Preparation of the HIV Attachment Inhibitor BMS-663068. Part 3. Mechanistic Studies Enable a Scale-Independent Friedel-Crafts Acylation

Article abstract of DOI:10.1021/acs.oprd.7b00115

During the development of a Friedel-Crafts acylation for the preparation of a key pyrrole intermediate in the synthesis of the HIV attachment inhibitor, BMS-663068-03, a significant scale dependence was found. A precipitous drop in yield was observed for

Full text of DOI:10.1021/acs.oprd.7b00115

Article
pubs.acs.org/OPRD
Preparation of the HIV Attachment Inhibitor BMS-663068. Part 3.
Mechanistic Studies Enable a Scale-Independent Friedel−Crafts
Acylation
Gregory L. Beutner,* Jacob Albrecht, Junying Fan, Dayne Fanfair, Michael J. Lawler, Michael Bultman,
Ke Chen, Sabrina Ivy, Richard L. Schild, Jonathan C. Tripp, Saravanababu Murugesan,
Konstantinos Dambalas, Douglas D. McLeod, Jason T. Sweeney, Martin D. Eastgate,
and David A. Conlon
Chemical & Synthetic Development, Bristol-Myers Squibb Company, One Squibb Drive, New Brunswick, New Jersey, 08903-0191,
United States
*
S Supporting Information
ABSTRACT: During the development of a Friedel−Crafts acylation for the preparation of a key pyrrole intermediate in the
synthesis of the HIV attachment inhibitor, BMS-663068-03, a significant scale dependence was found. A precipitous drop in yield
was observed for the acylation of a protected pyrrole with chloroacetyl chloride upon scale-up. Spectroscopic studies to mitigate
this scale dependence led to the identification of the complex effect of dissolved hydrogen chloride (HCl) as well as the poor
reactivity of the acylating agent, chloroacetyl chloride. At this point, the counterintuitive choice to switch to a longer, but scale-
independent, three-step route was made. By changing the acylating agent to acetyl chloride, a more robust process was obtained.
Rapid development of a high yielding α-chlorination then provided the common α-chloroketone intermediate required to
generate the desired α-amide ketopyrrole. The improved yield and scalability of this three-step process supported the addition of
one linear step to the route, and it was demonstrated successfully at scale.
INTRODUCTION
RESULTS AND DISCUSSION
■
■
The practical synthesis of substituted pyrroles, prominent
heterocycles in pharmaceutical products, is limited at large
Step 1: Friedel−Crafts Acylation of Pyrrole 1. The
1
Friedel−Crafts reaction between 1 and chloroacetyl chloride
(
ClCH COCl) is known to give moderate yields and high
2
6
scale due to the lack of availability of simple pyrrole building
blocks. Two main strategies are available to prepare these
compounds, namely, (a) the de novo synthesis of the pyrrole
ring, and (b) the selective functionalization of the parent,
unsubstituted pyrrole. Although numerous strategies exist for
de novo construction of pyrroles with a variety of substitution
C3:C2 selectivity (Scheme 1). In our hands at 100 g scale, the
reaction was homogeneous and took 4−6 h to reach full
conversion with no significant exotherm (adiabatic temperature
increase <10 °C). Attempts to optimize the process revealed
that the order and rate of reagent addition had a significant
impact on yield and selectivity. The optimal results were found
2
patterns, the functionalization of low-cost, readily available
by first charging AlCl and chloroacetyl chloride to a reactor
3
pyrroles remains an attractive alternative. However, due to the
containing CH Cl and then rapidly adding a solution of
2
2
3
pyrrole 1 in CH Cl . The reaction required 5 h to reach full
fact that pyrroles are ambident nucleophiles, their selective
2
2
4
conversion and afforded 2 in 74% yield and 16:1 C3:C2
selectivity (Table 1, entry 1). The remaining 21% of the input
pyrrole 1 could not be accounted for. All other permutations of
this charge sequence resulted in lower and variable yields. For
functionalization still presents major challenges. In light of
these limitations we set out to prepare 3-substituted pyrrole 5
(Scheme 1). The 3-aminoacetyl group in 5 was a key functional
group handle for further elaboration into the pyrrolopyridine
core of the target molecule, BMS-663068. Initial plans to
synthesize 5 started with readily available N-benzenesulfonyl
pyrrole 1 (Besyl, Bs = C H SO −). We envisioned that
example, a 5% yield was obtained when a mixture of AlCl and
3
ClCH COCl in CH Cl was charged to a solution of pyrrole 1
2
2
2
in CH Cl (entry 2).
2
2
6
5
2
The rate of addition of pyrrole 1 to the mixture of AlCl and
3
Friedel−Crafts acylation of 1 with chloroacetyl chloride
followed by amidation with the sodium salt of N-tosylforma-
ClCH COCl was also critical for achieving good yields and
2
selectivity. Thus, charging the solution of 1 over 180 min
resulted in a decreased yield and selectivity relative to faster
5
mide (4) would provide efficient access to 5. However, in the
course of this work, we realized that this two-step protocol
presented major concerns for scale-up. Herein, we describe
efforts to develop a high-yielding, scale-independent synthesis
of 5 as well as lessons learned about balancing empirical and
mechanistic methods to accelerate the development process.
