Discovery and Chemical Development of Suvorexant - A Dual Orexin

Dec 9, 2016 - This chapter outlines the discovery and chemical development of the dual orexin antagonist, suvorexant, which was recently approved as B...
2 downloads 12 Views 3MB Size
Chapter 1

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

Discovery and Chemical Development of Suvorexant - A Dual Orexin Antagonist for Sleep Disorder Debra J. Wallace,*,1 Ian Mangion,1 and Paul Coleman2 1Department

of Process Chemistry, Merck and Co, Rahway, New Jersey, 07065, United States 2Department of Medicinal Chemistry, Merck and Co, Westpoint, Pennsylvania, 19486, United States *E-mail: [email protected]. Phone1-732-594-3041.

This chapter outlines the discovery and chemical development of the dual orexin antagonist, suvorexant, which was recently approved as Belsomra® in the US and Japan for treatment of sleep disorders. The biological evidence that orexin signaling plays a central role in maintaining wakefulness and hence was a viable target for developing therapeutic entities is outlined, followed by details of the medicinal chemistry efforts to identify hit-molecules and ultimately the development candidate. The requirements of an ideal process chemistry synthesis, suitable for generating kilogram-, and ultimately manufacturing- scale quantities are explained. We then show how a staged approach to the synthetic development of suvorexant was employed focusing on an appropriate synthesis based on the scale of implementation and stage of development. We discuss how a number of alternative routes were evaluated before the final manufacturing route was chosen based on our guiding process chemistry principles. In particular, the importance of complete mechanistic understanding to allow for final reaction optimization for a key transformation is illustrated.

© 2016 American Chemical Society Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Introduction

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

Suvorexant (Belsomra®) is a first-in-class orexin receptor antagonist approved for the treatment of insomnia in the U.S. and Japan (1). Suvorexant 1 (Figure 1) has demonstrated efficacy in helping patients fall asleep and maintain sleep and is the first approved therapeutic agent that selectively blocks wake-promoting orexin signaling in the brain (2). This novel mechanism of action differs from other commonly used hypnotic agents that promote sleep through general suppression of central nervous system (CNS) activity. In this account, we describe the discovery and initial synthesis of suvorexant as well as the eventual conception and implementation of a large scale manufacturing route for this molecule.

Figure 1. Structure of suvorexant.

Role of Orexin Peptides Orexin peptides are central regulators of CNS arousal and vigilance and were discovered by two independent research teams in 1998 (3, 4). Orexin neuropeptides, orexin-A (OX-A) and orexin-B (OX-B) are generated in the hypothalamus from a common pre-pro-peptide precursor. OX-A and OX-B bind to and activate two transmembrane G-protein coupled receptors (OX1R and OX2R). Both OX1R and OX2R transactivate cells primarily through a Gq-mediated signaling leading to an increase in intracellular Ca2+ levels in neuronal cells (5–8). There are reports that both receptors can also alter intracellular cyclic adenosine monophosphate levels through a Gs/Gi/o-dependent mechanism. Orexin receptors are highly conserved across mammalian species with greater than 90% sequence identity between rodents and humans (9). Neurons that secrete excitatory orexins are restricted within the CNS and highly localized to the hypothalamus. The orexin-secreting neurons activate centers of the brain involved in arousal and wakefulness. The nerve terminals of orexin neurons project into the regions of the CNS that govern wakefulness and ascending arousal. Both OX2 and OX1 receptors are found in the laterodorsal tegmentum (LDT), pedunculopontine tegmentum (PPT), and dorsal raphe (DR) while the histaminergic tuberomammillary nucleus (TMN) and noradrenergic locus coeruleus (LC) have preferential expression of OX2R and OX1R, respectively (10–13). 2

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

A landmark event in the field of sleep research was the discovery that loss-of-function mutations in canine OX2R were the primary cause of inheritable narcolepsy and cataplexy in dogs (14). Following this discovery, it was learned that humans with narcolepsy, a sleep disorder marked by excessive daytime sleepiness and rapid transitions between sleep and wake, have significant loss of orexin-producing neurons resulting in depleted levels of excitatory orexins in the cerebrospinal fluid (CSF) (15, 16). Further corroboration for the role of orexins in maintaining wakefulness was established by the generation of rodent genetic strains. In mice, the knockout of the OX2R gene produces a narcolepsy-like phenotype similar to orexin peptide knockouts whereas OX1R knockouts present a milder phenotype with limited increases in sleep fragmentation (17, 18). Orexin peptides are secreted in the brain and their expression oscillates in a circadian manner peaking during the daytime hours in primates with levels falling during normal sleep (19, 20). Expression of orexins is thought to offset normal homeostatic sleep drive which builds during the day and provide stabilization of the wake state in the presence of increasing sleep drive. Aberrant signaling of this system during the normal sleep phase could drive undesired wakefulness and disrupted sleep. The discovery of this new neuropeptide signaling system, its anatomical restriction within the CNS, and its clear role in driving wakefulness provides a compelling rationale for the design of orexin receptor antagonists that selectively target this system. Orexin receptor antagonism represents a novel pharmacotherapy for the treatment of insomnia that has generated broad interest (21). Since a small molecule antagonist would presumably specifically target arousal-promoting signaling without attenuating other CNS functions, it is anticipated that a therapeutic molecule might have an improved tolerability profile and differentiated effects on sleep efficacy.

Discovery of Suvorexant We recognized the compelling evidence that orexin signaling plays a central role in maintaining wakefulness and, as part of a broader effort in developing therapeutic entities for treating wake/sleep dysregulation, we initiated a program to discover potent orexin antagonists with clinical utility. We embarked on a high-throughput (HTS) campaign of our sample collection to identify novel chemotypes that could function as antagonists of both OX1R and OX2R. We chose to focus on dual orexin antagonists (DORAs) due to the more robust sleep phenotype observed in dual receptor knockouts as well as the recognition that orexin tone is significantly reduced in human narcoleptics. This screening campaign was productive and identified multiple, diverse chemical series with promising attributes (Figure 2) (22). One of the hits of interest from our screen was diazepane amide 2 which bound with good affinity to both OX1R and OX2R and blocked orexin-A excitatory signaling in a cell based assay. This ligand while potent, lacked favorable drug-like properties. For example, compound 2 suffered from poor physicochemical properties (low solubility and high cLogP) and high rates of 3

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

hepatic metabolism. Our initial campaign focused on identifying molecules with reduced lipophilicity and improved metabolic stability. Extensive SAR exploration efforts were rewarded by the discovery of 3 which was found to maintain good receptor potency on both OX1R/OX2R and favorable brain penetration, while having reduced lipophilicity (23). Orexin antagonist 3 was active in vivo reducing spontaneous locomotor activity and promoting sleep in rats.

Figure 2. Hits from HTS.

While compound 3 displayed many favorable attributes, it had a poor pharmacokinetic profile with low oral bioavailability (F < 5%) and high clearance in the rat and dog. Significant reductions in plasma clearance could be achieved with further alteration of the diazepane structure and heteroaryl substitution to provide analogs such as 4 (24). Detailed metabolic studies on orexin antagonist 4 indicated a strong propensity for this molecule to generate electrophilic, reactive metabolites following metabolic activation. Specifically, we hypothesized that oxidative metabolism at methylene sites on the diazepane ring or on the heteroaryl moieties could generate reactive species that would be trapped by glutathione (GSH). Subsequent metabolic studies revealed that the fluoroquinazoline in 4 was a major site for bioactivation and trapping by glutathione (GSH) after incubation of 4 in microsomes treated with GSH. Replacement of the fluoroquinazoline in 4 with a chlorobenzoxazole suppressed the formation of undesired glutathione adducts and provided an overall favorable balance of potency, physicochemical properties, and pharmacokinetic profile. These efforts produced 1 (OX1R Ki = 50 nM; OX2R = 56 nM) which was subsequently advanced into preclinical development as MK-4305 and later named suvorexant. Suvorexant (1) maintains excellent potency against both OX1R and OX2R, has improved oral bioavailability and pharmacokinetics, and demonstrates potent in vivo activity in promoting sleep in preclinical species. As predicted by earlier conformational studies (25, 26), suvorexant adopts a U-shaped bioactive conformation when bound to OX2R and OX1R. When bound to the transmembrane helices in OX2R and OX1R, this conformation of suvorexant allows it to maximize key hydrogen bond and van der Waals contacts with critical residues within the ligand binding site thereby blocking neuropeptide binding and 4

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

preventing transmembrane helix motions required for activation. High resolution protein-ligand crystal structures have been solved for suvorexant bound to OX1R and OX2R (27, 28).

