Synthesis of Rovafovir Etalafenamide (Part IV): Evolution of the Synthetic Process to the Fluorinated Nucleoside Fragment

Article abstract of DOI:10.1021/acs.oprd.1c00026

Fluorinated nucleoside 1 is a key starting material in the synthesis of rovafovir etalafenamide (2), a novel nucleotide reverse transcriptase inhibitor under development at Gilead Sciences for the treatment of HIV. While an initial manufacturing route enabled the production of 1 to support clinical development, alternative approaches were explored to further enhance manufacturing effectiveness, improve processing time, reduce cost, and minimize the environmental impact. Toward this end, two new routes were developed to a key synthetic intermediate, which was converted to 1 using a new protecting group strategy. The new chemistry led to improvements in the manufacturing process while reducing the overall process mass intensity (PMI).

Full text of DOI:10.1021/acs.oprd.1c00026

pubs.acs.org/OPRD
Article
Synthesis of Rovafovir Etalafenamide (Part IV): Evolution of the
Synthetic Process to the Fluorinated Nucleoside Fragment
David A. Siler,* Selcuk Calimsiz, Ian J. Doxsee, Bernard Kwong, Jeffrey D. Ng, Keshab Sarma, Jinyu Shen,
Jonah W. Curl, Jason A. Davy, Jeffrey A. O. Garber, Sura Ha, Olga Lapina, Jisung Lee, Lennie Lin,
Sangsun Park, Mary Rosario, Olivier St-Jean, and Guojun Yu
Cite This: Org. Process Res. Dev. 2021, 25, 1263−1274
Read Online
sı
*
Metrics & More
Article Recommendations
Supporting Information
ACCESS
ABSTRACT: Fluorinated nucleoside 1 is a key starting material in the synthesis of rovafovir etalafenamide (2), a novel nucleotide
reverse transcriptase inhibitor under development at Gilead Sciences for the treatment of HIV. While an initial manufacturing route
enabled the production of 1 to support clinical development, alternative approaches were explored to further enhance manufacturing
effectiveness, improve processing time, reduce cost, and minimize the environmental impact. Toward this end, two new routes were
developed to a key synthetic intermediate, which was converted to 1 using a new protecting group strategy. The new chemistry led
to improvements in the manufacturing process while reducing the overall process mass intensity (PMI).
KEYWORDS: fluorinated nucleoside, glycosylation, fluorination, enzymatic hydrolysis
several steps7) was brominated at the anomeric position. From
INTRODUCTION
Synthetic nucleoside and nucleotide analogues have played a
■
bromide 5, a subsequent glycosylation reaction with N6-
benzoyladenine (6) afforded 7, which was then selectively
debenzoylated to give the first isolated intermediate, diol 8. Bis-
silylation of 8 followed by a moderately selective deprotection of
the 5′-silyl ether yielded 10 in a telescoped sequence. Following
oxidation to 11, a final global deprotection afforded 1. This
manufacturing route provided 1 with an overall yield of 41%,
enabling the production of sufficient amounts of 1 to support the
clinical development of 2.
prominent role in the treatment of a variety of conditions,
including viral infections.1 Fluorinated nucleosides and
nucleotides, in particular, have received attention as the
incorporation of fluorine into these structures can improve
their biological activity and stability.2,3 Rovafovir etalafenamide
(2, Scheme 1), a fluorinated nucleotide analogue of 2′,3′-
Scheme 1. Retrosynthetic Approach to 2
Several areas for improvement of the route, however, were
noted. Specifically, the selective ester hydrolysis of 7 required
careful control of the stoichiometry and reaction temperature to
avoid overhydrolysis of the benzamide moiety, while the
subsequent workup and crystallization of 8 were solvent- and
time-intensive. As selective direct oxidation of the 5′-OH group
was found to be challenging, the two-step sequence consisting of
bis-silylation followed by deprotection was developed; this
approach contributed to lengthy manufacturing operations and
high overall process mass intensity (PMI)8 for 10. Additionally,
the lability of the silyl ether functionality resulted in the
formation of significant amounts of 8 along with 10, which not
only led to an erosion in yield but also impacted the downstream
oxidation reaction, workup, and isolation.
didehydro-2′,3′-dideoxy adenosine (d4AP), maintains the anti-
HIV activity of the nonfluorinated analogue while exhibiting
lower mitochondrial toxicity and is being developed for the
treatment of HIV.4 The accompanying articles in this issue
describe the synthesis of 2 via a convergent synthesis which
combines the fluorinated nucleoside core 1 and the
phosphonamidate ester 3.5 The development of a manufactur-
ing process for the nucleoside core is described herein.
Received: January 21, 2021
Published: May 7, 2021
An initial manufacturing route for 1 is shown in Scheme 2.6 In
this route, the advanced fluoro glycosyl benzoate 4 (prepared via
https://doi.org/10.1021/acs.oprd.1c00026
Org. Process Res. Dev. 2021, 25, 1263−1274
© 2021 American Chemical Society
1263

Organic Process Research & Development
pubs.acs.org/OPRD
Article
a
Scheme 2. Initial Manufacturing Route for 1
a
Reagents and conditions: (a) 33 wt % HBr in AcOH, CH2Cl2, 0 °C; (b) tetrahydrofuran (THF), N-methyl-2-pyrrolidone (NMP), 68 °C; (c)
NaOH, THF, H2O, 0 °C, 59% from 4; (d) Et3SiCl, iPr2NEt, PhMe, 50 °C; (e) TsOH·H2O, MeOH, 3 °C, 73% from 8; (f) PhI(OAc)2, (2,2,6,6-
tetramethylpiperidin-1-yl)oxidanyl (TEMPO), Na2HPO4, MeCN, H2O, 22 °C; and (g) 25 wt % NaOMe in MeOH, MeOH, 22 °C, 96% from 10.
Scheme 3. New Synthetic Strategy for the Manufacturing of 1
a
RESULTS AND DISCUSSION
Scheme 4. Telescoped Route to 12 from 4
■
New Route Design. To address the difficulties with the
initial manufacturing process described above, a new synthetic
route to 1 was developed that leveraged the process knowledge
gained from the initial route. In this new route (Scheme 3),
global hydrolysis of 7 to new intermediate 12 would obviate the
need for careful control of a selective deprotection. From 12,
protection of the 3′-OH as an ester in place of the silyl ether was
envisioned for several reasons: (1) the acylation reaction would
be compatible with the free amino group on the adenine moiety;
(2) the resulting ester functionality is expected to have improved
chemical stability under the subsequent processing conditions
relative to the analogous silyl ether; and (3) enzymatic
hydrolysis could be leveraged to access intermediate 14.
a
Reagents and conditions: (a) 33 wt % HBr in AcOH, CH2Cl2, 0 °C;
(b) THF, NMP, 68 °C; and (c) LiOH, MeOH, H2O, 22 °C, 61%
from 4.
Improved Glycosylation Process/Global Debenzoyla-
tion to 12. The new route development began with the
refinement of the telescoped glycosylation step that forms 7
(Scheme 4). Following this sequence, 4 was first converted to
glycosyl bromide 5 with excellent diastereoselectivity (only the
α-diastereomer is observed via 19F nuclear magnetic resonance
(NMR)). After aqueous workup and drying, 5 was reacted with a
large excess of 6 (2.2 equiv) which also acts as a base to scavenge
the HBr produced during the reaction. The glycosylation
reaction is very sluggish, partly due to the low solubility of 6 in
the reaction mixture. Efforts to increase the solubility of 6, and
therefore increase the reaction rate, using more polar reaction
solvents led to a decrease in diastereoselectivity (Table 1), likely
due to a change in the reaction mechanism.9 For this reason, the
glycosylation reaction medium remained unchanged in the new
route (32:1 THF/NMP), affording a 95:5 mixture of anomers
favoring the β-diastereomer on scale.10 Inefficiencies in the
reaction workup, however, provided opportunities for develop-
1264
https://doi.org/10.1021/acs.oprd.1c00026
Org. Process Res. Dev. 2021, 25, 1263−1274

