Development of an Organometallic Flow Chemistry Reaction at Pilot

Feb 22, 2018 - We report the development of an organolithium addition to a chiral ketimine at pilot-plant scale using flow chemistry. Given the mixing...
1 downloads 5 Views 353KB Size
Subscriber access provided by MT ROYAL COLLEGE

Communication

Development of an organometallic flow chemistry reaction at pilot plant scale for the manufacture of verubecestat David A. Thaisrivongs, John R Naber, Nicholas Rogus, and Glenn Spencer Org. Process Res. Dev., Just Accepted Manuscript • DOI: 10.1021/acs.oprd.7b00385 • Publication Date (Web): 22 Feb 2018 Downloaded from http://pubs.acs.org on February 23, 2018

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.

Organic Process Research & Development 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 18 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

Development of an organometallic flow chemistry reaction at pilot plant scale for the manufacture of verubecestat David A. Thaisrivongs*, John R. Naber*, Nicholas J. Rogus, and Glenn Spencer Process Research and Development, Merck & Co., Inc., P.O. Box 2000, Rahway, New Jersey, 07065 USA

ACS Paragon Plus Environment

1

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 2 of 18

TABLE OF CONTENTS GRAPHIC

ACS Paragon Plus Environment

2

Page 3 of 18 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

KEYWORDS Flow chemistry, continuous processing, process development, organometallics

ABSTRACT

We report the development of an organolithium addition to a chiral ketimine at pilot plant scale using flow chemistry. Given the mixing-sensitivity of this process, established previously at both lab and kilogram scale, we conducted critical experiments to understand the impact of flow rate, temperature, and mixing on the performance of the reaction in a factory-sized static mixer. Online process analytical technology was used to collect real-time data on batches to demonstrate steady-state operation over multiple hours, and also to confirm the robustness of the continuous process against fouling. These experiments at pilot plant scale provided improved operating parameters that enable the implementation of this chemistry for the commercial manufacture of verubecestat, currently being evaluated in Phase III trials for the treatment of Alzheimer’s disease.

MAIN TEXT Verubecestat1 (1), an inhibitor of BACE12 that dramatically reduces the levels of amyloid β peptides in the central nervous system, is being evaluated in Phase III trials for the treatment of Alzheimer’s disease.3 We recently published an efficient chemical synthesis of 1 that has been implemented on pilot plant scale to support activities related to its late stage clinical program.4 Subsequent to that report, we disclosed the development of a novel continuous process for the pivotal Mannich-type ketimine addition that sets the tertiary carbamine stereocenter of 1 (Figure

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 18

1).5 This work was influenced by mechanistic studies which revealed that proton transfer between the nucleophile and electrophile, as well as between the unquenched product and the electrophile, is competitive with the desired transformation. Hypothesizing that, in batch reaction conditions, the true kinetic selectivity between the desired and undesired reactions was disguised by the mixing rate of the reactive species,6 we were able to realize significant improvements in yield when the process was conducted in flow with fast micromixing.7 Performing the process in a flow reactor also eliminated the need for cryogenic cooling, and we were able to further demonstrate the optimized process on kilogram scale before proceeding to even larger batches. Herein we describe the development and execution of this process on pilot plant scale.

Figure 1. Using flow to outpace fast proton transfer in an organometallic reaction for the synthesis of verubecestat (1) on kilogram scale

Despite the control we had established over the flow process at kilogram scale through reaction parameter optimization and equipment design, we remained unsatisfied with the robustness of the continuous deprotonation of 2 by n-hexyllithium. Conducting that operation at

