Route selection and process development for a 5-piperazinylquinaldine derivative for the treatment of depression and anxiety

Article abstract of DOI:10.1021/op200140v

1-(3-{2-[4-(2-Methyl-5-quinolinyl)-1-piperazinyl]ethyl}phenyl) -2-imidazolidinone, 1, was identified as a potential drug for the treatment of depression and anxiety. Herein is described the work carried out to select the manufacturing route and the proces

Full text of DOI:10.1021/op200140v

ARTICLE
pubs.acs.org/OPRD
Route Selection and Process Development for a 5-Piperazinylquinaldine
Derivative for the Treatment of Depression and Anxiety
Zadeo Cimarosti,*,† Nicola Giubellina,† Paolo Stabile,† Gilles Laval,† Francesco Tinazzi,† William Maton,†
Roberta Pachera,† Paola Russo,† Ramona Moretti,† Sara Rossi,† Jason W. B. Cooke,‡ and Pieter Westerduin†
†Chemical Development, GlaxoSmithKline Medicines Research Centre, Via Fleming 4, 37135 Verona, Italy
‡Chemical Development, Medicines Research Centre, GlaxoSmithKline, Stevenage, U.K.
ABSTRACT: 1-(3-{2-[4-(2-Methyl-5-quinolinyl)-1-piperazinyl]ethyl}phenyl)-2-imidazolidinone, 1, was identified as a potential
drug for the treatment of depression and anxiety. Herein is described the work carried out to select the manufacturing route and the
process research studies to optimize the key stages of route B. A particular focus is given to the genotoxic impurities, related to this
route, as one of the intermediates of the manufacturing route was genotoxic and many genotoxic impurities can be formed in the
process. Quality by Design principles were applied for the definition of the control strategy of these impurities.
Scheme 1. Early process for 1-aryl-2-imidazolidinone 1
1. INTRODUCTION
A new inhibitor of serotonin reuptake was recently identified as
antidepressant.11-(3-{2-[4-(2-Methyl-5-quinolinyl)-1-piperazinyl]
ethyl}phenyl)-2-imidazolidinone, 1, is a presynaptic inhibitor
5-HT1 receptor antagonist selected for the cure of depression
and anxiety that reached early phase II trials.2
Herein is described the work carried out to select the
manufacturing route and the process research studies to optimize
stages 1 and 2 of this route (route B, see Scheme 7). A particular
focus is given to the genotoxic impurities, related to this route, as
one of the intermediates of the manufacturing route was geno-
toxic and many genotoxic impurities (see Table 1) can be formed
in the process.
The high level of the genotoxic mesylate 3 (>150 ppm) in the
FTIH batch compelled us to purify it in order to meet the
required specifications (in absence of a defined dose, 150 mg/day
was considered, therefore an initial target specification limit of
10 ppm in the drug substance was adopted) as this mutagen had
to be controlled down to an acceptable level. Two recrystallisa-
tions of the drug substance 1 from methyl ethyl ketone (MEK)
were needed to lower the amount of the genotoxic impurity 3
down to an acceptable value (less than 10 ppm).
The principles of Quality by Design3 have been applied to
define the control strategy of these impurities. The concept of the
control strategy was introduced in the ICH Q10 and consists of a
set of controls, derived from current product and process under-
standing that assures process performance for obtaining drug
substance that meets the Critical Quality Attributes (the measur-
able properties that are critical to ensuring patient safety and
efficacy).
Although the synthesis depicted in Scheme 1 appealed to us
for its convergent approach, we speculated that the extremely low
tolerance of the genotoxic impurity 3 in the drug substance 1
might have represented a severe issue in view of manufacturing
campaigns. Hence, we decided to explore alternative, but still
convergent, routes to drug substance 1. Looking at the nature of
the starting materials shown in Scheme 1 (compounds 2 and 3),
and considering that the mesylate 3 was prepared from the ester
4, two different approaches were envisaged (Schemes 2 and 3).
The first approach (Approach A, Scheme 2) consists of a
reductive amination reaction. As a matter of fact, the required
arylacetaldehyde 5 can be obtained either from the correspond-
ing ester 4 or the alcohol 6.
2. EARLY PROCESS
The early process to synthesise the 1-aryl-2-imidazolidinone 1
is depicted in Scheme 12 and allowed to produce the first batches
of drug substance for toxicological use and to support the first
time in human (FTIH) studies. In particular, the last stage of the
Medicinal Chemistry synthesis involved the alkylation of 5-pi-
perazinylquinaldine 2 with the mesylate 3 (Scheme 1). Accord-
ing to the GSK internal risk assessment process for genotoxins,
the mesylate 3 is a structure of concern, proved to be mutagenic
in the Ames test,4 and therefore had to be controlled at a level
below the threshold of toxicological concern (TTC, 1.5 μg/day)
as at the time that these activities were undertaken the staged
TTC had not yet been introduced, a default TTC approach was
undertaken.5
Received: May 30, 2011
Published: September 05, 2011
r
2011 American Chemical Society
1287
dx.doi.org/10.1021/op200140v Org. Process Res. Dev. 2011, 15, 1287–1296
|

Organic Process Research & Development
ARTICLE
Scheme 2. Approach A
Figure 1. Impurities formed inthe DCC/HOBt-mediatedsynthesisof7.
Scheme 4. T3P-mediated synthesis of amide 7
Scheme 3. Approach B
Scheme 5. Reduction of amide 7 by Red-Al
shown to be slower and did not go to completion. The reaction was
best performed by adding the commercially available 50 wt % T3P
in EtOAc to a stirred mixture of acid 8, quinaldine 2 and
triethylamine (TEA) in dichloromethane (DCM) at room tem-
perature. After a careful setup of the whole process, the desired
amide 7 was thus recovered in very high yields (>90%) and with
high purity profile (>95% by HPLC and 1H NMR), see Scheme 4.
