Beyond organic solvents: synthesis of a 5-HT4receptor agonist in water

Article abstract of DOI:10.1039/d0gc03316b

Reducing or eliminating organic solvent use in pharmaceutical manufacturing is perhaps the most effective way to reduce the environmental, health, and safety impacts of drug substance manufacturing. With this in mind, we have developed a process to manufacture an investigational 5-HT₄receptor agonist that is conducted almost entirely in water, including multiple controlled isolations. Key transformations carried out in aqueous media include a benzimidazole cyclization, amide bond formation, reductive amination, and a selective oxidation of an aliphatic alcohol. Compared to the first-generation manufacturing process using organic solvents, the aqueous process described here uses 77% less material inputs, 94% less organic solvent, and, surprisingly, 48% less water, while improving overall yield from 35% to 56%.

Full text of DOI:10.1039/d0gc03316b

Green Chemistry
View Article Online
View Journal
PAPER
Beyond organic solvents: synthesis of a 5-HT4
receptor agonist in water†
Cite this: DOI: 10.1039/d0gc03316b
J. Daniel Bailey, *a Edward Helbling,b Amey Mankar,b Matthew Stirling,a
Fred Hicksa and David K. Leahya
Reducing or eliminating organic solvent use in pharmaceutical manufacturing is perhaps the most
effective way to reduce the environmental, health, and safety impacts of drug substance manufacturing.
With this in mind, we have developed a process to manufacture an investigational 5-HT4 receptor agonist
that is conducted almost entirely in water, including multiple controlled isolations. Key transformations
carried out in aqueous media include a benzimidazole cyclization, amide bond formation, reductive
amination, and a selective oxidation of an aliphatic alcohol. Compared to the first-generation manufactur-
ing process using organic solvents, the aqueous process described here uses 77% less material inputs,
94% less organic solvent, and, surprisingly, 48% less water, while improving overall yield from 35% to 56%.
Received 1st October 2020,
Accepted 3rd November 2020
DOI: 10.1039/d0gc03316b
rsc.li/greenchem
Moreover, a qualitative analysis suggests that water has signifi-
Imagining a future without organic
solvents
cantly lower life cycle impacts than organic solvents.2
A
growing body of research has demonstrated that a wide variety
of organic reactions can be successfully adapted to aqueous
media using micelle-forming surfactants or “on-water” meth-
odology.3 Researchers at Novartis have expanded on this work,
demonstrating that aqueous micellar media can serve as the
reaction medium for a variety of transformations during
pharmaceutical manufacturing on multi-kg scales, often pro-
viding superior outcomes over conventional solvents.4 A key
innovation enabling scale-up of reactions in aqueous micellar
media involves using small amounts of an organic co-solvent.5
Although there has been progress in adapting organic
chemistry to aqueous reaction conditions, replacing organic
solvents with water during workup, purification, and isolation
of reaction products has received comparatively little attention.
On average, reactions make up just 25% of the unit operations
used in API manufacturing, while the remaining 75% of unit
operations consist of extractions, crystallizations, filtrations,
and other operations related to product purification and iso-
lation.1 It is unsurprising then that an informal survey of
Takeda’s API manufacturing processes indicates that the
majority of organic solvent used during manufacturing is
devoted to purification and product isolation. Therefore, in
order to realize the full environmental and safety benefits of
conducting chemistry in water, there is an urgent need to
develop isolations from aqueous media that further reduce our
reliance on organic solvents while providing the physical pro-
perties and high product purities required for pharmaceutical
use. With this in mind, we set out to re-develop one of
Takeda’s API manufacturing processes with the goal of
running the entire process, including isolations, in water.
Since the first mass-produced drugs were introduced more
than 130 years ago, organic solvents have played a key role in
pharmaceutical manufacturing. Large quantities of solvent are
used at every stage of a typical drug manufacturing process to
solubilize organic reaction components, separate and purify
chemical mixtures, and isolate products. By one estimate,
organic solvents make up 56% by mass of the material inputs
used to manufacture active pharmaceutical ingredients (API).1
Despite a long history of solvent use, the pharmaceutical
industry has only recently begun to reckon with the full costs
of its dependence on organic solvents, costs that include a sub-
stantial environmental burden and serious risks to the health
and safety of workers. A full accounting of these costs has
prompted some in the pharmaceutical industry to begin
working toward a future where life-changing medicines are
manufactured without the use of organic solvents.
