Development of a scaleable process for the synthesis of a next-generation statin

Article abstract of DOI:10.1021/op100010n

This manuscript details the process research and development of a convergent and safe approach to 1 on a multikilo scale. Specific highlights of the process development efforts will be described, including the development of a dehydrogenation method for d

Full text of DOI:10.1021/op100010n

Organic Process Research & Development 2010, 14, 441–458
Development of a Scaleable Process for the Synthesis of a Next-Generation Statin
Lindsay A. Hobson,* Otute Akiti, Subodh S. Deshmukh,† Shannon Harper, Kishta Katipally, Chiajen J. Lai,
Robert C. Livingston,‡ Ehrlic Lo, Michael M. Miller, Srividya Ramakrishnan, Lifen Shen, Jan Spink,§ Srinivas Tummala,
Chenkou Wei, Kana Yamamoto,⊥ John Young, and Rodney L. Parsons, Jr.*
Department of Chemical Process Research and DeVelopment, Bristol-Myers Squibb Company, One Squibb DriVe, New
Brunswick, New Jersey 08903, U.S.A.
Abstract:
This manuscript details the process research and development of
a convergent and safe approach to 1 on a multikilo scale. Specific
highlights of the process development efforts will be described,
including the development of a dehydrogenation method for
dihydropyrimidines and a thermochemically safe synthesis of a
1,2,4-aminotriazole fragment. A key feature of the synthesis is the
use and optimization of a modified Julia-Kocienski olefination
reaction. Specifically, we report an unprecedented dependence of
the product olefin geometry on reaction temperature, where an
E:Z ratio as high as 200:1 can be obtained. Initial insights into
the mechanistic rationale for this observation are also provided.
Finally, a purity upgrade sequence via an intermediate crystalline
form is highlighted as a method of controlling the final API quality.
Figure 1. Structure of target compound 1.
644950, 1, was identified for advancement into clinical develop-
ment (Figure 1).5
Results and Discussion
The starting point for the synthetic route development was
the Discovery synthesis outlined in Scheme 1.5 Their approach
involved the preparation of intermediate 6 in a three-step
sequence from the previously reported pyrimidine 2.6 Di-
isobutylaluminum hydride reduction of ester 2 followed by
TEMPO oxidation afforded the corresponding aldehyde 4 in
60% overall yield. Aldehyde 4 was converted to intermediate
6 via the employment of a Julia-Kocienski olefination7 using
the previously described sulfone 5.8 Next, treatment of inter-
mediate 6 with the anion of 1-methyl-1H-1,2,4-triazol-5-amine,
7, generated via deprotonation with lithium hexamethyldisilazide
in THF/DMF, afforded intermediate 8a which on exposure to
methyl iodide provided intermediate 8b. Subsequent deprotec-
tion of the acetonide with HCl and removal of the tert-butyl
group with sodium hydroxide afforded 1a as the sodium salt.
While this sequence met the Discovery material requirements
and was amenable to preparing analogues of 6, consideration
from the process perspective revealed a number of synthetic
challenges.
Introduction
Coronary heart disease is a leading cause of death worldwide.
The risk of coronary heart disease is increased in individuals
having elevated concentrations of plasma low density lipoprotein
cholesterol (LDL-C).1 In the fight against coronary heart disease
statins,2 HMGR inhibitors represent the gold standard in treating
hypercholesterolemia and mixed dyslipidemia. Although many
marketed statins are effective in lowering levels of LDL-C they
have also been linked to a variety of skeletal muscle related
problems such as cramping, myalgia,3 and rhabdomyolysis.4 The
Discovery Chemistry group within Bristol-Myers Squibb fo-
cused on identifying a potent and efficacious HMGR inhibitor
with a potentially superior safety profile; to that end BMS-
* To whom correspondence should be addressed. E-mail: lindsay.hobson@
bms.com; rodney.parsons@bms.com.
First, the synthesis was a long linear sequence requiring the
extensive use of column chromatography, primarily for impurity
† Current address: Chemical and Pharmaceutical Development, Wyeth
Research, Pearl River, NY 10965.
‡ Current address: Process Development, Gilead Sciences, Inc., 333 Lakeside
Drive, Foster City, CA 94404.
(5) Ahmad, S.; Madsen, C. S.; Stein, P. D.; Janovitz, E.; Huang, C.; Ngu,
K.; Bisaha, S.; Kennedy, L. J.; Chen, B.-C.; Zhao, R.; Sitkoff, D.;
Monshizadegan, H.; Yin, X.; Ryan, C. S.; Zhang, R.; Giancarli, M.;
Bird, E.; Chang, M.; Chen, X.; Setters, R.; Search, D.; Zhuang, S.;
Nguyen-Tran, V.; Cuff, C. A.; Harrity, T.; Darienzo, C. J.; Li, T.;
Reeves, R. A.; Blanar, M. A.; Barrish, J. C.; Zahler, R.; Robl, J. A.
J. Med. Chem. 2008, 51, 2722–2733.
§ Current address: Johnson & Johnson, PRD, L.L.C., Welsh and McKean
Roads, P.O. Box 776, Spring House, PA 19477-0776.
⊥ Current address: Department of Chemistry, University of Toledo, 2801 W.
Bancroft St., Toledo, OH 43606.
(1) Third Report of the National Cholesterol Education Program (NCEP)
Expert Panel on Detection, Evaluation, and Treatment of High Blood
Cholesterol in Adults (Adult Treatment Panel III) final report.
Circulation 2002, 106, 3143-421.
(6) Watanabe, M.; Koike, H.; Ishiba, T.; Okada, T.; Seo, S.; Hirai, K.
Bioorg. Med. Chem. 1997, 5, 437–444.
(2) Grundy, S. M. J. Intern. Med. 1997, 241, 295–306. Ross, S. D.; Allen,
I. E.; Connelly, J. E.; Korenblat, B. M.; Smith, M. E.; Bishop, D.;
Luo, D. Clinical Outcomes in statin treatment trials: a meta-analysis.
Arch. Intern. Med. 1999, 159, 1793–1802.
(7) Blakemore, P. R.; Cole, W. J.; Kocienski, P. J.; Morley, A. Synlett
1998, 1, 26–28. Baudin, J. B.; Hareau, G.; Julia, S. A. Tetrahedron
Lett. 1991, 32, 1175. For recent reviews, see: Plesniak, K.; Zarecki,
A.; Wicha, J. Top. Curr. Chem. 2007, 275, 163–250. Blakemore, P. R.
J. Chem. Soc., Perkin Trans. 1 2002, 2563–2585.
(3) For a review, see: Rosenson, R. Am. J. Med. 2004, 116, 408–416.
(4) Graham, D. J.; Staffa, J. A.; Shatin, D.; Rade, S. E.; Schech, S. D.;
La Grenade, L.; Gurwitz, J. H.; Chan, K. A.; Goodman, M. J.; Platt,
R. JAMA, J. Am. Med. Assoc. 2004, 292, 2585–2590.
(8) Brodfuehrer, P. R.; Sattelberg, T. R., Sr.; Kant, J.; Qian, X. Process
for preparing chiral diol sulfones and dihydroxy acid HMG-CoA
reductase inhibitors. PCT Int. Appl. WO/2002/098854.
10.1021/op100010n  2010 American Chemical Society
Published on Web 03/04/2010
Vol. 14, No. 2, 2010 / Organic Process Research & Development
•
441

Scheme 1. Discovery chemistry route designed for SAR exploration
removal and also some of the intermediates were found to be
noncrystalline. A primary goal would be to eliminate the
requirement for chromatographic purification in order to
maximize the throughput of the sequence. The achievement of
this goal would be realized through the identification of
crystalline intermediates throughout the sequence that could
serve as control points for purity. Second, control of the
stereochemical integrity would be of significant concern during
the synthesis due to the acid lability of the acetonide and the
susceptibility of the C-5 hydroxyl to undergo epimerization.
Control would also be needed in the preparation of the trans-
olefin geometry. The synthesis utilized by the Discovery
Chemistry group to prepare aminotriazole, 7, generated meth-
anethiol, a highly flammable and toxic gas which would present
safety concerns for pilot scale manufacture therefore an alterna-
tive synthesis of 7 would be required for multikilogram-scale
production.
Toward this end retrosynthetic analysis of 1 suggests the
late-stage installation of the olefin via sulfone 5, preceded by
disconnection to aminotriazole 7 leading to the known pyrmi-
dine core 10, Scheme 2.9 Strategically, forming the trans-olefin
using a modified Julia-Kocienski olefination7 would provide
a complementary method to other olefination methods previ-
ously utilized on these types of substrates,9 as well as utilizing
established technology from within our department.8 Introduc-
tion of chirality through sulfone 5 at the penultimate step of
the synthesis would minimize the potential for epimerization
at the C-5 hydroxyl position as well as allow for a simple
transformation at the final step, removal of the acetonide and
tert-butyl groups. Sulfone 5 can be prepared from a known
procedure in three steps from the commercially available
Kaneka alcohol as outlined by Brodfuehrer et al.8 The synthesis
of pyrimidine core 10 could be readily achieved in three steps
from the commercially available 4-fluorobenzaldehyde, urea,
and isobutyryl acetate, through a sequence utilized to prepare
a key intermediate in the rosuvastatin synthesis.10
Synthesis of Chloropyrimidine, 10. Utilizing a previously
described procedure chloropyrimidine 10 could be prepared
through a three-component Biginelli coupling reaction,11 fol-
lowed by oxidation and electrophilic activation, Scheme 3.10
The Biginelli three-component coupling reaction effectively
provides pyrimidinone 11 after refluxing the three reagents in
methanol with catalytic copper(I) chloride and sulfuric acid for
24 h. The product crystallized directly from the reaction mixture;
however, during development of the process an exotherm was
observed upon crystallization. To circumvent this exotherm, on-
scale seeding was employed prior to the onset of crystallization.
Conversion of 11 into 12 required an oxidation, which in the
literature had been performed by adding compound 11 directly
into a solution of 65% nitric acid.10 In developing this process
we were cognizant of potential safety risks that could occur
from this reaction, and we sought to investigate the thermo-
chemical aspects of the process. Upon evaluation of the safety
of the reaction it was found that the oxidation chemistry
exhibited the potential for a temperature-dependent uncontrol-
lable runaway as outlined in the graphs in Figure 2.
The graph in Figure 2a shows the kinetic safety profile,
specifically the time-temperature relationship for a thermo-
(9) Koike, H.; Kabaki, M.; Taylor, N. P. Process for the production of
tert-butyl (E)-(6-[2-[4-(4-fluorophenyl)-6-isopropyl-2-[methyl(meth-
ylsulfonyl)amino]pyrimidin-5-yl]vinyl]-(4R,6S)-2,2-dimethyl[1,3-di-
oxan-4yl)acetate. PCT Int. Appl. WO/2000/49014, 2000.
(10) Matsushita, A.; Oda, M.; Kawachi, Y.; Chika, J. Preparation of
aminopyrimidine compounds. PCT Int. Appl. WO/2003/006439 A1,
2003.
(11) Biginelli, P. Gazz. Chim. Ital. 1893, 23, 360–413.
442
•
Vol. 14, No. 2, 2010 / Organic Process Research & Development

