Continuous Process Improvement in the Manufacture of Carfilzomib, Part 2: An Improved Process for Synthesis of the Epoxyketone Warhead

Article abstract of DOI:10.1021/acs.oprd.0c00052

The development and kilogram-scale demonstration of an improved process for the synthesis of the epoxyketone warhead of carfilzomib is described. Critical to the success of this process was: (1) development of a scalable asymmetric epoxidation protocol; (2) identification of a crystalline intermediate with improved physical properties for isolation; (3) discovery and optimization of epimerization conditions to set the target stereochemistry; and (4) introduction of a seeded-bed coaddition crystallization to facilitate isolation of the final low-melting target. The results of kilogram-scale demonstration runs are shared, including details of a continuous process for the safe execution of an exothermic Barbier-type Grignard process.

Full text of DOI:10.1021/acs.oprd.0c00052

pubs.acs.org/OPRD
Article
Continuous Process Improvement in the Manufacture of
Carfilzomib, Part 2: An Improved Process for Synthesis of the
Epoxyketone Warhead
Matthew G. Beaver,* Xianqing Shi, Jan Riedel, Parth Patel, Alicia Zeng, Michael T. Corbett,
Jo Anna Robinson, Andrew T. Parsons, Sheng Cui, Kyle Baucom, Michael A. Lovette, Elci̧ n Icţ en,
Derek B. Brown, Ayman Allian, Tawnya G. Flick, Wendy Chen, Ning Yang, and Shawn D. Walker
Cite This: https://dx.doi.org/10.1021/acs.oprd.0c00052
Read Online
sı
*
Metrics & More
Article Recommendations
Supporting Information
ACCESS
ABSTRACT: The development and kilogram-scale demonstration of an improved process for the synthesis of the epoxyketone
warhead of carfilzomib is described. Critical to the success of this process was: (1) development of a scalable asymmetric epoxidation
protocol; (2) identification of a crystalline intermediate with improved physical properties for isolation; (3) discovery and
optimization of epimerization conditions to set the target stereochemistry; and (4) introduction of a seeded-bed coaddition
crystallization to facilitate isolation of the final low-melting target. The results of kilogram-scale demonstration runs are shared,
including details of a continuous process for the safe execution of an exothermic Barbier-type Grignard process.
KEYWORDS: asymmetric epoxidation, continuous manufacturing, carfilzomib
INTRODUCTION
■
As described in the first part of this series (10.1021/
acs.oprd.0c00051), efforts toward a detailed understanding of
the commercial manufacture of epoxyketone 1 highlighted
several opportunities for improvement in the life-cycle
management phase of this program. Central to our approach
was identification and development of a scalable asymmetric
epoxidation method that could address challenges associated
with the existing bleach epoxidation process and eliminate the
requirement for column chromatography. Described herein are
our efforts toward an improved route to (S,R)-epoxyketone 1,
an intermediate in the synthesis of carfilzomib (Figure 1).1
RESULTS AND DISCUSSION
■
Development of an Improved Synthesis of 1. The
established oxidation procedure to prepare (S,R)-epoxyketone
1 employs a substrate-controlled bleach epoxidation to
generate a mixture of epoxide diastereomers (ca. 2:1 dr)
Figure 1. Epoxyketone intermediate in the synthesis of carfilzomib.
from enone 2 (Scheme 1). The poor stereoselectivity, coupled
with the low melting point of the desired epoxyketone
stereoisomer (41 °C), rendered crystallization of the product
from the crude reaction stream impractical. A chromatographic
purification step was required to reduce impurity content prior
to the final isolation, resulting in a significant increase in
solvent consumption, processing time, and analytical testing
burden. Efforts to deliver an improved manufacturing process
focused on the development of a stereoselective epoxidation
strategy to minimize formation of the undesired diastereomer
and facilitate crystallization of the target compound 1.
Scheme 1. Bleach Epoxidation to Synthesize Epoxyketone 1
Received: February 10, 2020
Despite the breadth of literature precedent, rendering the
epoxidation of olefin 2 asymmetric proved to be a significant
challenge.2 The electron-deficient nature of the substrate
https://dx.doi.org/10.1021/acs.oprd.0c00052
Org. Process Res. Dev. XXXX, XXX, XXX−XXX
© XXXX American Chemical Society
A

Organic Process Research & Development
pubs.acs.org/OPRD
Article
necessitates a nucleophilic epoxidation method, which
discounts many of the available asymmetric epoxidation
methods. In addition, the substitution pattern adjacent to the
ketone of olefin 2 challenges iminium ion and Lewis acid
activation methods, which have proven to be useful approaches
for the asymmetric epoxidation of α,β-unsaturated carbonyls.3
Leveraging phase-transfer catalysis4 resulted in poor con-
versions or epimerization of the labile amino-acid side chain.
Approaches relying on redox manipulations of the ketone were
evaluated but ultimately not selected due to the added
synthetic steps and limited improvement to the overall control
strategy.
Scheme 3. Optimized Conditions for Oxidation of Enone 2
Bioinspired nonheme manganese and iron complexes
bearing tetradentate N-donor ligands have emerged as a class
of bench-stable catalysts capable of oxidizing a wide range of
olefins, including electron-deficient and sterically congested
enones.5 Of direct relevance to this effort, Sun and coworkers
reported a manganese-catalyzed epoxidation protocol to
prepare the target epoxyketone in high yield and good
stereoselectivity, albeit with a preference for the undesired
diastereomer.6 The protocol utilizes a Mn-catalyst (C1) in the
presence of H2O2 and acetic acid to prepare (S,S)-epoxyketone
3 in >95% assay yield for the combined epoxide stereoisomers
and 88:12 dr, in excellent agreement with the literature report
(Scheme 2). The enantiomer of enone 2 was then subjected to
with Mn(OTf)2 provided a crystalline and bench-stable
catalyst complex.8
Despite an inability to overcome the substrate bias for
formation of the undesired diastereomer, the solid-state phase
behavior of (S,S)-epoxyketone 3 differed significantly from that
of the target stereoisomer. Notably, the undesired diastereomer
had an increased melting point (78 °C), rendering the
compound amenable to a wider range of crystallization
conditions and providing the opportunity to eliminate the
use of chromatography in the synthetic route. The synthesis
illustrated in Scheme 4 was proposed and served as the basis
Scheme 2. Mn-Catalyzed Asymmetric Epoxidation of Enone
2 by the Procedure of Sun and Coworkers
Scheme 4. Proposed Route to Prepare (S,R)-Epoxyketone 1
identical conditions to determine the influence of catalyst
control on stereoselectivity. This experiment also delivered the
undesired diastereomer as the major product but with
decreased selectivity and yield, indicative of a mismatched
catalyst−substrate combination (Scheme 2). These prelimi-
nary results encouraged further examination of the ligand
architecture and reaction conditions to increase process
efficiency.
