Development of a scaleable process for the synthesis of the A2a agonist, UK-371,104

Article abstract of DOI:10.1021/op700248t

The development and utilization of a scaleable process for the manufacture of the A2a agonist UK-371,104 (1) is described. Key steps in the synthesis include (i) a palladium-catalyzed cyanation reaction to prepare the nitrile 10, (ii) a telescoped convers

Full text of DOI:10.1021/op700248t

Organic Process Research & Development 2008, 12, 575–583
Development of a Scaleable Process for the Synthesis of the A2a Agonist,
UK-371,104
Stephen Challenger, Yann Dessi, Darrell E. Fox,† Lynsey C. Hesmondhalgh, Pierre Pascal, Alan J. Pettman, and
Julian D. Smith*
Chemical Research and DeVelopment, Pfizer Global Research and DeVelopment, Sandwich Laboratories, Ramsgate Road,
Sandwich, Kent CT13 9NK, United Kingdom
Abstract:
The development and utilization of a scaleable process for the
manufacture of the A2a agonist UK-371,104 (1) is described. Key
steps in the synthesis include (i) a palladium-catalyzed cyanation
reaction to prepare the nitrile 10, (ii) a telescoped conversion of
the acid 11 to the glycosidation substrate 9, and (iii) the stereo-
selective coupling of 8 with 9 in a glycosidation reaction mediated
by TMS triflate in 1,2-dimethoxyethane followed by conversion
through to 1.
Figure 1. Structure of UK-371,104 (1).
Introduction
The search for novel anti-inflammatory agents has led to
as a dry-powder formulation without the need to resort to the
preparation of a salt form. In order to supply preclinical and
phase 1 clinical trials, it was necessary for us to find a scaleable
process for the manufacture of multikilogram quantities of 1.
Initially we conducted a paper evaluation of the synthetic route
used by our Discovery Chemistry colleagues (Scheme 1),
considering key attributes and potential vulnerabilities if it were
to be used as a basis for a large-scale manufacturing process.
The route was a linear sequence, using guanosine (2) as the
starting material, that elegantly installed the key carbonyl
function at position 2 on the adenine ring Via palladium-
catalyzed aminocarbonylation of the iodide 7 in the final step.
The conversion of 2 to the 2-amino-6-chloro derivative 4 Via
the 2-amino-6-hydroxy derivative 3 was accomplished using
the procedure published by Robins and Uznan´ski.3 Formation
of the 2-iodo intermediate 5 followed the method described by
Matsuda et al., employing a diazotization reaction.4 The 6-chloro
substituent of the iodide 5 was then selectively displaced by
2,2-diphenylethylamine to give the triacetate 6, which was then
deprotected to the carbonylation substrate 7. This procedure was
used by the Discovery Chemistry Department to provide UK-
371,104 on multigram scale.
When considering the Discovery Chemistry route as a
possible way to manufacture multikilogram amounts of 1 to
supply clinical development, we felt that several aspects would
cause significant problems on scale-up. The route lacked
convergency, utilized potentially hazardous diazotization chem-
istry, and carried a potentially labile aryl iodide through two
stages of the synthesis. In addition, the last stage of the process
involved a palladium-catalyzed carbonylation reaction, which
would have been technically challenging for us to carry out
the discovery of a wide variety of biological targets that have
provided medicinal chemists with opportunities to develop novel
synthetic ligands of therapeutic relevance. The adenosine
receptor is one such target. The A2a subtype has been the
subject of significant interest in the search for drugs that agonize
this receptor which could lead to new therapies for the treatment
of diseases such as chronic obstructive pulmonary disease
(COPD), for which there is a high unmet medical need.1
Colleagues in the Pfizer Discovery Chemistry Department
at Sandwich undertook a program of research to identify A2a
agonists for the inhaled treatment of COPD. Such compounds
were designed to selectively agonize the A2a receptor sites on
the surface of neutrophils in the lungs of patients. The delivery
of the active agent directly to the site of action would require
the successful candidate to be formulated for inhaled delivery,
using dry-powder technology. A fundamental requirement for
this formulation approach was for the active pharmaceutical
ingredient (API) to possess a suitable crystalline solid form.
Research in our Discovery Chemistry Department resulted in
the identification of UK-371,104 (1, Figure 1), a crystalline
nucleoside derivative, as a development candidate.2
UK-371,104 (1) is a functionalized nucleoside that is
structurally related to adenosine and contains an unusual amide
function at position 2 of the adenine ring together with a
diphenylethylamino substituent at position 6. As the free base,
UK-371,104 is a crystalline material with a melting point of
182 °C, and its physical properties are suitable for development
* To whom correspondence should be addressed. E-mail: julian.smith@
pfizer.com.
† Present address: Chemical Research and Development, Pfizer Global
Research and Development, Groton Laboratories, Eastern Point Road, Groton,
CT 06340, U.S.A.
(3) Robins, M. J.; Uznan´ski, B. Can. J. Chem. 1981, 59, 2601.
(4) Matsuda, A.; Shinozaki, M.; Yamaguchi, T.; Homma, H.; Nomoto,
R.; Miyasaka, T.; Watanabe, Y.; Abiru, T. J. Med. Chem. 1992, 35,
241.
(1) Sullivan, G. W. Curr. Opin. InVest. Drugs 2003, 4 (11), 1313.
(2) Mantell, S. J.; Monaghan, S. M. WO 0077018, 2000.
10.1021/op700248t CCC: $40.75
Published on Web 05/21/2008
2008 American Chemical Society
Vol. 12, No. 4, 2008 / Organic Process Research & Development
•
575

Scheme 1. Discovery Chemistry synthesis of UK-371,104
under cGMP in the facilities available to us at the time, and
could also carry toxic palladium residues into the final API.