With regard to the reduction of the amide 7 to prepare the
drug substance 1, both commercially available 1 M BH3 in THF
and 65 wt % sodium bis(2-methoxyethoxy)aluminum hydride
(Red-Al) in toluene were tested. Since the latter furnished cleaner
reaction profiles, this approach was explored in detail. The efficiency
of the reduction reaction proved to be strongly affected by different
parameters such as temperature, solvent, equivalents and addi-
tion mode of the reducing agent. Reverse addition of a solution of
the amide 7 in CH2Cl2 to the Red-Al solution in toluene at room
temperature gave eventually a clean and reproducible reaction.
However, some residual starting material 7 (10À15%) and a
major impurity that was identified as the compound 11 (5À10%)
were still contaminating the drug substance 1 (Scheme 5). As a
result, the crude compound 1 needed to be recrystallised at least
twice from methyl ethyl ketone (MEK) in order to obtain pure
material (>96% by HPLC and 1H NMR).
The success of Approach A depended, of course, on the ease of
access to the arylacetaldehyde 5. Unfortunately, both reduction
of the ester 4 (e.g., with DIBAL-H) and oxidation of the alcohol 6
(e.g., with PDC, NaClO/TBAB, NaClO/TBAB/TEMPO,
TBAB/TEMPO/NaIO4, and under Swern conditions) resulted
only in extremely complex reaction mixtures.
Alternatively, the desired 1-aryl-2-imidazolidinone 1 could be
obtained by reduction of the corresponding amide 7, that in turn
could be synthesized from the acid 8 and the 5-piperazinylqui-
naldine 2 (Approach B, Scheme 3).
Several amide coupling systems were tested to synthesise the
amide 7. For example, the latter was formed in 80À90% yield by
using the DCC/HOBt coupling system. However, the resulting
material was contaminated by 5À10% of dicyclohexylurea 9, and
by ∼5% of a byproduct which was tentatively identified as
compound 10 (Figure 1).
The use of propane phosphonic acid anhydride (T3P) as the
coupling agent gave complete conversion and a very clean reaction
profile. A few solvents (DMF, EtOAc and dichloromethane) and a
few bases (triethylamine and diisopropylethylamine) were screened
at room temperature. The amide coupling was also investigated in
absence of an added base, taking advantage of the three basic
nitrogens in the piperazinylquinaldine 2, but the reaction rate was
The option of telescoping the amide coupling reaction and the
amide reduction was then implemented in view of a large scale
production of the 1-aryl-2-imidazolidinone 1. After fine-tuning of
1288
dx.doi.org/10.1021/op200140v |Org. Process Res. Dev. 2011, 15, 1287–1296

Organic Process Research & Development
ARTICLE
Scheme 6. Route A
the reaction conditions, 1.1 kg of drug substance 1 was prepared
in 30% yield after 3 recrystallisations of the crude 1. Having
successfully produced drug substance, this approach was defined
as Route A for regulatory purposes and is detailed in Scheme 6.
route, with the potential of being suitable for manufacturing
production. After reviewing the literature, it was decided to work
on the route with the same key step reported in Scheme 1 and
detailed inScheme7, asitintroduceda straightforwardapproach to
compound 1, this route was then defined as Route B for regulatory
purposes. In addition, the known concerns about the genotoxic
impurity 3 were somewhat discharged by a lowering of the
projected effective dose in clinical studies. The therapeutic dose
was lowered from 150 mg to 10 mg and to 3 mg after PhI clinical
studies. As a result, the maximum level of mesylate 3 allowed in
the API by the TTC approach increased from the initial 10 ppm to
500 ppm. This new limit was not considered a critical hurdle for this
route and development studies started. The development studies
for Stages 1 and 2 are detailed in this paper. The process research
for Stage 3 is not discussed here, even though the experimental
procedure is reported in the Experimental Section.
4.1. Genotoxic Risk Assessment. Once Route B was devel-
oped, a complete assessment of all the potential genotoxic
compounds generated in the chemical steps summarized in
Scheme 7, was carried out. A number of structures were identified
as potential contaminants of the drug substance 1 (Table 1),
including the previously mentioned mesylate 3. The next step
was to further assess the toxicity of these structures by means of
the Ames test. In case of a positive result (the structure of
concern is mutagenic) the molecule has to be controlled in the
drug substance following the TTC approach, otherwise it has
to be treated as a general impurity by following the ICH Q3A
guideline. The results of the Ames test are also reported in
Table 1.7
3. DEVELOPMENT AND SCALE-UP OF ROUTE A
The low overall yields and the poor processability of the
telescoped synthesis described above, prompted us to investigate
alternative methods for the reduction of the amide 7. Among the
reducing agents screened, LiAlH4 was superior in terms of
conversion to the drug substance 1 and reaction profile. A
telescoped coupling/amide reduction process was thus opti-
mized on a laboratory scale (Route A). In particular, it involved
a careful aqueous basic quench of the excess of hydride reagent to
transform the aluminium-containing coproducts into inorganic
salts that were easily removed by filtration over Celite (Scheme 6).
By means of Route A several small batches (up to dozens of
grams) of intermediate grade IG-16 and drug substance 1 were
successfully prepared in 50À60% overall yield.
Unfortunately, the process depicted in Scheme 6 was not
robust when scaled up on pilot plant scale. As a matter of fact,
even though4.65 kg of IG-1wereprepared asshown inScheme6,
the overall yield of the process (31.2%) was pretty low as
compared to the laboratory scale experiments. The low recovery
was mainly due to coprecipitation of 1 together with the
aluminium-salts during the filtration of the quenched reaction
mixture over Celite.
On the basis of these data, we were forced to seek a more
robust, reproducible and efficient synthesis of the 1-aryl-2-
imidazolidinone 1 in view of supplying a convenient and scalable
process for future manufacturing.
