Takeda Researchers Achieve Kilogram-scale GMP Production of a TORC1/2 Inhibitor

Unsuitable chiral separation
Based on their experience during the discovery phase, researchers Erin O'Brien, Scientist II, and colleagues in the Process Chemistry R&D department at Takeda Cambridge US realized that the installation of a quaternary stereogenic center would be key. The initial discovery synthesis, while convergent, resulted in the production of a racemic mixture of the TORC 1/2 inhibitor, which was then separated via chiral supercritical fluid chromatography to provide the desired enantiomer. This approach was suitable for discovery work, but the loss of the racemate in the final separation step was not practical for larger-scale production. The researchers thus set out to develop a process in which the quaternary carbon center was established in high enantiopurity, thus avoiding the need for a chiral separation.

Early installation of chirality
The researchers reported that the new approach involved installation of the chiral center at the very early stages of the synthesis (1). First, a chiral (S)-(+)-morpholine derivative was prepared in four steps from an N-benzyl protected serin derivative via ring closure with 2-chloroacetyl chloride followed by reduction with borane and then N-tert-butoxycarbonyl (BOC) protection. Although it was expected that methylation would lead to loss of enantiopurity of the (S)-(+)-morpholine derivative, the researchers were hopeful that examples of “memory of chirality” reported in the literature for amino ester derivatives similar to the protected morpholine derivative under investigation suggested the possibility of such behavior in their system (1).

Making use of “memory of chirality”
Memory of chirality refers to a phenomenon that occurs with trisubstituted carbons, in which the chirality of the carbon center is preserved in the quaternary carbon final product, even though the key intermediate (anion, carbenium ion, radical) does not retain the chirality. Such an intermediate, however, is conformationally chiral, and this form of chirality is a prerequisite for memory of chirality. In addition, the conformationally chiral intermediate must undergo racemization at a very slow rate while the rate of product formation is very rapid. Memory of chirality has been demonstrated for a wide variety of reaction chemistries, including inter- and intramolecular alkylations of various substrates via enolate formation, radical quenching and cyclization, and substitution reactions via carbocations (2) .

The researchers reported that when the α-methylation reaction using methyl iodide was carried out at -78 °C using sodium or potassium bis(trimethylsilyl)amide (NaHMDS or KHMDS, respectively) as the base in tetrahydrofuran (THF), the desired methyl-substituted morpholine derivative was obtained in 99% enantiomeric excess (ee), thus with nearly complete retention of chirality–even on a 100-g scale (1). It was found, however, that the low temperature was critical to maintaining the high level of selectivity. Even raising the temperature to -40 °C led to erosion of the enantiopurity to unacceptable levels (74%) (1). The choice of solvent was also found to be important; in toluene rather than THF using KHMDS, a lower ee was obtained (1).  

The researchers reported that notably, in the plant, the (S)-(+)-methyl-substituted protected morpholine intermediate was obtained in 97% ee on a 16-kg scale, the researchers reported . After removal of the BOC group, the enantiopurity of the free morpholine derivative was improved to 99.5% by forming its salt with (+)-camphorsulfonic acid (CSA) and separating the isomers (1).

Addition, reduction/cyclization and alkylation
The researchers reported that nucleophilic addition of the CSA salt to commercially available and inexpensive 2,4-dichloro-5-nitropyrimidine proceeded to give a 2.5:1 mixture of C4 and C2 addition products generated using diisopropylethylamine as the base at -30 °C (1). Although this ratio was lower than expected based on reported aromatic substitution reactions of this nitropyrimidine (likely due to steric interference), the researchers found that if the crude product was used directly in the next reduction/cyclization step, the reduced C2 regioisomer ended up in the filtrate (1). The reduction/cyclization step was carried out using platinum on carbon (Pt/C) to provide the desired lactam (1). The researchers noted that the reaction was run in THF at a low catalyst loading of from 0.5 mol% to 0.25 mol%, 35 °C, and a hydrogen pressure of 25 psi in order to maintain a reaction rate that was safe for scale up. The product was then isolated after filtration of the catalyst followed by precipitation. Both undesired impurities and residual platinum were removed in downstream processes (1).

Difluoropropylnosylate (prepared from difluoropropanol) was then used to introduce the difluoropropyl substituent to the lactam (1). The reaction was run in dimethylacetamide (DMAc) at 80 °C using potassium carbonate as the base to give a 5.8:1 ratio of the N- and O-alkylated products. The researchers found that treatment with methanesulfonic acid at 60 °C in THF led to precipitation of the mesylate salt of the desired N-alkylated product.

Dealing with palladium
The researchers reported that final step involved Suzuki coupling of the mesylated salt with commercially available 4-cyclopropyl-urea-pinacol boronate (1). “This step was of concern not because of the coupling chemistry, but because it meant the use of a palladium complex, which meant we would need to identify an effective method for reducing the residual palladium to acceptable levels,” notes O'Brien.

 The reaction proceeded as expected, but the researchers reported that the addition of L-cysteine was necessary for removal of some of the residual palladium. The desired TORC1/2 inhibitor was isolated as a 1:1 DMAc solvate and then dissolved in dimethylforamide (DMF) and treated with a silica-supported scavenger for further palladium removal. With this dual approach, the researchers successfully tackled the residual palladium issue (1). The final product was crystallized from DMF via addition of ethanol, which provided the desired thermodynamically stable polymorph with low levels of palladium (1).

References
1. F. Hicks, et. al., Org. Process Res. Dev. 17 (5) 829−837 (2013).
2. F. Voica, Baran Group Meeting Handout, 6/21/2008.