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

Medicinal Chemistry Synthesis of Suvorexant The synthesis of suvorexant began with heteroconjugate addition of N-Boc-1,2-diaminoethane 5 to methyl vinyl ketone followed by in situ trapping of 6 with benzyl chloroformate to provide the ketone adduct 7. Selective removal of the Boc group followed by intramolecular reductive amination and reprotection of the secondary amine as a Boc carbamate provides the racemic diazepane 8 in 38% overall yield from 5. At early stages of medicinal chemistry exploration, both the R- and S- antipodes were independently studied. Subsequently, it was found that the R-isomer had superior potency for orexin receptor binding, and this isomer of 8 could be resolved by chiral stationary phase HPLC to produce 9 after deprotection. Separately, the 2-(2H-1,2,3-triazol-2-yl)-5-methylbenzoic acid 10 was prepared via a microwave mediated amination of 2-iodo-4-methylbenzoic acid 11 with triazole/CuI to afford a 55:45 mixture of the 2-triazolyl and 1-triazolyl regioisomers 10 and 12, respectively in good yield. The two regioisomers were separated chromatographically from each other. Standard amide coupling of diazepane 9 with acid 10 provided intermediate 13 in good yield. Hydrogenolysis of the CBz group in 13 with Pd(OH)2 on carbon under 1 atm of hydrogen provided the unpurified diazepane 14 which could be subsequently treated with 2,5-dichloro-1,3-benzodioxazole 15 to afford suvorexant 1 (Scheme 1).

Scheme 1. Medicinal Chemistry Synthesis of Suvorexant (MK-4305) 5 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Based on suvorexant’s favorable profile, it was selected to enter preclinical development. Suvorexant showed robust dose-dependent promotion of sleep in rodents, dogs, and nonhuman primates (29). Further evaluation of this molecule showed that it was not genotoxic and was well-tolerated in preclinical toxicology assessments. Suvorexant had a safe cardiovascular profile in dogs and it showed good tolerability in longer duration toxicology studies in both rats and dogs. Given this promising preclinical profile, suvorexant was rapidly accelerated into clinical development.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

Initial Kilogram Scale Synthesis Once MK-4305 (suvorexant) was chosen as the compound to enter development, attention turned to evaluation of the chemical synthesis and in particular its suitability for larger scale implementation. The requirements for a manufacturing scale synthesis differ from those ideal for basic research purposes, where a modular and flexible approach is preferred allowing for many analogues to be prepared from common intermediates. However as scale increases other considerations become key as outlined below. Requirements of an ideal manufacturing route • • • • • • • • • • •

Control of impurity profile and physical properties Robust and reproducible reactions Appropriate isolations and purifications (no chromatography) Substrates and reagents must be available on large scale Cost of goods Maximize yield, minimize number of steps and operations, minimize costly reagents Safety concerns (for example, exotherms are more significant on scale) Minimal environmental impact (green chemistry targets) Freedom to operate Awareness of large-scale equipment and capabilities Regulatory considerations

Synthesis Assessment In reviewing the medicinal chemistry approach to MK-4305 (Scheme 1) a few areas were flagged as concerns for immediate or longer term use. 1.

2. 3.

Fairly lengthy linear synthesis with extensive use of protecting groups, in particular operations to afford the differentially protected and enantiomerically pure diamine 8 appeared inefficient Racemic synthesis relying on chromatographic separation of racemic 8 to afford a single enantiomer Final API isolated as a weakly crystalline form with low melting point 6

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

4.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

5.

Difficulty preventing over addition in reaction of Boc-ethylene diamine 5 with methyl vinyl ketone leading to an inseparable mixture of products (see Scheme 2). Triazole acid 10 formation was non-selective and involved chromatographic purification

However the team was aware that the resources and time required to address all of the challenges above might not be appropriate in early development, and hence initial efforts focused on bond forming improvements, but accepted that some form of enantiomer separation would still be used. The inspiration for a shorter route came from observing the undesired over addition product 16, which led us to suggest that the desired conjugate addition to methyl vinyl ketone should be possible starting from a secondary amine, and that secondary amine could include the desired benzoxazole western portion of the molecule, rather than a temporary protecting group. As such we postulated that addition of 17 to methyl vinyl ketone would afford the desired 18, avoiding the use of one of the protecting group, significantly shortening the route and giving higher yield in the conjugate addition step, Scheme 2. In turn 17 should be accessible by reaction of the previously used Boc-ethylenediamine 5 with an activated benzoxazole fragment.

Scheme 2. Proposed Shorter Synthesis for Reductive Amination Precursor 18

Synthesis of 22 With this proposed change the preparation of compound 18 was explored. At time of project initiation, a reliable source of an activated benzoxazole such as chloride 15 was not available so preparation from 2-amino-4-chlorophenol 19 was evaluated. Direct formation of the benzoxazole chloride 15 from 19 7 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

proved challenging, but the corresponding thiol 20 could be prepared by reaction with thiophosgene and isolated directly from the reaction mixture in high yield, (Scheme 3).

Scheme 3. Preparation of 18

In the medicinal chemistry route, the benzoxazole chloride 15 was prepared from thiol 20 by reaction with PCl3/POCl3, however the team rapidly identified oxalyl chloride as a milder alternative to these conditions. The initially formed chloride 15 was not isolated, but on treatment with triethylamine and Boc-ethylene diamine, the desired 17 was formed. Although an isolation of 17 was initially used, ultimately a through process was developed whereby after aqueous quench and solvent swap to acetonitrile, straightforward addition of DBU and methyl vinyl ketone allowed for excellent conversion to 18 in 75% isolated yield over the three step sequence from 20 (chlorination, amidation, conjugate addition) (30). The desired Boc-deprotection/intramolecular reductive amination sequence (Scheme 4) appeared to have much in common with the medicinal chemistry approach, however the presence of the benzoxazole in 18 vs the stable CBz group in 7 led to some challenges.

Scheme 4. Proposed Intramolecular Reductive Amination

Deprotection of the Boc group in 18 could be achieved with HCl or TFA, and the associated salts were stable as solids or in solution, but did not participate directly in the reductive amination reaction. Neutralization of the salts to the free amine 21 allowed for reductive amination promoted by sodium triacetoxyborohydride (STAB), however decomposition of both starting amine 8 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

and product 22 was seen on the time scales that would be associated with such large scale operations. Switching the deprotection reagent to methanesulfonic acid (MSA) allowed for the resulting MSA salt solution to be used directly in the reductive amination in modest yields. Alternatively the bis-MSA salt could be isolated by direct filtration, giving further purity upgrade and this salt was used in the reductive amination, again accompanied by formation of a number of impurities, including 23, 24 and 25 (31) (Figure 3).

Figure 3. Impurities formed in the reductive amination reaction.

Impurities 23 and 24 appear to result from hydrolysis of the benzoxazole moiety, suggesting that pH control in the mixture might be needed. Indeed, better results were obtained when the bis-MSA salt 26 was converted in situ to the monoMSA salt by addition of 1 equivalent of sodium acetate prior to addition of the STAB. In this way ring opened impurities were minimized and a near quantitative solution yield of amine racemic 22 was obtained (Scheme 5).

