Real-Time Monitoring of Solid-Liquid Slurries: Optimized Synthesis of Tetrabenazine

Yusuke Sato, Junliang Liu, Andrew J. Kukor, Je ﬀ rey C. Culhane, John L. Tucker, David J. Kucera,

Article

Real-Time Monitoring of Solid−Liquid Slurries: Optimized Synthesis

of Tetrabenazine

Brian M. Cochran, and Jason E. Hein*

Cite This: https://doi.org/10.1021/acs.joc.1c01098

Read Online

ACCESS

Metrics & More

Article Recommendations

sı Supporting Information

ABSTRACT: Solid−liquid slurries are vital and increasingly

prevalent in the pharmaceutical and chemical industries. Despite

the importance of these heterogeneous systems, process control

and optimization are fundamentally hindered by a lack of

compatible real-time analytical techniques. We present herein an

online HPLC monitoring platform enabling access to real-time

compositional information on slurries. We demonstrate the system

by investigating the heterogeneous synthesis reaction of

tetrabenazine. Furthermore, we integrated our online HPLC

platform with the orthogonal monitoring techniques of a pH

probe and a microscopic imaging probe to provide additional

mechanistic insight. These combined insights enable the

optimization of tetrabenazine synthesis in terms of reaction time,

byproduct formation, and diastereomeric purity of the ﬁnal product.

INTRODUCTION

Moreover, proving that an optimized synthetic process is

robust and reproducible is paramount for ﬁling and regulatory

approval.

■

Solid−liquid heterogeneous conditions are extensively used in

the chemical and pharmaceutical industries, particularly in

However, most conventional real-time reaction analytical

process chemical scale-up. These conditions can oﬀer high

reaction rates due to saturation, aﬀord signiﬁcant solvent

reductions, and in turn deliver improvements to process mass

intensity. Slurry batches are also readily telescoped into

isolation and puriﬁcation processes by leveraging crystalliza-

6−8

9,10

tools, such as NMR, IR,

Raman,

and mass spectrosco-

py face two major obstacles to implementation with solid−

liquid reaction slurries. First, these techniques cannot easily

deconvolute mixtures of reaction components such as starting

materials, intermediates, products, and byproducts. Second,

conventional tools are designed to study well-behaved systems

that tend to be homogeneous. Therefore, very limited

spectroscopic techniques are capable of understanding

component compositions of solid−liquid heterogeneous

tion. Finally, additional beneﬁts may be realized when the

targeted product has incompatible solubility with the reaction

mixture. In this case, crystallization enables the product to be

protected through phase segregation. For these reasons,

interest in slurry process batch reactions is growing relating

to chemical manufacturing and represents a correspondingly

signiﬁcant set of challenges facing industrial discovery and

process chemistry.

As new clinical candidates move forward, reﬁning and

optimizing a ﬁt-to-purpose synthetic scheme requires data-

intensive reaction information. Optimization relies on a

detailed understanding of how reaction outcomes (e.g., yield,

selectivity, time) are aﬀected by operational parameters (e.g.,

reactions.

HPLC analysis is an appealing technique due to its excellent

deconvolution capabilities, data richness, and abundance of

detection methods (e.g., UV, MSD, and ELSD). Adapting

this gold-standard analytical method to monitor solid−liquid

reaction slurries therefore presents a signiﬁcant opportunity.

To access this potential, we have previously reported the use of

HPLC platforms to monitor heterogeneous reactions. In our

reagent concentration, temperature, pH). Since the FDA

published its process analytical technology (PAT) guidelines in

004, real-time reaction monitoring has been highlighted as a

Special Issue: Enabling Techniques for Organic Syn-

thesis

means of providing data-rich reaction information for eﬃcient

experimental design. There is also further guidance that in

Received: May 10, 2021

order to adopt Quality by Design manufacturing practices,

data-rich experimentation should be every institution’s goal.

XXXX American Chemical Society

J. Org. Chem. XXXX, XXX, XXX−XXX

The Journal of Organic Chemistry

Article

ﬁrst iteration, the use of a Mettler-Toledo EasySampler and a

liquid handling robot enabled automated sampling, but HPLC

analysis was decoupled from the reaction and performed oﬀ-

line to allow for slurry dissolution. While this approach avoided

the manual labor endemic to oﬀ-line analysis, the intrinsic time

delay between sampling and analysis remained at play. This

time delay provides an opportunity for postsampling reactions

and/or the decay of short-lived intermediates, which can

Scheme 1. (a) Original Synthetic Procedure for (± )-TBZ

and (b) Potential Mechanistic Pathways to Access (± )-TBZ

15,16

compromise data integrity.

More recently, we developed an online HPLC platform to

delineate the reaction proﬁles of the dissolved components in a

heterogeneous Suzuki reaction containing poorly dissolved

CsCO₃salt.

Monitoring both the solid and liquid

components in this system, however, was hindered by two

major issues: compounds with low solubility risked line

clogging and the homogeneous and heterogeneous compo-

nents were found to have diﬀerent delivery kinetics during

sampling. We therefore identiﬁed the need for a sampling

method to eﬃciently dissolve heterogeneous components prior

to analysis. In this paper, we are reporting a new robust online

HPLC platform equipped with a dynamic mixer for

heterogeneous reaction monitoring.

shown in Scheme 1b (highlighted in blue/red and green,

respectively). Alternatively, elimination from the β-amino

ketone to generate enone 5 and dimethylamine 6 may precede

a concerted 4 + 2 cycloaddition (black pathway, Scheme 1b)

However, despite the industrial-scale production of TBZ and

the beneﬁts of understanding its reaction mechanism to

optimize synthetic procedures, a comprehensive mechanistic

study has not been reported to date.

RESEARCH TARGET

■

To test our platform, we chose to monitor the synthesis of

tetrabenazine (TBZ, 1 ; Figure 1 ). TBZ is an active

Herein, we present mechanistic insights into the synthesis of

TBZ under solid−liquid slurry conditions. These insights were

gained by combining online HPLC reaction monitoring with

an automated sampling platform. Though online HPLC

provides access to reaction composition data in heterogeneous

mixtures, it provides no data on two other aspects of the slurry

reaction. First, since Michael additions are typically catalyzed

by acids or bases, the pH could play an important role in this

reaction. Additionally, online HPLC involves heterogeneous

component dissolution by the diluent solvent; therefore, real-

time crystal information during the reaction is not obtained

(e.g., crystal shape, size, nucleation). Therefore, we applied

Mettler-Toledo’s pH probe and EasyViewer to complement

our compositional data obtained via HPLC and obtain a

comprehensive understanding of the TBZ synthesis mecha-

nism.

