Synthesis of the Nakanishi Ring-Locked Retinoid

An optimized synthetic route to prepare ring-locked retinoid 1a has been developed. We fully describe a purification protocol that provides isomerically pure 1a in support of on-going proof of concept studies for the development of therapeutic agents to treat human ADRP. Additionally, we have found that isomerically pure 1a can be stored in amber vials under argon at −20°C for use over time (up to six months) without degradation. Thus, enabling 1a to be an accessible and valuable biological tool.

Previous studies using P23H mutant opsin as an in vitro model of human Autosomal Dominant Retinitis Pigmentosa (ADRP) found that ring-locked retinoid 1a (Figure 1, carbon numbering shown based on 11-cis-retinal) was able to act as a pharmacological chaperone [1][2][3]. Compound 1a induced the mutant protein to fold properly and undergo normal cellular transport and degradation suggesting 1a could be a potential therapeutic agent for the prevention of ADRP [1][2][3]. This ring-locked analog of 11-cis-retinal has also been used extensively to elucidate the mechanism of photoactivation of rhodopsin [4][5][6].
We required a reliable synthesis of isomerically pure compound 1a. Its synthesis, along with the spectral data, has been reported twice in the literature [1,7], in addition to the syntheses of a number of related ring-locked retinoids [4][5][6]8]. The original synthetic route reported by Nakanishi resulted in the preparation of mixtures of multiple geometric isomers of 1 (E/Z-isomers among C7-C8, C9-C10, and C13-C14) and did not describe specific HPLC conditions for resolution of the final isomerically pure products. Our synthetic strategy differs from the two reported syntheses in two fundamental ways (discussed below) enabling ready access to 1a. Two key bond-forming events distinguish this synthesis: the C7-C8 double bond was originally established via a Julia olefination, while the connection between C13 and C14 was accomplished via a Petersen olefination. In each case, both E-and Z-isomers of the desired products were reported to have formed [7]. In contrast, our use of a Wittig reagent in the formation of the C7-C8 double bond afforded the E-isomer as the sole product in the former olefination; and in the case of the latter, the requisite Petersen reagent, a silylated acetaldehyde tert-butylamine, was not readily accessible in our hands, and an alternate methodology was instead utilized [7,9]. A Horner-Emmons-Wadsworth strategy (utilizing diethyl cyanomethylphosphonate) was found to be an advantageous replacement for the penultimate formation of the C13-C14 double bond in two steps (as opposed to three as previously reported) [1]. In particular, using the nitrile Horner-Emmons-Wadsworth (Scheme 2) reagent obviated the need for a final manganese dioxide oxidation which resulted in a more efficient and reliable synthesis endgame for the preparation of 1a. The key changes we have introduced to the synthetic route as described by Noorwez et al. [1] allowed us to increase the yields and geometric purity of key intermediates (7,11). Herein, we report the full details of the synthesis, purification, characterization, and stability of isomerically pure 1a.
Starting from commercially available cycloheptenone 2, the allylic acetate 4 was prepared via a two-step sequence of radical bromination with NBS and 1,1azobis(cyclohexanecarbonitrile) (ACHN) followed by displacement of the allylic bromide 3 with potassium acetate under phase transfer catalyst conditions with tetrabutylammonium bromide (TBAB, Scheme 1). The ketone underwent Horner-Emmons-Wadsworth olefination with diethyl(1cyanoethyl)phosphonate to provide a 50% yield of 5 and 6 with an E : Z ratio of 2 : 1. The isomers were separated by column chromatography, and the geometry of the double bond was established by NOE ( Figure 2). The allylic acetate 5 was hydrolyzed with potassium carbonate in methanol to furnish the allylic alcohol 7 in excellent yield and purity. The alcohol was protected as the TBS ether 8 in the presence of imidazole and DMAP, and then the nitrile was reduced with DIBAL-H to provide aldehyde 9 in 77% yield over two steps. The aldehyde 9 was carried forward without purification. Importantly, the double-bond geometry (between C9 and C10) remained fixed during this three-step reaction sequence. The trans geometry was confirmed by NOE experiments (key interactions are shown in Figure 2).
