹³ neurodegeneration,¹⁴ and neuropathic pain,¹⁵ among others.

Targeting the ATX-LPA signaling pathway for therapeutic potential has gained increasing interest from both academic and industrial settings and been reviewed extensively.^16,17 Both lipid-like and nonlipid ATX inhibitors have been discovered, with X-ray cocrystal structures disclosed for zinc binder LPA (14:0, 1)¹⁸ itself, and HA-155 (2).¹⁹ Nonacid potent inhibitors like 3 (ATX IC₅₀ = 2.2 nM)²⁰ are emerging, and so far, only one ATX inhibitor GLPG1690 (4, ATX IC₅₀ = 131 nM) has moved to clinic trials directed toward idiopathic pulmonary fibrosis (IPF) (Figure 1),^21,22 but it has stopped at phase 3 trials recently. Given the increasing interest of targeting ATX for neurology indications, a brain-penetrable, orally bioavailable ATX inhibitor is desirable but has yet to be reported.

Figure 1.

Chemical structures of LPA 14:0 (1), HA-155 (2), Lilly’s ATXi (3), and GLPG1690 (4).

Encouraged by the evidence that ATX and LPA levels are increased upon neuronal injury and that production of LPA up-regulates pain-related proteins through the LPA1 receptor,²³ we were particularly interested in developing an ATX inhibitor for neuropathic pain. Very recently, it was reported that an ATX inhibitor ONO-8430506 could ameliorate neuropathic pain symptom in CD rat model at 30 mg/kg,^24,25 which further strengthens this hypothesis. Our initial screening of an internal lipid mimetic chemical library derived from our previous S1P work²⁶ and subsequent optimization identified a hit compound 5 (Figure 2), a phosphonic acid-based inhibitor that inhibits ATX in the FRET assay with good potency (IC₅₀ = 28 nM). Extensive optimization guided by structural-based rational design led to the discovery of a potent, selective, orally bioavailable, brain-penetrable tool compound BIO-32546 (6, IC₅₀ = 1 nM), a novel nonzinc binding reversible ATX inhibitor. Herein we describe the synthesis, structure activity relationship, as well as X-ray cocrystal structure, pharmacokinetics, and in vivo efficacy.

Figure 2.

Chemical structures of hit 5 and BIO-32546 (6).

The preparation of this series of compounds was exemplified with the synthesis of compound 6 as depicted in Scheme 1. Mesylation of alcohol 7 followed by phenol displacement led to ether 8, which was iodinated and converted further to the CF₃ intermediate 9. Transmetalation of the bromide followed by quenching with DMF gave aldehyde 10, which was methylated to yield secondary alcohol 11 with good yield. Sequential mesylation, followed by amine displacement and ester hydrolysis gave the racemate 14, which was eventually separated under chiral SFC conditions to give compound 6 and its enantiomer. A single crystal X-ray of compound 6 (Figure 3) was obtained as an HCl salt which unambiguously established the absolute configuration. Its analogues were all prepared in a similar way.

Scheme 1. Synthesis of Compound 6.

Reagents and conditions: (a) i. MsCl, NEt₃, CH₂Cl₂, rt, 99%; ii. 7-bromonaphthalen-2-ol, Cs₂CO₃, DMF, 80 °C, 55%; (b) i. NIS, ZrCl₄, CH₂Cl₂, reflux, 94%; ii. FSO₂CF₂CO₂Me, CuI, HMPA, DMF, 80 °C, 98.5%; (c) BuLi, DMF, THF, −78 °C, 97.3%; (d) MeMgBr, THF, 0 °C, 92.6%; (e) MsCl, DIPEA, CH₂Cl₂, rt; (f) Cs₂CO₃, DMF, 80 °C; (g) LiOH, THF, rt, (34.2% combined from e–g); (h) SFC separation, 30%.

Figure 3.

Single-crystal X-ray structure of compound 6-HCl salt.

As our initial hit 5 is a phosphonic acid, which typically lacks good oral bioavailability,²⁷ our first move was replacing the phosphonic acid with a carboxylic acid with the aim to reduce the polar surface area (tPSA = 96 for 5 and 76 for 15). A significant drop in potency was observed with compound 15 (IC₅₀ = 3700 nM). Switching to amino acid 16 (IC₅₀ = 1300 nM) further lowered tPSA to 59 and showed a modest improvement on potency. Encouraged by this result, different amino acid linkers were screened, and eventually we found that the restricted amino acid 17 showed reasonable activity (IC₅₀ = 370 nM) with a low tPSA of 50 and cLogP of 3.43. Tackling the whole head moiety turned out to be quite fruitful as depicted in Table 1. Replacing the piperidine in compound 17 with a more steric hindered azabicyclo[3.3.1]nonane and replacing the tail CF₃ with CH₃ further improved the potency (18, IC₅₀ = 38 nM), while maintaining the molecular weight below 500.²⁸ Switching the regiosubstitution from 6- to 7- led to a potency breakthrough to the single digit nm range with compound 19 (IC₅₀ = 8.2 nM).

