Development of an Efficient Manufacturing Process to GSK2248761A API

Article abstract of DOI:10.1021/acs.oprd.7b00357

Amidation of indole 2-carboxylate 1 with ammonia gas via the imidazolide 2 gave GSK2248761A API 3, which was in development for the treatment of HIV. Three significant impurities, namely the phosphinic acid 4, the N-acyl urea 8, and the indoloyl carboxamide 6, were formed during the reaction, and the original process was unable to produce API within clinical specification when run at scale. Investigation into the origin, fate, and control of these impurities led to a new process which was able to produce API within clinical specification.

Full text of DOI:10.1021/acs.oprd.7b00357

Article
pubs.acs.org/OPRD
Cite This: Org. Process Res. Dev. XXXX, XXX, XXX−XXX
Development of an Efficient Manufacturing Process to GSK2248761A
API
Celine Bret,†,§ Elaine Smith,† Erica Vit,†,§ Jerome F. Hayes,*,† John D. Hayler,† Alan Ironmonger,†
David Pascoe,† Neil Hodnett,† Jonathan Stanway,† Stephen Etridge,† Alistair Stewart,‡,⊥ Jingyang Wang,‡,⊥
Benjamin A. Mayes,‡,⊥ and Adel Moussa‡,⊥
†GlaxoSmithKline, Medicines Research Center, Gunnels Wood Road, Stevenage, Hertfordshire, SG1 2NY, U.K.
‡Merck & Co. Inc., 126 East Lincoln Avenue, Rahway, New Jersey 07065, United States
ABSTRACT: Amidation of indole 2-carboxylate 1 with ammonia gas via the imidazolide 2 gave GSK2248761A API 3, which
was in development for the treatment of HIV. Three significant impurities, namely the phosphinic acid 4, the N-acyl urea 8, and
the indoloyl carboxamide 6, were formed during the reaction, and the original process was unable to produce API within clinical
specification when run at scale. Investigation into the origin, fate, and control of these impurities led to a new process which was
able to produce API within clinical specification.
GSK2248761A (IDX899, Fosdevirine, 3) is a potent non-
nucleoside reverse transcriptase inhibitor that was in develop-
ment as a potential treatment for HIV.1 It was discovered by
Idenix Pharmaceuticals, and a synthesis is described in the
preceding paper. It used an unusual palladium catalyzed
phosphorus−carbon coupling reaction to generate the entire
molecular backbone. The final step of the synthesis involved
the initial formation of an imidazolide intermediate 2 using
CDI, followed by an aminolysis on addition of ammonia gas to
the solution and was ostensibly straightforward (Scheme 1).2
of the indole acid 1 with CDI, variable amounts of the
diketopiperazide 5 were formed (Scheme 2). This intermediate
was prepared independently and was shown to react with
ammonia to give the indoloyl carboxamide 6. Presumably this
occurs via ring opening followed by acyl migration. Impurity 6
was found subsequently to react slowly with further ammonia
to give GSK2248761A (Scheme 2).
Impurity 8 was not detected when the aminolysis was
performed on a small scale (2 g). In contrast, the first pilot
plant batch gave API containing 0.30% a/a of impurity 8 by
HPLC, while the second batch gave 0.22% a/a impurity 8 by
HPLC (Table 1, entries 1 and 2). It was strongly suspected that
the ammonia gas supply rate was influencing impurity 8
formation. Traces of ammonia could promote N-acylation of
indole with CDI. The resultant adduct 9 would then react with
excess ammonia to form the hydantoin intermediate 7 which
would then react with further ammonia to give 8 (Scheme 3).
Alternatively ammonia might react with CDI to form
carboxamide 10 which could react with 23 to give the
hydantoin 7.
Scheme 1. Preparation of GSK2248761A API from the
Indole Acid 1
Addition of the required molar equivalents of ammonia took
approximately 100 min in the pilot plant. The ammonia supply
rate was limited by the constraints of the gas supply
configuration in the plant. In a process stretching experiment,
ammonia gas was added to a solution of the imidazolide 2 in
THF and was then removed after a few seconds. Very high
levels of impurity 8 (35% a/a by HPLC) were formed. In order
to quantify the effect of ammonia gas supply rate on impurity 8
formation, two experiments were performed in an RC1 reactor
on a 15 g scale (Figure 2). Ammonia was supplied via a mass
flow controller and the supply rate was monitored using a mass
flow meter, while the vessel occupancy was maintained at the
same level as the pilot plant reactor. At an ammonia flow rate of
30 mL/min the solution level of impurity 8 was 0.54% a/a, and
at 60 mL/min, 0.35% a/a 8 was obtained.
