Synthesis of filibuvir. Part III. Development of a process for the reductive coupling of an aldehyde and a β-keto-lactone

Article abstract of DOI:10.1021/op400237j

Development of a reductive coupling of a β-keto-lactone and an aldehyde is described, in which the Hantzsch ester serves as an inexpensive and convenient reducing agent. Structural features in the β-keto-lactone rendered standard reductive coupling conditions ineffective, requiring development of a specific addition and temperature protocol. Identification of one of the reactants as Ames positive required a single-digit parts per million control strategy for this impurity in the final active pharmaceutical ingredient (API).

Full text of DOI:10.1021/op400237j

Article
pubs.acs.org/OPRD
Synthesis of Filibuvir. Part III. Development of a Process for the
Reductive Coupling of an Aldehyde and a β‑Keto-lactone
Nathan D. Ide,*,† John. A. Ragan,*,† Gabriel Bellavance,† Steve J. Brenek,† Eric M. Cordi,†
Grace O. Jensen,‡ Kris N. Jones,† Danny LaFrance,§ Kyle R. Leeman,‡ Leo J. Letendre,† David Place,†
Corey L. Stanchina,† Gregory W. Sluggett,‡ and Holly Strohmeyer‡
‡
†Chemical Research and Development, Analytical Research and Development, Pfizer Worldwide Research and Development,
Eastern Point Road, Groton, 06340 Connecticut
Jon Blunt, Kevin Meldrum, and Stuart Taylor∥
Chemical Research & Development, Pfizer Process Development Facility, Ramsgate Road, Sandwich, Kent, U.K.
Ciaran Byrne, Denis Lynch, Sandra Mullane, Maria M. O’Sullivan, and Marcella Whelan
Pfizer Global Supply, Ringaskiddy, Cork County, Ireland
ABSTRACT: Development of a reductive coupling of a β-keto-lactone and an aldehyde is described, in which the Hantzsch
ester serves as an inexpensive and convenient reducing agent. Structural features in the β-keto-lactone rendered standard
reductive coupling conditions ineffective, requiring development of a specific addition and temperature protocol. Identification of
one of the reactants as Ames positive required a single-digit parts per million control strategy for this impurity in the final active
pharmaceutical ingredient (API).
alkene 5 were observed by HPLC, IR, or mass spectrometry.
The nucleophilicity of this particular β-keto-lactone was such
that, in any process in which Knoevenagel alkene was generated
in its presence, the bis-adduct was rapidly formed.
Given this realization, three general strategies for API
formation were considered: (i) generate an equilibrating
mixture of bis-adduct and Knoevenagel alkene with reduction
of the transient Knoevenagel alkene to form API; (ii) trap the
transient Knoevenagel alkene with a nucleophile which could
then be hydrogenolyzed to form API; (iii) avoid formation of
the bis-adduct altogether by reducing the Knoevenagel alkene
immediately upon formation. In short, we either needed to
work with the bis-adduct or avoid its formation.
For the first general strategy, the procedure was to combine
aldehyde 3 and β-keto-lactone 2 and allow equilibrium to be
established prior to introduction of a reducing agent.
Numerous reducing agents were examined, including Na-
(OAc)3BH, triethylammonium formate,7 hydrogen, 2-phenyl-
imidazoline,8 2-phenylthiazoline,9 and the Hantzsch ester
(12).10−13 While several of these reducing agents showed a
reasonable amount of product formation (particularly the
Hantzsch ester), the isolated yields suffered due to the
temperatures required to establish equilibrium between the
bis-adduct and Knoevenagel alkene (>60 °C). At these
temperatures, the API underwent decomposition via an
elimination/decarboxylation pathway (Scheme 4). Additionally,
INTRODUCTION
■
In the previous two articles (Synthesis of Filibuvir, Parts I and
II, the preceding two articles in this journal) we described two
routes to β-keto-lactone 2 that use Evans aldol technology for
the control of chirality. The development of a reductive
coupling of this β-keto-lactone (2) with aldehyde 3 to provide
filibuvir (1) and methods to meet rigorous purity specifications
are described herein (Scheme 1).
BACKGROUND
■
The initial medicinal chemistry1 and cGMP lots2,3 of 1 were
manufactured using a borane−amine-mediated reductive
coupling (Scheme 2). While this approach was effective and
provided over 250 kg of pharmaceutical ingredient (API), the
yield was modest on scale (55−60%), and a subsequent
recrystallization was required. Additionally, the borane−amine
complex posed safety, cost, and sourcing issues. These
considerations led us to seek an alternative reductive coupling
protocol.
Development of an Alternative Reductive Coupling.
The simplest approach for this coupling would be to combine
β-keto-lactone (2) and aldehyde (3) to generate Knoevenagel
alkene (5) and then to add a reducing agent to generate
filibuvir (1). While there is precedent for this direct approach
with related 1,3-dicarbonyl nucleophiles,4−6 initial experiments
identified a feature of our system that precluded this direct
approach. As shown in Scheme 3, when β-keto-lactone 2 and
aldehyde 3 were combined, they quickly reacted to form bis-
adduct 6; neither the initial aldol adduct 4 nor Knoevenagel
Received: August 27, 2013
© XXXX American Chemical Society
A
dx.doi.org/10.1021/op400237j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 1. β-Keto-lactone and aldehyde precursors to filibuvir (1)
Scheme 2. Reductive coupling by a borane−amine complex
Scheme 3. Reaction pathway of β-keto-lactone 2 and aldehyde 3 to bis-adduct 6
lack of chemoselectivity at the required temperatures led to
competitive reduction of the aldehyde, requiring the use of an
excess of this starting material.
in situ yield. Unfortunately, the yield could not be increased
further, and competitive aldehyde reduction continued to be
problematic.