ACS Paragon Plus Environment

4

Page 5 of 18 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

−20 °C, as we previously reported, provides stable steady-state performance on the timescale of roughly an hour, but there were indications that it would be challenging to execute the deprotonation for longer periods of time if required for commercial scale batches. Specifically, deposition of unidentified insoluble material at the point of contact between 2 and nhexyllithium8 eventually led to an increased pressure drop through the first mixer.9 Cooling the two feed streams slowed, but did not eliminate, the tendency of the system to foul, and we were limited in our ability to pursue this option as the solubility of 2 in THF at low temperatures is very poor. Notably, these concerns centered on the deprotonation process; no data suggested that there were similar risks of fouling at either the ketimine addition or quench mixers. One way to circumvent this issue would be to generate lithium anion 4 in a batch reactor and subsequently flow that solution into a mixer where it would be combined with ketimine 3 (Figure 2). This design would eliminate the fouling event associated with the continuous deprotonation, but would involve unavoidable trade-offs. Firstly, the feed solutions for the previous process design (2 in THF, n-hexyllithium in hexanes, and 3 in THF) were stable at ambient temperature while anion 4 is not. Thus it would be necessary to maintain 4 at low temperature (i.e., −20 °C) to minimize reagent decomposition over the course of a commercial scale flow batch.10 Secondly, additional temperature-controlled jacketed lines would be required to transfer anion 4 from its feed vessel to the mixer.11 Importantly, however, this approach would enable us to realize all of the value of performing the ketimine addition in flow while ensuring the deprotonation of 2 was as robust as possible. Lab experiments designed to mimic the redesigned process demonstrated that it was sufficiently robust to enable steady state operation for multiple hours and, importantly, the incorporated changes did not result in a diminution in the performance of the reaction.

ACS Paragon Plus Environment

5

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 6 of 18

Another important aspect of the new process was the order of reagent addition. Cooling a solution of 2 in THF to −20 °C before charging n-hexyllithium led to crystallization of 2 on the vessel wall, reducing the cooling efficiency of the jacket by effectively insulating the batch. By charging THF to the vessel, cooling to −20 °C, and then charging 2, the insoluble 2 remained a free-flowing slurry. The subsequent addition of n-hexyllithium proceeded several times more quickly than when the vessel wall was coated with 2, improving our plant time cycle and minimizing the length of time anion 4 was aged before the start of the continuous process.

Figure 2. Redesigned continuous process that incorporates a batch deprotonation of 2 to improve chemical robustness

In preparation for our first pilot plant scale batch, we were unexpectedly confronted by another robustness issue. In all previous experiments, including many at kilogram scale, the lithium anion 4 was a homogeneous solution in THF. Nevertheless, preparation of 4 at pilot scale resulted in the discovery of a crystalline phase of 4 that was thermodynamically favored over the soluble anion in our desired operating window.12 It was clear that the resulting slurry of 4 was incompatible with the planned flow process, so our focus turned to establishing modified

ACS Paragon Plus Environment

6

Page 7 of 18 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

conditions that would solubilize this new polymorph. Dilution did not appear promising, as at half of the planned concentration of 4, some of the new crystalline phase still remained out of solution.13 Neither could we employ a more polar solvent (e.g., DMSO, DMF, or NMP), as this resulted in a substantially reduced yield of the desired product 5. Early experiments had shown that warming this solution led to decomposition of the anion (vide supra), suggesting that the insolubility of 4 could not be overcome by simply adjusting the temperature upwards. Given the proclivity of organolithium species to form higher-order aggregates in solution,14 we hypothesized that additives able to perturb such species might positively impact the solubility of 4. A panel of structurally-diverse and readily available additives (tetramethylethylenediamine, N,N’-dimethylpropyleneurea (DMPU), N,N’-dimethylethyleneurea (DMEU), pyridine, trimethylamine, and dimethoxyethane) were tested and DMPU was most effective in solubilizing 4 at the desired operating concentration and temperature. Optimization experiments showed that between 0.8 and 1.0 equivalents of DMPU were required to generate a homogeneous solution of 4 which resisted crystallization even upon addition of isolated 4 as a seed. Importantly, the addition of DMPU did not measurably impact the performance of the reaction, nor did it significantly alter the stability of the lithium anion feed solution. In the optimized process, n-hexyllithium is charged to a mixture of 2 in THF to generate 4, at which point 1.0 equivalent of DMPU is added and the homogeneity of the batch was visually confirmed before the flow reaction was initiated.15 Having established at both lab and kilogram scale that mixing efficiency was a critical process parameter, one of the main objectives of our first pilot plant batch was to evaluate the reaction’s performance at a range of flow rates through a commercial scale mixer. A Y-shaped mixer was fabricated from 1/4 inch Schedule 40 pipe (316L stainless steel) and a Koflo static

