Rationally Designed, Short-Acting RPE65 Inhibitors for Visual Cycle-Associated Retinopathies

Abstract

The visual cycle is a metabolic pathway essential for visual function. The bisretinoid byproducts of this pathway can induce retinal toxicity, as occurs in Stargardt disease type 1 (STGD1). Emixustat, which inhibits bisretinoid production, is a visual cycle modulator (VCM) that targets RPE65. However, it causes visual impairment due to its unfavorable duration of action. Here, we report ester-containing analogs of emixustat that are susceptible to hydrolytic clearance and function as short-acting VCMs. We show that the esterase-mediated metabolism of these compounds can be tuned while maintaining high-affinity RPE65 targeting. Compounds 6 (EYE-002) and 7 (EYE-003) containing diethyl acetate and valproate esters, respectively, allowed faster recovery of visual cycle function compared to emixustat. These molecules protected against retinal degeneration in mouse models of photic retinopathy and STGD1. These data demonstrate that shorter attenuation of the visual cycle can therapeutically intervene in retinal diseases with fewer visual side effects compared to emixustat.

Introduction

The function of photoreceptor cells hinges on two pathways: phototransduction, which initiates visual perception beginning with the photoisomerization of retinaldehyde (RAL)-containing visual pigments, and the visual cycle, which ensures timely clearance and trans–cis recycling of RAL released from visual pigments after their photoactivation. While the molecular effectors of the first pathway are confined to photoreceptor cells, those of the latter are distributed both in the photoreceptor and the adjacent retinal pigment epithelium (RPE). Notably, when RAL production overwhelms the available clearance mechanisms or when trans to cis recycling is inadequate, photoreceptors can be degenerate leading to permanent vision loss. −

Stargardt disease 1 (STGD1), the most common inherited juvenile macular dystrophy, features loss-of-function mutations in the ABCA4 gene, which encodes a transporter critically involved in RAL clearance within photoreceptors. , In the absence of ABCA4 function, persistently high RAL levels accelerate the spontaneous formation of RAL condensation products termed bisretinoids. These molecules contribute to the formation of lipofuscin, a fluorescent material that accumulates in the retina and is toxic to the photoreceptors and RPE. Despite the complex genotype-phenotype relationships found in STGD1 patients, including variable lipofuscin accumulation, experimental evidence indicates that STGD1 progression is primarily driven by visual cycle activity. , Presently, there is no approved treatment for STGD1.

Emixustat is an investigational small-molecule inhibitor of the visual cycle isomerohydrolase, RPE65, − is currently being developed for STGD1 and other retinopathies. RPE65 is an enzyme found exclusively in the RPE where it catalyzes the synthesis of 11-cis-retinol, the precursor to 11-cis-retinal (11cRAL) bound to ground state visual pigments. , By competitively inhibiting RPE65, emixustat reduces the turnover of visual pigment regeneration, diminishing the exposure of the retina to free RAL. Additionally, emixustat has been shown to sequester retinaldehyde released from visual pigments by forming a Schiff base with the carbonyl moiety of retinaldehyde. These pharmacological actions manifest as a slower rate of recovery of the signal from scotopic electroretinography (ERG) recordings after a photobleach. − The rate of scotopic ERG recovery reflects the concentration of rhodopsin and therefore visual cycle activity. ,

Emixustat has advantages over other small molecules being investigated for STGD1 treatment including a well-defined and eye-specific molecular target, an apparent lack of systemic side effects, and a high pharmacodynamic ceiling. However, emixustat treatment is associated with visual adverse events (i.e., delayed dark adaptation and night blindness), which prompted dose reductions and the withdrawal of a significant number of patients from clinical trials.

It has been observed in human studies that emixustat suppresses the scotopic ERG for several days after treatment cessation, even though this molecule has a plasma elimination half-life of 4–6 h. One possible explanation for this observation is due to the accumulation of emixustat in ocular tissues based on its interaction with melanin. , Another possibility is that amidation by the lecithin:retinol acyltransferase in RPE cells results in amides that may be slowly hydrolyzed to produce sustained RPE65 inhibition. Regardless of the mechanism, the pharmacological properties of emixustat treatment do not allow a drug-free period in the retina during which visual cycle activity can bring the rate of dark adaptation back to the physiological range. The unusual pharmacology of emixustat complicates dosing and hinders its translation into a long-term treatment for retinopathies associated with the activity of the visual cycle, such as STGD1. Therefore, modifications to the emixustat molecule that shorten the duration of RPE65 suppression may lead to a more clinically viable VCM. Previously, we ,,− and others conducted extensive structure–activity relationship (SAR) studies on the emixustat scaffold, which has provided insights into sites of the molecule that can be targeted for deliberate metabolic lability.

Herein, we present data on rationally designed short-acting emixustat derivatives obtained by an ether-to-ester substitution that renders these new saVCMs susceptible to inactivation via esterase cleavage. We demonstrate that single doses of the lead compounds 6 (EYE-002) and 7 (EYE-003), prevented retinal damage induced by visual cycle activity in wild-type and Abca4 ^–/– Rdh8 ^–/– mice, respectively. Our data indicate that ester-based VCMs retain desirable properties of emixustat in selectively targeting RPE65 but feature a strategically limited duration of action. The limited duration of action of 6 and 7 allows a daily period during which the visual cycle can operate at a normal rate thereby minimizing disturbances of visual perception.

Results

Medicinal Chemistry

The use of ester derivatization to limit the duration of action of lead molecules is an example of a well-known strategy known as “soft drug design”. − Importantly, there are several esterases in the RPE, our target tissue of interest, that could act to limit ester-based emixustat derivative activity , (Figure A). The challenge in the synthesis of ester-containing VCMs is that, based on the emixustat pharmacophore defined in previous publications, they must contain a primary or secondary γ-amino-α-aryl alcohol. While necessary for potent RPE65 inhibition, , the amine group was incompatible with most ester synthetic methods due to the autoreactivity with the ester functional group. To overcome this conflict, we utilized an enamine-based Heck coupling, which used a commercially available benzyl chloroformate (Cbz)-protected 3-aminopropionaldehyde and 3-iodophenol to synthesize the Cbz-protected phenol (compound 4) in one step (Scheme ). , Conversion of the phenol to an ester was attained via direct reaction with the acyl chloride or activation of the corresponding carboxylic acid with 1-ethyl-3-(3-(dimethylamino)­propyl)­carbodiimide (EDC). The collection of carboxylic acids used for the synthesis were chosen to either mimic emixustat through a methyl cyclohexyl ester (5, also referred to as EYE-001) or another potent VCM, MB-004, in the form of a diethyl acetate ester (6, also referred to as EYE-002) or a valproate ester (7, also referred to as EYE-003), Figure B. The choices of the ester functional groups were based on defined SAR with the corresponding ether derivatives, but also on the lack of toxicity displayed by the carboxylic acid cleavage products at doses ∼30–50 times higher (∼300–500 mg/kg) than the dose we planned to use in this study (10 mg/kg). −

1.

Ester derivatives of emixustat inhibit RPE65 and are metabolized by esterases. (A) Strategy utilized to achieve short-acting RPE65 inhibition. (B) Chemical structure of test compounds. (C) Dose–response curves for RPE65 inhibition. Data are plotted as mean ± SD and represent n = 3 replicates per point. IC₅₀ values are given in Table . (D) Time courses of loss of test compound in the presence of porcine liver esterase; data are plotted as mean (solid lines) and 95% confidence interval (dotted lines), n = 3 replicates per point. The data obtained from the hydrolytic cleavage of test compounds were fit to either a linear or exponential decay function using GraphPad Prism. The values of the initial rates of VCM loss are given in Table .

1. Synthetic Scheme to Obtain Ester-Based Emixustat Analogs .

^a Abbreviations: 1,3-bis­(diphenylphosphino)­propane (DPPP); bis­(dibenzylideneacetone)­palladium(0) (Pd­(DBA)₂); N,N-diisopropylethylamine (DIPEA); molecular sieve (MS); 1-ethyl-3-(3-dimethylaminopropyl)­carbodiimide (EDC); acid chloride (RCOCl); palladium on carbon (Pd/C); dimethylformamide (DMF); methanol (MeOH); tetrahydrofuran (THF).

