Synthesis of Vibegron Enabled by a Ketoreductase Rationally Designed for High pH Dynamic Kinetic Reduction

Article abstract of DOI:10.1002/anie.201802791

Described here is an efficient stereoselective synthesis of vibegron enabled by an enzymatic dynamic kinetic reduction that proceeds in a high-pH environment. To overcome enzyme performance limitations under these conditions, a ketoreductase was evolved by a computationally and structurally aided strategy to increase cofactor stability through tighter binding.

Full text of DOI:10.1002/anie.201802791

A Journal of the Gesellschaft Deutscher Chemiker
Angewandte
Chemie
International Edition
www.angewandte.org
Accepted Article
Title: Synthesis of Vibegron Enabled by a Ketoreductase Rationally
Designed for High pH Dynamic Kinetic Reduction
Authors: Feng Xu, Birgit Kosjek, Fabien Cabirol, Haibin Chen, Richard
Desmond, Jeonghan Park, Anupam Gohel, Steven Collier,
Derek Smith, Zhuqing Liu, Jacob Janey, John Chung, and
Oscar Alvizo
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: Angew. Chem. Int. Ed. 10.1002/anie.201802791
Angew. Chem. 10.1002/ange.201802791
Link to VoR: http://dx.doi.org/10.1002/anie.201802791
http://dx.doi.org/10.1002/ange.201802791

10.1002/anie.201802791
Angewandte Chemie International Edition
COMMUNICATION
Synthesis of Vibegron Enabled by a Ketoreductase Rationally
Designed for High pH Dynamic Kinetic Reduction
Feng Xu,*[a] Birgit Kosjek,*[a] Fabien L. Cabirol,[b] Haibin Chen,[b] Richard Desmond,[a] Jeonghan Park,[a]
Anupam P. Gohel,[b] Steven J. Collier,[b] Derek J. Smith,[b] Zhuqing Liu,[a] Jacob M. Janey,[a] John Y. L.
Chung,[a] and Oscar Alvizo[b]
Abstract: Described here is an efficient stereoselective synthesis of
vibegron, enabled through an enzymatic dynamic kinetic reduction in
a high pH process environment. To overcome enzyme performance
limitations under these conditions, a ketoreductase has been
evolved using a computationally and structurally aided strategy to
increase cofactor stability via tighter binding.
convergence to the first generation route, but the low atom-
economy and the phosphorus waste were of concern in the
context of an environmentally responsible synthesis for large
scale preparation. Furthermore, the transformation of hydroxy
amino ester 4 to imine 6 introduced a reduction-oxidation-
reduction sequence and protection-deprotection,[5] contributing to
synthetic inefficiency.
Overactive bladder syndrome (OAB) is a condition of urinary
frequency and urgency that negatively affects quality of life. The
prevalence of OAB is expected to increase by 18% from 2008 to
2018.[1] Before 2012, muscarinic antagonists were the standard-
of-care prescribed pharmacological therapy, but these agents
lack sufficient efficacy and have side effects due to blockade of
M3 muscarinic receptor.[2] To address the unmet medical need,
3-adrenergic receptor (β3-AR) agonists have recently emerged
as a promising new class of agents for the treatment of OAB.[3]
Vibegron[4] (1) is a potent and selective β3-AR agonist
currently in Phase 3 clinical trials. It structurally differs from
acyclic β-hydroxylamine β3-AR agonists,[3a-b,4a] and possesses a
unique cis pyrrolidine moiety linking the C2 and C5 substituents.
The discovery of the cis pyrrolidine backbone improved the
pharmacological properties,[4a] but raised the synthetic
complexity.
Scheme 1. Key steps of the first generation process of vibegron (1).
