Commercial synthesis of (S,S)-reboxetine succinate: A journey to find the cheapest commercial chemistry for manufacture

Article abstract of DOI:10.1021/op200181f

The development of a synthetic process for (S,S)-reboxetine succinate, a candidate for the treatment of fibromylagia, is disclosed from initial scale-up to deliver material for registrational stability testing through to commercial route evaluation and subsequent nomination. This entailed evaluation of several alternative routes to result in what would have been a commercially attractive process for launch of the compound.

Full text of DOI:10.1021/op200181f

ARTICLE
pubs.acs.org/OPRD
Commercial Synthesis of (S,S)-Reboxetine Succinate: A Journey To
Find the Cheapest Commercial Chemistry for Manufacture
Stewart T. Hayes,* Georges Assaf, Graham Checksfield, Chi Cheung, Doug Critcher, Laurence Harris,
Roger Howard, Suju Mathew, Christian Regius, Gemma Scotney, and Adam Scott
Chemical Research and Development, Pfizer Inc., Sandwich Laboratories, Sandwich, Kent, CT13 9NJ, United Kingdom
ABSTRACT: The development of a synthetic process for (S,S)-reboxetine succinate, a candidate for the treatment of fibromylagia,
is disclosed from initial scale-up to deliver material for registrational stability testing through to commercial route evaluation and
subsequent nomination. This entailed evaluation of several alternative routes to result in what would have been a commercially
attractive process for launch of the compound.
commercial synthetic route to an active pharmaceutical
ingredient (API) needs to meet many criteria in terms of
A
safety, the environment, legality, control, and throughput.1 Cost
is also an important factor, and a commercial route to an API that
addresses the cost of goods is no longer sufficient to compete in
the long term with patent exclusivity and generic competition. It
is therefore important to develop the cheapest commercial route
as early as possible in the lifecycle of a compound, where even
modest cost improvements can result in large savings in the cost
of manufacturing a drug substance.2 Savings can also be realized
in the short term through cheaper raw material costs, higher
utilization factors of input materials, and shorter manufacturing
times, which all unite to produce a simpler supply chain in both
the short and long term.
Figure 1. Reboxetine mesylate (1) and succinate (2).
the advanced chiral diol 11) in gram quantities using the standard
three-step morpholine synthesis.
As (S,S)-reboxetine succinate (2) progressed through late
stage development, we had to balance meeting the short-term
clinical demand against the long-term goal of identifying and
developing the cheapest commercial synthesis. Herein this paper
describes our recent efforts towards achieving this.
Reboxetine mesylate (1), a selective norepinephrine reuptake
inhibitor (NRI), is an antidepressant currently approved and
marketed in over 60 countries with the trade name Edronax. The
(S,S)-enantiomer is significantly more active than the racemate in
a number of biological assays,3 and (S,S)-reboxetine succinate
(2) has been under late stage development at Pfizer for the
treatment of fibromyalgia (Figure 1).
There are several published syntheses of the reboxetine struc-
ture in both racemic and enantioenriched forms, and many of
these share the same common bond disconnections with a three-
step process to build up the morpholine ring in 7 from the pre-
requisite amino alcohol (3 or 4) (Scheme 1).4 The majority of
the published routes also introduce the 2-ethoxyphenol fragment
as the penultimate step,5 requiring the use of the stoichiometric
aryl-tricarbonylchromium reagent 8. These syntheses are 8 or 9
linear steps from raw materials (not including preparation of 8)
and deliver the free base 7 in overall low yields ranging from 6%
to 30%.
The first formal synthesis of 2 was published in 1985, utilizing
the same bond-forming chemistry as used for the racemate
(Scheme 2), with introduction of the chirality through the chiral
epoxide 14, which was accessed from the known trans-[(2R,3R)-
3-phenyloxiran-2-yl]methanol (10).6 An improved approach was
published in 2007 from Pfizer,7 with selective introduction of the
requisite chiral epoxide 10 utilizing the Sharpless oxidation
protocol (Scheme 2). This delivered (S,S)-reboxetine succinate
(2) in 19% overall yield over three telescoped steps (33% from
To support the phase 2 clinical demand, the first three batches
of (S,S)-reboxetine succinate (2) were prepared via a classical
resolution of the racemic commercial product, Edronax (1),
using (S)-mandelic acid (Scheme 3). The mesylate salt (1) was
broken to generate the requisite free amine, and this was resolved
to the crystalline (S,S)-reboxetine-(S)-mandelate (18) in ∼40%
recovery and 92À95% ee. The material was recrystallised to
upgrade the enantiomeric purity to >99.7% followed by a further
salt switch to succinate to deliver the desired product 2 in an
overall yield of 31%. Whilst this route was able to deliver more
than 100 kg of (S,S)-reboxetine succinate (2) using relatively
simple unit operations, it suffered from the high cost and supply
issues of using an established drug substance as a starting material.
In order to prepare 1 kg of 2, 3 kg of Edronax (1) was used, and
with the nontrivial route required to synthesize 1, this impacted
process throughput and resulted in high waste generation.
Early investigations examined a direct resolution of reboxetine
free base 19 without the need to form or break the mesylate salt
(Scheme 4). This approach was used to deliver 40 kg of material,
and whilst several improvements were made to the early steps to
improve robustness and decrease waste (e.g., replacing excess
Vitride with stoichiometric LiAlH4), the route suffered the same
Received: July 4, 2011
Published: August 18, 2011
r
2011 American Chemical Society
1305
dx.doi.org/10.1021/op200181f Org. Process Res. Dev. 2011, 15, 1305–1314
|

Organic Process Research & Development
ARTICLE
Scheme 1. Common 3-Step Morpholine Synthesesa
a (i) chloroacetyl chloride, NEt3; (ii) KOtBu, ROH; (iii) Vitride.
disadvantages, yielding API in 19À21% yield from the amino
alcohol rac-15. The key disadvantages are
• Approximately 5 kg of amino alcohol rac-15 was required to
deliver 1 kg of API.
should be realized and with minimal development as the same
morpholine ring construction could be employed (Scheme 5).
A screen of a wide range of resolving agents and solvent
systems for the amino alcohol rac-15 identified several hits
(Table 1), and upon verification only (R)-camphanic acid proved
to be robust and scalable (entry 3).
