The development of a manufacturing route for the GPIIb/IIIa receptor antagonist SB-214857-A. Part 1: Synthesis of the key intermediate 2,3,4,5-tetrahydro-4-methyl-3-oxo-1H-1,4-benzodiazepine-2-acetic acid methyl ester, SB-235349

Article abstract of DOI:10.1021/op034024c

The development of an efficient manufacturing route to 2,3,4,5-tetrahydro-4-methyl-3-Oxo-1H-1,4-benzodiazepine-2-acetic acid methyl ester SB-235349, a key intermediate in the synthesis of lotrafiban is described. The synthesis starts with 2-nitrobenzyl alcohol which is mesylated, reacted with methylamine and then dimethylacetylene dicarboxylate followed by reduction of the nitro group. Treatment of the resultant aniline with acid gives an intermediate quinazoline which rearranges on treatment with base to give a 1,4-benzodiazapine. Reduction of the exocyclic double bond affords SB-235349. The process can be run without isolation of any of the intermediates and has been used to prepare several tons of SB-235349.

Full text of DOI:10.1021/op034024c

Organic Process Research & Development 2003, 7, 655−662
The Development of a Manufacturing Route for the GPIIb/IIIa Receptor
Antagonist SB-214857-A. Part 1: Synthesis of the Key Intermediate
2,3,4,5-Tetrahydro-4-methyl-3-oxo-1H-1,4-benzodiazepine-2-acetic Acid Methyl
Ester, SB-235349
Ian P. Andrews, Richard J. Atkins, Gary F. Breen, John S. Carey, Michael A. Forth, David O. Morgan,* Amin Shamji,
Andrew C. Share, Stephen A. C. Smith, Timothy C. Walsgrove, and Andrew S. Wells†
Synthetic Chemistry, GlaxoSmithKline Pharmaceuticals, Old Powder Mills, Nr. Leigh, Tonbridge, Kent TN11 9AN, UK
Abstract:
All of the synthetic strategies adopted by Synthetic
Chemistry during the lotrafiban 1 supply campaigns relied
on a common intermediate having the basic racemic 1,4-
benzodiazepine skeleton 2, SB-235349 (Scheme 1). The
7-bipiperidyl amide was to be introduced via aminocarbo-
nylation of a suitably 7-substituted 1,4-benzodiazepine with
a mono-protected 4,4-bipiperidine.7 The chirality could be
introduced via a late-stage resolution or, more efficiently,
by introducing the chirality directly after the synthesis of
SB-235349, 2. We report here how an efficient and eco-
nomical synthesis of SB-235349, 2 was identified, developed,
and refined to produce the target molecule using seven
reactions in a “one pot” process (intermediates not isolated)
with an overall yield of 70%.8
The development of an efficient manufacturing route to 2,3,4,5-
tetrahydro-4-methyl-3-oxo-1H-1,4-benzodiazepine-2-acetic acid
methyl ester SB-235349, a key intermediate in the synthesis of
lotrafiban is described. The synthesis starts with 2-nitrobenzyl
alcohol which is mesylated, reacted with methylamine and then
dimethylacetylene dicarboxylate followed by reduction of the
nitro group. Treatment of the resultant aniline with acid gives
an intermediate quinazoline which rearranges on treatment with
base to give a 1,4-benzodiazapine. Reduction of the exocyclic
double bond affords SB-235349. The process can be run without
isolation of any of the intermediates and has been used to
prepare several tons of SB-235349.
Introduction
Results and Discussion
Lotrafiban, SB-214857-A, 1, acts as a potent nonpeptidic
glycoprotein IIb/IIIa receptor antagonist to prevent platelet
aggregation and thrombus formation.1,2 The molecule has the
(S)-stereochemistry, and only this enantiomer is active as a
GPIIb/IIIa antagonist. To support the clinical development
programme,3 the Synthetic Chemistry group at Tonbridge
had to prepare multi-ton quantities of homochiral SB-214857-
A. A number of routes to SB-214857-A, 1, were known
starting from (S)-aspartic acid,4-6 but for various reasons,
none were suitable for scale-up into an efficient, safe, and
economical manufacturing process.
Initial Synthesis of SB-235349, 2. The initial synthesis
of SB-235349, 2, is outlined in Scheme 2. 2-N-Methylami-
nomethylnitrobenzene 5 was prepared from 2-nitrobenzyl
alcohol 3 by bromination and reaction with aqueous methy-
lamine. The latter stages were adapted from a sequence
devised by GlaxoSmithKline Discovery chemists to prepare
racemic 1,4-benzodiazepines of structure similar to that of
1. This sequence was used to prepare multigram amounts of
high quality 2 in the laboratory in about 40% overall yield
from 2-nitrobenzylbromide 4.9 Whilst scaling up the syn-
thesis, several deficiencies became apparent. The overall
sequence was long with a modest yield of product. At each
stage, the intermediates were isolated, and some were viscous
oils. Bromide 4 was a very unpleasant material to handle
(Caution: lachrymatory!), and we had concerns about its
lack of thermal stability. Also, we wanted to avoid the use
of BOC protection/deprotection if at all possible. The BOC
deprotection stage required an excess of trifluoroacetic acid
(TFA) in dichloromethane. The excess TFA had then to be
removed. This left the bis-TFA salt of the acyclic 10 solvated
with at least 2 equiv of TFA. In the final ring-closing
reaction, up to 5 equiv of sodium methoxide had to be used
to neutralise the TFA before the benzodiazepine ring would
* Author for correspondence. Telephone: +44(0)1732 372361. Fax: +44-
(0)1732 372355. E-mail: David_O_Morgan@gsk.com.
