Development of a safe and scalable oxidation process for the preparation of 6-hydroxybuspirone: Application of in-line monitoring for process ruggedness and product quality

Article abstract of DOI:10.1021/op049918r

The development of a safe and scalable oxidation process for the hydroxylation of the azapirone psychotropic agent buspirone (1) to furnish 6-hydroxybuspirone (2) is described. A mechanistic understanding of how key process factors affected product quality led to the successful application of FTIR as a process analytical technology (PAT) tool. This enabled real time quality assurance and the development of an effective and efficient manufacturing process. The identification of impurities and the development of recrystallization methods to provide active pharmaceutical ingredients (API) with optimal purity will also be addressed.

Full text of DOI:10.1021/op049918r

Organic Process Research & Development 2004, 8, 616−623
Development of a Safe and Scalable Oxidation Process for the Preparation of
6-Hydroxybuspirone: Application of In-Line Monitoring for Process Ruggedness
and Product Quality
Daniel J. Watson,* Eric D. Dowdy, Jeffrey S. DePue, Atul S. Kotnis, Simon Leung, and Brian C. O’Reilly
Bristol-Myers Squibb Pharmaceutical Research Institute, Department of Process Research and DeVelopment,
One Squibb DriVe, New Brunswick, New Jersey 08903
Abstract:
Process research and development were therefore pre-
sented with the challenge to develop a safe, robust, and
scalable process to enable the multikilogram production of
2 to support toxicology studies, formulation development,
and clinical trials.
The development of a safe and scalable oxidation process for
the hydroxylation of the azapirone psychotropic agent buspirone
(1) to furnish 6-hydroxybuspirone (2) is described. A mecha-
nistic understanding of how key process factors affected product
quality led to the successful application of FTIR as a process
analytical technology (PAT) tool. This enabled real time quality
assurance and the development of an effective and efficient
manufacturing process. The identification of impurities and the
development of recrystallization methods to provide active
pharmaceutical ingredients (API) with optimal purity will also
be addressed.
Results and Discussion
Initial Experimentation. One of the classical oxidation
processes for the conversion of a ketone to the corresponding
R-hydroxyketone was first published in 1962 by Barton et
al..5 It involves the formation of the enolate followed by
oxidation employing oxygen. After isolation, the intermediate
hydroperoxide is then reduced with zinc dust in acetic acid
to furnish the desired alcohol. Later, in 1968, Gardner and
co-workers6 reported that the oxidation and reduction could
be achieved in a one-pot process by performing the reaction
in the presence of trialkyl phosphites.
Early experiments for the R-hydroxylation of 1 employed
KHMDS in THF for enolate formation, followed by sparging
with air in the presence of 1 equiv of triethyl phosphite at
low temperature (-78 °C). Conversion, monitored by
analytical HPLC, showed 74% 2, 20% starting material 1,
and varying amounts of several impurities. The variation of
the impurities formed throughout processing presented a
formidable challenge during process optimization of this drug
candidate. Purification using silica gel chromatography
provided 2 in a moderate yield of 44%.
Introduction
Buspirone (1), manufactured and marketed by Bristol-
Myers Squibb as Buspar (hydrochloride salt of buspirone),
is employed for the treatment of anxiety disorders and
depression.1 Studies show Buspar is extensively metabolized
in the body to several compounds.2 One such metabolite was
identified as the hydroxylated derivative, (()-6-hydroxy-
buspirone (2), which, like buspirone, demonstrates a strong
affinity for the human 5-HT1A receptor (Figure 1).3
Process Optimization. Typically this reaction was run
in THF at concentrations of 20 mL/g. This was the preferred
solvent owing to the solubility of the starting material,
product, and the intermediate enolate at low temperatures.
The THF process also provided a cleaner reaction profile
compared to other solvents. Research efforts therefore
focused on three areas: phosphite stoichiometry, the workup
and isolation protocol, and the control of enolate 3 formation.
Triethyl Phosphite Concentration. Product quality was
shown to be dependent on the phosphite stoichiometry. In
the absence of triethyl phosphite, a 1:1 mixture of starting
Figure 1.