Special Issue: From Invention to Commercial Process Definition:
The Story of the HIV Attachment Inhibitor BMS-663068
Received: March 24, 2017
©
XXXX American Chemical Society
A
DOI: 10.1021/acs.oprd.7b00115
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 1. Synthesis of Pyrrole 5 en Route to BMS-663068
Table 1. Conditions for Friedel−Crafts Acylation of Pyrrole
1
were not entirely reproducible, especially as the scale of the
reaction increased. Initial attempts at kilogram scale led to
yields far below 50% with the remainder of the material being
unaccounted for. Since the Friedel−Crafts acylation of 1 with
ClCH COCl is a relatively slow, homogeneous reaction
2
without a significant exotherm, the usual causes of scale
dependent performance, such as mass-transfer limitations or
poor heat transfer, did not appear to be at issue here. A Design
of Experiments (DoE) study was undertaken to investigate the
effect of temperature, reactor headspace, and excess of
ClCH COCl relative to AlCl on reaction yield and selectivity.
b
addition time
(min)
yield
(%)
2:3
a
b
entry
addition order
1 to AlCl /ClCH COCl
ratio
2
3
1
2
3
1
1
74
5
16:1
ND
3
2
The inclusion of temperature and reactor headspace as main
factors was deemed important since these are common causes
AlCl /ClCH COCl to 1
3
2
1 to AlCl /ClCH COCl
180
50
2.3:1
3
2
7
of issues on scale-up. Analysis of the DoE data revealed that
a
Reactions run at 16 L/kg DCM relative to 1 with 1.1 equiv of AlCl
and 1.2 equiv of ClCH COCl at 20 °C. Yield and C3:C2 ratio
3
b
reaction yield was strongly dependent on the charges of
2
ClCH COCl and AlCl but insensitive to temperature and
determined by quantitative HPLC measurements.
2
3
headspace.
Guided by the findings of the DoE and the observed
importance of the order and rates of reagent charges, the
reaction was run under conditions that would maximize yield
(1.2 equiv of AlCl and 1.4 equiv of ClCH COCl), minimize
additions (compare entries 3 and 1). Again, full conversion was
observed, but mass recovery was poor. This was common in
many reactions of ClCH COCl with 1 in our hands.
2
Although we identified conditions that could provide
3
2
acceptable yield and selectivity at 100 g scale, the results
the volume of the solution charge of pyrrole 1 (2.0 L/kg), add
Figure 1. In situ IR monitoring of the reaction of pyrrole 1 and ClCH COCl.
2
B
DOI: 10.1021/acs.oprd.7b00115
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
the solution of 1 as rapidly as possible subsurface (<15 min),
and ensure that the agitation rate was always set to maximum.
Despite these changes, the yield decreased from 71% at a 120
kg scale to 35% at a 450 kg scale, even with identical addition
times for the solution of pyrrole 1. These results suggested that
there were additional scale-dependent factors that had a
8
significant impact on the performance of this reaction. At
this point, further optimization by DoE could be performed,
but unless this unknown factor was included in the design, it
seemed unlikely that empirical optimization would provide a
path forward to a more robust process. Therefore, we turned to
mechanistic studies to better define the reaction pathway and
perhaps elucidate the root cause of the scale dependence.
Monitoring the reaction by in situ IR revealed some
intriguing trends (Figure 1). Upon mixing ClCH COCl with
2
9
AlCl in 1,2-dichloroethane (DCE), a slow reaction occurred
3
that led to the dissolution of AlCl . This phenomenon
3
corresponded to the disappearance of the free ClCH COCl
2
−
1
(
1817 cm ) and the formation of a new signal assigned as the
−1
AlCl complex of ClCH COCl (i, 1683 cm ). No signals that
3
2
Figure 2. Quantitative HPLC monitoring of reaction conversion and
mass recovery.
−1
could be attributed to an acylium ion at ∼2200 cm were
1
0
observed. The addition of 1 to this homogeneous solution led
to the instantaneous disappearance of the complex i and
the headspace of the reactor was aggressively swept with dry N2
to purge HCl. This resulted in complete consumption of
pyrrole 1, but <5% yield of the desired product 2! Conversely,
performing the reaction in a sealed vessel with DCE that had
been saturated with anhydrous HCl completely inhibited both
regeneration of free ClCH COCl. No free pyrrole 1 was
2
−1
observed at 1182 cm , but rather a slightly shifted signal
−1
appeared at 1126 cm . Control experiments performed where
was mixed with AlCl showed a similar pattern with a major
1
3
−
1
absorbance at 1126 cm , suggesting the formation of an AlCl -
3
the desired reaction and the decomposition of 1. Mixing AlCl3
complex of the pyrrole ii. The formation of product 2 could be
−1
and pyrrole 1 in a saturated solution of anhydrous HCl in
tracked through the growth of a band at 1573 cm , which was
11
DCE did not lead to formation of complex ii but gave a
different, unidentified species evidenced by a characteristic IR
band at 1200 cm− that was indefinitely stable at RT. The
dramatic difference in reaction performance observed between
these two extremes attested to the strong effect of HCl.
In fact, these experiments support the hypothesis that HCl
may be the scale-dependent factor which was missing from the
initial DoE optimization. To further assess the influence of HCl
and its interaction with the other variables, a four-factor
consistent with the AlCl complex iii. The identity of this signal
3
was confirmed by observation of an analogous spectrum when
1
isolated 2 was mixed with AlCl . It was also observed by IR and
3
HPLC that this complex iii was stable for prolonged periods of
time, indicating that the poor mass recovery observed in the
reaction was not a consequence of poor product stability under
the reaction conditions.
The fact that the addition of pyrrole 1 leads to a rapid and
complete consumption of complex i without the formation of
product 2 suggests that binding of AlCl to ClCH COCl is
12
Definitive Screening Design was conceived to ascertain the
3
2
relative importance of the reagent charges (AlCl3 equiv,
fairly weak compared to binding with 1. We wanted to
determine if the desired reaction occurs through the AlCl3
complex of the pyrrole ii or through a typical Friedel−Crafts
13
ClCH COCl equiv, CH Cl volume, and HCl equiv ). An
2
2
2
analysis of the results identified HCl as the key factor in
controlling yield (Figure 3). An unexpected nonlinear depend-
ence of yield on HCl amount suggested that at low conversion
dissolved HCl could be beneficial, whereas at higher
mechanism via the AlCl -activated complex i. To assess this
3
question, we attempted to prepare complex ii and study its
reactivity. However, it was observed that the mixture of 1 and
AlCl was unstable and decomposed rapidly to afford pyrrole
3
oligomers and polymers as observed by LC-MS. Based on this
finding, we hypothesized that the low mass recovery of the
reaction could be due to the rapid formation of the unstable
complex ii, which undergoes unproductive polymerization at a
rate competitive with the desired Friedel−Crafts acylation over
the long reaction times.
