Development of a Manufacturing Process for an HCV Protease

Dec 14, 2014 - Vadims Dvornikovs, Young C. Jung, Valdas Jurkauskas, Elaine C. Lee, Adam R. Looker, Hoa Luong,. Theodore A. Martinot, David B. Miller, ...
0 downloads 0 Views 776KB Size
Article pubs.acs.org/OPRD

Development of a Manufacturing Process for an HCV Protease Inhibitor Candidate Molecule Benjamin J. Littler,* Michael Aizenberg, Narendra B. Ambhaikar, Todd A. Blythe, Timothy T. Curran, Vadims Dvornikovs, Young C. Jung, Valdas Jurkauskas, Elaine C. Lee, Adam R. Looker, Hoa Luong, Theodore A. Martinot, David B. Miller, Bobbianna J. Neubert-Langille, Pieter A. Otten, Peter J. Rose, and Piero L. Ruggiero Downloaded via EASTERN KENTUCKY UNIV on January 29, 2019 at 09:49:39 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

Chemical Development, Vertex Pharmaceuticals Incorporated, 50 Northern Avenue, Boston, Massachusetts 02210, United States, and 11010 Torreyana Road, San Diego, California 92121, United States ABSTRACT: The scale-up of a prototype HCV protease inhibitor (1) from gram scale in the laboratory to kilogram scale in the pilot plant is described. Key features of the optimization included the synthesis of bulk quantities of exomethylene proline intermediate 6, separation of the diastereomers of spirocycle 2 without chromatography, isolation of the precursor to 1 to purge byproducts that might raise genotoxic structural alerts, and purification of an amorphous drug substance via a crystalline acetic acid solvate.



INTRODUCTION It is estimated that about 150 million people worldwide are chronically infected with the hepatitis C virus (HCV), with 3−4 million new infections every year. More than 350 000 people are thought to die every year from hepatitis C-related diseases. Following initial HCV infection, approximately 80% of people exhibit no symptoms, but eventually approximately 75−85% of patients progress to a chronic infected state. An estimated 60− 70% of these chronically infected people will eventually develop chronic liver diseases, such as cirrhosis or liver cancer. In 2011, two HCV protease inhibitors, telaprevir and boceprevir, were approved for treatment of hepatitis C in combination with interferon and ribavirin, affording treatment protocols that are more readily tolerated and have higher cure rates.1 Compound 1 is a prototype HCV protease inhibitor that required rapid scale-up to enable the completion of toxicology and early-phase clinical studies. As outlined in Figure 1, the pseudopeptide structure of compound 1 can be disconnected to spirocycle 2 and commercially available cyclohexaneacetic acid (3), N-Boc-tert-leucine (4), and chiral amine 5. Spirocycle 2 can be derived from exomethylene proline 6 and chloro-oxime 7 via a 3 + 2 cycloaddition. Our colleagues in Medicinal Chemistry had prepared up to a few hundred grams of compound 1 using the chemistry shown in Scheme 1.2 Three major issues were identified as needing to be resolved before kilogram quantities of material could be produced. First, an economical synthesis of exomethylene proline 6 was required. Second, we needed to replace the chromatography that was used to separate the 4:1 mixture of diastereomers produced in the 3 + 2 cycloaddition used to form spirocycle 2. Finally, we needed to avoid the chromatography that was used to purify compound 1. In this paper, we first describe the processes developed to make the first few kilograms of material in the kilo lab, which enabled the project to move rapidly to the clinic. We then describe the fit-for-purpose modifications that were made to the synthesis in readiness for the pilot-plant manufacture of batches © 2014 American Chemical Society

Figure 1. Retrosynthesis of compound 1.

of over 50 kg of compound 1. The whole synthesis development program was completed under tight timelines required to support the wider compound development program schedule.3



RESULTS AND DISCUSSION Synthesis of Exomethylene Proline 6. As outlined in Scheme 2, the synthesis of exomethylene proline 6 on a 10 g scale without chromatography has been reported. The synthesis starts with oxidation of commercially available N-Boc-4hydroxyproline (8) with sodium periodate and catalytic ruthenium dioxide to afford intermediate ketone 9. This is followed by a Wittig reaction of 9 with the ylide derived from Received: June 26, 2014 Published: December 14, 2014 270

