Pilot Plant Preparation of an αvβ3 Integrin Antagonist. Part 1. Process Research and Development of a (S)-β-Amino Acid Ester Intermediate: Synthesis via a Scalable, Diastereoselective Imino-Reformatsky Reaction

Article abstract of DOI:10.1021/op034094j

Described are four process research investigations directed toward discerning a scalable, enantioselective method for preparing (S)-β-amino acid ester 3, a key intermediate to the α_vβ₃ integrin antagonist 1. Reported are an asymmetric Michael reaction approach, attempts to enantioselectively hydrogenate an enamine, resolution of (±)-3 via diastereomeric salt formation, and a synthetic route employing a novel, diastereoselective imino-Reformatsky reaction. This last research initiative proved successful and was employed as the enabling route to initial API supply. Process development of this enabling chemistry is reported. The technical issues researched and optimized were (1) the necessity of employing MEM-protection for high yield and diastereoselectivity in the imino-Reformatsky reaction, (2) the reaction kinetics of MEM chloride hydrolysis and the application of these data to an on-scale quench procedure, (3) the efficient formation of the (S)-phenylglycinol imine 15 in NMP and a dehydration of this product solution on-scale, employing molecular sieves, (4) a calorimetric study of the Reformatsky reaction and the application of these data, (5) the replacement of Pb(OAc)₄ with NaIO₄ and the use of methylamine to sequester competing oxazolidine formation, and (6) further development of the isolation and purification protocol for the ethyl ester, p-TsOH salt of (S)-3. The results and challenges associated with two campaigns in which the potential commercial process was practiced are discussed.

Full text of DOI:10.1021/op034094j

Organic Process Research & Development 2004, 8, 51−61
Pilot Plant Preparation of an rvâ3 Integrin Antagonist. Part 1. Process
Research and Development of a (S)-â-Amino Acid Ester Intermediate: Synthesis
via a Scalable, Diastereoselective Imino-Reformatsky Reaction
Jerry D. Clark,*,† Gerald A. Weisenburger,‡ D. Keith Anderson,§ Pierre-Jean Colson,‡ Albert D. Edney,§
Donald J. Gallagher,‡ H. Peter Kleine,| Carl M. Knable,‡ Melissa K. Lantz,| Christine M. V. Moore,‡ James B. Murphy,§
Thomas E. Rogers,| Peter G. Ruminski,| Ajit S. Shah,| Neil Storer, and Bruce E. Wise‡
Global Research & DeVelopment, Pfizer Corporation, 2800 Plymouth Road, Ann Arbor, Michigan 48105, U.S.A.,
Global Research & DeVelopment, Pfizer Corporation, 4901 Searle Parkway, Skokie, Illinois 60077, U.S.A.,
Global Research & DeVelopment, Pfizer Corporation, 800 North Lindbergh BouleVard, St. Louis, Missouri 63167, U.S.A.,
Research and DeVelopment, Pfizer Corporation, 700 Chesterfield Parkway, Chesterfield, Missouri 63017, U.S.A., and
Oxford Asymmetry International, 151 Milton Park, Abington, Oxon OX14 4SD, UK
Abstract:
1 was required.2 As part one of a three part report describing
the successful pilot-plant preparation of 1, herein we wish
to detail the chemical process research and development of
a scalable and enantioselective method of preparing (S)-â-
amino ester 3, a key intermediate in the convergent synthesis
of 1 (Scheme 1).
A process research and development team was formed
to evaluate strategies by which 1 could be prepared on-scale.
Envisioned was a convergent synthesis in which tetrahydro-
pyrimidine 2 would be linked via glycine to the â-amino
acid ester 3. With regard to the preparation of 3, four research
programs3 were completed: (1) an asymmetric Michael
addition reaction to an appropriately substituted cinnamic
acid ester, (2) the enantioselective hydrogenation of a
corresponding enamine, (3) the resolution of (()-3 via
diastereomeric salt formation and (4) evaluation of a synthetic
route developed in-house employing a novel, diastereose-
lective imino-Reformatsky reaction as the key reaction.4
Described are four process research investigations directed
toward discerning a scalable, enantioselective method for
preparing (S)-â-amino acid ester 3, a key intermediate to the
rvâ3 integrin antagonist 1. Reported are an asymmetric Michael
reaction approach, attempts to enantioselectively hydrogenate
an enamine, resolution of (()-3 via diastereomeric salt forma-
tion, and a synthetic route employing a novel, diastereoselective
imino-Reformatsky reaction. This last research initiative proved
successful and was employed as the enabling route to initial
API supply. Process development of this enabling chemistry is
reported. The technical issues researched and optimized were
(1) the necessity of employing MEM-protection for high yield
and diastereoselectivity in the imino-Reformatsky reaction, (2)
the reaction kinetics of MEM chloride hydrolysis and the
application of these data to an on-scale quench procedure, (3)
the efficient formation of the (S)-phenylglycinol imine 15 in
NMP and a dehydration of this product solution on-scale,
employing molecular sieves, (4) a calorimetric study of the
Reformatsky reaction and the application of these data, (5) the
replacement of Pb(OAc)4 with NaIO4 and the use of methy-
lamine to sequester competing oxazolidine formation, and (6)
further development of the isolation and purification protocol
for the ethyl ester, p-TsOH salt of (S)-3. The results and
challenges associated with two campaigns in which the potential
commercial process was practiced are discussed.
Process Research
Asymmetric Michael Addition Route to 3. Studies
directed toward the synthesis of 3 began by exploring the
feasibility of employing an asymmetric Michael reaction
(1) (a) Eliceri, B. P.; Cheresh, D. A. Cancer 2000, 6S3, 245. (b) Roberto, R.;
Angelo, V.; Domenico, R.; Francesco, D. R.; Francesca, M.; Franco, D.
Haematologica 2002, 87, 836. (c) Ruminski, P. G.; Rogers, T. E.; Rico, J.