Extensive evaluation of the reaction parameters, including
ligand structure, metal and counterion, acid additive, oxidant,
temperature, and solvent, were unsuccessful in identifying a
system capable of overturning the apparent substrate-biased
diastereoselectivity.7 These optimization efforts culminated
with the identification of an improved catalyst complex C2
(Scheme 3) allowing for significantly decreased catalyst
loading (0.04 mol %) while retaining high reaction yield
(>95% assay yield) and diastereoselectivity (91:9 dr); however,
the preference for the undesired (S,S)-epoxyketone 3
remained. The ligand can be prepared in a single step from
commercially available materials. Complexation of the ligand
for further process development efforts, including: (1)
preparation of (R)-enone 2 from unnatural Boc-D-leucine;
(2) diastereoselective epoxidation and isolation by crystal-
lization of (R,R)-epoxyketone 4 with high stereochemical
purity; and (3) epimerization of the leucine side chain to
prepare (S,R)-epoxyketone 1 with high stereoselectivity (target
≥95:5 dr) and purity, enabling a final isolation by
crystallization of this low-melting compound.
As detailed in the previous manuscript of this series
(10.1021/acs.oprd.0c00051), precedent existed for the stereo-
chemical lability of the leucine side-chain of epoxyketone 1 in
the presence of basic additives. Base identity impacted the
selectivity of the epimerization and product stability, as
summarized in Figure 2. Inorganic bases resulted in
epimerization to form the desired stereoisomer but were
associated with significant decomposition. A range of organic
B
https://dx.doi.org/10.1021/acs.oprd.0c00052
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
pubs.acs.org/OPRD
Article
Figure 2. Effect of base on the epimerization and stability of (R,R)-epoxyketone 4.
Scheme 5. Existing Route to Prepare Enone 2
bases was investigated, demonstrating the requirement for a
small, strong (i.e., pKBH+ > 20 as measured in acetonitrile),9
non-nucleophilic base. Importantly, decomposition was sup-
pressed, and a preference for the desired stereoisomer was
observed, providing the requisite proof-of-concept required to
advance this route. Use of triethylamine (TEA, pKBH+ = 18.8)
did not result in formation of the desired isomer, even after
extended aging at elevated temperatures. The use of 1,8-
diazabicyclo[5.4.0]undec-7-ene (DBU, pKBH+ = 24.3) was
productive in effecting the desired epimerization event while
retaining near quantitative assay yields for the process. 1,5,7-
Triazabicyclo[4.4.0]dec-5-ene (TBD, pKBH+ = 26.0) was
similarly productive in generating the desired stereoisomer,
whereas 7-methyl-triazabicyclo[4.4.0]dec-5-ene (MTBD,
pKBH+ = 25.5) and polystyrene-bound TBD (pKBH+ = 26.0)
were completely ineffective for this transformation, supporting
the requirement for a nonhindered base. Sterically hindered
phosphazene bases did not result in the desired epimerization,
and use of less-hindered variants was accompanied by
decomposition. Overall, a range of bases was found to be
effective for this transformation, with diastereoselectivity
governed by thermodynamic equilibrium and favoring the
desired stereoisomer (ca. 95:5 dr at 20 °C). Importantly, the
diastereoselectivity for epimerization met the required target
purity to enable a final isolation by crystallization.
throughput efficiency while retaining the overall bond-
formation strategy.
Development efforts to prepare enone 2 initiated with
optimization of the step 1 process, involving synthesis of
Weinreb amide 5 via isobutylchloroformate-mediated amide-
bond formation. The morpholine amide variant (6) was
evaluated in preliminary experiments and found to perform
comparably in both the coupling step and downstream
processes; therefore, all subsequent studies focused on the
preparation and use of this intermediate.10 From the myriad of
commercially available and well-precedented amide-bond
forming reagents, N,N′-carbonyldiimidazole (CDI) provided
the highest yield and purity profile in an initial evaluation to
prepare morpholine amide 6.11 Formation of acyl imidazole
intermediate 7 proceeded rapidly, as determined by in situ
ReactIR measurements and corroborated by off-line UPLC
analysis (Figure 3).12 Of note, increasing the equivalence of
CDI (from 1.2 to 2.0 equiv) enabled the direct use of Boc-Leu-
OH·H2O without drying via azeotropic distillation. Addition of
morpholine to the reaction mixture resulted in rapid and
complete consumption of intermediate 7 and generation of the
desired morpholine amide 6. A range of solvents were found to
be effective for this transformation (e.g., 2-MeTHF, toluene,
MTBE, THF, etc.); ultimately, 2-MeTHF was chosen for its
ability to solubilize reaction byproducts (e.g., imidazole),
effective partitioning through the aqueous wash sequence, and
the ability to remove water via azeotropic distillation thus
ensuring compatibility with the downstream process. Intro-
duction of an aqueous HCl wash was sufficient to remove
imidazole and excess morpholine from the organic stream, and
an aqueous NaHCO3 or Na2CO3 wash was effective in purging
>5 mol % remaining starting material in the case of incomplete
reaction conversion. Details of a representative kilogram-scale
demonstration using MTBE as solvent to generate the desired
morpholine amide 6 in 97% assay yield and >99.5 A% HPLC
With lab-scale proof of concept established for the synthetic
sequence illustrated in Scheme 4, efforts turned to the
development of a robust, safe, and scalable solution for the
manufacture of (S,R)-epoxyketone 1 from raw materials
through final isolation in the absence of column chromatog-
raphy.
