Fluorinated Isoindolinone-Based Glucosylceramide Synthase Inhibitors with Low Human Dose Projections

Abstract

Inhibition of glucosylceramide synthase (GCS) has been proposed as a therapeutic strategy for the treatment of Parkinson’s Disease (PD), particularly in patients where glycosphingolipid accumulation and lysosomal impairment are thought to be contributing to disease progression. Herein, we report the late-stage optimization of an orally bioavailable and CNS penetrant isoindolinone class of GCS inhibitors. Starting from advanced lead 1, we describe efforts to identify an improved compound with a lower human dose projection, minimal P-glycoprotein (P-gp) efflux, and acceptable pregnane X receptor (PXR) profile through fluorine substitution. Our strategy involved the use of predicted volume ligand efficiency to advance compounds with greater potential for low human doses down our screening funnel. We also applied minimized electrostatic potentials (V_min) calculations for hydrogen bond acceptor sites to rationalize P-gp SAR. Together, our strategies enabled the alignment of a lower human dose with reduced P-gp efflux, and favorable PXR selectivity for the discovery of compound 12.

Parkinson’s disease (PD) is a progressive neurodegenerative disease resulting from depletion of dopamine-producing neurons in the substantia nigra. The cardinal motor features of PD include resting tremor, rigidity, and gait disturbances, while cognitive impairment, depression and anxiety, sleep disturbances, and constipation are among the common nonmotor symptoms. Among the most known genetic risk factors are the loss-of-function mutations of the gene encoding the lysosomal enzyme glucocerebrosidase (GCase) responsible for metabolizing the glycolipids glucosylceramide (GlcCer) and glucosylsphingosine (GlcSph).^1,2 Build-up of these glycolipids disrupts lysosomal function and induces formation of the misfolded protein α-synuclein aggregates, a main neuropathological hallmark of PD.^3,4 A proposed therapeutic strategy for slowing disease progression in PD involves reducing GlcCer and GlcSph levels via treatment with a brain-penetrant glucosylceramide synthase (GCS) inhibitor to restore proper lysosomal function and clear neuronal inclusions that are positive for α-synuclein. Support for this approach comes from recent studies that have shown GCS inhibitors can attenuate formation of pathological α-synuclein aggregates in PD patient-derived induced pluripotent stem cell (iPSC) neurons^5,6 and in a GBA mutant mouse model.⁷ The announcement of Venglustat’s negative results from Sanofi’s Phase 2 MOVES-PD trial paused our efforts to develop a novel, orally bioavailable and brain penetrant GCSi for the treatment of PD.⁸

In this report, we describe our fine-tuning of an advanced series of GCS inhibitors containing an isoindolinone core structure.⁹ While the details of our progression of an attractive screening hit with good CNS properties to lead compound 1 (Figure 1) will be the subject of another disclosure, this work focuses on designs to further lower the projected human dose and reduce susceptibility to P-glycoprotein (P-gp) efflux to improve central nervous system (CNS) penetration.^10,11 Herein, we present results from the systematic introduction of electron-withdrawing fluorine substituents on the core as well as the side chain of 1. We also report our discovery of a modification that attenuated pregnane X receptor (PXR) activation in this series and the challenge of identifying molecules with both low CYP3A induction potential and low susceptibility to P-gp transport. The culmination of these efforts was the identification of 12, a leading compound with an improved human dose projection, lower P-gp susceptibility, and similar PXR selectivity compared to that of compound 1.

Figure 1.

Evolution of Screening Hit to Lead Compound 1.

Because there was good agreement between the QD human dose projections derived from allometric scaling (AS derived dose = 35 mg) and in vitro to in vivo extrapolation (IVIVE derived dose = 65 mg) for compound 1, our strategy to improve human dose projections in this series centered on the identification of fluorination patterns that provided improvements in potency and metabolic stability. Fluorination has been shown to be an effective way of improving binding affinity and addressing potential metabolic soft spots without significantly altering the molecular size and impact on physiochemical properties such as lipophilicity, solubility, and permeability. For compound 1, the methylene group of the isoindolinone core was identified as a potential site of metabolism from hepatocyte incubation studies. As such, we sought to leverage fluorination in the vicinity of this potential soft spot to slow the rate of oxidative metabolism.