Scheme 2. Structural challenges and retrosynthetic analysis of 1
Scheme 3. Synthesis of pyrimidine 10
chemical event. From this plot it can be determined how long
the reaction can be held without cooling before it becomes an
uncontrollable event. What is most concerning from Figure 2a
is the slope of the plot versus the logarithmic scale for time
since this indicates that if cooling were to be lost on the reactor
then the mixture would self-heat and produce an adiabatic
temperature rise of 197 °C, which is the experimental adiabatic
temperature rise (125 °C) corrected by the Phi factor (1.58).12
If this reaction were to reach >40 °C, then it would become
unstoppable until all the reagents are consumed as shown by
Figure 2. Safety evaluation of the oxidation - ARC results.
Vol. 14, No. 2, 2010 / Organic Process Research & Development
•
443

Scheme 4. Synthesis of 12 using metal-catalyzed conditions
Scheme 5. Synthesis of aminotriazole 7
the graph in Figure 2b. Recall that the procedure was to add
11 to the nitric acid, so that, if active cooling was lost and the
reaction reached >40 °C, there would be no controls in place
to stop the reaction. As a result, the nitric acid oxidation was
not viewed as a viable, safe process for scale up, and thus we
sought to identify alternative conditions for scale up. To achieve
that goal many established oxidation protocols were attempted
but were found to be ineffective. For example, the use of DDQ,
CAN, or MnO2 required the use of stoichiometry-based metal
reagents and would not be ideal for use on scale. There were
also issues encountered with the isolation of the product with
respect to removal of the oxidant byproduct. The use of reagents
such as Pd/C, Mn(OAc)3, or NaBrO3 led to low-yielding
reactions with benzylic oxidation of the side-chain as a major
byproduct. After reviewing the literature we turned our attention
to investigating metal-catalyzed conditions which had been
outlined in a Japanese patent and had been employed on a
related series of substrates.13 The protocol involved the use of
a metal catalyst (copper, in particular) together with an oxygen-
based oxidant. Evaluation of these copper-catalyzed conditions
led to the identification of tert-butylhydroperoxide as the
preferred stoichiometric oxidant for the reaction, Scheme 4.
Furthermore, it was found that a base was an essential additive
for the reaction.14 Following a safety evaluation of the copper-
catalyzed conditions we were gratified to find that they were
much safer than the nitric acid conditions, since the self-heating
event did not occur until 170 °C. This procedure consistently
provided 12 in >85% yield and with >99 area % and >97
wt %.
nucleophilic displacement of methanethiol from S-methyl-
isothiourea sulfate with methylhydrazine followed by treatment
of the intermediate guanidine with formic acid.15 From the
process perspective there were a number of concerns identified
with this sequence which precluded it from further scale up,
primarily the evolution of the highly toxic and flammable
methanethiol byproduct together with the need for chromato-
graphic purification. In reviewing the literature for alternative
approaches it was found that many of the known methods for
preparing 1,2,4-triazoles would not be applicable for the
preparation of the 5-substituted aminotriazole 7. A report by
Selby in 1984 outlined a protocol which used dimethylcyano-
carbonimidate and methylhydrazine to prepare 3,5-substituted
triazoles, and this ultimately provided the basis for a modified
synthesis of 7.16 Based on this procedure, a modification of the
cyanocarbonimidate would be needed, and this could readily
be achieved through the use of dimethylformamide dimethy-
lacetal. It was gratifying to find that reaction of cyanamide with
dimethylformamide dimethylacetal in methanol provided uniso-
lated intermediate 13, which on reaction with methylhydrazine
at 30-40 °C formed the desired 5-aminosubstituted triazole 7,
in high overall yield, 75-85%, Scheme 5. The process
development to enable the successful preparation of 280 kg of
7 in high purity (99.9 area%, 99.4 wt %) will be reported in
due course.17
There is, however, one disadvantage to this approach in that
it does not allow for the simultaneous installation of both of
the required methyl groups. To accomplish this in the current
synthesis of 7 a modification of cyanamide to incorporate a
methyl group on the nitrile nitrogen would be required;
unfortunately, this was not feasible. Since the methyl group
could not be directly incorporated during the synthesis of the
Having addressed the thermochemical issues associated with
the oxidation and developed an alternative robust, scaleable
process, the next step was to perform electrophilic activation
of the pyrimidine ring. This could be readily achieved through
treatment of 12 with neat phosphorus oxychloride following
the literature procedure to provide 10.10 Through incremental
modifications, mechanistic understanding, and development of
the dehydrogenation method we were able to synthesize >400
kg of 10 in 70-75% overall yield to enable development of
the subsequent steps.
(12) Data collected in collaboration with SAFC Gillingham, Dorset, United
Kingdom.
(13) Metal-catalyzed oxidations in related systems: Kuwabe, T.; Okuyama,
S.; Hashimoto, S. Jpn. Kokai Tokkyo Koho; JP 10114755 A2
19980506 Heisei, 1998.
(14) For scope and details of the mechanistic study performed on this
reaction see: Yamamoto, K.; Chen, Y. G.; Buono, F. G. Org. Lett.
2005, 7, 4673–4676.
(15) Kroger, C.; Schoknecht, G.; Beyer, H. Chem. Ber. 1964, 97 (2), 396–
404.
Synthesis of Aminotriazole 7. The method utilized by the
Discovery Chemistry group to prepare aminotriazole 7 used a
procedure reported by Kroger in 1964 which involved the
(16) Selby, T. P.; Lepone, G. E. J. Heterocycl. Chem. 1984, 21, 61.
(17) Katipally, K. Synth. Commun. Manuscript in progress.
444
•
Vol. 14, No. 2, 2010 / Organic Process Research & Development

Scheme 6. Completion of pyrimidine core 18
triazole core, attempts were made to methylate compound 7.
Various bases and methylating reagents were screened but with
limited success due to difficulties in separating the desired
compound from a mixture of regioisomers and dimethylated
products. The most promising approach was through a reductive
amination with formaldehyde; however, this met with challenges
ranging from a mixture of products, low yields, and difficulties
in isolating clean product. As a result of these challenges
together with the time constraints in which this fragment was
needed it was decided to investigate the installation of the
methyl group at a later stage in the synthesis.
Completion of Pyrimidine Core 18. With a robust, scale-
able synthesis capable of producing multikilo quantities of 7
the next step required coupling to chloropyrimidine 10. From
the literature it was known that nucleophiles such as N-
methylsulfonamide can displace the chloride;10 however, the
key question for this synthesis to be viable was whether the
coupling could be accomplished with a compound such as 7
which bears a poorly nucleophilic amino function.
reaction mixture by the addition of water, thus eliminating the
need for an extractive workup. The product was isolated in 85%
yield with typical purities of 99.9 area %, and 99.4 wt % on a
260 kg scale. The next step required methylation of the
secondary amine. Having described earlier the previous attempts
to methylate aminotriazole 7 with limited success we were
pleased to find a procedure developed by Shieh18 for the
N-methylation of indoles using dimethylcarbonate as reagent
with catalytic DABCO at 90 °C, Scheme 6. On testing these
conditions with substrate 14 we obtained successful methylation;
however, due to the long reaction time needed to reach
completion several impurities formed which were difficult to
purge during the isolation. Simply increasing the amount of
DABCO to a stoichiometric charge significantly shortened the
reaction time and minimized the impurity formation, whilst
successfully achieving methylation to provide compound 15.
The next challenge was in the isolation of 15; after extensive
solvent screening it was found that 15 could not be rendered
crystalline, and therefore, the crude product would need to be
taken directly into the ester reduction step.
To effect the coupling of 10 with 7, various bases such as
sodium or potassium hexamethyldisilazide and sodium or
potassium tert-butoxide were tried. Of these bases sodium or
potassium tert-butoxide in THF was found to be the most
effective for the coupling although some issues were identified
with this reaction. First, the reactions performed in THF were
heterogeneous due to insolubility of aminotriazole 7. Second,
an impurity derived from the displacement of the chloride with
tert-butoxide was generated in significant amounts, 2-8 area
% in-process, and third, these conditions required an extensive,
volume-inefficient extractive workup followed by crystallization
to isolate the product. To address these issues and streamline
the process alternative reaction conditions were screened
including solvent selection and concentration, and this led to
the identification of DMF as an essential cosolvent. The addition
of DMF led to the solubilization of aminotriazole 7 and also
suppressed impurity formation, presumably due to the more
facile deprotonation of aminotriazole 7 in solution compared
to the heterogeneous mixture. Moreover, following reaction
completion product 14 could be directly crystallized from the
To this end, a telescoped procedure was developed which
involved the addition of toluene to the methylation reaction
mixture. Following an aqueous workup the toluene solution was
dried via azeotropic distillation prior to reduction of the ester
group. Attempts to reduce ester 15 directly to aldehyde 9 were
met with little to no success, and therefore, it was necessary to
reduce ester 15 to alcohol 16 and then reoxidize. To accomplish
this, the previously dried toluene solution of 15 was treated with
di-isobutylaluminum hydride (1.5 M in toluene) at -60 °C to
effect the complete reduction to alcohol 16. Upon reaction
completion the excess hydride was quenched with IPA and the
reaction mixture neutralized by inverse addition into HCl. After
an aqueous workup the product was crystallized from n-heptane/
toluene. During the crystallization of 16 it was found that the
compound had a tendency to entrain toluene which proved
difficult to remove during drying. To displace toluene from the
wet cake a heptane wash was incorporated into the filtration
(18) Shieh, W.-C.; Dell, S.; Bach, A.; Repic, O.; Blacklock, T. J. Org.
Chem. 2003, 68, 1954–1957.
Vol. 14, No. 2, 2010 / Organic Process Research & Development
•
445