† Present address: Process R&D, AstraZeneca Charnwood, Bakewell Road,
Loughborough, Leicestershire, LE11 5HT, UK.
(1) Samanen, J. M.; Ali, F. E.; Barton, L. S.; Bondinell, W. E.; Burgess, J. L.;
Callahan, J. F.; Calvo, R. R.; Chen, W.; Chen, L.; Erhard, K.; Feuerstein,
G.; Heys, R.; Hwang, S.-M.; Jakas, D. R.; Keenan, R. M.; Ku, T. W.; Kwon,
C.: Lee, C.-P.; Miller, W. H.; Newlander, K. A.; Nichols, A.; Parker, M.;
Peishoff, C. E.; Rhodes, G.; Ross, S.; Shu, A.; Simpson, R.: Takata, D.;
Yellin, T. O.; Uzsinskas, I.; Venslavsky, J. W.; Yuan, C.-K.; Huffman, W.
F. J. Med. Chem. 1996, 39, 4867-4870.
(2) Scarborough, R. M.; Gretler, D. D. J. Med. Chem. 2000, 43, 3453-3473.
(3) Topol, E. J.; Easton, J. D.; Amarenco, P.; Califf, R.; Harrington, R.;
Graffagnino, C.; Davis, S.; Diener, H. C.; Ferguson, J.; Fitzgerald, D.;
Shuaib, A.; Koudataal, P. J.; Theroux, P.; Van de Werf, F.; Willerson, J.
T.; Chan, R.; Samuels, R.; Ilson, B.; Granett, J. Am. Heart J. 2000, 139,
927-933.
(4) Miller, W. H.; Ku, T. W.; Ali, F. E.; Bondinell, W. E.; Calvo, R. R.; Davis,
L. D.; Erhard, K. F.; Hall, L. B., Huffman, W. F.; Keenan, R. M.; Kwon,
C.; Newlander, K. A.; Ross, S. T.; Samanen, J. M.; Takata, D. T.; Yuan,
C.-K. Tetrahedron Lett. 1995, 36, 9433-9436.
(5) Ku, T. W.; Ali, F. E.; Bondinell, W. E.; Erhard, K. F.; Huffman, W. F.;
Venslavsky, J. W.; Yuan, C. C-K. Tetrahedron Lett. 1997, 38, 3131-
3134.
(7) Etridge, S. K.; Hayes, J. F.; Walsgrove, T. C.; Wells, A. S. Org. Process
Res. DeV. 1999, 3, 60-63.
(8) Andrews, I. P.; Atkins, R. J., Badham, N. F.; Bellingham, R. K.; Breen, G.
F.; Carey, J. S.; Etridge, S. K.; Hayes, J. F.; Hussain, N.; Morgan, D. O.;
Share, A. C.; Smith, S. A. C.; Walsgrove, T. C.; Wells, A. S. Tetrahedron
Lett. 2001, 42, 4915-4917.
(6) Hayes, J. F. Synlett 1999, 865-866.
(9) Golec, F. A.; Reilly, L. W. J. Heterocycl. Chem. 1988, 25, 789-790.
10.1021/op034024c CCC: $25.00 © 2003 American Chemical Society
Published on Web 07/26/2003
Vol. 7, No. 5, 2003 / Organic Process Research & Development
•
655

Scheme 1. Synthetic Strategy
Scheme 2. Initial Synthesis of 2a
a Reagents: i) 48% aq HBr; ii) aq MeNH2; iii) BOC2O; iv) cyclohexene, Pd-C; v) DMAD; vi) Et2NH2CO2, Pd-C; vii) TFA; viii) NaOMe.
form. During the scale-up of the BOC deprotection reaction
it became evident that another product was being generated,
and this was increasing with scale. This compound was the
trifluoromethyl acetamide derivative of the secondary ali-
phatic amine 11 and was formed at 20-40% in larger-scale
reactions. This compound 11 could be readily cleaved back
to the secondary amine using potassium carbonate in
methanol, but it meant adding another step to an already
lengthy process.
Whilst this sequence was being scaled up and optimised,
a chance observation lead to the identification of a much
more direct and high-yielding process for the preparation of
2. To explore new routes to lotrafiban 1 we needed some of
the C-2 exocyclic double bond compound 16, Scheme 3, to
investigate chiral reduction as a potential way of introducing
the chirality at C-2. A sample of the N-BOC acrylamide 8
was taken and deprotected before reduction of the double
bond. It was thought that simple treatment with base would
yield 16. This did occur eventually, but it was observed that
the unprotected acrylamide never directly cyclised to 16 but
was converted to an intermediate that was then converted to
16 via a base-catalysed process. The intermediate compound
was isolated and identified as the quinazoline 15. Presumably,
the ring expansion mechanism occurs via deprotonation at
the carbon alpha to the methyl ester, followed by ring
cleavage to produce an amine anion that then reacts with
the ester group to produce the unsaturated diazepine.
Interestingly, no eight-membered rings were seen and no
products could be detected that would result from the initial
anion fragmenting the other way to produce an anilino anion.