In 2013 we reported an asymmetric synthesis[5] of cis-2,5-
disubstituted pyrrolidine 2, the core scaffold of vibegron. In this
approach (Scheme 1), the R,R (C1’,C2) stereochemistry of 1
was efficiently established through an enzymatic dynamic kinetic
(DK) reduction of -keto ester 3. The third stereogenic center,
C5, was established by diastereoselective hydrogenation of
imine 6. Vibegron (1) was obtained after coupling with
pyrimidinone acid sodium salt 7.[6]
Scheme 2. Retro-synthetic analysis of pyrrolidine 2
We envisioned that a more concise synthesis of vibegron
could arise from reducing a corresponding pyrrolidine imine
formed by intramolecular Michael addition to aryl alkyne 8
(Scheme 2). The inclusion of an electron-withdrawing nitro group
in precursor 8 not only promotes the Michael addition, but also
allows for unveiling of the aniline functionality in 2 through a
global reduction.
The first generation route provided a robust means to
prepare large quantities (>100 kg) of vibegron to support early
safety and clinical studies. Despite the acceptable overall yield
(32%), we envisioned that the atom economy could be improved.
The central problem of the first generation synthesis lies in the
arduous nature of creating the pyrrolidine imine moiety in
intermediate 6 with multiple transformations. The application of a
Another synthetic challenge of preparing vibegron is its
unique R,R (C1’,C2) stereochemistry. The methodologies of
choice to establish syn stereochemistry of 1,2-amino alcohols in
an open-chain system have been lacking.[7] Our previous
enzymatic DK reduction approach prepared optically pure syn
1,2 amino alcohol 4 effectively.[5] However, our new strategy
(Scheme 2) eliminates the activating ester group that facilitated
the epimerization of keto ester 3 at pH 7.5.[5] In contrast,
epimerization studies on 10, prepared by treating amino nitrile
11[8] with phenyl Grignard reagent (Scheme 3), confirmed that
fast racemization could only be realized at elevated temperature
(≥ 45C) and high pH (≥ pH 10). These conditions are generally
recognized as harsh for ketoreductases (KREDs), challenging
the limits of enzyme performance.
HornerWadsworthEmmons reaction using
5
offered
[a]
[b]
Dr. F. Xu, Dr. B. Kosjek, R. Desmond, Dr. J. Park, Dr. Z. Liu, Dr. J.
Janey, Dr. J. Y. L. Chung
Department of Process Research and Development, MRL, Merck &
Co., Inc., Rahway, NJ 07065, USA
E-mail: feng_xu@merck.com; birgit_kosjek@merck.com
Dr. F. L. Cabirol, Dr. H. Chen, A. P. Gohel, Dr. S. J. Collier, Dr. D. J.
Smith, Dr. O. Alvizo
Codexis, Inc., 200 Penobscot Drive, Redwood City, CA 94063, USA
Supporting information and the ORCID identification number(s) for
the author(s) of this article can be found under: https://
doi.org/10.1002/anie.201802791
This article is protected by copyright. All rights reserved.

10.1002/anie.201802791
Angewandte Chemie International Edition
COMMUNICATION
Table 1. Selected results of DK reduction through KRED evolution
entry
KRED
evolution
round
temp.
(C)
pH
substrate
concentration (g/L)
enzyme loading
(wt% relative to 10)
conv. (%)[a]
ee (%)[b]
dr (syn/ anti)[b]
1
2
3
4
5
KRED-P1B2
KRED-P1B2
KRED-p101
KRED-p201
KRED-p301
0
0
1
2
3
30
30
45
45
45
7.0
25
15
50
50
50
35
100
200
25
1
43[c]
5[d]
52
>99
>99
>99
>99
>99.4
16:1
n.d.
10.0
10.0
10.0
10.0
28:1
64:1
>100:1
38
95
[a] Unless otherwise noted, reactions were carried out in 50% aqueous i-PrOH with 0.2 M borate buffer at pH 10.0 and 0.1 g/L NADP+. [b] Determined by
HPLC analysis. [c] In aqueous i-PrOH with 0.1 M phosphate buffer at pH 7.0 and 1.0 g/L NADP+. [d] 2g/L NADP+
At the inception of this work, the literature precedence for
ketoreductase reactions was limited to pH <8 and ambient
temperature, reflecting typical physiological conditions.[9]
Examples of enzymatic DK reductions have been reported on -
keto esters or activated ketones with a pKa range of 712[5,10]
with the exception of rare reports using native extremophile
KREDs at pH 9.[10a] Alternatively, a strategy of modifying
functional group(s) in precursor 10 to lower the pKa and achieve
epimerization at pH <8 was unattractive and would reduce
synthetic efficiency. To pursue our retro-synthetic target
(Scheme 2), we first needed to identify a KRED that could
produce alcohol 9 from 10. Given the advances in protein
engineering technologies,[11] we envisioned tailoring this enzyme
to the required reaction conditions by employing directed
evolution with computationally and structurally designed
mutations to rapidly create an optimal catalyst.