Organic Process Research & Development
pubs.acs.org/OPRD
Article
a
effectively purged to levels of 0.1−0.4% AN in the crystallization
of 12. Based on X-ray crystallography data (Figure 1), a
Table 1. Impact of the Solvent on the Glycosylation of 5
entry
solvent
ratio (β/α)
1
2
3
4
5
THF
95.2:4.8
94.0:6.0
93.8:6.2
92.6:7.4
91.6:8.4
THF/NMP (73:1)
THF/NMP (23:1)
THF/NMP (14:1)
THF/NMP (7:1)
a
Reagents and conditions: (a) THF, NMP (solvent ratio indicated in
the table), 68 °C.
ment. In the initial manufacturing process, a lengthy workup
procedure was used, including (1) filtration of the reaction
mixture to remove the majority of the excess 6 as the
corresponding HBr salt, (2) solvent exchange of the filtrate
from THF to dichloromethane to facilitate the subsequent
aqueous workup, (3) acidic aqueous workup to remove the
remaining 6, and (4) solvent exchange of the organic layer from
dichloromethane back to THF for compatibility with the
downstream debenzoylation conditions. Overall, these solvent
exchanges contributed to an increase in processing time, cost,
and waste. To streamline the workup, it was found that the THF-
rich filtrate (after glycosylation and filtration) could be washed
directly with a mixture of dilute HCl and brine to efficiently
remove residual 6, obviating the need for solvent exchanges.
With the workup addressed, the hydrolysis reaction
conditions were modified to allow for the complete hydrolysis
of 7. Simply increasing the reaction temperature from 0 to 50 °C
led to full conversion to 12, but product decomposition was also
observed. After additional screening, methanol was identified as
a beneficial cosolvent, improving the reaction rate while allowing
for the reaction to be performed at a lower temperature (22 °C).
Upon reaction completion, the mixture was acidified with
aqueous HCl, with 12 partitioning into the aqueous layer due to
its high solubility in water at low pH. This operation enabled the
rejection of many process impurities in the organic layer,
including the stoichiometric byproduct benzoic acid. With these
conditions, however, precipitation of sodium chloride was
observed during the workup due to the presence of methanol in
the aqueous layer. A change in the base from sodium hydroxide
to lithium hydroxide in the hydrolysis step improved the workup
as lithium chloride is more soluble in methanol-rich water;
gratifyingly, this change provided an increase in the reaction rate
as well. After workup, 12 could be crystallized from the aqueous
layer by first neutralizing to pH 6−8 and then removing residual
methanol by distillation. The new route provided 12 in 61%
isolated yield from 4 on kilogram scale (compared to 59%
isolated yield of 8 from 4 in the initial route). While the PMI for
this three-step telescoped process was increased by 27% relative
to the original route, this increase is primarily due to an
increased use of water; the solvent contribution to the PMI is
reduced by 15%. In addition, the simplification of the workup for
intermediate 7 and the direct crystallization of 12 from the
aqueous medium reduced the processing time of the telescoped
process. The glycosylation step produces 5−6% AN (area
normalized) of the α-anomer impurity, but this impurity is
Figure 1. Three-dimensional hydrogen bond network for 12 (single-
crystal X-ray structure).
significant hydrogen bonding network exists between molecules
of 12 in the solid state, including H-bonds between the 3′-OH
and a neighboring N3 and between the 5′-OH and a neighboring
6-NH2. This high degree of intermolecular bonding interactions
provides a possible explanation for the high crystallinity of 12 as
well as the effective purge of the α-anomer impurity, which is
unable to participate in this H-bonding network to the same
extent.
Alternative Route to 12 from Adenosine. Although the
modifications to the glycosylation sequence enabled a robust
synthesis of 12, an alternative route was investigated (Scheme 5)
to eliminate the time-consuming glycosylation sequence and
reduce the number of chemical steps (including the steps
required to prepare starting material 4). This alternative route
would make use of adenosine (15), a relatively inexpensive
starting material, but the challenges of selective protection of the
3′-OH and 5′-OH and subsequent fluorination of the 2′ position
would need to be addressed.
At the outset, the protection of the 3′ and 5′ hydroxy groups as
the corresponding benzoyl esters was explored to eventually
intercept intermediate 12. This could be accomplished via
selective deprotection of tri-O-benzoyladenosine (16)11 or a
selective di-acylation of adenosine.12 Whereas 16 was easily
prepared (benzoic anhydride, triethylamine, 4-dimethylamino-
pyridine (DMAP), and acetonitrile at 50 °C), the selective
mono-deprotection to 17 was considerably more challenging.
Reagents such as hydrazine and hydroxylamine were evaluated,
but low conversion was generally observed. Under these
conditions, attempts to push the conversion led to indiscrimi-
nate deprotection.
On the other hand, the direct di-acylation of adenosine
showed more promise (Table 2). Despite the fact that initial
screens indicated a tendency toward the formation of the
unwanted di-benzoyl isomer 21, this route was pursued with
confidence that 21 could be converted to 17 via acid- or base-
mediated isomerization.12 Overacylation was initially observed
even at low temperature (−5 °C); this could be reduced by
charging an initial 0.8 equiv of benzoyl chloride followed by
additional reagent charges as needed, although incomplete
1265
https://doi.org/10.1021/acs.oprd.1c00026
Org. Process Res. Dev. 2021, 25, 1263−1274

Organic Process Research & Development
pubs.acs.org/OPRD
Article
Scheme 5. Strategy for the Alternative Synthesis of 12
Table 2. Screening Results for the Direct Bis-Acylation of Adenosine
mono-Bz
di-Bz
tri-Bz
entry
pyridine (vol)
DMF (vol)
BzCl (equiv)
temp. (°C)
20 (% AN)
21 (% AN)
17 (% AN)
16 (% AN)
a
1
2
3
4
5
6
40
15
15
15
15
10
25
10
10
10
10
5
2.8
2.8
2.8
3.0
2.0
2.0
−5
−5
10
10
10
10
24
28
29
16
13
54
48
49
53
58
55
8
8
9
9
7
6
10
11
9
17
18
21
a
a
a
b
b
14
a
b
An initial 0.8 equiv of benzoyl chloride was charged, followed by charges of 0.2 equiv. An initial 1.6 equiv of benzoyl chloride was charged over 2
h.
Table 3. Purity of 17 through the Stages of the Benzoylation Process
entry
operation
20 (% AN)
21 (% AN)
17 (% AN)
16 (% AN)
1
2
3
4
end of reaction
after workup
liquors
13
3
6
56
43
6
6
26
20
98
21
23
55
0.6
isolated solids
1
0.3
conversion of the monobenzoylated intermediate 20 was
observed in this case (entry 1). With portion-wise addition,
the solvent charge could be reduced with minimal impact to the
reaction profile (entry 2), and the reaction could be warmed to
10 °C as well (entry 3). To increase the reaction conversion,
additional benzoyl chloride could be charged (entry 4), although
this primarily decreased 20 while increasing the tri-benzoylated
impurity 16. Further improvement was gained by slowly
charging 1.6 equiv of benzoyl chloride over 2 h and then
charging an additional 0.4 equiv as needed (entry 5), reducing
the total quantity of benzoyl chloride required. With these
improvements, di-benzoylated isomers 17 and 21 were formed
in a combined 60−65% AN (entry 6) favoring regioisomer 21.13
As expected, partial isomerization of 21 to 17 via benzoate
migration was observed following acidic and basic aqueous
washes (Table 3, entries 1 and 2). An additional benefit of
washing with acidic solutions is that a significant portion of the
monobenzoylated impurity 20 can be purged into the aqueous
layer. As potential crystallization conditions were evaluated, it
was found that 17 could be crystallized from acetonitrile while
rejecting isomer 21. This discovery, combined with the observed
propensity for isomerization in the presence of acid or base, led
to the development of a reactive crystallization. After a solvent
exchange to acetonitrile, DMAP was added to facilitate
isomerization propelled by crystallization of 17 (entries 3 and
4). A purge of the tri-benzoylated impurity 16 was also observed
in this crystallization, which provided 17 in 52% yield and >98%
AN purity.
1266
https://doi.org/10.1021/acs.oprd.1c00026
Org. Process Res. Dev. 2021, 25, 1263−1274