DOI: 10.1021/op500210w Org. Process Res. Dev. 2015, 19, 270−283

Article

Organic Process Research & Development Scheme 1. Discovery synthesis of compound 1

Scheme 2. Literature preparation of exomethylene proline 6

Scheme 3. Pilot-plant synthesis of exomethylene proline 6

methyltriphenylphosphonium bromide, with subsequent extraction of the desired product into a basic aqueous phase and washing with Et2O to remove neutral organic byproducts such as triphenylphosphine oxide. Acidification followed by extraction and concentration was reported to afford 6 as a foam.4 Oxidation Process Development. We wanted to avoid using ruthenium or other metals to oxidize 8 to 9, so a TEMPO/ bleach oxidation was examined. No reaction occurred using the conventional TEMPO/bleach reaction system containing catalytic KBr and buffered with NaHCO3,5 but ketone 9 formed when KBr and NaHCO3 were omitted from the procedure (Scheme 3). We postulate that the free carboxylate group in 8 acts as a buffer that allows bleach to survive for use in the TEMPO oxidative cycle. Catalyst stability studies were performed to determine the addition time and temperature range for the reaction. In these experiments, a total of 1.0 equiv of bleach was added in seven smaller portions over 24 h at fixed temperatures between 0 and 20 °C in an automated reactor while the reaction and jacket

temperatures were monitored. When performed below 10 °C, the reaction went to completion and the heat flow (reaction temperature minus the jacket temperature) was relatively constant for each portion added over a 24 h period, indicating that the catalyst was not being significantly deactivated over the course of the experiment. When the reaction was performed at 20 °C, there was incomplete conversion to ketone 9 and the height of the heat flow spikes decreased in intensity over the course of the experiment, suggesting that the catalyst was decomposing. The data also showed that there was no heat flow between additions, so the process temperature could be readily controlled by varying the rate of the bleach addition to maintain the reaction temperature between 0 and 5 °C. An expedient procedure for isolating ketone 9 was achieved by adding the completed reaction mixture to a solution of aqueous NaHSO4 to crystallize the product. It was observed that the solid product must be filtered and washed with water as quickly as possible to obtain the highest product yield and quality. Prolonged exposure of ketone 9 to acid leads to 271

DOI: 10.1021/op500210w Org. Process Res. Dev. 2015, 19, 270−283

Article

Organic Process Research & Development

remove water from the EtOAc layer after the aqueous phase was removed. The largest single batch performed in the pilot plant converted 26.0 kg of ketone 9 into 30.9 kg of exomethylene proline 6·t-BuNH2 in 91% yield (Scheme 3). The t-BuNH2 salt was broken before the next step by forming a slurry in methyl tert-butyl ether (MTBE), washing with citric acid, and azeotropic drying. Synthesis of Chloro-oxime 7. The discovery procedure for the alkylation of phenol 12 used excess dimethyl sulfate with K2CO3 in acetone (Scheme 1). Methyl tosylate was selected as the alkylating agent for early development (Scheme 4) because its consumption was readily measured using a