Synthesis of (R)-Enone 2. The existing route to enone 2 is
a two-step telescoped process from Boc-leucine−OH·mono-
hydrate, as illustrated in Scheme 5. Development efforts for
these steps focused on improving yield, purity profile, and
C
https://dx.doi.org/10.1021/acs.oprd.0c00052
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
pubs.acs.org/OPRD
Article
Figure 4. RC1 data for Barbier-type Grignard process to prepare
enone 2.
presence of a catalytic quantity of iodine, the initiation event
typically occurs within 0.2−0.8 equiv of charged 2-
bromopropene; in this specific experiment, the initiation
proceeded upon addition and extended aging of 0.3 equiv 2-
bromopropene. The total heat of reaction was calculated to be
−1182 kJ/mol, corresponding to a 203 °C adiabatic temper-
ature rise. ReactIR and UPLC methods were developed to
monitor both the consumption of 2-bromopropene and the
formation of product to mitigate the potential safety issues
associated with accumulation of the Grignard precursor.
The high exothermicity of this process, together with
variability observed for the initiation step, prompted develop-
ment of a continuous process for the safe execution of this
chemistry. Translation of the batch process to a continuous
process utilizing continuously stirred tank reactors (CSTR) in
series was accomplished with only minor changes to the
procedure; for example, the THF/2-MeTHF ratio was
optimized to facilitate improved solubility of the Grignard
reagent to avoid clogging of the transfer lines or fouling of the
magnesium turnings.13 Proof-of-concept was completed at kg/
day productivities utilizing a dual CSTR reaction platform
followed by inline quench of the basic reaction medium with a
25 wt % citric acid solution and continuous phase separation.
As depicted in Figure 5, a solution of morpholine amide 6 in
THF/2-MeTHF was combined with 2-bromopropene (3.0
equiv) in CSTR1 (20 min, 40 °C) containing an excess of
magnesium turnings.14 The reaction solution was transferred
from CSTR1 to CSTR2 (20 min, 40 °C) to achieve the
required level of conversion (>95%) utilizing a peristaltic
pump and particle-segregation tube to prevent transfer of solid
magnesium turnings.15,16 The crude reaction mixture from
CSTR2 could then be combined with a 25 wt % citric acid
solution and MTBE in CSTR3 to facilitate neutralization of the
stream and phase separation, respectively. Through imple-
mentation of a continuous process, the reaction temperature of
both reaction CSTRs could be controlled to within 2.5 °C
over the course of a 26 h demonstration, ensuring safe
operation and delivery of kg-quantities of enone 2 in 89% yield
with the requisite achiral (>90 A%) and chiral (>99.9 A%)
purity profile for downstream processing.
Figure 3. (a) Reaction progress as determined by quantitative in situ
ReactIR measurements. (b) 3-D surface plot of IR spectra as a
function of time and wavenumber.
achiral and chiral purity are available in the Experimental
Section.
The step 2 process to prepare (R)-enone 2 involves addition
of isopropenylmagnesium bromide to Weinreb amide 5 in
THF (Scheme 5). The use of this commercially available
Grignard reagent presents a significant bottleneck as its low
solubility in THF (0.5 M) results in high dilution at the wide
point of the existing batch process (50 volumes). The
successful design, development, and implementation of a
Barbier-type Grignard process avoids the cycle times associated
with sourcing and titration of the preformed Grignard reagent
and eliminates solubility considerations as the reagent is
consumed on-demand.
Critical to both process safety and product quality for the
Barbier process was control of the exothermic Grignard
formation and a sequence of reactions including deprotonation
(carbamate N−H) and nucleophilic addition to the carbonyl of
morpholine amide 6. Figure 4 illustrates the heat flow (W)
associated with the portion-wise addition of 2-bromopropene
to a reaction calorimeter containing magnesium turnings and
morpholine amide 6 in THF/2-MeTHF at 40 °C. In the
The kilogram-scale demonstration of this process high-
lighted an additional safety benefit of continuous operation;
that is, a single small-scale magnesium activation event was
suitable to initiate Grignard formation and support the entirety
D
https://dx.doi.org/10.1021/acs.oprd.0c00052
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
pubs.acs.org/OPRD
Article
Figure 5. Details of the continuous process to prepare enone 2. (a) Process flow diagram including reaction, inline quench, and phase separation.
(b) Conversion values for enone 2 as determined by UPLC analysis of CSTR2.
of processing. For direct comparison, initiation of the multiday
continuous process used 6.0 g of magnesium turnings, whereas
the corresponding batch process would require a single charge
of over 1 kg. Of the myriad magnesium activation options
available,17 prestirring the magnesium turnings with a catalytic
quantity of iodine was demonstrated to provide a robust and
well-controlled initiation process. Subsequent additions of 2-
bromopropene resulted in immediate and reproducible
exotherms (see Figure 4).
Synthesis of (S,R)-Epoxyketone 1. With an improved
two-step telescoped process to deliver enone 2, we next
focused on the final steps to prepare the target (S,R)-
epoxyketone 1. The foundation for this two-step sequence,
Mn-catalyzed epoxidation and epimerization, was established
through the proof-of-concept efforts discussed in the previous
section. Continued development was then required to optimize
the reaction conditions, aqueous workup procedures, and
intermediate and final isolations to achieve the target of
eliminating column chromatography.
a cooled (−20 °C) solution of enone 2 and catalyst C2 in
ACN (10 volumes) and AcOH (5.0 equiv) generated the
desired epoxide in >95% assay yield and a 91:9 mixture of
diastereomers. Upon reaction completion, 5 volumes of a 25
wt % aqueous NaHSO3 solution was charged to quench
residual hydrogen peroxide and facilitate a phase separation. As
illustrated in Table 1 below, the concentration and identity of
the inorganic salt was critical for the efficient reduction of
hydrogen peroxide, quality of phase cut with the water-miscible
reaction milieu, and minimization of new impurities.
Critical to the success of this process was identification of
conditions for isolation of this intermediate, (R,R)-epoxyke-
tone 4. A solubility screen identified several solvent systems
capable of purging impurities generated in the telescoped
upstream process. Among these options, crystallization from
IPA/water was evaluated for ease of operation and
demonstrated capacity to reject the undesired diastereomer
generated from the epoxidation procedure. Solvent exchange
from the reaction solvents to IPA could be performed with a
single distillation to achieve ≤1.0 wt % ACN. The water
content could then be adjusted to the target of 60 wt %, as
informed by the IPA/water solubility profile (Figure 6) and
capacity for rejection of the undesired epoxide diastereomer.