Monitoring of the reaction by quantitative HPLC revealed
that a significant portion of this unproductive reaction occurred
at low conversion (Figure 2). Considering the balanced
equation for the Friedel−Crafts reaction, we wondered what
role HCl could play in the decomposition process. If HCl
present in the reaction affects the stability of 1, it could help
explain the variability at scale. For example, the method of
Figure 3. Impact of dissolved HCl and AlCl equivalents on reaction
3
charging moisture-sensitive AlCl could lead to increased HCl
3
performance. Observed assay yields (%, boxed values) and regressed
predictions (contour) revealed a nonlinear relationship between HCl
and AlCl3.
levels due to uncontrolled hydrolysis. To test the role of HCl
on reaction performance, we conducted an experiment where
C
DOI: 10.1021/acs.oprd.7b00115
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
conversions its presence could be detrimental. Having
identified the key factor, we were confident that DoE could
provide meaningful answers since we were now posing the
correct questions.
On the basis of these results, we propose the reaction
mechanism depicted in Figure 4. Although complex i is initially
2,6-di-tert-butylpyridine, led to decreased reactivity as did the
addition of various salts such as NaAlCl4.
An alternative approach was to consider changing the
structure of the electrophile. In general, Friedel−Crafts
reactions of sulfonyl pyrroles are reported to be high-yielding
15
transformations, and only in the case of α-chlorinated acid
6
chlorides are lower yields reported. A wide variety of
electrophiles were investigated that would still allow us to
access an intermediate which would readily lead to 5 (Figure
5
).
Figure 5. Survey of alternative electrophiles.
The anhydride of chloroacetic acid did not yield any tangible
improvements over ClCH COCl. Several α-amino acid
2
Figure 4. Proposed mechanism for the formation of 2.
chlorides were investigated but did not prove successful
owing to low conversions. In stark contrast, acetyl chloride
(AcCl) showed rapid and quantitative conversion to the desired
formed upon mixing AlCl and ClCH COCl, after addition of
3
2
16
pyrrole 1 this complex exists in an equilibrium that strongly
product. No polymerization of pyrrole 1 could be detected,
and in initial scale-up experiments, no loss in performance was
observed. The dramatic difference in reactivity between
favors pyrrole−AlCl complex ii. We propose that the desired
3
reaction takes place between complex i and 1 via a typical
Friedel−Crafts mechanism, rather than the carbometalation
ClCH COCl and AcCl is likely a function of the stronger
2
14
route through ii as proposed previously. The formation of
complex ii is unproductive and opens a pathway for
polymerization of 1. The formation of HCl as the reaction
proceeds slows the reaction rate but also slows the rate of
decomposition of 1, possibly through stabilization of pyrrole−
binding of AlCl to AcCl, as can be observed in NMR studies of
3
17
the binding to AlCl to a variety of acid chlorides. This
3
removes the issues caused by the unproductive AlCl3
equilibrium between 1 and ClCH COCl and allows the
2
Friedel−Crafts acylation to proceed more rapidly than pyrrole
polymerization.
AlCl complex ii. Under low [HCl], the pyrrole 1 is unstable,
3
while under conditions with higher [HCl], particularly under
conditions where the solution is saturated in HCl, pyrrole 1 is
stable, but unreactive. We believe that the dual role of soluble
HCl in the reaction is a key factor that influences the scale
dependence of the process. Variations in the liquid−gas
equilibrium of HCl caused by changes in the volume of the
With this promising result in hand, the reaction of AcCl with
1 was optimized using a similar DoE to that employed
previously during optimization of the ClCH COCl reaction.
2
Using a slight excess of AlCl and AcCl relative to 1 allowed for
3
nearly quantitative yields of ketone 6 with greater than 99:1
C3:C2 selectivity and excellent in-process purity (Scheme 2).
Furthermore, this reaction was essentially addition-controlled
and thus not subject to the scale dependence observed in the
reactor headspace, the rate of exchange of N in the headspace,
2
or the handling of AlCl could affect the amount of HCl and
3
thus the reaction rate and the stability of 1. Controlling the
amount of HCl in the reaction would prove difficult since it is a
reaction byproduct, an input related impurity, and strongly
affected by the liquid−gas equilibrium in the system.
original ClCH COCl process. However, the integration of this
2
reaction into our synthetic route to 5 would then require a high
yielding α-chlorination of ketone 6, which did not exhibit
competitive chlorination of the pyrrole ring.
Having identified the equilibrium between the
ClCH COCl−AlCl (i) and pyrrole−AlCl (ii) complexes
Step 2. Ketone α-Chlorination. There are numerous
literature reports of the α-halogenation of carbonyl com-
2
3
3
1
8
18a
and the complex role of HCl, we hypothesized that shifting this
equilibrium toward the productive complex i could improve the
reaction yield and rate, and possibly mitigate the scale
dependence of the reaction. Toward this end, a wide variety
of the reaction conditions were explored including solvents,
Lewis acids, and additives. However, these studies revealed that
the reaction only tolerates minor changes from the original
conditions. Chlorinated solvents were optimal, and only group
pounds. Many reports have utilized copper halides,
18b
trimethyammonium dichloroiodate,
dimethylhydantoin (DCDMH),
and 1,3-dichloro-5,5-
among other methods.
