Development of a fully telescoped synthesis of the S1P1 agonist GSK1842799

Article abstract of DOI:10.1021/op2000095

The development of a fully telescoped synthesis of the potent and selective S1P1 agonist GSK1842799 is described. Key features in the synthesis, which has been implemented on a multikilogram scale, include a nucleophilic aromatic substitution to install a

Full text of DOI:10.1021/op2000095

ARTICLE
pubs.acs.org/OPRD
Development of a Fully Telescoped Synthesis of the S1P1 Agonist
GSK1842799
Michael S. Anson, Jonathan P. Graham,* and Alastair J. Roberts
Chemical Development, GlaxoSmithKline Research and Development Ltd., Gunnels Wood Road, Stevenage,
Hertfordshire SG1 2NY, U.K.
S Supporting Information
b
ABSTRACT: The development of a fully telescoped synthesis of the potent and selective S1P1 agonist GSK1842799 is described.
Key features in the synthesis, which has been implemented on a multikilogram scale, include a nucleophilic aromatic substitution to
install a lipophilic 1-octyloxy chain, introduction of a chiral quaternary centre, and the use of Lawesson’s reagent to form a thiadiazole
ring. Due to the lack of crystalline intermediates, workup protocols that took advantage of the lipophilic nature of the compounds
were developed. This allowed full combination of the five chemistry stages and a salt formation, with the only isolation being that of
the final hemifumarate salt of the drug substance. The synthesis of the O-phosphorylated active metabolite is also described.
’ INTRODUCTION
reasoned that we might be able to use it to our advantage in
designing workup procedures, in a fashion analogous to the use
of fluorous tags.4 Despite the synthetic route to 1 being far from
the notion of the ‘ideal synthesis’,5 with excess reagents required
and byproduct generated at each step, the lipophilic ‘tag’ might
allow us to approach ‘ideal purifications’, in which the product is
separated into a different phase from everything else that is in the
reaction mixture.6 In this manner, we might control the purity
during the synthesis through design of workup procedures rather
than relying on crystallisations or chromatography.
As well as the challenges posed by the physical properties of
the intermediates and drug substance, other issues also required
attention. Significant frothing had been reported in the nucleo-
philic aromatic substitution reaction between 4-fluoro-3-trifluor-
omethylbenzoic acid 4 and 1-octanol, with the volume of the
froth many times that of the reaction mixture. Control of the
frothing would be necessary in order to avoid poor throughput in
the preparation of 5 due to the requirement for substantial
headspace availability above the reaction mixture. We were also
concerned about the potential for damage to the reaction vessels
due to the fluoride byproduct, particularly upon acidification in
order to isolate the benzoic acid product. Conversion of benzoic
acid 5 to hydrazide 6 posed safety concerns, since a large excess of
hydrazine was employed and upon completion the reaction
mixture was concentrated to low volume by distillation prior to
extraction into dichloromethane. Although two (S)-2-methyl-
serine derivatives, 7a and 7b, had been successfully coupled
with hydrazide 6 in the gram-scale preparations of 1, a multi-
kilogram supply would be required. Most of the preparations of
1 had used the dimethyl oxazolidine 7a, which had been
prepared from (S)-2-methylserine for which there is not a com-
mercial source on multikilogram scale. The tert-butyl oxazoli-
dine 7b had been prepared from readily available L-serine
methyl ester hydrochloride using Seebach’s self-regeneration
GSK1842799 (Scheme 1, 1) is a potent and selective sphin-
gosine-1-phosphate receptor subtype 1 (S1P1) agonist and, as
such, has potential utility for the treatment of diseases or
conditions associated with inappropriate immune responses,
including transplant rejection and autoimmune diseases such as
multiple sclerosis and psoriasis.1,2 Phosphorylation of 1 in vivo
generates the active metabolite 2, which shows structural simi-
larity to naturally occurring S1P (3), a bioactive lipid mediator.
There are five subtypes of S1P-responsive receptors, and
GSK1842799 was designed to be selective for the S1P1 subtype.
In order to evaluate the clinical potential of GSK1842799, multi-
kilogram quantities of 1 were required. Additionally, gram-
quantities of the phosphate 2 were required to support preclinical
activities.
The original synthetic route to GSK1842799, used to prepare
the initial gram quantities of 1, is depicted in Scheme 2.3 The
general synthetic approach for the construction of the drug
substance was deemed suitable for the manufacture of multi-
kilogram quantities of 1, but several issues needed addressing
prior to further scale-up. The six-step synthesis provided 1 as the
hydrochloride salt in an overall yield of ∼20-30%. Although the
hydrochloride salt was initially isolated as a crystalline solid, it is
hygroscopic and deliquescent; thus, the identification and isola-
tion of a stable, crystalline version of 1 was required. Further-
more, all of the intermediates were isolated by stripping to dryness
under vacuum to yield waxy solids, gums, or viscous oils. The
greater than 100% uncorrected yields reported at two stages
reflect the challenge in isolating these compounds free from
residual solvents and impurities. Presumably, the degrees of
rotational freedom provided by the octyloxy side chain are in
large part responsible for the lack of crystalline intermediates. In
addition to handling issues, the lack of solid intermediates meant
that purification by crystallisation was not an available option,
and so chromatography had been required to control purity at
two stages. Although we recognised that the lipophilic side chain
would make the isolation of solid intermediates challenging, we
Received: January 13, 2011
Published: March 10, 2011
r
2011 American Chemical Society
649
dx.doi.org/10.1021/op2000095 Org. Process Res. Dev. 2011, 15, 649–659
|

Organic Process Research & Development
Scheme 1. GSK1842799 (1) and the active metabolite 2, which shows structural similarity to naturally occurring S1P (3)
ARTICLE
Scheme 2. Discovery Route to GSK1842799 1a
a All reported yields are uncorrected for purity.
of stereocentres approach,7 but when we analysed these samples by
chiral GC, they only had a modest purity of ∼93% ee and on
processing through to drug substance 1 did not produce material
with sufficient chiral purity. Furthermore, we had concerns over
the ability to successfully scale up the chemistry to 7b, which
includes a diastereoselective alkylation under cryogenic con-
ditions.8 In completing the synthesis of 1, the coupling of the
(S)-2-methylserine derivative with hydrazide 6 used the expen-
sive and thermally unstable coupling reagent HATU; the thia-
diazole formation used Lawesson’s reagent and so would require
the control of the associated stench, and the formation of
potentially genotoxic impurities in the deprotection stage would
need controlling or avoiding.
The development of the chemistry to address the above issues
and allow the manufacture of multikilogram quantities of 1
without recourse to any intermediate isolation is described
below. An approach to convert the alcohol 1 to the correspond-
ing phosphate 2 is also described.