We therefore elected to pursue a more convergent strategy for
the large-scale manufacture of 1, avoiding the issues associated
with the Discovery Chemistry route, in order to support bulk
demands for toxicology and clinical trials.
dihydro-2H-pyran in ethyl acetate in the presence of catalytic
p-toluenesulfonic acid at 50 °C.6 The product 13 was isolated
in 85% yield following workup and crystallization from ethyl
acetate. Regioselective displacement of the 6-chloro substituent
in 13 was accomplished by adaptation of the method of Bhakuni
et al. by treating 13 with 2,2-diphenylethylamine at reflux in
2-propanol in the presence of Hu¨nig’s base.7 The chloride 14
Results and Discussion
Retrosynthetic Analysis. Our retrosynthetic analysis for 1
is shown in Figure 2 and involves a late-stage, convergent
coupling of 1,2,3,5-tetra-O-acetyl-ꢀ-D-ribofuranose (8) with an
adenine unit 9 that is suitably functionalized at C-2 and C-6.
Such glycosidation reactions are well-known in the literature
to proceed with silylated heterocyclic bases under the influence
of Lewis acids to afford products with high levels of stereose-
lectivity at the anomeric position on the sugar.5 We elected to
utilize the tetrahydropyranyl (THP)-protected adenine nitrile 10
as a precursor to 9 since a synthesis to this compound was
known from our colleagues in Discovery Chemistry,2 and its
hydrolysis to the derived carboxylic acid 11 was expected to
be straightforward. Disconnection of 10 then led back to
commercially available 2,6-dichloropurine (12).
Synthesis of the Nitrile 10. In common with work carried
out in the Discovery Chemistry Department on analogues related
to 1, we began our synthesis from 2,6-dichloropurine (12) as
shown in Scheme 2. The steps to the protected chloroadenine
14 did not require any significant development prior to scale-
up. Protection of the nitrogen at N-9 of the purine ring using a
tetrahydropyran group was accomplished using the method of
Robins et al. by treating 2,6-dichloropurine (12) with 3,4-
(5) (a) Vorbru¨ggen, H; Ruh-Pohlenz, C. Org. React. 2000, 55, 1. (b)
Vorbru¨ggen, H Acta Biochim. Pol. 1996, 43 (1), 25. (c) Vorbru¨ggen,
H.; Ho¨fle, G. Chem. Ber. 1981, 114, 1256.
Figure 2. Retrosynthetic analysis for UK-371,104 (1).
576
•
Vol. 12, No. 4, 2008 / Organic Process Research & Development

Scheme 2. Synthesis of the Nitrile 10
nitrile 10 to the carboxylic acid 11 using lithium hydroxide in
a refluxing mixture of water and Industrial Methylated Spirit
(IMS) gave the desired compound, but the formation of a very
thick suspension of the lithium salt of the product made agitation
difficult. Fortunately, substitution of lithium hydroxide for
sodium or potassium hydroxide avoided this problem, and since
sodium hydroxide gave a slightly cleaner reaction profile, this
base was selected for further development.
It was necessary to convert the sodium salt of the product
11 into the crystalline free acid since direct processing of a metal
salt in the next step was found to be unsuccessful. In breaking
the salt, we were concerned that prolonged contact of 11 with
the low pH conditions required to generate the free acid could
potentially result in loss of the THP group in the product, and
we reasoned that this would be more likely in the presence of
the IMS reaction solvent. Thus, we thought that it would be
advantageous to isolate the sodium salt of the product 11 from
the reaction mixture in order to remove it from the alcoholic
solvent prior to acid treatment. We found that once the nitrile
hydrolysis was complete, the solvent could be exchanged with
water, leading to direct crystallization of the sodium salt of 11,
which could then be isolated by filtration and fed into the salt-
break stage of the workup process.
The isolated sodium salt of 11 was then converted to the
free acid of 11 which could be crystallized from ethyl acetate.
Initially, it was found that suspending the sodium salt between
water and dichloromethane and then adjusting the pH with
hydrochloric acid effectively gave a solution of the free acid
which could then simply be isolated Via a solvent exchange
from dichloromethane into ethyl acetate followed by crystal-
lization. However, emulsions were commonly encountered
during the extractive part of the process, leading to poor and
time-consuming phase separations. The cause of this problem
was found to be the presence of small amounts of solid material
that was present in the extraction mixture. This solid was found
to be the highly insoluble unprotected carboxylic acid 17, which
was thought to be derived from unprotected nitrile 15 present
in the starting material 10 or from loss of the THP group in 11
during the acid treatment. As detailed earlier, we were able to
suppress the level of the impurity 15 during the synthesis of
the nitrile 10 but were unable to eliminate it completely. A
screen of extraction solvents failed to find anything that was
superior to dichloromethane in terms of phase separation. In
cases where phase separation was impossible, we found that
filtration of the mixture through a filter aid removed a sufficient
amount of the insoluble fine solid to allow a stable interface to
form.
was isolated in 90% yield by direct crystallization from the
reaction mixture on cooling.
The method used by our colleagues in Discovery Chemistry
to convert the chloride 14 to the nitrile 10 utilized a protracted
sequence involving displacement of the chloro substituent by
methanethiolate, followed by oxidation to the sulfone and
subsequent displacement with cyanide.2 We chose to follow a
more direct method to convert the chloride 14 to the nitrile 10
by employing a palladium-catalyzed reaction with zinc cyanide
using the method described by Gundersen.8 We found that the
best conditions were to use Pd(PPh3)4 (3 mol%), zinc cyanide
(0.60 equiv) in DMF at 80-85 °C with triethylamine as a base
to deactivate the acidic byproduct generated during the reaction.
It was necessary to strike a balance between reaction conversion
and the formation of impurities 15 and 16 arising from loss of
the THP group from the starting material and product, respec-
tively. A screen of alternative bases failed to lead to any
improvement. Under optimum conditions on scale, the nitrile
10 was isolated in 63% yield as a crystalline solid. The key
impurities present in the product were the chloride starting
material 14 together with the THP-deprotected compounds 15
and 16. The latter impurities in particular had the potential to
negatively impact the phase separations carried out during
aqueous workup of the subsequent step because of their very
low solubility; thus, their presence in 10 required some degree
of control.
Since 15 and 16 were much less soluble than 10, they could
not be easily removed by selective solvent extractions or
recrystallization. The most effective way to remove these
impurities from 10 was to dissolve the impure material in
refluxing 2-propanol, then add a filter aid and filter the hot
mixture. By carrying out this procedure, the amount of 15 and
16 in 10 could be controlled to acceptable levels.