• High dilutions were required in the first step equating to
110 L of solvent per kg of API.
• High waste generation
Because of its manufacture from camphor, camphanic acid
proved difficult to access as both enantiomers, and unfortunately
the widely available (S)-configuration led to precipitation of the
undesired enantiomer of 15 as the requisite salt. A work around
was duly developed utilising 0.55 equiv of the resolving agent,
where the undesired enantiomer could be removed by filtration
as the corresponding salt (at 97.8% ee). The filtrate contained the
desired enantiomer as the free base (∼88% ee), and this could be
successfully precipitated as the corresponding benzoate salt 20
with concomitant upgrade of the ee to >99% (Scheme 5). This
process robustly scaled to 150 kg starting material input and deli-
vered material in exceptional yields for a resolution step (average
of 47%). The introduction of the new resolution resulted in a
3-fold increase in throughput due to carrying this out early in the
synthetic route versus late in the route as previously conducted.
• Low throughput
It was clear that in order to meet the demand for material
moving into phase 3 and beyond, a more efficient synthesis
would be required. As detailed in Scheme 2, Edronax is synthe-
sized from a late stage racemic amino alcohol intermediate rac-15
through the standard morpholine ring formation chemistry.
Significant quantities rac-15 were available compared to Edronax
(1), and this provided us with a more amenable point from which
to prepare (S,S)-reboxetine succinate (2) to support further
clinical demands.
An effective solution was quickly identified starting with the
amino alcohol rac-15 used in the late resolution approach. By
shifting the resolution from the end to the beginning of the
synthesis, a significant improvement in throughput and yield
Scheme 2. First Synthesis of (S,S) Reboxetine Succinate (2)a
a (i) L-DIPT, Ti(i-PrO)4, TBHP, EtOAc; (ii) 2-ethoxyphenol, NaOH [58% over 2 steps]; (iii) TMSCl, NEt3 then MeSO3Cl, NEt3; (iv) H2O, HCl,
EtOAc; (v) NaOH, Bu3MeNCl, PhMe [60% over 4 steps]; (vi) NH3, then MeSO3H; (vii) chloroacetyl chloride, NEt3; (viii) t-AmONa; (ix) Vitride,
PhMe then succinic acid or MeSO3H, solvent, [55% over 4 steps].
1306
dx.doi.org/10.1021/op200181f |Org. Process Res. Dev. 2011, 15, 1305–1314

Organic Process Research & Development
ARTICLE
Scheme 3. Resolution of Edronax (1)a
a (i) aq NaOH, DCM then (S)-Mandelic acid, EtOH, [43%]; (ii) EtOH recrystallisation, [92%]; (iii) aq NaOH, DCM then succinic acid, EtOH, [88%].
Scheme 4. Late Resolution Approacha
a (i) Chloroacetyl chloride, THF, aq NaOH; (ii) KOtBu, THF [81.0% 50 kg batch size]; (iii) LiAlH4, THF then aq sodium tartrate, PhMe, EtOH;
(iv) (S)-mandelic acid, EtOH [34.5%]; (v) EtOH recrystallisation [91.5% 50 kg batch size]; (vi) TBME, EtOH, aq NaOH then succinic acid, EtOH
[79.9% 40 kg batch size].
Scheme 5. Optimised Early Resolution Approach to (S,S)-Reboxetine Succinate (2)a
a (i) aq NaOH, toluene, (S)-camphanic acid then aq NaOH, toluene, benzoic acid [47.4% over 3 steps, 75 kg batch size]; (ii) chloroacetyl chloride, THF,
aq NaOH; (iii) KOtBu, tBuOH, cyclohexane [75.8% over 2 steps, 60 kg batch size]; (iv) LiAlH4, THF then N(CH2CH2OH)3, cyclohexane, EtOH;
z(v) succinic acid, EtOH, [71.8% over 2 steps, 50 kg batch size].
1307
dx.doi.org/10.1021/op200181f |Org. Process Res. Dev. 2011, 15, 1305–1314

Organic Process Research & Development
ARTICLE
Table 1. Resolution Screen for the Amino Alcohol rac-15
Scheme 6. Step 1 of the Commercial Route to (S,S)-Rebox-
etine Succinatea
% ee of
liquors
entry
1
optically active acid
solvent
(S)-(À)-4-(2-chlorophenyl)-2-hydroxy-
5,5-dimethyl-1,3,2-dioxaphosphorinane
2-oxide
IPA
59
2
3
(S)-(+)-2-hydroxy-5,5-dimethyl-4-phenyl- IPA
57
1,3,2-dioxaphosphorinane
(1R)-(+)-camphanic acid
toluene
THF
89
90
97
75
96
100
100
69
94
92
61
81
77
79
76
72
57
55
88
EtOAc
acetone
IPA
MeCN
MeTHF
IPA
a (i) Novozym 435 (Candida antarctica lipase B), isopropenyl acetate,
PhMe; (ii) MsCl, NEt3, PhMe; (iii) aq NaOH, PhMe, MeN(n-Bu)3Cl,
PhMe; (iv) H2NCH2CH2OSO3H, DBU, PhMe, EtOH, [86% over 4
steps].
4
5
N-acetylhydroxy-L-proline
THF
EtOAc
Acetone
THF
N-(carbobenzyloxy)-L-phenylalanine
epoxide via the use of 2-amino ethyl hydrogen sulfate (2-
AEHS), as a two carbon subunit at the correct oxidation
state. Using a nucleophile such as 2-AEHS eliminates the
necessity for subsequent redox operations to achieve the
desired oxidation state of the eventual morpholine.
MeTHF
5% water/THF
EtOAc
IPA
6
7
8
9
N-acetyl-L-aspartic acid
(Z)-L-serine
A commercial route was hence selected and developed
that incorporated six chemical transformations in a total of
two synthetic steps with one isolated intermediate, Scheme 6.
With the enantioenriched diol (S,R)-11 in hand, the next step
involved regioselective protection of the primary alcohol.
Our initial attempts utilized a trimethylsilyl protecting group
(Scheme 2); unfortunately, these suffered from moderate
yields and moderate regiocontrol and were found to be non-
robust when processing larger batch sizes. An alternative
protecting group was sought, and after investigation of several
alternatives, an acetyl group was identified as a suitable
alternative.