Organic Process Research & Development
pubs.acs.org/OPRD
Article
With access to 17, the 2′-OH was readily activated as the
imidazole-1-sulfonate ester14 (22, Scheme 6). Direct fluorina-
Subjecting 24 to the fluorination reaction provided 7 in 60%
yield with no trace of any migration byproducts, albeit with a
small amount of the corresponding 2-chloro analogue (the
protected precursor of this impurity is formed during the
synthesis of 24); this impurity, however, could be suppressed by
conducting the benzoylation at low temperature. Intermediate 7
was then telescoped into the hydrolysis step using the conditions
from the route described above to access 12, with a three-step
yield of 46% from 22 (24 can be crystallized prior to the
fluorination reaction, if required). Overall, 12 was obtained in
23% yield (five steps, three to four isolations) from adenosine
(15) with a purity of >98% AN on several hundred grams scale.
Further development of this route will focus on reducing the
PMI and the use of chlorinated solvents to lessen the
environmental impact of the processes.
a
Scheme 6. Fluorination from 17
a
Revised Protecting Group Strategy for the Final Steps.
With two independent routes to 12 in hand, the development of
the downstream processes was addressed. To selectively protect
the 3′-OH group, an approach via a diester intermediate was
evaluated. From this vantage, enzyme-mediated ester hydrolysis
could provide an environmentally friendly route to the desired
3′-ester, particularly as high site-selectivity has been demon-
strated in the case of nucleoside diesters.15 While the use of a
benzoyl protecting group would have been ideal as it would
allow for the direct use of 7 without the need for hydrolysis and
re-protection, initial screening revealed low productivity for this
substrate. For the purposes of screening, 8 was acetylated, and
the corresponding product 25 was screened across a panel of
enzymes (Table 4). Consistent with the known selectivity of
these enzymes for nucleoside diester substrates, lipase A from
Candida antarctica and Amano lipase from Burkholderia cepacia
primarily formed the regioisomer 27 (entries 1 and 2), while
lipase B from C. antarctica exhibited a preference for the
formation of 26, albeit with significant overhydrolysis (entry 4).
Switching to an immobilized lipase improved the conversion to
26 (entry 6), with Lipolase 100T (entry 7) and NS 99013 (entry
8) giving high conversion and selectivity. The drawback to this
approach, however, is that 26 was found to be poorly soluble in
most organic solvents. This finding was particularly problematic
as the best results for the hydrolysis were achieved using
immobilized enzymes and removal of the enzyme after reaction
completion via filtration was desired. Fortunately, an improve-
ment in product solubility could be achieved via the replacement
of the acetate protecting groups with longer-chain esters. This
Reagents and conditions: (a) (1) SO2Cl2, pyridine, MeTHF, DMF,
−10 °C; then, (2) imidazole, MeTHF, −10 to 0 °C, 94%; (b) Et3N·
3HF, Cy2NMe, THF, 60 °C, 30−40%, 5:1 (19:23).
tion of 22 with Et3N·3HF, however, was complicated with a low
yield and the formation of an unexpected migration product
(23), so alternative approaches were investigated. It was
hypothesized that modulation of the electron density in the
adenine moiety might suppress its migratory tendency. To test
this hypothesis, the N6 nitrogen was acylated with benzoyl
chloride, giving 24 (along with a bis-benzoylated impurity at the
N6 position; not shown) in 89% isolated yield (Scheme 7).
a
Scheme 7. Revised Fluorination Sequence to 12
a
Reagents and conditions: (a) BzCl, NMI, CH2Cl2, 0−10 °C, 89%;
(b) Et3N·3HF, Cy2NMe, EtOAc, PhMe, 60 °C; and (c) LiOH,
MeOH, H2O, 20 °C, 52% from 24.
a
Table 4. Enzyme Screening with Substrate 25
entry
enzyme
25 (% AN)
26 (% AN)
27 (% AN)
8 (% AN)
1
2
3
4
5
6
7
8
lipase A, Candida antarctica
6.1
9.4
13.8
26.4
72.6
0
0
0.2
0
16.1
24.3
54.2
93.0
96.8
82.8
83.9
80.9
0.3
0.8
0.2
0
10.8
6.5
1.7
55.7
0.4
45.6
2.4
Amano lipase, Burkholderia cepacia, PS
cholesterol esterase, Pseudomonas fluorescens
lipase B, Candida antarctica
lipase, Thermomyces lanuginosus, solution
lipase, Thermomyces lanuginosus, Immobead
b
Lipolase 100T
NS 99013
4.3
1.4
b
0
1.8
a
Reagents and conditions: (a) enzyme, dimethyl sulfoxide (DMSO), 0.1 M aqueous sodium phosphate buffer (pH 6.8), ambient temperature, 20 h.
b
Received from Novozymes A/S.
1267
https://doi.org/10.1021/acs.oprd.1c00026
Org. Process Res. Dev. 2021, 25, 1263−1274

Organic Process Research & Development
pubs.acs.org/OPRD
Article
a
Scheme 8. Synthesis of 29 from 12 via Enzyme-Mediated Hydrolysis
a
Reagents and conditions: (a) hexanoic anhydride, pyridine, MeCN, 70 °C; (b) Lipolase 100T, MIBK, 0.1 M aqueous sodium phosphate buffer
(pH 7), 35 °C, 80% from 12.
Figure 2. Solvent screen for the hydrolysis of 28 to 29. Lipolase 100T (20 wt %), organic solvent (10 mL/g), 0.1 M aqueous sodium phosphate buffer
(20 mL/g), 35 °C; note: charges are based on the input of 12.
change in protecting groups had minimal to no impact on the
conversion and site-selectivity in the hydrolysis reaction, and
ultimately the hexanoate protecting group was selected for
further development based on the ready availability of hexanoic
anhydride.
With a suitable protecting group selected, reaction conditions
were developed to carry out the esterification of 12 and
enzymatic hydrolysis of the primary ester as a telescoped
sequence (Scheme 8). Intermediate 12 has an additional
reactive functional group with the 6-NH2 moiety as compared
to 8; fortunately, this could be protected concomitantly with the
3′-OH and 5′-OH groups, using hexanoic anhydride (3.6 equiv)
and the combination of pyridine as base and acetonitrile as
solvent. The selection of pyridine as a base for the reaction was
crucial, as use of common alkyl amine bases such as
triethylamine led to bis-acylation of the amine.
After aqueous workup, intermediate 28 was subjected to the
enzymatic hydrolysis reaction without further purification.
During early development, 2-MeTHF was used to facilitate
the aqueous workup of 28, and combining the 2-MeTHF stream
with 0.1 M aqueous sodium phosphate buffer (pH 7)16 and
Lipolase 100T (10 wt %) enabled productive conversion to 29
with high site-selectivity (the isomer resulting from hydrolysis at
the 3′-position was not detected). The reaction, however,
required >40 h to achieve >90% conversion. An evaluation of the
initial pH of the reaction (from 6 to 8) revealed an important
dependence of the reaction rate on the pH, with lower pH
providing a boost to the rate (>90% conversion after 17 h at pH
6). Upon scale-up, however, the reaction conversion ranged
from 85 to 95% across trials. To reduce this variability, several
alternative organic solvents were tested (Figure 2). From this
study, DMSO, tert-butyl methyl ether (MTBE), and methyl
isobutyl ketone (MIBK) were identified as potential solvents
based on the rate of conversion in these cases. DMSO was not
pursued further as its miscibility with water would complicate
the workup; likewise, MTBE was not evaluated as 29 was found
to have low solubility and would precipitate during the reaction,
so MIBK was selected for further development. Optimization
studies demonstrated that the reaction with MIBK was less
sensitive to the initial pH across the range of 6−8. Using the
preferred reaction conditions, the solution of 28 in MIBK was
combined with 0.1 M aqueous sodium phosphate buffer (pH 7),
and then the entire mixture was adjusted to a pH of 7.5 prior to
enzyme charging. This protocol allowed for the pH to drift
during the reaction (hexanoic acid is produced as a
stoichiometric byproduct) while remaining within the desired
range. The loading of the enzyme could be reduced to as low as 1
wt %, providing >95% conversion with overnight aging, but a 5
wt % loading was preferred to ensure the robustness of the
process. The amount of solvent was also found to influence the
reaction conversion, with increased conversion observed at a
higher dilution. This trend is likely due to the reaction being an
equilibrium between 28 and 29/hexanoic acid; the hypothesis
was supported by the observation that ca. 2% 28 was formed
when 29 and hexanoic acid were combined under the reaction
conditions. Using 8 mL/g MIBK and 16 mL/g pH 7 buffer
consistently provided 95−98% conversion to 29. After removing
the enzyme via filtration and performing an aqueous workup, 29
was crystallized from MIBK and heptane; gratifyingly, this
crystallization efficiently purges the unreacted starting material,
providing 29 in >98% AN purity. As observed with 12, hydrogen
bonding plays an important role in the crystal lattice of 29
(Figure 3), which may explain the excellent purge of 28 in
addition to other process impurities. Overall, 29 was obtained in
80% isolated yield on kilogram scale with a 56% PMI reduction
(relative to the synthesis of 10 from 8 in the initial route).
Safe and scalable oxidation conditions were developed for the
conversion of silyl ether 10 to intermediate 11 (see the
1268
https://doi.org/10.1021/acs.oprd.1c00026
Org. Process Res. Dev. 2021, 25, 1263−1274