Conversion of the Nitrile 10 to the Glycosidation Sub-
strate 9. With the nitrile 10 in hand, we examined its conversion
to the glycosidation substrate 9 (Scheme 3). Hydrolysis of the
Scale-up of this process in the laboratory using up to 180 g
of the nitrile 10 gave the carboxylic acid 11 in 84-88% yield.
Some problems were encountered during the crystallization of
the sodium salt of the product where oiling was observed, but
these were overcome by carrying out the initial crystallization
at 50 °C following a slow cooling ramp from reflux temperature.
Since the two batches of the nitrile 10 used contained relatively
high levels of the insoluble impurity 16, phase separations
during the aqueous workup presented some challenges, and it
was found necessary to filter mixtures through a filter aid in
order to remove the insoluble materials. In one case this resulted
(6) Cassidy, F.; Olsen, R. K.; Robins, R. K. J. Heterocycl. Chem. 1968,
5, 461.
(7) Bhakuni, D. S.; Gupta, P. K.; Chowdhury, B. L. Ind. J. Chem., Sect.
B. 1984, 23, 1286.
(8) Gundersen, L.-L. Acta Chem. Scand. 1996, 50, 58.
Vol. 12, No. 4, 2008 / Organic Process Research & Development
•
577

Scheme 3. Conversion of nitrile 10 to glycosidation substrate 9
in initial loss of product on the filter aid, but we were able to
recover this material by rewashing the filter cake. We also found
that during the crystallization of the free acid of 11 from ethyl
acetate, a very thick slurry was produced that was very difficult
to stir. Fortunately, dilution with additional ethyl acetate gave
mobile mixtures without negatively impacting the yield. Further
scale-up in our kilo laboratory facility proceeded without major
issue, leading to the isolation of two 1.7-kg batches of 11, both
in 86% yield, although the product was found to be less pure
than previous batches made in the laboratory. We attributed
this to the different source of the starting nitrile 10, which was
subsequently found to be impure. However, the impurities were
found to purge well in the next stage of the synthesis. Once
again the main processing problem was the phase separation,
which took longer than expected upon scale-up to reveal a clean
interface, but it was not necessary to pass the mixture through
a filter aid.
The next stage of the process was to convert the carboxylic
acid 11 to the glycosidation substrate 9. At the outset of this
work, we envisioned that the coupling of 11 with 1-(2-
piperidinyl)ethylamine could be simply achieved by using N,N′-
carbonyldiimidazole to activate the carboxyl function to give
the intermediate acyl imidazolide 18 followed by addition of
the amine. Using dry THF as the solvent, we found that this
protocol worked well, but we did find that monitoring the
formation of the acyl imidazolide 18 was somewhat trouble-
some. Initially, after screening various reagents to quench the
active species prior to HPLC analysis, we elected to quench
samples of the reaction mixture with concentrated aqueous
ammonium hydroxide, and then examined the resultant car-
boxamide by reverse-phase HPLC. It is interesting to note that
we did not find the competing reaction of the acyl imidazolide
with water to be a significant problem using this method.
Unfortunately, the coupled product 19 was not crystalline
and could only be isolated by evaporation to dryness to give a
gum. However, the glycosidation substrate 9 was crystalline,
and since this is the product of the next stage of the synthetic
sequence, we elected to telescope the two steps together.
Deprotection of THP groups in compounds such as 19 is
generally carried out under aqueous acidic conditions. Using
excess aqueous hydrochloric acid in IMS gave the desired
product 9 as a hydrochloride salt, but this salt-form proved to
have poor filtration properties and to be insufficiently soluble
in organic solvents to make it useable directly in the subsequent
step. Use of 2 equivalents of p-toluenesulfonic acid in aqueous
ethanol also worked but suffered from similar issues. We also
had concerns about the reaction of such strong acids with
alcohols leading to the presence of genotoxic impurities13 in
the product 9; thus, we proceeded to examine trifluoroacetic
acid in aqueous 2-propanol. Whilst these conditions gave the
monotrifluoroacetate salt of 9 as a monohydrate by a direct-
drop process in 76% yield, we found that this form of 9 failed
to react in the next step of the synthetic sequence. We therefore
chose to directly break the salt and isolate the free base form
from the reaction. This was accomplished by simple basification
of the reaction mixture to approximately pH 10 using aqueous
sodium hydroxide to give a slurry of the free base which could
be filtered directly. Care was taken since under more basic
conditions the solubility of the product in the solvent mixture
increased, which we attributed to deprotonation at the ring
nitrogen in 9. Since the filtration rates of the slurries generated
in this manner were slow, we found it to be advantageous to
subject the slurry to a heat-cool cycle. This had the effect of
modestly improving the rate of filtration, presumably through
slightly increasing the particle size of the bulk material Via an
Ostwald ripening process.
Whilst it was pleasing to find that the acid 11 could be
converted through to the glycosidation substrate 9 in good yield,
we subsequently found that the free base of 9 was actually a
monohydrate based on Karl Fischer analysis of the product.
Since the next step in the synthetic route required the use of a
very moisture-sensitive reagent (TMS triflate), we needed to
find a way to remove the water from 9 prepared using this
process. Prolonged drying reduced the water content in 9, but
not sufficiently to be useful. Poor solubility of 9 in water-
immiscible organic solvents limited the opportunity to control
water through an extractive workup. However, we found that
recrystallizing hydrated 9 from n-propanol, incorporating azeo-
tropic removal of water from the system by distillation, gave
crystalline 9 free base that was suitable for use in the next
578
•
Vol. 12, No. 4, 2008 / Organic Process Research & Development

Scheme 4. Conversion of the glycosidation substrate 9 to
UK-371,104 (1)
silylated heterocyclic bases under the influence of Lewis acids.
With regard to regioselectivity (i.e., alkylation at N-7 vs N-9
of 9) experience from our colleagues in Discovery Chemistry
highlighted that selectivity for N-9 alkylation was complete in
the glycosidation reactions of related adenine derivatives. In
such cases, it appears that regioselectivity is controlled by the
bulky diphenylethylamino substituent at the C-6 position
sterically blocking the N-7 position towards reaction with the
oxonium ion.