E.; Nickols, G. A.; Westlin, W. F.; Engleman, V. W.; Settle, S. L.; Keene,
J. L.; Freeman, S. K.; Carron, C. P.; Meyer, D. M. “Book of Abstracts”,
218th National Meeting of the American Chemical Society, New Orleans,
1999; American Chemical Society: Washington, D. C., 1999. (d) Ruminski,
P. G.; Rogers, T. E.; Rico, J. E.; Nickols, G. A.; Engleman, V. W.; Settle,
S. L.; Keene, J. L.; Freeman, S. K.; Carron, C. P.; Meyer, D. M. “Book of
Abstracts”, 217th National Meeting of the American Chemical Society,
Anaheim, Calif., 1999; American Chemical Society: Washington, D. C.,
1999. (e) Flynn, D. L.; Abood, N.; Bovy, P.; Garland, R.; Hockerman, S.;
Lindmark, R. J.; Schretzman, L.; Rico, J.; Rogers, T. Book of Abstracts,
210th National Meeting of the American Chemical Society, Chicago, IL,
1995; American Chemical Society: Washington, D. C., 1995.
(2) Collins, J. T.; Devadas, B.; Lu, H.-F.; Malecha, J. W.; Miyashiro, J. M.;
Nagarajan, S.; Rico, J. G.; Rogers, T. E. U.S. Patent 6100423, 2000.
(3) The results of a bio-resolution route to (S)-3 will be the subject of a separate
report.
Introduction
The Rvâ3 integrin plays an essential role in angiogenesis,
the process by which new blood vessels form from preexist-
ing blood vessels.1 Angiogenesis is required for tumor
growth, and therefore, antagonists of Rvâ3 are being studied
for the treatment of cancer. In a program directed toward
the discovery of such antagonists, an improved synthesis of
* To whom correspondence should be addressed. Telephone: 734-622-4411.
Fax: 734-622-3294. E-mail: jerry.d.clark@pfizer.com.
† Global Research & Development, Pfizer Corporation, Ann Arbor, MI.
‡ Global Research & Development, Pfizer Corporation, Skokie, IL.
§ Global Research & Development, Pfizer Corporation, St. Louis, MO.
| Research and Development, Pfizer Corporation, Chesterfield, MO.
Oxford Asymmetry International.
(4) (a) Colson, P.-J., Awashti, A. K.; Nagarajan, S. R PCT Int. Appl. 1999,
WO 9944985. (b) Colson, P-. J.; Boys, M. L.; Farid, P. N.; Awasthi, A. K.;
Ainsworth, O., unpublished results. (c) Colson, P-. J.; Boys, M. L.; Cain-
Janicki, K. J.; Doubleday: W. W.; Farid, P. N.; Duran, J. E. unpublished
results.
10.1021/op034094j CCC: $27.50 © 2004 American Chemical Society
Published on Web 12/18/2003
Vol. 8, No. 1, 2004 / Organic Process Research & Development
•
51

Scheme 1. Key intermediates toward preparation of 1
Scheme 2. Asymmetric Michael reaction approach to 3
(Scheme 2). Envisaged was the preparation of 7, a precursor
to 3, via the 1,4-conjugate addition of a homochiral lithium
R-methylbenzyl alkyl amide to cinnamic acid ester 6.5
Indeed, after extensive work, 7 was prepared with high
diastereoselectivity. Starting from commercially available
salicylaldehyde 4, the phenol was protected as the MOM
ether 5 in 74% isolated yield. The tert-butyl cinnamate 6
was prepared by Wadsworth-Emmons6 chemistry to afford
exclusively the trans-olefin in 82% isolated yield. Michael
addition at -30 °C employing lithium R-methylbenzyl allyl
amide7 gave diastereoselectively 7 in 75% yield and with
>98% de. Noteworthy aspects of this chemistry were the
use of a MOM protecting group and the tert-butyl ester.
Other esters and protecting groups were explored, but only
this combination diminished competing 1,2-nucleophilic
addition and gave rise to high diastereoselectivity. Also
noteworthy was completing the desired asymmetric Michael
reaction at -30 °C, a temperature much higher than generally
reported.5
Removal of the allyl group was accomplished successfully
on small scale, employing Pd(PPh3)4 and N,N-dimethylbar-
bituric acid in dichloromethane.8 This afforded 8 as an orange
oil in 91% yield. However, removal of the R-methylbenzyl
auxiliary also gave rise to dehalogenation of the aromatic
ring. For example, hydrogenation using Pearlman’s catalyst
in MeOH and acetic acid under 1 atm hydrogen resulted in
deprotection as well as dehalogenation to afford 9 in
quantitative yield. Multiple conditions were surveyed in an
attempt to produce the desired â-amino acid ester, but
competing dehalogenation always interfered. Therefore,
despite the high diastereoselectivity of this asymmetric
Michael reaction, this research effort was terminated.
Asymmetric Enamine Reduction Approach to 3. Chiral
reduction of an appropriately functionalized prochiral enam-
ine9 was also investigated as a logical method to access a
single enantiomer of 3. The chemistry employed to ascertain
the feasibility of such an approach is outlined in Scheme 3.
Preparation of the desired prochiral enamine began by
brominating commercially available acetophenone 10 em-
ploying NBS in DMF. Exposure of this to excess sodium
hydride and diethyl carbonate in refluxing toluene afforded
11. Reaction of 11 with excess ammonium acetate in DMF
(5) See for example, (1) Davies, S. G.; Ichihara, O.; Lenior, I.; Walters, I. A..
J. Chem. Soc., Perkin Trans. 1 1994, 1411, (2) Davies, S. G.; Ichihara, O.
Tetrahedron Asymmetry 1991, 183.
(6) For reviews see: (a) Izod, K. Coordination Chemistry ReViews 2002, 227(2),
173. (b) Bruckner, R. Current Organic Chemistry 2001, 5 (6), 679. (c)
Nicolaou, K. C.; Harter, M. W.; Gunzner, J. L.; Nadin, A. Liebigs Annalen/
Recueil 1997, 1283.
(8) Garro-Helion, F.; Merzouk, A.; Guibe, F. J. Org. Chem. 1993, 58, 6109.
(9) Abdel-Magid A. F.; Cohen J. H.; Maryanoff C. A. Current Medicinal
Chemistry 1999, 6, 955.
(7) (S)-R-methylbenzyl allylamine was available and therefore employed to
expedite ascertaining the feasibility of this synthetic route.
52
•
Vol. 8, No. 1, 2004 / Organic Process Research & Development

Scheme 3. Chiral reduction approach to 3
volumes. If full dissolution occurred, the reaction was
allowed to cool to room temperature and any solid formed
analyzed by chiral HPLC. Ten chiral acids in 10 different
solvents were tested. Resolution of (()-3 into (S)-3 by
diastereomeric salt formation was completed with (-)-(1R)-
camphor-10-sulfonic acid in either ethanol or 2-propanol.
