Process development and scale-up of a pparγ agonist: Selection of the manufacture route

Article abstract of DOI:10.1021/op800211c

Two short, high-yielding routes to the selective PPARγ agonist GSK376501A were developed and carried out on scale. The key bond -forming reaction in each synthesis was the substitution of a 3,5-difluoro or 3,5-dibromo aryl intermediate with 2-methoxy-etha

Full text of DOI:10.1021/op800211c

Organic Process Research & Development 2009, 13, 303–309
Process Development and Scale-Up of a PPARγ Agonist: Selection of the
Manufacture Route
Kae M. Bullock, Delphilia Burton, John Corona, Ann Diederich, Bobby Glover, Kim Harvey, Mark B. Mitchell,
Mark D. Trone, Robert Yule, Yong Zhang, and Jennifer F. Toczko*
Chemical DeVelopment, GlaxoSmithKline, FiVe Moore DriVe, P.O. Box 13398, Research Triangle Park,
North Carolina 27709, U.S.A.
Scheme 1. GSK376501A retrosynthesis
Abstract:
Two short, high-yielding routes to the selective PPARγ agonist
GSK376501A were developed and carried out on scale. The key
bond -forming reaction in each synthesis was the substitution of a
3,5-difluoro or 3,5-dibromo aryl intermediate with 2-methoxy-
ethanol. A nucleophilic aromatic (SNAr) substitution reaction
under basic conditions was developed for the aryl fluoride
substrate. The 3,5-dibromo aryl halide intermediate required
copper-catalyzed conditions to achieve substitution of the second
bromide. In addition, a 2-methoxyethanol decomposition pathway
to generate methanol and ethylene oxide under basic reaction
conditions as well as its effect on impurity formation, was
elucidated. Both the difluoro and dibromo intermediates were
considered as the basis for the final route of manufacture. The
difluoro substrates were chosen due to straightforward chemistry,
controllable impurity profile, and ease of fluoride removal.
Introduction
indole-2-carboxylate (3) with a commercially available diha-
lobenzyl bromide (4a, 4b). The key transformation in the
synthesis is an aryl halide displacement on 2a or 2b with
2-methoxyethanol. The final step is the recrystallization of
intermediate grade 1 (IG-1) to generate final active pharma-
ceutical ingredient (API) (1).
Both potential manufacture route starting materials (4a and
4b) were carried through to final API on large scale in
comparable overall yield (82%) and with comparable raw
materials costs. The decision regarding which dihalo intermedi-
ates to utilize would be based on the development of efficient,
high-yielding conditions for the coupling reaction and aryl halide
displacement. Moreover, impurity generation, process simplicity,
and waste stream differences would need to be taken into
consideration.
The Stage 1 development work was initially performed using
difluoro starting material 4a, which was coupled to indole 3, a
synthesis of which has been carried out in our laboratories and
described previously4 (Scheme 2). Original reaction conditions
utilized Cs2CO3 in DMF at elevated temperature to generate
5a, followed by hydrolysis with 50% NaOH to yield 2a.
However, scale-up of these conditions highlighted potential
mixing issues, as it was difficult to suspend the solid Cs2CO3
with overhead stirring, resulting in inconsistent reaction times
and reaction stalling.
Activation of the nuclear receptor, peroxisome proliferator
activated receptor gamma (PPARγ), has been shown to enhance
insulin sensitivity and reduce plasma glucose levels, which have
made it an attractive target for treatment of type 2 diabetes
mellitus (T2DM). 1 Full agonists of the PPARγ receptor such
as rosiglitazone and pioglitazone have been widely used to treat
T2DM; however, potential side effects have limited their
efficacy in some patient populations. Data in preclinical animal
models suggest that selective PPARγ modulators (SPPARMγ)
may provide the enhanced insulin sensitivity required for
effective treatment of T2DM while avoiding or greatly dimin-
ishing the undesired side effects observed in some patients
treated with a PPARγ full agonist.2 GSK376501A (1) is a
SPPARMγ molecule in development in our laboratories.3
Scheme 1 illustrates a convergent route to 1 which begins
with the reaction of ethyl 3-[4-(1,1-dimethylethyl)phenyl]-1H-
Author to whom correspondence may be sent. E-mail: * jennifer.f.toczko@
gsk.com.
(1) (a) Kelly, D. E.; Killian, D. Curr. Opin. Endocrinol. Diabetes 1998,
5, 90. (b) Johnson, M. D.; Campbell, L. K.; Campbell, R. K. Ann.
Pharmacother. 1998, 32, 337. (c) Leutenegger, M.; Sacca, L.; Alderton,
C.; Eckland, D.; Lettis, S. Curr. Ther. Res. 1997, 58, 403.