The protocol that is usually employed when attaching sugar
units to heterocyclic bases is the presilylation of the heterocycle
with a reagent such as hexamethyldisilazane or N,O-bis-
trimethylsilylacetamide, followed by reaction with a peracylated
sugar under the catalysis of a perfluoroalkyl triflate, which can
be generated in situ.9 As such, the Vorbru¨ggen modification of
the Hilbert-Johnson method has found great utility in the
preparation of nucleoside derivatives on preparative scale. We
examined these conditions and other related methods for their
potential to be used in the large-scale synthesis of 1, but found
that the reactions were capricious and often difficult to drive to
completion. Of the conditions initially screened, we found that
a combination of excess TMS triflate (to silylate 9 as well as
promoting ionization of 8) and DBU in acetonitrile10 showed
some promise, but further work highlighted that these conditions
were not sufficiently robust to be considered as scaleable. A
screen of a range of Lewis acids as alternatives to TMS triflate
did not yield any useful alternatives.
In a related project previously worked on by some colleagues
in the Pfizer laboratories in Groton, U.S.A., a similar glycosi-
dation process had been carried out by combining a heterocycle
with stoichiometric amounts of TMS triflate in 1,2-dimethoxy-
ethane (DME) followed by addition of an acetylated sugar
derivative.11 Our colleagues found that using DME as the
solvent was uniquely good for this type of reaction. In our hands,
the DME/TMS triflate combination performed extremely well,
and despite DME being a nonideal solvent for large-scale
manufacture, we found that its unique advantages outweighed
the issues associated with its use on scale.
reaction. Testing of the product by Karl Fischer analysis
indicated that the product still contained low levels of water
(up to 1% w/w), which suggests that either 9 crystallizes from
n-butanol with some retained water or that it slowly absorbs
water on exposure to a moist atmosphere prior to testing.
Although we were unable to fully examine the propensity of 9
to hydrate, low levels of water in 9 could be tolerated, and
material made in this way performed acceptably well in the
subsequent step, enabling us to proceed with our synthesis.
Scale-up of the conversion of the acid 11 to the glycosidation
substrate 9 in the laboratory proceeded without issue, and two
249-g batches of 11 were processed to give a total of 457 g of
9 in an average yield of 77%. Two 1.7-kg batches of 11 were
then processed in our kilo laboratory facility to give two 1.1-
kg batches of 9 in an average yield of 62%. The lower yield
obtained compared to the yields from the laboratory batches
was attributed to the higher impurity burden present in the
starting material (as mentioned previously), with the impurities
being purged in this step.
Conversion of the Glycosidation Substrate 9 to UK-
371,104 (1). The coupling of 9 with the peracetylated sugar 8
was viewed at the outset as being the most challenging step
for the development of a scaleable process. Moreover, we had
found that the coupled product 20 (Scheme 4) was not
crystalline and could not be purified by nonchromatographic
methods. Therefore a high-yielding and clean glycosidation
process was sought so that this intermediate could be generated
and then processed directly to 1 without any intervening
purification.
There are many methods that have been used for the stereo-
and regioselective attachment of sugar moieties to heterocyclic
bases to give nucleoside derivatives, and they generally rely
on the formation of an oxonium ion (or equivalent) through
ionization at the anomeric position of a suitably activated sugar
followed by addition of the heterocycle, often being rendered
more nucleophilic via silylation.5 The level of stereoselectivity
formed at the new anomeric centre can be a problem in these
reactions, but we were fortunate that peracylated ribose deriva-
tives such as the commercially available tetraacetate 8 are well-
precedented to undergo highly stereoselective addition of
Treatment of a suspension of 9 in DME with 2 equiv of
TMS triflate at 50-60 °C resulted in the formation of a clear
yellow solution. Below 50 °C a solid crystallized out from the
reaction mixture, which we presumed was the silylated triflate
salt 21 (Scheme 5), although we did not isolate or characterize
this intermediate. Subsequent addition of the peracetylated sugar
8 as a solution in DME resulted in the formation of the triflate
salt 22 of the product over a period of a few hours at 50-60
°C. The reaction was very highly stereoselective, and we did
not observe any of the R-anomer when analyzing the reaction
mixtures or the downstream products in the synthetic route.
The use of alternative silylating agent/acid combinations in
DME was briefly examined. Neither trimethylsilyl chloride/
TMS triflate, hexamethyl disilazane/triflic acid, N,O-bis(trim-
ethylsilyl)acetamide/BF3 ·OEt2 nor N,O-bis(trimethylsilyl)ace-
tamide/SnCl4 gave efficient reactions, but use of N,O-
bis(trimethylsilyl)acetamide/triflic acid did give clean reaction
(9) Vorbru¨ggen, H.; Bennua, B. Chem. Ber. 1981, 114, 1279.
(10) Kristinsson, H.; Nebel, K.; O’Sullivan, A. C.; Struber, F.; Winkler,
T.; Yamaguchi, Y. Tetrahedron 1994, 50, 6825.
(11) Fox, D. E.; Scott, R. W. Unpublished results.
Vol. 12, No. 4, 2008 / Organic Process Research & Development
•
579

Scheme 5. Working hypothesis for the mechanism of the
of 1, even from solvents such as DME. This led us to consider
the possibility that 20 could be hydrolyzed directly after the
glycosidation step, thereby giving 1 that could be crystallized
from the reaction mixture by the addition of water. It was found
that 1 only dissolved sufficiently well in two solvents, tert-amyl
alcohol and dichloromethane, to allow some sort of extractive
workup to be considered. However, use of dichloromethane
resulted in the formation of emulsions, and tert-amyl alcohol
gave poor recoveries following workup and crystallization.