Figure 1. Structures of (±)-tetrabenazine and valbenazine.

pharmaceutical ingredient prescribed for the management of

neuromuscular symptoms arising from Huntington’s disease.

The TBZ derivative valbenazine (VBZ, 2; Figure 1) received

FDA approval in 2017 and is marketed as Ingrezza (Neuro-

crine Biosciences Inc., San Diego, CA) to treat tardive

19,20

dyskinesia, a similar neuromuscular disorder.

VBZ is a

particularly potent compound because it is metabolized in vivo

to the single, highly bioactive stereoisomer (+)-α-dihydrote-

trabenazine ((2R,3R,11bR)-HTBZ), while the metabolic

RESULTS AND DISCUSSION

■

21,22

pathway of TBZ is not stereoselective.

These factors

Development of Online HPLC for Slurries. Our initial

sampling platform (Figure 2) was based on modiﬁed hardware

adapted from a system previously reported for monitoring a

have incited interest in optimizing the synthesis of TBZ as a

route to VBZ.

(

±)-Tetrabenazine is synthesized by formal annulation

heterogeneous Suzuki reaction. Using this device, a slurry

aliquot is obtained from the reaction vessel and delivered to an

HPLC injection loop. Critically, our new system incorporates

an inline-dynamic mixer (Akta M-925 mixing cartridge, 0.6

mL) between the reaction sampler and injection valve, which

acts to both quench and dilute the sample with the push

solvent in preparation for HPLC analysis. Further, the dynamic

mixer serves to homogenize the heterogeneous slurry,

preventing line clogging while ensuring accurate analysis of

all reaction components (Figure S1 in the Supporting

Information). A comparison of the capability of the two

setups with or without a dynamic mixer was visualized by

generating TBZ mixture calibration curves (Figure 3). The

dynamic mixer installation allowed us to generate a linear

calibration curve of the slurry TBZ mixture without the clog.

between 3,4-dihydroisoquinoline 3 and β-amino ketone 4

Scheme 1). The optimized reaction conditions include

(

substituting the original ethanol solvent for 30% water in

isopropyl alcohol (IPA; v/v), allowing favorable reaction

kinetics and high diastereoselectivity to be achieved while the

reaction temperature is reduced from reﬂux to 45 °C. While

this modiﬁcation furnished an enhanced yield, reduced

byproducts, and overall advantageous process conditions,

TBZ exhibits remarkably low solubility (<0.13 M), yielding a

heterogeneous reaction mixture incompatible with conven-

tional reaction monitoring techniques.

Several mechanisms have been proposed for this formal 2 +

annulation reaction (Scheme 1a), including a Mannich−aza-

Michael pathway and an aza-Michael−Mannich pathway as

The Journal of Organic Chemistry

J. Org. Chem. XXXX, XXX, XXX−XXX

this reaction (Figure 4 ). The slopes of these trends represent

Article

Supporting Information). However, the standard addition

method requires accurate reaction volume measurements.

Speciﬁcally, the volume changes due to the addition of solids

were ambiguous. To this end, we conducted experiments to

test the volume changes of the solid reagents and product of

Figure 2. Schematic diagram of the automated sampling platform

used for reaction monitoring.

Figure 4. Relationship of volume changes against solid doses of TBZ,

isoquinoline 3, and NaI in a solution of 30% water in isopropyl

alcohol at room temperature.

the reciprocal of the density of each chemical. As expected, NaI

has the lowest slope, as its mass density was the highest, while

TBZ and isoquinoline 3 have higher slopes as organic

compounds. The trends of these chemicals are linear before

and after their solubility limits, allowing us to calibrate the

reaction volume using these slopes. Notably, the solid dose can

signiﬁcantly change the total volume of the reaction. Under the

conditions of this reaction, the total volume increases 12.7%

due to the solid dose of isoquinoline and NaI. Taking these

considerations into account and using the measured volume

calibrations (Figure 4), the standard addition method enabled

us to convert our raw signal vs time data to concentration vs

time trends.

Figure 3. Calibration curves of TBZ generated using sampling

platforms with (a) or without (b) a dynamic mixer. TBZ precipitates

at concentrations above 0.13 M in a solution of 30% water in

isopropyl alcohol at 45 °C.

However, when we tested this system to sample a slurry

mixture consisting of isoquinoline 3 as a homogeneous

component and TBZ as a heterogeneous component, we

observed diﬀerent delivery kinetics (Figure S4a in the

Supporting Information). This indicated that the dynamic

mixer could not dissolve the heterogeneous component

immediately. Therefore, a wait time is required to achieve a

complete dissolution after pushing a certain volume (referred

to as the “mixer volume”) to deliver the heterogeneous sample

aliquot into the dynamic mixer. To this end, we varied the

amount of mixer volume followed by a 30 s wait time. Since

the measured volume between the EasySampler pocket to

dynamic mixer is 0.225 mL, we observed that less than 0.2 mL

mixer volume was not enough to obtain the same delivery

kinetics both of homogeneous and heterogeneous components.

In contrast, using 0.4 mL as the mixer volume ﬁxed this

inconsistency and the two phases followed the same delivery

kinetics after 0.6 mL dilution volume (Figure S4e in the

Supporting Information).

Reaction Monitoring. First, we monitored the reaction

under previously identiﬁed conditions, which involves reacting

isoquinoline salt 3 with β-amino ketone 4 using NaI in iPrOH/

H O (Figure 5). Reaction monitoring by online HPLC

revealed three new components: the major product TBZ

Standard Addition. Now we are capable of collecting

HPLC spectra of heterogeneous reactions using our robust

sampling platform. We further need to transform peak area

into concentration according to a Beer’s law plot. In most

cases, precalibration of all the reaction components were

needed to make quantitative measurements (oﬀ-line calibra-

tion curve). This requires multiple data points of concentration

vs signal to acquire the slope. Instead, we took advantage of

our in situ reaction monitoring technique and used standard

Figure 5. Reaction proﬁle under the initial conditions. Reaction

conditions: [3·HCl] = 0.64 M, [4] = 0.8 M, [NaI] = 0.25 M in a

solution of 30% water in isopropyl alcohol at 45 °C.

addition as an internal calibration method to transform signal

peak area) into concentrations (Figures S6 and S7 in the

(

J. Org. Chem. XXXX, XXX, XXX−XXX

The Journal of Organic Chemistry

1), the minor product diastereomer 7, and the intermediate

with enone 5 and dimethylamine 6 (Figure 6 ). Interestingly,

Article

(

enone 5. The presence of enone 5 as a steady-state

intermediate conclusively demonstrates that the reaction is

initiated via elimination of dimethylamine (DMA, 6), as

opposed to direct C−C bond formation via a Mannich-like

coupling involving a substrate enolate (blue pathway, Scheme

). Moreover, the relatively high concentration of enone 5

(

∼0.2 M) suggests that the elimination is rapid and not the

rate-determining step.