Pursuing our goal to inhibit the GCS within the CNS, we also set out to minimize substrate recognition by P-gp, a major efflux transporter at the blood brain barrier (BBB) that can limit the extent of brain penetration. To assess susceptibility of new designs to P-gp transport, we elected to utilize our rat Mdr1a P-gp assay.¹² Additionally, for selected analogs, we determined the B → A:A → B efflux ratio (P-gp ER) at two concentrations to evaluate the concentration dependence on P-gp efflux. In our assay, a P-gp substrate is defined by having an ER > 2.5. Strategic fluorination can help to mitigate a compound’s P-gp recognition by modulating key recognition elements including hydrogen bonding capacity, LogD, polar surface area, and basicity.¹³ In our series, where the hydrogen bond donors/acceptors were kept constant, the individual H-bond acceptor and donor strengths became particularly important. In addition to P-gp susceptibility, another counterscreen that we monitored was activation of PXR. PXR activation leads to the transcription of transporter and metabolizing genes, such as MDR1 and CYP3A. CYP induction in turn can perpetrate drug–drug interactions (DDIs) that can complicate drug development.

To summarize, we sought to leverage fluorination at key positions to improve potency, attenuate metabolism, and mitigate P-gp liability without significantly altering physiochemical properties and off-target selectivity. Our screening funnel for compound advancement is described in Figure 2. Compounds were screened for advancement based on GCS enzyme potency and a new metric being explored in our laboratories, called volume ligand efficiency (VLE = −logIC₅₀ – logVd_u).¹⁴⁻¹⁶ VLE represents the relative balance between potency and the unbound volume of distribution (Vd_u). Vd_u is an estimate of the overall drug binding in the body and is a primary determinant of unbound concentrations that can be achieved at a given dose to the body. Since unbound concentration in the body drives pharmacological efficacy, the combination of potency and Vd_u can be used to estimate the dose needed to achieve acute target engagement (i.e., the dose required to achieve the desired concentration at C_max). We relied on predicted VLE (iVLE) using measured enzyme potency and the in silico QSAR model predicted Vd_u as a guide to select compounds for downstream pharmacokinetic (PK) studies. Focusing on compounds with iVLE > 6.0 helped us guard against advancing those whose gains in potency were offset by large increases in unbound volume of distribution (Vd_u), and thus, it would be unlikely to offer a lower human dose projection. Once the potential for a low dose has been established, further ADME optimization may be required to ensure appropriate PK/PD coverage for the desired dosing interval (e.g., optimization of effective half-life for a C_trough target following QD dosing).

Figure 2.

Screening funnel for compound advancement.

Initially, we explored the systematic introduction of a single fluorine atom on the isoindolinone core of lead compound 1, providing compounds 2-4 (Table 1). Fluorination at either the C6 or C5 position, 2 and 3 respectively, improved both enzyme and cellular potency by about 3-fold,¹⁷ whereas addition of a single fluorine at the C4-position, 4, decreased potency almost 2-fold. Encouragingly for 2 and 3, the gain in enzyme potency was accompanied by improvement in iVLE, suggesting the potential for improved human dose as predicted Vd_u (calculated from QSAR Vd and plasma protein binding (f_u) models) was largely unaffected. The potency enhancing effect of fluorination at the C5 and C6 position was additive for difluorinated analogue 5, providing a substantial improvement in potency and iVLE. As anticipated, difluorination anchored at C4, compounds 6 and 7, offered no potency advantage over 1. The addition of a fluorine to the core had only modest effects on P-gp susceptibility with all compounds being categorized as moderate P-gp substrates; a small increase in P-gp ER was determined for 5 (9.0) while 7 showed a slight decrease (P-gp ER = 2.8) relative to 1. Overall, fluorination modestly increased the PXR activity except for compound 3. Compounds from this set with an iVLE > 6.0 and PXR EC₅₀ > 6 μM were selected for rat PK studies.