A variety of new technologies and approaches will be
needed to achieve this goal, but one of the most promising
existing strategies for removing organic solvents from drug
manufacturing processes involves using water as an alternative
reaction medium. Water is non-toxic and non-flammable,
making it inherently safer than many organic solvents.
aProcess Chemistry Development, Takeda Pharmaceuticals, 35 Landsdowne St.,
Cambridge, MA 02139, USA. E-mail: Dan.Bailey@Takeda.com
bWork carried out while at Process Chemistry Development, Takeda Pharmaceuticals,
35 Landsdowne St., Cambridge, MA 02139, USA
†Electronic supplementary information (ESI) available. See DOI: 10.1039/
d0gc03316b
This journal is © The Royal Society of Chemistry 2020
Green Chem.

View Article Online
Paper
Green Chemistry
properties and difficulty of isolation. In addition to problems
with process efficiency, five separate organic solvents are used
Results and discussion
TAK-954, the target API, is a 5-HT4 receptor agonist currently in the manufacturing sequence, most of which are categorized
under investigation by Takeda for the treatment of post-operat- as hazardous or problematic in the CHEM21 Solvent Guide.7
ive gastrointestinal dysfunction. The first-generation manufac-
In thinking about how to adapt this sequence to aqueous
turing route used to supply clinical material through Phase 2 conditions, we envisioned several changes to the synthetic
is shown in Scheme 1. The synthetic sequence begins with route, which are summarized in the retrosynthetic analysis in
condensation of the preformed bisulfite adduct of aldehyde 2 Scheme 2. These changes were intended to both improve the
with diamine 1 to give benzimidazole 3,6 which is then directly route’s overall efficiency and to ensure that the chemistry used
coupled with amine 4 in the presence of DBU to form amide 5. in the route is amenable to aqueous conditions. The most sig-
Boc-deprotection of 5 yields amine 6, which is isolated as a nificant change involved installing the second piperidine ring
bis-HCl salt. Piperidine 7 is then installed via reductive amin- with the desired methyl carbamate already in place, providing
ation followed by a second Boc-deprotection to give amine 9. direct access to TAK-954 and removing two steps from the
Finally, methyl chloroformate is used to form the methyl car- overall route. Because aldehyde 14 is not readily available, we
bamate moiety of TAK-954. A cooling recrystallization from planned to prepare it through carbamoylation and oxidation of
acetonitrile controls the crystal form and purity of TAK-954.
hydroxymethylpiperidine 12. We then hoped to derivatize alde-
This manufacturing sequence was executed multiple times hyde 14 in order to obtain a crystalline solid form suitable for
on production scale without issue, providing high-purity API use as a registrational starting material. The bond disconnec-
for clinical studies. The overall yield is an acceptable 35%. On tions in the earlier steps would remain unchanged. However,
the other hand, the overall process mass intensity (PMI), at preliminary experiments indicated that direct amidation of
350, is quite high given the modest complexity of the target methyl ester 3 with amine 4 is not feasible in aqueous media
molecule. An obvious synthetic inefficiency at the center of the due to competitive ester hydrolysis. Instead, we decided to
sequence contributes to the poor PMI: two non-productive pursue a more conventional amide coupling of carboxylic acid
steps are devoted to converting t-butyl carbamate 8 to methyl 11 and amine 4 mediated by an activating agent. The corres-
carbamate TAK-954. This inefficient protection/deprotection ponding diamine starting material was therefore changed
strategy was originally implemented to avoid using the methyl from methyl ester 1 to carboxylic acid 10.
carbamate analogue of 7, which was assessed to make an unac-
Initially, the first step, benzimidazole cyclization, was run
ceptable registrational starting material15 due to poor material as a slurry-to-slurry process due to the poor solubility of
Scheme 1 First-generation TAK-954 manufacturing sequence.
Green Chem.