’ RESULTS AND DISCUSSION
Nucleophilic Aromatic Substitution Reaction to Prepare
Octyloxy Benzoic Acid 5. The insertion of the octyloxy side
chain by a nucleophilic aromatic substitution reaction between
4-fluoro-3-trifluoromethylbenzoic acid 4 and 1-octanol under
basic conditions will inevitably generate a soaplike compound,
650
dx.doi.org/10.1021/op2000095 |Org. Process Res. Dev. 2011, 15, 649–659

Organic Process Research & Development
ARTICLE
Table 1. Nucleophilic aromatic substitution reaction for the
preparation of 5
and changing the base from sodium tert-butoxide to NaHMDS
led to a slightly improved purity profile and higher conversion
over 24 h (entry 8). The improvement in the reaction profile
might be due to full deprotonation of the 1-octanol with the
stronger base or due to changes in aggregation states or enhanced
solubility of the reaction components. Following a rapid optimi-
sation study, reaction completion could be achieved within 24 h
by using 2.8 equiv of NaHMDS and 2 equiv of 1-octanol (entry
9). Since a 1 M solution of NaHMDS in THF was initially used,
the increased base stoichiometry meant that the total solvent
volume increased to 13 vol. For the plant campaign we used a
more concentrated NaHMDS solution, but diluted with further
THF to maintain the same overall reaction dilution. Frothing in
the reaction was minimised by running 3-5 °C below the reflux
temperature and using minimum agitation to avoid ingress of
headspace gas into the reaction mixture. By using this protocol
the volume of froth observed was routinely estimated as sig-
nificantly less than 10% of the reaction volume. No etching of the
glassware during the extended reaction time at elevated tem-
perature was observed.
HPLC purity
@ 1 hb (%)
HPLC purity
@ 20 hb (%)
entrya
solvent
THFc
base
1
2
3
4
5
6
7
8
9f
KOtBu
57
0
60
0
THF
LiOtBu
THF
NaOtBu
NaOtBu
NaOtBu
NaOtBu
NaOtBu
NaHMDSe
NaHMDSe
25
75
<10
20
60
70
85
95
2-MeTHF
toluene
1,4-dioxane
1-octanold
THF
<10
<10
40
<10
20
THF
-
a Standard conditions: Benzoic acid 4 (1 equiv), 1-octanol (1.2 equiv),
and base (2.5 equiv) were heated in solvent (12 vol) to 65 °C for 20 h.
b Analysis by HPLC at 220 nm. % area product, adjusting for the solvent
c
peak where applicable, is reported. Diluted to 30 vol THF after 2 h.
d Heated to 130 °C after 2 h. e 1 M solution of NaHMDS in THF used.
No further THF added. f 2.0 equiv 1-octanol, 2.8 equiv NaHMDS (1 M
in THF).
With conditions now identified for the SNAr reaction, our
attention focused on the isolation of 5 from the reaction mixture,
which we required in high purity and in a suitable form for
onward processing. To do this, we needed to separate the octy-
loxy benzoic acid product 5 from the reaction solvent, excess
reagents (1-octanol and NaHMDS), byproduct (HMDS, NaF),
and impurities (including residual starting benzoic acid 4 and the
phenol analogue). Since precipitation of the product was not a
feasible option, due to its physical properties, and we wanted to
avoid the requirement for purification by chromatography, we
developed an extractive workup that utilised the lipophilicity of
the product imparted by the octyloxy chain (Figure 1). An initial
solvent swap into water by distillation was performed to remove
the THF. It was important to charge some water prior to
commencing the distillation in order to avoid excessive foaming.
Upon removal of the THF an aqueous slurry of the sodium
carboxylate product was obtained. Due to the high ionic strength
of the aqueous phase and the lipophilicity of 5 we were able to
extract the sodium carboxylate into an organic solvent, with
TBME being chosen for this purpose. The sodium fluoride
byproduct together with less lipophilic carboxylate impurities,
including unreacted 4, were removed to the aqueous waste. By
removing the fluoride waste whilst the reaction mixture was still
basic meant that we were able to avoid the generation of HF in
the workup, and no etching of glassware was observed in onward
processing. Upon treatment of the organic solution of the
sodium carboxylate product with fresh water, the salt preferen-
tially partitioned into the aqueous phase. The choice of TBME as
solvent for the initial extraction was important, since very little
(<1%) of the product was lost to the organic phase on extraction
into water. Organic impurities, including the HMDS and 1-octa-
nol, were removed to the organic waste, and further TBME
washes were performed to provide a clean, aqueous solution of
the sodium carboxylate. Upon acidification with aqueous HCl,
the carboxylic acid 5 could be extracted into an organic solvent.
For our first lab-scale batches we extracted into TBME and
isolated the product as a waxy solid by removal of the solvent
under vacuum. However, we wished to avoid processes involving
strips to dryness for transfer to our pilot plant, and thus switched
to extracting 5 into 2-MeTHF, since this should be compatible
with the downstream processing. The 2-MeTHF solution was
washed with water to remove any residual inorganics prior to
with a lipophilic tail and a charged, polar carboxylate headgroup.
If this ‘soap’ is generated under conditions that allow gas ingress
into the liquid phase, such as heating under reflux or with
vigorous agitation, then it would be expected that a significant
generation of froth or foam will occur. The conditions that had
been used in the initial gram-scale preparations used 2.5 equiv of
base (KOtBu) in refluxing THF, and therefore, it is unsurprising
that this led to foam generation. To avoid the formation of a
soaplike molecule, we could have considered protecting the
carboxylic acid (such as by converting to an ester) or inserting
the octyloxy chain later in the synthesis, but for expediency our
first choice was to retain the same synthetic route that had already
been used by the Discovery Group to produce 1. As such, we
began by screening the reaction conditions in an attempt to
identify a process whereby the product 5 could be formed
without a foam being generated.
In our hands, the initial conditions gave a rapid initial conver-
sion to 5, but as the reaction progressed the mixture became
gelatinous and did not proceed to completion, presumably due to
the very poor mixing (Table 1, entry 1). Since the counterion is
known to affect the rate of SNAr reactions9 and is also likely to
affect the physical properties of the reaction through differing
solubilities and potentially different aggregation states, we com-
pared the use of KOtBu with the lithium and sodium salts
(entries 2 and 3). No desired reaction was observed with LiOtBu,
whereas with NaOtBu the reaction did proceed. The initial rate
was slower compared to the that with the use of KOtBu, but
the reaction mixture remained a mobile slurry throughout and
proceeded further to completion in 24 h. An attempt to change
solvent from THF to 2-MeTHF, which would allow a higher
reaction temperature and potentially facilitate an easier workup,
was unsuccessful, since the reaction became gelatinous and
suffered from poor mixing from the start (entry 4). Other
solvents screened (entries 5-7) showed no advantage over
THF, including the use of 1-octanol as solvent (entry 7), which
required a higher reaction temperature to proceed. The reaction
in 1-octanol also resulted in significant levels of the phenol,
4-hydroxy-3-(trifluoromethyl)benzoic acid, which was only seen
at low levels in the other reactions. However, returning to THF
651
dx.doi.org/10.1021/op2000095 |Org. Process Res. Dev. 2011, 15, 649–659

Organic Process Research & Development
ARTICLE
Figure 1. Workup protocol for the isolation of 5 from the SNAr reaction.
azeotropic distillation to provide a dry solution of 5 in 2-MeTHF
with typically ∼95% solution yield and >95% area HPLC purity.