Table 1. Structure Activity Relationships (SAR) for Fluorination of Compound 1.

^aCompds 8–14 assumed (S) configuration.

^biVLE = -Log(Enz IC₅₀) – Log(predicted Vd_u).

^cND = no data.

Concurrent with our exploration of core fluorination, we investigated several modifications to the alpha substituent on the isoindolinone side chain of 1. Among these we discovered that while replacement of the cyclopropyl group with a CF₃ substituent resulted in a moderate loss of potency, the substitution surprisingly mitigated P-gp susceptibility for compound 8 (P-gp ER = 0.9), rendering it a non- substrate. It is worth noting that the PXR activity for compound 8 remained unchanged relative to 1. In order to increase the potency, we employed the previous strategy of fluorinating the isoindolinone core of 8. Similar potency enhancing trends were observed upon mono- and difluorination, as seen for compounds 9-14. To our gratification, difluorination at C6 and C5 afforded 12 as a nonsubstrate for P-gp with similar potency to 1 and increased iVLE to 6.1. From this set, only compound 12 fit our iVLE selection criteria and was advanced to rat PK studies.

Another major focus of our SAR exploration was to identify modifications that attenuated PXR activity (Table 2). Toward this end, we utilized QSAR PXR predictions to guide the design of libraries directed at replacing of the phenyl oxadiazole biaryl motif. From this effort we identified the isoindolinone substituent in 15 as a replacement that not only retained GCS potency but ablated PXR activity. However, this improvement in PXR selectivity came at the expense of an increased P-gp susceptibility. Attempting to leverage our recently discovered P-gp SAR finding, substituting the cyclopropyl group for a CF₃ substituent mitigated P-gp susceptibility but reduced GCS potency for compound 16. In order to increase potency, we again employed the previous strategy of fluorinating the isoindolinone core at the C5 or C6 positions for compounds 17–19. Unfortunately, even though we were able to attenuate PXR activity and restore potency with compound 19, we were unable to adequately address the P-gp liability for this subseries.

Table 2. SAR of Fluorination on Compound 15.

^aCompds 15–19 assumed (S) configuration.

For preliminary predictions of QD human dose, we employed allometric scaling using IV/PO PK data from rat. These early predictions were based on several assumptions that included 1) human plasma protein binding (PPB) being equal to rat PPB, 2) the human unbound brain to unbound plasma ratio (Kp_uu) was equal to 1, 3) human F = 100%, and 4) the unbound plasma concentration target at trough (C_24h) was equal to the in vitro IC₅₀ value from the cell-based assay. Support for the PK trough target came from pharmacokinetic/pharmacodynamic experiments in mice that showed the desired pharmacodynamic effect of 70% reduction of GlcCer in the brain was achieved when unbound brain levels approximated the cellular IC₅₀.

Table 3 summarizes the rat PK properties and preliminary QD human dose projections based on rat single species allometry for the selected set of four analogs from Table 1. Not unexpectedly, that these initial human dose estimates show an inverse relationship with rat VLE; i.e., compound 5 has the highest VLE and lowest projected human dose while compound 7 has the lowest VLE and highest dose prediction for this set of analogs. Because these compounds all showed preliminary QD human dose projections lower than those of lead compound 1, they were further characterized in dog PK studies and in hepatocyte metabolic stability studies to determine the in vitro intrinsic clearance (CL_int). The additional dog PK and hepatocyte CL_int data were then utilized to refine human dose projections via two methods: 1) allometric scaling of the rat and dog PK data, and 2) in vitro to in vivo extrapolation of the intrinsic clearance data from human hepatocytes. Confidence in the second method was gained by observing good in vitro to in vivo correlations between projected clearance values derived from metabolic scaling of rat and dog hepatocyte intrinsic clearance data and the measured clearance values determined from in vivo rat and dog PK studies.

