Synthesis of a bicyclic piperazine from l-aspartic acid and application of a fluoride-promoted SNAr coupling

Article abstract of DOI:10.1021/op2001854

The process development is reported of a pivotal C-N bond formation involving ((7R,9aS)-octahydro-1H-pyrido[1,2-a]pyrazin-7-yl)methanol (2) undergoing nucleophilic aromatic substitution with 3-chlorobenzo[d]isoxazole (3) to furnish ((7R,9aS)-2-(benzo[d]isoxazol-3-yl)octahydro-1H-pyrido[1,2-a] pyrazin-7-yl)methanol (4) as a key intermediate for a family of compounds (1). Essential to the success of the coupling is the use of fluoride in combination with a phase transfer catalyst. The development of an alternative route to bicyclic piperazine 2 that uses l-aspartic acid (20) as a starting material to avoid the need for a classical salt resolution is described.

Full text of DOI:10.1021/op2001854

ARTICLE
pubs.acs.org/OPRD
Synthesis of a Bicyclic Piperazine from L-Aspartic Acid and Application
of a Fluoride-Promoted SNAr Coupling
Janice E. Sieser,‡ Robert A. Singer,*,‡ Jason D. McKinley,‡ Dennis E. Bourassa,‡ John J. Teixeira,‡ and
James Long†
‡Chemical Research and Development, Pfizer Worldwide Research and Development, Pfizer Inc., Eastern Point Road, Groton,
Connecticut 06340, United States
†Chemical Research and Development, Pfizer Worldwide Research and Development, Sandwich Laboratories, Ramsgate Road,
Sandwich, Kent CT13 9NJ, United Kingdom
ABSTRACT: The process development is reported of a pivotal CÀN bond formation involving ((7R,9aS)-octahydro-1H-
pyrido[1,2-a]pyrazin-7-yl)methanol (2) undergoing nucleophilic aromatic substitution with 3-chlorobenzo[d]isoxazole (3) to
furnish ((7R,9aS)-2-(benzo[d]isoxazol-3-yl)octahydro-1H-pyrido[1,2-a]pyrazin-7-yl)methanol (4) as a key intermediate for a
family of compounds (1). Essential to the success of the coupling is the use of fluoride in combination with a phase transfer catalyst.
The development of an alternative route to bicyclic piperazine 2 that uses L-aspartic acid (20) as a starting material to avoid the need
for a classical salt resolution is described.
’ INTRODUCTION
shift higher-yielding steps toward the end of the synthesis. This
would also have the advantage of minimizing the introduction of
impurities toward the end of the route where control of
impurities is essential. The revised route began with the coupling
between racemic bicyclic piperazine 11 and 3 by heating to
130À140 °C in pyridine and DBU (1 equiv).8 During our
laboratory development efforts of this coupling between 11
and 3, we had observed yields ranging from 35 to 60%. It was
only after unsuccessfully scaling up the first batch that we realized
the extreme sensitivity of the reaction to residual solvents such as
methylene chloride and water. To improve the reliability of the
coupling, potassium carbonate (2 equiv) was introduced to the
process to act as a water scavenger. Not only does water impede
the coupling reaction, but bicyclic piperazine 11 can undergo
alkylation with methylene chloride to afford an adduct.9 Residual
methylene chloride was present during processing of the first
batch due to 3 being isolated as a solution in methylene chloride,
and the solvent had not been completely removed by distillation
prior to the coupling. In later batches, we avoided this problem by
substitution of methylene chloride with a blend of tert-butyl
methyl ether and heptane for the isolation of 3 as a solution.10
With these process changes the yield was increased from 10% in
the first batch up to 65% in the second batch.
After the coupling of 11 and 3 to furnish 16, a classical salt
resolution with D-tartaric acid provided 17 in 45% (90% theo-
retical) yield, effectively removing the undesired enantiomer that
had been carried along for several steps. The sequence of
oxidation, epimerization, and reduction resulted in a yield of
70% to provide the penultimate intermediate 4, with a minimal
amount of the yield loss being attributed to a methyl thiomethyl
ether side product of the oxidation step.11 These steps were
performed following protocols previously reported7 with the
exception of removing boron by temporarily forming an HCl salt
Bicyclic piperazines such as 2 are frequently incorporated as a
rigid scaffold into the framework of pharmaceutical targets.1
While Pd-catalyzed methodologies have recently dominated
the literature as a reliable means of coupling piperazine build-
ing blocks such as 2 with aryl halides,2 there are occasions
where conventional nucleophilic aromatic substitution (SNAr)
approaches can be competitive in efficiency with the metal-
catalyzed strategies.3 Some substrates, such as electron-rich aryl
chlorides, are more amenable to undergo SNAr reactions than
cross-couplings due to electronic and steric effects of the
substrates4 that impede the ability to undergo oxidative addition
with a Pd catalyst.5
While carrying out process research and development for a
route to a family of potential antipsychotic agents,6 1 (R = aryl or
heterocycle), we encountered a challenging coupling between
bicyclic piperazine 2 and 3-chlorobenzo[d]isoxazole (3). The
original route provided to us by the Medicinal Chemistry team
utilized an SNAr approach for coupling 2 and 3 (Scheme 1).1d
The coupling required high temperatures (130À140 °C) to
proceed and was not reproducible in yield or rate due to
sensitivities to residual water levels and other factors. The route
lacked efficiency in that oxidation state adjustments were re-
quired to correctly obtain the desired stereochemistry of the
bicyclic piperazine and a protecting group was employed to mask
a secondary amine that would eventually need to be coupled to 3.
