Novel Tricyclic Pyroglutamide Derivatives as Potent RORγt Inverse Agonists Identified using a Virtual Screening Approach

Abstract

Employing a virtual screening approach, we identified the pyroglutamide moiety as a nonacid replacement for the cyclohexanecarboxylic acid group which, when coupled to our previously reported conformationally locked tricyclic core, provided potent and selective RORγt inverse agonists. Structure–activity relationship optimization of the pyroglutamide moiety led to the identification of compound 18 as a potent and selective RORγt inverse agonist, albeit with poor aqueous solubility. We took advantage of the tertiary carbinol group in 18 to synthesize a phosphate prodrug, which provided good solubility, excellent exposures in mouse PK studies, and significant efficacy in a mouse model of psoriasis.

The retinoic acid-related orphan receptor (ROR) family comprises three nuclear hormone receptors, RORα, RORβ, and RORγ. These receptors are widely expressed in many tissues and are involved in numerous regulatory functions.¹ In contrast, the expression of RORγt, a 498 amino acid splice variant of RORγ, is restricted to lymphoid cells. Acting as an important transcription factor in thymic tissue, RORγt mediates the differentiation of naive T cells into Th17 cells and regulates their maintenance and production of pro-inflammatory cytokines, including IL-17, IL-22, and GM-CSF.²⁻⁴ IL-17 has been implicated in the pathology of multiple autoimmune diseases, including rheumatoid arthritis, psoriasis, inflammatory bowel disease, and multiple sclerosis. The utility of down-regulation of IL-17 as a treatment for autoimmunity has been validated in animal models as well as clinically by anti-IL-17 antibodies (such as secukinumab and ixekizumab) which have been approved for clinical use.⁵ Thus, considerable effort has been expended to identify small molecules, which would down-regulate IL-17 through modulation of RORγt.⁶⁻⁸

Previous reports from our laboratories have detailed the discovery of several series of small-molecule inverse agonists with potential utility in the treatment of autoimmune diseases.⁹⁻¹¹ In particular, compounds bearing carboxylic acid groups have proven to be very potent and selective RORγt inverse agonists, culminating in identification of the clinical candidate BMS-986251 (1).¹¹ As part of our continued pursuit of RORγt inverse agonists, we wanted to identify compounds lacking the carboxylate moiety due to the potential liability of acyl glucuronide formation with carboxylic acids.^12,13 A previously reported nonacidic compound 2 showed good potency and selectivity, along with useful pharmacokinetics and pharmacodynamics.¹⁰ Here, we report our continued efforts to find additional nonacid-containing analogues as inverse agonists of RORγt with in vitro profiles suitable for evaluation in animal models of autoimmune diseases (Figure 1).

Figure 1.

Previously reported RORγt inverse agonists

Virtual screening, wherein a library of hypothetical molecules is evaluated in silico for interaction with a biological target of interest, can be an efficient way to narrow the focus of a discovery effort and guide the actual preparation of compounds more likely to show interesting biological activity. Because of the huge number of possible nonacidic substituents which could be envisioned to replace the cyclohexylcarboxylate moiety of 1, we undertook a virtual screening effort by creating a virtual library of compounds with nonacidic moieties linked to our previously reported and optimized (S)-3-((4-fluorophenyl)sulfonyl)-3-(4-(perfluoropropan-2-yl)phenyl)pyrrolidine moiety.⁹ Potential lead structures were docked into the ligand binding domain (LBD) of RORγt from an in-house X-ray crystal structure with a related compound in the ligand binding pocket using Glide SP.¹⁴ Constraints and post minimization were not required, and binding poses were ranked by the Glide docking score. Of the ∼60 000 virtual analogues screened, the top-scoring 400 were evaluated visually. Many were discarded for lack of additional binding interactions with the LBD, or for lack of synthetic versatility to enable SAR studies. One of the most attractive candidates was the (S)-pyroglutamide analogue 3 (Figure 2). In the binding model, the left-hand portion of 3 was predicted to bind in the hydrophobic pocket, making good contacts with the side chains of Trp317 on helix 3, His479 on helix 11, Ile400 and Phe401 on helix 7, and Met365 and Arg364 on helix 5, as previously reported (Figure 3).⁹ In addition, (a) the benzamide moiety was predicted to project into a hydrophobic pocket, (b) the benzamide carbonyl was predicted to form a hydrogen bond to the backbone NH of Glu379, and (c) the side chain of Arg367 shifted to form a hydrogen bond to the pyroglutamate carbonyl. In the event, preparation of this compound confirmed inverse agonist potency for the (S)-pyroglutamide in the RORγt GAL4-Luciferase reporter assay (GAL4, Figure 2). In contrast, the corresponding (R)-pyroglutamide analogue was over 30-fold less potent (EC₅₀ 1.5 μM, compared to 46 nM for 3). The ease of synthesis of pyroglutamic acid derivatives and the potential for additional pyroglutamide ring substitution were also appealing features of this lead, which was selected for further study.