(2) (a) Rosenstock, J.; American Diabetes Association Annual Meeting,
June 10-14, 2005, San Diego, CA,Abstract No. 44-OR. (b) Feldman,
P. L.; Lambert, M. H.; Henke, B. R. Curr. Top. Med. Chem. 2008, 8,
728.
(3) Spearing P. K.; Lambert, M. H.; Ray, J. A.; Laudeman, C. P.;
Szewczyk, J. R.; Banker, P. N-(Arylmethyl)-1H-indole-2-carboxylic
Acid Derivatives for Use As PPAR Modulators. PCT Int. Appl. WO
2008028118 A1 20080306, CAN 148:331538, 2008.
(4) Bullock, K. M.; Mitchell, M. B.; Toczko, J. F. Org. Process Res. DeV.
2008, 12, 896.
10.1021/op800211c CCC: $40.75
Published on Web 12/05/2008
2009 American Chemical Society
Vol. 13, No. 2, 2009 / Organic Process Research & Development
•
303

Scheme 2. Stage 1
conditions. While Stage 1a proceeded with good conversion in
numerous solvents, especially MeCN (entry 5), Stage 1b
proceeded well in only N,N-dimethylacetamide (DMAc) and
N-methylpyrrolidine (NMP), with NMP giving superior results
(entries 1 and 2). Scale-up of the NMP conditions (entry 6),
however, demonstrated that NMP remained in isolated 2a, (up
to 18 wt %) even when large volumes of water were used to
wash the solid. Therefore, the reaction in DMAc was examined
more closely, and an increase in the equivalents of 4a (from
1.0 to 1.2) resulted in improved reaction performance in DMAc
(Table 1, entry 7).
The Stage 1 product (2a) could be isolated from the reaction
mixture by precipitation upon the addition of water followed
by concentrated HCl. However, this resulted in a tendency for
2a to form large lumps that were difficult to remove from fixed
equipment. This issue was resolved by the slow addition of
reaction mixture to 6 N HCl with maximum stirring in the
reaction vessel.5 The optimized reaction conditions and reverse
addition were successfully carried out on 200-gal scale in 97%
yield, and 2a was generated as a free-flowing solid that could
easily be drained from the reactors.
Not surprisingly, 4b proved to have reactivity very similar
to that of 4a in Stage 1 (Scheme 2), and reaction conditions
and workup were developed to synthesize 2a successfully
generated 2b. Optimization work for the dibromo substrate
included an increase in the amount of KOH (2.5 equiv) to
achieve consistent hydrolysis reaction times and a solvent
change to NMP. NMP provided a better yield (97% with NMP
vs 84% with DMAc) and removal of NMP from the isolated
solid could be easily accomplished when either a normal or
reverse addition was used to precipitate the product. The
optimized reaction conditions performed well to generate 2b
on 50-L scale.
Table 1. Solvent screen results using KOtBu and KOH for
Stage 1
4a
5
2a
entry solvent equiv (HPLC area %) (HPLC area %)
1
2
3
4
5
6
7
DMAc
NMP
1.0
1.0
1.0
1.0
1.0
1.0
1.2
85
90
83
74
94
82
89
87
96
9
36
13
90
96
DME
MTBE
MeCN
NMPa
DMAc
The Stage 1 development and scale-up work demonstrated
that both 2a and 2b could be generated in an analogous manner.
Therefore, both were progressed into Stage 2 where the reaction
conditions required for the aryl halide displacements differenti-
ated the intermediates, and it was determined 2a had significant
advantages over 2b.
Stage 2. Synthesis of IG-1 via Difluoro Intermediate 2a.
Our approach to the synthesis of IG-1 via 2a involved an aryl
fluoride displacement under basic conditions with the potassium
alkoxide of 2-methoxyethanol (Scheme 3). As expected, the
displacement of one fluoride to generate 6a was facile since
the fluoride in the Ar-5 position activates the ring towards the
SNAr reaction. Displacement of the second fluoride, on the now
deactivated ring, proved to be more challenging; however,
reaction conditions for the displacement of aryl fluorides on
deactivated rings, similar to those previously developed in our
laboratories,6 could be employed on this system to generate
intermediate grade 1 (IG- 1) (Table 2).
a Scale-up of conditions identified in screen.
Therefore, alternate bases were considered, and KOH, which
provided 97% conversion to 2a, was initially identified as
promising. However, because of the potential for hydrolysis of
3 prior to the coupling reaction and the fact that the acid of 3
will not react with 4a, a better choice of base for Stage 1a
proved to be KOtBu. In DMF with KOtBu, the coupling was
complete in 2 h at room temperature. The advantage of using
KOtBu in place of high-molecular weight and relatively
expensive Cs2CO3 was a homogeneous reaction mixture, which
resulted in consistent, reproducible reaction times and obviated
the need for mixing studies that would have been required on
the heterogeneous reaction mixture if Cs2CO3 were to be used.
