Pilot-plant preparation of an αβ3 integrin antagonist: Process development of a carbonyldiimidazole peptide coupling

Article abstract of DOI:10.1021/op8002213

The process development of a peptide coupling with CDI is discussed. Various solvents, addition orders, stoichiometries, and reaction temperatures were investigated. A reliable crystallization procedure was also developed. The new process was piloted to p

Full text of DOI:10.1021/op8002213

Organic Process Research & Development 2009, 13, 60–63
Pilot-Plant Preparation of an rvꢀ3 Integrin Antagonist: Process Development of a
Carbonyldiimidazole Peptide Coupling
Gerald A. Weisenburger,*, D. Keith Anderson,O Jerry D. Clark,O Albert D. Edney,† Puneh S. Karbin,‡ Donald J. Gallagher,¶
Carl M. Knable,# and Mark A. Pietz4
Chemical Research and DeVelopment, Pfizer Global Research DiVision, Pfizer Inc., Eastern Point Road, Groton,
Connecticut 06340, U.S.A., and Portfolio Resource Management, Pfizer Global Research DiVision, Pfizer Inc.,
700 Chesterfield Parkway, Chesterfield, Missouri, 63017, U.S.A.
Scheme 1. Original route used to prepare supplies of 3 for
preclinical development
Abstract:
The process development of a peptide coupling with CDI is
discussed. Various solvents, addition orders, stoichiometries, and
reaction temperatures were investigated. A reliable crystallization
procedure was also developed. The new process was piloted to
provide 342 kg of product in two batches with an average 85%
yield and 99% assay.
Introduction
hydroxysuccinimide ester for 8 h at 20 °C. Vacuum distillation
of the DMF, addition of ethyl acetate, a series of aqueous
extractions, a second distillation for drying, back addition of
ethyl acetate, and addition of heptane antisolvent provided 3 in
66% average yield. Our goals for improving the synthesis of 3
were (1) to lower the cost by replacing the expensive N-t-Boc-
glycine hydroxysuccinimide ester, (2) to reduce the cycle time,
and (3) to develop a more reliable process for the crystallization
of 3.
The integrin Rvꢀ3 plays an essential role in angiogenesis,
the process by which new blood vessels form from pre-existing
blood vessels.1 Angiogenesis is required for tumor growth, and
therefore, antagonists of Rvꢀ3 are being studied for the treatment
of cancer. We have previously reported the synthesis of the
left-hand tetrahydropyrimidine portion of antagonist 1 and a
diastereoselective imino-Reformatsky reaction to install the (S)-
ꢀ-amino acid functionality in the right-hand portion.2 As part
of a report describing the successful pilot-plant preparation of
1, herein is detailed the chemical process research and develop-
ment for installation of the glycine linker connecting the right-
and left-hand portions of the molecule.3 This paper describes
improvements in the synthesis of 3, a key intermediate in the
synthesis of 1.
Results and Discussion
The high cost of N-t-Boc-glycine hydroxysuccinimide ester
($315/mol) prompted us to investigate reactions of 2 with N-t-
Boc-glycine ($51/mol) in the presence of several standard
coupling agents. Carbonyldiimidazole (CDI) was chosen for
several reasons: CDI is relatively inexpensive ($24/mol);
coupling of amino acid salts with amines has been reported;
the release of CO2 provides a driving force for reaction; and
the imidazole·pTSA byproduct could be easily removed by an
aqueous extraction.4
Initial experiments were carried out at 25 °C by adding
N-methylpyrrolidinone (NMP) to a 1:1 mixture of CDI and N-t-
Boc-gylcine to generate the activated glycine derivative 4 as
shown in Scheme 2.5 Addition of 1.0 equiv of the solid amine
Original Process. Initial supplies of 3 for preclinical
development employed a process in which a DMF solution of
2 was stirred with N-methylmorpholine and N-t-Boc-glycine
(2) (a) Clark, J. D.; Collins, J. T.; Kleine, H. P.; Weisenburger, G. A.;
Anderson, D. K. Org. Process Res. DeV. 2004, 8 (4), 571. (b) Clark,
J. D.; Weisenburger, G. A.; Anderson, D. K.; Colson, P.-J.; Edney,
A. D.; Gallagher, D. J.; Kleine, H. P.; Knable, C. M.; Lantz, M. K.;
Moore, C. M. V.; Murphy, J. B.; Rogers, T. E.; Ruminski, P. G.; Shah,
A. S.; Storer, N.; Wise, B. E. Org. Process Res. DeV. 2004, 8 (1), 51.
(3) Cunningham, J.; Gordon, G. B.; Nickols, G. A.; Westlin, W. F.; Rogers,
T. E.; Ruminski, P. G. PCT Int. Appl. WO Patent 2000051686, CAN
133:223050, 2000.
