Kilogram synthesis of a LFA-1/ICAM inhibitor

Article abstract of DOI:10.1021/op9003168

The process development and the kilogram-scale synthesis of BMS-587101 (1) are described. The synthesis features a [3 + 2] azomethine ylide cycloaddition to efficiently build the spirocyclic core in a diastereoselective fashion followed by a classical res

Full text of DOI:10.1021/op9003168

Organic Process Research & Development 2010, 14, 553–561
Kilogram Synthesis of a LFA-1/ICAM Inhibitor
Albert J. DelMonte,* Robert E. Waltermire, Yu Fan, Douglas D. McLeod, Zhinong Gao, Kirsten D. Gesenberg,
Kevin P. Girard, Miguel Rosingana, Xuebao Wang, Jennifer Kuehne-Willmore, Alan D. Braem, and John A. Castoro
Process Research and DeVelopment, Bristol-Myers Squibb Company, One Squibb DriVe, P.O. Box 191,
New Brunswick, New Jersey 08903-0191, U.S.A.
Abstract:
Results and Discussion
The original synthesis of 1 by our Discovery colleagues
provided the multigram quantities required for preclinical
development studies (Scheme 1).4,5 This route featured a
stereospecific [3 + 2] cycloaddition of olefin 3 with N-
(methoxymethyl)-N-(trimethylsilylmethyl)benzylamine to afford
a single diastereomer of the spirocyclic core in a racemic
fashion, with the olefin geometry establishing the relative stereo-
chemistry of 4. Deprotection of the resulting N-benzyl pyrro-
lidine afforded the racemic secondary amine 5.6 The desired
single enantiomer was obtained by a chiral chromatographic
separation using supercritical fluid chromatography (SFC). The
final step in the synthesis was reductive amination with
5-formyl-3-thiophenecarboxylic acid.
In order to support the synthesis of kilogram quantities of
API for clinical development, a shorter and more efficient route
to 1 was desired. As shown in the retrosynthetic analysis (Figure
1) our strategy centered on retaining the powerful [3 + 2]
cycloaddition approach to assemble the spirocyclic core and,
at the same time, eliminating or minimizing the employment
of protecting groups and chromatography. Ideally, the unpro-
tected spirocyclic hydantoin 5 would be directly obtained by a
The process development and the kilogram-scale synthesis of BMS-
587101 (1) are described. The synthesis features a [3 + 2]
azomethine ylide cycloaddition to efficiently build the spirocyclic
core in a diastereoselective fashion followed by a classical resolution
which affords the desired enantiomer in >98% enantiomeric
excess. The target was prepared in four steps in an overall yield
of 22%.
Introduction
Interaction between leukocyte function-associated antigen-1
(LFA-1), expressed on the surface of cytokine-stimulated cells,
and intercellular adhesion molecule (I-CAM), found on the
surface of both leukocytes and endothelium, plays a key function
in the intercellular immune response, causing T-cell adhesion
and subsequent migration through the blood vessel wall to the
inflamed area.1 Small molecules which inhibit the LFA-1/I-
CAM interaction are targeted as potential drugs for the treatment
of a variety of autoimmune and inflammatory diseases such as
rheumatoid arthritis and psoriasis.2,3 The LFA-1 receptor antag-
onist, BMS-587101, 1,4,5 was selected for clinical development,
and we required a synthesis that would reliably generate
kilogram quantities of API. This paper details the identification
and development of a synthesis which enabled the realization
of this goal.
(4) The Discovery work towards this target compound BMS-587101 is
described in: Potin, D.; Launay, M.; Monatlik, F.; Malabre, P.;
Fabreguettes, M.; Fouquet, A.; Maillet, M.; Nicolai, E.; Dorgeret, L.;
Chevallier, F.; Besse, D.; Dufort, M.; Caussade, F.; Ahmad, S. Z.;
Stetsko, D. K.; Skala, S.; Davis, P. M.; Balimane, P.; Patel, K.; Yang,
Z.; Marathe, P.; Postelneck, J.; Townsend, R. M.; Goldfarb, V.; Sheriff,
S.; Einspahr, H.; Kish, K.; Malley, M. F.; DiMarco, J. D.; Gougoutas,
J. Z.; Kadiyala, P.; Cheney, D. L.; Tejwani, R. W.; Murphy, D. K.;
Mcintyre, K. W.; Yang, X.; Chao, S.; Leith, L.; Xiao, Z.; Mathur, A.;
Chen, B.-C.; Wu, D.-R.; Traeger, S. C.; McKinnon, M.; Barrish, J. C.;
Robl, J. A.; Iwanowicz, E. J.; Suchard, S. J.; Dhar, M. T. G. J. Med.
Chem. 2006, 49, 6946.
* Author for correspondence.
(1) For a discussion on the inhibition of LFA-1/ICAM-1as an approach
to treating autoimmune diseases see: Yusuf-Makagiansar, H.; Ander-
son, M. E.; Yakovleva, T. V.; Murray, J. S.; Siahaan, T. J. Medicinal
Research ReViews 2002, 22, 146.
(2) For a discussion of therapeutic options for treatment of psoriasis, see:
Gottlieb, A. B. J. Acad. Dermatol 2005, 53, S3. Larson, R. S.; Davis,
T.; Bologa, C.; Semenuk, G.; Vijayan, S.; Li, Y.; Oprea, T.; Chigaev,
A.; Buranda, T.; Wagner, C. R.; Sklar, L. A.
(5) For additional information related to this compound see: (a) Chen,
B.-C.; DelMonte, A. J.; Dhar, T. G. M.; Fan, Y.; Gougoutas, J. Z.;
Malley, M. F.; McLeod, D. D.; Waltermire, R.; Wei, C. Crystalline
Forms and Process for Preparing Spiro-Hydantoin Compounds.
(Bristol-Myers Squibb). U.S. Patent 7,381,737 B2 . (b) Dhar, T. G. M.;
Potin, D.; Maillet, M.; Launay, M.; Nicolai, E.; Iwanowicz, E. Spiro-
cyclic compounds useful as anti-inflammatory agents. Bristol-Myers
Squibb and Cerep). U.S. Patent 7,199,125 B2. (c) Launay, M.; Potin,
D.; Maillet, M.; Nicolai, E.; Dhar, T. G. M.; Iwanowicz, E. Hydantoin
compounds useful as anti-inflammatory agents. (Bristol-Myers Squibb).