18c,d
Our initial focus was to find a high-yielding and atom
economical method for performing the chlorination. For that
reason, we were intrigued by two simple reagents reported in
1
9
the literature: Oxone/NH Cl and N-chlorosuccinimide
4
20
(NCS). The combination of Oxone and NH Cl appeared
4
1
3 Lewis acids such as AlCl and GaCl gave reasonable levels
to fit our criteria and gave promising results (Table 2).
Optimization studies first focused on the reaction solvent
3
3
of reactivity. The addition of sterically hindered amines, like
D
DOI: 10.1021/acs.oprd.7b00115
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 2. Preparation of Pyrrole 5 from Pyrrole 1
Table 2. Conditions for the α-Chlorination of Ketone 6 with
Oxone
achieve complete conversion of 6. The minimum agitation rate
for complete suspension (Njs) of Oxone was calculated as 110
rpm ±20% for a 3000 L reactor that is half full. This agitation
rate is within 8% of the maximum agitation rate for this
particular reactor (120 rpm); therefore, it is likely that all of the
Oxone will not be completely suspended in the reaction stream.
Second, the reaction kinetics were greatly affected by the
particle size of Oxone, which means that a size specification
would have to be set for this reagent to consistently achieve
reaction completion within the desired time. All of these factors
were a major challenge for the translation of this chemistry into
a robust process.
entry
solvent
MeOH
chloride source conversion (%) 2:8 (ratio)
1
2
3
4
5
6
7
8
9
NH Cl
99
78
69
68
51
49
20
20
12
7
11
113
9
4
AcOH
NH Cl
4
MeOH/H O (9:1)
NH Cl
NCS was explored in the next attempt at devising a robust,
high-yielding chlorination. Literature reports describe it as a
versatile and convenient reagent, but in our hands it was found
to be far less selective than the combination of Oxone and
NH Cl (Table 3). In CH Cl , THF, or MeCN, low yields of
2
4
IPA
NH Cl
6
4
DPMU
DMF
NH Cl
5
4
NH Cl
29
179
2
4
CH CN
NH Cl
3
4
4
2
2
propylene glycol
sulfolane
IPAc
NH Cl
4
NH Cl
35
25
11
17
65
48
18
18
13
10
8
4
Table 3. Conditions for the α-Chlorination of 6 with NCS
10
11
12
13
14
15
16
17
18
19
NH Cl
4
MeTHF
DCE
NH Cl
5
4
NH Cl
2
4
MeOH
MeOH
MeOH
MeOH
MeOH
MeOH
MeOH
Bu NCl
66
73
98
72
52
92
89
4
LiCl
CsCl
CaCl2
MgCl2
NaCl
KCl
entry
solvent
conversion (%)
yield (2, %)
1
2
3
4
CH Cl2
53
56
9
18
10
93
2
THF
MeCN
MeOH
78
>98
(entries 1−12). In general, protic and polar solvents gave
higher conversions, likely due to the enhanced solubility of
Oxone, although no clear trends were apparent. Since MeOH
was found to be an optimal solvent for the chlorination, the
effect of chloride source was next examined to see if further
improvement in terms of selectivity was possible (entries 13−
the desired chloroketone 2 were obtained along with
dichlorinated side product 8 and additional ring-chlorinated
impurities. However, much to our surprise, the use of MeOH
allowed for an efficient chlorination with high selectivity for 2
and an excellent impurity profile. Less than 5% of 8 was formed
with no detectable amounts of the ring chlorination byproducts.
Two key factors were found for achieving a high yield in this
reaction: NCS equivalents and water content. Elevated levels of
8 were obtained when the NCS charge exceeded 1.1 equiv.
Significant amounts of overchlorination products were
1
9). In MeOH, the ratio of mono- to dichlorination was
significantly higher for Bu NCl and LiCl as compared to
4
NH Cl, but neither gave high conversion. Therefore, NH Cl
4
4
was selected for further investigation since it gave full
conversion and acceptable levels of the dichlorination side
product 8.
Further development of the α-chlorination with Oxone and
observed when 1.0 equiv of H O was added, which also led
2
NH Cl in MeOH showed that the formation of chloroketone 2
to a reduced reaction rate. Thus, it was important to control the
NCS charge between 1.05 and 1.10 equiv and water content to
below 0.08 equiv. Additional screening revealed that the
reaction rate is enhanced by the addition of catalytic amounts
of acid. In the presence of 0.1 equiv of methanesulfonic acid
(MSA), a significant improvement in rate and yield was
observed. Under these optimized conditions, complete
conversion was achieved in 6 h at 30 °C with a 90% in-process
yield. The safety of this process was evaluated with an
4
could be achieved in a 55−60% solution yield at 60 °C. The
reaction produced the dichloro compound 8 as the major side
product along with several other low level impurities.
Unfortunately, the reaction stream was not stable, and the
level of 8 and other byproducts increased with reaction time
and during the subsequent workup. From an engineering
perspective, there were several challenges associated with this
process. First, Oxone is a relatively dense solid (∼1.2 g/mL),
and a high loading (3 kg/kg of substrate) was required to
21
Advanced Reactive System Screening Tool (ARSST), and no
E
DOI: 10.1021/acs.oprd.7b00115
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
self-heating or self-pressurization events were observed up to
1
60 °C. Since the reaction was run in MeOH, several enol ether
intermediates could be detected by HPLC (Scheme 3). At the
Scheme 3. Intermediates Observed in the NCS Chlorination
of 6
Figure 6. Response of the crystallization of 5 to IPA and water volume
%
.
a much stronger function of IPA concentration. The isolated
yield and potency could be maximized by first charging 30 vol
%
of water to dissolve the NaCl generated during the reaction
and then charging 70 vol % of IPA to maximize the
crystallization yield (green circle in Figure 6).