(4) Paul, R.; Anderson, G. W. J. Am. Chem. Soc. 1960, 82, 4596.
(5) The addition of solvent to the two solids resulted in rapid reaction
and release of CO2. On small scale (1g) we did not experience any
foaming or pressurization problems because we had sufficient head-
space and a relatively large vent. Obviously, this mode of addition
would need to be very carefully evaluated prior to scaling up.
* gerald.a.weisenburger@pfizer.com.
Chemical Research and Development, Pfizer Global Research Division,
Pfizer Inc.
O Portfolio Resource Management, Pfizer Global Research Division, Pfizer
Inc.
† Current Address: Abbott Laboratories, 431 Parkshire Place Drive, Dardenne
Prairie, MO 63368.
‡ Current Address: 1495 N. Clybourn, Unit E, Chicago, IL 60610.
¶ Current Address: Neurocrine Biosciences, 12780 El Camino Real, San Diego,
CA 92130.
# Current Address: Abbott Laboratories, 1401 Sheridan Road, North Chicago,
IL 60064.
4 Current Address: Eli Lilly, 1400 W. Raymond, Indianopolis, IN 46285.
(1) Folkman, J. Sci. Am. 1996, 275, 150.
60
•
Vol. 13, No. 1, 2009 / Organic Process Research & Development
10.1021/op8002213 CCC: $40.75
2009 American Chemical Society
Published on Web 11/26/2008

Scheme 3. Impurities formed during the synthesis of 3
using N-t-Boc-glycine and 1,1′-carbonyldiimidazole
Figure 1. Formation and subsequent consumption of 5 during
a 2 h addition of 4 to 2.
Scheme 2. Preparation of 3 by the reaction of the
imidazolide 4 with 2; O-acylation of 3 gives the diglycine
impurity 5
therefore, the reaction yield. For example, a reaction temperature
of 60 °C gave significant amounts of the lactone 6 through
intramolecular cyclization of 3 as shown in Scheme 3. Excess
2 combined with elevated reaction temperatures gave 7 through
reaction of 2 with either 3 or 6. Since significant amounts of 6
and 7 were observed within 90 min at 60 °C, we chose to run
the reaction at 25 °C and accept the longer reaction time in
exchange for lower impurity levels. Careful control of the
stoichiometry was necessary to minimize impurites. Excess CDI
led to the formation of the carbamate 8 by intramolecular
cyclization of 2.8 Even though excess 4 leads to 5, in practice,
a nominal excess of 4 was necessary for high conversion of
the expensive 2.9 Small amounts of unreacted 2 were easily
removed by an acidic extraction.
Our next set of experiments focused on replacing NMP as
the reaction solvent. Ethyl acetate, tetrahydrofuran, and toluene
were screened as alternative solvents. In each case the solvent
was added to a 1:1 mixture of CDI and N-t-Boc-glycine
followed by the addition of 1.0 equiv of 2.5 The heterogeneous
reactions were stirred at room temperature for 5 h and monitored
by HPLC. A significant increase in the amount of 7 was
observed when toluene was employed as the solvent. The
reaction rates and impurity profiles with ethyl acetate or THF
as solvent were essentially identical to those observed with
NMP. Therefore, replacement of NMP with ethyl acetate or
THF appeared promising, but one more issue had to be
addressed. For ease of reaction screening in the laboratory, we
had been adding solid 2 to a solution of 4, but addition of the
agglomerated solid was not practical for our pilot plant.
