Optimization of the manufacturing route to PF-610355 (1): Synthesis of intermediate 5

Article abstract of DOI:10.1021/op300341n

Tertiary carbinamine 5 is an isolated intermediate in the synthesis of a novel, inhaled β-2 adrenoreceptor agonist PF-610355. Process development for the key amide-formation and Ritter reactions, together with reaction understanding studies are discussed in context of the synthesis of 5. The optimized process employed to manufacture 140 kg of 5 is described, and was shown to have superior metrics to the preliminary commercial route.

Full text of DOI:10.1021/op300341n

Article
pubs.acs.org/OPRD
Optimization of the Manufacturing Route to PF-610355 (1): Synthesis
of Intermediate 5
Stephane Content, Thomas Dupont, Nicolas M. Fedou, Julian D. Smith, and Steven J. R. Twiddle*
́
́
Chemical Research and Development, Pfizer Global Research and Development, Ramsgate Road, Sandwich, Kent, U.K., CT13 9NJ
ABSTRACT: Tertiary carbinamine 5 is an isolated intermediate in the synthesis of a novel, inhaled β-2 adrenoreceptor agonist
PF-610355. Process development for the key amide-formation and Ritter reactions, together with reaction understanding studies
are discussed in context of the synthesis of 5. The optimized process employed to manufacture 140 kg of 5 is described, and was
shown to have superior metrics to the preliminary commercial route.
hydroxy-2-methylpropyl)phenyl]acetic acid 1 and the hydro-
chloride salt of 3′-(aminomethyl)biphenyl-4-ol 2 to give the
tertiary alcohol 3 which, after an aqueous workup, was
progressed as a solution into a Ritter reaction with
chloroacetonitrile. This allowed the substitution of the alcohol
function by an amine protected as the chloroacetamide.6 The
resulting product 4 was again not isolated but progressed, after
workup, through to the deprotection reaction as a solution in
acetic acid. Removal of the chloroacetamide with thiourea
generated the desired amine 5 which was isolated in a
reasonable 58% yield (average) and high purity.
1. INTRODUCTION
Respiratory diseases such as asthma and chronic obstructive
pulmonary disorder (COPD) are highly prevalent and affect
millions of people across the globe. Common treatments
include β-adrenoreceptor agonist inhalers, antimuscarinic
inhalers, steroid inhalers, or combination treatments. Current
marketed β-adrenoreceptor agonists salmeterol and formoterol
possess a duration of action which is adequate for twice-daily
administration.1 A key focus area has been the identification
and development of long acting β-2 adrenoreceptor agonists
(LABAs) which would provide a single daily dose through a dry
powder inhaler device. A summary of the discovery and
synthesis of a range of β-2 adrenoreceptor agonists culminating
in the identification of PF-610355 as a clinical candidate has
recently been published.1 The candidate progressed through
the development phases with publications documenting the
enabling chemistry2 and alternative route investigations.3 This
paper documents the development work performed to afford a
synthetic process suitable for commercial scale manufacture of
intermediate 5.
Although the preliminary commercial route had provided the
desired material of high purity on 35 kg scale, on commencing
development work a number of areas were identified as targets
for further optimization. First, the amide coupling between 1
and 2 utilized undesirable reagents N-ethyl-N′-(3-
dimethylaminopropyl)carbodiimide hydrochloride
(EDC·HCl)/hydroxybenzotriazole (HOBt)) and an undesir-
able solvent (dichloromethane) and involved a protracted
aqueous workup.7 Second, the Ritter reaction utilized a large
excess of the genotoxic chloroacetonitrile8 and trifluoroacetic
acid. The reaction workup was again protracted, employed
DCM and large volumes of concentrated aqueous ammonia,
and delivered a solution of 4 that still contained several
equivalents of chloroacetonitrile. As a result, a large excess of
thiourea was required to cleave the chloroacetamide group to
give the tertiary carbinamine 5. This afforded significant
quantities of solid waste products which were removed through
a slow and problematic filtration. Amine 5 was isolated from
the filtrate through a convoluted workup process utilizing
DCM, expensive 2-MeTHF, and concentrated aqueous
ammonia affording the point of highest dilution in the process
(83 L/kg wrt 1). From an environmental aspect the existing
synthetic process generated a large quantity of waste (E-Factor
= 302 [total waste {kg} per kg of 5]) and employed large
quantities of nonenvironmentally friendly reagents. Our goals
were therefore to streamline the process, eliminate/minimize
the use of undesirable reagents and solvents, and increase
throughput, particularly in the workup stages.
2. RESULTS AND DISCUSSION
2.1. Background. Upon the identification of PF-610355 as
a clinical candidate, enabling work performed led to the
development of a synthetic process which built upon the
medicinal chemistry route and developed effective syntheses of
key building blocks.4 Although this provided the required
quantity of material, the route was considered suboptimum for
commercial scale manufacture in terms of processing and yield
(36%). Process research investigations resulted in the
nomination of a preliminary commercial route in which the
API was constructed from three advanced starting materials.
This process was employed to generate 35 kg of PF-610355
(Scheme 1) to support early clinical development.3 As the
candidate progressed, continual development work was
performed on the synthetic process and is discussed herein.
This article focuses solely on the synthesis of 5, the first isolated
intermediate in the original process.
