A Next Generation Synthesis of BACE1 Inhibitor Verubecestat (MK-8931)

Article abstract of DOI:10.1021/acs.orglett.8b00259

The development of a commercial manufacturing route to verubecestat (MK-8931) is described, highlights of which include the application of a continuous processing step to outcompete fast proton transfer in a Mannich-type ketimine addition, a copper-cataly

Full text of DOI:10.1021/acs.orglett.8b00259

Letter
Cite This: Org. Lett. XXXX, XXX, XXX−XXX
pubs.acs.org/OrgLett
A Next Generation Synthesis of BACE1 Inhibitor Verubecestat
MK‑8931)
(
,†
,†
,†
†
†
David A. Thaisrivongs,* William J. Morris,* Lushi Tan,* Zhiguo J. Song, Thomas W. Lyons,
†
†
†
‡
‡
‡
Jacob H. Waldman, John R. Naber, Wenyong Chen, Lu Chen, Baoyun Zhang, and Jun Yang
†
Process Research and Development, Merck & Co., Inc., P.O. Box 2000, Rahway, New Jersey 07065, United States
‡
WuXi AppTec (Shanghai) Pharmaceutical Co. Ltd., 288 Fute Zhong Road, Waigaoqiao Free Trade Zone, Shanghai 200131, China
S Supporting Information
*
ABSTRACT: The development of a commercial manufactur-
ing route to verubecestat (MK-8931) is described, highlights of
which include the application of a continuous processing step
to outcompete fast proton transfer in a Mannich-type ketimine
addition, a copper-catalyzed amidation reaction, and an
optimized guanidinylation procedure to form the key iminothiadiazine dioxide core.
1
erubecestat (MK-8931, 1), currently being evaluated in
Scheme 1. Second Generation Synthesis of Verubecestat (1)
V
Phase III clinical trials, is an inhibitor of BACE1 that
dramatically reduces the levels of amyloid β peptides in the
2
central nervous system of Alzheimer’s disease (AD) patients.
Our laboratories have recently disclosed the first generation
3
synthesis of 1, a route that enabled the discovery, preclinical, and
early development research programs. We have also reported a
highly efficient second generation synthesis that triples the
4
overall yield of the first. The key features of this latter effort are a
diastereoselective Mannich-type addition of a methyl sulfona-
5
mide (3) to a chiral Ellman sulfinyl ketimine (2) to generate the
stereogenic tertiary carbamine, a copper-catalyzed C−N
coupling of the resultant aryl bromide (4) and 4-fluoropicolina-
mide (5), and a late-stage guanidinylation of the penultimate
intermediate (7) with cyanogen bromide (Scheme 1). The
conciseness of this improved synthesis, supported by a
minimization of functional group manipulations and protecting
group chemistry, provides access to 1 at half the cost of the first
generation route and with a more than 70% reduction in the6
waste generated as judged by the process mass intensity (PMI).
The revised synthesis was successfully executed at scale to
support activities related to the verubecestat (1) late develop-
ment program. Nonetheless, several chemical transformations
drew our attention as steps that should be able to be conducted
with even greater synthetic efficiency. The first was the Mannich-
type addition of the lithium anion of 3 to ketimine 2. Only a
moderate assay yield of 4 is obtained at pilot plant scale (70−
process for this step that controls the selectivity between the
various reaction pathways through mixing characteristics in
flow. Not only does operating in a continuous instead of batch
7
mode provide a substantially higher assay yield of the desired
product (89−91%), but it relieves the reaction of the cryogenic
operating requirement; the optimal process temperature is now
only −20 °C (Scheme 2).