With optimized Stage 1a conditions in hand, we next turned
our attention to Stage 1b, saponification of ethyl ester 5a. A
switch from 50% NaOH to aqueous KOH (1.2 equiv in 2 vol
water at 60 °C) not only resulted in consistent Stage 1b reaction
times but also prevented the precipitation of 2a as a gummy
solid during the isolation, which was an issue when NaOH was
used.
First generation conditions to synthesize IG-1 via 2a
involved treatment of 2-methoxyethanol (20 equiv) with KOtBu
(18 equiv) in toluene. Starting material 2a was then added as a
solution in toluene and 1,3-dimethyl-3,4,5,6-tetrahydro-2(1H)-
The final Stage 1 variable examined was the effect of solvent
on reaction performance using the KOtBu/KOH base system
(Table 1). In all cases, 3 and KOtBu were premixed prior to
addition of 4a; which decomposes upon exposure to basic
(5) It was later discovered that the reaction could be performed in NMP,
and a reverse addition allowed for easy removal of NMP from the
isolated material.
(6) Kim, A.; Powers, J. D.; Toczko, J. F. J. Org. Chem. 2006, 71, 2170.
304
•
Vol. 13, No. 2, 2009 / Organic Process Research & Development

developed (Table 2, entry 2).7 However, a distillation step after
alkoxide generation and prior to addition of the 2a solution was
necessary to reduce the reaction volume in order to achieve
complete consumption of intermediate 6a at these lower
equivalents. Although these conditions accomplished the goal
of lowering alkoxide loading and were successful at 50-L scale,
the distillation step increased both complexity and the cycle
time of the process.
Scheme 3. Stage 2
Third-generation conditions eliminated the distillation and
reduced cycle times by reducing the toluene volume during the
alkoxide formation and slightly increasing alkoxide equivalents,
with complete conversion to IG-1 in 3 h (Table 2, entry 3).
Although the alkoxide loading was still rather high, it was
deemed that these conditions were a good compromise between
alkoxide loading versus cycle time and process simplicity.
The final issues in the reaction development were in the
workup and isolation. On pilot-plant scale, to avoid generation
of HF and potentially etching glass-lined reactors, fluoride levels
needed to be <20 ppm before the reaction mixture could be
acidified. Therefore, highly efficient and effective workup
conditions needed to be developed to remove the fluoride
generated during the reaction, and a series of brine washes on
the reaction mixture proved to be extremely effective. On
laboratory scale, three washes at room temperature brought the
fluoride in the organic layer to <0.5 ppm. The organic layer
could then be heated to 60 °C and neutralized with 6 N HCl,8
followed by addition of heptane to crystallize IG-1. Further
development work focused on minimizing emulsion issues
during the brine washes and ensuring that IG-1 did not
crystallize out of solution during the HCl wash. By increasing
the temperature of the brine washes to 60 °C and diluting the
organic layer with additional toluene prior to 6 N HCl addition,
both the emulsion and crystallization issues were addressed. A
final wash of IG-1 with iPrOH removed any residual DMPU
that was present in the isolated solid.9 Optimized reaction and
workup conditions were carried out on 100- and 200-gal scale
in 83-85% yield. 10
Table 2. Optimization of fluoride displacement reaction
alcohol
(equiv)
KOtBu
(equiv)
temp
(°C)
time
(h)
yield
(%)
entry
1
2
3
20
5
7
18
4
6
80
110
110
12
12
3
80
76a
83
a Volume of alkoxide formation reaction reduced by 10-20% prior to addition
of 2a solution.
Figure 1. Methoxide displacement impurity.
pyrimidinone (DMPU) to the alkoxide solution with additional
heating at 80 °C for 12 h (entry 1). The high alkoxide loading
and long cycle time for the displacement were deemed necessary
in order to keep the reaction temperature at 80 °C and still
achieve complete conversion. At the time, we thought the lower
temperature was required in order to minimize a byproduct
which was tentatively identified as decarboxylated 1. While not
ideal, the advantage of the first generation process was its
operational simplicity, and these conditions were successfully
carried out on 50-L scale. Further development efforts focused
on maintaining the operational simplicity while lowering the
alkoxide loading. First, however, the structure of the purported
decarboxylated impurity needed to be confirmed.
Mass spectrometric analysis and the synthesis of a marker
showed that the impurity was, in fact, not decarboxylated
product, but rather methoxy analog 7 (Figure 1) (Vide infra).
Since decarboxylation appeared not to be a major side reaction,
we could now explore higher reaction temperatures. By increas-
ing the reaction temperature, we reasoned that equivalents of
base and alcohol could be significantly lowered, and this was
indeed the case.