’ RESULTS AND DISCUSSION
For the first bulk campaign we elected to introduce benzisox-
azole 3 at the start of the processing steps to include a
chromophore for analytical tracking purposes (Scheme 2) and
to facilitate isolations with the increased crystallinity.7 From an
economical perspective, it was advantageous to carry out the low-
yielding coupling between 3 and bicyclic piperazine 11 earlier to
avoid carrying bulk along that would eventually be lost and to
Received: July 8, 2011
Published: September 04, 2011
r
2011 American Chemical Society
1328
dx.doi.org/10.1021/op2001854 Org. Process Res. Dev. 2011, 15, 1328–1335
|

Organic Process Research & Development
ARTICLE
Scheme 1. Medicinal Chemistry route to 1
of 4 and then breaking the salt.12 Containment of the dimethyl-
sulfide byproduct of the oxidation was managed by using a bleach
scrubber during vacuum distillation in the workup. The final
coupling between 4 and 5 was often carried out in DMAC with a
base such as potassium tert-butoxide or KHMDS to afford a series
of potential drug candidates (1).
Optimization of the Coupling Between Bicyclic Piperazine
2 and 3 Using Fluoride and PTC. After this first scale-up
experience we considered new routes that would avoid the need
to adjust oxidation states and directly provide bicyclic piperazine
2 with the correct stereochemistry. A more immediate goal was
to further optimize the coupling between 2 and 3 to realize a
more robust and higher-yielding process. As a priority, we set out
to study the coupling, since any route would require this essential
bond formation. Metal-catalyzed cross-coupling strategies were
investigated, but chlorobenzisoxazole 3 was found to be unreac-
tive with all Pd catalysts examined.13 Recognizing that this
avenue was not viable, we refocused our efforts on nucleophilic
aromatic substitution reactions.
sources.1b,15 Initially, 3 was coupled with N-methylpiperazine as
a model substrate for 2. The trials with N-methylpiperazine and
3 using TBAF showed that the coupling proceeded well in
THF, DMAC, DMF, DMSO, acetonitrile, ethanol, isopropanol,
and acetone. When examining the coupling of 2 and 3, the
reaction was more sensitive to water as well as the choice of
solvent to facilitate the coupling, while also requiring higher
temperatures. The coupling was found to work optimally in
acetonitrile since it is suitable to remove water as an azeotrope
by distillation. As an alternative to TBAF which is highly
hygroscopic and often purchased in a hydrated form, KF was
used in combination with a PTC such as tetrabutylammonium
chloride, tetraethylammonium chloride or benzyltriethylam-
monium chloride.16 We favored using tetraethylammonium
chloride since it has better stability at higher temperatures
and can be dried with heating under vacuum to remove water
prior to use without decomposition. The bromide salts of the
PTCs failed to perform the desired coupling, limiting the
selection of PTCs to chlorides. While precedent for a fluoride-
catalyzed amination by Senanayake and co-workers14 showed
evidence for conversion of a heteroaryl chloride to a more
reactive heteroaryl fluoride intermediate, we were never able to
observe a 3-fluorobenzisoxazole intermediate under the reac-
tion conditions we used.
Fluoride has been reported to accelerate arylation reactions
and various other couplings, often in a solubilized form such as
TBAF.14 Considering that use of fluoride may promote the SNAr
coupling of 2 and 3 under milder conditions, various solvents
were examined in combination with TBAF and other fluoride
1329
dx.doi.org/10.1021/op2001854 |Org. Process Res. Dev. 2011, 15, 1328–1335

Organic Process Research & Development
ARTICLE
Scheme 2. First-generation process route to 1
Scheme 3. Incorporation of PTC coupling into process route
The optimized procedure for this process involved dissolving
the substrates in acetonitrile with potassium carbonate, DBU,
KF, and tetraethylammonium chloride and distilling acetonitrile
to remove the water azeotrope. The inclusion of DBU was not
essential to the success of the coupling, however, as in the original
process (Schemes 1 and 2), DBU was found to increase the rate
of the coupling. The coupling was carried out relatively concen-
trated so that the salts raised the boiling point of the reaction
mixture to the range of 90À96 °C. Using an excess of 3 was found
to result in coupling at both the amine and primary alcohol
moieties of bicyclic piperazine 19. Consequently, the stoichio-
metry of 3 was limited to 1.2À1.3 equiv which limited coupling at
the alcohol group to a level of less than 4% in solution. After
heating the reaction at a reflux for 12À24 h, the product (17) was
extracted into aqueous HCl and neutralized for extraction into
ethyl acetate. The product (17) was crystallized as a free base in
76% yield. This process was implemented in our next campaign
(Scheme 3) and provided 9.00 kg of advanced intermediate 17
from 10 kg of 19. The steps enabling the conversion of 17 to 4
were essentially identical to what had been carried out during the
prior campaign.
Development of an Alternative Route to the Bicyclic
Piperazine (2). With the couplings of the side chains developed,
the next objective was to find a more direct route to the bicyclic
piperazine core, 2. The existing route to the racemic core (11) uses
many hazardous reagents with low-yielding steps (Scheme 1).
Even after preparation of 11, a resolution and an epimerization are
necessary to install the requisite stereochemistry. While it is likely
1330
dx.doi.org/10.1021/op2001854 |Org. Process Res. Dev. 2011, 15, 1328–1335

Organic Process Research & Development
ARTICLE
Scheme 4. Retrosynthetic analysis of bicyclic piperazine core 2
Scheme 5. Coupling, cyclization, reduction, and Michael addition to afford piperazine 24
Scheme 6. Tandem phosphate formation, cyclization, and salt formation
that a vendor could handle the sulfur oxidation, epimerization, and
subsequent reduction to furnish 2, we wished to demonstrate a
route avoiding these operations.
additive to LiBH4 to form borane in situ. This approach seemed
to work best in our hands, but the vendor wished to avoid the
use of fluoride salts for the scale up.19 After heating to complete
the reduction, the reaction mixture was quenched into methanol
and distilled to remove some of the boron as trimethylborate.