U.S. Patent 6,710,064 B2. For the radiolabelled synthesis of BMS-
587101 see: Tran, S. B.; Maxwell, B. D.; Chen, S.-Y.; Bonacorsi,
S. J.; Leith, L.; Ogan, M.; Rinehart, J. K.; Balasubramanian, B.
J. Labelled Compd. Radiopharm. 2009, 52, 236.
(3) For other small molecule LFA-1/ICAM-1 antagonists as potential drugs
please see: (a) Pei, Z.; Xin, Z.; Liu, G.; Li, Y.; Reilly, E. B.; Lubbers,
N. L.; Huth, J. R.; Link, J. T.; von Geldern, T. W.; Cox, B. F.; Leitza,
S.; Gao, Y.; Marsh, K. C.; DeVries, P.; Okasinski, G. F. J. Med. Chem.
2001, 44, 2913. (b) Liu, G.; Huth, J. R.; Olejniczak, E. T.; Mendoza,
R.; DeVries, P.; Leitza, S.; Reilly, E. B.; Olasinski, G. F.; Fesik, S. W.;
von Geldern, T. W. J. Med. Chem. 2001, 44, 1202. (c) Wu, J.-P.;
Emeigh, J.; Gao, D. A.; Goldberg, D. R.; Kuzmich, D.; Miao, C.;
Potocki, I.; Qian, K. C.; Sorcek, R. J.; Jeanfavre, D. D.; Kishimoto,
K.; Mainolfi, E. A.; Nabozny, G.; Peng, C.; Reilly, P.; Rothlein, R.;
Sellati, R. H.; Woska, J. R.; Chen, S.; Gunn, J. A.; O’Brien, D.; Norris,
S. H.; Kelly, T. A. J. Med. Chem. 2004, 47, 5356. (d) Last-Barney,
K.; Davidson, W.; Cardozo, M.; Frye, L. L.; Grygon, C. a.; Hopkins,
J. L.; Jeanfavre, D. D.; Pav, S.; Qian, C.; Stevenson, J. M.; Tong, L.;
Zindell, R.; Kelly, T. A. J. Am. Chem. Soc. 2001, 123, 5643. (e) Wang,
G. T.; Wang, S.; Gentles, R.; Sowin, T.; Leitza, S.; Reilly, E. B.; von
Geldern, T. W. Bioorg. Med. Chem. Lett. 2005, 15, 195. (f) Wattanasin,
S.; Albert, R.; Ehrhardt, C.; Roche, D.; Savio, M.; Hommel, U.;
Welzenbach, K.; Weitz-Schmidt, G. Bioorg. Med. Chem. Lett. 2003,
12, 499.
(6) While effective at removing the benzyl group, avoiding the use of
1-chloroethyl chloroformate in dichloroethane was highly desirable
since use of these reagents would raise concerns about genotoxic
impurity formation.
(7) Throughout this paper when the molecule is a racemic mixture of a
single diastereomer, solid and dashed rectangular bonds are used to
convey relative stereochemistry. When the compound is a single
enantiomer, solid and dashed wedged bonds are used to convey
absolute stereochemistry.
10.1021/op9003168  2010 American Chemical Society
Published on Web 03/10/2010
Vol. 14, No. 3, 2010 / Organic Process Research & Development
•
553

Scheme 1. Discovery synthesis of 1a
a Reagents and conditions: (a) Sarcosine ethyl ester, NaOH, THF/H2O, 96%; (b) 4-cyanobenzaldehyde, ꢀ-alanine, AcOH, reflux, 35% or 4-cyanobenzaldehyde,
pyrrolidine/EtOH, 78 °C 18 h, 85%; (c) N-(methoxymethyl)-N(trimethylsilylmethyl)benzylamine, THF, TFA, 0 °C to RT, 18 h, MeOH, reflux, 2 h, 50-85%;7 (d)
1,2-dichloroethane, 1-chloroethyl chloroformate, 5 °C, to rt, 18 h, MeOH, reflux, 3 h, 50-85%; (e) Chiralpak-AD column, CO2/MeOH, ∼100% desired enantiomer;
(f) 5-formyl-3-thiophenecarboxylic acid, Na(OAc)3BH, Na2SO4, 1,2-dichloroethane, 75-90%.
Scheme 2. Synthesis of olefin 3
choice of base and solvent had a strong influence on the E:Z
preference and had selected pyrrolidine as the base and ethanol
as the solvent.4 We conducted a solvent screen and the results
Figure 1. Retrosynthetic analysis of 1.
confirmed ethanol was the optimal solvent as the E-stereoisomer
was extremely insoluble producing isolated 3 in high recovery
[3 + 2] cycloaddition of olefin 3 with a simple azomethine
and stereoselectivity (>100:1 E:Z).10 The condensation of
ylide derived from glycine.8 The identification and development
4-cyanobenzaldehyde with the commercially available hydan-
toin 2 (Scheme 2) was initially conducted at 13 mL/g11 in
of an efficient chemical resolution of 5 using a chiral acid was
critical to the success of this approach. We also felt it would
ethanol; however, a very thick paste typically formed and
be advantageous to introduce the thiophene intermediate at the
adequate mixing was not consistently achieved. By diluting to
penultimate step of the synthesis as this would provide an
17 mL/g and controlling the onset of crystallization with a
timely addition of seeds,12 mixing was improved dramatically.
This procedure was demonstrated multiple times at kilogram
scale and gave excellent yield, reproducibility and purity (84.6
kg of 3, 83.9% average yield and >99.9 HPLC area % purity).
Azomethine Ylide Cycloaddition To Afford Spirocyclic
Hydantoin 5. As mentioned above, we felt it important to
develop a [3 + 2] cycloaddition process that directly generated
spirocycle 5, without the need for a deprotection step as
performed in the original synthesis. In this vein, examples of
direct [3 + 2] azomethine ylide cycloadditions utilizing glycine
additional opportunity to remove structurally related thiophene-
derived impurities.9 Having a simple saponification as the final
step was also an attractive feature of this synthetic approach
from an impurity control perspective.
Preparation of Olefin (3). Generation of 3 with high
stereoisomeric purity was an important objective during the
development of this reaction as only the E-stereoisomer affords
the desired diastereomer in the subsequent cycloaddition. Our
Discovery colleagues had previously demonstrated that the
(8) An asymmetric cycloaddition would have presented a preferred
approach, and we carefully considered this option. Unfortunately, in
reviewing the known systems, we found none that were easily
adaptable to our system as they would all form a pyrrolidine product
with a substitution pattern that would require substantial functional
group transformations to afford our target molecule. For a review of
asymmetric azomethine ylide [3 + 2] cycloadditions, please see:
Pandey, G.; Banerjee, P.; Gadre, S. R. Chem. ReV. 2006, 106, 4484.
Tsuge, O.; Kanemasa, S. AdV. Heterocycl. Chem. 1989, 45, 231.
(9) There were no specific thiophene derived impurities of concern.
Incorporation of the thiophene as an ester was implemented proactively
as a conservative approach to minimize the potential for an impurity
to be introduced into the API step.