An additional structure of concern was the propyl mesylate,
potentially formed from traces of methanesulfonic acid (from
methanesulfonic anhydride used in Stage 1) and n-propanol (used
as solvent in Stage 3). However, it was decided not to test this
compound in the Ames test. As a matter of fact, the formation of
this mesylate was considered negligible, as methanesulfonic acid, if
present, would have been neutralised by the excess of the base and
4. DEVELOPMENT OF ROUTE B
As highlighted previously, Route A suffered from significant
drawbacks. Some attempts were made to select and scale-up a new
1289
dx.doi.org/10.1021/op200140v |Org. Process Res. Dev. 2011, 15, 1287–1296

Organic Process Research & Development
ARTICLE
Scheme 7. Route B
Table 1. Structures of concern
in these conditions it was proved that esterification reactions with
the involvement of methanesulfonic acid do not occur.8
4.2.1. Control of Genotoxin 13 in Compound 6. Compound 6
is the input material for the generation of the mesylate 3 and
one of the starting materials of the route. As for the risk as-
sessment described in Table 1, the genotoxic compound 13
might also be formed if residual chloride ions are contained in
the starting material 6, therefore, to avoid any contamination of
the starting material:
• A control of chloride ions in compound 6 was introduced as
part of the control strategy by means of a specification.
• A change control procedure for the manufacturing process
of the starting material 6 was put in place in agreement with
the starting material supplier.
As for the application of the Quality by Design principles,
these impurities are drug substance-CQAs as, due to their
toxicity, they have a direct impact on patient safety and thus,
need to be controlled at the TTC level. Therefore, a control
strategy3 needs to be developed to ensure their control, and
therefore the process development studies took into account the
root causes for the formation of compounds 12, 13, and 14.
A significant effort was expended in trying to identify and
optimize analytical methods capable of detecting the mutagenic
compounds 3, 12, 13, and 14 at the TTC level (500 ppm), as
detection in the drug substance 1 was necessary.
4.2.2. Selection of the Solvent and the Base. The mesylation
reaction of Scheme 7 involved the combination acetonitrile and
pyridine but propionitrile and 2-picoline were finally selected.
The rationale for the change is described below. Pyridine was the
4.2. Development of Stage 1 of Route B. The studies carried
out to develop a suitable process for Stage 1 and the approach followed
to control the related genotoxins are reported in this paragraph.
1290
dx.doi.org/10.1021/op200140v |Org. Process Res. Dev. 2011, 15, 1287–1296

Organic Process Research & Development
ARTICLE
Figure 2. Result of the multivariate study.
quantity of 2-picoline and the reaction volume. The main factors
affecting the formation of the impurity 12 were the temperature, the
quantity of 2-picoline and methanesulfonic anhydride, along with
the interaction between these two parameters. Hence, as target
value for the process parameters, a compromise between the higher
yield and the lower amount of bis-mesylate 12 was sought and the
centre points of the ranges shown in Table 2 were selected. Only in
the case of the equivalents of methanesulfonic anhydride, the
amount was reduced to 1.15 to minimize the formation of the
impurity 12.
Some work was then carried out on the workup conditions. As
introduced in Table 1, the presence of compound 3 and chloride
ions resulted in the formation of the genotoxin 13. It was
speculated that the hydrochloric acid previously used in the
hot workup (40 °C) of stage 1 was a source of chloride ions.
Because of that, aqueous sulphuric acid was used in place of
hydrochloric acid to avoid the formation of the chloride impurity
13. The use of sulphuric acid might in turn generate the
corresponding sulphate impurity. This was not formed or was
undetected in our analytical method.
4.3. Development of Stage 2 of Route B. The studies carried
out to develop a suitable process for Stage 2 and the approach
followed to control the related genotoxins are reported in this
paragraph.
4.3.1. Control of Genotoxin 14 in Compound 2. The risk
assessment identified the presence of hydrobromic acid in the
starting material 2 as a root cause for the formation of the
genotoxin 14, therefore a basic wash was introduced in the key
step of the formation of the starting material 2.
4.3.2. Selection of the Solvent and the Base. A multivariate
analysis started to identify some combinations of solvents and
bases. A Principal Components Analysis (PCA)10 was employed
to ensure diversity in bases and solvents for the screening. Four-
teen solvents and eleven bases were selected, knowing that below
70 °C the conversion was low and the reaction rate slow. In
particular the choice of the solvents was restricted to those with a
boiling point higher than 70 °C. As for the bases, primary and
secondary amines were not selected, because they might react with
the mesylate 3. In Table 3 the solvents selected are shown.
4.3.2.1. Experimental Studies. It is worth noting that this
approach would have required 154 experiments (14 solvents  11
bases) to test all the possible combinations. In order to reduce the
number of experiments the Partial Least Square (PLS) approach
was used11 and only 20 experiments were run. In these experiments
a mixture of equimolar compounds 2 and 3 and the base (2 equiv)
in the solvent (12.5 vol) was stirred at ∼80 °C. Each reaction was
Table 2. Ranges for the Stage 1 multivariate study
process parameter
range
centre point
temperature
20À40 °C
30 °C
methanesulfonic anhydride
2-picoline
1.05À1.35 equiv
1.2À3 equiv
7À13 vol
1.2 equiv
2.1 equiv
10 vol
concentration
selected reagent according to a base screening for stage 1 but
2-picoline represents a safer, less toxic alternative with similar
physical propertiesand comparable reactionprofile. Proprionitrile is
similar to acetonitrile and gave complete conversion to the
mesylate 3. Moreover, propionitrile is immiscible with water,
allowing a simple workup without the use of any additional
cosolvent. Despite the fact that propionitrile is not classified by
ICH and recommended limits are not available, this solvent was
not a problem for the process as the drying conditions (vacuum,
50À55 deg) were selected to allow its complete removal prior to
the last step (the Stage 3 recrystallization).9 Finally, methane-
sulfonic anhydride was preferred over methanesulfonyl chloride
as it does not produce chloride ions in the reaction that can be
responsible for the formation of genotoxin 13 (see Table 1).