The Wittig olefination reagent, β-cyclogeranyl triphenylphosphonium bromide, was prepared following established methodology [10,11]. Reaction conditions using n-BuLi as the base resulted only in decomposition. However, the desired tetraene 11 was obtained when potassium tertbutoxide was employed as the base in the presence of a catalytic amount of 18-crown-6 [12]. In addition, the newly established C7-C8 double bond was solely of the trans geometry. The cis isomer of 11 was not obtained or detected due to judicious choice of conditions, optimization of the reaction time/temperature, and the inherent instability of the C7/C8 cis isomer due to the unfavorable sterics between the C5/C9 methyl groups. Interestingly, compound 11 was found to be quite stable when stored at 4 • C in the dark under argon (no decomposition or isomerization detected after 18 months). The TBS protecting group was cleaved using TBAF to provide the corresponding allylic alcohol. In contrast to 11, we found the allylic alcohol readily decomposed under normal atmospheric conditions; thus, it was more convenient to oxidize it directly using MnO 2 to provide ketone 12 which, in our hands, was more stable to manipulation. The Horner-Emmons-Wadsworth reaction on 12 with diethyl cyanomethylphosphonate provided the corresponding nitrile in excellent yield as a 1 : 1 mixture of Eand Z-isomers which, as reported on similar analogues, were not separable by HPLC [8]. Thus, reduction of the nitrile was achieved by treatment with DIBAL-H to the intermediate imine, and silica gel-mediated hydrolysis furnished the aldehydes 1a and 1b in high yield. The observed ratio of C13-C14 double-bond isomers and reaction yield is related to the conditions of the final step of synthesis. Hence, if hydrolysis of the intermediate imine from the conversion of 12 to 1 on silica gel is left longer than 30 minutes at room temperature, more isomerization and decomposition takes place (resulting in the appearance of three other peaks in the HPLC trace), and the yield of 1a/b decreases. Under the conditions reported herein, the combined percentage of the other impurities in addition to the C13-C14 E/Z-isomers was minimal (4% AUC at 350 nm). Both the C13-C14 E/Zisomers of 1 were readily separated by normal phase HPLC in a dark room (see Experimental). The isomers were identified by comparison of key chemical shifts in the 1 H NMR spectra to literature data (Table 1) [1,7].
It was observed that the isomerization of the C13-C14 double bond of 1 occurs readily in solution under normal atmospheric conditions regardless of exposure to light. Hence, a crude sample isolated from the reaction as a 1 : 1 mixture of 1a and 1b would isomerize between the two compounds.
To probe the isomers' stability to light and storage conditions, 2 mg of each purified isomer was dissolved in hexane (2 mL) in a clear glass vial. Half of this solution was transferred into an amber glass vial. Each of the vials was flushed with argon and left to stand at room temperature in ambient light. After one hour, aliquots were removed from each vial and analyzed by HPLC. The samples in the amber vials remained unchanged, whereas the samples in the clear vials had isomerized to 13% (AUC at 350 nm). The vials were flushed with argon again and left overnight at room temperature. Again, aliquots from each vial were analyzed by HPLC, the samples in the amber vials remained unchanged, but the samples in the clear vials had isomerized further to 31% ( Figure 3).
Importantly, in a separate experiment, a purified sample of 1a was stored neat at −20 • C in an amber vial flushed with argon. It was found that the sample remained unchanged from isomerization or decomposition (via HPLC monitoring) after six months.
In conclusion, we have developed a synthetic route to prepare the ring-locked retinoid 1a and described a reproducible purification protocol by normal phase HPLC to isolate the isomerically pure C13-C14 Z-isomer of 1. Finally, we have found that under proper conditions 1 can be stored for extended periods of time (up to six months) without degradation. 1 H, 13 C spectra were recorded on a Bruker ARX 300, 400, or 600 MHz spectrometers. Low-resolution mass spectra (MS) were obtained on a Thermo Finnigan LCQ Classic or LCQ Duo spectrometer. HPLC purifications were conducted on a Waters Prep LC 4000 equipped with a Waters plus 717 autosampler and a Waters 2487 dual λ Absorbance Detector.