The resolution efficiency was 91 and 81% ee, respectively.
Resolution of (() 3 into (R)-3 was consummated by (R)-
mandelic acid in either acetyl acetate or acetonitrile. The
resolution efficiency was 90 and 83% ee, respectively.
Scale-up of these experiments revealed two major draw-
backs. First, the products’ ee values on-scale were substan-
tially lower than those acquired during the screening process.
For example, the ee obtained for the CSA salt of (S)-3 was
53%, compared to 91%. The ee could be increase to g98%
with two additional crystallizations, but the overall yield was
too low. This was due in part to the second major problem.
The product salts were gelatinous solids, making it difficult
to manipulate them. Therefore, this approach toward prepar-
ing enantiomerically enriched 3 was halted.
Scheme 4. Preparation of (()-3
Diastereoselective Imino-Reformatsky. Prior to and
concomitant with the aforementioned research efforts, work
directed toward preparing an analogue of 3 via a novel,
diastereoselective imino-Reformatsky reaction was proving
to be successful. Drawing upon technology developed in-
house,4 the route shown in Scheme 5 successfully prepared
the ethyl ester of 3 as the p-TsOH salt. The sequence entailed
protecting salicylaldehyde 4 as the methoxyethoxymethyl
(MEM) ether. This was accomplished with MEM chloride
affording 14 in 92% isolated yield. Reaction of this with
(S)-(+)-2-phenylglycinol in THF gave imine 15 as a noniso-
lated intermediate. After removal of the water and in vacuo
concentration, dissolution of 15 in NMP and reaction at <-5
°C with an excess of the Reformatsky reagent prepared from
tert-butyl bromoacetate in THF gave rise to the â-amino acid
ester derivative 16. This intermediate was isolated and
characterized in our lab research, but the diastereoselectivity
was not determined until 16 was converted to the desired
(S)-â-amino acid ethyl ester analogue of 3. In this regard,
removal of the phenylglycinol chiral auxiliary was completed
in a two-step procedure. Oxidative cleavage of the hy-
droxymethyl moiety with lead tetraacetate in methanol
afforded imine 17 after neutralization with NaOH and
concentration. Treatment of this crude oil with a slight excess
of p-TsOH‚H2O in refluxing ethanol effected imine hydroly-
sis, deprotection, and transesterification. This provided 3 as
the ethyl ester and p-TsOH salt in 65% overall from 4. An
ee > 99% was obtained as measured by chiral HPLC
analysis.
at room temperature followed by acetylation with acetic
anhydride and pyridine provided the requisite enamine 12
as a white solid in 70% overall yield from 10.
Unfortunately asymmetric reduction of 12 was only
moderately successful. Six commercially available catalysts
and multiple conditions were investigated. The best result
obtained was 70% ee using catalytic (+)-1,2-bis-((2S,5S)-
2,5-diethylphospholano)benzene(cyclooctadiene) rhodium (I)
trifluoromethanesulfonate in dichloromethane under 5 atm
of H2. Also investigated were other analogues of 12 wherein
the ester and hydroxyl-protecting groups were either removed
and/or changed, but no improvement was observed. There-
fore, on the basis of the moderate success of this approach,
this research effort was also discontinued.
Resolution via Diastereomeric Salt Formation. Resolu-
tion of (()-3 via diastereomeric salt formation was inves-
tigated. This would represent a viable method of obtaining
a single enantiomer of 3 if preparation of the racemate could
be implemented on-scale. This was accomplished by em-
ploying methodology developed in-house11 as shown in
Scheme 4. Starting with the aforementioned salicylaldehyde
4, Perkin reaction12 in refluxing acetic anhydride in the
presence of triethylamine provided coumarin 13 in 70% yield.
Exposure of 13 to ammonia at 80 °C in a pressure reactor
for 24 h accomplished the desired Michael reaction, giving
rise to the â-amino acid amide analogue of 3. After
evaporation of the excess ammonia, acidic ethanolysis
afforded (()-3 as the HCl salt in approximately 60% yield.
Experiments to discern the best conditions for resolution
of (() 3 were accomplished by subjecting the free base to a
screen of commercially available chiral acids in a number
of solvents. The substrate and chiral acid (1 equiv) were
added to the appropriate solvent (5 vols) and heated to reflux.
If dissolution was incomplete, the solvent volume was
increased in 5-volume portions up to a maximum of 50
With the success of this technology and the demand for
clinical supplies of 1, this chemistry was used as the enabling
route. Indeed, 7.7 kg of (S)-3 as the ethyl ester, p-TsOH salt
were prepared from salicylaldehyde 4 via the chemistry
shown in Scheme 5 in 57% overall yield with an ee > 99%.
This was a remarkable achievement. Nevertheless, practicing
this process on-scale accentuated several development op-
portunities.
(10) See Scheffer, J. R. Can. J. Chem. 2001, 79, 349 and references therein.
(11) Rogers, T. E.; Ruminski, P. G. US Patent 6013651, 2000.
(12) Rosen, T. Compr. Org. Synth. 1991, 2, 395.
Vol. 8, No. 1, 2004 / Organic Process Research & Development
•
53

Scheme 5. Preparation of (S)-3 via a diastereoselective Reformatsky reaction
Scheme 6. Feasibility studying employing MOM protection
Process Development
Although the aforementioned enabling methodology yielded
excellent quality product in very good yield, the following
issues were identified. The process (1) employed MEM
chloride, a hazardous reagent, (2) involved the isolation of
an unstable imine intermediate 15, (3) engaged a highly
energetic imino-Reformatsky reaction, (4) utilized Pb(OAc)4,
and (5) displayed inconsistency in the purification and
isolation of 3 as the ethyl ester, and p-TSA salt. Furthermore
the enabling technology involved twelve phase separations,
six distillations, three isolations, two MgSO4 dehydrations,
one crystallization, and two drying steps. These features tied
closely with issues such as safety, engineering, cost of goods,
resource allocation, raw materials sourcing as well as the
business strategy were the emphasis behind further develop-
ment of the imino-Reformatsky route to (S)-3.
The MEM-Protecting Group: Essential for Success.
The first stage in the enabling process to supply (S)-â-amino
acid ethyl ester 3 for preclinical development entailed the
conversion of 4 to the corresponding MEM protected
aldehyde 14. An alternative to this protecting group was
desired because of the toxicity and expense of MEM chloride.