In our pilot-plant campaign, using a 7.2 kg input 4 batch size,
we achieved a lower than expected yield, with an average of 80%
across two batches. The drop in yield was attributed to physical
losses during the phase separations, with several litres of the
organic phase lost to waste in each phase split. We performed
both batches of the nucleophilic aromatic substitution chemistry
prior to commencing the next step, and the 2-MeTHF solution of
5 proved chemically and physically stable for more than one week
at ambient temperature (no change in HPLC profile and no
material out of solution). The ability to split the telescoped
process by isolating the intermediate as a solution greatly increased
the processing flexibility during the plant campaign.
Preparation of Hydrazide 6. Given the toxicity and safety
concerns with the use and handling of hydrazine, a modified
procedure was required for the formation of hydrazide 6. The
original process used six equivalents of hydrazine hydrate, and
the reaction was concentrated to low volumes upon reaction
completion, which gives a safety risk due to the potential to con-
centrate dry hydrazine. Following dilution with dichloromethane
and subsequent aqueous washes, the product 6 was isolated by
removal of the solvent under vacuum. Ideally, we wished to
isolate 6 without using a strip to dryness procedure and we were
also concerned about the compatibility of the hydrazide with
dichloromethane. The original process used N,N0-carbonyldii-
midazole (CDI) in THF for the activation of acid 5, and we found
that the activation also proceeded smoothly in 2-MeTHF at
ambient temperature. We retained a slight excess (1.3 equiv) of
CDI to ensure that the reaction would proceed to completion in
the presence of adventitious water. By performing a reverse
addition of the activated acid into a solution of hydrazine hydrate
in 2-MeTHF we were able to reduce the excess of hydrazine
hydrate required to two equivalents. Reducing the equivalents of
hydrazine hydrate further led to generation of two impurities,
which we assigned as the hydroxyoxadiazole 10 and the N,N0-
carbonyldihydrazide 11 (Figure 2). Both of these impurities
would form by reaction of the product hydrazide 6 with CDI,
followed by either an intramolecular cyclisation to form 10 or
reaction with a further equivalent of 6 to form 11. With sufficient
hydrazine hydrate, the excess CDI remaining after the acid
activation step will react to give predominately the acylimidazole
hydrazine 12, and thus, impurities 10 and 11 are not observed.
Effective mixing of the reaction and a controlled addition of the
activated acid into the hydrazine hydrate solution are also vital in
controlling these impurities. Using 2-MeTHF as the reaction
solvent, the hydrazine hydrate solution is initially biphasic. If the
activated acid is charged too quickly or without effective stirring,
then a local excess of hydrazine at the point of addition is not
achieved. We set an addition time of at least 20 min which, coupled
with modest agitation, allowed full conversion to 6 without
formation of impurities 10 and 11.
Isolation of 6 required the separation of the product from the
excess hydrazine, imidazole byproduct, and the acylimidazole
hydrazine 12. We were again able to utilise the lipophilicity of our
product to gain the required level of purity control in the workup.
Washing the 2-MeTHF solution twice with 5 volumes of 1 M
HCl ensured the removal of hydrazine, imidazole, and acylimi-
dazole hydrazine 12 without loss of any hydrazide 6 as the HCl
salt. A subsequent wash with aqueous potassium carbonate free-
bases the product and removes any residual benzoic acid 5. A dry
solution of 6 in 2-MeTHF was obtained by azeotropic distilla-
tion, with typically >95% solution yield and >95% area HPLC
purity. The 2-MeTHF solution of 6 is chemically and physically
652
dx.doi.org/10.1021/op2000095 |Org. Process Res. Dev. 2011, 15, 649–659

Organic Process Research & Development
ARTICLE
Figure 2. Impurities in the preparation of hydrazide 6.
Scheme 3. Approaches to (S)-(2)-methylserine derivatives
7a and 7b: (A) via diastereoselective salt resolution of racemic
2-methylserine; (B) from L-serine methyl ester via a diaster-
eoselective alkylation
The original conditions for the coupling of hydrazide 6 with
the dimethyl oxazolidine 7a used the expensive and potentially
hazardous coupling agent HATU in dichloromethane and DMF.
Since we had isolated hydrazide 6 as a solution in 2-MeTHF, we
ideally wished to remain in this solvent for the coupling, and
we wanted to move to a safer and less expensive coupling agent.
We reasoned that the use of CDI in 2-MeTHF would enable a
facile workup to provide a clean solution of the dicarbonyl 8a,
since the imidazole byproduct should be easy to remove with an
aqueous acidic wash, and if a small excess of the oxazolidine 7a
was used, then the excess could be removed with an aqueous
basic wash. However, the use of CDI had been investigated for
the coupling by the Discovery groups, but formation of the N,N0-
carbonyldihydrazide impurity 11, by reaction of hydrazide 6 with
CDI, had proved an issue and would need to be addressed. The
activation of acid 7a with CDI in 2-MeTHF proceeded cleanly,
but required gentle warming to 40 °C. Again, we chose to employ
a slight excess of CDI to ensure full activation of 7a. As expected,
addition of hydrazide 6 into this reaction mixture produced
impurities 10 and 11, as had been seen in the hydrazide forma-
tion reaction. This could be avoided by using a deficit of CDI, but
we wanted to avoid the sacrifice of the protected (S)-2-methyl-
serine whilst ensuring all hydrazide 6 was consumed, since we
considered that 6 would be more challenging to separate away
from our desired product. Given the steric hindrance around the
activated acid, we reasoned that it should be relatively stable and
that it might be possible to destroy the excess CDI without
reacting with the activated acid of 7a. To achieve this, we added
one molar equivalent of 2-propanol to the activated acid solution
and heated it at 40 °C for 30 min prior to addition of the
hydrazide 6. This resulted in the conversion to dicarbonyl 8a in
3-4 h at 40 °C, with none of the impurities 10 or 11 detected,
indicating that the IPA quench had successfully destroyed the
excess CDI. Sequential washes with 1 M HCl, 5% w/w aqueous
NaHCO3 solution, and water, followed by azeotropic distillation,
resulted in a dry 2-MeTHF solution of 8a. We assumed 100%
yield for this step and used the 2-MeTHF solution directly in the
next stage.
stable, and so we were able to run the hydrazide formation chem-
istry in plant and hold the output solution at ambient tempera-
ture for a month before coupling with the (S)-2-methylserine
derivative.
Coupling of Hydrazide 6 with a (S)-2-Methylserine Deri-
vative. There are many methods available for accessing quaternary
R-amino acids,10 including specific methods for 2-methylserine,11
but their preparation on a multikilogram scale can remain a
challenge. Indeed, we initially contracted a CRO partner to
synthesise 4 kg of the dimethyl oxazolidine 7a, but they were
unable to develop a suitable process that could deliver beyond a
gram scale. Given the difficulties our initial CRO partner had in
identifying a suitable process and in order to reduce supply risk, we
chose to explore two complementary methodologies (Scheme 3).12
These approaches were chosen for investigation on the basis of
potential for rapid scale-up and probability of delivering sufficient
material of appropriate quality.