One attractive strategy for the synthesis of 6-hydroxy-
buspirone involves a one-step conversion from buspirone,
owing to its long-term availability as an inexpensive starting
material. Several synthetic approaches to 6-hydroxybuspirone
have been disclosed by our colleagues in Drug Discovery;
however, these were not amendable to development because
of reagent costs and purification challenges.4
(1) Bailey, E. J.; Barton, D. H. R.; Elks, J.; Templeton J. F. J. Chem. Soc.
1962, 1578.
(2) Gardner, J. N.; Carlon, F. E.; Gnoj, O. J. Org. Chem. 1968, 3294.
(3) New, J. S. Med. Res. ReV. 1990, 10, 283-326.
(4) Mayol, R. F.; et al. Clin. Pharmacol. Ther. 1985, 37, 210
(5) Mayol, R. F. U.S. Patent 2000/6150365.
(6) (a) Yevich, J. P.; Mayol, R. F.; Li, J.; Ruediger, E. H. US Patent 2003/
0069251 A1. (b) Yevich, J. P.; Mayol, R. F.; Li, J.; Yocca, F. WO 03/
009851 A1; (c) Yevich, J. P.; Mayol, R. F.; Li, J.; Yocca, F. U.S. Patent
2003/0022899 A1.
* Corresponding author. E-mail: daniel.watson@bms.com.
616
•
Vol. 8, No. 4, 2004 / Organic Process Research & Development
10.1021/op049918r CCC: $27.50 © 2004 American Chemical Society
Published on Web 06/12/2004

Scheme 1. Impurity formation in the absence of triethyl phosphate
material 1 and the corresponding keto acid 6 was observed
presumably through the pathway shown in Scheme 1.
Early laboratory experiments for the oxidation of 1 to 2
therefore employed 1.1 equiv of triethyl phosphite resulting
in improved conversion, but the product was still contami-
nated with a number of unidentified impurities. A systematic
study of the phosphite stoichiometry revealed that a charge
of triethyl phosphite greater than 3.0 equiv provided reactions
with significantly cleaner profiles but that further increase
of the level of phosphite led to higher losses of product
during the isolation. The decision was therefore made to
perform this reaction with 3.5 equiv of triethyl phosphite to
balance purity and product loss.
One additional concern with regard to safety was the
formation of presumably unstable hydroperoxide 4 as an
intermediate in this process. The presence of an excess of
the reductant, triethyl phosphite, should ensure that the
concentration of 4 did not build up throughout the course of
the reaction. In situ testing for peroxides throughout the
oxidation process with peroxide test strips indicated low
levels at or below 1 mg/L suggesting the peroxide intermedi-
ate 4 is immediately consumed by the phosphite.
Evolution of the Isolation. The initial procedure for the
workup and isolation of 2 involved neutralization, phase
separation, back-extraction with ethyl acetate, brine washes,
and isolation by precipitation with heptanes. This expedient
process provided good product recovery; however, most
impurities coprecipitated leading to an inferior quality
product. We subsequently found that an exchange of the THF
reaction solvent with ethyl acetate after neutralization,
followed by slow addition of heptanes and seeding to effect
crystallization, gave crystalline material with improved
purity. The drawbacks to this crystallization procedure were
high mother liquor losses, on the order of 15 mol %, and a
tendency for the product to oil before crystallization. A more
robust protocol was developed based on earlier experiences
learned while developing a process for the HCl salt of 2.
The mixture was quenched, neutralized by the addition of
aqueous HCl, and diluted with MTBE. This provided several
advantages to the previous workups. The addition of MTBE
afforded better phase separations during the back-extractions
and aqueous washes, and the low boiling point of MTBE
enabled a facile exchange of the MTBE/THF organic layer
to isopropyl alcohol until the IPA/THF ratio was >25:1 and
KF < 1.0% by GC and Karl-Fisher titration, respectively.
Upon cooling and seeding, 2 crystallized reliably from the
mixture. The mother liquor losses during this isolation were
generally lower (6-8 mol %) than those by other direct
crystallization techniques and provided product with an
improved purity profile.