In the long term, the high volumes of brine required for
fluoride removal could be an issue in regard to waste stream
disposal; therefore, preliminary work, focused on reducing
fluoride levels by filtration of KF and/or CaF2 salts, was carried
out. In laboratory-scale experiments, it was demonstrated that
filtration of the reaction mixture, followed by treatment of the
filtrate with aqueous CaSO4 and filtration of the resulting CaF2
salts reduced fluoride levels in the organic layer to <20 ppm.
Acidification and isolation of IG-1 could be performed as
before.
Concurrent with reaction optimization work, experiments
were performed to identify the source of impurity 7. Under-
standing and controlling the factors that influence its formation
(7) Minimizing the amount of 2-methoxyethnaol used in the process was
important from a green chemistry perspective since the compound is
a tetratogen and reproductive hazard.
(8) Elevated temperature was required to keep IG-1 in solution during
the wash.
(9) 1H NMR was used to monitor DMPU levels in isolated product.
(10) In some batches, a fourth brine wash was required to meet the <20
ppm target. This was likely due to a rag layer at the organic/aqueous
interface during the phase splits that was retained in the reactor until
the final wash.
Second-generation conditions in the toluene/DMPU solvent
system with a dramatically lower alkoxide loading were
Vol. 13, No. 2, 2009 / Organic Process Research & Development
•
305

Scheme 4. Generation of Impurity 7
were particularly important since 7 had been identified in final
API synthesized by all processes at variable levels (0.3% to
1.1 area %).
The obvious first variable examined was the possibility that
MeOH was present in the 2-methoxyethanol starting material.
GC analysis of numerous 2-methoxyethanol batches showed
that, in some cases, small amounts of MeOH were present,
whereas in others none was detected. Surprisingly, when a
MeOH-free lot was used in Stage 2, 7 was still observed by
HPLC.
Based on these results, it seemed probable that MeOH was
being generated over the course of the reaction, likely through
the decomposition of 2-methoxyethanol (Scheme 4). Careful
GC monitoring of the reaction mixture confirmed that MeOH
is indeed produced during alkoxide formation. Additionally,
ethylene oxide, the other decomposition product of 2-meth-
oxyethanol, was detected using headspace GC/MS during
alkoxide generation. Further evidence that the methoxide was
generated from decomposition of 2-methoxymethanol was
provided by treatment of 2a with CD3OCD2CD2OD. Mass
spectrometric analysis indicated that 8 (Figure 1) was generated.
With a good understanding of the source of 7, we next turned
out attention to identifying the variables that control its
formation. Both reaction temperature and nitrogen sweep were
examined in a series of small-scale reactions at 80 and 110 °C,
both in sealed tubes and with a nitrogen sweep.
Figure 2. Potential impurities generated in Stage 2 with 2b
input.
These conditions provided IG-1 (74%) in addition to a des-
bromo impurity (9) (Figure 2). Although promising, the reaction
required high copper loading for good conversion, and a
significant amount of 9 (∼8 area %) was present. Additionally,
we wanted to telescope both aryl bromide displacements into a
single reaction, and when the above conditions were performed
using 2b as input material, incomplete conversion was observed
after 17 h of reaction time. These reaction conditions, however,
provided a good starting point with which to use design of
experiments (DoE) to quickly and efficiently optimize the
chemistry (Table 3).
At both temperatures 7 was generated; however, the sealed
experiments showed higher levels of impurity than those with
a nitrogen sweep (2-4 area % higher at 110 °C, and ∼1 area
% higher at 80 °C)11 The importance of nitrogen sweep was
further confirmed on pilot-plant scale where it was demon-
strated, with a static nitrogen blanket on the reactor headspace,
levels of 7 in isolated IG-1 were 1.7 area %. When performed
with a headspace nitrogen sweep, levels of 7 in isolated IG-1
were 0.13 to 0.65 area %.
Stage 2. Synthesis of IG-1 via Dibromo Intermediate 2b.
Our initial approach to the synthesis of IG-1 via intermediate
2b was to perform the aryl bromide displacement under basic
conditions similar to those employed for 2a. However, while
the displacement of the first bromide to generate 6b was facile,
conversion of 6b to IG-1 required very forcing conditions (20
equiv of alkoxide, toluene/DMPU, 65 h).12
The DoE studies started with a solvent and base screen, with
the goal of maximizing formation of IG-1 and minimizing 9.
Additionally, impurity 7 would be monitored since it was also
generated under the reaction conditions. Solvents and bases were
varied, while holding CuI (0.7 equiv) and 2-methoxyethanol
(10 equiv) constant. Starting material 2b was added to all
reactions as a solution in DMF, since previous experiments in
DME and DMAc demonstrated the copper-catalyzed reaction
would not proceed in good conversion without it. The results
of the screen indicated sodium tert-pentoxide was the best base,
and the best solvents were diglyme and iPrOAc, which showed
complete consumption of 2b with no 6b present (Table 3, entries
1 and 2). However, in iPrOAc, mass spectrometric analysis
indicated a side product (10) (Figure 2), resulting from addition
of isopropyl alkoxide to the aryl ring, was present. Because of
the undesired side product, we chose to progress the diglyme/
DMF solvent system into the next screen.