In Vitro Characterization of Short-Acting VCMs

We tested the ability of 5–7 to inhibit RPE65 in vitro using bovine RPE microsomes as an enzyme source (Figure C). All three compounds displayed dose-dependent RPE65 inhibition with IC₅₀ values (∼70–164 nM) comparable to that of emixustat (∼98 nM). Importantly, the amine-containing product of ester hydrolysis, 4, did not significantly inhibit RPE65. Next, we measured the susceptibility of the ester derivatives to porcine liver esterase-catalyzed hydrolysis (Figure D). Equal concentrations of 5–7 and emixustat (negative control) were incubated with the esterase, and the time course of the reaction was followed for 60 min by HPLC. Compounds 5–7 showed time-dependent enzymatic hydrolysis, with the hydrolytic rate being correlated with the degree of steric hindrance exerted by the ester group substituents (Table ). Compound 5, with its conformationally restricted cyclohexyl group, was hydrolyzed most rapidly. Compounds 6 and 7 exhibited greater stability with ∼ 4-fold and ∼ 13-fold slower initial rates of hydrolysis. This greater stability can be attributed in part to their more conformationally flexible diethyl acetate and valproate esters, which can adopt conformations that sterically protect the carbonyl group from nucleophilic attack during the esterase-catalyzed reaction. Nonenzymatic aminolysis also showed that 5 is substantially less stable compared to 6 and 7 (Figure S1).

1. Values of Half-Maximal Inhibitory Concentrations (IC₅₀) and Rates of Esterase-Mediated Hydrolysis Measured In Vitro .

test compound	IC50 toward RPE65 (μM)	initial rate of VCM loss (% compound/min)
emixustat	0.098 ± 0.014	0.069
4	>100	not tested
5	0.164 ± 0.027	5.611
6	0.069 ± 0.015	1.512
7	0.102 ± 0.012	0.440

To gain an understanding of how the ester functionality can be accommodated in the RPE65 active site, we determined the crystal structure of RPE65 in complex with 6 (Table ). Crystals of the RPE65/6 complex were obtained by cocrystallization. The unbiased omit electron density maps revealed characteristic density for the γ-amino-α-aryl alcohol VCM core together with a bound palmitate molecule as observed in other RPE65 crystal structures (Figure A). ,,, The diethyl acetate tail of the molecule was well-defined but appeared to adopt different conformations in the two different chains in the asymmetric unit. Flexibility in this region of the molecule was due to a smaller number of stabilizing interactions in the membrane-proximal region of the active site and is consistent with the variability seen in the tail region in published RPE65-ligand complexes. Comparison of the mode of 6 active-site binding to that of emixustat and MB-004 revealed an overall similar docking location but with some small shifts in orientation (Figure B).

2. X-ray Diffraction Data Collection, Processing, and Refinement.

Data Collection and Processing
crystal	RPE65/compound 6
X-ray source	SSRL 12–2
wavelength (Å)	0.979460
space group	P65
unit cell lengths (Å)	a = 175.81, c = 86.46
resolution (Å)	50.0–2.10 (2.22–2.10)
unique reflections	87,750 (12,591)
multiplicity	10.0 (4.7)
completeness (%)	98.8 (92.6)
⟨I/σI⟩	8.5 (0.8)
R merge I(%)	15.3 (202.5)
CC 1/2 (%)	99.7 (29.8)
Wilson B factor (Å2)	50

Refinement
resolution (Å)	47.9–2.1
no reflections	83,253 (4,497)
R work/R free (%)	18.0/20.9
no atoms	9396
protein	8333
metal	2 FE2
water	987
ligand	36 (6), 38 (PLM)
⟨B-factor⟩ (Å2)	52.7
protein	51.5
metal	43.9
water	61.6
ligand	75.4 (6), 67.6 (PLM)
RMS deviations
bond lengths (Å)	0.002
bond angles (°)	0.914
Ramachandran plot (% favored/outliers)	97.7/0
all-atom clashscore (%)	1.69 (100th percentile)
PDB accession code	pdb_00009DQA

^aFinal data set obtained by merging data from three isomorphous crystals.

^bValues in parentheses are for the highest resolution shell of data.

^cValue in parentheses indicates the number of reflections used for cross-validation.

^dEvaluated using Molprobity.

2.

Crystal structure of compound 6 in complex with RPE65. (A) Illustration of RPE65 residues within 4.5 Å from the bound ligands. The corresponding omit 2F ₀- F _c electron density map, contoured at 1 RMSD, is shown as a blue mesh within 2 Å of the bound ligands (PDB accession code 9DQA). (B) Comparison of the binding orientations between 6, emixustat (PDB accession code 4RSC), and MB-004 (PDB accession code 5UL5) in the RPE65 active site.

In Vivo Pharmacodynamic Properties of Short-Acting VCM

Next, we examined the inhibitory effects of 5–7 on visual chromophore regeneration in unanesthetized mice following a light exposure that bleached >90% of the visual pigment. In this experiment, dark-adapted BALB/cJ mice received a single intraperitoneal (IP) injection of either vehicle (DMSO) or test compounds (10 mg/kg) in the dark. Thirty min later, the animals were subjected to a 10 min, 10,000 lx photobleach and then placed back in darkness for 2 h to allow visual pigment regeneration. Afterward, the animals were euthanized, and their eyes collected for retinoid analysis by HPLC. The spectral properties of the white LED cluster used for the photobleach and those of the fluorescent white light in our vivarium are shown in Figure S2. Compounds 6 (red trace) and 7 (blue trace) reproduced the retinoid pattern observed after emixustat treatment (purple trace), featuring a buildup of atREs (a) and minimal levels of 11cRAL (b) (Figure A). By contrast, 5 (green trace), like 4 (gray trace), led to the same pattern as vehicle treatment (black trace), with minimal atREs and high 11cRAL levels, respectively. The absolute quantification of these retinoids is shown in Figure B,C. Emixustat led to a 93% reduction of 11cRAL synthesis while 6 and 7 led to a 78 and 71% reduction, respectively. This result demonstrated that the new ester chemistry was compatible with RPE65 inhibition in vivo. The lack of RPE65 inhibition by 5 was attributed to faster hydrolysis based on the results from the in vitro esterase assay (Figure D).

3.

Ester analogs of emixustat are short-acting RPE65 inhibitors. (A) Representative HPLC chromatograms showing the effects of vehicle or test compounds on the retinoid extracted from whole mouse eye homogenates. Dark-adapted, 6–8 weeks old mice received a single IP injection of vehicle or test compounds in the dark. Thirty min after the IP injection, the mice were photobleached and then housed in the dark for 2 h before HPLC analysis. The peaks corresponding to atREs (b) and 11cRAL oxime (syn) (a) are indicated by solid black lines and their UV–vis spectra is shown in the insets. (B and C) Absolute 11cRAL and atREs quantification from the screening of test compounds. ANOVA showed a significant effect of the test compounds (p < 0.0001). Dunnett’s multiple comparison test showed a significant effect of emixustat, 6 and 7 in reducing 11cRAL and augmenting atREs, respectively, compared to the vehicle (*p < 0.05, ^§ p < 0.0001). Data represent mean ± SD, n = 3 animals. (D) Raw HPLC chromatograms from mouse eye retinoid extracts illustrating short-acting RPE65 inhibition by 6 and 7. Dark-adapted, 6–8 weeks old mice were administered a single IP injection of vehicle or test compounds in the dark. At different points in time after the IP injection (0.5–4 h), the mice were photobleached and then housed in the dark for 2 h before HPLC analysis. (E and F) Absolute quantification of 11cRAL synthesis and atREs clearance at different points is time after IP injection of vehicle or test compounds. ANOVA analysis showed a significant effect for the test compounds (p < 0.0001). The results of Dunnett’s multiple comparison test are reported in the figure. (*p < 0.05, ^# p < 0.01, ^‡ p < 0.001, ^§ p < 0.0001).

Next, we examined the duration of RPE65 inhibition following single doses of 6 and 7. In this experiment, dark-adapted mice received a single IP injection of either vehicle or the test compounds (10 mg/kg) and were housed in the dark for different periods of time (0.5–4 h) before being subjected to a deep photobleach. Following a 2 h dark recovery period, the animals were euthanized and their ocular retinoids quantified by HPLC. Figure D shows the raw chromatograms obtained from this experiment. After a period of 4 h in the dark, the treatment with 6 (red traces) and 7 (blue traces) led to the recovery of 11cRAL and atREs levels comparable to those obtained with vehicle treatment (black traces) (Figure E,F). By contrast, emixustat treatment (purple trace) led to significant RPE65 inhibition after 4 h. Importantly, this result confirmed that 6 and 7 were short-acting RPE65 inhibitors.

Electrophysiological measurements of rod function confirmed the short-acting effect of 6 (Figure S2). The mice were administered either vehicle or test compounds (emixustat or 6) via a single IP injection (10 mg/kg) in the dark. Thirty min later, the animals were photobleached, and the course of dark adaptation was monitored between 0.5–8 h after photobleach (Figure S2A). After 4 h of dark adaptation, the animals treated with vehicle (black trace) showed a similar ERG pattern compared to dark-adapted animals (gray trace). As expected, emixustat treatment (purple trace) abolished the a- and b-waves after 8 h of dark-adaptation. Conversely, treatment with 6 (red traces) featured timed recovery of both a- and b-wave amplitudes. After 8 h of dark adaptation, the mice treated with 6 recovered ∼80% of the a- and b-wave amplitudes that were measured in the vehicle group (Figure S2B,C).