Organic Process Research & Development
pubs.acs.org/OPRD
Article
usage (18%). The increased yield, reduced PMI, and enhance-
ments in operational ease are hallmarks of the new process.
EXPERIMENTAL SECTION
■
General Information. All reagents were commercially
available and used as received. Liquid chromatography (LC)
analyses were performed on a Waters Acquity H-Class UPLC
system equipped with a photodiode array detector or a Waters
Alliance e2695 HPLC system equipped with a photodiode array
detector. Purities are reported on an area normalized basis (%
AN) as determined by UV detection at the indicated
wavelength. NMR analyses were performed on a Bruker 400
MHz AVANCE spectrometer equipped with a 5 mm BBO
broadband probe. Chemical shifts for H and 13C spectra are
1
reported in ppm relative to tetramethylsilane and were
referenced to the solvent residual signal from the deuterated
NMR solvent.
LC Methods. Method A (UPLC)Acquity HSS T3, 1.8
μm, 2.1 mm × 150 mm; 98−10% gradient of 0.1% trifluoroacetic
acid (TFA) in water and 0.1% TFA in acetonitrile; flow rate of
0.4 mL/min; column temperature of 30 °C; acquisition time of
24 min; and UV detection at 260 nm.
Method B (HPLC)Agilent Poroshell 120 EC-C18, 2.7 μm,
4.6 mm × 100 mm; 90−10% gradient of 0.025% TFA in water,
and 0.025% TFA in acetonitrile; flow rate of 1.0 mL/min;
column temperature of 25 °C; acquisition time of 14 min; and
UV detection at 200 nm.
Method C (UPLC)Acquity HSS T3, 1.8 μm, 2.1 mm × 100
mm; 70−10% gradient of 0.1% formic acid in water and 0.1%
formic acid in acetonitrile; flow rate of 0.5 mL/min; column
temperature of 30 °C; acquisition time of 13 min; and UV
detection at 260 nm.
Method D (UPLC)Acquity CSH C18, 1.7 μm, 2.1 mm ×
150 mm; 95−5% gradient of 0.1% formic acid in water and 0.1%
formic acid in acetonitrile; flow rate of 0.4 mL/min; column
temperature of 50 °C; acquisition time of 16 min; and UV
detection at 260 nm or 274 nm (as indicated).
Method E (UPLC)Acquity HSS T3, 1.8 μm, 2.1 mm × 100
mm; 70−30% gradient of 0.1% TFA in water and 0.1% TFA in
acetonitrile; flow rate of 0.5 mL/min; column temperature of 30
°C; acquisition time of 27 min; and UV detection at 274 nm.
Preparation of (2R,3R,4S,5R)-5-(6-amino-9H-purin-9-yl)-4-
fluoro-2-(hydroxymethyl)tetrahydrofuran-3-ol (12) from 4.
To a reactor were charged 4 (1.5 kg, 3.2 mol) and
dichloromethane (5.7 L, 3.8 L/kg). The solution was cooled
to 0 °C, and then 33 wt % hydrobromic acid in acetic acid (3.1
kg, 13 mol HBr, 4.0 equiv) was charged while not exceeding an
internal temperature of 3 °C; the charge line was rinsed forward
with dichloromethane (0.23 L, 0.15 L/kg). The reaction mixture
was stirred overnight. Upon reaction completion (determined
by 19F NMR), the reaction mixture was diluted with dichloro-
methane (0.65 L, 0.43 L/kg), and then water (7.5 L, 5.0 L/kg)
Figure 3. Dimers of 29 via hydrogen bonding (single-crystal X-ray
structure).
Supporting Information for reaction calorimetry data), and
these conditions were expected to be suitable for the conversion
of ester 29 to intermediate 30; verification was conducted in the
laboratory on 65 g scale (Scheme 9). In the oxidation step, the
improved solubility of 29 enabled substantial improvement in
the reaction throughput via a reduction in solvent usage (60%
reduction in acetonitrile, 40% reduction in water). Moreover,
the enhanced stability of intermediate 30 allowed for a
straightforward aqueous workup. As further scale-up is
warranted, additional reaction calorimetry studies will be
performed to verify the safety of this revised process.
Intermediate 30 is not isolated and is instead carried forward
in a telescoped fashion. In the deprotection step, the 3′-ester was
found to rapidly undergo solvolysis, followed by cleavage of the
hexanamide within 1 h (18−24 h were required for the cleavage
of the N6-benzamide in the initial route). With these process
improvements, 1 was obtained in 96% isolated yield
(comparable to the initial route) and with a 37% reduction in
PMI for the two-step telescoped process.
CONCLUSIONS
■
In summary, two routes to intermediate 12 were developed and
demonstrated on 500 g (or greater) scale, providing
manufacturing and supply chain flexibility while reducing the
environmental impact of the processes. From 12, the develop-
ment of a scalable enzyme-mediated ester hydrolysis enabled a
more robust synthesis of 1. Using the route starting from 4
(Scheme 10), the process improvements increased the yield of 1
by 15% (from 41 to 47%) while reducing the overall PMI by 32%
(Figure 4). Organic solvent usage was dramatically reduced
(59%), which was partially off-set by a marginal increase in water
a
Scheme 9. Synthesis of 1 from 29
a
Reagents and conditions: (a) PhI(OAc)2, TEMPO, Na2HPO4·7H2O, MeCN, H2O, 22 °C; (b) 25 wt % NaOMe in MeOH, MeOH, 22 °C, 96%
from 29.
1269
https://doi.org/10.1021/acs.oprd.1c00026
Org. Process Res. Dev. 2021, 25, 1263−1274

Organic Process Research & Development
pubs.acs.org/OPRD
Article
a
Scheme 10. Revised Route to 1
a
Reagents and conditions: (a) 33 wt % HBr in AcOH, CH2Cl2, 0 °C; (b) THF, NMP, 68 °C; (c) LiOH, MeOH, H2O, 22 °C, 61% from 4; (d)
hexanoic anhydride, pyridine, MeCN, 70 °C; (e) Lipolase 100T, MIBK, 0.1 M aqueous sodium phosphate buffer (pH 7), 35 °C, 80% from 12; (f)
PhI(OAc)2, TEMPO, Na2HPO4·7H2O, MeCN, H2O, 22 °C; and (g) 25 wt % NaOMe in MeOH, MeOH, 22 °C, 96% from 29.
Figure 4. PMI comparison between the initial and revised routes from 4 to 1.
was charged while not exceeding an internal temperature of 15
°C. The layers were separated, and then the organic layer was
washed sequentially with water (7.5 L, 5.0 L/kg) and a solution
of sodium carbonate (0.75 kg, 0.50 kg/kg) in water (6.8 L, 4.5 L/
kg) while not exceeding an internal temperature of 15 °C. The
organic layer was diluted with THF (1.5 L, 1.0 L/kg) and then
distilled to a final volume of 3 L (2 L/kg) under vacuum at 35
°C. THF (5.0 L, 3.3 L/kg) was charged, and the resulting
solution was distilled to a final volume of 3 L (2 L/kg) under
vacuum at 35 °C; this operation was repeated one additional
time. The solution was transferred to a separate glass-lined
reactor, precharged with 6 (1.7 kg, 7.1 mol, 2.2 equiv), THF (13
L, 8.7 L/kg), and NMP (0.73 L, 0.49 L/kg); the transfer line was
rinsed forward with THF (11 L, 7.3 L/kg). The resulting slurry
was heated to 68 °C and stirred for 54 h. Upon reaction
completion (determined by 19F NMR), the slurry was cooled to
ambient temperature and filtered, rinsing forward with THF
(1.7 L, 1.1 L/kg), and then the filtrate was returned to the
reactor. A solution of sodium chloride (4.0 kg, 2.7 kg/kg) and
concentrated hydrochloric acid (1.8 kg, 1.2 kg/kg) in water (33
L, 22 L/kg) was prepared, and then the filtrate was washed three
times with the aqueous solutions (the solution was divided into
three equal portions). Methanol (7.6 L, 5.1 L/kg) was charged
to the organic layer and the solution was adjusted to 22 °C. A
solution of lithium hydroxide (0.39 kg, 16 mol, 5.0 equiv) in
water (7.8 L, 5.2 L/kg) was charged, and the resulting solution
was stirred for 21 h. Upon reaction completion (determined by
LC method A), a solution of concentrated hydrochloric acid
(2.0 kg, 1.3 kg/kg) in water (1.7 L, 1.1 L/kg) was charged,
followed by toluene (7.5 L, 5.0 L/kg). The layers were separated,
and the aqueous layer was adjusted to a pH of 6−8 with 25 wt %
aqueous sodium hydroxide (1.1 kg, 0.74 kg/kg). The aqueous
solution was distilled to a final volume of 14 L (9 L/kg) under
vacuum at 50 °C, and a slurry formed during the distillation. The
slurry was stirred for 1 h, cooled to 20 °C over 1 h, stirred
overnight, and then filtered. The wet cake was washed with
water (3.0 L, 2.0 L/kg). The wet cake was dried under vacuum at
50 °C to afford 12 (0.53 kg, 98.7% AN, LC method A) as an off-
white solid in 61% yield. 1H NMR (400 MHz, DMSO-d6) δ 8.24
(d, J = 2.1 Hz, 1H), 8.16 (s, 1H), 7.34 (s, 2H), 6.41 (dd, J = 14.5,
4.6 Hz, 1H), 5.95 (d, J = 5.0 Hz, 1H), 5.27 (t, J = 4.2 Hz, 0.5H),
5.18−5.07 (m, 1.5H), 4.45 (dtd, J = 18.9, 5.2, 3.8 Hz, 1H), 3.86
(q, J = 4.8 Hz, 1H), and 3.76−3.59 (m, 2H). 13C NMR (101
MHz, DMSO-d6) δ 156.02, 152.80, 149.16, 139.44 (d, J = 4.1
1270
https://doi.org/10.1021/acs.oprd.1c00026
Org. Process Res. Dev. 2021, 25, 1263−1274