Once formed, 16 would not revert back to 15 under basic
conditions. This is contrary to literature precedent that shows
656
•
Vol. 7, No. 5, 2003 / Organic Process Research & Development

Scheme 3. Final Synthesis of 2a
a Reagents: i) MeSO2Cl, Et3N, THF; ii) aq MeNH2 92%; iii) DMAD, EtOAc, 100%; iv) cyclohexene, Pd-C, 92% or H2, Raney Ni, 89%; v) AcOH, MeOH, 94%;
vi) DBU or NaOMe, MeOH, 95%; vii) NH4HCO2 or NaBH4, Pd-C, MeOH, H2O, 93%.
that other 1,4-benzodiazepine-3-ones rearrange to quinazo-
lines under base catalysis.10
First Pilot Synthesis of SB-235349, 2. Having identified
this novel and irreversible ring expansion, we were able to
devise a simplified process to prepare 2 on scale (Scheme
3). The process was designed to be a “one-pot” process, with
no isolation of intermediates until crystallisation of 2.
2-Nitrobenzyl bromide was replaced by the corresponding
mesylate, 12, derived from 2-nitrobenzyl alcohol, 3. To
minimise handling and exposure, the mesylate was prepared
and added directly to a large excess of aqueous methylamine.
The methylamine product, 5, was extracted into dichlo-
romethane after acid-base partitioning, and then the solvent
was exchanged for ethyl acetate. The Michael reaction
between 5 and dimethylacetylene dicarboxylate (DMAD) was
straightforward. The selective reduction of the nitro group
in 13 was more problematic. When the process was first
piloted, the only clean reduction conditions found were a
transfer hydrogenation with Pd/C and cyclohexene as the
hydrogen donor. Other hydrogen donors gave complex
mixtures or also reduced the acrylate double bond. The latter
problem was also found with hydrogen and Pd/C catalysts
without careful control of the reaction conditions. Prolonged
heating cyclised the resulting aniline 14 to the quinazoline
15. The ring expansion was performed with 2 equiv of 1,8-
diazabicyclo[5.4.0]undec-7-ene (DBU). The reduction of 16
to 2 was readily accomplished using gas-phase hydrogena-
tion. However, the large demands for supplies of 2 out-
stripped the R&D hydrogenation capacity available; there-
fore, a transfer hydrogenation was developed which utilised
ammonium formate as the hydrogen donor.
Figure 1.
The mesylation of 3 went cleanly in 7 volumes of THF
and an excess of mesyl chloride (1.2 mol equiv) and
triethylamine (1.1 mol equiv). The mesylate, 12 was formed
in essentially quantitative yield from 2-nitrobenzyl alcohol,
3. Water had to be rigourously excluded to avoid hydrolysis
of the mesyl chloride. Prolonged contact of the product with
the Et3N‚HCl formed led to slow conversion to 2-nitrobenzyl
chloride, which reacted as the mesylate with aqueous
methylamine in the subsequent step. To minimise handling
the mesylate, the Et3N‚HCl byproduct was not removed. The
resulting slurry from the mesylation reaction was transferred
to an excess of aqueous methylamine. A large excess of
methylamine is used to suppress the formation of the
dialkylated product 17, Figure 1. Apart from the loss in yield
attributed to the formation of this byproduct, this compound
could also act as a poison in the cyclohexene reduction; thus,
it was deemed desirable to form as little of this material as
possible. The dialkylated product 17 was usually formed at
3-5 mol % when 12 mol equiv of methylamine was used.
The level of dialkylation was found to vary with the charge
of methylamine and cosolvent. Optimisation eventually
reduced the amount of methylamine added to 9 mol equiv.
After the reaction the product, 5, was originally extracted
into dichloromethane. Initially, this presented no problem
since the solutions were carried forward straight away.
However, during one pilot-plant run, the dichloromethane
solution of 5 had to be stored for a few days. The yields in
subsequent steps were lower than expected. It was found
that 5 had a lot of the HCl salt present as well as other
decomposition products resulting from reaction with dichlo-
romethane.11 After this, the solvent was exchanged for ethyl
acetate or toluene. Before further use, 5 was purified by
Process Development. The route as shown in Scheme 3
worked well in the laboratory and scaled reasonably well in
the plant, producing 2 in 50-55% overall yield from 2-
nitrobenzyl alcohol, 3. Some reactions were, however, a little
capricious in the plant, and there were obvious areas where
increases in throughput and efficiency could be achieved.
(10) Heindel, N. D.; Fish, V. B.; Lemke, T. F. J. Org. Chem. 1968, 33, 3997-
4004.
(11) Nevstad, G. O.; Songstad, J. Acta Chem. Scand. B 1984, 38, 469-477.
Vol. 7, No. 5, 2003 / Organic Process Research & Development
•
657

extraction into aqueous sulphuric acid to remove traces of
2-nitrobenzyl alcohol, 3, that could act as a catalyst poison
in the cyclohexene reduction. After washing the aqueous
acid, the product was extracted into ethyl acetate, and reaction
with dimethylacetylene dicarboxylate, DMAD, could take
place.
run under Dean and Stark conditions to remove the water
added via the 13 solution and the Pd/C catalyst. Investigations
to find the cause of the variability revealed that dehydration
of the catalyst was the cause of reactions stopping prema-
turely. This transfer reduction seems to be very sensitive to
metal sintering on the catalyst surface (this probably occurs
as a consequence of dehydration). The procedure for running
the reaction in the plant was therefore modified. After the
cyclohexene was added, the reaction was run under reflux
until the reduction was about 80% complete. The reactor
was then set to Dean and Stark to begin the drying process.
This new procedure gave very reliable reaction times of about
4 h, and once the procedure was adopted, additional catalyst
never had to be added. During the process-development
phase, a major goal was to eliminate cyclohexene from the
process and to identify conditions that could tolerate a
feedstock solution of 13 that had not undergone extensive
extractions to clean it up. Several grades of Pd/C catalyst
were identified that gave good selectivity in the reduction
using hydrogen gas; however, the best solution was to use
washed Raney nickel as the catalyst with hydrogen gas. This
gave excellent selectivity and could tolerate crude 13
solutions. The Raney nickel reduction was not scaled up onto
the GlaxoSmithKline pilot plant but was later successfully
scaled up at a contract facility.