decomposition. Many batches of this oxidation were completed with isolated yields typically between 75 and 90%, with the lower yields observed on a larger scale in the pilot plant as a result of the extended filtration times. In the largest single batch performed, 35.0 kg of N-Boc-4-hydroxyproline 8 was converted to 26.7 kg of ketone 9 in 77% yield (Scheme 3). Installation of the Exocyclic Methylene Group. We initially considered forming the carboxylic acid ester before the Wittig reaction to reduce the amount of base required, but we decided to retain the carboxylic acid group so that we could use an aqueous workup to remove the triphenylphosphine oxide byproduct without chromatography. The decision to isolate the carboxylic acid was further supported by the fact that our preliminary development studies found that crystalline 6 could be obtained from a slurry in i-PrOAc, whereas all attempts to crystallize the tert-butyl (10) or methyl (11) carboxylic acid ester derivatives of olefin 6 were unsuccessful. The reported procedures for preparing olefin 6 from ketone 9 used both methyltriphenylphosphonium bromide and base in at least 2-fold excess,4,6 so modification was needed to improve throughput before the process would be suitable for kilogramscale production. The major change that we made was to reduce the amount of methyltriphenylphosphonium bromide to 1.0 equiv. We reasoned that only an excess of base was required because the first equivalent of base would be needed to remove the carboxylic acid proton so that the second equivalent of base could generate the ylide. Gratifyingly, complete conversion of ketone 9 to olefin 6 was achieved in a trial experiment when 2.0 equiv of KOtBu in 2-MeTHF was added to 1.0 equiv of methyltriphenylphosphonium bromide followed by treatment of the resulting solution with 1.0 equiv of ketone 9. For the initial kilo-lab runs, ketone 9 was added to the reaction mixture as a solid under a blanket of nitrogen to maximize the process throughput because of the relatively low solubility of 9 in 2MeTHF (73 g/L). For the pilot plant, ketone 9 was added as a concentrated solution in N-methylpyrrolidone (NMP) (500 g/ L). At the end of the reaction, the mixture was quenched with water, and the carboxylate anion partitioned into the basic aqueous phase. A screen of solvents showed that CH2Cl2 was the most effective solvent at selectively removing both triphenylphosphine oxide and excess methyltriphenylphosphonium salts from the aqueous phase containing the desired product. CH2Cl2 also minimized the processing time because it is heavier than water, which allowed multiple washes to be rapidly performed and drained from a single vessel. Olefin 6 was then isolated by acidification of the aqueous layer with citric acid followed by extraction with EtOAc. In the initial kilolab procedure, the EtOAc was removed using a rotary evaporator to afford an amorphous solid that was crystallized from only 1.5 volumes of toluene to minimize yield losses. This isolation process was used to make many batches of 6 as a palebrown crystalline solid in the kilo lab in about 70−80% yield. In the pilot plant it would not have been practical to concentrate the mixture to dryness and then triturate with only 1.5 volumes of solvent, so we screened a series of inexpensive, lowmolecular-weight amines to find a crystalline salt that could be readily isolated. Morpholine and tert-butylamine were identified as possible crystallization partners. Morpholine was rejected because the presence of ≥1% water dissolved in EtOAc dramatically increased the solubility of 6·morpholine, whereas 6·t-BuNH2 crystallized directly from the workup solution in excellent yield and purity without requiring additional drying to

Scheme 4. Kilo-lab synthesis of chloro-oxime 7

simple HPLC method. Small-scale studies showed that any unreacted phenol 12 was easily removed by basic aqueous washes, so methyl tosylate was used as the limiting reagent (0.95 equiv). The reaction was best performed in MeCN at 60−70 °C for 16 h since the reaction was faster than in acetone at reflux (56 °C). The completed reaction in MeCN was quenched with water (10 L/kg of 12), and a phase split occurred, allowing the excess phenol to be simply removed in the aqueous layer. In the kilo lab, methyl ether 13 was obtained as an oil after a solvent exchange from MeCN to MTBE using a rotary evaporator followed by an aqueous workup and removal of the solvent. Conversion of 13 to oxime 14 was then accomplished by adding NH2OH·HCl and NaOAc in a mixture of THF, EtOH, and water. After 16 h of stirring at ambient temperature, the volatile organic solvents were removed using a rotary evaporator, and then oxime 14 was extracted using toluene. Oxime 14 was obtained as an oil containing 5−10 wt % residual toluene. The oil partially solidified on standing, but all attempts to develop a true crystallization of 14 were unsuccessful. Fourteen kilo-lab batches converted 55.9 kg of phenol 12 into 69.3 kg of oxime 14. Chloro-oxime 7 was initially prepared by a procedure in which a solution of N-chlorosuccinimide (NCS) in MeCN was added to a heated mixture of oxime 14 in MeCN while the temperature was maintained between 45 and 50 °C. At the end of the reaction, the mixture was concentrated to remove MeCN and then worked up in i-PrOAc, followed by the addition of nheptane to afford chloro-oxime 7 as a solid. This process was scaled up to prepare 1 kg of chloro-oxime 7, but differential scanning calorimetry (DSC) showed that the solid was thermally unstable above 120 °C. As a result, we decided that it was safer not to isolate the solid but to telescope a solution of chloro-oxime 7 in DMF directly into the 3 + 2 cycloaddition (Scheme 4). In the kilo lab it was possible to add solid NCS 272