ACS Paragon Plus Environment

7

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 8 of 18

mixing element16 was installed in the downstream segment (Table 1). The experimental plan was to measure the reaction conversion and diastereoselectivity over a range of flow rates (from 55.5 kg/h to 250.0 kg/h total), maintaining the necessary flow rate ratio to always deliver 1.7 equivalents of 4 relative to 3, while also varying the temperature set points of the 3 feed stream (from 22 to −62 °C).17 Given the relative instability of 4 at temperatures greater than −10 °C, that feed solution was both held and transferred to the mixer between −10 and −20 °C. At all temperatures studied, the reaction conversion consistently improved as the total flow rate, and thus the quality of mixing, was increased. This effect was most pronounced at the warmest temperature, wherein the conversion increased from 64.2% to 82.1% when the total flow rate was raised from 55.5 kg/h to 250.0 kg/h, holding all other parameters constant (entries 1 to 5). As the feed temperature of 3 was lowered, the overall reaction temperature was consequently reduced leading to higher reaction conversions and a progressive decrease in the impact of the total flow rate on conversion. At the coldest temperature, the improvement in conversion increased from 80.5% to 85.6% when the total flow rate was raised from 55.5 kg/h to 250.0 kg/h (entries 16 and 20). Table 1. The impact of flow rate and temperature at pilot plant scale O O -10 to -20 °C Li S H2C NPMB Me

Temperature Br

Br Temperature and total flow

3 (°C)

t-Bu

Me

Koflo® static mixer t-Bu

total flow (kg/h)

O S

F 3 (0.8 M) 1.0 equiv

4 (1.0 M) 1.7 equiv and 1.0 equiv DMPU 1/4 inch Schedule 40 Pipe

entry

N

postmixer (°C)

O S

NH O O S NPMB Me Me F 5

conversion to 5 (%)a

dra

ACS Paragon Plus Environment

8

Page 9 of 18 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

1 55.5 21 2 111.1 21 3 166.6 21 4 222.3 21 5 250 21 6 55.5 −6 7 111.1 −4 8 166.6 −2 9 222.3 −1 10 250 0 11 55.5 −32 12 111.1 −28 13 166.6 −24 14 222.3 −20 15 250 −19 16 55.5 −57 17 111.1 −50 18 166.6 −43 19 222.3 −37 20 250 −35 a Determined by HPLC analysis

11.0 10.5 10.9 10.5 10.4 4.7 4 3.6 3.9 4.1 −2.5 −3 −2.7 −2 −1.6 −8.2 −9 −7.8 −6.5 −6.2

64.2 73.1 82.4 81.8 82.1 76.6 81.4 82.8 84 84.1 79 83.3 84.6 84.9 85.3 80.5 84.5 85.7 85.6 85.6

91.5:8.5 90.0:10.0 89.5:10.5 89.7:10.3 89.7:10.3 90.6:9.4 90.5:9.5 90.4:9.6 90.3:9.7 90.3:9.7 91.1:8.9 91.0:9.0 91.0:9.0 90.8:9.2 90.8:9.2 91.7:8.3 91.3:8.7 91.4:8.6 91.3:8.7 91.1:8.9

Before the product stream was transferred to the quench vessel, it was passed through a flow cell equipped with an IR probe allowing for real-time analysis of the reaction’s performance. Satisfyingly, once the measured IR response was adjusted based on the temperature of each measurement, we were able to confirm that a high degree of correlation existed between the offline HPLC and online IR analyses (Figure 3). Importantly, this latter method is well-suited to real-time monitoring of a continuous reaction in a commercial manufacturing environment, where such process analytical technology can provide instantaneous feedback on the reaction, enable a rapid response to maintain the target design space, and reduce the safety and environmental hazards associated with sampling.18

ACS Paragon Plus Environment

9

Organic Process Research & Development

Figure 3. Correlation of reaction conversion by HPLC and real-time flow IR responsea 90

HPLC conversion (%)

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 10 of 18

80 R² = 0.98

70

60 0.065

0.07 0.075 0.08 IR response (arbitrary units)

0.085

a

IR response values are normalized with respect to temperature and are given in arbitrary units by the online IR software used to record this data.