Optimizing the amidation step was the final component to
successfully develop a high-yielding, selective, and scale-
independent route to 5 (Scheme 4). The commercial viability
end of the reaction, 40% of the product exists as the enol ether
1
Scheme 4. Synthesis of 5
11 as determined by H NMR. We believe that the presence of
1 at the end of the reaction provides a rationale for the high
1
selectivity for the desired mono chlorination product. The
intermediate 11 can easily be converted back to 2 upon the
addition of water during workup, allowing for the isolation of
pure 2.
Step 3. Amidation of Chloroketone 2. Having devised a
scale-independent alternative process to access chloroketone 2,
our focus turned to the final amidation step to provide 5
2
2
(
Scheme 2). Due to the high in-process purity of the
reactions, the first two steps of the sequence were telescoped,
23
thus avoiding the isolation of compounds 6 and 2. This
allowed for streamlined access to chloroketone 2 via the new
route, despite the addition of another step relative to the
original Friedel−Crafts acylation with ClCH COCl. Conditions
2
24
for the final amidation of α-chloro ketone 2 involved the use
of the sodium salt of N-formyl tosylamide 4 in the presence of
25
Bu NBr in 2-MeTHF. The amidation was optimized by
4
carrying out DoE studies focused on the effect of concentration,
temperature, water content, equivalents of Bu NBr (TBABr)
4
and 4 on the reaction rate, and the final purity profile. Careful
analysis of the DoE data revealed that using high substrate
concentrations (6 L 2-MeTHF/kg of 2) and 0.2 equiv of
Bu NBr at 40 °C allowed for reaction completion in <4 h with
4
an excellent purity profile. Crystallization of 5 demonstrated
excellent purging of hydrolysis impurities (typically formed in
<
1 HPLC AP (area percent) relative to 5) along with
succinimide, allowing for the isolation of 5 in ≥98 HPLC AP.
The amidation reaction results in a reactive crystallization
which also generates 1 equiv of NaCl as a byproduct. A
crystallization was devised that utilized the addition of an
antisolvent mixture of isopropanol/water to the Me-THF
process stream to maximize the yield and potency of 5 while
also purging NaCl. Figure 6 shows a map of the expected
isolated yields versus the volumes (L/kg) of IPA and water.
The expected isolated yields are based on solubility data and
only account for material lost to the mother liquors. The plot
shows the isolated yield is a weak function of water content and
of this three-step process has been demonstrated on a 420 kg
scale and has consistently achieved a 62−75% overall yield for
the three-step telescope with excellent purity (>98 HPLC AP)
and potency (>98 wt %). Compared to the original two-step
process, this represents a significant improvement despite the
addition of a linear step to the sequence (Table 4).
CONCLUSIONS
■
Much of process chemistry is focused on streamlining reactions
to improve efficiency. For example, improvements in yield, the
reduction of solvent usage, and the removal of unit operations
F
DOI: 10.1021/acs.oprd.7b00115
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Table 4. Comparison of the Two- and Three-Step Routes to
5
The resulting biphasic mixture was stirred at 20−25 °C for
0.5−1 h. The organic layer was separated and washed
sequentially with 5 wt % sodium carbonate solution (1141 L,
original route
commercial route
1
0 L/kg) and water (1141 L, 10 L/kg). The organic layer was
step 1
ClCH COCl
AcCl
2
concentrated under vacuum to 3−4 L/kg while maintaining the
batch temperature below 40 °C. To this solution was added
THF (1141 L, 10 L/kg), and then it was concentrated to 3−4
L/kg under vacuum. After the batch was cooled to 25−30 °C,
THF (685 L, 6 L/kg) was added. The water content of the
number of steps
yield
2
3
34%
75%
purity (HPLC)
process robustness
97−98%
scale dependent
98−99%
scale independent
crude solution was 2 wt % H O as determined by Karl Fischer
2
26
are key deliverables in late phase process development.
titration. The solution yield of 2 was 71.3%, and in-process
purity was 90.3 AP. To this solution of 2 was added the sodium
salt of N-formyl-p-toluene sulfonamide 4 (110.5 kg, 499.97
mol, 0.9 equiv), followed by tetra-n-butyl ammonium bromide
(13.4 kg, 41.56 mol, 0.07 equiv). The resulting suspension was
heated to 60 °C and aged for 4 h or until full conversion was
obtained as confirmed by HPLC. The suspension was then
cooled to 25 °C and washed with 0.5 wt % HCl solution
containing 5% brine (685 L, 6 L/kg) and then brine (685 L, 6
L/kg). The organic layer was concentrated under vacuum to 7
L/kg. To perform the crystallization, isopropanol (650.6 L, 5.7
L/kg) was charged at 25 °C. The batch was then seeded with
0.05 wt % 5 and stirred for 30 min to generate a suspension. A
second charge of isopropanol (650.6 L, 5.7 L/kg) was then
added, and the mixture was stirred at 25 °C for 2 h. A thick
slurry was obtained, which was cooled to 10−15 °C and stirred
for 2 h prior to filtration. The cake was washed with chilled
1.75:1 IPA/THF (10−15 °C, 342.5 L, 3 L/kg) followed by 2:1
IPA/THF (10−15 °C, 456.6 L, 4 L/kg). The wet cake was
Design of Experiments (DoE) studies are a popular method for
performing this kind of optimization. Although this technique
has proven powerful and insightful for such fine-tuning of
reaction conditions, it may have limitations if the conditions are
at a local, rather than a global, optimum. Step changes in terms
of reagent choice or route are difficult to predict or support
with the statistical outputs of a DoE study. In this report, we
describe a scenario where scale-dependent factors led to
significant challenges that could not be eliminated through
simple variations in the original reaction conditions. While DoE
studies led to specific improvements for a given scale and
reactor, no general solutions could be found to define a high-
yielding process. In an attempt to better understand the root
cause of the scale dependence, a series of spectroscopic
investigations were undertaken which identified key reaction
intermediates and revealed limitations that are inherent to the
mechanism of the reaction. These studies also pointed to a path
forward, but only if an additional step could be added to the
route. Although the proposal to add a step to a synthetic route
in planning for a large-scale campaign is not a decision that can
be made lightly, a mechanistic explanation of the issues allowed
us to build a strong, data-driven case for this change. In the end,
the decision to switch from a highly variable and low yielding
two-step route to a high-yielding and robust three-step route
was made, and the experience at the plant scale has validated
that choice.