A second set of experiments was devised with the goal of
optimizing reagent loadings and investigating the reaction space
to determine if varying the reagent equivalents affected the
generation of 9. The amounts of 2-methoxyethanol, sodium tert-
pentoxide, CuI, and temperature were varied while holding
solvent (diglyme/DMF) constant. The DoE data, followed by
a small number of single experiments, identified the best
We therefore turned our attention to metal-catalyzed pro-
cesses to promote aryl bromide displacement on 6b. Initially,
the best procedure involved isolation of 6b, followed by a
copper-catalyzed reaction in DMF/DME using CuI (0.7 equiv),
KOtBu (5.5 equiv), and 2-methoxyethanol (9 equiv) at 100 °C.
(11) The reactions carried out at 80 °C did not achieve complete conversion.
(12) The displacement reaction was also evaluated on the corresponding
dichloro substrate. The reaction did not proceed to completion despite
high alkoxide loadings and long cycle times.
306
•
Vol. 13, No. 2, 2009 / Organic Process Research & Development

Table 3. Optimization of bromide displacement
sodium tert-
pentoxide
(equiv)
CuI
(equiv)
alcohol temp time
IG-1
9
10/11
entry
solvent
(equiv)
(°C)
(h)
(HPLC area %) (HPLC area %) (HPLC area %)
1
2
3
4
5
6
diglyme/DMF
iPrOAc/DMF
diglyme/DMF
iPrOAc
6
6
7
7
7
7
0.7
0.7
0.1
0.1
0.1
0.1
10
10
8
90
90
90
80
85
90
16
16
2
3.5
12
2
88.7
72.0
88.2
82.6
90.6
84.4
2.1
5.4
1.1
2.1
1.1
3.1
NA
107.7
NA
108.2
112.1
NA
8
tBuOAc
8
8
diglyme/DMFa
a Scale-up of optimized conditions.
conditions as sodium tert-pentoxide (7 equiv), CuI (0.1 equiv),13
and 2-methoxyethanol (8 equiv) (Table 3, entry 3). Moreover,
under all conditions screened, the level of 9 impurity was <2.5
area %, and impurity 7 was present at levels comparable to
those observed in the SNAr reaction of 2a.
With optimized reagent loadings in hand, a final set of
experiments was carried out using the conditions in a variety
of solvents without DMF. Interestingly, in the absence of DMF,
the only solvents that provided good conversion to IG-1 were
iPrOAc and tBuOAc.14 However, the corresponding alkoxide
addition products were also generated (Table 3, entries 4 and
5). While changing to a less water-miscible solvent was
attractive due to the expanded options for reaction workup, we
chose to advance the diglyme/DMF solvent system for scale-
up because it eliminated the potential for solvent addition
impurities (Table 3, entry 6).
Upon completion of the reaction in diglyme/DMF, IG-1 was
isolated by cooling to 80 °C followed by the addition of iPrOH
and water to dissolve the salt of the product, which had
precipitated out of solution during the reaction. The inorganic
solids were filtered off, and acidification of the filtrate resulted
in crystallization of IG-1. The optimized reaction conditions
and workup were carried out on 50-L scale in 75% yield.15
While our goals of optimizing the generation of IG-1 and
minimizing impurity 9 were accomplished and the process
performed well on scale, utilizing 2b in the manufacture route
had several drawbacks including variability in levels of impurity
9 and potential long-term issues with dissolved copper levels
in the waste stream.
Impurity 9 arises from reductive dehalogenation, a known
side reaction in many copper-catalyzed processes.16 With our
substrate, the amount of 9 produced is highly variable, depend-
ing on the choice of base and solvent. For example, when
NaOtBu is used in place of sodium tert-pentoxide in digylme/
DMF, very high levels of 9 are observed (27 area %). However,
in iPrOAc, 9 is present in ∼2 area % (in addition to solvent-
related impurity 11 in 13 area %) whether sodium tert-pentoxide
or NaOtBu is used as the base. Therefore, although the
formation of 9 appears to be controlled under the sodium tert-
pentoxide, DMF/diglyme conditions, if future work required a
solvent or base change, the impurity levels could also change.
This is a particularly important consideration since 9 is not
completely removed in the final recrystallization step.
In the reaction workup, >95% of the copper is filtered out
of the reaction as a solid; however, dissolved copper levels in
the waste stream would likely need to be lowered if the process
progresses into a manufacturing setting. While Stage 2 can be
carried out with lower CuI loadings, the resulting reaction rates
are slower. Slower reaction rates are not ideal since experiments
have indicated some variability in reaction time with the 0.1
equiv CuI loading, and in some cases the reaction required up
to 3 h to proceed to completion. Changing the workup will be
the most prudent solution to further reduce dissolved copper in
the waste stream; however, the majority of the copper is
precipitated from the solution using the current process, and
reducing levels further, while maintaining minimal workup unit
operations, will be difficult.