Identifying and Minimizing Impurities. Identification
of process impurities and a mechanistic understanding of their
origin often lead to improved process conditions and a quality
product. Impurities associated with this process are shown
in Figure 2. Two major impurities common to this transfor-
mation were the starting material 1 and the over-oxidized
product 6,10-dihydroxybuspirone 8. Excess base produced
impurity 8 in a 10:1 ratio of isomers favoring the trans-
diol. Experimentally, we observed that the amount of excess
base correlated well with the amount of diol 8 produced (i.e.,
5% excess base resulted in 5% diol production). On the other
hand, an undercharge of the base gave an incomplete reaction
leading to product contamination with 1. Since the amount
of base is critical to both yield and quality, we titrated the
base in triplicate and measured water in all the THF solutions
to ensure an accurate charge.
The isolation procedure incorporated a rugged crystal-
lization of the product directly from the process stream.
Occasionally, however, batches were produced where the
levels of these known impurities (Figure 2) were greater than
those observed in the qualifying toxicology batch. The
process therefore required a recrystallization to ensure that
API of the desired purity could be reliably obtained.
Determination of the solubility of these impurities in various
solvents using automated instrumentation enabled a rapid
development of rework procedures (Figure 3).
After screening several solvents, anisole appeared to
provide the greatest difference in solubility between product
2 and the major impurities starting material 1 and diol 8. To
purge the diol impurity 8, a reverse crystallization was
developed wherein the material was dissolved in anisole at
90 °C, cooled to ambient temperature, and seeded with 0.25
wt % diol seeds to crystallize 8. After removal of 8 by
filtration, the anisole was concentrated to half its initial
Vol. 8, No. 4, 2004 / Organic Process Research & Development
•
617

Figure 2. Structures of impurities identified in the oxidation of buspirone (1).
Figure 3. Solubility measurements of 2 and key process impurities.
volume and heptane was charged slowly to crystallize the
product 2 with excellent purity. A recrystallization protocol
from isopropyl alcohol was also developed that could purge
up to 4% starting material 1, thus providing a lower limit to
the amount of base charged. It became apparent that a slight
undercharge of the base was the preferred route as the
removal of the diol impurity 8 was time-consuming and both
labor- and equipment-intensive. Consequently, tight control
of the base titer and water content proved crucial to the
success of this reaction to avoid undercharging or overcharg-
ing the base (a typical base charge was 0.95-0.97 equiv).
Pilot Plant Campaign. The optimized process employed
NaHMDS as the preferred base with triethyl phosphite as
the reductant owing to a combination of cost, availability,
and convenience. To obtain a clean reaction profile, the
optimal triethyl phosphite charge and temperature were 3.5
equiv and -70 °C, respectively. In addition, lab experiments
showed the rate of oxidation was significantly faster using
pure oxygen instead of air. However, from a safety perspec-
tive this could not be readily scaled up owing to our decision
to limit the level of oxygen in the headspace to less than
6%. To keep the headspace oxygen level low, we sparged
618
•
Vol. 8, No. 4, 2004 / Organic Process Research & Development

Figure 4.
the reaction with air and employed a 3:1 volumetric ratio of
nitrogen to the air sparge. This effectively kept the oxygen
level in the headspace at or below 5.5% which was monitored
using an explosion proof oxygen monitor (N-TRON).7 When
the reaction was complete, the oxygen sparge was replaced
with a nitrogen sparge until the off-gas was below 0.5%
oxygen.
air and moisture sensitivity. As a result, each lot of phosphite
was use tested prior to running the process on scale.
In-Process Monitoring Using React-IR. Subsequent
work in the laboratory focused on introducing Process
Analytical Technology (PAT)8 using an in-line monitoring
tool to achieve process ruggedness and consistently ensure
acceptable and reproducible product quality. Specifically, in
situ FTIR was employed to monitor the deprotonation of 1.
This enabled a more precise determination of the endpoint
for enolate formation. The FTIR profile for the deprotonation
of 1 is shown in Figure 4. Signals at 1677 cm-1 and 1627
cm-1 correspond to the starting material 1 and the enolate
3, respectively.