alcohol 9 due to steric interactions with the Boc or alkyne group
(Figure 1B).
In parallel, we targeted to recover overall conversion at the
desired conditions. The degradation rate of NADP+ increases
significantly above pH 8, in particular at elevated temperature.[12]
We concluded that poor performance of the reduction in the
basic environment was the result of cofactor rather than enzyme
instability. Since the catalytic nicotinamide cofactor cycles
between its oxidized and reduced form throughout the reaction
(Scheme 3), decomposition of NADP+ eventually exhausts all
available cofactor and shuts down the reaction. We
hypothesized that improved binding of the cofactor to the
KRED’s active site would provide a protective environment for
the cofactor and increase its stability.
We screened a Codexis® KRED panel under standard
aqueous conditions at pH 78. The results not only showed
excellent reactivity, but also gave us a hit on stereoselectivity.
Although epimerization of 10 did not occur at pH 7 and 30C,
kinetic resolution with KRED-P1B2 gave desired alcohol 9 with
high dr (16:1) and excellent ee (>99%) (Table 1, entry 1) while
the undesired ketone enantiomer S-10 remained intact. As
expected, the reaction underperformed with increasing pH,
producing only minimal amounts of 9 at pH 10 (Table 1, entry 2).
Clearly, variant P1B2 required improvements in both
diastereoselectivity and productivity at higher temperature and
pH. To address stereoselectivity, we focused on seven
positions[8] in and around the substrate binding pocket. These
positions underwent saturation mutagenesis in an attempt to
identify improved amino acids. Screening of the libraries
revealed seven unique mutations at five of the targeted positions
that were 1.5 to 13.8-fold improved in activity over P1B2, and a
mutation to phenylalanine at position 206 (M206F) which
improved the enzyme selectivity. The phenylalanine was
structurally predicted to have favorable interaction with the
phenyl group of the substrate when bound by the enzyme in a
conformation that yields the desired (R,R) alcohol 9 (Figure 1A).
In contrast, the substrate is disfavored from binding in the
opposite conformation that would result in the (S,R) or (S,S)
Scheme 3. Enzymatic reduction of ketone 10.
Figure 1. M206F beneficial mutation. The NADPH cofactor, substrate 10, and
M206F mutation are depicted in yellow, green, and salmon, respectively. The
docked 10 is shown in a conformation that results in the desired (R,R) alcohol
9 at the targeted carbon center in panel A and in the corresponding undesired
(R,S)-9 in panel B. The predicted steric clash between the Boc group and the
phenylalanine group at position 206 is circled in red.
This article is protected by copyright. All rights reserved.

10.1002/anie.201802791
Angewandte Chemie International Edition
COMMUNICATION
conversion at decreased enzyme loading. In the second round
of evolution, 20 positions[8] around the substrate binding pocket
were targeted for saturation mutagenesis. Screening of the
library identified ten unique mutations at eight positions that
were beneficial over the parent. The best mutations G94P,
I144V, and L196M were improved in activity by 1.7-, 2.1-, and
2.1-fold, respectively. The three mutations were combined in
KRED-p201, resulting in a round 2 variant more active and
selective than KRED-p101 (Table 1, entry 4). The final round of
evolution recombined all beneficial diversity identified in the
previous two rounds in one library to fully optimize the enzyme.
