Synthesis and Optimization of Canagliflozin by Employing Quality by Design (QbD) Principles

Article abstract of DOI:10.1021/acs.oprd.7b00281

Efforts toward a synthesis and process optimization of canagliflozin 1 are described. Canagliflozin synthesis was accomplished via purified open ring intermediate 12. The process was optimized by employing quality by design (QbD) methodologies, and a telescopic strategy was executed for the first three and last two steps in a total six-step sequence. Optimization of the Friedel-Craft acylation reaction followed by Lewis acid mediated reductive elimination, n-BuLi mediated C-arylation, and reductive demethoxylation was performed to develop a robust process. These steps were found to be critical; therefore, critical process parameters (CPPs) were identified by employing design of experiment (DoE) methodology. In addition, control strategies for dealing with impurities are described.

Full text of DOI:10.1021/acs.oprd.7b00281

Subscriber access provided by READING UNIV
Full Paper
Synthesis and Optimization of Canagliflozin by Employing QbD Principles
Dattatray S. Metil, Swapnil P. Sonawane, Sharad S. Pachore, Aaseef Mohammad, Vilas H.
Dahanukar, Peter J. McCormack, Ch. Venkatramana Reddy, and Rakeshwar Bandichhor
Org. Process Res. Dev., Just Accepted Manuscript • DOI: 10.1021/acs.oprd.7b00281 • Publication Date (Web): 06 Dec 2017
Downloaded from http://pubs.acs.org on December 6, 2017
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Organic Process Research & Development is published by the American Chemical
Society. 1155 Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Synthesis and Optimization of Canagliflozin by Employing QbD Principles
ab
a
a
a
Dattatray S. Metil, Swapnil P. Sonawane, Sharad S. Pachore, Aaseef Mohammad, Vilas
a
c
b
H. Dahanukar, Peter J. McCormack, Ch. Venkatramana Reddy and Rakeshwar
a*
Bandichhor
a
API R & D, IPDO, Dr. Reddy’s Laboratories Ltd. Bachupally, Hyderabad, Telangana, India 500090.
b
Department of Chemistry, JNTU college of Engineering, JNT university, Kukatpally, Hyderabadꢀ500072,
Telangana, India.
c
Dr. Reddy’s Laboratories, Chirotech Technology Center, 410 Cambridge Science Park Milton Rd, Cambridge
CB4 0PE, U.K.
*
To whom correspondence should be addressed. Eꢀmail: rakeshwarb@drreddys.com
Tel: 91 40 4434 6429, Fax: 9140 4434 6285
.
1
ACS Paragon Plus Environment

Organic Process Research & Development
Page 2 of 38
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Abstract
Efforts towards a synthesis and process optimization of canagliflozin 1 are described.
Canagliflozin synthesis was accomplished via purified open ring intermediate 12. The
process was optimized by employing quality by design (QbD) methodologies and a telescopic
strategy was executed for the first three and last two steps in total six step sequence.
Optimization of the FriedelꢀCraft acylation reaction followed by Lewis acid mediated
reductive elimination, nꢀBuLi mediated Cꢀarylation and reductive demethoxylation were
performed to develop a robust process. These steps were found to be critical therefore critical
process parameters (CPPs) were identified by employing design of experiment (DoE)
methodology. In addition, control strategies for dealing with impurities are described.
Keywords
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Canagliflozin, typeꢀ2 diabetes, synthesis, QbD, DoE, impurity control strategy
2
ACS Paragon Plus Environment

Page 3 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
Introduction
Typeꢀ2 diabetes mellitus is long term metabolic disorder and it is a chronic disease with
worldwide prevalence. It is characterized by high blood sugar, resistance to insulin and a
1
a
deficiency in insulin secretion. Symptoms of typeꢀ2 diabetes include frequent urination,
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1b
increased thirst and hunger and weight loss. Patients having typeꢀ2 diabetes are found to
have a higher risk of heart disease, strokes and poor blood flow in limbs. Sudden onset of
1(cꢀd)
hyperosmolar hyperglycemic state also can occur.
Canagliflozin 1 is an antidiabetic drug of the gliflozin class or subtype 2 sodiumꢀglucose
2
(aꢀb)
cotransporter 2 (SGLTꢀ2) inhibitors,
it is used to improve glycemic control in patients
2
a,c
with typeꢀ2 diabetes. SGLTꢀ2 is an enzyme in the kidney tubule that causes glucose to be
2b
reabsorbed from urine. Canagliflozin blocks the action of SGLTꢀ2. Thus, canagliflozin
reduces the reabsorption of glucose from renal tubules leading to greater excretion of glucose
2d
in urine. Canagliflozin is the first SGLTꢀ2 inhibitor approved by FDA in March 2013. It is
developed by Mitsubishi Tanabe Pharma Corporation and marketed as Invokana® under
2
e
license by Janssen Pharmaceutical, Inc. a division of Johnson and Johnson for the treatment
3
of typeꢀ2 diabetes.
Figure 1. Structure of canagliflozin 1.
2c,4,5,6
There are many synthetic schemes reported for the synthesis of canagliflozin.
Most of
the reported synthesis start from the common halide intermediate 2a or 2b. More recently,
elegant βꢀselective Cꢀarylation of silyl protected 1,6 anhydroglucose with arylalane and other
methods that feature Cꢀarylation with diisobutylaluminium hydride modified 1,6
5
(aꢀd)
anhydroglucose with arylalane to obtain canagliflozin were reported.
In another report,
stereoselective direct Cꢀglycosylation of glycosyl bromide with arylzinc reagent was
3
ACS Paragon Plus Environment