However, three significant impurities were formed during the
reaction, the phosphinic acid 4, the N-acyl urea 8, and the
indoloyl carboxamide 6 and the original process was unable to
produce API within clinical specification when run on an 8 kg
scale (Figure 1). In order to meet the challenging timelines for
supply of API for clinical studies, it was necessary to scale-up
the original process with minimal opportunity for process
development. An understanding of the origin, control, and fate
of these impurities was critical to enable us to accomplish this.
Origin and Control of Impurities. It was found that the
phosphinic acid 4 was formed by overexposure of the product
GSK2248761A to ammonia once the required reaction was
complete. This was not of major concern, as 4 formed slowly
on the reaction time scale and was readily removed by
extraction with a pH 10 aqueous buffer. During the activation
Received: November 10, 2017
© XXXX American Chemical Society
A
DOI: 10.1021/acs.oprd.7b00357
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 1. Structures of the main impurities.
Scheme 2. Formation of Indoloyl Carboxamide 6 and Its Conversion to GSK2248761A
and as predicted no impurity 8 was formed provided the
equipment was configured to prevent ingress of ammonia gas
into the bulk imidazolide feed solution. The inverse addition
process was then performed on a pilot plant scale (20 kg), and
consistently no impurity 8 was formed (Table 1 batches 3−6).
However, to enable safe operation within the plant vessel
pressure limits while avoiding the risk of starving the reaction of
ammonia, several interruptions to the imidazolide addition
were required in order to repressurize the vessel with ammonia.
For later batches this issue was circumvented by evacuating the
vessel headspace prior to charging ammonia. Given the
variation in operating limits between individual vessels and
production facilities, it was felt that this process was not
amenable to straightforward changes in scale and equipment
and therefore an improved process was sought.
Table 1. Pilot Plant Data for the Conversion of Indole Acid
1 to GSK2248761A via Aminolysis of Imidazolide 2
a
a
b
b
b
b
Batch
1
2
3
4
5
6
c
scale (kg)
8.1
21.0
20.0
26.1
87.5
100
17.2
17.5
yield (%)
82.2
99.24
0.10
0.30
<0.05
<0.05
83.7
89.0
87.1
97.8
d
purity (% area)
4 (% area)
8 (% area)
1 (% area)
6 (% area)
99.67
<0.05
0.22
99.83
ND
99.88
ND
99.88
<0.05
ND
ND
ND
ND
ND
<0.05
ND
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
a
b
Addition of ammonia to the imidazolide. Addition of the
c
d
imidazolide to ammonia. Scale wrt indole acid 1. See ref 4 for
HPLC conditions for API impurity profile.
Process Optimization. Having successfully supplied the
API required for clinical studies we were able to focus on
improvements to this process with respect to long-term
manufacturability. Of specific concern was the use of ammonia
gas and the overall volume inefficiency of the process
(maximum 90 volumes).
Addition of a solution of imidazolide 2 to a mixture of dilute
aqueous ammonia (less than 28% v/v) (8 equiv) in THF gave
an increased amount of hydrolysis (i.e., unreacted indole acid
1) and an increased amount of indoloyl carboxamide 6.
However, using 28% aqueous ammonia (8 equiv) we were able
to isolate GSK2248761A within specification. In addition,
raising the temperature to 40−50 °C gave an increased rate of
conversion of intermediate 6 to 3, whereas with ammonia gas in
The 30 mL/min supply rate was a direct scale down of the
ammonia gas supply rate achieved on plant scale where the
solution level of impurity 8 was 0.54% a/a. The results obtained
strongly support the correlation between the level of impurity 8
formed in the reaction and ammonia supply rate. It was evident
therefore that the ammonia supply rate would need to be
increased significantly to deliver API within clinical specification
(0.15% a/a 8) using the original process.
To circumvent this engineering challenge, we considered
inverse addition, i.e. addition of the imidazolide solution to an
ammonia gas/THF mixture. This would ensure that the
imidazolide was always exposed to a large excess of ammonia.