Figure 2.

From virtual lead to tricyclic pyroglutamide series 4.

Figure 3.

Model of compound 3 in the binding pocket of RORγt.

We opted to explore the pyroglutamide structure–activity relationship (SAR) in our tricyclic series (generic structure 4, Figure 2), since conformational restraint significantly improved the RORγt potency in both GAL4 and a human whole blood assay (hWB) by about 50-fold over that of the unconstrained series.¹⁰ We initially explored variation of the pyrrolidinone N-substituent, as shown in Table 1. Since the binding model for 3 suggested that lipophilic substituents would be beneficial (vide supra), the 4-fluorobenzyl compound 5 was prepared and found to be very potent in GAL4 (EC₅₀ 7 nM) and moderately potent in hWB with an EC₅₀ of 57 nM. However, it had poor metabolic stability, as indicated by short half-lives following incubation with human, mouse, and rat liver microsomes (LM). A second benzyl analogue 6 with a nitrile replacing the fluorine substituent to temper the lipophilicity (clogP 6.06, vs 6.35 for 5) improved the metabolic half-life. Since the clogP for this compound was still quite high, and the benzylic methylene linker was felt to be a metabolic liability, smaller and less lipophilic substituents were explored. The propiononitrile derivative 7 (clogP 4.25) was still potent in the GAL4 assay with slightly improved hWB potency. Importantly, the human metabolic half-life was significantly improved, although there was very little improvement for this parameter in rodents. The acidic compound 8, however, lost GAL4 potency dramatically. This was not surprising since this part of the LBD of RORγt is hydrophobic, based on the aforementioned model (Figure 3). Smaller substituents (isopropyl 9 and methyl 10) on the nitrogen atom of pyroglutamide revealed that larger groups were not required for good GAL4 potency, while moderate hWB potency and some degree of metabolic stability (particularly in human microsomes) could be maintained. Overall, the SAR of the N-substituent suggested that a wide variety of groups were tolerated, leading to compounds with moderate to good potency in the GAL4 and hWB assays, although metabolic stability remained a significant issue, particularly in mouse which was the species for preclinical efficacy studies.

Table 1. Pyrrolidinone N-Substitution.

compd	R	RORγt GAL4 EC50 (nM)a	IL-17 hWB EC50 (nM)a,b	LM t1/2 (min) (h, m, r)b,c	clogPd
5	CH2-4-F-Ph	7.0 ± 1.4	57 ± 4.1	24, 29, 35	6.35
6	CH2-4-CN-Ph	30 ± 29	nt	58, 74, 61	6.06
7	CH2CH2CN	18 ± 11	30 ± 15	>120, 79, 45	4.25
8	CH2COOH	1300 ± 1500	nt	nt	3.93
9	iPr	13 ± 1.1	61 ± 28	50, 52, 45	5.25
10	Me	8.5 ± 6.5	81 ± 5.4	>120, 75, 55	4.47

^aMean EC₅₀ values ± standard deviation for two or more experiments, except as noted.