The conversion of pure isolated 20 to 1 under transesteri-
fication conditions (e.g., EtOH/EtONa) was a clean reaction,
and the UK-371,104 (1) could be crystallized directly from the
reaction mixture, which suggested to us that this type of direct-
drop approach might be useful on scale. However, for such a
process to be used on reaction mixtures from the glycosidation
stage, at least two additional equivalents of base would be
needed to quench the triflic acid/TMS triflate present in the
glycosidation reaction mixture. In addition, it was likely that
removal of at least some of the DME would be required to
ensure that the concentration of the alcohol solvent in the
mixture was high enough to ensure reaction completion, given
that transesterification is an equilibrium process. Disappoint-
ingly, when a reaction mixture from the glycosidation step was
quenched with ethanol and then treated with sodium ethoxide,
no crystallization of 1 occurred despite a clean reaction profile.
These factors, together with a desire to subject the glycosidation
reaction to some sort of aqueous quench and neutralization to
deactivate any residual TMS triflate or triflic acid generated
prior to subsequent processing led us to consider an alternative
approach.
Conversion of purified 20 to 1 was clean and rapid under
standard hydrolytic conditions in DME, requiring at least 3
equiv of dilute NaOH to remove the acetate groups. Interest-
ingly, use of aqueous K2CO3 or aqueous Na3PO4 was not as
effective as NaOH in these reactions. Following completion of
the hydrolysis reaction, dilution of the resulting mixture in DME
with water afforded crystalline UK-371,104 that was filtered
from the mixture in approximately 60% yield over the two
stages from 9. When more concentrated NaOH (5 M) was used,
it was noticed that two liquid phases formed, which was
presumably a consequence of the high ionic strength of the
aqueous phase. This suggested to us that it might be possible
to remove aqueous-soluble impurities from the process by
carrying out a phase separation, and that this might lead to a
more reliable crystallization step when the organic phase was
diluted with water.
glycosidation reaction
product. However, since workup of such reactions would
generate stoichiometric amounts of acetamide, a category 2B
carcinogen that would need to be controlled to ppm levels in
the API, we elected to develop and scale the process using TMS
triflate alone.
Our working hypothesis for the mechanism of the glycosi-
dation reaction is shown in Scheme 5. We assumed that 9 was
silylated at N-9 to give the intermediate 21 which then reacts
with the stabilized oxonium ion 19 to deliver the triflate salt
22. The high level of stereoselectivity for the ꢀ-anomer obtained
through the Lewis acid-promoted reaction of silylated hetero-
cycles with acyl-protected ribose derivatives such as 8 is well
precedented and is thought to be a consequence of the stabilizing
acyloxy group shielding the R-face of the sugar ring.5 Attack
of the incoming nucleophile occurs from the top face to give
the desired diastereoisomer, resulting in TMS triflate being
regenerated. In theory, one equivalent of TMS triflate is required
to silylate 9, and then a catalytic amount would then be required
to ionize 8 to the oxonium ion. However, in practice we found
that charges of less than 2 equiv tended to lead to incomplete
reactions, at least on laboratory scale. It was not clear whether
the second equivalent of TMS triflate was necessary to install
a second silyl group onto the purine heterocycle (e.g., on the
diphenylethylamino nitrogen), or whether the extra reagent was
required merely to overcome the presence of adventitious water
that would inevitably be present in the small-scale reactions
that were run during laboratory development of this step.
Following neutralization of the triflate salt 22, we wished
to develop a workup process that avoided the need to isolate
the noncrystalline glycosidation product 20. Concurrent with
this work we were developing the final API isolation process
and found that, whereas the final target 1 was poorly soluble in
dry water-miscible organic solvents, the presence of small
amounts of water dramatically increased solubility. Continued
addition of water to the solution then resulted in recrystallization
Having demonstrated that this process was viable using
isolated 20, it was then assessed using unquenched reaction
mixtures obtained from the glycosidation step. Once the
formation of 20 was complete, the reaction was quenched by
the addition of 5 M NaOH. It was found that 7.5 equiv of NaOH
were required, and measurement of the pH of the reaction
mixture demonstrated that a pH greater than 13 units was
required in order for the reaction to reach completion. We were
initially anticipating that approximately 5 equiv of NaOH would
be required, calculated from the 3 mol needed for the hydrolysis
of the three acetate groups plus 2 mol required to quench the
triflic acid generated from the 2 equiv of TMS triflate used in
580
•
Vol. 12, No. 4, 2008 / Organic Process Research & Development

the glycosidation reaction. The extra two equiv of NaOH appear
to be consumed by the two of trimethylsilyl alcohol (pKa )
11) that are also formed in the reaction.12
However, the reagent used for kilo-scale manufacture was
received in a dark amber bottle and was found to be a black
liquid just prior to being used in the manufacturing campaign.
A sample of this material was tested in a laboratory-scale
reaction and was found to work as expected, with no obvious
impact on impurity profile although, not surprisingly, the
reaction mixtures were significantly darker than was usual. In
the plant, the bond-forming chemistry proceeded as expected
using the black TMS triflate, apart from the dark reaction color,
but additional problems were encountered during the workup.
In the first batch that was run, no liquid-liquid phase separation
occurred following the hydrolysis with 5 M NaOH, although it
was noticed that some solid had precipitated on the walls of
the reactor vessel. After removing the solids, the solution was
diluted with water and was heated, but the product did not
crystallize over the time period expected. Fortunately, prolonged
agitation of the mixture did result in crystallization, and the
crude 1 was isolated by filtration. In the second batch, the
liquid-liquid phase separation this time did occur, but again
crystallization of the product required prolonged stirring. A total
of 1.7 kg of crude 1 was isolated in an average 61% yield, but
both batches were a dark brown color and a new impurity was
visible by thin-layer chromatography. Since the quality of these
two batches was so poor, we did not anticipate that taking it
directly into the API recrystallization step would provide
sufficient cleanup. We therefore decided to reslurry the com-
bined batches in a 1:1 mixture of DME and water, and this
procedure returned 1.5 kg of 1 in 94% yield with significant
purge of impurities and improvement to the color of the product.
This material was then processed without issue through the
aqueous 2-propanol filtration/recrystallization process to give
1.28 kg of clinical quality 1 with 85% recovery.