Scheme 7. Impurities observed in aldehyde 9
step, and following this 16 was isolated in 80% yield from 10.
During the course of the development activities >500 kg of 16
were produced using the described six-step sequence in 58%
overall yield with high quality and purity (>99.4 area % and
>98.4 wt %).
minimized the formation of these side products. In addition,
process controls were implemented for NaOCl concentration
prior to charging, reaction conversion, and also for acid impurity
17 during the workup.
Typically the TEMPO reaction is buffer dependent due to
the equilibrium composition of aqueous solutions of NaOCl.
When the bleach (Clorox) or the reaction mixture is buffered
(pH 8.5 to 9.5), HOCl is the co-oxidant and is believed to be
distributed between the biphasic reaction mixture. In the
presence of KBr, HOCl is likely converted to HOBr, which is
thought to more efficiently oxidize TEMPO to the oxoammo-
nium ion. One issue observed with buffering the bleach was
that the potency decreased by nearly 50% over a 12 h time
frame. It is also known that higher concentrations of NaOCl
can decrease in potency by 20% per month during storage,22
thus making a titration of the NaOCl essential when needing
to limit overoxidation. To eliminate the observed instability of
the buffered NaOCl and increase the general robustness of the
process, alternative reaction conditions were explored. It was
found that, when nonbuffered/nondiluted NaOCl was charged
in a portionwise manner, the product formation was controlled
on the basis of reagent charge without a latent exothermic event,
as shown by the flat heat signature line in Figure 3. In addition,
one could track the formation of product (on the basis of HPLC
conversion) directly with the theoretical %NaOCl dispensed.
Completion of the Synthesis and Formation of 1. With
the core of the molecule completed the next step was to perform
the oxidation to aldehyde 9, the precursor needed for the
Julia-Kocienski olefination reaction. Taking advantage of
internal knowledge it was decided to develop a TEMPO
oxidation since this had not only been employed by the
Discovery Chemistry team but had also been used within the
process group at scale for a similar compound. During early
development efforts when the typical catalytic system19 involv-
ing TEMPO, KBr as a cocatalyst, aqueous buffered NaOCl,
and DCM as solvent was utilized, a number of processing issues
were immediately highlighted. These included an unsatisfactory
impurity profile, the stability of the NaOCl reagent in the
buffered solution, and the need for a more robust workup and
isolation protocol.
During the development of this step several impurities were
identified, which included the formation of the corresponding
acid 17 as a result of overoxidation and generation of halide
impurities 18 which could potentially form from a decarboxy-
lation pathway of the acid, Scheme 7.20 Another side reaction
observed during the conversion of 16 to 9, was a base-catalyzed
disproportionation of two aromatic product aldehydes 9 into
the corresponding alcohol 16 and acid 17. This Cannizzaro
reaction21 accounts for the instances where small amounts of
alcohol 16 are observed in the isolated solid, though not in the
in-process control sample. Screening reaction conditions using
DOE led to an optimized reaction stoichiometry (0.1 equiv of
TEMPO, 0.05 equiv of KBr, 1.12 equiv of NaOCl) which
(20) Interestingly, removal of 2,2,6,6-tetramethylpiperidin-1-oxyl (TEMPO)
from the reaction does not prevent the formation of this impurity. In
fact, it seems as though the alcohol, aldehyde, and acid all afford this
late-eluting impurity upon treatment with bleach and KBr in CH2Cl2.
A mechanistic understanding for these processes is still being
developed. It should be noted that, although this putative halide is an
impurity in the oxidation of 18, its overall abundance is significantly
low when employing the conditions developed. The crystallization
conditions used to isolate 9 have also been shown to purge most, if
not all, of this side product.
(21) Cannizzaro, S. Ann. 1853, 88, 129. List, K.; Limprich, H. Ann. 1854,
90, 180.
(19) Anelli, P. L.; Biffi, C.; Montanari, F.; Quici, S. J. Org. Chem. 1987,
52, 2559–2562. Anelli, P. L.; Montanari, F.; Quici, S. Org. Synth.
1990, 69, 212–219.
(22) Mohrig, J. R.; Nienhuis, D. M.; Linck, C. F.; Van Zoeren, C.; Fox,
B. G.; Mahaffy, P. G. J. Chem. Educ. 1985, 62, 519–521.
446
•
Vol. 14, No. 2, 2010 / Organic Process Research & Development

Figure 3. Graph of pH versus addition time.
Figure 4. Raman profile to show conversion of 16 into 9.
The nonbuffered conditions also led to reduced levels (<0.15
AP) of the overoxidized product, 17.
was an induction period during the addition of the NaOCl
solution, and it could also be used to avoid reagent accumulation.
The development of a crystallization protocol for 9 employed
a standard screen of sixteen (16) solvents with six (6) antisol-
vents and identified two solvent systems: 1:1 EtOH/H2O and
1:1 IPA/H2O. Investigation of the EtOH/H2O crystallization
protocol provided excellent yield and purity; however, needle-
like crystals were produced which showed a strong tendency
to agglomerate, Figure 5. These agglomerated crystals made
In order to minimize sampling, an in-line Raman spectros-
copy method was investigated to monitor the oxidation of 16
to 9. The in-line Raman was able to effectively track both
starting material and product during the reaction and the signal
correlated well with the HPLC data taken intermittently during
the course of the reaction, Figure 4. The Raman could be further
used to ensure safe execution as it could detect whether there
Vol. 14, No. 2, 2010 / Organic Process Research & Development
•
447