Scheme 5. Optimized Reductive Amination Procedure

With racemic diamine in hand, attention turned to separation of the enantiomers. In the medicinal chemistry route this was achieved by chiral column chromatography of the differentially protected diamine racemic 8. However for a target delivery of around five kilograms of final material, the time, solvent and stationary phase requirements to separate enantiomers of racemic 22 were clearly impractical and instead a classical resolution approach was pursued. A screen of readily available chiral acids and common solvents led to identification of a 1:1 L-dibenzoyltartaric acid salt from THF as the best lead, giving about 76% ee in the initial salt, and further work concentrated on this system. The two diastereomerically pure salts 27 and 28 were prepared from samples of each enantiomer of amine 22 (32) to study solubilities of the pure compounds. This confirmed that THF would be the optimum solvent for resolution based on solubility differences (Table 1). 9 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

Table 1. Solubility Data for 1:1 and 2:1 L-DBT Salts

However the process was further complicated by competing formation of a 2:1 amine:acid salt 29 leading to racemic material. Solubility of this hemi-salt 29 in THF was comparable to the desired 1:1 salt 27 and an addition protocol to minimize its formation was developed. Hence the amine racemic 22 was added slowly to 1.5 equivalents of L-di-benzoyltartaric acid thereby maintaining an excess of acid to suppress formation of the hemi-salt. In this way a 38% yield of salt 27 containing 74% ee amine (87% ds salt) was obtained on multi-kilogram scale. Based on the modest ee obtained in the initial salt formation, an upgrade of the optical purity of the isolated salt was sought via re-crystallization or re-slurry protocols. Surprisingly the previously employed THF did not offer any further improvement. This led to the proposal that the enantiopurity of the product was compromised by partial formation of the hemi-salt 29, rather than the diastereomeric 28. 1H NMR confirmed the presence of excess amine in these isolated salts giving further credence to this theory. Given the similar solubility of these two species in THF (3.25 mg/mL vs 4.1 mg/mL, Table 2) an efficient upgrade in this solvent appeared unlikely and attention switched to use of methanol which offered both reasonable absolute solubility, and a significant difference between the desired mono salt 27 and hemi-salt 29. Ultimately the use of IPAc as co-solvent with methanol attenuated the solubility and reduced losses to the filtrate. The final upgrade procedure was carried out by re-slurrying the solids in a 3:1 IPAc:MeOH mixture proceeding in 70% yield to afford 96% ee amine (Scheme 6). 10

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Scheme 6. Resolution and Upgrade of Amine 22

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

Triazole Acid Synthesis The medicinal chemistry route to triazole acid 10 involved a microwave promoted amination of iodide 11 with 1,2,3-triazole in NMP at 120 °C, using a diamine promoter which proceeded to give a 55:45 ratio of regioisomers 10 and 12. A lengthy extraction of the water soluble products and chromatographic separation using high solvent volumes was then required to afford the desired isomer in around 40% final yield (Scheme 7). For large scale processing both the use of microwave and the chromatographic purification needed to be addressed.

Scheme 7. Medicinal Chemistry Preparation of Triazole Acid 10 After an initial screen of all reaction variables, we found that complete reaction and higher selectivity could be achieved using an excess of 1,2,3-triazole, copper iodide, without the use of microwave or a diamine promoter, at 65 °C in THF/DMF with potassium carbonate as base. Under these conditions the reaction typically reached >98% conversion and produced an 81:19 ratio of regioisomers with 99.5% regioisomeric purity.

Scheme 8. Optimized Triazole Acid Preparation

Amide Coupling and API Form In preparation for the final amide bond formation, the L-DBT salt 27 was converted to the free amine form 22 with aqueous sodium hydroxide and following extraction with dichloromethane, was used as a solution in the subsequent step. Coupling of the two penultimate fragments was initially attempted using traditional coupling reagents (EDC, DCC, etc), however rate of reaction was very low, likely due to steric constraints. Additionally an isomeric impurity 31 was seen to form in the final compound which probably results from impurity 25 (see Figure 3). Ultimately it was found that coupling via the acid chloride gave a faster reaction rate and cleaner reaction profile. The acid chloride 32 was prepared by treatment of acid 10 with oxalyl chloride and DMF in dichloromethane, to which was added a solution of amine 22 and trimethylamine. This produced clean conversion to the final compound within half an hour. After an extractive work up and concentration of the CH2Cl2 solution, suvorexant 1 was isolated by the slow addition of heptane to the dichloromethane solution (Scheme 9) in 88% yield and 99.5% purity with 96.4% ee.

Scheme 9. Coupling To Give Crude MK-4305 12 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

MK-4305 isolated from the above reaction was shown to have the same crystal form (form 1) as that obtained in Medicinal Chemistry. This form was confirmed to be anhydrous and weakly crystalline, but had a low, broad melting point of 123 °C (+/- 3°C), and poor physical properties leading to slow filtration. High throughput polymorph screens revealed one new form (form 2) which was obtained by heating MK-4305 in water at close to 100 °C. This provided material which was more crystalline as assessed by the sharper peaks in the XRPD spectrum (Figure 4).

Figure 4. XRPD spectrum of the original (form 1) and new form (form 2). The melting point of the new form 2 was 154 °C. It was non-hygroscopic, absorbing only 0.06 wt% water at 95% RH and 25 °C, stable after aging overnight at 30, 40, 50 and 60 °C, and TGA confirmed it to be anhydrous. This form was considered appropriate for further development. Somewhat surprisingly stability studies showed conversion of form 2 to the less crystalline form 1 after stirring at room temperature in most solvents, with water being an exception. On large scale, the form turnover was achieved by slurrying the initially isolated form 1 in water at just below 100 °C, and after cooling, form 2 solid was isolated by filtration in 97% yield, 99.6% chemical purity and 97.4% ee.

Summary of Initial GMP Delivery Using the chemistry as described above, a rapid delivery of over 3.0 kg of material in around 12% overall yield from the commercial chloro-aminophenol starting material 19 was achieved (Scheme 10). The process generated material with very high chemical purity (99.6%) and moderate enantiomeric purity (97.4%) which was sufficient to support the program through initial toxicology studies, early formulation development and into PhI clinical studies (33). Key successes of the approach were as follows: 13 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

1.

2.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

3. 4.

Use of the benzoxazole fragment in place of the temporary protecting group on the ethylene diamine gave a much shorter and more convergent synthesis avoiding a number of protecting group manipulations A classical resolution of racemic 22 replaced the time consuming chiral separation A new and highly crystalline form of MK-4305 was discovered Higher selectivity and more scalable conditions for the triazole acid synthesis, and a method to remove the undesired regioisomer via salt formation rather than chromatography.

Scheme 10. Summary of First GMP Delivery of MK-4305

PhII Synthetic Approach As clinical data became available from PhI studies, it became clear that larger amounts of API would be needed to support the project moving forward into PhII and beyond. In reviewing the approach used for the initial kilogram scale delivery, the team felt that the synthesis offered a convergent approach and the bond forming reactions were efficient and high yielding. For the most part only minor modifications to these reactions would be necessary. However the classical resolution to afford a single enantiomer of 22 was low yielding, afforded only moderate ee material even after an upgrade step, and required kinetic control for good results due to facile formation of the thermodynamically favored 2:1 29. This would not be appropriate for API deliveries in excess of a few kilograms. As such, preparation of diamine 22 in high enantiomeric purity and high yield became the main objective for the next stage of development with a view to supporting API deliveries of 10-100 kg scale. 14

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Asymmetric Reductive Amination

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

While other synthetic approaches to a single enantiomer of 22 could be envisaged, given the success of the intramolecular reductive amination from the first delivery and the high yielding process to prepare the precursor, our efforts focused on an asymmetric version of this reaction. The value of high throughput experimentation for such an endeavor quickly became apparent as a number of reagent combinations were evaluated, as outlined in Scheme 11.