Several inverse addition experiments were run in a 1 L reactor
maintaining the ammonia pressure between 0.2 and 0.4 barg,
B
DOI: 10.1021/acs.oprd.7b00357
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 3. Proposed Mechanism for the Formation of the N-Acyl Urea Impurity 8
Figure 2. Experimental configuration used to study the aminolysis reaction. Nitrogen was used to purge and inert the vessel. For the inverse addition
experiments a nonreturn valve was installed on the imidazolide feed line to prevent ingress of gaseous ammonia. Process sensors are shown as labeled
circles (PI = pressure indicator, TT = temperature transmitter, ST = stirrer speed transmitter, WIT = weight indicator and transmitter).
THF, an increase in temperature led to an increase in formation
of impurity 8. This could be rationalized on the basis of
decreased solubility of ammonia at elevated temperatures under
anhydrous conditions. Substituting THF with 2-methyltetra-
hydrofuran also significantly improved the process. Due to the
immiscibility of 2-methyltetrahydrofuran with water, it was no
longer necessary to use TBME in the aqueous workup.
Furthermore, it was possible to increase the reaction
concentration from 40 volumes to 15 volumes with no adverse
effect on product purity. At this higher concentration,
imidazolide 2 dimerized to the diketopiperazide 5 in greater
amounts than was previously observed. Subsequent reaction
monitoring demonstrated that under these conditions
diketopiperazide 5 reacted efficiently with concentrated
aqueous ammonia to give GSK2248761A within specification.
These observations enabled us to design a significantly more
C
DOI: 10.1021/acs.oprd.7b00357
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
volume efficient process and avoided the use of gaseous
ammonia entirely.
0.6 m2 agitated filter drier and was obtained as a white
crystalline solid. Yield = 15.6 kg, 87%. HPLC (99.88%);4 1H
NMR (500 MHz, CDCl3) δ ppm 2.37 (s, 3H), 3.86 (d, J = 15.0
Hz, 3H), 5.86 (d, J = 15.0 Hz, 1H), 5.94 (s, 1H), 7.33 (dd, J =
9.0 Hz, J = 2.0 Hz, 1H), 7.34 (d, J = 15.5 Hz, 1H), 7.39 (s, 1H),
7.49 (dd, J = 9.0 Hz, J = 1.5 Hz, 1H) 7.60 (d, J = 13.5 Hz, 1H),
7.64 (d, J = 13.5 Hz, 1H), 7.65 (d, J = 1.5 Hz, 1H), 10.40 (s,
1H), 10.88 (s, 1H); 13C NMR (126 MHz, CDCl3) δ 21.3, 52.1,
98.1, 100.5 (d, J = 152.5 Hz), 113.9, 117.6, 120.9, 126.2, 126.5
(d, J = 11.3 Hz) 128.7, 129.9 (d, J = 10.1 Hz), 131.7, 133.0 (d, J
= 151.2 Hz), 133.2 (d, J = 8.8 Hz), 133.4 (d, J = 10.1 Hz),
134.1 (d, J = 15.1 Hz), 138.7, 139.9, 149.2 and 161.2; 31P NMR
(202 MHz, CDCl3) δ 31.4.
CONCLUSION
■
A pilot plant process which was capable of producing the
potent non-nucleoside reverse transcriptase inhibitor
GSK2248761A within clinical specification was successfully
developed. This was subsequently optimized on a laboratory
scale with an intricate knowledge of impurity origin to give a
significantly more volume efficient process which avoided the
operational challenges associated with gaseous ammonia.
IR ν (cm−1) 3280, 3065, 1679, 1619, 1402, 1195 and 1010.
HRMS calcd for C20H18ClN3O3P: 414.0769; HRMS found
[M + H]+: 414.0760.
EXPERIMENTAL SECTION
■
General Experimental. Reactions were monitored using
HPLC on a Luna C18(2), 50 mm × 2 mm, 3 μm, eluting with
gradient 0% to 95% (0.05%v/v TFA in water to 0.05%v/v TFA
in acetonitrile) over 8 min at 40 °C detecting at 220 nm. NMR
spectra were recorded on a Bruker 400 mHz Ultrashield, or a
Bruker AV500 spectrometer. HRMS data were recorded on an
LQT Orbitrap spectrometer. Infrared spectra were recorded on
a PerkinElmer Spectrum One spectrometer.