It was recognized, however, that the high diastereoselectivity
(>99% de) and yield of the subsequent conversion of imine
15 to (S)-3 might be adversely affected by this change.4,13
The first alternate protecting group investigated was
methoxymethyl (MOM). This group might provide the
aforementioned complexation site and may have the added
advantage in that it could be introduced by using dimethoxy-
methane, avoiding the use of the carcinogenic reagent
chloromethyl methyl ether. For these feasibility studies,
however, MOM chloride was employed to expedite evalu-
ation of this idea. Accordingly, 4 was converted to the
corresponding MOM ether 18 in 74% isolated yield and
transformed to imine 19 in near quantitative yield. Reaction
of 19 with the Reformatsky reagent derived from tert-butyl
bromoacetate unfortunately afforded 20 in only 33% overall
yield from 4 after column chromatography (Scheme 6).
Attempts to optimize the desired reaction were unsuccessful;
therefore, results obtained from the MOM protecting ana-
logue 18 could not compete with those obtained from MEM
ether 14.
(13) Two equivalents of the Reformatsky were essential for complete reaction.
It is envisaged that at least one of the two oxygen atoms of the MEM moiety
assists in the observed diastereoselectivity by complexation with one
equivalent of reagent.
54
•
Vol. 8, No. 1, 2004 / Organic Process Research & Development

Figure 1. FT-IR data for the hydrolysis of MEM chloride with DMF/H2O.
Other protecting groups were explored and found to fail
in comparison. For example, base labile groups such as
acetate did not hold up to the downstream chemistry. Base
stable groups such as pivaloyl resulted in low yields and
poor diastereoselectivity. Groups requiring hydrogenation for
removal such as benzyl or benyloxycarbonyl resulted in
dehalogenation of the aromatic ring. Thus, it was established
that the MEM protecting group was required for successful
completion of this chemistry.
Recognition of this fact focused our research toward full
development of the safety protocols necessary for use MEM
chloride on-scale as well as the destruction of excess reagent
postreaction. Accordingly, the rate constant for hydrolysis
of MEM chloride in DMF/H2O was measured by FT-IR and
1H NMR analysis. In the FTIR experiment, water was added
to a solution of MEM chloride in DMF and the hydrolysis
monitored using the ReactIR.14 Spectra were recorded every
1.4 s for approximately 10 min. A ConcIRT analysis15 of
the data was performed to extract the individual components.
A plot of the data is shown in Figure 1.
) 1 × 10-2 M-1 s-1. The data and fit are shown in Figure
2.
The rate of MEM chloride hydrolysis as measured by 1H
NMR analysis started by adding it to a small vial containing
DMF-d7 and mesitylene as an internal standard. The solution
was transferred to an NMR tube and the 1H NMR spectrum
acquired. D2O was added and the tube inverted to allow for
mixing. Fifteen spectra (eight acquisitions each) were
sequentially recorded. The time from D2O addition to the
first acquisition was approximately 1 min. Twenty-five
seconds were required for eight acquisitions. The first
spectrum was complete within 85 s, and 3% of the MEM
chloride remained. The second-order rate constant was
calculated to be k ) 1 × 10-2 M-1 s-1 using the integrated
rate equation shown in eq 1.
k )
[H2Oinitial][MEMClfinal
]
1
ln
[MEMClinitial] - [H2Ofinal
]
t([MEMClinitial] - [H2Oinitial])
(1)
The initial MEM chloride and water concentrations are
known from the amounts initially charged. The final con-
centrations and final time were estimated from Figure 1.
Essentially all of the MEM chloride was destroyed within 8
min. The data were fit using Scientist,16 a modeling program,
and the second-order rate constant17 was calculated to be k
Thus, it was shown by both spectroscopic techniques that
the second-order rate constant for hydrolysis of MEM
chloride is k ) 1 × 10-2 M-1 s-1. The experimental result,
therefore, suggest that essentially all remaining MEM
chloride would be quenched by water if the mixture was
allowed to stir at ambient temperature for approximately 1
h.
This result was confirmed in the laboratory by direct
quantitative GC analysis of the MEM chloride in DMF and
water solutions. For example, after 83 min the MEM chloride
was not destroyed by DMF alone. However, the addition of
water to the DMF greatly increases the rate at which MEM
chloride is degraded. In a 1:1 mixture of DMF/water, MEM
chloride could not be detected after 21 min.
(14) All in situ infrared spectra were collected by means of an ASI ReactIR
1000 equipped with a 5/8-in. DiComp ATR probe. ASI Applied Systems,
Inc., Millersville, MD 21108.
(15) ConcIRT is an algorithm used by ASI ReactIR software to pull individual
components out of a complex mixture.
(16) Scientist for Windows, Version 2.01, MicroMath Scientific Software, P.O.
Box 21550, Salt Lake City, UT 84121, (801) 943-0290.
(17) R-Chloroethers have previously been shown to hydrolyze by an SN1
mechanism.18 Since the mechanism has not been rigorously established under
these conditions, we chose to use a second-order model which led to a more
conservative estimate of the MEM chloride concentration.
Vol. 8, No. 1, 2004 / Organic Process Research & Development
•
55

Figure 2. Mathematical fit of the MEM chloride hydrolysis FT-IR data.
However, additional evidence was obtained by analyzing
isolated samples of 14 for MEM chloride. When the reaction
mixture was stirred with water for 1 h or more, no MEM
chloride was detected by gas chromatographic analysis. On
the other hand, traces for MEM chloride in 14 were detected
if the reaction mixture quench was allowed to stir for only
10 min.
procedure seemed to perform effectively in all laboratory
runs, but during the first piloting run the total batch was
lost due to decomposition. Accordingly, a study was initiated
to solve this heretofore unknown problem.
The laboratory work began with attempts to repeat the
decomposition observed in the pilot plant. Employing
supplies of (S)-phenylglycinol prepared from two third-party
manufacturers (TPM), we found that one source did not give
rise to decomposition for a period of up to 6 days. In fact,
accelerated decomposition studies on neat 15 revealed no
change at temperatures as high as 90 °C for periods of 1-15
The mode of quench addition also affected the MEM
chloride levels entrapped in 14. No MEM chloride was
detected if the reaction mixture was charged to an excess of
water; however, if water was added to the reaction mixture,
MEM chloride could be detected in isolated 14 even after
stirring for 1 h. It was found that the addition of water to
the reaction immediately precipitated small amounts of 14
enriched in MEM chloride. Therefore, the practiced pilot-
plant procedure for the conversion of salicylaldehyde 4 to
its MEM protected analogue 14 was to add powered
potassium carbonate (1.06 equiv) to a solution of 4 in DMF.
