Nucleosidic phosphoramidite synthesis via phosphitylation: Activator selection and process development

Article abstract of DOI:10.1021/op050077d

Nucleosidic phosphoramidites are key building blocks for the automated, solid supported syntheses of oligonucleotide-based drugs. A safe, industrially viable process for preparing nucleosidic phosphoramidites 5-Me-MOE-U and MOE-A has been developed and ut

Full text of DOI:10.1021/op050077d

Organic Process Research & Development 2005, 9, 730−737
Full Papers
Nucleosidic Phosphoramidite Synthesis via Phosphitylation: Activator
Selection and Process Development
Chaoyu Xie,* Michael A. Staszak,* John T. Quatroche, Christa D. Sturgill, Vien V. Khau, and Michael J. Martinelli†
Global Chemical Product Research and DeVelopment, Lilly Research Laboratories, Eli Lilly and Company,
Indianapolis, Indiana 46285-4813, U.S.A.
Scheme 1. Synthesis of nucleosidic phosphoramidites
Abstract:
Nucleosidic phosphoramidites are key building blocks for the
automated, solid supported syntheses of oligonucleotide-based
drugs. A safe, industrially viable process for preparing nucleo-
sidic phosphoramidites 5-Me-MOE-U and MOE-A has been
developed and utilized on multikilogram scales. The optimiza-
tion of this process is described in detail. Emphasis is placed
on the search for activators to replace the hazardous 1H-
tetrazole, the development of an extractive workup, and the
advancement of a precipitation method to avoid both chro-
matographic purification and product foaming issues.
Scheme 2. 2′-O-MOE nucleosidic phosphoramidites
Introduction
Numerous antisense oligonucleotide-based drugs are cur-
rently being evaluated in preclinical and clinical studies for
the treatment of a variety of diseases including cancer,
cardiovascular disease, diabetes, and inflammatory disorders.1
A successful clinical evaluation and marketing strategy for
such drugs requires that oligonucleotides become available
in large quantities. Presently, clinical oligonucleotides are
produced using solid-phase synthesis and employing nucleo-
sidic phosphoramidites as the basic building blocks. How-
ever, synthesizing these nucleosidic phosphoramidites re-
mains challenging, especially on large scale. With the
growing success of clinical programs and the potential market
demand, the development of efficient, robust syntheses of
nucleosidic phosphoramidites is essential. Herein, we report
a safe, scalable process for preparing nucleosidic phosphora-
midites from nucleosides via phosphitylation (Scheme 1).
The chemically modified and structurally more complex
oligonucleotides such as the 2′-O-methoxyethyl (MOE)
derivatives have recently emerged as the second generation
antisense drugs.2 These compounds exhibit improved hy-
bridization properties, more favorable nuclease resistance,
and improved pharmacological profiles.3 Given the medicinal
importance of the 2′-O-MOE oligonucleotides, considerable
efforts have been directed towards synthesizing the corre-
sponding nucleosidic phosphoramidites. The processes we
describe focus on two of the targets shown in Scheme 2.
These targets are 5-Me-MOE-U amidite (5), a representative
pyrimidine nucleosidic phosphoramidite class, and MOE-A
amidite (6), a representative of the purine nucleosidic
phosphoramidites.
Issues with the Initial Process.4 Initially, the desired
phosphoramidites were prepared by treating the appropriate
nucleosides with 2-cyanoethyl N,N,N′,N′-tetraisopropyl phos-
phorodiamidite (1),5 1H-tetrazole, and N-methylimidazole in
DMF at ambient temperature. When the nucleosides were
consumed, the reaction mixtures were diluted with DMF/
H2O, washed with hexanes, and extracted with hexanes/
(2) (a) Sewell, K. L.; Geary, R. S.; Baker, B. F.; Glover, J. M.; Mant, T. G. K.;
Yu, R. Z.; Tami, J. A.; Dorr, F. A. J. Pharmacol. Exp. Ther. 2002, 303,
1334. (b) Garay, M.; Gaarde, W.; Monia, B. P.; Nero, P.; Cioffi, C. L.
Biochem. Pharmacol. 2000, 59, 1033-1043. (c) Bost, F.; McKay, R.; Dean,
N. M.; Potapova, O.; Mercola, D. Methods Enzymol. 2000, 314 (Antisense
Technology, Part B), 342-362.
(3) Altman, K.-H.; Dean, N. M.; Fabbro, D.; Frier, S. M.; Geiger, T.; Haner,
R.; Husken, D.; Martin, P.; Monia, B. P.; Muler, M.; Natt, F.; Nicklin, P.;
Phillips, J.; Pieles, U.; Sasmor, H.; Moser, H. H. Chimia 1995, 50, 168-
176.
* To whom correspondence should be addressed. E-mail: xiech@lilly.com.
† Present address: Amgen Inc., One Amgen Center Dr., Thousand Oaks, CA
91320.
(1) (a) Holmlund, J. T. Ann. N. Y. Acad. Sci. 2003, 1002, 244. (b) Crooke, S.
T. Oncogene 2000, 19, 6651. (c) Ravichandran, L. V.; Dean, N. M.;
Marcysson, E. G. Oligonucleotides, 2004, 14, 49. (d) Yu, R. Z.; Su, J. Q.;
Grundy, J. S.; Sewell, K. L.; Dorr, A.; Levin, A. A. Antisense Nucleic Acid
Drug DeV. 2003, 13, 57.
(4) Procedures of the initial process, see Experimental Section.
(5) Preparation of reagent 1 and its application in antisense oligonucleotide
synthesis: (a) Barone, A. D.; Tang, J. Y.; Caruthers, M. H. Nucleic Acids
Res. 1984, 12, 4051. (b) Kraszewski, A.; Norris, K. E. Nucleic Acids Res.
1987, 18, 177. (c) Song, Q.; Ross, B. WO 03/087130 A2.
730
•
Vol. 9, No. 6, 2005 / Organic Process Research & Development
10.1021/op050077d CCC: $30.25 © 2005 American Chemical Society
Published on Web 09/08/2005

Scheme 3.13 Activation mechanism
toluene. Those extracts were washed with water and subse-
quently concentrated to a foam to provide the corresponding
nucleosidic phosphoramidites in 75-95% yields.