The productive intermediacy of enone 5 was conﬁrmed by

repeating the synthesis of TBZ, but replacing β-amino ketone 4

Figure 7. TBZ formation trends using diﬀerent initial concentrations

of enone 5 and DMA 6. Reaction conditions: [3·HCl] = 0.64 M, [5]

0.8 or 0.4 M, [6] = 0.8 or 0.4 M, [NaI] = 0.25 M in a solution of

30% water in isopropyl alcohol at 45 °C.

utilized, a decreasing pH trend, from 9 to 8, was observed

when the reaction was performed with enone 5 and DMA 6.

This latter pH trend, dropping from 9 to ∼8, was also observed

when the initial concentration of enone 5 was decreased to 0.4

M. When they are taken together, these observations suggest

that the reaction displayed a positive kinetic order with respect

to enone 5 but that the rate of the reaction is suppressed at

high pH. This antagonistic behavior between enone 5 and pH

is responsible for the fortuitous similarity in reaction proﬁles

for the two experiments (Figure 6).

Figure 6. Reaction proﬁle substituting β-amino ketone 4 with enone 5

and DMA 6. Reaction conditions: [3·HCl] = 0.64 M, [5] = 0.8 M, [6]

0.8 M, [NaI] = 0.25 M in a solution of 30% water in isopropyl

alcohol at 45 °C.

This hypothesis spurred us to modify the reaction pH by

reducing the initial concentration of DMA 6 to 0.4 M, while

the reaction proﬁle and rate of TBZ formation closely parallel

the optimized conditions initiated with β-amino ketone 4

conditions. This equivalence in reaction rates implies that

enone 5 exerts no kinetic driving force on the overall reaction

cascade and could be classiﬁed as possessing a zeroth order in

the overall reaction. This conclusion is borne out by the

identical rates of TBZ generation over the ﬁrst 10 h, despite

the two reactions having dramatically diﬀerent concentrations

of the enone 5 within this time regime.

setting [5] at 0.8 M. In line with our expectations, the initial

pH of the reaction dropped to ∼7 and displayed a similar

decrease in pH over the course of the reaction, resulting in a

solution at pH ∼5.5 at the reaction’s end. More importantly,

this modiﬁcation also led to a 5-fold increase in the reaction

rate over the initial test conditions (Figure 7, red vs purple

trend). Further investigation revealed that an initial pH of ∼7

was ideal. In fact, removing DMA entirely, which gave a

solution at an initial pH of ∼3.3, nearly arrested the formation

of TBZ entirely (Figure 7, red vs green trend). Finally, by

To unambiguously conﬁrm the reaction order for enone 5,

its initial concentration was varied. Reducing [5] to 0.4 M

decreasing [5] to 0.4 M, we observed a concomitant reduction

from 0.8 M resulted in a concomitant reduction in the rate of

product formation (Figure 7). This suggests that enone 5 has a

positive order in this reaction, but there are notable deviations

from ideal ﬁrst-order behavior. In particular, the initial rate of

the reaction is identical for T = 0−1.5 h then deviates sharply

after this time point, indicating that the initial concentration of

enone does exert a signiﬁcant role in the observed rate of the

reaction and thus is coupled to the rate-limiting step but that

other factors are still at play.

in the reaction rate, conﬁrming that the overall rate of product

formation is sensitive to the concentration of enone 5 and thus

is likely involved in the rate-limiting step.

While executing the reaction without added DMA did not

produce an appreciable quantity of TBZ, a new intermediate

was observed. After isolation and characterization, this species

was conﬁrmed to be iminium 8 (Scheme 2), which is likely

formed via an acid-catalyzed aza-Michael addition between

isoquinoline 3 and enone 5.

The observation of this complex behavior spurred us to

reexamine our initial reaction proﬁles from Figures 5 and 6.

While the rates of product formation were nearly identical, we

noted that the pH vs time trends between reaction 1 in Figure

Scheme 2. Formation of Iminium Intermediate 8

(with β-amino ketone 4) and reaction 2 in Figure 6 (with

enone 5 and DMA 6) showed signiﬁcant diﬀerences. While the

pH increased from 7 to 8 when the β-amino ketone 4 was

J. Org. Chem. XXXX, XXX, XXX−XXX

The Journal of Organic Chemistry

Article

After acquiring an authentic sample of this intermediate, we

were able to conﬁrm that iminium 8 is detectable at low

concentrations in all of the reactions. These data conﬁrm that

the synthesis of TBZ likely proceeds via a sequential aza-

Michael addition, followed by annulation involving the

corresponding enol or enolate (Scheme 1, green pathway).

Moreover, we found that this iminium intermediate is readily

cyclized via treatment with various amines. However, this

directly opposes the observation that the product formation

accelerated with decreasing initial amine concentration from

proposed mechanism, a derived rate law including information

in terms of pH, the pK of isoquinoline, and the pK of HA is

therefore necessary. This can be achieved by assuming the two

major acid−base processes involving the conjugate acid of

isoquinoline 3 (3·H ) and HA are in quasi-equilibrium with

one another, and thus we can use their K values:

[

3][H ]

K =

[

3·H ]

(2)

(3)

−

[

A ][H ]

.8 to 0.4 M. This result suggests that the ﬁnal cyclization step

K_a₂

is not the rate-determining step, as decreasing the initial amine

concentration would slow down product formation if this were

the case. From this, we propose that the rate-determining step

of the TBZ 2 + 4 annulation is instead the aza-Michael

addition.