The first pilot-plant batch with a 10 kg input of 1
employed 0.95 equiv of 1 N NaHMDS in THF, providing
7.35 kg of 2 in 71% yield and >98 HPLC area percent (area
%) purity. The next batch again employed 0.95 equiv of base
(a different lot than the base used in the first batch) with a
12.46 kg input of 1 providing 8.96 kg of 2 in 69% yield.
However, this material did not meet specifications for purity
due to the formation of undesired side products, namely the
6,10-dihydroxybuspirone (8) (13 area %). This suggested an
approximated additional 13% (1.08 equiv) of base had been
charged. Among other issues that accounted for the discrep-
ancy in the overcharge of base, it was theorized that the
NaHMDS (1 N in THF, stored in a pressure bomb) had
crystallized from the THF resulting in a nonhomogeneous
solution, therefore providing a nonrepresentative sample for
the titration. This would result in a lower apparent titer and
an overcharge of the base. The rework from anisole purged
the diol 8 affording material in excellent purity (>99 area
%) in 70% recovery. This, however, demonstrated that the
process did not possess suitable robustness. The initial
optimized procedure (vida infra) relied on two corrections
to estimate the amount of base needed to deprotonate 1,
namely adventitious water and base titer. Although this
methodology was sufficient to produce several high-quality
batches of 2, there was the occasional failure as evidenced
above, as excess base could easily be added, leading to the
production of 6,10-dihydroxybuspirone (8). This impurity
was difficult to purge and therefore best controlled with
processing conditions. Additionally, it was observed in the
laboratory that certain lots of triethyl phosphite accounted
for lower conversions in the oxygenation reaction due to its
Incremental charges of base show a decrease of the SM
1 peak at 1677 cm-1 and a corresponding increase in the
enolate 3 peak at 1627 cm-1 (Figure 5).
Other interesting observations were (i) deprotonation of
the buspirone (1) was rapid and usually complete within 5
min after addition of the base, (ii) the enolate anion 3 was
stable at or below -25 °C over a 12 h period (temperatures
above -25 °C led to decomposition of the enolate [-15 °C,
3.6% decomposition of the enolate 6; -5 °C, 7.6%; 5 °C,
18%; and 20 °C, 25%]), and (iii) addition of triethyl
phosphite before addition of the base had no impact on the
IR signal.
To ensure that excess base was not present, an improved
process for the formation of enolate 3 was developed.
Monitoring with FTIR, we charged the base until the signal
for the starting material (1677 cm-1) no longer declined. This
indicated that we had achieved complete consumption of
the starting material. A solution of starting material in THF
was then charged back to the vessel in incremental amounts
until the signal increased therefore indicating a slight excess
of the starting material. Quantifying the signals while
correcting for changes in concentration enabled a back charge
to provide a 1-3% excess of starting material. It is important
to note that monitoring the deprotonation by FTIR allowed
for Variations in the base titer, water content, and phos-
phite quality. This led to the successful production of sev-
(7) If the oxygen off-gas concentration had crept near 6%, the oxygen sparge
would have been replaced with a nitrogen sparge and the combination of
the two nitrogen streams would have quickly reduced the oxygen concentra-
tions. It should also be noted that one can purchase custom filled cylinders
containing 5% oxygen in nitrogen.
(8) FDA Draft Guidance for Industry; PAT-A Framework for Innovative
Pharmaceutical Manufacturing and Quality Assurance, August 2003.
Vol. 8, No. 4, 2004 / Organic Process Research & Development
•
619

Figure 5.
Figure 6. IR signal of buspirone indicating the backcharges of the starting material to reach desired target of 1-3% excess buspirone.
eral batches of 2 in the laboratory with low levels of diol
impurity 8.
employment of the FTIR to control the enolization endpoint
proVided lower leVels of diol impurities and therefore
exemplified its effectiVeness as an in-process control for
process optimization.