Contemporaneous with this discovery, we continued to study
the copper-catalyzed amidation reaction between 5-fluoro-
picolinamide (5) and aryl bromide 4, as it too proceeded in
only modest yield (70%) at pilot plant scale (Scheme 1). No
75%), and the reaction must be performed under cryogenic
operating conditions (less than −65 °C), a temperature that not
only is highly energy-intensive to achieve and maintain but also
imposes significant constraints on the flexibility of both the siting
and vessel selection of commercial product manufacture. We
have recently reported deuterium quenching studies of this
reaction that revealed the competition between the desired 1,2-
addition and deprotonation of the electrophile by both the
organometallic nucleophile and the unquenched product,
experiments which led us to the discovery of a continuous
Received: January 28, 2018
©
XXXX American Chemical Society
A
DOI: 10.1021/acs.orglett.8b00259
Org. Lett. XXXX, XXX, XXX−XXX

Organic Letters
Letter
Scheme 2. Using Flow To Outpace Fast Proton Transfer in the
Mannich-Type Addition of 9 to 2
Table 1. Selected Catalyst Loading and Temperature
Optimization Experiments for the C−N Coupling of Aryl
Bromide 11 with 5-Fluoropicolinamide (5)
CuI
(equiv)
ligand
(equiv)
temp
(°C)
conversion
a
a
entry
13 (%)
(%)
1
2
3
4
5
6
0.3
0.3
0.3
0.3
0.2
0.1
0.6
0.6
0.6
0.3
0.2
0.1
88
107
125
110
110
110
3.0
1.6
1.3
0.6
0.9
0.7
90.5
99.3
99.4
99.7
99.5
92.0
significant improvement was realized through exhaustive high-
throughput reaction parameter screening. In considering
alternative fragment couplings that might proceed in higher
efficiency, we became interested in the reactivity and selectivity
of a partially deprotected analog of 5 under similar catalytic
conditions. We treated the unquenched product stream from the
continuous process (Scheme 2) with excess HCl to selectively
and rapidly remove the N-tert-butanesulfinyl group in the
presence of the more acid-stable para-methoxybenzyl sulfona-
mide, which delivered a solution of amine 11 in high assay yield
a
Determined by HPLC analysis.
significant cost-savings at commercial manufacturing scale. This
observation was consistent with the hypothesis that the substrate
itself can serve to ligate copper, although an exogenous ligand is
still necessary to effect this reaction. The decrease in ligand
loading also slightly increased the assay yield of 12, as we found a
proportional decrease in the amount of 13 formed (e.g., entries 3
and 4). The catalyst loading can be decreased to only 0.2 equiv
with no significant decrease in performance (entry 5), but 0.1
equiv was insufficient for full conversion (entry 6). Under the
optimal conditions, 12 was obtained in 90% isolated yield on
(Scheme 3). Amine 11 is only weakly crystalline, and so the
Scheme 3. Synthesis of Amine 11 and Subsequent C−N
Coupling with 5-Fluoropicolinamide (5)
9
pilot plant scale.
Together, the novel continuous process for the Mannich-type
ketimine addition to prepare 10 and the revised C−N coupling
chemistry provided a significant improvement in the overall yield
compared to the second generation route. The main challenge to
integrating these innovations into a next generation synthesis of
verubecestat (1) was that, in the previous route, crystallization of
intermediate 6 provided the pivotal point of stereochemical
purity upgrade (typically from 91:9 dr to >99:1 dr). This
isolation was the only stage in the synthesis at which the
separation of diastereomers through crystallization was possible,
and downstream rejection of the corresponding undesired
enantiomers through crystallization was far more modest.
Intermediate 4 is a weakly crystalline solid, preventing it from
serving as a similar point of diastereomeric control. Thus, we
chose to evaluate chiral acid salts of our new C−N coupling
precursor 11, seeking to upgrade the stereochemical purity as
early as possible in the synthesis and, critically, to identify an
isolable solid intermediate. In high-throughput fashion, we
looked for crystalline salts of a wide array of commercially
available chiral acids in the presence of 11; (S)-mandelic acid
isolation of pure material without resorting to chromatography
proved challenging. Nevertheless, we were delighted to discover
that, unintuitively, the desired amidation reaction of 11 with 5
proceeded with near-quantitative conversion of the starting
material and in higher isolated yield (85%) than the previously
employed fully protected aryl bromide 4. We speculate that 11
could serve as a monodentate ligand for copper, accelerating the
aryl amidation reaction by favoring the catalytically active
8
10
copper−amide species over the unreactive cuprate complex.