Therefore, a process employing a slurry or a solution of 2 was
needed. Attempts to slurry 2 in ethyl acetate or THF were not
salt 2 to the solution of 4 provided 3 in 93% crude yield. The
diglycine impurity 5, derived from O-acylation of 3, was also
observed in 3 area %. Reaction monitoring of the homogeneous
solution 30 min after the addition revealed a 7:2:1 ratio of 3:2:
5.6 Aging of the reaction mixture for 4.5 h led to a 37:2:1 ratio
of 3:2:5. The observation that the ratio of 2 to 5 remains constant
during the aging period suggests that 2 reacts with the phenolic
glycine moiety of 5 to give 2 equiv of the desired product 3.7
These observations suggest that the rate-limiting step for
complete conversion of 2 to 3 is the reaction of 2 with 5.
Having identified the feasibility of performing the desired
conversion, process optimization studies were completed. Since
the reaction of 2 with 5 to give 3 is relatively slow compared
to the reaction of 2 with 4, we wanted to design the process to
minimize the reaction pathway involving 5. We felt that this
could be accomplished by reversing the addition order. As
shown in Figure 1, 5 was formed in less than 1 mol % during
the first 40 min of a 2 h addition of 4 to 2. An exponential
increase in 5 occurred as the concentration of 2 decreased and
the concentration of 3 increased. Use of a large excess of the
expensive starting material 2 was not a practical solution to
overcome this concentration effect.
Reaction temperature and stoichiometry were identified as
two critical process parameters affecting impurity formation and,
(6) We assume that 4 has been completely consumed at this point in the
reaction. We did not develop an analytical method to confirm this
assumption.
(7) An alternative mechanism is rate-limiting equilibration of 5 with
imidazole to give 4 followed by reaction with 2 to give 3.
(8) The urea resulting from addition of 2 equiv of 2 to CDI was not
observed under these reaction conditions.
(9) Due to time constraints, we did not develop an assay for CDI or 4.
All bulk materials were use-tested in the lab prior to the scale-up.
Vol. 13, No. 1, 2009 / Organic Process Research & Development
•
61

Table 1. Solubility of 3 in various mixtures of EtOAc,
toluene, and heptane at 55, 25, and -5 °C
successful because the solids stuck together to form a solid mass
that was not readily stirred. Therefore, we could not completely
replace NMP because it was needed to dissolve 2, but we were
able to significantly reduce the amount by preparing 4 in ethyl
acetate.10
solvent composition (vol %)
solubility of 3 (mg/mL)
EtOAc
toluene
heptane
55 °C
25 °C
-5 °C
20
30
10
10
26.5
33.3
100
0
20
10
30
10
26.5
33.3
0
100
0
0
60
60
60
80
47
33.4
0
50.9
81.5
22.4
7.0
116
168
380
141
0.18
193
10.8
340
20.2
20.7
10.3
3.0
30.2
63.0
244
95.0
0.05
117
7.67
167
8.1
13.5
3.7
Having defined the method of interaction of the substrate
with the reagents, we now focused our studies on the transfer
of the technology to a two-reactor workstation in the pilot plant.
This was successfully accomplished employing the following
protocol. First, a solution of 4 in ethyl acetate was prepared by
adding a solution of N-t-Boc-glycine from reactor 1 to a slurry
of CDI in reactor 2. Reactor 1 was rinsed with ethyl acetate,
and a solution of 2 in NMP was prepared in it. The ethyl acetate
solution of 4 in reactor 2 was quickly added to the solution of
2 in reactor 1, and the heterogeneous mixture was stirred at
room temperature for 3.5 h.11 The reaction rates and impurity
profiles with the mixed solvent system using inverse addition
were essentially identical to those observed with NMP in the
normal addition mode. The reaction was diluted with ethyl
acetate, washed with 1 M HCl, and washed with water to
provide an ethyl acetate solution of 3.
2.1
13.8
22.2
166
12.1
0.01
35.3
2.15
101
0
0
100
50
50
0
50
0
50
50
50
Table 2. Results from kilo lab and pilot-plant preparations
of 3 from reaction of 2 with 4 using an optimized process
2
4
3
3
3
(kg) (equiv) (% yield) (% assay) (mol % in mother liquor)
2.75 1.05
74.5
84.8
86.4
98.7
98.9
99.7
11.4
7.1
8.0
203
203
1.10
1.13
Our third goal was to improve the crystallization of 3 from
the ethyl acetate solution. The original process called for a
vacuum distillation of the ethyl acetate solution to remove water.