DOI: 10.1021/op500210w Org. Process Res. Dev. 2015, 19, 270−283

Article

Organic Process Research & Development

order to better control the exotherm, we added a catalytic amount of concentrated hydrochloric acid since this has been reported to activate NCS.7 An initial small-scale reaction went to >95% conversion within 5 min of adding NCS to a mixture of oxime 14 and 0.1 equiv of concentrated HCl in MTBE and DMF at 20 °C. Adding the solution of oxime 14 in MTBE to a mixture of NCS and concentrated HCl in DMF was equally effective and allowed the exotherm to be readily controlled. This approach looked like it might be suitable for the pilot plant but was not pursued further because an accelerating-rate calorimetry experiment on the product solution of 14 in DMF showed an exothermic event starting at 82 °C.8 Ultimately, for the pilot plant we chose to implement an inherently safer process in which the chloro-oxime was generated using bleach in the presence of olefin 11 so that the chloro-oxime would react immediately to form cycloadduct 15 (see Scheme 7).9 Heat flow measurements showed a readily controlled exotherm during the addition of bleach to an organic solution of oxime 14 and exomethylene proline 11 and no significant exotherm during the rest of the reaction. Studies indicated that 1.5 equiv of NaOCl could be used to complete the reaction, but because of the variability of the concentration in aqueous bleach solutions, typically 2.0 equiv of NaOCl was used to be certain that enough base was present to convert the chloro-oxime to the reactive dipolar nitrile oxide. Development of the Cycloaddition To Synthesize Spirocycle 2. A key modification that the development team made at the outset of the project was to change the protecting group on the carboxylic acid in the cycloaddition from a tertbutyl ester to a methyl ester. The medicinal chemistry team had used the tert-butyl ester protecting group in an attempt to improve the stereoselectivity of the cycloaddition by maximizing the steric bulk of the chiral element on the proline ring (see Scheme 1). Unfortunately, extensive studies by our colleagues had shown that the stereoselectivity of the cycloaddition with the tert-butyl ester protecting group was approximately 80:20 across a wide range of aryl substituents and reaction conditions, which suggested that the steric influence of the ester moiety on the cycloaddition selectivity was minimal. The extra bulk of the tert-butyl ester did improve the chromatographic separation of the diasteroemeric product mixture, but this added an extra step later in the sequence in order to convert the tert-butyl ester to a more readily removed methyl ester for the endgame. We initially prepared exomethylene proline methyl ester 11 under relatively neutral conditions using a combination of 1-ethyl-3-(3(dimethylamino)propyl)carbodiimide hydrochloride (EDCI), DMAP, and MeOH in CH2Cl2 (Scheme 6).10 A screen of solvents for the cycloaddition was performed, and we were

into the reactor in small portions under a blanket of N2 while maintaining the solution of oxime 14 in 4 volumes of DMF at 33−38 °C. This allowed over 20 batches of chloro-oxime 7 in DMF to be produced on up to a 4 kg scale in the kilo lab. For the pilot-plant campaign, the process to make oxime 14 was further simplified to remove the need to concentrate the mixture to dryness between steps (Scheme 5). Intermediate 13 Scheme 5. Pilot-plant synthesis of chloro-oxime 7

was not isolated in the plant: once the alkylation was complete, aqueous NaOH and NH2OH·HCl were added to the reaction mixture and the heating was continued. Once oxime 14 had completely formed, the mixture was cooled to ambient temperature and extracted with three portions of MTBE to afford a solution of oxime 14 that was telescoped directly into the cycloaddition. With this procedure, a single batch of 80 kg of phenol 12 was converted to 94 kg of oxime 14 in 98% yield and 93% area under the curve (AUC) by HPLC. The mass of oxime 14 was estimated by quantitative HPLC analysis of the MTBE solution. Initial attempts to use the kilo-lab procedure to make chlorooxime 7 in the pilot plant demonstrated that maintaining the temperature of the reaction mixture between 33 and 38 °C during the addition of NCS proved to be much more difficult in the pilot plant because of the slower response that varying the jacket temperature had on the temperature of the reaction mixture and the increased difficulty of adding the solid NCS at a rate to ensure a consistent exotherm. We did not want the temperature to exceed 40 °C because of the thermal instability of chloro-oxime 7, but we found that the reaction rate decreased dramatically when the reaction temperature fell below 33 °C. Addition of a significant quantity of NCS with the reaction mixture below 33 °C resulted in a strong and rapid exotherm once the mixture was heated again above 35 °C. In