Activator Selection. Literature surveys revealed several
potential replacements for 1H-tetrazole as the activator for
the phosphitylation chemistry. These included 4,5-dicy-
anoimidazole (DCI),11 and various imidazolium, anilinium,
and pyridinium salts.12 Our lab trials on the phosphitylation
of 5′-O-(4,4′-dimethoxytriphenylmethyl)-N4-benzoyl-2′-deoxy-
5-methylcytidine (5-Me-dC) using DCI (0.8 equiv in DMF
at ambient temperature for 5 h) afforded an 84% yield of
the desired phosphoramidite, which was ∼5% lower than
trials using 1H-tetrazole. A major drawback to using DCI
was the need for a silica gel plug to remove the DCI residue.
In the 1H-tetrazole case, an aqueous workup removes all
activator residues, which is advantageous. Hence, we decided
to focus on ammonium, pyridium, and azolium salts in our
activator screening studies. These derivatives were widely
available, had mild acidity, and were water soluble.
Ammonium Salts. The mechanisms described in the
literature12,13 indicate that the activator must be acidic to
allow protonation of the diisopropylamine group of phos-
phitylating reagent 1 (Scheme 3). The rate-determining step
in these mechanisms is the displacement of the diisopropyl
amino group with the activator to form 2b. A substitution
reaction on this activated intermediate (2b) by the nucleoside
(Nuc-OH) generates the nucleosidic phosphoramidite. An
ideal activator thus not only requires mild acidity and good
nucleophilicityy but should also serve as a good leaving
group to complete the substitution reaction.
To better understand the requirement of the activator as
a leaving group, we chose diisopropylammonium chloride
as the activator, since the phosphitylation reaction itself
produces one equivalent of diisopropylamine. Consequently,
5′-O-(4,4′-dimethoxytriphenylmethyl)-N4-benzoyl-2′-deoxy-
5-methylcytidine (5-Me-dC nucleoside) was treated with 1
equiv of diisopropylamine hydrochloride and 1.5 equiv of
phosphitylating reagent 1 in DMF at room temperature.
Interestingly, the reaction went smoothly but required a
longer reaction time than that of 1H-tetrazole. After 24 h,
approximately 30% of the starting nucleoside remained in
the reaction mixture. Longer reaction times increased con-
Despite reasonable yields, the above procedure had a
number of processing issues which made it impractical for
pilot-plant operations. The major issue centered on the use
of 1H-tetrazole as the activator for the relatively nonreactive
phosphitylating agent 1. 1H-tetrazole is classified as an
explosive and must be stored in sealed magazines prior to
use.6 By Indiana State law, a Federal ATF7 permit is required
to purchase and hold the material. Thus, our first develop-
ment goal was to eliminate this reagent from the synthesis.
A second problem with the original process was its
lengthy and cumbersome workup protocol. A total of nine
extractions were used to remove excess reagents and byprod-
ucts. Consequently, the workup volumes were very large,
and the operation times were extensive, leading to potential
decomposition of the phosphoramidites. Typically, the crude
material from this procedure required further purification and
manipulations to reach a suitable purity and form.
The third major issue with the procedure was the isolation
of the desired product as an intractable foam. To our
knowledge, none of the MOE-substituted nucleosidic phos-
phoramidites have been found to be crystalline entities, and
many do not even form manageable amorphous solids. In
our case, the final workup solution was concentrated to
dryness, and the resulting foam was collapsed to fine particles
via scraping with a spatula. Such operations are not viable
on pilot-plant scale, and thus development of a reliable
precipitation method was another major goal.
The process development of the nucleosidic phosphora-
midite synthesis is further complicated by a number of other
issues. Since the compounds are acid sensitive, they are
subject to de-tritylation and de-purination reactions.8 Since
they are also base sensitive, hydrolysis reactions are also
factors. Finally, the materials exhibit poor ARC9 profiles with
high risk of thermal liability. A hazards review by Lilly
Chemical Hazards Laboratory indicated that the 5-Me-
MOE-U (5) and MOE-A (6) phosphoramidites should not
be exposed to temperatures above 45 and 35 °C, respec-
tively.10
The above stated characteristics greatly limit the available
workup and solvent-exchange conditions for these com-
pounds and thus necessitate the development of highly
efficient reaction conditions which provide very clean crude
product as a starting point for isolation. With all of these
issues in mind, we initiated the process development of the
MOE nucleosides, while striving to rapidly determine
practical reaction/isolation conditions which could facilitate
the multikilogram production of phosphoramidites in our
pilot plant.
(6) (a) Stull, D. R. Fundamentals of Fire and Explosion; AIChE Monograph
Series; 1977; Vol. 73(10), p 22. (b) See Material Safety Data Sheet in Sigma-
Aldrich database: www.sigmaaldrich.com.
(7) ATF stands for Bureau of Alcohol, Tobacco and Fire Arms.
(8) Krotz, A. H.; Klopchin, P. G.; Walker, K. L.; Srivatsa, G. S.; Cole, D. L.;
Ravikumar, V. T. Tetrahedron Lett. 1997, 38, 3875.
(9) ARC abbreviates as Accelerated Rate Calorimetry. The ARC data were
accumulated by Mr. D. Cheatham of the Lilly Hazards Laboratory.
(10) Applicable to near atmospheric pressure and quantities of less than 500 kg.
(11) Vargeese, C.; Carter, J.; Yegge, J.; Krivjahsky, S.; Settle, A.; Kropp, E.;
Peterson, K.; Pieken, W. Nucleic Acids Res. 1986, 26, 1046.
(12) (a) Hayakawa, Y.; Kawai, R.; Hirata, A.; Sugimoto, J.; Kataoka, M.;
Sakakura, A.; Hirose, M.; Noyori, R. J. Am. Chem. Soc. 2001, 123 (34),
8165-8176. (b) Sanghvi, Y.; Guo, Z.; Pfundheller, H.; Converso, A. Org.
Process Res. DeV. 2000, 4, 175.
(13) Dahl, B. H.; Nielsen, J.; Dahl, O. Nucleic Acids Res. 1987, 15, 1729.