After stirring the reaction until the hydrolysis was complete,
the aqueous phase (which was a relatively small volume
compared to the organic phase) was removed, and the organic
solution was diluted with 2 vol of water (relative to the charge
of DME). After the mixture warmed to 50 °C with stirring,
UK-371,104 (1) crystallized from the mixture and was isolated
by filtration and was dried to give the product in 75-95% yield
over the two steps from the glycosidation substrate 9, with
acceptable quality. We therefore concluded that this approach
was viable for scale-up, and our attention turned to developing
a method to purify crude 1 generated by this process to give
clinical quality API.
Purification of UK-371,104 API. Although we had no
scale-up experience and no detailed purity data for the formation
of crude 1 as detailed above, we considered that it was unlikely
that the material so-formed would be of suitable quality for use
in regulatory toxicology studies and clinical trials. We were
concerned about residual levels of the DME together with low-
level impurities and the potential presence of insoluble particu-
lates. We therefore devised a method to dissolve crude 1, filter
the resulting solution, and then crystallize the product to give
clinical quality material.
The solubility of 1 was low in all ICH class 3 solvents.
However, the observation that the addition of water to water-
miscible solvents dramatically increased solubility led us to
devise a system whereby crude 1 was dissolved in a 20% v/v
mixture of water in 2-propanol. The resulting solution was
filtered to remove particulate matter, and the water was removed
from the mixture by azeotropic distillation. On cooling, the
product crystallized and, following filtration and drying, afforded
clinical-grade 1 in approximately 85% isolated yield on labora-
tory scale. Use of other solvents in this step was found to give
less reproducible crystallizations, and we therefore selected this
process for use on scale.
The conversion of the glycosidation substrate 9 through to
UK-371,104 1 was scaled-up, first in the laboratory and then
in our kilo laboratory facility. In the laboratory, the process
successfully gave two 200-g batches of crude 1 in an average
77% yield from 9, with no major processing issues. However,
two minor problems did arise during the reaction. First there
was some precipitation of inorganic materials during the
hydrolysis stage, although these were easily removed with the
aqueous phase. Second, the crystallization of 1 was problematic,
due to initial product deposition as an oil on cooling to 30 °C.
Fortunately, reheating the mixture to 50 °C resulted in crystal-
lization, and the product was isolated without any impact on
yield.
Summary
We have developed a scaleable route to the prototype inhaled
A2a agonist UK-371,104 (1) that overcame the issues associated
with the method used during medicinal chemistry research. Key
steps in the synthesis are (i) a palladium-catalyzed cyanation
reaction to prepare the nitrile 10, (ii) a telescoped conversion
of the acid 11 to the glycosidation substrate 9, and (iii) the highly
stereoselective coupling of 8 with 9 mediated by TMS triflate
in DME and conversion through to clinical quality 1. The route
was developed on laboratory scale, and then scaled into a kilo-
scale facility where 1.28 kg of API was manufactured. Further
scale-up of this process beyond a few kilograms of 1 would
require additional research in order to develop more effective
solutions to some of the problems described in this account.
An improved synthesis of the acid 11 that avoided the
processing issues associated with the cyanation/hydrolysis
protocol would be advantageous. In addition, further process
development of the glycosidation and API crystallization steps
would further enhance process robustness and efficiency for
larger-scale manufacture.
Upon scale-up into our kilo laboratory facility further
problems were encountered, which we largely attributed to the
lower quality of the TMS triflate that was used. During the
development of the process in the laboratory we had always
used TMS triflate that was purchased in sealed glass ampoules,
and this material was always a colorless or pale-yellow liquid.
Experimental Section
All reactions involving air-sensitive reagents were performed
under dry nitrogen. Melting points are uncorrected. 300 MHz
1H NMR spectra were recorded using CDCl3 as the solvent
(12) Kagiya, T.; Sumida, Y.; Tachi, T. Bull. Chem. Soc. Jpn. 1970, 43,
3716.
(13) Snodin, D. J. Regul. Toxicol. Pharmacol. 2006, 45, 79.
Vol. 12, No. 4, 2008 / Organic Process Research & Development
•
581

unless otherwise indicated. Chemical shifts are reported in ppm
(δ) relative to residual protons in the deuterated solvent. All
materials obtained from commercial suppliers were used without
further purification.
experiments, and the combined material was recrystallized from
2-propanol, including a hot filtration step, in a similar fashion
to that described above. A total of 7.10 kg (16.4 mol) of the
chloride 14 was converted to a total of 4.40 kg (63%) of 10
using this process. 1H NMR (400 MHz, CDCl3) δ: 8.00 (1H,
s), 7.40-7.20 (10H, m), 6.00-5.75 (1H, br s), 5.70 (1H, d),
4.40-4.20 (3H, m), 4.20-4.10 (1H, m), 3.80-3.70 (1H, m),
2.20-1.90 (3H, m), 1.90-1.60 (3H, m); LRMS (AP-) m/z
[M - H]- 423.
2,6-Dichloro-9-(tetrahydro-2H-pyran-2-yl)-9H-purine (13).
To a suspension of 2,6-dichloropurine (1.00 kg, 5.29 mol) in
ethyl acetate (10 L) was added p-toluenesulfonic acid mono-
hydrate (8.0 g, 0.042 mol), and the resultant mixture was heated
to 50 °C under an atmosphere of nitrogen. To the stirred mixture
was added 3,4-dihydro-2H-pyran (750 mL, 8.25 mol) over a
period of 2 h causing the reaction temperature to increase from
50 to 62 °C. During the addition, the solids dissolved to give a
clear solution. The reaction was then stirred for a further 30
min, and was then allowed to cool to ambient temperature over
18 h. The resultant solution was washed with water (2 L then
1 L), concentrated to a low volume, and was seeded to induce
crystallization. After being left to crystallize at ambient tem-
perature overnight, the solid was collected by filtration and was
dried at 50 °C under vacuum to give 13 (1.27 kg, 88%) as a
colorless solid, mp 122-124 °C. A second crop of 13 (0.14
kg, 10%) was obtained on cooling the filtrate. 1H NMR δ: 8.34
(1H, s), 5.77 (1H, dd), 4.21 (1H, br d), 3.77 (1H, br t),
2.25-1.58 (6H, m).