[HA]

The conservation of mass for both isoquinoline and the

general acid provide their mass balance equations:

[3]_t_o_t_a_l= [3] + [3·H⁺]

(4)

Eﬀect of pH on Michael Addition. We observed that this

reaction has an optimal pH of around 7, as the reaction rate

was observed to decrease when the pH was both lower and

higher than 7. This suggests that the rate-determining step (the

aza-Michael addition) requires both the deprotonation of a

[A]total = [HA] + [A⁻]

(5)

Combining the deﬁnitions for the K values with the

conservation law results in the following expressions:

nucleophile (i.e., the isoquinoline HCl salt, pK = 5.8) as well

[3]_t_o_t_a_lK_a₁

[

3] =

as the protonation of an electrophile (i.e., the enone 5) or of a

general acid. Since the oxygen atom on the enone 5 is known

K + [H ]

(6)

(7)

to have a very low pK (around −2), it is unlikely that it will be

[A]_t_o_t_a_l[H ]

[

HA] =

protonated at the optimal pH of 7. We therefore hypothesize

that the aza-Michael addition step is catalyzed by some general

K_a₂+ [H ]

acid (HA) with a p K higher than 7 (Figure 8 ). Balancing the

Finally, combining these equations (eq 6 and 7) with the

reaction rate law (eq 1) allows for the rate to be expressed

strictly in terms of the pH and pK of species present:

[

3]_t_o_t_a_lK_a₁[A]_t_o_t_a_l[H ]

rate = k [5]

K + [H ] K + [H ]

(8)

K_a₁

[H ]

rate = k [5][3]total^[^A^]

total

K + [H ] K + [H ]

K_a₁

[H ]

k′

K + [H ] K + [H ]

(9)

This allowed for us to model the reaction rate on the basis of

the solution pH and pK of the nucleophile and general acid.

This model (Figure 8) showed excellent agreement with our

observed reaction rates at various pHs, strongly supporting our

proposed mechanism. Furthermore, if the reaction occurs at a

very low pH, [H ] ≫ K and K and so eq 8 simpliﬁes into a

Figure 8. Proposed trimolecular transition structure and modeled pH

proﬁle: the initial rate, free isoquinoline ratio, and free HA ratio were

more succinct rate law:

plotted. The initial rate was calculated by rate = k′*(K /

_r_a_t_e₌_k_′_[K

[H] ))*([H ]/(K [H] )). When we applied pK = 5.8 for

H ]

(10)

(11)

the conjugate acid of isoquinoline and pK = 8.5 for HA, the modeled

pH proﬁle ﬁt the experimental data well.

^l^o^g^r^a^t^ehigh pH = log k′ − pK + pH

Similarly, if the reaction occurs at a very high pH, [H ] ≪

protonation of the general acid with the deprotonation of the

nucleophile (isoquinoline) therefore leads to the observed

K_a₁and K and thus an alternate simpliﬁed equation is

obtained:

optimal pH of 7 (modeled in Figure 8, with the pK of HA

being 8.5). This modeled behavior closely ﬁts the observed

rates for the various pHs tested.

The rate equation for the aza-Michael addition can therefore

be expressed as

[

H ]

rate = k′

K_a₂

(12)

(13)

^l^o^g^r^a^t^elow pH = log k′ + pK_a₂− pH

rate = k [enone 5][isoquinoline 3][HA]

(1)

This hypothesis can explain the reaction proﬁles. When the

pH is high (around 9), most of the general acid HA is

deprotonated. This would slow down the aza-Michael addition.

However, this equation shows no immediate relation to the

reaction pH or the pK of any species. To support our

J. Org. Chem. XXXX, XXX, XXX−XXX

The Journal of Organic Chemistry

both 0.8 and 0.4 M conditions (Figure 9 ).

Article

On the other hand, when pH is even lower (approximately

Scheme 3. Proposed Reaction Acceleration Mechanism

under Diﬀerent Amine Concentrations with DMA 6 and

DIPEA

3.3), the isoquinoline 3 is fully protonated, thus impeding the

ability of the isoquinoline 3 to nucleophilically attack the

enone 5.

Secondary vs Tertiary Amine. A secondary amine such

as dimethylamine 6 can catalyze the reaction as either a simple

base catalyst or an organocatalyst through enamine or iminium

chemistry. To investigate the catalytic manner, we also tried

N,N-diisopropylethylamine (DIPEA) as a tertiary amine under

was a relationship between the crystallization and TBZ ratio

over the diastereomers 7 (Figure 10 ). This ratio could be

Figure 10. Trends of TBZ ratio over TBZ and TBZ diastereomers at

diﬀerent amine concentrations with DMA or DIPEA under the same

conditions as in Figure 7.

aﬀected by kinetic control of the Mannich cyclization step,

epimerization between TBZ and TBZ diastereomers 7, and

crystallization of TBZ (Scheme 4). The initial ratio indicated

that the kinetic control of the Mannich reaction preferably

generated TBZ. This was probably due to the 1,3-diaxial

interaction of the isobutyl group presented in the assumed

transition state to form the diastereomers 7. The data showed

Figure 9. (a) TBZ formation and (b) pH trends using diﬀerent initial

amine concentrations with DMA 6 and DIPEA. Reaction conditions:

[

3·HCl] = 0.64 M, [5] = 0.8 or 0.4 M, [amine] = 0.8 or 0.4 M, [NaI]

0.25 M in a solution of 30% water in isopropyl alcohol at 45 °C.

Scheme 4. Proposed Kinetic Control of the Mannich

Cyclization Step via Chairlike Transition State and

Thermodynamic Epimerization Mechanisms

Interestingly, the observed TBZ formation rate was

signiﬁcantly lower with the 0.8 M tertiary amine, suggesting

that a secondary amine accelerates the reaction through

iminium chemistry (Scheme 3). In contrast, with 0.4 M amine,

the TBZ formation trend and pH proﬁle with DIPEA were

identical with those with dimethyl amine. This implied that the

Michael addition is simply catalyzed by protonation rather

than iminium formation at lower pH.

Crystallization. While online HPLC demonstrated the

capability of both heterogeneous and homogeneous compo-

nent proﬁles, EasyViewer provided a method to measure the

physical properties of reaction mixtures such as dissolution,

primary crystallization, and crystal growth. We found that the

primary crystallization occurred when the TBZ concentration

reached approximately 0.25 M. Moreover, we found that there

J. Org. Chem. XXXX, XXX, XXX−XXX

The Journal of Organic Chemistry

Article

that all reactions had a 96% initial ratio except for 91%

observed under 0.4 M DMA conditions. This suggested that

the ﬁnal cyclization proceeded through diﬀerent mechanisms

under 0.4 M DMA conditions (enamine assisted). On the

other hand, TBZ and its diastereomers 7 can interchange with

each other through an epimerization process. The epimeriza-

tion equilibrium was a result of thermodynamic control. It was

legitimate to assume that the epimerization equilibrium would

stop at a diﬀerent TBZ/diastereomer ratio from the kinetically

controlled result of the Mannich reaction. Theoretically, the

TBZ/diastereomer ratio would be driven to the equilibrium

point by the diﬀerence in concentrations. Under 0.8 M DIPEA

conditions, epimerization from TBZ to diastereomers 7

resulted in a decreasing TBZ ratio over the reaction. This

result implied that the epimerization was a slow process,

requiring over 20 h. In contrast, reactions with 0.8 M DMA

and 0.4 M DIPEA followed the same decreasing trends until

the crystallization of TBZ. The crystallization signiﬁcantly

suppressed the epimerization or inverted the epimerization

direction. This was because the crystallization process dropped

the TBZ concentration from the supersaturation state to the

saturation state. Moreover, while the produced TBZ

immediately precipitated out, the diastereomer 7 concentration

kept increasing. The ﬂat trend seen under 0.4 M DMA

conditions illustrated that either there was no epimerization or

it already reached the equilibrium.