The reaction of 5 with DMAD was straightforward. The
exothermic Michael reaction was readily controlled to below
30 °C by slowly adding the solution of DMAD to 5 with
glycol cooling and control of addition rate. After the reaction,
the methylamine can be replaced with the dialkylated
material, 17, and residual 5 could be removed by extraction
into aqueous acid. This was done by washing with dilute
sulphuric acid, aqueous sodium sulphate-sulphuric acid,
dilute sodium carbonate, then water. Due to the instability
of 13 to aqueous acids (fairly rapid hydrolysis occurs to give
5 and dimethyl 2-oxosuccinate), contact with the acid
solutions had to be kept to a minimum and the solutions
kept cold (5 °C). Whilst this removed all of the dimer, the
process was very lengthy, and it was felt that the instability
of 13 towards aqueous acids may prove problematic if control
of the process was lost on scale-up. One of the long-term
goals was to remove the acid-base manipulations of 13 and
5, which would eliminate six extraction steps from the overall
process.
The selective reduction of the nitro group in 13 was
initially achieved via a transfer hydrogenation using Pd-C
catalyst and cyclohexene. When the process was first piloted,
this was the only set of conditions found that gave excellent
selectivity for the reduction of the nitro group and not the
double bond in 13. A large drawback was that 5 mol equiv
of cyclohexene was required to achieve a reasonable reaction
rate. Another undesirable feature was the production of
benzene as a byproduct. The reaction was very sensitive to
low levels of poisons such as 2-nitrobenzyl alcohol, 3, or
the dialkylated methylamine, 17. Hence, the lengthy extrac-
tion processes to ensure that these were totally removed prior
to the transfer hydrogenation. Despite its limitations, the
transfer reduction was used in several campaigns and
performed consistently well in the pilot plant, provided
several precautions were observed. The 13 feedstock had to
be free of the poisons as already mentioned.
Great care had to be taken to ensure that traces of acids
were not introduced via the 13 solution or this could
contaminate the hydrogenation vessel inadvertently. If the
reduction medium became acidic, then considerable amounts
of byproducts resulted from hydrogenolysis of the N-benzyl
bond. One curious observation during the plant campaigns
was the very variable length of the reaction (4-24 h). On
occasion, the reaction would stop before complete consump-
tion of 13 and would have to be restarted by the addition of
a small amount of fresh catalyst or by filtering off the old
catalyst and starting again with a fresh charge. Use-testing
batches of Pd/C catalyst and solutions of 13 gave no
indication of when this problem was likely to arise. Since
the downstream reactions had to be anhydrous, the standard
procedure was to charge the solution of 13, add the catalyst,
heat the reaction to reflux then add the cyclohexene and then
During the course of the reduction, 20-60% of the aniline
product 14 would cyclise to the quinazoline, 15. When the
reduction was complete and the Pd/C catalyst removed by
filtration, the resulting mixture of 14 and 15 was refluxed
to convert remaining aniline to 15. The time for completion
of this process was always very variable, on occasions up
to 24 h. It was discovered that the cyclisation was much
faster in protic solvents and could be catalysed by traces of
weak anhydrous acids. It was found to be very important to
prevent solutions of 15 from coming into contact with
aqueous acid to prevent rapid hydrolysis of the aminal
function. Since the solutions had to be azeotropically dried
and the solvent replaced with methanol for the ring-expansion
reaction, it was overall more efficient to run a solvent
exchange immediately after the reduction of 13 had finished
and the catalyst had been filtered off. Once the ethyl acetate
had been removed and replaced with methanol, a small
amount of acetic acid was added. This procedure ensured
complete conversion of the aniline 14 to the quinazoline 15
in less than an hour. The ring-expansion reaction to give 16
was originally piloted using 2 mol equiv of DBU in ethyl
acetate. This gave complete reaction in about 6 h at 80 °C.
The DBU was then washed out into aqueous acetic acid.
Whilst this gave a clean and high-yielding conversion, the
relative high cost of DBU made this reagent unattractive for
large-scale manufacture. A range of solvents and bases was
screened. Overall, the best combination was found to be
methanol with a few mol % of sodium methoxide as the
basic catalyst. This fitted in well with the best conditions
for the quinazoline-forming reaction. After the formation of
the quinazoline, acetic acid was neutralised with sodium
methoxide and a small excess added to act as a catalyst for
the ring-expansion reaction. This typically took about 1 h at
658
•
Vol. 7, No. 5, 2003 / Organic Process Research & Development

Scheme 4. Alternative Synthesis of 16a
a Reagents: i) aq MeNH2, MeOH; ii) NaBH4, H2O; iii) H2, Pd-C, MeOH, iv) DMAD, MeOH; v) AcOH, MeOH; vi) NaOMe, MeOH.
60 °C. The key to maintaining high yield and purity in this
stage was to ensure the solution was dry before the sodium
methoxide was added.
borohydride would appear to offer a significant advantage
in terms of cost. The reaction was shown to proceed very
efficiently with only 1.3 equiv of amine necessary for
complete reaction and with no dialkylation being observed.
Subsequent studies showed that 5 could readily be reduced
via catalytic hydrogenation over Pd-C in methanol to furnish
the diamine 19 in virtually quantitative yield over the two
stages from 18. Reaction of 19 with DMAD in methanol
gave the conjugate adduct 14, which upon standing partially
isomerised to a mixture containing regioisomer 20 and
quinazoline 15. Treatment of this mixture with a catalytic
quantity of acetic acid afforded the tetrahydoquinazoline 15.