6-[(2,2-Diphenylethyl)amino]-9-(tetrahydro-2H-pyran-2-
yl)-9H-purine-2-carboxylic Acid (11). To a solution of 10
(1.89 kg, 4.45 mol) in industrial methylated spirits (IMS, 8.3
L) was added aqueous sodium hydroxide (1.9 L of a 6.6 M
solution, 12.5 mol), and the mixture was heated at reflux for
2 h at which point the reaction was complete. Approximately
4.3 L was distilled from the reaction mixture at atmospheric
pressure, and deionized water (4.3 L) was then added. This
operation was repeated, removing 4 L of distillate and then
adding 4.5 L of water. The mixture was cooled to 25 °C over
a period of 5 h during which time the sodium salt of the product
crystallized to give a thick slurry. The mixture was stirred at
25 °C for 9.5 h and cooled to 5 °C, and the solid product was
isolated by filtration and washed with a mixture of water (1.4
L) and IMS (0.2 L). The damp product was suspended in a
stirred mixture of dichloromethane (10.6 L) and water (10.6
L), and the pH of the aqueous phase was adjusted to a stable
pH of between 1.2 and 1.4 by the addition of concentrated
hydrochloric acid (approximately 0.55 L) whilst maintaining
good agitation. The phases were allowed to separate, and the
aqueous layer was extracted with dichloromethane (2.5 L). The
combined organic phases were washed with water (4 L) and
were concentrated to an approximate volume of 5 L by
distillation at atmospheric pressure. Ethyl acetate (11.5 L) was
added, and distillation was continued until the vapour temper-
ature reached 73 °C during which time the mixture became a
thick slurry. Additional ethyl acetate (8 L) was then added, and
the mixture was cooled to 25 °C over 14 h. The mixture was
then cooled to 6 °C, and the solid product was collected on a
filter, washed with ethyl acetate (1.5 L), and dried at 60 °C
under vacuum to give 11 as a colourless solid (1.70 kg, 86%),
mp 155 °C dec. 1H NMR (300 MHz, CDCl3) δ: 8.10 (1H, s),
7.40-7.10 (10H, m), 6.30 (1H, br s), 5.90 (1H, d), 4.50-4.20
(3H, m), 4.15 (1H, br d), 3.80 (1H, br t), 2.20-1.60 (6H, m).
LRMS (AP-) m/z [M - H]- 442.
2-Chloro-N-(2,2-diphenylethyl)-9-(tetrahydro-2H-pyran-
2-yl)-9H-purin-6-amine (14). To a suspension of 13 (0.93 kg,
3.42 mol) in 2-propanol (10.3 L) under an atmosphere of
nitrogen was added diisopropylethylamine (0.96 mL, 5.5 mol)
and 2,2-diphenylethylamine (0.705 kg, 3.57 mol), and the
resultant suspension was heated to reflux for 3 h to give a clear
solution. The solution was then cooled to ambient temperature,
and some seed crystals were added to induce crystallization.
The mixture was then stirred at -10 °C for 18 h, and the solid
was collected by filtration. The filter cake was washed with
2-propanol (2 × 1 L), and the solid was dried at 50 °C under
vacuum to give 14 (1.33 kg, 90%) as a colourless solid, mp
1
117-119 °C. H NMR δ: 7.90 (1H, br s), 7.42-7.16 (10H,
m), 5.90 (1H, br s), 5.69 (1H, d), 4.43-4.07 (4H, m), 3.78 (1H,
br t), 2.19-1.52 (6H, m).
6-[(2,2-Diphenylethyl)amino]-9-(tetrahydro-2H-pyran-2-
yl)-9H-purine-2-carbonitrile (10). The chloride 14, (1.10 kg,
2.54 mol), N,N-dimethylformamide (4 L), triethylamine (0.814
L, 5.84 mol), zinc cyanide (0.184 kg, 1.57 mol), and tetraki-
s(triphenylphosphine)palladium(0) (0.092 kg, 0.08 mol) were
charged to a stirred autoclave, and the contents were purged
with argon and then pressurized with argon to 10 bar. The
contents were heated to 80-82 °C with stirring for 21 h, and
the reaction was cooled to ambient temperature. The reaction
mixture was added to deionized water (20 L) over 15 min with
stirring to give a suspension. Stirring was continued for a further
0.5 h, after which time, the solid was collected by filtration.
The damp filter cake was suspended in 2-propanol (10 L), and
the suspension was heated to reflux with stirring. The mixture
was then filtered through a filter aid pad whilst still hot, and
the filtrate was allowed to cool, with stirring, to ambient
temperature over 16 h, during which time crystallization
occurred. The solid was collected by filtration, washed with
2-propanol (2 L), and was dried in Vacuo at 50 °C. The solid
so obtained was combined with material obtained from similar
6-[(2,2-Diphenylethyl)amino]- N-[2-(1-piperidinyl)ethyl]-
9H-purine-2-carboxamide (9). To a mixture of 11 (1.70 kg,
3.83 moles) in dry THF (17 L) was added N,N′-carbonyldi-
imidazole (0.75 kg, 4.60 moles), and the mixture was stirred
for approximately 3 h at 22 °C. The mixture was then cooled
to 10 °C, and a solution of 1-(2-aminoethyl)piperidine (0.59
kg, 4.60 mol) in dry THF (0.7 L) was added to the reaction
mixture over a period of 0.5 h whilst maintaining the temper-
ature below 20 °C. The mixture was stirred for approximately
1 h at 20 °C, treated with deionized water (0.1 L), and then
concentrated by distillation under a slight vacuum to remove
approximately 12 L of solvent. To the residue was then added
2-propanol (22 L), and distillation was continued until the
reaction temperature reached 83 °C. The volume of the mixture
582
•
Vol. 12, No. 4, 2008 / Organic Process Research & Development

was then adjusted to approximately 21 L by the addition of
2-propanol (4 L), and the mixture was cooled to 20 °C.