Scheme 11. Screening for Asymmetric Version of the Reductive Amination

Low ee’s and/or conversions were obtained using chiral phosphoric acid reagents and chirally modified sodium borohydride derivatives. Some encouraging selectivities were seen with rhodium- or iridium-catalyzed hydrogenations, but the most promising lead was using a ruthenium-catalyzed transfer hydrogenation promoted by complex 33. After further optimization of the ligand system, we were able to achieve 90% conversion to give 22 in 85% ee. Thiswould be appropriate for the next stage of development, especially as it was anticipated an upgrade in enantiopurity should be possible via salt formation with the previously used resolving agent di-benzoyl tartaric acid. The reductive amination proceeded smoothly on multi-kilogram scale however unexpected challenges were encountered during the salt formation to upgrade the enantiomeric purity. Use of L-di-benzoyl tartaric acid led to material with very poor filtration properties. As with the previous delivery, upgrade to 15 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

better than around 97% ee proved elusive. Additionally impurity 25 was seen to form over time during the upgrade, which was further exacerbated by slow filtration rates. Screening for other salt forms led to the discovery that a chiral acid was not needed. Formation of the acetate salt 34 gave more efficient upgrade of enantiomeric purity with better filtration properties. However, the longer cycle times associated with large-scale processing once again led to significant formation (>10%) of the isomeric 25. Further operations were then required to remove this impurity and improve the ee, which in turn led to further degradation, reducing the overall yield. Despite these limitations, this process allowed a 62% yield of the acetate salt in 98.5% ee on 10-50 kg scale, supporting API deliveries to initiate PhII studies (Scheme 12).

Scheme 12. Initial Asymmetric Reduction Route

While the discovery of the asymmetric reductive amination was a breakthrough step and allowed for rapid generation of larger amounts of material on the 10-50 kg scale, further optimization would be required before implementation of 100s kilogram scale. In particular, conversion, enantiomeric excess, stability and purity upgrade of the product all required attention. Efforts initially focused on optimization of the asymmetric transformation. Extensive screening confirmed the triisopropyl ligand 37 and p-cymene aryl group as optimum, however changes to the base, formic acid equivalents, reduction in reaction temperature and change of solvent to dichloromethane all led to an improvement in both conversion and enantiomeric excess (Table 2). Under these optimal conditions, the reaction proceeded in 98% yield to give diamine 22 of 94% ee. With a higher yielding process in place, attention turned to the best way to stabilize and isolate the product. The impurity formed during the previous process was confirmed to be the amine isomer 25, presumably generated by transient formation of the bicyclic compound 38 (Scheme 13), and was common to an impurity previously seen in the achiral version of the process (see Figure 3). 16

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

Table 2. Optimization of Reductive Amination Conditions

Scheme 13. Isomerization of Amine 22 17 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

To understand conditions under which this isomerization would be problematic, a stability study of amine 22 was carried out under conditions and pH which could feasibly be encountered during the reductive amination reaction, work-up, and subsequent amide coupling. Somewhat surprisingly the amine showed good stability under strongly acidic or basic conditions, however at neutral pH, and particularly in weak acids (with pKa in range 3-6) extensive isomerization was seen giving what appeared to be an equilibrium ratio of 22:25 60:40, (Table 3).

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

Table 3. Stability of Amine 22 at Various pHs Temperature Time

10 °C 8 h 48 h

pH1

1

45 °C 8 h 48 h

60 °C 8 h 48 h

Amount of 25 (based on HPLC) (%)

12

0.3

0.2

0.2

0.2

0.2

1.7

10

0.4

0.6

1.6

3.6

1.6

4.4

8

0.5

1.2

3.5

12

8.2

16

5

4.5

11.3

37

41

41

41

2

0.5

0.7

1.5

6.9

5.1

21

target pH obtained by addition of dilute HCl or NaOH.

These results explain the rapid isomerization during the previous salt formation procedures. Both acetic acid and di-benzoyltartaric acid have pKa’s in the range which the desired product is the least stable. Additionally the isomerized product 31 seen in early EDC promoted couplings (see Scheme 9) is thought to arise from prior isomerization of the amine under the mildly acidic pH conditions in that reaction, which was mitigated by switching to the acid chloride prepared under a more acidic regime. As such to ensure good yields, operating conditions in the pH range 3-6 would need to be avoided and this was addressed in two ways: 1) Modification of work-up conditions: Once the target conversion was obtained, the reaction was quenched with aqueous sodium hydroxide solution to ensure strongly basic conditions were maintained during the subsequent extractions and solvent reduction. 2) Alternative salts for ee upgrade and isolation: To ensure pH remained below 2 during salt formation, stronger acids were evaluated. In the presence of HCl, degradation was seen to be minimal. Appropriate ee upgrade was not seen in many common extraction solvents, however a mixture of DMAc:toluene produced the HCl salt 39 in excellent enantioselectivity after a single isolation, with minimal loss to the liquors. 18

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

With the above changes to the reaction conditions and isolation procedure, the asymmetric transfer hydrogenation now proceeded in almost quantitative conversion, and 94% ee as measured in the reaction mixture. After standard extractive work-up, HCl salt formation, and a single isolation allowed a 90% yield of the amine HCl salt in essentially complete enantiomeric purity (Scheme 14).

Scheme 14. Optimized Reductive Amination and Isolation Conditions

Early and End-Game Steps Modifications As MK-4305 continued to move through PhII trials and successively larger synthetic campaigns were required, improvements to the other synthetic steps were also sought to address robustness, yields and environmental (Green Chemistry) concerns. To avoid the use of thiophosgene on large scale, an alternative synthesis of thiol 20 was developed using potassium ethylxanthate as reagent. Initially pyridine was employed as solvent for this transformation, but ultimately the less toxic ethanol was suitable, providing a 96% yield. In the next sequence, dichloromethane was replaced with THF for the chlorination of thiol 20, K2CO3 was used in place of triethylamine for the amination to give 17, and DBU was replaced with catalytic sodium hydroxide to promote the conjugate addition reaction. These changes had a beneficial impact on the amount and type of waste generated for preparation of 18 and moreover improved the yield and robustness (Scheme 15).

Scheme 15. Improved Synthesis of 18 In reviewing the final coupling and form conversion steps for MK-4305, a number of issues from the initial kilogram scale delivery required further development, as outlined below. 19 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

1. 2. 3.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

4.

High volumes of dichloromethane used for coupling reaction. An off-line salt break and extraction with dichloromethane was needed to prepare the free amine for coupling. A form turnover step is required. Although in some cases a “purification step” is desirable, in this case essentially no purity improvement is seen due to the very low solubility in water. Lack of understanding of the API phase, and conversion to the undesired form 1 at room temperature in most organic solvents.

Initial efforts focused on optimization of the coupling reaction. Although coupling via the acid chloride was high yielding, both the acid chloride 32 and free amine 22 needed preparation in separate operations, with dichloromethane being used to efficiently extract the amine. Moreover despite screening for other suitable solvents, the very thick nature of the slurry seen during the reaction, likely due to formation of triethylamine hydrochloride, necessitated use of high volumes of dichloromethane as reaction solvent to retain mobility and good reaction rates. An alternative approach was explored using a bi-phasic system to allow for reaction of the acid chloride with the amine HCl salt 39 in the presence of an aqueous inorganic phase. After some experimentation, optimum conditions involved preparation of the chloride with oxalyl chloride in isopropyl acetate (IPAc) with catalytic DMF leading to a homogeneous solution. This was added to a two phase mixture of the amine salt 39 (no separate salt break needed) in more IPAc and aqueous potassium carbonate with vigorous stirring. Coupling to form the desired final compound 1 was essentially complete by the end of the acid chloride addition and after aqueous work up a 98% solution yield was obtained. At this stage crude isolation and form turnover in water at 100 °C as previously described would afford the desired form 2, however to minimize operations direct isolation of the desired form was now evaluated. To aid this, further characterization and understanding of the relationship between form 1 and 2 was undertaken. Although form 2 has the higher melting point and thus was initially assumed to be more stable, conversion to form 1 was seen at room temperature in a range of solvents, especially if form 1 seed was added. Further studies indicated that the two forms were enantiotropic with a transition temperature between 35 and 40 °C and form 1 was more stable at 25 °C. Despite the potential challenges of isolating the less stable form at room temperature, the favorable morphology of form 2 led to its selection for longer term use. In the solid state no conversion of form 2 to form 1 was seen during storage at 25 °C/60% RH for up to 36 months. With the new knowledge of the relationship between form 1 and 2, generation of form 2 would need to occur above 40 °C. Isolation either above that temperature, or in a solvent in which conversion to form 1 is very slow was sought. The previously used process involved heating in water at 100 °C which converted all material to form 2. Although the slurry was then cooled to below the transition temperature for isolation, the very low solubility of final material in water prevented any significant conversion to form 1 under the time cycles employed. Similar behavior was seen in other solvents where the intrinsic 20

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

solubility of MK-4305 was low, such as heptane and MTBE, whereas in a 1:1 IPA:water mixture, form 2 converts to form 1 below 40 °C and re-converts to form 2 above 40 °C.