The second general strategy involved evaluation of a variety
of nucleophiles to trap the transient Knoevenagel alkene and, in
the presence of hydrogen and a catalyst, form API via
hydrogenolysis (see Scheme 5). There is precedent for this
strategy with cyclopentanedione Knoevenagel alkenes and
pyrrolidine.14 Nucleophiles examined included NaSO2Ph,
NaI/Et3N, amines (e.g., n-butylamine, benzylamine, and 2-
aminopyridine), and thiols (e.g., n-octylthiol).15 These experi-
ments were performed in EtOH solvent with hydrogen and
either Pd/C or Raney Ni catalysts. The most promising of these
was NaSO2Ph, which after catalyst optimization provided a 67%
A strategy related to this in situ nucleophile trapping was
preformation of an imminium salt (10) and trapping with β-
keto-lactone. Hydrogenolysis of the resulting amine (11) would
then provide API 1 (Scheme 6). This approach was examined
and showed promise with dimethylamine and a thiol-poisoned
Pd/S catalyst (amine 11 likely benefits from a H-bond between
the amine lone pair and the enolic OH of the β-keto-lactone).
However, with the parallel development of the semi-batch
Hantzsch process (vide infra), it was ultimately found to be less
attractive in terms of reaction yield, workup complexity,
product purity, and equipment train flexibility.
B
dx.doi.org/10.1021/op400237j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 4. Decomposition of API via elimination and decarboxylation
Scheme 5. Nucleophile-intercepted Knoevenagel alkene strategy
The third general strategy (avoiding bis-adduct formation
altogether) provided the most promising results. It was hoped
that by slowly adding a solution of β-keto-lactone 2 to a
mixture of aldehyde and Hantzsch ester (HEH, 12), a low
concentration of β-keto-lactone could be maintained (Scheme
7). Under these conditions, it was envisioned that the transient
Knoevenagel alkene would react faster with the excess
Hantzsch ester (to form 1) than it would with the low
concentration of unreacted β-keto-lactone (to form bis-adduct
6). Indeed, slow (semi-batch) addition of β-keto-lactone 2 over
4 h to a slurry of aldehyde 3 and Hantzsch ester (HEH) 12
provided a dose-controlled reaction with an in situ yield >95%.
On the basis of initial solvent screens, a solvent mixture of
methanol, THF, and acetic acid was selected (methanol helped
to solubilize the β-keto-lactone 2). To minimize reaction
variables, the same solvent mixture was used in both the β-keto-
lactone charge solution and the HEH−aldehyde slurry in the
primary reaction vessel. These initial optimization studies led to
a reaction temperature of 40 °C, an addition time of 6 h, and a
stoichiometry of 1.2 equiv each of Hantzsch ester and aldehyde
relative to β-keto-lactone. Stir rates were examined and found
to have no impact on the in situ yield.
Development of an effective workup posed several
challenges, including removal of unreacted Hantzsch ester
and its pyridine byproduct 13, appropriate pH control for the
acidic (β-ketolactone) and basic (pyridine, aminoguanidine)
functionalities in the API, and control of residual aldehyde to
parts per million (ppm) levels (vide infra). We found that
extraction with aqueous citric acid pulled the API into the
aqueous phase via protonation of the pyridine nitrogen of 1,
which is more basic than the Hantzsch pyridine. Thus, addition
of n-propyl acetate and 10% aqueous citric acid provided a
cleanly separated, biphasic mixture. The aqueous phase
contained API and unreacted aldehyde, while the organic
phase retained the Hantzsch pyridine 13. The organic phase
(containing 13) was discarded, THF was added to the aqueous
phase, and the pH was adjusted to 5.5−6.5. This pH range was
based on the experimentally determined pH/solubility profile
and is consistent with the titrametrically determined pKa values
of 7.26 and 4.61 (pyridine and β-ketolactone, respectively). At
pH values between these two numbers, the API will exist as a
C
dx.doi.org/10.1021/op400237j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 6. β-Keto-lactone/imminium condensation strategy
Scheme 7. Semi-batch addition of β-keto-lactone (2) to aldehyde (3) and HEH (12)
neutral compound (possibly zwitterionic), and its aqueous
solubility will be minimized. Addition of n-propyl acetate then
provided a two-phase system in which the API resided primarily
in the organic phase, and the aqueous phase contained
unreacted aldehyde and inorganic salts. The organic phase
was separated, and subjected to distillation with addition of n-
propyl acetate to provide crystalline API in 60−70% yield.
At this point, further optimization of the workup was
hindered due to an incomplete understanding of mass balance.
In an attempt to clarify this, 14C-labeled β-keto-lactone (labeled
at the lactone carbonyl) was prepared. By tracking the
radioactivity in the workup solutions, we learned that ∼20%
of material was being lost to the initial organic waste layer (after
acidification with citric acid), a fact that was not apparent
earlier, due to variability in the HPLC potency measurements.
Addition of a second extraction of this layer with aqueous citric
acid provided a yield improvement to 80−90%.
Identification of a Mutagenic Impurity. At this juncture
a critical toxicology finding changed the course of development
for this process. Ames testing revealed that aldehyde 3 is
mutagenic. This aldehyde is considered an OEB3 (Occupa-
tional Exposure Band 3) compound and does not pose any
atypical worker safety issues, on the basis of negative
chromosomal damage assays. However, the positive Ames
test triggered a staged TTC (Threshold of Toxicological
Concern) limit for this impurity in the API.16,17 The phase 2
limit was 7 ppm, based on a staged TTC of 10 μg/day and a
maximum daily dose of 1.5 g. For phase 3 the limit was 4 ppm,
D
dx.doi.org/10.1021/op400237j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
based on a staged TTC of 5 μg/day and maximum daily dose of
1.2 g. The commercial specification was not yet determined,
but was likely to be in the low ppm range.
of n-propyl acetate (variations in the time required to reach
final temperature led to variations in the amount of product lost
to decomposition, per the pathway shown in Scheme 4). This
issue was addressed by introduction of a lower-temperature
vacuum distillation at pressures from 100 to 500 mbar.
In addition to the yield variability, the citric acid heat cycle
was generating two impurities (Figure 1). These impurities
were present in the isolated product at levels below the clinical
specification, but above the likely commercial specification of
NMT 0.08% (Figure 1).