Modified Workup Procedure: Understanding the Fate
of the Triethyl Phosphite and Its Impact on Product
Recovery. During the pilot plant campaign, losses of 2 to
the mother liquor during the crystallization was about 10-
15%. Experimentally, we have traced this rather high loss
to the use of excess triethyl phosphite during the oxidation
step. The excess triethyl phosphite (theoretical 2.5 equiv)
remaining after the completion of the reaction hydrolyzes
during the workup and crystallization protocol to diethyl
This process was ultimately implemented in the pilot
plant. Buspirone (1) (13.42 kg) was converted into 2 (8.96
kg) in 69% yield and >99 area %. For this plant run, we
used the same lot of base that had caused problems in the
previous batch (vida supra). NaHMDS (1.01 equiv of
apparent titer of 1 M in THF) was charged, followed by a
total back charge of 12 mol % 1. This provided a stable,
positive buspirone signal corresponding to a 1.3% excess of
starting material by Process-IR (Figures 6 and 7). This
generated API with low levels of starting material and diol.
Fortunately, despite the ambiguous base concentration, the
620
•
Vol. 8, No. 4, 2004 / Organic Process Research & Development

Figure 7. IR data for the oxidation process.
phosphite [(EtO)2P(OH)], and we have determined that 2 has
a very high solubility in both diethyl phosphite and diethyl
phosphite/IPA mixtures. The amount of diethyl phosphite
present in the crude reaction mixture corresponds to ap-
proximately 1 mL/g of 2.
Conclusions
In summary, we have demonstrated a practical and safe
method for the aerobic oxidation of azapirones. The process
is suitable for implementation on pilot plant scales. Opti-
mization of the process conditions and workup provided
cleaner reaction profiles and lower losses to the waste
streams, respectively. Further process control afforded by
the use of the in-line FTIR monitoring tool ultimately ensured
product quality and prevented the need for batch reprocess-
ing.
To overcome these high losses to the mother liquor, an
alternative workup and isolation protocol was developed.
When the oxidation is complete, the reaction mixture is
treated with acid to lower the pH to approximately 2.0,
whereupon the mixture is heated to hydrolyze the residual
triethyl phosphite to phosphorous acid and monoethyl phos-
phorous acid. Subsequent neutralization with base, followed
by aqueous extraction, removes the phosphorous acids.
Solvent exchange of tetrahydrofuran with 2-propanol at
reduced pressure followed by crystallization afforded 2 in
yields of 70% with excellent purity (>99 area %). Using
this new protocol, the losses to the mother liquor returned
to their initial values observed in the laboratory of 6-8%.
Experimental Section
General. NaHMDS (1 N) and HCl (1 N) are com-
mercially available and were used as is. THF, isopropyl
alcohol, anisole, and methyl tert-butyl ether were used
without further purification. Moisture content was deter-
mined by coulometric titration on a Mitsubishi CA-06
moisture meter on a weight/weight basis. Proton and carbon
Vol. 8, No. 4, 2004 / Organic Process Research & Development
•
621

NMR were run on a Bruker AC-300 spectrometer at 300
MHz for proton and 75 MHz for carbon in DMSO-d6. In
situ measurements were performed using a React-IR spec-
trometer (instrument model: Mettler-Toledo AutoChem
ReactIR 1000 and ProcessIR) equipped with an immersion
probe having a diamond internal reflection element. The
probe was purged with nitrogen. The titration of the
NaHMDS solution was performed in triplicate with 1,3-
diphenylacetone tosylhydrazone. HPLC was run under the
following conditions: column, YMC ODS-AQ 4.6 × 150
mm2 s-3 µ; column temperature, ambient; flow rate, 1.0
mL/min; 236 nm (8 nm band); injection volume, 10 µL;
mobile phase A, 0.05% NH4OAc in 80:20 H2O/MeOH;
mobile phase B, 0.05% NH4OAc in 75:20:5 ACN/MeOH/
H2O; recording time, 25 min. Gradient program: initial (30%
B), 15 min (55% B), 17 min (100% B), 20 min (100% B),
22 min (30% B). RRT (retention times in minutes): bus-
pirone 1 (starting material), 1.35 (typical Tr ) 12.65 min);
6-hydroxybuspirone 2 (BMS-528215), 1.00 (typical Tr )
9.38 min); lactone 13, 0.80 (typical Tr ) 7.47 min);
6,10-dihydroxybuspirone 8, 0.74 (typical Tr ) 6.91 min).