Classical chemical acetylation conditions suffered from sig-
nificant levels of di-acetylation and poor regiocontrol, and hence
a biocatalytic acylation was investigated and met with great
success. A highly active and commercially available immobilized
enzyme preparation, Novozym 435 (Candida antartica lipase B),
was identified along with conditions that were amenable to
telescoping, allowing the selective mono-acetylation of the
primary alcohol as the corresponding acetate 22 with >98%
regioselectivity in >99% in situ yield. The regioselectivity of this
transformation could further be improved to 99% whilst main-
taining the excellent in situ yield by the incorporation of 10%
THF in the solvent system on a laboratory scale. Simple filtration
removed the biocatalyst, and the solution was successfully
progressed onto the next transformation. It was also noted that
on laboratory scale the recovered enzyme catalyst could be
recycled and reused successfully with no loss in regioselectivity
or in situ yield over several cycles.
EtOAc
IPA
0.9 equiv di-P-toluoyl-L-tartaric acid
0.1 equiv L-tartaric acid
EtOAc
0.9 equiv di-P-toluoyl-L-tartaric acid
0.1 equiv dipivoyl-L-tartaric acid
EtOAc
87
The second major change in the early resolution approach was to
telescope the reduction step directly into the salt formation,
which placed significant challenges on the use of LiAlH4 and the
removal of the aluminum and lithium waste. To address these an
efficient and scalable workup need to be developed. Initially an
aqueous disodium tartrate wash was developed; however, this
suffered from mass transfer issues upon scaleup, and hence an
alternative was quickly developed utilizing triethanolamine as a
fully soluble quench procedure to completely eliminate the pre-
viously observed issues. Althoughdevelopmentof the route shown
in Scheme 5 gave an efficient process that was capable of delivering
several hundreds of kilograms of clinical grade material, ultimately
this process was still far from ideal in a manufacturing setting in the
long term.
The key areas to address with a commercial route for (S,S)-
reboxetine succinate are
• Control of the stereochemistry. The best methodology for
introducing the chirality proved to be that already sup-
ported in the literature, via Sharpless asymmetric epoxida-
tion of cinnamyl alcohol (Scheme 2).
• Introduction of the 2-ethoxy phenol subunit. The best
methodology for introducing the 2-ethoxyphenol subunit
proved to be via a nucleophilic based approach to glycidyl
epoxide (10, Scheme 2).
The major process impurity observed in this transformation
was the acetylated secondary alcohol 25. This impurity was of
some concern as the formation of the secondary acetylated
product 25 would ultimately result in producing the diastereoi-
somer of the API (S,R)-2. There are two proposed mechanisms
for the formation of this impurity: via isomerisation of the
•
Best methodology to build the morpholine ring. The stan-
dard three-step morpholine synthesis was substituted for
by a streamlined two step procedure from the prerequisite
1308
dx.doi.org/10.1021/op200181f |Org. Process Res. Dev. 2011, 15, 1305–1314

Organic Process Research & Development
ARTICLE
Scheme 7. Formation of (S,R)-Reboxetine through Uncontrolled Diol Acetylation Regioselectivitya
a (i) Novozym 435 (Candida antarctica lipase B), isopropenyl acetate, PhMe [quant].
Scheme 8. Formation of the Zwitterion 24 via Opening of the Epoxide (S,S)-14a
a (i) H2NCH2CH2OSO3H, DBU, PhMe, EtOH [86%].
acetylated primary alcohol 22 or via the imperfect regioselectivity
of the enzyme (Scheme 7). Closer investigation of the formation
of 25 suggested both pathways were playing a role. Optimal
control of the regioisomer formation could be achieved through
low reaction temperature and short reaction times whereby
carrying out the acetylation below 25 °C and for less than 16 h
proved highly beneficial. This process gave a solution of 22 in
toluene for progression into the later stages of the synthesis.
With 22 in hand the next transformation involved activation of
the secondary alcohol for SN2 displacement with the primary
alcohol upon acetyl deprotection. The methanesulfonyl leaving
group was utilized resulting in a quantitative yield of 23. After a
standard workup procedure the toluene solution containing the
product was telescoped into the next transformation.
The terminal epoxide (S,S)-14 was delivered in a highly
concise manner by treatment of the toluene solution containing
23 with 12 M sodium hydroxide solution in the presence of a
catalytic phase transfer catalyst. The combination of base and
catalyst were extensively investigated, and the best combination
of reaction time (less than 6 h) and ambient reaction tempera-
tures along with the low cost of reagents delivered a highly
efficient and economical chemical transformation. Methyltri-
butylammonum chloride was chosen as this was the cheapest
phase transfer catalyst on scale and could also be sourced as
an aqueous solution for ease of charging in a manufacturing
environment.
Because of concerns over the stability of the terminal epoxide
(S,S)-14 and the difficulties involved in isolating an oil in a manu-
facturing environment, it was decided to telescope this inter-
mediate as a solution in toluene into the next step, ring opening
with 2-aminoethyl hydrogen sulphate (2-AEHS) (Scheme 8). As
2-AEHS exists as a zwitterion, treatment with a base was required
to liberate the amine functionality for the reaction. After ex-
tensive screening of suitable bases the optimum proved to be 1,8-
diazobicyclo[5.4.0]undec-7-ene (DBU).8 The use of DBU max-
imised product formation whilst simultaneously minimising the
formation of the zwitterionic amine adducts such as 27 and 28
when employing DBU. The major impurity, dimer 26, forms
during this transformation as a result of the desired product 24
reacting with additional epoxide (S,S)-14. The key parameters
that were important to controlling the formation of 26 were
found to be charge of DBU, charge of 2-AEHS, mode of addition
of the reagents, cumulative addition time, and reaction tempera-
ture. A quick and efficient isolation of 24 was developed by pH
adjustment of a basic solution containing the zwitterion 24 with
hydrochloric acid to the isoelectric point of 24 of 5.1, and then
1309
dx.doi.org/10.1021/op200181f |Org. Process Res. Dev. 2011, 15, 1305–1314

Organic Process Research & Development
ARTICLE
Scheme 9. Morpholine Formation in the Commercial Route of (S,S)-Reboxetine Succinate (2)a
a (i) NaOH, THF, EtOH; (ii) succinic acid, EtOH [78À80%].
simple filtration and washing with water yielded the desired
product containing the dimer 26.