Organic Process Research & Development
Page 4 of 38
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
5
(eꢀg)
described.
Although numerous reports on the synthesis of canagliflozin have appeared, to
the best of our knowledge, Cꢀarylation of readily available TMS protected gluconolactone 3
with the lithiated aglycon 2a or 2b followed by reduction or methoxylation and subsequent
reduction to canagliflozin is the most convenient and commonly practiced approach. This
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
c,4(aꢀh),6
approach is shown in Scheme 1.
4
ACS Paragon Plus Environment

Page 5 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
Route 1
3 (0.91 equiv)
S
F
1
) n-BuLi (1.0 equiv)
:1 THF:Toluene
70 °C 1 h
X
1
S
-
OMe
OH
F
2
) MsOH (2.73 equiv)
MeOH, r.t., 17 h
O
2
2
a : X = Br
b : X = I
HO
OH
91%
OH
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
O
TES (3.02 equiv)
BF3 Et2O (2.99 equiv)
OTMS
OTMS
O
1
TMSO
DCM, 0 °C, 2.0 h
5
6%
OTMS
3
Route 2
S
F
3
(1.3 equiv)
TES (2.66 equiv)
2
a : X = Br (Me)3SiCH2Li (1.3 equiv)
BF3 Et O (2.43 equiv)
2
OH
crude 1
2b : X = I
THF
OTMS
DCM
O
-
30 °C, 0.5 h
25 °C, 1.5 h
- 40 °C
TMSO
OTMS
OTMS
5
S
F
NMM (5.2 equiv)
Ac O (5.2 equiv)
LiOH (0.48 equiv)
2
1
DMAP (0.10 equiv)
DCM
OAc
OAc
THF, MeOH
25 °C, 12 h
O
AcO
2
5 °C. 1.0 h
OAc
6
Route 3
S
F
3
(1.2 equiv)
1
) i-PrLiCl.LiCl (1.0 equiv)
THF, 0-5 °C, 1 h
NMM (5.2 equiv)
Ac O (5.2 equiv)
2
2
a : X = Br
b : X = I
2
OMe
4
2
) MsOH (2.73 equiv)
MeOH, r.t., 17 h
DMAP (0.10 equiv)
Ethyl acetate
O
OAc
AcO
2
5 °C, 15 h
OAc
OAc
7
TES (2.66 equiv)
BF3 Et O (2.43 equiv)
2
LiOH (1 equiv)
THF, MeOH, water
5 °C, 19 h
6
1
acetonitrile, 0 °C, 4 h
2
Scheme 1. Reported Cꢀarylation approaches to the synthesis of canagliflozin 1.
Most of the reported processes have limitations with respect to the methodology used to
attain the desired purity (>99.50%) of the final canagliflozin API. Typically this is done by
some form of solvent crystallization. However, due to the nature of the final canagliflozin
5
ACS Paragon Plus Environment

