The fit for purpose development of S1P1 receptor agonist GSK2263167 using a robinson annulation and saegusa oxidation to access an advanced phenol intermediate

Article abstract of DOI:10.1021/op400162p

A fit for purpose approach has been adopted in order to develop a robust, scalable route to the S1P₁ receptor agonist, GSK2263167. The key steps include a Robinson ring annulation followed by a Saegusa oxidation, providing rapid access to an ad

Full text of DOI:10.1021/op400162p

Article
pubs.acs.org/OPRD
The Fit For Purpose Development of S1P1 Receptor Agonist
GSK2263167 Using a Robinson Annulation and Saegusa Oxidation to
Access an Advanced Phenol Intermediate
Robert M. Harris,* Benjamin I. Andrews, Stacy Clark, Jason W. B. Cooke, John C. S. Gray,
and Stephanie Q. Q. Ng
Chemical Development, GlaxoSmithKline Research and Development Ltd., Gunnels Wood Road, Stevenage, Hertfordshire SG1 2NY,
U.K.
S
* Supporting Information
ABSTRACT: A fit for purpose approach has been adopted in order to develop a robust, scalable route to the S1P1 receptor
agonist, GSK2263167. The key steps include a Robinson ring annulation followed by a Saegusa oxidation, providing rapid access
to an advanced phenol intermediate. Despite the use of stoichiometric palladium acetate for the Saegusa oxidation, near complete
recovery of the palladium has been demonstrated. The remaining steps have been optimised including the removal of all
chromatography. An alternative to the Saegusa oxidation is described as well as the development of a flow process to facilitate
further scale-up of the amidoxime preparation using hydroxylamine at elevated temperature.
including α-halogenation/elimination,6 benzoquinone oxida-
INTRODUCTION
■
tion,7 epoxidation/aromatisation8,9 and Pd/C dehydrogen-
Sphingosine 1-phosphate (S1P, 1) has a number of physiological
ation10 were unsuccessful with acidic conditions and high
roles. Agonism of sphingosine-1-phosphate receptor subtype 1
temperatures resulting in undesired Boc deprotection. This
(S1P1) is a precedented mechanism designed to reduce
inflammation by the sequestration of lymphocytes.1 However,
prompted investigation into the Saegusa oxidation, which avoids
both acidic conditions and high temperatures.11 Pleasingly,
it has been shown that agonism of S1P3 can lead to unwanted
cardiovascular effects.2 GSK2263167 (2, Figure 1) is a potent
oxidation of enone 9 could be achieved via reaction of the
corresponding TMS enol ether 10 with 1.2 equiv of Pd(OAc)2
(Scheme 3).
S1P1 receptor agonist and shows a high degree of selectivity over
S1P3.3 In order to progress through early development activities,
a route capable of delivering kilogram quantities of compound 2
was required.
The initial medicinal chemistry route to compound 2 involved
a 15-step linear synthesis from 2-methyl-3-methoxy benzoic acid
3 (Scheme 1).4
While this route was suitable for making small quantities of
material, the number of steps, large amount of chromatography,
capricious chemistry, and poor yields (overall <2%) made it
unsuitable for further progression. Therefore, an alternative
route was sought that could deliver the active pharmaceutical
ingredient (API) required for precandidate selection and safety
assessment studies.
A route was identified which could potentially provide access
to phenol 4 in just two steps from N-Boc piperidone, 8.3 This
involved a Robinson annulation reaction to install the
tetrahydroisoquinoline carbon skeleton followed by an aroma-
tisation to give phenol 4, significantly reducing the number of
steps (Scheme 2).
Given the advantages of being able to rapidly access phenol 4,
it was deemed worthwhile to commit resources to evaluate this
alternative route. Although direct literature precedent for
Robinson annulation on the benzyl protected piperidone
existed,5 reaction of the Boc derivative would feed directly into
the existing route. Following the literature conditions, enone 9
was prepared in 79% crude yield. However, oxidation to phenol 4
proved more challenging. A number of aromatisation conditions
Using the Saegusa conditions, phenol 4 was isolated in 41%
yield from piperidone 8 over two steps after purification by
column chromatography. With the new route demonstrated, 116
g of phenol 4 was readily prepared and converted to API
following the existing medicinal chemistry route to give 44 g of
GSK2263167 2b for precandidate selection studies. This
corresponded to an increase in overall yield from <2% to 10%
and a reduction in the number of steps from 15 to 9.
Despite the improved yield and reduction in step count, a
number of opportunities for further optimisation remained, and a
fit for purpose approach to process development was adopted. In
general, the amount of effort expended on route development
and optimisation of early phase compounds is often limited by
the pressure of tight delivery timelines. There can be a reluctance
to over resource early phase projects in an attempt to minimise
wasted effort resulting from compound attrition. Therefore, a fit
for purpose approach is often taken, where an appropriate
amount of work is performed at each stage of development in
order to progress the project to the next milestone. This was the
strategy employed for GSK2263167 in order to deliver the
material required for further safety assessment studies and
clinincal use.