Table 1. SAR on R1 and R2 Positions.

^aHuman ATX FRET assay with FS-3 as substrate (mean, n ≥ 2).

^bChantest fastpatch.

At this stage, we began to profile this series of compounds, and immediately we noticed that 17–19 all showed more than 50% hERG inhibition at 10 μM. The slightly less lipophilic analogue 20 showed somewhat improved hERG mitigation compared with 19 (65% vs 84% @ 10 μM). Eventually, α-methylation on the benzyl amine position satisfied the hERG issue with compound 21 showed 24.4% inhibition at 10 μM with slightly improved potency (IC₅₀ = 3.6 nM). We were delighted to see that α-methylated azabicyclo[3.2.1]octane (compound 23) not only addressed the hERG problem (21.9% @ 10 μM) but also improved the potency significantly (IC₅₀ = 1.4 nM). Interestingly, the other enantiomers 22 and 24 showed much reduced potency, together with abrogated hERG activity. All 4 compounds (21-24) demonstrated excellent selectivity over receptors S1P1–5 and LPA1–3, 5 with IC₅₀ more than 10 μM.

Compound 23 was further profiled and showed reasonable solubility (27 μg/mL at pH 7), good permeability (Papp (A-B/B-A) 6.29/14.47 × 10^–6 cm/s in CACO-2 cell), and high plasma protein binding (0.22% and 0.18% free fraction in rat and human, respectively) and negative in in vitro GSH trapping studies. However, it showed high in vitro clearance in rat liver microsomes (equivalent of 73% hepatic blood flow), though low in human hepatocytes (12%). The in vivo PK in rat confirmed its high systemic clearance (CL = 94 mL/min/kg), but it showed 0.8 total brain/plasma ratio at 4 h with 103 ng/mL brain concentration after oral dosing at 10 mg/kg. Encouraged by this result, we looked at the metabolic profile of compound 23. Metabolite identification (MetID) in rat and human liver microsomes and hepatocytes identified monooxygenation as a major pathway and predominantly on the substituted cyclohexane. Acyl-glucuronidation was identified as a minor pathway, and no oxidation on the naphthalene ring and bicyclic group was observed. This directed our focus to improving the metabolic stability of the cyclohexyl ring, which led to the identification of the CF₃ substitution²⁹ (compound 6), coincidental with the original hit, affording a moderate blocking effect with lowered clearance (CL = 29 mL/min/kg) in rat PK, while maintaining good ATX potency (IC₅₀ = 1.0 ± 0.5 nM, N = 5) and low human microsomal clearance (CL = 7.6 mL/min/kg). Compound 6 maintained low hERG inhibition (21.3% @ 10 μM) and excellent selectivity over receptors S1P1–5 and LPA1–3,5 (IC₅₀ > 10 μM).

Compound 6 was cocrystallized with mouse ATX, and the X-ray structure was solved at 2.3 Å. As shown in Figure 4A, the tail of compound 6 binds to the hydrophobic pocket and the bicyclic head portion binds to the hydrophobic channel with hydrogen bond between carboxylate and His251. Unlike LPA, compound 6 does not bind to the zinc ion,¹⁸ but shares a similar binding mode with GLPG1690 (Figure 4B).²²

Figure 4.

(A). X-ray cocrystal structure of compound 6 (PDB: 7MFH) with mATX (Cyan: 6, Yellow: LPA 14:0, overlay from PDB: 3NKN(18)). (B). Overlay of 6 (Yellow) with GLPG1690 (Magentas, PDB: 5MHP(22)).

Compound 6 exhibits good solubility (73.2 μg/mL at pH 7), good permeability (Papp (A-B/B-A) 16.8/40.3 × 10^–6 cm/s in CACO-2 cell). It has an ELogD of 4.73 at pH 7.4. It showed improved plasma protein binding (0.45% and 0.66% free fraction in rat and human, respectively) compared with its methyl tail analogue 23. Its IC₅₀s at CYP isoforms (3A4, 1A2, 2C19, 2C9, 2D6) were all greater than 10 μM, and its hERG IC₅₀ is greater than 10 μM. Compound 6 is negative in in vitro GSH trapping studies.

Furthermore, compound 6 shows good pharmacokinetics across species as depicted in Table 2. In rat, it has a half-life of 3 h with plasma AUC of 594 ng/mL*hr after IV dosing of 1 mg/kg. It shows good oral bioavailability (F = 66%) with total brain/plasma ratio = 0.2 at 4 h with 97 ng/mL brain concentration after oral dosing at 10 mg/kg. Compound 6 also demonstrated good pharmacokinetics in mice with a low clearance and good bioavailability (F = 51%) (Table 3). Its mice brain protein binding is slightly higher than plasma (0.599% free fraction in plasma and 0.096% free fraction in brain), similar to rat (0.45% in plasma vs 0.12% in brain). The total brain exposure is 8840 ng/mL after oral dosing at 10 mg/kg to mice, and the corresponding unbound C_max in brain is 1.5 nM, well above ATX IC₅₀ (1.0 nM). It was speculated that the brain penetrability of those zwitterionic compounds is likely transporter-mediated; however, no proof was obtained yet.