The first approach, which involved the diastereomeric salt
resolution of racemic 2-methylserine methyl ester,13 was devel-
oped by Chirotech.14 The six-step synthesis provided over 15 kg
of 7a in >99% ee and about 9% overall yield from hydroxyac-
etone.12 The chemistry was simple, robust, and inexpensive and
hence suitable for rapid scale-up for large-scale production. The
second approach, developed in-house at GSK, involved the diaster-
eoselective alkylation of an L-serine derivative using self-regenera-
tion of stereocentres (SRS) methodology7,8,15 to access the tert-
butyl oxazolidine 7b. The four-step telescoped synthesis was used
to prepare over 20 kg of 7b in 98% ee and about 50% overall yield
from L-serinemethyl ester hydrochloride.12 The chemistry required
cryogenic conditions for the alkylation, but the route was shorter
and higher yielding than the resolution route. Both routes demon-
strate a scalable approach to N,O-protected (S)-2-methylserine and
allowed us to continue with the manufacture of 1. The remaining
discussion will focus on the use of the dimethyl oxazolidine 7a, but
the tert-butyl oxazolidine 7b can be used analogously.
Thiadiazole 9 Formation Using Lawesson’s Reagent. De-
spite the often clean conversion of carbonyl compounds to the
analogous thiacarbonyl using Lawesson’s reagent,16 the techni-
que often suffers difficulties in product isolation and purification
due to the formation of many byproducts. The use of Lawesson’s
reagent, particularly on an industrial scale, also requires the
control of stench, both during the reaction and in associated
waste streams. Thermal decomposition of Lawesson’s reagent
(13) generates two equivalents of the dithiophosphine ylide 14,
and it is this that is believed to be the active species (Scheme 4).
Carbonyl attack at the electrophilic phosphorous of the dithio-
phosphine ylide leads to the generation of a thiaoxaphosphetane,
which can then collapse to form the thiocarbonyl and an equi-
valent of the metathiophosphonate byproduct 15. It is assumed
653
dx.doi.org/10.1021/op2000095 |Org. Process Res. Dev. 2011, 15, 649–659

Organic Process Research & Development
ARTICLE
Scheme 4. Thiadiazole formation
that the formation of a thiadiazole ring proceeds via thiation
of both carbonyls prior to cyclisation, since one equivalent of
Lawesson’s reagent (13) is required and hydrogen sulfide is
generated (Scheme 4).17
system, were removed in the aqueous phase. Later running
impurities, which might include various oligomeric species,
remained in the organic phase. No efforts were made to identify
the impurities in the 2-MeTHF solution of 9a, but if they were
oligomers of 14 and 15 then we considered that they might be
hydrolysed under acidic conditions to the (di)thiophosphonate
species and removed in subsequent basic washes. Given that our
final chemistry step in the synthesis of 1 is an acidic deprotection,
we chose to proceed with the crude thiadiazole 9a solution and
assumed a 100% yield.
Deprotection of 9 and Isolation of a Crystalline Salt of 1.
Conversion of N-Boc oxazolidine 9a to 1 proceeded smoothly
upon heating the 2-MeTHF solution with aqueous HCl. We also
observed that many of the Lawesson’s reagent byproducts were
converted to the same polar species that we had removed in the
previous step, based on HPLC retention times. Washing the
2-MeTHF solution first with water then with aqueous sodium
hydroxide removed these impurities and resulted in a solution of
1 with >90% area purity by HPLC. Using the t-butyl oxazolidine
analogue 9b it was required to perform a distillation step after the
reaction to remove the pivalaldehyde byproduct, otherwise the
oxazolidine ring reformed upon neutralisation. Although 2-MeTHF
is more stable than THF to acid-catalysed ring-opening with
HCl, we were still concerned about the possibility of forming
potentially genotoxic alkyl chloride impurities upon prolonged
reaction time at elevated temperature.18 Use of dilute aqueous
sulfuric acid could replace HCl in the deprotection step and so
minimise the ring-opening of 2-MeTHF, and, by using aqueous
conditions, hydrolysis of any resulting alkyl sulfonate would
further minimise the level of any potential genotoxic impurity
formed.19 Alternative solvents, including isopropyl acetate, could
also replace 2-MeTHF in the deprotection step.
However, it is not just the hydrogen sulfide and metathiopho-
sphonate 15 that will need to be separated from the thiadiazole
product 9. Both the metathiophosphonate byproduct 15 and any
excess dithiophosphine ylide 14 can form various oligomers, and
they can also be hydrated, by adventitious water or during an
aqueous quench, to give the corresponding thiophosphonic acid
and dithiophosphonic acid, respectively. Indeed, when we ran the
original conditions for the conversion of the dicarbonyl 8a to
the thiadiazole 9a - 1.2 equivalents of Lawesson’s reagent in
toluene at 80 °C for 3-4 h - we saw complete conversion by
HPLC to the desired product, but also over twenty other UV active
components that we assigned as Lawesson’s reagent byproducts.
Some of these components were out of solution in toluene and could
be removed by filtration, but their removal on scale would
be problematic due to the high level of odour associated with
the solids. Attempts to perform aqueous washes on the toluene
solution led to the formation of emulsions, and whilst clean
thiadiazole 9a could be obtained by column chromatography,
this was not a viable option to perform on a multikilogram scale
due to the high quantities of odorous waste that would be
generated. Gratifyingly, when we treated the input 2-MeTHF
solution of 8a direct from the coupling reaction with 1.1 equiv of
Lawesson’s reagent at reflux, complete conversion to 9a was
achieved within 3-4 h. A similar HPLC profile was obtained in
2-MeTHF as for the reaction in toluene, but the reaction in
2-MeTHF remained in solution even on cooling at the end of the
reaction. We reasoned that treatment of this reaction mixture
with aqueous sodium hydroxide would allow removal of some of
the polar byproducts to the aqueous waste, such as the metathio-
phosphonate byproduct 15 by conversion to the corresponding
sodium thiophosphonate. Generation of the sodium salts should
also reduce the volatility of the compounds and so minimise the
issue of stench. Treatment of the 2-MeTHF solution with 5 M
NaOH gave clean phase separations and the polar reaction
components, which eluted early on the reverse-phase HPLC
Attempts to crystallise the free-base of 1 were all unsuccessful,
so a salt screen using 16 diverse and pharmaceutically acceptable
acids in six different solvent systems was performed. Crystalline
solids were obtained with benzoic acid, 2,5-dihydroxybenzoic
acid, 1,5-naphthalenedisulfonic acid, and fumaric acid, but only
fumaric acid gave a high melting (>100 °C) solid with a single
thermal event by DSC. Initially the salt was assigned with 1:1
654
dx.doi.org/10.1021/op2000095 |Org. Process Res. Dev. 2011, 15, 649–659

Organic Process Research & Development
ARTICLE
Scheme 5. New six-step telescoped synthesis of the hemifumarate salt of 1
stoichiometry, but later analysis showed it was the hemifumarate
salt and initial samples were contaminated with free fumaric acid.