HPLC sample preparation: (1) For in-process samples of 2,
dilute ∼10 µL of reaction mixture with ∼1.5 mL of mobile
phase B; (2) for isolated 2, dissolve ∼0.1 mg of solid in 1.5
mL of ACN (note: sonication may be necessary for
solvation)
Procedures. A. 6-Hydroxybuspirone (2). Buspirone (8-
[4-[4-(2-pyrimidinyl)-1-piperazinyl]butyl]-8-azaspiro(4.5)-
decane-7,9-dione) (1) (10 kg, 25.9 mol) was charged to a
500 L vessel under nitrogen. Tetrahydrofuran (142.4 kg,
161 L) was charged, and the mixture was agitated at am-
bient temperature until homogeneous. The water content
of the solution was measured by Karl-Fischer titration
(KF ) 0.009%) (this enabled the base charge to be ad-
justed, compensating for the water present in the solu-
tion). Triethyl phosphite (15.1 kg, 90.9 mol, 15.6 L, 3.5
equiv) was added, and the mixture was cooled to <-60 °C.
Sodium bis(trimethylsilyl)amide, 1.0 M in THF (22.9 kg,
25.4 mol, 25.4 L, 0.98 equiv) was charged to the mixture
at such a rate so as to maintain the temperature at less
than -60 °C. Ultra zero grade air was sparged into the
reaction mixture. The initial rate of sparging was con-
trolled so as to maintain the temperature of the reaction
mixture at less than -60 °C. The sparging was continued
until the reaction was complete by HPLC (<3% buspirone).
Methyl tert-butyl ether (16.1 kg, 21.7 L) was added fol-
lowed by hydrochloric acid (52.3 kg, 51.8 mol, 51.8 L,
1.0 M), and the solution was warmed to ambient temper-
ature. The pH was adjusted to 6.0-7.0 (preferred range pH
6.5-6.9) using 1.0 M hydrochloric acid and/or 0.5 M
Na3PO4 solution (final pH 6.6). The phases were separated,
and the upper organic phase was washed twice with 25 wt
% brine (30 kg, 27.3 L). The rich organic layer was then
solvent swapped into isopropyl alcohol using three charges
of 79.0 kg of IPA providing an IPA/THF ratio of 38:1 and
KF ) 0.37%. A final concentration of 8-10 mL/g was
achieved. The solution was cooled to ambient temperature
over 4 h to crystallize the material. The crystalline slurry
was then filtered by centrifugation, and the wet cake was
washed twice with isopropyl alcohol (2 × 7.7 kg, 9.8 L)
and dried at 40 °C under vacuum to provide 6-hydroxybus-
pirone (2) (7.35 kg, 18.3 mol, 71%). Mp 109.5 °C; 1H NMR
(CDCl3, ppm) δ 1.23-1.20 (m, 1H), 1.41-1.36 (m, 1H),
1.76-1.54 (m, 10H), 2.09-2.03 (m, 1H), 2.55-2.40 (m,
7H), 2.80 (d, J ) 17.5 Hz, 1H), 3.55 (s, 1H), 3.83-3.72 (m,
5H), 4.20 (s, 1H), 6.48 (t, J ) 4.7 Hz, 1H), 8.30 (d, J ) 4.7
Hz, 2H). 13C NMR (CDCl3, ppm): δ 24.3, 25.9, 26.2, 26.5,
30.6, 36.2, 40.5, 43.7, 44.8, 44.9, 53.2, 58.2, 72.8, 74.3,
109.6, 157.0, 160.9, 170.1, 173.0, 174.4. Anal. Calcd for
C21H31N5O3: C, 62.82; H, 7.78; N, 17.44. Found: C, 61.41;
H, 7.52; N, 17.13.