For these reasons, it was determined that the difluoro
intermediate 2a would be the optimal choice for incorporation
into the final route of manufacture, and IG-1, synthesized via
the fluoride displacement chemistry, was carried into the Stage
3 recrystallization studies.
Stage 3. Crystallization of IG-1. The Stage 3 recrystalli-
zation converts the intermediate grade 1 (IG-1) to final API
(1) and was incorporated into the synthesis to ensure control
over final API particle formation. Only a single anhydrous
crystal form of 1 had been identified; therefore, recrystallization
studies focused on throughput, mobility of the slurry upon
crystallization of 1, and the effect of impurities on the
crystallization.
A small-scale solvent screen identified five promising
solvents for the recrystallization: n-propanol, methyl isobutyl
ketone (MIBK), MeOH, toluene, and 2-butanone. Two-point
solubility data (Figure 3), volume of solvent required for
dissolution, and the calculated expected yield were obtained
(Table 4).
In the case of 2-butanone, reasonable yields could not be
expected; therefore, it was eliminated from further study.
Moreover, MeOH was eliminated due to the high number of
volumes required for dissolution. A scale-up experiment using
n-propanol required additional solvent volumes to facilitate
mobility of the thick slurry. Because the throughput would be
significantly decreased, n-propanol was also eliminated, and
further work focused on toluene and MIBK.
(13) The reaction could tolerate as little as 0.05 equiv of CuI, with a reaction
time of 4 h. The reaction also performs well with Cu(OAc)2 as the
catalyst.
(14) Toluene, THF, DMPU, 2-methoxyethanol, diglyme, NMP/diglyme all
proceeded in <55% conversion.
(15) Copper levels in isolated IG-1 were <1 ppm.
(16) (a) Lindley, J. Tetrahedron. 1984, 40, 1433. (b) Bacon, R. G. R.;
Rennison, S. C. J. Chem. Soc. (C) 1969, 308. (c) Bacon, R. G. R.;
Rennison, S. C. J. Chem. Soc. (C) 1969, 312. (d) Keegstra, M. A.;
Peters, T. H. A.; Brandsma, L. Tetrahedron. 1992, 48, 3633.
Solubility curves in toluene and MIBK were generated using
high purity 1 (>99 area %), and the data showed effective
Vol. 13, No. 2, 2009 / Organic Process Research & Development
•
307

Table 5. Effect of Impurities on Recrystallization Yield
IG-1 input: 96.9 area %
yield (%)
purity (area %)
toluene
MIBK
95
84
98.2
98.7
window, and yield were employed as when pure 1 was used.
As a result, in order to ensure a robust process regardless of
variation in the impurity profile of the IG-1, toluene was
selected as the solvent of choice for the crystallization.
Figure 3. Recrystallization. Two-point solubility data.
Conclusions
Table 4. Recrystallization screen results
Two efficient, high-yielding syntheses of GSK376501A were
developed and carried out on scale. The key step in both
syntheses was the development of conditions for the Stage 2
aryl halide displacement reaction, and the data collected during
development and scale-up were used to select the API manu-
facture route.
dissolution
volumes
temp range (°C)
yielda (%)
n-PrOH
MeOH
MIBK
2.0-68
1.5-49
1.9-68
2.2-78
22
61
15
16
98
85
89
98
toluene
The Stage 2 reaction conditions required for the conversion
of 2b into IG-1 had several drawbacks including control of
impurity profile and the presence of low levels of copper in
the waste stream which would be unacceptable in the long term.
For intermediate 2a, Stage 2 is a straightforward aryl fluoride
displacement that is operationally simple, highly reproducible,
and has performed well on a variety of scales ranging from 20
L to 200 gal. The workup is extremely effective in removing
the fluoride generated in the reaction, and the impurity profile
of 1 synthesized via 2a has been consistent and controllable
once the variables affecting the generation of impurity 7 were
understood. For these reasons, the difluoro intermediates were
selected for incorporation into the final route of manufacture.
a Calculated.
Figure 4. Solubility curves in toluene and MIBK using high
purity (>99 area %) 1.