Ethyl acetate was back added, and the solution was warmed to
55 °C. Crystallization of the product was achieved by addition
of heptane and seed crystals, and cooling to 0 °C. The procedure
was unreliable, and oiling of the product was periodically
observed after the heptane charge or during the cool down.
Substitution of heptane with hexane helped alleviate the oiling
problem, but we wanted to avoid handling hexane on large scale.
Because the solubility of 3 was very low in heptane and very
high in ethyl acetate, 3 had a propensity to become highly
supersaturated and oil out during crystallization attempts from
a two-solvent system of heptane/ethyl acetate. It appeared that
an alternative antisolvent was needed to provide a bridge from
high solubility to low solubility and enable a controlled
crystallization of 3. Since 3 had intermediate solubility in
toluene, it was a logical choice, but on the basis of solubility
data, it became clear that a toluene/ethyl acetate system would
result in low yields and heptane was necessary for recovery.
Therefore, we decided to try a three-solvent system for the
crystallization of 3. Solubility data at 55, 25, and -5 °C for 3
in various ratios of ethyl acetate, toluene, and heptane are shown
in Table 1.
Response surfaces were generated, and from these data we
determined that an end point consisting of 7 volumes of a 1:1:3
ethyl acetate:toluene:heptane mixture would allow for maximum
recovery of 3 and good operability in a pilot plant. We found
that decreasing the relative amount of ethyl acetate led to oiling
out of the product. The desired ratio was achieved through an
atmospheric distillation of ethyl acetate followed by addition
of toluene at 55 °C. Quantitation of ethyl acetate was achieved
by means of GC analysis with toluene providing an internal
standard. Ethyl acetate could be back added if too much was
removed by distillation, or toluene could be added to obtain
the desired 1:1 ratio in the case of under-distillation. Heptane
was added slowly during three charges because of the slow
desaturation rate of 3. The first heptane charge was designed
to give 20% supersaturation of 3 since oiling had been observed
at 40% supersaturation. Seeding, two additional heptane charges,
and controlled slow cooling to -5 °C resulted in consistent
crystallization of the product.
The new process was run once in a kilo laboratory and twice
in a pilot plant. The data are summarized in Table 2. The initial
scale up to 2.75 kg of 2, and 1.05 equiv of 4 provided 3 in
74.5% yield with a 98.7% assay. During the crystallization
portion of the process the ethyl acetate was under-distilled and,
therefore, additional toluene and heptane were used. This
resulted in an 11 mol % loss of product to the mother liquor.
The first 203 kg batch of 2 run in the pilot plant used 1.10
equiv of 4 based on laboratory use tests. The desired product 3
was obtained in 84.8% yield with an assay of 98.9%. We
increased 4 in the second batch to 1.13 equiv to increase the
conversion of 2 to 3. The product was obtained in 86.4% yield
with an assay of 99.7%. Since overdistillation of ethyl acetate
was achieved during the crystallization in both pilot plant
batches, ethyl acetate was back added and product loss to the
mother liquor was 7.1 and 8.0 mol%. We attribute the lower
mass balance of the kilo laboratory preparation of 3 to the lower
purity 2 used in that synthesis.12
In summary, a procedure was developed and demonstrated
on a pilot-plant scale that N-t-Boc-glycine in the presence of
CDI is a cost-effective replacement for N-t-Boc-glycine hy-
droxysuccinimide ester for the synthesis of 3. Coupled with
improvements in the crystallization procedure, the new process
provided 342 kg of product in two batches with an average
85% yield and 99% assay.
(10) We chose ethyl acetate over THF because of the cost difference.
(11) Slow addition of 4 to 2 resulted in longer reaction times due to the
slow build up of 5 and subsequent slow reaction of 5 with 2 to give
3.
(12) The purity of 2 was 83.1, 88.1, and 91.6% in the kilo lab prep and
two piloting preps, respectively.