MEM chloride (1.05 equiv) was added via pressure canister,
while maintaining a temperature of 20-25 °C. After
complete addition, the reaction was stirred for 3 h at ambient
temperature. The product mixture was charged to a reactor
containing water (4.6 L/L DMF). The slurry was stirred for
30-60 min and the white solid (mp 32.5-33 °C) isolated
by either centrifugation or by pressure filter. The wet cake
was washed with water and dried in vacuo at ambient
temperature to afford 14 in 92% yield.
Imine 15, an Unstable Intermediate. In the enabling
technology imine 15 was prepared by reaction of 14 with
(S)-phenylglycinol in THF. After the desired conversion was
complete, the reaction mixture was concentrated to dryness
under reduced pressure. The distillation served mainly to the
remove the water generated during imine formation to levels
of less than 0.1 wt. %. The product concentrate was then
diluted with NMP and reacted with the Reformatsky reagent
prepared form tert-butyl bromoacetate to provide 16. This
1
h as determined by H NMR spectrum analysis. It was not
until 15 was exposed to 90 °C for 65 h that signs of
decomposition began to emerge.
On the other hand, the decomposition phenomenon could
be repeated in the lab when employing the same phenylg-
lycinol employed in the pilot run, reagent sourced from a
second TPM. Surprisingly, in contrast to the aforementioned
phenylglycinol, this lot gave rise to decomposition in the
THF solution itself, before concentration! Therefore, on the
basis of these preliminary laboratory findings, it was evident
that the sources of phenylglycinol were behaving differently.
Our focus turned toward finding the cause of the
1
decomposition. A H NMR study was completed in which
the two different sources of phenylglycinol were employed
and compared directly. An empirical observation was made.
The different sources of phenylglycinol gave rise to vastly
different rates of imine formation. For example, the reaction
was complete within 30 min using the TPM’s phenylglycinol.
Conversely, it took the same reaction approximately 6 h to
go to completion when employing the lab lot of phenylg-
lycinol.
This prompted us to hypothesize that an agent in the
TPM’s phenylglycinol, perhaps an acid or a metal, caused
the differences in the observed rates of reaction and
decomposition. Unfortunately, no definitive cause could be
56
•
Vol. 8, No. 1, 2004 / Organic Process Research & Development

Table 1. Summary of thermokinetic data for the conversion
of 14 to 15 and 15 to 16
found within the accelerated timeline of the project. In fact,
on the basis of all analytical information received, the two
lots of phenylglycinol appeared compositionally identical.
Titration, trace metals analysis, elemental analysis, and
extensive liquid chromatography/mass spectral analysis of
the two lots could not resolve the difference. Thinking that
the method by which the phenylglycinol was manufactured
may shed some light on a possible difference, a request was
made of both manufacturers to provide a process description.
Unfortunately, the manufacturer of the phenylglycinol that
did not promote decomposition would not release their
preparative method because it was considered a trade secret.
Nevertheless, a robust method to effect the desired
reaction and avoid the aforementioned decomposition was
found. The answer was NMP, the solvent used in the
forthcoming Reformatsky reaction. In this solvent, the rate
of formationof 15 was slower (20-30 h), but acceptable,
and the aforementioned decomposition was eliminated. For
example, the lot of phenylglycinol that caused decomposition
in THF was applied using NMP. No decomposition of imine
15 was observed for 8 days, even at elevated temperatures.
NMP had an additional advantage in that its use eliminated
one distillation used in the enabling process.
Having determined that NMP would resolve the decom-
position problem and simplify the process, our studies turned
toward dehydrating the product solution. Spiking studies on
the effect of water in the forthcoming Reformatsky reaction
revealed that the presence of 1 mol equiv of water, the
theoretical amount, led to a 75% decrease in product yield,
whereas, only 5% loss was observed when the equivalent of
water present was e0.4 with respect to 15. These investiga-
tions also revealed that the type of dehydrating agent was
important. The use of an inorganic salt or other solid
dehydrating agent in NMP always led to filtration problems.
However, a column of 4 Å molecular sieves proved suc-
cessful. Studies found that the water level could be decreased
to less than 0.3 wt % by simply passing the NMP solution
through a column of untreated sieves.
As it turns out, this method of dehydrating the reaction
mixture was very robust. The technology was employed
successfully on 12 different occasions and at three different
pilot plants. Dehydration was effected either by employing
a column of sieves or a Sparkler filter filled with sieves. In
a properly designed column system the level of water in the
reaction mixture was diminished to well below the effective
level by passing the solution through only one time. In the
Sparkler filter system, the solution was recirculated through
the unit until the desired level of dehydration was achieved.
Thus, the general procedure employed for preparation of an
NMP solution of 15 entailed dissolution of aldehyde 14 into
NMP. Phenylglycinol (1.005 equiv) was charged in one
portion and the reaction mixture agitated for 20-30 h. The
product solution was then pumped through a 15-cm column
of untreated 8-12 mesh, 4 Å molecular sieves and used
without further purification.
total molar adiabatic
Cp initial Cp final heat
heat
temp
[°C]
step
[J/kg/K] [J/kg/K] [kJ] [kJ/mol]
14 f 15
“BrZnCH2CO2 Bu” 1601
15 f 16
quench
1806
1713
1728 149
1600
2300
11
37
205
56
20
173
13
t
950
1600
17
29
183
1R
Pfizer on a number of occasions.4 This experience led us to
appreciate the energy of this conversion, and therefore,
studies directed toward understanding the thermo- and
reaction kinetics of the conversion of 15 to 16 were
completed. Such data would help in designing a safe and
effective protocol for practicing this chemistry on-scale.
The study was performed employing a Mettler RC1 SV01
1L reaction Calorimeter equipped with a ReactIR13 to obtain
FTIR-derived kinetic information. Spectra were recorded
every 5 min for approximately 500 min. A ConcIRT
analysis14 of the data was performed in order to extract the
individual components. A probe for GC-MS analysis of the
reactor headspace was also installed so that the rate of
ethylene generation during zinc activation with 1,2-dibro-
moethane could be monitored and quantified.