In this same pot was added triethylamine and benzoic anhydride
which selectively protects the less hindered nitrogen of 22 and
sets up for a Michael addition with acrylate at the unprotected
nitrogen. The protection proceeded during the slow dosing
of benzoic anhydride over 3 h under ambient conditions in
methanol and limited the formation of a bis-amide side product
to a level of no more than 3%.20 Once the benzamide was formed,
tert-butyl acrylate was added to introduce the remaining
carbons for the bicyclic piperazine scaffold. Heating to
50À60 °C enabled the Michael addition to reach completion
within 24À48 h. The yield of 24 on lab scale was often 60À65%
for steps 4À6, but the yield was only 49À58% in the pilot plant
on a 25À50 kg scale.
One option that seemed appealing involved using L-aspartic
acid (20) as the source of chirality for the bicyclic piperazine
(Scheme 4). Coupling L-aspartic acid with glycine and subse-
quent cyclization to afford 21 followed by reduction to give
piperazine 22 has been demonstrated.17 We speculated that 22
could be used to prepare the bicyclic piperazine core (2) by
undergoing addition with an acrylate followed by ring closure
through an intramolecular enolate alkylation. Such a process was
realized as shown in Schemes 5 and 6. The opening steps
followed procedures analogous to those reported to furnish
21.18 For the reduction of 21, LiBH4 was used initially to ensure
complete reduction of the ester, and boraneÀTHF was subse-
quently charged to complete reduction of the two amides to
afford 22 (Scheme 5). As an alternative to purchasing borane, we
had demonstrated the use of boron trifluoride etherate as an
1331
dx.doi.org/10.1021/op2001854 |Org. Process Res. Dev. 2011, 15, 1328–1335

Organic Process Research & Development
ARTICLE
Scheme 7. Scale up of new route to 4
The conversion of 24 to the bicyclic piperazine was accom-
plished by activation of the primary alcohol as a phosphate
followed by intramolecular alkylation of an enolate
(Scheme 6).21 After formation of phosphate 25 and removal of
the triethylammonium salt by filtration, the resulting solution of
25 was added at 0 °C to a solution of LHMDS in hexane.
Surprisingly, no cyclization occurred until THF was charged to
the solution of LHMDS and 25, probably due to a need for
changing the aggregation state of the LHMDS; however, if THF
was added to the LHMDS prior to the charge of 25, then more
side products tended to form.22 The mode of the quench of the
cyclization was essential for dictating the diastereoselectivity in
the product. The optimal quench was carried out by charging the
reaction mixture to chilled water, which typically provides a 20:1
ratio of diastereomers, whereas quenching with addition of water
to the reaction mixture allowed opportunity for epimerization
during the quench. After workup and distillation of solvent to
remove water, the product (26) was crystallized as the D-tartrate
salt in methanol and MTBE to furnish 26a in ∼50% yield from
24. The yield of 26a was much lower on production scale than in
the lab since more epimerization occurred during the preparation
of 21 in the pilot plant due to longer stir times at low pH.17
Fortunately, the isolation of 26a as a salt did enable the purge of
the undesired enantiomer of 26 and the mother liquors of the
filtercake contained the enantiomer of 26 in 93% optical purity,
suggesting that this is an effective diastereomeric salt resolution.
On lab scale about 10À13% of the undesired enantiomer was
present prior to salt formation of 26, but on pilot-plant scale this
amount rose to ∼20%. It was found subsequently that use of a
slightly milder acid, such as TFA, avoided epimerization during
conversion of 30 to 21. As a result of the partial racemization, the
yield of 26a over the three steps was only 39À43% in the pilot
plant. The level of the undesired enantiomer of 26 was 0.7À1.0%
in each batch after isolation of crude 26a. With recrystallization of
26a from methanol and MTBE in 97% yield, the level of the
enantiomer of 26 was reduced to 0.3%, with total chemical
impurities (other than the enantiomer) at a level of 0.3% (by
HPLC area %).
Scheme 8. Preparation of 3
avoid precipitation of the product. After displacing the methanol
with acetone, 2 crystallized as a free base in 59% yield.
In preparation for the coupling between 2 and 3, it was
necessary to convert 3-hydroxybenzo[d]isoxazole (27)24 to 3
using POCl3 (Scheme 8). The addition of phosphoric acid to
POCl3 enabled the reaction to be homogeneous and allowed the
chemistry to proceed at a lower temperature,25 which avoided
side products.26 We treated 3 as a potential irritant since related
compounds have proven to be hazardous in this respect.27 For
this reason we had planned to prepare 3 as a solution to avoid
isolation and use the crude product solution in the subsequent
coupling reaction.
With both 2 and 3 in hand, we proceeded with the most pivotal
bond-forming step. The coupling was carried out by combining 2
with the solution of 3 as well as KF, tetraethylammonium
chloride, and DBU in acetonitrile. The residual solvent intro-
duced with 3 was distilled along with some of the acetonitrile to
remove any water. While on lab scale the use of potassium
carbonate was beneficial to exclude water; we found that
potassium carbonate was unnecessary on manufacturing scale
due to the efficiency of removing water as an azeotrope by
distillation with acetonitrile. After the reaction had been heated
for ∼19 h, the coupling was worked up, and the organic phase
was diluted with acetone to crystallize 4 as the di-HCl salt in 73%
yield and >99% purity on 12-kg scale. Throughout the extraction
sequence of the workup, it was important to maintain the batch at
35 ° C to avoid precipitation of the product prior to the intended
crystallization.
’ CONCLUSIONS
Scale Up of the New Route to 4. After preparing 40 kg of 26a,
this intermediate was reduced to provide the bicyclic piperazine
core, 2 (Scheme 7). Originally Vitride23 was used to reduce 26 to
the N-benzyl-protected core and the benzyl group was subse-
quently cleaved by Pd/C-catalyzed hydrogenolysis to provide 2.