Organic Process Research & Development
pubs.acs.org/OPRD
Article
Hz), 118.23, 95.42 (d, J = 192.1 Hz), 83.49 (d, J = 5.3 Hz), 81.43
(d, J = 17.0 Hz), 72.70 (d, J = 23.3 Hz), and 60.43. 19F NMR
(376 MHz, DMSO-d6) δ −197.65 to −197.92 (m). HRMS (ESI
+) calculated for C10H13FN5O3+ ([M + H]+) 270.09969, found
270.09882.
and the organic layer was distilled to a final volume of 2 L (3 L/
kg) under vacuum at 40 °C. 2-MeTHF (5.8 L, 10 L/kg) was
charged, and the solution was filtered through a 6 μm filter,
rinsing forward with 2-MeTHF (1.2 L, 2.1 L/kg). The filtrate
was distilled to a final volume of 5 L (8 L/kg) under vacuum at
40 °C, and then the reaction mixture was adjusted to 30 °C. The
solution was seeded with 22 (0.58 g, 0.001 kg/kg). A slurry
formed which was stirred for 20 min, and then it was cooled to
20 °C over 30 min. Heptane (2.6 L, 4.5 L/kg) was charged over
8 h, and then the slurry was stirred overnight and filtered. The
wet cake was washed with 1:1 2-MeTHF/heptane (2.9 L, 5.0 L/
kg). The wet cake was dried under vacuum at 40 °C to afford 22
(0.70 kg, 95.5% AN, LC method C, 6.3 wt % 2-MeTHF) as a
light yellow solid in 94% yield. 1H NMR (400 MHz, DMSO-d6)
δ 8.23 (s, 1H), 8.12 (s, 1H), 8.07−8.03 (m, 2H), 8.00−7.96 (m,
2H), 7.82 (s, 1H), 7.77−7.71 (m, 1H), 7.71−7.65 (m, 1H), 7.60
(t, J = 7.8 Hz, 2H), 7.53 (t, J = 7.8 Hz, 2H), 7.49 (t, J = 1.6 Hz,
1H), 7.38 (s, 2H), 6.88−6.83 (m, 1H), 6.49 (dt, J = 10.5, 5.1 Hz,
2H), 6.17 (t, J = 5.2 Hz, 1H), 4.80−4.68 (m, 2H), and 4.58 (dd, J
= 12.0, 4.2 Hz, 1H). 13C NMR (101 MHz, DMSO-d6) δ 165.34,
164.42, 156.11, 152.62, 148.76, 140.25, 137.30, 134.09, 133.49,
130.99, 129.62, 129.32, 129.24, 128.89, 128.66, 128.29, 119.32,
117.98, 85.18, 80.13, 79.10, 69.71, and 62.91. HRMS (ESI+)
calculated for C27H24N7O8S+ ([M + H]+) 606.14016, found
606.13784.
Preparation of (2R,3S,4R,5R)-5-(6-amino-9H-purin-9-yl)-2-
((benzoyloxy)methyl)-4-hydroxytetrahydrofuran-3-yl ben-
zoate (17). To a reactor were charged adenosine (0.65 kg, 2.4
mol), pyridine (5.9 L, 9.1 L/kg), and DMF (3.3 L, 5.1 L/kg).
The resulting slurry was cooled to 9 °C, and then a solution of
benzoyl chloride (0.55 kg, 3.9 mol, 1.6 equiv) in pyridine (0.65
L, 1.0 L/kg) was charged over 3 h. The resulting solution was
stirred for 1 h; then, additional benzoyl chloride (0.14 kg, 0.98
mol, 0.40 equiv) was charged over 25 min. The reaction mixture
was stirred for an additional 1 h. Upon reaction completion
(determined by LC method B), the reaction mixture was
distilled to a final volume of 3 L (4 L/kg) under vacuum at 75
°C. The reaction mixture was then cooled to 20 °C and
dichloromethane (11 L, 17 L/kg) was charged. The reaction
mixture was washed with 8 wt % aqueous sodium bicarbonate
(6.5 L, 10 L/kg). The layers were separated, and then the
organic layer was washed three times with 0.5 M aqueous citric
acid (13 L, 20 L/kg; then 2 × 6.5 L, 2 × 10 L/kg). The organic
layer was then washed with 8 wt % aqueous sodium bicarbonate
(6.5 L, 10 L/kg) followed by water (6.5 L, 10 L/kg). The organic
layer was distilled to a final volume of 3 L (4 L/kg) under
vacuum at 40 °C. Acetonitrile (3.3 L, 5.1 L/kg) was charged to
the reaction mixture, and the resulting solution was distilled to a
final volume of 4 L (6 L/kg) under vacuum at 50 °C; this
operation was repeated one additional time. The reaction
mixture was then heated to reflux (85 °C) and stirred for 1 h
giving a slurry. The slurry was cooled to 20 °C, and then DMAP
(0.15 kg, 1.2 mol, 0.50 equiv) was charged. The slurry was stirred
for 5 h, and then it was cooled to −5 °C over 1 h, stirred
overnight, and filtered. The wet cake was washed with cold (−5
°C) acetonitrile (1.3 L, 2.0 L/kg). The wet cake was dried under
vacuum at 50 °C to afford 17 (0.60 kg, 97.4% AN, LC method
B) as a white solid in 52% yield. 1H NMR (400 MHz, DMSO-
d6) δ 8.37 (s, 1H), 8.12−8.06 (m, 3H), 8.01 (dd, J = 8.3, 1.4 Hz,
2H), 7.75−7.64 (m, 2H), 7.55 (dt, J = 21.4, 7.8 Hz, 4H), 7.34 (s,
2H), 6.07 (dd, J = 11.0, 6.1 Hz, 2H), 5.75 (dd, J = 5.6, 3.5 Hz,
1H), 5.29 (q, J = 6.0 Hz, 1H), and 4.75−4.57 (m, 3H). 13C NMR
(101 MHz, DMSO-d6) δ 165.49, 165.02, 156.14, 152.69,
149.38, 140.09, 133.57, 133.48, 129.53, 129.40, 129.35, 129.29,
128.73, 128.70, 119.32, 87.90, 79.24, 73.19, 70.89, and 63.95.
Preparation of (2R,3R,4R,5R)-4-(((1H-imidazol-1-yl)-
sulfonyl)oxy)-5-(6-benzamido-9H-purin-9-yl)-2-
((benzoyloxy)methyl)tetrahydrofuran-3-yl benzoate (24). To
a reactor were charged 22 (0.67 kg, 1.1 mol), dichloromethane
(6.7 L, 10 L/kg), and N-methylimidazole (0.41 L, 5.1 mol, 4.6
equiv). The resulting slurry was cooled to 0 °C, and then benzoyl
chloride (0.39 L, 3.3 mol, 3.0 equiv) was charged over 1 h. The
slurry was warmed to 10 °C and stirred for 3 h, becoming a
solution. Upon reaction completion (determined by LC method
D at 274 nm), 8 wt % aqueous sodium bicarbonate (6.7 L, 10 L/
kg) was charged over 1 h. The mixture was warmed to 20 °C, and
then the layers were separated. The aqueous layer was filtered
through a pad of diatomaceous earth (0.14 kg, 0.21 kg/kg),
rinsing forward with dichloromethane (0.30 L, 0.45 L/kg) to
recover an additional organic layer. The combined organic layers
were washed with water (6.7 L, 10 L/kg) and the layers were
separated. The aqueous layer was filtered through a pad of
diatomaceous earth (0.14 kg, 0.21 kg/kg), rinsing forward with
dichloromethane (0.30 L, 0.45 L/kg) to recover an additional
organic layer. Sodium sulfate (0.67 kg, 1.0 kg/kg) was charged to
the combined organic layers, and the resulting slurry was stirred
for 30 min. The slurry was filtered, rinsing forward with
dichloromethane (0.70 L, 1.0 L/kg). The filtrate was distilled to
a final volume of 1 L (2 L/kg) under vacuum at 40 °C.
Acetonitrile (3.4 L, 5.1 L/kg) was charged, and the resulting
solution was distilled to a final volume of 1 L (2 L/kg) under
vacuum at 50 °C; this operation was repeated one additional
time. The mixture was cooled to 25 °C, and then water (2.7 L,
4.0 L/kg) was charged over 1 h, giving a slurry. The slurry was
stirred for 2 h and filtered. The wet cake was washed with 1:2
acetonitrile/water (1.3 L, 2.0 L/kg) and then water (1.3 L, 2.0
L/kg). The wet cake was dried under vacuum at 50 °C to afford
24 (0.70 kg, 81.9% AN, LC method D at 274 nm) as a light
HRMS (ESI+) calculated for C24H22N5O6 ([M + H]+)
+
476.15646, found 476.15445.
Preparation of (2R,3R,4R,5R)-4-(((1H-imidazol-1-yl)-
sulfonyl)oxy)-5-(6-amino-9H-purin-9-yl)-2-((benzoyloxy)-
methyl)tetrahydrofuran-3-yl benzoate (22). To a reactor were
charged 17 (0.58 kg, 1.2 mol), DMF (0.58 L, 1.0 L/kg), pyridine
(0.48 kg, 6.1 mol, 5.0 equiv), and 2-MeTHF (3.2 L, 5.5 L/kg).
The resulting solution was cooled to −10 °C; then, sulfuryl
chloride (0.25 kg, 1.8 mol, 1.5 equiv) was charged over 1 h. The
reaction mixture was stirred for an additional 3 h, then a solution
of imidazole (0.50 kg, 7.3 mol, 6.0 equiv) in 2-MeTHF (2.6 L,
4.5 L/kg) was charged over 2 h. The solution was warmed to 0
°C and stirred for 14 h. Upon reaction completion (determined
by LC method C), 5 wt % aqueous citric acid (2.9 L, 5.0 L/kg)
was charged over 30 min. The mixture was warmed to 22 °C, and
then the layers were separated. The organic layer was washed
with a mixture of water (2.9 L, 5.0 L/kg) and saturated aqueous
sodium chloride (0.29 L, 0.50 L/kg). The layers were separated,
1
yellow solid in 89% yield. H NMR (400 MHz, DMSO-d6) δ
11.25 (s, 1H), 8.59 (s, 1H), 8.29 (s, 1H), 8.14 (t, J = 1.1 Hz, 1H),
8.10−7.97 (m, 6H), 7.79−7.52 (m, 9H), 7.48 (t, J = 1.5 Hz,
1H), 6.84 (dd, J = 1.8, 0.8 Hz, 1H), 6.66 (d, J = 5.0 Hz, 1H), 6.48
(t, J = 5.4 Hz, 1H), 6.22 (t, J = 5.3 Hz, 1H), 4.87−4.74 (m, 2H),
1271
https://doi.org/10.1021/acs.oprd.1c00026
Org. Process Res. Dev. 2021, 25, 1263−1274