Despite extensive investigations the formation of the dimeric
impurity 26 could not be prevented, but it could be minimized
and subsequently purged further by the introduction of an
ethanol reslurry of the isolated material. This resulted in mini-
mising the levels of 26 in the isolated product from 8% to less
than 3%. Such levels were well tolerated and purged in the down-
stream chemistry prior to API formation. It was later found that
the dimeric impurity could be tolerated up to 8% in the down-
stream chemistry with only an impact on the yield of the isolated
API (81% vs 74%). Avoiding the reslurry would ultimately make
the process more streamlined and ultimately cheaper as the
reslurry and subsequent drying of the zwitterion 24 is a time-
consuming operation on a manufacturing scale.
With the zwitterion 24 in hand the final chemical transforma-
tion involved morpholine formation and then subsequent succi-
nate salt formation to furnish the API 2. A base-mediated ring
closure of 24 was explored with a focus on screening a wide range
of bases in solvents where salt formation could be carried out
following reaction completion and workup (Scheme 9). Both
solid NaOH and KOH were quickly identified as effective bases,
which at high concentrations suppress the formation of the major
impurity observed in such chemistry, the amino diol 29. The use of
solid sodium hydroxide in THF in combination with an additive
significantly increased the rate of reaction as well as also improving
the yield from 65% to 90%. Several solvent additives were identi-
fied that yielded homogeneous reaction conditions whilst main-
taining the chemoselectivity, reaction rates, and high yield of this
chemical transformation (water, MeOH, EtOH, t-BuOH, i-PrOH,
t-AmylOH). It was found that alcohols offered a far greater robust
operating range compared to water. The use of water required less
than 1% incorporation into the solvent matrix. At these levels this
proved highly beneficial, whereas at levels greater than 3% both
a retardation of the reaction rate as well as an increase in the
production of the amino diol 29 were observed.9
Figure 2. Powder pattern highlighting additional peak at 13.99° 2θ
(black line) compared to reference standard of (S,S)-reboxetine succi-
nate 2 (pink line).
development of the commercial route. When you compare the
use of the diol (S,R)-11 as a starting material through the pre-
viously published synthetic route7 to the amino alcohol (S,S)-15
coupled with the optimized three-step morphline formation
highlighted earlier in this paper, which furnished (S,S)-rebox-
etine succinant (2) in an overall yield of 33%, you can see that a
nearly double improvement in overall yield has also been realized
by developing the commercial route.
With the change in synthetic route, several batches of API
were synthesised to facilitate stability studies. The initially
synthesized material prepared via the new commercial route
suffered from out of specification residue on ignition (ROI)
levels. Powder X-ray diffraction was used as a detection technique
to identify monosodium succinate as the impurity causing the
ROI issue (Figure 2). This was found to be a result of sodium
hydroxide carryover from the morpholine ring closure into the
succinate salt formation.
The new process delivered clinical grade API in 78À80% yield
that equates to an overall yield of 67À69% over six chemical
transformations. Compared to the previously optimized 33%
for the early resolution route (Scheme 3) an almost double
improvement has been realized through the identification and
The problem was quickly rectified by the implementation of a
water wash prior to salt formation. The volume of water used was
crucial so as to avoid yield losses to the aqueous phase. To
monitor this crucial phase separation and to ensure that all the
sodium hydroxide was removed from the organic phase prior to
1310
dx.doi.org/10.1021/op200181f |Org. Process Res. Dev. 2011, 15, 1305–1314

Organic Process Research & Development
ARTICLE
salt formation, an online conductivity-based process analytical
technique (PAT) was developed.10
Table 2
column
Chiralpak AD-H
With the successful laboratory development of the commer-
cial route to (S,S)-reboxetine succinate (2) the process was trans-
ferred to our large scale laboratory, and the chemistry was
successfully scaled to a 1 kg input of the chiral diol (S,R)-11.
No deviations were observed and clinical grade API was manu-
factured in 70% yield over the two steps and six chemical trans-
formations. The first step was accomplished in 86% yield
followed by ring closure and salt formation in 81% yield.
With the demonstration of the new commercial route, com-
parison with the previous manufacturing routes could be made.
The implementation of a new commercial route had the opportu-
nity to realize large savings for the launch campaign whilst simulta-
neously reducing the waste generated, with estimations showing a
>500 MT reduction from just early development to product launch
and >1300 MT from product launch to peak volume.
The commercial manufacture of (S,S)-reboxetine succinate
has undergone several route changes to meet the clinical de-
mands. The utilization of a late stage resolution approach enabled
fast delivery of drug substance early in the project. Later the early
resolution based approach allowed delivery of bulk material to
fund large clinical trials. It was quickly identified that in order to
meet the long-term demands of the project a new commercial
route was required. Identification of the commercial route was
realized after evaluating several alternatives and selecting one for
further development to allow demonstration on a kilogram scale.
The changeto the commercial route afforded a 58% improvement
in overall yield and a significant saving in API cost along with a
potential to reduce the waste generated by over 1300 MT/yr at
peak drug substance volumes.
mobile phase
93:7 heptane/isopropy alcohol +
0.3% TFA + 0.2% DEA
1.5 mL/min
flow
detector
276 nm
run time
12.0 min
column temperature
retention times
45 °C
(2R,3R)-3-(2-ethoxyphenoxy)-2-
hydroxy-3-phenylproylamine
(2S,3S)- 3-(2-ethoxyphenoxy)-2-
hydroxy-3-phenylproylamine
6.9 min
9.5 min
compound as a white solid (52.7 kg, 52.6%). This material
1
assayed at 97.8% ee. Mp 194À196 °C. H NMR (400 MHz,
d4-MeOD) δ 7.46À7.44 (m, 2H), 7.40À7.31 (m, 3H), 6.98 (dd,
J = 8.2, 1.0 Hz, 1H), 6.88 (ddd, J = 8.2, 6.4, 2.7 Hz, 1H),
6.74À6.69 (m, 2H), 5.20 (d, J = 5.1 Hz, 1H), 4.21À4.11 (m,
3H), 3.08À2.98 (m, 2H), 2.48 (ddd, J = 13.2, 10.5, 4.2 Hz, 1H),
2.00À1.83 (m, 2H), 1.55 (ddd, J = 13.0, 9.0, 4.1 Hz, 1H), 1.48 (t,
J = 7.0 Hz, 3H), 1.09 (s, 3H), 1.06 (s, 3H), 0.93 (s, 3H). For chiral
HPLC assay details, see Table 2.