A more efficient route was developed by adding a solution of 26
to chilled LAH in THF. By maintaining a cold reaction mixture,
the ester was rapidly reduced as expected, and the benzamide was
semi-reduced to provide a hemi-aminal which upon quench was
hydrolyzed to give bicyclic piperazine 2 and benzaldehyde. After
the quench, sodium sulfate was added to the mixture to assist
with filtration of the aluminum salts and methanol was added to
In summary we have demonstrated an approach to carrying
out a challenging bond formation between bicyclic piperazine 2
and chlorobenzisoxazole (3) using a fluoride promoter that is
solubilized through the use of a phase transfer catalyst. With this
process improvement, the yield was improved from a highly
variable range of 35À60% to 73À76%. This methodology should
translate well to related systems and illustrates how the combina-
tion of a phase transfer catalyst with fluoride can facilitate
nucleophilic aromatic substitution reactions. In addition, we
developed an alternative approach to preparing 2, which is often
used as a rigid scaffold in pharmaceutical targets. This new
1332
dx.doi.org/10.1021/op2001854 |Org. Process Res. Dev. 2011, 15, 1328–1335

Organic Process Research & Development
ARTICLE
synthesis of 2 avoids the need for a classical salt resolution and
provides the desired stereochemistry using L-aspartic acid to
set all stereochemistry without the need for oxidations or
epimerizations, thus wasting less of 3 in subsequent adjustments
to stereochemistry. A reduction step is required to transform an
ester into a primary alcohol, but this simultaneously cleaves a
benzoyl protecting group that provides a chromophore for
analytical tracking of intermediates as well as regiochemical
control of Michael addition with tert-butyl acrylate. Furthermore,
the undesired stereoisomers are well controlled during the
isolation of 26a as a D-tartrate salt, prior to the reduction step
to afford 2. All of the materials used in the preparation of 2 are
inexpensive commodities, including the L-aspartic acid that is
used to define the stereochemistry. The new synthesis of 2 was
applied to the preparation of 4 on 12-kg scale.
residue in ethanol (∼140 L) was treated with methanesulfonic
acid (23.3 kg). The resulting reaction mixture was heated to
60À70 °C for ∼1 h; then triethylamine (61.5 kg) was added, and
heating was continued at a reflux for 5 h. The reaction mixture
was gradually cooled to 0 °C to crystallize the product and stirred
for 4 h. The crystallized product was isolated by filtration. The
filtercake was washed with ethanol (17.5 L) followed by heptane
(35 L). After drying under vacuum at 50À55 °C, 21 was isolated
as a white, crystalline solid (27.4 kg, 68%). Mp 187À188 °C; 13C
NMR (DMSO-d6, 100 MHz): δ 170.7, 167.7, 166.6, 60.8, 51.4,
45.2, 37.2, 14.7; Anal. Calcd for C8H12N2O4: C, 48.00; H, 6.04;
N, 13.99; Found: C, 47.77; H, 5.97; N, 13.90; [α]22D = +48 ̊ (c =
0.90; methanol). Analysis of the spectroscopic data matched
reported data.17
(S)-tert-Butyl 3-(4-Benzoyl-2-(2-hydroxyethyl)piperazin-
1-yl)propanoate (24). To a suspension of 21 (50.0 kg, 1 equiv)
and THF (250 L) was charged 2.0 M lithium borohydride in
THF (288 L, 2.3 equiv) at 20À25 °C. The reaction mixture was
stirred for 1 h at 20À25 °C. The reaction mixture was heated to
50 °C, held for 1 h and was cooled to 40 °C. To the reaction
mixture was added 1.0 M boraneÀTHF complex (500 L, 2 equiv)
at 40À50 °C over 1.25 h. The reaction mixture was stirred for
3 h at 40À50 °C. The reaction was cooled to ∼0 °C. The
reaction was quenched with methanol (40 L) added over 1 h at
0 °C and was allowed to warm to 20 °C. The reaction mixture
was stirred for an additional h. A methanol solution of HCl
(118 kg of HCl in 1000 L of methanol) was charged at 0À20 °C
over 50 min. The reaction mixture was heated to ∼60 °C for
distillation of the volatiles (∼1250 L distilled). Added methanol
(2500 L) and resumed distillation (distilled ∼2500 L). Triethy-
lamine (425 L, 12 equiv) was added at 20À40 °C. The reaction
was heated to 65À75 °C and held for 1 h. The reaction was
cooled to 20 °C for addition of a solution of benzoic anhydride
(52 kg, 1 equiv) in THF (150 L) over 3 h. After stirring for
another hour at 20 °C tert-butyl acrylate (96 kg, 2.0 equiv) was
added, and the resulting mixture was heated to 50À60 °C for 48 h.
The reaction mixture was concentrated (distilled ∼265 L).
The reaction mixture was diluted with MTBE (1000 L) and
aqueous citric acid (483 kg of citric acid in 1500 L of water). The
mixture was stirred for 1 h, and then the organic layer was
discarded. The aqueous layer was treated with 50% aqueous
NaOH (500 L) to adjust the pH to 13À14. After addition of
toluene (1250 L) to extract the product, the mixture was stirred
for 1 h. After discarding the aqueous phase, the organic layer was
washed with an aqueous brine solution (100 kg of NaCl in 750 L
of water). The organic layer was concentrated (final volume of
∼350 L). The solution of 24 (50 kg, 55% yield) was used directly
in the next sequence without isolation. 13C NMR (CDCl3,
100 MHz): δ 171.9, 170.9, 135.8, 130.0, 128.7, 127.2, 60.5,
57.2, 49.0, 48.0, 46.7, 44.7, 40.5, 33.9, 30.0, 28.3.