This journal is © The Royal Society of Chemistry 2020

View Article Online
Green Chemistry
Paper
Scheme 2 Retrosynthetic analysis of TAK-954 showing proposed second-generation process in water.
diamine starting material 10 and benzimidazole product 11 in fite adduct exists in equilibrium with the free aldehyde in
water at neutral pH. However, these conditions led to the for- aqueous solution, an excess of sodium bisulfite (3 equiv.) was
mation of two regioisomeric impurities at levels as high as 20 required to push this equilibrium completely toward the bisul-
HPLC area %. The two impurities were identified as 15a and fite adduct. Next, we looked at controlling the relative amounts
15b (Fig. 1), resulting from bis addition of aldehyde 2 to of bisulfite adduct and diamine available to react during the
diamine 10 followed by cyclization and a 1,3-hydride shift. reaction. Under the initial reaction conditions, the bisulfite
Two factors were found to contribute to the formation of 15a adduct of 2 is fully dissolved while diamine 10 is poorly
and 15b: (1) free aldehyde 2 reacts with diamine 10 to exclu- soluble, meaning a large excess of bisulfite adduct is available
sively give the two bis addition impurities, and (2) the presence to react with diamine 10 in the solution phase throughout the
of an excess of the bisulfite adduct of 2 favors formation of the reaction, precisely the situation we need to avoid in order to
bis addition impurities. In order to suppress the formation of suppress the formation of the bis-addition impurities.
15a and 15b, we addressed each of these factors separately. Attempts to improve the solubility of diamine 10 through the
First, we optimized the amount of sodium bisulfite used to addition of a surfactant (2 wt% TPGS-750-M) failed to impact
preform the bisulfite adduct of 2 to ensure that diamine 10 the formation of the bis-addition impurities. However, the
was not exposed to any residual free aldehyde. Since the bisul- acid functionality of 10 and the aqueous reaction medium pro-
vided an opportunity to use pH to control solubility. Addition
of sodium hydroxide (2 equiv.) to the reaction mixture led to
full dissolution of acid 10. Finally, we implemented a slow
addition of a solution of the bisulfite adduct of 2 to the solu-
tion of 10 in aqueous hydroxide in order to maintain an excess
of diamine 10 throughout the reaction. These changes, sum-
marized in Fig. 1, resulted in suppression of 15a and 15b to
undetectable levels. Next, we implemented a direct isolation of
benzimidazole 11, again using pH to control solubility. Slow
addition of aqueous HCl to the product mixture over two
hours at room temperature led to crystallization of 11, which
was then isolated via filtration in 79% yield. The step PMI is
24.5, and no organic solvents or surfactants are used during
the reaction or isolation.
The next step, coupling of acid 11 and amine 4 to form
amide 5, requires an activating agent. Published examples of
Fig. 1 Benzimidazole cyclization in water (top). Process flow diagram
showing optimized conditions (bottom).
amide bond formation in aqueous media feature a variety of
This journal is © The Royal Society of Chemistry 2020
Green Chem.

View Article Online
Paper
Green Chemistry
different activating agents.8 Without a clear sense of which tion kinetics are so fast, this reactive precipitation process was
activating agent was best suited to the formation of amide 5, rapid and uncontrolled and initially resulted in product
we screened several different classes of activating agents, gumming and oiling, making isolation via filtration imposs-
including those that form acid chlorides, mixed anhydrides, ible. Rather than introducing an immiscible organic solvent to
activated esters, and other activated species (Table 1). extract the oiled-out product, we focused our efforts on devel-
Ultimately, conditions employing TCFH and NMI (Table 1, oping a controlled reactive crystallization to allow for direct
entry 9) proved to be most effective, giving nearly complete isolation of the product. TCFH was initially added in a single
conversion to amide 5 with faster reaction kinetics than the portion, but in order to reduce the degree of product supersa-
next most effective activating agent, DMTMM, while also turation over the course of the reaction we instead
offering superior atom economy and lower cost. TCFH/NMI implemented addition of TCFH in three equal portions,
conditions were developed by chemists at Bristol-Meyers waiting 15 minutes between each addition. Next, we intro-
Squibb, who originally identified acetonitrile as the optimal duced a small amount of THF co-solvent (15 vol%) to the reac-
solvent for this transformation.9 However, we found that for- tion mixture in order to modestly improve the product solubi-
mation of amide 5 using the published TCFH/NMI conditions lity, again with the goal of reducing the degree of product
proceeds cleanly and rapidly in water, provided that TCFH is supersaturation. Finally, experiments with and without surfac-
added to the reaction mixture last. We believe this is the first tant (2 wt% TPGS-750 M) revealed that the presence of surfac-
example of a TCFH/NMI mediated amide bond formation tant had no impact on the reaction profile, but inhibited the
under aqueous conditions. Both TCFH itself and the activated reactive crystallization, exacerbating product gumming.