(2S,3S)-3-(2-Ethoxyphenoxy)-2-hydroxy-3-phenylpro-
pylamine Benzoate
’ EXPERIMENTAL SECTION
1H and 13C NMR spectra were recorded on a Bruker Ultra-
shield 400 Plus spectrometer at 400 and 100.6 MHz, respectively.
(2R,3R)-3-(2-Ethoxyphenoxy)-2-hydroxy-3-phenylpro-
pylamine (S)-(À)-Camphanate
Stage 3: (S,S)-Benzoate Salt Formation. (iii) The combined
organic liquors obtained at the end of step (ii) above were heated
to 45 °C and treated with a solution of NaOH (3.8 kg, 93.8 mol,
0.45 equiv) in water (1.0 L). The resulting biphasic solution was
agitated at 45 °C for 3 h after which the lower aqueous phase was
removed. The upper organic layer was dried by distillation under
DeanÀStark conditions, cooled to ∼65 °C, and subsequently
used in the next step without further purification. Alternatively,
the title compound can be isolated by the addition of an anti-
solvent such as cyclohexane that results in crystallization of the
desired compound. This material assayed at 88% ee via chiral
HPLC method as before. Mp 98À101 °C. 1H NMR (400 MHz,
d4-MeOD) δ 7.44À7.41 (m, 2H), 7.38À7.26 (m, 3H), 6.94 (dd,
J = 8.1, 1.5 Hz, 1H), 6.87À6.83 (m, 1H), 6.77À6.68 (m, 2H),
5.08 (d, J = 6.3 Hz, 1H), 4.15À4.08 (m, 2H), 3.94 (q, J = 6.3 Hz,
1H), 2.63 (d, J = 5.7 Hz, 2H), 1.46 (t, J = 7.03 Hz, 3H)
Stage 1: Mesylate Salt Break. (i) Racemic (2RS,3RS)-3-(2-
ethoxyphenoxy)-2-hydroxy-3-phenylpropylamine mesylate (75.0
kg, 207.0 mol, 1.0 equiv) was agitated in toluene (750 L) at 55 °C.
A solution of NaOH (15.0 kg, 375.0 mol, 1.8 equiv) in water (150 L)
was added. The slurry turned into a clear biphasic solution within
30 min, and the reaction mixture was agitated for an addi-
tional 1 h at 55 °C. The lower aqueous layer was removed and
the upper organic layer was progressed to the next stage.
Stage 2: R,R-Camphanate Salt Formation. (ii) To the organic
solution was added (S)-camphanic acid (20.4 kg, 102.8 mol, 0.55
equiv) at 55 °C. A white precipitate appeared within 30 min. The
temperature was increased to 75 °C, and the slurry was agitated
for an additional 2 h after which the temperature was decreased
to 30 °C over 3 h and maintained for 16 h. The slurry was filtered
to remove the undesired enantiomer, and the filter cake was
washed with toluene (150 L) and dried to yield the desired
A previously prepared solution of benzoic acid (3.8 kg, 30.8
mol, 0.15 equiv) in toluene (1 L) was added to the organic
solution obtained at the end of step (iii). Nucleation was initiated
by seeding (1 wt %) and was ripened at 65 °C for 5 h. Further
benzoic acid (7.5 kg, 61.5 mol, 0.30 equiv) in toluene (150 L) was
added over ∼3 h while maintaining the temperature between 60
and 70 °C. The slurry was agitated for 16 h, cooled to 20 °C over
4 h, and agitated for a further for 4 h at 20 °C. The slurry was
filtered, and the filter cake was washed with toluene (150 L) and
dried to yield (2S,3S)-3-(2-ethoxyphenoxy)-2-hydroxy-3-phe-
nylpropylamine benzoate (41.0 kg, 47.4%) as a white solid.
1
This material assayed at >99% ee. Mp 148À150 °C. H NMR
1311
dx.doi.org/10.1021/op200181f |Org. Process Res. Dev. 2011, 15, 1305–1314

Organic Process Research & Development
ARTICLE
Table 3
Table 4
column
Chiralpak AD-H
93:7 heptane/isopropy alcohol +
0.3% TFA + 0.2% DEA
1.5 mL/min
column
Chiracel OJ
mobile phase
mobile phase
25:75 ethanol/heptane +
0.05% diethylamine
0.70 mL/min
275 nm
flow
flow
detector
276 nm
detector
run time
12.0 min
run time
retention times
30 min
column temperature
retention times
45 °C
(2R,3R)-2-[α-(2-ethoxyphenoxy)
9.3 min
(2R,3R)-3-(2-ethoxyphenoxy)-2-
hydroxy-3-phenylproylamine
(2S,3S)- 3-(2-ethoxyphenoxy)-2-
hydroxy-3-phenylproylamine
6.9 min
9.5 min
benzyl]morpholine
(2S,3S)-2-[α-(2-ethoxyphenoxy)
benzyl]morpholine
12.4 min
6.74À6.70 (m, 1H), 5.33 (d, J = 6.1 Hz, 1H), 4.29À4.18 (m,
3H), 4.12À4.04 (m, 2H), 3.35 (dd, J = 12.1, 10.5 Hz, 1H), 2.99
(dd, J = 12.1, 2.9 Hz, 1H), 1.43 (t, J = 7.0 Hz, 3H). 13C NMR
(100 MHz, d4-MeOD) δ171.3, 151.3, 149.0, 138.7, 129.7, 129.6,
128.8, 123.8, 122.2, 119.4, 115.8, 83.0, 77.0, 68.3, 66.0, 43.7, 15.5.
m/z [M + H]+ = 328.