Scheme 6. Initial approach to isolate spirocycle 2 as a single diastereomer

273

DOI: 10.1021/op500210w Org. Process Res. Dev. 2015, 19, 270−283

Article

Organic Process Research & Development happy to find that the stereoselectivity remained at approximately 80:20 regardless of the solvent or the ester protecting group, so DMF was initially selected as the reaction solvent for the cycloaddition because it gave the fewest byproducts (13% AUC by HPLC). We used the methyl ester protecting group throughout the rest of the program since this was readily removed later in the synthesis using aqueous base. We also stayed with the combination of EDCI, DMAP, and MeOH to prepare the methyl ester throughout the program because the water-soluble byproducts allowed for a simple and fast workup. In the kilo lab, the esterification was performed in CH2Cl2 and methyl ester 11 was concentrated to an oil on the rotary evaporator, but for the pilot-plant campaign, we formed 11 in MTBE and telescoped the product solution directly into the 3 + 2 cycloaddition after a water wash and azeotropic drying. As shown in Scheme 6, our initial approach to separate the spirocyclic diastereomers was to first remove the Boc group from 15, then to isolate the free-base mixture of diastereomers, and finally to form a salt that we hoped would preferentially crystallize the desired diastereomer. The Boc group was removed using methanesulfonic acid in THF,11 and after an aqueous workup, the mixture of free-base spirocyclic diastereomers was treated with 12 different acids. This screen showed that maleic acid in EtOAc or i-PrOAc was able to upgrade the diastereomeric ratio from 80:20 in solution to >98:2 in the solid. This procedure was successfully demonstrated on a 200 g scale, but further scale-up to 2 kg surprisingly afforded only a 94:6 mixture. Subsequent studies showed that the >98:2 selectivity was obtained only when the solid was rapidly filtered after addition of the maleic acid. Prolonged stirring led to a thermodynamic mixture that could not be raised above 95:5 without an unacceptable yield loss. Additional equilibrium solubility studies indicated that the maleic acid salt of 2 has higher thermodynamic solubility than its diastereomeric counterpart in all of the organic solvents screened, so a new salt with a favorable thermodynamic solubility profile was required. During the scale-up to 200 g it became clear that the aqueous workup of the deprotection procedure using methanesulfonic acid led to significant losses of the desired product. Treatment of the initial Boc-protected cycloadduct product 15 with anhydrous 4 M HCl in 1,4-dioxane was therefore examined as an alternative method to remove the Boc group in order to avoid an aqueous workup. We expected to isolate a solid with an 80:20 mixture of diastereomers, but remarkably we found that salt 2·HCl crystallized directly from the reaction mixture with a diastereomeric ratio of >99:1! HPLC analysis showed that almost all of the reaction byproducts, including the undesired diastereomer, remained in the mother liquor and that almost all of the desired diastereomer 2 had crystallized. At this point, further studies on identifying other salts to separate the diastereomers were abandoned. Approximately 30 batches of the cycloaddition, deprotection, and crystallization sequence were performed in the kilo lab to convert olefin 11 and oxime 14 to supply over 100 kg of spirocycle 2·HCl. In all but two cases, 2·HCl crystallized directly from the reaction mixture with a diastereomeric ratio of >98:2. No clear cause could be identified as to why two batches crystallized with a lower diastereomer ratio of 89:11, but a recrystallization process using MeOH and MTBE upgraded the diastereomeric ratio of these two batches to >99:1 in high yield. This recrystallization process was ultimately incorporated into