Received: June 17, 2013
© XXXX American Chemical Society
A
dx.doi.org/10.1021/op400162p | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 1. GSK2263167 and the phospholipid S1P.
a
Scheme 1. Medicinal chemistry route to GSK2263167
a
Reagents and conditions: (a) BH3·THF, THF, 97%; (b) MnO2, DCM; (c) MeNO2, NH4OAc, 100 °C, 66% (2 steps);(d) LiAlH4, THF, 50%; (e)
CH3CO2CHO, reflux, 50%; (f) (CH2O)n, HCO2H, AcOH, 100 °C, 58%; (g) BBr3, DCM then MeOH, 94%; (h) Boc2O, DCM, 70%; (i) Tf2O,
pyridine, DCM, −30 °C, 95%; (j) Zn(CN)2, Pd(PPh3)4, DMF, 120 °C, 80%; (k) NH2OH·HCl, NaHCO3, EtOH, reflux; (l) 3-cyano-4-
isopropoxybenzoyl chloride 6, NEt3, DMF, 120 °C, 58% (2 steps); (m) HCl, dioxane, DCM; (n) ethyl acrylate, DBU, MeCN; (o) NaOH, EtOH,
85% (3 steps).
a
removed the residual LiCl prior to treatment with Pd(OAc)2 and
was shown to be beneficial to the reaction, possibly due to the
prevention of the formation of a palladium chloride containing
intermediate formation. Azeotropic distillation reduced the
amount of residual hexamethyldisilazane which was also
beneficial to the reaction. These improvements enabled a
reduction in the required amount of Pd(OAc)2 from 1.2 to 1
equivalents resulting in a significant reduction in cost.
The workup of the Saegusa reaction involved the addition of
triethylamine to reduce any residual palladium(II) to palla-
dium(0) which then precipitated and was removed by filtration
through Celite. This reduction to palladium(0) was not 100%
efficient and the residual palladium(II) in solution caused oiling
during attempts to crystallise product 4. Replacing triethylamine
with aqueous potassium formate in the workup gave a more
efficient reduction of palladium resulting in large particles of
palladium(0) that were removable by simple filtration (<80 ppm
Pd by ICP). The efficient removal of palladium allowed the
development of a crystallisation of phenol 4 from MeTHF with
heptane as the antisolvent. The optimised process was carried
out in the pilot plant to prepare 11 kg of phenol 4 in 42% yield
from compound 8 with a further 2 kg product remaining in the
liquors.
Scheme 2. Improved route to phenol 4
a
Reagents and conditions: (a) pyrrolidine, toluene, ethyl vinyl ketone,
79% crude.
a
Scheme 3. Saegusa oxidation of enone 9
a
Reagents and conditions: (a) LiHMDS, TMSCl, THF, −78 °C; (b)
Pd(OAc)2, THF, 0−20 °C, 52%.
RESULTS AND DISCUSSION
■
Preparation of Phenol 4. Since the Robinson annulation of
compound 8 worked well, efforts towards optimising the
synthesis of phenol 4 were focused on the Saegusa oxidation.
It was shown that by substituting trimethylsilyl for triethylsilyl,
the enol ether formation could be carried out at 0 °C avoiding the
requirement for cryogenic conditions and also giving a cleaner
reaction. This is presumably due to the increased stability of the
enol ether compared to its trimethylsilyl equivalent. Unfortu-
nately, initial attempts to use substoichiometric palladium acetate
using benzoquinone as co-oxidant (0.5 equivalents of each) gave
poor conversion and purity compared to the stoichiometric
process.
The potassium formate workup not only avoided the
requirement for chromatography but also allowed the efficient
recovery of palladium from the process. Of the 10.5 kg Pd metal
used, 10.3 kg was recovered and sold back to Johnson Matthey
for a profit of £62.5 K as a result of the doubling of the price of
palladium during the campaign. However, due to the volatility of
palladium prices an alternative route avoiding stoichiometric
Pd(OAc)2 was desired prior to further scale-up.
Changing solvent from THF to MeTHF in the enol ether
formation enabled the introduction of a water wash. This
B
dx.doi.org/10.1021/op400162p | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Preparation of Benzonitrile 5. Conversion of phenol 4 to
benzonitrile 5 was achieved via cyanation of triflate 11 with
Zn(CN)2 and Pd(PPh3)4 (Scheme 4).
Due to the improved reaction profile and the insolubility of
amidoxime 12 in aqueous ethanol, the product was found to
crystallise out of the reaction mixture and could be isolated by
filtration. In contrast, the residual triphenylphosphine oxide from
the cyanation reaction was readily soluble and efficiently purged
into the liquors. In order to push the reaction to completion,
extended reaction times in refluxing ethanol were required,
presumably due to the steric bulk imparted by the ortho-methyl
group. Given that 50% aqueous hydroxylamine solution can
decompose violently at elevated temperatures (eq 1)12 and the
rate of decomposition can be significantly increased by low levels
of metal contamination, there were process safety implications to
consider when running this reaction on scale.13−15
a
Scheme 4. Conversion of phenol 4 to benzonitrile 5
a
Reagents and conditions: (a) Tf2O, pyridine, 5 °C, 95%; (b)
Zn(CN)2, Pd(PPh3)4, DMF, 120 °C, 87%.
The medicinal chemistry conditions for the triflation required
low temperatures (−30 °C), purification by chromatography, use
of DCM and gave a thick, poorly stirred reaction slurry. It was
found that the thick slurry could be avoided by performing the
reaction in low volumes of pyridine (2.6 vol) and by carefully
controlling the addition of the trifluoroacetic anhydride the
reaction could be carried out at 5 °C with a much improved
reaction profile. As a result the crude triflate could be taken
directly into the cyanation reaction without further purification.
After addition of ethyl acetate, the pyridine which was
detrimental to the cyanation reaction, could be removed by
aqueous washes with 1 M HCl. No Boc-deprotection was
observed, and the process could be telescoped by solvent
swapping to DMF. This process was run in the pilot plant to give
16.5 kg of crude triflate 11 as a solution in DMF in similar yield
and purity.