Figure 5. Stability of Form 2 in 90:10 and 85:15 Heptane:IPAc based on XRPD.

Figure 6. Stability of Form 2 in 90:10 and 85:15 Heptane:IPAc based on melting point. To allow for direct isolation of form 2 from the final IPAc stream, solubility and form stability studies of various IPAc:anti-solvent mixtures were carried out. This led to identification of 85:15 n-heptane:IPAc mixture as suitable to give 21 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

high recovery, good impurity rejection, and form stability. In this solvent ratio, conversion of the desired form 2 to form 1 was essentially negligible even at room temperature in the absence of any form I seed as assessed by both XRP and melting point (Figure 5 and 6). With 10% of form 1 remaining in the initial mixture, further change was only observed after 48 h at room temperature and hence a robust operating window was assured. Ultimately an isolation process was designed whereby the final IPAc layer containing product was reduced in volume above 40 °C, leading to crystallization of form 2 upon further addition of heptane. Filtration was also carried out at above 40 °C, leading to a 96% isolated yield of >99.6% purity final API (Scheme 16).

Scheme 16. Optimized coupling and isolation of MK-4305 Form 2

Summary of PhII Synthetic Approach With the improvements detailed in the above sections for the asymmetric hydrogenation, early steps, and end game process, the suvorexant synthesis became suitable for implementation on several hundred kilogram scale and was reliable and reproducible to support the project through PhII clinical trials (Scheme 17) (34).

Scheme 17. Summary of PhII Synthetic Approach to Suvorexant 22 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Commercial Synthesis Development To this point a robust and readily scalable process had been developed that was amenable to the synthesis of hundreds of kilograms of MK-4305. However, the team now faced the question of whether this synthesis was appropriate for execution on the commercial scale, in which metric ton quantities might be required. Although the discovery of an asymmetric reductive amination had greatly improved overall yield and reduced waste, there remained two considerations for that reaction that merited further evaluation of the process:

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

1.

2.

The asymmetric reduction used a high loading (3 mol %) of a ruthenium catalyst that impacts cost of goods and long term sustainability of the process This reaction is run preferentially in dichloromethane, which is disfavored as a process solvent due to its environmental impact.

Therefore the team embarked on two parallel efforts to solve the perceived liabilities of the existing synthesis. First we will discuss efforts to revise the synthesis to remove the metal-catalyzed asymmetric reductive amination in favor of an enzymatic approach, which entailed an entirely new sequence to the chiral diazepine core. Second we will review investigations into the mechanism of the existing reductive amination, in order to understand what factors might be limiting to catalyst loading and solvent selection. Development of an Enzymatic Approach The key structural feature of MK-4305 is the core chiral diazepane ring 22, which had been assembled using the aforementioned ruthenium-catalyzed asymmetric reductive amination. This method achieved high levels of enantioselectivity (94% ee), but required the use of a transition metal catalyst and dichloromethane as solvent, both of which we hoped to eliminate to lessen the environmental impact of the process. Therefore an alternative bond disconnection was envisioned taking advantage of biocatalytic transamination technology (Figure 7) (35–37). Specifically, an asymmetric transamination of a ketone 40 that bears a suitable leaving group could set the stage for a cascade transamination (TA)/medium ring annulation, completing the diazepane system (38). By this time there were already some reports demonstrating the utility of enantioselective biocatalytic reactions in pharmaceutical applications, in particular an improved synthesis of sitagliptin using this same enabling transamination reaction (Scheme 18), which inspired us in our thinking (39). The hope was that this enzymatic approach might offer advantages both in reaction performance and environmental sustainability. Furthermore, if the leaving group for the annulation could be derived from activation of an alcohol, then compound 40 could be constructed from inexpensive starting materials including ethanolamine (41) and the previously employed methyl vinyl ketone. This would then offer some advantage over the existing process, where the analogous intermediate was made using N-Boc ethylene diamine 5, a more expensive raw material. 23

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

Figure 7. Proposed cascade transamination/ring annulation strategy.

Scheme 18. Enantioselective Transamination Approach to Sitagliptin To effectively compete with the ruthenium-catalyzed asymmetric reductive amination process, we needed a rapid and scalable approach to ketone 40 that would also allow us to evaluate several possible leaving groups to optimize the ring annulation, presuming a suitable transamination could be achieved. We were able to apply a strategy similar to the existing process, in which an acid catalyzed condensation of commercially available phenol 19 with trimethyl orthoformate yielded benzoxazole 42 (Scheme 19) (40, 41). The reaction stream of 42 was then partially distilled to remove methanol and used directly in a subsequent lithiation/ bromination sequence to furnish bromide 43. This activated benzoxazole was then treated with ethanolamine to provide alcohol 44 without the use of chromatography or intermediate isolations in 90% overall yield from phenol 19 (42–46).

Scheme 19. Chromatography-Free Synthesis of Ketone 46 The next step to the targeted ketone substrate required an aza-Michael addition to methyl vinyl ketone. While this was similar to an analogous reaction for the existing process, there is one critical distinction – whereas in the existing process 24 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

the conjugate addition can effectively occur only on one nitrogen due to the use of a Boc protecting group on the diamine backbone. Here chemoselectivity between nitrogen and oxygen addition was targeted in the absence of protecting groups. An additional complexity is that the Michael addition itself is reversible, and while chemoselectivity may be high at low conversions, it would invariably worsen as the reaction progressed. Additionally, under basic conditions the benzoxazole heterocycle itself is subject to hydrolysis over time. To overcome this problem, a wide variety of bases and solvents were surveyed in a high-throughput multi-well format, including organic and inorganic bases with varying pKas. These initial screens identified alkali carbonates in polar solvents (e.g. DMF) as being near optimal displaying >10:1 regioselectivity. However results on scale were capricious, likely due to the low solubility of the carbonates under the conditions studied, with most of the activity ascribed to a dilute soluble population of base. We soon realized the variability of results tracked roughly with residual water in our laboratory solvents, and that the actual active base was likely to be likely trace hydroxide from adventitious water. We evaluated concentrated sodium hydroxide as a catalyst and found that it effectively promoted the aza-Michael in 95% yield and 42:1 selectivity (N to O) as a 1 mol % additive. Gratifyingly, this aza-Michael addition could be telescoped with functionalization of free alcohol 45 as the corresponding mesylate (46) without any intermediate workup or isolation. This was a principal benefit of the low loading of base in the aza-Michael addition, such that only a small excess of mesyl chloride was required to consume the sodium hydroxide present to enable a through process. These conditions were then applied to a one step synthesis of mesylate 46 from alcohol 44, following the aza-Michael with mesylation and crystallization of the product in 81% yield (Scheme 19). We now have an efficient method to access sulfonate esters, such as mesylate 46, from inexpensive starting materials. This put us in a position to evaluate the crucial transamination reaction with an opportunity to match or improve the economics of the existing process. However, we came to quickly realize that a highly selective transamination of the ketone (40 to 47, Scheme 20) is only the first of a number of outcomes that might occur in the course of a tandem transamination and ring annulation sequence. We had seen from earlier research on the asymmetric reductive amination process that the benzoxazole is labile to nucleophilic attack and postulated that the transamination event might initiate an intramolecular ring opening (47 to 48). These same studies had demonstrated the propensity of diazepane 22 to undergo an unusual rearrangement of the diazepane ring (22 to 25), which in our studies occurred in a pH range of 3 to 12. We reasoned that at a low pH (< 3) diazepane 22 predominantly exists in a doubly protonated form, and that above pH 12 it exists as a free base. As such, extreme pH regimes suppress intermediate 49, which arises from a singly protonated species, and therefore the undesired rearrangement of the diazepane ring is also suppressed under these conditions. However, based on literature precedent and our own experiences, we expected that the enzymatic transamination would perform poorly outside of near-neutral pH ranges, due in part to instability of the enzyme itself. 25

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

In addition we expected that the sulfonate ester leaving group in the annulation reaction would be susceptible to hydrolysis under basic reaction conditions, however non-acidic conditions would be required for the annulation step to proceed. A successful process would have to balance these competing factors.