Table 2. Single-Dose Pharmacokinetic Parameters of 6 in Rats, Dogs, and Cynomolgus Monkeys^a.

species	route	dose (mg/kg)	Clp (mL/min/kg)	Vss (L/kg)	t1/2 (hr)	AUC0–24 h (ng/mL*hr)	Cmax (ng/mL)	Tmax (hr)	F (%)
rat	IVb	1	29 ± 4	3.0 ± 0.3	3 ± 0.4	594 ± 98
	POb,c	10				3969 ± 235	813 ± 144	1.6 ± 0.4	66
dog	IVb	0.2	1.7 ± 0.3	2.0 ± 0.2	14 ± 2	1390 ± 138
	POb	1				6591 ± 455	376 ± 19	2 ± 1	69
monkey	IVb	0.2	9.3 ± 1.3	2.4 ± 0.2	4 ± 0.1	369 ± 54
	POb	1				614 ± 110	89 ± 4	4 ± 1	34

^aMean ± SD (n = 3).

^bFormulation IV, and PO, solution in 15%HPCD.

^cBrain/plasma ratio at 4 h = 0.2.

Table 3. Single-Dose Plasma and Brain Pharmacokinetic Parameters of 6 in mice.^a.

route	dose (mg/kg)	tissue	Cl (mL/min/kg)	Vss (L/kg)	t1/2 (hr)	AUC0–24 h (ng/mL*hr)	Cmax (ng/mL)	Tmax (hr)	F (%)
IVb	2.32	plasma	6.09	2.36	4.72	6170	1270
		brain				3020	274	3.00
POb,c	10	plasma				12200	1070	3.00	51
		brain				8840	823	7.00

^aIn male C57BL/6 mice (n = 3).

^bFormulation IV, and PO, solution in 15%HPCD.

^cKp,uu = 0.11.

Compound 6 demonstrated significant LPA reduction (IC₅₀ = 53 ± 26 nM, N = 5) in a human plasma LPA assay (Figure S1). A similar LPA reduction (IC₅₀ = 47 ± 20 nM, N = 4) was observed in a rat plasma LPA assay. Supported by this data, we then tested compound 6 in rat CFA model, an acute pain model.³⁰ As shown in Figure 5A, compound 6 demonstrated dose-dependent efficacies with sign of effect as low as 0.3 mg/kg, and sustained effect for 24 h with 3 mg/kg and 10 mg/kg oral dose once. Dose dependent LPA reduction was also observed (Figure 5B). Remarkably, LPA level was reduced to 48% and 61% in plasma at 6 h after oral dosing at 10 mg/kg and 3 mg/kg, respectively. It should be noted that since T_max of compound 6 in rat was 2 h and maximal LPA reduction was observed at 2 h based on a 3 mg/kg rat PK/PD correlation study (Figure S2), the maximal efficacy observed at 6 h time point for 3 mg/kg dose with LPA level recovered back to 61% implies that a sustained LPA reduction might be required. The unbound LPA EC₅₀ (1.1 nM) was also consistent with ATX IC₅₀ (1.0 nM).

Figure 5.

(A). Efficacy in rat CFA model with single doses of 6 at 0.1, 0.3, 3, and 10 mg/kg. (B). LPA 20:4 reduction and drug concentration at 6 h after oral dosing of 6.

In summary, we have discovered a novel series of nonzinc binding ATX inhibitors starting from a phosphonic acid hit. The hERG liability was successfully mitigated by introduction of a methyl group to the benzylic position. A potent, selective, orally bioavailable, brain-penetrable tool compound BIO-32546 has been identified and has demonstrated in vivo efficacy in a model of acute pain with good PK/PD correlation. Further profiling of the compound in other neurological disease models will be disclosed in due course.

Glossary

ABBREVIATIONS

CL

clearance

GSH

glutathione

HPCD

hydroxypropyl-beta-cyclodextrin

IV

intravenous

PK

pharmacokinetics

PO

oral

S1P

sphingosine 1-phosphate

SAR

structure activity relationship

SFC

supercritical fluid chromatography

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.1c00211.

• Experimental details for synthetic procedures and analytical data for key compounds and assay conditions, crystallographic data for compound 6 (PDF)

Accession Codes

Structure data associated with this study have been deposited to the RCSB Protein Data Bank (http://www.rcsb.org) and can be accessed with the following codes: mATX bound to 6 is 7MFH. Authors will release the atomic coordinates and experimental data upon article publication.

Author Contributions

All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.