Figure 11. TBZ formation trends with or without NaI under reaction

or reaction 2.

Overall, we found that the epimerization process decreased

the kinetically obtained TBZ/diastereomer ratio. This process

required time to reach the equilibrium and was signiﬁcantly

suppressed by crystallization events. Therefore, crystallization

at an early reaction time can increase the ratio.

Role of Sodium Iodide. Ions in aqueous solutions have an

important role in the solubility of organic compounds.

Typically, this phenomenon is known as salting out

undissolved solid. We applied this platform with orthogonal

monitoring techniques to reveal comprehensive information on

slurry TBZ synthesis reactions from 3,4-dihydroisoquinoline 3

and β-amino ketone 4. The identiﬁcation of two important

intermediates, the enone 5 and the iminium 8, supported that

this reaction proceeds via an aza-Michael−Mannich pathway

(

Scheme 5). The reaction begins by the rapid formation of the

(

decreasing the solubility of solute in water) and salting in

increasing the solubility of solute in water) and is utilized for

Scheme 5. Proposed Mechanism of Acid-Catalyzed TBZ

Synthesis

puriﬁcation and extraction. While chloride ion is used as a

salting-out reagent due to its hardness, iodide is known to

increase the solubility of organic compounds in organic media

due to its high polarizability. We predicted that the addition of

NaI into the TBZ synthesis reaction changes the counteranion

for the isoquinoline from Cl⁻to I , which increases the

isoquinoline’s solubility. This change in solubility was

conﬁrmed by using the Mettler-Toledo EasyViewer. Upon

elimination of NaI isoquinoline remained insoluble when the

reaction was initiated, thus reducing the solution-phase

concentration of 3.

−

While the presence or absence of NaI revealed minor

changes in the observed reaction rates (Figure 11), the nature

of reagents used in the reaction also had an eﬀect. While

adding NaI appears to marginally accelerate the rate when

enone 5 is employed (Figure 11, blue vs yellow), the opposite

is evident when amino ketone 4 is used (Figure 11, purple vs

green). This observation may indicate that NaI has a minor but

complicated role when the reaction is performed on a small

bench scale.

enone intermediate 5 and dimethylamine 6. The following aza-

Michael addition forms the iminium intermediate 8. The fact

that there is an optimal pH supports that this step is the rate-

determining step. The subsequent cyclization step from

iminium intermediate 8 to TBZ is facilitated by a base. By

tuning the reaction with diﬀerent concentrations of amines, an

optimal amine concentration of 0.4 M demonstrated a 5-fold

increase in the reaction rate over previously optimized TBZ

syntheses. Furthermore, assuming that the aza-Michael

addition is the rate-determining step in this reaction, we

explained the bell-shaped pH proﬁle with a model. On the

basis of this model, the aza-Michael addition has an optimal

CONCLUSION

■

In summary, we have developed a robust sampling platform

enabling us to monitor heterogeneous components in a slurry

reaction. The installation of a dynamic mixer enabled the

preparation of a well-mixed homogeneous sample prior to

sample loading and prevented the clogging of lines by the

J. Org. Chem. XXXX, XXX, XXX−XXX

The Journal of Organic Chemistry

Article

pH of 7 derived from the pK values of isoquinoline (5.8) and

HA (8.5).

In addition to online HPLC analysis, with the assistance of

the Mettler-Toledo EasyViewer, we found that NaI helped the

isoquinoline dissolution in aqueous solvent, and the crystal-

lization process signiﬁcantly changed TBZ ratio over its

diastereomers 7.

The solid remaining was washed with Et

methylpropyl)-3-oxobutyl]azanium iodide (43.5 g, 92.6% yield) as a

white crystal.

Procedure for the Synthesis of 5. Trimethyl[2-(2-methyl-

propyl)-3-oxobutyl]azanium iodide (43.5 g, 233 mmol) was charged

O to aﬀord trimethyl[2-(2-

in water (350 mL) and Et O (466 mL). To the stirred solution was

added 10 M KOH (140 mL, 1.40 mol). The mixture was stirred

vigorously for 18 h at rt and then was transferred to a separatory

funnel. The mixture was washed with water (3 × 300 mL). The

organic layer was dried over Na SO and concentrated in vacuo to

EXPERIMENTAL SECTION

■

aﬀord 5-methyl-3-methylidenehexan-2-one (18.3 g, 62.1% yield) as a

colorless oil.

The 6,7-dimethoxy-1,2,3,4-tetrahydroquinoline hydrochloride salt 3

and 3-[(dimethylamino)methyl]-5-methylhexan-2-one oxalate 4 were

supplied by Neurocrine Biosciences. Optima-grade UHPLC solvents

and sodium iodide were purchased from Fisher. Iodomethane was

purchased from Alfa Aesar. Potassium hydroxide, dimethylamine

solution, and N,N-diisopropylethylamine were purchased from Sigma.

All other reagents and solvents were purchased from conventional

suppliers. All reagents and solvents were used as received. NMR

spectra were recorded on a Bruker Avance III 300 (300 and 75 MHz

Procedure for the Synthesis of rac-7. In a 100 mL EasyMax

ﬂask equipped with a stir bar was placed the crude, unﬁltered reaction

mixture from a TBZ 1 synthesis reaction. The mixture was stirred

vigorously at 60 °C (temperature controlled by an EasyMax

theromowell) for 4 days to produce the diastereomers and then

cooled to 5 °C at 0.1 K/min and ﬁltered to separate out the solid

TBZ. The ﬁltrate containing TBZ and the TBZ diastereomers then

had sthe olvent removed in vacuo, and the residue was dissolved in 30

mL of DCM. Silica gel (5 g) was then added the and solvent again

removed in vacuo. The dry-loaded solids were then puriﬁed via ﬂash

column chromatography (0−70% EtOAc in hexanes) using a

RediSepRf gold column (24 g) to aﬀord 30 mg of TBZ diastereomers

for H and C, respectively) or a Bruker AV III HD 400 spectrometer

(

400 and 100 MHz for H and C, respectively) and were calibrated

with the solvent (CDCl : 7.26 ppm for H NMR and 77.16 ppm for

C NMR). The abbreviations s, d, t, and m signify singlet, doublet,

triplet, and multiplet, respectively. NMR spectra were analyzed by

using the software MNova. High-resolution mass spectrometry data

were acquired by the UBC Department of Chemistry Mass

Spectrometry Center.