dried at 50 °C under N /vacuum sweep to reach LOD < 0.5%,
2
and 5 was isolated as a white crystalline solid (85 kg, 34.3%
1
yield, 98.0 AP). Mp 145 °C. H NMR (DMSO-d ) 9.18 (br,
6
1H), 8.43 (m, 1H), 8.09 (m, 2H), 7.90 (m, 2H), 7.81 (m, 1H),
7.70 (m, 2H), 7.50−7.42 (m, 3H), 6.67 (m, 1H), 4.91 (br, 2H),
13
2.41 (s, 3H). C NMR (DMSO-d ) 184.6, 160.8, 145.8, 137.9,
6
134.9, 134.8, 130.3, 129.9, 127.9, 127.4, 125.6, 124.4, 121.9,
112.2, 48.3, 21.8. HRMS elemental calculated for
C H N O S : 447.0679; found: 447.0677.
In summary, we hope that this report emphasizes the
importance of deep, fundamental process understanding at
scale if robust and reliable processes are the end goal. A first-
principles approach allows us to view processes in a more
holistic sense and drive decisions which will have significant
payoffs in terms of cost and productivity. Exclusive focus on
simple metrics like step count may in fact hinder the effort to
identify more efficient processes.
20
18
2
6 2
Preparation of N-(2-Oxo-2-(1-(phenylsulfonyl)-1H-
pyrrol-3-yl)ethyl)-N-(phenylsulfonyl)formamide (5) by
the Three-Step Route. To a reactor under N was charged
2
CH Cl (3384 L, 8 L/kg), followed by addition of aluminum
2
2
chloride (229.3 kg, 1719.6 mol, 1.1 equiv). The resulting
suspension was held at 20 °C. Acetyl chloride (133.3 L, 147.2
kg, 1875.15 mol, 1.2 equiv) was added to the suspension, and it
was stirred at 20 °C until a clear solution was obtained (∼1 h).
To this solution was then added a solution of benzenesulfonyl-
EXPERIMENTAL SECTION
Preparation of N-(2-Oxo-2-(1-(phenylsulfonyl)-1H-
pyrrol-3-yl)ethyl)-N-(phenylsulfonyl)formamide (5) by
■
pyrrole (1) in 850 L of CH Cl (423 kg, 1563.27 mol) over 60
2 2
min to control off-gassing (1 equiv HCl). The homogeneous
reaction was stirred at 20 °C for 1 h or until >99% conversion
was obtained as confirmed by HPLC. The solution was cooled
to 10−20 °C and then slowly transferred to another reactor
containing water (2115 L, 5 L/kg based on 1) at 5−10 °C, at
such a rate as to maintain an internal temperature below 20 °C.
The biphasic mixture was stirred vigorously at 20 °C for at least
1 h. The organic phase was separated, and a vacuum distillation
the Two-Step Route. To a reactor under N was charged
2
CH Cl (1826 L, 16L/kg based on 1) followed by addition of
2
2
aluminum chloride (87.4 kg, 655.46 mol, 1.19 equiv). The
suspension was stirred at 15−20 °C for 30 min. Chloroacetyl
chloride (62.48 L, 88.6 kg, 784.48 mol, 1.42 equiv) was then
added, and the mixture was stirred at 25 °C for 1−2 h until a
clear solution was obtained. To this solution was then added a
solution of benzenesulfonylpyrrole (1) in 228 L CH Cl2
was performed to obtain a water content of <0.05 wt % H O as
2
2
(
114.15 kg, 550.81 mol) over 11 min. The resulting solution
determined by Karl Fischer titration. Methanol (2538L, 6 L/
kg) was then added, and CH Cl was removed by vacuum
was heated to 35 °C and aged for 5 h or until full conversion
was obtained as confirmed by HPLC. The solution was cooled
to 10−20 °C and then slowly transferred to another reactor
containing water (1369 L, 12 L/kg based on 1) at 5−10 °C, at
such a rate as to maintain an internal temperature below 20 °C.
2
2
distillation until less than <0.05 wt % CH Cl could be detected
2
2
by GC analysis. Methanol was added to adjust the final volume
to 18 L/kg (∼7600 L), and the water content of the solution
was controlled at <0.15 wt % H O as determined by Karl
2
G
DOI: 10.1021/acs.oprd.7b00115
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Fischer titration. The solution was cooled to 20 °C. The
solution yield of 6 was >99%, and the in-process purity was 99
AP. To this MeOH solution of 6 was added N-chlorosuccini-
mide (219.2 kg, 1641.57 mol, 1.05 equiv) followed by
methanesulfonic acid (13.64 L, 18 kg, 187.28 mol, 0.12
equiv). The mixture was heated to 30 °C and aged for 6 h or
until >97% conversion was obtained as confirmed by HPLC.
The solution was cooled to 20 °C, and water (65 L, 0.2 L/kg)
was added. The solution was concentrated under vacuum to 5
L/kg for addition of 2-MeTHF (3250 L, 10 L/kg). The
resulting solution was stirred at 20 °C for 1 h to effect cleavage
of the enol ether 11. Methanol was removed by vacuum
distillation to achieve a 10 L/kg 2-MeTHF solution of 2
containing less than 1 wt % MeOH while maintaining the
internal temperature below 40 °C. The solution was washed
with 20 wt % NaCl solution (1625 L, 5 L/kg) containing 0.15
equiv of sodium bicarbonate (19.7 kg, 234.49 mol). The
organic layer was distilled under vacuum to obtain a water
on the content. We thank the Chemical & Synthetic
Development senior management for support during the
preparation of this manuscript.