Organic Process Research & Development
Page 6 of 38
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
API its crystallization in pure form is problematic. Thus the purification is accomplished in
many ways e.g. by converting crude canagliflozin into its Oꢀacetyl derivative 6 followed by
4
c,e,f
deprotection (Scheme 1 Route 2),
or converting methyl glycoside 4 to corresponding Oꢀ
acetyl derivative 7 followed by reduction to obtain 6 and subsequent deprotection to furnish
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
d
canagliflozin 1 (Scheme 1 Route 3). In another approach crude canagliflozin is purified by
6
coꢀcrystal formation followed by breaking the coꢀcrystal. All the above methods involve
additional protection/deprotection steps, tedious workup processes followed by isolation of
the product using solvent and antiꢀsolvent technique which led to moderate overall yield.
Additional protection and deprotection steps also generate comparatively more waste.
Considering this, it is important to develop a new and telescoped process which should be
viable at industrial scale. In order to gain significant cost advantage herein, we opted to
improvise the synthetic route for canagliflozin presented in Scheme 1 (Route 1). In this
article, we describe an improved and commercially viable telescoped process for the
synthesis of canagliflozin 1. The simple modification of isolating highly pure open ring
intermediate 12 in the reported process enabled us to achieve high purity of canagliflozin 1,
which is substantially free from all process impurities with an overall yield of ~47% for the
six steps, including telescoped synthesis of aglycon 2b, Cꢀarylation, methoxylation and
subsequent reduction to canagliflozin 1 as shown in Scheme 2. In contrast, intermediate 4
was a low melting tacky intractable solid in our hands that did not allow for any useful
purification. This new improvised route leverages the useful solid form features of 12 and
thus greatly eases the purification problems encountered in the conventional syntheses.
We also report the optimization of critical Cꢀarylation and reduction steps by employing DoE
methodologies for the identification of CPPs. The control strategies for the impurities are also
described.
Results and discussion
6
ACS Paragon Plus Environment

Page 7 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
Our improved synthesis (Scheme 2) commenced with the synthesis of aglycon 2b using
commercially available starting materials 8b and 10.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Scheme 2. Improved process for the synthesis of canagliflozin 1.
Process development of aglycon 2b
2c,4(dꢀf),7
Though there are many reports available for the synthesis of aglycon 2a-2b.
The
FriedelꢀCrafts acylation of 5ꢀhaloꢀ2ꢀmethylbenzoyl chloride 9a or 9b followed by reduction
2
c,4(dꢀf),7(aꢀf)
is the convenient and the most common route (Scheme 3).
a
Scheme 3. Reported FriedelꢀCrafts acylation approach to the synthesis of aglycons 2a-2b.
The reported synthetic sequence started with either 5ꢀbromoꢀ2ꢀmethylbenzoic acid 8a or 5ꢀ
iodoꢀ2ꢀmethylbenzoic acid 8b. In this approach, the acids were converted into the
corresponding acid chlorides followed by FriedelꢀCrafts acylation with 4ꢀfluorophenyl
thiophene 10 in presence of Lewis acid to furnish corresponding precursor halo ketones 11a
or 11b. The halo ketones 11a or 11b thus obtained were subsequently reduced by using
hydride source in presence of Lewis acid to afford corresponding aglycon 2a or 2b.
7
ACS Paragon Plus Environment

Organic Process Research & Development
Page 8 of 38
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
The first two steps of the most of the reported processes are carried out by using DCM as a
solvent. In addition, in the third step, the reduction reaction is carried out either in dimethoxy
ethane, THF, DCM, acetonitrile or in a mixture of DCM and acetonitrile. The reported prior
art processes involve synthesis of halo ketones 11a-11b and subsequent reduction using
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2c,7(dꢀe),4f
hydride reagents such as TES in presence of BF ꢁEt O,
1,1,3,3ꢀtetramethyldisiloxane
3
2
(bꢀc)
4e,7f
7a
(TMDS) in presence of AlCl
3
,7
NaBH
4
in presence of AlCl
3
3 2
or BF ꢁEt O.
In order to develop an improved process, we have screened various solvents, chlorinating
reagents, Lewis acids and hydride reagents (Table 1). After examining the solvent screening
results, chlorobenzene was found to be suitable solvent for both FriedelꢀCrafts and hydride
reduction reactions. With the intention to telescope the process into a single solvent,
chlorobenzene was selected as a solvent, which eventually avoided usage of low boiling
DCM or hazardous acetonitrile. An advantage in the use of chlorobenzene as a solvent over
the reported solvents is that its higher boiling point allows excellent recovery to be achieved.
This high boiling property of chlorobenzene also helps to reduce the vapor losses during the
extractive workup at large scale production, unlike DCM. This also helps in reducing the cost
and reduces issues around solvent abatement. After completion of third step, the reaction
mass was quenched with water followed by layer separation and pH adjustment of organic
layer towards basic to remove trace of acid.
Screening results indicated that, for the first two steps 1.3 equiv of SOCl and 1.1 equiv of
2
AlCl were found to be suitable coupling reagents and provided 97% conversion in
3
chlorobenzene (Table 1, entry 5). For the third step, TMDS was found to be an appropriate
hydride reagent for the reductive elimination reaction in the presence of Lewis acid, AlCl
3
(Table 1, entry 8).
Table 1. Screening of conditions for FriedelꢀCrafts reaction followed by reductive
elimination reaction
8
ACS Paragon Plus Environment