’ EXPERIMENTAL SECTION
General Procedures. All materials were purchased from
commercial suppliers and used without further purification. All
reactions were conducted under an atmosphere of nitrogen. 1H
and 13C NMR spectra were recorded on a Varian Unity Plus 400
spectrometer at 400 and 100.6 MHz, respectively. Intermediates
were analyzed by reverse phase HPLC on an Agilent 1100 series
instrument according to the following conditions: column
Atlantis-C18 4.6 mm  150 mm i.d., 3 μm; eluent A, 0.1%
aqueous phosphoric acid; eluent B, acetonitrile; flow rate
1.0 mL/min; wavelength, 287 nm; column temperature, 30 °C;
injection volume, 10 μL; at t = 0 min, 10% eluent B; at t = 25 min,
70% eluent B; at t = 26 min, 85% eluent B; at t = 27 min, 10%
eluent B.
(S)-Diethyl 2-Aminosuccinate Hydrochloride (28). To etha-
nol (150 L) cooled to 0À10 °C was added acetyl chloride (60 kg)
over 1 h at no more than 30 °C. This was added to a solution of
L-aspartic acid (20) (50.0 kg) and ethanol (150 L) over 15 min.
The resulting solution was heated to a reflux for 3 h. The reaction
mixture was concentrated by atmospheric distillation until the
reaction pot reached a temperature of 88 °C (∼275 L of solvent
distilled). Isopropyl acetate (350 L) was added and redistilled
until the reaction pot reached a temperature of 88À89 °C (∼350 L
of solvent distilled). Isopropyl acetate (350 L) was added, and
the resulting mixture was cooled to 22 °C over 1 h. After stirring
for 18 h, the crystals were cooled to 0 °C and stirred an additional
3 h before filtering. The vessel and filtercake were washed with
isopropyl acetate (25 L). The crystals were dried under vacuum
at 40À50 °C to give 28 as white crystals (80.0 kg, 94%). mp
78À80 °C; [α]22 = +11.5 ̊ (c = 1.25, water); 13C NMR
D
(DMSO-d6, 100 MHz): δ 169.7, 168.9, 62.6, 61.6, 49.1, 34.8,
14.6, 14.5. Analysis of the spectroscopic data matched reported
data.17
(S)-Ethyl 2-(3,6-dioxopiperazin-2-yl)acetate (21). A sus-
pension of 28 (52 kg) and triethylamine (26.6 kg) in ethyl
acetate (200 L) was stirred for 30 min at 35À45 °C. To a solution
of 29 (35 kg) in ethyl acetate (90 L) was added carbonyldiimi-
dazole (35.65 kg) as a solution in ethyl acetate (90 L) over 1 h at
20À30 °C. This reaction mixture was stirred at 20À30 °C for
30 min and then was transferred to the suspension of 28 over
30 min. The combined reaction mixture was stirred for 1 h and
then was quenched with water (50 L) and methanesulfonic acid
(74 kg). The aqueous layer was discarded. The organic layer was
concentrated under vacuum at 15À35 °C and displaced with
ethanol so that the final volume was ∼200 L. The product
(7R,9aS)-tert-Butyl 2-Benzoyloctahydro-1H-pyrido[1,2-a]-
pyrazine-7-carboxylate D-Tartrate (26a). Triethylamine (28 kg,
38.5 L, 2 equiv) was added to a solution of 24 (50 kg, 1 equiv)
in toluene (300 L) and cooled to 0 °C. To the solution was added
diphenylchlorophosphate (44.5 kg, 1.2 equiv). The reaction was
stirred for 1 h and was warmed to 20À25 °C. The reaction
developed a suspension and was stirred for 15 h. The reaction
mixture was filtered to remove salts, and the filtercake was rinsed
with toluene (50 L). This solution was charged over 1 h to a
1.0 M solution of LHMDS in hexane (392 kg, 550 L, 4 equiv)
at 0 °C. THF (125 L) was added to the cold reaction mixture over
30 min at 0 °C. The reaction mixture was held at 0À10 °C for 2 h.
1333
dx.doi.org/10.1021/op2001854 |Org. Process Res. Dev. 2011, 15, 1328–1335

Organic Process Research & Development
ARTICLE
The reaction mixture was quenched into aqueous sodium hydroxide
(18.4 kg 30% NaOH and 395 L water) at 0À10 °C. After the
quench, the mixture was warmed to 25 °C and stirred for 1 h.
The mixture was settled and the aqueous layer was discarded.
The organic phase was concentrated under vacuum (to a volume
of ∼150 L) and was diluted with isopropanol (650 L). This solution
was concentrated under vacuum again (to a target volume of
∼100À150 L). Methanol (300 L) was added, and again the
solution was concentrated under vacuum (until ∼100À150 L
remained). The solution was heated to 45À55 °C and a
solution of D-tartaric acid (20.8 kg) in methanol (125 L) was
added to the warm reaction mixture. The reaction mixture was
cooled to 20 °C and crystals formed. The suspension was stirred
for 1 h and then MTBE (250 L) was added. This mixture
was stirred for 1 h and then was cooled to À10 °C to stir for 1 h.
The crystals were filtered and washed with MTBE (50 L).
The product was dried in the vacuum oven at 40À50 °C for 24 h
to afford 26a as white crystals (27.3 kg, 40%). mp 147À150 °C.
Anal. Calcd for C24H34N2O9: C, 58.29; H, 6.93; N, 5.66; Found:
C, 58.07; H, 7.05; N, 5.35. Optical purity was analyzed by
normal phase HPLC on an Agilent 1100 series instrument
according to the following conditions: column Chiralpak AD
4.6 mm Â250 mm; eluent, diethylamine:isopropanol:hexanes
(1:150:1850); flow rate 0.5 mL/min.; wavelength, 240 nm;
column temperature, 30 °C; injection volume, 50 μL; 26 elutes
at 20.1 min and ent-26 elutes at 21.6 min. The level of ent-26 was
0.7À1.0% initially in crude 26a. With recrystallization of 26a
from methanol and MTBE, the level of ent-26 was reduced
to 0.3%.