(400 MHz, d4-MeOD) δ 7.96À7.93 (m, 2H), 7.45À7.30 (m,
8H), 6.97 (dd, J = 8.0, 1.0 Hz, 1H), 6.88 (ddd, J = 8.3, 6.3, 2.4 Hz,
1H), 6.74À6.68 (m, 2H), 5.19 (d, J = 5.1 Hz, 1H), 4.20À4.10
(m, 3H), 3.06À2.97 (m, 2H), 1.47 (t, J = 7.0 Hz, 3H). 13C NMR
(100 MHz, d4-MeOD) δ 175.4, 150.7, 146.8, 138.9, 138.7, 131.6,
130.4, 129.7, 129.6, 128.9, 128.6, 123.7, 122.1, 118.4, 114.9, 85.5,
72.2, 65.8, 43.5, 15.4. m/z [M + H]+ = 288. For chiral HPLC
assay details, see Table 3.
(2S,3S)-2-[α-(2-Ethoxyphenoxy)benzyl]morpholine Suc-
cinate Salt (S,S)-Reboxetine Succinate)
(2S,3S)-2-[α-(2-Ethoxyphenoxy)benzyl]morpholine-5-one
(2S,3S)-2-[α-(2-Ethoxyphenoxy)benzyl]morpholine-5-one
(50.0 kg, 152.7 mol, 1 equiv) was dissolved in anhydrous THF
(250 L) and added over 4 h to a 2.4 M solution of lithium
aluminium hydride (63.5 L, 152.7 mol, 1 equiv) stirring at
20À25 °C. The reaction mixture was stirred for 2 h and
quenched with addition of a solution of triethanolamine (53.6
kg, 360.0 mol, 2.4 equiv) in THF (100 L) over 2 h. The reaction
was agitated for 30 min, and NaOH in water (25.7 kg in 250 L)
was added. The reaction mixture was agitated for an additional
15 min, then cyclohexane (150 L) was added, and the phases
were separated. The upper organic phase was washed twice with
aqueous NaOH (2 Â 12.8 kg NaOH in 125 L water) followed by
water (50 L). The organic phase was azeotropically dried and
distilled to ∼125 L (∼2.5 L/kg). Ethanol denatured with 3%
cyclohexane (250 L) was added and distilled to ∼125 L (∼2.5
L/kg). The ethanol charge/distillation process was repeated to
yield (S,S)-reboxetine free base as an ethanol solution.
Succinic acid (19.9 kg, 168.7 mol, 1.1 equiv) was dissolved in
ethanol denatured with 3% cyclohexane (175 L) at 55 °C, and the
solution was added to the solution of (2S,3S)-2-[α-(2-ethoxy-
phenoxy)benzyl]morpholine in ethanol from above. The reac-
tion mixture was heated to reflux for 30 min and subsequently
cooled to 0 at 0.3 °C/min. The reaction mixture was granulated
for 1 h at 0 °C, and the solid was isolated by filtration. The solid
was washed twice with ethanol denatured with cyclohexane (2 Â
50 L) and dried at ∼55 °C to yield (S,S)-reboxetine succinate as a
white crystalline solid (43.1 kg, 71.8%). This material assayed at
greater than 99% ee. Mp 149À152 °C. [α]32D = (c 37, ethanol)
+18.37. 1H NMR (400 MHz, d6-DMSO) δ 7.22À7.54 (m, 5H),
6.66À6.96 (m, 4H), 5.27 (d, J = 6.0 Hz, 1H), 4.01 (q, J = 7.1 Hz,
(2S,3S)-3-(2-Ethoxyphenoxy)-2-hydroxy-3-phenylpropyl-
amine benzoate (60.0 kg, 146.0 mol, 1.0 equiv) was dissolved
in tetrahydrofuran (300 L) and water (60 L) at 20À25 °C.
An aqueous solution of sodium hydroxide (17.6 kg, 440.0 mol,
3.0 equiv) in water (60 L) was added, and the biphasic mixture
was agitated for 30 min. The reaction mixture was cooled to
10 °C, and a solution of chloroacetyl chloride (19.8 kg, 176.8
mol, 1.2 equiv) in tetrahydrofuran (15 L) was added over 1 h.
The reaction mixture was warmed to ambient temperature, and
cyclohexane (120 L) was added. The phases were separated, and
the upper organic phase was washed with 15% w/v brine (30 L)
and azeotropically dried by distillation to give an organic solution
containing the amide 16.
The organic solution from above was added over 30 min to a
5 °C solution of potassium tert-butoxide (33.0 kg, 294.6 mol,
2.0 equiv) in THF (24 L) and tert-butanol (240 L). The reaction
was agitated for 15 min and then quenched with 10 wt % aqueous
acetic acid (120 L). The reaction mixture was allowed to warm to
room temperature over 1 h, cyclohexane (120 L) was added, and
the phases were separated. The upper organic phase was
azeotropically dried by distillation and then distilled to ∼150 L
(∼2.5 L/kg). Isopropyl alcohol (144 L) was added, and the
reaction mixture was distilled to 2.4 L/kg. The solution was
cooled to 10 °C, and the resulting slurry was granulated for 8 h
and isolated by filtration. The solid was washed with isopropyl
alcohol (60 L) and dried at 50 °C to yield the title compound as a
1
white crystalline solid (36.2 kg, 75.8%). Mp 126À127 °C. H
NMR (400 MHz, d4-MeOD) δ 7.46À7.43 (m, 2H), 7.37À7.28
(m, 3H), 6.92 (dd, J = 8.2, 1.6 Hz, 1H), 6.88À6.82 (m, 2H),
1312
dx.doi.org/10.1021/op200181f |Org. Process Res. Dev. 2011, 15, 1305–1314

Organic Process Research & Development
ARTICLE
2H), 3.83 (m, 2H), 3.50 (m, 2H), 2.61À2.82 (m, 3H), 2.34 (br s,
4H), 1.33 (t, J = 7.1 Hz, 3H). 13C NMR (100 MHz, d6-DMSO) δ
174.4, 149.0, 147.3, 137.8, 128.2, 127.3, 120.7, 116.7, 114.4, 80.8,
77.5, 65.9, 64.1, 45.8, 44.1, 39.7, 39.5, 39.3, 39.1, 30.2, 14.9. m/z
[M + H]+ = 314. For chiral HPLC assay details, see Table 4.