B. In Situ Monitoring with FTIR for the Conversion
of Buspirone (1) to 6-Hydroxybuspirone (2). Buspirone
(8-[4-[4-(2-pyrimidinyl)-1-piperazinyl]butyl]-8-azaspiro(4.5)-
decane-7,9-dione) (1) (246.5 g, 639.6 mmol) was charged
to a 12 L flask equipped with a mechanical stirrer and a
React-IR probe under inert gas. Tetrahydrofuran (4.383 kg,
60.8 mol, 4.930 L) was charged, and the mixture was agi-
tated at ambient temperature until homogeneous. Triethyl
phosphite (371.9 g, 2.238 mol, 383.8 mL, 3.5 equiv) was
added, and the mixture was cooled to -68 to -75 °C. The
mixture was agitated at this temperature for at least 10 min
to allow the React-IR signal to stabilize. Sodium bis-
(trimethylsilyl)amide, ∼1.0 M in THF (1.00 equiv) was
charged to the mixture at such a rate to maintain the
temperature at less than -60 °C. Small amounts of so-
dium bis(trimethylsilyl)amide were charged to the mixture
until the IR signal for buspirone reached a minimum,
indicating complete deprotonation of buspirone. Additional
buspirone in THF (10-20 mL/g) was then charged to the
reaction mixture in small increments until the IR signal
indicated a 0.5%-5% excess of buspirone. Air was sparged
into the reaction mixture, controlling the initial rate of
sparging so as to maintain the temperature of the reaction
mixture at less than -60 °C. The sparging was continued
until the reaction was complete as indicated by HPLC.
Methyl tert-butyl ether (384.8 g, 4.365 mol, 520.0 mL) was
added followed by 1 M hydrochloric acid (1350.8 g, 1.328
mmol, 1328 mL), and the solution was warmed to ambient
temperature. The pH was adjusted to between 6.5 and 6.9
using hydrochloric acid and Na3PO4. The phases were
separated, and the organic phase was washed twice with brine
(542.3 g, 493 mL). The solvent of the rich organic layer
was then replaced by isopropyl alcohol, and the solution was
cooled to crystallize the reaction product (there is an option
to seed with 0.01-5% 6-hydroxybuspirone (2) at ap-
proximately 54-56 °C). The crystalline slurry was then
filtered, and the wet cake was washed with isopropyl alcohol
and dried to provide 6-hydroxybuspirone 2 (220.0 g, 82%),
mp 109.5 °C.
C. Crystallization of 6-Hydroxybuspirone (2) to Re-
duce 6,10-Dihydroxybuspirone (8) and Buspirone (1).
6-Hydroxybuspirone (2) (8.0 kg, 20.8 mol) was slurried with
anisole (80 L, 10 L/kg, 10-15 L/kg may be used). The
mixture was heated to 80-100 °C and stirred to obtain a
clear solution. The solution was then cooled to 75-85 °C
622
•
Vol. 8, No. 4, 2004 / Organic Process Research & Development

before 6,10-dihydroxybuspirone (8) seeds (20 g, 0.25 wt %,
0-2 wt % may be used) were added. The mixture was then
cooled to ambient temperature over 2-6 h and stirred
overnight. The resulting slurry was filtered, and the filtrate
was concentrated to approximately half its initial volume.
Heptane (1 volume relative to the initial volume of anisole)
was then added over ∼5 h, and the resulting slurry was stirred
at ambient temperature overnight. The slurry was filtered,
and the resulting filter cake was washed with heptane. After
drying, 5.56 kg of 6-hydroxybuspirone (2) was obtained
(70% recovery). The level of diol 8 was reduced from 13
area % to 0.6 area % and buspirone (1) from 1.35 to 0.17
area %.
Acknowledgment
We thank Shufang Niu and Andy Wang for analytical
support, Frank Okuniewicz and Dr. Steve Wang for discus-
sions on safety-related issues, and Dr. Chenkou Wei for
advice on the art of crystallizations. We also thank Dr.
Wendel Doubleday, Dr. Edward Delaney, and Dr. David
Kronenthal for reviewing the manuscript and Dr. Brian Kaller
for helpful discussions. We extend our sincere appreciation
to the BMS Glass Plant and Pilot Plant Operations for their
assistance with the scale-up campaign.
Received for review April 20, 2004.
OP049918R
Vol. 8, No. 4, 2004 / Organic Process Research & Development
•
623