emerged as the most promising. In the optimized crystal-
trans-N,N′-Dimethylcyclohexane-1,2-diamine remained the
preferred bidentate ligand, as it afforded the best yield while
minimizing the amount of ligand-aryl bromide coupled by-
product (which we observed with other less sterically hindered
diamine ligands such as N,N′-dimethylethylenediamine and
trans-1,2,-diaminocyclohexane). Toluene was the optimal
solvent to suppress proto-debromination of 11 to amine 13
lization process, the desired salt (15) could be isolated in 98.5%
yield (relative to the desired enantiomer present in the starting
material), 99.9% purity, and 98.8% ee (Scheme 4). Notably,
being able to use a water-immiscible solvent such as toluene for
the crystallization after an aqueous deprotection of 10 provided a
Scheme 4. Optimized Conditions for the Crystallization of 15
(
Scheme 3). The coupling must be conducted above 100 °C to
achieve optimal conversion and avoid significant formation of 13
Table 1, entries 1 and 2), but diminishing returns were observed
(
at 125 °C (entry 3). Unlike the C−N coupling of the second
generation synthesis, only an equimolar amount of ligand was
required relative to the copper salt (entry 3), which provides
B
DOI: 10.1021/acs.orglett.8b00259
Org. Lett. XXXX, XXX, XXX−XXX

Organic Letters
Letter
seamless connection between an organic phase extraction of free-
based 11 (the end of the continuous process) and the
downstream salt isolation (the starting material for the revised
C−N coupling).
Scheme 6. Optimized Conditions for the Deprotection of 12
With these synthetic improvements in place, we turned to the
deprotection of 12. In the second generation synthesis, the para-
methoxybenzyl protecting group is removed with neat trifluoro-
acetic acid (Scheme 1). Faced with the prospect of continuing to
use such an excess (24 equiv) of an expensive reagent, we
returned to reaction parameter screening in an attempt to
uncover a practical and cost-efficient alternative. We overcame
the very poor solubility of 7 in most organic solvents by using
acetic acid as the bulk media, and although acetic acid was not
strong enough to perform the desired transformation efficiently,
stronger acids could be added to accelerate the deprotection.
Both sulfuric and methanesulfonic acid in the presence of acetic
acid provided 7 in equivalently high yield and purity as obtained
with the neat trifluoroacetic acid protocol (Scheme 5); we chose
to advance the latter acid since sulfuric acid led to intractable
emulsions in a variety of aqueous workups.
In the second generation synthesis, a reactive crystallization of
hydrogen bromide salt 8 provided the basis for an efficient
formation of the iminothiadiazine dioxide ring from 7 and
cyanogen bromide (Scheme 1). Our laboratories have recently
disclosed an unprecedented formation of a 1:1:2 cocrystal of 1, 7,
and hydrogen bromide, respectively, under these reaction
conditions, a singular physical phenomenon that renders the
existing process untenable. In renewed efforts to wholly redesign
the end game to circumvent this liability, we initially sought to
identify basic conditions that would provide a homogeneous
reaction and neutralize the problematic hydrogen bromide. In
general, we observed low assay yields due to the reaction of 1 (no
longer protected as hydrogen bromide salt 8) with cyanogen
bromide. For example, when 0.5 equiv of N,N-diisopropyl-
ethylamine was used we observed a 8:1 mixture of 1 to
overcyanated 16 (Scheme 7).
11
Scheme 5. Previous Conditions for the Deprotection of 12
Scheme 7. Over-cyanation of 1 with Cyanogen Bromide To
Form 16 under Basic Reaction Conditions
In lab-scale experiments, once the deprotection was complete,
the hazy solution was cooled to 20 °C, diluted with water, washed
with toluene, and then slowly added to aqueous ammonium
hydroxide to effect a pH-driven crystallization of 7. In our first
kilogram-scale batch of the new methanesulfonic acid based
procedure, we did not, surprisingly, observe the same hazy
solution at the end of reaction. Instead, the solution became clear
and the hazy material amassed to a sticky gumball that could not
be redispersed into solution with agitation or heat. Analysis of the
solid material by MALDI-TOF mass spectrometry revealed
many species, the largest of which were over 6000 Da. Each
observed peak differed in mass from the others by a multiple of
We hypothesized that we could avoid generating either the
aforementioned cocrystal or 16 if we developed a process that
featured a chemoselective conversion of intermediate 7 to
cyanamide 17, a workup step to remove residual cyanogen
bromide, and then a final intramolecular cyclization event to
generate 1. Amine bases weaker than N,N-diisopropylethylamine
provided cyanamide 17 as the exclusive product, but in only 50−
1
20 Da (Figure 1), implicating para-methoxybenzyl cation
6
3% conversion (Table 2, entries 1−3). A breakthrough occurred
polymerization as the source of the gumball. Conducting the
deprotection in the presence of 1.0 equiv of the electron-rich
arene 1,3-dimethoxybenzene (DMB) completely prevented the
formation of higher-order PMB polymers, and the resulting
DMB-PMB byproducts were readily removed during the workup
in the existing toluene wash (Scheme 6).