Vol. 9, No. 6, 2005 / Organic Process Research & Development
•
731

Figure 1. Substituted imidazolium salts.
Scheme 4. Oxidation of 5-Me-MOE-U under acidic
condition
Therefore, any potential advantages of Py‚TFA as activator
were offset.
A second problem associated with Py‚TFA was the
formation of a dimer 414 (Scheme 5). Phosphitylation of the
5-Me-MOE-U nucleoside with 1.2 equiv of reagent 1 and
1.0 equiv of Py‚TFA in 5 vols of CH2Cl2 at room temperature
yielded ∼30% of the dimer 4. The mechanism of the dimer
formation is similar to that of the 5-Me-MOE-U amidite
formation. Py‚TFA protonates the diisopropylamine group
of the newly formed phosphoramidite product, and the dimer
forms when another molecule of nucleoside displaces the
activated amidite. Ideally, the activator should be active
enough to trigger the protonation of the phosphitylating
reagent without allowing subsequent protonation of the
desired product. This requirement led us to investigate the
azolium salts.
version, but led to higher levels of undetermined related
substances, likely due to the instability of the phosphora-
midite product over time. Similar reactions with 1H-tetrazole
take 3-4 h to reach completion. This experiment showed
that a better leaving group than diisopropylamine was
required to increase the reaction rate.
Azolium Salts. Figure 1 shows a number of different
substituted imidazolium salts which could serve as potential
activators for nucleosidic phosphitylation. N-Methyl benz-
imidazolium salts and N-phenylimidazolium (NPI) salts have
been reported to be efficient promoters for the synthesis of
oligodeoxyribonucleotides and oligoribonucleotides via the
phosphoramidite approach.12a Our investigations on the use
of these salts were not fruitful. Both reaction efficiency and
chemical yield with N-methyl benzoimadazole triflate were
acceptable, but the compound did not meet our requirement
of being easily extractable by aqueous media. Laboratory
trials using N-phenyl imidazolium triflate (NPI‚Tf) and
N-phenyl imidazolium tetrafluoroborate (NPI‚HBF4) afforded
much slower reaction rates15 and ∼10% lower yields than
desired.
Pyridinium Salts. Based upon the availability, cost
considerations, and literature precedent,12b pyridinium tri-
fluoroacetate (Py‚TFA) was the next activator tested in our
5-Me-MOE-U reactions. This salt worked well in a number
of acceptable solvents (e.g., acetonitrile), with reasonable
reaction rates (3-12 h). One problem observed when using
Py‚TFA as activator was residual pyridine in the crude
product. Although the TFA salt was water soluble and thus
easily removed during the aqueous workup, free pyridine
was not readily washed from the system. We were unable
to perform even mild acid washes due to product decomposi-
tion. For example, an aqueous wash of the crude 5-Me-
MOE-U with a 1% solution of citric acid (2 × 5 vols) led to
the formation (34%) of the phosphorus(V)-related impurity
3 (Scheme 4). Activated carbon (Darco, 20-40 mesh)
treatment of the water-washed product decreased the pyridine
amount to a minimum, but with a product yield loss of 20%.
We also tested N-methyl imidazolium trifluoroacetate, a
relatively cheap salt, which met our water solubility require-
Scheme 5. Dimer formation with Py‚TFA
732
•
Vol. 9, No. 6, 2005 / Organic Process Research & Development

ment. This activator afforded good yields with acceptable
reaction rates in the nucleosidic phosphitylation reaction.
However, its hygroscopicity gave us reservations about its
large-scale use.
by aqueous extraction of acetonitrile (and other impurities)
during workup. Further development of this extractive solvent
exchange led to the very efficient system of MTBE with
1:1 DMF/H2O washes for the workup protocol.
According to the CRC Handbook,16 N-methyl imidazole
has infinite solubility in water. N-Methyl imidazolium triflate
is not only highly water soluble but also has excellent
solubility in polar organic solvents such as DMF and
acetonitrile. Consequently, this salt was our next choice to
test as an activator for our phosphitylation reactions. A small
amount of N-methyl imidazolium triflate (NMI‚Tf) was
prepared using a literature example.12b The compound was
tested in the 5-Me-MOE-U amidite formation and was found
to perform better than 1H-tetrazole. The reaction activated
with NMI‚Tf consistently provided >90% chemical yield
To this point, all development for the MOE nucleosides
had been performed using small, laboratory-scale experi-
ments (<1 g scale). On such scale, the reagents were
typically combined with no concern for exothermic events.
When a 5-g trial was performed in this manner, a 15-20
°C exotherm ensued, with concomitant decrease in overall
product purity. This exotherm would likely be more signifi-
cant on larger scale. With the ARC data indicating the MOE
amidites to be unstable at only moderately elevated temper-
atures, exothermic events were unacceptable.
Subsequently, an experiment was performed whereby
reagent 1 was dissolved in a small amount of acetonitrile,
and that solution was added slowly to the 5-Me-MOE-U
substrate. This method controlled the exotherm and afforded
good-quality product. However, 31P NMR analysis of the
products indicated a significant impurity at 141 ppm.17 LC/
MS data indicated the new peak to have a mass of 1337 (vs
the desired mass of 818). These data, along with the 31P NMR
results, supported the conclusion that dimer 4 was again
being produced. The adjusted reaction conditions (i.e. slowly
adding the reagent 1 to the MOE nucleoside), which had
been invoked to control exothermic events, had led to the
new problem of dimer formation. We also observed that the
amount of dimer 4 formed (20-40 mol %) depended on the
addition rate. This result suggested that the solution to the
problem might simply be inverse addition of the reagents.
Thus, an experiment was performed whereby the 5-Me-
MOE-U starting material was dissolved in acetonitrile and
added dropwise to a solution of reagent 1 and activator NMI‚
Tf. This protocol, which maintains an excess of 1 throughout
the addition, both controlled the exotherm and eliminated
the dimer 4 formation.
The above modifications to the reaction procedures
resulted in large strides towards the goal of consistently
obtaining good-quality 5-Me-MOE-U amidite. However,
even that product quality was still not considered high enough
to afford reproducible results in a precipitation method.
Further investigation into this reaction was clearly needed.