In the existing process, the synthesis of 5 was via a three-
reaction sequence in which intermediates were not isolated but
were carried through as solutions following workups. The
sequence commenced with the amide coupling of [3-(2-
Received: November 27, 2012
Published: February 5, 2013
© 2013 American Chemical Society
193
dx.doi.org/10.1021/op300341n | Org. Process Res. Dev. 2013, 17, 193−201

Organic Process Research & Development
Article
a
Scheme 1. PF-610355 preliminary commercial route
a
Reagents and conditions: (a) EDC·HCl, HOBt, Et3N, DCM, 20 °C then THF and 6 aq. washes. (b) ClCH2CN, TFA, 50 °C, then DCM and 4 aq.
washes. (c) (i) Thiourea, AcOH, 70 °C then byproduct filtration and 2-MeTHF and aq. washes (NH3); (ii) Crystallization from MeCN, 58%. (d)
NaHCO3, nBuOAc, 125 °C then EtOAc and aq. wash. (e) Et3N(HF)3, nBuOAc, EtOAc, MeOH, 20 °C then 2 aq. washes (NH3). (f) Fumaric acid,
nBuOAc, MEK, 82%. (g) MeCN, 50 °C, 95%. (h) (i) THF/water, 2 aq. washes (NH3); (ii) H2, Pd/C; (iii) Crystallization from MeCN, 77%. (i)
MeOH/water, 50 °C, 91%.
a
2.2. Development of the Amide Coupling. The first
change made to the amide coupling reaction was to use the
free-base form rather than the hydrochloride salt of 2 to avoid
the need for an amine base. This change was facilitated by the
identification of a new synthetic route to the free base
outsourced to an external vendor.9
Scheme 2. Optimized amide coupling
N,N-Carbonyldiimidazole (CDI) was rapidly identified as a
suitable coupling agent, which led us to concentrate on using
this reagent. This decision was derived from experience using
the reagent, knowledge of the typical side reactions, and the
opportunity to simplify the workup, as the imidazole byproduct
is much easier to purge than those generated through an
EDC·HCl mediated coupling.10,11 A range of solvents were
screened with a large proportion disregarded based on
solubility. The superior solvents were ethyl acetate and THF
with the former selected due to lower water miscibility
(facilitating the workup) and the fact a reaction concentration
of 5 L/kg with respect to 1 could be employed.
a
Reagents and conditions: (a) (i) CDI, EtOAc; (ii) 2, 50 °C then 2 aq.
washes.
reaction with the amine 2 (Scheme 3). However, we anticipated
the compounds 11 and 12 would react in the Ritter reaction in
Scheme 3. Reactions of 1 and CDI
The new set of reaction conditions allowed the design of an
aqueous workup which was superior to the protracted
procedure employed in the initial route. The aim was to, in a
minimum number of operations, remove the imidazole
byproduct and any of the unreacted starting materials. After
screening a number of approaches, a 2 M citric acid wash
followed by a 1 M sodium hydrogen carbonate wash was
implemented. The end product was a solution of the tertiary
alcohol 3 in ethyl acetate (Scheme 2).
There are potential side reactions associated with the use of
CDI which can lead to impurity formation. The desired
reaction is imidazolide 10 formation through reaction with the
carboxylic acid function of 1. In addition to the desired
transformation, the CDI can also react with the tertiary alcohol
function of 1 to give 1H-imidazole-1-carboxylate derivatives 11
and 12 which could potentially lead to carbamates upon
a similar way to 1 and the imidazolide alcohol 10 respectively.
Furthermore, if any unreacted CDI is present upon the addition
of 2, the urea would rapidly form and it was known that this
compound was difficult to purge in the downstream process.
Therefore, the CDI charge is critical for controlling the desired
194
dx.doi.org/10.1021/op300341n | Org. Process Res. Dev. 2013, 17, 193−201

Organic Process Research & Development
Article
pathway and hence minimizing impurity formation. It was
believed that the dual reactivity of 1 could provide an advantage
in terms of controlling the CDI charge through the use of
process analytical technology (PAT).
The in situ monitoring of the CDI activation of 1 was
investigated using mid-IR. Two absorption peaks for CDI
(1406−1389 cm−1 and 880−860 cm−1) could be followed to
assess the level present in the reaction. The portionwise
addition of a slight excess of CDI allowed differentiation of the
k1 and k4 reaction rates; the consumption of CDI slowed
resulting in a different decay profile in the mid-IR compared to
the substoichiometric charges (Figure 1). This is due to the
showed that the conditions did not generate any violent
exotherms associated with alternative systems such as sulfuric
acid.14 A protocol was therefore developed to allow the solvent
exchange of the solution of 3 in ethyl acetate to the high boiling
chloroacetonitrile (124−126 °C).
With the reagents for the Ritter reaction selected, attention
was focused on increasing the reaction understanding in order
to optimize the conditions used in the transformation. In doing
so there was a particular emphasis on reducing the quantities of
the chloroacetonitrile and TFA employed. The original reaction
conditions of 12 equiv of chloroacetonitrile and 25 equiv of
TFA at 50 °C were scrutinized using HPLC monitoring over
the course of the reaction to develop a better understanding of
the reaction profile, kinetics, and any transient species formed
(Chart 1).
Chart 1. Ritter reaction profile
Figure 1. Mid-IR profile of CDI consumption.
slower reaction of the tertiary alcohol function with CDI
becoming predominant following consumption of all the
carboxylic acid groups. This experimental data allowed kinetic
modeling using Dynochem12 which suggested that the rate
constants k1 and k3 ≫ k2 and k4, and it is these rate differences
that account for the desired product being formed under
normal conditions.