Scheme 20. Competing Pathways during the TA/Ring Annulation Sequence The transamination of ketone 46 was evaluated with a panel of commercially available enzymes, though unfortunately most provided no significant conversion. One exception was ATA-117 (Codexis), which gave a modest 8% conversion of the ketone. However, this encouraged us to evaluate the (R)-selective transaminase developed for the aforementioned sitagliptin process, as this was an evolved variant of ATA-117 developed specifically to accommodate a sterically larger substrate than the natural enzyme (47–49). Indeed, this enzyme provided both good conversion and high enantioselectivity in the transamination of 46 (Table 4, entry 1, >99% ee). To our delight, we also established proof of concept that the desired ring annulation to form 22 could occur in situ without further manipulations. However, we observed significant amounts of guanidine 48 (nonproductive path a depicted in Scheme 20) under a variety of conditions, suggesting that competition between SN2 attack on the sulfonate ester and ring opening of the benzoxazole was finely poised. In addition we also observed the diazepane rearrangement product 25, which was expected in light of observations previously discussed, but which imposed limits on the yield of the reaction. Finally, yield was also impacted by hydrolysis of the sulfonate ester activating group. To mitigate these side reactions, a range of sulfonate esters was examined under identical conditions (Table 4, entries 2-5), but in the end mesylate 46 was still superior in overall profile. Interestingly, the tosylate variant displayed better selectivity against the formation of guanidine 48 (X = OTs) (entry 2), but nevertheless slower reaction kinetics and poorer substrate stability provided diminished yield. Replacing the sulfonate ester with an alkyl chloride eliminated the substrate stability issue and indeed provided clean reactivity in the transamination (entry 6). But only the undesired guanidine 48 (X = Cl) arose as a product because the chloride was insufficiently reactive to compete with the benzoxazole ring opening pathway. 26

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Modulating the pH of the reaction to encompass more basic conditions did lessen the formation of 48 and 25 (entries 7-10), but a practical upper limit of pH 10 was encountered due to stability issues for both the sulfonate ester and indeed the enzyme itself (entries 9-10). The final conditions chosen in this study leveraged both a moderate pH and a slow addition protocol for 46 to minimize substrate decomposition pathways, ultimately delivering diazepane 22 in 71% yield and greater than 99% ee (entry 11) (50).

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

Table 4. Optimization of the Transamination Protocol

With this unusual tandem transamination/medium ring annulation approach to produce the core diazepane in hand, we now had a new standard by which to judge the existing supply route. That is, with the use of halogenated solvents or heavy metal catalysts eliminated in this alternative route, a business case for a more sustainable large-scale commercial-supply based on the ruthenium-catalyzed

27 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

asymmetric reductive amination would require greener solvent choice and a lower loading of the catalyst. We will next discuss our efforts to achieve these goals.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

Revisiting the Asymmetric Hydrogenation: Mechanism and Optimization We have earlier described efforts to identify and improve the rutheniumcatalyzed asymmetric reductive amination (Scheme 21), which had proven competent for furnishing clinical supply during development. However, the focus for a potential commercial route using this approach relied on lowering the amount of the ruthenium catalyst required under the existing conditions, perhaps by improving reaction rate or by eliminating unproductive reaction pathways of the catalyst. In so doing, we hoped to then demonstrate applicability of the optimized chemistry in non-halogenated solvents. To accomplish this, we began a series of studies meant to examine the kinetic behavior of the reaction network (33).

Scheme 21. Reductive Amination Approach

A general mechanism for the reductive amination based on analogy to literature is shown in Figure 8. Briefly, we thought of the reaction as occurring in three stages, from insertion of formic acid into catalyst 33 to yield Ru-formate 49 to extrusion of carbon dioxide forming Ru-H species 50, and ultimately reduction of the substrate (21) regenerating the catalyst. We hoped that a kinetic analysis would reveal the rate order of substrate and reagents while providing insight into what reaction features might play into the rate-limiting step of this process. Two observations became immediately apparent upon conducting the reaction under standard conditions: • •

the overall reaction displayed first-order kinetics. the reaction appeared to be zero-order in the substrate (21).

This would require that regeneration of the ruthenium hydride were rate limiting rather than the reduction of the substrate. There was precedence using related tethered ligand derivatives of 33 in which it was shown that the reaction may be zero-order in substrate at low conversions when the substrate is at a high concentration, but then become first-order at higher conversions when the reduction of the substrate becomes rate-limiting rather than catalyst regeneration (51). In this system, reaction rate increases with a decreasing initial concentration of substrate, a phenomenon that is consistent with product inhibition. 28 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

Figure 8. Initially proposed catalytic cycle for asymmetric transfer hydrogenation.

However, it was quickly shown that this was not the case, and in general secondary amines had no impact on the performance of the reaction. The kinetic phenomena might also be explainable by reversibility in the transfer hydrogenation; however enantiopure product resubjected to reaction conditions did not racemize. Finally, it was considered that the extrusion of carbon dioxide (compound 49 to compound 50) might be reversible. This was in contrast to the general understanding from the literature at that time (52). However by adding carbon dioxide into the reaction system, a substantial loss in reactivity could be induced seemingly validating our last standing hypothesis. The apparent equilibrium between formate 49 and hydride 50 was studied further by 1H NMR spectroscopy, in which it was revealed that in the absence of substrate they quickly formed in a 95:5 ratio (49:50) that did not change even upon extended aging. Again, this was counterintuitive as it had been anticipated that the entropic gain realized by extrusion of carbon dioxide would render formation of 50 irreversible in practical conditions, but empirical results continued to suggest otherwise. Conclusive evidence for the equilibrium process was established when the known RuH(p-cymene)(TsDPEN) (51) (53) was generated and put in a CD2Cl2 solution that was then exposed to CO2 (Scheme 22). 1H NMR immediately showed resonances fully consistent with the Ru-formate complex (52) together with complete consumption of the Ru-hydride species. 29

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

Scheme 22. Conversion of Ru-Hydride 51 to Ru-Formate 52 under CO2.

If we then incorporate this finding into the initially proposed catalytic cycle, we arrive at a modified scheme shown below (Figure 9). Here the concentration of CO2 will influence the rate of the reverse reaction indicated as k-2, and indeed this was verified experimentally. As such, the team implemented a simple engineering solution in order to improve reaction rate: by purging CO2 continuously with nitrogen throughout the process, regeneration of the catalyst is no longer limiting at higher conversions.

Figure 9. Modified catalytic cycle reflecting CO2 catalyst inhibition.

30 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

With the improved reaction rate a number of other reaction benefits were realized (Scheme 23):-

1. 2.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

3.

Reduction of the catalyst loading to 2 mol %, reducing cost and material consumption. An alternative solvent system of acetonitrile and toluene rather than dichloromethane can be used. The improved fundamental reactivity allowed us to select solvents that were slightly inferior with regard to overall reaction rate but substantially superior with regard to environmental impact. No requirement to isolated the MSA salt of 21, instead Boc removal from 18 was carried out using HCl in the desired reaction solvent, toluene, and a solution of 21 taken directly into the hydrogenation reaction after phase separation.