Page 9 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
Silane
Lewis
Purity Yield
Entry
Solvent
Reagent
Product
(hydride
acid
(%)
(%)
source)
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
c
c
a
1
.
.
oꢀXylene
oꢀXylene
SOCl
(COCl)
SOCl
(COCl)
SOCl
(COCl)2
2
AlCl
AlCl
AlCl
AlCl
AlCl
3
ꢀꢀ
11b
11b
11b
11b
11b
11b
2b
84
ꢀꢀ
ꢀꢀ
ꢀꢀ
ꢀꢀ
ꢀꢀ
ꢀꢀ
a
2
2
2
3
3
3
3
ꢀꢀ
ꢀꢀ
83
c
a
3.
Cyclohexane
2
98
b,c
4
.
.
Cyclohexane
ꢀꢀ
ꢀꢀ
c
a
5
Chlorobenzene
Chlorobenzene
DCM
2
ꢀꢀ
97
c
a
6.
AlCl3
ꢀꢀ
85
d
7
.
.
SOCl
SOCl
SOCl
SOCl
SOCl
2
2
2
2
2
AlCl
AlCl
AlCl
AlCl
3
3
3
3
TMDS
TMDS
TES
TMDS
TES
96
99
ꢀꢀ
82
d
8
Chlorobenzene
2b
83
b
9.
Chlorobenzene
2b
ꢀꢀ
ꢀꢀ
b
1
0.
1.
2b
ꢀꢀ
Chlorobenzene, acetonitrile
DCM, acetonitrile
d
1
BF
3
ꢁEt
2
O
2b
92
81
a
b
c
Area % by HPLC of crude compound. Product formation is very low, therefore not isolated. Solvent was
d
evaporated to obtain crude compound. Isolated solid.
After fixing the reaction conditions for the first two steps, we studied effect of TMDS equiv,
reaction temperature and chlorobenzene volume on the reductive elimination reaction. The
experimental results are presented in Table 2. It was observed that TMDS equiv and reaction
temperature had little impact on purity and yield, however chlorobenzene volume plays
crucial role to obtain higher purity and yield (Table 2). At lower volume of chlorobenzene
poorer yield and purity were observed, while at higher volume of chlorobenzene (>8 vol), the
obtained purity and yield was >90% and >80% respectively. The reason for the lower
conversion and yield may be explained by the fact that the reaction mass is heterogeneous
during the FriedelꢀCrafts reaction and at lower solvent volumes it becomes a thick slurry,
which is difficult to stir. This results in lower conversion and lower yield (Table 2, entry 2
and 5), therefore we opted to use 12 volume of chlorobenzene and 4.25 equiv of TMDS for
this reaction.
Table 2. Optimization of conditions for the reductive elimination reaction
9
ACS Paragon Plus Environment

Organic Process Research & Development
Page 10 of 38
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Entry TMDS equiv
Chlorobenzene volume
Temperature (ºC) Purity (%) Yield (%)
1
2
2.5
2.5
2.5
4
12
5
30
70
70
50
70
70
40
70
45
70
91
73
91
94
85
93
92
91
93
91
85
70
83
82
77
86
83
82
87
85
3
12
8
4
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
5
6
5
6
6
12
10
10
12
12
7
4
8
4
9
4.25
4.5
10
For this stage we have obtained overall >83% yield for the three steps. Having accomplished
an expedient synthesis of 2b, the next task was to optimize a suitable reaction conditions for
Cꢀarylation reaction.
Process development of intermediate 12
Various solvents and combinations of solvents including 2ꢀmethyl THF, THF and toluene,
were examined for Cꢀarylation of TMS protected gluconolactone 3. Apart from combination
of THF and toluene all other solvents furnished either lower yield or less purity, as shown in
Table 3. Besides the solvent screening, various bases and their combinations including nꢀ
BuLi, nꢀBuLi and nꢀBuMgCl, nꢀhexyllithium, isopropyl MgClꢁLiCl, lithium
bis(trimethylsilyl)amide and (trimethylsilyl)methyllithium were studied at different reaction
temperatures to accomplish coupling reaction. Eventually, nꢀBuLi along with mixture of THF
and toluene at ꢀ78 ºC was found to be the best combination for this reaction (Table 3 entry 5).
Hence we selected this combination for further optimization by using DoE, as described in
the subsequent section.
Table 3. Screening of the reaction conditions for Cꢀarylation reaction
1
0
ACS Paragon Plus Environment