Impurity 15 results from an acid-mediated aldol condensa-
tion between the Hantzsch pyridine 13 and unreacted aldehyde
3, followed by lactonization and elimination (Scheme 9). The
four remaining impurities (16a−d) are regio- and stereo-
isomers arising from a condensation between the nucleophilic
methyl groups on the triazolopyrimidine ring and residual
aldehyde 3. All of these impurities form during the citric acid
heat cycle, and are partially responsible for the destruction of
the aldehyde 3 observed during that process (reduction to
alcohol 14, Scheme 8, is the primary destruction pathway). The
challenge posed by impurity control was amplified due to the
high daily dose of API. In this case, the ICH Q3A(R2)
guideline indicates an identification and qualification threshold
of 1.0 mg/day which corresponds to an unspecified impurity
limit of NMT 0.08%.18
The requirement that a starting material in the final bond-
forming step be purged into the single digit ppm range
presented a significant challenge. Additionally, once ppm level
assays were developed, it was found that aldehyde 3 was also an
oxidative degradant of the API. Exposure of API to open air and
light conditions for several days led to a ∼10-fold increase in
aldehyde levels (from 2 to 4 to 15−30 ppm). API stored in
sealed containers did not show increases in aldehyde levels, and
drug product samples were sufficiently stable. Nevertheless, this
presented further handling challenges in terms of API isolation.
Our first strategy, developed in preparation for a Pilot Plant
campaign to deliver phase 2 clinical supplies (7 ppm aldehyde
limit), was two-fold: add a heat cycle prior to crystallization to
reduce residual aldehyde to the alcohol 14 (which was Ames
negative) (Scheme 8), and optimize the crystallization process
to maximize the purge of residual aldehyde.
Scheme 8. Reduction of aldehyde 3 to alcohol 14
Because these impurities all resulted from the presence of
excess aldehyde and the citric acid−heat cycle required to
reduce it to alcohol 14, we decided to revisit the reaction
stoichiometry. A DoE based on equivalents of aldehyde (3) and
Hantzsch ester (12) relative to β-keto-lactone (2) indicated
that we could reduce these values to 1.00 and 1.07 equiv,
respectively. As expected, this led to a small decrease in in situ
yield and a small increase in the level of the bis-adduct (6) (1−
2% each), but the isolated yield was essentially unchanged,
presumably due to the increased efficiency of the crystallization
in the absence of the heat cycle-generated impurities. This
provided a significant improvement in overall product purity
prior to isolation, with lower levels of residual aldehyde, and
undetectable levels of the impurities described above (Figure
1), due to the removal of the heat cycle.
This process was evaluated in an RC-1 reactor on 85 g scale,
and in two runs led to yields of 79% and 76%, with aldehyde
levels below the limit of quantitation (≤0.4 ppm). All other
impurities were within specified limits. This process was further
scaled to >100 kg in the SEG facility; 329 kg of API was
prepared in two batches, meeting all purity specifications, in a
total yield of 77.5%. This material was earmarked for phase 3
clinical trials, and thus the limit on residual aldehyde was ≤4
ppm, a specification that both batches met (1.0 and 2.2 ppm).
This slight increase in aldehyde levels (≤0.4 ppm to 1.0−2.2
ppm) is not unexpected for such a large increase in reaction
scale.
An optional reprocessing procedure was also developed, so
that remediation was available (see Experimental Section for
details). This reprocessing sequence was found to offer a
significant (∼20×) purge of residual aldehyde 3, as well as a
general enhancement of API purity. On scale, it was required
for 2 out of 15 batches.
A more streamlined workup was subsequently developed that
reduced the number of unit operations and phase separations.
This process, described in detail in the Experimental Section,
involved omission of the citric acid and use of n-heptane and
water in the initial workup. The acidic aqueous product phase
The reduction of unreacted aldehyde was achieved by adding
a 60 °C heat cycle after the addition of aqueous citric acid.
While aldehyde reduction is slow under the standard reaction
conditions, we knew from earlier work that both heat and acid
accelerate the rate of aldehyde reduction by Hantzsch ester. A 6
h hold at 60 °C reduced the majority of the excess aldehyde
(1.2 equiv was used, so that at 100% conversion our starting
point prior to reduction would be 0.2 equiv or ∼200,000 ppm).
Experimental evaluation of the maximum temperature, hold
time, cooling rate, seed temperature, and seed quantity led to
the following optimized conditions. At the end of the
atmospheric distillation, the solution was cooled from 102 to
80 °C at a rate of 0.5 °C/min, seeded with 0.2 wt %/wt 1, and
held at 80 °C for 1 h. The slurry was then cooled to 10 °C at a
rate of 0.5 °C/min, stirred at that temperature for 1 h, isolated
by filtration, and rinsed with 10 °C n-propyl acetate (2 mL/g).
With these improvements in place, product containing <1 ppm
of aldehyde 3 was isolated in 76.0 and 76.4% yields in two RC-1
runs (85 g).
This procedure was effective on multikilogram scale, both in
Pilot Plant (Sandwich, UK) and Small Equipment Group
(SEG, Ringaskiddy, Cork, Ireland) campaigns. There were
several issues encountered, however, including significant yield
variability (Table 1). In postcampaign analyses, we determined
that the primary cause for this variability was thermal
decomposition of the API during the atmospheric distillation
Table 1. Yields for bulk campaigns
campaign (kg API/batch)
yields (%)
pilot plant (10−40 kg)
SEG (100−130 kg)
73.3, 64.9, 74.2, 63.6
64.7, 71.1, 80.4
E
dx.doi.org/10.1021/op400237j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 1. Impurities formed during citric acid heat cycle.