The change in shape of the mid-IR profiles on addition of the
portions of CDI can be used to give an indication of when an
excess of CDI has been added. This could allow maximization
of conversion when considering variability in CDI potency and
also avoid the need for a separate off-line reaction completion
test. On a 25 kg batch size in our pilot plant, the CDI was
added portionwise and acquired mid-IR data broadly reflected
the lab observations shown in Figure 1.
2.3. Development of the Ritter Reaction. In the
enabling process the conversion of the coupled product 3
had been performed with chloroacetonitrile (12 equiv) and
trifluoroacetic acid (25 equiv) at 50 °C, and we wished to
determine whether these conditions were optimum. A wide
range of reagents and reaction conditions were screened and
afforded the desired transformation.13 However, when
considering the toxicity of reagents such as sulfonic acids, the
requirement of a solvent such as acetic acid, and how the
reaction could be integrated into the telescoped process, it was
clear the alternative reaction conditions explored offered no
significant advantage. Because a solvent was required for the
Ritter reaction, the decision was made to focus on the
development and optimization of the TFA/chloroacetonitrile
conditions with chloroacetonitrile acting as both the reagent
and the solvent. Furthermore, process safety investigations
The consumption of the starting material was very rapid,
affording not only the desired product chloroacetamide 4 but
also two other species, the alkene 13 and the trifluoroacetate
ester 14. Note that styrene 13 undergoes the Ritter reaction
and that 14 appears unstable and eliminates to afford the
styrene. Scheme 4 shows the desired transformation following
the readily accepted SN1 mechanism; also depicted are a
number of pathways for which evidence was collated.
Protonation of the alcohol 3 and subsequent loss of water
lead to the formation of the tertiary carbocation 15. Following
the standard mechanism of nitrile trapping and addition of
water to the nitrillium ion, the desired chloroacetamide 4 is
produced. The styrene 13 is observed in the reaction and is
formed by elimination from either the alcohol or the derived
TFA ester and undergoes the Ritter reaction after protonation
of the double bond to afford 4 allowing conversion of the
transient styrene species into the desired product. This
mechanism raised the possibility of forming a regioisomer of
4 through chloroacetonitrile trapping of a benzylic carbocation;
however, even when starting from pure 13, investigations failed
to identify this potential impurity in the reaction mixture or
isolated material. The synthesis of the chloroacetamide from 13
raised an additional issue around the reaction mechanism, the
source of the oxygen atom of the product chloroacetamide.
Unlike alcohol 3, water, the most likely source, is not generated
on formation of the carbocation, and analysis of the reagents
suggested the water levels were too low to allow the observed
level of conversion. Evidence for an alternative or additional
source came from the identification of an impurity formed in
the reaction, 2-chloroacetamide 16, the hydrolysis product of
chloroacetonitrile. Further experiments suggested that the TFA
195
dx.doi.org/10.1021/op300341n | Org. Process Res. Dev. 2013, 17, 193−201

Organic Process Research & Development
Article
of 2 h to give a solution of the desired product 4 in
chloroacetonitrile and TFA. Attention then focused on the
optimum approach to progress the chloroacetamide 4 to the
target molecule 5.
Scheme 4. Ritter reaction mechanistic pathways
A major drawback of the preliminary commercial route was
the use of a large thiourea excess (4.2 equiv employed) in the
chloroacetamide deprotection step. This was necessary to
achieve completion of the desired transformation due to the
competing reaction of thiourea with the chloroacetonitrile
carried over from the Ritter reaction. This side reaction
afforded the potentially genotoxic16 diaminothiazole 18
(Scheme 5) which was removed, together with some of the
Scheme 5. Undesirable side reaction of chloroacetonitrile
and thiourea
pseudothiohydantoin 20 generated as a byproduct from the
desired reaction (Scheme 6), through a time-consuming
filtration upon completion of the reaction. This generated
approximately 3.9 kg of solid waste per kg of target compound
5, and so there was a significant waste stream to minimize.
Work therefore focused on the optimum method for the
removal of chloroacetonitrile from the system prior to the
deprotection stage in order to allow a reduced thiourea charge
and minimize waste generation.
At this point the optimum solvent for the deprotection
reaction had been identified as 2-butanol (see section 2.4.); the
first area of investigation was a solvent exchange of the
chloroacetonitrile/TFA solution into 2-butanol. Exploratory
practical investigations of this solvent exchange by distillation
demonstrated that TFA could be easily removed but complete
removal of chloroacetonitrile was not possible. Computer
modeling-based prediction of the distillation characteristics for
this solvent system was employed to help identify a more
efficient process. Initially binary parameters for each solvent
pair were studied along with a ternary diagram for the trisolvent
mixture (Figure 2).5 Note that T-x (blue) and T-y (green) are
the liquid and vapour phase boundaries respectively.
The diagrams illustrate that the TFA pairings are the most
problematic because of a maximum boiling azeotrope near the
end of TFA removal from both 2-butanol and chloroacetoni-
trile. This is further demonstrated in the ternary map for the
combination of all three solvents where the contour lines all
tend toward the maximum boiling azeotrope of TFA and
chloroacetonitrile. Furthermore, because chloroacetonitrile has
a higher boiling point than 2-butanol, and the relative volatility
between them is low, the distillation is inefficient heading
towards the 2-butanol boiling point. The results suggested that
the most efficient approach to maximize the removal of
chloroacetonitrile was to remove as much TFA as possible in an
initial concentration by distillation and then to remove the
chloroacetonitrile by successively adding 2-butanol. Models
were constructed and compared to evaluate the concentration−
dilution−concentration and constant volume distillation
modes; however neither offered a process that was of the
efficiency desired, and so an alternative protocol was sought.
reacts with the chloroacetonitrile to form acylating species 17.