Page 11 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
Entry
Solvent
2ꢀmethyl THF
THF
Base
nꢀBuLi
Temperature (ºC) Purity (%)
Yield (%)
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
d
1.
ꢀ78
ꢀ78
ꢀ20
ꢀ5
96
76
a
2.
nꢀBuLi
43
53
20
ꢀꢀ
ꢀꢀ
ꢀꢀ
a
a
3
.
.
THF
nꢀBuLi, nꢀBuMgCl
isopropyl MgClꢁLiCl
nꢀBuLi
4
THF, toluene
THF, toluene
THF, toluene
d
5.
ꢀ78
ꢀ78
ꢀ75
ꢀ50
98
99
ꢀꢀ
82
d
6
.
.
nꢀhexyllithium
31
b
7
THF
ꢀꢀ
ꢀꢀ
hexamethyldisilazanelithium
b
8.
THF
ꢀꢀ
(
trimethylsilyl)methyllithium
a
b
d
Area % by HPLC of crude compound. Product formation is very low, therefore not isolated. Isolated solid.
During screening of the reaction conditions, as expected around 10% of potential deshalo
impurity 13 (Figure 2) was observed in this reaction.
Figure 2. Structure of deshalo impurity 13.
Formation of impurity 13 during Cꢀarylation reaction clearly led to loss of yield and less
purity of the intermediate 12. In order to identify the vital process parameters which are
responsible for the formation of deshalo impurity 13 as well as to obtain maximum purity and
yield of 12 a screening DoE was performed. The purpose of this screening DoE is to evaluate
the number of input process parameters that have impact on response variables or in other
words to identify vital few parameters from the many insignificant others. Based on our
preliminary process understanding, we have screened different input variables (process
parameters) such as compound 3 equiv, nꢀBuLi equiv, reaction temperature, THF and toluene
volume that have an impact on response variables such as purity, amount of deshalo impurity
13 and amount of unreacted starting material 2b. A five factor definitive screening factorial
DoE design with one center point was performed. The screening DoE parameters, their
1
1
ACS Paragon Plus Environment

Organic Process Research & Development
Page 12 of 38
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
ranges and results are presented in the supporting information Table S1 and S2. ANOVA
analysis and contour plots of response variables for screening DoE for Cꢀarylation reaction
are presented in the supporting information Table S3ꢀS5 and Figure S1ꢀS3 respectively.
Based on ANOVA analysis and subsequent contour plots, it is evident that, purity of 12 is a
function of input variables such as reaction temperature and amount of nꢀBuLi equiv.
Whereas the amount of formation of impurity 13 is influenced by temperature and THF
volume. Other process parameters such as amount of compound 3 equiv and toluene volume
dont have such a significant impact on these response variables. It was also evident that a
superꢀstoichiometric equiv of nꢀBuLi is required to consume the starting material 2b. The
contour plot for the variation in impurity 13 with respect to THF volume and temperature
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(see supporting information Figure S2) revealed that medium amount of THF volume (~7
volume) is sufficient to control the formation of impurity 13 up to <6 %. Thus as an outcome
of the screening DoE, process parameters such as reaction temperature and amount of nꢀBuLi
equiv were identified as CPPs, therefore to optimize the ideal quantity of nꢀBuLi as well as
reaction temperature with respect to the formation of 13, these two parameters were subjected
to DoE optimization study with their redefined ranges. Low and high ranges of these two
parameters used in the DoE optimization study are summarized in Table 4.
a
Table 4. Two process parameters were considered for optimization DoE
Range
Symbol
Parameter
Unit
Low
ꢀ75
High
ꢀ50
A
B
Temperature
ºC
Quantity of nꢀBuLi
Equiv
1.4
1.8
a
2
A 2 full factorial design with two center points was planned to study
the effect of the two important input process parameters on the response
variables with all of the other parameters kept at predefined levels.
The process parameters such as compound 3 equiv, THF and toluene volume were considered
constant during DoE optmization study considering their predefined limits as shown in Table
5.
1
2
ACS Paragon Plus Environment