Figure 5. Comparison of crystals from different solvent
Figure 6. Impact of warming on the selectivity and yield of
the modified Julia olefination.
systems.
Scheme 8. Improved TEMPO process
to be unstable even at cryogenic temperatures, conducting the
reaction under Barbier conditions7 provided satisfactory conver-
sion with only 1.1 equiv of the sulfone required. Procedurally,
the reaction was executed by addition of LiHMDS (1 M solution
in THF) to a mixture of sulfone 5 and aldehyde 9 in THF below
-70 °C, followed by cold quenching with aqueous ammonium
chloride, to provide the desired alkene 8 in 85% yield. Under
these conditions, the E-olefin was generated with 15:1 selectivity
over the Z-isomer, which was effectively purged during the final
crystallization of 8b from ethanol and water. Implementation
at the kilogram scale (3 × 3.5 kg of aldehyde 9) provided olefin
8b in consistently high yield (85-86%) and quality (99.7-99.8
area %).
filtration and washing operations difficult; therefore, optimiza-
tion of the IPA/H2O conditions were pursued. The IPA/H2O
solvent system provided more granular crystals that did not
agglomerate and thus facilitated filtration and drying. An
additional benefit of the IPA/water system was that it was
superior in purging process impurities to <0.1 area %, while
yielding compound 9 in 94% yield with >99.36 wt %.23
The oxidation process underwent several process improve-
ments of which the most important from a process perspective
was to use nonbuffered NaOCl. As noted earlier this is outside
the experiences of what has been generally discussed in the
literature and circumvents the reagent instability of buffered
NaOCl. The use of Raman analysis in conjunction with HPLC
for reaction monitoring was demonstrated during pilot-plant
implementation at 30 kg scale to produce 9 in 93.5% yield,
99.88 area %, and 99.63 wt %. Scheme 8.
The final bond-forming reaction in our convergent synthesis
was to couple chiral fragment 5 with aldehyde 9 via an
appropriate olefination method. To utilize previously developed
technology from within our department we focused our efforts
on the modified Julia olefination,7 specifically on phenyltetra-
zolyl sulfone 5 derived in three steps from the commercially
available Kaneka alcohol. While the lithium salt of 5 was found
Although 15:1 selectivity was sufficient for initial purposes,
we sought to improve the performance of the olefination
reaction. To accomplish this goal, a number of reaction
parameters (temperature, deprotonating base, counterion of base,
and reaction solvent) were evaluated to ascertain their influence
on the observed E:Z selectivity in our system. On the basis of
these studies, temperature was found to have the most significant
impact on the observed E:Z ratio. Surprisingly, it was not the
temperature at which the initial reaction was performed but
rather the temperature post reagent addition (i.e., reaction age
temperature) that impacted the observed E:Z ratio. Specifically,
our temperature-selectivity dependence studies focused on three
specific process phases: the reagent addition temperature, the
reaction age temperature, and the quench temperature. When
the deprotonating base is added at higher temperatures (>-78
°C) the selectivity for the E-olefin is reduced, and the impurity
profile for the reaction is also negatively impacted. To assess
the influence of reaction age temperature on selectivity we
conducted a series of parameter matrix studies where the age
temperature and hold periods were varied. The temperatures
evaluated ranged from -70 to 10 °C and hold times from 30
min to 2 h. These studies clearly showed an increase in the
observed E:Z ratio when the reaction mixture was aged at
noncryogenic temperatures, Figure 6. The reaction age duration
also impacted both the observed E:Z ratio and the measured
product HPLC solution yield. Specifically, longer hold times
(>1 h) led to a significant decay in solution product content
and final isolated yield. This degradation process provided a
complex mixture of side products, but the primary pathway
involved ꢀ-elimination of the acetonide. The final optimized
process conditions which provided the best balance of E:Z ratio
and product yield involved reagent additions at -70 °C, then
warming the reaction mixture to -10 °C and aging for 1 h at
-10 °C, followed by reaction quench. These conditions afforded
an E:Z selectivity of >100:1 in favor of the E-isomer.
(23) In addition to these optimization studies, a great deal of effort was
placed on finding a alternative isolation/crystallization possibility in
order to avoid the solvent swap from EtOAc to IPA. After a subsequent
screen of solvents and anti-solvents, a crystallization mixture of EtOAc/
heptane was chosen for further refinement and development. Solubility
of 9 was mapped at different temperatures and various solvent
compositions, and 3:1 heptane/EtOAc was chosen as a potential
crystallization solvent mixture. Although this provided a crystallization
directly from the reaction solvent, an easier filtration due to the larger
granular crystals generated, and improved drying efficiency, the
recovery was plagued with issues during discharge from the reactor
(i.e., a semi-stickiness of the resulting solid or possible nucleation/
crystallization on the walls of the vessel). Diminished yields (63 M%
for 1:1 heptane/EtOAc; 73-79 M% for 3:1 heptane/EtOAc) due to
material loss and inherent solubility of 9 were also observed. Further
exploration into temperature effects on the crystallization, different
ramping protocols, temperature cycling variants, and seeding regimes
were found to be successful and to dramatically reduce the amount
of fines present in the final solid. These improvements furnished
very large granular crystals and an increased yield (9: 92 M%, 100
wt%).
448
•
Vol. 14, No. 2, 2010 / Organic Process Research & Development

Scheme 9. Formation of 8b through a modified Julia olefination
To the best of our knowledge this specific temperature-
influencing phenomenon has not been previously reported in
the literature. Due to the unexpected nature of these findings
and the dramatic enhancement of selectivity, we were prompted
to try to elucidate its mechanistic origins. To further evaluate
and understand the mechanistic underpinnings and scope of this
phenomenon we investigated the reactions with a simplified
model system, specifically reactions of benzaldehyde with
compound 5. The parameters studied included the reaction
solvent, counterion of the deprotonating base, and reaction
temperature. When benzaldehyde was reacted with sulfone 5
(LiHMDS, THF, -78 °C) an E:Z ratio of 60:40 in favor of the
Z-isomer was observed when the mixture was aged and
quenched at -78 °C. In contrast, warming the reaction mixture
to -10 °C and holding for 1 h, resulted in a selectivity reversal
and enhancement to afford an 80:20 mixture in favor of the
E-isomer. Next, we evaluated the impact of solvent polarity.
Aging of the reaction mixtures at noncryogenic temperatures
in less polar solvents resulted in an increased E:Z selectivity
(solvent selectivity trend was DMF < THF < DME < MTBE <
toluene) Changing the counterion of the deprotonating base (Li,
Na, K evaluated) also influenced the observed selectivity.
Overall, the results indicated that weakly coordinating coun-
terions provide lower E:Z ratios (1.5:1 versus 6:1).
implemented in the pilot plant (2 × 27 kg of aldehyde 9) to
provide 8b in 80% average yield with 99.9 area % purity,
Scheme 9.
The final stage in the synthesis required the double depro-
tection of the acetonide and tert-butyl ester groups. The chiral
fragment of the molecule was sensitive and susceptible to
epimerizations, oxidations, annulations and lactonizations,
therefore conditions would not only need to be carefully selected
to minimize impurity formation but also the order in which the
protecting groups were to be removed would be critical. A key
consideration for the deprotection sequence was the identifica-
tion of a crystalline compound for isolation, since it was known
that the free acid was not crystalline. Initial work showed that
diol 19 was also not crystalline therefore a suitable salt form
would need to be identified. Although the salt would not need
to be the final salt form for development, the following attributes
would be required: crystalline; robust preparation; and remove
process impurities. From the perspective of stability it was found
that the acetonide needed to be removed first under acidic
conditions, followed by removal of the tert-butyl ester group
under basic conditions since this minimized the formation of
impurities due to epimerization and lactonization, Scheme 10.
Utilization of this strategy would provide a quality control
intermediate, which could then be converted into the final
pharmaceutically acceptable salt form.
The conclusions from this work are that less polar solvents
provide improved E-selectivity under thermodynamic conditions
and that the selectivity and isomerization efficiencies are
dependent upon the counterion of the base. On the basis of
results from the optimization and mechanistic model studies
we postulated that the increased selectivity may be due to an
equilibration of the initial kinetic intermediate ꢀ-alkoxysulfone
via a retroaddition/addition pathway. Other mechanistic path-
ways could be considered, further mechanistic and scope studies
are ongoing and will be reported in due course.
Other improvements to the process included replacement of
the ammonium chloride quench and sodium bicarbonate wash
with two potassium bicarbonate washes. This streamlined
processing and was more effective in removing the phenyltet-
razolone byproduct, which was not well tolerated in downstream
processing. With potassium bicarbonate solutions of sufficiently
high ionic strength, it was found that clean phase separations
could be achieved without the addition of ethyl acetate as
cosolvent to the THF reaction mixture, which improved volume
efficiency and greatly shortened the cycle time of the subsequent
solvent exchange. The streamlined process was ultimately
The initial conditions attempted were previously utilized on
a similar motif and involved hydrochloric acid (6 N, 3 equiv)
in tetrahydrofuran at 20-25 °C to remove the acetonide
followed by sodium hydroxide (5 equiv) to remove the tert-
butyl group.24 Employing these conditions with compound 8b
three impurities formed: the epimer at C-5, 20, and the
corresponding lactones 21 and 22 as shown in Figure 7. Using
methanol and 2N HCl (2 equiv)25 at 0-20 °C, provided the
intermediate diol ester containing epimer (0.5-1 area %) 20
and a mixture of the tert-butyl ester 19 and methyl ester, both
of which could be converted to the carboxylate during the
sodium hydroxide step. Reducing the concentration of acid (1
N) provided compound 1 with reduced levels of 20.26 Condi-
tions using a catalytic amount of HCl with concentrations of
0.01 N and heating to 40 °C have also been demonstrated on
similar compounds.27 Due to the vast number of combinations
for the conditions for the deprotection sequence, it would be
difficult to select processing conditions through a typical
(24) Private communication with X. Qian.
(25) Private communication with R. Waltermire.
(26) Private communication with S. Ahmad.
Vol. 14, No. 2, 2010 / Organic Process Research & Development
•
449

