Subscriber access provided by UNIV AUTONOMA DE COAHUILA UADEC
Full Paper
Systematic optimization of a robust telescoped process for a BTK inhibitor with atropisomer control by high-throughput experimentation, design of experiments, and linear regression. Federico Lora Gonzalez, Steven R. Wisniewski, Kishta Katipally, Jason M Stevens, Victor W. Rosso, Brendan Mack, and Thomas M. Razler Org. Process Res. Dev., Just Accepted Manuscript • DOI: 10.1021/acs.oprd.8b00398 • Publication Date (Web): 08 May 2019 Downloaded from http://pubs.acs.org on May 8, 2019
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 12 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Organic Process Research & Development
Systematic optimization of a robust telescoped process for a BTK in‐ hibitor with atropisomer control by high‐throughput experimenta‐ tion, design of experiments, and linear regression. Federico Lora Gonzalez*, Steven R. Wisniewski, Kishta Katipally, Jason M. Stevens, Victor Rosso, Brendan Mack, Thomas M. Razler Chemical and Synthetic Development, Bristol‐Myers Squibb Company, One Squibb Drive, New Brunswick, New Jersey 08903,United States. KEYWORDS: Atropisomer, Bruton’s Tyrosine Kinase, Chiral Axis, Optimization, High‐throughput Experiments ABSTRACT: The development of a telescoped process for the API step in the route to a reversible inhibitor to BTK is reported. Although a robust and effective reaction and crystallization was previously implemented, concerns around the storage and stability of the penultimate material led to investigating a telescoped process. Utilizing design of experiments (DoE) to understand impacts to the API reaction and high‐throughput extractions (HTEx) to optimize the work‐up, a pro‐ cess was rapidly developed to achieve the same quality API with an overall 13% improvement to yield and 41% improvement in process mass intensity (PMI).
INTRODUCTION Since the introduction of the Food and Drug Admin‐ istration’s (FDA) initial guidance on Quality by Design (QbD), the use of statistical design of experiments, multi‐ variate analysis, and study of chemical kinetics for process understanding has become commonplace.1 The QbD framework as outlined by the FDA proposed to shift the paradigm from product testing to more thorough and fun‐ damental understanding of the chemical synthesis and unit operations involved in the manufacture of active phar‐ maceutical ingredients (APIs). The QbD proposal, along with subsequent publications by the International Council for Harmonisation of Technical Requirements for Pharma‐ ceuticals for Human Use (ICH) emphasized the use of ex‐ perimental design, statistical methods, and fundamental understanding to mitigate risk to quality in the production of API.2 The use of statistical design of experiments (DoE), model‐based optimization, and parallel experimentation has also enabled faster development timelines with less cost, improved use of resources, smaller environmental footprints for both the experiments and the resulting pro‐ cesses, and more robust control strategies for critical qual‐ ity attributes (CQAs).3 Herein, we present a case study of parallel development of a multi‐step process (multiple re‐ actions and workups) using statistically designed screen‐ ing design of experiments, high‐throughput screening for extractions (HTEx), and parallel experimentation to opti‐ mize yield, safety, process mass intensity (PMI), and cost while maintaining an atropisomer impurity to acceptable
levels. This work fits in to the FDA’s vision of QbD by uti‐ lizing different methods to drive down risk to quality by understanding the parameter space and the critical aspects of the operations to achieve acceptable CQA levels. The case study presented here is focused on the devel‐ opment of a telescoped process of the final two steps in the synthesis of a Bruton’s Tyrosine Kinase (BTK)4 inhibitor. The goal of the design of the telescoped process was to maximize yield and minimize cost while maintaining ac‐ ceptable quality. The penultimate step in the route5 to a re‐ versible inhibitor of BTK involves an EDAc (1‐ethyl‐3‐(3‐ dimethylaminopropyl) carbodiimide) mediated amidation between a highly functionalized aniline 1 and carboxylic acid 2 (Eq 1). Optimization of the reaction parameters and charge ranges of the reagents led to a robust transfor‐ mation in which variation of reagent equivalents (1‐me‐ thyl‐imidazole, methanesulfonic acid, EDAc) or reaction temperature (0‐20 °C) had little impact to the yield and no impact to the quality of 3. The reaction was demonstrated with six batches at 63 kg scale each, affording amide 3 in an average yield of 86%. H2N H 2N
O H N
HO O
F Me NH2 1
1-Me-Imidazole (1.3 equiv) MSA (0.5 equiv) EDAc*HCl (1.4 equiv)
OH
PrO
N Me
1.2 equiv. 2
F
DMF (7 vol) 10 °C, 4 h, 86%
O H N
O F
Me
OH (1)
O
N H PrO2C N Me 3
F
The crystallization of 3 initially proved to be challenging. Exploration of numerous solvents and conditions yielded only one form that could be reproducibly crystallized and
ACS Paragon Plus Environment
Organic Process Research & Development 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
easily filtered.3 This form was the heptane solvate, crystal‐ lized by the addition of heptane to a solution of 3 in iso‐ propyl acetate (IPAc) and dichloromethane. The crystalli‐ zation was extremely efficient by purging all but one im‐ purity to below reporting limits with low loss to the mother liquor. Further, residual heptane after drying did not have an impact on the final step of the synthesis. Although an effective crystallization was developed, long‐term storage of the heptane solvate posed a signifi‐ cant stability risk. In the first iteration of the process, the solvate was isolated with ~13 wt % heptane. Over 6 months of storage without desiccant, the amount of heptane de‐ creased from 11.5 wt% to 3.9 wt%. The potential release of a significant amount of a flammable solvent (~4.8 kg of heptane based on the 63 kg scale) led us to investigate whether we could dry the solvate to a stable level before discharging. Drying studies completed on lab scale showed that the levels of heptane became stable at ~6 wt % upon isolation and storage with desiccant, which would alleviate our concerns around solvent release. Upon implementa‐ tion of this drying process on plant scale (45 kg), the time required to reach a stable level of heptane was ~40 h, and the drying temperature had to be increased from 55 °C to 60 °C to reach the desired endpoint. The extended drying time, however, posed a significant quality risk as the only impurity, a product degradant,6 formed at a rate of 0.01 AP (area percent by HPLC)/h when drying at 60 °C. This prod‐ uct degradant results in the formation of an undesired at‐ ropisomer in the API step, and therefore it must be care‐ fully controlled during the penultimate step. Increased amount of this impurity increases the risk of failing a CQA, and therefore is the highest risk that needed to be miti‐ gated in the optimization of the process to arrive at the API. A second quality risk appeared when storing the heptane solvate, as it is hygroscopic. In two weeks at ambient con‐ ditions without desiccant, an increase in the water content from 0.2 wt % to 2 wt % was observed. Water has a signifi‐ cant effect on the subsequent step, greatly increasing the formation of one critical impurity. Therefore, residual wa‐ ter has to be tightly controlled, and any significant increase in water content requires an additional pre‐reaction azeo‐ tropic distillation in the subsequent step, increasing pro‐ cessing time, solvent demands, and process complexity. To alleviate the concerns around the stability of the hep‐ tane solvate of amide 3 and potential risks to quality, we envisioned developing a telescoped process that subjected a solution of 3 directly to the final reaction. In addition to circumventing the issues associated with isolating the pe‐ nultimate compound 3, a telescoped process could consid‐ erably lower the overall API cost and lower the PMI by re‐ ducing both solvent demands and cycle time via the re‐ moval of the isolation (filtration, drying, etc.). Further, a telescoped process could increase the yield by elimination of mother liquor losses in the isolation of 3. To prevent increasing quality risks, the impurity profile of the telescoped process must be controlled to the same limits as the original process. The ultimate goal of the tel‐
Page 2 of 12
escoped process was to develop the process in a safe, ro‐ bust, and efficient manner with optimum control over the CQAs of the API. Due to the complexity and scale of the problem, a com‐ binatorial approach to experimental optimization would be inefficient and impractical. As such, we opted to use de‐ sign of experiments and statistical models to optimize three separate unit operations in parallel. The API reac‐ tion, the penultimate workup, and the penultimate reac‐ tion were independently and contemporaneously opti‐ mized, then integrated into a single process. Previous scale‐up knowledge of the process and established CQAs informed the experimental designs for each operation; fur‐ thermore, critical process parameters (CPPs) identified in downstream operations were considered as responses for optimizing upstream operations, and upstream responses were regarded as input factors to downstream operations (Figure 1). Using this approach constrained the scope of the optimization for each unit operation and enabled tele‐ scoping the two reactions. Ultimately, the optimization significantly increased the yield and purity and minimized cost and PMI of the API in the shortest amount of time.
Figure 1. Schematic representation of the experimental plan and constraints guiding the use of DoE and HTEx.
Results and Discussion Part I: Optimization of the API Reaction The first step in the work flow was evaluation of the API reaction to determine which CPPs needed to be addressed during the work‐up of the penultimate step (Figure 1). The API‐forming reaction (Eq. 2) involves a base‐catalyzed, di‐ astereoselective cyclization to afford the quinazolinone 4.7 The key impurity formed in the reaction is the atropisomer of the API 5. The selectivity in the API reaction is con‐ trolled by the use of lithium as a counterion of the base, presumably through a preferred lithium aggregate that leads to the desired product.8 Therefore, we investigated residual N,N‐dimethylformamide (DMF) from the up‐ stream steps in the organic stream and residual water of both the penultimate solution and the base solution (the API step is a slow addition of a solution of penultimate to a solution of base) as coordination of both of these to a lithiated intermediate may impact selectivity. We also
ACS Paragon Plus Environment
2
Page 3 of 12 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Organic Process Research & Development
looked at reaction concentration as a factor to determine if dilution would impact the selectivity of the reaction. Fur‐ ther, we investigated temperature of the reaction because faster isomerization of the atropisomer 5 is observed when increasing the temperature.5
A screening DoE9 was used to better understand the im‐ pact of these parameters on the API reaction, and to opti‐ mize the robustness of the unit operation (Table 1). Twenty experiments were run, none of which were center point conditions.10 The original design (n = 16) was run, then augmented (n = 4) based on the measured values of resid‐ ual water input (experimental values shown in Table 1). The experiments were run on a Mettler‐Toledo EasyMax workstation in blocks of 4 experiments. The pa‐ rameter ranges were chosen based on previous experience with the process. We also included mixing rate and equiv‐ alents of base as factors to see if they would affect the tel‐ escoped API reaction. Of the parameters tested, mixing rate, base equivalents, and input stream concentration were found to be not significant to the formation of atro‐ pisomer 5 within the ranges tested. The other factors were analyzed by fitting the experimentally measured atropiso‐ mer levels with a simple linear regression.11 Table 1. Factors in the API DoE Factor
Low
High Impact
Temperature (°C)
15
35
DMF (mL/g)
0
0.045 High
Vol. Pot (mL/g)
15
25
Med
H2O, pen sol. (ppm)
58
4061
Med
H2O, base sol. (ppm)
51
715
High
Mixing rate (rpm)
100
500
Low
Base (mol %)
4
10
Low
High
Figure 2. Impact of DMF and water on the API Reaction. The effect of water and DMF on the atropisomer level can be re‐ duced by increasing the temperature.