Scheme 23. Final Hydrogenation Conditions

With an improved asymmetric transfer hydrogenation in hand that addressed both catalyst loading and solvent choice concerns, we were now prepared to select this approach as a commercial route to MK-4305. Also, the observations that arose from this mechanistic study appear more broadly applicable to this class of Noyori-type formic acid-driven transfer hydrogenations (54), exemplifying the unexpected value that can come from a detailed reaction study.

Summary of Process Development Efforts Herein we have described a series of efforts to design and optimize syntheses of MK-4305, in order to reduce waste and cost of goods, eliminate chromatography and reduce isolations, increase yield and process robustness, and ultimately identify the best route to supply the API commercially with minimal environmental impact (Scheme 24). 31 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

Scheme 24. Final Manufacturing Synthesis of MK-4305

As MK-4305 advanced from Medicinal Chemistry to early clinical trials, then to larger clinical studies, and finally to commercialization, the challenges and basic requirements of a suitable synthesis evolved. However, as we continued to study the chemistry, our knowledge evolved with the synthesis and we were able to deliver improvements that overcame the needs of each phase of development. Critical milestones in the development of the commercial process include the following: 1. 2. 3. 4. 5.

6.

Replacing chiral chromatography from the Medicinal Chemistry synthesis with a classical resolution. Streamlining step count and protecting groups by using the benzoxazole heterocycle as the starting point for a new synthesis of the diazepine ring. Doubling yields of the triazole acid while eliminating chromatographic isolation. Installing a direct isolation of the desired form of MK-4305 in the final step, removing an intermediate form turnover operation. Discovering and optimizing an unprecedented catalytic asymmetric reductive amination, and continually improving the process through mechanistic investigation. Developing alternative synthetic approaches, including an asymmetric, enzymatic transamination to deliver the chiral diazepine.

The path to development of the final manufacturing process involved several unexpected discoveries, unusual diversions, and dedicated efforts to continue to improve the chemistry to its maximum even when a workable solution was already in hand. It is the defining challenge to the process chemist to devise a synthesis that will withstand the test of time; that will continue to serve as the benchmark for efficiency and cost even as chemical technology evolves and new chemists in different parts of the world consider alternative approaches (55). A manufacturing process is never perfect, and we still wonder whether this reaction could have 32 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

been more selective or if that reaction could give a better yield. However, these questions are just one facet of the joy of chemical synthesis, and we hope the journey of the MK-4305 process will inspire others who engage with the demands of contemporary chemistry.

Acknowledgments

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

We would like to thank our many colleagues in the Merck Research Labs and Merck Manufacturing Division who contributed tirelessly to this work, in addition to those colleagues named in references below.

References 1. 2.

3.

4.

5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

Belsomra® (suvorexant) label and prescribing information: https:// www.merck.com/product/usa/pi_circulars/b/belsomra/belsomra_pi.pdf. Herring, W. J.; Connor, K. M.; Ivgy-May, N.; Snyder, E.; Liu, K.; Snavely, D. B.; Krystal, A. D.; Walsh, J. K.; Benca, R. M.; Rosenberg, R.; Sangal, R. B.; Budd, K.; Hutzelmann, J.; Leibensperger, H.; Froman, S.; Lines, C.; Roth, T.; Michelson, D. Biol. Psych. 2016, 79, 136–148. De Lecea, L.; Kilduff, T. S.; Peyron, C.; Gao, X.-B.; Foye, P. E.; Danielson, P. E.; Fukuhara, C.; Battenberg, E. L. F.; Gautvik, V. T.; Bartlett, F. S., II; Frankel, W. N.; Van Den Pol, A. N.; Bloom, F. E.; Gautvik, K. M.; Sutcliffe, J. G. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 322–327. Sakurai, T.; Amemiya, A.; Ishii, M.; Matsuzaki, I.; Chemelli, R.; Tanaka, H.; Williams, S. C.; Richardson, J. A.; Kozlowski, G. P.; Wilson, S.; Arch, J. R. S.; Buckingham, R. E.; Haynes, A. C.; Carr, S. A.; Annan, R. S.; McNulty, D. E.; Liu, W.; Terrett, J. A.; Elshourbagy, N. A.; Bergsma, D. J.; Yanagisawa Cell 1998, 92, 573–585. Holmqvist, T.; Akerman, K. E.; Kukkonen, J. P. FEBS Lett. 2002, 526, 11–14. Kukkonen, J. P.; Akerman, K. E. NeuroReport 2001, 12, 2017–2020. Lund, P. E.; Shariatmadari, R.; Uustare, A.; Detheux, M.; Parmentier, M.; Kukkonen, J. P.; Akerman, K. E. J Biol. Chem. 2000, 275, 30806–30812. Zhu, Y.; Miwa, Y.; Yamanaka, A.; Yada, T.; Shibahara, M.; Abe, Y.; Sakurai, T.; Goto, K. J Pharmacol. Sci. 2003, 92, 259–266. Wong, K. K.; Ng, S. Y.; Lee, L. T.; Ng, H. K.; Chow, B. K. Comp Endocrinol. 2011, 171, 124–130. Marcus, J. N.; Aschkenasi, C. J.; Lee, C. E.; Chemelli, R. M.; Saper, C. B.; Yanagisawa, M.; Elmquist, J. K. J. Comp. Neurol. 2001, 435, 6–25. Trivedi, P.; Yu, H.; MacNeil, D. J.; Van der Ploeg, L. H. T.; Guan, X. M. Febs Lett. 1998, 438, 71–75. Mochizuki, T.; Scammell, T. E. Curr. Biol. 2003, 13, R563–R564. Torrealba, F.; Yanagisawa, M.; Saper, C. B. Neuroscience 2003, 119, 1033–1044. Lin, L.; Faraco, J.; Li, R.; Kadotani, H.; Rogers, W.; Lin, X.; Qiu, X.; de Jong, P. J.; Nishino, S.; Mignot, E. Cell 1999, 98, 365–376. 33