The final variant,[8] KRED-p301, contained six additional
mutations relative to KRED-p201 and was shown to be at least
50-fold improved overall, resulting in an efficient and selective
high pH DK reduction biocatalyst (Table 1, entry 5). In practice,
treatment of 10 with 1wt% KRED-p301 and only 0.1g/L NADP in
pH 10 aqueous borate buffer containing 50% i-PrOH at 45C
afforded 9 in 95% assay yield with >99% ee and >100:1 dr.
Sonogashira coupling of crude 9 with p-iodonitrobenzene in the
presence of <1 mol% (PPh3)2PdCl2 and 0.52 mol% CuI in THF
and i-PrOH at ambient temperature gave the desired coupling
product 12 in 95% yield.
Figure 2. H40R beneficial mutation. The NADPH cofactor and the H40R
mutation are depicted in yellow and salmon respectively. The mutation to
arginine at position 40 is shown interacting with the phosphate of the cofactor.
We applied a rationally guided strategy, analyzing the KRED
cofactor binding pocket for improvements. Position 40 was
quickly identified as a good candidate for mutation based on
computational structural analysis, and showed benefits from
mutating it to an arginine. The H40R mutation displayed
favorable electrostatic interactions with the phosphate of
NADPH (Figure 2) and likely stabilized binding of the cofactor.
Engineering of the H40R mutation into the selectivity hit M206F
produced the first round hit KRED-p101. Experimental
characterization of this double mutant showed KRED-p101 to be
successfully improved in cofactor binding, validating the
structural predictions. In the absence of the H40R mutation,
variant M206F converts over 40% of 5 g/L 10 in 24 hours at 0.5
g/L cofactor (Figure 3A). However, at 0.1 g/L cofactor, the
activity drops by an order of magnitude while the variant with the
H40R mutation retained its activity and showed close to 70%
conversion (Figure 3B). The result is consistent with the
hypothesis that tighter binding serves to protect the cofactor
from prolonged exposure to the higher pH environment and thus
supresses its breakdown.
Carrying out the Sonogashira reaction in alcohol solvents
allowed significant reduction in catalyst loading, presumably
because the reduction capability of alcohol solvents maintained
the Pd(0) catalyst reactivity.[13] Without isolation, intermediate 12
was treated with conc. HCl in i-PrOH to afford HCl salt 13.
Combining directed evolution approaches and rational
mutagenesis revealed key mutations improving both
stereospecific control in the reduction of 10 and stabilization of
the cofactor in the enzyme, effectively increasing the enzyme’s
operating window to the target conditions in a single round of
evolution.
Scheme 4. Preparation of vibegron
Figure 3. Impact of the H40R mutation on productivity (5 g/L of 10) at
decreased cofactor loading and pH 10 at 35C. a) The performance of the
parent P1B2 vs. the M206F mutant at 0.5 g/L cofactor. b) The performance of
P1B2, mutant M206F and the double mutant M206F, H40R (KRED-p101) at
0.1 g/L cofactor.
Upon treatment of HCl salt 13 with i-Pr2NEt in THF-AcNMe2
at 60C, the desired intramolecular Michael addition proceeded
smoothly under an atmosphere of nitrogen to afford imine 14.
The cyclization was highly selective, and Michael addition of the
benzylic OH to the electron deficient carbon-carbon triple bond
was not observed. Isolation of 14 was undesirable since the free
base was prone to benzylic oxidation in air. However,
The subsequent protein engineering focused on further
improving selectivity and specific activity to achieve higher
This article is protected by copyright. All rights reserved.

10.1002/anie.201802791
Angewandte Chemie International Edition
COMMUNICATION
maintaining an inert atmosphere to prevent oxidation was
straightforward at plant scale. Therefore, the crude stream 14
was carried forward to the hydrogenation step and treated in situ
with TMSCl to form the corresponding silyated intermediate 15,
which was required to enhance the diastereoselectivity of the
hydrogenation.[5] Methanol was added to consume excess
TMSCl before the reaction stream 15 was subjected to global
hydrogenation conditions. In the presence of Pt/Al2O3,
pyrrolidine 2 was obtained in 95:5 dr. Surprisingly, preliminary
results showed that the initial hydrogenation lacked
reproducibility in terms of conversion and reaction rate. In
addition, the high catalyst loading (>3 mol%) was not cost
effective, even though the heterogeneous catalyst could be
recovered easily on large scale. After several experiments, we
found that residual palladium carried through from the
Sonogashira coupling was responsible for catalyst poisoning,
which was easily rejected to <150 ppm by a slurry wash of the
wet cake of 13 during isolation. The resulting robust
hydrogenation was then achieved with low catalyst loading (<1
mol%). Upon aqueous workup, pyrrolidine 2 was isolated in 80%
yield and the undesired diastereomer was rejected to <1%.