Page 13 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
Table 5. List of parameters that were held constant during optimization DoE studies
S. No.
Parameter
Limit
Justification
1
2
3
Quantity of 3
1.1 equiv
7 volume
minimum equiv studied during screening DoE
Quantity of THF
medium volume is preferable
no effect on responses
Quantity of toluene 3 volume
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
A 2 full factorial DoE design and the results of the same are captured in Table 6. In this
instance, the amount of starting material 2b as another response was not detected in the
experimental runs hence it is considered below detection limits.
2
Table 6. The 2 full factorial design for optimization of the Cꢀarylation reaction
Factors (process parameters) Response variable
S. No.
Impurity 13 (%)
Temperature n-BuLi equiv
1
2
3
4
5
6
ꢀ62.50
ꢀ75
1.6
1.8
1.8
1.4
1.4
1.6
6.14
6.29
6.44
6.21
3.96
5.07
ꢀ50
ꢀ75
ꢀ50
ꢀ62.5
Effects of process parameters on deshalo impurity 13
The halfꢀnormal plot, which displays significant effects to the right of the origin and Pareto
chart, where potentially important effects are shown in decreasing order of significance are
shown in Figure 3 and Figure 4 respectively. These plots indicat that the amount of formation
of deshalo impurity 13 was directly affected by the amount of nꢀBuLi, whereas variation in
the reaction temperature did not have any impact on this impurity within the studied range.
The same conclusion can be drawn from analysis of variance (ANOVA) (Table 7) and this
fact was augmented by the contour plot (Figure 5). Pareto chart (Figure 4), also revealed that
an interaction effect between these two factors was negligible for the formation of impurity
13.
1
3
ACS Paragon Plus Environment

Organic Process Research & Development
Page 14 of 38
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 3. Halfꢀnormal plot for deshalo impurity 13.
Figure 4. Pareto chart for deshalo impurity 13.
Figure 5. Contour plot for the variation of deshalo impurity 13 with respect to nꢀBuLi equiv
and temperature. Blue region indicates deshalo impurity content <5%, while progression
towards green and red region, indicates increasing amount deshalo impurity by increasing nꢀ
BuLi equiv.
1
4
ACS Paragon Plus Environment

Page 15 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
Table 7. ANOVA analysis for impurity 13
Sum of
squares
Degrees of
freedom
Mean
squares
5.1984
p-value prob
Source
F-value
> F
Model
5.1984
1
25.75111973
25.75111973
0.007107726 significant
B (nꢀBuLi
equiv)
5.1984
1
5.1984
0.007107726
Residual
Cor total
0.807483333
6.005883333
4
5
0.201870833
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Finally, a design space was generated from the contour plot and the model was validated by
conducting three validation experiments with 1.4 and 1.6 equiv of nꢀBuLi, 1.1 equiv of
compound 3 and at ꢀ75 ºC temperature. The results of the validation were found to be as
expected (yield >80% and purity >99%), and results were within the 95% confidence interval
(CI).
Identification of CPPs of
C
-arylation reaction
Based on understandings from the DoE studies, reaction temperature and nꢀBuLi equiv were
identified as CPPs for the Cꢀarylation reaction. In order to obtain better conversion, enhanced
purity and control over the formation of deshalo impurity 13, the reaction mass temperature
should be maintained between ꢀ75 to ꢀ50 ºC and amount of nꢀBuLi should be maintained
between 1.4−1.6 equiv. These advances have been demonstrated successfully at 40 kg scale
on plant.
During initial optimization of Cꢀarylation stage, we have synthesized methyl glycoside 4 by
quenching reaction mass in the mixture of MeOH and MsOH followed by extraction and
evaporation of solvent to obtain methyl glycoside 4 as an oily crude mass with ~87% of
purity by HPLC. The entire attempts for the purification to obtain high purity of methyl
glycoside 4 through solvent crystallization method was unsuccessful. Further carrying the
impure methyl glycoside 4 into the reduction step gave impure canagliflozin 1 and our
attempts to purify and achieve the desired quality canagliflozin from this were unsuccessful.
Hence we shifted our focus to the isolation of precursor 12 by hydrolyzing Cꢀarylation
reaction mass in the mixture of TFA and water. Eventually crystallization of 12 from a
1
5
ACS Paragon Plus Environment

Organic Process Research & Development
Page 16 of 38
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
mixture of THF and toluene afforded high HPLC purity (>99%) of compound 12 (see
supporting information Figure S21) with 84% of yield. Most gratifyingly all unwanted
impurities formed during Cꢀarylation reaction were purged into the mother liquor. The
deshalo impurity 13 which is the major impurity that forms during this stage is highly soluble
in the selected crystallization solvent. The crystallization process is highly efficient and
capable enough to wash up to 15% of deshalo impurity 13 to the extent of less than 0.3%.
The PXRD analysis confirms crystalline form of 12 (see supporting information Figure S19).
The pure intermediate 12 thus obtained was converted cleanly to methyl glycoside 4 which
was subsequently subjected to the reduction step to obtain canagliflozin 1. With this modified
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
process as described above, we were able to obtain high purity of canagliflozin 1. The prior
4f,8(aꢀb)
art has reported six membered pyranose structure (Figure 6)
in equilibrium, up to some
8
c
extent, with open chain structure (Figure 7) of 12.
Figure 6. Six membered pyranose form of 12.
To ascertain the exact chemical structure, compound 12 was subjected to 1D and 2D NMR
spectroscopy. The comprehensive structural elucidation of 12 was carried out unambiguously
1
13
using the H NMR spectroscopy, C NMR spectroscopy with and without proton decoupling,
COSY, HSQC, HMBC, NOESY spectroscopy (see supporting information Figure S8,
13
S10−S15). In a C NMR and HMBC correlations spectra, the chemical shift at 199.75 ppm
was observed. The analysis of the NMR data (see supporting information Figure S10 and
9
S13) confirmed that the structure of 12 is a ring opened keto form (Figure 7).
1
6
ACS Paragon Plus Environment