acyl imidazolium intermediate would appear to be prone to Surfactant was therefore removed from the reaction mixture.
hydrolysis in an aqueous environment, but in the presence of With these changes, amide 5 was isolated in 91% yield as a
amine 4 the desired reaction pathway is evidently favored over crystalline solid by direct filtration once the reaction was com-
hydrolysis. The reaction is extremely fast at ambient tempera- plete. As an added benefit to conducting the isolation in
ture, typically reaching >95% conversion within ten minutes of aqueous media, water-soluble reaction by-products, including
TCFH addition. The starting materials, acid 11 and amine 4, tetramethyl urea and NMI salts, along with any residual start-
are fully soluble in the reaction mixture after addition of NMI. ing materials, 11 and 4, are rejected during filtration. The step
The amide product, however, is poorly soluble under the reac- PMI is 9.2. No surfactants and a minimal amount of organic
tion conditions and precipitates as it forms. Because the reac- solvent are used during the reaction and isolation.
Next, the Boc protecting group was removed from piper-
idine 5.10 Use of 3 equiv. of aqueous HCl led to complete de-
protection of 5 within 1 hour at 50 °C. 4 Volume % of aceto-
nitrile co-solvent was included to avoid product oiling during
isolation. Both starting material 5 and product 6 are fully
Table 1 Activating agent screen for amide coupling of intermediates 11
and 4a
soluble under the acidic reaction conditions. Once the de-
protection is complete, addition of aqueous sodium hydroxide
leads to crystallization of piperidine 6, which is directly iso-
lated via filtration in 88% yield. The step PMI is 24.0, and no
surfactant and a minimal amount of organic solvent is used.
With the first three steps of the synthesis complete, we
turned our attention to preparing aldehyde 14, which is
Entry Activating agent
Base
Conversion to 5b (%)
required for formation of TAK-954 via reductive amination
with piperidine 6. Because methyl carbamate 13 and aldehyde
14 are both water-soluble oils, we envisioned a telescoped reac-
tion sequence where hydroxymethylpiperidine 12 is carbamoy-
lated at the piperidine nitrogen followed by oxidation of the
primary alcohol to an aldehyde, all without intermediate iso-
lation. Finally, we hoped that conversion of aldehyde 14 to its
corresponding bisulfite adduct would allow for isolation of the
product as a crystalline solid, which could serve as a stable,
well-defined registrational starting material. Carbamoylation
of 12 proceeded smoothly and rapidly in water using potass-
ium carbonate as base and methyl chloroformate as the carba-
moylating reagent at ambient temperature. Surfactant and
organic co-solvent were tolerated but were not required due to
the high aqueous solubility of the reaction components and
products. The next step, oxidation of primary, aliphatic alcohol
13 to aldehyde 14 proved more challenging. TEMPO/bleach
1c
2
3
4
5
6
7
8
9
Thionyl chloride
Isobutyl chloroformate NMM
Cyanuric chloride
EDCI/HOBt
TNTU
EEDQ
COMU
DMTMM
TCFH
K2CO3
59
30
19
34
16
22
NMM
NMM
DIPEA
None
2,6-Lutidine 47
NMM
NMI
97
>98
EDCI: 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide; HOBt: 1-Hydroxy-
benzotriazole; TNTU: O-(5-Norbornene-2,3-dicarboximido)-N,N,N′,N′-
tetramethyluronium tetrafluoroborate; EEDQ: N-Ethoxycarbonyl-2-ethoxy-
1,2-dihydroquinoline; COMU: (1-Cyano-2-ethoxy-2-oxoethylidenaminooxy)
dimethylamino-morpholino-carbenium hexafluorophosphate; DMTMM:
4-(4,6-Dimethoxy-1,3,5-triazin-2-yl)-4-methylmorpholinium
chloride;
TCFH: Chloro-N,N,N′,N′-tetramethylformamidinium hexafluorophosphate;
NMM: 4-methylmorpholine; NMI: 1-Methylimidazole. a Reaction con-
ditions: 1.2 mmol 11, 1.2 mmol 4, 2.5 mL TPGS-750-M (2 wt% in H2O).
b HPLC area %. c The acid chloride of 11 was pre-formed in neat thionyl
chloride before addition to an aqueous solution of 4 and K2CO3.
Green Chem.