A known culprit in phosphitylation reactions is adventi-
tious water. Several advantages of choosing acetonitrile as
reaction solvent were mentioned earlier. Another major
advantage stems from the solvent’s known azeotrope with
water.18 When the starting 5-Me-MOE-U nucleoside was
dissolved in an excess (6 vols) of acetonitrile and then
vacuum distilled to the desired reaction concentrations (4
vols) prior to any other reagent additions, a significant jump
in product purity was realized. Consequently, the acetonitrile
1
with product purity >98%. Careful H NMR and HPLC
analyses of our crude products were performed to ensure
that all residual salt and free amine had been removed by
the aqueous washes.
As confidence increased that this salt was the activator
of choice for our chemistry, the literature preparation of
N-methyl imidazolium triflate was quickly modified to
replace diethyl ether with MTBE and to increase concentra-
tion. Using the adjusted conditions, we were able to produce
over 27 kg of the pure salt to meet our longer-term
production needs.
To date, NMI‚Tf stands as the best activator among all
those tested. This compound met all of our processing
requirements and allowed us to proceed with the development
of the nucleosidic phosphoramidites described in the fol-
lowing sections. Unless specifically indicated otherwise,
NMI‚Tf was employed as the activator in those studies.
Synthesis of 5-Me-MOE-U (5). With the activator
identified, our attention turned to optimizing the reaction
conditions and developing a dependable isolation/purification
method. Because of the ready availability of the starting
nucleoside, 5-Me-MOE-U was chosen to start our studies.
However, the solubility characteristics of 5 indicated that a
precipitative isolation would be difficult.
Our studies on activation with Py‚TFA had demonstrated
that solvent played an important role in the phosphitylation
of 5-Me-MOE-U. For comparative purposes, CH2Cl2 was
tested in the phosphitylation of 5-Me-MOE-U while using
NMI‚Tf as the activator. These trials were successful,
producing high quality 5 in 90% yield. However, this solvent
did not lend itself to precipitation methods for the 5-Me-
MOE-U amidite, requiring vacuum distillation for product
isolation as a foam.
Acetonitrile was tested next and was found to exhibit a
number of advantages over other solvents. First, its high
polarity afforded good solubility for all the substrates and
reagents. The phosphitylation reactions on 5-Me-MOE-U
were both fast and very clean relative to those using other
solvents. Additionally, a solvent exchange could be affected
easily by MTBE extraction of the crude product followed
(16) CRC Handbook of Chemistry and Physics, 58th ed.; CRC PRESS: West
Palm Beach, Florida, 1977; p 347.
(17) The 31P NMR (CDCl3) data show a single peak at 141 ppm, not two peaks.
This fact indicated that the diastereomers that are normally produced around
the phosphorous atom were not present and meant that the impurity might
have two identical groups on the P atom.
(14) Dimer 4 was identified by 31P NMR. The chemical shifts of 5-Me-MOE-U
amidite and 4 are very different. 5-Me-MOE-U amidite absorbs at δ ∼150
ppm, while 4 absorbs at δ ∼140 ppm (CDCl3).
(18) Azeotrope Data-III; Advances in Chemistry Series 116; Gould, R. F., Ed.;
American Chemical Society: Washington, D.C., 1973. The binary azeotrope
between water and acetonitrile forms at 81.6 °C and is 16.5% water in the
distillate at atmospheric pressure.
(15) Reactions using NPI‚Tf and NPI‚HBF4 required 16 h to go completion.
Vol. 9, No. 6, 2005 / Organic Process Research & Development
•
733

Scheme 6. Formation of O-methoxy-Derived byproduct
azeotrope was incorporated into the procedure to control the
solution water content to <0.1%.19
Stoichiometric Ratio. After finalizing reaction parameters
such as solvent, concentration, and temperature, we turned
our attention to determining the optimum ratio of nucleoside/
reagent 1/activator (NMI‚Tf). Separation of the excess P(III)
reagent 1 from amidite could be achieved using heptane/
DMF/H2O extractions, with 1 distributed in the heptane layer.
However, since earlier ARC9 studies had demonstrated that
the amidite was sensitive at temperatures g45 °C, a solvent
exchange from acetonitrile with heptane/DMF was not an
option and was consequently abandoned.
Control experiments had demonstrated that 1 was unstable
at room temperature under the phosphitylation conditions.
Furthermore, 1 is both moisture sensitive and tends to easily
oxidize to P(V) compounds. We originally employed an
excess amount (1.5 equiv) of reagent 1 to both promote
reaction completion and prevent the formation of dimer 4.
With the difficulty of removing excess 1, we decided to
decrease the reagent level to 1.05 equiv.20
Initially, stoichiometric amounts of NMI‚Tf were tested
in conjunction with 1.1 equiv of reagent 1. Subsequent
experiments showed that 0.5 equiv of NMI‚Tf was sufficient
to attain reaction completion. However, the reaction time
increased from 4 to 12 h when using the lowered activator
amounts. Additionally, trials using 0.5 equiv of NMI‚Tf
typically had residual levels of 1 after workup. Adjusting
the activator load to 0.75 equiv consumed essentially all
reagent 1 with full conversion of the starting material to
product. Therefore, the optimum ratio of nucleoside/1/
activator was determined to be 1.00:1.05:0.75.
Extractive Purification. The impurities produced in the
phosphitylation reaction include imidazole, imidazole triflate,
diisopropylamine salts, low levels of 1, and the P(V)
impurities derived from both 1 and the product. The first
three of these are easily removed by the aqueous extraction.
However, separating the P(V) impurities from product by a
traditional extractive workup was troublesome. We did
determine that MTBE/DMF/H2O extractions could remove
the bulk of these P(V) impurities. The optimum ratio of
DMF/H2O is 1:1, with higher ratios causing unacceptable
yield losses. Employing two such washes during the workup
typically removed 80% of the P(V) impurities while losing
< 5% of the product.
A precipitation trial using MeOH/H2O provided filterable
solids with reduced P(V) impurity levels. However, the
residual MeOH and H2O were difficult to remove by the
usual drying methods. Since these two solvents adversely
affect the efficiency of the downstream chemistry (oligo-
nucleotide synthesis), they needed to be eliminated. Ad-
ditionally, a new byproduct formed when using this system.