Initially the process consisted in the portionwise addition of the
solid methanesulfonic anhydride to the slurry of 2-picoline and
alcohol 6, this was decided to avoid an excess of methanesulfonic
anhydride in presence of the starting material 6 and 2-picoline as it
was considered critical for the formation of the bis-mesylate 12.
4.2.3. Optimisation of the Reaction Parameters. A DoE study
(Fractional Factorial Design) was carried out to identify the main
factors that affected the yield and the formation of the impurity
12 in the synthesis of compound 3. Since strong curvature was
observed (Figure 2), an upgrade to a Central Composite Design
was performed as second step of the investigation. Previous
process knowledge and a risk assessment allowed the selection
of a series of process parameters and ranges for the DoE study
(see Table 2).
The main impurity observed in this step was the bis-mesylate
impurity 12 (Table 1). This impurity was also a drug substance-
CQA, as it was carried through the entire process contaminating
the drug substance, and therefore it needs to be controlled
following the TTC approach.5
Analysis of the data showed a strong curvature of the yield
(ranging from 78 to 97%), that was mainly due to the quantity of
2-picoline. Other factors affecting the yield were the quantity of
methanesulfonic anhydride and the interaction between the
1291
dx.doi.org/10.1021/op200140v |Org. Process Res. Dev. 2011, 15, 1287–1296

Organic Process Research & Development
ARTICLE
Table 3. Solvent and base screening
Table 4. Factor Ranges for the definition of the stoichiometry
of stage 2
conversion
run
solvent
base
(24 h)
note
range
1
2
3
4
5
6
7
8
9
isopropyl acetate DBU
54.2
85.8
69.9
93.2
62.9
68.6
65.0
58.6
21.4
95.0
45.5
51.6
76.1
48.0
68.3
66.6
88.8
95.2
74.7
95.6
gum
factor
unit
low
high
toluene
TEA
suspension/gum
solution
A
B
compound 2
DIPEA
equiv
equiv
vol
1
1.4
1.8
10
butanone
1-butanol
2-methyl-THF
MIBK
N-methylpiperidine
tributylamine
1.2
5
solution
C
D
concentration
temperature
2,3-lutidine
suspension
suspension
suspension/gum
solid/gum
solution
°C
kept constant (at reflux)
4-methylmorpholine
pyridine
n-butyl acetate
isooctane
4-methylmorpholine
Table 5. Hydrolysis of the mesylate 3 in aqueous Na2CO3
(5% wt)
ethylene glycol DABCO
10 NMP
DIPEA
DMAP
DBU
solution
time (h)
alcohol 6 (%)
mesylate 3 (%)
11 isooctane
12 DMSO
gum
0
0
4
100
96
91
75
30
solution
0.5
1.5
4.5
24
13 ethylene glycol N-methylpiperidine
solution
9
14 n-butyl acetate
15 MIBK
4-phenylmorpholine
2,3-lutidine
pyridine
suspension
suspension/gum
suspension/gum
solution
25
70
16 2-methyl-THF
17 anisole
tributylamine
TEA
18 propionitrile
19 ethyl acetate
20 1-butanol
solution
minimal number of experiments (20 versus 154 potentially
required to test all the combinations) and identifying a successful
combination of solvent and base (propionitrile/DIPEA).
4.3.3. Optimisation of the Reaction Parameters. A DoE was
carried out to identify the main factors that could affect the final
yield of Stage 2 (see Table 4). The initial process knowledge and
some risk assessment activities allowed to identify three factors
for the multivariate study: the amount of starting material 2, the
amount of DIPEA and the concentration of the mesylate 3 in
propionitrile. The mesylate 3 was used in substoichometric quantity
in the study to increase the chance to complete the reaction and
minimize the risk of contamination of the residual mesylate 3 in the
compound IG-1. The temperature was kept at reflux to increase the
solubility of the reactants and the reaction rate.
A 2 Level Factorial design (Full design) with one block, two
center points and ten reactions, was used to study the reaction.
Analysis of the obtained data, carried out by using the software
Design Expert 7.0, did not give a significant model, mainly
because the yield range was narrow (between 67.6% and 73.0%
calculated with respect to the alcohol 6 by using phenanthrene as
internal standard) and the variability of the two centre points
quite high, if compared to the yield range. Thus, the reaction was
considered robust enough in the measured range.
Verification experiments confirmed these data and hypoth-
eses. It was demonstrated that the concentration was mainly
affecting the reaction rate, with more concentrated reaction
conditions giving a faster conversion of the mesylate 3. More-
over, the reaction was stable with time. The following target value
was then selected, based on the latter results:
compound 2: 1.1 mol equiv versus the mesylate 3
DIPEA: 1.8 mol equiv versus the mesylate 3
temperature: reflux of propionitrile
concentration of the mesylate 3: 10 volumes (this parameter
was selected to minimise the likelihood of precipitation of the
IG-1 during the workup procedure)
4.3.4. Studies on the Workup Procedure. As the reaction was
optimised in propionitrile, a series of aqueous washings were used to
remove all the reaction components that could have affected the
N-methylpiperidine
DIPEA
solution
solution
sampled and analysed by HPLC after 3 and 6 h; after ∼24 h the
reactions were cooled to room temperature and each reaction
mixture treated with DMF until complete dissolution, sampled
and analysed. The obtained data are reported in Table 3, only the
conversion after 24 h is reported as in the previous check points
(3 and 6 h) the reaction was not complete. The conversion of
each reaction is calculated by HPLC comparing the % area of the
peak of compound 1 versus compound 2.