Efforts to crystallise the hemifumarate salt of 1 from a 2-MeTHF
solution, with and without cosolvents, resulted in either low
recovery or the formation of noncrystalline materials. However,
with pure input free-base 1 and 0.5 equiv of fumaric acid a
solution was obtained in isopropyl acetate at 70 °C, with the
hemifumarate salt of 1 crystallising in good yield upon cooling.
To this end, a distillative solvent swap from 2-MeTHF to
isopropyl acetate was performed prior to the deprotection step.
The deprotection did not occur with fumaric acid but was facile
with aqueous sulfuric acid to yield, after aqueous sodium hydro-
xide washes, a solution of free-base 1 with >90% area purity by
HPLC. However, heating this crude solution to 70 °C with
fumaric acid gave a hazy, rather than clear, solution, and upon
cooling the product 1 was isolated with variable salt stoichiom-
etry (0.6-0.9 equiv of fumaric acid to 1). We hypothesized that
carry-over of base from the aqueous washes in the previous step
led to the formation of sodium fumarate, which contaminates the
output hemifumarate salt. Performing further water and brine
washes prior to the salt formation led to an improved situation,
but the stoichiometry of the salt remained >0.5 equiv. However,
the hot, hazy solution of the hemifumarate salt of 1 can be washed
with water, without loss of product, leading to a clear solution.
This solution can be clarified for the preparation of clinical grade
material by passing through a e5 μm filter, and the product
crystallises on cooling to provide the hemifumarate salt of 1 with
the correct stoichiometry. This process allowed the direct
isolation of the hemifumarate salt of 1 with >99% area HPLC
purity and >99.9% ee, in an overall yield of about 50% from
4-fluoro-3-(trifluoromethyl) benzoic acid (4) (Scheme 5). The
six-step telescoped process was successfully run in the pilot plant
to generate 10.4 kg of the hemifumarate salt of 1 across two
batches.
Synthesis of Phosphate Metabolite 2. The preparation of
20 g of the phosphate metabolite 2 was required to support
preclinical activities. Milligram quantities of 2 had previously
been prepared in the Discovery groups, in a four-step process
655
dx.doi.org/10.1021/op2000095 |Org. Process Res. Dev. 2011, 15, 649–659

Organic Process Research & Development
ARTICLE
Scheme 6. Initial route to phosphate 2
Scheme 7. New route for the preparation of phosphate 2
from GSK1842799 1 (Scheme 6).1a Protection of the amino
group provided the N-Boc compound 16, which was phosphory-
lated with bis-tert-butyldiethylphosphoramidite, oxidised with
hydrogen peroxide, and subsequently deprotected with TFA
to provide phosphate 2. This method had been reported to be
problematic, and in our hands we found that the oxidation and
deprotection steps were low yielding and the product difficult to
purify; as a result, we sought an alternative approach. Attempts to
directly phosphorylate 1 with POCl3 led to significant decom-
position, whereas treatment of the N-Boc compound 16 with
POCl3 with subsequent hydrolysis by treatment with water did
result in the preparation of 2. However, on scaling this chemistry
to the gram-scale the hydrolysis proved slow and generated 2 in
low purity, and again purification proved difficult.
Due to the difficulties experienced in purifying 2, we wished to
prepare a stable intermediate that we could purify before
performing a final deprotection step to yield 2 directly in high
purity (Scheme 7). We found that we could use the hemifuma-
rate salt of 1 directly in the N-Boc protection step, by treatment
with three equivalents of Boc-anhydride in a biphasic mixture of
toluene and aqueous sodium bicarbonate. After removal of the
aqueous phase, excess Boc-anhydride was destroyed by the
treatment with N-methylpiperazine. The resulting N-Boc-N0-
methylpiperazine, together with unreacted N-methylpiperazine,
was readily removed with a dilute HCl wash, providing a toluene
solution of 16 with 96% solution yield and >99% area purity
by HPLC on a 90-g scale. The toluene solution of 16 was
treated with diethylchlorophosphate and triethylamine at 50 °C,
to provide the diethylphosphate 17 with 96% area purity by
HPLC. Since we wanted high-purity material to go into the
final deprotection step, the crude 17 was purified by column
chromatography, with the pure fractions pooled to provide a
64% yield of 17 with >99% area purity by HPLC. Global
deprotection of 17 proved quicker and cleaner in dichloro-
methane rather than continuing in toluene, and the product
was isolated by solvent swapping to acetonitrile and crystal-
lising by addition of water to provide 51 g of 2 in 90% yield and
99% area purity by HPLC.
’ CONCLUSIONS
The development of a fully telescoped six-step synthesis
allowed the rapid and efficient manufacture of over 10 kg of
the hemifumarate salt of GSK1842799 (1). Gaining an under-
standing of the byproducts and impurities at each stage of the
chemistry and utilising the lipophilic nature of the compounds
allowed the design of efficient workup protocols that gave appro-
priate control of purity during the synthesis. As well as avoiding
all chromatography and intermediate isolations, the new process
controls the frothing in the nucleophilic aromatic substitution,
demonstrates the safe use of hydrazine, avoids expensive cou-
pling agents, and controls stench issues with the use of Law-
esson’s reagent. Furthermore, two scaleable approaches to (S)-2-
methylserine derivatives have been developed,12 and a multi-
gram-scale preparation of the phosphate active metabolite 2 has
been performed.
656
dx.doi.org/10.1021/op2000095 |Org. Process Res. Dev. 2011, 15, 649–659

Organic Process Research & Development
ARTICLE
’ EXPERIMENTAL SECTION
1.20-1.46 (19H, m), 1.49 (3H, s), 1.54 (3H, s), 1.64 (3H, s),
1.69-1.78 (2H, m), 3.84 (1H, d, J = 8.8 Hz), 4.12 (1H, d, J =
8.8 Hz), 4.19 (2H, t, J = 6.2 Hz), 7.37 (1H, d, J = 9.3 Hz), 8.17 (2H,
d, J = 7.6 Hz), 9.52 (1H, d, J = 26.8 Hz), 10.50 (1H, d, J = 58.2 Hz).
1,1-Dimethylethyl-(2R,4S)-2-(1,1-dimethylethyl)-4-methyl-
4-[(2-{[4-(octyloxy)-3-(trifluoromethyl)phenyl]carbonyl}-
hydrazino)carbonyl]-1,3-oxazolidine-3-carboxylate (8b).
A 2-MeTHF solution of 8b was prepared in an analogous fashion
4-(Octyloxy)-3-(trifluoromethyl)benzoic Acid (5). A 40%
solution of NaHMDS in THF (46 kg, 2.8 equiv) was diluted with
THF (51 L), and 1-octanol (9.5 kg, 2.0 equiv) was charged and
the reaction warmed to 55-60 °C. A solution of 4-fluoro-
3-(trifluoromethyl)benzoic acid (4) (7.2 kg, 1 equiv) in THF
(15 L) was charged, washing through with further THF (2 L).