Improved Process to GSK2248761A. To a stirred
suspension of N,N-carbonyldiimidazole (7.8 g, 48.2 mmol) in
2-methyltetrahydrofuran (150 mL) was added GSK2326299A
(10.0 g, 24.1 mmol) over 5−10 min at 25 °C. The temperature
was adjusted to 34−37 °C, and the solution was stirred for 2 h.
The resultant yellow solution was then cooled to 25 °C, and it
was added to a separate, pressure-rated vessel containing a 28−
30% aqueous solution of ammonium hydroxide (12.9 mL,
193.3 mmol) at 3−10 °C over 15 min. The resultant mixture
was warmed to 25 °C over 15 min and was then warmed to
40−42 °C. The mixture was stirred for 2 h after which time the
solution was cooled to 5 °C and water (40 mL) was added.
Concentrated hydrochloric acid (40 mL) was added slowly at
5−20 °C to adjust to pH 1.0−1.5. The organic phase was
separated and was washed with brine (50 mL of a 1% w/w
aqueous solution). The organic phase was then washed with a
pH 10 sodium bicarbonate buffer containing 1% w/w sodium
chloride (2 × 50 mL). The organic phase was diluted with 2-
methyltetrahydrofuran (60 mL), and solvent (120 mL) was
distilled off. The resultant azeotropically dried solution was
allowed to cool to room temperature and was filtered.
Methanol (60 mL) was added, and solvent (60 mL) was
distilled off. This was repeated twice to give a final volume of 60
mL. The solution was cooled to rt after which point the
product crystallized. Water (26 mL) was then added, and the
resultant slurry was stirred for 30 min. The product was filtered,
washed with MeOH/water 2:1, and dried under a stream of
nitrogen at room temperature. Yield = 8.0 g (81%), HPLC
99.85%.
Pilot Plant Preparation of GSK2248761A 1 (Batch 5).
To a solution of N,N-carbonyldiimidazole (13.34 kg, 82.27
mol) in THF (229 kg) was added GSK2326299A (17.2 kg, 41.5
mol) as a solution in THF (153 kg) at 19−25 °C over 30 min.
Further THF (38 kg) was then added. The resultant mixture
was warmed to 38 °C over 30 min and was maintained at this
temperature for 2 h. The resultant clear yellow solution was
then cooled to 5−15 °C over 1 h. The reactor was then
evacuated 4 times and purged with nitrogen to remove carbon
dioxide.
A second reactor was charged with THF (153 kg), and the
solvent was cooled to 0−10 °C. The reactor was then evacuated
to −0.85 barg and then was charged with ammonia gas (4.9 kg,
287.7 mol) over 100 min. The imidazolide solution in THF was
then transferred to the reactor containing ammonia in THF
over 30 min at 0−10 °C. A differential pressure of 0.5 barg was
maintained between the two reactors to prevent ammonia
transfer to the bulk imidazolide solution. A 38 kg line wash of
THF was then transferred from the first reactor to the
aminolysis reactor. The reaction mixture was then warmed to
17−27 °C over 1 h. The reaction mixture was stirred at this
temperature for 2.5 h after which time further ammonia (0.7 kg,
41.1 mol) was added. The reaction mixture was stirred at 17−
27 °C for a further 17 h after which time the level of imide 6 fell
below the target of 0.2a/a %. The reactor was then purged with
nitrogen, and the reaction mixture was cooled to 10 °C.
Hydrochloric acid (171 kg of a 2 M aqueous solution) was then
added at 10−15 °C over 80 min. tert-Butylmethyl ether (508
kg) was then added, and the mixture was stirred at 15−25 °C
for 25 min. The organic layer was separated and was washed
with water (172 kg), aqueous pH 10 buffer solution (172 kg
containing 0.96 kg of sodium hydroxide and 4.7 kg of sodium
bicarbonate in 166 kg water), and water (172 kg). The organic
layer was then concentrated by vacuum distillation to a total
weight of 72 kg. Methanol (97 kg) was then added, and a
further 103 L of solvent were removed by vacuum distillation.