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

15. Nishino, S.; Ripley, B.; Overeem, S.; Lammers, G. J.; Mignot, E. Lancet 2000, 355, 39–40. 16. Peyron, C.; Faraco, J.; Rogers, W.; Ripley, B.; Overeem, S.; Charnay, Y.; Nevsimalova, S.; Aldrich, M.; Reynolds, D.; Albin, R.; Li, R.; Hungs, M.; Pedrazzoli, M.; Padigaru, M.; Kucherlapati, M.; Fan, J.; Maki, R.; Lammers, G. J.; Bouras, C.; Kucherlapati, R.; Nishino, S.; Mignot, E. Nat. Med. 2000, 6, 991–997. 17. Chemelli, R. M.; Willie, J. T.; Sinton, C. M.; Elmquist, J. K.; Scammell, T.; Lee, C.; Richardson, J. A.; Williams, S. C.; Xiong, Y.; Kisanuki, Y.; Fitch, T. E.; Nakazato, M.; Hammer, R. E.; Saper, C. B.; Yanagisawa, M. Cell 1999, 98, 437–451. 18. Willie, J. T.; Chemelli, R. M.; Sinston, C. M.; Tokita, H.; Williams, S. C.; Kisanuki, Y. Y.; Marcus, J. N.; Lee, C.; Elmquist, J. K.; Kohlmeier, K. A.; Leonard, C. S.; Richardson, J. A.; Hammer, R. E.; Yanagisawa, M. Neuron 2003, 38, 715–730. 19. Fujiki, N.; Yoshida, Y.; Ripley, B.; Honda, K.; Mignot, E.; Nishino, S. NeuroReport 2001, 12, 993–997. 20. Gotter, A. L.; Winrow, C. J.; Brunner, J.; Garson, S. L.; Fox, S. V.; Binns, J.; Harrell, C. M.; Cui, D.; Yee, K. L.; Stiteler, M.; Stevens, J.; Savitz, A.; Tannenbaum, P. L.; Tye, S. J.; McDonald, T.; Yao, L.; Kuduk, S. D.; Uslaner, J.; Coleman, P. J.; Renger, J. J. BMC Neurosci. 2013, 14, 90–106. 21. Roecker, A. J.; Cox, C. D.; Coleman, P. J. J. Med. Chem. 2015, 59, 504–530. 22. Coleman, Paul J.; Cox, C. D.; Roecker, A. J. Curr. Top. Med. Chem. 2011, 11, 696–725. 23. Whitman, D. B.; Cox, C. D.; Breslin, M. J.; Brashear, K. M.; Schreier, J. D.; Bogusky, M. J.; Bednar, R. A.; Lemaire, W.; Bruno, J. G.; Hartman, G. D.; Reiss, D. R.; Harrell, C. M.; Kraus, R. L.; Li, Y.; Garson, S. L.; Doran, S. M.; Prueksaritanont, T.; Li, C.; Winrow, C. J.; Koblan, K. S.; Renger, J. J.; Coleman, P. J. ChemMedChem. 2009, 4, 1069–1074. 24. Cox, C. D.; Breslin, M. J.; Whitman, D. B.; Schreier, J. D.; McGaughey, G. B.; Bogusky, M. J.; Roecker, A. J.; Mercer, S. P.; Bednar, R. A.; Lemaire, W.; Bruno, J. G.; Reiss, D. R.; Harrell, C. M.; Murphy, K. L.; Garson, S. L.; Doran, S. M.; Prueksaritanont, T.; Anderson, W. B.; Tang, C.; Roller, S.; Cabalu, T. D.; Cui, D.; Hartman, G. D.; Young, S. D.; Koblan, K. S.; Winrow, C. J.; Renger, J. J.; Coleman, P. J. J. Med. Chem. 2010, 53, 5320–5332. 25. Cox, C. D.; McGaughey, G. B.; Bogusky, M. J.; Whitman, D. B.; Ball, R. G.; Winrow, C. J.; Renger, J. J.; Coleman, P. J. Bioorg. Med. Chem. Lett. 2009, 19, 2997–3001. 26. McGaughey, G.; Bayly, C. L.; Cox, C. D.; Schreier, J. S.; Breslin, M. J.; Pitzenberger, S.; Ball, R.; Coleman, P. J. J. Comput. Aided. Mol. Design 2014, 28, 5–12. 27. Yin, J.; Mobarec, J. C.; Kolb, P.; Rosenbaum, D. M. Nature 2015, 519, 247. 28. Yin, L.; Babaoglu, K.; Brautigam, C.; Clark, L.; Shao, Z.; Scheuermann, T.; Harrell, C. M.; Gotter, A. L.; Roecker, A. J.; Winrow, C. D.; Renger, J. J.; Coleman, P. J.; Rosenbaum, D. M. Nat. Struct. Mol. Biol. 2016, 23, 293–299. 34

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

29. Winrow, C. J.; Gotter, A. L.; Cox, C. D.; Doran, S. M.; Tannenbaum, P. L.; Breslin, M. J.; Garson, S. L.; Fox, S. V.; Harrell, C. M.; Stevens, J.; Reiss, D. R.; Cui, D.; Coleman, P. J.; Renger, J. J. J. Neurogenet. 2011, 25, 52–61. 30. Stewart, G. W.; Baxter, C. A.; Cleator, E.; Sheen, F. J. J. Org. Chem. 2009, 74, 3229–3231. 31. We will discuss a possible mechanism for the formation of impurity 25 in a later part of this chapter. 32. The enantiomerically pure samples were prepared by small scale chiral separation of racemic 22. 33. Baxter, C. A.; Cleator, E.; Brands, K. M. J.; Edwards, J. S.; Reamer, R. A.; Sheen, F. J.; Stewart, G. W.; Strotman, N. A.; Wallace, D. J. Org. Process Res. Dev. 2011, 15, 367–375. 34. Strotman, N. A.; Baxter, C. A.; Brands, K. M. J.; Cleator, E.; Krska, S. W.; Reamer, R. A.; Wallace, D. J.; Wright, T. J. J. Am. Chem. Soc. 2011, 133, 8362–8371. 35. For a review on biocatalytic transformations see: De Wildeman, S. M. A.; Sonke, T.; Schoemaker, H. E.; May, O. Acc. Chem. Res. 2007, 40, 1260–1266. 36. Moore, J. C.; Pollard, D. J.; Kosjek, B.; Devine, P. N. Acc. Chem. Res. 2007, 40, 1412–1419. 37. Reetz, M. T. Ang. Chem., Int. Ed. 2011, 50, 138–174. 38. Mangion, I. K.; Sherry, B. D.; Yin, J.; Fleitz, F. Org. Lett. 2012, 14, 3458–3461. 39. Savile, C. K.; Janey, J. M.; Mundorff, E. C.; Moore, J. C.; Tam, S.; Jarvis, W. R.; Colbeck, J. C.; Krebber, A.; Fleitz, F. J.; Brands, J.; Devine, P. N.; Huisman, G. W.; Hughes, G. J. Science 2010, 329, 305–309. 40. For a related heterocyclic condensation see: Vechorkin, O.; Hirt, N.; Hu, X. Org. Lett. 2010, 12, 3567–3569. 41. For a related heterocyclic condensation see: Cioffi, C. L.; Lansing, J. J.; Yüksel, H. J. Org. Chem. 2010, 75, 7942–7945. 42. The synthesis of aminobenzoxazoles from benzoxazoles has also been reported via direct oxidative methods, see: Guo, S.; Qian, B.; Xie, Y.; Xia, C.; Huang, H. Org. Lett. 2011, 13, 522–525. 43. Froehr, T.; Sindlinger, C. P.; Kloeckner, U.; Finkbeiner, P.; Nachtsheim, B. J. Org. Lett. 2011, 13, 3754–3757. 44. Wertz, S.; Kodama, S.; Studer, A. Angew. Chem., Int. Ed.. 2011, 50, 11511–11515. 45. Li, Y.; Xie, Y.; Zhang, R.; Jin, K.; Wang, X.; Duan, C. J. Org. Chem. 2011, 76, 5444–5449. 46. Lamani, M.; Prabhu, K. R. J. Org. Chem. 2011, 76, 7938–7944. 47. For other applications of (R)-selective transaminases see: Truppo, M. D.; Turner, N. J.; Rozzell, D. Chem. Commun. 2009, 2127–2129. 48. Koszelewski, D.; Clay, D.; Rozzell, D.; Kroutil, W. Eur. J. Org. Chem. 2009, 2289–2292. 49. Koszelewski, D.; Tauber, K.; Faber, K.; Kroutil, W. Trends Biotechnol 2010, 28, 324–325. 35

Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.

Downloaded by 80.82.78.170 on December 12, 2016 | http://pubs.acs.org Publication Date (Web): December 9, 2016 | doi: 10.1021/bk-2016-1239.ch001

50. The mass balance consists of 45 from hydrolysis of the sulfonate and oligomers derived from intermolecular alkylation of amine 47. 51. Cheung, F. K.; Lin, C.; Minissi, F.; Criville, A. L.; Graham, M. A.; Fox, D. J.; Wills, M. Org. Lett. 2007, 9, 4659–4662. 52. Fujii, A.; Hashiguchi, S.; Uematsu, N.; Ikariya, T.; Noyori, R. J. Am. Chem. Soc. 1996, 118, 2521–2522. 53. Noyori, R.; Hashiguchi, S. Acc. Chem. Res. 1997, 30, 97–102. 54. Applying the CO2 purging approach to the transfer hydrogenation of acetophenone using this catalyst system produced a nearly 10-fold rate increase for a sealed reaction. 55. Zhang, T. Y. Chem. Rev. 2006, 106, 2583–2595.

36 Abdel-Magid et al.; Comprehensive Accounts of Pharmaceutical Research and Development: From Discovery to Late-Stage ... ACS Symposium Series; American Chemical Society: Washington, DC, 2016.