Table 3. PK and Human Dose Projections.

^aRat Wistar HAN; 24 h end point; IV: 2 mg/kg dose, DMSO/PEG400/H₂O – 20/60/20 formulation, 1 mL/kg dose volume; PO: 5 mg/kg dose, PEG400/Tween80/H₂O – 40/10/50 formulation; 2 mL/kg dose volume.

^bRat PO: 4 mg/kg dose.

^cAssumes human PPB = rat PPB, Kp_uu = 1.0, F = 100%, [trough]_u= cell IC₅₀.

^dDog Beagle; 24 h end point; IV: 0.5 mg/kg dose, DMSO/PEG400/H₂O – 20/60/20 formulation, 0.5 mL/kg dose volume; PO: 1 mg/kg dose, 30% Captisol formulation, 2 mL/kg dose volume.

^eDog PO: 10% Tween 80 formulation.

^fAssumes Kp_uu = 1.0, F = 50%, [trough]_u= cell IC₅₀

While all compounds showed an improved human dose projection relative to 1 based on rat and dog allometry, the IVIVE-derived doses for 3 and 5 had diverged significantly as a result of higher turnover in human hepatocyctes, which eliminated these compounds from further consideration. Not unexpectedly, the increase in P-gp ER for 5 appeared to translate to a lower rat Kp_uu versus 1 (0.2 vs 0.5), potentially necessitating an even larger upward adjustment to its projected human dose. Moreover, this compound showed potential to perpetrate drug–drug interactions (DDIs) via CYP3A induction at projected clinical exposures based on preliminary modeling of single donor hepatocyte induction data (not shown). Compounds 7 and 12 were identified as the most interesting compounds of the set. They demonstrated excellent oral bioavailability in rat and dog with long mean residence time (MRT) values. The PK parameters and human CL_int data supported low human dose projections based on our two methods. In addition, 7 and 12 exhibited moderate to low P-gp susceptibility, respectively, and sufficient PXR selectivity. However, in a rat tissue distribution study, the Kp_uu for compound 7 was determined to be 0.1, similar to compound 5, thus potentially requiring an upward correction for the human dose. On the other hand, the Kp_uu for compound 12 was determined to be 0.4, similar to compound 1. The Kp_uu results for 7 and 12 were lower than expected given their low in vitro P-gp efflux ratios and suggested that other transporters at the BBB may be playing a role in limiting brain penetrance for these compounds. We had previously determined that 7, 12 and other compounds in this series were not substrates of breast cancer resistant protein (BCRP), another efflux transporter located at the BBB. While fluorination appeared to have minimal effect on the dog PK parameters for these analogs, difluorination of the isoindolinone core at C5 and C6 improved rat MRT for compound 5 which was extended further with the CF₃ modification in 12. The CF₃ substituent also improved human hepatocyte stability for 12 by ∼3 fold, relative to the other analogs, which translated to a low human QD dose based on IVIVE methodology.

The isoindolinone core was readily accessed by using the synthetic routes illustrated for compounds 7 and 12 in Scheme 1. No unexpected or unusually high safety hazards were encountered. The appropriately difluorinated 2-methylbenzoic acids, 23 and 35, were iodinated to give 24 and 36, respectively. Esterification followed by bromination provided intermediates 26 and 38. Displacement of the alkyl bromide with the amines 22 and 34 and condensation furnished the iodo-isoindolinone cores 27 and 40, respectively. Suzuki coupling with commercially available 2-methyl-5-(4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)phenyl)-1,3,4-oxadiazole provided compounds 7 and 12. The remaining compounds were synthesized in a similar manner starting with the appropriately substituted benzoic acids and completing with the appropriate boronate. To establish the absolute configuration of these fluorinated chiral compounds, a single crystal X-ray structure was obtained of compound 3, as well as vibrational circular dichroism (VCD) spectroscopy of compound 12, both of which confirmed the (S) configuration.