2-{[(2S,3S)-3-(2-Ethoxyphenoxy)-2-hydroxy-3-phenyl-
propyl]amino}ethyl Hydrogen Sulfate
with water (3.0 L), and dried to yield the desired product as an off
white solid (1.2 kg, 86% yield).
The zwitterion product from above (1.2 kg, 3.0 mol, 1.0 equiv)
was slurried in ethanol denatured with cyclohexane (6.1 L),
heated to 50 °C, and then subsequently agitated for 2 h. The
slurry was then cooled to 10 °C over 2 h and granulated for 16 h
after which the slurry was filtered, washing with ethanol dena-
tured with cyclohexane (2.5 L). A second wash was performed
with ethanol denatured with cyclohexane (2.5 L), and the filter
cake was deliquored and then dried in vacuo at 50 °C to yield the
desired product (1.1 kg, 84%) as a white solid. Mp 183À185 °C.
1H NMR (400 MHz, d4-MeOD) δ 7.42 (m, 2H), 7.34 (m, 2H),
7.28 (m, 1H), 6.93 (dd, J = 8.1, 1.6 Hz, 1H), 6.84 (ddd, J = 8.1,
7.3, 1.6, 1H), 6.74 (dd, J = 8.1, 1.6 Hz, 1H), 6.68 (ddd, J = 8.1, 7.3,
1.6 Hz, 1H), 5.14 (d, J = 5.3 Hz, 1H), 4.26 (ddd, J = 9.8, 5.3, 3.4
Hz, 1H), 4.21 (m, 2H), 4.11 (dq, J = 11.8, 7.1, 1H), 4.10 (dq, J =
11.8, 7.1, 1H), 3.32 (m, 2H), 3.18 (dd, J = 12.6, 9.8 Hz, 1H), 3.10
(dd J = 12.6, 3.4 Hz, 1H), 1.43 (t, J = 7.1 Hz, 3H). 13C NMR (100
MHz, d6-DMSO) δ 149.9, 147.6, 138.0, 128.7, 128.4, 127.9,
122.8, 121.2, 118.1, 114.8, 82.7, 69.4, 64.6, 61.8, 49.6, 47.7, 15.3.
m/z [M + H]+ = 412 and [M À SO3]+ = 332.
(2S,3S)-2-[α-(2-Ethoxyphenoxy)benzyl]morpholine Suc-
cinate Salt (S,S)-Reboxetine Succinate
Stage 1: Acetylation. Toluene (5.0 L) and isopropenyl acetate
(765.0 mL, 6.9 mol, 2.0 equiv) followed by (2R,3S)-3-(2-
ethoxyphenoxy)-3-phenylpropane-1,2-diol (1.0 kg, 3.5 mol,
1.0 equiv) were added to a 25 L fixed rig. The reaction mixture
was agitated and heated to an internal temperature of 30 °C.
Novozymes CaL B (20.0 g, 2 mol %) was added, and the mixture
was agitated for 17 h. The enzyme was removed by filtration, and
the filtrate was progressed to stage 2.
Stage 2: Mesylation. To a stirred solution from stage 1 at
20 °C was added triethylamine (822 mL, 5.9 mol, 1.7 equiv) in
one portion. Methanesulfonyl chloride (362 mL, 4.7 mol, 1.4
equiv) in toluene (1.0 L) was added dropwise over ∼2 h. The
reaction mixture was agitated for a further 1 h to ensure reaction
completion. Aqueous hydrogen chloride (1 M, 3.0 L, 3.0 mol)
was added in a single portion. The biphasic mixture was agitated
for 30 min. The phases were separated, and the upper organic
phase was progressed to stage 3.
Stage 3: Epoxide Formation. A solution of sodium hydroxide
(1.4 kg, 36.0 mol) in water (3.0 L) and methyltributylammonium
chloride (81.8 g, 347 mmol, 0.1 equiv) was added to the organic
stream from stage 2. The resulting mixture was agitated for
3.5 h to ensure reaction completion. The phases were allowed to
separat,e and the lower aqueous phase was discarded. The upper
organic phase was washed with water (2.0 L) and progressed to
stage 4.
Stage 4: Zwitterion Formation. In a separate vessel, 2-ami-
noethyl hydrogen sulfate (1.2 kg, 8.7 mol, 2.5 equiv) was slurried
in toluene (2.0 L) and ethanol (2.0 L). 1,8-Diazabicyclo-
[5.4.0]undec-7-ene (1.3 L, 8.6 mol, 2.48 equiv) was added in a
single portion, and the contents were heated to 65 °C and
agitated for 1 h to ensure complete deprotonation. The organic
solution from stage 3 was diluted with toluene (5.0 L) and added
to the activated amine mixture over 1 h. The mixture was agitated
at 70 °C for 2 h post end of addition. The reaction mixture was
subsequently cooled to ambient temperature and extracted into
an aqueous solution of sodium hydroxide (416 g in 8.0 L of water,
3.0 equiv NaOH) for 2 h. The upper organic phase was discarded,
and the lower aqueous phase was treated with conc HCl to adjust
the pH to 5.1À5.2 (1.2 L, 36% conc HCl) resulting in crystal-
lization of the product, which was isolated by filtration, washed
To a 25 L fixed rig were introduced tetrahydrofuran (7.1 L) and
ethanol (denatured with 3% cyclohexane) (210 mL) followed
by 2-{[(2S,3S)-3-(2-ethoxyphenoxy)-2-hydroxy-3-phenylpro-
pyl]amino}ethyl hydrogen sulfate (1.1 kg, 2.6 mol, 1 equiv).
The slurry was agitated at 18 °C for 1 h, and then sodium
hydroxide (318 g, 7.1 mol, 3.0 equiv) was added in a single
portion. The reaction mixture was subsequently heated to reflux
(65 °C) for ∼5 h. The reaction mixture was cooled to room
temperature and then treated with water (5.3 L). The reaction
mixture was agitated for 16 h after which cyclohexane (4.2 L) was
added, and the phases were separated. The upper organic phase
was washed with water (2.2 L) and distilled at atmospheric
pressure under DeanÀStark conditions to decrease the water
levels to <0.2% (as determined by Karl Fischer analysis). The
organic phase was concentrated by distillation under atmospheric
pressuretoa volumeof ∼2.0L. Ethanol(denatured with3%cyclo-
hexane) (5.1 L) was added, and a distill and replace operation was
performed using ethanol (denatured with 3% cyclohexane) (3 Â
5.1 L) to achieve not more than 1% tetrahydrofuran and a total
volume of 2À2.5 L.