3-Chlorobenzo[d]isoxazole (3). Phosphorus oxychoride
(22.9 kg, 2 equiv) and 85% phosphoric acid (1.02 L, 0.20 equiv)
were sequentially added to 3-hydroxybenzo[d]isoxazole (27)
(10.0 kg). Pyridine (6.10 L, 1 equiv) was added over 1 h at
20À50 °C. The reaction mixture was heated to 80À85 °C and
held for 12 h. The reaction was cooled to 35À50 °C and
quenched over 1 h into a mixture of MTBE (35 L), heptanes
(45 L), and water (50 L) at 35À50 °C. The reaction mixture was
stirred for 1 h at 35 °C, then cooled to 20À25 °C before
discarding the aqueous layer. Stirred the organic layer with an
aqueous brine solution (5 kg of sodium chloride in 40 L of water)
for 1 h and then discarded the aqueous layer. The organic layer
was transferred to a drum to be used in the next step without
isolation. Assumed theoretical yield of 3 (11.4 kg). The product
may be isolated by concentration of the mixture to yield a low-
melting solid. 13C NMR (CDCl3, 100 MHz): δ 164.0, 150.1,
131.4, 124.7, 121.0, 112.0, 110.7; 1H NMR (CDCl3, 400 MHz):
δ 7.66 (d, 1H, J = 7.9 Hz), 7.61À7.55 (m, 2H), 7.41 (d, 1H, J =
8.7 Hz), 7.38 (t, 1H, J = 7.9 Hz), 5.8 (bs, 1H). Analysis of the
spectroscopic data matched reported data.7
((7R,9aS)-Octahydro-1H-pyrido[1,2-a]pyrazin-7-yl)methanol
(2). A solution of 26 (82.3 mol) in MTBE and THF was distilled
atmospherically to remove water as an azeotrope until the
temperature reached 66À68 °C (all MTBE and water were
removed). The solution of 26 was added over 1 h to a solution of 10%
lithium aluminum hydride solution in THF (69.4 L; 164.62 mol)
diluted with additional THF (194 L) at À5 to 5 °C. The
reaction was stirred for 1 h. The reaction mixture was warmed to
15À20 °C over 30 min once the reaction was determined to be
complete by GC/MS (used a DB-5MS column 30 m  0.25 mm Â
0.25 μm film, flow rate of 1 mL/min, temperature ramp from
100 to 280 °C at a rate of 5 °C/min, 26 elutes at 24 min and 2
elutes at 13 min). A mixture of water (7 L) in THF (63.5 L) was
added to the reaction mixture over 1 h at 025 °C. Gas evolution
was observed (be cautious of hydrogen evolution). After stirring
the mixture 1 h an aqueous solution of sodium hydroxide (1 kg;
25 mol in 7 L water) was added over 15 min followed by water (7 L)
which was added over 15 min. The quenched reaction mixture
was stirred at 15À25 °C for 3 h. Sodium sulfate (40.7 kg) and
methanol (65.1 L) were added to the quenched reaction mixture.
The suspension was filtered after stirring 2 h. The cake was
washed with a mixture of THF (204 L) and methanol (20.4 L).
The combined filtrates were concentrated by atmospheric dis-
tillation at 50À67 °C (until ∼120 L remain). A constant volume
(120 L) was maintained with gradual addition of THF (81.4 L)
while atmospheric distillation continued to ensure complete
removal of methanol (temperature reached 66À68 °C). Solids
crystallized from solution duringthe distillation once the methanol
was removed. Acetone (204 L) was gradually added maintaining a
constant volume (120 L) during the atmospheric distillation to
ensure complete removal of THF. Distillation was continued until
the distillation temperature was between 56 and 58 °C. The slurry
was cooled to 20 °C over 1 h and stirred as a suspension for 4 h
before filtering. The filtered off-white solids were washed with
acetone (3 Â 40.7 L) and dried under vacuum at 40 °C for 16 h to
afford 2 (8.3 kg, 59%). [α]20D = +8.4 ( 0.1 (c = 1.00; CH3OH);
mp 157À158 °C; 13C NMR (CD3OD, 100 MHz): δ 65.1, 62.3,
59.0, 55.4, 50.7, 44.8, 38.7, 28.7, 26.9. Analysis of the spectroscopic
data matched reported data.7
((7R,9aS)-2-(Benzo[d]isoxazol-3-yl)octahydro-1H-pyrido-
[1,2-a]pyrazin-7-yl)methanol Dihydrochloride (4). Potas-
sium fluoride (27.3 kg; 469.9 mol), tetraethylammonium
chloride (21.6 kg; 117.5 mol), DBU (8.6 kg; 56.4 mol), and 2
(8.0 kg; 47.0 mol) are charged to a dry Hastelloy vessel with
acetonitrile (152 L). The suspension was concentrated by
atmospheric distillation (until approximately 65 L remain),
and then cooled to 45 °C. A solution of 3 in MTBE/heptane
(47.6 kg of solution; 7.9 kg of 3; 51.7 mol) was added to the
reaction mixture, and atmospheric distillation was resumed
(until approximately 50 L remained). The thick suspension
remaining was stirred at reflux (90À96 °C) for 15 h, then cooled
to 35À40 °C. Charged MTBE (56 L) and THF (24 L), followed
by water (120 L) and stirred the mixture for 30 min at 35À40 °C.
The lower aqueous layer was discarded. Acetone (80 L) was
added to the organic layer followed by the slow addition of 37 wt%
hydrochloric acid (23.6 L; 28.2 mol) over 30 min at 15À20 °C
to precipitate the product as the dihydrochloric acid salt. The
suspension was stirred 4 h before the solids were filtered, washed
with acetone (40 L), and dried under vacuum at 40 °C for 18 h.