The reaction mixture was stirred at 55-60 °C until deemed
complete by HPLC (∼24 h, <4% area 4 remaining). The reaction
was then solvent exchanged into water (75 L) by distillation to
afford a tan slurry that was extracted into TBME (75 L). Water
(75 L) was charged to the organic layer and the product extracted
into the aqueous layer. The aqueous layer was washed with
TBME (4 ꢀ 37 L), then acidified with 5 M HCl (22 L) and
extracted into 2-MeTHF (37 L). The 2-MeTHF solution was
washed with water (3 ꢀ 37 L) and dried azeotropically by put-
and-take distillation with 2-MeTHF (3 ꢀ 37 L) to provide a dry
2-MeTHF solution of 5 (26.8% w/w; 34.4 kg, 84%) with 95%
area purity by HPLC for use directly in the next stage. 1H NMR
(400 MHz, DMSO-d6) δ 0.85 (3H, t, J = 7.0 Hz), 1.18-1.36
(8H, m), 1.37-1.47 (2H, m), 1.69-1.78 (2H, m), 4.18 (2H, t,
J = 6.3 Hz), 7.35 (1H, d, J = 8.8 Hz), 8.09 (1H, d, J = 2.0 Hz), 8.16
(1H, dd, J = 8.8, 2.0 Hz).
4-(Octyloxy)-3-(trifluoromethyl)benzohydrazide (6). CDI
(5.86 kg, 1.3 equiv) was charged to the 2-MeTHF solution of 5
(26.8% w/w; 34.2 kg, 1 equiv), rinsing in with 2-MeTHF (8 L).
The mixture was stirred at 25 ( 5 °C until activation was deemed
complete by HPLC (∼15 min, e2% a/a 5 remaining). The
solution was then added over about 20 min to hydrazine hydrate
(3.69 kg, 2 equiv) in 2-MeTHF (29 L), rinsing in with further
2-MeTHF (6 L). The resultant reaction mixture was stirred at
25 ( 5 °C until the reaction was deemed complete by HPLC
(∼30 min, e2% a/a activated intermediate remaining). The
reaction was washed successively with 1 M HCl (2 ꢀ 58 L) and
13% w/w aqueous potassium carbonate solution (35 L), then
dried azeotropically by put-and-take distillation with 2-MeTHF
(3 ꢀ 29 L) to provide a dry 2-MeTHF solution of 6 (16.1% w/w;
55.9 kg, 93%) with 96% area purity by HPLC for use directly in
the next stage. 1H NMR (400 MHz, DMSO-d6) δ 0.84 (3H, t, J =
6.8 Hz), 1.18-1.35 (8H, m), 1.36-1.46 (2H, m), 1.66-1.76
(2H, m), 4.14 (2H, t, J = 6.3 Hz), 4.49 (2H, bs) 7.31 (1H, d,
J = 9.5 Hz), 8.07-8.12 (2H, m), 9.86 (1H, bs).
1,1-Dimethylethyl-(4S)-2,2,4-trimethyl-4-[(2-{[4-(octyloxy)-
3-(trifluoromethyl)phenyl]carbonyl}hydrazino)carbonyl]-
1,3-oxazolidine-3-carboxylate (8a). Carboxylic acid 7a (7.31 kg,
1.05 equiv) was dissolved in 2-MeTHF (45 L) and CDI (5.18 kg,
1.2 equiv) was charged, washing in with 2-MeTHF (2 L). The
reaction was heated to 40 ( 5 °C and stirred at this temperature for
1 h, when HPLC analysis showed the reaction was complete.
2-Propanol (1.8 L) was charged and the reaction stirred at 40 (
5 °C for 30 min. The 2-MeTHF solution of 6 (16.1% w/w; 55.4 kg,
1 equiv) was charged, washing in with 2-MeTHF (2 L), and the
resulting solution was stirred for 3 h at 40 ( 5 °C, when HPLC
analysis showed the reaction was complete. The mixturewas cooled
to 20 ( 5 °C, and washed successively with 1 M HCl (2 ꢀ 27 L),
5% w/w aqueous NaHCO3 solution (36 L) and water (18 L). The
organic layer was concentrated to 80 L by distillation and dried
azeotropically by put-and-take distillation with 2-MeTHF (3 ꢀ
45 L) to provide a dry 2-MeTHF solution of 8a (assume 100%
yield; 92% area purity by HPLC) for use directly in the next stage.
1H NMR (400 MHz, DMSO-d6) δ 0.85 (3H, t, J = 6.7 Hz),
1
as above, using carboxylic acid 7b input. H NMR (400 MHz,
DMSO-d6) δ 0.85 (3H, t, J = 6.7 Hz), 0.96 (9H, s), 1.22-1.37
(10H, m), 1.46 (9H, s), 1.63 (3H, s), 1.71-1.86 (2H, m), 3.86
(1H, d, J = 8.6 Hz), 4.18 (2H, t, J = 6.2 Hz), 4.35 (1H, d, J = 8.8
Hz), 5.09 (1H, s), 7.36 (1H, d, J = 8.6 Hz), 8.16-8.22 (2H, m),
9.63 (1H, s), 10.65 (1H, s).
1,1-Dimethylethyl-(4S)-2,2,4-trimethyl-4-{5-[4-(octyloxy)-
3-(trifluoromethyl)phenyl]-1,3,4-thiadiazol-2-yl}-1,3-oxazo-
lidine-3-carboxylate (9a). Lawesson’s reagent (11.9 kg, 1.1 equiv)
was charged to the above solution of 8a in 2-MeTHF, washing in
with 2-MeTHF (2 L). The reaction was heated to 70 ( 5 °C for
4 h, when HPLC analysis showed the reaction was complete. The
reaction was cooled to 20 ( 5 °C and washed with 5 M NaOH
(2 ꢀ 30 L), and then with 13% w/w brine (20 L). The organic
phase was solvent exchanged into isopropyl acetate by vacuum
distillation and adjusted to 80 L to provide a solution of 9a in
isopropyl acetate (assume 100% yield; 80% area purity by HPLC)
for use directly in the next stage. 1H NMR (400 MHz, DMSO-d6)
δ 0.85 (3H, t, J = 7.1 Hz), 1.15-1.45 (19H, m), 1.57 (3H, s), 1.67
(3H, s), 1.71-1.78 (2H, m), 1.88 (3H, s), 4.08-4.24 (4H, m),
7.41 (1H, d, J = 8.9 Hz), 8.10-8.24 (2H, m).
1,1-Dimethylethyl-(2R,4S)-2-(1,1-dimethylethyl)-4-methyl-
4-{5-[4-(octyloxy)-3-(trifluoromethyl)phenyl]-1,3,4-thiadiazol-
2-yl}-1,3-oxazolidine-3-carboxylate (9b). An isopropyl acet-
ate solution of 9b was prepared in a fashion analogous to that
above, using 8b input. 1H NMR (400 MHz, DMSO-d6) δ 0.85
(3H, t, J = 7.0 Hz), 0.93(9H, s), 1.22-1.37 (17H, m), 1.39-1.48
(2H, m), 1.71-1.78 (2H, m), 1.87 (3H, s), 4.15 (1H, d, J =
8.7 Hz), 4.20 (2H, t, J = 6.2 Hz), 4.37 (1H, bd, J = 8.5 Hz), 5.14
(1H, s), 7.41 (1H, d, J = 8.9 Hz), 8.16 (1H, d, J = 2.1 Hz), 8.23
(1H, dd, J = 8.9, 2.1 Hz).