Further ethanol (denatured with 3% cyclohexane) (4.9 L) was
charged to the reaction vessel followed by succinic acid (301 g,
2.60 mol, 1.0 equiv), and the contents were heated to reflux to
ensure full dissolution. The temperature was lowered to ∼65 °C,
and the solution was seeded with (S,S)-reboxetine succinate
(5.5 g, 12.8 mol, 0.05 equiv) followed by granulation for 3 h. The
slurry was cooled to 0 at 0.5 °C/min and granulated for a further
16 h. The slurry was filtered, washing with ethanol (denatured
with 3% cyclohexane) (2.1 L/kg), and dried at 50 °C to yield (S,
S)-reboxetine succinate (897 g, 82%) as a white solid. 1H NMR
1313
dx.doi.org/10.1021/op200181f |Org. Process Res. Dev. 2011, 15, 1305–1314

Organic Process Research & Development
ARTICLE
Table 5. (S,S)-Reboxetine Succinate HPLC Assay
(2) For a medium volume product and an annual tonnage of 20 MT,
reducing the cost of the drug substance by as little as $500/kg can equate
to annual savings of $10 million per year.
(3) Zampieri, M.; Airoldi, A.; Martini, A. Patent WO/2003106441.
Hughes, B.; McKenzie, I.; Stoker, M. J. Patent WO/2006000903.
Ellis, A. J.; Junor, R. W. J.; Stoker, M. J.; Whelan, L. J. Patent WO/
2010044016.
(4) Henegar, K. E.; Ball, C. T.; Horvath, C. M.; Maisto, K. D.;
Mancini, S. E. Org. Process Res. Dev. 2007, 11, 346–353.
(5) Brenner, E.; Baldwin, R. M.; Tamagnan, G. Org. Lett. 2005, 7,
937–939.
(6) Melloni, P.; Della Torre, A.; Lazzari, E.; Mazzini, G.; Meroni, M.
Tetrahedron 1985, 41, 1393–1399.
column
gradienta
0
Phenomenex Luna C8 (2) 150 m  4.6 mm, 5 μm
75.0 MPA, 25.0 MPB
75.0 MPA, 25.0 MPB
30.0 MPA, 70.0 MPB
30.0 MPA, 70.0 MPB
75.0 MPA, 25.0 MPB
75.0 MPA, 25.0 MPB
1 mL/min
4
26.5
30
31
36
flow
temp
injection
detection
sample prep
45 °C
50 μL
(7) Henegar, K. E.; Cebula, M. Org. Process Res. Dev. 2007, 11,
354–358.
276 nm
(8) Bulk cost of DBU on a metric tonne scale is $15/kg.
(9) Assaf, G.; Cansell, G.; Critcher, D.; Field, S.; Hayes, S.; Mathew,
S.; Pettman, A. Tetrahedron Lett. 2010, 51, 5048–5051. Brown, G. R.;
Forster, G.; Foubister, A. J.; Stribling, D. J. Pharm. Pharmacol. 1990,
42, 797–799. Wijtmans, R.; Vink, M. K. S.; Schoemaker, H. E.; van Delft,
F. L.; Blauww, R. H.; Rutjes, F. P. J. T. Synthesis 2004, 5, 641–662.
(10) Conductivity measurements were taken of the aqueous phases
prior to drug substance isolation and then calibrated against measured
ROI levels in the drug substance. This allowed a calibration model to be
developed to allow a conductivity ceiling to be set that would ultimately
lead to the isolation of drug substance that met the ROI specification.
0.2 mg/mL (27.5 mg in 100 mL
[25% MeCN in buffer])
a MPA = 0.02 M KH2PO4 at pH 6.8 (0.5 M KOH). MPB = acetonitrile.
Table 6. (S,S)-Reboxetine Succinate Chiral HPLC Analysisa
column
MP
Diacel chiracel OJ 250 m  4.6 mm
heptane/ethanol/diethylamine 1500:500:1
flow
0.7 mL/min
30 °C
temp
injection
detection
run time
sample prep
20 μL
276 nm
20 min
1.0 mg/mL in MP
a (R,R)-Reboxetine enantiomer measured at not more than 0.05% (none
detected).
(400 MHz, d6-DMSO) δ 7.22À7.54 (m, 5H), 6.66À6.96 (m,
4H), 5.27 (d, J = 6.0 Hz, 1H), 4.01 (q, J = 7.1 Hz, 2H), 3.83 (m,
2H), 3.50 (m, 2H), 2.61À2.82 (m, 3H), 2.34 (br s, 4H), 1.33 (t,
J = 7.1 Hz, 3H). 13C NMR (100 MHz, d6-DMSO) δ 174.4, 149.0,
147.3, 137.8, 128.2, 127.3, 120.7, 116.7, 114.4, 80.8, 77.5, 65.9,
64.1, 45.8, 44.1, 39.7, 39.5, 39.3, 39.1, 30.2, 14.9. m/z [M + H]+ =
314. [α]32D = +18.37 (c 37, ethanol). HPLC assay: 72.2% as is,
99.5% salt corrected. Impurities: no impurities detected at levels
of not more than 0.05%. For HPLC assay details, see Tables 5
and 6.
’ AUTHOR INFORMATION
Corresponding Author
*E-mail: stewart.hayes@pfizer.com.
’ ACKNOWLEDGMENT
We would like to thank the broader team based in Sandwich,
U.K. for all the work and effort put in by all on the reboxetine
project over its varied and lengthy history.
’ REFERENCES
(1) Butters, M.; Catterick, D.; Craig, A.; Curzons, A.; Dale, D.;
Gillmore, A.; Green, S. P.; Marziano, I.; Sherlock, J. P.; White, W. Chem.
Rev. 2006, 106, 3002–3027.
1314
dx.doi.org/10.1021/op200181f |Org. Process Res. Dev. 2011, 15, 1305–1314