Early studies on the cyanation reaction showed it to be very
capricious, with incomplete conversion being a problem. This
was attributed to oxidation of the phosphine ligands and was
overcome by efficient degassing of the solvent prior to addition of
the Pd(PPh3)4. On a 20-L scale this was achieved by a number of
nitrogen/vacuum purging cycles. However, on pilot plant scale it
was shown that simply heating to 90 °C for 2-h prior to catalyst
addition reduced the oxygen levels in solution to less than 0.1%
which was sufficiently low to give a robust reaction. The addition
of toluene and incorporation of aqueous washes to remove DMF
enabled crude nitrile 5 to be isolated in 87% yield after a
concentration to dryness. It was demonstrated that the residual
triphenylphosphine oxide byproduct was tolerated and readily
purged in the next step, removing the requirement for
chromatography. Whilst a concentration to dryness is acceptable
on a 20-L scale, it is not practical on scale-up. As a result, in the
pilot plant, the solution of nitrile 5 in toluene was concentrated to
3 volumes and telescoped directly into the amidoxime formation.
Preparation of Amidoxime 12. The original conditions for
the conversion of benzonitrile 5 into amidoxime 12 used
hydroxylamine hydrochloride and yielded a 1:1 mixture of the
desired product 12 and undesired amide impurity 13 which were
inseparable by column chromatography. By using a 50% w/w
aqueous solution of hydroxylamine, the amount of amide
impurity could be reduced to just 2% (by HPLC) (Scheme 5).
Decomposition of hydroxylamine
157NH2OH(l) → 61NH3(g) + 35N2(g) + 12N2O
(g)
+ 2NO + 143H2O + H2(g)
(g)
(g)
(1)
Samples of the reaction mixture were therefore tested for
thermal stability by differential scanning calorimetry (DSC) and
under adiabatic conditions using accelerating rate calorimetry
(ARC). On the basis of the ARC experiments, a maximum
recommended processing temperature of 80 °C was set.16 Prior
to the scale-up of this chemistry, a series of controls was put into
place to minimise any potential safety issues including the
filtration of all solvents into the reaction vessels to avoid potential
ingress of metals; an automatic control system to put the
contents of the vessel into a soft crash cooling should the reaction
temperature reach 90 °C or upon failure of a bursting disc; and
appropriate handling and storage of hydroxylamine and any
waste generated. With these controls in place the chemistry was
scaled up to deliver first, 550 g of amidoxime 12 (84% yield), and
then, 8 kg (75% yield), in excellent purity (97−98% by HPLC).
The difference in yields was due to a difference in the isolation
temperature. In the 20-L reactors the amidoxime 12 was isolated
at 0 °C, whereas at scale the temperature of isolation was
restricted to 10 °C. This was due to 50% aqueous hydroxylamine
having a melting point of 8 °C and concerns that the aqueous
hydroxylamine could crystallise out of solution.
Preparation of Oxadiazole 7. The medicinal chemistry
conditions for the coupling of amidoxime 12 and a solution of
acid chloride 6 in toluene to give oxadiazole 7 suffered from a
poor yield (60%) due to the formation of two major impurities,
14 and 15 (Scheme 6).
Impurity 14 was observed at levels up to 20% (by HPLC) and
arises from the high-temperature decomposition of DMF to give
dimethylamine which can react with acid chloride 6. By replacing
DMF with toluene, the formation of impurity 14 was avoided.
The high reaction temperature resulted in some Boc
deprotection of either compound 12 or compound 7. The
resulting free amine can then react with the acid chloride to form
impurity 15. By reducing the reaction temperature to 80 °C the
amount of this impurity was reduced from 8% to 4% (by HPLC).
Unfortunately, reducing the reaction temperature increased the
reaction time from 2.5 to 24 h. The reaction was further
optimised by using pyridine as solvent. This resulted in a cleaner
reaction profile and a faster reaction rate with the reaction time
reduced to 3.5 h. In the workup, ethyl acetate was added followed
by aqueous washes to remove the pyridine hydrochloride. Using
this optimised process 7.6 kg of oxadiazole 7 was produced in an
a
Scheme 5. Conversion of benzonitrile 5 to amidoxime 12
a
Reagents and conditions: (a) aq NH2OH, EtOH, reflux, 84%.
C
dx.doi.org/10.1021/op400162p | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
a
Scheme 6. Formation of oxadiazole 7 and associated impurities
a
Reagents and conditions: (a) Et3N, DMF, 120 °C, 60%.
a
Scheme 7. Telescoping of Boc deprotection, alkylation, and hydrolysis to prepare 2b
a
Reagents and conditions: (a) aq HCl/EtOH; (b) aq NaOH, ethyl acrylate; (c) aq NaOH, EtOH, 91%.
a
Scheme 8. Preparation of 2a anhydrate from sodium salt 2b
a
Reagents and conditions: (a) MeOH, clarification then AcOH; (b) EtOH, 65 °C, 86% (2 steps).
improved 89% yield and 94% purity (by HPLC). The solution
was then solvent swapped to ethanol and telescoped into the Boc
deprotection reaction.
Preparation of GSK2263167. In the initial route to
GSK2263167 conversion of oxadiazole 7 to sodium salt 2b
required three separate stages of chemistry (see Scheme 1,
reagents and conditions). After each stage the product was
isolated and purified by column chromatography. Since all three
stages were high yielding, it was decided to telescope them to
save on plant operating time and costs (Scheme 7).
For the Boc deprotection of compound 7 the HCl/dioxane/
DCM medicinal chemistry conditions were replaced with the
more environmentally friendly aqueous HCl/ethanol. The pH of
the resulting solution was then adjusted to pH 8 with 2 M NaOH
in the presence of ethyl acrylate to promote the alkylation
reaction and give ester 17. It was observed that the rate of
alkylation was optimal when the pH was above 8 and that the pH
dropped slowly during the reaction. The alkylation profile was
improved by carefully monitoring the pH and by the gradual
addition of 2 M NaOH during the reaction to keep the pH
around 8. Ester hydrolysis was effected by addition of 32%
aqueous NaOH. The improved process was run in the pilot plant,
and 9.4 kg of sodium salt 2b was isolated in 91% yield and 99%
purity (by HPLC).