Isolated 4 as white crystals of the dihydrochloride salt (12.3 kg,
(7R,9aS)-tert-Butyl 2-Benzoyloctahydro-1H-pyrido[1,2-a]-
pyrazine-7-carboxylate (26). The tartrate salt, 26a, (40.7 kg;
82.3 mol) was suspended in MTBE (203.5 L) and water (285 L).
An aqueous solution of sodium hydroxide (9.9 kg; 246.9 mol in
41 L water) was added to the suspension over 45 min at
15À22 °C. The mixture was stirred 1 h to dissolve all solids
and then was settled. The light-yellow aqueous phase (pH should
be 14) was discarded. The light-yellow organic layer was stirred
for 1 h with added sodium sulfate (40.7 kg). The mixture was
filtered and the sodium sulfate filtercake was washed with THF
(40.7 L). The wash was added to the filtrate containing 26 in a
MTBE/THF solution. Used the solution of 26 directly in the
reduction to afford 2 without isolation.
73%). [α]20D = À1.5 ( 0.1 (c = 1; H2O); mp >280 °C dec; 13
C
NMR (DMSO-d6, 100 MHz): δ 164.0, 160.7, 131.0, 123.5,
1334
dx.doi.org/10.1021/op2001854 |Org. Process Res. Dev. 2011, 15, 1328–1335

Organic Process Research & Development
ARTICLE
123.4, 115.8, 110.9, 63.6, 61.1, 56.3, 52.3, 50.7, 45.6, 36.9, 26.0,
25.1; 13C NMR (CDCl3, 100 MHz): δ 164.2, 161.3, 129.8, 122.5,
122.4, 116.3, 110.7, 66.3, 60.5, 58.9, 54.4, 53.8, 48.4, 39.2, 29.1,
26.9. Analysis of the spectroscopic data matched reported data.7
(9) Cyclic secondary amines are known to undergo alkylation with
methylene chloride, see: Mills, J. E.; Maryanoff, C. A.; McComsey, D. F.;
Stanzione, R. C.; Scott, L. J. Org. Chem. 1987, 52, 1857.
(10) Compound 3 may be isolated as a solid but is very low melting
(∼35 °C) and is likely a sensitizer or irritant. For these reasons we often
elected to prepare 3 just prior to its use in a coupling and to store it
temporarily as a solution to avoid handling of this material.
(11) Parikh, J. R.; Doering, W. v. E. J. Am. Chem. Soc. 1967, 89, 5505.
(12) To remove boron from the sodium borohydride reduction, 4
was treated with anhydrous HCl in methanol, distilled to a low volume,
and crystallized as the HCl salt. This operation reduced the boron levels
from ∼40,000 ppm down to below 20 ppm.
’ AUTHOR INFORMATION
Corresponding Author
*robert.a.singer@pfizer.com.
’ ACKNOWLEDGMENT
We thank Frank Urban for providing helpful insight to our
process research and development. We thank John Salisbury for
development of HPLC analytical methods. We thank the tech-
nicians of the Pilot Plant (Groton and Sandwich) and Kilo Lab
for their help in the scale up work. We acknowledge PCAS
(France) for the preparation of 26a on 40-kg scale.
(13) In our hands 3 does not appear to undergo oxidative insertion
by Pd readily. One test examined the coupling of 3 with phenyl boronic
acid. Even using this outstanding coupling partner failed to give more
than trace product using electron-rich ligands such as Xphos.
(14) For reviews on fluoride-promoted couplings, see: (a) Clark,
J. H. Chem. Rev. 1980, 80, 429. (b) Yakobson, G. G.; Akhmetova, N. E.
Synthesis 1983, 169. (c) Fringuelli, F.; Lanari, D.; Pizzo, F.; Vaccaro, L.
Curr. Org. Synth. 2009, 6, 203.
(15) Senanayake, C. H.; Hong, Y.; Xiang, T.; Wilkinson, H. S.;
Bakale, R. P.; Jurgens, A. R.; Pippert, M. F.; Butler, H. T.; Wald, S. A.
Tetrahedron Lett. 1999, 40, 6875.
(16) Carpino, L. A.; Sau, A. J. Chem. Soc., Chem. Commun. 1979, 514.
(17) For reports of the preparation of 21 and 22, see: (a) Leeming,
P.; Fronczek, F. R.; Ager, D. J.; Laneman, S. A. Top. Catal. 2000, 13, 175.
(b) Hubbs, J. C. U.S. Patent 5,144,073 A, 1992; (c) Toshihisa, K.;
Tadashi, T. Eur. Pat. Appl. EP 493812, 1992.
(18) The process began with esterification of L-aspartic acid and
coupling with BOC-glycine using CDI followed by deprotection under
acidic conditions and cyclization with neutralization using triethylamine.
It was later realized that the deprotection of the BOC group with MsOH
could lead to some epimerization of the stereocenter, depending on the
temperature control of the MsOH addition as well as the total stir time at
elevated temperature prior to adding triethylamine for cyclization. Use
of TFA or a milder acid than MsOH resulted in no epimerization.
(19) The vendor had concerns of fluoride etching the glass-lined
vessel and would require the use of a steel or Hastelloy vessel which was
not as readily available for the scale up.