General Procedure for Reaction Monitoring. Monitored

experiments were performed with a Mettler-Toledo EasyMax 102

Advanced Thermostat System to enable online control of temperature

and stirring rate via iControl software. A Mettler-Toledo 50 mL one-

piece glass reactor was used, allowing the attachment of multiple

monitoring equipments with ﬁve openings.

(

30 mg, 0.4%) as an oﬀ-white solid.

Procedure for the Synthesis of 8. In a 25 mL three-necked,

pear-shaped ﬂask equipped with a stirrer bar were placed IPA (5.1

mL), water (2.1 mL), 5-methyl-3-methylidenehexan-2-one 4 (1.01 g,

8.00 mmol), 6,7-dimethoxy-1,2,3,4-tetrahydroquinoline hydrochloride

salt 3 (1.46 g, 6.40 mmol), and NaI (384 mg, 2.56 mmol). The

mixture was stirred vigorously for 24 h at 45 °C (temperature

controlled by an EasyMax theromowell) and then was transferred to a

separatory funnel. The mixture was extracted with DCM (100 mL)

and washed with water (3 × 100 mL), which retains most of

isoquinoline salt 3 in the water layer. The organic layer was

concentrated in vacuo to remove the remaining enone 5.

Recrystallization with AcOEt gave 6,7-dimethoxy-2-[2-(2-methyl-

propyl)-3-oxobutyl]-3,4-dihydroisoquinolin-2-ium salt (54 mg, 2.7%

In a 50 mL EasyMax ﬂask equipped with a stir bar were placed

isopropyl alcohol (IPA), water, and NaI. Isoquinoline 3 was added in

two portions as standard additions followed by β-amino ketone 4 (or

enone 5/amine) as a single portion. The mixture was stirred

vigorously at 45 °C (temperature controlled by an EasyMax

theromowell) under monitoring with online HPLC, EasyViewer,

and a pH probe. After the reaction was judged to be completed, rac-1

and enone 5 were added in two portions as standard additions.

An online HPLC sample was obtained using a modiﬁed reaction

monitoring platform similar to that which has been reported in our

yield) as yellow crystals.

Tetrabenazine (TBZ) (1): H NMR (400 MHz, CDCl ) δ 6.60 (s,

H), 6.54 (s, 1H), 3.84 (s, 3H), 3.81 (s, 3H), 3.49 (d, J = 11.5 Hz,

H), 3.27 (dd, J = 11.5 Hz, 6.3 Hz, 1H), 3.15−3.03 (m, 2H), 2.88

(

dd, J = 13.5, 3.0 Hz, 1H), 2.79−2.67 (m, 2H), 2.63−2.47 (m, 2H),

2.34 (t, J = 11.6 Hz, 1H), 1.84−1.73 (m, 1H), 1.70−1.59 (m, 1H),

group previously. Acetonitrile was used as the diluent solvent.

13 1

1.07−0.96 (m, 1H), 0.89 (dd, J = 5.4, 4.6 Hz, 6H); C{ H} NMR

101 MHz, CDCl ) δ 210.1, 147.9, 147.6, 128.5, 126.1, 111.5, 107.9,

Procedure for the Synthesis of rac-1. In a 25 mL three-

(

necked, pear-shaped ﬂask equipped with a stirrer bar were placed IPA

2.5, 61.5, 56.1, 56.0, 50.7, 47.6, 47.6, 35.1, 29.4, 25.5, 23.3, 22.2; MS

(5.1 mL), water (2.1 mL), 3-[(dimethylamino)methyl]-5-methylhex-

ESI (calcd for C H NO , 317.20) m/z 318.2 [M + H] .

an-2-one 4 (1.37 g, 8.00 mmol), 6,7-dimethoxy-1,2,3,4-tetrahydro-

quinoline hydrochloride salt 3 (1.46 g, 6.40 mmol), and NaI (384 mg,

19 27

-[(Dimethylamino)methyl]-5-methylhexan-2-one (4): H NMR

(

400 MHz, CDCl ) δ 2.82−2.70 (m, 1H), 2.57 (t, J = 11.6 Hz, 1H),

.56 mmol). The mixture was stirred vigorously for 24 h at 45 °C

.13 (s, 3H), 2.17−2.10 (m, 1H), 2.12 (s, 6H), 1.56−1.40 (m, 2H),

(temperature controlled by an EasyMax theromowell). After 24 h,

13 1

.25−1.14 (m, 1H), 0.88 (t, J = 7.0 Hz, 6H); C{ H} NMR (101

IPA was added to the mixture until the yellow precipitate was

completely dissolved at 70 °C, and then the reaction mixture was

cooled to rt to aﬀord tetrabenazine (1.20 g, 59.0%) as white crystals.

MHz, CDCl ) δ 212.3, 62.4, 49.7, 45.8, 39.5, 28.4, 26.2, 23.1, 22.3;

MS ESI (calcd for C H NO, 171.16) m/z 172.2 [M + H] .

10 21

Trimethyl[2-(2-methylpropyl)-3-oxobutyl]azanium iodide: ¹H

Procedure for the Salt Braking of 4. 3-[(Dimethylamino)-

NMR (400 MHz, acetone-d ) δ 4.22−4.06 (m, 1H), 3.71−3.60 (m,

methyl]-5-methylhexan-2-one oxalate (41.9 g, 160 mmol) was

H), 3.38 (s, 9H), 2.41 (s, 3H), 1.92−1.75 (m, 1H), 1.66−1.47 (m,

suspended in Et O (80 mL) and water (200 mL). A 2 M KOH

H), 1.42−1.25 (m, 1H), 1.05 (d, J = 6.5 Hz, 3H), 0.96 (d, J = 6.6

solution (200 mL, 400 mmol) was charged, generating a white

precipitate. The mixture was stirred vigorously for 1 h at rt and then

Hz, 3H); MS ESI (calcd for C H NO , 186.19) m/z 186.2 [M] .

11 24

-Methyl-3-methylidenehexan-2-one (5): H NMR (400 MHz,

was transferred to a separatory funnel with Et O (200 mL). The

CDCl ) δ 5.98 (s, 1H), 5.71−5.65 (m, 1H), 2.28 (s, 3H), 2.09 (dd, J

mixture was washed with water (2 × 200 mL). The organic layer was

7.2, 1.1 Hz, 2H), 1.74−1.61 (m, 1H), 0.81 (d, J = 6.7 Hz, 6H);

dried over Na SO and concentrated in vacuo to aﬀord 3-

C{ H} NMR (101 MHz, CDCl ) δ 200.0, 148.3, 125.9, 40.1, 27.2,

[

(dimethylamino)methyl]-5-methylhexan-2-one (25.8 g, 93.9%

yield) as a colorless oil.

26.0, 22.4.

Procedure for the Salt Braking of Trimethyl[2-(2-methyl-

TBZ diastereomers (7): H NMR (400 MHz, CDCl ) δ 6.62 (s,

propyl)-3-oxobutyl]azanium Iodide.

3-[(Dimethylamino)-

1H), 6.56 (s, 1H), 3.86 (s, 3H), 3.83 (s, 3H), 3.44 (br d, J = 11.5 Hz,

1H), 3.19−3.08 (m, 1H), 3.07−2.93 (m, 2H), 2.82 (ddd, J = 14.2,

3.3, 1.5 Hz, 1H), 2.72 (dd, J = 11.6, 4.0 Hz, 1H), 2.68−2.59 (m, 2H),

2.58−2.49 (m, 2H), 1.85−1.73 (m, 1H), 1.60−1.46 (m, 2 H), 0.93

methyl]-5-methylhexan-2-one 4 (25.7 g, 150 mmol) was charged in

methanol (150 mL). To the stirred solution was added iodomethane

(

25.56 g, 180 mmol) dropwise. The mixture was stirred vigorously for

h at rt, and then the solvent was removed under reduced pressure.

(d, J = 6.2 Hz, 3H), 0.88 (d, J = 6.3 Hz, 3H); C{ H} NMR (100

J. Org. Chem. XXXX, XXX, XXX−XXX

The Journal of Organic Chemistry

S.; Orr, J. D.; Reid, G. L.; Thompson, D. R.; Ward, H. W. Industry

Article

MHz, CDCl ) δ 212.6, 147.7, 147.7, 128.9, 126.6, 111.4, 107.8, 62.2,

Engineering Research Council of Canada (RGPIN-2016-

9.8, 56.1, 56.0, 51.8, 49.2, 45.4, 40.7, 29.4, 25.9, 22.8, 22.4; HRMS

ESI-TOF) m/z [M + H] calcd for C H NO 318.2069, found

04613, NSERC CGS-M to A.J.K.).

(

19 28 3

18.2065.

REFERENCES

■

,7-Dimethoxy-2-[2-(2-methylpropyl)-3-oxobutyl]-3,4-dihydroi-

(1) Chanda, A.; Daly, A. M.; Foley, D. A.; LaPack, M. A.; Mukherjee,

soquinolin-2-ium (8): H NMR (400 MHz, CDCl ) δ 9.67 (s, 1H),

Perspectives on Process Analytical Technology: Tools and Applica-

tions in API Development. Org. Process Res. Dev. 2015 , 19 , 63 −83.

(2) Gherras, N.; Fevotte, G. On the Use of Process Analytical

Technologies and Population Balance Equations for the Estimation of

Crystallization Kinetics. A Case Study. AIChE J. 2012, 58 (9), 2650.

3) Yu, L. X.; Lionberger, R. A.; Raw, A. S.; Costa, R. D.; Wu, H.;

Hussain, A. S. Applications of Process Analytical Technology to

Crystallization Processes. Adv. Drug Delivery Rev. 2004, 56 (3), 349−

.60 (s, 1H), 6.85 (s, 1H), 4.22 (d, J = 8.9 Hz, 2H), 4.05 (t, J = 7.8

Hz, 2H), 3.98 (s, 3H), 3.89 (s, 3H), 3.61−3.50 (m, 1H), 3.23 (t, J =

.1 Hz, 2H), 2.32 (s, 3H), 1.74−1.60 (m, 2H), 1.41−1.33 (m, 1H),

13 1

.00 (d, J = 6.2 Hz, 3H), 0.95 (d, J = 6.3 Hz, 3H); C{ H} NMR

(101 MHz, CDCl ) δ 210.0, 165.5, 158.0, 149.1, 132.5, 117.3, 116.1,

10.8, 59.9, 57.0, 56.9, 50.0, 49.8, 39.0, 29.7, 26.2, 26.0, 23.0, 23.0;

(

HRMS (EI-TOF) m/z [M] calcd for C H NO 318.2069, found

318.2059.

369.

ASSOCIATED CONTENT

■

4) Chen, Z.; Lovett, D.; Morris, J. Process Analytical Technologies

sı Supporting Information

The Supporting Information is available free of charge at

https://pubs.acs.org/doi/10.1021/acs.joc.1c01098.

and Real Time Process Control a Review of Some Spectroscopic

Issues and Challenges. J. Process Control 2011 , 21 (10), 1467 −1482.

(5) Gomez, M. V.; De La Hoz, A. NMR Reaction Monitoring in

Flow Synthesis. Beilstein J. Org. Chem. 2017, 13, 285−300.

(6) Foth, P. J.; Malig, T. C.; Yu, H.; Bolduc, T. G.; Hein, J. E.;

Sammis, G. M. Halide-Accelerated Acyl Fluoride Formation Using

Sulfuryl Fluoride. Org. Lett. 2020, 22 (16), 6682−6686.

Details of reaction monitoring and NMR data ( H

NMR, C NMR) (PDF)

AUTHOR INFORMATION

(7) Barrios Sosa, A. C.; Conway, R.; Williamson, R. T.; Suchy, J. P.;

Edwards, W.; Cleary, T. Application of PAT Tools for the Safe and

Reliable Production of a Dihydro-1 H -Imidazole. Org. Process Res.

Dev. 2011, 15 (6), 1458 −1463.

■

Corresponding Author

Jason E. Hein − Department of Chemistry, The University of

British Columbia, Vancouver, British Columbia V6T 1Z1,

Canada; orcid.org/0000-0002-4345-3005;

Email: jhein@chem.ubc.ca

8) Leadbeater, N. E. In Situ Reaction Monitoring of Microwave-

Mediated Reactions Using IR Spectroscopy. Chem. Commun. 2010,

46 (36), 6693−6695.

9) Johansen, L. N.; Kloster, L.; Andreasen, A.; Kucheryavskiy, S.;

(

Authors

Nielsen, R. P.; Maschietti, M. Raman Spectroscopy for Monitoring

Aqueous Phase Hydrogen Sulphide Scavenging Reactions with

Triazine: A Feasibility Study. Chem. Eng. Trans. 2019 , 74 , 541 −546.

(10) Hamlin, T. A.; Leadbeater, N. E. Raman Spectroscopy as a

Tool for Monitoring Mesoscale Continuous-Flow Organic Synthesis:

Equipment Interface and Assessment in Four Medicinally-Relevant

Reactions. Beilstein J. Org. Chem. 2013, 9, 1843−1852.

11) Ray, A.; Bristow, T.; Whitmore, C.; Mosely, J. On-Line

Reaction Monitoring by Mass Spectrometry, Modern Approaches for

the Analysis of Chemical Reactions. Mass Spectrom. Rev. 2018, 37 (4),

Yusuke Sato − Department of Chemistry, The University of

British Columbia, Vancouver, British Columbia V6T 1Z1,

Canada

Junliang Liu − Department of Chemistry, The University of

British Columbia, Vancouver, British Columbia V6T 1Z1,

Canada

Andrew J. Kukor − Department of Chemistry, The University

of British Columbia, Vancouver, British Columbia V6T 1Z1,

Canada

Jeﬀrey C. Culhane − Neurocrine Biosciences, San Diego,

California 92130, United States

John L. Tucker − Neurocrine Biosciences, San Diego,

California 92130, United States

Hoffman, R. M.; Cook, K. D.; Song, L. Evaluation of Direct Analysis

65−579.

12) Cho, D. S.; Gibson, S. C.; Bhandari, D.; McNally, M. E.;

in Real Time Mass Spectrometry for Onsite Monitoring of Batch

Slurry Reactions. Rapid Commun. Mass Spectrom. 2011, 25 (23),

575−3580.

13) Görög, S. The Changing Face of Pharmaceutical Analysis.

TrAC, Trends Anal. Chem. 2007, 26 (1), 12−17.

14) Rougeot, C.; Situ, H.; Cao, B. H.; Vlachos, V.; Hein, J. E.

(

David J. Kucera − Neurocrine Biosciences, San Diego,

California 92130, United States

Brian M. Cochran − Neurocrine Biosciences, San Diego,

California 92130, United States

Automated Reaction Progress Monitoring of Heterogeneous Re-

actions: Crystallization-Induced Stereoselectivity in Amine-Catalyzed

Aldol Reactions. React. Chem. Eng. 2017, 2 (2), 226−231.

Complete contact information is available at:

https://pubs.acs.org/10.1021/acs.joc.1c01098

(

15) Christensen, M.; Adedeji, F.; Grosser, S.; Zawatzky, K.; Ji, Y.;

Liu, J.; Jurica, J. A.; Naber, J. R.; Hein, J. E. Development of an

Automated Kinetic Profiling System with Online HPLC for Reaction

Optimization. React. Chem. Eng. 2019 , 4 (9), 1555 −1558.

Author Contributions

All authors have given approval to the ﬁnal version of the

manuscript.

Notes

16) Chung, R.; Vo, A.; Hein, J. E. Copper-Catalyzed Hydrogen/

Iodine Exchange in Terminal and 1-Iodoalkynes. ACS Catal. 2017, 7

The authors declare no competing ﬁnancial interest.

(4), 2505−2510.

(17) Malig, T. C.; Koenig, J. D. B.; Situ, H.; Chehal, N. K.; Hultin, P.

ACKNOWLEDGMENTS

G.; Hein, J. E. Real-Time HPLC-MS Reaction Progress Monitoring

Using an Automated Analytical Platform. React. Chem. Eng. 2017, 2

■

The authors wish to thank Mettler-Toledo Autochem for their

generous donation of process analytical equipment (ReactIR,

Easymax, and Easysampler) to J.E.H. Financial support for this

work was provided by Neurocine Biosciences Inc., the

University of British Columbia, the Canada Foundation for

Innovation (CFI-35883), and the Natural Sciences and

(

3), 309−314.

18) Mestre, T.; Ferreira, J.; Coelho, M. M.; Rosa, M.; Sampaio, C.

Therapeutic Interventions for Symptomatic Treatment in Hunting-

ton ’s Disease. Cochrane Database Syst. Rev. 2009, 3, CD006456.

(19) Skor, H.; Smith, E. B.; Loewen, G.; O ’Brien, C. F.; Grigoriadis,

D. E.; Bozigian, H. Differences in Dihydrotetrabenazine Isomer

J. Org. Chem. XXXX, XXX, XXX−XXX

The Journal of Organic Chemistry

Concentrations Following Administration of Tetrabenazine and

Article

Valbenazine. Drugs R&D 2017, 17 (3), 449−459.

(20) Grigoriadis, D. E.; Smith, E.; Hoare, S. R. J.; Madan, A.;

Bozigian, H. Pharmacologic Characterization of Valbenazine (NBI-

21) Paek, S. M. Synthesis of Tetrabenazine and Its Derivatives,

8854) and Its Metabolites. J. Pharmacol. Exp. Ther. 2017, 361 (3),

54−461.

Pursuing Efficiency and Selectivity. Molecules 2020, 25 (5), 1175.

22) Yao, Z.; Wei, X.; Wu, X.; Katz, J. L.; Kopajtic, T.; Greig, N. H.;

(

Sun, H. Preparation and Evaluation of Tetrabenazine Enantiomers

and All Eight Stereoisomers of Dihydrotetrabenazine as VMAT2

Inhibitors. Eur. J. Med. Chem. 2011, 46 (5), 1841 −1848.

(23) Openshaw, H. T.; Whittaker, N. The Synthesis of Emetine and

,2,3,4,6,7-Hexahydro-9,10-Dimethoxy-2-Oxo-11bH-Benzo[a]-

Quinolizines. J. Chem. Soc. 1963, 1449−1460.

24) Hutchinson, G.; Welsh, C. D. M.; Burés, J. Use of Standard

Addition to Quantify In-Situ FTIR Reaction Data. J. Org. Chem. 2021,

6 (2), 2012−2016.

25) Hyde, A. M.; Zultanski, S. L.; Waldman, J. H.; Zhong, Y. L.;

Shevlin, M.; Peng, F. General Principles and Strategies for Salting-Out

Informed by the Hofmeister Series. Org. Process Res. Dev. 2017, 21

(

9), 1355−1370.

(

26) Yu, Q.; Luo, W.; Deschamps, J.; Holloway, H. W.; Kopajtic, T.;

Katz, J. L.; Brossi, A.; Greig, N. H. Preparation and Characterization

of Tetrabenazine Enantiomers against Vesicular Monoamine Trans-

porter 2. ACS Med. Chem. Lett. 2010, 1 (3), 105−109.

(27) Cormier, R. A.; Schreiber, W. L.; Agosta, W. C. Photochemical

Rearrangements of.Alpha.-Methylene Ketones. J. Am. Chem. Soc.

973, 95 (15), 4873.