The API sodium salt 2b used in early development studies was
found to be highly hygroscopic, making it unsuitable for further
development. A form screen highlighted free acid anhydrate 2a as
the most suitable physical form. Unfortunately, due to the
insolubility of different salts of compound 2 in a number of
solvents it was not possible to find a suitable solvent from which
acid 2a could be isolated directly. Instead sodium salt 2b was
dissolved in methanol, clarified, and then acidified with acetic
acid to precipitate out acid 2a as the hydrate (Scheme 8).
Slurrying the hydrate in ethanol at 65 °C for 1 h facilitated the
desired form change to free acid anhydrate which was
successfully monitored by Raman spectroscopy.17 The process
to convert sodium salt 2b into free acid 2a was scaled up to
produce 7.7 kg of anhydrate 2a in 86−88% yield and excellent
purity (>99% by HPLC).
PREPARATIONS FOR FURTHER SCALE-UP
■
After successful completion of the pilot plant campaign, two
areas were identified as requiring development prior to further
scale-up: avoiding of the use of stoichiometric palladium acetate
in the Saegusa oxidation to give phenol 4 and overcoming
limitations due to process safety restrictions in the conversion of
nitrile 5 to amidoxime 12.
D
dx.doi.org/10.1021/op400162p | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
A number of oxidation conditions were investigated as
alternatives to the Saegusa oxidation. The most promising
conditions involved the oxidation of enone 18 with copper(II)
bromide (Scheme 9).18
conversion, showing that lower amounts of hydroxylamine could
be tolerated.
With minimal further optimisation, both the continuous flow
process and the copper(II) bromide oxidation could provide
suitable alternatives should further scale-up be required.
Scheme 9. Copper(II) bromide oxidation of benzyl protected
enone 18
a
CONCLUSION
■
The original medicinal chemistry route to the S1P1 receptor
agonist GSK2263167 was unsuitable for further scale-up, and as
such, an alternative route to advanced intermediate 4 utilising a
two-step Robinson annulation/Saegusa oxidation sequence from
N-Boc piperidone 8 was evaluated and developed. The remaining
steps were optimised and all chromatography steps removed. A
number of stages were telescoped to avoid isolation of
intermediates, and a final stage was added to isolate the API as
the free acid anhydrate. The resulting eight stages of chemistry
were scaled up, first to prepare 470 g of compound 2a in 16.5%
overall yield and then 7.7 kg of compound 2a in 18.3% overall
yield. In addition, the obstacles to further scale-up were
addressed with the development of a CuBr2 oxidation process
avoiding the requirement for stoichiometric Pd(OAc)2 and a
continuous flow process for the preparation of amidoxime 12,
avoiding the process safety issues of heating hydroxylamine.
a
Reagents and conditions: (a) CuBr2, LiBr, MeCN, reflux, 86%; (b)
Pd/C, H2, EtOH, 80%.
Good conversion to phenol 19 could be achieved in just 15
min by refluxing benzyl protected enone 18 with 2 equiv of
CuBr2 in acetonitrile with LiBr as an additive, although lower
yields were observed using Boc-protected enone 9. Pleasingly,
the HBr salt crystallised out of solution, and the product could be
isolated by filtration in 86% yield. The residual copper and
lithium levels were efficiently reduced to <1 ppm by washing with
acetonitrile. The benzyl protecting group could be removed by
hydrogenation to give compound 20, the precursor to phenol 4
in the original medicinal chemistry route. These alternative
conditions not only significantly improved the yield of the
oxidation reaction but also replaced Pd(OAc)2 with cheaper
CuBr2 and simplified the workup, offering a viable alternative to
the Saegusa oxidation.
Further scale-up of the process to convert benzonitrile 5 to
amidoxime 12 was prevented due to the safety issues of using
hydroxylamine on scale. In order to prepare larger quantities of
material, either multiple batches would be required or an
alternative process would need to be developed. As a result, the
reaction of benzonitrile 5 with hydroxylamine as a continuous
flow process was investigated. Ideally flow reactions are fast (<1
h) and homogeneous. The reaction of aqueous hydroxylamine
with bezonitrile 5 in refluxing ethanol, however, is slow (22 h),
and the product crystallises out of solution. By running the
reaction in Vapourtec continuous flow equipment in n-butanol
with 20 equiv of hydroxylamine at 120 °C, and with a residence
time of 1 h under 8−9 bar pressure, excellent conversion to
amidoxime 12 was achieved (Figure 2). Pleasingly, the product
was shown to be soluble at the reaction temperature, addressing
the concerns around heterogeneity.
EXPERIMENTAL SECTION
■
tert-Butyl-5-methyl-6-oxo-3,4,6,7,8,8a-hexahydroiso-
quinoline-2(1H)-carboxylate (9). To a solution of N-Boc-4-
piperidone 8 (19.8 kg, 99.2 mol, 1 equiv) in toluene (87 L) was
added pyrrolidine (10.9 kg, 1.55 equiv). The solution was heated
to reflux for 2 h and the water removed by azeotropic distillation.
The solution was concentrated to 36 L by atmospheric
distillation then cooled to 20 °C. NMR analysis confirmed the
formation of the enamine and the removal of excess pyrrolidine.
Toluene (81 L) was charged, followed by a slurry of
hydroquinone (64 g, 0.5 mol %) in toluene (0.3 L), and the
reactor lights were switched off. The reactor jackets were set to
10 °C, and ethyl vinyl ketone (8.30 kg, 0.99 equiv) was charged.