Scheme 10. Strategy for removal of the acetonide and tert-butyl groups
univariate approach. An initial screen of conditions was pursued,
followed by a more focused multivariate design of experiments
to further define the parameters.
sodium hydroxide (2 N, 1.5 equiv) at 20-25 °C.28 These
conditions also resulted in the saponification of lactones, 21
and 22, which formed during the acetonide deprotection. Having
developed the deprotection process, all that remained was to
identify an acceptable method to isolate the product, Vide infra,
by selection of a suitable salt form of the carboxylic acid. The
free acid was a noncrystalline solid which on storage would
degrade to epimer 20 and lactones 21 and 22; the proposed
mechanism for decomposition is shown in Scheme 11. Efforts
were focused on identifying a salt of 1 since a salt would not
be expected to lactonize or epimerize as readily as the free acid.
Initial salt screening focused on pharmaceutically acceptable
salts and identified three potential counterions (Na, Ca, and Zn)
for development; however, they demonstrated poor crystalliza-
tion properties, and moreover none of these salts provided any
purification upgrade for removing the process-related impurities
formed during the deprotection steps. During a screen of amine
bases several crystalline leads were identified, methylamine,
ammonia, dicyclohexylamine, glycine esters, which were then
further investigated for use as an intermediate salt form for
purification and isolation. To our surprise the ammonium salt
was found to be unusually stable. Single crystal X-ray analysis
revealed an array of multiple hydrogen bonds in the crystal
structure (which also contained one molecule of water), and
the unexpected stability is attributed to this. Ultimately, the
ammonium salt was selected as the intermediate salt form to
enable isolation from the deprotection sequence due to its ability
to purge impurities (Table 2). Only one impurity, 19, remained
at a higher level than the typical qualification threshold, and
this was acceptable since it had been qualified in toxicological
studies.
To gain more understanding of the conditions for the
deprotection sequence, a screen of different acids (HCl, MSA,
sulfuric, acetic, and phosphoric) and solvents (MeCN, THF,
acetone, IPA, MeOH, EtOH, DMF, and NMP) was performed.
This showed that HCl, sulfuric, and MSA in MeCN or MeOH
provided the best results. A larger multivariable screen inves-
tigating the design space for the acid equivalents, molarity,
temperature, and concentration was then performed, focusing
on hydrochloric acid with MeCN and MeOH as outlined in
Table 1.
From this study it was found that temperature had the most
impact on reaction rate with acid concentration having the next
greatest effect. Temperature was shown to impact the degree
of epimerization at the C-5 hydroxyl, however this could be
minimized through the use of catalytic, dilute acid. Based on
these results the deprotection step was further optimized to use
a catalytic amount of dilute acid at elevated temperature
conditions. The final protocol to remove the acetonide used 0.05
equiv of 0.02 N HCl in acetonitrile (10 vol) at 40-45 °C. After
3 h, <2 area % of 8b typically remained. Saponification of the
tert-butyl ester was then readily achieved through the use of
The final chemical sequence was then optimized into a
telescoped process such that after deprotection of the acetonide
and hydrolysis of the tert-butyl ester, the pH was adjusted to
∼2 with 1 N hydrochloric acid, and the product was extracted
into isopropyl acetate. The isopropyl acetate layer was then
water washed and treated directly with either a solution of
ammonia in ethanol (2 N) or with ammonia gas. Since the
Figure 7. Process impurities observed in the final step.
Table 1. Ranges of conditions tested for the deprotection
reaction
concentration,
mL/g
acid
concentration
acid
equiv
temp,
°C
35-49
35-49
(27) Taylor, N. P.; Okada, T. Crystalline salts of 7-[4-(4-Fluorophenyl)-
6-isopropyl-2-[methyl(methylsulfonyl)amino]pyrimidin-5-yl]-(3R,5S)-
3,5-dihydroxyhept-6-enoic acid. (AstraZeneca UK Limited and Shion-
ogi & Co., Ltd.). PCT WO/2001/60804 A1 2001.
(28) The removal of tert-butyl esters is typically performed under acidic
conditions. The basic conditions had been used previously on a similar
compound. Acidic conditions for this molecule would have led to the
formation of epimer 20 and lactones 21 and 22.
solvent
MeOHa
MeCN
15-30
7.5-12.5
0.01-0.04
0.015-0.03
0.02-0.08
0.03-0.07
a Due to the solubility characteristics of compound 8b higher volumes of
MeOH were needed relative to MeCN.
450
•
Vol. 14, No. 2, 2010 / Organic Process Research & Development

Scheme 11. Proposed mechanism for the formation of 20, 21 and 22
Table 2. Typical purification upgrade during the formation
of the ammonium salt with respect to impurities 19-23
Scheme 12. Formation of ammonium salt 1b from 8b
area
area
area
area
area
area
sample
%a 1 % 19 % 20 % 21 % 22 % 23
aqueous phase
98.65
0.92
0.89
0.44
0.19
0.21
0.09
0.09
<0.03
<0.03
<0.03
<0.03
<0.03
<0.03
postdeprotection
organic phase prior 98.44
to salt formation
dried cake
99.56
<0.03
<0.03
a Normalized area %.
ammonium salt forms as a monohydrate, it was also essential
to have a stoichiometric amount of water present prior to
crystallization. By determining the amount of water needed as
well as the solubility of water in the organic phase, it was found
that additional water charges post aqueous washes were not
needed to consistently deliver the salt as its monohydrate. In
order to control the nucleation event and crystallization process,
seeding with 0.5% of 1b was implemented. Scheme 12 shows
the final sequence from 8b to 1b.
Final Form Selection. As noted previously, a significant
process challenge for this API was the lack of a suitable
crystalline form for development. Despite the beneficial proper-
ties of the ammonium salt, (high degree of crystallinity, excellent
purity upgrade, reproducible preparation), it was not suitable
for use as a final salt form. One major concern was the potential
odor due to loss of ammonia in the solid state during long-
term storage and further processing. Data from ICH solid-state
stability studies performed on ammonium salt 1b indicated three
issues: loss of ammonia when stored in open conditions at
elevated temperatures with significant formation of lactone 21;
instability to high-intensity light, with the formation of diaster-
eomeric electrocyclic annulation adducts 23; and a susceptibility
to oxidation when exposed to air with the C-5 hydroxyl being
oxidized to form an enone 24, Figure 8.
In comparison to the ammonium ion, the calcium ion is a
pharmaceutically acceptable counterion, and the calcium salt
1c exhibited acceptable solid-state chemical stability with no
lactone formation observed in samples stored at elevated
temperatures and humidity, see Table 3. The calcium salt was
selected as the final salt form for oral solid dosage development,
and this selection was further supported by the prevalence of
other marketed statin drugs as calcium salts (e.g., Atorvastatin
and Rosuvastatin).29,30
Vol. 14, No. 2, 2010 / Organic Process Research & Development
•
451

appeared to be a degree of amorphous content as indicated by
the halo and broad peak shape in the pXRD pattern, Figure
10b. It is important to note that the slurry pattern can be obtained
reproducibly when dried powder is reslurried in water; therefore,
the hydration/dehydration of the crystals appears to be a
dynamic process.
Further evidence for 1c exhibiting dynamic hydration was
obtained from dynamic vapor sorption studies indicating that
1c was hygroscopic as shown by the gradual weight gain with
increasing relative humidity, Figure 11. The lack of a step-
isotherm in both adsorption and desorption curves can be
indicative of a loose water channel structure for the solid or
due to amorphous content in the solid.
Figure 8. Degradation products observed on stability.
Table 3. Solid state stability of 1c and 1b at specified
storage conditions
From a processing perspective the calcium salt presented
significant challenges to the development of a reproducible
process for scale up. Two possible approaches were identified
for the preparation of 1c. The first approach involved preparing
the free acid from 1b followed by treatment with a source of
calcium ions, and the second approach involved a direct salt
exchange from the ammonium to the calcium salt. Upon
evaluation of the first approach two issues were immediately
identified which precluded further development: the formation
of an unstirrable, amorphous gum phase and the instability of
the free acid with respect to degradation and impurity formation
as discussed earlier. Since this approach was not viable, we
turned our attention to the second approach of the direct salt
exchange process, Scheme 13. After determining the aqueous
solubilities of the ammonium and calcium salts it was found
that there was a 2-fold difference in their solubilities which
would be advantageous for using an aqueous system for the
preparation of 1c. Ammonium salt 1b was dissolved in water
(20 vol) at 50 °C and treated with a source of calcium ions.
Various calcium ion sources such as calcium hydroxide, calcium
chloride, and calcium acetate were evaluated, and all provided
conversion to the calcium salt. Calcium chloride was selected
since it could be added as an aqueous solution. During the early
stages of the calcium ion addition, an amorphous gum phase
formed which eventually converted to a crystalline material.
After exploration of a variety of different addition methods,
(calcium ion source added to 1b or vice versa), it was found
that no method was devoid of gum formation. Although this
approach appeared straightforward, in practice there were
1c amount remaining 1b amount remaining
area %
area %
storage
condition
1 week
3 weeks
1 week
3 weeks
50 °C (closed)
40 °C/75% RH (open)
98.1
98.6
98.2
98.6
98.6
96.7
98.0
85.0
The calcium salt 1c did, however, present challenges with
regard to its characteristics and preparation. Water was an
essential component of the crystal lattice; however, the isolated
crystals were found to lose water readily, resulting in an
amorphous phase. Overall, the calcium salt was characterized
as a weak hydrate with the differential scanning calorimetry
(DSC) trace showing a broad endotherm at ∼85 °C due to
dehydration and a weak endotherm at ∼130 °C due to melting
of the crystals (confirmed by hot stage microscopy), Figure 9.
Thermal gravimetric analysis (TGA) of 1c shows a weight loss
between 50 and 110 °C which was attributed to dehydration of
the hydrate form, Figure 9. The weight loss of different batches
was shown to be variable, depending on the drying conditions
used in the preparation. Under controlled drying conditions
material having a weight loss between 4 and 7% could be
consistently obtained.
Upon examination of the pXRD patterns of various batches
of the slurry, wet cake, and dried solids it was found that the
patterns varied, depending upon the degree of hydration, Figure
10a. These discrepancies could be attributed to either the
existence of a different phase in the slurry or a different
hydration state in the solid. Of the calcium salt batches prepared
a reproducible pXRD pattern was obtained, although there
Figure 9. Typical DSC and TGA thermograms of 1c.
452
•
Vol. 14, No. 2, 2010 / Organic Process Research & Development

Figure 10. PXRD patterns of 1c Figure 10b PXRD pattern of 1c dried solid.
Figure 11. Adsorption/desorption curves of 1c.
several aspects which hampered its further development: (1)
the high volume of water required for dissolution of 1b
negatively impacted the yield of the reaction, 76% due to the
aqueous solubility of calcium salt 1c; (2) dissolution of 1b at
50 °C resulted in impurities 20, 21, and 22; and (3) the formation
of an amorphous gum phase even when a large excess of seeds
were used. The formation of the gum phase is due to product
precipitation as a result of the steep solubility profile of the
calcium salt in this aqueous system. Furthermore, the crystal-
lization demonstrated slow nucleation kinetics and crystal
Vol. 14, No. 2, 2010 / Organic Process Research & Development
•
453