REFERENCES
■
(
1) (a) Li, J. J. Heterocyclic Chemistry in Drug Discovery; Wiley:
Hoboken, NJ, 2013; pp 18−53. (b) Young, I. S.; Thornton, P. D.;
Thompson, A. Nat. Prod. Rep. 2010, 27, 1801−1839. (c) Lipkus, A. H.;
Yuan, Q.; Lucas, K. A.; Funk, S. A.; Bartelt, W. F., III; Schenck, R. J.;
Trippe, A. J. J. Org. Chem. 2008, 73, 4443−4451.
(
2) (a) Estevez, V.; Villacampa, M.; Menendez, J. C. Chem. Soc. Rev.
2010, 39, 4402−4421. (b) Bergman, J.; Janosik, T. In Comprehensive
Heterocyclic Chemistry III; Jones, G., Ramsden, C. A., Eds.; Elsevier:
Amsterdam, 2008; Vol. 3, pp 269−351.
(
3) (a) Anderson, H. J.; Loader, C. E. Synthesis 1985, 1985, 353−364.
b) Jones, R. A.; Bean, G. P. The Chemistry of Pyrroles; Academic Press:
London, 1977; pp 115−208.
4) (a) Gutierrez, E. G.; Wong, C.J.; Sahin, A. H.; Franz, A. K. Org.
(
(
Lett. 2011, 13, 5754−5757. (b) de Haro, T.; Nevado, C. J. Am. Chem.
Soc. 2010, 132, 1512−1513. (c) Tobisu, M.; Yamaguchi, S.; Chatani,
N. Org. Lett. 2007, 9, 3351−3353. (d) Pirrung, M. C.; Zhang, J.;
Lackey, K.; Sternbach, D. D.; Brown, F. J. Org. Chem. 1995, 60, 2112−
content of <0.03 wt % H O as determined by Karl Fischer
2
titration. The batch was cooled to 20 °C, and 2-MeTHF was
added to a final volume to 6 L/kg (∼1900 L). The solution
yield of 2 was 90%, and the in-process purity was 93 AP. To
this 2-MeTHF solution of 2 was added the sodium salt of N-
formyl-p-toluene sulfonamide 4 (345.8 kg, 1564.6 mol, 1.0
equiv) and TBABr (100.8 kg, 312.68 mol, 0.2 equiv). The
resulting suspension was heated to 40−50 °C and aged 1 h or
until >99% conversion was obtained as confirmed by HPLC.
To the solution was added 2:1 IPA/water (8400 L, 20 L/kg)
over 1 h while maintaining its temperature at 40 °C. The
mixture was aged at 40 °C for 1 h then cooled to 20 °C over 1
h and held at 20 °C for no less than 2 h. The resulting slurry
was filtered and washed with IPA (2100 L, 2 × 5 L/kg). The
2
124. (e) Bray, B. L.; Mathies, P. H.; Naef, R.; Solas, D. R.; Tidwell, T.
T.; Artis, D. R.; Muchowski, J. M. J. Org. Chem. 1990, 55, 6317−6328.
(5) (a) Chen, K.; Risatti, C.; Bultman, M. S.; Soumeillant, M.;
Simpson, J.; Zheng, B.; Fanfair, D.; Mahoney, M.; Mudryk, B.; Fox, R.
J.; Hsaio, Y.; Murugesan, S.; Conlon, D. A.; Buono, F. G.; Eastgate, M.
D. J. Org. Chem. 2014, 79, 8757−8767. (b) Eastgate, M. D.; Bultman,
M. S.; Chen, K. C.; Fanfair, D. D.; Fox, R. J.; La Cruz, T. E.; Mudryk,
B. M.; Risatti, C. A.; Simpson, J. H.; Soumeillant, M. C.; Tripp, J. C.;
Xiao, Y. Methods for the Preparation of HIV Attachment Inhibitor
Piperazine Prodrug Compound. U.S. Patent 2013/0203992. (c) See
1
(
0.1021/op7b00121.
6) (a) Silvestri, R.; Pifferi, A.; De Martino, G.; Massa, S.; Saturnino,
wet cake was dried at 50 °C under N /vacuum sweep to reach
C.; Artico, M. Heterocycles 2000, 53, 2163−2174. (b) De Micheli, C.;
2
5
LOD < 0.5%, and 5 was isolated as a white crystalline solid
De Amici, M.; Locati, S. Il Farmaco 1984, 39, 277−288.
(529.1 kg, 75.8% yield, 98.8 AP).
(7) Atherton, J. H., Carpenter, K. J. In Process Development:
Physiochemical Concepts; Oxford University Press: Oxford, England,
2
(
005; pp 36−53.
ASSOCIATED CONTENT
* Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/acs.oprd.7b00115.
■
8) Due to time constraints, the initial campaign was completed using
S
the ClCH COCl procedure, and a total of thirteen 120 kg scale
batches were processed that consistently gave a ∼70% yield of 2.
(
evaporation of DCM during long experiment times.
(10) Germain, A.; Commeyras, A.; Casadevall, A. J. Chem. Soc. D
1
(
mmol/mL; see: Ahmed, W.; Gerrard, W.; Maladkar, V. K. J. Appl.
Chem. 1970, 20, 109−116.
(
2
9) ReactIR experiments were performed in DCE to minimize
1
Experimental procedures, characterization data, and H
and 13C NMR data (PDF)
971, 0, 633−635.