4.3.2.2. Analysis of the Data. Analysis of the data using the
PLS approach showed that the yield was mainly affected by the
factors correlated to the bases; in particular trialkylamine, such as
TEA, DIPEA, tripropylamine (not in the original design) and
tributylamine, gave the most promising results and, among these,
DIPEA and tributylamine were chosen. Moreover, looking at the
solvents, a correlation between the polarity and the yield was
observed. According to this, among the analysed solvents,
1-butanol, anisole and propionitrile were chosen, because they
permitted precipitation of the product after cooling to room
temperature, allowing an easy isolation of the desired product
(DMSO, NMP and ethylene glycol gave a solution after cooling).
4.3.2.3. Selection of Propionitrile and DIPEA. Further experi-
ments, where all the combinations of the three solvents and the
two bases were tested, gave very similar results (almost complete
conversion of the mesylate 3 was observed in all the cases).
However propionitrile gave the highest yield after precipitation
(88% with respect to the mesylate 3). In addition it was selected
as it allowed the direct precipitation of the compound IG-1 at
room temperature. DIPEA was selected as more commercially
available with respect to tributylamine. This combination was
considered for the optimisation of the reaction conditions.
It has to be highlighted once again that the application of PCS/
PLS approach is extremely useful for the investigation of a
complex experimental space, in this case it is described how a
complex experimental space generated by the selection of 14
solvents and 11 bases, could be successfully explored with a
1292
dx.doi.org/10.1021/op200140v |Org. Process Res. Dev. 2011, 15, 1287–1296

Organic Process Research & Development
ARTICLE
Scheme 8. Route B, final process
Table 6. Level of genotoxins in the kilo-lab and pilot-plant batches
level of genotoxins (ppm)b
run
batcha
scale
3
12
13
14
total impurities (% a/a)
yieldc (%)
1
2
3
4
5
6
IG-1
DS-1
IG-2
DS-2
IG-3
DS-3
kilo lab
<5
<5
<5
<5
<50
<5
<5
<5
<5
<5
<5
<5
1.3
0.6
1.7
1.1
0.9
1.0
76
kilo lab
86
kilo lab
<50
<5
<5
<500
<250
<50
78
kilo lab
<5
88.5
70
pilot plant
<50
<50
<5
pilot plant
NA
<50
80
a IG is Intermediate grade and DS is Drug Substance. DS-1 was prepared from IG-1, DS-2 from IG-2, and DS-3 from IG-3. b The method for the
detection of these genotoxins was developed as a limit test by LC/MS (ESI positive, SIM mode). c Yield from compound 2 for entries 1, 3, and 5. Yield
from IG-1 for entries 2, 4 and 6.
crystallization of 1. Since the mesylate 3 was the most critical to be
eliminated due to its toxicity, a washing with sodium carbonate was
introduced with the objective of hydrolyzing 3 (see Table 5). The
reaction was studied over 24 h. Aqueous sodium carbonate (5% wt)
was added to a representative sample of mesylate 3 (1 equiv)
reaction mixture in propionitrile (10 vol) at 83 °C. The biphasic
mixture was kept for 24 h at 83 °C and evaluated against an internal
standard (anthracene). It was found that 70% of the mesylate was
hydrolyzed in the workup conditions in 24 h. Thus, the workup can
be used as a step to significantly reduce the mesylate 3.
of the alcohol 6 and methanesulfonic anhydride at 30 °C
was introduced in order to have an easier process for the pilot
scale-up. In addition to a lower quantity of methanesulfonic
anhydride and the reverse addition previously introduced, tem-
perature control kept the bismesylate impurity 12 below the level
of concern. This required the slow and controlled addition of
2-picoline in order to keep the mesylation reaction at a tempera-
ture below 40 °C. Under these conditions complete conversion
(>99%, HPLC) of 6 to the mesylate 3 was achieved in 30 min,
and the bismesylate impurity was 1.4% or lower (HPLC). Work-
up was performed by diluting with propionitrile (10 vol) and
extracting with aqueous sulphuric acid solution and water.
Removal of water via azeotropic distillation (<1 wt %) gave a
red solution ready for the subsequent nucleophilic substitution
(Stage 2). It was found that water content higher than 1 wt % in
propionitrile resulted in slow conversion rates for the formation
of IG-1.
Considering the rate of reaction and the reaction conditions,
15 min stirring at 83 °C were selected for the optimised process.
5. ROUTE B, FINAL PROCESS AND SCALE-UP RESULTS
The final process after the development work is reported in
Scheme 8
For Stage 2, piperazinylquinaldine 2 (1.1 equiv) and DIPEA
(1.8 equiv) were added to the solution of the mesylate 3, the
reaction mixture was heated at reflux (about 94 °C) for at least
10 h to achieve complete conversion. A quench with a 5% wt
Na2CO3 aqueous solution at 83 °C was introduced to remove
any residual mesylate 3 and methanesulfonic acid, produced by
the mesylation reaction (Stage 1). After separation, the dark red
residue was finally washed with water at 83 °C. The implementa-
tion of a warm workup was needed to keep the compound IG-1
in solution. When IG-1 crashed out accidentally in the presence
of water, for instance during the washes at temperatures lower
than 60 °C, it crystallized in the undesired monohydrate form.12
Instead, the correct anhydrous form was obtained after the warm
aqueous washes and azeotropic removal of water via propionitrile
distillation.
This process was scaled up to kilo lab (50À200 g input of
starting material 2) and pilot plant (up to 5 kg input of starting
material 2), the results obtained are reported in the following
paragraphs
5.1. Final Process Details. The process depicted in Scheme 8
was implemented with some minor changes. Stage 1 was operated
with the amount of solvent/reagents suggested by the DoE
studies. The alcohol 6 was dissolved in propionitrile (10 vol)
and submitted to the mesylation reaction in the presence of
methanesulfonicanhydride(1.15equiv) and2-picoline(2.1 equiv).