We thank our colleagues of Merck & Co., Inc., Rahway, NJ, USA,
Dr. S. Hoerrner, J. Hill and M. Weisel of the High Pressure
Laboratory for experimental assistance, R. Reamer, Dr. P.
Dormer and L. DiMichele for assistance with NMR studies, T.
Nowak for HRMS analysis, and Dr. G. Hughes for useful
discussion.
Keywords: enzyme • evolution • ketoreductase • synthesis
[1]
a) K. S. Coyne, C. C. Sexton, V. Vats, C. Thompson, Z. S. Kopp, I.
Milsom, Urology 2011, 77, 1081; b) D. E. Irwin, Z. S. Kopp, B. Agatep,
I. Milsom, P. Abrams, BJU Int. 2011, 108, 1132.
[2]
[3]
P. Abrams, K.-E. Andersson, BJU Int. 2007, 100, 987.
a) E. Sacco, R. Bientinesi, D. Tienforti, M. Racioppi, G. Gulino, D.
D’Agostino, M. Vittori, P. Bassi, Expert Opin. Drug Discovery 2014, 9,
433; b) R. Bragg, D. Hebel, S. M. Vouri, J. M. Pitlick, Consult.
Pharmacist 2014, 29, 823; c) K.-E. Andersson, N. Martin, V. Nitti, J.
Urol. 2013, 190, 1173; d) P. Tyagi, V. Tyagi, N. Yoshimura, M.
Chancellor, O. Yamaguchi, Drugs Future 2009, 34, 635.
a) S. D. Edmondson, et al., J. Med. Chem. 2016, 59, 609; b) M.
Yoshida, M. Takeda, M. Gotoh, S. Nagai, T. Kurose, Eur. Urol. 2018,
73, in press.
[4]
[5]
[6]
[7]
F. Xu, et al., Org. Letters 2013, 15, 1342.
The initial coupling of 2 with pyrimidione acid 9 was carried
out in the presence of EDC·HCl in aqueous AcNMe2 (Scheme
1).[6] However, an undesired solvate form of 1 was isolated in
77% yield, which required a recrystallization to obtain the
desired, more bioavailable crystal form. To improve the
volumetric productivity and eliminate the recrystallization for
crystal form correction, a process to carry out the EDC coupling
in pH controlled aqueous i-PrOH was developed. The pyrrolidine
nitrogen in 2 was selectively protonated with HCl, leaving the
less basic aniline nitrogen available to effectively couple with the
free acid of 7. Worthy of note is the operational ease of this
endgame process at large scale. After adjusting the pH of a
solution of 2 and sodium salt 7 in aqueous i-PrOH with HCl to
pH 3.13.7, EDC·HCl was charged at 510C to cleanly afford
desired product 1. Direct addition of aqueous NH4OH to the
reaction stream followed by filtration completed the synthesis of
vibegron, which was obtained in its desired crystal form in 93%
yield with 99.8% purity and nearly perfect ee.
N. Yasuda, et al., PCT Int. Appl. 2013, WO 2013062878 A1.
a) Z. Liu, et al., Tetrahedron Lett. 2011, 52, 1685; b) B. Seashore-
Ludlow, P. Villo, C. Häcker, P. Somfai, Org. Lett. 2010, 12, 5274.
For details, see the Supporting Information.
[8]
[9]
a) T. S. Moody, S. Mix, G. Brown, D. Beecher in Science of Synthesis.
Biocatalysis in Organic Synthesis 2 (Eds.: K. Faber, W.-D. Fessner, N.