’ REFERENCES
(1) The bicyclic piperazine motif has been used in various families of
pharmaceutical targets, see for example: (a) Salama, I.; Schlotter, K.; Utz,
W.; Hubner, H.; Gmeiner, P.; Boeckler, F. Bioorg. Med. Chem. 2006,
14, 5898. (b) Peng, H.; et al. J. Med. Chem. 2004, 47, 6218. (c) Sanner,
M. A.; et al. Bioorg. Med. Chem. Lett. 1998, 8, 725. (d) Bright, G. M.;
Desai, K. A. U.S. Patent 5,122,525, 1992; Chem. Abstr. 1991, 114, 81886;
(e) Saleh, M. A.; Compernolle, F.; Toppet, S.; Hoornaert, G. J. Tetra-
hedron 1994, 50, 1811. (f) Saleh, M. A.; Compernolle, F.; Van den
Braden, S.; De Buysser, W.; Hoornaert, G. J. Org. Chem. 1993, 58, 690.
(g) Compernolle, F.; Saleh, M. A.; Toppet, S.; Hoornaert, G. J. Org.
Chem. 1991, 56, 5192.
(2) The Pd-catalyzed amination has been highly developed, see: (a)
Surry, D. S.; Buchwald, S. L. Angew. Chem., Int. Ed. 2008, 47, 6338.
(b) Schlummer, B.; Scholz, U. Adv. Synth. Catal. 2004, 346, 1599.
(c) Hartwig, J. F.; Shekar, S.; Shen, Q.; Barrios-Landeros, F. In The
Chemistry of Anilines; Rappoport, Z., Ed.; Wiley: New York, 2007; p 455.
(d) Hartwig, J. F. In Modern Amination Methods; Ricci, A., Ed.;
Wiley-VCH: Weinheim, 2000; p 195.
(20) The slow dosing of the benzoic anhydride ensures that no more
than about 3% of bis-benzoylated amide side product forms. Adding the
benzoic anhydride over 1 h resulted in ∼7% or more of the bis-
benzoylated amide forming.
(21) Aside from the phosphate, we had prepared the mesylate,
chloride, and tosylate derivatives of 24; however, none of these cyclized
as readily as the phosphate.
(22) In addition to controlling the addition of THF to LHMDS and
25, it is important to maintain a lower temperature during the addition of
the THF when cyclization occurs.
(23) Vitride is also sold as Red-Al, which is a 65 wt % solution of
bis(2-methoxyethoxy)aluminum hydride in toluene.
(24) While 27 is commercially available, we typically prepared this
building block as follows: methyl salicylate was treated with hydro-
xylamine hydrochloride in aqueous base to obtain a hydroxamic acid
adduct which is crystallized upon treatment with aqueous HCl. This
intermediate is cyclized to 27 by slow addition of CDI in THF/
acetonitrile with heating and is crystallized from water/acetonitrile by
addition of aqueous HCl.
(25) The addition of phosphoric acid to POCl3 for improving the
efficiency of the conversion to the chloride has been reported
previously, see: Andersen, K.; Begtrup, M. Acta Chem. Scand. 1992, 46, 1130.
(26) At higher temperatures (100À120 °C) chlorination of the
benzisoxazole ring occurred, and these side products were found to be
difficult to purge downstream in the process.
(3) Electron-deficient aryl halides are more susceptible to undergo
oxidative addition and SNAr. Many heterocycles are intrinsically acti-
vated to these pathways as well, see: Joule, J. A.; Mills, K.; Smith, G. F.
Heterocyclic Chemistry, 3rd ed.; Chapman and Hall: New York, 1995.
(4) Some heterocyclic substrates are more challenging to optimize in
cross-coupling approaches due to inhibition of the Pd catalyst from the
substrate or product. A number of reports have focused on this
challenging class of substrates, see: Hartwig, J. F. J. Am. Chem. Soc.
2008, 130, 6586. Anderson, K. W.; Tundel, R. E.; Ikawa, T.; Altman,
R. A.; Buchwald, S. L. Angew. Chem., Int. Ed. 2006, 45, 6523. Charles,
M. D.; Schultz, P.; Buchwald, S. L. Org. Lett. 2005, 7, 3965. Yin, J. J.;
Zhao, M. M.; Huffman, M. A.; McNamara, J. M. Org. Lett. 2002, 4, 3481.
Shen, Q.; Shekhar, S.; Stambuli, J. P.; Hartwig, J. F. Angew. Chem., Int. Ed.
2005, 44, 1371. Shen, Q.; Ogata, T.; Hooper, M. W.; Utsunomiya, M.;
Hartwig, J. F. J. Org. Chem. 2003, 68, 2861. Wolfe, J. P.; Tomori, H.;
Sadighi, J. P.; Yin, J.; Buchwald, S. L. J. Org. Chem. 2000, 65, 1158.
Wagaw, S.; Buchwald, S. L. J. Org. Chem. 1996, 61, 7240.
(5) Aryl chlorides are not as facile to undergo oxidative addition to Pd
catalysts as aryl bromides or iodides, especially electron-rich systems, see:
Littke, A. F.; Fu, G. C. Angew. Chem., Int. Ed. 2002, 41, 4176.
(6) Bright, G. M.; Brodney, M. A.; Wlodecki, U.S. Patent 7,345,038
B2, 2008; Chem. Abstr. 2004, 141, 296048.
(7) For experimental procedures on the route leading up to 4, see:
(a) Urban, F. J. U.S. Patent 5,719,286 A, 1998; Chem. Abstr. 1994, 121,
108844; (b) Urban, F. J. J. Heterocycl. Chem. 1995, 32, 857. A related
route also leading to 17 has been disclosed, see: Urban, F. J.; Breiten-
bach, R.; Murtishaw, C. W.; Vanderplas, B. V. Tetrahedron: Asymmetry
1995, 6, 321.
(27) We have had experience handling 3-chlorobenzo[d]isothiazole
which is a known irritant sold by numerous suppliers and suspected that
3-hydroxybenzo[d]isoxazole may possess similar hazards in handling.
(8) For this campaign, Carbogen had prepared 11 for us.
1335
dx.doi.org/10.1021/op2001854 |Org. Process Res. Dev. 2011, 15, 1328–1335