(10) Solvents and bases screened included IPAC, EtOAc, MTBE, 1-BuOH,
IPA, THF, t-BuOH, MeOH, EtOH, toluene, DMSO, NMP, acetone
and Et3N, 2,6-lutidine. Of these, EtOH was deemed superior as the
alternative conditions suffered from either poor reaction conversion,
high levels of impurities, or stirring issues.
(11) mL/g refers to mL of solvent/g of substrate.
(12) The optimal seeding point was determined via lab experiments. Seeding
prior to the reaction was unsuccessful as the initial mixture was not
yet saturated with product and the seeds simply dissolved.
554
•
Vol. 14, No. 3, 2010 / Organic Process Research & Development

Scheme 3. [3 + 2] cycloaddition under modified Tsuge conditions
Scheme 4. Synthesis of resolved spirocyclic hydantoin 5b
and paraformaldehyde are known.13 Our initial investigations
of this cycloaddition with a model substrate, dimethyl maleate,
were conducted in toluene following the experimental proce-
dures of Tsuge.9 While the Tsuge conditions afforded the desired
cis-disubstituted pyrrolidine product 8, it quickly became
apparent that this chemistry was very challenging. Obtaining
consistent conversions was problematic, and we attributed this
to the heterogeneous nature of the reaction in toluene or xylenes.
Upon further evaluation of the reported Tsuge reaction condi-
tions, we found it interesting that almost all examples were
performed in toluene with the sole exception being a reaction
that was conducted in DMF. We were intrigued as to why the
heterogeneous toluene conditions were chosen over the homo-
geneous DMF conditions for the majority of the substrates. We
repeated the cycloaddition in a mixture of 2:1 N-methylpyr-
rolidinone (NMP)/toluene which, similarly to DMF, afforded
a nearly homogeneous reaction. While we were pleased to
discover the NMP/toluene conditions did provide consistent
reaction conversions, this modified system afforded primarily
the undesired trans-pyrrolidine 9 (Scheme 3). It was apparent
that lack of stereospecificity could make this homogeneous
system impractical for systems with olefin substrates susceptible
to isomerization or pyrrolidine products susceptible to epimer-
ization. These potential liabilities would have to be considered
when investigating our substrate. It should also be noted that,
regardless of the solvent system utilized, the initial products
afforded were N-methylene-bridged dimers 7 which required
cleavage.14 In order to apply this direct [3 + 2] azomethine
ylide cycloaddition chemistry to the substrate of interest, control
of olefin isomerization and product epimerization as well as an
efficient cleavage of the N-methylene-bridged dimer would be
required.
Initial azomethine ylide cycloadditions of 3 in toluene were
heterogeneous and, as predicted from our experience with the
dimethyl maleate substrate, suffered from inconsistent reaction
conversions and thus poor yields. However, cycloadditions of
3 which were conducted in 2:1 mixtures of NMP/toluene with
paraformaldehyde did provide consistent reaction conversion.
Isomerisation of the olefin 3 or epimerization of the [3 + 2]
cycloaddition adduct would have negated this synthetic ap-
proach; thus, we were very pleased to observe that after acidic
hydrolysis of the initial N-methylene-bridged dimer 1015
(Scheme 4) we found almost exclusively the desired diastere-
omer.16 We evaluated other formaldehyde sources and replaced
paraformaldehyde with the more soluble hexamethylenetetra-
mine (HMTA).17 The reaction could be performed at temper-
atures as low as 100 °C, but much cleaner purity profiles and
faster rates were obtained at higher temperatures (140-145 °C).
It was also imperative to cool the reaction mixture once full
conversion to product was attained since prolonged heating
dramatically increased impurity levels.18 Impurities that were
identified (Figure 2) include 3,5-dichloroaniline, the mixed ureas
11, and minor amounts of the diastereomer 13 which results
(13) (a) Mortier, J.; Joucla, M. J. Chem. Soc., Chem. Commun. 1985, 22,
1566. (b) Borsato, G.; Della Negra, F.; Gasparrini, F.; Misiti, D.;
Lucchini, V.; Possamai, G.; Villani, C.; Zambon, A. J. Org. Chem.
2004, 69, 5785. (c) Joucla, M.; Mortier, J. Bull. Soc. Chim. Fr. 1988,
3, 579. (d) Tsuge, O.; Kanemasa, S.; Ohe, M.; Takenaka, S. Chem.
Soc. Jpn. 1987, 60, 4079. (e) Tsuge, O.; Kanemasa, S.; Ohe, M.;
Takenaka, S. Chem. Lett. 1986, 6, 973. (f) Tsuge, O.; Kanemasa, S.;
Ohe, M.; Yorozu, S.; Takenaka, S.; Ueno, K. Chem. Lett. 1987, 60,
4067.
(14) The currently accepted mechanism suggests the azomethine ylide
precursor is the methylene-bridged dimer 3,3′-methylenedioxazolidin-
5-one.
Vol. 14, No. 3, 2010 / Organic Process Research & Development
•
555

Figure 2. Impurities formed in the [3 + 2] cycloaddition reaction.
from epimerization of the desired compound 5 during prolonged
aging under the basic reaction conditions.19
spirocyclic hydantoin 5 and the corresponding N-methylated
spirocyclic hydantoin, both detectable by reverse phase HPLC.
React-IR was ultimately demonstrated to be an effective
approach for monitoring the presence and consumption of the
N-methylene bridged dimer.22
The procedure initially developed and implemented on
multikilogram scale employed an acidic workup to break the
N-methylene-bridged dimer which enabled isolation of 5 as a
racemic HCl salt 5a. Although effective, one unappealing
feature of this approach, which we needed to address for
subsequent campaigns, was the slow rates of filtration and
washing of the resulting HCl salt (12 h on 42 kg scale). The
successful implementation of this process was demonstrated on
three batches at kilogram scale and afforded good yields and
purities (41.8 kg of 3, 77.1% average yield, and 98.6-99.4
HPLC area % purity).