Reaction conditions for the Mn-catalyzed epoxidation to
prepare epoxyketone 4 were established through the proof-of-
concept studies (Scheme 3, vide infra). In summary, slow
addition of a 50 wt % aqueous solution of H2O2 (2.0 equiv) to
E
https://dx.doi.org/10.1021/acs.oprd.0c00052
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
pubs.acs.org/OPRD
Article
Table 1. Representative Survey of Aqueous Work-up
Conditions
Scheme 6. Epimerization Process to Prepare Crude (S,R)-
Epoxyketone 1
peroxide
test strip
result
losses in aq
phase (%)
entry
1
conditions
comments
5 wt % NaHSO3 no phase separation
(10 volumes)
positive
2
3
20 wt %
NaHSO3 (10
volumes)
clean phase cut
<1
<1
negative
25 wt %
NaHSO3 (5
volumes)
clean and rapid phase
separation
negative
(S,R)-epoxyketone 1, meeting targets for chiral and achiral
purity.
4
5
a
5 wt % Na2SO3
(10 volumes)
no phase separation
positive
negative
negative
The high solubility of epoxyketone 1 in a wide range of
organic solvents prompted the use of water as antisolvent for
isolation at reasonable processing temperatures. Unfortunately,
an evaluation of aqueous-based solvent systems quickly
identified a propensity for oiling (liquid−liquid phase
separation). For example, addition of water to solutions of
epoxyketone 1 in water-miscible cosolvents (e.g., NMP or
MeOH) resulted in oiling and reactor fouling for a wide range
of temperatures, solvent compositions, substrate concentra-
tions, and addition rates. Attempts to determine solubility in a
range of solvent systems highlighted increased temperature, vol
% organic solvent, and epoxyketone 1 concentration as
contributing to liquid−liquid phase separation (Figure 7). At
20 wt % Na2SO3 precipitation of inorganic
(10 volumes) salts
35 wt % Na2SO3 clean phase separation,
(3 volumes) new impurity at 2.2 A%
<1
Figure 6. Solubility of epoxyketone 4 in IPA/water (20 °C) and
representative polarized-light microscope image of isolated solids.
Cooling the batch to 5 °C served as an efficient strategy to
minimize losses without impact to product quality. As
designed, the epoxidation reaction, aqueous workup, and
IPA/water crystallization could tolerate and reject process
impurities generated in the preceding steps without the aid of
column chromatography. Details of a representative kilogram-
scale procedure to prepare (R,R)-epoxyketone 4 in 77%
potency adjusted yield containing <0.5 A% epoxide diaster-
eomer are available in the Experimental Section.
The high-purity crystalline solid isolated from step 3 could
be subjected to the base-promoted epimerization procedure to
generate a reaction solution of adequate purity to enable a final
crystallization of the low-melting target compound 1.
Optimization of the epimerization process identified sub-
stoichiometric quantities of DBU to be an effective promoter
in a range of nonpolar organic solvents (e.g., MTBE, MeTHF,
toluene, heptane, etc.). MTBE was selected due to the high
solubility of both starting material and product, compatibility
with aqueous workup, and facile exchange to the crystallization
solvent (vide infra). Continued development of this process
resulted in the procedure shown in Scheme 6. Addition of
DBU (20 mol %) to a solution of (R,R)-epoxyketone 4 in 10
volumes MTBE results in a 95:5 dr favoring the target
compound 1 after 10 h at 20 °C. A wash with 5 wt % aqueous
NaHSO4 served to remove DBU from the organic layer, and a
final water wash purged residual inorganics. Representative
material from this optimized process was used to support
development of the final crystallization procedure to isolate
Figure 7. Solubility of epoxyketone 1 in MeOH/water as a function
of solvent composition and temperature.
20 °C, equilibrating epoxyketone 1 in solvent compositions
>45 vol % MeOH resulted in oils, whereas at 5 °C oils were
not observed through 70 vol % MeOH. Based on these results
and the requirement to purge 5 mol % diastereomer from the
upstream process, a crystallization operating at an end-point of
50 vol % water at 5 °C was developed.
The solubility of epoxyketone 1 in MeOH/water and
propensity for oil formation suggested the substrate should be
isolated using a seeded batch coaddition process. This variety
of process enables a constant solvent composition and
supernatant concentration when addition rates are controlled
to match desupersaturation kinetics. In this approach, two
solutions (A) epoxyketone 1 in organic solvent and (B) water
are added simultaneously and at the same rate to a preformed
slurry of 5 wt % seed in 5 volumes 1:1 (v/v) MeOH/water at 5
°C (Figure 8). Water content and supernatant concentration
can be monitored throughout the crystallization process and
adjusted based on addition rates of the two streams. Keeping
these values constant throughout the addition allows the
execution of this process avoiding conditions where liquid−
F
https://dx.doi.org/10.1021/acs.oprd.0c00052
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
pubs.acs.org/OPRD
Article
(4.8 L, 2.5 L/kg) was added, and the distillation was repeated.
MTBE (9.6 L, 5.0 L/kg) was added; the solution was
concentrated to 2 L/kg under vacuum while maintaining the
temperature at <35 °C, and the resulting concentrate was
reconstituted with MTBE (9.6 L, 5.0 L/kg). The solution was
cooled to 0 °C, and a slurry of N,N′-carbonyldiimidazole (1.51
kg, 9.28 mol) in MTBE (7.7 L, 4.0 L/kg) was added over 25
min while maintaining the internal temperature at <5 °C. The
reaction mixture was aged for an additional 1 h at 0 °C. To the
cooled reaction mixture was added morpholine (1.01 kg, 11.6
mol) over 30 min while maintaining the internal temperature
at <10 °C. The reaction mixture was aged for 1 h at 0 °C. A
solution of 1 M aqueous HCl (6.8 L, 3.5 L/kg) was added, and
the biphasic mixture was warmed to 20 °C. The layers were
allowed to separate, and the bottom aqueous layer was
removed. The organic layer was washed sequentially with 1 M
aqueous HCl (2.9 L, 1.5 L/kg), 8 wt % aqueous NaHCO3 (1.9
L, 1.0 L/kg), and 6 M aqueous NaCl (5.8 L, 3.0 L/kg) to
afford an MTBE solution of morpholine amide 6 (2.25 kg, >
99.5 LCAP, 97% solution assay yield). A sample was
characterized after concentration of the solution to provide
Figure 8. Illustration of the seeded batch coaddition crystallization
procedure.
liquid phase separation occur, facilitating the isolation of this
low-melting compound in a robust manner. Details of a
representative kilogram-scale procedure to prepare (S,R)-
epoxyketone 1 utilizing an analogous NMP/water seeded
batch coaddition crystallization procedure are available in the
Experimental Section.