The put and take distillation was repeated with two further 81
kg batches of methanol. Water (13 kg) was then added over 30
min, and the mixture was stirred at 15−25 °C for 1 h. The
resultant slurry was filtered and washed with methanol/water
3:2 (51 kg). The product was dried at 50 °C under vacuum in a
Preparation of Phosphinic Acid 4. To a solution of
GSK2248761A 1 (0.66 g, 1.59 mmol) in methyl isobutyl
ketone (25 mL) was added tetrabutylammonium bromide (2.0
g, 6.2 mmol). The reaction mixture was heated to 100 °C to
give a colorless solution. After 3 h, further tetrabutylammonium
bromide (2.0 g, 6.2 mmol) was added, and the reaction was
continued for a further 2 h. The reaction mixture was then
cooled to rt and was washed with sulfuric acid (50 mL of a 1 M
aqueous solution and 100 mL of a 1 M aqueous solution). The
organic layer was then extracted with sodium bicarbonate (2 ×
50 mL of a 6% w/w aqueous solution). The combined aqueous
extracts were acidified to pH 1.0 to give phosphinic acid 4 as a
white precipitate. This was filtered off and was recrystallized
1
from methanol (5.0 mL). Yield = 0.15 g, 24%. H NMR (500
MHz, DMSO) δ 2.23 (s, 3H), 6.38 (d, J = 16.5 Hz, 1H), 7.18
(dd, J = 8.7 Hz, J = 2.5 Hz, 1H), 7.43 (m, 1H), 7.46 (dd, (br),
1H), 7.58 (s, 1H), 7.61 (d, 1H), 7.66 (d, 1H), 7.69 (d, J = 1.8
Hz, 1H), 7.79 (s (br), 1H), 10.49 (s (br), 1H) and 12.51 (d, J =
D
DOI: 10.1021/acs.oprd.7b00357
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
2.4 Hz, 1H). IR ν (cm−1) 3440, 2920, 1668, 1620, 1402, 1054,
and 1028.
Notes
The authors declare no competing financial interest.
§C.B. and E.V. are former GlaxoSmithKline employees.
⊥A.S., J.W., B.A.M., and A.M. are former Idenix employees.
HRMS calcd for C19H16ClN3O3P: 400.0612; HRMS found
[M + H]+: 400.0608
Preparation of N-Acyl Urea 8. To a solution of
GSK2248761A (0.57 g, 1.38 mmol) in acetonitrile (7.5 mL)
and THF (7.5 mL) was added N-chlorosulfonyl isocyanate
(0.12 mL, 1.38 mmol). The reaction mixture was stirred at rt
for ca. 3 h and was then quenched with water (15 mL). Organic
solvent was then removed in vacuo, and the aqueous mixture
was extracted with ethyl acetate (2 × 75 mL). The organic
extract was concentrated to dryness in vacuo, and the residue
was chromatographed on silica gel using ethyl acetate/heptane
4:1 as eluant. The required product was obtained as a white
REFERENCES
■
(1) (a) Klibanov, O. M.; Kaczor, R. Current opinion in investigational
drugs (Biomed central) 2010, 11 (2), 237−245. (b) Zhou, X.-J.;
Pietropaolo, K.; Damphousse, D.; Belanger, B.; Chen, J.; Sullivan-
Bolyai, J.; Mayers, D. Antimicrob. Agents Chemother. 2009, 53, 1739.
(c) Storer, R.; Dousson, C.; Alexandre, F.-R.; Roland, A. US 2006
0074054 A1. (d) Castellino, S.; Groseclose, M. R.; Sigafoos, J.;
Wagner, D.; de Serres, M.; Polli, J. W.; Romach, E.; Myer, J.;
Hamilton, B. Chem. Res. Toxicol. 2013, 26, 241.
(2) (a) Storer, R.; Alexandre, F. R.; Dousson, C.; Moussa, A. M.;
Bridges, E. PCT International Application WO2008/042240A2, 2008.
(b) Alexandre, F. R.; Amador, A.; Bonaric, S.; Bot, S.; Caillet, C.; Da
Costa, D.; Lioure, M. P.; Mayes, B. A.; Parsy, C.; Roland, A.;
Rosinovsky, E.; Stewart, A.; Wang, J.; Moussa, A.; Dousson, C. 236th
ACS National Meeting in Philadelphia 17−21 August 2008.