Scheme 9. Mechanism for the Formation of Impurity 15
Scheme 10. Deuterium-labeling experiments
was then pH-adjusted, the product extracted into n-propyl
acetate, and vacuum distillation performed as before. This has
not been evaluated on larger scale, but is expected to be the
superior process. The Experimental Section includes an 80 g
run of this process, as well as a multikilogram run of the earlier
workup procedure.
was initially assumed to be due to oxidative degradation of the
API. The most likely alternate explanation, oxidation of residual
alcohol 14, was ruled out on the basis of unchanging levels of
the alcohol as aldehyde levels increased. Alcohol 14 was
typically present at 100−150 ppm in API. In samples in which
aldehyde levels were purposefully increased (e.g., by extended
exposure to air and light during product filtration), the alcohol
levels did not decrease in correlation with increasing aldehyde
Mechanistic Studies on Oxidative Degradation. The
increase in aldehyde levels in API samples upon exposure to air
F
dx.doi.org/10.1021/op400237j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 11. Two proposed mechanisms of oxidative degradation
Scheme 12. Comparison of nucleophile/aldehyde partners under ambient batch conditions
Scheme 13. Reaction of Meldrum’s acid with aldehyde 3 and β-keto-lactone 2 with benzaldehyde
levels. Nevertheless, definitive evidence that the source of
increasing aldehyde levels was oxidative degradation of the API
(as opposed to some other trace impurity) was lacking until a
deuterium-labeling experiment was performed (Scheme 10).
API with monodeuterio labeling at C3′ (D1-1) was prepared
through use of dideuterio Hantzsch ester.19 Subjection of this
deuterium-labeled API to air for several days at 80 °C generated
monodeuterio aldehyde D1-3. Additionally, the ratio of deuterio
to protio aldehyde allowed the determination of an observed
kinetic isotope effect of kH/kD = 2.7 at 80 °C.20 Given the
G
dx.doi.org/10.1021/op400237j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 14. Simple β-keto-lactones under batch and semi-batch conditions (60 °C)
reaction temperature (80 °C), this is consistent with a primary
kinetic isotope effect, suggesting that breaking of the benzylic
C−H/D bond occurs in or prior to the rate-determining step.
This supports mechanism A over mechanism B (Scheme 11).21
The latter mechanism was considered on the basis of the
observation of α-hydroxy ketone 19 as a degradant during
forced degradation studies of API. Additionally, this study
supported our hypothesis that the source of increasing aldehyde
levels was direct oxidative degradation of the API.
(1) The performance difference between cyclic diester
nucleophiles (e.g., Meldrum’s acid) and β-keto-lactones
(e.g., 2, 26) is due to the enhanced nucleophilicity of the
β-keto-lactone. Under ambient batch conditions (EtOH,
25 °C), this enhanced reactivity leads to formation of bis-
adduct by addition of the β-keto-lactone to the
intermediate Knoevenagel alkene. Under these condi-
tions the bis-adduct is a dead end, and thus yields of the
desired product are poor.
Comparison of β-Keto-lactone (2) and Aldehyde (3) to
Related Coupling Partners. With a viable commercial
process in hand, we turned to the question of why our reaction
partners behaved differently than many other Hantzsch ester-
mediated reductive couplings. For example, in work from the
Ramachary group,22 a variety of 1,3-dicarbonyl nucleophiles
and aldehydes are effectively coupled by simply combining in
the presence of the Hantzsch ester (HEH) and catalytic proline
at ambient temperatures (“ambient batch conditions”). A
specific comparison (Scheme 12) highlights the difference with
our system, in which Meldrum’s acid and benzaldehyde couple
in 87% yield, whereas our β-keto-lactone/aldehyde combina-
tion delivers a 1:1 mixture of desired product and bis-adduct
under the same conditions.
In order to further probe the reason for this dichotomy, we
systematically varied each reacting partner (Scheme 13).
Meldrum’s acid and aldehyde 3 provided a good yield under
ambient batch conditions (72%). In contrast, β-keto-lactone 2
and benzaldehyde provided a complex mixture under the same
conditions. These data suggest that β-keto-lactone 2 is the
primary culprit for the poor results observed in our system
under ambient batch conditions.
Further light was shed by studying simpler β-keto-lactones
(Scheme 14). These nucleophiles (e.g., 26), which should be
similar to our β-keto-lactone except for the lack of the
diethylpyridine substituent, were also poor performers under
ambient batch conditions. However, at 60 °C, simple batch
conditions in ethanol−acetic acid (i.e., direct combination of
aldehyde, Hantzsch ester, and β-keto-lactone) were effective for
the reductive coupling. No benefit was realized with semi-batch
addition of the β-keto-lactone (i.e., the conditions required for
high yields with our coupling partners), and in fact some yield
erosion was observed at longer addition times. We also
evaluated coupling of 2 + 3 under these 60 °C simple batch
conditions, and confirmed that our coupling partners do not
react cleanly under these conditions.
(2) With simple β-keto-lactones (e.g., 26), this bis-adduct
can be rendered productive by adding AcOH and
increasing the reaction temperature to 60 °C. These
conditions allow for equilibration of the bis-adduct with
the Knoevenagel alkene, which is then reduced to the
desired product.
(3) In the case of β-keto-lactone 2, the presence of the basic
pyridine nitrogen slows this acid-catalyzed equilibration
and requires higher temperatures (>60 °C) or more acid
to establish equilibrium between bis-adduct and
Knoevenagel alkene. At these higher temperatures,
reduction of the aldehyde and product decomposition
(e.g., Scheme 4) begin to compete with the desired
reaction, and thus yields are attenuated. This can be
addressed by use of a semi-batch addition mode at 40 °C,
in which concentrations of β-keto-lactone are kept low,
favoring reduction of the Knoevenagel alkene by the
Hantzsch ester rather than formation of the bis-adduct.
These studies show that relatively subtle variations in
substrate structure, particularly in the nucleophile, can lead to
significantly different outcomes in Hantzsch ester-mediated
reductive couplings. Our particular reaction partners required
development of specific temperature and addition protocols
based on β-keto-lactone reactivity, bis-adduct formation/
equilibration, and the potential for aldehyde reduction. With
simpler β-keto-lactones this level of optimization may not be
necessary, but each set of reacting partners will require some
individual optimization in order to maximize yield.