Scheme 1. Synthetic Schemes for Compounds 7 and 12.

Reagents and conditions: (a) 3 M MeMgBr in Et₂O, THF, 0 °C; (b) TFA, DCM, rt; (c) I₂, PhI(OAc)₂, Pd(OAc)₂, DMF, 100 °C; (d) TMSCHN₂, 3:1 THF/MeOH, 0 °C; (e) NBS, BPO, DCE, 90 °C; (f) K₂CO₃, ACN, 90 °C; (g) 2-methyl-5-(4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)phenyl)-1,3,4-oxadiazole, PdCl₂(dppf), 1 M aq. K₃PO₄, 1,4-dioxane, 80 °C; (h) O-benzyl carbamate, DCM, rt; (i) (CF₃CO)₂O, Pyridine/Et₂O, rt; (j) NaBH₄, Et₂O, 0 °C; (k) Mg, MeI, Et₂O, rt; (l) Pd/C, EtOAc, rt; (m) oxalic acid, EtOAc, rt; (n) PhI(OAc)₂, Pd(OAc)₂, DMF, 100 °C; (o) TMSCHN₂, 3:2 THF/MeOH, 0 °C; (p) NBS, BPO, DCE, 90 °C; (q) K₂CO₃, ACN, 90 °C; (r) LiOH, 5:1 THF/H₂O, 40 °C; (s) 2-methyl-5-(4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)phenyl)-1,3,4-oxadiazole, XPhos Pd G3, Cs₂CO₃, 10:1 1,4-dioxane/H₂O, 90 °C; (t) SFC and crystallization.

The effect of the CF₃ substituent on decreasing the P-gp ERs in our in vitro assay was intriguing and warranted further investigation. We initially hypothesized that the CF₃ substituent might be altering the hydrogen bond acceptor (HBA) strengths of the isoindolinone carbonyl group and hydroxyl moiety in the side chain. To explore this possibility, we calculated minimized electrostatic potentials (V_min) associated with each acceptor site, which has been shown to be a predictor of HBA strength and has been used to explain trends in P-gp susceptibility.¹⁸ Specifically, we examined the influence of fluorination on the HBA strength of the carbonyl oxygen and hydroxyl oxygen by calculating the average V_min across the 20 lowest energy conformations for compounds 1, 5, and 8, excluding the biaryl motif for simplicity (Chart 1).¹⁹ The calculations showed that the median V_min for the carbonyl oxygen is lower than that for the hydroxyl oxygen, suggesting that the former is a stronger hydrogen bond acceptor. Overall, core and side chain fluorination were predicted to decrease HBA strength for the carbonyl and hydroxyl groups, respectively, relative to 1. Compound 8 showed the highest predicted V_min value for the hydroxyl group where the CF₃ substituent appears to render a greater change to V_min compared to core fluorination in 5. These V_min predictions taken together with the corresponding P-gp values suggest that the hydroxyl group may be the primary recognition element for P-gp. However, it is important to consider that there could be differences in hydrogen bond donor (HBD) strength and intramolecular H bonding that is impacting P-gp susceptibility for this set of compounds.

Chart 1. Box Plot of V_min Calculations for Estimating HBA Strength.

^a Molecules truncated.

^b Rat P-gp ER dose = 100 nM (1, 5), = 1.0 μM (8).