Experimental Section
Recrystallization of 1-[(3,5-Bis{[2-(methyloxy)ethyl]oxy}-
phenyl)methyl]-3-[4-(1,1-dimethylethyl)phenyl]-1H-indole-
2-carboxylic Acid (1). A reaction vessel was charged with IG-1
(50.0 kg, 1.00 equiv) followed by toluene (630 kg, 14.5 vol)
and heated to 78-87 °C. When dissolution was observed, the
reaction mixture was cooled to 70-74 °C, and a slurry of IG-1
seeds in toluene (100 g of GSK376501A in 1.0 kg of toluene)
was added. The reaction mixture was stirred for 1 h and then
cooled to 0 °C over 2.5 h. The solids were filtered off, and the
cake was washed with cold toluene (43.5 kg, 1.00 vol × 2)
and dried under vacuum to give 49.0 kg (98% yield, 99.37 area
%) of 1.
dissolution at 80 °C could be achieved with 15 volumes of
toluene or 9 volumes of MIBK (Figure 4). Additionally, yields
of >95% upon isolation at 0 °C were expected under these
conditions, and both solvents gave a mobile slurry upon cooling
to 0 °C. Using these set solvent volumes, metastable zone widths
(MSZWs) for both solvents were approximated for these
conditions by fully dissolving the solid, then cooling until the
solids spontaneously nucleated. Both solvents had large MSZWs,
17 °C in toluene and 26 °C in MIBK, which indicated a wide
temperature range for seeding.
At this point, both solvents were still considered good
candidates for the final recrystallization, with MIBK at a small
advantage due to slightly better throughput. Yields for both
solvents using the best conditions were 94-96%, and experi-
ments to examine the effect of impurities present in IG-1 were
necessary to make the final decision.
A batch of IG-1 was selected and tested with both toluene
and MIBK (Table 5). It was observed that a drop in the purity
of IG-1 (from >99 to 96-97 area %) had a noticeable effect
on the solubility in MIBK, greatly increasing the solubility
throughout the temperature range. This caused the seeds to
dissolve, and the yield at 0 °C was much lower than the
expected at 95%. Solubility in toluene was not greatly affected
by the impurities, and the same dissolution volume, seed
1H NMR: δ 1.34 (s, 9H), 3.25 (s, 6H), 3.57-3.59 (t, 4H, J
) 4.5 Hz), 3.98-4.00 (t, 4H, J ) 4.0 Hz), 5.75 (s, 2H), 6.23
(s, 2H), 6.39 (s, 1H), 7.11-7.14 (t, 1H, J ) 7.5 Hz), 7.29-7.35
(ddd, 1H, J ) 0.9, 7.0, 8.3 Hz), 7.38-7.49 (m, 4H), 7.48-7.51
(m, 1H), 7.58-7.61 (d, 1H, J ) 8.5 Hz), 13.1 (s, 1H).
Preparation of 1-[(3,5-Bis{[2-(methyloxy)ethyl]oxy}-
phenyl)methyl]-3-[4-(1,1-dimethylethyl)phenyl]-1H-indole-
2-carboxylic Acid (IG-1). Fluoride Displacement. A reaction
vessel was charged with KOtBu (80.5 kg, 6.00 equiv), toluene
(130.5 kg, 3.00 vol), and 2-methoxyethanol (63.5 kg, 7.00
equiv). After the mixture was heated to 80 °C for 30 min, a
solution of 2a (50.0 kg, 1.00 equiv) in DMPU (106 kg, 2.00
vol) and toluene (22 kg, 0.50 vol) was added. The reaction
308
•
Vol. 13, No. 2, 2009 / Organic Process Research & Development

mixture was heated at 110 °C for 3 h and then cooled to 60
°C. Toluene was added to bring the reactor contents to ∼10
vol. The solution was then washed at 60 °C with 5% brine (250
kg × 3) followed by 6 N HCl (165 kg, 3.30 vol). Heptane (170
kg, 3.00 vol) was added to the organic layer, and the reaction
mixture was cooled to 0 °C over 1.5 h and stirred for several
hours. The solids were filtered off, and the cake was washed
with iPrOH (60 kg, 5.0 vol) and dried under vacuum to give
52.8 kg (83% yield, 99.14 area %) of IG-1.
Preparation of 1-[(3,5-Bis{[2-(methyloxy)ethyl]oxy}-
phenyl)methyl]-3-[4-(1,1-dimethylethyl)phenyl]-1H-indole-
2-carboxylic Acid (IG-1). Bromide Displacement. A reaction
vessel was charged with sodium tert-pentoxide (2.8 kg, 7.0
equiv), CuI (70 g, 0.10 equiv), and diglyme (8.0 L, 4.0 vol).
2-Methoxyethanol (2.2 kg, 8.0 equiv) was added, maintaining
the reaction temperature below 55 °C. The reaction mixture
was heated to 55 °C for 30 min, and a solution of 2b (2.0 kg,
1.0 equiv in 4.0 L, 2.0 vol DMF) was added. The slurry was
heated at 90 °C for 2 h and then cooled to <80 °C. Water (2.0
L, 1.0 vol) and iPrOH (10 L, 5.0 vol) were added and stirred at
60 °C for 30 min. The solids were filtered off, and the filtrate
was warmed to 60 °C. A solution of 6 N HCl (10 L, 5.0 vol)
was added, and the slurry was cooled to 20 °C over 1.5 h and
stirred overnight. The solids were filtered off, and the cake was
washed with a water/iPrOH mixture (10.6 L, 5.30 vol water,
5.4 L, 2.7 vol IPA) and dried under vacuum to give 1482 g
(76% yield, 96.7 area %) of IG-1.
reactor rinse followed by a direct cake wash with 116.4 L),
and dried under vacuum to give 48.9 kg (97% yield, 97.23 area
%) of 2a.