Interestingly, during the course of the reaction a water-
soluble, white solid slowly collected in the condenser. This was
a concern as a potential safety liability should the condenser
become completely occluded.20 Analysis indicated this solid was
not paraformaldehyde as we first postulated but rather am-
monium carbamate, resulting from condensation between the
liberated CO2 and ammonia. We rationalized that incorporation
of a nitrogen sweep would aid in dilution of these gases, thus
minimizing accumulation of the solid. We were pleased to
observe this modification not only substantially decreased the
amount of ammonium carbamate collected but also favorably
impacted the reproducibility of the reaction rate. Early in our
development efforts, reaction times were highly variable and
ranged from 7 to 20 h. However, after incorporation of a
vigorous nitrogen sweep, the reaction was consistently complete
in 4-6 h.21 As expected, the initial product from the [3 + 2]
cycloaddition is the N-methylene-bridged dimer 10. Because
the [3 + 2] cycloaddition is achiral and both enantiomers of
the spirocycle are produced, 10 can, in principle, be generated
as either a single enantiomer (DL, not shown) or meso-
compound.15 Monitoring levels of this intermediate was chal-
lenging as 10 was not directly detectable by reverse phase HPLC
due to hydrolytic instability and was observed as a broad peak
under normal phase HPLC conditions. However, we found that
levels of 10 could be monitored indirectly by treating a sample
of the crude reaction mixture with sodium triacetoxyborohydride
(STAB) which converts 10 to a mixture of the desired,
Diastereomeric Salt Resolution of 5. A key objective in
the design of the synthetic process was to identify a diastere-
omeric salt resolution to avoid the need for large scale
(17) During the development of the [3 + 2] cyclopropanation, both the
HMTA and glycine equivalents were optimized with the highest in-
process yield resulting from 2.0 equiv of glycine and 0.7 equiv of
HMTA. However, in an attempt to minimize build-up of ammonium
carbonate in the condenser and epimerization of product (see ref 16)
as well as increase the ease in developing a telescoped reaction,
ultimately 1.9 equiv of glycine and 0.33 equiv of HMTA were utilized.
Our initial attempts at scaling up paraformaldehyde conditions
indicated that a portionwise addition was required on ∼10-100 g scale
to achieve a reasonable yield. The same protocol when implemented
on a 1-5 kg scale afforded substantially decreased yields. The yield
significantly improved when HMTA was utilized and there was no
requirement to add in small portions to achieve reproducibility on small
or large scale.
(15) Although the [3 + 2] cycloaddition product is primarily a single
diastereomer, the initial product is a mixture of D,L- and meso-
methylene-bridged dimer as shown below. Each was isolated as a
single crystal solvate form, and X-ray structures were determined. Full
crystallographic data sets have been deposited to the Cambridge
Crystallographic Data Center (CCDC reference numbers 756962 and
756963). Copies of the data can be obtained free of charge via the
internet at http://www.ccdc.cam.ac.uk.
(18) When the isolated spirocyclic hydantoin 5a was resubjected to the
reaction conditions for 8h at 140 °C, only 62% of 5a was recovered.
Diastereomer 13 was present in 12%, 3,5-dichloroaniline in 4% and
mixed urea 11 in 2%. Under the same conditions without HMTA
present, 92% of 5a was recovered.
(19) All impurities were identified by either NMR or MS and were either
isolated from the reaction mixture or synthesized independently.
(20) Prior to scale-up, a full safety evaluation was performed including
reaction calorimetry, measurement of off-gassing and dust-explosivity
testing. Although CO2 and ammonia are released during the reaction
conditions, no foaming or frothing was observed. However, a control
experiment in which a large excess of ammonium carbamate (2 g)
was initially added to the reaction mixture (10 g of starting material
3) did result in vigorous off-gassing as the reaction temperature reached
90-100 °C during the heat-up to 130 °C, further emphasizing the
need to avoid the accumulation of this solid in the condenser. The
excess ammonium carbamate had no effect on the reaction conversion.
(21) We postulate that incorporation of a nitrogen sweep reduces the
concentration of ammonia in the reaction mixture. This has a positive
impact on the reaction rate because high levels of ammonia lead to
reformation of HMTA as opposed to the desired formation of the
azomethine ylide precursor (3,3′-methylenedioxazolidin-5-one). While
one could easily envision a loss of solvent under refluxing conditions
with a nitrogen sweep, we found that a properly functioning condenser
kept losses to a minimum.
.
(16) During the course of the development work a full set of control
experiments were run prior to scale-up. No isomerization of the starting
material 3 was observed under the reaction conditions. The product
5a, however, was found to be susceptable to base-catalyzed epimer-
ization and equilibrium between 5a and 13 could be readily achieved
by treatment with DBU. Additional evidence of base-catalyzed
epimerization of the benzylic center was obtained during stress tests
of the hydrolysis reaction in which diastereomer was formed at pH
∼14 in THF/alcohol mixtures.
(22) For React-IR data please see the Supporting Information.
556
•
Vol. 14, No. 3, 2010 / Organic Process Research & Development

Scheme 5. Telescoped cyclization/resolution for the direct preparation of 5b
chromatographic separation of the enantiomers. Our initial salt
screens were conducted utilizing free base generated in situ by
addition of diisopropylethylamine (DIPEA) to 5a. Our first salt
screen identified (S)-(+)-camphorsulfonic acid (CSA) as a
promising candidate for resolution in methylethyl ketone
(MEK). In initial laboratory experiments, effective resolution
was observed with ee >95%. However, when slurries of the
CSA salt were held for extended periods, the enantiomeric
excess (ee) eroded and it became obvious that the CSA salt
was not suitable for kilogram-scale synthesis. Fortunately,
studies with a second lead from our screening efforts, (+)-di-
p-toluyl-D-tartaric acid ((+)-DTTA) in CH2Cl2 were more
successful. We found that resolution of 5 as its hemi-(+)-DTTA
salt 5b also afforded enantiomeric excess of >95%. Fortunately,
5b was found to be thermodynamically stable and not subject
to the same deterioration of ee as observed with the CSA salt
upon aging of the crystallization slurry. The successful imple-
mentation of this process was demonstrated in two batches on
kilogram scale and gave excellent yields and purities (48.1 kg
of 5b, 30.6% average yield, 98.1-99.2% ee and 98.5-98.6
HPLC area % purity). The overall yield from 3 was 23.6%.
Direct Preparation of Chiral Spirocyclic Hydantoin 5b
from 3. Although the two-step process for producing 5b was
effective, the slow filtration of 5a and the low overall yield of
5b (23.6%) suggested a telescoped procedure, which avoided
isolation of the HCl salt, could increase the overall efficiency
of this process. A crucial component of a telescoped process is
the efficient cleavage of the N-methylene-bridged dimer 10. We
were pleased to observe 10 was readily cleaved in the crude
reaction mixture by treatment with ethylene diamine at 50 °C
for 2 h. A one-pot process was demonstrated by subjecting the
resulting solution containing monomeric racemic pyrrolidine
to an extractive workup, solvent swap into methylene chloride,
and treatment with (+)-DTTA. The earliest attempts to develop
this process proved challenging since the ee’s and yields were
highly variable. However, a critical observation was made
during the laboratory development studies that imparted con-
sistency and robustness to the resolution process. Reactions
which were azeotropically distilled with MeOH and contained
residual methanol in the crystallization matrix consistently
yielded material of very high purity and enantiopurity.23 It was
subsequently determined that MeOH (0.5-1 mL/g) was re-
quired to achieve a robust resolution that reliably produced high
quality (purity and ee) 5b with consistent yield. On scale, the
MeOH content was adjusted to the desired stoichiometry
immediately prior to the crystallization. In addition, the
introduction of one equivalent of water was also shown to be
advantageous as it also facilitated the formation of crystals with
high enantiopurity 5b (>98% ee).24 Using these optimized
conditions, the process was extremely tolerant of excess (+)-
DTTA so stoichiometry adjustments were not required based
on in-process yield of 5. The successful implementation of this
process (Scheme 5) was demonstrated on kilogram scale and
gave excellent yield and purity (33.2 kg of 5b, 32.2% yield,
98.7% ee and 99.9 HPLC area % purity). Interestingly, in the
original process when the HCl salt 5a was isolated, thus
affording higher purity input material for the resolution, the
addition of MeOH and water were not required.