At first, the process included the portionwise addition of the solid
methanesulfonic anhydride to the slurry of 2-picoline and alcohol
6. However, methanesulfonic anhydride is a very hygroscopic
material that is better manipulated only once it is charged into a
reactor. Then, the reverse addition of the 2-picoline to a slurry
1293
dx.doi.org/10.1021/op200140v |Org. Process Res. Dev. 2011, 15, 1287–1296

Organic Process Research & Development
ARTICLE
5.2. Process Scale-Up Results. The results of the scale-up of
the optimised process in kilo lab and pilot plant are summar-
ized in Table 6. Regarding the yield a small decrease of the
yield was obtained at a plant scale, but this was not considered
critical for this project phase. The impurity profile gave a slight
improvement from kilo lab batches versus pilot plant (runs 1
and 3 versus 5) while the difficulty of the crystallization pro-
cess in Stage 3 to remove general impurities was confirmed (run 1
compared with 2, run 3 compared with 4, run 5 compared
with 6).
As can be seen from Table 6, the level of the genotoxins (3, 12,
13, 14) were controlled properly by the optimised Stage 1 and 2
process (entries 1, 3 and 5) as the data referred to the analyses of
intermediate grade material (IG-1 in Scheme 8). In addition, the
final crystallization in 1-propanol carried out as the last step
(Stage 3 in Scheme 8) seemed to further reduce the level
of mutagens 3 and 13 (run 1 compared with 2, run 3 compared
with 4). Regarding the mutagens 12 and 14 their level was not a
concern.
For genotoxin 14, the process selected to prepare the starting
material 2 was considered to be appropriate for its control, so any
process change introduced would have been risk assessed and
tested on the appropriate scale before approval.
The same approach was used for the starting material 6 where
the amount of chloride ions was introduced as a test in the release
specification. In addition, any change introduced in the process
to 6 would have been risk assessed and tested on the appropriate
scale before approval.
It is recommended that analytical data on the level of these
mutagens in the drug substance 1 are collected for the develop-
ment and scale up batches so that, for the commercial product,
the removal of the testing of these mutagens in the drug
substance can be considered after the appropriate risk assessment
in agreement with the current regulatory requirements.
6. CONCLUSION
The activities for the identification of the manufacturing route
and the optimisation of the process for stages 1 and 2 to the drug
substance 1 are discussed in this paper. The process allowed the
preparation of the target compound in good yield, in addition no
scale up problems were encountered confirming full control on
the impurities present in the process. Once again, the importance
of the use of a DOE approach coupled with the PLC/PLS tools to
build the knowledge in support of a robust and scalable manu-
facturing process was confirmed.
Finally, a particular focus is given to the genotoxic impurities
as they needed to be controlled below the TTC (lower than
1.5 μg/day). The studies carried out to build a robust control
strategy for genotoxic impurities are reported. As for the general
impurities, the optimised process allows complete control of
their level in the drug substance. In spite of these promis-
ing initial data, it is recommended that data on the performance
of the process in controlling these genotoxic impurities are
collected throughout the development phase. If the robustness
of the control strategy is confirmed and the risk of contamina-
tion is fully under control, there is the opportunity, according to
the current regulatory requirements, to remove these drug
substance-CQAs (or some of them) from the drug substance
specification through the full application of the QbD principles
to manufacturing processes.
In conclusion, the optimised process allowed us to prepare the
target compound in acceptable yield, confirming full control of
the impurities and of the genotoxins present in the process.
5.3. Control Strategy for Genotoxins. In summary, the
following control points have been defined, these are elements
of the control strategy for genotoxins in the drug substance 1
For the intermediate mesylate 3:
7. EXPERIMENTAL SECTION
NMR spectra were recorded on a 600 MHz Varian Inova 600
spectrometer. HPLC analysis of the intermediates and reaction
monitoring were carried out on Agilent Series 1200 (Agilent).
Generic acidic HPLC method used: column type Luna C18; mobile
phase A: 0.05% TFA/water and B: 0.05%. TFA/acetonitrile;
gradient: 0 min 100% A to 8 min 95% B; flow 1 mL/min;
column temperature 40 °C; detector UV DAD @ 220 nm.
Mass spectra analyses were performed on Agilent 1100 LC/MS
equipped with a Waters ZQ single quadrupole, operating in
ESI positive mode, SIM acquisition. Commercially available
reagents and solvents were purchased from ordinary chemical
suppliers and used without purification. The starting materials,
i.e. 5-piperazinylquinaldine 2, methyl arylacetate 4, arylethanol
6, and arylacetic acid 8, were supplied by external contractors
on multikilogram scale.
Preparation of 5-Piperazinylquinaldine 1 (IG-1). Stage 1.
The alcohol 6 (1 equiv) and the methanesulfonic anhydride
(1.15 equiv, 0.96 wt) were suspended in propionitrile at 20 °C
(10 vol). The slurry was warmed up to 30 °C and kept at 30 °C
with vigorous stirring and under a flow of nitrogen for at least
15 min. 2-Picoline (2.1 equiv, 1.01 vol) was added dropwise,
maintaining the internal temperature below 40 °C. At the end of
the 2-picoline addition, a solution was obtained, and within a few
minutes a suspension appeared. The mixture was heated to
40 °C. After 30 min the reaction mixture was quenched at
40 °C with 6% vol/vol aqueous sulphuric acid (3 vol) and the
temperature increased to 50 °C; after separation the organics
• Stage 2 stoichiometry: the mesylate 3 is substoichiometric
with respect to compound 2 in Stage 2 (see section 4.3.3).
• Stage 2 workup: the sodium carbonate washing temperature
is 83 °C (see section 4.3.4).
For the bismesylate impurity 12:
• Stage 1 reaction parameters: the excess of methanesulfonic
anhydride is minimised (1.15 equiv), the amount of 2-pico-
line is 2.1 equiv, and the temperature is 30 °C (see 4.2.2).
• Stage 1 addition order: 2-picoline is added to a slurry of the
alcohol 6 and methanesulfonic anhydride at 30 °C (see
section 4.2.2).
For the chlorinated impurity 13:
• Specification: the control of chloride ions in starting ma-
terial 6 is introduced (see section 4.2.1).