Ethylene gas was detected during reflux only. The
maximum rate of ethylene formed was measured at 3% of
the ethylene volume per minute, assuming complete conver-
sion. Extrapolation of this information indicated that a pilot
plant’s scrubber or flume system or both would adequately
handle this rate of ethylene release.
A summary of the thermokinetic data obtained for the
conversion of 14 to 15 and then 15 to 16 is given in Table
1. The information indicates that the formation of the reagent
itself is the most energetic component of the reaction
sequence, giving rise to a calculated ∆Hrxn ) -205 kJ/mol
tert-butyl bromoacetate and an adiabatic temperature rise of
173 °C.
The data were acquired by the addition of tert-butyl
bromoacetate over 2 h to a suspension of activated zinc in
THF at 50 °C. The heat generation profile for this experiment
is shown in Figure 3. Note a significant, but not surprising,
observation. A 15-min induction period was observed; thus,
important information for controlling the rate of tert-butyl
bromoacetate addition and predicting reaction behavior,
particularly on-scale, was obtained.
After the reaction was complete, the THF used to generate
t
BrZnCH2CO2 Bu was partially distilled off. It was important
to monitor the amount of THF removed because eliminating
too much led to precipitation of the THF solvate of the
Reformatsky reagent. On-scale this could result in the agitator
freezing up. Studies in this regard found that optimum results
were obtained when 40 wt % of the original THF charge
was removed. In cases where 45-50% of the THF was
removed, the agitator became sluggish and started to stall.
The thermal data obtained for the desired Reformatsky
reaction is shown in Figure 4. In this experiment, 88% of
imine 15 (1.0 equiv) in NMP was added via pump over a
1.5-h period to a -10 °C NMP/THF solution. The remaining
The Diastereoselective Imino-Reformatsky Reaction.
The preparation of chiral â-amino acid esters via an imino-
Reformatsky reaction has been employed successfully by
Vol. 8, No. 1, 2004 / Organic Process Research & Development
•
57

t
Figure 3. Preparation of “BrZnCH2CO2 Bu”. Heat released and accumulation data, watts vs time.
Figure 4. Conversion of 15 to 16. Heat released and accumulation data, watts vs time.
12% was added over a 6-min period using an addition funnel.
The overall adiabatic temperature rise of the Reformatsky
reaction was calculated to be 13 °C, with a molar heat of
reaction of approximately -56 kJ/mol of 14, assuming 100%
conversion to 16. The heat flow data indicated that the
conversion of 15 to 16 is dose controlled. Reaction time
based on the FTIR data indicated that the reaction was nearly
dosed controlled and complete once all of 15 was added.
The thermokinetic data for the quench reaction (1:1
aqueous NH4Cl/concentrated HCl) revealed an adiabatic
temperature rise of 16 °C and a molar heat of reaction of
183 kJ/mol HCl. On the basis of the recorded heat profile,
the reaction time was dosed controlled and nearly complete
within the first few percent of the quench solution addition.
To supplement the calorimetry study, temperature experi-
ments were completed to ascertain this effect on the
conversion of 15 to 16. This revealed that 16 was obtained
in near quantitative yield when the substrate was combined
with the reagent in such a manner that the internal temper-
ature remained between -10-0 °C. If the reaction temper-
ature was allowed to remain within a temperature range of
0-5 °C, a small decrease in yield to approximately 95%
was observed. A loss of approximately 40% was observed
if the reaction temperature reached 20 °C. Interestingly, no
loss in diastereoselectivity19 was observed within this tem-
perature range, a true testimony to the robustness of this
chemistry.
From the aforementioned data, the following standard
protocol for large-scale conversion of 15 to 16 was developed
and practiced. Zinc (4.0 equiv/mol 15) was charged to a
reactor, followed by THF and 1,2-dibromoethane (0.077
equiv/mol 15). The contents were heated to reflux (∼66 °C)
and held for at least 1 h. The mixture was cooled and tert-
butyl bromoacetate (3.83 equiv/mol 15) added in ap-
proximately five equal aliquots within a temperature range
(18) (a) Jones, T. C.; Thornton, E. R. J. Am. Chem. Soc. 1967, 89, 4863. (b)
Tou, J. C., Kallos, G. J. Anal. Chem. 1974, 46, 1866. (c) Richard, J. P.;
Amyes, T. L.; Rice, D. J. J. Am. Chem. Soc. 1993, 115, 2523.
(19) Diastereoselectivity measured by chiral HPLC analysis of the product
mixture and by analysis of the product derived from conversion to (()-3
ethyl ester, p-TsOH salt.
58
•
Vol. 8, No. 1, 2004 / Organic Process Research & Development

of 48-50 °C. An exotherm of 5-7 °C was generally
observed with each aliquot. The contents were then held at
50 ( 5 °C for at least 1 h. The contents of the reactor were
cooled to 0 ( 5 °C, and vacuum was applied to the system.
Approximately 40% of the THF was removed by vacuum
distillation. Following this, NMP was charged, and the
contents were cooled to -15 ( 5 °C.
The NMP solution of imine 15 in a second reactor was
added to the Reformatsky reagent while maintaining the
batch temperature below -5 °C. The contents were held for
at least 3 h at e0 °C and then filtered to remove the excess
zinc. The zinc complex of 16 in NMP was quenched by the
addition of an ammonium chloride and hydrochloric acid
solution while maintaining the temperature at e10 °C. The
contents of were warmed to 20 ( 5 °C and MTBE added to
extract 16. Following a series of aqueous washes, the MTBE
was removed under reduced pressure until minimum stir
volume was reached. Methanol was charged and this solution
used without further purification.
Replacing Pb(OAc)4 with NaIO4. The enabling proce-
dure for the conversion of the imino-Reformatsky reaction
product 16 to benzyl imine 17 entailed treating a methanol
solution of 16 with a molar equivalent of Pb(OAc)4 at 0-5
°C. Sodium hydroxide was then added to the reaction mixture
and the methanol removed by vacuum distillation. The
product 17 was extracted from the mixture into ethyl acetate,
dried, concentrated to an oil, and taken into the next step
without further purification. Although this procedure pro-
vided quality product in good yield, the waste-handling issues
associated with the use of lead gave rise to an effort to
replace this oxidant. A number of alternatives to Pb(OAc)4
were therefore examined. Only sodium periodate in methanol
gave the desired product. Noteworthy was the fact that
methanol was required for successful conversion. A slow
reaction rate was observed in ethanol. It is believed that this
subtle difference is related to the relative solubility of
periodate in ethanol versus that in methanol.