^bnt = not tested.

^cHalf-life in human, mouse, and rat liver microsomes

^dlog P calculated using the ChemAxon calculator¹⁵

Additional substitution on the pyrrolidinone ring was explored in order to examine effects on potency and metabolic stability (Table 2). The (4S)-hydroxy substituted compound 11 lost potency by 3-fold in both GAL4 and hWB without improving metabolic stability. The corresponding methoxy derivative 12 regained whole blood potency but at the expense of reduction in metabolic stability, probably because of oxidative demethylation. Inversion of the hydroxyl group of 11 with a fluoro moiety provided the (4R) derivative 13, with potency in hWB improved over the unsubstituted analogue 10. Gratifyingly, the metabolic stability was restored to the level seen for 10. On the basis of the well-precedented improvement in stability toward oxidative dealkylation seen with deuterium substitution,¹⁶ the trideuteromethyl analogue 14 was prepared and showed a dramatic improvement in stability in both rodent species. Curiously, the epimeric fluoride 15, while retaining potency, showed a drop-off in stability, which was even more pronounced in human in the case of difluoro analogue 16. The latter may be due to the increase in lipophilicity (clogP 5.32, vs 4.41 for the corresponding monofluoro analogues.)

Table 2. Pyrrolidinone Ring Substitution.

^aMean EC₅₀ values ± standard deviation for two or more experiments, except as noted.

^bHalf-life in human, mouse, and rat liver microsomes

^clog P calculated using the ChemAxon calculator¹⁵

While 14 was a promising compound with desirable potency and excellent microsomal stability, very poor aqueous solubility (<1 μg/mL) imposed a significant hurdle for in vivo evaluation. Because of the beneficial effect of the (4R)-fluoro substituent on both hWB potency and rodent metabolic stability, we briefly re-examined variation of the pyrrolidinone N-substituent in the context of this improvement. A pendant hydroxyl group was considered as a possible solution to the solubility issue, since this would provide a handle for potential solubilizing prodrugs¹⁷ should that become desirable. The easily prepared 2-hydroxyethyl derivative 17 was intriguing in that no significant change in potency relative to 14 was seen, although metabolic stability in rodents was adversely effected. However, replacement of the primary carbinol of 17 with the dimethyl tertiary carbinol provided 18, with both potency and metabolic stability virtually identical to those achieved with 14. An X-ray cocrystal structure of 18 bound in the LBD of RORγt was obtained (Figure 4, PDB ID 7JYM), and it was gratifying to see the binding pose of this compound to be as predicted for compound 3 (Figure 3). In addition, the pyroglutamide fluorine substituent makes favorable contacts with Cα and the side chain methylenes of Arg364. The gem-dimethyl moiety makes beneficial lipophilic contacts with the side chains of Gln286, Leu287, and Phe377, directing the hydroxyl group toward solvent. Although no appropriate water molecules were observed in the crystal structure, the distances from the pyrrolidinone oxygen to the side chain of Arg367 (4.3 Å) and to the backbone carbonyl oxygen of Arg364 (3.7 Å) suggest the possibility of water-mediated hydrogen bonding between the lactam and the protein.

Figure 4.

Co-crystal structure of compound 18 with the LBD of RORγt (PDB ID 7JYM).

Because of its promising profile, 18 was further profiled in vitro, and these results are summarized in Table 3. Compound 18 was very selective against other nuclear hormone receptors (RORα, RORβ, PXR, LXRα, and LXRβ) and was devoid of recombinant cytochrome P450 enzyme inhibition (rCYP), except for weak inhibition of the 2C9 isozyme. In addition, 18 showed no effect on HepG2 liver cells (EC₅₀ > 100 μM), suggesting a reduced liability for liver toxicity relative to another nonacidic compound (2, HepG2 EC₅₀ 42 μM). Compound 18 had reasonable permeability in CACO-2 cells, with no significant efflux. Pharmacokinetic studies of 18 in BALB/c mice (Table 4) showed good oral exposure, with excellent bioavailability (94%), low clearance (CL), long T_1/2, and a good peak-to-trough ratio (3.5). However, 18 had very poor aqueous solubility (Table 3), which would severely limit dosing for in vivo efficacy and tolerability studies.