This can react with a range of nucleophiles such as TFA to
form the anhydride, the tertiary alcohol 3, to afford the
observed transient ester species 14 or the phenolic oxygen of
the R group, all of which liberate 2-chloroacetamide. This
mechanism could account for the hydrolysis of the nitrillium
ion along the desired pathway which affords compound 4 when
starting from either the alcohol 3 or alkene 13.15
Building on the initial findings further work was performed
to understand how the telescoped nature of the process
influenced the reaction. It was found that the presence of water
reduced the reaction rate, presumably by shifting the
equilibrium of carbocation formation. This supported the
selection of ethyl acetate over THF as the amide coupling
solvent, since it would minimize the potential carryover of
water into the solution of 3 in chloroacetonitrile.
The focus was then turned toward the optimization of the
reagent quantities. By considering the accepted SN1 mechanism
for the Ritter reaction, where the rate should depend solely on
the concentration of 3, it was confirmed that the reduction of
the chloroacetonitrile charge from 12 to 6 equiv was possible
with a negligible effect on the rate of the Ritter reaction for a
given charge of TFA. Importantly, this reduced charge still
maintained stirrable solutions in chloroacetonitrile’s second
role as a solvent. Reducing the TFA charge had a more
profound effect, slowing the reaction and resulting in increased
levels of the alkene and TFA adduct intermediates during the
course of the reaction. Although the reduction in reagent
charges resulted in a slower reaction (completion after 2 h,
previously 1 h), the advantage of reducing their quantities with
only a relatively small time extension was a significant overall
improvement to the process. Concentrating the reaction with
respect to the chloroacetonitrile charge appeared to be limited
by the solubility of 3 and also the volume required to ensure an
efficient solvent exchange from the solution of 3 in EtOAc from
the amide coupling stage. From the investigation the conditions
selected for the Ritter reaction were to use chloroacetonitrile (6
equiv) and TFA (16 equiv) at 50 °C affording a reaction time
196
dx.doi.org/10.1021/op300341n | Org. Process Res. Dev. 2013, 17, 193−201

Organic Process Research & Development
Article
Scheme 6. Chloroacetamide 4 deprotection
Figure 2. Distillation phase diagrams for chloroacetonitrile, TFA, and 2-butanol.
Although early stage development of the route indicated that
4 was not crystalline, a solid form was discovered during later
studies. This presented the opportunity to purge chloroacetoni-
trile into the mother liquors resulting from a crystallization
process. A solubility screen was completed for 4 which
identified 2-MeTHF, tert-butyl methyl ether (TBME), and 2-
butanol as potential crystallization solvents. More importantly
the screen showed that, due to the impact its presence had on
increasing solubility, removal of TFA from the system was
necessary to achieve crystallization with an acceptable level of
recovery. First, an effective method of TFA removal was
required. From the distillation data (vide supra) it was
concluded this would be best achieved by simply concentrating
the product mixture by distillation upon completion of the
Ritter reaction. To achieve this, the distillation had to be
performed at reduced pressure so as to maintain the contents
below 50 °C to avoid degradation of the product that was
known to occur when such mixtures were subjected to higher
temperatures for extended periods of time.
Whilst, 2-MeTHF failed to give acceptable recoveries and
TBME afforded several different solvated forms, preliminary
studies showed that a highly crystalline form of 4 could be
obtained from the chloroacetonitrile/2-butanol solvent system,
although crystallization required an extended time period to
occur. In order to maximize yield and expedite nucleation by
increasing the level of supersaturation initially in the system,
further investigation of the 2-butanol crystallization was made
in conjunction with an antisolvent, cyclohexane. When
investigating the process further it was known that after the
TFA distillation and dilution with 2-butanol the solution was
supersaturated, and so the solution was held at 60 °C until self-
nucleation occurred. The isolated 4 was granular and dense and
was filtered very rapidly and easily dried. PXRD studies showed
the material to be highly crystalline.
The crystallization was optimized further to maximize the
yield through the addition of cyclohexane antisolvent without
compromising the process throughput or purity. To avoid
reliance on self-nucleation, the highly crystalline material so-
formed was used to seed the subsequent crystallizations after
dilution with 2-butanol. The optimized process gave a
consistent yield in the laboratory of 85−88% of high purity 4
The crystallization of the chloroacetamide 4 was then
investigated by the addition of the potential crystallization
solvents noted previously for the chloroacetonitrile solution.
197
dx.doi.org/10.1021/op300341n | Org. Process Res. Dev. 2013, 17, 193−201

Organic Process Research & Development
Article
Scheme 7. Amphoteric behavior of 5
and solved the problem of chloroacetonitrile carryover into the
next stage of the process. Whilst the introduction of an isolation
stage into a previously telescoped process could be viewed as a
step backwards in terms of cost and process efficiency, the
advantages that it gave in terms of process robustness,
throughput, and product quality significantly outweighed
these perceived disadvantages and resulted in its inclusion for
scale-up.
base of 5 and to maximize the purge of the pseudothiohy-
dantoin 20, which is a potentially genotoxic compound16 and
hence needed to be controlled to low levels.
The salt break of the acetate salt of 5 was not straightforward
since this compound is amphoteric by virtue of the structure
containing both an amine and a phenol group. The salt break
was investigated with 2 M sodium hydroxide, using 1H NMR to
determine the optimum pH window for generating the neutral
species. As can be seen in Scheme 7, the desired range was pH
9.6−12.2 with a target for the process set at pH 10.8 at 25 °C.