Systemic Administration of Compound 6 Protects the Retina from Light-Induced Retinal Damage

Here, we explored whether 6 reproduced the effect of emixustat in protecting the retina of BALB/cJ mice from damage induced by a prolonged exposure to white light. First, dark-adapted animals received a single IP dose of either vehicle or test compounds (10 mg/kg) in the dark. Thirty min later, the mice were illuminated with 15,000 lx of constant white LED light for 8 h. Next, the animals were moved back to the standard light/dark cycle for 7 days after which their retinal structure was assessed by optical coherence tomography (OCT) and scanning laser ophthalmoscopy (SLO) and retinal function assessed by scotopic ERG. Eventually, the animals were sacrificed, their eyes processed for histological analysis, and the number of rod nuclei across the thickness of the outer nuclear layer (ONL) was counted at fixed distances from the optic nerve.

SLO analysis (Figure A) showed that vehicle treatment was associated with the presence of autofluorescent puncta distributed across the whole mouse retina. By contrast, 6 and emixustat treatments were associated with rare autofluorescent puncta. These autofluorescent puncta were attributed to lesions in the neurovascular tissue. OCT analysis (Figure B) revealed that vehicle treatment did not prevent the ablation of the ONL between ± 750 μm from the optic nerve head. Instead, 6 and emixustat treatments preserved the thickness of the ONL (black brackets). The thickness of the ONL was quantified as shown in Figure C, showcasing an equivalent efficacy between emixustat and 6 in preventing retinal damage induced by prolonged exposure to bright light.

4.

A single administration of 6 prevents photic retinal damage in BALBc/J mice. (A) SLO images illustrate the effect of vehicle or test compounds on retinal autofluorescence. Dark-adapted, 6–8 weeks old mice received an IP injection in the dark. Thirty min later the mice were photobleached for 8 h (15,000 lx) and then housed 1 week in regular light/dark cycles before assessment. Scale bars 500 μm. (B) Corresponding retinal OCT images. Test compounds maintain an intact ONL (black brackets). The white arrows indicate the optic nerve head. Scale bars 200 μm. (C) ONL thickness as measured in OCT images at 500 μm from the optic nerve head. Data represent means ± SEM, and circles individual eyes, n = 12. ANOVA analysis showed a significant effect of test compounds (p < 0.0001). The results of Dunnett’s test are reported in the figure. (D) Images of retinal cross sections stained with H&E, scale bars 100 μm. The white arrows indicate the optic nerve head. The white rectangles correspond to the area magnified in the insets; scale bars 50 μm. Abbreviations: retinal ganglion cell layer (RGC); inner nuclear layer (INL). (E) Spider plot showing the thickness of ONL. Data represent means ± SEM, n = 10. ANOVA analysis showed that emixustat and 6 have a significant effect (p < 0.0001). The results of Dunnett’s test are shown in the figure. (F) Corresponding scotopic ERG swipes. The black arrowheads indicate the timing of test flash delivery. (G and H) Quantification of scotopic a-wave and b-wave amplitudes, respectively. ANOVA analysis showed a significant effect of the compounds on the amplitude of a- (p = 0.0005) and b- (p = 0.0016) waves. The results of Dunnett’s multiple comparisons are reported in the figure (*p < 0.05, ^# p < 0.01, ^‡ p < 0.001, ^§ p < 0.0001).

Consistent with the data gathered by live imaging techniques, histological analysis of retinal cross sections (Figure D) showed the presence and absence of retinal damage in the retina treated with vehicle or test compounds, respectively. Notably, ablation of the ONL was found in an area within ± 1.5 mm from the optic nerve head (Figure E), suggesting that the phototoxicity affects the retinal regions where light is directly focused on the retina.

Next, we investigated whether these structural changes translated to a loss of rod function as measured by scotopic ERG (Figure F). In this experiment, the mice were dark-adapted for 24 h before recording scotopic ERG. Vehicle treatment (black traces) did not afford the protective effect on the amplitude of scotopic ERG responses because they were largely reduced compared to 6 (red traces) and emixustat (purple traces) treatment. Quantification of the scotopic a- (Figure G) and b- (Figure H) waves showed significant differences between the vehicle and treatment groups, highlighting that the observed structural damage translated into a diminished capability of the retina to respond to light.

Together, these results demonstrated that the efficacy of a single systemic administration of 6 is equal to that of emixustat in the prevention of photic retinal damage in albino BALB/cJ mice, which are vulnerable to photic retinal damage.

Oral Administration of Compound 7 Protects a STGD1 Mouse Model from Retinal Damage

Scotopic ERG was also used to measure the effect of enterally delivered 6 and 7 in BALBc/J mice. Vehicle or test compounds (emixustat, 6, or 7) (10 mg/kg) were administered to dark-adapted mice by oral gavage. After 30 min, the mice were exposed to a 10 min, 10,000 lx photobleach and housed in the dark for 2 h before measuring the scotopic ERG. As expected, the vehicle treatment (black) led to a complete recovery of the scotopic ERG response while emixustat (purple) suppressed it (Figure S3A,B). Compounds 6 (red) and 7 (blue) suppressed significantly the scotopic ERG a-wave but only 7 suppressed also the b-wave amplitude to about ∼ 50% that of the vehicle treatment. This result was expected (Figure D and Table ) and attributed to the slower hydrolysis of the ester bond of 7 before absorption into the systemic circulation. Consequently, 7 was selected to test the efficacy of short-acting RPE65 inhibitors in preventing the retinal degeneration associated with the dysfunctional visual cycle of the Abca4 ^–/– Rdh8 ^–/– mouse model of STGD1.

Next, 7 and the vehicle control were administered by either IP injection or by oral gavage to light-adapted Abca4 ^–/– Rdh8 ^–/– mice. Thirty min after the administration, the mice were illuminated with white LED light (10,000 lx) for 30 min and then housed under normal vivarium lighting for 1 week before assessing the structure and function of their retinas (Figure ). SLO analysis showed the presence of autofluorescent puncta in the retina of vehicle-treated Abca4 ^–/– Rdh8 ^–/– mice (Figure A) that were like those observed previously in BALB/cJ mice (Figure A). In addition, Abca4 ^–/– Rdh8 ^–/– mice displayed large areas of homogeneous autofluorescent signal in the background arising from lipofuscin of which bisretinoids are a major component. Instead, a single IP injection of 7 (10 mg/kg) prevented the development of these autofluorescent signals. However, oral administration of 7 at doses of 10, 20, and 50 mg/kg did not afford complete protection and both fluorescent signals were detected by SLO.

5.

Oral administration of EYE-003 prevents retina degeneration in a mouse model of STGD1. (A) SLO images illustrate the effect of vehicle or 7 on retinal autofluorescence. Light-adapted, 6–8 weeks old Abca4 ^–/– Rdh8 ^–/– mice received a single oral or IP dose of vehicle or 7 (5 mg/kg). Thirty min after the administration of the vehicle or 7, the mice were photobleached for 30 min (10,000 lx) and then housed in regular light/dark cycles for 1 week before assessment. Scale bars 500 μm. (B) Corresponding retinal OCT images. 7 maintains the ONL (black brackets). The white arrows indicate the optic nerve head. Scale bars 200 μm. (C) ONL thickness as measured in OCT images at 500 μm from the optic nerve head. Data represented as means ± SEM, the individual data points represent single eyes, n = 6 or more eyes per group. ANOVA analysis showed a significant effect of 7 administration (p < 0.0001). Dunnett’s multiple comparison test shows that 7 has a significant effect when dosed by IP injection (10 mg/kg) and orally at doses of 20 and 50 mg/kg (^§ p < 0.0001). (D, E, and F) Quantification of ERG b-wave amplitudes from the rods, S-cones, and M-cones, respectively. ANOVA analysis showed a significant effect of 7 dosing on the ERG b-wave amplitude of rods (p = 0.0002), S-cones (p = 0.002), and M-cones (p = 0.0199). The results of Dunnett’s multiple comparisons test are reported in the figure (*p < 0.05, ^# p < 0.01, ^‡ p < 0.001, ^§ p < 0.0001).