1
the following analytical data: H NMR (400 MHz, CDCl3) δ
5.26 (d, J = 8.9 Hz, 1H), 4.62 (m, 1H), 3.45−3.72 (m, 8H),
1.71 (m, 1H), 1.42 (m, 11 H), 0.96 (d, J = 6.7 Hz, 3H), 0.92
(d, J = 6.5 Hz, 3H); 13C NMR (125 MHz, CDCl3) δ 171.6,
155.6, 79.6, 66.8, 66.6, 48.2, 46.0, 42.8, 42.4, 28.4, 24.6, 23.3,
22.0; HRMS (ESI-TOF) m/z calcd for C15H29N2O4 (M + H)+
301.2127, found 301.2126.
Preparation of tert-Butyl (R)-(2,6-Dimethyl-3-oxohept-1-
en-4-yl)carbamate 2. Two solutions were prepared for
continuous addition to CSTR1: (A) a 0.658 M solution of
morpholine amide 6 (5.22 kg, 17.4 mol) in 2.5:1 (v/v) THF/
2-MeTHF (21.9 L, 4.20 L/kg) and (B) a 0.886 M solution of
2-bromopropene (6.61 kg, 51.9 mol) in THF (53.8 L, 10.3 L/
kg).
Batch start-up of CSTR1 was performed to activate the
magnesium turnings prior to continuous operation. To an
inerted 1 L jacketed vessel (CSTR1) was charged magnesium
turnings (6.04 g, 0.248 mol), a catalytic quantity of iodine, and
morpholine amide solution A (0.120 L, 0.0776 mol). The
reaction mixture was warmed to 40 °C, and 2-bromopropene
(5.94 g, 0.0466 mol) was added in three equivalent portions
and aged until initiation was confirmed by observation of an
exothermic event and corroborated by in situ ReactIR. To the
reaction mixture was added an additional quantity of 2-
bromopropene (23.8 g, 0.186 mol) while maintaining the
internal temperature at <45 °C.
CONCLUSIONS
■
An improved four-step process to prepare the epoxyketone
warhead of carfilzomib was developed, which culminated in the
elimination of column chromatography. The identification of a
stereoselective epoxidation reaction, coupled with an under-
standing of the phase behavior of each stereoisomer, were
critical to advance this approach. The process has been
demonstrated on kilogram-scale to generate the target
compound in over 50% yield with an E-factor reduction of
88% (2639 to 304).18,19
EXPERIMENTAL SECTION
■
General. All reactions were performed under an atmos-
phere of nitrogen under anhydrous conditions unless otherwise
noted. All solvents and reagents were commercially obtained
1
and used without further purification. H NMR spectra were
recorded at ambient temperature at 400 MHz using a Bruker
AVANCE-400 spectrometer, and 13C NMR spectra were
recorded at 125 MHz using a Bruker AVANCE-500
1
spectrometer. The H NMR data are reported as follows:
A single bulk addition of magnesium turnings (59.2 g, 2.44
mol), corresponding to 12.0 equiv per CSTR turnover, was
charged to CSTR1, and then continuous operation was
commenced. To CSTR1 was pumped simultaneously feed
solution A (15.4 mL/min, 10.1 mmol/min) and feed solution
B (35.0 mL/min, 31.0 mmol/min). Both CSTR1 and CSTR2
were 1 L jacketed reactors with mean residence times of 20
min each, as dictated by the level of transfer tube. The internal
reaction temperature was controlled at 40 °C, and the vessels
were swept with nitrogen. A tube-in-tube transfer line was used
between CSTR1 and CSTR2 to prevent solid magnesium
transfer.15 Additional magnesium turnings were charged in
bulk portions at regular intervals to maintain 6−18 equiv
magnesium with respect to morpholine amide 6 (per CSTR
turnover).
chemical shift in parts per million (ppm) from an internal
standard of residual CHCl3 in CDCl3 (7.27 ppm) on the δ
scale, multiplicity (s = singlet, d = doublet, t = triplet, q =
quartet, m = multiplet), coupling constants in hertz (Hz), and
integration (H). Chemical shifts of 13C NMR spectra are
reported in ppm from the central peak of CDCl3 (77.0 ppm).
High resolution mass spectra (HRMS) were obtained using
Agilent 1100 systems. Melting points were measured by
differential scanning calorimetry (DSC) using a TA Instru-
ments Q200 DSC at 10 °C/min.
Preparation of tert-Butyl (R)-(4-Methyl-1-morpholino-1-
oxopentan-2-yl)carbamate 6. A solution of (tert-butoxycar-
bonyl)-D-leucine·mohonohydrate (1.93 kg, 7.73 mol) in THF
(4.8 L, 2.5 L/kg) was concentrated to 2 L/kg under vacuum
while maintaining the temperature at <35 °C. Additional THF
G
https://dx.doi.org/10.1021/acs.oprd.0c00052
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
pubs.acs.org/OPRD
Article
The product stream from CSTR2 was transferred con-
tinuously to CSTR3 and combined with 25 wt % aqueous citric
acid (28 mL/min, 9.0 L/kg) and MTBE (15 mL/min, 5.0 L/
kg). CSTR3 was a 1 L glass-lined jacketed reactor with internal
reaction temperature controlled at 20 °C and mean residence
time of 5 min. The rapidly agitated biphasic mixture from
CSTR3 was transferred to a settling vessel where the aqueous
solution was diverted to waste and the organic layer was
transferred to a collection vessel to afford a solution of enone 2
(3.64 kg, 94.4 LCAP, 89% solution assay yield) over 26 h of
continuous operation. A sample was characterized after
concentration of the solution to provide the following
HRMS (ESI-TOF) m/z calcd for C14H25NNaO4 (M + Na)+
294.1681, found 294.1680.