Table 3. In Vitro Profile of Compound 18.

RORγt GAL4 EC50 (nM)	16 ± 9
RORα GAL4 EC50 (nM)	>10 000
RORβ GAL4 EC50 (nM)	8400 ± 2100
IL-17 hWB EC50 (nM)	43 ± 24
mouse Th17 EC50 (nM)	25 ± 20
PXR EC50 (nM)	>50 000
LXRα EC50 (nM)	>7500
LXRβ EC50 (nM)	>7500
rCYP IC50 (nM)
1A2	>20 000
2C8	>20 000
2C9	4900
2C19	>20 000
2D6	>20 000
3A4	>20 000
HepG2 EC50 (nM)	>100 000
LM t1/2 (min) h, m, r	>120, >120, >120
protein binding, % free (h, m)	2.5, 1.3
CACO-2 A-B, B-A (nm/s)	93, 190
CACO-2 efflux ratio	2.0
solubility (mg/mL):
@pH 1 (0.1 N HCl solvent)	<0.001
@pH 4 (50 nM acetate buffer)	<0.001
@pH 6.5 (50 nM acetate buffer)	<0.001

Table 4. PK Profile of Compound 18 in BALB/c Mice^a.

route	IVb	POc
dose (mg/kg)	2	4
Cmax (μM)		2.2 ± 0.6
C24h (μM)		0.62 ± 0.21
Tmax (h)		4.3 ± 2.3
AUC72h (μM-h)	23.0	43.9
AUCtot (μM-h)	23.7	44.4
T1/2 (h)	15
CL (mL/min/kg)	2.0
Vss (L/kg)	2.7
F (%)		94

^a12 time points, 3 mice/time point/route.

^bSolution vehicle: 2.5% NMP; 67.5% PEG 400; 4.5% Pluronic F-68; 25.5% water.

^cSolution vehicle: 5% NMP; 76% PEG 400; 19% TPGS.

As a preliminary approach to improving solubility, the phosphate ester 19 (Figure 5) was prepared as a potential prodrug. Prodrug 19, as expected, provided a dramatic increase in solubility at pH 6.5 (≥4.7 mg/mL). This enabled in vivo studies in BALB/c mice with 19 using higher doses than were possible with 18 itself, while also avoiding the use of N-methylpyrrolidine (NMP) and emulsifier in the dosing solution (vehicle: 20% of 0.1 M acetate buffer, 80% PEG 400). At 5 mg/kg (parent equivalent), 19 showed roughly the same trough concentration of 18 (0.76 ± 0.13 μM) as when 18 was dosed directly (1.08 ± 0.15 μM), with 67% exposure of the parent compound. At 25 mg/kg (parent equivalent), a dose which could not be achieved with the poorly soluble parent, the prodrug provided a near-dose proportional increase in exposure (C_24h 4.26 ± 0.63 μM).

Figure 5.

Phosphate prodrug of 18.

These promising results prompted us to evaluate compound 19 in the mouse acanthosis model, a preclinical model of psoriasis.^18,19 The prodrug was dosed at 6, 17, and 50 mg/kg, equivalent to 5, 15, and 45 mg/kg of parent, calculated on the basis of the PK results for 19. Acanthosis was induced with recombinant humanized IL-23 injected every other day into the right ear of C57bl/6 female mice, until the last injection on day 9. A starting baseline measurement of ear thickness (before the first injection) was made on day 0, and ear thickness was then measured every other day, prior to the next ear injection. Compound 19 was dosed orally approximately 2 h before the first IL-23 injection and continued once daily until day 9. The placebo/control group was dosed with blank vehicle, and a recombinant dual chain human anti-IL-23 adnectin was administered SC as a positive control at doses of 3 mg/kg on days 0, 3, and 7 (which gave maximal efficacy based on other studies). No significant weight loss was seen at any of the doses, suggesting that the compound was well tolerated. At the two higher doses, trough blood levels of the parent 18 (Table 5) exceeded the protein binding adjusted IC₉₀ for this compound in a mouse Th17 reporter assay by 3- and 6-fold, respectively. (No intact phosphate prodrug was observed in blood samples.) Gratifyingly, as shown in Figure 6, the 15 and 45 mg (parent equivalent) doses of prodrug 19 resulted in significantly reduced ear swelling (38% and 67%, respectively, on day 9), with the highest dose comparable in efficacy to the IL-23 antibody.