OCT analysis (Figure B) revealed that the vehicle treatment did not afford protection against light-induced ONL thinning and retinal detachment. Instead, the ONL remained intact in the group treated with a single IP dose of 7 (black bracket). Notably, oral treatment with 7 provided a dose-dependent preservation of the ONL layer. The measurement of the ONL thickness (Figure C) showed that the effect of oral administration of 7 was significant starting from 20 mg/kg, which was twice the dose administered by IP injection, although the efficacy of 50 mg/kg treatment was less than 50% of that achieved with the IP control.

The function of rods was measured by scotopic ERG and that of S-cones and M-cones was measured by photopic ERG. As expected, Rod-specific ERG recordings (Figure D) showed that the vehicle treatment (black rhombs) did not afford the preservation of rod ERG responses while the IP administration of 7 (red hexagons) had a significant effect in maintaining rod function. The oral administration of 7 led to a dose-dependent preservation of rod ERG b-wave amplitudes although the significance of these responses was not consistent throughout the whole range of flash intensities tested. A similar trend was observed for S-cone (Figure E) and M-cone (Figure F) b-wave amplitudes. However, cones are resilient to the damage induced by prolonged exposure to bright light, and this property contributed to the lack of consistent significance of the drug effect explored in this study.

These data, on one hand, confirmed that 7 protected the retina of Abca4 ^–/– Rdh8 ^–/– mice from photic damage and, on the other hand, it demonstrated that the inferior efficacy of the oral treatment was likely due to the suboptimal oral bioavailability of 7. Altogether, these results indicated that the oral administration of 50 mg/kg of 7 preserved the soma of a significant number of rod photoreceptors in the ONL of a STGD1 mouse model, the function of these rods was partially impaired while that of cones was maintained.

Discussion and Conclusions

Visual cycle modulation is a therapeutic concept that has been under development for over two decades but, despite its promise in animal models, is one that has yet to yield an FDA-approved molecule. Although highly selective molecules targeting proteins of the visual cycle (e.g., emixustat inhibition of RPE65) or those mediating retinoid delivery to the RPE (e.g., tinlarebant antagonism of RBP4) have been identified, clinical investigations of these molecules have revealed problematic on-target (i.e., visual) side effects that have limited subject acceptance and often complicated clinical trials due to high attrition rates. In general, these adverse effects are believed to originate from the chronic visual cycle suppression that is achieved with these agents, which is not easily overcome by dosing modifications. In the present investigation, we pursued the hypothesis that rational redesign of the RPE65 inhibitor, emixustat, could generate a molecule capable of short-term inhibition of RPE65 thereby allowing visual cycle recovery each day. The key innovation of this study is the introduction of an ester functionality within the emixustat scaffold that confers susceptibility to metabolism by nonselective esterases, including those found within the RPE. The data demonstrate that ester analogs of emixustat are cleaved by one such esterase at a rate that is governed by the steric bulk of the ester substituent. The ester functionality allows hydrolytic depletion of compound stores while preserving high-affinity RPE65 active site targeting. This novel metabolic susceptibility manifested in vivo as an ability to therapeutically suppress the visual cycle but with a duration of action much shorter than that of emixustat.

Based on our findings, we argue that ester-containing, short-acting RPE65 inhibitors hold significant potential for use in the treatment and prevention of retinopathies associated with visual cycle activity. The foremost among these is STGD1, which currently lacks effective therapies but is among the most common recessive Mendelian disorders affecting retinal structure and function. The emixustat ester derivatives exhibited protective effects against light damage in a mouse model of STGD1, providing proof of concept for the therapeutic properties of these molecules. Beyond their suitability to treat STGD1, these metabolically labile RPE65 inhibitors would be ideal for indications where transient visual cycle suppression could be indicated; for example, in situations where the retina is susceptible to light damage, as demonstrated by our light damage protection studies in albino BALB/cJ mice. Ocular surgeons are aware of the delicate trade-off between bright illumination of the field of operation and retinal safety, as highlighted by the history of vision loss caused by surgical illumination. − Although iatrogenic retinal damage with surgical lights is now a rare event, there appear to be vulnerable individuals that may benefit from the prophylactic administration of a short-acting RPE65 inhibitor before ocular surgery. , Our proposition considers the growing interest in ocular gene therapies, where the eye may be exposed to light for extended periods of time (hours). Indeed, there are more than 30 gene therapy trials for retinal diseases actively recruiting or enrolling, several of which require vitrectomy for delivery. Although there are 3D visualization systems for ocular surgery that reduce the amount of illumination needed for the operation, we foresee that the combined use of 3D visualization systems and saVCMs could offer an opportunity for ophthalmologists to create safer protocols for ocular surgeries and ocular gene therapy administration.

The key advantage of using ester-containing RPE65 inhibitors is that they combine a very rapid onset of the pharmacological effect (minutes) with a tunable duration of action (hours), allowing the recovery of visual cycle activity during drug-free periods. By contrast, the effectiveness of RBP4 antagonists relies on sustained RBP4 depletion, which is achieved with a delay of at least 4–6 h from the time of administration. , Furthermore, physiological levels of RBP4 cannot be recovered for days or even weeks after treatment cessation. These pharmacodynamic properties of RBP4 antagonists limit the range of applications compared to that of short-acting RPE65 inhibitors.

The introduction of an ester moiety is expected to diminish the excellent oral bioavailability exhibited by emixustat. Indeed, we observed that 7 had weaker, albeit significant, retinal protective activity when administered orally as compared to by IP injection. However, the medicinal chemistry strategy we elucidate in this work provides a clear path for future development of emixustat ester analogs that are less prone to hydrolysis during the absorption phase after oral administration. Indeed, the pharmacopeia provides several examples of ester-containing drugs (e.g., aspirin, felodipine, and methylphenidate) that avoid first-pass metabolism to reach their sites of action intact. It is well-known that drug clearance can differ significantly between mice and humans. Future research will be required to establish which ester group will provide ideal bioavailability and duration of action in humans.

We acknowledge that mice lack some of the critical features of the human retina, namely a cone-rich macula, which is lost in STGD1, making them an imperfect model of this condition. Nonetheless, the association of visual cycle activity with STGD1 progression, the history of safe emixustat administration to various vulnerable populations, and the shared mechanism of action between this VCM and the new short-acting RPE65 inhibitors compel further development of saVCMs.

Experimental Section

Quantification and Statistical Analysis

The ages of mice used for experiments are given in figure legends. The experiments were not randomized, and the investigators were not blinded to allocation prior to data analysis. Statistical analyses and graph generation were carried out using GraphPad Prism. Statistical methods and details of descriptive statistics are provided in the figure legends. A p value of less than 0.05 was considered significant.

Animal Study Approvals

All animal procedures were approved by the Institutional Animal Care and Use Committees (IACUC) at the University of California Irvine (approval number AUP-22-140) and the Tibor Rubin VA Long Beach Medical Center (approval numbers 1774515 and 1618962). All experimental protocols were conducted following the NIH Guide for the Care and Use of Laboratory Animals, the recommendations of the American Veterinary Medical Association Panel on Euthanasia, and the Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Ophthalmic and Visual Research.

Sex as a Biological Variable

Our study examined male and female animals. No sex-related differences in experimental outcomes were observed.

Animal Husbandry

BALB/cJ mice (Jackson, strain #000651) and Abca4 ^–/– Rdh8 ^–/– (local animal colony) were housed in a standard 12/12-h light/dark cycle environment, fed a standard soy protein-free rodent chow diet (Envigo Teklad 2020X), provided water ad libitum, and housed in plastic cages with standard corncob rodent bedding and 6 g nestlets (Ancare). The spectral information on the white, fluorescent vivarium lights are shown in Figure S4.