This byproduct, shown in Scheme 6, was identified by LC/
MS as the methoxy derivative and was formed by MeOH
displacement of diisopropylamine. Due to these issues, work
on a MeOH-H2O precipitation process was abandoned.
Our efforts next turned to finding a nonaqueous precipita-
tion method. Adding the antisolvent hexane to MTBE
solutions of 5 only led to glue-like mixtures of the product.
Reverse addition of an MTBE solution of 5 to cold hexanes
(-9 °C) was promising, providing filterable solids. Unfor-
tunately, these solids oiled-out on the filter.
Another issue with the MTBE/hexane reverse addition
method, was the large solvent volumes required to achieve
precipitation. With the typical concentrations of 5 in MTBE
after reaction workup, upwards of 100 vols of hexane were
needed to induce precipitation. Further studies indicated that
there was an optimum ratio of hexane/MTBE which would
induce precipitation while preventing the filtered product
from transforming into an oil. Modifying the process to
include a vacuum distillation with simultaneous addition of
hexane to maintain constant tank volume afforded the desired
result. Using that process while maintaining a 20/1 hexane/
MTBE ratio resulted in well-behaved, filterable solids.
A final processing challenge for 5 arose when oven drying
the filtered solids. The material expanded to ∼3 times its
initial volume while self-converting into a foam. Fortunately,
a simple switch from hexane to heptane as the antisolvent
solved this problem without adversely affecting the precipita-
tion behavior of the product.
Synthesis of MOE-A Phosphoramidite (6). With the
significant progress achieved on the 5-Me-MOE-U amidite
process, our attention turned to the MOE-A amidite 6. Much
of the knowledge acquired during the process development
of the 5-Me-MOE-U derivative was successfully applied to
6. However, an initial laboratory trial using the identical
protocol developed for 5-Me-MOE-U (acetonitrile, 1.1 equiv
of phosphitylating reagent 1, 0.75 equiv of activator salt)
did not go to completion after stirring for 16 h. Obviously,
further process development was needed to optimize the
reaction conditions for the MOE A amidite.
Precipitation. As stated previously, foaming is a major
problem when isolating amidites on scale. Development of
a suitable crystallization or precipitation method was required
for our production levels. Solubility studies on the nucleosidic
amidites had indicated that developing precipitation or
crystallization methods would be very challenging. Antisol-
vents were limited to heptane, hexane, and water. Attempts
to actually crystallize 5-Me-MOE-U under numerous condi-
tions were unsuccessful.
(19) Water content was analyzed by the Karl-Fisher method.
(20) The purity of 2-cyanoethyl N,N,N′,N′-tetraisopropylphosphorodiamidite (1)
is required to be >98%. 2-Cyanoethyl N,N,N′,N′-tetraisopropylphospho-
rodiamidite used at the pilot plant was purchased from Digital Specialty
Chemicals, Inc. Canada.
A major difference between the “A” and “U” starting
nucleosides was solubility. The N6-benzoyl adenosine portion
of the molecule afforded a decreased solubility profile for
734
•
Vol. 9, No. 6, 2005 / Organic Process Research & Development

Table 1. Effect of DMF/H2O ratio for MTBE/DMF/H2O
Fortunately, those same solubility characteristics which
had caused problems with the MOE-A starting material were
found to be advantageous for isolation of the MOE-A
product. We rapidly determined precipitation conditions
using an EtOAc/hexanes system with a reverse addition
protocol. Thus, adding an EtOAc solution of the product to
stirring hexanes resulted in the immediate formation of solids.
This method also reduced the levels of P(V) impurities to
an acceptable level. The only drawback of the EtOAc/
hexanes system was the large solvent volumes. The minimum
hexanes/EtOAc ratio needed for successful precipitation of
MOE-A amidite was 50:1, making scale-up difficult.
Further investigations revealed MTBE be a good replace-
ment for EtOAc. Using a 5:1 ratio of hexanes (or heptane)
to EtOAc, and employing a reverse addition protocol,
resulted in easily filtered MOE-A amidite. This solvent
system efficiently removed the P(III) and P(V) impurity
species, providing crude products of >98% purity.
extractiona
DMF:H2O
0:0
1:5
1:2
1:1
2:1
remaining P(V)22
recovery
30.0%
100%
9.3%
91%
7.6%
4.6%
4.4%
89.4% 88.2% 85.5%
a Workup conditions: 15 vols MTBE/10 vols DMF-H2O.
MOE-A versus that for MOE-U. Consequently, the com-
pound did not fully dissolve when using the same concentra-
tion of acetonitrile employed for the 5-Me-MOE-U nucleo-
side.
DMF is certainly one of the most common solvents
employed in the nucleoside phosphitylation literature, and
the MOE-A nucleoside was quite soluble in this solvent.
Reactions using DMF performed well. Our processing goal
was to maintain the azeotropic drying capabilities of aceto-
nitrile while keeping the amount of DMF in the system to a
minimum. Towards that end, a set of experiments was
designed to determine the best combination of DMF and
acetonitrile for this reaction. The ending conditions utilized
2 vols of DMF to dissolve the starting material and 8 vols
of acetonitrile for the azeotropic drying. When this initial
acetonitrile was removed, the remaining DMF solution of
the starting material was added slowly to an acetonitrile
solution (2 vols) of 1 and the activator salt. This protocol
resulted in very clean MOE-A amidite reaction profiles after
16 h.
A reaction which employed the DMF/acetonitrile condi-
tions using 1.00 equiv of activator and 1.25 equiv of
phosphitylating reagent, exhibited an impurity at ∼140 ppm21
in the 31P NMR spectrum. A subsequent experiment where
unadulterated reaction samples were checked by 31P NMR
showed the impurity was forming during the reaction, not
during workup. Returning to a full equivalent of the activator
suppressed these impurities.