Treatment of the solution of 15 with sodium methoxide in
methanol as per the current synthesis furnished 16 which
was isolated in an overall yield of 68% from 18. This
simplified procedure appears to circumvent the nitro group
reduction of the current synthesis, and furthermore there is
no solvent change when proceeding from 19 to 16 with all
reactions being conducted in methanol.
In the initial laboratory investigation into the synthesis
of 2, simple hydrogenation of 16 was shown to be feasible
(Pd/C catalyst, 50 °C, 50 psi H2). A transfer hydrogenation
method was developed. The methanol solution of 16 was
treated with a small amount of acetic acid to neutralise the
sodium methoxide, Pd/C catalyst was added, and the mixture
was heated to 65 °C. A solution of ammonium formate in
water was added to act as the hydrogen source. This worked
well in the laboratory, but scale-up was problematical. Often
in the pilot plant, a very large excess of ammonium formate
had to be added to drive the reaction to completion (up to
20 mol equiv). This was attributed to unproductive decom-
position of the ammonium formate on the Pd catalyst. As a
consequence of this, the process was very volume inefficient
and could take up to 18 h to complete. To address this, a
screen of other hydrogen donors was run. It was found that
the best solution was to add aqueous sodium borohydride to
a methanol solution of 16 and Pd/C catalyst at about 45 °C.
This procedure gave a fast and clean reduction virtually as
soon as the borohydride solution was added (reaction time
was reduced from 18 to 1 h). When the reduction was
complete, the mixture was neutralised with acetic acid.
Attempts to process the basic reaction mixture lead to
decomposition of 2. The solution was then diluted with
dichloromethane. During the reduction, the presence of water
causes most of the 2 to crystallise out as it is formed. The
dilution with dichloromethane was necessary to solubilise
the product.
Thus, a chance observation in the laboratory led to a fast
and efficient multistep process to produce 2. At the end of
the process-development phase, we could prepare 2 in 120-
kg batches of starting from 100 kg of 2-nitrobenzyl alcohol,
3. The process was suitable for multi-ton campaigns if
required.
Later Process Refinements. While the route described
above was successfully being run in a manufacturing plant,
further laboratory investigations led to the evolution of a
simplified version of the chemistry, and it is shown in
Scheme 4. The commercially available 2-nitrobenzyl alcohol,
3, used in the current process is, in fact, manufactured by
reduction of 2-nitrobenzaldehyde, 18. Accordingly, conver-
sion of 18 to the current intermediate 5 via reductive
amination with methylamine in the presence of sodium
During the course of these investigations work on the
lotrafiban project was suspended, and so this simplified
process just described was never subjected to scale-up.
Summary
An efficient synthesis of the key intermediate, 2,3,4,5-
tetrahydro-4-methyl-3-oxo-1H-1,4-benzodiazepine-2-acetic
acid methyl ester 2, for the synthesis of SB-214857-A is
described. The starting material 2-nitrobenzyl alcohol, 3, is
converted into the required product 2 in seven chemical
transformations. Since this process would be required to
produce many hundreds of kilograms of ester 2 the process
was adapted such that a “one-pot” procedure could be run,
i.e., none of the six intermediates need be isolated en route
to the product. Not isolating the intermediates leads to a high
overall yield, since no loss upon crystallisation is encoun-
tered. Also, the process is more expedient to run on a pilot
plant since isolations by filtration and subsequent drying tend
to be time-consuming processes. This process was performed
on a 100-kg scale in the GSK pilot plant where the molar
isolated yield of ester 2 was 70%. Subsequently, larger
quantities of ester 2 were produced by out-sourced contrac-
tors using this chemistry.
A modified route starting from 2-nitrobenzaldehyde, 18,
is also described. The modified procedure requires much less
methylamine as reagent, and all reactions can be performed
in a single solvent, methanol.
Vol. 7, No. 5, 2003 / Organic Process Research & Development
•
659

Experimental Section
allowing the temperature to rise to 7 °C. After addition of
the second bicarbonate wash, the headtank and transfer lines
were washed through with demineralised water (10 L).
Ten percent palladium on charcoal (Pd-C), 59% water
wet (30 kg, source Johnson Matthey, 87L) was charged and
the mixture heated to 66 °C. Cyclohexene (240 kg, 2921
mol) was added over 35 min, maintaining reflux. The
headtank and transfer lines were washed through with ethyl
acetate (9 kg). The reaction mixture was stirred at reflux
(68 °C) for 4.5 h, and then the water was separated off by
Dean-Stark distillation over a further 4 h 20 min. The total
volume of water collected was 54 L.
2,3,4,5-Tetrahydro-4-methyl-3-oxo-1H-1,4-benzodiaz-
epine-2-acetic Acid Methyl Ester (2) from Nitrobenzyl
Alcohol (3). A solution of methanesulphonyl chloride (87
kg, 795 mol) in tetrahydrofuran (THF) (35 kg) was added
to a solution of 2-nitrobenzyl alcohol 3 (100 kg, 653 mol)
and triethylamine (71.5 kg, 706 mol) in THF (620 kg) over
2 h, allowing the temperature to rise from 17 °C to a
maximum of 34 °C. The headtank and transfer lines were
washed through with THF (10 kg). The reaction mixture was
stirred at 32 °C for 1 h. The mixture was added to 40% w/w
methylamine in water (620 kg, 7985 mol) over 45 min,
allowing the temperature to rise from 20 to a maximum of
33 °C. The transfer line was washed through with THF (89
kg), and stirring continued at 32 °C for 1.5 h. The phases
were allowed to separate for 15 min. The phases were
separated, and the aqueous phase was extracted with toluene
(256 kg) and then discarded. The organic phases were
combined, and solvent (900 L) was distilled off under
vacuum, maximum base temperature 22 °C. The resulting
concentrate was cooled to 8 °C (pH 10.3) and the pH adjusted
to 4.3 by the addition of 12% v/v aqueous sulphuric acid
solution (177 L), taken from a bulk solution of concentrated
sulphuric acid (340 kg) and demineralised water (1500 L).