An additional objective of this pilot plant scale experimentation was to demonstrate the robustness of the overall process with respect to clogging. Both reactants were delivered by pressurizing the feed vessels and delivering the solutions through independent and automatic mass flow controllers. Feeding from pressurized vessels rather than using pumps (as we had on lab scale) eliminated any pulsing that could vary the feed ratio, which we believed might deleteriously affect the reaction conversion given the mere millisecond mixing timeframe. By continuously measuring both the mass flow of each reactant and the controller valve position, we were able to assess whether any fouling along either flow path was occurring, as that would result in an increase in the pressure drop and a concurrent opening of the respective control valve to maintain the mass flow set point. Measurements for both the lithium anion 4 (Figure 4) and ketimine 3 (Figure 5) feeds established robust steady-state performance at each mass flow set

ACS Paragon Plus Environment

10

Page 11 of 18

point and temperature (Table 1). Significantly, multiple experiments were performed at the same mass flow set points (e.g., Table 1, entries 1, 6, 11, and 16 were all performed at 55.5 kg/h) and the same controller valve position was observed (e.g., Figures 4 and 5 at 0-10, 27-37, 53-64, and 82-91 minutes). Similar robustness has been demonstrated in many subsequent batches that each involved a single continuous run of this process for many hours. Figure 4. Pilot plant mass flow and controller valve measurements for lithium anion 4a 120

250

Valve Flow

100

200

80 150 60 100

Mass flow (kg/h)

Controller valve position (%)

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

40 50

20

0

0 0

30

60 Time (min)

90

120

a

Measurements throughout the 20 experiments described in Table 1. The vessel containing lithium anion 4 emptied after approximately 115 minutes, and the vessel containing ketimine 3 emptied after approximately 113 minutes.

Figure 5. Pilot plant mass flow and controller valve measurements for ketimine 3a

ACS Paragon Plus Environment

11

Organic Process Research & Development

120

120

Valve Flow

100

100

80

80

60

60

40

40

20

20

0

Mass flow (kg/h)

Controller valve position(%)

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 12 of 18

0 0

30

60 Time (min)

90

120

a

Measurements throughout the 20 experiments described in Table 1. The vessel containing lithium anion 4 emptied after approximately 115 minutes, and the vessel containing ketimine 3 emptied after approximately 113 minutes.

While the demonstration of system robustness under steady state operation was a significant accomplishment, the start-up and shut-down of a continuous flow system is another aspect of process robustness that needed to be evaluated. System startup was designed to both maximize process robustness and to ensure that the more valuable reagent was processed with the highest level of conversion. Both flow lines were fitted with divert valves upstream of the static mixer to allow priming. Once primed, the control valves were manually set to the target positions previously determined during solvent dummy runs and both shutoff valves were opened, starting with the anion stream. The total time that elapsed from the start of fluid transfer to the point at which process control was attained was less than a minute, minimizing the amount

ACS Paragon Plus Environment

12

Page 13 of 18 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

of excess anion preceding the start of the reaction. Once both streams were flowing the mass flow controller was switched to automatic mode for the remainder of the reaction. The flow reactor used on pilot plant scale was fabricated from ¼ inch Schedule 40 pipe with static mixing elements in the tail piece (6 inches in length). The internal volume of the static mixer was about 10 mL, and when compared to the total volume of the flow reaction (1000 L), the total number of reactor volumes (100,000) makes the diversion of any startup material unnecessary. Similarly, when the total reaction flow rate is 4.5 L/min and the residence time is 0.13 seconds, the time to steady state is less than one second once process control is established. For shutdown, the process was allowed to run to completion with very little intervention. Based on the small excess of anion prepared ahead of the batch, the ketimine stream was consumed first and was allowed to run to dryness with the vessel pressure blowing the line dry. At this point the flow controller for the anion stream was switched to manual and the control valve opened to the maximum to allow for fast emptying of the anion vessel. All material was pushed into the quench vessel for downstream processing. Encouraged by the experimental data that had been collected in the first pilot plant batch, we sought to further optimize and streamline the process design. Enhancements to the equipment train enabled rigorous temperature control at each point along the flow path and improvements in our ability to execute the entire process rapidly enabled us to significantly shorten the length of time between the start of the n-hexyllithium charge to 2 and the beginning of the flow reaction.19 Since solutions of 4 slowly decrease in purity over time even at −20 °C, the best results are obtained with freshly prepared anion. In addition to assessing the impact of flow rate and temperature, we were also able to test the performance of several static mixers at pilot plant scale.20 When the Y-shaped mixer (mixer A) we had employed in the first batch was used with