The conversion of 16 to 17 was monitored carefully by
both
1H NMR and LC analysis to determine which formal-
dehyde-sequestering agent would be best suited for scale-
up. This revealed no 21 at any time with methylamine
present, but the oxazolidine content increased to a maximum
(approximately 16% by HPLC area) and then decreased to
near zero in the presence of acetic acid. Comparative analysis
of the two methods further revealed that a higher-quality
product and more efficient workup were obtained when using
methylamine. Thus, methylamine was used on-scale.
The following general procedure for preparation of benzyl
imine 17 was employed in the pilot plant. To a vessel were
charged NaIO4 (2.0 equiv), methanol, and methylamine (1.2
equiv). The contents were heated to 30 °C. The methanol
solution of 16 was then added to the reactor containing the
NaIO4 mixture and held for 12 h. Upon complete reaction,
ethyl acetate was charged and this slurry filtered through a
precoated filter system. The filter cake was washed with ethyl
acetate, and the combined organic layers were distilled in
vacuo until the concentrate contained e2.5% methanol by
GC analysis. This was important to maximize product yield
in the following aqueous treatments. The distillation was
repeated if necessary. The concentrate was then taken back
up in ethyl acetate and washed sequentially with 20%
aqueous sodium thiosulfate and brine. Vacuum distillation
afforded the crude product, which was taken up into ethanol
and used without further purification.
Purification and Isolation of (S)-3 Ethyl Ester, p-TSA
Salt. The enabling route for conversion to (S)-3 entailed
exposure of an ethanol solution of 17 to a slight excess of
TsOH‚H2O in refluxing ethanol. The ethanol was then
removed in vacuo and the residue taken up into tetrahydro-
furan. This was reconcentrated and taken up again in THF.
After heating to reflux, hexane was added and the mixture
cooled to allow crystallization. The product was filtered and
the cake washed with acetone. Thus, four solvents and two
distillations were employed to carry out this operation.
Additionally, it was observed that during vacuum drying of
(S)-3 ethyl ester, p-TsOH salt, a color change from white to
brown took place and the product solid appeared to be
nonhomogeneous. Recrystallization from EtOH/EtOAc re-
moved the color and gave rise to a homogeneous white solid.
Analysis of the mother liquor revealed trace amounts of
benzylaldehyde and p-TsOH. Interestingly it was found
through spiking studies that neither of these contaminants
effected on their own the color change but together created
the browning phenomenon. The origin of this effect was not
investigated further.
Although reaction of 16 with NaIO4 in methanol gave
the desired product 17, this reaction also afforded oxazolidine
21 as 23% of the product mixture. Evidently the liberated
formaldehyde was reacting with substrate.
Refinement of the aforementioned isolation of (S)-3 ethyl
ester, p-TsOH salt was studied. Precipitations directly from
EtOH using an antisolvent were investigated, but scale-up
results were not consistent. The use of a combination of
MTBE/EtOH was also investigated, and looked promising
since precipitation of a white solid took place at reflux, but
the product was identified as the tert-butyl ester analogue
of 3. Although this was not a bad result, the application of
the tert-butyl ester of 3 in the synthesis of 1 led to lower
yields in the ensuing chemistry.
Attempts to minimize the formation of 21 met with limited
success at first. Factors such as solvent, periodate counterion,
equivalents of periodate, concentration, temperature, and
order of addition were investigated, all giving only incre-
mental improvements. The problem of oxazolidine 21
formation was solved, however, when we looked into
sequestering the formaldehyde formed. A number of acids
and bases were scanned, and it was established that the
formation of 21 was terminated with either acetic acid or
methylamine.
Vol. 8, No. 1, 2004 / Organic Process Research & Development
•
59

Therefore, despite extensive investigation, our efforts
turned toward simplifying the enabling procedure. In this
regard, three significant observations were noted. First, the
yield of (S)-3 ethyl ester, p-TsOH salt was highly dependent
upon the ethanol content of the crystallization solvent system.
For example, if the ethanol content was g5 wt % before
antisolvent addition, a yield loss of >10% resulted. Second,
if the product mixture was concentrated too much, an
extremely viscous mixture formed that stopped agitation!
And third, high temperature (65 °C) addition of the antisol-
vent afforded better product quality and color.
These procedural modifications led to the following pilot-
plant procedure. The ethanol of solution of 17 and p-TsOH‚
H2O (1.3 mol equiv) were charged to a steam-jacketed vessel.
The solution was heated to reflux, maintained at this
temperature for 8 h, and then cooled to 10 °C. The contents
were distilled under vacuum until either the minimum stir
volume of the reactor was reached or the batch temperature
reached 40 °C. THF was charged, and the contents were
again concentrated under vacuum until the batch temperature
reached 40 °C. The contents were diluted again with THF
and heated to 65 °C, and heptane was added while maintain-
ing temperature. During heptane addition, (S)-3 ethyl ester,
p-TsOH salt precipitates out of solution. The contents were
cooled to 5 °C and held at this temperature for 2 h. The
solids were filtered and washed with cold acetone three times
to yield (S)-3 ethyl ester, p-TsOH salt in 48-57% overall
yield from salicylaldehyde 4. Analysis of the free base of
the product by chiral HPLC measured an enantiomer excess
of >99.7%.
adjusted yield). However, the fourth batch gave the desired
product in only 32% overall yield. Additionally troubling
was the average assay for all product preparedsonly 77%.
The pilot-plant for Campaign 2 was chartered to produce
approximately 350 kg of (S)-3 ethyl ester, p-TsOH salt. This
facility used three 750-gal glass-lined reactors, one 100-gal
stainless steel reactor, a 750-gal glass-lined receiver, a
molecular sieve column, a bag filter system, and a Nustche
filter. The aforementioned throughput restrictions within this
equipment and at this scale necessitated a tiered approach
to making product. First 240 kg of 14 was converted to the
Reformatsky product 16. This batch was then split in two
and processed to product separately. The split lot was held
in cold storage until it was ready for use. This pattern was
repeated two more times. All of this work was completed
within 12 weeks, giving rise to 350 kg of (S)-3 ethyl ester,
p-TsOH salt with an average assay of 90 wt %. Although
this result was superior to that obtained for Campaign 1, the
assay was still somewhat lower than anticipated.