Page 17 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 7. The structure and atom numbering used in the assignment of 12.
This was further confirmed by FTꢀIR analysis. From the FTꢀIR spectra (see supporting
information Figure S18), it can clearly be seen that the spectra of 12 contains a signal at 1674
ꢀ1
cm , this signal is strongly indicative of a carbonyl ketone being present in 12.
Sugars and sugar derivatives like compound 12 when dissolved in solution, exist as a mixture
of tautomeric forms viz six membered pyranose form (Figure 6), openꢀchain form (Figure 7)
and five membered furanose form (Figure 8). This phenomenon is known as ringꢀchain
tautomerism. It is possible since the NMR based structural work was carried out in solution
that some of this tautomerism had occurred to give the ring opened/closed intermediate 12.
However, the FTꢀIR work was done on solid 12 as a KBr disc and the existence of the clear
carbonyl signal was a real indication that the isolated solid was in fact the ring opened
intermediate 12.
Figure 8. Furanose form of 12.
To further investigate the behavior of 12 in solution and the existence of mixtures of
1
tautomers in the solution, we have performed H NMR analysis of 12 in DMSOꢀd6 at
different time intervals such as immediately after dissolution of solid, 6 h, 2 days and 1 week
1
7
ACS Paragon Plus Environment

Organic Process Research & Development
Page 18 of 38
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
(see supporting information Figure S8−S9). It was found that, the compound 12 slowly
converts into a mixture of species indicating ongoing equilibration of ringꢀchain tautomerism.
After accomplishing facile synthesis of an intermediate 12, our focus was shifted towards
optimization of the final API stage.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Process development of canagliflozin 1
Synthesis of canagliflozin 1 starting from intermediate 12 involves two stages, in order to
simplify the process, two stages were telescoped. The penultimate stage is methoxylation of
an intermediate 12, which was accomplished by reacting 12 with MeOH in the presence of
MsOH (catalyst) at 10 °C for 4 h. After completion of the reaction, the reaction mixture was
3
quenched with saturated NaHCO solution and methyl glycoside 4 was extracted into DCM.
The organic layer was washed with water and excess solvent was distilled under vacuum to
obtain crude methyl glycoside 4 having HPLC purity >96%. The crude residue thus obtained
was dissolved in fresh DCM and then carried into the reduction stage without isolation.
The sixth and final stage of the synthesis is Lewis acid catalyzed reductive elimination
reaction. This is the key step in this synthesis; typically, this transformation has required
1
0
sterically hindered silane in order to obtain high βꢀselectivity. We have used a relatively
less expensive and readily available silane, TES for this reduction reaction to obtain
11
canagliflozin 1 with high selectivity for the βꢀanomer.
The cascading transformations for the stereoselective reduction of 4 to obtain canagliflozin 1
are shown in the Figure 9.
1
8
ACS Paragon Plus Environment

Page 19 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 9. Cascading transformations for the stereoselective reduction of 4 to 1.
Many alternative solvents and combinations including DCM, acetonitrile, dimethoxyethane
and 2ꢀmethyl THF were evaluated with respect to reaction conversion and impurity profile in
this stereoselective reduction reaction and experimental results are presented in Table 8. The
reaction conversion was very poor in all the solvents except DCM and mixtures of DCM and
acetonitrile. Hence, the use of DCM is considered appropriate for the reduction reaction.
Apart from solvent screening, other key parameters such as various silanes and Lewis acids
(catalysts) at different reaction temperatures were also studied. Using TMDS in presence of
AlCl , very little product formation was observed, whereas TMDS in presence of BF ꢁEt O
3
3
2
resulted in low purity. On the other hand higher yield and purity were obtained using
combination of TES and BF ꢁEt O (Table 8 entry 4). On the basis of these results, TES and
3
2
3 2
BF ꢁEt O were found to be appropriate combination for the reduction reaction.
Table 8. Screening of reaction conditions for reduction
1
9
ACS Paragon Plus Environment