• Change control procedure: a change control procedure for
the process of the starting material 6 was put in place (see
section 4.2.1).
• Stage 1 reactant: methanesulfonic anhydride is used instead
of methanesulfonyl chloride as reagent in this step (see
section 4.2.2).
• Stage 1 work up conditions: a sequence of washing, avoiding
chloride ions, was optimised (see section 4.2.3).
For the brominated impurity 14:
• Starting material process: a basic wash was introduced in the
key step of the formation of the starting material 2 (see
section 4.3.1).
1294
dx.doi.org/10.1021/op200140v |Org. Process Res. Dev. 2011, 15, 1287–1296

Organic Process Research & Development
ARTICLE
were washed with water (3 vol) (during the wash the internal
temperature must be kept at 48 ( 2 °C). To the washed organic
phase propionitrile (10 vol) was added, and the resulting solution
was concentrated to 10 vol by distillation at atmospheric
pressure. The propionitrile solution was cooled to 60 ( 2 °C.
Stage 2. The piperazine quinaldine 2 (1.1equiv, 1.2 wt) and N,
N-diisopropylethylamine (1.8 equiv, 1.5 vol) were added to the
solution of the mesylate 3 at 60 °C. The resulting mixture was
heated at reflux (approximately 94 °C) for at least 10 h. The
solution was sampled for analysis by HPLC (IPM expected:
>99% conversion to drug substance 1 with respect to the
mesylate). The solution was cooled to 83 ( 2 °C and diluted
with proprionitrile (10 vol). The solution was washed at 83 (
2 °C with 5 wt % aqueous sodium carbonate (5 vol) and stirred
for 15 min, and the lower aqueous phase was separated. To the
warm organic phase was added slowly water (5 vol), keeping the
internal temperature at 83 ( 2 °C, and was stirred at 83 ( 2 °C
for at least 5 min. The lower aqueous phase was separated. The
warm organic phase (about 20 vol) was concentrated by distilla-
tion at atmospheric pressure to 10 vol with a DeanÀStark (about
4 h) (KF < 0.5 wt/wt), while the internal temperature reached
95 ( 2 °C. The slurry was cooled to 83 ( 2 °C and diluted
with propionitrile (5 vol) and the temperature was increased to
95 ( 2 °C to give a solution. Then the temperature was adjusted
to 83 ( 2 °C and seeded with drug substance 1 (0.005 wt). The
suspension was cooled to 20 ( 2 °C over approximately 1 h. The
slurry was stirred for 1 to 2 h at 20 ( 2 °C then the solid was
collected by filtration. The cake was washed with propionitrile
(2 Â 3 vol). All the washes were carried out at 20 ( 2 °C. The
damp cake was dried in a vacuum oven at 50À55 °C overnight
(yield 70% mol).
Preparation of 5-Piperazinylquinaldine 1. Stage 3. IG-1
(1 wt) was dissolved in 1-propanol (15 vol) by heating to 85 °C;
then the solution was passed through a clarification line-filter
(5 μm). The line-filter was washed with hot 1-propanol (2 vol).
The collected filtrates were concentrated to 9 vol by distillation at
atmospheric pressure. The resulting solution was then cooled to
70 °C (internal temperature) and seeded with drug substance 1
(0.0025 wt). The suspension was cooled down to 20 °C (internal
temperature) over approximately 1 h. The slurry was stirred for
2 h at 20 °C (internal temperature), then the solid was collected
by filtration. The cake was washed successively with 1-propanol
(2 vol), 1-propanol/isooctane 1:1 (2 vol), and isooctane (2 vol).
All the washes were carried out at 20 °C. The damp cake was
dried in a vacuum oven at 50À55 °C overnight. (yield 80% mol).
1H NMR (600 MHz, DMSO-d6) δ 2.64 (m, 2 H), 2.63 (s, 3 H),
2.74 (br s, 4 H), 2.78 (m, 2 H), 3.04 (br s, 4 H), 3.39 (m, 2 H),
3.84 (dd, J = 8.94, 7.01 Hz, 2 H), 6.89 (d, J = 7.70 Hz, 1 H), 6.91
(m, 1 H), 7.11 (dd, J = 6.60, 1.92 Hz, 1 H), 7.21 (t, J = 7.97 Hz,
1 H), 7.39 (m, 2 H), 7.46 (t, J = 1.65 Hz, 1 H), 7.59 (m, 2 H), 8.34
(d, J = 8.80 Hz, 1 H); m/z 416.
Costa, Jim Harvey, David Elder, Paul Ravenscroft and Vance
Novack for helpful discussions.
’ REFERENCES
(1) (a) Oficialdegui, A. M.; Martinez, J.; Perez, S.; Hears, B.; Irurzun,
M.; Palop, J. A.; Tordera, R.; Lasheras, B.; del Rio, J.; Monge, A. Farmaco
2000, 55, 345. (b) Evrard, D. A.; Zhou, P.; Yi, S.; Zhow, D.; Smith, D.;
Sullivan, K. M.; Hornby, G. A.; Scheckter, L. E.; Andree, T. H.;
Mewshaw, R. E. Bioorg. Med. Chem. Lett. 2005, 15, 911.
(2) (a) Bergauer, M.; Bertani, B.; Biagetti, M.; Bromidge, S. M.;
Falchi, A.; Leslie, C. P.; Merlo, G.; Pizzi, D. A.; Rinaldi, M.; Stasi, L. P.;
Tibasco, J.; Vong, A. K.; Ward, S. E. Quinoline and quinazoline
derivatives having affinity for 5HT1-type receptors. WO/2005/
014552, 17022005, 2005. (b) Bentley, J.; Bergauer, M.; Bertani, B.;
Biagetti, M.; Borriello, M.; Bromidge, S. M.; Gianotti, M.; Granci, E.;
Leslie, C. P.; Pasquarello, A.; Zucchelli, V. Fused tricyclic derivatives for
the treatment of psycotic disorders. WO/2006/024517, 09032006,
2006. (c) Serafinowska, H. T.; Blaney, F. E.; Lovell, P. J.; Merlo, G.;
Scott, C. M.; Smith, P. W.; Starr, K. R.; Watson, J. M. Bioorg. Med. Chem.