This evaluation and pH target selection was later validated
through further experimentation and Dynochem modeling.12
The model, based on estimated pKa values, identified that pH
10.6 was optimum to achieve the optimum ratio of free base.
However, based on the equilibria the model suggested that a
maximum of only 95% of 5 could exist in the desired free base
form.
Although conditions for the salt break had been identified,
the workup procedure was complicated by the presence of the
reaction byproduct pseudothiohydantoin 20. This byproduct
exhibited some water solubility which provided encouragement
that the desired purge could be achieved; however it
precipitated on cooling the reaction mixture to 25 °C, and
solids remained during the salt break and phase separation. To
overcome this problem, two modifications were made to the
workup. First the reaction solvent system was changed from
neat 2-butanol to 2-butanol/water whilst maintaining the same
concentration. This change did not significantly impact the
reaction rate, but it prevented precipitation of the pseudothio-
hydantoin on cooling the reaction mixture. Second the aqueous
workup was performed at 40 °C to prevent solids from
precipitating during the aqueous washes and so maximized the
partition of 20 into the aqueous phase. Two further water
washes ensured pseudothiohydantoin purged to acceptable
levels.
The isolation developed in the preliminary commercial route
involved a cooling crystallization from acetonitrile following a
solvent exchange from the workup solvent. After evaluation of
alternative strategies for the isolation of 5, such as formation
and crystallization of a salt, crystallization after azeotropic
drying of a solution, and cooling crystallizations with a range of
solvents, the acetonitrile approach was retained. Therefore, the
process that needed to be developed for the crystallization was
an exchange from 2-butanol to acetonitrile before seeding to
initiate crystallization.
Crystallization and solubility studies were completed to
identify the optimum acetonitrile/2-butanol composition from
which to isolate 5 based on yield and purity. This was shown to
be 90/10% w/w acetonitrile/2-butanol. Solutions richer in
acetonitrile were found to be less effective for purging
impurities, although yields were slightly higher. Further
investigation indicated that the presence of water had a
negative impact on nucleation by increasing the metastable
zone width. Above a water level of 5% w/w, self-nucleation
2.4. Development of the Deprotection Reaction. In the
initial exploratory commercial route the use of a solution of 4
resulted in a number of species, impurities, residual solvents,
and excess reagents, such as chloroacetonitrile, entering the
deprotection reaction step. The conditions employed for the
desired reaction required an excess of thiourea (4.2 equiv) in
neat acetic acid (11 L/kg wrt 1) at 70 °C, conditions that were
not ideal for large scale manufacture.3 This was in addition to
the problematic workup involving filtration of the solid side-
product 18 (diaminothiazole) and byproduct 20 (pseudothio-
hydantoin) as discussed earlier. The isolation of 4 as a
crystalline solid eliminated the problem associated with the
carryover of chloroacetonitrile, and this allowed the thiourea
charge to be lowered to 1.2 equiv and simplified the workup
considerably through the minimization of waste. There was also
a strong desire to avoid the use of acetic acid as the solvent
because managing this solvent contributed to the high volume
of the reaction workup. In order to develop a more efficient
process for the deprotection reaction, gaining a better
understanding of the reaction and workup was the initial focus.
The desired transformation is depicted in Scheme 6 and
illustrates the intermediacy of the thiourea adduct 18 and the
formation of the pseudothiohydantoin 20 byproduct that needs
to be purged during the workup. A literature review for
chloroacetamide deprotection reactions highlighted the opti-
mum solvent system as being an ethanol/acetic acid 4:1 volume
ratio.6 Initial investigations confirmed the use of alcoholic
solvents provided an increased reaction rate over acetic acid for
the initial nucleophilic attack of thiourea on 4. However, the
reactions stalled at the thiourea adduct stage, 19, unless acetic
acid was present in the reaction mixture, in which case the
desired product 5 was formed as the acetate salt. Significantly,
the laboratory data confirmed that acetic acid could be
employed as a reagent rather than solvent to guarantee
complete reaction conversion.
Attention was then focused on the identification of the
optimum solvent. The reaction could be successfully performed
in a range of solvents, primarily alcohols but also THF, 2-
MeTHF, and MEK. However, the solvent also had to provide
solubility for both the acetate salt and free base of 5 as well as a
good phase separation during an aqueous workup. After
significant experimentation and evaluation, 2-butanol was
selected since it was the solvent that most closely matched
the requirements. With the solvent selected and very different
reaction conditions developed, the reaction workup required
redesigning. The key goals were to efficiently generate the free
198
dx.doi.org/10.1021/op300341n | Org. Process Res. Dev. 2013, 17, 193−201

Organic Process Research & Development
Article
would not occur on cooling. This was important since the
solution of 5 in 2-butanol generated following the water-
washing stage contained high and variable levels of water (44−
59% w/w). Thus the water content had to be controlled and
reduced as much as possible to achieve the target solvent
composition for crystallization. The best opportunity to remove
water was through distillation in the initial concentration
operation of the solvent exchange process, in order to utilize
the water/2-butanol azeotrope. Analysis of the 2-butanol/water
phase diagram showed that a water content greater than 28%
w/w would, during any initial concentration, result in the vapor
phase actually containing a greater composition of 2-butanol
effectively increasing the water content of the bulk solution
(Figure 3).
the pilot plant for the four batches and consequently filtered
and dried rapidly on 50 kg scale. The isolated intermediate 4
was progressed as four separate batches through the
deprotection stage with no deviations from the expected
process to afford 140 kg of the amine 5 in excellent yield (84−
88%) and again with high reproducibility. The overall yield for
the synthesis of 5 from 1 was 74%, an increase of 16% on the
58% obtained from the previous process. The purity of the
isolated product was extremely high with a maximum total
impurity level in any batch of 0.16% and was suitable for
progression into the downstream process for the manufacture
of the PF-610355 API.