Conclusions
In summary, we have achieved our goal of developing a
safe, reliable process for the synthesis of nucleosidic phos-
phoramidites 5-Me-MOE-U and MOE-A. We identified a
cheap, water-soluble activator to replace the hazardous 1H-
tetrazole typically used in these reactions. Significantly,
dependable precipitation and isolation conditions were
developed for the traditionally problematic phosphoramidite
products. The finely tuned process affords products of such
purity as to avoid the typical chromatographic purifications
common in the literature. Finally, these two key nucleosidic
phophoramidites have been successfully tested on pilot-plant
scale. The results represent a step forward in the synthesis
of these key building blocks for the oligonucleotide-based
drug candidates.
Experimental Section
General. All reagents and solvents were purchased and
1
used without further purification. H and 13C NMR spectra
Extractive Workup. The MTBE/DMF/H2O workup
procedure, developed earlier for the 5-Me-MOE-U amidite,
was successfully applied to compound 6 with only minor
modifications. To determine the best extraction method to
remove P(V) impurities, we started with crude material
containing 30% of the P(V) species. We then varied the ratio
of DMF/H2O in the extractions and analyzed the products
for the amount of residual P(V) impurities. The results,
summarized below in Table 1, indicate that the use of 1:1
DMF/H2O afforded the best balance of product recovery and
impurity rejection for the MOE-A amidite.
Precipitation. Attempts to crystallize MOE-A utilizing
binary solvent systems such as CH2Cl2/hexane, acetone/
water, acetonitrile/MTBE, acetonitrile/hexanes, EtOAc/hex-
ane, acetone/heptane, and toluene/heptane afforded thick oils
or gummy solids. Accordingly, we turned our attention to
the development of a feasible precipitation method compar-
able to that developed for the 5-Me-MOE-U amidite.
were recorded respectively at 500 and 125 MHz on a Varian
Inova spectrometer with the solvents as internal standard
(δH: CDCl3 7.26 ppm; δC: CDCl3 77.0 ppm). J values are
given in Hz. 31P NMR spectra were recorded at 202 MHz
on a Varian Inova spectrometer with H3PO4 as an external
standard (δP: H3PO4 0.00 ppm). Product purities and in-
process controls were measured by HPLC employing a
Discovery HS C18 column (7.5 cm × 4.6 mm i.d., 3 um,
s/n: 40962-03), plus a 2 cm × 4.0 mm, 3 um guard, at a
column temperature of 30 °C and UV wavelength of 260
nm. The gradient elution was 60% acetonitrile/10 mM NH4-
OAc ramping to 90% acetonitrle/10 mM NH4OAc over 15
min at flow rate of 1.0 mL/min, then kept at 90% acetonitrle/
10 mM NH4OAc for 8 min.
Initial Process Procedure.22-23 [5′-O-(4,4′-Dimethoxy-
triphenylmethyl)-2′-O-(2-methoxyethyl)-5-methyluridin-3′-O-
yl]-2-cyanoethyl-N,N-diisopropylphosphoramidite (5). 5′-O-
(22) Mole percentage of the remaining P(V) impurities is determined by the
sum of integration of all P(V) peaks which is <60 ppm in 31P NMR (CDCl3).
(23) The initial process was operated in 50-L three-necked flasks.
(21) The impurity was assumed to be dimer, with a similar chemical shift as 4
(
31P NMR).
Vol. 9, No. 6, 2005 / Organic Process Research & Development
•
735

(4,4′-Dimethoxytriphenylmethyl)-2′-O-(2-methoxyethyl)-5-
methyluridine (1237 g, 2.0 mol) was dissolved in anhydrous
DMF (2.5 L). The solution was co-evaporated with toluene
(200 mL) at 50 °C under reduced pressure. After cooling to
room temperature, 2-cyanoethyl tetraisopropylphosphorodia-
midite (900 g, 3.0 mol) and tetrazole (70 g, 1.0 mol) were
added. The mixture was shaken until all tetrazole was
dissolved, and N-methylimidazole (20 mL) was added. After
stirring at room temperature for 5 h, triethylamine (300 mL)
was added. The resulting mixture was diluted with DMF (3.5
L) and water (600 mL) and then was extracted with hexane
(3 × 3 L). The aqueous layer was further diluted with water
(1.6 L) and extracted with a mixture of toluene (12 L) and
hexanes (9 L). The two layers were separated, and the upper
layer was washed with DMF/water (7:3, v/v, 3 × 3 L) and
water (3 × 3 L). The upper layer was dried (Na2SO4),
filtered, and concentrated in vacuo. The residue was co-
evaporated with acetonitrile (2 × 2 L) under reduced pressure
and dried to a constant weight (25 °C, 0.1 mmHg, 40 h) to
give the product as an off-white foam.
was also transferred through the line to the 8-gal vessel. The
end solution was held under nitrogen for further processing.
Acetonitrile (10 L), NMI‚Tf (1.41 kg, 6.07 mol), and
2-cyanoethyl tetraisopropylphosphorodiamidite (2.56 kg, 8.49
mol) were subsequently charged to a separate 30-gal glass-
lined reactor. The vacuum-dried, acetonitrile solution of (5′-
O-(4,4′-dimethoxytriphenylmethyl)-2′-O-(2-methoxyethyl)-
5-methyluridine from above was added to this second reactor
over 30-60 min, keeping the tank temperature at 20-25
°C. The resulting mixture was stirred at 20-25 °C for 12-
20 h until the starting material was <0.1% by HPLC. The
tank contents were then transferred to a 50-gal glass-lined
reactor, diluted with MTBE (75 L), washed with water (25
L), 1:1 (v/v) DMF:H2O (2 × 50 L), water (25 L), and 13%
brine (25 L). The MTBE layer was dried with Na2SO4 (10
kg) and filtered into a clean 30-gal glass-lined reactor, and
the filter cake rinsed with MTBE (20 L). This product
solution was then concentrated via vacuum distillation to a
volume of 17 L while maintaining the liquid temperature at
<35 °C.
[5′-O-(4,4′-Dimethoxytriphenylmethyl)-2′-O-(2-methoxy-
ethyl)-N6-benzoyladenosin-3′-O-yl]-2-cyanoethyl-N,N-diiso-
propylphosphoramidite (6). 5′-O-(4,4′-Dimethoxytriphenyl-
methyl)-2′-O-(2-methoxyethyl)-N6-benzoyladenosine (1098
g, 1.5 mol) was dissolved in anhydrous DMF (3 L). The
resulting solution was coevaporated with toluene (300 mL)
at 50 °C under reduced pressure. After cooling to ambient
temperature, 2-cyanoethyl tetraisopropylphosphorodiamidite
(680 g, 2.26 mol) and tetrazole (78.8 g, 1.24 mol) were
added. This mixture was shaken until all tetrazole was
dissolved and then N-methylimidazole (30 mL) was added.