Table 5. Exposure of Compound 18 following Dosing Prodrug 19 in Mouse Acanthosis.

dose (mg/kg QD)a	5	15	45
day 1 AUCtot (μM-h)	13.9	30.9	274.3
day 9 AUC24 (μM-h)	9.0	45.2	162.9
day 9 Cmax (μM)	0.65 ± 0.19	2.58 ± 0.36	8.34 ± 2.26
day 1 C24h (μM)	0.21 ± 0.08	0.48 ± 0.14	0.91 ± 0.33
day 9 C24h (μM)	0.15 ± 0.08	1.36 ± 0.56	5.82 ± 1.10
day 9 C24h fold mouse Th17 IC90b	0.4×	3.2×	13.9×

^aProdrug was dosed to provide the indicated amount of parent after phosphate hydrolysis; solution vehicle: 20% of 0.1 M acetate buffer; 80% PEG 400.

^bProtein binding adjusted mouse Th17 IC₉₀ = 0.42 μM

Figure 6.

Oral efficacy of prodrug 19 in mouse acanthosis. Doses are of parent equivalent.

Three general routes (Scheme 1) were used to make the N-substituted pyroglutamic acid intermediates for coupling to the tricyclic core (the preparation of which has been previously described).¹⁰ Alkylation of commercially available methyl or ethyl (S)-5-oxopyrrolidine-2-carboxylate 20 by an alkyl halide followed by saponification of ester 21 provided substituted (S)-pyroglutamic acid 22. The N-substituted pyroglutamic acid derivatives could also be obtained by the cyclization of N-alkyl glutamate esters or N-alkylglutamic acid,²⁰⁻²² which were obtained by reductive alkylation of diethyl l-glutamate hydrochloride or l-glutamic acid with an aldehyde or ketone, or by another N-alkylation method. For example, conjugate addition of glutamic acid 23 to acrylonitrile provided 24, which was cyclized to provide 25. Pyroglutamic acids bearing an additional ring substituent were prepared from commercially available Boc-protected hydroxyproline ester 26.²³⁻²⁵ Following protection of the hydroxy group as the TBDMS ether, ruthenium(IV) oxide oxidation to lactam 27 and deprotection provided 28. Monomethylation using cesium carbonate as base, or dimethylation with sodium hydride provided, after saponification, 29 or 30, respectively. The ring fluorine was introduced by treatment of 26 with DAST, followed by oxidation, deprotection, and alkylation as above to provide 34. Coupling of the acids to the tricyclic amine core to give final products was generally accomplished using HATU or BOP.

Scheme 1. General Synthesis of N-Substituted Pyroglutamic Acid Intermediates.

Reagents and conditions: (a) NaH, THF or DMF, rt; (b) LiOH, THF:MeOH:H₂O (3:1:1), rt, 32-69% for 2 steps; (c) NaOH, H₂O, 50 °C; (d) concentrated HCl, acetone, reflux, 60% for 2 steps; (e) TBDMS-Cl, Et₃N, imidazole, THF-DMF, rt, quantitative yield; (f) RuO₂·H₂O, NaIO₄, EtOAc-H₂O, rt, 89-98%; (g) TFA, DCM, 0 °C - rt, quantitative yield; (h) Cs₂CO₃, CH₃I, 50 °C, 37%; (i) NaH, CH₃I, DMF, 50 °C, 26%; (j) DAST, DCM, 0 °C to rt, 34%; (k) 4M HCl in dioxane, DCM, rt, 63%.