This journal is © The Royal Society of Chemistry 2020

View Article Online
Green Chemistry
Paper
Table 2 Optimization of aldehyde regeneration protocola
conditions led to a product mixture consisting of the desired
aldehyde, overoxidized carboxylic acid, and unreacted alcohol in
a roughly 1 : 1 : 1 ratio. Notably, we obtained similarly poor
selectivity when using anhydrous dichloromethane as solvent
for this transformation, suggesting that the poor selectivity is
related to the substrate and oxidation conditions rather than to
the aqueous reaction medium. Given the poor selectivity of the
TEMPO/bleach conditions, we screened a variety of other oxi-
dation conditions, including several metal-catalyzed aerobic oxi-
dation conditions, which, unfortunately, showed little to no
reactivity.‡ Ultimately, the conditions that gave the best conver-
sion and selectivity involved catalytic TEMPOL with diacetoxyio-
dobenzene as terminal oxidant, which cleanly converted alcohol
13 to aldehyde 14 with no overoxidation, even after prolonged
reaction times. Because diacetoxyiodobenzene is poorly water
soluble, 2 wt% TPGS-750 M and THF co-solvent were needed for
the oxidation to proceed. Once the oxidation was complete,
addition of sodium metabisulfite to the reaction mixture led to
formation of bisulfite adduct 16. However, due to its high solu-
bility in water, an organic antisolvent was required to crystallize
Method of aldehyde
Entry regeneration
Conversion to
Base
Full regeneration prior to addition NaOH
TAK-954b (%)
1
2
>99%
>99%
of other reaction components
In situ regeneration in the
presence of other reaction
components
(3 equiv.)
NaOAc
(2 equiv.)
3
In situ regeneration in the
presence of other reaction
components
None
>99%
a Reaction conditions: 0.17 mmol 6, 0.25 mmol bisulfite adduct of 14,
0.25 mmol α-picoline-borane, base as noted, 0.5 mL TPGS-750-M
(2 wt% in H2O), 0.1 mL MeOH, 60 °C, 8 h. b HPLC area %.
and isolate bisulfite adduct 16. Ethanol was selected for this 1). Exposure of bisulfite adduct 16 to aqueous sodium hydrox-
purpose both for its suitability as an antisolvent and for its ide led to full conversion to the parent aldehyde, which was
environmental and safety profile. Addition of ethanol led to then used as a solution without further manipulation in the
crystallization of bisulfite adduct 16, which was isolated via fil- subsequent reductive amination. Micellar media (2 wt%
tration. Because they share similar solubility profiles, several TPGS-750 M in water) and an organic cosolvent (20 vol%
reaction by-products, presumably including acetate salts, methanol) were used in the reductive amination both to
residual sodium bisulfite, and residual potassium carbonate, provide a lipophilic environment within the micelles where
tend to precipitate along with the bisulfite adduct product. As a imine formation would be favorable and to overcome the poor
result, the bisulfite adduct is isolated in 70–80% assay potency. aqueous solubility of piperidine 6 and TAK-954 at neutral to
The impurities making up the remaining 20–30% of the assay basic pH. α-Picoline-borane was selected as reducing agent for
value, however, are well tolerated in the next step. The three- its aqueous stability. Under these conditions, piperidine 6 and
step telescoped sequence delivers the bisulfite adduct in 38% a solution of aldehyde 14 in aqueous hydroxide reacted cleanly
assay corrected yield with a PMI of 27.1.
to form TAK-954. No reduction of aldehyde 14 was observed.
The final step of the manufacturing sequence involves the Next, we attempted to combine regeneration of the parent
reductive amination of piperidine 6 with bisulfite adduct 16 to aldehyde with the reductive amination (Table 2, entry 2).
produce TAK-954. In thinking about how to approach this Adding exogenous base (sodium acetate) to the reaction
transformation, we foresaw two challenges. First, reductive mixture allowed for the direct use of bisulfite adduct 16, giving
amination requires the loss of an equivalent of water during a reaction profile that was comparable to experiments where
imine formation, presenting
a potential thermodynamic the parent aldehyde was regenerated separately. However,
barrier to reaction progress in an aqueous environment.§ And while optimizing the amount of exogenous base, we found
second, we needed to devise a protocol that would allow for that the base could be omitted entirely without affecting the
use of the bisulfite adduct in the reductive amination, presum- progress or purity of the reductive amination (Table 2, entry 3).