After stirring at ambient temperature for 5 h, triethylamine
(300 mL) was added. The reaction mixture was diluted with
DMF (1 L) and water (400 mL), and then extracted with
hexanes (3 × 3 L). The aqueous layer was further diluted
with water (1.4 L) and extracted with a mixture of toluene
(9 L) and hexanes (6 L). The two layers were separated and
the upper layer was washed with DMF/water (60:40, v/v; 3
× 3 L) and water (3 × 2 L). The organic layer was dried
(Na2SO4), filtered and evaporated to a sticky foam. The
residue was coevaporated with acetonitrile (2.5 L) under
reduced pressure and dried to a constant weight (25 °C, 0.1
mmHg, 40 h) to give the product as an off-white foam.
Improved Pilot-Plant Procedures. [5′-O-(4,4′-Dimethoxy-
triphenylmethyl)-2′-O-(2-methoxyethyl)-5-methyluridin-3′-O-
yl]-2-cyanoethyl-N,N-diisopropylphosphoramidite (5). A 30-
gal glass-lined reactor was charged with (5′-O-(4,4′-
dimethoxytriphenylmethyl)-2′-O-(2-methoxyethyl)-5-
methyluridine (5.00 kg, 8.08 mol) and acetonitrile (30 L).
The resulting solution was concentrated via vacuum distil-
lation (54-31 mmHg) to ∼15 L at 20-25 °C (vapor
temperature) until the water content was <0.05% by Karl
Fischer titration. The tank contents were filtered through a
5 µ cartridge filter24 into a portable 8-gal vessel. The original
tank was rinsed with 3 L of dry acetonitrile, and this rinse
Heptane (150 L) was charged to a 50-gal glass-lined
reactor and held for use during the following precipitation
procedure. Heptane (125 L) was charged to a second 50-gal
glass-lined reactor which was set up for vacuum reflux/
distillation. Sufficient vacuum and heat were applied to
achieve reflux of the heptane at a liquid temperature of 20-
25 °C. Once the heptane began to reflux, the reactor was
switched to distillation mode. Next, the simultaneous co-
addition of fresh heptane from the first 50-gal reactor and
the MTBE solution of crude 5 from the 30-gal reactor was
initiated (at a ratio of ∼8:1 heptane:product-MTBE solution)
while maintaining a constant 125-L volume in the precipita-
tion reactor. (Note: The product begins to precipitate
immediately with its addition to the heptane solution). After
the addition of the product solution was complete, the
distillation was continued until the solution volume was
reduced to 50 L. Additional heptane was then added to adjust
the volume of the slurry to 125 L. Heating was stopped, and
the resulting precipitate was allowed to stir at ambient
temperature for 1.5 h.
The precipitate was vacuum filtered, and the solids were
rinsed with hexane (3 × 25 L) and then allowed to pull dry.
The filter cake was removed to a tared tray and vacuum-
dried at 35 °C to constant weight. A total of 7.61 kg of white
solids was recovered (93.2%) with an HPLC purity of 99.1%
as a 1:1 mixture of a pair of diastereomers at the phosphorus
center: 1H NMR (500 MHz, CDCl3) δ 8.88-8.94 (br, 1),
7.69 (d, 1 × 0.5, J ) 1.1), 7.62 (d, 1 × 0.5, J ) 1.1), 7.44-
7.40 (m, 2), 7.33-7.22 (m, 7), 6.85-6.81 (m, 4), 6.08 (d, 1
× 0.5, J ) 5.0), 6.03 (d, 1 × 0.5, J ) 4.5), 4.54 (ddd, 1 ×
0.5, J ) 10.0, 5.0, 5.0), 4.48 (ddd, 1 × 0.5, J ) 11.0, 5.0,
5.0), 4.30-4.28 (m, 1), 4.23-4.20 (m, 1), 3.98-3.50 (m,
9), 3.79 (s, 6), 3.33 (s, 3), 3.35-3.28 (m, 1), 2.65 (dd, 1, J
) 6.0, 6.0), 2.38 (dd, 1, J ) 6.0, 6.0), 1.34 (d, 3, J ) 3.2),
1.18-1.16 (m, 9), 1.01 (d, 3, J ) 7.0); 13C NMR (125 MHz,
CDCl3) δ (164.1, 164.0), 158.7, (150.6, 150.5), (144.3,
144.2), 135.7, 135.4, (135.30, 135.2), (130.30, 130.26),
(130.19, 130.17), (128.4, 128.2), (127.98, 127.95), 127.2,
(24) Filtration is required to remove unreactive solid particles contained in the
custom-made raw material (5′-O-(4,4′-dimethoxytriphenylmethyl)-2′-O-(2-
methoxyethyl)-5-methyluridine. This unit of operation could be ignored if
a high-quality raw material is employed.
736
•
Vol. 9, No. 6, 2005 / Organic Process Research & Development

(117.9, 117.4), (113.24, 113.20), (87.6, 87.1), (87.1, 86.9),
(81.8, 81.2), (72.4, 72.2), (70.7, 70.6), (70.4, 70.3), (70.0,
70.0), (62.7, 61.9), (58.97, 58.84), (57.9, 57.8), 55.3, (43.34,
43.23), (43.20, 43.12), (24.62, 24.57), (20.45, 20.40), (20.22,
20.17), (11.69, 11.63); 31P NMR (202 MHz, CD2Cl2) δ
(152.6, 152.6).
then filtered. The solids were washed with hexanes (3 × 13
L) and pulled dry. The filter cake was removed to a tared
tray and dried under vacuum at 35 °C to constant weight. A
total of 7.31 kg of white solids was recovered (95.0%) with
a HPLC purity of 98.9% as a 1:1 mixture of a pair of
1
diastereomers at the phosphorus center: H NMR (500 MHz,
[5′-O-(4,4′-Dimethoxytriphenylmethyl)-2′-O-(2-methoxy-
ethyl)-N6-benzoyladenosin-3′-O-yl]-2-cyanoethyl-N,N-diiso-
propylphosphoramidite (6). A solution of 5′-O-(4,4′-di-
methoxytriphenylmethyl)-2′-O-(2-methoxyethyl)-N6-benzoyl-
adenosine (2.50 kg, 3.42 mol) in DMF (5 L) and acetonitrile
(1 L) was prepared in a 22-L round-bottom Buchi flask. The
flask was place on a Buchi rota-vapor, and its contents were
concentrated in vacuo at 20-35 °C (vapor temperature) while
dry acetonitrile (17 L) was simultaneously added at a
sufficient rate to maintain a constant flask volume. (This
vacuum distillation was continued until the water content of
the final solution was <0.04% by Karl Fischer analysis.)