every kilo-lab batch of 2·HCl because the initial colorless to offwhite solid sometimes led to poor phase cuts and low recoveries in subsequent steps despite having >98% AUC purity. The recrystallized solid was typically colorless with >99.5% AUC purity by HPLC and gave clean phase splits in all subsequent steps. The isolated yield of 2·HCl was typically 50− 55% from olefin 11 and oxime 14, which is excellent considering that the cycloaddition yield of both diastereomers was estimated by quantitative HPLC to be 75%. The maximum possible yield of 2·HCl from 11 and 14 was approximately 60% given that there was an 80:20 mixture of diastereomers. This demonstrates just how effective the crystallization was in affording 2·HCl in excellent yield and purity from a complex mixture. For the pilot-plant campaign, the in situ chloro-oxime generation and subsequent cycloaddition reaction could be readily run in MTBE (18 h reaction time, 78:22 diastereomer ratio) or CH2Cl2 (8 h reaction time, 75:25 diastereomer ratio) at ambient temperature (Scheme 7). MTBE was selected as the Scheme 7. Pilot-plant synthesis of spirocycle 2·HCl

preferred solvent because the sequence could be telescoped through all of the steps from exomethylene proline 6·t-BuNH2 and oxime 14 to cycloadduct 2·HCl without needing to swap solvents. Another key advantage of this procedure over the kilolab process was that the reaction mixture after the cycloaddition was much less colored and no polymeric materials were generated, eliminating the need to recrystallize 2·HCl from MeOH and MTBE. The largest batch performed converted 33.2 kg of methyl ester olefin 11 and 25.9 kg of oxime 14 to 22.5 kg of 2·HCl in 46% yield from 11 with 99% AUC purity. The slightly lower molar yield for the pilot-plant process was outweighed by the improved throughput, simplicity of 274

DOI: 10.1021/op500210w Org. Process Res. Dev. 2015, 19, 270−283

Article

Organic Process Research & Development operation, and reduced waste along with the fact that it was inherently safer. Optimization of the Conversion of Spirocycle 2 to Compound 1. The discovery methods for converting spirocycle 2 to compound 1 proceeded in excellent yield using standard peptide coupling chemistry (see Scheme 1), so the initial goals of the development effort were to improve the process throughput and to remove the chromatography that had been used to improve the purity of compound 1. Of particular concern was the fact that no neat crystalline form of the final product had been obtained despite extensive screening, so we expected that robust processes and intermediate purifications would be needed to ensure a consistent purity profile of drug substance 1.12 Coupling of Spirocycle 2 with N-Boc-tert-leucine (4) Followed by Boc Deprotection To Afford 17·HCl. The coupling of spirocycle 2·HCl with N-Boc-tert-leucine (4) was readily accomplished in the kilo lab using EDCI and HOBt with N-methylmorpholine (NMM) as the base and EtOAc as the solvent (Scheme 8).13 Spirocycle 2 was used directly as its

Table 1. Product ratio of the desired product 17 to diketopiperazine impurity 18 formed under different aqueous base workup procedures base

17:18 ratio (HPLC area %)

K2CO3 (25%) NaHCO3 (sat) NaOH (1 M) 1 N NH3 in MeOH

75:25 82:18 43:57 0:100

Scheme 8. Pilot-plant synthesis of amine 17·HCl from spirocycle 2·HCl

Figure 2. Impurities generated in the conversion of spirocycle 2 to amine 17·HCl.

The original medicinal-chemistry Boc removal procedure directly precipitated the hydrochloride salt of 17 from the reaction mixture using 6 N HCl in dioxane, and diketopiperazine 18 was not observed as a major process impurity, so we reverted back to this direct-drop process to avoid the aqueous workup. The coupling and deprotection reactions were telescoped in the kilo lab using a rotary evaporator to swap the solvent containing intermediate 16 from EtOAc to 1,4dioxane. Commercially available 4 N HCl in dioxane was then added, and the mixture was stirred at 25 °C for 12 h, during which time solid started to precipitate. Heptane was added at the end of the reaction to complete the precipitation of 17· HCl. This process was used to prepare a number of batches in the kilo lab, and typically the product was obtained as a colorless powder in approximately 90% yield on a kilogram scale with a purity of >98% AUC. However, the following issues were identified during the kilo-lab runs that meant that a new process was needed for the pilot plant: 1. One batch of the coupling in the kilo lab starting with 2.5 kg of spirocycle 2·HCl became so thick that it could not be mixed completely and eventually broke the pin holding the PTFE paddle onto the glass stirring rod. This led to the formation of many new byproducts, the most abundant being impurity 19 containing an additional tertleucine residue (Figure 2). The new byproducts were mostly purged after the Boc deprotection in the crystallization of 17·HCl, but the yield of the isolated