General Synthetic Methods

The detailed synthetic methods and the characterization of the chemicals is reported in the Supporting Information. All reactions were performed in oven-dried glassware under inert atmosphere. The reagents were purchased from various vendors (Sigma-Aldrich, Fisher, Oakwood, Combi Blocks, and AA Blocks) and were used as supplied without purification. Thin-layer chromatography (TLC) was performed on 0.25 mm glass-backed EMD Millipore 60 F254 plates. Visualization of TLC plates was accomplished with UV light (254 nm) and final amines were stained with permanganate (KMnO₄). Purification of all intermediates was achieved by using the CombiFlash Nextgen 100 (Teledyne Isco). All final target amines were purified manually by forced air-flow on silica gel (Merck, 230–400 mesh) using eluting solvents (reported as V/V ratio mixture) that were dried with sodium sulfate, filtered and evaporated under reduced pressure. The ¹H, ¹³C NMR nuclear magnetic resonance (NMR) spectra were recorded at 25 °C on Bruker AVANCE NMR spectrometer operating at 500 MHz. Chemical shifts were reported in δ units, part per million, with reference to the residual solvent peak CDCl₃ (δ 7.26), DMSO-d ₆ (δ 2.50), and methanol-d ₄ (δ 3.33) for ¹H and CDCl₃ (δ 77.3) and methanol-d ₄ (δ 49.5) for ¹³C NMR spectra. NMR data are presented in the following order: chemical shift, peak multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, dd = doublet of doublet, dt = doublet of triplet, ddd = doublet of doublet of doublet, dq = doublet of quartet), coupling constant (in Hz). Analytical HPLC analysis of final targets was carried out on an Agilent 1260 series system consisting of a G4204A quaternary pump, a G4226A ALS autosampler, and a G1316C column compartment. The separation was performed on a Shimadzu Premier C18 (5 μm, 100 mm × 4.5 mm) column using a mobile phase consisting of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B) at a flow rate of 1 mL/min, and the mobile-phase gradients and time course were as follows: 0 – 5 min, 95% A/5% B; 5–10 min, 95–5% A/5–95% B; 10–20 min, 5% A/95% B; 20–25 min, 5–95% A/95–5% B; 25–30 min, 95% A/5% B. The final biologically tested compounds displayed ≥ 95% purity by HPLC analysis. The synthesis of the precursors and target esters is detailed in the following section. Chemical structures and NMR spectra for all synthesized compounds and representative HPLC traces of final targets are shown in Figures S5–S22.

Synthetic Chemistry

Benzyl (3-(3-Hydroxyphenyl)-3-oxopropyl)­carbamate (Compound 2)

To a dried flask was added benzyl (3-oxopropyl)­carbamate (0.500 g, 2.4 mmol), Pd­(DBA)₂ (0.037 g, 0.04 mmol), DPPP (0.034 g, 0.06 mmol), 4 Å molecular sieve (2.0 g) and DMF (8 mL), followed by pyrrolidine (0.2 mL, 2.4 mmol) and DIPEA (2 mL). After the mixture turned yellow, a solution of 3-iodophenol (0.440 g, 2.0 mmol) in DMF (5 mL) was injected. Then the reaction was stirred at 115 °C for 2 days. After cooling, the mixture was diluted with EA (250 mL), washed with water (2 × 200 mL) and brine (100 mL), dried over anhydrous Na₂SO₄ and concentrated under reduced pressure. The crude product was further purified by column chromatography (hexanes/EA = 4/1 to 1/1) and concentrated in vacuo to give a product as a deep yellow liquid (0.200 g, 33%). ¹ H NMR (500 MHz, CDCl3) δ 7.43 (s, 1H, Ar–H), 7.43 (d, J = 10.2 Hz, 1H, Ar–H), 7.34–7.26 (m, 5H, Ar–H), 7.26 (t, J = 8.3 Hz), 7.06 (dd, J = 8.3, 3.1 Hz, 1H, Ar–H), 5.57 (t, J = 6.4 Hz, 1H, N–H), 5.08 (s, 2H, CH₂), 3.58 (dt, J = 6.4, 5.7 Hz, 2H, CH₂), 3.15 (t, J = 5.7 Hz, 2H, CH₂). This ¹H NMR spectra is shown in Figure S6. ¹³ C NMR (126 MHz, CDCl₃) δ 199.4, 157.0, 156.8, 137.7, 136.2, 130.0, 128.6, 128.2, 128.1, 121.1, 120.2, 114.7, 67.1, 38.5, 36.1. This ¹³C NMR spectra is shown in Figure S7.

3-(3-(((Benzyloxy)­carbonyl)­amino)­propanoyl)­phenyl cyclohexanecarboxylate (Compound 3a)

To a dried flask was added benzyl (3-(3-hydroxyphenyl)-3-oxopropyl)­carbamate (0.400 g, 1.34 mmol), cyclohexanecarboxylic acid (0.192 g, 1.50 mmol), EDC·HCl (0.326 g, 1.70 mmol), DMAP (0.208 g, 1.70 mmol), THF (25 mL) and DCM (10 mL), then stirred at r.t. overnight. The reaction was directly concentrated under reduced pressure, purified by column chromatography (hexanes/EA = 4/1) and concentrated in vacuo to give a product as a yellow viscous liquid (0.370 g, 68%). ¹ H NMR (500 MHz, CDCl3) δ 7.80 (d, J = 7.8 Hz, 1H, Ar–H), 7.65 (t, J = 2.6 Hz, 1H, Ar–H), 7.47 (t, J = 8.0 Hz, 1H, Ar–H), 7.38–7.31 (m, 5H, Ar–H), 7.30 (dd, J = 8.0, 2.6 Hz, 1H, Ar–H), 5.50 (t, J = 6.4 Hz, 1H, N–H), 5.10 (s, 2H, CH₂), 3.61 (dt, J = 6.4, 5.7 Hz, 2H, CH₂), 3.21 (t, J = 5.7 Hz, 2H, CH₂), 2.64–2.55 (m, 1H, CH), 2.12–2.04 (m, 2H, CH₂), 1.89–1.80 (m, 2H, CH₂), 1.75–1.68 (m, 1H, CH₂), 1.67–1.55 (m, 2H, CH₂), 1.44–1.25 (m, 3H, CH₃). This ¹H NMR spectra is shown in Figure S8. ¹³ C NMR (126 MHz, CDCl₃) δ 198.1, 174.3, 156.4, 151.2, 137.9, 136.6, 129.7, 128.5, 128.1, 128.0, 126.8, 125.3, 121.2, 66.6, 43.1, 38.7, 35.9, 28.9, 25.7, 25.3. This ¹³C NMR spectra is shown in Figure S9.

3-(3-(((Benzyloxy)­carbonyl)­amino)­propanoyl)­phenyl 2-ethylbutanoate (Compound 3b)

To a dried flask was added benzyl (3-(3-hydroxyphenyl)-3-oxopropyl)­carbamate (0.500 g, 1.67 mmol) and THF (15 mL), followed by 2-ethylbutyryl chloride (0.25 mL, 1.80 mmol). The mixture was stirred under 0 °C and NEt₃ (1 mL) was injected. After stirring for 15 min, the reaction was allowed to warm to r.t. and stirred overnight. The reaction was directly concentrated under reduced pressure, purified by column chromatography (hexanes/EA = 4/1) and concentrated in vacuo to give a product as a yellow viscous liquid (0.450 g, 68%). ¹ H NMR (500 MHz, CDCl₃) δ 7.69 (d, J = 8.0 Hz, 1H, Ar–H), 7.53 (t, J = 2.0 Hz, 1H, Ar–H), 7.36 (t, J = 8.9 Hz, 1H, Ar–H), 7.26–7.19 (m, 5H, Ar–H), 7.19 (dd, J = 8.0, 2.0 Hz, 1H, Ar–H), 5.39 (t, J = 6.5 Hz, 1H, N–H), 4.98 (s, 2H, CH₂), 3.50 (dt, J = 6.5, 5.7 Hz, 2H, CH₂), 3.09 (t, J = 5.7 Hz, 2H, CH₂), 2.42–2.35 (m, 1H, CH), 1.74–1.54 (m, 4H, CH₂), 0.94 (t, J = 7.4 Hz, 6H, CH₃). This ¹H NMR spectra is shown in Figure S10. ¹³ C NMR (126 MHz, CDCl₃) δ 198.1, 174.5, 156.4, 151.1, 137.9, 136.5, 129.7, 128.5, 128.1, 128.0, 126.9, 125.4, 121.2, 66.6, 48.9, 38.7, 35.9, 25.0, 11.9. This ¹³C NMR spectra is shown in Figure S11.

3-(3-(((Benzyloxy)­carbonyl)­amino)­propanoyl)­phenyl 2-propylpentanoate (Compound 3c)

To a dried flask was added benzyl (3-(3-hydroxyphenyl)-3-oxopropyl)­carbamate (0.393 g, 1.31 mmol), valpiroic acid (0.23 mL, 1.40 mmol), EDC·HCl (0.345 g, 1.80 mmol), DMAP (0.220 g, 1.80 mmol), THF (25 mL) and DCM (10 mL), then stirred at r.t. overnight. The reaction was directly concentrated under reduced pressure, purified by column chromatography (hexanes/EA = 4/1) and concentrated in vacuo to give a product as a yellow viscous liquid (0.309 g, 56%). ¹ H NMR (500 MHz, CDCl₃) δ 7.66 (d, J = 7.9 Hz, 1H, Ar–H), 7.51 (t, J = 2.1 Hz, 1H, Ar–H), 7.33 (t, J = 7.9 Hz, 1H, Ar–H), 7.22–7.17 (m, 5H, Ar–H), 7.16 (dd, J = 7.9, 2.1 Hz, 1H, Ar–H), 5.48 (t, J = 6.4 Hz, 1H, N–H), 4.96 (s, 2H, CH₂), 3.46 (dt, J = 6.4, 5.7 Hz, 2H, CH₂), 3.06 (t, J = 5.7 Hz, 2H, CH₂), 2.56–2.48 (m, 1H, CH), 1.69–1.60 (m, 2H, CH₂), 1.49–1.40 (m, 2H, CH₂), 1.38–1.28 (m, 4H, CH₂), 0.87 (t, J = 7.3 Hz, 6H, CH₃). This ¹H NMR spectra is shown in Figure S12. ¹³ C NMR (126 MHz, CDCl₃) δ 198.1, 174.8, 156.5, 151.1, 137.9, 136.6, 129.7, 128.5, 128.1, 128.0, 126.9, 125.4, 121.2, 66.6, 45.3, 38.7, 35.9, 34.6, 20.8, 14.1. This ¹³C NMR spectra is shown in Figure S13.