Scheme 13. Preparation of 1c from 1b
theoretical yield (based on solubility of 1c) could be increased
significantly (76-94% based on a total water charge of 30 L/kg
input). For the heterogeneous protocol 1b was suspended in
water at a concentration of 10 L/kg input, and a solution of
CaCl2 (11 equiv) in water (10 L/kg) was added to the slurry.
The mixture was stirred at 20 to 25 °C for 1 h, heated to 50 °C
over 2 h to complete the salt exchange, and aged at 50 °C for
12 to 24 h. This protocol provided several advantages: the
material was crystalline throughout the entire salt formation
process with no gum or amorphous phase observed, as verified
by optical microscopy, Figure 12; there was no requirement
for dissolution of 1b prior to the addition of CaCl2; therefore,
additional water could be used for the CaCl2 charge (10 L/kg
input) alleviating the abrupt supersaturation change that had
been observed during the direct crystallization process. Seeding
was not needed; however, it was employed to ensure polymorph
control during the crystallization.
During this crystallization process the heterogeneous mixture
goes through a very thick stage similar to that of the direct
crystallization process due to the intrinsic fine-needle morphol-
ogy of the crystals. The thickness of the slurry presents potential
challenges for agitation, and insufficient agitation has been
shown to lead to a stagnant mixing zone near the walls and
bottom of the vessel. Over time the material in this stagnant
zone can build into a shell or crust-like solid mass which
hampers isolation. Another potential issue with crust formation
is that the material may not mix with the bulk mixture and can
lead to the presence of residual ammonium salt 1b in the
product. These problems were addressed by employing a two-
stage water addition strategy, which involved a second water
charge (6 L/kg input) after the slurry goes through its thickest
stage. This allows any crust formed to be dislodged from the
vessel walls into the bulk mixture as well as facilitating agitation
by making the mixture less viscous. The additional water charge
has an insignificant impact on yield since the solubility curve
is essentially flat when using 3-5 wt % CaCl2. During the
transformation, Raman technology was employed to monitor
the rate of formation of 1c as well as to ensure complete
conversion of 1b. Compounds 1b and 1c have distinct peaks
in the 980-1180 cm-1 and 1480-1550 cm-1 range that makes
growth rate. The crystal morphology further exacerbated the
situation due to being a thin, short needle shape which
demonstrated poor filtration rate properties at scale.
In an attempt to address these issues, modifications of this
procedure were investigated. A cubic addition profile was
employed for the charge of the aqueous CaCl2 solution in an
attempt to relieve the rapid supersaturation buildup that plagued
linear addition protocols. A cooling step from 52 to 40 °C over
1 h was added after 30% of the CaCl2 charge, in order to
establish a seed bed prior to completing the CaCl2 addition.
Several temperature cycles between 20 and 50 °C (2 h heating,
15 h aging at 50 °C, followed by a 2 h cooling to 20 °C) were
also added to the process in an attempt to increase the particle
size by removing fines and thus facilitating filtration.
Although these modifications eliminated the formation of
the gum phase through a specific CaCl2 addition and seeding
protocol, amorphous material was still evident during the initial
stage of the process, and the crystal morphology was still thin,
short needles. Also, the yield for these protocols was only
60-65%, due to a significant amount of material lost to the
mother liquor and washes due to the relatively high solubility
of 1c in water (10.6 mg/mL). Due to the limitations of these
protocols we focused the development efforts on a heteroge-
neous transformation versus the traditional direct crystallization.
During the development of the heterogeneous protocol it was
found that the presence of excess CaCl2 dramatically reduced
the aqueous solubility of 1c. A 5-6-fold solubility reduction
of 1c (10.6 mg/mL in pure water versus 2 mg/mL in 3-5 wt
% aqueous CaCl2) could be achieved through the addition of
extra CaCl2. Based on this finding, a modified protocol was
developed using 10-12 equiv of CaCl2, and as such the
Figure 12. Evolution of crystals during the formation of 1c via the heterogeneous crystallization protocol.
454
•
Vol. 14, No. 2, 2010 / Organic Process Research & Development

Figure 13. (a) Raman spectra for the conversion of 1b to 1c (b) Kinetic monitoring of the transformation of 1b to 1c by Raman
technology.
Raman monitoring possible, Figure 13a. Figure 13b shows that
the salt exchange from 1b to 1c starts at 20 to 25 °C and is
completed once the batch temperature reaches 50 °C. The
additional aging at 50 °C is employed to improve the particle
size of the product but is not a requirement for complete
transformation.
Upon filtration, the cake was found to be highly compressible
and retained mother liquor, thus making efficient deliquoring
and washing difficult. After evaluating several filtration options,
it was found that the use of a plate filter with a minimal head
pressure together with charging the slurry in portions to build
up the cake thickness was the preferred method for filtration.
To displace the mother liquor retained in the cake, water and
(29) Roth, B. D. Preparation of trans-6-[(carbamoylpyrrolyl)alkyl]-4-
hydroxypyranones as hypocholesterolemics. U.S. Patent 4,681,893,
1987.
(30) Hirai, K.; Ishiba, T.; Koike, H.; Watanabe, M. Pyrimidine derivatives
as HMG-CoA reductase inhibitors. Eur. Pat. Appl. 1993, EP 521471,
1993.
Vol. 14, No. 2, 2010 / Organic Process Research & Development
•
455

heptane washes were incorporated, and the product was then
dried in an agitated dryer until the KF was less than 10%. By
utilizing the improvements described above the heterogeneous
crystallization protocol was demonstrated on a 28 kg scale to
provide the product in 80% yield with 99.3 area % and 82.8
wt % versus the free acid theoretical of 89.7 wt %.
11 by HPLC). The solution was cooled to 20-25 °C and treated
with a mixture of aqueous Na2S2O3 (0.5 M solution, 7.0 L)
and 25 wt % aqueous NH4Cl (3.5 L), and the resulting biphasic
mixture was stirred vigorously for 60 min. The pH of the
aqueous and organic phases should be ∼7.5-8.0. The two
phases were separated, and the absence of remaining oxidant
was checked using a peroxide test strip. If needed, an additional
thiosulfate wash can be performed. The organic phase was
concentrated to minimum volume via reduced pressure distil-
lation. During the distillation, product 12 precipitated from the
solution as a white solid. A solvent exchange to heptane was
performed until the amount of residual CH2Cl2 was <10 vol %
by GC. The product was filtered and washed with filtrate
followed by heptane (1.0 L), and the wet cake was dried under
vacuum (∼27 Torr) at 40 °C until an LOD of <1.0% was
achieved. 12 was obtained in 95.8% isolated yield (666.1 g) as
a white powder with HPLC purity of 99.6 area % and 97.6 wt
%.
Methyl 4-(4-Fluorophenyl)-6-isopropyl-2-(1-methyl-1H-
1,2,4-triazol-5-yl)amino)pyrimidine-5-carboxylate (14). A
400-L glass-lined reactor was purged with nitrogen and charged
with dimethylformamide (93.5 kg), chloropyrimidine 10 (19.8
kg, 64.13 mol), and aminotriazole 7 (7.55 kg, 76.95 mol). The
suspension was cooled to 0-5 °C, and potassium tert-butoxide
(1.0 M solution in THF, 115.7 kg, 128.26 mol) was charged
over 1 h, maintaining the batch temperature <15 °C. At the
end of the potassium tert-butoxide addition, the reaction mixture
temperature was raised to 18-25 °C, stirred for 1 h, and then
sampled for HPLC analysis. HPLC typically indicated less than
0.5 area % of 10 remaining in the reaction mixture. If the
amount of 10 remaining is g0.5%, an additional amount of
base should be added. At this point the majority of the THF
was removed by distillation at <40 °C under reduced pressure
(100-130 Torr). The product was precipitated by the addition
of water (198 kg) over 2 h while maintaining the temperature
at 25-35 °C. During the water addition, the initial cloudy
reaction mixture became clear and homogeneous. After ∼2/3
of the water has been added the product begins to crystallize.
The slurry was stirred for >20 h at 20-25 °C and was then
filtered. The wet cake was washed with a mixture of DMF/
water (60 L 1:3 by volume) followed by 60 L of water. The
wet cake was dried under vacuum at <50 °C. Compound 14
was obtained in 86% yield, ∼20.38 kg, with an HPLC purity
of 99.9 area % and 99.4 wt %. 1H NMR (400 MHz, CDCl3):
δ 1.21 (d, J ) 6.59 Hz, 6H), 3.12 (m, 1H), 3.65 (s, 3H), 3.78
(s, 3H), 7.08 (dd, J ) 8.33, 8.83 Hz, 2H), 7.55 (dd, J ) 5.12,
8.66 Hz, 2H), 7.85 (s, 1H), 8.97 (s, 1H). 13C NMR (100 MHz,
CDCl3): δ 22.3, 33.7, 36.7, 53.0, 115.6 (d, J ) 21.7 Hz), 118.1,
129.9 (d, J ) 8.3 Hz), 133.7 (d, J ) 3.3 Hz), 148.2, 149.1,
159.2, 163.5 (d, J ) 250.9 Hz), 163.8, 168.4, 175.0. 19F NMR
(282 MHz, CDCl3): δ -112.1. ESIMS m/z 371 [M + H+], mp
186.8 °C (dec). Anal. Calcd for C18H19FN6O2: C, 58.37; H, 5.17;
N, 22.69. Found: C, 58.27; H, 4.90; N, 22.82.
Conclusion
In summary, this manuscript discloses the development of
a robust, efficient and scalable synthesis of the HMGR inhibitor
1c. Over the course of the development efforts numerous
process improvements were achieved: chromatography was
eliminated and replaced with selective and efficient crystalliza-
tion protocols; safe and scalable syntheses of the aminotriazole
and pyrimidine cores were developed and executed at scale.
The development of the Julia-Kocienski olefination reaction
led to the unprecedented dependence of product geometry on
reaction temperature, such that an E:Z ratio in excess of 200:1
was obtained. The incorporation of an ‘intermediate’ salt drop
was leveraged to ensure API quality prior to a heterogeneous
salt-to-salt transformation. Key considerations in the develop-
ment strategy and route selection included the careful evaluation
of thermochemical safety and impurity control via the selection
of process parameters, which was attained through the use of
mechanistic first principles and process understanding. The
result of these aspects culminated in the ability to prepare >70
kg of 1c in 35% overall yield.
Experimental Section
The experimental outlined below provide representative
procedures for what was run on scale in the kilo/pilot-plant
facilities. Reagents and solvents were acquired from a variety
of sources and employed without further purification. 1H NMR
spectra were obtained at 400 MHz, 13C NMR spectra were
obtained at 100 MHz and 19F NMR spectra at 376.5 MHz at
25 °C in CDCl3 except as indicated. Chemical shifts are reported
in ppm downfield from an internal tetramethylsilane standard
or relative to the residual solvent signal. Coupling constants
(J) are given in hertz. Melting points were measured with a
capillary melting point apparatus and are uncorrected. Moisture
determinations were performed with a Karl Fischer titrator.
HPLC analyses were performed using a reverse phase technique.
Gas chromatographic analyses were performed with a J&W
DB-1 column (30 m × 0.32 mm × 5 µm film thickness). A
general GC method was employed: 1.6 mL/min flow rate (He
carrier), FID, 1 µL injection volume, 250 °C injector port,
running a gradient of 5 min at 50 °C, then to 225 °C at a rate
of 20 °C/min, followed by a 10 min hold at 225 °C.
Methyl 4-(4-Fluorophenyl)-2-hydroxy-6-isopropylpyri-
midine-5-carboxylate (12). A 20 L reactor was purged with
nitrogen and charged with dihydropyrimidine 11 (700.0 g, 2.39
mol), CuCl2 (3.25 g, 0.024 mol), K2CO3 (33.1 g, 0.243 mol),
and CH2Cl2 (7.0 L). The starting material does not completely
dissolve under these conditions. The suspension was heated to
35 °C and treated with tert-butyl hydroperoxide (70% aqueous
solution) (635.0 g, 677.2 mL) over 120 min with vigorous
agitation. After 24 h, HPLC indicated the consumption of the
starting material (criterion:<1.0 area % of residual pyrimidine
(4-(4-Fluorophenyl)-6-isopropyl-2-(methyl(1-methyl-1H-
1,2,4-triazol-5-yl)amino)pyrimidin-5-yl)methanol (16). A
1200-L Hastelloy reactor was purged with nitrogen and charged
with dimethyl carbonate (104.7 kg), ester 14 (19.6 kg, 52.9 mol),
and 1,4-diazabicyclo[2,2,2]octane (6.34 kg, 56.5 mol). The
456
•
Vol. 14, No. 2, 2010 / Organic Process Research & Development