The solution was heated to 40 °C for 1 h and analysed by HPLC
to confirm consumption of the ethyl vinyl ketone. The solution
was heated to 105 °C for at least 18 h before being cooled to 20
°C and sampled for analysis (<1% uncyclised Robinson
intermediate remaining by HPLC).
The organic phase was washed with saturated NH4Cl solution
(28.4 kg NH4Cl dissolved in 75 L water) followed by water (96
kg). The organic phase was concentrated to 65 L to give a toluene
solution of enone 9 (23.04 kg of crude product corrected for
toluene, 88% yield based on N-Boc piperidone 8) which was used
directly in the next stage. NMR data are based on a sample that
was concentrated to dryness to remove toluene. 1H NMR (400
MHz, CDCl3) δ 1.50 (s, 9H), 1.52−1.62 (m, 1H), 1.79 (s, 3H),
1.98−2.10 (m, 1H), 2.24−2.64 (m, 5H), 2.74 (dt, J = 15.87, 4.06
Hz, 1H), 3.05 (ddd, J = 13.20, 10.30, 3.70 Hz, 1H), 3.90−4.26
(m, 2H); ESI-MS (m/z) 210 [M − isobutylene + H]+, 192 [M −
tBuO−]+.
The reaction reached steady state after ∼2 residence times
with <2% residual bezonitrile 5. After 210 min the flow of
hydroxylamine was reduced to adjust the equivalents of
hydroxylamine to 16 equiv. This change had little effect on the
tert-Butyl-6-hydroxy-5-methyl-3,4-dihydroisoquino-
line-2(1H)-carboxylate (4). The solution of enone 9 in toluene
(23.0 kg in 65 L, 86.8 mol) was dried by adding toluene (115 L)
and performing an azeotropic distillation under reduced pressure
to concentrate back down to 65 L. Anhydrous 2-MeTHF (115 L)
was added to give a 14.4% w/w solution of enone 9 calculated by
NMR. This solution was divided and processed in two equal
batches.
Figure 2. Reaction profile of amidoxime 12 preparation using Vapourtec
continuous flow equipment.
E
dx.doi.org/10.1021/op400162p | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
To a solution of enone 9 (12 kg, 45.2 mol, 1 equiv) in 2-
MeTHF/toluene was added more 2-MeTHF (54 L) and the
reaction mixture cooled to 10 3 °C. Chlorotriethylsilane (7.57
kg, 1.11 equiv) was charged followed by 24% LiHMDS in THF
(45.6 kg, 1.45 equiv), keeping the internal temperature below 20
°C. The reaction was stirred for 30 min at 20 °C and analysed by
LC/MS which showed complete consumption of enone 9. The
organic phase was washed with water (2 × 54 L), and then the
hexamethyldisilazine was removed by concentrating to 30 L
under reduced pressure, adding 2-MeTHF (120 L) and
concentrating to 30 L again. 2-MeTHF (178 L) was added
followed by Pd(OAc)2 (11.5 kg, 1.13 equiv) and the reaction
stirred at 20 °C for 6 h.
A solution of 29% aqueous potassium formate (20.16 kg, 1.5
equiv) was added and the reaction stirred for 2 h (caution: gas
evolution was observed). A 1 M solution of TBAF in THF (20.4
kg, 0.5 equiv) was added and the reaction stirred for 1 h to cleave
the silyl protected product. The solid palladium was removed by
filtration through a 1-μm cloth (caution: highly flammable),
washing the solid with 2-MeTHF (2 × 24 L). The aqueous phase
was removed, and the organic phase was washed with water (60
L) and 1% w/w brine (60 L). The organic phase was dried by
azeotropic distillation under reduced pressure by concentrating
to 36 L, adding 2-MeTHF (180 L), and concentrating again to 36
L.
tert-Butyl-6-cyano-5-methyl-3,4-dihydroisoquinoline-
2(1H)-carboxylate (5). The solution of triflate 11 (7.90 kg, 20.0
mol, 1 equiv) in EtOAc from the previous stage was distilled
under vacuum (100 mbar, jacket set point 55 °C) and solvent
swapped into anhydrous DMF (40 L). The solution was then
heated to 90 °C for at least 2 h before being cooled to 70 °C.
Zn(CN)2 (3.05 kg, 1.3 equiv) and Pd(PPh3)4 (2.3 kg, 10 mol %)
were added, and the solution was then heated to 120 °C and
stirred for 5 h. The contents were cooled to room temperature.
Toluene (79 L) and water (79 L) were added and the contents
stirred before filtration. The filtrate was washed with toluene
(63.2 L), and the combined mixture was stirred before separation
of the phases. The aqueous phase was discarded, and the organic
phase was washed with 15% w/w aqueous brine (79 L) followed
by water (79 L) to give benzonitrile 5 as a ∼4% w/w solution in
toluene (85% yield) with 50% HPLC purity and 42%
triphenylphosphine oxide which was used directly in the next
stage. 1H NMR (400 MHz, CDCl3) δ 1.50 (s, 9 H), 2.47 (s, 3 H),
2.76 (t, J = 5.75 Hz, 2 H), 3.69 (t, J = 5.87 Hz, 2 H), 4.61 (s, 2 H),
7.04 (d, J = 8.07 Hz, 1 H), 7.45 (d, J = 7.82 Hz, 1 H); m/z (CI+)
t
216 [M − Bu + H+]; 172 [M − Boc + H+]; HRMS (CI+)
calculated 273.1591, found 273.1598 (Δ = −2.6 ppm); 910 ppm
Pd, 100 ppm Zn by ICP-OES.