Preparation of tert-Butyl ((S)-4-Methyl-1-((R)-2-methylox-
iran-2-yl)-1-oxopentan-2-yl)carbamate 1. To a 20 °C
solution of (R,R)-epoxyketone 4 (1.03 kg, 3.79 mol) in
MTBE (10 L, 10 L/kg) was charged l,8-diazabicyclo[5.4.0]-
undec-7-ene (0.116 kg, 0.762 mol) in a single portion. The
reaction mixture was aged for 24 h at 20 °C, and a solution of 5
wt % aqueous NaHSO4 (4.6 L, 4.5 L/kg) was added. The
layers were allowed to separate, and the bottom aqueous layer
was removed. The organic layer was washed with water (5.5 L,
5.4 L/kg) and then polish filtered to remove particulate matter.
The solution was concentrated to 3 L/kg under vacuum while
maintaining the temperature at <35 °C. NMP (2.1 L, 2.0 L/
kg) was added, and the solution was concentrated under
vacuum while maintaining temperature at <35 °C until a target
of <1 wt % MTBE remained (ca. 3 L/kg). The solution was
diluted with additional NMP (3.1 L, 3.0 L/kg) and transferred
to a holding vessel. Simultaneous addition of the organic
solution and an equal volume of water (6.1 L, 5.9 L/kg) to a
precooled (5 °C) seed slurry of (S,R)-epoxyketone 1 (0.0515
kg, 0.200 mol) in 1:1 (v/v) NMP/water (5.1 L, 5.0 L/kg)
proceeded over 9 h. The resulting slurry was then filtered, and
the wet cake was washed twice with a cooled (5 °C) solution
of 1:1 (v/v) NMP/water (2.1 L, 2.0 L/kg), twice with cooled
(5 °C) water (2.1 L, 2.0 L/kg), and dried at 20 °C under
vacuum to afford (S,R)-epoxyketone 1 (0.89 kg, 98.6 LCAP,
1
analytical data: H NMR (400 MHz, CDCl3) δ 6.09 (s, 1H),
5.89 (s, 1H), 5.10 (m, 2H), 1.91 (s, 3H), 1.74 (m, 1H), 1.49
(m, 1H), 1.44 (s, 9H), 1.34 (m, 1H), 1.01 (d, J = 6.5 Hz, 3H),
0.92 (d, J = 6.6 Hz, 3H); 13C NMR (125 MHz, CDCl3) δ
155.5, 142.4, 126.0, 105.0, 79.6, 52.6, 43.2, 28.4, 25.0, 23.4,
21.8, 17.8; HRMS (ESI-TOF) m/z calcd for C14H25NNaO3
(M + Na)+ 278.1732, found 278.1731.
A total of 75 kg of organic solution from continuous
operation was subjected to batch workup and distillation.
Aqueous wash of the product solution was performed in two
identical batches: the organic solution was washed sequentially
with 5 wt % aqueous NaCl (10 L, 4.0 L/kg), 8 wt % aqueous
NaHCO3 (13 L, 5.0 L/kg), and 5 wt % aqueous NaCl (10 L,
4.0 L/kg). The resulting organic solutions were combined and
then concentrated to 2 L/kg under vacuum while maintaining
the temperature at <30 °C. Acetonitrile (24 L, 5.0 L/kg) was
added, and the solution concentrated to 2 L/kg under vacuum
while maintaining temperature at <30 °C. The potency of the
organic solution was then adjusted with acetonitrile in
preparation for the downstream process.
1
97.4 wt %, 79% yield) as a white crystalline solid: H NMR
(400 MHz, CDCl3) δ 4.86 (d, J = 8.5 Hz, 1H), 4.31 (m, 1H),
3.29 (d, J = 4.9 Hz, 1H), 2.88 (d, J = 5.0 Hz, 1H), 1.72 (m,
1H), 1.51 (s, 3H), 1.48 (m, 1H), 1.41 (s, 9H), 1.17 (m, 1H),
0.96 (d, J = 6.5 Hz, 3H), 0.93 (d, J = 6.6 Hz, 3H); 13C NMR
(125 MHz, CDCl3) δ 209.5, 155.6, 79.7, 59.0, 52.3, 51.4, 40.5,
28.3, 25.1, 23.4, 21.3, 16.8; melting point = 39−41 °C; HRMS
(ESI-TOF) m/z calcd for C14H25NNaO4 (M + Na)+ 294.1681,
found 294.1681.
Preparation of tert-Butyl ((R)-4-Methyl-1-((R)-2-methylox-
iran-2-yl)-1-oxopentan-2-yl)carbamate 4. To a reactor
containing a solution of enone 2 (1.09 kg, 4.28 mol) in
acetonitrile (12 L, 11 L/kg) was charged catalyst C2 (0.00139
kg, 0.00171 mol) and acetic acid (1.29 kg, 21.4 mol). The
solution was cooled to −20 °C; a solution of 50 wt % aqueous
hydrogen peroxide (0.583 kg, 8.57 mol) was added over 1.5 h
while maintaining temperature at <−15 °C, and the reaction
mixture was aged for 2 h at −20 °C. Upon reaction
completion, the solution was warmed to 0 °C, and 35 wt %
aqueous Na2S2O3 (5.1 L, 4.7 L/kg) was added over 20 min
while maintaining the temperature at <10 °C. The biphasic
mixture was warmed to 20 °C; the layers were allowed to
separate, and the bottom aqueous layer was removed. The
organic layer was concentrated to 3 L/kg under vacuum while
maintaining the temperature at <35 °C. Isopropanol (12 L, 11
L/kg) was added, and the solution was concentrated to 5 L/kg
under vacuum while maintaining the temperature at <35 °C.
Water (5.6 L, 5.1 L/kg) was added over 1.5 h, and the resulting
slurry was cooled to 0 °C over 4 h. The slurry was then filtered,
and the wet cake was washed twice with a cooled (0 °C)
solution of 2:3 (v/v) isopropanol/water (2.4 L, 2.2 L/kg) and
dried at 20 °C under vacuum to afford ( R,R)-epoxyketone 4
ASSOCIATED CONTENT
■
sı
* Supporting Information
The Supporting Information is available free of charge at
https://pubs.acs.org/doi/10.1021/acs.oprd.0c00052.