The temperature was allowed to rise to a maximum of 17
°C. The mixture was stirred for 5 min, and the phases were
allowed to separate for 15 min. The phases were separated,
and the organic phase was discarded. The aqueous phase
was diluted with demineralised water (150 L) and then
basified to pH 11.7 with 50% w/w sodium hydroxide solution
(110 kg). Compound 5 was extracted into ethyl acetate (270
kg), and the mixture was warmed to 27 °C. The aqueous
phase was separated off and extracted with further ethyl
acetate (135 kg), maintaining the temperature at 25-30 °C,
and then discarded. The organic phases were combined.
A solution of dimethylacetylene dicarboxylate (86 kg, 605
mol) in ethyl acetate (90 kg) was added over 1 h with cooling
applied, allowing the temperature to rise to a maximum of
29 °C. The headtank and transfer lines were washed through
with ethyl acetate (9.0 kg). The reaction mixture was stirred
at 23-24 °C for 40 min. The mixture was cooled to 4 °C,
and demineralised water (50 L) was charged, followed by
ethyl acetate (180 kg). The pH was adjusted from 10.9 to
2.6 by the careful addition of 12% v/v aqueous sulphuric
acid (45 L), maintaining the temperature between 0 and 5
°C. The mixture was stirred for 5 min, and the phases were
allowed to separate for 15 min. The aqueous phase was
separated off and discarded, and the organic phase washed
with an aqueous solution of sulphuric acid and sodium
sulphate (135 kg), taken from a bulk solution of sodium
sulphate (50 kg), concentrated sulphuric acid (9.0 kg), and
demineralised water (650 L), maintaining the temperature
between 0 and 5 °C. The headtank and transfer lines were
washed through with demineralised water (10 L). The
mixture was stirred for 5 min, and the phases were allowed
to separate for 15 min. The aqueous phase was separated
off and discarded and the organic phase washed twice with
10% w/w sodium bicarbonate solution (2 × 125 kg),
The mixture was cooled to 25 °C and filtered, washing
the filters and lines through with ethyl acetate (89 kg). A
mixture of ethyl acetate and unreacted cyclohexene was
distilled off at atmospheric pressure (950 L), the temperature
rising to a maximum of 82 °C. The remaining contents of
the vessel were cooled to 62 °C, and methanol (550 kg) was
added. The distillation was continued at atmospheric pressure
until a further 200 L of solvent had been distilled out
(maximum temperature 65 °C). Glacial acetic acid (2.1 kg,
35 mol) was charged, washing through with methanol (8 kg),
and the mixture stirred at reflux for 1.5 h.
A 30% solution of sodium methoxide in methanol (19.7
kg, 109 mol) was added, maintaining reflux, and the headtank
and lines were washed through with methanol (8 kg). The
mixture was stirred at reflux for 2 h. The mixture was cooled
to 33 °C and the pH (10.0) adjusted to 7.7 by the addition
of glacial acetic acid (5.1 kg, 85 mol). The mixture was
stirred at 30-35 °C for 33 min (final pH 7.5). Ten percent
Pd-C, 59% water wet (15 kg, source Johnson Matthey 87L)
was charged, and a solution of sodium borohydride in
demineralised water containing sodium hydroxide [sodium
borohydride (20.0 kg, 528.7 mol) in demineralised water (265
L) containing sodium hydroxide (0.12 kg)] was added over
2 h 45 min, allowing the temperature to rise from 33 °C and
maintaining at 40-45 °C. The headtank and transfer lines
were washed through with demineralised water (5 L). The
mixture was stirred at 40-42 °C for 55 min.
The pH was adjusted from 9.6 to 7.0 by the addition of
glacial acetic acid (25 kg, 416 mol). The headtank and
transfer lines were washed through with demineralised water
(10 L). Dichloromethane (930 kg) was added and the slurry
(35 °C) filtered to a clean vessel. The filters and lines were
washed through with dichloromethane (266 kg) at 30 °C. A
mixture of dichloromethane and methanol was distilled off
at atmospheric pressure to a base temperature of 50 °C
(volume of distillate 800 L). The distillation was continued
under vacuum to a base temperature of 53 °C (volume of
distillate 511 L). Demineralised water (500 L) was added to
the concentrate and the mixture cooled to 20 °C to crystallise
the product. After stirring the slurry for 1 h at 17-20 °C,
the product was filtered off in a filter drier, washed with
demineralised water (200 L), and dried at 35-40 °C under
vacuum for 14 h to give 2 as a white solid, 117 kg 70%
1
from 2-nitrobenzyl alcohol, 3: mp 174-176 °C; H NMR
(400 MHz, CDCl3) δ 7.07 (td, 1 H, J ) 7.7, 1.4 Hz), 6.93
660
•
Vol. 7, No. 5, 2003 / Organic Process Research & Development

(dd, 1 H, J ) 7.5, 1.1 Hz), 6.67 (td, 1 H, J ) 7.4, 1.0 Hz),
6.57 (dd, 1 H, J ) 8.1, 0.9 Hz), 5.42 (d, 1 H, J ) 16.3 Hz),
5.00 (q, 1 H, J ) 6.1 Hz), 4.09 (d, 1 H, J ) 5.4 Hz), 3.73
(s, 3 H), 3.72 (d, 1 H, J ) 16.4 Hz), 3.07 (s, 3 H), 2.98 (dd,
1 H, J ) 16.0, 6.8 Hz), 2.65 (dd, 1 H, J ) 16.0, 6.5 Hz);
13C NMR (100 MHz, CDCl3) δ 171.9, 169.6, 145.1, 129.5,
129.0, 119.8, 118.2, 117.5, 53.3, 51.9, 51.8, 35.9, 34.6; IR
(neat, cm-1) 1728, 1650; LRMS (CI +ve) m/z 249 (M+ +
H).