Preparation of the Penultimate Intermediate 6. Reductive
amination (Scheme 6) was utilized to effect the convergent
coupling of 5b and the methyl ester-protected thiophenecar-
boxaldehyde 14. To obtain a homogeneous solution, a suspen-
sion of 14 and 5b in THF was treated with triethylamine and
acetic acid. The resulting solution was then treated with 4 equal
portions of STAB over 2 h.25 After addition of MTBE, three
extractions, one back extraction, and a subsequent solvent swap
to isopropanol, the desired product was isolated as the HCl salt,
6. It is interesting to note that no enhancement of the enantio-
(23) During the development work, a MeOH azeotropic distillation was
investigated to aid in removal of toluene. Ultimately an alternative
procedure which did not involve an MeOH distillation was developed
and optimized for scale-up.
(24) The expected water content for a monohydrate is 2.87%. While
combustion analysis of the initial sample characterized from the non-
telescoped process indicates a hemi-hydrate, when a variety of
developmental samples were analyzed, the KF ranged from 1.1 to
5.5%. This may indicate that while a hemi-hydrate is achievable, it is
possible to easily over-dry or under-dry the isolated solid. It may also
suggest that alternative forms are available or that the isolated product
is hygroscopic.
(25) It should be noted that no significant exotherm was observed; however,
low levels of hydrogen off-gassing were evident during the STAB
additions.
Vol. 14, No. 3, 2010 / Organic Process Research & Development
•
557

Scheme 6. Reductive amination of 5b
meric purity was observed during the crystallization. Further
process development successfully reduced the workup protocol
to only one sodium bicarbonate wash with no sacrifice to yield
or purity. This process was demonstrated on eight batches at
kilogram scale and gave excellent yields and purities (22.9 kg
of 5b, 89.8% average yield, 97.8-98.9% ee and 99.1-99.5
HPLC area % purity).
H2O mixture and subsequent azeotropic distillation resulted in
crystallization of 1 (Scheme 7).27 This approach afforded two
opportunities to provide purity enhancement with the final
crystallization used to control the crystalline form. In order to
achieve the desired particle size (d90 < 66 µm) wet milling was
performed post-distillation. This process was successfully
performed on three batches at kilogram scale and provided 1a
of excellent quality (25.0 kg of 6, 89.6% average yield and
98.7-99.2 HPLC area % purity, and 90.5-92.9 area weight
% purity). The three THF solvate 1a batches were converted
to 1 in good yield and excellent quality (80.9% average yield,
d90 < 60 µm, 99.7-99.9 HPLC area % purity, and 99.5 area
weight % purity). The overall yield of 1 from 6 was 72.5%.
One-Step Preparation of BMS-587101 (1). The obvious
potential efficiency gains of a one-drop process dictated
exploration of alternative crystallization protocols. A key
procedural step which enabled the one-step process was a
heptane extraction immediately after the hydrolysis of 6 was
complete. This heptane extraction proved to be effective in
removing ester impurities such as 6, 17, and 18. Substitution
of acetic acid with citric acid, for acidification after the
saponification, was also implemented to address odor concerns.
A solvent swap into isopropyl acetate after the hydrolysis in
THF allowed for direct crystallization of 1. In addition to an
improved yield over the two-drop process (90% versus 72%),
the cycle time was significantly reduced (68 h versus 135 h).
To further reduce cycle time, simultaneous wet milling during
the distillative crystallization was demonstrated in the final batch
with no impact to purity or particle size. This process was
demonstrated on kilogram scale (six batches) and gave excellent
yield and purity (232.3 kg of 1, 90% average yield, d90 < 60
µm, and >99.3 HPLC area % purity).
Two-Step Preparation of BMS-587101 (1). Our initial
synthetic strategy intentionally incorporated a simple hydrolysis
as the final step to facilitate impurity control. However,
deprotection of 6 was more challenging than initially anticipated
since impurities resulting from hydration of the nitrile (15) or
hydrolysis of the hydantoin core (16) were readily formed
(Figure 3). Screening a variety of reagents, solvents, and reaction
conditions indicated that a low-temperature (10 °C) hydrolysis
in THF with KOH effectively controlled the level of impurity
formation; however, the rate of reaction was extremely slow.
We were pleased to observe that the addition of 2 equiv of
1,2-propanediol accelerated the hydrolysis reaction. We pos-
tulate this rate acceleration is via transesterification with 1,2-
propanediol to initially form 17 and 18 which provide intramo-
lecular anchimeric assistance in the ester hydrolysis.26 A concern
with using this approach was the presence of residual 17 and
18. Fortunately, these impurities were observed in very low
levels and were easily removed via the crystallization process
as was the methyl ester starting material.
The isolation of the active pharmaceutical ingredient (1) was
complicated by a propensity to form solvates with a broad
diversity of solvents. This property was strategically employed
to better control purity levels in our initial process. Specifically,
we developed a two-drop process, in which THF solvate 1a
was first isolated. Dissolution of 1a in an 95:5 isopropyl acetate/
Conclusions
In summary, we have described an efficient, four-step
synthesis of BMS-587101 (1) from commercially available
hydantoin 2 as shown in Scheme 8. All key intermediates were
crystalline, and no exotic processing equipment was required,
providing versatility in future manufacturing site selection. The
(26) For literature precedent see: (a) Balakrishnan, M.; Rao, G. V.;
Venkatasubramanian, N. Tetrahedron Lett. 1972, 45, 4617. (b) In side-
by-side reactions at 7 °C, in 21 h the conversions for the reactions
with the 1,2-propanediol and without any additives were 84.6% and
57.2%, respectively.
(27) After a thorough screen, isopropyl acetate was identified as one of
the few solvents which did not form solvates with 1. However, in
order to develop a practical procedure, the low solubility of 1a in
isopropyl acetate would need to be addressed. It was determined that
low levels of water significantly increased 1a solubility in isopropyl
acetate. Solubility data: Compounds 1 and 1a in isopropyl acetate with
low water content (KF ) 0.023%) have a solubility of 2.2 wt %.
Compounds 1 and 1a in isopropyl acetate with 1.2 wt % water have
a solubility of 3.6 wt %.