CONCLUSION
■
We have developed an efficient and scaleable reductive
coupling of β-keto-lactone 2 and aldehyde 3 to form filibuvir
(1). The Hantzsch ester serves as an inexpensive, chemo-
selective, and easily handled reducing agent in this process. The
tendency of our reacting partners to form bis-adduct 6 required
careful optimization of the reaction conditions to minimize its
These data support the following mechanistic rationale:
H
dx.doi.org/10.1021/op400237j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
formation. Minimization of the Ames positive aldehyde 3 in the
final API was a key goal and was realized on multikilogram scale
to meet single digit ppm specifications.
The flow rate was 0.4 mL/min, and the column temperature
was 40 °C. The injection volume was 2 μL, and UV detection
was performed at 265 nm. Solid samples of 1 were prepared at
1 mg/mL in 0.1% formic acid/acetonitrile (50/50, v/v).
Reaction mixtures were prepared for analysis by diluting
approximately 0.1 mL of the mixture to 50 mL with 0.1%
formic acid/acetonitrile (50/50, v/v).
EXPERIMENTAL SECTION
■
Filibuvir (1) is commercially available from Sigma-Aldrich
(Sigma Catalog #PZ0030).
The approximate retention times are listed in Table 3.
HPLC-MS Analysis. The HPLC-MS method used for
determination of low levels of 3 in filibuvir (1) employed an
Agilent 1100 series LC/MSD instrument (Agilent Technolo-
gies, Santa Clara, CA) and a Zorbax SB-CN column (100 mm
× 3.0 mm, 3.5 μm particle size). Mobile phase solution A was
10 mM ammonium acetate in water, and mobile phase solution
B was acetonitrile. Gradient elution was performed as shown in
Table 4 with a run time of 12 min and a re-equilibration time of
approximately 4 min. The flow rate was 0.6 mL/min, the
column temperature was 35 °C, and the injection volume was 5
μL. Mass spectrometric detection was performed using positive
ion electrospray and a group of three ions (m/z 177, 179, 195)
in selected-ion monitoring (SIM) mode. Solid samples of 1
were prepared at 100 mg/mL in methanol for analysis. Under
these conditions compounds 3 and 14 elute at approximately
3.5 and 5.5 min, respectively.
Table 2. Gradient program for UPLC analysis
time (min)
% mobile phase A
% mobile phase B
0
15
17
19
90
65
5
10
35
95
95
5
Table 3. Approximate retention times
cmpd
retention time (min)
1
5.6
4.7
2
6
11.1
13.0
11.6
13.7
12
13
15
Preparation of (R)-6-Cyclopentyl-6-(2-(2,6-diethylpyr-
idin-4-yl)ethyl)-3-((5,7-dimethyl-[1,2,4]triazolo[1,5-a]-
pyrimidin-2-yl)methyl)-4-hydroxy-5,6-dihydro-2H-
pyran-2-one (1). Two procedures are provided below. The
first is the most optimized process executed on multikilogram
scale (15 kg scale Pilot Plant run with n-propyl acetate/aq citric
acid workup, no heat cycle, vacuum distillation). The second is
an 80 g laboratory run with a simplified workup (n-heptane/aq
NaOH/ n-propyl acetate, vacuum distillation). Project
prioritization prevented full evaluation and optimization of
this latter process on multikilogram scale.
16a−d
6.4−6.9 (3 peaks)
Table 4. Gradient program for HPLC-MS analysis
time (min)
% mobile phase A
% mobile phase B
0
5
100
100
90
0
0
0
8
10
8.1
12
100
100
0
Pilot Plant procedure (15 kg scale). THF (28.5 kg) and
methanol (6.3 kg) were combined in reactor #1 and stirred for
5 min, and 3 kg of this mixture was transferred to a drum for
use as a transfer rinse. To reactor #1 were charged β-keto-
lactone 2 (14.6 kg, 42.5 mol, 1.00 equiv, corrected for HPLC
potency of 88.8%) and AcOH (14.1 kg); this mixture was
stirred for at least 15 min at ambient temperature. Reactor #2
was charged with THF (28.5 kg), AcOH (14.1 kg), methanol
Analytical Methods. The UPLC method used for reaction
monitoring and analysis of filibuvir (1) employed a Waters
Acquity series instrument (Waters Corp., Milford, MA) and an
Acquity BEH Shield RP18 column (100 mm × 2.1 mm, 1.7 μm
particle size). Mobile phase solution A was 0.1% formic acid in
water, and mobile phase solution B was acetonitrile. Gradient
elution was performed as shown in the Table 2 with a run time
of 19 min and a re-equilibration time of approximately 2 min.
Figure 2. Structure 16c with arrows indicating key 2D long-range (nJCH, n = 2, 3) heteronuclear couplings (shown in red) and homonuclear
couplings (shown in blue).