62
•
Vol. 13, No. 1, 2009 / Organic Process Research & Development

Experimental Section
1,1′-carbonyldiimidazole (4.10 g, 25.3 mmol), and the solution
was stirred for 1 h. 2 (5.00 g, 10.1 mmol) was added, and the
solution was stirred for 3 h. Reaction monitoring revealed a
1.3:1 ratio of 3:5. The addition of N-methylmorpholine (0.71
mL, 6.4 mmol) did not significantly change the product ratio
after stirring for 15 h. In a separate flask N-methylpyrrolidinone
(8 mL) was added to N-t-Boc-glycine (2.66 g, 15.2 mmol) and
carbonyldiimidazole (2.46 g, 15.2 mmol), and the solution was
stirred for 1.25 h. The reaction mixture containing 3 and 5 was
added to the solution of 4 and stirred for 1 h. The ratio of 3:5
did not significantly change. Ethyl acetate (50 mL) was added,
and the organic was washed with brine (50 mL), 1 M HCl (2
× 50 mL), sat. NaHCO3 (50 mL), dried over NaSO4, and
concentrated to give a mixture of 3, 5, and NMP (4.93 g) as a
yellow foam. The foam was dissolved in 50% ethylacetate/
heptane (15 mL), washed with water (15 mL), dried over
NaSO4, and concentrated. Separation of 3 and 5 was achieved
by reverse phase chromatography using 50% ACN in water to
provide 1.98 g (30.8% yield) of analytically pure 5. 1H NMR
(CDCl3, 400 MHz) δ 1.20 (t, J ) 7.5, 3H), 1.44 (s, 9H), 1.47
(s, 9H), 1.73 (s, 1H), 2.84 (br d, J ) 7.5, 2H), 3.75 (br d, J )
7.5, 2H), 4.11 (q, J ) 7.5, 2H), 4.22 (br d, J ) 5.0, 2H),
5.22-5.32 (br s, 1H), 5.38-5.47 (m, 2H), 7.36 (d, J ) 3.5,
1H), 7.52 (d, J ) 3.5, 1H). Anal. Calcd for C25H35BrClN3O9:
C, 47.14; H, 5.54; Br, 12.55; Cl, 5.57; N, 6.60. Found: C, 46.74;
H, 5.54; Br, 12.03; Cl, 5.86; N, 6.49.
(S)-Ethyl 2-(8-bromo-6-chloro-2-oxo-3,4-dihydro-2H-
benzo[e][1,3]oxazin-4-yl)acetate (8). N-methylpyrrolidinone (5
mL) was added to a flask containing 2 (2.00 g, 4.04 mmol)
and 1,1′-carbonyldiimidazole (0.3277 g, 2.02 mmol), and the
solution was stirred for 7 h. Triethylamine (0.56 mL, 4.02
mmol) was added, and the solution was stirred overnight. Note:
We were initially attempting to make the urea resulting from
addition of 2 equiv of 2 to CDI, but none was observed under
these reaction conditions. A second charge of 1,1′-carbonyldi-
imidazole (0.3277 g, 2.02 mmol) was added and the solution
was stirred overnight. Ethyl acetate (20 mL) was added and
the organic was washed with brine (20 mL), 1 M HCl (20 mL),
sat. NaHCO3 (20 mL), dried over NaSO4, and concentrated to
give 1.06 g (75% yield) of 8 as a white solid. 1H NMR (CDCl3,
400 MHz) δ 1.29 (t, 3H), 2.78 (d, 2H), 4.22 (q, 2H), 4.94
(td, 1H), 6.27 (br d, 1H), 7.07 (d, 1H), 7.56 (d, 1H). 13C
NMR (CDCl3, 125 MHz)14.1, 43.0, 49.9, 61.7, 111.5,
122.0, 124.6, 130.0, 132.9, 145.5, 148.7, 170.2. Anal.
Calcd for C12H11BrClNO4: C, 41.35; H, 3.18; Br, 22.92;
Cl, 10.17; N, 4.02. Found: C, 41.50; H, 3.23; Br, 24.19;
Cl, 9.56; N, 4.09.