suspension was heated to reflux (88-90 °C), stirred for 3-4
h, and then sampled for HPLC analysis. If the amount of
residual starting material is >0.5 area %, the reaction mixture
should be heated for another 2 h. Upon reaction completion,
the reaction mixture was cooled to 20-25 °C, and water (20
L) and toluene (40 L) were charged. Any solids observed in
the reaction mixture typically go into solution upon addition of
the water. The phases were separated, and the upper organic
phase was washed sequentially with 25 wt % aqueous phos-
phoric acid (52.2 kg), 5 wt % aqueous sodium bicarbonate
solution (35 L), and saturated aqueous sodium chloride solution
(52 L). The organic solution was concentrated to minimum
volume via reduced pressure distillation, while maintaining the
internal temperature at <50 °C. The solvent exchange to toluene
was performed by repeated toluene addition (39.7 kg) and
distillation until the amount of residual dimethyl carbonate and
water were below <0.1 area % by GC and <500 ppm by KF,
respectively. The concentrated mixture was diluted with toluene
to a volume of ∼125 L and then cooled to -15 °C. To the
reaction mixture was added DIBAl-H (74.7 kg, 1.5 M solution
in toluene) over 2-3 h, maintaining the temperature below -5
°C. After the addition, the mixture was agitated for an additional
30 min at -15 to -5 °C and then sampled for HPLC analysis.
If compound 15 is g0.5 area % HPLC, then additional
DIBAl-H should be added (0.1 equiv portions). Upon comple-
tion the reaction was quenched by the addition of isopropanol
(7.9 kg, 1 equiv per equiv of DIBAl-H added) to the solution,
while maintaining the internal temperature <-5 °C. The addition
was exothermic, and hydrogen gas was formed. The reaction
mixture was heated to 20-25 °C and transferred into a solution
of hydrochloric acid (132 L, 1 N) over 2-4 h, while maintaining
the internal temperature between 15-20 °C. A precipitate
formed during the quench which became a thick slurry. Toluene
(113.0 kg) was charged, and the reaction mixture was heated
to ∼30 °C and held with high agitation for 1 h. The phases
were separated, and the organic phase was washed sequentially
with water (65 L), 5 wt % aqueous sodium bicarbonate (65 L),
and brine (77 L). All washes were performed at 30 °C. The
organic solution was concentrated to ∼50 L via reduced
pressure distillation, while maintaining the internal temperature
<50 °C. During the distillation the product crystallized, and the
suspension was stirred for 1 h at ∼40 °C prior to cooling to
20-25 °C. n-Heptane (33 L) was added over 1 h and the slurry
stirred for an additional 2 h before cooling to 0-5 °C. The
slurry was stirred at 0 °C for 2 h and then filtered. The wet
cake was washed with a mixture of n-heptane/toluene (2:1, 22
L), followed by a wash with n-heptane (177 L). The wet cake
was dried under vacuum at <50 °C until an LOD of <1.0%
was achieved. The product 16 was obtained in 78% yield, 14.71
kg, with HPLC purity of 99.6 area % and 99.1 wt %. 1H NMR
(400 MHz, CDCl3): δ 1.19 (d, J ) 6.32 Hz, 6H), 2.20 (br s,
1H), 3.40 (m, 1H), 3.55 (s, 3H), 3.64 (s, 3H), 4.56 (br s, 2H),
7.10 (dd, J ) 8.31, 8.79 Hz, 2H), 7.69 (dd, J ) 5.38, 8.78 Hz,
2H), 7.82 (s, 1H). 13C NMR (100 MHz, CDCl3): δ 22.7, 31.9,
36.0, 37.2, 57.8, 115.1 (d, J ) 21.6 Hz), 118.8, 131.0 (d, J )
8.5 Hz), 134.1, 149.2, 152.9, 159.3 163.1 (d, J ) 247.5 Hz),
166.1, 177.3. 19F NMR (282 MHz, CDCl3): δ -112.0. ESIMS
m/z 357 [M + H+], mp 129.6 °C dec. Anal. Calcd for
C18H21FN6O: C, 60.66; H, 5.93; N, 23.58. Found: C, 60.63; H,
5.88; N, 23.82.
4-(4-Fluorophenyl)-6-isopropyl-2-(methyl(1-methyl-1H-
1,2,4-triazol-5-yl)amino)pyrimidine-5-carbaldehyde (9). To
a 2000-L Hastelloy reactor, equipped with a 3-set pitch blade
impeller and previously inerted with N2, was charged 16 (29.63
kg, 83.12 mol, 98.75 wt %), TEMPO (1.30 kg, 8.32 mol), and
KBr (500 g, 4.20 mol) as solids. This was followed by the
addition of EtOAc (270.6 kg) and H2O (78.28 kg). The batch
was agitated at 80 rpm, and a recirculation loop with in-line
Raman and pH probes was started. The batch was cooled to
between 0 and -5 °C, and a titrated/nonbuffered sodium
hypochlorite (NaOCl) solution (64.03 kg, 10.96 wt %) was
added slowly over ∼2 h, at such a rate that the internal
temperature of the batch did not exceed 5 °C. Upon complete
addition, the batch temperature was maintained between -5
and 5 °C for 20 min before sampling the biphasic mixture for
reaction completion. Upon reaching the in-process control
criterion of <0.5 area % of 16 by HPLC, a 1.0 M sodium
thiosulfate solution (173.4 kg) was prepared and charged over
∼20 min in order to quench the reaction mixture. The batch
was warmed to 20-25 °C, stirred for ∼15 min, and analyzed
for residual oxidant using a potassium iodide-starch test. Upon
obtaining a negative test result, the agitation was ceased, and
the biphasic mixture was allowed to settle for 10 min. The
phases were separated, and the organic-rich layer was treated
with a 0.5 M sodium hydroxide solution (240.5 kg). After
agitating the biphasic mixture at 25 °C for 15 min, the layers
were separated, and the organic layer was checked by HPLC
analysis for the presence of acid impurity 17 (criterion <1.0
area %). The organic phase was washed with H2O (300.4 kg).
Following separation of the layers the organic layer was
transferred to a clean reactor followed by an ethyl acetate rinse
(30.2 kg). The organic layer was concentrated via distillation
to a target volume of 60 L (Tbatch ) 62 °C; vac ) 300-400
mmHg; Tjacket ) 83 °C). Isopropyl alcohol (141.3 kg) was
charged in portions (5-10 kg at a time in order to maintain a
level of 60-90 L in the tank, semibatch mode), and the
distillation/solvent swap was continued until a final target
volume of 60 L (Tbatch ) 62-65 °C; vac ) 350-400 mmHg;
Tjacket ) 83 °C) was reached. An additional charge of isopropyl
alcohol (23.8 kg) was added, and the sample was assayed for
residual EtOAc (criterion: <1 vol %). Water (70.0 kg) was
added, maintaining the batch temperature between 60-65 °C
in order to reach 30 vol % of water in IPA; a slurry was
observed prior to the completion of the H2O charge. The
resulting thick slurry was cooled to 10 °C over 2 h and held
overnight. The slurry was filtered in two loads on a centrifuge,
and each load was washed with 1:3 IPA/H2O (56.8 kg) and
H2O (41.1 kg). The resulting white crystalline solid was dried
at ∼25 °C, under full vacuum until the KF was <0.5%,
furnishing 27.64 kg of 9 in 93.5% yield with 99.88 area % and
99.63 wt %.
tert-Butyl 2-((4R,6S)-6-((E)-2-(4-(4-fluorophenyl)-6-iso-
propyl-2-(methyl(1-methyl-1H-1,2,4-triazol-5-yl)amino)py-
rimidin-5-yl)vinyl)-2,2-dimethyl-1,3-dioxan-4-yl)acetate (8b).
To a 600-L cryogenic reactor was charged compound 9 (27.5
kg, 77.6 mol), compound 5 (38.4 kg, 84.86 mol), and THF
Vol. 14, No. 2, 2010 / Organic Process Research & Development
•
457