hydrochloride salt by adding an extra equivalent of NMM base (3.0 equiv total). Robustness testing showed that the reaction could be run at 20, 30, 40, or 50 °C with no observed major side reactions, so the reaction was typically run in the kilo lab by adding NMM as the last component at 30 °C and stirring the resulting suspension for 12 h at 35 °C. After a standard aqueous workup with acid and base, the solvent was removed on the rotovap in the kilo lab. This afforded intermediate 16 as an oil that was directly used in the Boc deprotection step. We initially planned to remove the Boc group from 16 using standard conditions of methanesulfonic acid in THF11 followed by neutralization with base and an extractive workup to afford 17. The deprotection was complete within 30 min at 40 °C, with no major side products detected by HPLC. The neutralization was studied with a range of bases, but as summarized in Table 1, significant quantities of diketopiperazine impurity 18 (Figure 2) formed during the workup under all of the conditions studied. 275

DOI: 10.1021/op500210w Org. Process Res. Dev. 2015, 19, 270−283

Article

Organic Process Research & Development

water. The use of 6 N HCl in methanol, ethanol, or isopropanol was rejected because these solvents could not be fully removed from the product 17·HCl and could react in the subsequent coupling reaction to generate ester byproducts. The use of 1.5 N HCl in EtOAc was unsuccessful because it gave an unacceptably slow reaction at 20−25 °C and 6% diketopiperazine 18 formed when the reaction was heated to 45−50 °C. For comparison, the N-Boc deprotection in 1,4-dioxane employing a 4 N solution of HCl in 1,4-dioxane gave very high conversion at 20−25 °C. The rate of reaction was increased at 45−50 °C. Notably, no diketopiperazine 18 formed even at 45−50 °C. Furthermore, very little degradation was observed when the deprotection reaction in 1,4-dioxane was stirred for 5 days at 50 °C. Thus, anhydrous 1,4-dioxane and 4 N HCl solution in 1,4-dioxane were selected for the conversion of 16 and isolation of 17·HCl in the pilot plant. Notably spike-and-purge studies showed that both diketopiperazine 18 and amino acid 21 (Figure 2) were purged in the crystallization of 17·HCl. The product 17·HCl was dried on a filter at ambient temperature by simply applying simultaneous vacuum and nitrogen purge. Drying of the solid at elevated temperature was not performed because of the increased risk of degradation. Isolated 17·HCl contained 8−12% residual 1,4-dioxane, which increased the solid’s hygroscopicity, thereby promoting the formation of diketopiperazine 18 and amino acid 21. However, we found that samples of 17·HCl could be stored for up to a month at 5 °C under a blanket of nitrogen to prevent formation of the impurities. Attempts were made to remove the residual 1,4-dioxane by slurrying 17·HCl in various solvents. The 1,4dioxane content was reduced below 2% when the product was slurried in t-BuOH, EtOAc, or THF, but in each case at least 15% of the mass of 17·HCl was lost to the supernatant and >7% of the new solvent remained in the solid. On the basis of these results, it was deemed that removal of the residual 1,4dioxane by slurrying solid 17·HCl in other organic solvents was ineffective and that the best course of action was to plan the pilot-plant processing schedule in such a way that 17·HCl would be stored under nitrogen at 5 °C and used in the next step within a week of being isolated. Four batches were scaled up in the pilot plant (Scheme 8). The scales ranged from 2 to 45 kg input of 2·HCl. The stage 1 product 16 was not isolated in the plant; instead, the solvent was exchanged from MTBE to 1,4-dioxane by distillation, and the solution of 16 in 1,4-dioxane was used directly in stage 2 to afford 17·HCl. Each batch of 17·HCl contained 8−11% residual 1,4-dioxane. The uncorrected crude yields of 17·HCl were 98−104% over two steps from 2·HCl. The purity of the product in each batch was ≥98.0% AUC as determined by HPLC. Coupling of 17·HCl with Cyclohexaneacetic Acid (3) Followed by Hydrolysis To Afford Acid 23. Our main goal for developing the coupling of 17·HCl with cyclohexaneacetic acid (3) for the initial kilo-lab campaign was to replace DMF with a lower-boiling solvent that would provide a phase cut in an aqueous workup. A solvent screen was performed on a gramscale in which cyclohexaneacetic acid was activated with NMM, HOBT, and EDCI and then 17·HCl was added. These studies showed that about 30−40% of the product mixture was diketopiperazine 18 when the reaction was performed in iPrOAc or EtOAc. The reaction in DMF afforded approximately 5% diketopiperazine 18. No diketopiperazine was detected in CH2Cl2, so this was selected as the reaction solvent. Further