1H NMR: δ 1.34 (s, 9H), 5.86 (s, 2H), 6.77-6.78 (d, 2H, J
) 6.3 Hz), 7.10-7.17 (m, 2H), 7.33-7.36 (t, 1H, J ) 7.8 Hz),
7.40-7.49 (m, 4H), 7.49-7.51 (d, 1H, J ) 8.9 Hz), 7.59-7.61
(d, 1H, J ) 8.3 Hz), 13.0 (s, 1H).
Preparation of 1-[(3,5-Dibromophenyl)methyl]-3-[4-(1,1-
dimethylethyl)phenyl]-1H-indole-2-carboxylic Acid (2b). A
reaction vessel was charged with 3 (3.0 kg, 1.0 equiv), KOtBu
(1.26 kg, 1.20 equiv), and NMP (15 L, 5.0 vol) and stirred at
20 °C for 45 min. Add a solution of 4b (3.65 kg, 1.20 equiv)
in NMP (7.5 L, 2.5 vol) was added and the reaction mixture
was stirred at 20 °C for 1-2 h. After heating to 60 °C, aq.
KOH (1.31 kg, 2.50 equiv in 6 L, 2 vol water) was added,
maintaining the reaction temperature below 65 °C and the
reaction mixture was heated at 60 °C for 1 h. The reaction
mixture was added to 6 N HCl (13.8 L, 4.60 vol) at 60 °C over
1 h, and cooled to 20 °C. The solids were filtered off, and the
cake was washed with water (12 L, 4.0 vol then 18 L, 6.0 vol),
and dried under vacuum to give 5.65 kg (112% yield, 97.57
area %)18 of 2b.
1H NMR: δ 1.35 (s, 9H), 5.82 (s, 2H), 7.14-7.17 (t, 1H, J
) 7.1 Hz), 7.30-7.31 (d, 2H, J ) 1.6 Hz), 7.34-7.38 (t, 1H,
J ) 7.1 Hz), 7.39-7.49 (m, 4H), 7.50-7.52 (d, 1H, J ) 8.0
Hz), 7.62-7.64 (d, 1H, J ) 8.4 Hz), 7.73-7.74 (t, 1H, J )
1.8 Hz), 13.1 (s, 1H).
Preparation of 1-[(3,5-Difluorophenyl)methyl]-3-[4-(1,1-
dimethylethyl)phenyl]-1H-indole-2-carboxylic Acid (2a). A
reaction vessel was charged with DMAc (292 kg, 8.00 vol)17
followed by 3 (38.8 kg, 1.00 equiv), and the solution was added
to a second reactor containing KOtBu (16.3 kg, 1.20 equiv)
and stirred at 20 °C for at least 45 min. 3,5-Difluorobenzyl
bromide (4a) (29.9 kg, 1.20 equiv) was then added, and the
reaction mixture was stirred at 20 °C for 1.5-2 h. A solution
of KOH (12.2 kg, 1.80 equiv KOH in 79.3 L, 2.0 vol water)
was added slowly, maintaining a reaction temperature of 20
°C. The reaction mixture was heated to 60 °C for 1 h and then
slowly added to 6 N HCl (198 kg, 4.50 vol) at 60 °C. The
slurry was cooled to 20 °C and held at least 6 h. The solids
were filtered off, and the cake washed with water (77.6 L of
Acknowledgment
We thank David Black, John Grimes, Byron Johnson, Eric
Kraft, Mark Saulter, and Hamid Shafiei (GlaxoSmithKline,
Analytical Sciences) for NMR, mass spectrometry, ICP-OES,
and ion chromatography support.
Supporting Information Available
Characterization and purity data for compounds 1, IG-1, 2a,
2b. Characterization data for compounds 7, 8, 9, 10, 11. This
material is available free of charge via the Internet at
http://pubs.acs.org.
Received for review August 29, 2008.
OP800211C
(17) After DMAc addition, the batch was analyzed for water content. For
safety reasons, water levels needed to be <0.1% prior addition of the
solution to KOtBu. This also minimized the risk of hydrolysis of the
indole starting material.
(18) The high yield is likely due to inorganic solids that were isolated with
the product. This material was used in the subsequent stage on 50-L
scale which performed as expected. In small-scale experiments
contamination of the product inorganic solids was not observed.
Vol. 13, No. 2, 2009 / Organic Process Research & Development
•
309