upon the evaluation of inorganic bases, where higher levels of
conversion to cyanamide 17 were observed in a range of solvents
using either KH PO or NaHCO (entries 5, 6, 11, and 12). We
2
4
3
chose to optimize the reaction with NaHCO rather than
3
KH PO , as the latter presented handling challenges on scale due
2
4
to its hygroscopicity. When employing NaHCO , some product
3
decomposition was observed under extended reaction times in
MeOH or DMAc (entries 8 and 9). NMP, THF, and 2-MeTHF
were differentiated with respect to conversion and stream
stability, and similarly high levels of reaction conversion were
achieved using 2-MeTHF when the process was conducted at
only 45 °C. Following a reductive workup using aqueous sodium
thiosulfate to destroy the small amount of residual cyanogen
bromide, the 2-MeTHF stream containing cyanamide 17 could
be simply treated with aqueous sodium hydroxide to facilitate
intramolecular cyclization to 1, which was accompanied by only
trace amounts of overcyanated 16 (Scheme 8). The product was
isolated as its para-toluenesulfonate salt (18) by addition of a
solution of para-toluenesulfonic acid; a subsequent free-basing
with potassium carbonate and crystallization from EtOAc and n-
heptane provided 1 in 86% isolated yield from 7.
Figure 1. MALDI spectra of PMB polymers from the deprotection of 12
to 7.
C
DOI: 10.1021/acs.orglett.8b00259
Org. Lett. XXXX, XXX, XXX−XXX

Organic Letters
Letter
Table 2. Selected Optimization Experiments for the
Conversion of Amine 7 to N-Cyano Intermediate 17
AUTHOR INFORMATION
Corresponding Authors
■
*
*
*
E-mail: david.thaisrivongs@merck.com.
E-mail: william.morris2@merck.com.
E-mail: lushi_tan@merck.com.
ORCID
entry
base
solvent
conversion (%) ,
ab
David A. Thaisrivongs: 0000-0002-0387-8885
William J. Morris: 0000-0003-4322-4509
Wenyong Chen: 0000-0002-7337-0876
Notes
1
2
3
4
5
6
7
8
9
HMDS
MeCN
MeCN
MeCN
THF
50
diethylaniline
56
2,6-di-tert-butylpyridine
63
K HPO4
2
58
The authors declare no competing financial interest.
K HPO4
2
DMAc
IPA
73
K HPO4
2
86
ACKNOWLEDGMENTS
K HPO4
2
NMP
MeOH
DMAc
THF
60
■
NaHCO3
NaHCO3
NaHCO3
NaHCO3
NaHCO3
63 (54)
80 (71)
53 (86)
96
We thank Xiaodong Bu, Li Zhang, Erik Regalado, Daniel Zewge,
Chunyu Li, and Heather Wang for analytical chemistry support
and Becky Ruck, Paul Bulger, and Louis-Charles Campeau for
helpful discussions. All are members of Process Research and
Development, Merck & Co., Inc.
1
1
1
0
1
2
NMP
c
2-MeTHF
>99
a
b
Determined by HPLC analysis after 24 h. Values in parentheses
refer to conversion determined after 48 h. Reaction conducted at 45
C.
c
REFERENCES
■
°
(1) Scott, J. D.; Stamford, A. W.; Gilbert, E. J.; Cumming, J. N.; Iserloh,
U.; Misiaszek, J. A.; Li, G. PCT Int. Appl., WO 2011/044181 A1, PCT/
US2010/051553, 2011.
Scheme 8. Revised End Game Synthesis of Verubecestat (1)
(2) Kennedy, M. E.; Stamford, A. W.; Chen, X.; Cox, K.; Cumming, J.