product fell to 64% and the purity decreased from >98% to 96% AUC. 2. Diketopiperazine 18 was found to slowly form upon prolonged standing of isolated 17·HCl. This would be a significant challenge when the process was transferred from the kilo lab to the plant, where the longer operating cycles would mean that we might have to store samples of 17·HCl for over a month. 3. When the EtOAc was not fully removed in the solvent exchange between the coupling and deprotection steps, it inhibited the precipitation of 17·HCl and afforded a sticky product containing acetamide 20 (Figure 2) produced from the acidic hydrolysis of EtOAc. Acetamide 20 and its downstream products were found to be poorly rejected in the subsequent synthesis and purification operations, leading to contamination of the final product 1, so the formation of acetamide 20 had to be prevented. 2-Propanephosphonic acid anhydride (T3P),14 oxalyl chloride, and isobutyl chloroformate were briefly assessed as alternative coupling reagents for the pilot-plant conditions, but none was as effective as EDCI/HOBt in dipolar aprotic solvents. Notably, the coupling reaction in NMP, DMF, or DMA gave greater than 97% conversion in 18−24 h at ambient temperature. DMA was ruled out as a solvent because the purity of the product in the reaction mixture was only 92% AUC, compared with 97−98% AUC for DMF and NMP at 20−25 °C. In order to improve the process throughput, the reactions were heated to 35 °C in DMF and NMP. Both reactions were complete within 6−8 h. However, significant degradation of the product was observed in DMF at 35 °C, so NMP was selected for the EDCI/HOBt-mediated coupling reaction. A study of the effect of the reaction temperature on the product purity profile showed that it was necessary to charge NMP first, cool it to 0−5 °C, and then charge the reagents. If the reagents were charged at 20−25 °C, 1−2% AUC impurity 19 formed via double addition of the tert-leucine unit. Only 0.2−0.4% impurity 19 formed when reagents were added to NMP at 0−5 °C. It was demonstrated that the optimal amount of NMM for the coupling reaction was 3.0 equiv with respect to 2·HCl. Slower reaction was observed when 2.5 equiv of NMM was employed, and there was no benefit from charging 4.0 equiv of NMM. In order to avoid the potential formation of acetamide 20 during the Boc deprotection step, the solvent in the workup after the coupling needed to be changed from EtOAc to a more chemically inert solvent. Also, a solvent with a relatively low boiling point was required in order to facilitate the solvent swap to 1,4-dioxane, so CH2Cl2 and MTBE were tested to solubilize intermediate 16 while washing the reaction with water to remove NMP. It was determined that more NMP partitioned into the aqueous phase from MTBE, so this was selected as the organic phase for the aqueous workup of the coupling reaction for the pilot-plant campaign. Prior to the pilot-plant campaign, we re-examined our choice of 4 N HCl in 1,4-dioxane for the Boc deprotection step since 1,4-dioxane is classified by the ICH as a class 2 solvent with an option 1 limit of 380 ppm. We sought to use commercially available solutions of HCl in ICH class 3 solvents (option 1 limit