ably by regenerating the parent aldehyde.¶ In order to investi- This result suggests that in an aqueous environment, the bisul-
gate the direct use of bisulfite adduct 16 in the reductive amin- fite adduct exists in equilibrium with the parent aldehyde,
ation, we initially chose to separate the regeneration of parent allowing the reductive amination to proceed without the need
aldehyde 14 from the reductive amination itself (Table 2, entry to fully regenerate the parent aldehyde before the start of the
reaction. This discovery made for an operationally simple reac-
tion protocol where the bisulfite adduct is used directly as an
aldehyde surrogate: after all reaction components are com-
‡See ESI† for details of alternative oxidation conditions that were screened.
§Literature examples of reductive aminations in aqueous media were limited
bined, the mixture is warmed to 50 °C and reaches full conver-
but encouraging. Alinezhad and coworkers reported several high-yielding reduc-
tive aminations in aqueous micellar media using sodium borohydride as redu- sion to TAK-954 within 8 hours.
cing agent and CTAB as surfactant,11 while Kikugawa and coworkers reported
The pH-dependent solubility of TAK-954 in water provided
“on water” reductive aminations of poorly soluble substrates using α-picoline-
borane as reducing agent.12
an opportunity to design a controlled crystallization directly
from the reductive amination product mixture. However, upon
addition of sodium hydroxide, TAK-954 oiled out of solution.
Although the product oil droplets eventually crystallized, a
¶Researchers at Amgen have developed reductive amination conditions in
methanol that allow for the direct use of bisulfite adducts by introducing
exogenous base to regenerate the parent aldehyde in situ.13
This journal is © The Royal Society of Chemistry 2020
Green Chem.

View Article Online
Paper
Green Chemistry
crystallization that proceeds through an oil phase risks solvent for organic chemistry due to the poor aqueous solubi-
entraining impurities and does not allow for control of the lity of many organic compounds, but the first three steps of
physical properties of the crystalline solid. Initial attempts to the TAK-954 sequence challenge this assumption. Each of
mitigate product oiling by seeding the product mixture and these steps is conducted entirely in water, and, aside from a
slowing the rate of hydroxide addition failed. Instead, we small amount of organic co-solvent in steps 2 and 3, no surfac-
implemented a reverse addition crystallization where the tants or other solubilizing additives are used, yet all three reac-
product mixture is slowly added to a TAK-954 seed bed in tions occur in the solution phase. Surfactant is used in sub-
aqueous hydroxide, leading to direct crystallization of TAK-954 sequent steps in order to overcome the limited aqueous solubi-
as a hydrate and avoiding product oiling. Reductive amination lity of certain reaction components, and, in the case of the
followed by direct isolation provides the TAK-954 hydrate in reductive amination, to overcome the thermodynamic limit-
91% yield with a step PMI of 21.7. However, the pharmaceuti- ations of imine formation in an aqueous environment.
cally desired crystal form of TAK-954 is an anhydrate. Earlier Throughout the manufacturing sequence, pH adjustments are
form screening and competitive slurry experiments indicated used to manipulate the solubility of intermediates and the API
that it is not possible to isolate the anhydrate form from in aqueous media, allowing reactions to occur in the solution
aqueous media. Therefore, a non-aqueous recrystallization was phase while providing a means to purify and isolate products
required to convert the TAK-954 hydrate to the desired anhy- through direct crystallization without the need for organic
drate form. The acetonitrile cooling crystallization developed solvent. As pharmaceutically relevant compounds often
for the first-generation route was used for this purpose, deli- contain a high proportion of heteroatoms and one or more
vering the anhydrate in 97% yield. This recrystallization rep- acidic or basic centers, manipulating pH to control the
resents the only operation carried out in non-aqueous media aqueous solubility of intermediates and APIs during drug
in the entire manufacturing sequence and was unavoidable manufacturing may be a generally applicable strategy. In the
given crystal form requirements.
step 2 amide bond formation, however, we took a different
approach to direct product isolation, implementing a reactive
crystallization, a strategy that may be applicable in other
Analysis of process improvements
The re-developed TAK-954 manufacturing process is shown in instances where starting materials are water soluble while the
Scheme 3. Water is commonly assumed to be an impractical product is not.
Scheme 3 Re-developed TAK-954 manufacturing sequence in water.
Green Chem.