3. CONCLUSION
A new process for the manufacture of 5 has been developed
that overcomes many of the issues associated with the original
process. Process optimization focused on elimination/mini-
mization of the use of undesirable materials, reproducibility,
time reduction, throughput increase, waste minimization, and
yield optimization. The key change made to the process was the
introduction of the isolation of the chloroacetamide 4. The
seeded crystallization of 4 from 2-butanol/chloroacetonitrile
allowed the isolation of a highly crystalline, granular form that
provided a number of advantages. Of primary significance was
the purge of chloroacetonitrile which prevented carryover to
the deprotection reaction and avoided the production of 3.9 kg
per kg of 5 of side-product waste. This in turn allowed the
reduction of the maximum volume of the step by 61% from 83
L/kg to 32 L/kg (wrt 1) through simplification of the
chloroacetamide deprotection step and workup. This dramatic
reduction in volume swing allowed each step to be performed
in a single reactor. The rapid filtration of the intermediate, and
the streamlining of the process to reduce unnecessary unit
operations, afforded a 33% reduction in cycle time from 180 to
120 h. All of the improvements contributed to an overall yield
increase from 58% to 74% and vastly superior green metrics.
The E-factor (total waste [kg] per kg of 5) was reduced from
302 to 91 with a significant proportion of the reduction
attributed to the optimization of the aqueous workups
(aqueous waste was reduced by 100 kg per kg of 5). The
new process was demonstrated on scale to be robust, leading to
the manufacture of 140 kg of 5.
Figure 3. Phase diagram for 2-butanol/water.
Thus it was essential to ensure that the water content of the
initial 2-butanol solution was below the level of 28% w/w. The
simplest and most efficient way to achieve this was to partially
dry the organic phase by the introduction of a brine wash as the
final step of the aqueous workup. This guaranteed the water
content to be below the 28% w/w target, allowing its efficient
removal through the solvent exchange process, to finally
achieve the seeding solvent composition of approximately
80:20% w/w acetonitrile/2-butanol necessary for crystallization.
This solution was seeded (1% w/w) to induce a controlled
crystallization prior to dilution with additional acetonitrile
antisolvent to achieve the desired 90:10% w/w isolation solvent
composition and maximize the yield and purity on cooling.
Across numerous trial runs in the laboratory the maximum
water content observed at the seeding point was 0.2% w/w
affording 5 in a typical yield of 85% (Scheme 8).
2.5. Transfer to the Pilot Plant. The process, developed
as outlined as above, was transferred to the pilot plant to
generate 140 kg of 5. Four batches of 4 were produced in the
pilot plant with no deviations from the laboratory process,
producing 203 kg with a reproducible 86−88% yield. The
crystallization process of the highly crystalline, granular form of
chloroacetamide 4 observed in the laboratory was replicated in
4. EXPERIMENTAL SECTION
General. The starting materials 1 and 2 were synthesized by
external vendors following procedures developed and supplied
by Pfizer. All reagents and solvents were used as received from
1
commercial suppliers. H NMR was performed using either a
Bruker Avance III spectrometer at 600 MHz or a Bruker
Ultrashield Plus spectrometer at 400 MHz. 13C NMR was
performed using either a Bruker Avance III spectrometer at 150
MHz or a Bruker Ultrashield Plus spectrometer at 100 MHz.
a
Scheme 8. Optimized deprotection process
a
Reagents and conditions: (a) (i) Thiourea, acetic acid, 2-butanol, water, reflux; (ii) 2 M NaOH, 2 water washes and a brine wash; (iii)
Crystallization from MeCN, 85%.
199
dx.doi.org/10.1021/op300341n | Org. Process Res. Dev. 2013, 17, 193−201

Organic Process Research & Development
Article
Mass spectrometry was performed on a Bruker MaXis QTOF
under positive ion conditions. Combustion analyses were
performed by Warwick Analytical Service, University of
Warwick Science Park, The Venture Centre, Sir William
Lyons Road, Coventry, UK, CV4 7EZ.