(Z)-tert-Butyl 6-(N′-hydroxycarbamimidoyl)-5-methyl-
3,4-dihydroisoquinoline-2(1H)-carboxylate (12). The sol-
ution of benzonitrile 5 (9.12 kg, 1 equiv, 29.7 mol) in toluene
from the previous stage was solvent swapped into EtOH by
reduced pressure distillation. To the slurry of benzonitrile 5 in
EtOH (46 L) was charged 50% w/w aqueous NH2OH solution
in one portion (17.9 kg, 8 equiv) and the reaction heated to 80 °C
for at least 22 h. The slurry was cooled to 0 °C, aged for 3 h, and
filtered. The product was washed with cold EtOH (2 × 27 L),
deliquored, and dried under vacuum at 45 °C to yield amidoxime
12 as a crystalline white solid (7.95 kg, 75% yield) with 96%
HPLC purity. 1H NMR (400 MHz, DMSO-d6) δ 1.42 (s, 9 H),
2.21 (s, 3 H), 2.67 (t, J = 5.75 Hz, 2 H), 3.58 (t, J = 5.75 Hz, 2 H),
4.48 (br s, 1 H), 5.66 (br s, 1 H), 6.99 (d, J = 8.07 Hz, 1 H), 7.07
(d, J = 7.82 Hz, 1 H), 9.23 (s, 1 H); 13C NMR (101 MHz,
DMSO-d6) δ 15.7, 25.9, 28.1, 41.4, 45.1, 79.0, 123.2, 126.6, 132.9,
133.3, 133.8, 134.9, 152.5, 153.9; m/z (CI+) 306 [M + H+];
HRMS (CI+) calculated 306.1823, found 306.1808 (Δ = −4.9
ppm); 160 ppm Pd, <1 ppm Zn, <1 ppm Na by ICP-OES.
tert-Butyl-6-(5-(3-cyano-4-isopropoxyphenyl)-1,2,4-
oxadiazol-3-yl)-5-methyl-3,4-dihydroisoquinoline-
2(1H)-carboxylate (7). To a suspension of 3-cyano-4-
isopropoxybenzoic acid (5.64 kg, 1.05 equiv) in anhydrous
toluene (55 L) was added oxalyl chloride (4.3 kg, 1.30 equiv)
over 15 min. The reaction mixture was then stirred at 55 °C for 7
h until all the 3-cyano-4-isopropoxybenzoic acid was consumed
by HPLC. The resulting solution of acid chloride 6 was
concentrated to 28 L by distillation.
To a mixture of amidoxime 12 (7.95 kg, 1 equiv, 26.0 mol) in
anhydrous pyridine (39 kg) was added the solution of acid
chloride 6 in toluene over 30 min followed by a line wash of
toluene (8 L). The reaction mixture was heated to 40 °C and
stirred for 30 min until amidoxime 12 had been consumed by
HPLC. The reaction was then heated to 100 °C and stirred for
3.5 h until the level of the uncyclised intermediate dropped below
1%.
The reaction mixture was cooled to 25 °C, and EtOAc (96 L)
was added. The organic phase was washed once with water (40
L), four times with 1 M aqueous HCl (4 × 80 L) to remove the
pyridine, and finally with 5% w/w aqueous Na2CO3 (84 L) to
Heptane (12 kg) was added and the solution seeded with
phenol 4 (12 g). Crystallisation was promoted by adding heptane
(116 L) over 2 h, concentrating under reduced pressure to 60 L,
adding more heptane (144 L), cooling to 0 °C over 30 min, and
aging for 1 h. The solid was isolated by filtration, washed with
heptane (2 × 36 L), and dried in a vacuum oven at 45 °C to give
phenol 4 (11.05 kg from two batches, 48% yield) as an off-white
1
solid with >99% area purity by HPLC. H NMR (400 MHz,
CDCl3) δ 1.50 (s, 9H), 2.15 (s, 3H), 2.74 (t, J = 5.99 Hz, 2H),
3.65 (t, J = 5.87 Hz, 2H), 4.50 (s, 2H), 5.34 (br s, 1H), 6.67 (d, J =
8.07 Hz, 1H), 6.81 (d, J = 8.31 Hz, 1H); 13C NMR (101 MHz,
CDCl3) δ 11.0, 26.5, 28.6, (40.8, 41.9 rotamer pair), (45.3, 46.0
Boc rotamer pair), 80.1, 113.2, 122.2, 124.2, 125.4, 134.5, 152.6,
155.1; ESI-MS (m/z) 264 [M + H]+, 208 [M − isobutylene +
H]+; <80 ppm Pd by ICP-OES.
tert-Butyl-5-methyl-6-(((trifluoromethyl)sulfonyl)oxy)-
3,4-dihydroisoquinoline-2(1H)-carboxylate (11). Triflic
anhydride (14.3 kg, 1.2 equiv) was added to a solution of phenol
4 (11.1 kg, 1 equiv, 27.9 mol) in pyridine (29 L) at 0 °C over at
least 1.5 h, whilst maintaining the temperature of the contents
between 5 and 15 °C. Following addition the contents were
warmed to 25 °C over at least 30 min. EtOAc was added (111 L),
followed by water (55 L), and the mixture was stirred vigorously
for at least 5 min. The aqueous phase was removed, followed by
the organic layer. The aqueous phase was returned to the vessel,
and EtOAc (33 L) was added. The mixture was stirred vigorously
for at least 5 min, and the aqueous phase was discarded. The
organics from the previous steps were combined and washed
with aqueous 1 M HCl (3 × 66.3 L), and the aqueous phases
were discarded. The organic solution was washed with aqueous
5% w/w sodium carbonate solution (55 L), followed by water
(55 L) to give triflate 11 as a ∼16.7% w/w solution in EtOAc
(94% yield) with 97% HPLC purity which was used directly in
the next stage. 1H NMR (400 MHz, CDCl3) δ 1.50 (s, 9 H), 2.26
(s, 3 H), 2.77 (t, J = 5.87 Hz, 2 H), 3.69 (t, J = 5.87 Hz, 2 H), 4.58
(s, 2 H), 7.02 (d, J = 8.56 Hz, 1 H), 7.10 (d, J = 8.56 Hz, 1 H); m/
z (CI+) 339 [M − tBu + H+], 295 [M − Boc + H+]; HRMS (CI+)
calculated 396.1076, found 396.1087 (Δ = −2.7 ppm).