This journal is © The Royal Society of Chemistry 2020

View Article Online
Green Chemistry
Paper
differences between the two routes, the individual impurities
present in the API produced using each route are slightly
different, but both routes deliver the API with an acceptable
overall purity of approximately 99.8 HPLC area %.
Conclusions
While a handful of examples of applications of aqueous reac-
tion media to API synthesis have been published in recent
years,14 we believe this is the first reported example of an API
manufacturing process conducted nearly exclusively in
aqueous media, including multiple controlled crystallizations
and isolations. We view this work as a proof of concept and
expect that its benefits will extend beyond improving the sus-
tainability of this particular drug. We hope to apply the tech-
niques and approaches developed here to other API manufac-
turing processes at Takeda in order to continue to reduce our
reliance on organic solvents.
Fig. 2 Comparison of PMI between the first-generation TAK-954
process in organic solvent and the re-developed, 4-step linear process
in water.
The four-step linear sequence to TAK-954 was successfully
demonstrated on 100 g scale, sufficient to satisfy current clini-
cal demand. However, we do not anticipate any barriers in
further scaling this process if necessary. Notably, the process
can be run using existing reactor trains and equipment found
in typical manufacturing plants. In addition to the environ-
mental, health, and safety benefits that come with replacing
organic solvents with water, comparison of the aqueous four-
step linear sequence to the organic-solvent-mediated first-
generation route reveals dramatic improvements in process
efficiency.∥ Overall yield was improved from 35% to 56%.
Reductions in process mass intensity are illustrated in Fig. 2.
Overall process mass intensity was reduced from 350 to 79,
representing a 77% reduction in the amount of materials
required to manufacture TAK-954 API. By shifting the reaction
and isolation media from organic solvents to water, the
portion of the process mass intensity attributable to organic
solvents was reduced from 223 to 14, representing a 94%
reduction in solvent use. Perhaps more surprisingly, the
portion of the process mass intensity attributable to water was
reduced from 106 to 55, meaning that the manufacturing
process in water uses 48% less water than the organic-solvent-
based process. Like many conventional API manufacturing
processes employing organic solvents as reaction media, the
first-generation TAK-954 process used large quantities of water
during product workups and isolations. By improving the
overall efficiency of the synthetic route and implementing
direct isolations for each intermediate, we were able to signifi-
cantly reduce water use, even as we shifted the bulk reaction
and isolation medium from organic solvents to water. Due to
Conflicts of interest
There are no conflicts to declare.
Acknowledgements
We would like to thank Justin Quon for his advice in develop-
ing a crystallization of the crude API. Izumi Takagi contributed
analytical support, and we are indebted to Ivan Dai for NMR
analysis and interpretation.
Notes and references
1 C. Jiménez-González, P. Poechlauer, Q. B. Broxterman,
B. Yang, D. am Ende, J. Baird, C. Bertsch, R. E. Hannah,
P. Dell’Orco, H. Noorman, S. Yee, R. Reintjens, A. Wells,
V. Massonneau and J. Manley, Org. Process Res. Dev., 2011,
15, 900.
2 J. H. Clark and S. J. Tavener, Org. Process Res. Dev., 2007,
11, 149.
3 B. H. Lipshutz, S. Ghorai, A. R. Abela, R. Moser,
T. Nishikata, C. Duplais, A. Krasovskiy, R. D. Gaston and
R. C. Gadwood, J. Org. Chem., 2011, 76, 4379; B. H. Lipshutz,
S. Ghorai and M. Cortes-Clerget, Chem. – Eur. J., 2018, 24,
6672; G. La Sorella, G. Strukul and A. Scarso, Green Chem.,
2015, 17, 644; A. Steven, Synthesis, 2019, 51, 2632;
S. Narayan, J. Muldoon, M. G. Finn, V. V. Fokin, H. C. Kolb
and K. B. Sharpless, Angew. Chem., Int. Ed., 2005, 44, 3275;
Y. Hayashi, Angew. Chem., Int. Ed., 2006, 45, 8103; A. Chanda
and V. V. Fokin, Chem. Rev., 2009, 109, 725; H. C. Hailes,
Org. Process Res. Dev., 2007, 11, 114; J. Brals, J. D. Smith,
F. Ibrahim, F. Gallou and S. Handa, ACS Catal., 2017, 7,
7245; J. D. Smith, T. N. Ansari, M. P. Andersson, D. Yadagiri,
F. Ibrahim, S. Liang, G. B. Hammond, F. Gallou and
∥Because the preparation of aldehyde 7 was not included in metrics calculations
for the first-generation route, the preparation of the bisulfite adduct of aldehyde
14 was excluded from metrics calculations for the route in water to allow for a
direct comparison of the two routes.