3-(3-Amino-1-hydroxypropyl)­phenol (Compound 4)

To a sealed tube was added 3-hydroxy-3-(3-hydroxyphenyl)­propanenitrile (1.292 g, 7.9 mmol), tetrahydrofuran (60 mL). Then 1 M of lithium aluminum hydride solution in tetrahydrofuran (16 mL) was added by drop, then stirred at 50 °C overnight. The reaction mixture was quenched by MeOH, concentrated under reduced pressure, purified by column chromatography (EA/MeOH/NH₃·H₂O = 6/1/0.4) and concentrated in vacuo to give a product as a light yellow solid (0.765 g, 58%). ¹ H NMR (500 MHz, CD₃OD) δ 7.13 (t, J = 7.85 Hz, 1H, Ar–H), 6.82 (s, 1H, Ar–H), 6.80 (t, J = 7.3 Hz, 1H, Ar–H), 6.68 (dd, J = 8.0, 1.8 Hz, 1H, Ar–H), 4.65 (dd, J = 8.1, 5.2 Hz, 1H, HOCH), 3.35 (s, 1H, OH), 2.74 (m, 2H, CH2), 1.93 (s, 2H, NH2), 1.86 (m, 2H, CH2). This ¹H NMR spectra is shown in Figure S14. ¹³ C NMR (126 MHz, CD₃OD) δ 159.1, 147.8, 130.3, 117.1, 115.4, 114.0, 73.4, 41.9, 39.5. This ¹³C NMR spectra is shown in Figure S15. HRMS (ES): (m/z) calculated for C₉H₁₃NO₂ [M]⁺ 168.1024; found 168.1024.

3-(3-Amino-1-hydroxypropyl)­phenyl cyclohexanecarboxylate formate (Compound 5, EYE-001)

To a flask was added 3-(3-(((benzyloxy)­carbonyl)­amino)­propanoyl)­phenyl cyclohexanecarboxylate (0.370 g, 0.904 mmol), 10% Pd/C (0.056 g) and MeOH (25 mL). The flask was put into a high-pressure reactor, evacuated and backfilled the flask with H₂ (150 Psi), stirred at r.t. for 2 days. The Pd/C was filtered, and the reaction was concentrated under reduced pressure, purified by column chromatography (EA/MeOH/NH₃·H₂O = 12/1/0.4). After concentrated by rotavap, formic acid (0.17 mL) in EA (15 mL) was added. Then it was concentrated in vacuo to give a product as a yellow viscous liquid (0.149 g, 51%). ¹ H NMR (500 MHz, CDCl3) δ 8.54 (s, 1H, HCO₂ ^–), 7.40 (t, J = 7.9 Hz, 1H, Ar–H), 7.28 (d, J = 7.6 Hz, 1H, Ar–H), 7.14 (s, 1H, Ar–H), 6.99 (dd, J = 7.9,1.9 Hz, 1H, Ar–H), 4.87 (dd, J = 8.6, 4.2 Hz, 1H, HOCH), 3.15–3.02 (m, 2H, CH₂), 3.65–3.57 (m, 1H, CH), 2.12–1.94 (m, 4H, CH₂), 1.87–1.80 (m, 2H, CH₂), 1.75–1.68 (m, 1H, CH₂), 1.65–1.54 (m, 2H, CH₂), 1.48–1.28 (m, 3H, CH₂). This ¹H NMR spectra is shown in Figure S16. ¹³ C NMR (126 MHz, CDCl₃) δ 176.1, 152.5, 147.6, 130.5, 124.0, 121.7, 119.9, 72.4, 44.2, 38.4, 37.1, 30.1, 26.8, 26.3. This ¹³C NMR spectra is shown in Figure S17. HRMS (ES): (m/z) calculated for C₁₆H₂₃NO₃ [M]⁺ 278.1756; found 278.1746.

3-(3-Amino-1-hydroxypropyl)­phenyl 2-ethylbutanoate formate (Compound 6, EYE-002)

To a flask was added 3-(3-(((benzyloxy)­carbonyl)­amino)­propanoyl)­phenyl 2-ethylbutanoate (0.534 g, 1.34 mmol), 10% Pd/C (0.053 g) and MeOH (25 mL). The flask was put into a high-pressure reactor, evacuated and backfilled the flask with H₂ (150 psi), stirred at r.t. for 2 days. The Pd/C was filtered, and the reaction was concentrated under reduced pressure, purified by column chromatography (EA/MeOH/NH₃·H₂O = 12/1/0.4). After concentration in a rotavap, formic acid (0.3 mL) in EA (20 mL) was added. Then it was concentrated in vacuo to give a product as a yellow viscous liquid (0.309 g, 74%). ¹ H NMR (500 MHz, CDCl₃) δ 8.56 (s, 1H, HCO₂ ^–), 7.41 (t, J = 7.9 Hz, 1H, Ar–H), 7.29 (d, J = 7.8 Hz, 1H, Ar–H), 7.15 (s, 1H, Ar–H), 7.00 (dd, J = 8.0, 1.6 Hz, 1H, Ar–H), 4.88 (dd, J = 8.7, 4.1 Hz, 1H, HOCH), 3.16–3.04 (m, 2H, CH₂), 3.53–3.45 (m, 1H, CH), 2.11–1.97 (m, 2H, CH₂), 1.83–1.65 (m, 4H, CH₂), 1.05 (t, J = 7.4 Hz, 6H, CH₃). This ¹H NMR spectra is shown in Figure S18. ¹³ C NMR (126 MHz, CDCl₃) δ 176.3, 170.3, 152.4, 147.8, 130.5, 124.1, 121.7, 119.9, 72.3, 50.1, 38.4, 37.2, 26.2, 12.2. This ¹³C NMR spectra is shown in Figure S19. HRMS (ES): (m/z) calculated for C₁₅H₂₃NO₃ [M]⁺ 288.1576; found 288.1577.

3-(3-Amino-1-hydroxypropyl)­phenyl 2-propylpentanoate formate (Compound 7, EYE-003)

To a flask was added 3-(3-(((benzyloxy)­carbonyl)­amino)­propanoyl)­phenyl 2-propylpentanoate (0.309 g, 0.726 mmol), 10% Pd/C (0.036 g) and MeOH (18 mL). The flask was put into a high-pressure reactor, evacuated and backfilled the flask with H₂ (150 Psi), stirred at r.t. for 2 days. The Pd/C was filtered, and the reaction was concentrated under reduced pressure, purified by column chromatography (EA/MeOH/NH₃·H₂O = 12/1/0.4). Formic acid (0.3 mL) in EA (15 mL) was added. The precipitate was filtered and the filtrate was concentrated in vacuo to give a product as a light yellow viscous liquid (0.159 g, 75%). ¹ H NMR (500 MHz, CDCl₃) δ 8.42 (s, 1H, HCO₂ ^–), 7.28 (t, J = 7.9 Hz, 1H, Ar–H), 7.16 (d, J = 7.9 Hz, 1H, Ar–H), 7.00 (s, 1H, Ar–H), 6.85 (dd, J = 8.1, 3.2 Hz, 1H, Ar–H), 4.75 (dd, J = 8.7, 4.0 Hz, 1H, HOCH), 3.02–2.90 (m, 2H, CH₂), 2.56–2.49 (m, 1H, CH), 1.98–1.81 (m, 2H, CH₂), 1.67–1.57 (m, 2H, CH₂), 1.51–1.43 (m, 2H, CH₂), 1.39–1.29 (m, 4H, CH₂), 0.88 (t, J = 7.3 Hz, 6H, CH₃). This ¹H NMR spectra is shown in Figure S20. ¹³ C NMR (126 MHz, CDCl3) δ 176.6, 170.2, 152.4, 147.8, 130.6, 124.2, 121.7, 119.9, 72.4, 46.5, 38.5, 37.2, 35.9, 21.8, 14.4. This ¹³C NMR spectra is shown in Figure S21. HRMS (ES): (m/z) calculated for C₁₇H₂₇NO₃ [M]⁺ 294.2069; found 294.2066.