F
dx.doi.org/10.1021/op400162p | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
give oxadiazole 7 (10.58 kg, 89% yield) with 93.7% area purity by
HPLC as a solution in EtOAc/toluene. This solution was used
directly in the next stage. 1H NMR (400 MHz, CDCl3) δ 1.48 (d,
J = 6.11 Hz, 6H), 1.51 (s, 9H), 2.53 (s, 3H), 2.84 (t, J = 5.62 Hz,
2H), 3.72 (t, J = 5.75 Hz, 2H), 4.64 (s, 2H), 4.80 (spt, J = 6.11 Hz,
1H), 7.09−7.15 (m, 2H), 7.77 (d, J = 8.07 Hz, 1H), 8.33 (dd, J =
8.80, 2.20 Hz, 1H), 8.42 (d, J = 2.20 Hz, 1H); ESI-MS (m/z) 475
[M + H]+, 419 [M − isobutylene + H]+, 375 [M − Boc + H]+; <1
ppm Zn, <1 ppm Pd, 200 ppm Na by ICP-OES.
Sodium-3-(6-(5-(3-cyano-4-isopropoxyphenyl)-1,2,4-
oxadiazol-3-yl)-5-methyl-3,4-dihydroisoquinolin-2(1H)-
yl)propanoate (2b). The solution of oxadiazole 7 (10.1 kg, 1
equiv, 21.3 mol) in EtOAc/toluene from the previous stage was
solvent swapped into EtOH/water by concentrating to 30 L
under reduced pressure, adding absolute EtOH (150 L),
concentrating to 30 L again, and adding water (34 L). The
reaction was heated to 70 °C and treated with 5 M aqueous HCl
(21.7 kg) for 10 h.
Ethyl acrylate (3.1 kg, 1.45 equiv) was added and the solution
taken to pH 8 with 1.5 M aqueous NaOH (37.7 kg). The reaction
was stirred for 5 h at 70 °C, maintaining pH 7−8 by addition of
more 1.5 M aqueous NaOH (1.75 kg) until the alkylation was
complete by HPLC. Once the alkylation was complete, 32%
aqueous NaOH (3.20 kg) was added and the reaction stirred for
30 min at 70 °C.
The solution was cooled to 60 °C and aged for 1 h to allow the
product sodium salt to crystallise (the solution can be seeded if
necessary). The slurry was cooled to 0 °C over 6 h and aged for 1
h. The solid was filtered, washed with cold EtOH/water (2 × 30
L), and dried in the oven at 50 °C to give sodium salt 2b (9.40 kg,
92% yield) as an almost white solid with >99% purity by HPLC.
1H NMR (400 MHz, MeOD) δ 1.45 (d, J = 6.11 Hz, 6H), 2.47 (s,
3H), 2.52 (dd, J = 8.31, 7.34 Hz, 2H), 2.86−2.94 (m, 6H), 3.76
(s, 2H), 4.93 (spt, J = 6.10 Hz, 1H), 7.09 (d, J = 8.07 Hz, 1H),
7.42 (d, J = 9.05 Hz, 1H), 7.68 (d, J = 8.07 Hz, 1H), 8.36−8.42
(m, 2H); 13C NMR (101 MHz, methanol-d4) δ 16.9, 22.2 (2C),
28.3, 36.6, 52.1, 56.4, 57.3, 74.2, 104.7, 115.7, 116.3, 118.2, 125.8,
125.9, 128.9, 135.0, 135.4, 135.6, 137.8, 138.7, 164.4, 171.4,
174.7, 180.7; ESI-MS (m/z) 447 [M + H]+; 2.52% w/w water by
KF; <1 ppm Pd by ICP-OES.
3-(6-(5-(3-Cyano-4-isopropoxyphenyl)-1,2,4-oxadia-
zol-3-yl)-5-methyl-3,4-dihydroisoquinolin-2(1H)-yl)-
propanoic Acid (2a). A solution of sodium salt 2b (9.12 kg, 1
equiv, 20.4 mol) in MeOH (73 L) was filtered to remove
inorganics, and the vessel and line were washed with MeOH (9
L). AcOH was added (1.14 kg, 1.1 equiv), and the reaction was
stirred for 90 min, whereupon a slurry of the free acid 2a formed.
The solid was isolated by filtration and the cake washed with
MeOH (18 L). The isolated material was reslurried in EtOH (74
L) and aged at 78 °C for at least 1 h to turn over to the anhydrate.
The slurry was cooled to 0 °C and aged for at least 1 h before
filtration. The solid was washed with EtOH (18.4 L) and dried in
a vacuum oven overnight at 40 °C to yield acid 2a (7.65 kg, 88%
yield) as a crystalline white solid with >99%% HPLC purity. 1H
NMR (400 MHz, DMSO-d6) δ 1.40 (d, J = 6.11 Hz, 6 H), 2.44 (s,
3 H), 2.52 (s, 2 H), 2.71−2.84 (m, 6 H), 3.68 (s, 2 H), 4.99 (spt, J
= 6.11 Hz, 1 H), 7.12 (d, J = 8.07 Hz, 1 H), 7.56 (d, J = 9.05 Hz, 1
H), 7.67 (d, J = 8.07 Hz, 1 H), 8.40 (dd, J = 9.05, 2.20 Hz, 1 H),
8.49 (d, J = 2.20 Hz, 1 H); 13C NMR (101 MHz, DMSO-d6) δ
16.2, 21.5, 27.0, 31.9, 50.1, 52.9, 55.5, 72.5, 102.4, 114.8, 115.3,
116.0, 123.7, 124.3, 127.1, 133.7, 134.2, 134.5, 136.0, 137.7,
162.4, 169.3, 172.7, 173.5m/z (CI+) 447 [M + H+]; HRMS (CI+)
calculated 447.2027, found 447.2014 (Δ = −2.9 ppm); 3 ppm
Pd, <1 ppm Zn, <1 ppm Na by ICP-OES; ROI <0.1.