This journal is © The Royal Society of Chemistry 2020
Green Chem.

View Article Online
Paper
S. Handa, Green Chem., 2018, 20, 1784; T. N. Ansari,
Green Chemistry
G. A. Weisenburger, Tetrahedron Lett., 2017, 58, 4391;
C. M. Gabriel, M. Keener, F. Gallou and B. H. Lipshutz,
Org. Lett., 2015, 17, 3968; M. Parmentier, M. K. Wagner,
K. Magra and F. Gallou, Org. Process Res. Dev., 2016, 20,
1104; F. Gallou, P. Guo, M. Parmentier and J. Zhou, Org.
Process Res. Dev., 2016, 20, 1388; M. Shi, N. Ye, W. Chen,
H. Wang, C. Cheung, M. Parmentier, F. Gallou and B. Wu,
Org. Process Res. Dev., 2020, 24, 1543.
S. Handa and F. Gallou, Cross-Couplings in Water – A Better
Way to Assemble New Bonds, in Organometallic Chemistry in
Industry:
A Practical Approach, ed. T. J. Colacot and
C. C. C. J. Seechurn, Wiley, 2020, 203.
4 F. Gallou, N. A. Isley, A. Ganic, U. Onken and
M. Parmentier, Green Chem., 2016, 18, 14; M. Parmentier,
M. Wagner, R. Wickendick, M. Baenziger, A. Langlois and
F. Gallou, Org. Process Res. Dev., 2020, 24, 1536.
9 G. L. Beutner, I. S. Young, M. L. Davies, M. R. Hickey,
H. Park, J. M. Stevens and Q. Ye, Org. Lett., 2018, 20, 4218.
5 Organic co-solvents are known to improve the scalability of
micellar chemistry through a number of beneficial effects, 10 For
Boc-deprotections
in
aqueous
media
see:
including enlargement of the micellar core, improved solu-
bility of reaction components, and mitigation of product
C. Zinelaabidine, O. Souad, J. Zoubir, B. Malika and
A. Nour-Eddine, Int. J. Chem., 2012, 4, 73.
‘gumming’. For details, see: C. M. Gabriel, N. R. Lee, 11 H. Alinezhad, M. Tajbakhsh and F. Salehian, Monatshefte
F. Bigorne, P. Klumphu, M. Parmentier, F. Gallou and für Chemie, 2005, 136, 2029.
B. H. Lipshutz, Org. Lett., 2017, 19, 194; F. Gallou, P. Guo, 12 S. Sato, T. Sakamoto, E. Miyazawa and Y. Kikugawa,
M. Parmetier and J. Zhou, Org. Process Res. Dev., 2016, 20, Tetrahedron, 2004, 60, 7899.
1388; N. R. Lee, F. Gallou and B. H. Lipshutz, Org. Process 13 M. Faul, R. Larsen, A. Levinson, J. Tedrow and
Res. Dev., 2017, 21, 218; F. Gallou, P. Guo, M. Parmentier
and J. Zhou, Org. Process Res. Dev., 2016, 20, 1388.
6 B. Eren and Y. Bekdemir, Quim. Nova, 2014, 37, 643.
7 D. Prat, A. Wells, J. Hayler, H. Sneddon, R. McElroy,
F. Vounatsos, J. Org. Chem., 2013, 78, 1655.
14 B. S. Takale, R. R. Thakore, F. Y. Kong and B. H. Lipshutz,
Green Chem., 2019, 21, 6258; F. Gallou, N. A. Isley, A. Ganic,
U. Onken and M. Parmentier, Green Chem., 2016, 18, 14.
S. Abou-Shehada and P. J. Dunn, Green Chem., 2016, 18, 288. 15 Guidance for Industry: Q11 Development and Manufacture
8 M. Badland, R. Crook, B. Delayre, S. J. Fussell, I. Gladwell,
M. Hawksworth, R. M. Howard, R. Walton and
of Drug Substances. U.S. Dept. of Health and Human
Services, Food and Drug Administration. 2012.
Green Chem.
This journal is © The Royal Society of Chemistry 2020