(2S)-2-Amino-2-{5-[4-(octyloxy)-3-(trifluoromethyl)phenyl]-
1,3,4-thiadiazol-2-yl}-1-propanol Hemifumarate (Hemifu-
marate Salt of 1). Water (9 L) and conc. sulfuric acid (3.3 kg,
1.2 equiv) were added to the above solution of 9a in isopropyl
acetate and the reaction was warmed to 70 ( 5 °C, until deemed
compete by HPLC (∼13 h). Water (45 L) was added and the
reaction cooled to 20 ( 5 °C. The reaction was washed three
times with 2 M NaOH (27 L then 2 ꢀ 45 L), then 13% w/w brine
(20 L). The solution was adjusted to 80 L by addition of
isopropyl acetate (35 L), and fumaric acid (1.51 kg) was charged.
The mixture was heated to 70 °C to provide a hazy solution, then
cooled to 65 ( 3 °C and washed with water (2 ꢀ 18 L). The
solution was dried azeotropically by put-take distillation with
further isopropyl acetate (45 L), clarified through a 5 μm filter,
and the lines were washed through with further isopropyl acetate
(45 L). The resulting solution was concentrated to 89 L then
cooled to 60 ( 3 °C and seeded with crystals of the hemi-
fumarate salt of 1 (9 g). The resulting slurry was aged at 60 (
3 °C for 1 h and then allowed to cool to 20 ( 5 °C. The solids
were isolated by filtration, washed with isopropyl acetate (3 ꢀ
21 L), and dried under vacuum at 50 ( 5 °C to provide a white
solid of the hemifumarate salt of 1 (5.3 kg, 41%), with >99.9% ee
and >99% area purity by HPLC. Typical differential scanning
657
dx.doi.org/10.1021/op2000095 |Org. Process Res. Dev. 2011, 15, 649–659

Organic Process Research & Development
ARTICLE
calorimetry (DSC)/thermalgravimetric analysis (TGA) shows
melt at ∼111-114 °C. 1H NMR (400 MHz, DMSO-d6) δ 0.85
(3H, t, J = 6.9 Hz), 1.20-1.37 (8H, m), 1.37-1.49 (2H, m), 1.47
(3H, s), 1.70-1.79 (2H, m), 3.52 (1H, d, J = 10.5 Hz), 3.69 (1H,
d, J = 10.5 Hz), 4.19 (2H, t, J = 6.2 Hz), 4.90-6.35 (3H, bs), 6.58
(1H, s), 7.41 (1H, d, J = 8.8 Hz), 8.11 (1H, d, J = 2.0 Hz), 8.15
(1H, dd, J = 8.7, 2.1 Hz). HRMS (ES þve; MHþ) calcd for
C20H29F3N3O2S 432.1933, found 432.1935.
The reaction was performed in an analogous fashion using 9b
input, but following addition of water post reaction completion,
the trimethyl acetaldehyde byproduct was removed by distilla-
tion with further isopropyl acetate (5 vol).
dichloromethane to acetonitrile by distillation under vacuum at
20 °C, to provide 400 mL of an acetonitrile solution of 2. A 5%
aqueous acetonitrile solution (160 mL) was charged over 1 h and the
solids were collected by filtration, washed with acetonitrile (4 ꢀ
150 mL) and dried under vacuum at 40 °C to provide 2 (51.4 g,
1
90%) as a white solid with 99% area purity by HPLC. H NMR
(400MHz, CD3OD) δ0.91 (3H, t, J= 6.8 Hz), 1.25-1.44 (8H, m),
1.48-1.58 (2H, m), 1.80-1.89 (2H, m), 1.90 (3H, s), 4.20 (2H, t,
J=6.2Hz), 4.32(1H, dd, J= 11.2, 4.8 Hz), 4.39 (1H, dd, J= 11.2, 5.2
Hz), 7.35 (1H, d, J = 8.7 Hz), 8.18 (1H, dd, J = 8.7, 2.1 Hz), 8.22
(1H, d, J = 2.1 Hz). HRMS (ES þve; MHþ) calcd for
C20H30F3N3O5PS 512.1596, found 512.1593.
1,1-Dimethylethyl-((1S)-2-hydroxy-1-methyl-1-{5-[4-(octy-
loxy)-3-(trifluoromethyl)phenyl]-1,3,4-thiadiazol-2-yl}ethyl)
carbamate (16). The hemifumarate salt of 1 (90 g, 1 equiv) was
slurried in toluene (1.5 L) and Boc-anhydride (121 g, 3 equiv)
was charged, rinsing in with further toluene (0.3 L). A solution of
NaHCO3 (77 g, 5 equiv) in water (1 L) was charged and the
reaction warmed to ∼40 °C for 17 h then at 50 °C for a further
4 h, when the reaction was shown to be complete by HPLC. The
reaction was cooled to 20 °C and the lower aqueous phase
removed. The aqueous phase was back-extracted with toluene
(0.2 L), and the organic phases were combined. 1-Methylpiper-
azine (61 mL, 3 equiv) was charged and the reaction stirred for
30 min (gas evolution observed for the first 10 min after 1-methyl-
piperazine addition). The reaction was washed with 1 M HCl
(2 ꢀ 900 mL) then 6.5% w/w aqueous NaHCO3 solution
(0.5 L), and the toluene solution was dried over MgSO4 and
concentrated to 0.54 L by distillation to provide a solution of 16
in toluene (19.4% w/w; 484 g; 96%) with >99% area purity by
HPLC. 1H NMR (400 MHz, CDCl3) δ 0.89 (3 H, t, J = 7.0 Hz),
1.22-1.39 (8H, m), 1.42 (9H, s), 1.45-1.55 (2H, m), 1.79 (3H,
s), 1.79-1.86 (2H, m), 3.82 (1H, bs), 3.92 (1H, t, J = 9.3 Hz),
4.07 (2H, t, J = 6.5 Hz), 4.10 (1H, bs), 5.75 (1H, s), 7.01 (1H, d,
J = 8.7 Hz), 8.03 (1H, dd, J = 8.7, 2.1 Hz), 8.09 (1 H, d, J = 2.1 Hz).