ASSOCIATED CONTENT
* Supporting Information
■
S
NMR data and Raman spectra for dehydrate and anhydrate of
acid 2a. This material is available free of charge via the Internet at
http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
*E-mail: robert.m.harris@gsk.com
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank Lynne Burrell and Laura Hook for key analytical
support; Robert Willacy and Sarah Vallance for salt-screening
and Raman spectroscopy work; Jamie Russell, Graeme March-
bank, Hardeep Sahota, Alistair Greenlaw, Keith Meaney, Hilary
Logan, Surabhi Mishra for plant support; and Andy Payne for
Process Safety work.
REFERENCES
■
(1) For recent reviews on sphingosine 1-phosphate receptor
modulation and S1P1 agonists see: (a) Bolli, M. H.; Lescop, C.;
Nayler, O. Curr. Top. Med. Chem. 2011, 11, 726. (b) Hla, T.; Brinkmann,
V. Neurology 2011, 76, S3−. (c) Cusack, K. P.; Stoffel, R. H. Curr. Opin.
Drug Discovery Dev. 2010, 13, 481.
(2) Hale, J. J.; Doherty, G.; Toth, L.; Mills, S. G.; Hajdu, R.; Keohane,
C. A.; Rosenbach, M.; Milligan, J, J.; Shei, G.-J.; Chrebet, G.; Bergstrom,
J.; Card, D.; Forrest, M.; Sun, S.-Y.; West, S.; Xie, H.; Homura, N.;
Rosen, H.; Mandala, S. Bioorg. Med. Chem. Lett. 2004, 14, 3501.
(3) Demont, E. H.; Andrews, B. I.; Bit, R. A.; Campbell, C. A.; Cooke, J.
W. B.; Deeks, N.; Desai, S.; Dowell, S. J.; Gaskin, P.; Gray, J. R. J.;
Haynes, A.; Holmes, D. S.; Kumar, U.; Morse, M. A.; Osborne, G. J.;
Panchal, T.; Patel, B.; Perboni, A.; Taylor, S.; Watson, R.; Witherington,
J.; Willis, R. Med. Chem. Lett. 2011, 2, 444.
(4) Heer, J. P.; Heightman, T. D. WO/2009/080724, 2009.
(5) Barth, M. M.; Binet, J. L.; Thomas, D. M.; de fornel, D. C.; Samreth,
S.; Schuber, F. J.; Renault, P. P. J. Med. Chem. 1996, 39, 2302.
(6) Corgero-Vargas, A.; Quiclet-Sire, B.; Zard, S. Z. Tetrahedron Lett.
2004, 39, 7335.
(7) Banerjee, A. K.; Pineda, J. J.; Mora, H. D.; Laya, M. S. Synth.
Commun. 2007, 37 (22), 3917.
(8) Le Quesne, P. W.; Allmam, K.; Abdel-Baky, S.; Onan, K. D.; Purdy,
R. H. Steroids 1989, 53 (6), 649.
(9) Mihailovic, M. L.; Forsek, J.; Lorenc, L. J. Chem. Soc., Perkin Trans. 1
1982, 1, 1.
(10) Holt, D. A.; Levy, M. A.; Ladd, D. L.; Oh, H. J.; Erb, J. M.; Heaslip,
J. I.; Brandt, M.; Metcalf, B. W. J. Med. Chem. 1990, 33, 937.
(11) Ito, Y.; Hirao, T.; Saegusa, T. J. Org. Chem. 1978, 43, 1011.
(12) Cisneros, L. O.; Wu, X.; Rogers, W. J.; Mannan, M. S.; Park, J.;
North, S. W. Process Saf. Environ. Prot. 2003, 81 (2), 121.
(13) Cisneros, L. O.; Rogers, W. J.; Mannan, M. S.; Li, X.; Koseki, H. J.
Chem. Eng. Data 2003, 48 (5), 1164.
(14) Cisneros, L. O.; Rogers, W. J.; Mannan, M. S. J. Hazard. Mater.
2001, 82 (1), 13.
(15) Iwata, Y.; Koseki, H. J. Hazard. Mater. 2003, 104 (1−3), 39.
(16) An accepted industry-wide standard based on Arrhenius kinetics
was used. A 30 °C safety margin was imposed from the exotherm
detection (onset). On the basis of instrument sensitivity, under adiabatic
conditions this equates to a 10 K temperature rise over 66 h of
processing time.
(17) See Supporting Information for Raman spectra of 2a hydrate and
2a anhydrate.
G
dx.doi.org/10.1021/op400162p | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
(18) (a) Morales-Alanis, H.; Brienne, M. J.; Jaques, J.; Bouton, M. M.;
Nedelec, L.; Torelli, V.; Tournemine, C. J. Med. Chem. 1985, 28 (12),
1796. (b) Kosower, E. M.; Wu, G.-S. J. Org. Chem. 1963, 28, 633.
H
dx.doi.org/10.1021/op400162p | Org. Process Res. Dev. XXXX, XXX, XXX−XXX