Figure 3. Impurities formed in the API step.
558
•
Vol. 14, No. 3, 2010 / Organic Process Research & Development

Scheme 7. Two-drop and one-drop preparation of 1 from 6
Scheme 8. Efficient four-step synthesis of 1 (BMS-587101)
use of glycine and HMTA in the [3 + 2] azomethine ylide
cycloaddition provides quick and direct access to spirocycle 5.
In addition, an efficient telescoped resolution with (+)-DTTA
was developed for the direct conversion of 3 to 5b which
circumvented the isolation the racemic HCl salt 5a, increased
yields, and reduced processing time. The overall process has
been demonstrated on multikilogram scale and is amenable for
commercial-scale manufacture.
Preparation of (E)-4-((1-(3,5-Dichlorophenyl)-3-methyl-
2,5-dioxoimidazolidin-4-ylidene)methyl)benzonitrile (3). A
mixture of 2 (69.0 kg, 266.3 mol) and 4-cyanobenzaldehyde
(52.8 kg, 402.5 mol) was heated in EtOH (980 kg) to 60 °C.
Pyrrolidine (19 kg, 267.2 mol) and an ethanol line rinse (4.0
kg) were charged. The mixture was held at 65 °C, and after
0.5 h seeds (259 g) were charged. The mixture was then heated
at reflux for 16 h. The slurry was cooled to 55 °C and diluted
with THF (305 kg) to afford a more readily stirrable solution.
The mixture was then cooled to 45 °C. Ethanol was charged
(50 kg), and then the slurry was cooled over 2-3 h to 20 °C
and filtered. The cake was washed with ethanol (220 kg), and
upon drying, 3 was obtained as a fluffy, yellowish crystalline
solid (84.6 kg of 3, 85.4% yield, and >99.9 HPLC area %
impurity). Mp ) 239-241 °C. 1H NMR (DMSO-,d6): 8.07 (2H,
d, J ) 8.3 Hz), 7.86 (2H, d, J ) 8.4 Hz), 7.72 (1H, m), 7.59
(2H, m), 6.72 (1H, s), 3.35 (3H, s). 13C NMR (DMSO-,d6):
(159.80, 151.48, 137.64, 133.83, 133.70, 131.80, 130.80, 130.68,
127.71, 125.51, 118.83, 114.48, 110.32, 26.72). Anal. Calcd
for C18H11Cl2N3O2: C, 58.08; H, 2.97; N, 11.29; Cl, 19.05.
Found: C, 58.14; H, 2.72; N, 11.14; Cl, 19.15.
Experimental Section
All reactions were performed under a nitrogen atmosphere.
All reagents purchased from vendors were used as received
unless otherwise indicated. Chiral analysis was conducted on a
Shimadzu HPLC with a Chiracel AD or AD-H column unless
stated otherwise. Proton and Carbon NMR were run on a Bruker
AC-300 at 300 MHz for proton and 75 MHz for carbon or on
a Bruker AVANCE 400 at 400 MHz for proton and 100 MHz
for carbon. The melting points were obtained with a Mettler-
Toledo FP 62 melting point instrument by measurement of the
change of luminous intensity during the melting process.
Reported yields have not been corrected for impurity levels or
moisture content.
Vol. 14, No. 3, 2010 / Organic Process Research & Development
•
559

Preparation of 4-[(5S*,9R*)-3-(3,5-Dichlorophenyl)-1-
methyl-2,4-dioxo-1,3,7-triazaspiro[4.4]non-9-yl]-benzoni-
trile hydrochloride salt (5a). A mixture of 3 (41.8 kg, 112.3
mol), glycine (21.5 kg, 286.4 mol) and hexamethylenetetramine
(12.0 kg, 85.6 mol) in 1-methyl-2-pyrrolidinone (222 kg) and
toluene (92.0 kg) was heated at 140 °C for 10 h. Upon reaction
completion (<1% 3 remaining), the mixture was cooled to 20
°C and polish filtered. The filtrate was heated to 70 °C and 1
M HCl (580 kg) was charged (to extract the basic product into
the aqueous layer). The resulting biphasic mixture was stirred
for 10-15 min. Agitation was halted and the phases were
separated. The organic phase was washed with 1 M HCl (72
kg) at 65 °C. The aqueous phases were combined and were
stirred at 80 °C for 2 h. The solution was cooled to 50 °C and
seeded with 5a (200 g) followed by slow cooling to 15 °C with
gentle stirring. The resulting suspension was held at 15 °C for
4 h with no agitation and then filtered. The filter cake was
washed with 1 M HCl (2 × 110 kg) and reslurried in 250 kg
water at 100 °C for 0.5 h. The mixture was cooled to 85 °C
and seeded with 5a (200 g). The mixture was cooled to 5 °C
over 16 h with low agitation and the slurry was filtered. The
slurry was washed with water (2 × 17.5 kg) and dried under
vacuum at 35 °C. Spirocyclic hydantoin 5a was obtained as a
beige crystalline solid (41.5 kg of 5a, 82.6% yield, and 98.6
HPLC area % purity). Mp ) 183-185 °C. 1H NMR (DMSO-
d6): 7.87(2H, d, J ) 8.1 Hz), 7.61 (1H, m), 7.40 (2H, d, J )