resulting slurry was stirred at 70 °C for 1 h. The slurry was
diluted with additional acetonitrile (275 L) before cooling to 20
°C at 0.3 °C/min where it was left to stir for 4 h. The slurry was
diluted by the addition of acetonitrile (121 L). The solid was
collected by filtration, washed with acetonitrile (2 × 121 L),
and dried in vacuo at 50 °C to give 5 (39.8 kg, 86.7% yield
−85.7% when corrected for H2O content) as a white solid. Mp
2-Chloro-N-{2-[3-(2-{[(4′-hydroxybiphenyl-3-yl)-
methyl]amino}-2-oxoethyl)phenyl]-1,1-dimethylethyl}-
acetamide 4. [3-(2-Hydroxy-2-methylpropyl)phenyl]acetic
acid 1 (30 kg, 144 mol) was dissolved in ethyl acetate (150
L) at 25 °C. 1,1′-Carbonyldiimidazole (24.1 kg, 144 mol) was
added to the solution, and the reaction was left to proceed for 1
h. 3′-(Aminomethyl)biphenyl-4-ol 2 (31.6 kg, 159 mol) was
added, and the resulting slurry was heated to 50 °C. After 3 h
the reaction mixture was cooled 20 °C and was washed with 2
M aqueous citric acid (158 L) and 1 M aqueous sodium
bicarbonate solution (150 L). The organic layer was diluted
with chloroacetonitrile (56.1 L), and the solution was distilled
at atmospheric pressure; the end point was when the internal
temperature reached 110 °C. The chloroacetonitrile solution
was cooled to 50 °C, and trifluoroacetic acid (168 L) was
charged over 90 min. The reaction was allowed to proceed at
this temperature for an additional 2 h. The solution was
subjected to a reduced pressure (−0.950 barg) to allow the
distillation of trifluoroacetic acid, maintaining a temperature of
50 °C. Upon completion of the distillation, 2-butanol (225 L)
was added to the reaction and the solution was warmed to 60
°C. Seed crystals of 4 (0.67 kg, 1% w/w based on 100%
conversion) were added, and the slurry was held at 60 °C for 1
h. Cyclohexane (150 L) was added whilst maintaining the 60
°C temperature. The slurry was allowed to cool from 60 to 20
°C at a rate of 0.5 °C/min and stirred at 20 °C for 4 h. The
solid was collected by filtration, washed with cyclohexane (2 ×
90 L), and dried in vacuo at 45 °C to give 4 (58.7 kg, 87.6%
1
144 °C. H NMR (400 MHz, d6-DMSO) δ 0.96 (s, 6H), 2.54
(s, 2H), 3.48 (s, 2H), 4.33 (d, J = 4 Hz, 2H), 6.85 (d, J = 8 Hz,
2H), 7.06 (d, J = 8 Hz, 1H), 7.11−7.23 (m, 4H), 7.32 (t, J = 8
Hz, 1H), 7.38−7.44 (m, 4H), and 8.57 (bs, 1H). 13C NMR
(100 MHz, d6-DMSO) δ 30.0, 42.2, 42.5, 49.7, 50.5, 115.7,
124.3, 124.7, 125.2, 126.5, 127.5, 127.6, 128.4, 128.7, 130.7,
131.0, 135.8, 138.8, 139.9, 140.2, 157.3, and 170.2. HRMS
(ESI): calcd (M + H)+ 389.2224, found 389.2223. Anal. Calcd
for C25H28N2O2·0.25 H2O: C, 76.50; H, 7.33; N, 7.13. Found:
C, 76.44; H, 7.18; N, 7.15.
AUTHOR INFORMATION
■
Corresponding Author
*E-mail: Steven.Twiddle@pfizer.com.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank Stuart Field and Phil Peach for screening studies;
Wilfried Hoffman for Dynochem modeling; Eric Cordi for
distillation modeling; Mike Hawksworth for process safety
assessments; Andrew Fowler and Tom Moran for scale-up
support; Neil McDowall and May-Ling Yeow for PAT support;
and James MacGregor, Sam Morris, Rutjit Durve, Alex Paget,
and Griff Read for analytical support.
1
yield) as a white solid. Mp 157−159 °C. H NMR (600 MHz,
d6-DMSO) δ 1.22 (s, 6H), 2.96 (s, 2H), 3.48 (s, 2H), 4.00 (s,
2H), 4.34 (d, J = 5.9 Hz, 2H), 6.87 (m, 2H), 7.00 (dt, J = 7.5
and 1.5 Hz, 1H), 7.06 (t, J = 1.5 Hz, 1H), 7.15 (d, J = 7.7 Hz,
3H), 7.17 (d, J = 7.5 Hz, 1H), 7.21 (t, J = 7.5 Hz, 1H), 7.34 (t, J
= 7.7 Hz, 1H), 7.42 (m, 4H), 7.60 (s, 1H), 8.58 (t, J = 5.9 Hz,
1H) and 9.56 (br s, 1H). 13C NMR (150 MHz, d6-DMSO) δ
26.7, 42.3, 42.4, 43.3, 43.4, 53.5, 115.7, 124.4, 124.8, 125.3,
126.8, 127.6, 127.7, 128.5, 128.8, 130.8, 131.2, 135.9, 137.8,
139.9, 140.3, 157.2, 165.5, and 170.2. HRMS (ESI): calcd (M +
H)+ 465.1939, found 465.1950. Anal. Calcd for C27H29ClN2O3:
C, 69.74; H, 6.29; N, 6.02. Found: C, 69.64; H, 6.31; N, 5.99.
2-[3-(2-Amino-2-methylpropyl)-phenyl]-N-[(4′-hy-
droxybiphenyl-3-yl)methyl]acetamide 5. 2-Chloro-N-{2-
[3-(2-{[(4′-hydroxybiphenyl-3-yl)methyl]amino}-2-oxoethyl)-
phenyl]-1,1-dimethylethyl}acetamide 4 (55 kg, 118 mol) was
added to a stirred solution of 2-butanol (248 L) and water (275
L) at 20 °C forming a slurry. Acetic acid (14.9 L, 260 mol) and
thiourea (10.8 kg, 142 mol) were added, and the reaction was
heated at reflux. After 5 h the solution was cooled to 40 °C.
REFERENCES
■
(1) Glossop, P. A.; Lane, C. A. L.; Price, D. A.; Bunnage, M. E.;
Lewthwaite, R. A.; James, K.; Brown, A. D.; Yeadon, M.; Perros-
Huguet, C.; Trevethick, M. A.; Clarke, N. P.; Webster, R.; Jones, R.
