Organic Process Research & Development
Article
oil or a thin-walled PFA tubing (1/16″ or 0.106″ i.d., 1/8″ o.d.)
enclosed in aluminum clamshell reactor plates. Size 1/4−28
nuts, tee-junctions, and unions all made of poly-ether-
etherketone (PEEK) were used for assembling the tubing
components of the reactors and feed lines (IDEX Health &
Science, Lake Forest, IL). Back pressure regulators (Model BPR-
10) from Zaiput Flow Technologies (Waltham, MA) were
applied at the outlets of step 2 and step 5 (175 psi each). An
assortment of either Milligat (Global FIA; Fox Island, WA) or
Eldex (Napa, CA) pumps were applied to pump reagents for all
experiments. For the continuous extraction, the final version for
the gravity separation unit was a custom 100 mL glass vessel
manufactured by Ace Glass Inc. (Vineland, NJ).
Analytical Procedure. Reaction stream samples were
collected after approximately five residence times for each
experiment and diluted in a mixture of acetonitrile and aqueous
Samples were then analyzed via high-performance liquid
chromatography (HPLC) with ultraviolet detection as
analysis carried out using liquid chromatography−mass
spectrometry (LCMS) also as described in the Supporting
Information Starting material stability for the stocks of acyl
chloride 1, acrylate 2, cyclopropylamine 4, and piperazine 8 was
performed with gas chromatography (GC) as described and
The process analytical technology (PAT) was applied in two
locations; an in-line infrared spectrometer equipped with a fiber-
optic probe was used after reaction step 2 to monitor
intermediate and product peaks (Mettler-Toledo, Columbus,
OH), and an in-line Raman spectrometer equipped with a flow
cell was applied to monitor the final outlet product stream
(Marqmetrix, Seattle, WA). An in-depth overview of the PAT
data and strategy is beyond the scope of this article, but as a vital
component of the project, a separate article is being prepared
with further analysis.
Software. All data from Design of Experiment (DOE)
studies were analyzed using the software JMP 15 (SAS Institute,
Cary, NC). Milligat pumps and temperature control for the
clamshell reactors were controlled via LabVIEW process control
architectures (National Instruments, Austin, TX).
significantly increased the LCAP (Table 1; runs 2, 3, 6, and 8).
The proposed impurity species 12 and 13 were also observed to
be sensitive to relative quantities of 1 in the stream as their
LCAP levels were significantly increased when equivalents of 1
to 2 were 1.6:1 (Table 1; runs 5 and 7), whereas when 2 was in
excess, species 12 and 13 were observed at their lowest LCAP
values (Table 1; runs 2, 3, 6, and 8). Effects on purity,
conversion, and impurity formation with regard to the step 1
reactor temperature were essentially negligible compared to
effects of varying concentrations and equivalents of 1 and 2.
Further experiments isolating temperature as a variable
confirmed that an intermediate temperature was observed to
maximize the purity and yield of 5 (see the Supporting
Information). For the second DOE study described below, the
selected conditions for step 1 were 2 at 1.2 M with 1.25 equiv of
DIPEA (with respect to 2) and 1 at 1.0 M, carried out at 150 °C
for 2 min.
The second study, which focused on step 2 (Table 2), showed
considerably less variation in LCAP and yield of 5 across
experiments when compared to step 1, with no significant
difference observed between runs. Some degree of variation was
expected when the concentration of 4 was reduced to 1 M,
effectively becoming equimolar with regard to the theoretical
step 1 outlet concentration of enamine 3, and an observed
significant effect was that species 13 was observed to have LCAP
values >1.0% at this set point (Table 2; runs 3, 5, 6, and 9). With
4 at 1.5 M, levels of 13 were substantially reduced to ∼0.1%;
however, species 12 was seen to increase in some cases (Table 2;
runs 2 and 10). The presumption that the reaction was rapid
enough to perform at room temperature was verified as no clear
trend with temperature was observed going from ambient
conditions to 65 °C. This short time frame additionally
indicated future potential for reactor volume reduction to
decrease overall system residence time and plausibly reduce time
for further side reactions to occur. Furthermore, the effect of
DIPEA was determined to be negligible and was therefore
removed as a reagent from step 2.
Based off these studies and various one-factor validation
experiments, the step 1 optimized conditions are 1.2 M of 2 with
1.15 equiv of DIPEA (with respect to 2) and 1.0 M of 1 heated at
150 °C for 2 min. For step 2, the conditions selected are 1.25 M
of 4 with no additional DIPEA at ambient temperature for 1.3
min. These conditions ultimately offered a purity by LCAP for 5
of 95 1% and an overall yield of 91 2% across two chemical
transformations in two reactors, with a 5 stream concentration of
95 3 mg/mL.
Development of CLLE for Step 3. In the original process,
the DMA 6 byproduct was removed by reaction with acetyl
chloride 7 as depicted above in Scheme 1. However, this
procedure carried forward the resulting N,N-dimethylacetamide
(not shown) through the process and further diluted the process
stream. Additionally, acetonitrile was used as the solvent for
steps 1−2 but is not a favorable solvent for the remaining two
steps due to solubility concerns with the intermediates formed
therein. The process enhancement implemented was a CLLE
unit operation (Figure 1) that performed three essential tasks:
(1) adequate removal of DMA 6 and other aqueous impurities,
(2) increasing the concentration of the reaction intermediate 5
in the organic stream, and (3) a solvent exchange to 2-
methyltetrahydrofuran (2-MeTHF) which is believed to be a
more suitable solvent for the subsequent reactions.
RESULTS AND DISCUSSION
■
Steps 1 and 2. Step 1 in this synthesis is an acylation to
produce keto-ester 3, followed by a rapid amine substitution
(step 2), providing cyclopropyl-enamine 5. Due to the high
reactivity of 3 and the rapid nature of step 2, these first two steps
were carried out without analysis of intermediate 3 after step 1,
assuming that any perturbations to the process caused by
changes to either step would be captured by analyzing the step 2
outlet stream. Two DOE19,20 studies were performed for these
synthesis steps: the first focused on step 1 variables, holding step
2 variables constant (Table 1), and the second focused on step 2
variables, holding constant the previously optimized step 1
conditions (Table 2).
A potential source of impurities were side reactions between
reaction intermediates and excess acyl chloride 1, cyclopropyl-
amine 4, or the DMA 6 species displaced from 3 in step 2. To
test this, the study was designed so that 1 was the limiting
reagent for several of the runs (Table 1; runs 2, 3, 6, and 8). The
study results with respect to purity of 5, represented as area
percent by liquid chromatography (LCAP), indicated that
ensuring that using acyl chloride 1 as the limiting reagent
Various designs for the CLLE unit operation were considered
including a continuous membrane separation;21 however, the
1527
Org. Process Res. Dev. 2021, 25, 1524−1533