RPE Microsomal Preparations

Bovine RPE microsomes were isolated from RPE homogenates by differential centrifugation as previously described. The resulting microsomal precipitate was resuspended in 10 mM Bis-Tris propane/HCl buffer, pH 7.4, to achieve a total protein concentration of approximately 5 mg/mL. Then, the mixture was placed into a quartz cuvette and irradiated for 6 min at 4 °C with a ChromatoUVE transilluminator (model TM-15; UVP) to eliminate residual retinoids. After irradiation, DTT was added to the RPE microsomal mixture to achieve a final concentration of 5 mM.

In Vitro RPE65 Activity Assay

Compounds 4–7 and emixustat (2 μL in DMF, with the final concentration ranging between 0.1 and 10 μM) were added to 10 mM Bis-Tris propane/HCl buffer, pH 7.4, containing 150 μg RPE microsomes, 1% BSA, 1 mM disodium pyrophosphate, and 20 μM apo cellular retinaldehyde-binding protein (CRALBP). The resulting mixture was preincubated at room temperature for 5 min. Then, all-trans-retinol (1 μL in DMF), was added to a final concentration of 20 μM. The resulting mixture was incubated at 37 °C for 1 h. The reaction was quenched by adding 300 μL methanol, and the products were extracted with 300 μL of hexanes. Production of 11-cis-retinol was quantified by normal-phase HPLC with 10% (v/v) EtOAc in hexanes as the eluent at a flow rate of 1.4 mL/min. Retinoids were detected by monitoring absorbance at 325 nm and quantified based on a standard curve representing the relationship between the amount of synthetic 11-cis-retinol standard and the area under the corresponding chromatographic peak. IC₅₀ and relative SD were obtained by fitting the results from each inhibitor using the inhibitor concentration vs normalized response – variable slope function of GraphPad Prism software.

Shelf Life of Ester-Based VCMs

Ten mg of ester-based VCMs and 10 mg of acetaminophen (internal standard) were dissolved in 25 mL of PBS. This solution was dispatched in 1 mL aliquots into an autosampler vial and capped. The vials were incubated at 37 °C. Three vials were analyzed at each time point (0–16 days) by injecting 50 μL of the solution from individual vials in an HPLC equipped with a C18 column. The flow rate was 1 mL/min and the esters were detected at 265 nm. The test compounds were eluted with a mixture of water and methanol supplemented with 0.1% formic. The elution gradient was the following: 0 min, 100% A; 10 min, 50% A; 15 min, 0% A; 22 min, 0% A; 22.01 min, 100% A; 29 min, 100% A. The area under the peak of the VCMs was normalized with the area under the peak of the acetaminophen and the ratio was converted in % of VCM by normalizing to the time 0 min.

In Vitro Esterase Susceptibility Assay

Esterase from the porcine liver (E3019–20KU, Sigma) was dissolved in DBPS (SH30028.02, Cytiva) to achieve the concentration of 20 mU/mL. 500 μL aliquots of the enzyme solution were prewarmed to 37 °C for 10 min prior to adding 5 μL of 10 mM test compound in DMF. The sample was mixed and then incubated at 37 °C with 300 rpm shaking in an Eppendorf Thermomixer. 100 μL samples were taken at 0, 5, 15, 30, 60 min after the initiation of the reaction. At each time point, the reactions were immediately quenched with 100 μL of ice-cold methanol, vortexed for 3 s, and stored at −20 °C. Next, the solvent of each sample was evaporated in a SpeedVac vacuum concentrator. Each dried sample was reconstituted in 100 μL in 0.1% (v/v) trifluoroacetic acid in water, centrifuged to remove particulates, and then transferred to an HPLC vial. 50 μL of the sample was analyzed on an Agilent 1260 Infinity series HPLC equipped with a Proshell EC-18 column (Agilent) and a diode array detector. The sample was separated using a mobile phase consisting of 0.1% (v/v) trifluoroacetic acid in H₂O (A) and acetonitrile (B) at the following ratios and time intervals: B was increased from 0 to 50% between 0–10 min followed by a gradient to 100% B over 5 min; and then a gradient from 100 to 0% B over 5 min. The disappearance of the reaction substrates was assessed by monitoring absorbance at 265 and 275 nm. Test compounds were eluted at ∼11.75 min. Results were normalized by dividing the results by the value obtained at 0 min. Results were fit using the one-phase decay slope function of GraphPad Prism software.

HPLC Analysis of Retinoid Extracts from Whole Mouse Eye Homogenates

For retinoid analysis, both eyes from each animal were homogenized in 1 mL of 10 mM sodium phosphate buffer, pH 8.0, containing 50% v/v methanol (Sigma-Aldrich; 34860-1L-R) and 100 mM hydroxylamine, pH = 8 (Sigma-Aldrich; 159417-100G). After a 15 min incubation at room temperature, 2 mL of 3 M sodium chloride were added to the homogenate. The resulting sample was extracted twice with 3 mL of ethyl acetate (Fisher Scientific; E195-4). The combined organic phase was dried in vacuo and reconstituted in 450 μL hexane. Retinoids extracts (100 μL) were analyzed with an Agilent 1260 Infinity II HPLC system equipped with a diode array detector (DAD) and a Zorbax RX-Sil column (5 μm; 4.6 mm × 250 mm; Agilent Technologies). The mobile phase consisted of 0.6% v/v ethyl acetate in hexanes (Fisher Scientific; H302-4) flowing at 1.4 mL/min for 17 min followed by a step increase to 10% v/v ethyl acetate in hexanes flowing at 1.4 mL/min for an additional 25 min. Retinoids were detected by monitoring absorbance at 325 and 360 nm with continuous spectral recording. Peaks were identified based on their absorbance spectra and retention times relative to authentic standards. Absolute quantification of retinoids was conducted by peak integration with reference to authentic retinoid standards.

Extent and Duration of 5–7 Effects on Visual Chromophore Recovery in BALBc/J Mice

Six to eight-week-old BALBc/J Mice were dark adapted for 24 h prior experiments. All drug administration procedures were performed under dim red light. The mice were administered 50 μL of vehicle (DMSO) or test compounds (emixustat and 4–7) at doses of 10 mg/kg by IP injection. Next, the pupils of the animals were dilated by topical administration of 1% tropicamide ophthalmic solution (Akorn; 17478-102-12) and 10% phenylephrine ophthalmic solution (MWI Animal Health #054243). Thirty min after IP injection, the mice were placed in a cage with a light-reflective white coating. Each cage was divided into four equal compartments with transparent Plexiglas separators. To prevent grouping and consequent light shielding, single animals were placed into their own compartment for the duration of the experiment. Then exposed to 10,000 lx white LED light for 10 min to bleach their visual pigment. Next, the animals were allowed to recover in the dark for 2 h before euthanasia, eye collection, and retinoid analysis by HPLC. To study the drug duration effect of 6 and 7, the same procedure of above was carried out with only one difference: the mice were left in the dark for different amounts of time (0.5–4 h) between the IP administration and the photobleach. After the photobleach, the mice were dark-adapted for 2 h before euthanasia, eye collection, and retinoid analysis by HPLC.

In Vivo ERG Measurement of Rod Dark Adaptation

BALBc/J mice were dark-adapted for 24 h. Under a red safety light, the mice were administered test compounds a single IP injection or gavaged with a single dose of vehicle or test compounds (10 mg/kg). Thirty min later the animals were placed in a white cage and exposed to 10,000 lx white light for 10 min and then housed in the dark. Rod-driven ERG responses were recorded between 0.5–8 h after the bright light exposure to track the dark adaptation process. In the experiments using oral gavage as the method of drug administration, dark adaptation was assessed 2 h after the photobleach.

The scotopic ERG was measured as follows: mice were anesthetized by isoflurane inhalation and their pupils were dilated with topical administration of 1% tropicamide ophthalmic solution (Akorn; 17478-102-12) and 10% phenylephrine ophthalmic solution (MWI Animal Health #054243), followed by 0.3% hypromellose (Akorn; 9050-1) to maintain corneal hydration. Anesthetized mice were placed on a heated pad set to 37 °C to prevent hypothermia. ERG responses were measured using a Diagnosys Celeris rodent-ERG device (Diagnosys LCC, Lowell, MA). Ocular stimulator electrodes were placed on the corneas, the reference electrode was positioned subdermally between the ears, and a ground electrode was placed in the rear leg. The eyes were subjected to a green-light stimulus (peak emission 544 nm, bandwidth ∼160 nm) of 0.5 cd·s/m², which exclusively excites rod photoresponses. The responses obtained from 5 consecutive stimuli with an interstimulus interval of 30 s were averaged, and the a- and b-wave amplitudes were acquired from the averaged ERG waveform. Data were analyzed with Espion V6 software (Diagnosys LLC).