N.; Dockendorf, M. F.; Egan, M.; Ereshefsky, L.; Hodgson, R. A.; Hyde,
L. A.; Jhee, S.; Kleijn, H. J.; Kuvelkar, R.; Li, W.; Mattson, B. A.; Mei, H.;
Palcza, J.; Scott, J. D.; Tanen, M.; Troyer, M. D.; Tseng, J. L.; Stone, J. A.;
Parker, E. M.; Forman, M. S. Sci. Transl. Med. 2016, 8, 363ra150.
(3) Scott, J. D.; Li, S. W.; Brunskill, A. P. J.; Chen, X.; Cox, K.;
Cumming, J. N.; Forman, M.; Gilbert, E. J.; Hodgson, R. A.; Hyde, L. A.;
Jiang, Q.; Iserloh, U.; Kazakevich, I.; Kuvelkar, R.; Mei, H.; Meredith, J.;
Misiaszek, J.; Orth, P.; Rossiter, L. M.; Slater, M.; Stone, J.; Strickland, C.
O.; Voigt, J. H.; Wang, G.; Wang, H.; Wu, Y.; Greenlee, W. J.; Parker, E.
M.; Kennedy, M. E.; Stamford, A. W. J. Med. Chem. 2016, 59, 10435.
(4) Thaisrivongs, D. A.; Miller, S. P.; Molinaro, C.; Chen, Q.; Song, Z.
J.; Tan, L.; Chen, L.; Chen, W.; Lekhal, A.; Pulicare, S. K.; Xu, Y. Org.
Lett. 2016, 18, 5780.
(5) (a) Cogan, D. A.; Ellman, J. A. J. Am. Chem. Soc. 1999, 121, 268.
(b) Robak, M. T.; Herbage, M. A.; Ellman, J. A. Chem. Rev. 2010, 110,
3600.
The sum of this work constitutes the third generation synthesis
of verubecestat (1). The overall yield of 61% from the coupling of
and 3 through to 1 improves significantly on the second
2
(
6) Jimenez-Gonzalez, C.; Ponder, C. S.; Broxterman, Q. B.; Manley, J.
B. Org. Process Res. Dev. 2011, 15, 912 and references therein.
7) (a) Thaisrivongs, D. A.; Naber, J. R.; McMullen, J. P. Org. Process
generation route, for which the same sequence delivers a 44%
overall yield. Additionally, gains in efficiency across the overall
process produced a further 18% reduction in PMI. These results
were enabled by the discovery of a novel continuous process for
the Mannich-type ketimine addition used to form the chiral
tertiary carbamine, the identification of a new chiral salt (15) to
provide the key point of stereochemical upgrade, and a new C−N
coupling that takes advantage of an unexpected boost in
reactivity gained by employing a substrate (11) that has been
partially deprotected. Further improvements in the end game
that enhance the robustness of this process include the use of 1,3-
dimethoxybenzene to avoid risks posed by PMB-based polymers
in the final deprotection and a new orchestration of the
guanidylation sequence to synthesize the iminothiadiazine
dioxide ring that circumvents a recently identified cocrystal.
(
Res. Dev. 2016, 20, 1997. (b) Thaisrivongs, D. A.; Naber, J. R.; Rogus, N.
J.; Spencer, G. Org. Process Res. Dev. 2018, 10.1021/acs.oprd.7b00385.
(
8) For a report of a similar copper-catalyzed C−N coupling rate
enhancement in the presence of n-HexNH , see: Klapars, A.; Huang, X.;
2
Buchwald, S. L. J. Am. Chem. Soc. 2002, 124, 7421.
(9) See Supporting Information for details on how the Design of
Experiments method was used to further optimize this transformation.
(10) See Supporting Information for details.
(
11) Artino, L.; Varsolona, R.; Brunskill, A.; Morris, W. J.;
Thaisrivongs, D. A.; Waldman, J.; Lyons, T. W.; Xu, Y. Org. Process
Res. Dev. 2018, 10.1021/acs.oprd.8b00015.
ASSOCIATED CONTENT
Supporting Information
■
*
S
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/acs.orglett.8b00259.
Experimental procedures, compound characterization
data, and NMR spectra (PDF)
D
DOI: 10.1021/acs.orglett.8b00259
Org. Lett. XXXX, XXX, XXX−XXX