Organic Process Research & Development
pubs.acs.org/OPRD
Article
and 4.62 (dd, J = 12.2, 4.3 Hz, 1H). 13C NMR (101 MHz,
DMSO-d6) δ 170.30, 165.59, 165.33, 164.41, 151.53, 151.45,
150.66, 143.75, 137.38, 134.13, 133.54, 133.22, 132.51, 131.13,
129.65, 129.30, 129.27, 128.87, 128.73, 128.51, 128.46, 128.27,
125.97, 117.83, 85.29, 80.09, 79.31, 69.74, 62.87, 59.72, 20.73,
and 14.06. HRMS (ESI+) calculated for C34H28N7O9S+ ([M +
H]+) 710.16637, found 710.16327.
(10 L, 8.0 L/kg) was charged. The solution was washed three
times with saturated aqueous sodium bicarbonate (7.2 kg, 5.7
kg/kg, each). The organic layer was combined with 0.1 M
aqueous sodium phosphate buffer (pH 7, 20 kg, 16 kg/kg). The
resulting mixture was adjusted to 35 °C and to a pH of 7.5 with
50 wt % aqueous sodium hydroxide (1.7 kg, 1.4 kg/kg). Lipolase
100T (63.0 g, 0.05 kg/kg) was charged, and the reaction mixture
was stirred for 25 h. Upon reaction completion (determined by
LC method E), the reaction mixture was filtered through a pad
of diatomaceous earth (0.64 kg, 0.51 kg/kg), rinsing forward
twice with MIBK (1.3 L, 1.0 L/kg, each). The filtrate was
returned to the reactor, and the layers were separated. The
organic layer was washed sequentially with saturated aqueous
sodium bicarbonate (8.4 kg, 6.7 kg/kg) and 0.1 M aqueous
sodium phosphate buffer (pH 7, 7.5 kg, 6.0 kg/kg). The organic
layer was filtered through a 1.2 μm filter, rinsing forward with
MIBK (0.62 L, 0.50 L/kg). The filtrate was distilled to a final
volume of 5 L (4 L/kg) under vacuum at 50 °C. The solution
was adjusted to 45 °C, and heptane (2.6 L, 2.0 L/kg) was
charged over 1 h. The mixture was seeded with 29 (6.2 g, 0.005
kg/kg), and then additional heptane (7.6 L, 6.0 L/kg) was
charged over 3 h to give a slurry. The slurry was cooled to 20 °C
over 5 h, stirred overnight, and filtered. The wet cake was washed
with 1:4 MIBK/heptane (2.5 L, 2.0 L/kg). The wet cake was
dried under vacuum at 40 °C to afford 29 (1.7 kg, 98.8% AN, LC
method E) as a white solid in 80% yield. 1H NMR (400 MHz,
DMSO-d6) δ 10.72 (s, 1H), 8.68 (s, 1H), 8.59 (d, J = 2.4 Hz,
1H), 6.56 (dd, J = 17.1, 4.1 Hz, 1H), 5.60−5.55 (m, 1H), 5.49
(ddd, J = 24.1, 4.6, 2.6 Hz, 1H), 5.19 (t, J = 5.9 Hz, 1H), 4.11 (q,
J = 4.6 Hz, 1H), 3.71 (ddt, J = 23.6, 11.9, 6.5 Hz, 2H), 2.56 (t, J =
7.4 Hz, 2H), 2.42 (t, J = 7.4 Hz, 2H), 1.67−1.52 (m, 4H), 1.37−
1.23 (m, 8H), and 0.93−0.82 (m, 6H). 13C NMR (101 MHz,
DMSO-d6) δ 172.14, 171.59, 151.98, 151.44, 149.73, 142.70 (d,
J = 4.8 Hz), 122.96, 93.15 (d, J = 192.3 Hz), 82.08 (d, J = 16.9
Hz), 81.49 (d, J = 3.4 Hz), 74.86 (d, J = 27.5 Hz), 60.37, 36.08,
33.16, 30.78, 30.56, 24.37, 23.87, 21.88, 21.75, 13.83, and 13.77.
19F NMR (376 MHz, DMSO-d6) δ −197.26 to −197.53 (m).
Preparation of (2R,3R,4S,5R)-5-(6-amino-9H-purin-9-yl)-4-
fluoro-2-(hydroxymethyl)tetrahydrofuran-3-ol (12) from 24.
Compound 24 (0.52 kg, 0.73 mol) was divided equally among
four round-bottom flasks (0.13 kg, 0.18 mol, each). To each flask
was charged ethyl acetate (1.3 L, 10 L/kg, each), toluene (2.0 L,
15 L/kg, each), dicyclohexylmethylamine (79 mL, 0.37 mol, 2.0
equiv, each), and triethylamine trihydrofluoride (0.15 kg, 0.92
mmol, 5.0 equiv, each). The reaction mixtures were heated to 60
°C and stirred for 26 h each. Upon reaction completion
(determined by LC method D at 260 nm), the mixtures were
cooled to ambient temperature. To each flask, 8 wt % aqueous
sodium bicarbonate (1.3 L, 10 L/kg, each) was charged. The
layers were separated, and the organic layers were combined in a
reactor. The organic layer was washed sequentially with 1 M
aqueous hydrochloric acid (5.2 L, 10 L/kg), 8 wt % aqueous
sodium bicarbonate (5.2 L, 10 L/kg), and water (5.2 L, 10 L/
kg). The organic layer was distilled to a final volume of 0.5 L (1
L/kg) under vacuum at 50 °C. THF (2.6 L, 5.0 L/kg) was
charged, and the solution was distilled to a final volume of 0.5 L
(1 L/kg) under vacuum at 50 °C. THF (2.6 L, 5.0 L/kg) was
charged, and the solution was distilled to a final volume of 2 L (4
L/kg) under vacuum at 50 °C. The mixture was cooled to 20 °C,
and then methanol (1.0 L, 2.0 L/kg) was charged. A solution of
lithium hydroxide monohydrate (0.054 kg, 1.3 mol, 1.8 equiv) in
water (1.1 kg, 2.1 L/kg) was charged while not exceeding an
internal temperature of 25 °C. The reaction mixture was then
stirred at 20 °C for 16 h. Upon reaction completion (determined
by LC method A), 6 M aqueous hydrochloric acid (0.46 kg, 0.88
kg/kg) and toluene (1.0 L, 2.0 L/kg) were charged. The layers
were separated, and the aqueous layer was adjusted to a pH of
6−8 with 5 M aqueous sodium hydroxide (0.30 L, 0.57 L/kg).
The aqueous solution was distilled to a final volume of 2 L (4 L/
kg) under vacuum at 50 °C, and a slurry formed during the
distillation. The slurry was cooled to 20 °C over 1 h, stirred for
an additional 1 h, and then filtered. The wet cake was washed
with water (0.41 L, 0.79 L/kg). The wet cake was dried under
vacuum at 50 °C to afford 12 (0.10 kg, 98.6% AN, LC method
A) as a light tan solid in 52% yield. 1H NMR (400 MHz, DMSO-
d6) δ 8.24 (d, J = 2.1 Hz, 1H), 8.16 (s, 1H), 7.34 (s, 2H), 6.41
(dd, J = 14.6, 4.6 Hz, 1H), 5.95 (d, J = 5.1 Hz, 1H), 5.27 (t, J =
4.2 Hz, 0.5H), 5.15−5.09 (m, 1.5H), 4.45 (dtd, J = 18.9, 5.2, 3.8
Hz, 1H), 3.85 (q, J = 4.9 Hz, 1H), and 3.76−3.59 (m, 2H). 13C
NMR (101 MHz, DMSO-d6) δ 156.02, 152.79, 149.16, 139.44
(d, J = 4.0 Hz), 118.23, 95.41 (d, J = 192.0 Hz), 83.48 (d, J = 5.4
Hz), 81.43 (d, J = 17.0 Hz), 72.70 (d, J = 23.2 Hz), and 60.43.
19F NMR (376 MHz, DMSO-d6) δ −197.66 to −197.93 (m).
HRMS (ESI+) calculated for C22H33FN5O5 ([M + H]+)
+
466.24602, found 466.24390.
Preparation of (2S,3R,4S,5R)-5-(6-amino-9H-purin-9-yl)-4-
fluoro-3-hydroxytetrahydrofuran-2-carboxylic acid (1). To a
reactor were charged 29 (65 g, 140 mmol), sodium phosphate
dibasic heptahydrate (220 g, 810 mmol, 5.8 equiv), acetonitrile
(330 mL, 5.0 L/kg), and water (330 mL, 5.0 L/kg). To the
resulting slurry, iodobenzene diacetate (120 g, 380 mmol, 2.7
equiv) was charged followed by TEMPO (3.3 g, 21 mmol, 0.15
equiv), resulting in an exotherm from 15 to 29 °C over 10 min
and giving a biphasic solution. The reaction mixture was stirred
at 22 °C for an additional 40 min. Upon reaction completion
(determined by LC method E), a solution of sodium sulfite (13
g, 100 mmol, 0.74 equiv) in water (120 mL, 1.8 L/kg) was
charged over 5 min. The mixture was stirred for 15 min at 22 °C
and then tested for residual oxidant (potassium iodide starch test
paper). 2-MeTHF (260 mL, 4.0 L/kg) was charged and the
layers were separated. The aqueous layer was extracted with 2-
MeTHF (130 mL, 2.0 L/kg). The combined organic layers were
returned to the reactor, rinsing forward with 2-MeTHF (33 mL,
0.51 L/kg), and washed twice with 15 wt % aqueous sodium
chloride solution (330 mL, 5.0 L/kg, each). The organic layer
was distilled to a final volume of 450 mL (6 L/kg) under vacuum
at 50 °C. Methanol (420 mL, 6.4 L/kg) was charged and the
resulting solution was distilled to a final volume of 330 mL (5 L/
kg) under vacuum at 55 °C. The solution was cooled to 22 °C,
HRMS (ESI+) calculated for C10H13FN5O3 ([M + H]+)
+
270.09969, found 270.09867.
Preparation of (2R,3R,4S,5R)-4-fluoro-5-(6-hexanamido-
9H-purin-9-yl)-2-(hydroxymethyl)tetrahydrofuran-3-yl hexa-
noate (29). To a reactor were charged 12 (1.3 kg, 4.7 mol),
acetonitrile (3.8 L, 3.0 L/kg), pyridine (1.5 kg, 19 mol, 4.1
equiv), and hexanoic anhydride (3.7 kg, 17 mol, 3.7 equiv). The
resulting slurry was heated to 70 °C and stirred for 21 h, to give a
solution. Upon reaction completion (determined by LC method
E), the reaction mixture was cooled to 35 °C, and then MIBK
1272
https://doi.org/10.1021/acs.oprd.1c00026
Org. Process Res. Dev. 2021, 25, 1263−1274