ACS Paragon Plus Environment

13

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 14 of 18

the optimized equipment train, we observed a small but consistent increase in reaction performance relative to prior results (Table 1) when both feed solutions were cooled to at least −19 °C (Table 2, entry 1). Consistent with our previous observations, further cooling of ketimine 3 did not significantly improve the reaction conversion or diastereoselectivity (Table 2, entry 2). A corresponding T-shaped mixer (mixer B) containing the same Koflo static mixing element design as mixer A was fabricated out of 1/4 inch Schedule 40 pipe. Under similar operating conditions, conversion was slightly lower than the Y-shaped mixer (Table 2, entry 3). A third mixer (mixer C), also T-shaped, with a much smaller inner diameter (0.194 inches, compared with the inner diameter of 0.364 inches of 1/4 inch Schedule 40 pipe) was fabricated to evaluate whether a further increase in mixing efficiency gained by conducting the flow reaction in a mixer with a smaller cross-sectional area would lead to an improvement in conversion. In this experiment, the substantial increase in pressure drop across this narrower mixer limited the total flow rate we were able to achieve. Nevertheless, at 166.7 kg/h the reaction conversion and diastereoselectivity remained unchanged relative to the best results obtained with the larger mixer (Table 2, entries 2 vs 4); similar results were obtained at 188.9 kg/h, the maximum possible throughput based on the pressure applied to the feed vessels (Table 2, entries 5 and 6).21 Table 2. Comparison of performance of three pilot plant scale mixers

entry mixer

total flow (kg/h)

4 (°C)

3 (°C)

postmixer (°C)

conversion (%)a

dra

ACS Paragon Plus Environment

14

Page 15 of 18 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

90.8:9.291.7:8.3 91.7:8.32 A 250.0 88.1-88.5 −20 −38 −7 92.2:7.8 91.5:8.53 B 250.0 86.5-86.9 −29 −34 −12 92.1:7.9 4 C 166.7 88.0-88.5 91.1:8.9b −21 −21 −2 5 C 188.9 88.7b 91.4:8.6b −21 −20 −2 91.3:8.76 C 188.9 88.6-88.7 −26 −29 −8 91.5:8.5 a Determined by HPLC analysis; the minimum and maximum values at steady state operation are reported. bEach sample at steady state gave the same value, within error. 1

A

250.0

−19

−22

−2

87.5-88.2

Given that a narrower tube mixer provided no chemical benefit and significantly reduced productivity, we performed even longer test runs with the wider mixer A. Based on the accumulated data about the relationship between reaction conversion, diastereoselectivity, and temperature (Tables 1 and 2), we choose to deliver both reactants between −10 and −20 °C, a regime that did not require specialized cryogenic equipment. Further, matching the feed solution temperatures enabled a single external chiller unit to supply cooling fluid to both flow heat exchangers, simplifying the engineering requirements. With these changes, we were able to conduct the optimized process at steady state conditions for over three hours at pilot plant scale to produce more than 100 kg of product in a single batch with consistent assay yields of 88 to 89% (Figure 6). Figure 6. Optimized continuous process at pilot plant scale

ACS Paragon Plus Environment

15

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 16 of 18

In summary, we have conducted process development for an organometallic flow chemistry reaction in a pilot plant. To our knowledge, there are no literature reports of such experimentation on a fast, mixing-sensitive reaction at this scale (>100 kg batches). In addition to evaluating in detail the relationship between mass flow rate, temperature, and conversion, a primary focus of our piloting efforts was to understand the weak points in the overall design and subsequently introduce modifications which enhanced the overall robustness of the process. We have found this methodology to be particularly useful for the development of flow chemistry reactions for large scale applications, and hope that this publication serves as a guide to some of the considerations that are important when designing a continuous process for commercial manufacturing.