M.; Burrows, J. L.; Feeder, N.; Taylor, S. C. J.; Spence, F. J. J. Med.
Chem. 2010, 53, 6640.
(2) de Koning, P. D.; Gladwell, I. R.; Moses, I. B.; Panesar, M. S.;
Pettman, A. J.; Thomson, N. M. Org. Process Res. Dev. 2011, 15, 1247.
(3) de Koning, P. D.; Castro, N.; Gladwell, I. R.; Morrison, N. A.;
Moses, I. B.; Panesar, M. S.; Pettman, A. J.; Thomson, N. M. Org.
Process Res. Dev. 2011, 15, 1256.
(4) Synthesis of compounds 1, 2, and 6 can be found in ref 2 and:
Gladwell, I.; De Koning, P. D.; Moses, I. B.; Pettman, A. J.; Thomson,
N. M. Process for the Preparation of Sulfonamide Derivatives, WO 2007/
010356.
(5) Jirgensons, A.; Kauss, V.; Kalvinish, I.; Gold, M. R. Synthesis 2000,
12, 1709.
(6) Urben, P. G., Ed. Brethericks Handbook of Reactive Chemical
Hazards, 6th ed.; Butterworth-Heineman: Oxford, Boston, 1999; p
746.
(7) (a) Daniel, F. B.; Schenk, K. M.; Mattox, J. K.; Lin, E. L. C.; Haas,
D. L.; Pereira, M. A. Fundam. Appl. Toxicol. 1986, 6, 447. (b) Muller-
Pillet, V.; Joyeux, M.; Ambroise, D.; Hartemann, P. Environ. Mol.
Mutagen. 2000, 36, 52.
(8) The new synthetic route to 2 involved a Suzuki coupling reaction
between 4-bromophenol and (3-cyanophenyl)boronic acid followed
by reduction of the nitrile group by catalytic hydrogenation and
isolation of the free-base form.
(9) Dale, D. J.; Dunn, P. J.; Golightly, C.; Hughes, M. L.; Levett, P.
C.; Pearce, A. K.; Searle, P. M.; Ward, G.; Wood, A. S. Org. Process Res.
Dev. 2000, 4, 17.
The pH was adjusted to 10.8
0.2 (at 25 °C) using 2 M
NaOH, and the layers were then separated. The organic layer
was diluted with additional 2-butanol (165 L) before washing
with water (2 × 248 L) and brine (248 L). The organic layer
was distilled down to a volume of 143 L at atmospheric
pressure. Acetonitrile (242 L) was then added, and the solution
was distilled down to a volume of 180 L. Acetonitrile (198 L)
was then added. The solution was cooled to 70 °C at which
point seed crystals of the title compound were added (0.4 kg,
1% based on 100% yield of the title compound), and the
200
dx.doi.org/10.1021/op300341n | Org. Process Res. Dev. 2013, 17, 193−201

Organic Process Research & Development
Article
(10) Bradley, P. A.; Carroll, R. J.; Lecouturier, Y. C.; Moore, R.;
Noeureuil, P.; Patel, B.; Snow, J.; Wheeler, S. Org. Process Res. Dev.
2010, 14, 1326.
(11) Dynochem 2011, Scale-up Systems, www.scale-up.com, 2011.
(12) (a) Krimen, L. I.; Cota, D. J. Org. React. 1969, 17, 213.
(b) Yadav, J. S.; Subba Reddy, B. V.; Pandurangam, T.; Jayasudan
Reddy, Y.; Gupta, M. K. Catal. Commun. 2008, 9, 1297. (c) Salehi, P.;
Khodaei, M. M.; Zolfigol, M. A.; Keyvan, A. Synth. Commun. 2001, 31,
1947. (d) Polshettiwar, V.; Varma, R. S. Tetrahedron Lett. 2008, 49,
2661. (e) Firouzabadi, H.; Iranpoor, N.; Khoshnood, A. Catal.
Commun. 2008, 9, 529. (f) Callens, E.; Burton, A. J.; Barrett, A. G. M.
Tetrahedron Lett. 2006, 47, 8699. (g) Sanz, R.; Martinez, A.; Guilarte,
V.; Alvarez-Gutierrez, J. M.; Rodriguez, F. Eur. J. Org. Chem. 2007,
4642.
(13) (a) Fugate, W. O.; D’Errico, M. J. U.S. Patent 3,151,157, 1964.
(b) Chang, S.-J. Org. Process Res. Dev. 1999, 3, 232.
(14) The alternative mechanism has also been postulated in the
literature. (a) Glickmans, G.; Torck, B.; Hellin, M.; Coussement, F.
Bull. Soc. Chim. Fr. 1966, 4, 1376. (b) Roberts, S. W.; Shaw, S. M.;
Milne, J. E.; Cohen, D. E.; Tvetan, J. T.; Tomaskevitch, J., Jr.; Thiel, O.
R. Org. Process Re. Dev. 2012, 16, 2058.
(15) By analogy with aminothiazole when existing in the aromatic
form: McCarren, P.; Berbenitz, G. R.; Gedeck, P.; Glowienke, S.;
Grondine, M. S.; Kirman, L. C.; Klickstein, J.; Schuster, H. F.;
Whitehead, L. Bioorg. Med. Chem. 2011, 19, 3174. Thiourea, 18, and
20 were analyzed and controlled to trace levels at intermediate 5.
(16) Note that the binary parameters were estimated from
COSMOtherm calculations.
201
dx.doi.org/10.1021/op300341n | Org. Process Res. Dev. 2013, 17, 193−201