(3S)-N-[(1,1-Dimethylethoxy)carbonyl]glycyl-3-[3-bromo-
5-chloro-2-hydroxyphenyl]-ꢀ-alanine ethyl ester (3). A solu-
tion of N-t-Boc-glycine (81.2 kg, 464 mol) at 25 °C in ethyl
acetate (428 L) in reactor 1 was transferred to a slurry of CDI
(75.2 kg, 464 mol) at 25 °C in ethyl acetate (325 L) in reactor
2. Reactor 1 was rinsed with ethyl acetate (50 L), and the rinse
was transferred to reactor 2. Minor foaming was observed, but
the pressure did not increase. The clear yellow solution was
stirred for 21 h at 25 °C. Note: A reaction time of only 1 h is
necessary for complete conversion to 4. A solution of 2 (221.6
kg, 91.6% assay, 410.3 mol) in NMP (239 L) was prepared in
reactor 1 at 25 °C. The solution of 4 in reactor 2 was transferred
to the solution of 2 in reactor 1 over 19 min. Reactor 2 was
rinsed with ethyl acetate (50 L), and the rinse was transferred
to reactor 1. The reaction mixture was sampled after 4 h at 25
°C, and HPLC analysis showed 2.7 area % of 2 remained. The
reaction mixture was diluted with ethyl acetate (828 L), washed
once with 1 M HCl (aq) (1188 L), and washed three times with
water (3 × 1170 L). The organic layer was concentrated by
atmospheric distillation (71-84 °C pot temperature) of ethyl
acetate (1090 L) over a 7 h period, and then the solution was
cooled to 55 °C. Toluene (275 L) was charged at a rate to keep
the solution above 50 °C. The solution was sampled, and GC
analysis showed that the weight ratio of ethyl acetate:toluene
was 0.40. Ethyl acetate (169 L) was added to achieve the desired
1.03 weight ratio. Heptane (550 L) was charged at a rate to
keep the solution above 45 °C, and the solution was reheated
to 50 °C. Seeds of 3 (1.7 kg) were added, and the slurry was
stirred at 50 °C for 4 h. Heptane (137 L) was charged and the
slurry was stirred for 100 min at 50 °C. Heptane (137 L) was
added and the slurry was cooled to 45 °C over 50 min and
held for 75 min. The slurry was cooled to 30 °C over 3.25 h
and held for 1 h, and then cooled to -5 °C over 3.42 h and
held for 37.25 h. The slurry was filtered, washed with a mixture
of cold toluene/heptane (217 L/412 L), and dried in a tumble
dryer to give 3 (170.6 kg, 84.8% yield adjusted for seeds and
assay). The product was 98.9 wt % pure by HPLC analysis
against a standard. HPLC method: YMC basic 150 mm × 4.6
mm, 5 µm, 25 °C, 1.0 mL/min, 220 nm, 50% pH 7 buffer:
50% MeOH 20 min gradient up to 80% MeOH, retention times
(min), 2(7.48), 3 (10.98), 5 (12.29), 6 (8.89), 7 (11.97), 8 (8.29).
Characterization for a typical standard of 3 follows: 1H NMR
(CDCl3, 400 MHz) δ 1.20 (t, J ) 7.5, 3H), 1.43 (s, 9H),
2.81-2.95 (m, 2H), 3.75-3.87 (m, 2H), 4.13 (q, J ) 7.5, 2H),
5.09-5.17 (m, 1H), 5.53-5.98 (m, 1H), 7.13 (d, J ) 3.0 Hz,
1H), 7.22-7.35 (br s, 1H), 7.41 (d, J ) 3.0 Hz, 1H), 7.60 (d,
J ) 10 Hz, 1H). 13CNMR (CDCl3, 125 MHz) 14.0, 28.2, 38.0,
44.3, 46.1, 61.0, 80.5, 111.9, 125.5, 127.0, 129.2, 131.2, 149.4,
156.1, 170.0, 170.7.
(3S)-N-[(1,1-Dimethylethoxy)carbonyl]glycyl-3-[3-bromo-
5-chloro-2-(((1,1-dimethylethoxy)carbonyl)glycyl)oxy]-ꢀ-ala-
nine ethyl ester (5). N-methylpyrrolidinone (10 mL) was added
to a flask containing N-t-Boc-glycine (4.43 g, 25.3 mmol) and
Received for review September 10, 2008.
OP8002213
Vol. 13, No. 1, 2009 / Organic Process Research & Development
•
63