A second 22-L Buchi flask was charged with identical
amounts of starting material and solvents as those above and
the resulting solution subjected to the same vacuum distil-
lation conditions. The resulting two portions of dried starting
materials (5.00 kg, 6.83 mol total) were held under nitrogen
for further processing.
CDCl3) δ 8.95 (br, s, 1), 8.75 (s, 1 × 0.5), 8.72 (s, 1 × 0.5),
8.27 (s, 1 × 0.5), 8.23 (s, 1 × 0.5), 8.02-8.01 (m, 2), 7.62-
7.59 (m, 1), 7.55-7.51 (m, 2), 7.45-7.40 (m, 2), 7.35-
7.20 (m, 7), 6.82-6.75 (m, 4), 6.18 (d, 1 × 0.5, J ) 5.5),
6.18 (d, 1 × 0.5, J ) 5.5), 4.95 (dd, 1 × 0.5, J ) 5.5, 5.0),
4.91 (dd, 1 × 0.5, J ) 5.5, 5.0), 4.66 (ddd, 1 × 0.5, J )
10.5, 5.0, 5.0), 4.61 (ddd, 1 × 0.5, J ) 11.0, 4.1, 4.1), 4.45-
4.42 (m, 1 × 0.5), 4.38-4.35 (m, 1 × 0.5), 4.00-3.46 (m,
15), 3.35-3.32 (m, 1), 3.24 (s, 3), 2.67-2.63 (m, 1), 2.40-
2.37 (m, 1), 1.24-1.16 (m, 9), 1.08 (d, 3, J ) 7.0); 13C NMR
(125 MHz, CDCl3) δ 164.5, 158.6, 152.7, 151.8, 149.5,
(144.53, 144.46), (142.3, 142.2), 135.7, 135.6, 133.7, 132.7,
130.2, 130.1, 128.9, 128.3, 128.2, 127.8, 126.9, (117.8,
117.4), 113.1, (87.19, 87.09), (86.64, 86.54), (81.17, 80.69),
(72.1, 72.0), (71.5, 71.4), (70.9, 70.8), (70.4, 70.1), (63.1,
62.7), (59.0, 58.9), (58.0, 57.9), 55.2, (43.4, 43.3), (43.2,
43.1), (24.62, 24.59), (20.43, 20.38), (20.17, 20.11); 31P NMR
(202 MHz, CD2Cl2) δ (152.9, 152.6).
To a 30-gal glass-lined reactor at ambient temperature
was charged acetonitrile (10 L), NMI‚Tf (1.59 kg, 6.84 mol),
and 2-cyanoethyl tetraisopropylphosphorodiamidite (2.57 kg,
8.53 mol). The two dry DMF/acetonitrile solutions of 5′-O-
(4,4′-dimethoxytriphenylmethyl)-2′-O-(2-methoxyethyl)-N6-
benzoyladenosine from above were combined and added over
30-60 min to the 30-gal tank. The tank temperature was
maintained between 20 and 25 °C during this addition. The
reaction mixture was stirred at 20-25 °C for 19 h until the
starting material was <0.1% by HPLC. The ending mixture
was transferred to a 50-gal glass-lined reactor, diluted with
MTBE (75 L), washed sequentially with water (25 L), 1:1
DMF/H2O (2 × 50 L), water (50 L), and brine (50 L). The
resulting MTBE solution was dried with Na2SO4 (10 kg) and
filtered into a clean 30-gal glass-lined reactor. This filtrate
solution was concentrated to 20 L by vacuum distillation
while maintaining the tank temperature below 25 °C. MTBE
(50 L) was added to this product solution which was again
concentrated to a volume of 20 L. This MTBE dilution and
distillation sequence was repeated two more times, with the
final concentration stopping at 40 L total volume.
Preparation of N-Methylimidazole Triflate Salt. To a
30-gal glass-lined reactor was charged 5.32 kg of 1-meth-
ylimidazole and 36 L of methylene chloride. The solution
was stirred at room temperature for 30 min before the
addition of 9.16 kg of triflic acid; 36 L of MTBE was then
added for crystallization. The resulting mixture was stirred
at room temperature for 30 min and filtered. The solids were
rinsed with 18 L of MTBE, and dried to provide 13.7 kg
(96.6%) of the desired salt.
Acknowledgment
We thank Dr. Quanlai Song and Dr. Bruce Ross at Isis
Pharmaceuticals for providing the initial process procedures
and helpful discussions. We also thank Mr. Jason Cronin,
Mr. John Schafer, Mr. Joe Pawlak for providing NMI‚Tf
salts. We acknowledge Mr. David Robbins, Mr. Paul Dodson,
Ms. Lisa Zollars, and Dr. Chris Doecke for generous
analytical support, and Ms. Lynne Buccilli for her important
work with sourcing raw materials and obtaining pricing
information. We are grateful to Mr. Jeff Niemeier for sharing
the hazard information of 1H-tetrazole and Lilly Chemical
Hazards Laboratory for thermal stability tests.
The MTBE solution containing the crude, concentrated
product was added over 30-60 min to stirring hexane (95
L) in a second 50-gal glass-lined reactor. The resulting
precipitate was stirred at ambient temperature for 1 h and
Received for review May 23, 2005.
OP050077D
Vol. 9, No. 6, 2005 / Organic Process Research & Development
•
737