Research directed toward ascertaining the reason(s) behind
the lower than expected yields and assays were completed.
Postprocess analysis of reaction stream samples revealed an
accentuated level of an impurity in the Reformatsky product
mixture. These studies also concluded that this contaminant
carried through into product. This impurity was identified
as the amino acid itself, (S)-3. Further tests revealed that
the amount of (S)-3 present was proportional to the time the
Reformatsky product solution was held before quench. For
example, if the solution of 16 was held for 4.8 h before
quench, 17 area % of (S)-3 was present in the product
mixture. If this material was held for 8.6 h, the level of this
impurity increased to 32 area %. These data were compared
to results from laboratory trials and verified. Thus, prior to
quench, the tert-butyl ester of 16 was converting to the
corresponding acid (or precursor thereof) and carried through
to the isolation of the ethyl ester, p-TsOH salt of (S)-3, a
direct result of the tiered approach to processing this much
material.
Pilot-Plant Campaigns
The technology shown in Scheme 5 was transferred
directly to toll manufacturers. The procedure was practiced
on three different occasions. Learnings from these campaigns
were then incorporated into a potential commercial process.
Planned was the preparation of approximately 600 kg of (S)-3
ethyl ester, p-TsOH salt. Additionally, two aspects of the
process development effort were incorporated into the
operating instructions at this juncturesreplacing Pb(OAc)4
with NaIO4 and practicing the revised (S)-3 ethyl ester,
p-TsOH salt crystallization protocol.
In the course of transferring the technology to two
potential commercial manufacturing sites, significant engi-
neering limitations were realized. Centrifugation was not
available at either facility. Therefore, isolation of product
had to be carried out using some sort of pressure filtration
system. Second, a bottleneck was recognized when filtering
the NMP solution of the Reformatsky product 16 before
isolation. Regardless, appropriate strategies were devised, and
equipment was installed or modified at each facility.
The equipment employed for Campaign 1 utilized two
200-gal glass-lined reactors, one 200-gal stainless steel
reactor, two 100-gal glass-lined reactors, a 36-in. seven-plate
Sparkler stainless steel filter (for the molecular sieves) and
a 24-in. glass-lined pressure filter. The first three of four
batches at this location processed fairly uneventfully. Em-
ploying 87 kg of 4 in each run, 241 kg of (S)-3 ethyl ester,
p-TsOH salt was produced in 60% overall yield (assay-
Fortunately, a simple rework protocol for poor-quality
material was devised that entailed reslurry of the product in
acetone. This procedure provided product with an average
assay of 95 wt %. Approximately 95% of the available
product in this material was recovered. The poorer quality
batches from Campaigns 1 and 2 were reworked in this
manner to provide in total 591 kg of acceptable product.
Conclusions
Four process research investigations directed toward
discerning a scalable, enantioselective method for preparing
a key intermediate to the Rvâ3 integrin antagonist 1, (S)-â-
amino acid ester 3, were completed. A synthetic route
developed in-house employing a novel, diastereoselective
imino-Reformatsky reaction was selected for development.
Process development of this chemistry into a potential
commercial process addressed several technical issues.
Discovered was the necessity of employing the MEM-
protected salicylaldehyde 15 in the Reformatsky reaction.
This maintained high yield and diastereoselectivity. The
reaction kinetics of MEM chloride hydrolysis were utilized
60
•
Vol. 8, No. 1, 2004 / Organic Process Research & Development

to determine an amenable quenching procedure for the safe
and effective application of this reagent on-scale. The
efficient formation of the (S)-phenylglycinol imine 15 in
NMP and the dehydration of this product solution employing
molecular sieves were described. A calorimetry study on the
Reformatsky reaction and the application of this thermal and
kinetic data to the development of an effective, large-scale
procedure were completed. Pb(OAc)4 used in the enabling
process was replaced with NaIO4 in methanol. The use of
methylamine to sequester competing oxazolidine formation
was successful and employed in the commercial process.
Finally, an isolation and purification protocol for (S)-3 ethyl
ester, p-TsOH salt was developed.
David Addison, David Banaszak, Angela Heseman, Gary
Hosna, Vasilios (Bill) Iskos, Kim Kolbert, Randy Meyer,
John Schroeder, and Bruce Sorensen. David Brown, Stephanie
Camden, Mark LaPack, Jung-Min Park, Erwin Irdam, Shan
Lin, Jeff Wysocki, and Tony Thompson provided safety data
and/or consultation. Teddy Albano, Jan Gard, Mary Gasper,
Mason Hughes, Ed Sheldon, Don Stano, and Wei Kong are
recognized for their invaluable analytical work. Additionally,
we thank Thomas Muller of Carbogen, Aarau, Switzerland,
for his early contributions and helpful discussions, Ray Frame
for help with the 22 L work, and Thomas Fevig for advice
on the manuscript. We thank Joe Timko for helpful sugges-
tions. Kathy Barton, Jim Behling, John Dygos, Jim Miles,
and Sean Nugent are acknowledged for their assistance. Russ
Buchman and Virgil Coker are recognized for their efforts
in sourcing this project. Finally, due to the depth of this
project, many individuals cannot be addressed personally.
To these we express our gratitude.
The potential commercial process was practiced. Yields
and assays were lower than expected. It was found that the
Reformatsky product 16 degraded during processing. Nev-
ertheless, after rework, 591 kg of acceptable product was
successfully prepared.
Acknowledgment
Supporting Information Available
We thank the many operators and support personnel at
Synthetech, Inc., Portland, OR, RSA Corp., Banbury CT,
Pharmacia’s Skokie, IL, Pilot Plant, and Pharmacia’s St.
Louis, MO, Queeny Pilot Plant. We especially recognize Tim
Fitzpatrick, David Short, Jim Johnston, Jan Anthony, and
Ron Scott. We thank Frank Muellner for managing multiple
aspects of the research timeline with the business timelines.
We also gratefully acknowledge engineering support from
Experimental procedures for the preparation of most
compounds (PDF). This material is available free of charge
via the Internet at http://pubs.acs.org.
Received for review July 8, 2003.
OP034094J
Vol. 8, No. 1, 2004 / Organic Process Research & Development
•
61