Additional experimental details (PDF)
C2 crystallographic information file (CIF)
AUTHOR INFORMATION
■
Corresponding Author
Matthew G. Beaver − Process Development, Amgen, Inc.,
Cambridge, Massachusetts 02142, United States; orcid.org/
0000-0003-3816-4601; Email: mbeaver@amgen.com
Authors
Xianqing Shi − Process Development, Amgen, Inc., Thousand
Oaks, California 91320, United States
Jan Riedel − Process Development, Amgen, Inc., Cambridge,
Massachusetts 02142, United States
Parth Patel − Process Development, Amgen, Inc., Cambridge,
Massachusetts 02142, United States
Alicia Zeng − Process Development, Amgen, Inc., Cambridge,
Massachusetts 02142, United States
(0.91 kg, 95.9 LCAP, 97.2 wt %, 77% yield) as a white
1
crystalline solid: H NMR (400 MHz, CDCl ) δ 4.88 (m,
3
1H), 4.58 (m, 1H), 3.04 (d, J = 5.1 Hz, 1H), 2.86 (d, J = 5.1
Hz, 1H), 1.71 (m, 1H), 1.56 (s, 3H), 1.44 (s, 9H), 1.36 (m,
2H), 0.98 (d, J = 6.4 Hz, 3H), 0.93 (d, J = 6.6 Hz, 3H); 13C
NMR (125 MHz, CDCl3) δ 207.8, 155.2, 79.8, 58.7, 52.9, 52.8,
41.1, 28.3, 24.9, 23.3, 21.5, 17.6; melting point = 77−78 °C;
Michael T. Corbett − Process Development, Amgen, Inc.,
Thousand Oaks, California 91320, United States
H
https://dx.doi.org/10.1021/acs.oprd.0c00052
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
pubs.acs.org/OPRD
Article
́
(3) See, for example: (a) Marigo, M.; Franzen, J.; Poulsen, T. B.;
Zhuang, W.; Jørgensen, K. A. Asymmetric Organocatalytic Epox-
idation of α,β-Unsaturated Aldehydes with Hydrogen Peroxide. J. Am.
Chem. Soc. 2005, 127, 6964−6965. (b) Bondzic, B. P.; Urushima, T.;
Ishikawa, H.; Hayashi, Y. Asymmetric Epoxidation of α-Substituted
Acroleins Catalyzed by Diphenylprolinol Silyl Ether. Org. Lett. 2010,
12, 5434−5437. (c) Nemoto, T.; Ohshima, T.; Yamaguchi, K.;
Shibasaki, M. Catalytic Asymmetric Epoxidation of Enones Using
La−BINOL−Triphenylarsine Oxide Complex: Structural Determi-
nation of the Asymmetric Catalyst. J. Am. Chem. Soc. 2001, 123,
2725−2732. (d) Jacques, O.; Richards, S. J.; Jackson, R. F. W.
Catalytic asymmetric epoxidation of aliphatic enones using tartrate-
derived magnesium alkoxides. Chem. Commun. 2001, 2712−2713.
(4) See, for example: (a) Corey, E. J.; Zhang, F. Y. Mechanism and
Conditions for Highly Enantioselective Epoxidation of α,β-Enones
Using Charge-Accelerated Catalysis by a Rigid Quaternary
Ammonium Salt. Org. Lett. 1999, 1, 1287−1290. (b) Lifchits, O.;
Mahlau, M.; Reisinger, C. M.; Lee, A.; Fares, C.; Polyak, I.;
Gopakumar, G.; Thiel, W.; List, B. The Cinchona Primary Amine-
Catalyzed Asymmetric Epoxidation and Hydroperoxidation of α,β-
Unsaturated Carbonyl Compounds with Hydrogen Peroxide. J. Am.
Chem. Soc. 2013, 135, 6677−6693.
Jo Anna Robinson − Process Development, Amgen, Inc.,
Cambridge, Massachusetts 02142, United States
Andrew T. Parsons − Process Development, Amgen, Inc.,
Cambridge, Massachusetts 02142, United States; orcid.org/
0000-0002-0320-0919
Sheng Cui − Process Development, Amgen, Inc., Cambridge,
Massachusetts 02142, United States
Kyle Baucom − Process Development, Amgen, Inc., Thousand
Oaks, California 91320, United States
Michael A. Lovette − Process Development, Amgen, Inc.,
Thousand Oaks, California 91320, United States
Elci̧ n Icţ en − Process Development, Amgen, Inc., Cambridge,
Massachusetts 02142, United States
Derek B. Brown − Process Development, Amgen, Inc., Thousand
Oaks, California 91320, United States; orcid.org/0000-
0002-7909-0867
Ayman Allian − Process Development, Amgen, Inc., Thousand
Oaks, California 91320, United States
Tawnya G. Flick − Process Development, Amgen, Inc., Thousand
Oaks, California 91320, United States; orcid.org/0000-
0001-7266-8999
Wendy Chen − Process Development, Amgen, Inc., Thousand
Oaks, California 91320, United States
Ning Yang − Process Development, Amgen, Inc., Cambridge,
Massachusetts 02142, United States
Shawn D. Walker − Process Development, Amgen, Inc.,
Cambridge, Massachusetts 02142, United States
́
(5) For select reviews, see: (a) Cusso, O.; Ribas, X.; Costas, M.
Biologically inspired non-heme iron-catalysts for asymmetric epox-
idation; design principles and perspectives. Chem. Commun. 2015, 51,
14285−14298. (b) Bryliakov, K. Environmentally Sustainable Catalytic
Asymmetric Epoxidations; Taylor & Francis Group: New York, 2015.
(6) Wang, B.; Miao, C.; Wang, S.; Xia, C.; Sun, W. Manganese
Catalysts with C1-Symmetric N4 Ligand for Enantioselective
Epoxidation of Olefins. Chem. - Eur. J. 2012, 18, 6750−6753.
(7) A summary of the reaction parameters investigated is available in
the Supporting Information
Complete contact information is available at:
https://pubs.acs.org/10.1021/acs.oprd.0c00052
(8) Details of the synthesis are available in the Supporting
Information. The observed displacement of a triflate anion in the
single crystal X-ray structure likely occurred during the isolation from
ACN/water. No negative impact was observed upon subjecting this
catalyst to the epoxidation chemistry.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
(9) Kaljurand, I.; Kutt, A.; Soovali, L.; Rodima, T.; Maemets, V.;
̈
̈ ̈
We thank Yasmin Alaee, Jiemin Bao, and Guilong Cheng for
analytical contributions. We thank Kumar Ranganathan,
Simone Spada, and John Huckins for their contributions to
develop and demonstrate the process. We thank Richard
Staples for performing the included single crystal X-ray
analysis.