I
dx.doi.org/10.1021/op400237j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Table 5. Summary of 1H and 13C NMR data for impurity 16c
phases were combined in a clean reaction vessel, and THF
(38.9 kg) was added. Concentrated aqueous sodium hydroxide
(40 wt %) was then added slowly with continuous pH
monitoring, maintaining an internal temperature of NMT 60
°C, to a target pH of 6.0 (acceptable range 5.5−6.5). For this
batch, 49.2 kg of caustic was required to reach a final pH of
6.24. n-Propyl acetate was then added (130 kg), the
temperature was adjusted to 50 °C, and stirring was maintained
for 15 min. Stirring was stopped, and the phases were allowed
to settle. The lower aqueous phase was separated, and the
product-containing organic phase was washed with a second
portion of water (29 L). Stirring was maintained for 15 min,
then the phases were allowed to settle, and the pH of the
aqueous phase was checked (target pH 6.0, acceptable range
5.5−6.5, measured 5.79). The aqueous phase was then
separated, and the organic phase was speck-free filtered. The
organic phase was then concentrated by vacuum distillation to a
volume of 124 L (100 mbar, final temp ∼30 °C at this
pressure). The vessel was placed under nitrogen, and n-propyl
acetate was added (65 kg); then 100 mbar vacuum was applied,
and distillation was resumed, to a final volume of 124 L (during
the second distillation, pot temperature was 34−41 °C). Water
content was measured by Karl Fischer titration (spec NMT
0.15 wt %, found 0.01%); then the reactor was heated at
ambient pressure to 102−103 °C and held at that temperature
(mild reflux) for 15 min. The solution was cooled to 80 °C at a
rate of 0.5 °C/min, and seeded with 1 (29 g, 0.2 wt %; seeding
was performed to ensure nucleation and prevent occlusion of
impurities in the crystal lattice. (If crystallization had already
occurred, seeding could be omitted.) The slurry was stirred at
80 °C for 60 min, cooled to 10 °C at a rate of 0.5 °C/min, and
held at 10 °C for at least one hour. The solids were collected by
filtration, rinsing with two portions of cold (10 °C) n-propyl
acetate (26 kg each). Drying at 50 °C under a slow stream of
nitrogen provided filibuvir (1) as a white solid (16.8 kg, 33.4
mol, 78% yield).
Streamlined Workup Laboratory Procedure (80 g scale).
To a 2-L, jacketed glass reactor equipped with overhead
mechanical stirrer, temperature probe, condenser, and nitrogen
inlet was charged THF (147 mL) and aldehyde 3 (37.8 g, 206
mmol, 1.00 equiv, corrected for measured potency of 96.1%),
and the resulting slurry was stirred for 30 min at 20 °C.
Methanol (37 mL) and acetic acid (64.4 g, 1.07 mol) were then
added, and the mixture was heated to 70 °C with stirring for 30
min. After cooling to 40 °C, Hantzsch ester 12 (55.9 g, 221
mmol, 1.07 equiv) was charged, and the mixture was stirred at
40 °C for 30 min. A separate and similarly equipped glass
reactor (“charge vessel”) was charged with β-keto-lactone 2
(80.0 g, 206 mmol, 1.00 equiv, corrected for measured potency
of 88.6%) and THF (165 mL). With stirring at 20 °C,
methanol (41 mL) and acetic acid (71.9 g, 1.20 mol) were
added, and stirring was maintained until the solids were
dissolved. The solution of β-keto-lactone 2 was then added to
the aldehyde (3)−Hantzsch ester mixture over a period of 6.5
h, maintaining a temperature in the reactor of 40 °C. The
reaction mixture was then quenched by the addition of water
(354 mL) and n-heptane (709 mL). The resulting mixture was
stirred at 50 °C for 15 min. Stirring was halted, and the layers
were allowed to settle for 15 min. The lower phase (product-
containing aqueous phase) was transferred to the charge vessel,
maintaining a temperature of 50 °C during the transfer. With
stirring, THF (213 mL) was added, and the temperature was
reduced to 30 °C. While monitoring pH, 50% aqueous NaOH
atom
number
13C chemical shift
(ppm)
1H
integral
1H chemical shift
(ppm)
1H
a
multiplicity
2
3
82.25
36.66
−
1
−
2.44
−
d
1
2.69
d
4
174.05
0
−
−
5
95.60
0
−
−
6
7
169.68
0
−
2.35
−
quin
47.43
1
8 (11)
27.61 (27.78)
1 (1)
1.34 (1.56)
1.50 (1.64)
1.49
m (m)
1 (1)
m (m)
9/10
11 (8)
12
26.54/26.56
27.78 (27.61)
38.54
2
m
2
1.56
m
1 (1)
1 (1)
1
1.56 (1.34)
1.64 (1.50)
1.96
m (m)
m (m)
m
m
m
−
s
1
2.07
b
13
14
30.42
153.04
120.20
163.53
24.66
2
2.54
−
2
−
6.73
15/19
16/18
21
−
2
−
3.78
−
−
−
−
dd
−
−
−
s
22
169.19
156.41 (156.49)
147.67
112.51
165.69
31.78
−
−
−
1
25 (43)
27
28
7.00
29
−
4
−
2.57
−
q
b
32/34
33/35
36
14.39
6
1.13
3.56
3.64
2.53
5.52
−
−
−
6.93
t
38.50
1
dd
dd
s
1
37 (50)
38
25.16
65.90
3
1
dd
−
−
−
s
40
169.19
156.49 (156.41)
148.36
111.83
166.13
17.18
−
−
−
1
43 (25)
45
46
47
−
3
−
−
s
49
2.66
2.57
1.22
2.94
50 (37)
DEA
DEA
25.16
3
s
11.69
t
42.78
q
a
s = singlet, d = doublet, dd = doublet of doublet, t = triplet, q =
b
quintet. Overlapped signal.
(6.3 kg), aldehyde 3 (7.5 kg, 42.5 mol, 1.00 equiv, corrected for
HPLC potency of 98.6%), and Hantzsch ester 12 (11.5 kg, 45.5
mol, 1.07 equiv). Stirring was initiated, and the vessel was
warmed to 40 °C for at least 15 min. The contents of reactor
#1 were then added to reactor #2 over a period of at least 6 h,
maintaining an internal temperature of 40 °C. The 3 kg solvent
rinse was then used to rinse reactor #1. A sample was removed
for analysis (0.29% unreacted β-keto-lactone was detected). To
the reaction vessel were then added 10% aqueous citric acid (76
kg) and n-propyl acetate (130 kg), the vessel temperature was
increased to 50 °C, and stirring was maintained for at least 15
min. Stirring was then stopped, and the layers were allowed to
settle (30 min was required). The lower aqueous phase was
separated, and the organic phase was extracted with a second
portion of 10% aqueous citric acid (76 kg), following the same
stirring/settling protocol described above. The two aqueous
J
dx.doi.org/10.1021/op400237j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
(as required for desired pH, ∼124 g, 1.55 mol) was added at a
rate such that the temperature remained below 60 °C, to a final
pH of 6.0 (range 5.5−6.5). n-Propyl acetate (709 mL) was then
added, and the mixture was heated to 50 °C with stirring for 15
min. Stirring was halted, and the layers were allowed to settle
for 15 min. The lower phase (aqueous waste) was removed,
and the remaining product-containing organic phase was cooled
to 18 °C. Speck-free filtration was performed at this point.