8.1 Hz), 6.68 (2H, m), 4.17 (1H, m), 3.85 (2H, m), 3.76 (2H,
m), 3.43 (3H, s), 3.24(2H, s). 13C NMR (DMSO-d6): (170.84,
152.92, 137.35, 133.94, 132.87, 132.35, 128.01, 124.50, 118.12,
111.30, 71.42, 46.57, 45.11, 25.51). Anal. Calcd for
C20H17Cl3N4O2 + 1.3 H2O: C, 50.51; H, 3.91; N, 11.79; Cl,
22.39. Found: C, 50.56; H, 3.86; N, 11.58; Cl, 21.98; KF, 5.12.
Preparation of 4-[(5S,9R)-3-(3,5-Dichlorophenyl)-1-meth-
yl-2,4-dioxo-1,3,7-triazaspiro[4.4]non-9-yl]-benzonitrile Semi
(+)-DTTA Salt (5b). To a suspension of 5a (50.0 kg, 120.4
mol) in dichloromethane (650 kg) was added diisopropylethy-
lamine (17.4 kg, 134.6 mol). The mixture was stirred until a
clear solution formed, to which (+)-Di-p-toluoyl-D-tartaric acid
(10.5 kg, 27.2 mol) was added. The resulting solution was
heated to 35 °C and seeded with 5b (200 g). The mixture was
held for 30 min and cooled to 20 °C over 2 h. MTBE (175 kg)
was added over 0.5 h. The suspension was stirred at 20 °C for
5 h and filtered. The filter cake was washed with dichlo-
romethane (27 kg)/MTBE (8 kg), followed by an MTBE wash
(22 kg). The solid was dried at 40 °C. The resolved spirocyclic
hydantoin 5b was obtained as a white crystalline solid (29.7
kg of 5b, 33.1% yield, 98.6 HPLC area % purity, and 98.1%
ee). Mp ) 175-177 °C. 1H NMR (DMSO-d6): 7.86 (2H, d, J
) 8.1 Hz), 7.81 (2H, d, J ) 8.3 Hz), 7.61 (1H, m), 7.28 (2H,
d, J ) 8.1 Hz), 7.22 (2H, 8.5 Hz), 6.68 (2H, m), 5.71 (1H, s),
3.81(1H, m), 3.50 (4H, m), 3.06 (3H, s), 2.34 (3H, s). 13C NMR
(DMSO-d6): (171.45, 169.40, 165.04, 152.88, 143.61, 138.99,
133.88, 133.08, 132.16, 129.26, 129.20, 128.76, 127.84, 126.99,
124.51, 118.25, 110.78, 72.81, 73.38, 48.15, 47.51, 46.30, 24.90,
21.14). Anal. Calcd for C30H25Cl2N4O6 + 0.5 H2O: C, 58.40;
H, 4.17; N, 9.08; Cl, 11.49. Found: C, 58.58; H, 4.06; N, 8.94;
Cl, 11.38; KF, 1.59.
Preparation of 4-[(5S,9R)-3-(3,5-Dichlorophenyl)-1-meth-
yl-2,4-dioxo-1,3,7-triazaspiro[4.4]non-9-yl]-benzonitrile Semi
(+)-DTTA Salt (5b) Directly from 3. A mixture of 3 (68.8
kg, 184.4 mol), glycine (26.4 kg, 351.7 mol), hexamethylene-
tetramine (8.5 kg, 60.6 mol) in 355 kg N-methylpyrrolidinone
and 150 kg of toluene was heated at 144 °C under a nitrogen
sweep. After 6 h the reaction was cooled to 55 °C, THF (355
kg) and a line rinse of toluene (10 kg) were charged, followed
by ethylene diamine (18.6 kg, 309.5 mol) and 11.2 kg of THF.
The solution was held at 55 °C for 1 h and cooled to 20 °C. To
the mixture were charged 20 wt/wt % NaCl in water (1022 kg)
and toluene (20 kg), maintaining temperature <20 °C. The
mixture was stirred 15 min and allowed to settle for 30 min.
The aqueous phase was discarded, and toluene (150 kg) was
charged to the organic phase. The solution was distilled under
vacuum (<50 mbar) at 55 °C. The reaction was cooled to <25
°C, and CH2Cl2 (840 kg) was charged. The mixture was
sampled for KF (target <0.2%) and toluene (target <0.09 g/mL)
content and then polish filtered. and CH2Cl2 (272 kg), H2O (4.2
kg) and MeOH (26.8 kg) were charged. The mixture was stirred
15 min, seeds (175 g) were added. and (+)-DTTA (16.1 kg)
was charged. The slurry was held for 20 h, cooled (0-5 °C).
and held for 3.5 h. The slurry was filtered and washed three
times with CH2Cl2 (120 kg, 270 kg, and 120 kg). The isolated
solid was dried in an agitated filter dryer for 37 h. The resolved
spirocyclic hydantoin 5b was obtained as a white solid (36.2
kg of 5b, 32.2% yield, 99.7 HPLC area % purity, and 98.5%
ee).
Preparation of 5-[(5S,9R)-9-(4-Cyanophenyl)-3-(3,5-dichlo-
rophenyl)-1-methyl-2,4-dioxo-1,3,7-triazaspiro[4.4]non-9-yl-
methyl]-thiophene-3-carboxylic Acid Methyl Ester Hydro-
chloride Salt (6). To a suspension of 5b (23.3 kg, 38.3 mol),
methyl 5-formylthiophene-3-carboxylate (6.8 kg, 40.0 mol), and
280 kg THF was added triethylamine (7.0 kg, 69.2 mol) at
20-25 °C followed by a THF rinse solution (20 kg). The
mixture was stirred until a clear solution was obtained, to which
acetic acid (4.1 kg, 68.3 mol) was added followed by a THF
line rinse (20 kg). The resulting mixture was stirred at 20-25
°C for 1 h and then cooled to 15 °C. Solid sodium triacetoxy-
borohydride (4.1 kg, 19.3 mol) was added, and the reaction
mixture was stirred for 1 h. The addition of sodium triacetoxy-
borohydride was repeated three more times. After all of the
sodium triacetoxyborohydride (16.4 kg, 77.4 mol in total) was
added over 3 h, the reaction mixture was stirred at 20-25 °C
for an additional 4 h. Upon reaction completion (<1% 5b
remaining as monitored by HPLC), MTBE (177.6 kg) and 5%
sodium hydrogen carbonate solution (340 kg) were added to
the resulting mixture. The mixture was agitated for 45 min and
held for 40 min. The aqueous layer was discarded, the organic
layer was polish filtered, and the tank and filter were washed
with 39.5 kg MTBE. The mixture was distilled to a volume of
250 L, and isopropanol (550 kg) was charged. The mixture was
distilled until 220 L was collected, and the mixture was warmed
to 70 °C. A 33% hydrochloric acid solution (5.2 kg) was
charged followed by an isopropanol rinse (20 kg). Seed crystals
(100 g) were charged, and the mixture was cooled to 20 at 15
°C/h and held for 6 h. The slurry was filtered and washed with
isopropanol (50 kg). After drying the methyl ester 6 was
560
•
Vol. 14, No. 3, 2010 / Organic Process Research & Development

obtained as white crystalline solid (21.8 kg of 6, 94.0% yield,
99.3 HPLC area % purity and 98.6% ee). Mp ) 204-207 °C.
1H NMR (CDCl3): 14.22 (1H, b), 8.18 (1H, d, J ) 0.9 Hz),
7.86 (1H, m), 7.67 (2H, d, J ) 8.1 Hz), 7.24 (1H, m), 7.23
(2H, d, J ) 8.1 Hz), 6.67 (2H, m), 4.76 (2H, m), 4.46 (1H, m),
4.16 (1H, m), 4.02 (2H, m), 3.86 (3H, s), 3.75 (1H, m), 3.38
(3H, s). 13C NMR (CDCl3): (171.24, 162.32, 152.98, 136.05,
135.27, 134.03, 132.83, 131.94, 130.46, 128.85, 128.56, 123.92,
117.52, 113.43, 71.13, 52.43, 52.22, 46.73). Anal. Calcd for
C27H23Cl3N4O4S: C, 53.52; H, 3.83; N, 9.25; S, 5.29; Cl, 17.55.
Found: C, 53.07; H, 3.69; N, 9.08; S, 5.23; Cl, 17.20.