The synthesis of compound 18 and its prodrug 19 is outlined in Scheme 2. Following coupling of the intermediate unsubstituted pyroglutamic acid intermediate 33 (Scheme 1) with amine 35,⁶ the resulting intermediate 36 was alkylated with isobutylene oxide to give 18 as the major product. Compound 18 was converted to dibenzyl phosphite 37, which was then oxidized to dibenzyl phosphate 38 in a one-pot two-step reaction. Reductive removal of the dibenzyl groups afforded the phosphate prodrug 19.

Scheme 2. Synthesis of Final Compounds 18 and 19.

Reagents and conditions: (m) HATU, DIEA, DMF, rt, 35%; (n) K₂CO₃, iPrOH, 100 °C, 51%; (o) 5-methyl-1H-tetrazole, dibenzyl diisopropylphosphoramidite, DCM, 0 °C – rt; (p) 35% H₂O₂ in water, DCM, 0 °C, 89% for two steps o & p; (q) H₂, Pd–C, MeOH/EtOAc, rt, 78%.

In summary, from a virtual screening exercise, pyroglutamide analogues emerged as interesting leads for identifying novel nonacid containing RORγt inverse agonists. Optimization of the pyroglutamide moiety in the context of our conformationally locked tricyclic series led to the identification of compound 18 as a potent and selective compound with an attractive in vitro profile. In the mouse acanthosis model of psoriasis, 18 (dosed as the more soluble phosphate prodrug 19) provided good dose-dependent efficacy. Although a significant improvement was achieved in the in vitro HepG2 liver cell toxicity assay over a previous nonacidic lead, suggesting reduced potential for liver injury, further evaluation of this compound was precluded by the identification of a potential hERG inhibition issue (hERG patch clamp IC₅₀ 1.9 μM). However, the current work demonstrated that useful pharmacokinetics and efficacy could be achieved with nonacidic analogues of our clinical compound 1 and that a good solubility could be obtained in this lipophilic series without resort to incorporation of a carboxylic acid moiety. Further exploration based on these results will be reported elsewhere.

Glossary

ABBREVIATIONS

Arg

arginine

AUC

area under the curve

Boc

tert-butyloxycarbonyl

BOP

benzotriazol-1-yloxytris(dimethylamino)phosphonium hexafluorophosphate

C₂₄

concentration at 24 h

C_max

maximum concentration

CL

clearance

clogP

calculated log partition coefficient

DAST

dimethylaminosulfur trifluoride

DCM

dichloromethane

DIEA

diisopropylethylamine

DMF

N,N-dimethylformamide

EC₅₀

50% efficacious concentration

EtOAc

ethyl acetate

F

bioavailability

hWB

human whole blood assay

GAL4

GAL4-luciferase reporter assay

Gln

glutamine

Glu

glutamic acid

GM-CSF

granulocyte-macrophage colony-stimulating factor

HATU

hexafluorophosphate azabenzotriazole tetramethyl uronium

His

histidine

IL

interleukin

Ile

isoleucine

IV

intravenous

LBD

ligand binding domain

Leu

leucine

LM

liver microsomes

LXR

liver X receptor

Met

methionine

NMP

N-methylpyrrolidinone

PEG

polyethylene glycol

Phe

phenylalanine

PO

per os (oral)

PXR

pregnane X receptor

QD

per day

rCYP

recombinant cytochrome P450

ROR

receptor-related orphan receptor

Rt

room temperature

SAR

structure–activity relationship

TBDMS

tert-butyldimethylsilyl

TFA

trifluoroacetic acid

TH17

T helper 17 cell

THF

tetrahydrofuran

TPGS

tocopheryl polyethylene glycol succinate

Trp

tryptophan

Vss

volume of distribution.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.0c00496.

• Description of the virtual screening protocols and complete experimental details and characterization for all new compounds (PDF)

The authors declare no competing financial interest.