Deionized water (10.6 L) and trifluoroacetic acid (2.6 kg, 18
mol) were added, and the resultant mixture was heated at reflux
for approximately 2 h. The reaction mixture was cooled to 20
°C, and then the pH of the mixture was adjusted to between
pH 9.8 and 10.0 by the addition of aqueous sodium hydroxide
(1.5 L of 40% w/w solution) with the temperature below 30
°C and good agitation. The resultant slurry was heated to reflux,
cooled to 25 °C over 8 h, and then cooled further to 8 °C. The
solid was collected by filtration, and the filter cake was washed
twice with 2-propanol (2 L each wash). The damp product was
then suspended in 1-propanol (54 L), and the mixture was
heated to reflux to give a solution. Approximately 27 L of the
solvent was then removed from the mixture by distillation at
atmospheric pressure. The mixture was then cooled to 25 °C
over approximately 4 h, and then cooled further to 6 °C. The
solid was collected by filtration, and the filter cake was washed
with 1-propanol (6 L). The solid was dried at 60 °C under
vacuum to give 9 as a colourless solid (1.14 kg, 63%). 1H NMR
(300 MHz, CDCl3) δ: 15.25 (1H, br s), 8.55 (1H, br s), 8.30
(1H, s), 7.40-7.15 (10H, m), 5.90 (1H, br s), 4.50-4.25 (3H,
m), 3.60 (2H, q), 2.55 (2H, t), 2.50-2.30 (4H, m), 1.50-1.20
(6H, m). LRMS (AP+) m/z [MH+] 470.
6-[(2,2-Diphenylethyl)amino]- N-[2-(1-piperidinyl)ethyl]-
9-(ꢀ-D-ribofuranosyl)-9H-purine-2-carboxamide (1). To a
stirred suspension of 9 (200 g, 0.426 mol) in anhydrous 1,2-
dimethoxyethane (DME, 800 mL) under an atmosphere of
nitrogen was added a solution of trimethylsilyl trifluoromethane-
sulfonate (200 g, 0.900 mol) in anhydrous DME (200 mL) over
a period of 15 min. During the addition, the solid dissolved to
give a deep red/amber solution, and the reaction temperature
rose from 20 to 31.5 °C. The resultant mixture was heated to
55-60 °C, and a solution of 1,2,3,5-tetra-O-acetyl-ꢀ-D-ribo-
furanose (8, 163 g, 0.512 mol) in anhydrous DME (400 mL)
was added over a period of 40 min, followed by a rinse with
anhydrous DME (200 mL). The reaction mixture was heated
at 60 °C for 3 h, and was then allowed to cool to ambient
temperature. This crude reaction solution, containing the triac-
etate 20, was held at ambient temperature for 18 h, and then
treated with an aqueous solution of sodium hydroxide (640 mL
of a 5 M solution, 3.2 mol) over a 45 min period with cooling.
The resultant mixture was stirred at ambient temperature for
3 h, and the layers were separated. The stirred organic phase
was diluted with deionized water (1800 mL) with cooling, and
the resultant mixture was then heated to 50-55 °C whereupon
crystallization started. To this heated and stirred suspension was
added more deionized water (1800 mL) over a period of 50
min. Once the addition was complete, the resultant slurry was
cooled to 10 °C over a period of 45 min, and the solid was
collected by filtration. The filter cake was then washed with a
solution made up from DME (400 mL) and deionized water
(800 mL), and was then dried at 55 °C in Vacuo to give the
crude 1 as a brown solid (203 g). This material was combined
with material obtained from similar processes and was purified
in the following manner. To a suspension of crude 1 (398 g,
0.661 mol) in 2-propanol (7.05 L) was added deionized water
(1.76 L), and the resultant mixture was stirred and warmed until
a clear solution was obtained. The resultant solution was filtered,
and the filtrate was distilled under nitrogen at atmospheric
pressure with periodic addition of filtered 2-propanol to maintain
the reaction volume. Over the course of the distillation, a total
of 29.1 L of distillate was collected, and a total of 26.1 L of
filtered 2-propanol was added. Towards the end of the distil-
lation, the amount of water present in the distillate was measured
by Karl Fischer analysis to be <0.5% w/w. The mixture was
allowed to cool to 40 °C over 3.5 h with stirring, during which
time crystallization occurred. The resultant slurry was stirred
at ambient temperature for 12.5 h, and cooled to 2 °C in an ice
bath over 5.5 h. The solid was collected by filtration, washed
with chilled, filtered 2-propanol (2 × 1.5 L), and dried at 60
°C in Vacuo to give 1 as a pale beige-coloured solid (306 g),
mp 182 °C, that was found to contain 0.63% w/w (20 mol %)
1
of water by Karl Fischer analysis. H NMR (500 MHz, d6-
DMSO) δ: 8.50 (1H, br t), 8.40 (1H, s), 8.00 (1H, br t), 7.35
(4H, d), 7.26 (4H, t), 7.15 (2H, t), 5.91 (1H, d), 5.39 (1H, d),
5.14 (1H, d), 5.06 (1H, t), 4.64-4.50 (2H, m), 4.28-4.18 (2H,
m), 4.18-4.10 (1H, m), 3.96-3.90 (1H, m), 3.70-3.61 (1H,
m), 3.60-3.50 (1H, m), 3.46-3.37 (2H, m), 2.50-2.44 (2H,
m, partly obscured by DMSO peak), 2.40-2.32 (4H, m),
1.46-1.38 (4H, m), 1.36-1.28 (2H, m). LRMS (AP+) m/z
[MH+] 602. Anal. Calcd for C32H39N7O5 ·0.2 H2O: C, 63.50;
H, 6.56; N, 16.20. Found: C, 63.55; H, 6,56; N, 16.27.
Acknowledgment
We thank Sam Farenden and Christophe Lefeuvre for
analytical support, John Williams and Dave Clifford for
supervision of the kilo laboratory scale-up, Geoffrey Fuller for
his practical contributions to scale-up of the preparation of the
nitrile 10. We also thank Dr. Mike Williams and Dr. Alan
Happe for useful comments made during the preparation of this
manuscript.
Received for review November 1, 2007.
OP700248T
Vol. 12, No. 4, 2008 / Organic Process Research & Development
•
583