Organic Process Research & Development
pubs.acs.org/OPRD
Article
and then methanol (81 mL, 1.2 L/kg) and 25 wt % sodium
methoxide in methanol (69 mL, 300 mmol, 2.2 equiv) were
charged. The reaction mixture was stirred at 22 °C for 5 h. Upon
reaction completion (determined by LC method A), a solution
of sodium chloride (16 g, 0.25 kg/kg) in water (310 mL, 4.8 L/
kg) was charged followed by tert-butyl methyl ether (440 mL,
6.8 L/kg). The resulting mixture was heated to 42 °C, and the
pH was adjusted to 3 with 6 M aqueous hydrochloric acid
solution to give a slurry. The slurry was cooled to 20 °C over 1 h,
stirred overnight, and filtered. The wet cake was washed with
MTBE (88 mL, 1.4 L/kg), twice with water (65 mL, 1.0 L/kg,
each), and finally with THF (74 mL, 1.1 L/kg). The wet cake
was dried under vacuum at 50 °C to afford 1 (38 g, 99.4% AN,
Sura Ha − Pharmaceutical Process R&D Team, Research
Institute, Yuhan Corporation, Yongin-si 17084, Gyeonggi-do,
South Korea
Olga Lapina − Department of Process Chemistry, Gilead
Sciences, Inc., Foster City, California 94404, United States
Jisung Lee − Pharmaceutical Process R&D Team, Research
Institute, Yuhan Corporation, Yongin-si 17084, Gyeonggi-do,
South Korea
Lennie Lin − Department of Process Development, Gilead
Alberta ULC, Edmonton, Alberta T6S 1A1, Canada
Sangsun Park − Pharmaceutical Process R&D Team, Research
Institute, Yuhan Corporation, Yongin-si 17084, Gyeonggi-do,
South Korea
Mary Rosario − Department of Process Development, Gilead
Alberta ULC, Edmonton, Alberta T6S 1A1, Canada
Olivier St-Jean − Department of Process Chemistry, Gilead
Sciences, Inc., Foster City, California 94404, United States
Guojun Yu − Department of Process Development, Gilead
Alberta ULC, Edmonton, Alberta T6S 1A1, Canada
1
LC method A) as a white solid in 96% yield. H NMR (400
MHz, DMSO-d6) δ 13.83 (br, exchangeable), 8.34 (d, J = 2.6
Hz, 1H), 8.18 (s, 1H), 7.44 (br, 2H), 6.58 (dd, J = 23.7, 2.9 Hz,
1H), 6.48 (br, 1H), 5.12 (ddd, J = 50.3, 3.0, 1.5 Hz, 1H), 4.71−
4.65 (m, 1H), 4.61 (d, J = 1.2 Hz, 1H). 13C NMR (101 MHz,
DMSO-d6) δ 170.90, 156.11, 152.60, 148.72, 139.90 (d, J = 4.4
Hz), 118.27, 93.01 (d, J = 189.6 Hz), 84.97 (d, J = 15.3 Hz),
83.01, and 76.62 (d, J = 26.0 Hz). 19F NMR (376 MHz, DMSO-
d6) δ −197.61 (dddd, J = 50.3, 23.6, 10.1, 2.7 Hz). HRMS (ESI
+) calculated for C10H11FN5O4+ ([M + H]+) 284.07896, found
284.07815.
Complete contact information is available at:
https://pubs.acs.org/10.1021/acs.oprd.1c00026
Author Contributions
The manuscript was written through the contributions of all
authors. All authors have given approval to the final version of
the manuscript.
ASSOCIATED CONTENT
■
sı
* Supporting Information
Notes
The Supporting Information is available free of charge at
https://pubs.acs.org/doi/10.1021/acs.oprd.1c00026.
The authors declare the following competing financial
interest(s): The authors are employees of Gilead Sciences,
Inc., Foster City, California, United States; Gilead Alberta ULC,
Edmonton, Alberta, Canada; and Yuhan Corporation, Yongin,
Gyeonggi, South Korea.
Reaction calorimetry and representative experimental
procedure for the conversion of 10 to 1 (PDF)
Crystallographic data of 12 (CIF)
Crystallographic data of 29 (CIF)
ACKNOWLEDGMENTS
■
The authors would like to thank Li Wang for experimental
contributions, Amy Cagulada, Nisha Shah, and Lu Zou for
performing reaction safety assessments, Sunghee Cho, Teague
McGinitie, Sebin Park, and Alban Pereira for analytical support,
and Andrea Ambrosi and Dietrich Steinhuebel for helpful
feedback on the manuscript.
AUTHOR INFORMATION
■
Corresponding Author
David A. Siler − Department of Process Chemistry, Gilead
Sciences, Inc., Foster City, California 94404, United States;
orcid.org/0000-0002-2642-4392; Email: david.siler@
gilead.com
REFERENCES
■
(1) (a) Seley-Radtke, K. L.; Yates, M. K. The evolution of nucleoside
analogue antivirals: A review for chemists and non-chemists. Part 1:
Early structural modifications to the nucleoside scaffold. Antivir. Res.
2018, 154, 66−86. (b) Yates, M. K.; Seley-Radtke, K. L. The evolution
of antiviral nucleoside analogues: A review for chemists and non-
chemists. Part II: Complex modifications to the nucleoside scaffold.
Antivir. Res. 2019, 162, 5−21.
Authors
Selcuk Calimsiz − Department of Process Development, Gilead
Alberta ULC, Edmonton, Alberta T6S 1A1, Canada
Ian J. Doxsee − Department of Process Chemistry, Gilead
Sciences, Inc., Foster City, California 94404, United States
Bernard Kwong − Department of Process Development, Gilead
Alberta ULC, Edmonton, Alberta T6S 1A1, Canada
Jeffrey D. Ng − Department of Process Chemistry, Gilead
Sciences, Inc., Foster City, California 94404, United States
Keshab Sarma − Department of Process Chemistry, Gilead
Sciences, Inc., Foster City, California 94404, United States
Jinyu Shen − Department of Process Development, Gilead
Alberta ULC, Edmonton, Alberta T6S 1A1, Canada
Jonah W. Curl − Department of Process Development, Gilead
Alberta ULC, Edmonton, Alberta T6S 1A1, Canada
Jason A. Davy − Department of Process Development, Gilead
Alberta ULC, Edmonton, Alberta T6S 1A1, Canada
Jeffrey A. O. Garber − Department of Process Development,
Gilead Alberta ULC, Edmonton, Alberta T6S 1A1, Canada
(2) Zhou, Y.; Wang, J.; Gu, Z.; Wang, S.; Zhu, W.; Acena, J. L.;
̃
Soloshonok, V. A.; Izawa, K.; Liu, H. Next Generation of Fluorine-
Containing Pharmaceuticals, Compounds Currently in Phase II-III
Clinical Trials of Major Pharmaceutical Companies: New Structural
Trends and Therapeutic Areas. Chem. Rev. 2016, 116, 422−518.
(3) (a) Cavaliere, A.; Probst, K. C.; Westwell, A. D.; Slusarczyk, M.
Fluorinated nucelosides as an important class of anticancer and antiviral
agents. Future Med. Chem. 2017, 9, 1809−1833. (b) Liu, P.; Sharon, A.;
Chu, C. K. Fluorinated nucleosides: Synthesis and biological
implication. J. Fluorine Chem. 2008, 129, 743−766.
(4) (a) Boojamra, C. G.; Mackman, R. L.; Markevitch, D. Y.; Prasad,
V.; Ray, A. S.; Douglas, J.; Grant, D.; Kim, C. U.; Cihlar, T. Synthesis
and anti-HIV activity of GS-9148 (2′-Fd4AP), a novel nucleoside
phosphonate HIV reverse transcriptase inhibitor. Bioorg. Med. Chem.
Lett. 2008, 18, 1120−1123. (b) Ray, A. S.; Vela, J. E.; Boojamra, C. G.;
1273
https://doi.org/10.1021/acs.oprd.1c00026
Org. Process Res. Dev. 2021, 25, 1263−1274