11) The solubility of HCl in DCE at 20 °C is reported as 0.53
AUTHOR INFORMATION
Corresponding Author
E-mail: gregory.beutner@bms.com.
■
*
12) Jones, B.; Nachtsheim, C. J. J. Qual. Technol. 2011, 43, 1−15.
13) For the DOE study, MeOH was premixed with AlCl in DCM
(
3
ORCID
prior to the reaction to simulate the addition of HCl.
Gregory L. Beutner: 0000-0001-8779-1404
Michael Bultman: 0000-0002-0140-6120
Martin D. Eastgate: 0000-0002-6487-3121
(14) Huffman, J. W.; Smith, V. J.; Padgett, L. W. Tetrahedron 2008,
6
(
1
(
4, 2104−2112.
15) Kakushima, M.; Hamel, P.; Frenette, R.; Rokach, J. J. Org. Chem.
983, 48, 3214−3219.
16) For examples with acylating reagents, see: (a) Noland, W. E.;
Notes
The authors declare no competing financial interest.
Lanzatella, N. P. J. Heterocycl. Chem. 2009, 46, 1285−1295. (b) Harrak,
Y.; Rosell, G.; Daidone, G.; Plescia, S.; Schillaci, D.; Pujol, M. D.
Bioorg. Med. Chem. 2007, 15, 4876−4890. (c) Muratake, H.; Natsume,
M. Heterocycles 1990, 31, 683−690. (d) De Micheli, C.; De Amici, M.;
Platini, D. B. Il Farmaco 1986, 41, 913−925. (e) Anderson, H. J.;
Loader, C. E.; Xu, X. R.; Le, N.; Gogan, N. J.; McDonald, R.; Edwards,
L. G. Can. J. Chem. 1985, 63, 896−902. (f) Xu, X. R.; Anderson, H. J.;
Gogan, N. J.; Loader, C. E.; McDonald, R. Tetrahedron Lett. 1981, 22,
ACKNOWLEDGMENTS
■
The authors thank Dr. Frank A. Rinaldi for his assistance with
NMR spectroscopy, Dr. Wei Ding and Mr. Mike Peddicord for
assistance with mass spectroscopy, Drs. Brendan Mack and Jose
Tabora for assistance with DoE planning, Dr. Jacob Janey for
assistance with lab automation, and Dr. Sergei Kolotuchin for
helpful discussions. We thank Dr. Michael Schmidt for a careful
review of this manuscript and for providing helpful suggestions
4
899−4900.
(17) Bigi, F.; Casnati, G.; Sartori, G.; Predieri, G. J. Chem. Soc., Perkin
Trans. 2 1991, 2, 1319−1321.
H
DOI: 10.1021/acs.oprd.7b00115
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
(
18) (a) Shi, X.-X.; Dai, L.-X. J. Org. Chem. 1993, 58, 4596−4598.
(
b) Kajigaeshi, S.; Kakinami, T.; Moriwaki, M.; Fujisaki, S.; Maeno, K.;
Okamoto, T. Synthesis 1988, 1988, 545−546. (c) Chen, Z.; Zhou, B.;
Cai, H.; Zhu, W.; Zou, X. Green Chem. 2009, 11, 275−278. (d) Xu, Z.;
Zhang, D.; Zou, X. Synth. Commun. 2006, 36, 255−258.
(
19) Zhou, Z. S.; Li, L.; He, X. H. Chin. Chem. Lett. 2012, 23, 1213−
1
(
(
(
216.
20) (a) Lee, J. C.; Park, H. J. Synth. Commun. 2007, 37, 87−90.
b) Wang, C.; Tunge, J. Chem. Commun. 2004, 23, 2694−2695.
c) Meshram, M.; Reddy, P. N.; Sadashiv, K.; Yadav, J. S. Tetrahedron
Lett. 2005, 46, 623−626. (d) Sreedhar, B.; Surendra Reddy, P.;
Madhavi, M. Synth. Commun. 2007, 37, 4149−4156.
(
21) Advanced Reactive System Screening Tool (ARSST).
Burelbach, J. P. North American Thermal Analysis Society, 28th
Conference, Orlando, FL, Oct. 4−6, 2000.
(
22) Hoffmann, R. W.; Bruckner, D. New J. Chem. 2001, 25, 369−
3
(
(
73.
23) Compound 2 has been shown to be a potent skin sensitizer
local lymph node assay EC3 = 0.028%).
24) (a) Li, X.; Zhou, X.; Zhang, J.; Wang, L.; Long, L.; Zheng, Z.; Li,
(
S.; Zhong, W. Molecules 2014, 19, 2004−2028. (b) van Zupten, S.;
Kraus, M.; Driessen, C.; van der Marel, G. A.; Overkleeft, H. S.;
Reedijk, J. J. Inorg. Biochem. 2005, 99, 1384−1389. (c) Adam, I.; Orain,
D.; Meier, P. Synlett 2004, 11, 2031−2033. (d) Mitschke, U.;
Debaerdemaeker, T.; Baeuerle, P. Eur. J. Org. Chem. 2000, 2000, 425−
4
(
37.
25) For examples of the use of catalytic TBABr in chloroketone
amidations, see: Aldous, S.; Fennie, M. W.; Jiang, J. Z.; John, S.; Mu,
L.; Pedgrift, B.; Pribish, J.; Rauckman, B.; Sabol, J. S.; Stoklosa, G. T.;
Thurairatnam, S.; van Deusen, C. L. Pyrimidine Hydrazine Compounds
as PGDS Inhibitors. WIPO, WO 2008/121670 A1..
(
26) Weissman, S. A.; Anderson, N. G. Org. Process Res. Dev. 2015,
1
9, 1605−1633.
I
DOI: 10.1021/acs.oprd.7b00115
Org. Process Res. Dev. XXXX, XXX, XXX−XXX