1,1-Dimethylethyl ((1S)-2-{[bis(ethyloxy)phosphoryl]oxy}-
1-methyl-1-{5-[4-(octyloxy)-3-(trifluoromethyl)phenyl]-1,3,
4-thiadiazol-2-yl}ethyl)carbamate (17). The solution of 16 in
toluene (19.4% w/w; 484 g) was treated with triethylamine
(123 mL, 5 equiv) and diethylchlorophosphate (64 mL, 2.5 equiv)
and warmed to 50 °C for 6 h, when the reaction was shown to be
complete by HPLC. The reaction was cooled to 20 °C, and a
solution of NaHCO3 (37 g, 2.5 equiv) in water (560 mL) and
EtOAc (560 mL) was charged. The lower aqueous phase was
removed, and the organics were dried over MgSO4 and concen-
trated under vacuum to provide crude 17 (147 g) with 96% area
purity by HPLC. The crude product was purified in four runs by
column chromatography on SiO2, eluting with EtOAc/heptanes,
and the pure fractions were combined and concentrated under
vacuum to provide 17 (74.8 g, 64%) as an oil with >99% area purity
by HPLC. 1H NMR (400 MHz, CDCl3) δ 0.89 (3H, t, J = 7.0 Hz),
1.22-1.39 (8H, m), 1.34 (6H, t, J = 6.9 Hz), 1.43 (9H, s), 1.43-
1.54 (2H, m), 1.80-1.88 (2H, m), 1.89 (3H, s), 4.08 - 4.19 (6H,
m), 4.40 (1H, dd, J = 10.3, 6.2 Hz), 4.53 (1H, dd, J = 10.3, 6.2 Hz),
5.90 (1H, s), 7.07 (1H, d, J = 8.9 Hz), 8.05-8.12 (2H, m).
(2S)-2-Amino-2-{5-[4-(octyloxy)-
’ ASSOCIATED CONTENT
S
Supporting Information. Methods for in-process mon-
b
itoring and purity analysis for each reaction. This material is
available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION
Corresponding Author
jonathan.p.graham@gsk.com.
’ ACKNOWLEDGMENT
We thank Antonella Carangio for initial demonstration of the
CDI-activated coupling of the (S)-methylserine derivative 7a
with hydrazide 6, and Matthew Walker and Nick Shipley for
investigations into routes to the phosphate 2. We appreciate the
efforts of Tom Ramsey for pilot plant support; Paul Evans for
Process Safety work; and Kelly Goulston and Fiona King for key
analytical support. We acknowledge Caroline Day, Leanda Kindon,
and Grahame Woolham for the salt-screening activities.
’ REFERENCES
(1) (a) Evindar, G.; Deng, H.; Bernier, S.; Yao, G.; Coffin, A.; Yang,
H. WO/2008/016692, 2008. (b) Anson, M. S.; Day, C. J.; Graham, J. P.;
Kindon, L. J.; Woolham, G. R. WO/2009/098193, 2009.
(2) For a recent review on S1P1 agonists, see Cusack, K. P.; Stoffel,
R. H. Curr. Opin. Drug Discovery Dev. 2010, 13, 481.
(3) Initial discovery and preparation of GSK1842799 was performed
by Praecis Pharmaceuticals, Inc., and multigram preparations of 1 were
performed under contract by Albany Molecular Research, Inc.
(4) For a review on the use of fluorous tags in chemical synthesis, see
Zhang, W. Chem. Rev. 2009, 109, 749.
(5) Wender, P. A.; Miller, B. L. In Organic Synthesis: Theory and
Applications; Hudlicky, T., Ed.; Jai Press: Greenwich, 1993; Vol. 2, pp
27-66; Wender, P. A. Chem. Rev. 1996, 96, 1.
(6) Curran, D. P. Angew. Chem., Int. Ed. 1998, 37, 1174.
(7) For a review on the concept of self-regeneration of stereocentres,
see Seebach, D.; Sting, A. R.; Hoffmann, M. Angew. Chem., Int. Ed. 1996,
35, 2708.
(8) Decomposition of the lithium enolate of methyl (2R,4S)-2-(1,1-
dimethylethyl)-3-formyl-1,3-oxazolidine-4-carboxylate has been noted
even at -78 °C, see Seebach, D.; Aebi, J. D. Tetrahedron Lett. 1984,
25, 2545.
(9) Reinheimer, J. D.; Kieffer, W. F.; Frey, S. W.; Cochran, J. C.; Barr,
E. W. J. Am. Chem. Soc. 1958, 80, 164.
(10) For reviews on the synthesis of R,R-disubstituted R-amino
acids, see (a) Vogt, H.; Br€ase, S. Org. Biomol. Chem. 2007, 5, 406.
(b) Ohfune, Y.; Shinada, T. Eur. J. Org. Chem. 2005, 5127. (c) With, T.
In Organic Synthesis Highlights IV; Schmalz, H.-G., Ed.; Wiley: Wein-
heim, 2000; pp 26-33.
3-(trifluoromethyl)phenyl]-1,3,4-thiadiazol-2-yl}propyl Di-
hydrogen Phosphate (2). Bromotrimethylsilane (91.6 g, 5.3 equiv)
was charged to a solution of 17 (74.7 g) in dichloromethane
(750 mL) and the reaction stirred at 20 °C for 20 h. Further bromo-
trimethylsilane (8.1 g, 0.5 equiv) was charged and the reaction stir-
red for a further 3 h. The reaction was solvent exchanged from
658
dx.doi.org/10.1021/op2000095 |Org. Process Res. Dev. 2011, 15, 649–659

Organic Process Research & Development
ARTICLE
(11) For a review on the synthesis of 2-methylserine, see Nakano, K.;
Kotsuki, H.; Ichikawa, Y. Org. Prep. Proced. Int. 2008, 40, 67.
(12) Anson, M. S.; Clark, H. F.; Evans, P.; Fox, M. E.; Graham, J. P.;
Griffiths, N. N.; Meek, G.; Ramsden, J. A.; Roberts, A. J.; Simmonds, S.;
Walker, M. D.; Willets, M. Org. Process Res. Dev. 2011, 10.1021/
op100299d.
(13) Jones, M. A.; Hislop, A. D.; Snaith, J. S. Org. Biomol. Chem. 2006,
4, 3769.
(14) Chirotech Technology, Ltd., A subsidiary of Dr. Reddy’s Labora-
tories (EU) Ltd.
(15) (a) Seebach, D.; Aebi, J. D.; Gander-Coquoz, M.; Naef, R. Helv.
Chim. Acta 1987, 70, 1194. (b) Brunner, M.; Saarenketo, P.; Straub, T.;
Rissanen, K.; Koskinen, A. M. P. Eur. J. Org. Chem. 2004, 3879–3883.
(16) For recent reviews on Lawesson’s reagent, see Ozturk, T.; Ertas,
E.; Mert, O. Chem. Rev. 2007, 107, 5210. Jesberger, M.; Davis, T. P.;
Barner, L. Synthesis 2003, 13, 1929.
(17) Rasmussen, P. B.; Pedersen, U.; Thomsen, I.; Yde, B.; Lawesson,
S.-O. Bull. Soc. Chim. Fr. 1985, 62.
(18) Aycook, D. F. Org. Process Res. Dev. 2007, 11, 156.
(19) Teasdale, A.; Eyley, S. C.; Delaney, E.; Jacq, K.; Taylor-Worth,
K.; Lipczynski, A.; Reif, V.; Elder, D. P.; Facchine, K. L.; Golec, S.;
Oestrich, R. S.; Sandra, P.; David, F. Org. Process Res. Dev. 2009, 13, 429.
659
dx.doi.org/10.1021/op2000095 |Org. Process Res. Dev. 2011, 15, 649–659