All commercial reagents were used as received, unless otherwise noted. All reactions, work-up, and purifications for compounds 1, 9-12 were performed in a dark room utilizing a red spectrum light source. Final compounds and isolated key intermediates were stored in amber glass vials at −20 • C under argon. Analytical thin-layer chromatography (TLC) was performed on glass-backed TLC plates (silica gel 60 F 254 from Silicycle), visualized with a UV lamp (254 nm), and stained with a basic solution of KMnO 4 . Flash column chromatography was conducted on an ISCO    Figure 2: Key NOEs used to establish the geometry of the tetra-substituted double bond corresponding to the double bond between C9 and C10 of 1.    (5) (See [12]). Into a dried flask fitted with an addition funnel was placed NaH (60% dispersion in oil, 200 mg, 84.0 mmol) and THF (140 mL). The slurry was cooled in an ice bath, and the addition funnel was charged with a solution of diethyl(1cyanoethyl)phosphonate (1.06 mL, 6.00 mmol) in THF (45 mL). The phosphonate solution was added slowly over 5 minutes to the NaH slurry at 0 • C. The reaction mixture was then warmed to r.t. over 30 minutes. A solution of 4 (900 mg, 5.40 mmol) in THF (30 mL) was added via addition funnel to the phosphonate solution over 5 minutes. The reaction mixture was then stirred at r.t. for 18 h. The reaction was quenched by the addition of water (200 mL) and then extracted with ether (3 × 150 mL). The combined organic phases were dried over MgSO 4 , filtered, and concentrated in vacuo. The E-and Z-isomers were separated by column chromatography (120 g silica gel cartridge, isocratic elution 10% EtOAc-hexanes); yield of 5, 0.36 g (1.8 mmol, 33%) and 6, 0.23 g (1.1 mmol, 17%).

(E)-4-((E)-4-(2,6,6-Trimethylcyclohex-1-en-1-yl)but-3-en-2-ylidene)cyclohept-2-enone (12) (See
International Journal of Medicinal Chemistry 7 minutes at r.t., then the flask was wrapped in aluminum foil. A solution of the 12 (130 mg, 0.450 mmol) in THF (6 mL) was added to the reaction mixture, and the reaction was stirred for 3 h at r.t. in the dark. The reaction was poured into 30 mL of ice water and then extracted with ether (4 × 20 mL). The combined organic phases were washed with brine (20 mL), dried over MgSO 4 , filtered, and concentrated in vacuo. The intermediate nitrile was purified by column chromatography (40 g silica gel cartridge, isocratic 5% EtOAc-hexane); yield of pentaene nitrile 0.14 g (0.45 mmol, 99%). To a solution of the nitrile (140 mg, 0.450 mmol) in Et 2 O (60 mL) at −78 • C in the dark was added DIBAL-H (1 M in CH 2 Cl 2 , 3.00 mL, 3.00 mmol). The reaction mixture was stirred at −78 • C for 10 minutes, then warmed to r.t., and stirred for 30 minutes in the dark. It was then poured into a vigorously stirred slurry of silica gel (5.50 g, ∼40 times the mass of the nitrile) and ether (20 mL) under argon at r.t. The slurry was stirred for 30 minutes at r.t. The silica gel was removed by filtration under argon, and the silica gel was washed with ether (100 mL). The bright yellow filtrate was concentrated in vacuo, yield of E/Z-1 0.14 g (0.45 mmol, 96%, E : Z ratio ∼1 : 1). The product was purified by HPLC using a normal phase semiprep column (Phenomenex Luna 5 μ silica 100Å, 250 mm × 10 mm × 5 mm), isocratic elution 5% ether-hexanes, and flow rate of 10 mL min −1 and monitored at 350 and 210 nm. For optimal separation of isomers 1 mg aldehyde in 0.2 mL hexanes (HPLC grade) was injected per run. This provided 1a with an isomeric purity of 96% by HPLC (AUC at 350 nm), and 1b with an isomeric purity of >98% by HPLC (AUC at 350 nm) (see Figure 3, panels 1a and 1b at t 0 ).