1 H), 4.07 (d, 1 H, J ) 16.8 Hz), 3.81 (s, 3 H), 3.73 (d, 1
H, J ) 16.7 Hz), 3.68, (s, 3 H), 3.15, (d, 1 H, J ) 15.8 Hz),
2.93 (d, 1 H, J ) 15.8 Hz), 2.34 (s, 3 H); 13C NMR (100
MHz, CDCl3) δ 171.4, 170.9, 140.3, 127.6, 127.0, 118.4,
117.0, 115.1, 74.5, 52.9, 52.0, 51.8, 42.1, 38.3; IR (neat,
cm-1) 1732; LRMS (CI +ve) m/z 279 (M+ + H).
A sample of intermediate (1,3,4,5-tetrahydro-4-methyl-
3-oxo-2H-1,4-benzodiazepin-2-ylidene)acetic acid methyl
ester 16 was isolated in the laboratory and characterised as
a white solid: mp 115-116 °C; 1H NMR (400 MHz, CDCl3)
δ 10.61 (s, 1 H), 7.30 (td, 1 H, J ) 7.7, 1.5 Hz), 7.17 (d, 1
H, J ) 7.5 Hz), 7.06-7.02 (m, 2 H), 5.43, (s, 1 H), 4.27, (s,
2 H), 3.74 (s, 3 H), 3.10 (s, 3 H); 13C NMR (100 MHz,
CDCl3) δ 170.9, 162.5, 153.6, 139.2, 129.5, 128.3, 127.5,
123.6, 120.5, 89.4, 50.8, 50.8, 34.5; IR (neat, cm-1) 1650,
1613, 1582; LRMS (CI +ve) m/z 247 (M+ + H).
Procedure for the Reduction of 13 to 14 Using Raney
Nickel. Nitro-aromatic 13 (5.00 g, 16.2 mmol) was dissolved
in a mixture of toluene (30 mL) and methanol (20 mL) and
then added to Raney nickel (0.6 g dry weight, source W. R.
Grace and Co., previously washed with water until pH 7) in
a suitable pressure vessel. The system was purged with
hydrogen, pressurised to 40 psi with hydrogen, and then
heated to 45 °C. After 8 h the reaction was complete, and
the reaction mixture was cooled, purged with nitrogen, and
then filtered through Celite. The filtercake was washed with
toluene (30 mL) and water (50 mL). The layers were
separated, and the organic layer was dried over MgSO4 and
then concentrated by vacuum distillation to leave aniline 14
as a viscous yellow oil (4.00 g, 89%).
HPLC in process-control method; Luna or Prodigy ODS
column 75 mm × 4.6 mm, 3 µm. Eluent A: 0.05 M NaH2-
PO4 adjusted to pH 7.0 with aq NaOH. Eluent B: acetonitrile.
Isocratic with 35% B, run time 10 min, Flow rate 1.5 mL/
min. Detector at 254 nm. Injection volume 10 µL. Sample
preparation 0.1 mg/mL. Typical retention times; 2-nitroben-
zyl alcohol 3 1.6 min, mesylate 12 3.8 min, amine 5 0.7
min, nitro 13 7.0 min, aniline 14 3.6 min, quinazoline 15
2.9 min, unsaturated ester 16 3.3 min, ester 2 1.9 min.
A sample of intermediate (2-nitrophenyl)methyl meth-
anesulfonate 12 was isolated in the laboratory and charac-
1
terised as a light tan solid: mp 92-94 °C; H NMR (400
MHz, CDCl3) δ 8.18 (dd, 1 H, J ) 8.0, 0.8 Hz), 7.76-7.73
(m, 2 H), 7.58 (td, 1 H, J ) 8.0, 0.5 Hz), 5.67 (s, 2 H), 3.13
(s, 3 H); 13C NMR (100 MHz, CDCl3) δ 147.0, 134.3, 130.1,
129.7, 129.4, 125.3, 68.0, 37.8; IR (neat, cm-1) 1521, 1340,
1171.