Organic Process Research & Development
pubs.acs.org/OPRD
Article
Zhang, L.; Hui, H.; Callebaut, C.; Stray, K.; Lin, K.-Y.; Gao, Y.;
Mackman, R. L.; Cihlar, T. Intracellular Metabolism of the Nucleotide
Prodrug GS-9131, a Potent Anti-Human Immunodeficiency Virus
Agent. Antimicrob. Agents Chemother. 2008, 52, 648−654. (c) Boojamra,
C. G.; Lin, K.-Y.; Mackman, R.; Markevitch, D. Y.; Petrakovsky, O. V.;
Ray, A. S.; Zhang, L. Phosphonate Analogs of HIV Inhibitor
Compounds. U.S. Patent US7871991 B22011.
(5) (a) Standley, E. A.; Bringley, D. A.; Calimsiz, S.; Ng, J. D.; et al.
Synthesis of Rovafovir Etalafenamide (Part I): Active Pharmaceutical
Ingredient Process Development, Scale-Up, and Impurity Control
Strategy Org. Process Res. Dev. 2021, accepted for publication.
DOI: 10.1021/acs.oprd.1c00059. (b) Ambrosi, A.; Bringley, D. A.;
Calimsiz, S.; Garber, J. A. O.; et al. Synthesis of Rovafovir Etalafenamide
(Part III): Evolution of the Synthetic Process to the Phosphonamidate
Fragment Org. Process Res. Dev. 2021, DOI: 10.1021/ac-
s.oprd.0c00428.
(6) Yu, R. H. C.; Brown, B. H.; Polniaszek, R. P.; Graetz, B. R.; Sujino,
K.; Tran, D. D.-P.; Triman, A. S.; Kent, K. M.; Pfeiffer, S. Processes and
Intermediates for Preparing Anti-HIV Agents. WIPO Patent
WO2012US38615.
(7) (a) Tann, C. H.; Brodfuehrer, P. R.; Brundidge, S. P.; Sapino, C.,
Jr.; Howell, H. G. Fluorocarbohydrates in Synthesis. An Efficient
Synthesis of 1-(2-Deoxy-2-fluoro-β-D-arabinofuranosyl)-5-iodouracil
(β-FIAU) and 1-(2-Deoxy-2-fluoro-β-D-arabinofuranosyl)thymine (β-
FMAU). J. Org. Chem. 1985, 50, 3644−3647. (b) Chou, T. S.; Becke, L.
M.; O’Toole, J. C.; Carr, M. A.; Parker, B. E. Triethylamine
poly(hydrogen fluorides) in the synthesis of a fluorinated nucleoside
glycon. Tetrahedron Lett. 1996, 37, 17−20.
2009, 11, 855−862. (c) Jones, L. A.; Moorman, A. R.; Chamberlain, S.
D.; de Miranda, P.; Reynolds, D. J.; Burns, C. L.; Krenitsky, T. A. Di-
and Triester Prodrugs of the Varicella-Zoster Antiviral Agent 6-
Methoxypurine Arabinoside. J. Med. Chem. 1992, 35, 56−63.
(13) Although 10 volumes of pyridine are still required for this
reaction, a majority can be recovered via distillation and recycled.
(14) Yoshimura, Y.; Endo, M.; Miura, S.; Sakata, S. An Alternative
Synthesis of the Antineoplastic Nucleoside 4′-ThioFAC and Its
Application to the Synthesis of 4′-ThioFAG and 4′-Thiocytarazid. J.
Org. Chem. 1999, 64, 7912−7920.
(15) (a) García, J.; Fernández, S.; Ferrero, M.; Sanghvi, Y. S.; Gotor, V.
Building Blocks for the Solution Phase Synthesis of Oligonucleotides:
Regioselective Hydrolysis of 3′,5′-Di-O-levulinylnucleosides Using an
Enzymatic Approach. J. Org. Chem. 2002, 67, 4513−4519. (b) Taverna-
Porro, M.; Bouvier, L. A.; Pereira, C. A.; Montserrat, J. M.; Iribarren, A.
M. Chemoenzymatic Preparation of Nucleosides from Furanoses.
Tetrahedron Lett. 2008, 49, 2642−2645. (c) Martínez-Montero, S.;
Fernández, S.; Sanghvi, Y. S.; Gotor, V.; Ferrero, M. An Expedient
Biocatalytic Procedure for Abasic Site Precursors Useful in
Oligonucleotide Synthesis. Org. Biomol. Chem. 2011, 9, 5960−5966.
(16) Prepared from sodium phosphate monobasic and sodium
phosphate dibasic.
(8) Jimenez-Gonzalez, C.; Ponder, C. S.; Broxterman, Q. B.; Manley, J.
B. Using the Right Green Yardstick: Why Process Mass Intensity is
Used in the Pharmaceutical Industry to Drive More Sustainable
Processes. Org. Process Res. Dev. 2011, 15, 912−917.
(9) (a) Bauta, W. E.; Schulmeier, B. E.; Burke, B.; Puente, J. F.;
Cantrell, W. R., Jr.; Lovett, D.; Goebel, J.; Anderson, B.; Ionescu, D.;
Guo, R. A New Process for Antineoplastic Agent Clofarabine. Org.
Process Res. Dev. 2004, 8, 889−896. (b) Tann, C. H.; Brodfuehrer, P. R.;
Brundidge, S. P.; Sapino, C., Jr.; Howell, H. G. Fluorocarbohydrates in
Synthesis. An Efficient Synthesis of 1-(2-Deoxy-2-fluoro-β-D-arabino-
furanosyl)-5-iodouracil (β-FIAU) and 1-(2-Deoxy-2-fluoro-β-D-
arabinofuranosyl)thymine (β-FMAU). J. Org. Chem. 1985, 50, 3644−
3647. It is also possible that the increasing solvent polarity increases the
solubility of bromide which could racemize the anomeric position prior
to displacement; for example: (c) Kotick, M. P.; Szantay, C.; Bardos, T.
J. Synthesis of 5-S-Substituted 2′-Deoxyuridines. Study of the Factors
Influencing the Stereoselectivity of the Silyl Modification of the Hilbert-
Johnson Reaction. J. Org. Chem. 1969, 34, 3806−3813. For an
alternative strategy for glycosylation with poorly soluble bases, see:
(d) Gallou, F.; Seeger-Weibel, M.; Chassagne, P. Development of a
Robust and Sustainable Process for Nucleoside Formation. Org. Process
Res. Dev. 2013, 17, 390−396.
(10) The diastereomeric ratio typically observed in vial-scale
screening experiments was slightly lower than the ratio observed on
scale.
(11) (a) Kumar, P.; Emami, S.; Kresolek, Z.; Yang, J.; McEwan, A.;
Wiebe, L. Synthesis and Hypoxia Selective Radiosensitization Potential
of β-2-FAZA and β-3-FAZL: Fluorinated Azomycin β-Nucleosides.
Med. Chem.. 2009, 5, 118−129. (b) Kirschberg, T. A.; Mish, M.;
Squires, N. H.; Zonte, S.; Aktoudianakis, E.; Metobo, S.; Butler, T.; Ju,
X.; Cho, A.; Ray, A. S.; Kim, C. U. Synthesis of 1′- C -Cyano Pyrimidine
Nucleosides and Characterization as HCV Polymerase Inhibitors.
Nucleosides, Nucleotides and Nucleic Acids 2015, 34, 763−785. (c) Xu,
S.; Yao, P.; Chen, G.; Wang, H. A New Synthesis of 2-Chloro-2′-
Deoxyadenosine (Cladribine), CdA). Nucleosides, Nucleotides and
Nucleic Acids 2011, 30, 353−359.
(12) (a) Sivets, G. G. Regio- and Stereoselective Syntheses of l-
Pentose Derivatives from l-Arabinose. Tetrahedron 2018, 74, 920−931.
(b) Simeó, Y.; Sinisterra, J. V.; Alcántara, A. R. Regioselective
Enzymatic Acylation of Pharmacologically Interesting Nucleosides in
2-Methyltetrahydrofuran, a Greener Substitute for THF. Green Chem.
1274
https://doi.org/10.1021/acs.oprd.1c00026
Org. Process Res. Dev. 2021, 25, 1263−1274