J. Turner), Georg Thieme Verlag KG, Stuttgart New York, 2015, pp.
447-450. b) J. C. Moore, D. J. Pollard, B. Kosjek, P. N. Devine, Acc.
Chem. Res. 2007, 40, 1412.
[10] a) G. A. Applegate, D. B. Berkowitz, Adv. Synth. Catal. 2015, 357,
1619; b) D. Kalaitsakis, J. D. Rozzell, S. Kambourakis, I. Smonou, Org.
Lett. 2005, 7, 4799; c) B. Kosjek, D. M. Tellers, M. Biba, R. Farr, J. C.
Moore, Tetrahedron: Asymmetry 2006, 17, 2798; d) R. L. Hanson, Z.
Guo, F. Gonzáles-Bobes, M. D. B. Fenster, A. Goswami, J. Mol. Catal.
B: Enzym. 2016, 133, 20; e) J. Mangas-Sanchez, E. Busto, V. Gotor, V.
Gotor-Fernandez, Org. Lett. 2013, 15, 3872; f) A. M. Hyde, et.al., Org.
Lett. 2016, 18, 5888.
[11] a) U. T. Bornscheuer, G. W. Huisman, R. J. Kazlauskas, S. Lutz, J. C.
Moore, K. Robins, Nature 2012, 485, 185; b) C. K. Savile, et al.,
Science 2010, 329, 305. c) U. T. Bornscheuer, Phil. Trans. R. Soc. A
2018, 376, 20170063. d) F. Rudroff, M. D. Mihovilovic, H. Gröger, R.
Snajdrova, H. Iding, U. T. Bornscheuer, Nature Catal. 2018, 1, 12. e) R.
O. M. A. de Souza, L. S. M. Miranda, U. T. Bornscheuer, Chem. Eur. J.
2017, 23, 12040. f) M. Hönig, P. Sondermann, N. J. Turner, E. M.
Carreira, Angew. Chem. Int. Ed. 2017, 56, 8942; Angew. Chem. 2017,
129, 9068.
In summary, a highly efficient, asymmetric synthesis of
vibegron (1), which has been implemented on a manufacturing
scale, has been described. The synthesis features an enzymatic
DK reduction to establish the challenging C’1,C2
stereochemistry in excellent enantio- and diastereo-selectivity.
Incorporating structurally designed mutations into directed
enzyme evolution produced a ketoreductase suitable for high pH
and elevated temperature conditions in only three evolution
rounds. Subsequent intramolecular Michael addition followed by
diastereoselective hydrogenation concisely constructed the
backbone of vibegron. The second generation route to vibegron
is realized by integration of bioengineering technology into
organic synthesis and is atom efficient with minimal
transformations of functional groups, requiring only one
protecting group in the early intermediates.
[12] H. K. Chenault, G. M. Whitesides, Appl. Biochem. Biotechnol. 1987, 14,
147.
[13] M. J. Zacuto, F. Xu, J. Org. Chem. 2007, 72, 6298.
Acknowledgements
This article is protected by copyright. All rights reserved.

10.1002/anie.201802791
Angewandte Chemie International Edition
COMMUNICATION
COMMUNICATION
F. Xu,* B. Kosjek,* F.L. Cabirol, H.
Chen, R. Desmond, J. Park, A.P. Gohel,
S.J. Collier, D.J. Smith, Z. Liu, J. Janey,
J.Y.L. Chung, and O. Alvizo
Page No. – Page No.
Synthesis of Vibegron Enabled by a
Ketoreductase Rationally Designed
for High pH Dynamic Kinetic
Reduction
Evolution for synthetic efficiency: Target-driven evolution optimizes enzymes for
synthetically desired transformations under conditions that challenge and surpass
known limitations of enzyme performance. A biocatalyst, engineered for
stereoselective, dynamic kinetic reduction in a high pH environment, has enabled
an efficient synthesis of vibegron, a potent 3-adrenergic agonist for the treatment
of overactive bladder syndrome.
This article is protected by copyright. All rights reserved.