In the original process, residual water from the aqueous splits in the work‐up of the amidation step was completely removed through an azeotropic distillation and subse‐ quent crystallization. Residual DMF was also completely purged in the isolation of 3, and therefore neither DMF nor water could impact the API reaction. However, when the reaction is telescoped, any residual DMF and water after the work‐up will be carried into the API reaction. One ben‐ efit from statistically designed experiments is observing the presence of secondary interactions; in this case, the secondary interaction between DMF and temperature was significant (Figure 2) where, at low temperatures, the effect of residual DMF was amplified. Under the center point conditions (white marker in Figure 3), ~4.2 AP of the atro‐ pisomer is formed. At 20 °C, an increase to only 0.040 mL/g DMF could lead to a 0.8 AP increase in atropisomer. How‐ ever, at 30 °C, an increase in residual DMF to 0.040 mL/g did not have an effect on the atropisomer levels. Because of this effect and to mitigate risk, the reaction temperature was increased to from 20 °C to 30 °C.
Based on the results of the DoE, the most important fac‐ tors influencing the selectivity of the reaction are: temper‐ ature, water content of the input and base solution, and residual DMF in the input stream (Figure 2). An increase in the water content coupled with an increase in DMF re‐ sults in a significant increase in atropisomer 5. However, the data suggests that if the reaction is run at higher tem‐ perature, the effects of higher DMF and higher water can be reduced.
ACS Paragon Plus Environment
3
Organic Process Research & Development 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Page 4 of 12
Table 2: Factors impacting the EDAc coupling reaction Factor
Low
Center
High
DMF (vol%)
40
60
75
Total Vol (mL/g)
5
7
9
2 (equiv)
0.9
1.15
1.25
Temperature (°C)
0
10
20
1‐Me‐Imidazole (equiv)
1.15
1.30
1.45
MSA (equiv)
0.45
0.5
0.55
EDAc (equiv)
1.15
1.4
1.55
Figure 3. Predicted atropisomer 5 levels at the end of the cy‐ clization reaction as a function of DMF and reaction temper‐ ature. White marker = original center point conditions. Red marker = first generation telescoped conditions. Green marker = optimized telescoped process.
Optimization and understanding of the API reaction by the use of design of experiments and linear regression im‐ proved the robustness of the process by identifying new centerpoint conditions and CPPs that must be controlled in upstream operations as responses (residual DMF and water). Additionally, the use of DoE identified secondary interactions (temperature + DMF) that would have other‐ wise been missed and could affect the robustness of the re‐ action. Part II. Optimization of the EDAc Coupling Reaction. The next step in the workflow was analysis of the penul‐ timate reaction. Because residual DMF was potentially quality and yield impacting, we investigated changing the solvent ratio in the penultimate step. We envisioned using a mixture of the API reaction solvent 2‐methyltetrahy‐ droufuran (MeTHF) and DMF to not only decrease the amount of DMF from the initial 7 mL/g but also to improve process greenness. The EDAc reaction cannot be run in neat MeTHF due to insolubility of the reactants and rea‐ gents. A set of statistically designed experiment was run to ef‐ ficiently evaluate the impact of the mixed solvent system on the quality and performance of the penultimate reac‐ tion (Table 2). A total of twenty four experiments were run in parallel in a high‐throughput set‐up using robotic solids handling and robotic sampling (Freeslate CM3).12,13 The use of laboratory automation technology enabled the en‐ tire experimental setup to be initiated over the course of a single day to produce detailed and highly structured exper‐ imental data. The factors tested were: solvent composition (DMF vol %, balance MeTHF), EDAc (equiv), 2 (equiv), MSA (equiv), 1‐Me‐imidazole (equiv), total solvent vol‐ umes (mL/g), and temperature (°C). The automation pro‐ tocols allowed for programmed and evenly spaced kinetic data to be collected every hour for 8 hours.14 For the anal‐ ysis, data points from the early times (