A sample of intermediate N-methyl-1-(2-nitrophenyl)-
methanamine 5 was isolated in the laboratory and charac-
terised as the hydrochloride salt. The data obtained was in
accordance with the published data.9
A sample of intermediate dimethyl (2E)-2-[methyl(2-
nitrobenzyl)amino]but-2-enedioate 13 was isolated in the
laboratory and characterised as a yellow solid: mp 87-89
(1,3,4,5-Tetrahydro-4-methyl-3-oxo-2H-1,4-benzodiaz-
epin-2-ylidene)acetic Acid Methyl Ester (16) from Ni-
trobenzaldehyde (18). A solution of 2-nitrobenzaldehyde
18 (15.1 g, 100 mmol) in methanol (90 mL) was treated with
aqueous methylamine (40%, 10 mL, 130 mmol) and the
solution stirred at 25 °C for 15 min. A solution of sodium
borohydride (2.86 g) in water (15 mL) was added dropwise
at 25-30 °C over 10 min and the resulting mixture stirred
at 25 °C for a further 1 h. The solution was hydrogenated
over 10% Pd-C (0.6 g, 59% wet, 0.28 mmol, Johnson
Matthey 87L) at 60 psi and 25 °C for 1.5 h. The mixture
was filtered through Celite and the bed rinsed with methanol
(30 mL). The combined filtrate and rinse were partitioned
into ethyl acetate and water. The aqueous phase was
re-extracted with ethyl acetate, and the combined organic
phases were dried over magnesium sulphate and evaporated
to give the diamine 19 as a pale brown oil (13.24 g, 97.4%).
1H NMR (400 MHz, CDCl3) δ 7.07 (td, 1 H, J ) 7.6, 1.6
Hz), 7.01 (dd, 1 H, J ) 7.2, 1.2 Hz), 6.68-6.62 (m, 2 H),
3.73 (s, 2 H), 2.41 (s, 3 H); 13C NMR (100 MHz, CDCl3) δ
146.9, 129.8, 128.3, 124.1, 117.6, 115.6, 55.1, 36.0; IR (neat,
cm-1) 1614, 1493. A further sample was isolated as the
mixed oxalate/hydrochloride salt. The data obtained was in
accordance with the published data.9
1
°C; H NMR (400 MHz, CDCl3) δ 8.12 (dd, 1 H, J ) 8.2,
1.2 Hz), 7.67 (td, 1 H, J ) 7.6, 1.3 Hz), 7.48 (td, 1 H, J )
7.8, 1.9 Hz), 7.38 (dd, 1 H, J ) 7.8, 0.9 Hz), 4.75 (s, 2 H),
4.68 (s, 1 H), 3.91 (s, 3 H), 3.64 (s, 3 H), 2.92 (s, 3 H); 13
C
NMR (100 MHz, CDCl3) δ 167.7, 165.6, 154.8, 147.8, 134.2,
131.8, 128.5, 128.0, 125.4, 86.0, 53.8, 53.0, 50.8, 38.4; IR
(neat, cm-1) 1732, 1694; LRMS (CI +ve) m/z 309 (M+ +
H).
A sample of intermediate dimethyl (2E)-2-[(2-ami-
nobenzyl)(methyl)amino]but-2-enedioate 14 was isolated
in the laboratory and characterised as a viscous yellow oil.
1H NMR (400 MHz, CDCl3) δ 7.14 (td, 1 H, J ) 7.7, 1.4
Hz), 7.00 (dd, 1 H, J ) 7.5, 1.2 Hz), 6.72 (td, 1 H, J ) 7.4,
1.1 Hz), 6.66 (dd, 1 H, J ) 8.0, 0.9 Hz), 4.77 (s, 1 H), 4.19
(s, 2 H), 3.96 (s, 3 H), 3.65 (s, 3 H), 2.66 (s, 3 H); 13C NMR
(100 MHz, CDCl3) δ 168.1, 166.7, 154.9, 145.6, 130.6,
129.6, 118.3, 118.1, 116.1, 86.0, 53.7, 53.1, 50.9, 35.2; IR
(neat, cm-1) 1734, 1685; LRMS (CI +ve) m/z 279 (M+ +
H).
A sample of intermediate methyl 2-(2-methoxy-2-oxo-
ethyl)-3-methyl-1,2,3,4-tetrahydroquinazoline-2-carboxy-
late 15 was isolated in the laboratory and characterised as
A solution of the diamine 19 (13.2 g, 97 mmol) in
methanol (70 mL) at 0-5 °C was treated with a solution of
DMAD (13.8 g, 11.9 mL, 97 mmol) in methanol (30 mL),
added dropwise over 30 min. After stirring for a further 1 h,
1
viscous oil. H NMR (400 MHz, CDCl3) δ 7.06 (td, 1 H, J
) 8.0, 0.8 Hz), 6.91 (dd, 1 H, J ) 7.5, 0.7 Hz), 6.72 (td, 1
H, J ) 7.4, 1.1 Hz), 6.64 (dd, 1 H, J ) 8.0, 0.8 Hz), 5.13 (s,
Vol. 7, No. 5, 2003 / Organic Process Research & Development
•
661

glacial acetic acid (0.79 g, 0.75 mL, 13 mmol) was added
and the solution heated under reflux for 1 h. After cooling
to 50 °C, a solution of sodium methoxide in methanol (30%,
6.25 mL, 34.7 mmol) was added and the resulting solution
heated under reflux for 2 h. The solution was cooled to 50
°C, and glacial acetic acid (2.1 g, 2 mL, 35 mmol) was added
followed by water (120 mL) at 50 °C. The solution was
cooled to 0-5 °C over 1 h and stirred at this temperature
for 5 h. The resulting solid was collected via suction
filtration, washed with 3:2 water:methanol (75 mL), and dried
in air to give the benzodiazepine 16 as a white crystalline
solid (16.3 g, 68.3% from 18), identical to an authentic
sample described above.
Acknowledgment
We thank our R&D Chemical Development colleagues.
Kaye Cattanach and Richard Escott from Analytical Sciences
and John Lewis, Martyn Fice, and Emma Griva from Process
Technologies are thanked for all their assistance.
Received for review February 10, 2003.
OP034024C
662
•
Vol. 7, No. 5, 2003 / Organic Process Research & Development