ASSOCIATED CONTENT Supporting Information. Equipment details and a representative pilot plant batch description. This material is available free of charge via the Internet at http://pubs.acs.org. AUTHOR INFORMATION Corresponding Author

ACS Paragon Plus Environment

16

Page 17 of 18 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

*E-mail: [email protected] *E-mail: [email protected]

Notes The authors declare no competing financial interest. ACKNOWLEDGMENT We acknowledge Erik L. Regalado for analytical support, Ted Furman, Jarod Nappi, Sarah Pulicare, and Azzeddine Lekhal for engineering support, and thank the supervisors and chemical operators of the Merck & Co., Inc., Rahway, NJ 07065 USA Small Scale Organics Pilot Plant. REFERENCES

1 (a) Scott, J. D.; Stamford, A. W.; Gilbert, E. J.; Cumming, J. N.; Iserloh, U.; Misiaszek, J. A.; Li, G. PCT Int. Appl., WO 2011/044181 A1, PCT/US2010/051553, 2011. (b) Scott, J. D.; Li, S. W.; Brunskill, A. P. J.; Chen, X.; Cox, K.; Cumming, J. N.; Forman, M.; Gilbert, E. J.; Hodgson, R. A.; Hyde, L. A.; Jiang, Q.; Iserloh, U.; Kazakevich, I.; Kuvelkar, R.; Mei, H.; Meredith, J.; Misiaszek, J.; Orth, P.; Rossiter, L. M.; Slater, M.; Stone, J.; Strickland, C. O.; Voigt, J. H.; Wang, G.; Wang, H.; Wu, Y.; Greenlee, W. J.; Parker, E. M.; Kennedy, M. E.; Stamford, A. W. J. Med. Chem. 2016, 59, 10435. 2 For recent process development work on the synthesis of BACE1 inhibitors, see: (a) Znidar, D.; Hone, C. A.; Inglesby, P.; Boyd, A.; Kappe, C. O. Org. Process Res. Dev. 2017, 21, 878. (b) Kolis, S. P.; Hansen, M. M.; Arslantas, E.; Brändli, L.; Buser, J.; DeBaillie, A. C.; Frederick, A. L.; Hoard, D. W.; Hollister, A.; Huber, D.; Kull, T.; Linder, R. J.; Martin, T. J.; Richey, R. N.; Stutz, A.; Waibel, M.; Ward, J. A.; Zamfir, A. Org. Process Res. Dev. 2015, 19, 1203. (c) Hansen, M. M.; Jarmer, D. J.; Arslantas, E.; DeBaillie, A. C.; Frederick, A. L.; Harding, M.; Hoard, D. W.; Hollister, A.; Huber, D.; Kolis, S. P.; Kuehne-Willmore, J. E.; Kull, T.; Laurila, M. E.; Linder, R. J.; Martin, T. J.; Martinelli, J. R.; McCulley, M. J.; Richey, R. N.; Starkey, D. R.; Ward, J. A.; Zaborenko. N.; Zweifel, T. Org. Process Res. Dev. 2015, 19, 1214. (d) DeBaillie, A. C.; Jasper, P. J.; Li, S.; McCulley, M. J.; Vaidyaraman, S.; Zhang, Z. Org. Process Res. Dev. 2015, 19, 1244. (e) Henegar, K. E.; Lira, R.; Kim, H.; Gonzalez-Hernandez, J. Org. Process Res. Dev. 2013, 17, 985. 3 Kennedy, M. E.; Stamford, A. W.; Chen, X.; Cox, K.; Cumming, J. N.; Dockendorf, M. F.; Egan, M.; Ereshefsky, L.; Hodgson, R. A.; Hyde, L. A.; Jhee, S.; Kleijn, H. J.; Kuvelkar, R.; Li, W.; Mattson, B. A.; Mei, H.; Palcza, J.; Scott, J. D.; Tanen, M.; Troyer, M. D.; Tseng, J. L.; Stone, J. A.; Parker, E. M.; Forman, M. S. Sci. Trans. Med. 2016, 8, 363ra150. 4 Thaisrivongs, D. A.; Miller, S. P.; Molinaro, C.; Chen, Q.; Song, Z. J.; Tan, L.; Chen, L.; Chen, W.; Lekhal, A.; Pulicare, S. K.; Xu, Y. Org. Lett. 2016, 18, 5780. 5 Thaisrivongs, D. A.; Naber, J. R.; McMullen, J. P. Org. Process Res. Dev. 2016, 20, 1997. 6 (a) Rys, P. Acc. Chem. Res. 1976, 9, 345. (b) Rys, P. Angew. Chem. Int. Ed. Engl. 1977, 16, 807; Angew. Chem. 1977, 89, 847. 7 (a) Yoshida, J.; Nagaki, A.; Iwasaki, T.; Suga, S. Chem. Eng. Technol. 2005, 28, 259. (b) Nagaki, A.; Togai, M.; Suga, S.; Aoki, N.; Mae, K.; Yoshida, J. J. Am. Chem. Soc. 2005, 127, 11666. (c) J. Yoshida, Flash Chemistry. Fast Organic Synthesis in Microsystems, Wiley-Blackwell, 2008; (d) J. Yoshida, A. Nagaki and T. Yamada, Chem.– Eur. J., 2008, 14, 7450. (e) J. Yoshida, Chem. Rec., 2010, 10, 332. (f) H. Kim, A. Nagaki, J. Yoshida, Nat. Commun. 2011, 2, 264 and references therein. 8 Analysis by 1H NMR revealed the solid material contained fragments of decomposed 2, though no definitive identification was made. 9 We hypothesize that the apparent decomposition is due to the heat of reaction when 2 contacts n-hexyllithium. Although heat transfer is generally deemed efficient in flow, we could not completely dissipate the exotherm of this reaction, as downstream temperature measurements unequivocally demonstrated.