While stirring, a vacuum of 100 mbar was applied, and
distillation was performed by heating the mixture (at this
pressure the distillation range was 21−30 °C) to a volume of
∼600 mL (7.5 mL/g β-keto-lactone 2). An additional portion
of n-propyl acetate (709 mL) was added, and vacuum
distillation at 100 mbar was resumed (the distillation range
was 34−41 °C) to a volume of ∼600 mL (7.5 mL/g β-keto-
lactone 2). Water content was measured by Karl Fischer
titration (spec NMT 0.15 wt %); then the mixture was heated
at ambient pressure to 102−103 °C and held at that
temperature (mild reflux) for 15 min. The solution was then
cooled to 80 °C at a rate of 0.5 °C/min and seeded with API 1
(0.14 g, 0.2 wt %; seeding was performed to ensure nucleation
and prevent occlusion of impurities in the crystal lattice (if
crystallization had already occurred, seeding could be omitted).
The slurry was stirred at 80 °C for 60 min, then cooled to 10
°C at a rate of 0.5 °C/min, and held at 10 °C for at least one
hour. The solids were collected by filtration, rinsing with two
portions of cold (10 °C) n-propyl acetate (36 L each). Drying
at 50 °C under a slow stream of nitrogen provides the API (1)
as a white solid (74.2 g, 147 mmol, 71.6% yield).
h, then isolated by filtration, rinsing with two portions of cold
(10 °C) n-propyl acetate (28 mL each, 2 mL/g). After drying in
a vacuum oven with nitrogen bleed at 50 °C, product was
obtained in ∼90% yield.
Characterization of Impurities 15 and 16. . Preparative
HPLC isolation of these impurities provided the following
characterization data:
(E)-2-(2-(5,7-Dimethyl-[1,2,4]triazolo[1,5-a]pyrimidin-2-
1
yl)vinyl)-5-(ethoxycarbonyl)-6-methylnicotinic acid (15): H
NMR (700 MHz, pyridine-d5): δ 10.02 (d, 1H, J = 15.5), 9.22
(s, 1H), 8.66 (d, 1H, J = 15.5), 6.60 (s, 1H), 4.30 (q, 2H, J =
7.2), 3.05 (s, 3H), 2.60 (s, 3H), 2.53 (s, 3H), 1.21 (t, 3H, J =
7.2).
13C NMR (175 MHz, pyridine-d5): δ 170.75, 166.20, 164.97,
164.83, 160.42, 156.27, 155.13, 146.58, 142.05, 134.97, 129.25,
126.10, 124.50, 110.67, 61.30, 25.38, 24.84, 16.62, 14.25.
HRMS (ESI+): m/z calcd for C19H20N5O4 (M + H)
382.15098; found 382.15158.
(R)-6-Cyclopentyl-6-(2-(2,6-diethylpyridin-4-yl)ethyl)-3-((5-
(2-(5,7-dimethyl-[1,2,4]triazolo[1,5-a]pyrimidin-2-yl)-2-hy-
droxyethyl)-7-methyl-[1,2,4]triazolo[1,5-a]pyrimidin-2-yl)-
methyl)-4-hydroxy-5,6-dihydro-2H-pyran-2-one (16c). One
of the four isomers (two regioisomers, each of which is a
mixture of two diastereomers at the newly formed secondary
alcohol) was isolated by preparative chromatography for full
characterization:
HRMS (ESI+): m/z calcd for C24H30N4O3ClS (M + H)
489.17217; found 489.17274.
NMR (700 MHz, CD3CN):
The 1D proton spectrum shows the expected chemical shifts,
multiplicities, and integrations that are consistent with
1H NMR (500 MHz, (CD3)2SO): δ 10.90 (br s, 1H), 7.04 (s,
1H), 6.91 (s, 2H), 3.84 (d, 1H, J = 16.4), 3.72 (d, 1H, J = 16.1),
2.80 (d, 1H, J = 17.6), 2.64 (q, 4H, J = 7.5), 2.59−2.55 (m,
3H), 2.52 (s, 3H), 2.43 (s, 3H), 2.41 (m, 1H), 2.15 (m, 2H),
1.69 (m, 2H), 1.57 (m, 5H), 1.38 (m, 1H), 1.16 (t, 6H, J = 7.6).
13C NMR (125 MHz, (CD3)2SO): δ 166.91, 166.66, 165.12,
163.53, 161.98, 154.90, 151.24, 146.19, 119.17, 110.02, 98.17,
81.44, 45.72, 36.83, 33.01, 30.44, 29.12, 26.50, 26.28, 25.40,
25.34, 24.32, 23.20, 16.07, 13.86.
1
Structure 16c. The D carbon spectrum is consistent based
upon chemical shift. The 2D spectra show the expected
1H−13C and 1H−1H correlations. Key homonuclear and
heteronuclear correlations are indicated in Figure 2. Proton
and carbon assignments are tabulated in Table 5. The
secondary alcohol center at position 38 was not determined.
HRMS (ESI+): m/z calcd for C29H38N5O3 (M + H)
504.2975; found 504.2963.
Anal. Calcd for C29H37N5O3: C, 69.16; H, 7.40; N, 13.91.
Found: C, 69.18; H, 7.60; N, 13.93.