Preparation of 5-[(5S,9R)-9-(4-Cyanophenyl)-3-(3,5-dichlo-
rophenyl)-1-methyl-2,4-dioxo-1,3,7-triazaspiro[4.4]non-9-yl-
methyl]-thiophene-3-carboxylic Acid (1) with Isolation of
the THF Solvate. To a solution of 6 (20.12 kg, 33.2 mol) and
1,2-propanediol (6.4 kg) in tetrahydrofuran (226 kg) and water
(127.2) was added cold (0-10 °C) potassium hydroxide solution
(16.0 kg KOH and 127.1 kg H2O) at 8-12 °C over 0.5 h. The
resulting biphasic mixture was stirred at 8-12 °C for 18 h until
the reaction was complete. n-Heptane (208.7 kg) was charged
to the reactor, and after stirring 1 h, agitation was stopped. After
1 h, the aqueous layer was removed, combined with isopropyl
alcohol (39.9 kg), and neutralized to pH 7.5 with aqueous acetic
acid (only a portion of a solution of 20 kg acetic acid and 19.1
kg H2O was charged). Seeds (150 g in 2.1 kg H2O) were
charged to the reaction mixture, and the pH was adjusted to
pH ) 4.8 by addition of aqueous acetic acid. The reaction
mixture was stirred for 12 h, filtered, and washed with H2O
(127.1 kg). The THF solvate 1b was obtained as a white solid
(93.1% yield and 98.8 HPLC area % purity).
(1H, dd, J1 ) 9.6 Hz, J2 ) 6.2 Hz), 3.25 (3H, s), 3.24 (1H,
dd, J1 ) 9.6 Hz, J2 ) 11.7 Hz), 3.01 (1H, d, J ) 11.3 Hz).
13C NMR (acetone-d6): (172.69, 163.7, 153.98, 144.55, 142.23,
135.26, 135.09, 134.41, 133.89, 132.96, 130.33, 128.27, 126.98,
125.18, 119.07, 112.44, 74.28, 59.09, 56.45, 54.33, 50.73,
25.75). Anal. Calcd for C26H20Cl2N4O4S: C, 56.22; H, 3.62; N,
10.08; S, 5.77; Cl, 12.76. Found: C, 56.27; H, 3.20; N, 9.97; S,
5.65; Cl, 12.68.
Preparation of 5-[(5S,9R)-9-(4-Cyanophenyl)-3-(3,5-dichlo-
rophenyl)-1-methyl-2,4-dioxo-1,3,7-triazaspiro[4.4]non-9-yl-
methyl]-thiophene-3-carboxylic Acid (1) Directly from 6. To
a solution of 6 (46.9 kg, 77.6 mol) and 1,2-propanediol (11.8
kg) in tetrahydrofuran (41.7 kg) and water (266.8 kg) was added
cold (0-10 °C) potassium hydroxide solution (1 N, 244.5 kg)
at 8-12 °C in 0.5 h. The resulting biphasic mixture was stirred
at 8-12 °C for 18-24 h until the reaction was complete (<1%
6 remaining as monitored by HPLC). The reaction mixture was
washed with n-heptane (385.7 kg). The pH was adjusted to 7.5
with addition of 1.5 M citric acid (22.9 kg). Isopropyl acetate
(817.8 kg) was charged, and 1.5 M citric acid(aq) (∼22.9 kg)
was added until a pH of 6.5 was attained. After agitating for
15 min and holding for 30 min, the aqueous layer was discarded,
and the organic layer was washed with H2O (470 kg). The
solution was then polish filtered, and isopropylacetate (52.2 kg)
was used to rinse the polish filter assembly. The solution was
concentrated under reduced pressure (240 Torr) to a volume of
718 L at <45 °C. Seeds (500 g) were charged, and the distillation
was continued until a volume of ∼207 L was attained. Heptane
(117.8 kg) was charged, the slurry was cooled to 20 °C over
1.5 h and was subsequently wet milled until d90 < 60 µm. The
slurry was held for >2 h and filtered. The cake was washed
with a 1:1 isopropyl acetate/heptane solution (109.7 kg)
isopropyl acetate and dried in vacuum at 35-40 °C to a constant
weight. Acid 1 (39.6 kg, 91.5% yield and 99.33 HPLC area %
purity) was obtained as a white and sandy crystalline solid.
The THF solvate 1b (20.1 kg) was charged to the reactor
followed by isopropyl acetate (592.6 kg) and H2O (7.6 kg). The
mixture was heated to 68 °C and after complete dissolution,
polish filtered. The solution was distilled under vacuum (240
Torr) until approximately 10% of the volume was removed.
Seeds (250 g in 1.1 kg isopropylacetate) were charged and the
distillation was continued until only 160 L remained. The slurry
was cooled to 40 °C and sampled for KF (0.032%). The slurry
was cooled to 20-25 °C, passed through a wet mill for 1 h to
reduce the particle size to d90 < 60 µm. The crystal slurry was
stirred at 20-25 °C for g2 h and was then filtered. The cake
was washed with (5-10 °C) isopropylacetate (22.2 kg) and
dried in vacuum at 35-40 °C to a constant weight. Acid 1 (15.0
kg, 80.4% yield, 99.9 HPLC area % purity and 99.5 weight %
purity) was obtained as white and sandy crystalline solid. The
Acknowledgment
We thank the BMS pilot plant operations and Analytical
R&D staff for their valuable support during our scale-up runs.
In addition, we thank the safety group for hazard assessment
studies and Mary Malley for crystal structure analysis. We also
thank Drs. Rodney Parsons, David Kronenthal, and Akin
Davulcu for helpful discussion during the preparation of the
manuscript.
1
overall yield from 6 was 74.8%. Mp ) 209-230 °C.28 H NMR
Supporting Information Available
React-IR data for the ethylenediamine cleavage of 10 and
DSC of 1. This material is available free of charge via the
Internet at http://pubs.acs.org.
(acetone-d6): 8.19 (1H, d, J ) 1.3 Hz), 7.76 (2H, d, J)8.4 Hz),
7.49 (2H, d, J ) 8.2 Hz), 7.43 (1H, d, J ) 1.0 Hz), 7.41 (1H,
t, J ) 1.9 Hz), 6.87 (2H, d, J ) 1.9 Hz), 4.16 (1H, dd, J1 )
13.9 Hz J2 ) 0.8 Hz), 4.10 (1H, dd, J1 ) 11.7 Hz, J2 ) 6.2
Hz), 3.99 (1H, d, J ) 14.0 Hz), 3.48(1H, d, J ) 10.6 Hz), 3.47
Received for review December 4, 2009.
OP9003168
(28) For a copy of the DSC, please see the Supporting Information.
Vol. 14, No. 3, 2010 / Organic Process Research & Development
•
561