Lett. 2008, 18, 5581.
(3) (a) ICH Q8 Pharmaceutical Development, (R2); U.S. Depart-
ment of Health and Human Services, Food and Drug Administration,
Center for Drug Evaluation and Research (CDER): Rockville, MD, Aug
2009. (b) ICH Q9 Quality Risk Management; U.S. Department of Health
and Human Services, Food and Drug Administration, Center for Drug
Evaluation and Research (CDER): Rockville, MD, June 2006. (c) ICH
Q10 Pharmaceutical Quality System; U.S. Department of Health and
Human Services, Food and Drug Administration, Center for Drug
Evaluation and Research (CDER): Rockville, MD, April 2009.
(4) Mesylate 3 was tested in a bacterial mutation screening assay
(Ames test) with Salmonella typhimurium TA1535, TA1537, TA98,
TA100 and Escherichia coli WP2uvrA(pKM101) in the presence and
absence of S9-mix. Mesylate 3 was mutagenic in the Ames test in strains
TA100, TA1535 and WP2uvrA(pKM101) when tested in the presence and
absence of S9-mix. The maximum concentration tested was 5000 μg/plate,
in accordance with current guidelines.
(5) (a) Draft Guidance for Industry. Genotoxic and Carcinogenic
Impurities in Drug Substances and Products: Recommended Approaches.
U.S. Department of Health and Human Services, Food and Drug Admin-
istration, Center for Drug Evaluation and Research (CDER): Rockville,
MD, December 2008. (b) Guideline on the Limits of Genotoxic Impurities,
CPMP/SWP/5199/02, EMEA/CHMP/QWP/251344/2006; Committee
for Medicinal Products (CHMP), European Medicines Agency (EMEA):
London, 28 June 2006. (c) Question and Answers on the CHMP Guideline on
Genotoxic Impurities, CHMP/SWP/431994/2007; Committee for Medicinal
Products for Human Use (CHMP), European Medicines Agency (EMEA):
London, January 2008, . (d) Elder, D. P.; Harvey, J. S. Org. Process Res. Dev.
2010, 14, 1037–1045.
(6) IG-1 is the Intermediate grade drug substance, the acronym
indicates the compound before the final crystallization step.
(7) It is worth noting that the Ames test defines if a substance is
mutagenic (it changes the genetic material), and this concept is slightly
different from the concept of genotoxicity (a substance that has negative
effect on the cell’s genetic material). For the purposes of the discussion
in this contribution a positive result in the AMES test identifies a
mutagenic compound, and this compound is considered a genotoxin.
(8) Teasdale, A.; Eyley, S. C.; Delaney, E.; Jacq, K.; Taylor-Worth,
K.; Lipczynski, A.; Reif, V.; Elder, D. P.; Facchine, K. L.; Golec, S.;
Schulte Oestrich, R.; Sandra, P.; David, F. Org. Process Res. Dev. 2009,
13, 429–433.
’ AUTHOR INFORMATION
Corresponding Author
zadeo.cimarosti@aptuit.com.
(9) As mentioned above the wet cake drying conditions in Stage 2
(vacuum at 50À55 °C) were selected to allow full removal of propioni-
trile. This is confirmed by the NMR where, after drying, is not seen any
residual left in IG-1 (if the 13C satellite peaks of the 1H NMR resonances
of the IG-1are considered, up to 0.5% mol of propionitrile can be
detected, this limit corresponds to 0.1% w/w of propionitrile in IG-1,
that, on a dose of 3 mg/day is equal to a daily intake of 3 μg/day).
’ ACKNOWLEDGMENT
We thank Marco Lo Re, Fernando Bravo, Antonella Carangio,
Alberto Spezzaferri, Alessandro Lamonica, Fabio Fazio, Corinne
Leroi, Stefano Provera, Lucilla Turco, Simone Tortoioli, Laura
1295
dx.doi.org/10.1021/op200140v |Org. Process Res. Dev. 2011, 15, 1287–1296

Organic Process Research & Development
ARTICLE
In addition, Stage 3, the recrystallization step in n-propanol, introduces
an additional step where the amount of residual propionitrile is further
reduced.
(10) Each class of compounds (for example solvents and reagents)
can be described by relevant chemical descriptors (e.g., dielectric
constant, boiling point, log P, etc.) that generate vectors in a multi-
dimensional space. The PCA is a statistical approach that allows
reducing the descriptors to three principal components; in this way,
solvents that are expected to behave similarly are close in space. For a
more statistical discussion: Eriksson, L.; Johansson, E.; Kettaneh-Wold,
N., Wold, S. Multi- and Megavariate Data Analysis, Principles and
Applications; Umetrics Academy, 2001; pp 43À48.
(11) Partial Least Squares (PLS) regression is a multivariate data
analysis technique which can be used to relate several response variables
to several explanatory variables. In this context this is used to find
relations among chemical properties of solvent and base. The PLS model
can be used to predict the outcome of the experiment for other
combinations of solvent, reagent and/or catalyst not evaluated in the
initial array. See also Eriksson, L.; Johansson, E.; Kettaneh-Wold, N.,
Wold, S., Multi- and Megavariate Data Analysis, Principles and Applica-
tions; Umetrics Academy, 2001; pp 43À48.
(12) The IG-1 compound as monohydrate form was not a problem
for this process; even in case of its precipitation, it could be fully removed
in the following Step 3 where it was demonstrated that its formation was
due to the amount of water left in the n-propanol solution before starting
the precipitation.
1296
dx.doi.org/10.1021/op200140v |Org. Process Res. Dev. 2011, 15, 1287–1296