ACS Paragon Plus Environment

17

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 18 of 18

10

A solution of 4 held at −20 °C for three days lost between 2 and 4% of its potency each day. In addition, a certain amount of heat gain of the anion solution through a pump head or mass flow controller was similarly unavoidable, though at the high flow rates we hoped to achieve on plant scale, and the use of pressure-driven flow on scale, the contact time between the solution and such equipment would be minimized. 12 We have not been able to determine why previous lab scale experiments failed to elucidate this polymorph. Subsequent to this observation in the pilot plant, we were able to reproduce the crystallization event on lab scale. We believe this phenomenon is complementary to that of a “disappearing polymorph”; see Dunitz, J. D.; Berstein, J. Acc. Chem. Res. 1995, 28, 193. 13 Measuring the solubility of 4, an unstable reactive intermediate, in various solvent systems and at different temperatures proved very challenging, and in the end we were limited to qualitative analyses. 14 Reich, H. J. Chem. Rev. 2013, 113, 7130 and references therein. 15 See Supporting Information for details, including batch photos. 16 Model number: ¼-40-3-12L-1, purchased from Koflo Corporation, 309 Cary Point Drive, Cary, IL, 60013. 17 Based on equipment constraints, 250 kg/h was the maximum flow rate we could achieve and −62 °C was the lowest temperature. 18 Chanda, A.; Daly, A. M.; Foley, D. A.; LaPack, M. A.; Mukherjee, S.; Orr, J. D.; Reid, G. L.; Tompson, D. R.; Ward, H. W. Org. Process Res. Dev. 2015, 19, 63. 19 In our first pilot plant batch, the flow reaction began approximately 16 hours after the start of the n-hexyllithium charge; in subsequent batches we were able to more than halve this time. 20 See Supporting Information for details about the construction of mixers A, B, and C. 11

21

Subsequent to this study, a more rigorous characterization of these flow reactors was performed that is able to better distinguish their mixing performance. This work will be reported in due course.

ACS Paragon Plus Environment

18