AUTHOR INFORMATION
■
Corresponding Authors
*nathan.d.ide@pfizer.com
*john.a.ragan@pfizer.com
Reprocessing Procedure (if needed to meet purity
specifications). API 1 (14.0 g, 40 mmol) and THF (55 mL,
4 mL/g) were combined and the resulting slurry was heated to
60 °C with stirring until homogeneous. If desired, a speck-free
filtration could be performed at this stage. To the THF solution
was added n-propyl acetate (69 mL, 5 mL/g), and this solution
was cooled to 17 °C (crystallization may occur during this cool
down). Vacuum (100 mbar) was applied, and the solution was
distilled, with occasional addition of n-propyl acetate to
maintain a constant volume of 112−126 mL (8−9 mL/g)
during distillation. Approximately 140 mL (10 mL/g) of n-
propyl acetate was required during the distillation. Distillation
was halted when the solution had reached a temperature of ∼42
°C. When distillation was complete, additional n-propyl acetate
was added to a total volume of 140 mL (10 mL/g), and the
resulting solution was heated to 102−103 °C. This temperature
(reflux) was maintained for 15 min, then cooled to 80 °C at a
rate of 0.5 °C/min. Seeding with pure API (30 mg, 0.2 wt %)
was done unless crystallization had already initiated. The slurry
was maintained at 80 °C for 2 h, and then cooled to 10 °C at a
rate of 0.5 °C/min. The slurry was stirred at 10 °C for at least 3
Present Addresses
§Tetraphase Pharmaceuticals, 480 Arsenal Street, Suite 110,
Watertown, MA 02472.
∥Canterbury Christ Church University, North Holmes Road,
Canterbury, Kent, UK.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
The authors thank John Lucas and Brian Conway for
experiments related to the Mannich-Hydrogenation approach
(Scheme 6), Anil Rane and Donna Brown for preparation of
14C-labelled β-keto-lactone (2), and Silke Wunderwald, Ron
Morris, Mark Zell, and Todd Zelesky for characterization of
impurities 15 and 16.
REFERENCES
■
(1) Li, H.; Tatlock, J.; Linton, A.; Gonzalez, J.; Jewell, T.; Patel, L.;
Ludlum, S.; Drowns, M.; Rahavendran, S. V.; Skor, H.; Hunter, R.; Shi,
K
dx.doi.org/10.1021/op400237j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
S. T.; Herlihy, K. J.; Parge, H.; Hickey, M.; Yu, X.; Chau, F.;
Nonomiya, J.; Lewis, C. J. Med. Chem. 2009, 52, 1255.
(2) Camp, D.; Matthews, C. F.; Neville, S. T.; Rouns, M.; Scott, R.
W.; Truong, Y. Org. Process Res. Dev. 2006, 10, 814.
(3) Johnson, S.; Drowns, M.; Tatlock, J.; Linton, A.; Gonzalez, J.;
Hoffman, R.; Jewell, T.; Patel, L.; Blazel, J.; Tang, M.; Li, H. Synlett
2010, 796.
(4) Hutchins, R. O.; Rotstein, D.; Natale, N.; Fanelli, J.; Dimmel, D.
J. Org. Chem. 1976, 41, 3328.
(5) Xue, D.; Chen, Y.-C.; Cui, X.; Wang, Q.-W.; Zhu, J.; Deng, J.-G. J.
Org. Chem. 2005, 70, 3584.
(6) Jiang, S.; Li, P.; Lai, C. C.; Kelley, J. A.; Roller, P. P. J. Org. Chem.
2006, 71, 7307.
(7) Toth, G.; Molnar, S.; Tamas, T.; Borbely, I. Org. Prep. Proced. Int.
1999, 31, 222.
(8) Chikashita, H.; Miyazaki, M.; Ito, K. J. Chem. Soc., Perkin Trans. 1
1987, 699.
(9) Chikashita, H.; Itoh, K. Bull. Chem. Soc. Jpn. 1986, 59, 1747.
(10) You, S.-L. Chem.Asian J. 2007, 2, 820.
(11) Ramachary, D. B.; Kishor, M. J. Org. Chem. 2007, 72, 5056.
(12) Yang, J. W.; Fonseca, M. T. H.; List, B. J. Am. Chem. Soc. 2005,
127, 15036.
(13) Ouellet, S. G.; Walji, A. M.; Macmillan, D. W. C. Acc. Chem. Res.
2007, 40, 1327.
(14) Kennedy, J. W. J.; Vietrich, S.; Weinmann, H.; Brittain, D. E. A.
J. Org. Chem. 2008, 73, 5151.
(15) Fuchs, K.; Paquette, L. A. J. Org. Chem. 1994, 59, 528.
(16) Mueller, L.; Mauthe, R. J.; Riley, C. M.; Andino, M. M.; De, A.
D.; Beels, C.; De, G. J.; De, K. A. G. M.; Ellison, D.; Fagerland, J. A.;
Frank, R.; Fritschel, B.; Galloway, S.; Harpur, E.; Humfrey, C. D. N.;
Jacks, A. S.; Jagota, N.; Mackinnon, J.; Mohan, G.; Ness, D. K.;
O’Donovan, M. R.; Smith, M. D.; Vudathala, G.; Yotti, L. Regul.
Toxicol. Pharmacol. 2006, 44, 198.
(17) http://www.ema.europa.eu/docs/en_GB/document_library/
Scientific_guideline/2009/09/WC500002907.pdf
(18) http://www.ich.org/fileadmin/Public_Web_Site/ICH_
Products/Guidelines/Quality/Q3A_R2/Step4/Q3A_R2__Guideline.
pdf
(19) Norcross, B. E.; Klinedinst, P. E., Jr.; Westheimer, F. H. J. Am.
Chem. Soc. 1962, 84, 797.
(20) The measured ratio of deuterium/hydrogen in the product
aldehyde 3 corresponds to kH/kD, as breaking of the C−H bond in
D1-1 leads to deuterium incorporation in the aldehyde, and vice versa.
(21) Mechanism B could also be consistent with this data if the
elimination were E2, but this seemed unlikely under these conditions.
(22) Ramachary, D. B.; Kishor, M.; Reddy, V. Eur. J. Org. Chem.
2008, 975.
L
dx.doi.org/10.1021/op400237j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX