Understanding and control of dimethyl sulfate in a manufacturing process: Kinetic modeling of a fischer esterification catalyzed by H2SO 4

Article abstract of DOI:10.1021/op200323j

The formation and fate of monomethyl sulfate (MMS) and dimethyl sulfate (DMS) were studied by proton NMR for a sulfuric acid catalyzed esterification reaction in methanol. The kinetic rate constants for DMS and MMS were determined at 65 °C by fitting time-dependent experimental data to a model using DynoChem. In refluxing methanol, sulfuric acid was converted to monomethyl sulfate (MMS) in nearly quantitative yield within 45 min. Once formed, the MMS underwent a reversible esterification reaction to form DMS. Dimethylsulfate reacted with methanol to regenerate MMS and form dimethyl ether. A byproduct of the esterification reaction was water, which further consumed DMS through hydrolysis. On the basis of derived rate constants, in refluxing methanol, DMS would not be expected to exceed 4 ppm in the reaction mixture at equilibrium. In the presence of the carboxylic acid substrate, DMS was not detected in the reaction mixture. The reaction pathways of this system have been systematically investigated, and the results of this study will be presented.

Full text of DOI:10.1021/op200323j

Article
pubs.acs.org/OPRD
Understanding and Control of Dimethyl Sulfate in a Manufacturing
Process: Kinetic Modeling of a Fischer Esterification Catalyzed
by H2SO4
John P. Guzowski, Jr.,*,† Edward J. Delaney,‡ Michael J. Humora,† Erwin Irdam,† William F. Kiesman,†
Albert Kwok,† and Amy D. Moran†
†Biogen Idec, 14 Cambridge Center, Cambridge Massachusetts 02142, United States
‡Reaction Science, 11 Deerpark Drive Suite 202, Monmouth Junction, New Jersey 08852, United States
S
* Supporting Information
ABSTRACT: The formation and fate of monomethyl sulfate
(MMS) and dimethyl sulfate (DMS) were studied by proton
NMR for a sulfuric acid catalyzed esterification reaction in
methanol. The kinetic rate constants for DMS and MMS were
determined at 65 °C by fitting time-dependent experimental
data to a model using DynoChem. In refluxing methanol, sulfuric
acid was converted to monomethyl sulfate (MMS) in nearly
quantitative yield within 45 min. Once formed, the MMS
underwent a reversible esterification reaction to form DMS.
Dimethylsulfate reacted with methanol to regenerate MMS and form dimethyl ether. A byproduct of the esterification reaction
was water, which further consumed DMS through hydrolysis. On the basis of derived rate constants, in refluxing methanol, DMS
would not be expected to exceed 4 ppm in the reaction mixture at equilibrium. In the presence of the carboxylic acid substrate,
DMS was not detected in the reaction mixture. The reaction pathways of this system have been systematically investigated, and
the results of this study will be presented.
the gentle reflux of H2SO4 in methanol encountered during the
esterification process used to make the active pharmaceutical
ingredient (API). The question posed by the regulator was
easily addressed with a direct analysis method using gas chro-
matography (GC) to measure dimethyl sulfate in the isolated
drug substance. Multiple batches produced at commercial scale
were tested with a validated GC method (limit of detection,
LOD = 0.1 ppm), and DMS was not observed in the API.
From a quality-by-design (QbD) perspective, the more
relevant technical question for the chemistry team was “could
DMS be generated in the reaction mixture prior to API
isolation?”
The GC method used to analyze isolated drug substance
could not be applied to in-process solutions that contained sul-
furic acid, carboxylic acid ester, and methanol. Detailed method
development efforts revealed that DMS could actually be for-
med in the heated injection port when samples of H2SO4 in
methanol were subjected to GC analysis. It was not possible at
that time to successfully mitigate artifacts arising from the sample
matrix, and HPLC methods lacked the necessary sensitivity and
selectivity.
INTRODUCTION
■
There is significant interest in the pharmaceutical manufactur-
ing and regulatory communities regarding the generation and
analysis of genotoxic impurities.1 In 2010, 26 articles were
published that highlighted the phrase “genotoxic impurities” as
a keyword.2 We received the following inquiry during regu-
latory review of a commercial process to manufacture a drug
candidate: “Based on the synthetic pathway, including puri-
fication steps, of the drug substance, please address the potential
formation of methyl hydrogen sulfate and dimethyl sulfate
impurities with a validated analytical method, which is sensitive
enough to detect and quantitate the potential genotoxic impu-
rities. The maximum daily intake of the potential genotoxic
impurities should not exceed 1.5 μg per person per day by taking
the highest dose of your clinical candidate.”
The process in question consisted of the sulfuric acid-
catalyzed esterification of a dicarboxylic acid in methanol. A
large body of literature describes dimethyl sulfate (DMS) as a
known genotoxin; furthermore, its reactivity as an electrophilic
2
methylating agent in SN alkylations is greater than that of
methyl iodide.3 In contrast, monomethyl sulfate (MMS) is a
poor alkylating agent and is not genotoxic.3 The potential
formation of DMS as an impurity in the reaction matrix thus
became the primary focus of our investigation.
An orthogonal analytical method for DMS was also inves-
tigated using a well-known derivatizing agent, triethylamine
(TEA), followed by HPLC−MS analysis.4 Unfortunately, it was
Commercial DMS manufacture is typically performed with
SO3 and anhydrous methanol catalyzed by Pd or other tran-
sition metals.3 These forcing conditions are quite different than
Received: November 10, 2011
Published: December 23, 2011
© 2011 American Chemical Society
232
dx.doi.org/10.1021/op200323j | Org. ProcessRes. Dev. 2012, 16, 232−239

Organic Process Research & Development
Article
discovered that the diester drug substance also reacted with
A multistep reaction mechanism based on published work6 is
proposed in Scheme 1 for the generation and consumption of
MMS and DMS.
triethylamine, producing a false positive result for DMS.
Around this time, an investigation was published that
described the use of GC−MS to measure reaction kinetics
for the formation of methyl methanesulfonate.5 This compound
is also a genotoxic impurity and of regulatory interest because
of its potential to form under conditions used to manufacture
mesylate salts of an API.
Scheme 1. Proposed pathways for the formation and
degradation of DMS and MMS
1
Herein, we discuss the application of H NMR techniques
and kinetic modeling to the complex multistep equilibria
of dimethyl sulfate and monomethyl sulfate under reaction
conditions used to manufacture the drug substance.
RESULTS AND DISCUSSION
■
This investigation was undertaken to better understand the
potential to form sulfuric acid-derived impurities in a com-
mercial diester synthesis. The reaction employs methanol and
catalytic amounts of sulfuric acid to effect esterification of the
dicarboxylic acid starting material (eq 1). Sulfuric acid-derived
reaction byproduct could potentially include monomethyl sul-
fate and dimethyl sulfate.
In contrast to previously studied alkylsulfonic acids (such as
methanesulfonic acid and ethanesulfonic acid), sulfuric acid
can undergo sequential reactions with methanol to generate
two different sulfate esters. Monomethyl sulfate is a relatively
benign, nongenotoxic impurity that can be controlled like other
process-related contaminants.3 Conversely, DMS is a known
genotoxic species and must be controlled to very low levels.7
It was important to examine the formation and fate of these two
sulfate esters under actual process conditions to fully under-
stand the potential process risk presented by DMS.
In order to make the study of this complex multistep mech-
anism manageable, the pathway was broken into four discrete
steps, and the results of each were used to develop a compre-
hensive kinetic model. Where possible, a rate constant was
independently evaluated and determined. The effect of water
concentration on each corresponding rate constant was sepa-
rately evaluated and incorporated into the final model. Results
from each step allowed an overall mechanistic model to be
constructed. Early kinetic studies were conducted without the
dicarboxylic acid substrate, but it was included in subsequent
investigations.
The esterification reaction kinetics were first examined to
establish the baseline conditions needed to further study for-
mation of the two sulfate ester impurities (MMS, DMS). The
reaction mixture consisted of 1 equiv dicarboxylic acid,
16.5 equiv methanol (5.7 volumes), and 0.25 equiv concentrated
H2SO4 (0.12 volumes). The solution was heated to a gentle
1
reflux at 65 °C, and time-course samples were analyzed by H
NMR. The reaction was complete within 3 h, and integral peak
data were used to produce the reaction kinetic profile presented
in Figure 1 (millimoles vs time).
Monomethyl Sulfate: Rate of Formation and Degra-
dation. The initial sets of experiments were designed to inves-
tigate the formation (k1) and degradation (k−1) of monomethyl
sulfate, the key intermediate in the pathway to the formation of
dimethyl sulfate.
1. Formation of Monomethylsulfate (k1). The formation of
MMS is depicted in eq 2:
Dry methanol (<0.01% water) was mixed with concentrated,
dry sulfuric acid and heated to 65 °C. After waiting one minute
to reach temperature, seven discrete samples were collected
1
within 12 min and H NMR data (16 transients) obtained for
these solutions. Equilibrium was reached within 1 h, resulting in
an essentially quantitative conversion of H2SO4 into mono-
methyl sulfate. The heated sample remained unchanged after
several days’ storage in a sealed tube. Identity of the MMS
resonance in the NMR spectrum was confirmed by spiking an
authentic sample of monomethyl sulfate into the reaction mix-
ture. The MMS peak integral (CH3, 3.45 ppm) was normalized
to the methyl peak resonance of the methanol solvent (CH3,
3.18 ppm) and DynoChem8 used to calculate the forward
rate constant (k1) from the peak-integral data. The study was
Figure 1. Reaction profile for the esterification of a dicarboxylic acid
by methanol and sulfuric acid as measured by NMR. Experimental data
represented by individual points on the graph.
The NMR-derived data in Figure 1 were consistent with lab-
and production-scale sampling studies where the esterification
reaction was >95% complete within 1 h at 65 °C. Subsequent
kinetic studies focused on generation and consumption of the
two sulfate esters and completed within 60 min.
233
dx.doi.org/10.1021/op200323j | Org. ProcessRes. Dev. 2012, 16, 232−239

Organic Process Research & Development
Article
conducted twice with good agreement between the two derived
rate constants. The experimental (diamonds) and fitted (line)
data for the formation of monomethyl sulfate are presented in
panel a of Figure 2 (early time points in reaction) and panel b
999:1 in favor of the forward reaction, thus conservatively
defining k−1 as 6.5 × 10−8 L/mol s.
In related work, Wolfenden and Yuan9 measured the rate
constants for the hydrolysis of MMS in water over a range of
temperatures and pH and found the extrapolated value (at 25 °C)
to be 1.7 × 10−8 L/mol s (1 M HCl, T = 40−100 °C).
Results from both laboratories agree that MMS is rapidly
formed and stable over a wide range of temperatures and water
concentrations. These results are also consistent with the fact
that monomethyl sulfuric acid is a relatively poor alkylating
agent.
Dimethyl Sulfate: Rate of Degradation and Forma-
tion. Dimethylsulfate is formed and consumed in a complex set
of interrelated equilibria. Rates of DMS solvolysis (methanol-
ysis and hydrolysis) can be readily measured, and these are the
pathways that consume DMS. However, the amount of DMS
that is formed by the forward reaction between methanol and
MMS is very small. Hence, to simplify the experimental design,
the DMS methanolysis and hydrolysis rates were empirically
measured (eqs 4 and 5, respectively).
The equilibrium level of remaining dimethylsulfate following
methanolysis was then used to derive the forward rate of for-
mation for DMS.
1. Methanolysis of DMS (k3 ). Solutions of 1.5 mol % DMS
in methanol were heated in a sealed tube to 35 °C, and time-
dependent NMR spectra were collected for these mixtures.10
Data obtained over 60 min showed that dimethyl ether (DME)
and MMS resonances gradually increased over time with a
concurrent decline in the DMS signal. Representative 1H NMR
spectra from these experiments are presented in Figure 3. The
Figure 2. Experimental and predicted fit of data for the formation of
monomethyl sulfate at 65 °C. (a) Data points collected in the first
12 min, 0.379 M H2SO4. (b) Equilibrium data points obtained from
15 to 180 min, 0.365 M H2SO4.
of Figure 2 (later reaction time points). The H2SO4 concen-
trations were similar for both studies but they were run at
different scales (5 mL vs 50 mL). The second-order forward
rate constant (k1) for the formation of MMS at 65 °C was
experimentally determined to be 6.5 × 10−5 L/mol s with a
confidence interval of 7% RSD.
2. Hydrolysis of Monomethylsulfate (k−1). As illustrated in
Scheme 1, monomethyl sulfate can either hydrolyze back to
sulfuric acid (k−1) or further react with methanol to form DMS
(k2). The hydrolysis of MMS is presented in eq 3
Figure 3. Time-resolved spectra for the methanolysis of DMS at 35 °C.
chemical shift of dimethyl ether was comparable to that
reported in the literature and used to confirm its identity.11 Use
of NMR data helped verify the dynamic reaction pathways
proposed in Scheme 1.
The gradual downfield shift of the exchangeable OH reso-
nance (4.8 ppm) is consistent with the presence of a strong acid
(i.e., MMS) as outlined in eq 4.
The rate constant of this reaction was measured by spiking
water into solutions containing MMS (1.5 mol %) and
monitoring the heated, sealed vessel by H NMR for 45 h.
Water was spiked into the matrix at two different levels, −6 and
12 mol %. In both cases, the MMS level remained virtually
unchanged, confirming that the equilibrium for this reaction lies
far to the right (eq 3). In order to develop the larger model, the
MMS equilibrium constant K = (k1/k−1) was assigned a value of
1
A duplicate set of experiments were performed under typical
reaction conditions at 65 °C (again spiking 1.5 mol % DMS
234
dx.doi.org/10.1021/op200323j | Org. ProcessRes. Dev. 2012, 16, 232−239

Organic Process Research & Development
Article
into dry methanol with <0.01% w/w H2O and no sulfuric acid)
and monitoring the reaction for 60 min. The reaction profile
for the methanolysis of DMS (k3) is presented in Figure 4. The
Figure 4. Experimental reaction profiles for the methanolysis of DMS
at 65 °C for two replicate experiments. The DMS concentration was
measured after mixing dimethylsulfate with dry methanol.
diamonds in Figure 4 represent two replicate measurements
used by DynoChem to generate the second-order rate constant.
This reaction was assumed to be essentially irreversible under
nominal plant process conditions since the reaction vessels are
not pressurized and the resultant DME would bubble out of
solution.
The rate constant for the methanolysis of DMS (k3) was
derived from this experimental data using DynoChem and found
to be 3.1 × 10−5 L/mol s 10% RSD.
The rate measured by Kolesnikov and co-workers12 for this
reaction was reported as 2.27 × 10−5 s−1 at T = 35 °C.
Kolesnikov also presented the Arrhenius parameters for this
system, and they can be used to calculate the rate constant at
65 °C, which yields a value of 1.4 × 10−5 L/mol s (methanol =
23.4 M) and is comparable to the rate determined in this study.
2. Hydrolysis of DMS (k−2). The hydrolysis rate of dimethyl
sulfate was determined by spiking known amounts of water into
a mixture of DMS and methanol at three different levels −14.3,
18.1, and 21.9 mol %. The mol % of water was calculated as
[100(mol H2O/total moles in mixture)]. The reactions were
monitored for 60 min at 65 °C, and DMS peak integrals were
obtained from the NMR data. Time-dependent reaction pro-
files for the consumption of DMS are presented in Figure 5 as
the following: (a) 14.3 mol %, (b) 18.1 mol %, and (c) 21.9 mol %.
The predicted DMS content (mmol) in the presence of
different water spikes was performed using the k2 (hydrolysis)
and k3 (methanolysis) rate constants.
The rate constant for the hydrolysis of DMS (k−2) was
derived by fitting all of the data to a DynoChem model that
employed the time-dependent peak integral data presented
above (Figure 5 a−c). The experimentally determined value for
k−2 was found to be 1.3 × 10−4 L/mol s 20% RSD.
Kolensikov13 reported a value of 6.14 × 10−4 s−1 (T = 35 °C)
for this reaction along with the Arrhenius parameters that were
then used to determine the second-order rate constant at
55 °C, which was the upper temperature limit of their study. The
rate constant adjusted for water concentration (H2O = 54.8 M)
was 7.86 × 10−5 L/mol s and is in good agreement with the
result from this study.
Figure 5. (a) Experimental reaction profiles for the hydrolysis of DMS
at 65 °C, water spike −14.3 mol %. (b) Experimental reaction profiles
for the hydrolysis of DMS at 65 °C, water spike −18.1 mol %. (c)
Experimental reaction profiles for the hydrolysis of DMS at 65 °C,
water spike −21.9 mol %.
Profiles of the DMS methanolysis and hydrolysis experi-
ments revealed that MMS was formed more rapidly when the
methanol solution contained water, whereas dimethylether
formation was retarded by the addition of H2O. These results
suggest a bimolecular mechanism for DMS methanolysis (k3)
and hydrolysis (k−2). Methanol and water compete to consume
2
the available DMS in an SN -like displacement reaction. The
more nucleophilic water molecule is able to hydrolyze DMS
faster than dimethylsulfate can react with methanol (k−2 > k3).
However, in this reaction system, methanol is in much greater
abundance and will be the primary species that consumes DMS.
These two competing reactions are presented in Scheme 2.
235
dx.doi.org/10.1021/op200323j | Org. ProcessRes. Dev. 2012, 16, 232−239

Organic Process Research & Development
Article
Scheme 2. Competition between methanol and water for DMS
Teasdale et al. described similar findings in their study of
the methanolysis and hydrolysis of methyl methanesulfonate.5
Their investigation using O18 labeled methanol confirmed that
oxygen in the dimethyl ether, formed during methanolysis of
methyl methanesulfonate, came from methanol and not meth-
anesulfonic acid.
The DMS peak was measured against the MMS 13C satellite
peak because it more closely approximated the dimethyl sulfate
signal level and would better reflect the true solution concen-
tration. The natural abundance of the 13C isotope was 1.11%
and equally divided between two symmetrical resonances.
When corrections were applied for the number of chemically
equivalent protons in DMS (6 protons) and MMS (3 protons)
the steady-state DMS concentration in the reaction matrix
was calculated to be 157 ppm relative to MMS,16 resulting in a
derived k2 rate constant of 4.9 × 10−9 L/mol s.
3. Formation of Dimethylsulfate (k2). The formation of
DMS from MMS can be expressed as eq 6:
The forward rate constant for methyl methanesulfonate for-
mation, a closely related chemical system, has previously been
reported by Teasdale5 to be 7.1 × 10−8 s−1. In that case, the rate
constant was measured directly and assumed to be dependent
upon conversion of a single species (methanol protonated by
methanesulfonic acid) into methyl methanesulfonate and water.
The rate constant was therefore expressed as a first-order process.
Using the mechanistic pathway presented in Scheme 1, the
forward reaction rate for DMS (k2) can be calculated under
steady-state conditions as follows:
(a) The rate of formation and loss of DMS are in balance
d[DMS]
= 0
−
+
forward rate = k[CH SO MeOH ]
3
3
2
dt
In the current study, a second-order forward rate constant is
reported since the measurement was carried out indirectly from
a methanolysis reaction at equilibrium. However, a more direct
comparison can be made. Since the equilibrium for the
protonation of methanol by methyl methanesulfonic acid lies
far to the right, the rate can also expressed as a second-order
rate constant by employing the alternative convention used in
this work:
= k2[MMS][CH3OH] − k−2[DMS]
[H2O] − k3[DMS][CH3OH]
(b) Solving for k2
[DMS](k [H O] + k [CH OH])
−2
2
3
3
k =
2
[MMS][CH OH]
3
(c) Under dry conditions, the equation can be further simplified
forward rate = k2[CH3SO3H][MeOH]
k [DMS][CH OH]
[DMS]
[MMS]
3
3
By expanding the kinetic pre-equilibration step to include
the methanol concentration term (23.1 M concentration of
methanol in the earlier work employing 1 M methanesulfonic
acid in methanol), the first-order rate constant can be trans-
formed into an equivalent second-order rate constant of 3.1 ×
10−9 L/mol s. The DMS formation rate constant reported here
(4.9 × 10−9 L/mol s) is only slightly higher than the published
forward rate of methyl methanesulfonate formation.5
k =
= k
3
2
[MMS][CH OH]
3
Thus, the rate constant for the formation of DMS (k2) can
be calculated if [DMS] can be measured at steady state
since k3, [MMS], and [CH3OH] have been experimentally
determined.
In early attempts to measure DMS under typical esterifi-
cation conditions, the dimethyl sulfate resonance was not ob-
served, likely because the concentration was too low (ppm) to
be detected by NMR.14 Therefore, a H2SO4 solution was pre-
pared in large excess relative to the standard process conditions
(0.075 mol equivalent H2SO4 relative to methanol or a 5-fold
concentration increase) and heated for an extended time period
(7 h) at 65 °C. Under these forcing conditions, a small, but
clearly discernible peak was observed at a chemical shift of
3.7 ppmthe DMS resonance region. An authentic sample of
DMS spiked into this mixture confirmed the peak’s identity
(15). The DMS peak integral area was approximately 5.7% of
the upfield 13C satellite peak of MMS.
The rate constants for the entire reaction system are sum-
marized in Table 1, as determined using the DynoChem soft-
ware kinetics module.
Comparison of DynoChem Kinetic Model to Measured
DMS Concentrations in a Process Stream. The derived
rates constants resulted from statistical fitting of all data to a
kinetic model that was generated to predict the dynamics of
DMS formation/degradation. The model was subsequently
used to study the formation of DMS in a simulated esterifica-
tion process stream and demonstrated that steady-state levels of
dimethylsulfate were formed within the typical 3-h operating
window employed in the plant. Assuming anhydrous starting
conditions, the steady-state concentration of DMS in the reaction
236
dx.doi.org/10.1021/op200323j | Org. ProcessRes. Dev. 2012, 16, 232−239

Organic Process Research & Development
Article
Table 1. Rate constants for the diesterification process
1
2
Assigned to be 999× slower than k1 for purpose of model creation based on experimental data. Confidence interval was not determined as data
obtained directly from NMR experiment (not DynoChem) in a single measurement, value derived from experimental data collected for other rate
constant.
vessel was predicted to not exceed 4 ppm in the reaction mix-
ture.17 Typically, reagent grade methanol used in the manu-
facturing plant will not be anhydrous and adventitious water
would even further suppress the formation of DMS in this
reaction system because the rate of hydrolysis (k−2) is 5 orders
of magnitude faster than formation of the DMS (k2). However,
this calculation can be used as the worst case scenario in a
failure-mode evaluation of the process.
The API was also tested for DMS, and it was not detected in
the drug substance (<0.1 ppm, the method’s limit of detection).
Dimethyl sulfate that spiked into the reactor during these
experiments exceeded the level predicted by the kinetic model
(4 ppm in solution) by 250-fold. These results confirm that the
process conditions actually control the in situ level of generated
DMS. One possible explanation is the reaction between DMS
and carboxylic acid substrate, which is present in large excess
(3700×).18 This reaction is illustrated as eq 7:
The effective consumption of DMS in the reaction matrix
ensures that the level of this impurity does not approach the
threshold for toxicological concern.
The chemical systems initially used to predict the rate constants
did not contain the carboxylic acids that would be present during
drug substance manufacture. Dimethyl sulfate is a strong alkylating
agent and might also react with the carboxylic acid substrate,
further reducing the DMS concentration in the mixture. Teasdale
previously reported5 that sulfonate ester formation was effectively
quenched when the reaction mixture contained a weak base in
slight excess, which was used to simulate a drug substance.
An in-process reaction mixture containing all components
was tested for DMS content by GC−MS and the result was
<0.5 μg/mL (ppm) as presented in Table 2, entry 1. The DMS
concentration in the reaction solution was below the predicted value
from the kinetic model and confirms that dimethyl sulfate was not
generated at measurable levels under actual process conditions.
To further challenge the reaction system, a DMS stock solu-
tion was gravimetrically prepared and an aliquot spiked into the
reaction mixture at a concentration of 1000 ppm. Samples were
taken during the course of reaction and solutions analyzed for
DMS using a modified GC−MS sample extraction preparation
procedure. Dimethyl sulfate was present at 20 ppm after 1 h
and quickly decreased to <0.5 ppm after only 3 h (Table 2).
CONCLUSION
■
The formation and fate of monomethyl sulfate and dimethyl
sulfate were studied by proton NMR for a methanol−sulfuric
acid esterification reaction, and these results have been pre-
sented. The rate constants were determined using DynoChem to
model the time-dependent peak integral data collected from a
series of experiments probing each mechanistic step. Sulfuric
acid is converted into MMS in near quantitative yield in less
than 60 min under reaction conditions. Monomethyl sulfate
is the more persistent of these two species and appears quite
stable under process conditions. Dimethyl sulfate rapidly con-
verts into MMS and is readily hydrolyzed by water. The pre-
sence of adventitious water further serves to reduce DMS in the
reaction mixture and retards its formation. Under typical pro-
cess conditions, the amount of dimethylsulfate present in the
reaction solution is predicted to be less than 4 ppm. To date,
DMS has not been detected in any drug substance batches
(<0.1 ppm, LOD).
Table 2. Analysis of DMS in reaction solutions by GC−MS
using a liquid−liquid extraction procedure
DMS spiked in
reaction mixture
(μg/mL, ppm)
DMS measured in
reaction mixture
(μg/mL, ppm)
experimental
conditions
entry
1
T = 0 h (analysis
before DMS spike)
0
1000
1000
<0.5
2
3
T = 1 h (250× excess
DMS spike)
20
EXPERIMENTAL SECTION
■
NMR was chosen as the preferred method for this study, owing
to its directness of measurement for all species, its wide dynamic
T = 3 h (250× excess
DMS spike)
<0.5
237
dx.doi.org/10.1021/op200323j | Org. ProcessRes. Dev. 2012, 16, 232−239

Organic Process Research & Development
Article
range of measurement, and the fact that mechanistic assumptions
could be confirmed under the same set of conditions that rate
constants were being measured.
(100-g batch size). The effective DMS solution concentration
was 1000 ppm. The following conditions were used for this
spiking experiment:
All NMR spectra were obtained in, and referenced against,
DMSO-d6 at 2.5 ppm using a Varian 500 MHz NMR. Concen-
trated sulfuric acid, dicarboxylic acid (starting material), and
dimethyl ester (product) were obtained from the Sigma-Aldrich
Chemical Co. Authentic standards of MMS (Na salt) and DMS
were also obtained from Sigma-Aldrich. For spiking identification
purposes, an authentic a sample of MMS was prepared by pass-
ing a methanolic solution of the commercially available sodium
salt through Amberlite FPA 22 resin (H form, 10 mol equiv),
and concentrating to an oil on a flash evaporator.
1 Charge 100 g of dicarboxylic acid starting material.
2 Charge 580 mL of methanol and start agitation at 480 rpm.
3 Charge 21.3 g of sulfuric acid
4 Heat the reaction to 65 1 °C.
5 Charge DMS, ∼0.58 g, accurately weighed (0.55 wt %/wt,
or 5500 ppm, relative to the resulting API).
6 Maintain at this temperature for 3 h. Remove aliquots for
analysis during 3 h reaction time.
7 Cool to 22 °C in 8 h and hold for 2 h.
8 Filter and wash cake four times with 70 mL of methanol.
9 Dry the cake at 22 °C at 100 mmHg.
All small-scale reactions performed in the study were mag-
netically stirred and carried out using an insulated oil bath
10 Sample dry API for DMS content.
maintained at 65
1 °C. Methanol used in this study con-
tained no more than 0.01% water, and was further dried using 3 Å
zeolite molecular sieves that had been predried overnight at
175 °C. Concentrated sulfuric acid, purchased from Sigma-Aldrich,
was >99.9% nominal purity and was used as provided.
Sampling of reactions involved removal of a minimum of
150−400 μL of the reaction mixture and addition to DMSO-d6
(lock solvent). The final volume in the NMR sample tube was
650 μL. Samples were chilled in an ice/water bath, and ana-
lyzed within 5−10 min of their preparation. Reaction profiles
were tracked by plotting reaction completion (via reactant inte-
gral measurements) against time.
ASSOCIATED CONTENT
■
S
* Supporting Information
Experimental data and DynoChem results from the modeling
studies. This material is available free of charge via the Internet
at http://pubs.acs.org
AUTHOR INFORMATION
■
Corresponding Author
John.Guzowskijr@biogenidec.com.
Although DMS was present in only trace amounts upon
reaching equilibrium following extended methanolysis, the S/N
ratio of the DMS peak was >10:1 and thus could be measured
with confidence. However, MMS and its left satellite peak (nearby
in chemical shift to the DMS resonance) could not be accurately
integrated electronically due to the baseline deflections in this
region from the broad methyl resonance of methanol (solvent).
As a result, the spectrum was enlarged, and the peaks were
physically extracted and weighed in order to obtain the molar
ratio. The error associated with this measurement procedure was
expected to be no more than 10% and comparable to other
sources of experimental and computational uncertainties.
Fitting of all experimental data to generate the rate constant
data and an overall kinetic model was carried out using DynoChem
(version 3.3).
REFERENCES
■
(1) (a) Robinson, D. I. Org. Process Res. Dev. 2010, 14, 946−959. (b)
Genotoxic Impurities- Strategies for Identification and Control; Teasdale,
A., Ed.; John Wiley & Sons: NJ, 2010.
(2) Number of citations for genotoxic impurity studies for the year
2010 as obtained from a Scifinder search.
(3) (a) Friedli, F. E. In Rate of Quaternization as a Function of
the Alkylating Agent. World Conference on Oleochemicals into the
21st Century: Proceedings ; World Conference on Oleochemicals − Into
the 21st Century, Kuala Lumpur, Malaysia, October 8−12, 1990;
Applewhite, T. H., Ed.; American Oil Chemists’ Society: Champaign,
IL, 1990; pp 296−297. (b) Thayer, F. J. Am. Chem. Soc. 1924, 46,
1044−1046. (c) Tan, E.; Brimer, P.; Schenley, R.; Hsie, A. J. Toxicol.
Environ. Health 1983, 11, 373−380.
(4) An, J.; Sun, M.; Bai, L.; Chen, T.; Liu, D. Q.; Kord, A. J. Pharm.
Biomed. Anal. 2008, 48, 1006−1010.
A modified GC−MS method (liquid−liquid extraction) was
developed to analyse the in-process solutions and mitigate
artifacts related to the sample matrix. The extraction method
was also used to analyse isolated drug substance development
samples. An Agilent 6890N GC was fitted with a Supelco
Equity-1701 (30 m long × 0.32 mm ID, 1.0 μm film thickness)
column using helium carrier gas at 2 mL/min (constant flow).
Analyte detection was accomplished by an FID. The column
was heated from 50 °C to 280 °C over 21 min. Approximately
∼100 mg of the sample was accurately weighed into a 10 mL
conical tube, and 10.0 mL of 0.1 M NaCl was added to the
sample container and vortexed for 5 min. Then, 1.0 mL of
methyl tert-butyl ether (MBTE) was added to the mixture and
vortexed for another 5 min to extract the DMS. The solution
was centrifuged at 4000 rpm for 10 min, and 200 μL of the
MTBE top layer was removed for analysis by GC.
(5) Teasdale, A.; Delaney, E. J.; Eyley, S. C.; Jacq, K.; Taylor-Worth,
K.; Lipczynski, A.; Hoffmann, W.; Reif, V.; Elder, D. P.; Facchine,
K. L.; Golec, S.; Oestrich, R. S.; Sandra, P.; David, F. Org. Process Res.
Dev. 2010, 14, 999−1007.
(6) Teasdale, A.; Eyley, S. C.; Delaney, E.; Jacq, K.; Taylor-Worth, K.;
Lipczynski, A.; Reif, V; Elder, D.; Facchine, K. L.; Golec, S.; Oestrich,
R; Sandra, P.; David, F. Org. Process Res. Dev. 2009, 13, 429−433.
(7) Sobol, Z.; Engel, M.; Rubitski, E; Ku, W. W.; Aubrecht, J.;
Schiestl, R. Mutat. Res. 2007, 633, 80−94. (b) International Agency for
Research on Cancer (IARC) Monographs Program on the Evaluation of
Carcinogenic Risks to Humans (2004) for Dimethyl Sulfate; International
Agency for Research on Cancer: Lyon, France, 1999; Vol. 71, pp 575−
588.
(8) DynoChem, version 3.3; DynoChem, Inc.: Wilmington, DE,
19803.
(9) Wolfenden, R; Yuan, Y. Proc. Natl. Acad. Sci U.S.A. 2007, 104,
83−86.
(10) Dimethyl ether (DME) resonances were not observed when this
study was first attempted at 65 °C in a standard NMR tube.
(11) Nagai, Y.; Matubayasi, N.; Nakahara, M. J. Phys. Chem. A 2005,
109, 3558−3564.
(12) Kolensikov, V. A.; Efremov, R. V. Kinetika I Kataliz 1979, 20,
813.
DMS Spiking Experiment. A stock solution of DMS was
prepared and used to challenge the disposition of dimethylsul-
fate under process conditions (Table 2). The final concen-
tration of DMS in the reactor was 0.55% (5500 ppm), relative
to API, and spiked into the mixture at the start of the reaction
238
dx.doi.org/10.1021/op200323j | Org. ProcessRes. Dev. 2012, 16, 232−239

Organic Process Research & Development
Article
(13) Kolensikov, V. A.; Efremov, R. V.; Danov, S. M.; Gryaznova,
L. V. Kinetika I Kataliz 1977, 18, 1065−1066.
(14) It was also possible that the weak DMS signal (CH3OSO2O-
H3C, 3.7 ppm) was obscured by the presence of resonances arising
from the reaction products. The monoester and diester products yield
13C satellite signals that reside in the same spectral region as DMS.
(15) A second minor peak was observed at 3.8 ppm and later iden-
tified as a solvent-related impurity arising from the methanol used in
this study.
(16) [DMS] = (1H peak area DMS/1H peak area from total MMS) ×
(equivalent 1H for MMS)/(equivalent 1H for DMS) × (13C abundance
for MMS) = 0.057(3/6) × 0.0055 = 157 ppm.
(17) Assuming virtually anhydrous starting conditions, 2.39 × 10−5
moles of DMS will be present in the reaction mixture under steady-
state conditions as predicted by DynoChem. The DMS amount can be
normalized to the dicarboxylic starting material (1 mol) to yield a
value of 24 ppm expressed as (mol DMS/mol starting material). The
effective DMS solution concentration was determined to be 4 ppm by
correcting for the reaction volume (∼6 vols).
(18) The steady-state concentration of methyl ester intermediate was
determined by DynoChem to be 8.97 × 10−2 mol and compared to that
of DMS (2.39 × 10−5 mol) to determine molar excess (3700×).
239
dx.doi.org/10.1021/op200323j | Org. ProcessRes. Dev. 2012, 16, 232−239