Leito, I.; Koppel, I. A. Extension of the Self-Consistent Spectrophoto-
metric Basicity Scale in Acetonitrile to a Full Span of 28 pKa Units:
Unification of Different Basicity Scales. J. Org. Chem. 2005, 70, 1019−
1028.
(10) For select examples of morpholine amides used in ketone
synthesis, see: (a) Brown, J. D. Asymmetric synthesis of naproxen via
a pinacol-type reaction. Tetrahedron: Asymmetry 1992, 3, 1551−1552.
(b) Martin, R.; Romea, P.; Tey, C.; Urpi, F.; Vilarrasa, J. Simple and
Efficient Preparation of Ketones from Morpholine Amides. Synlett
1997, 12, 1414−1416. (c) Sengupta, S.; Mondal, S.; Das, D. Amino
acid derived morpholine amides for nucleophilic α-amino acylation
reactions: A new synthetic route to enantiopure α-amino ketones.
Tetrahedron Lett. 1999, 40, 4107−4110. (d) Jackson, M. M.; Leverett,
C.; Toczko, J. F.; Roberts, J. C. An Unexpected Equilibrium Process
Associated with a Standard Approach to Ynone Synthesis. J. Org.
Chem. 2002, 67, 5032−5035.
(11) For a recent review, see: Dunetz, J. R.; Magano, J.;
Weisenburger, G. A. Large-Scale Applications of Amide Coupling
Reagents for the Synthesis of Pharmaceuticals. Org. Process Res. Dev.
2016, 20, 140−177.
(12) UPLC analysis was accomplished by derivatization of the acyl
imidazole intermediate with a 10 wt % solution of morpholine in
acetonitrile.
REFERENCES
■
(1) Siegel, D. S.; Martin, T.; Wang, M.; Vij, R.; Jakubowiak, A. J.;
Lonial, S.; Trudel, S.; Kukreti, V.; Bahlis, N.; Alsina, M.; Chanan-
Khan, A.; Buadi, F.; Reu, F. J.; Somlo, G.; Zonder, J.; Song, K.;
Stewart, A. K.; Stadtmauer, E.; Kunkel, L.; Wear, S.; Wong, A. F.;
Orlowski, R. Z.; Jagannath, S. A phase 2 study of single-agent
carfilzomib (PX-171−003-A1) in patients with relapsed and refractory
multiple myeloma. Blood 2012, 120, 2817−2825.
(2) For select reviews, see: (a) Xia, Q.-H.; Ge, H.-Q.; Ye, C.-P.; Liu,
Z.-M.; Su, K.-X. Advances in Homogeneous and Heterogeneous
Catalytic Asymmetric Epoxidation. Chem. Rev. 2005, 105, 1603−
1662. (b) De Faveri, G.; Ilyashenko, G.; Watkinson, M. Recent
advances in catalytic asymmetric epoxidation using the environ-
mentally benign oxidant hydrogen peroxide and its derivatives. Chem.
Soc. Rev. 2011, 40, 1722−1760. (c) Zhu, Y.; Wang, Q.; Cornwall, R.
G.; Shi, Y. Organocatalytic Asymmetric Epoxidation and Aziridination
of Olefins and Their Synthetic Applications. Chem. Rev. 2014, 114,
8199−8256. (d) Davis, R. L.; Stiller, J.; Naicker, T.; Jiang, H.;
Jørgensen, K. A. Asymmetric organocatalytic epoxidations: reactions,
scope, mechanisms, and applications. Angew. Chem., Int. Ed. 2014, 53,
7406−7426. (e) Bryliakov, K. P. Catalytic Asymmetric Oxygenations
with the Environmentally Benign Oxidants H2O2 and O2. Chem. Rev.
2017, 117, 11406−11459.
(13) Braden, T. M.; Johnson, M. D.; Kopach, M. E.; McClary Groh,
J.; Spencer, R. D.; Lewis, J.; Heller, M. R.; Schafer, J. P.; Adler, J. J.
Development of a Commercial Flow Barbier Process for a
Pharmaceutical Intermediate. Org. Process Res. Dev. 2017, 21, 317−
326.
(14) Details of the magnesium charge schedule are noted in the
Experimental Section.
I
https://dx.doi.org/10.1021/acs.oprd.0c00052
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
pubs.acs.org/OPRD
Article
(15) Wong, S.-W.; Changi, S. M.; Shields, R.; Bell, W.; McGarvey,
B.; Johnson, M. D.; Sun, W.-M.; Braden, T. M.; Kopach, M. E.;
Spencer, R. D.; Flanagan, G.; Murray, M. Kinetics Model for
Designing Grignard Reactions in Batch or Flow Operations. Org.
Process Res. Dev. 2016, 20, 540−550.
(16) Reaction conversion in CSTR1 ranged from 90−95% upon
achieving a state-of-control during continuous operation. Empirical
evidence supported the inclusion of a second reactor to improve
reaction conversion; a CSTR was chosen instead of a plug-flow
reactor to minimize the risk of clogging if solid magnesium particles
were transferred from CSTR1. Detailed computational modeling was
performed to validate this approach and is the subject of a separate
communication.
(17) Tilstam, U.; Weinmann, H. Activation of Mg Metal for Safe
Formation of Grignard Reagents on Plant Scale. Org. Process Res. Dev.
2002, 6, 906−910.
(18) Environmental (E) factor = (kg of waste)/(kg of product). See:
Sheldon, R. A. The E Factor: fifteen years on. Green Chem. 2007, 9,
1273−1283.
(19) Beaver, M.; Cui, S.; Xianqing, S.Synthesis of (S)-2-Amino-4-
methyl-1-((R)-2-methyloxirane-2-yl)-pentan-1-one and Pharmaceuti-
cally Acceptable Salts Thereof. WO2018/027021A1, 2018.
J
https://dx.doi.org/10.1021/acs.oprd.0c00052
Org. Process Res. Dev. XXXX, XXX, XXX−XXX