(3) Macor, J. E.; Cuff, A.; Cornelius, L. Tetrahedron Lett. 1999, 40,
2733.
1
solid (0.12 g, 42%). H NMR (500 MHz, DMSO) δ 2.34 (s,
3H), 3.79 (d, 3H), 6.52 (d, 1H), 7.36 (dd, 1H), 7.55 (d, 1H),
7.56 (s br, 1H), 7.64 (dd, 1H), 7.66 (d, 1H), 7.67 (d, 1H), 7.74
(m, 1H), 7.84 (d, 1H), 7.94 (s br, 1H), 12.75 (s, 1H), and
13.05 (s, 1H).
IR ν (cm−1) 3401, 3207, 2944, 2860, 1714, 1660, 1588, 1400,
and 1231.
(4) HPLC method: Kinetrix 2.6 μm C18 100A, 4.6 mm × 100 mm;
flow: 2.4 mL/min; eluant A: water with 0.25% v/v TFA; eluant B:
acetonitrile with 0.25% v/v TFA; Gradient: 0 min: 80% A, 20% B; 55
min: 47% A, 53% B; 55.1min: 80% A, 20% B; 60 min: 80% A, 20% B.
UV detection at 275 nm.
HRMS calcd for C21H19ClN4O4P: 457.0827; HRMS found
[M + H]+: 457.0821.
Preparation of Indoloyl Carboxamide 6. To a solution
of diketopiperazide 5 (500 mg, 0.630 mmol) in 2-methyltetra-
hydrofuran (2-MeTHF (7.5 mL) was added ammonium
hydroxide (0.8 mL of a 25% w/w aqueous solution, 5.0
mmol). The resultant mixture was stirred at room temperature
until the bright yellow color had disappeared (ca. 30 min).
After this time the reaction mixture was added to a 1:1 mixture
of water/2-methyltetrahydrofuran (100 mL). The organic layer
was separated and was concentrated to dryness to give a 7:3
mixture of 6/GSK2248761A (455 mg, 63%). An analytically
pure sample was prepared using mass directed preparative
1
chromatography. H NMR (400 MHz, CDCl3) δ 2.34 (s, 6H),
3.86 (d, 6H), 5.88 (d, 2H), 7.26 (d, 2H), 7.36 (m, 4H), 7.60 (d,
2H), 7.65 (d, 2H), 7.80 (d, 2H), 7.89 (s, 2H), 11.56 (s, 2H),
and 13.90 (s, 1H).
IR ν (cm−1) 3239, 1730, 1543, 1403, 1317, 1202, and 1025.
HRMS calcd for C40H32Cl2N5O6P2: 810.1199; HRMS found
[M + H]+: 810.1197.
Preparation of Diketopiperazide 5. To a suspension of
CDI (7.82 g, 48.23 mmol) in 2-Me-THF (100 mL) was added
1 (10 g, 24 mmol) over 5 min at rt. The resultant mixture was
warmed to 39 °C and was stirred for 2 h. The reaction mixture
was then allowed to cool to rt, and the resultant suspension was
stirred for 20 h. The product was filtered and washed with 2-
MeTHF (ca. 20 mL) and was dried at 25 °C for 3 h. The
product was obtained as a bright yellow solid (7.89g, 82%). 1H
NMR (400 MHz, CDCl3) δ 2.39 (s, 6H), 3.89 (d, 6H), 5.96 (d,
2H), 7.38 (d, 2H), 7.40 (s, 2H), 7.62 (d, 2H), 7.76 (d, 2H),
7.93 (d, 2H), 8.52 (d, 2H), and 8.64 (s, 2H); IR ν (cm−1) 1709,
1540, 1447, 1359. 1167, 1120, 1066, 1029, and 965.
HRMS calcd for C40H29Cl2N4O6P2: 793.0934; HRMS found
[M + H]+: 793.0926
AUTHOR INFORMATION
■
Corresponding Author
*E-mail: Jerome.Hayes@gsk.com.
ORCID
Jerome F. Hayes: 0000-0001-9875-0464
John D. Hayler: 0000-0003-3685-3139
E
DOI: 10.1021/acs.oprd.7b00357
Org. Process Res. Dev. XXXX, XXX, XXX−XXX