(244.5 kg). The batch was stirred for 15 min to ensure
dissolution of the solids. The solution was cooled to -80 °C
and LiHMDS (63.7 kg, 1.2 equiv) was charged over 80 min
keeping the batch temperature below -70 °C throughout. After
a further 5 min HPLC analysis indicated 1.2 area % of
compound 9 remaining (criterion: 2 area %). The batch was
warmed to -10 °C, and after 1 h an E:Z product ratio of 91:1
was observed. The reaction was quenched with a 10% aqueous
potassium bicarbonate solution (137.5 kg), and the batch was
warmed to 15-25 °C. After agitating for 5 min and settling
for 15 min the aqueous phase was drained and a second portion
of 10% aqueous potassium bicarbonate (137.5 kg) was charged.
The mixture was agitated for 5 min and allowed to settle for
1 h to provide a clean phase split. Ethanol SDA3A (173.5 kg)
was added, and the solution was transferred to a 1900-L reactor
through a 10 µm polish filter. Additional ethanol SDA3A (174.6
kg) was introduced and the solution concentrated by vacuum
distillation (150 mmHg) from 766 to 270 L. Ethanol SDA3A
(195.3 kg) was charged at which point analysis by GC showed
4.4 vol % of THF remaining (criterion: 6 vol %). Water (330.0
kg) was added over 5 h, keeping the batch temperature at 38-42
°C, and the resulting slurry was held at this temperature for
2 h. The batch was cooled to 18-20 °C over 90 min and held
for 1 h, at which point analysis showed 0.19 wt % product in
the mother liquor. The slurry was filtered, and the cake was
washed with 60% ethanol SDA3A in water (300 kg). The wet
cake was dried without agitation under full vacuum for 32 h
(LOD ) 0.9%) followed by drying with periodic agitation for
24 h (final LOD ) 0.5%). The final product 8b was isolated as
a white solid, 33.58 kg with 99.92 area % and 99.2 wt % (74.2%
adjusted yield).
Ammonium (3R,5S,E)-7-(4-(4-Fluorophenyl)-6-isopropyl-
2-(methyl(1-methyl-1H-1,2,4-triazol-5-yl)amino)pyrimidin-
5-yl)-3,5-dihydroxyhept-6-enoate Monohydrate (1b). Com-
pound 8b (35.00 kg, 60.27 mol) was charged to a 1900-L
reactor, followed by acetonitrile (275 kg). To the resultant slurry
was added hydrochloric acid (0.02 N, 151.20 kg, 3.01 mol) at
20-25 °C, and an endotherm to 10 °C occurred. The slurry
was heated to 40-45 °C for 3 h until the area % of compound
8b remaining was <2 area %. The resultant solution was cooled
to 20-25 °C, and NaOH (2 N, 43.75 kg, 90.41 mol) was added.
The solution was stirred at 20-25 °C for 1 h until compound
19 had <1 area % remaining. Water (WPUR) (175.0 kg) was
charged to the reactor followed by heptanes (145.1 kg). The
mixture was agitated at 20-25 °C for ∼10 min. Agitation was
stopped, and the two phases were allowed to separate, ∼15 min.
The two phases were separated, and the product-rich aqueous
layer was recharged to the reactor. To this aqueous solution
was added hydrochloric acid (1 N) until the pH reached between
7 and 8, (amount of 1 N HCl required was 29.5 kg). Isopropyl
acetate (252.0 kg) was added, and the pH continued to be
adjusted with hydrochloric acid (1 N) until the pH was between
2.2 and 2.8, (amount of 1 N hydrochloric acid added was 59.9
kg). Agitation was continued for 10 min. The two phases were
allowed to separate. The organic layer was washed with a 10
wt % brine solution (192.5 kg). After agitating for 10 min the
two phases were allowed to separate. The phases were separated,
and the organic phase was transferred to a clean reactor through
a 0.5 µm polish filter. The temperature of the batch was set to
20-25 °C. Ammonia gas (0.12 kg, 6.63 mol) was added to
the solution over ∼30 min. The batch was seeded with a slurry
of compound 1b (35 g) in isopropyl acetate (274 g) and held
for 5 min at 20-25 °C. The remaining ammonia gas (1.01 kg,
59.67 mol) was added over a period of at least 2 h. A sample
of the slurry was taken to ensure that the pH was >8 by pH
paper, and the slurry was held at 20-25 °C for 15 h. A sample
was removed to ensure the liquor concentration was <0.5 wt
% of compound 1b. On meeting this criterion the batch was
filtered, and the cake was washed in portions with isopropyl
acetate (122.0 kg). The cake was dried at <40 °C under vacuum
until the wt % of isopropyl acetate was less than 0.2 wt %. 1b
was isolated as a white powder, 27.61 kg (89.5% yield, 99.85
wt %, 99.4 area %).
Hemi-calcium (3R,5S,E)-7-(4-(4-Fluorophenyl)-6-isopro-
pyl-2-(methyl(1-methyl-1H-1,2,4-triazol-5-yl)amino)pyrimi-
din-5-yl)-3,5-dihydroxyhept-6-enoate Dehydrate (1c). Com-
pound 1b (ammonium salt) (28.4 kg) was charged to an inerted
1900 L Hastelloy reactor. Water (284 kg, 10 L/kg input) was
added to produce an aqueous slurry of 1b which was held at
20 °C. In a separate Hastelloy vessel, anhydrous CaCl2 (44.6
kg, 11 equiv) was dissolved in water (284 kg 10 L/kg input).
The aqueous CaCl2 solution (13.6 wt %) was transferred into
the reactor containing the slurry of 1b using pressure via a
CUNO filter. Following the addition of the CaCl2 solution, the
slurry was held at 20 °C for ∼60 min. Heat was then applied
to the reactor jacket to heat the slurry to 50 °C over a 2 h period.
When the temperature reached ∼31 °C, the slurry was observed
to thicken, and the agitation was increased to maintain good
mixing. When the batch reached 50 °C, water (170.4 kg, 6 L/kg
input) was charged to the slurry, and the slurry was then aged
at 50 °C for 14 h. After aging, the batch was cooled to 20 °C.
The batch had a total volume of ∼800 L and was filtered in
three portions on a 36-in. Nutsche plate filter. Each load was
washed three times with water (79 kg) followed by a heptane
displacement wash (109 kg). The combined wet cake was then
dried in a 500 L conical dryer at 50 °C with full vacuum and
intermittent agitation until the KF was <10 %. The product 1c
was obtained as an off-white powder, 23.88 kg, in 79.9% yield
with 99.3 area % and 82.8 wt %.
Acknowledgment
We thank our Drug Discovery colleagues for useful discus-
sions. We also thank Drs. David Kronenthal, Robert Waltermire,
John Venit, Xinhua Qian, Ambarish Singh, San Kiang, and Mr.
Gerald Powers for helpful discussions, Victor Rosso for reaction
screens and DoE assistance, Merrill Davies for chromatography
screening, Francis Okuniewicz, Steven Chan, and Steve Wang
for safety testing. We also acknowledge the analytical support
provided by Shirley Wong, Peter Tattersall, Joan Ruan, Frank
Rinaldi, and Charles Pathirana as well as support from material
science during the development of this work, in particular
Shawn Yin, Jack Gougoutas, and Ray Scaringe.
Received for review January 14, 2010.
OP100010N
458
•
Vol. 14, No. 2, 2010 / Organic Process Research & Development