In summary, we demonstrated that fluorination was an effective strategy for the optimization of the projected human dose in this isoindolinone series of GCS inhibitors. Difluorination on the isoindolinone core at C4 and C5 provided compound 7 with good oral bioavailability in rat and dog and a projected human QD dose of 20 mg as determined by both AS and IVIVE but would most likely need a dose correction due to its low Kp_uu value. Pairing difluorination on the isoindolinone core at C5 and C6 with a CF₃ substituent on the isoindolinone side chain provided our lead compound 12 which showed excellent PK properties and stability in human hepatocytes, translating to a low projected human QD dose of 15 mg. Importantly, good CNS penetration and acceptable PXR selectivity were maintained for compound 12 (Table 4). Due to the stoppage of venglustat in the clinic, it was indicated that further study is needed on the mechanism of action before advancing any compounds further.

Table 4. Comparison of Properties of Compounds 1 and 12.

Glossary

Abbreviations

GCase

glucosylcerebrosidase

GlcCer

glucosylceramide

GlcSph

glucosylsphingosine

GCS

glucosylceramide synthase

GBA

glucocerebrosidase

V_min

value of molecular electrostatic strength at the minimum–predictive of H-bond basicity

PD

Parkinson’s disease

a-syn

alpha-synuclein

CNS

central nervous system

MPO

multiparameter optimization

MW

molecular weight

SAR

structure–activity relationship

P-gp ER

P-glycoprotein efflux ratio

BCRP

breast cancer resistance protein

QSAR

quantitative structure activity relationships

MRT

mean residence time

VLE

volume ligand efficiency

QD

“quaque die” or once a day

IC₅₀

conc of an inhibitor where response (or binding) is reduced by half

EC₅₀

conc of a drug that gives half-maximal response

DMSO

dimethyl sulfoxide

PEG400

polyethylene glycol 400

cLogP

calculated water/octanol partition coefficient

cLogD

calculated water/octanol distribution coefficient

PK

pharmacokinetics

PSA

polar surface area

pK_a

acid dissociation constant

Kp_uu

unbound partition coefficient

CYP

cytochrome P450

PXR

pregnane X receptor

HBD

hydrogen bond donor

PPB

plasma protein binding

VCD

Vibrational Circular Dichroism

M

molarity

MeMgBr

methylmagnesium bromide

Et₂O

diethyl ether

THF

tetrahydrofuran

TFA

trifluoroacetic acid

DCM

dichloromethane

I₂

iodine

PhI(OAc)₂

(diacetoxyiodo)benzene

DMF

dimethylformamide

TMSCHN₂

trimethylsilyldiazomethane

MeOH

methanol

NBS

N-bromosuccinimide

BPO

benzoyl peroxide

DCE

dichloroethane

K₂CO₃

potassium carbonate

ACN

acetonitrile

PdCl₂(dppf)

(1,1′-Bis(diphenylphosphino)ferrocene)palladium(II) dichloride

K₃PO₄

potassium phosphate

(CF₃CO)₂O

trifluoroacetic anhydride

NaBH₄

sodium borohydride

MeI

iodomethane

Pd/C

palladium on carbon

EtOAc

ethyl acetate

Pd(OAc)₂

palladium acetate

LiOH

lithium hydroxide

Cs₂CO₃

cesium carbonate

SFC

supercritical fluid chromatography

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.3c00436.

• SI 1: Chemistry experimental procedures and data, crystal data and structure refinement for compound 3, vibrational circular dichroism spectroscopy of compound 12, and high-resolution mass spectrometry (PDF)

• SI 2: Supporting ¹H NMR, ¹³C NMR, ¹⁹F NMR, and LCMS spectral data for compounds 1–12 and intermediates 13–43; Differential Scanning Calorimetry (DSC), Thermal Gravimetric Analysis (TGA), X-ray Powder Diffraction (XRD), and Polarized Light Microscopy (PLM) of compound 12 (PDF)

Author Present Address

^⊥ Vertex Pharmaceuticals, San Diego, CA 92121, USA

Author Present Address

^|| Bristol Myers Squibb Co., Princeton, NJ 08543, USA.

Author Present Address

^# Kymera Therapeutics, Watertown, MA 02472, USA.

The authors declare no competing financial interest.

Author Status

^△ R.B.F., X.L., and W.L. are retired.