Induction of Retinal Phototoxicity in Mice

Six to eight-week-old BALBc/J mice were dark-adapted for 24 h prior to the experiment, and all drug administration procedures were performed under dim red light. Animals were administered 50 μL of vehicle or test compounds (emixustat or 6) dissolved in DMSO at a dose of 10 mg/kg in DMSO vehicle by IP injection. Pupils were dilated by the application of a drop of 1% atropine sulfate to the eye (Ophthalmics Inc.). The mice were placed in a cage with a light-reflective white coating. Each cage was divided into four equal compartments with transparent Plexiglas separators. To prevent grouping and consequent light shielding, single animals were placed into their own compartment for the duration of the experiment. Mice were exposed to broad-spectrum 15,000 lx white light emitted by clusters of LEDs placed on top of each white cage for 8 h. The animals had full access to food and water during light exposure. During the light exposure, atropine sulfate was applied to mouse eyes every 2 h to maintain mydriasis. After light exposure, animals were returned to their normal cages and kept under standard lighting conditions for 1 week before SLO and OCT imaging and ERG.

The same procedure as above was used to induced photic retinal damage in Abca4 ^–/– Rdh8 ^–/– mice with minor modifications. Abca4 ^–/– Rdh8 ^–/– mice were adapted to the vivarium light conditions. Next, these mice were administered either 50 μL of vehicle (DMSO) or 7 dissolved in DMSO (10 mg/kg) by IP injection or 100 μL of an emulsion containing 10% DMSO – soybean oil by oral gavage. For the oral administration, 7 was dosed at 10, 20, or 50 mg/kg and compared to the 10% DMSO – soybean oil without test compound. Next, their pupils were dilated as mentioned above. Thirty min after the IP injection or oral administration, the mice were placed in the same cage with a light-reflective white coating used for inducing retinal damage in BALBc/J and photobleached at 10,000 lx for 30 min. After the photobleach, the Abca4 ^–/– Rdh8 ^–/– mice were returned to their normal cages and kept under standard lighting conditions for 1 week before SLO and OCT imaging and ERG.

In Vivo Retinal Imaging

Mice were anesthetized by an IP injection of ketamine (100 mg/kg) and xylazine (8.75 mg/kg) and pupils were dilated with 1% tropicamide and 10% phenylephrine before imaging. Ultrahigh-resolution spectral domain OCT (Bioptigen, Research Triangle Park, NC) was performed for cross-sectional imaging of mouse retinas as described previously. Briefly, five frames of OCT images were first acquired in the B-mode in two orthogonal directions and then averaged. Images were then analyzed using ImageJ software. Briefly, for each animal, the images collected from each eye were used to measure the thickness of the outer nuclear layer (ONL) in the four retinal quadrants at 0.5 mm from the optic nerve head. The values from the right and left eye of the same animal were averaged. During the same session the fundus was imaged by scanning laser ophthalmoscopy (SLO) (Heidelberg Engineering, Heidelberg, Germany) in autofluorescence mode as previously described.

In Vivo ERG Scotopic Intensity-Response Measurements after Light Damage

One week after induction of retinal damage induced by light, mice were dark-adapted overnight for rod ERG recordings. All procedures were performed like described above for the measurement of rod dark-adaptation with the difference that several intensities of green-light stimulus were assessed (0.002 to 100 cd·s/m²). The responses obtained from consecutive stimuli were averaged, and the a- and b-wave amplitudes were acquired from the averaged ERG waveform. Data were analyzed with Espion V6 software (Diagnosys LLC).

In Vivo ERG Photopic Intensity-Response Measurements after Light Damage

The eyes of mice were stimulated with a green light-emitting diode (LED) (peak 544 nm, bandwidth 160 nm) as steady rod-suppressing background light. To measure M-cone and S-cone function, stimulation was performed with alternating green and UV light at increasing intensities. Green light stimulation (peak emission 544 nm, bandwidth 160 nm) had intensity increments of 0.3, 0.5, 1, 3, 10, 30, 100 cd·s/m². UV light stimulation (peak emission 370 nm, bandwidth 50 nm) had intensity increments of 0.03, 0.05, 0.1, 0.3, 0.5, 1, 3, 10, 30 log cd·s/m². The responses for 20–25 stimuli were averaged together, and the a- and b-wave responses were acquired from the averaged ERG waveform. The ERGs were analyzed with the Espion V6 software (Diagnosys LLC).

Histology

The enucleated mouse eyes were kept in Hartman’s fixative (Millipore Sigma) for 24 h at RT, transferred to 70% ethanol, and embedded in paraffin. Sagittally cut 6-μm paraffin sections spanning the optic nerve head (ONH) were stained with hematoxylin and eosin (H&E) and imaged with light microscopy with BZ-X800 (Keyence) instruments. Manual counting of photoreceptor nuclei per row, every 500 μm starting from the edge of the ONH along both superior and inferior directions, was performed manually. Average values of ONL nuclei counts for each animal group represent data obtained from 6 eyes.

RPE65 Crystallization and Structure Determination

Crystals of RPE65 in the complex with 6 were obtained using previously described procedures. Briefly, isolated bovine RPE membranes were incubated with 2.5 mM 6 (delivered in DMF) for 15 min before solubilization with 24 mM hexaethylene glycol mono-octyl ether (C₈E₆). After anion-exchange chromatography, purified RPE65 was concentrated to 10–15 mg/mL and 6 was added again to a concentration of 1 or 10 mM before crystallization. Crystals were grown by the hanging-drop vapor diffusion method on siliconized coverslips by mixing the concentrated RPE65 sample with 2 μL of crystallization solution consisting of 100 mM Tris-HCl, pH 8.5, containing 30% (v/v) polyethylene glycol 200 and 200 mM ammonium phosphate dibasic. The drops were placed over a well solution consisting of 100 mM 2-(cyclohexylamino)­ethanesulfonic acid–NaOH, pH 9.5, containing 40% (v/v) polyethylene glycol 300 and 200 mM NaCl. After incubation for 1–2 weeks at 8 °C, crystals of ∼50 × 50 × 400 μm³ in size were observed. Mature crystals were harvested directly into liquid nitrogen for X-ray data collection. X-ray diffraction data were collected at the SSRL 12–2 beamline. Data were processed with XDS, , and the initial model was obtained by direct refinement using published RPE65 coordinates in which ligands had been removed (PDB accession code 4RSC). The structure was refined by alternating reciprocal space refinement in REFMAC and manual building and adjustments in Coot. Ligand coordinates and geometry dictionary files were generated using the grade server (http://grade.globalphasing.org/cgi-bin/grade/server). The models were validated using MolProbity and the wwPDB validation server.

Glossary

Abbreviations Used

11cRAL

11-cis-retinal

AMD

age-related macular degeneration

atREs

all-trans-retinyl esters

atRAL

all-trans-retinal

ERG

electroretinography

IP

intraperitoneal

OCT

optical coherence tomography

RPE

retinal pigment epithelium

SLO

scanning laser ophthalmoscopy

STGD1

Stargardt disease 1

VCM

visual cycle modulator

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jmedchem.5c01353.

• Aminolysis time course data; Rod-specific ERG recovery data; ERG suppression data after oral administration of short-acting VCMs; spectra of the white light used in experimental procedures; molecules synthesized during the course of this study; ¹H NMR spectra; ¹³C NMR spectra; HPLC traces from hydrolysis stability studies (PDF)

• Raw data used to produce figures (XLSX)

• Molecular formula strings of the target compounds (CSV)

M.B. designed research, conducted experiments, acquired data, analyzed data, wrote the manuscript. Y.H. medicinal chemistry and synthesis of EYE compounds. B.L. medicinal chemistry and synthesis of AHPP. X.C. conducted experiments, acquired data. V.S. conducted experiments, acquired data. F.D. conducted experiments, acquired data. R.S. conducted experiments, acquired data. K.P. analyzed data, and supervised the project. G.P.T. designed research, analyzed data, and supervised the project. P.D.K. designed research, conducted experiments, acquired data, analyzed data, supervised the project, and wrote the manuscript.

The authors declare the following competing financial interest(s): M.B., P.D.K., K.P., and G.P.T. are equity holders in Eyesomer Therapeutics Inc. K.P. is a consultant for Polgenix Inc. and AbbVie Inc., and serves on the Scientific Advisory Board of Hyperion Eye Ltd.