Organic Process Research & Development
Page 20 of 38
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Temperature
HPLC (%)
Yield
(%)
51
ꢀꢀ
Entry
Solvent
Lewis acid
Silane
(°C)
Purity
4
BDL
ꢀꢀ
Impurity 14
a
1.
DCM/acetonitrile
DCM
AlCl
AlCl
3
3
TES
TMDS
TMDS
TES
ꢀ5
93
3
ꢀꢀ
4
2.
30
ꢀꢀ
3
.
.
DCM
BF
3
ꢁEt
2
O
ꢀ45
ꢀ45
ꢀ45
ꢀ45
30
77
10
ꢀꢀ
a
4
DCM
BF ꢁEt O
97
BDL
14
1.5
6
80
ꢀꢀ
3
3
3
3
3
2
2
2
2
2
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
a
5.
Dimethoxyethane
Acetonitrile
Acetonitrile
2ꢀMethyl THF
BF
BF
BF
BF
ꢁEt
ꢁEt
ꢁEt
ꢁEt
O
O
O
TES
52
6
.
.
TES
15
36
ꢀꢀ
34
2
ꢀꢀ
7
TES
58
3
ꢀꢀ
8.
O
TES
ꢀ45
65
22
ꢀꢀ
a
b
c
Area % by HPLC of crude compound. Product formation is very low, therefore not isolated. Solvent
d
evaporated to obtain crude compound. Isolated solid, BDL=Below detection limit.
Initial process understanding studies based on screening experiments for this reduction stage
indicated that the major hurdles to be overcome with this stage were, to increase the reaction
conversion, reduce the formation of furanose impurity 14 (Figure 10), achieve the ICH
quality and obtain an improved yield. The formation of substantial amount of furanose
impurity 14 lead to loss of yield, due to the repeated purifications to achieve desired ICH
quality of the final API.
Figure 10. Structure of furanose impurity 14.
In order to develop a robust process for the reduction step, an optimization program was
planned by employing DoE. The DoE study was helpful to identify the CPPs which most
strongly affect the response variables such as levels of furanose impurity 14 and purity.
Firstly, based on initial process understanding, process parameters such as amount of
BF
3 2 3 2
ꢁEt O and TES, volume of DCM, reaction temperature and addition time of BF ꢁEt O
solution were identified as potential CPPs. A five factors definitive screening factorial DoE
2
0
ACS Paragon Plus Environment

Page 21 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
with three center points was performed. The screening DoE parameters, their ranges and
results are presented in the supporting information Table S6 and S7.
From the analysis of definitive screening DoE results, it is evident that one of the responses,
purity, is significantly affected by the amount of BF ꢁEt O and volume of DCM whereas
3
2
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
another response, the amount of furanose impurity 14, is influenced by the amount of
BF ꢁEt O and reaction temperature parameter. Lower reaction temperature (ꢀ45 to ꢀ35 °C) is
3
2
favorable to obtain higher purity and to control the formation of furanose impurity 14,
therefore the reaction temperature was kept constant during the optimization DoE (Table 9).
It is also apparent from the screening DoE that, both the response variables (purity and
furanose impurity level) were not affected by variation of two other parameters: the amount
3 2
of TES and the addition time of BF ꢁEt O solution, hence these two parameters were also
considered constant during optimization DoE with predefined limits (Table 9). Eventually to
3 2
optimize the desired quantities of BF ꢁEt O and DCM with respect to the above two response
2
variables, a 2 factorial DoE with three center points was planned. Based on the knowledge
gained from screening DoE, ranges for these two factors were redefined and are presented in
Table 10.
Table 9. List of parameters that were held constant for optimization DoE studies
S. No.
Parameter
Limit
ꢀ45 to ꢀ35 °C
2.25 equiv
1ꢀ2 h
Justification
1
2
3
Temperature
lower temeparture is favorable
Quantity of TES
Additon time
no effect on responses
no effect on responses
a
Table 10. Two process parameters were considered for optimization DoE
Range
Symbol
Parameter
Unit
Low
High
2
A
B
Quantity of BF
3
ꢁEt
2
O
Equiv
Vol
1
8
Volume of DCM
12
a
2
A 2 full factorial design with three center points was planned to study
the effect of the two important input process parameters on the response
2
1
ACS Paragon Plus Environment

Organic Process Research & Development
Page 22 of 38
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
variables with all of the other parameters kept at predefined levels.
2
A 2 full factorial DoE design and the results of the same are captured in Table 11. It is clear
from the results that both the response variables are significantly influenced by quantity of
3 2
BF ꢁEt O whereas selected range of DCM volume as well as interaction of these parameters
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
have negligible effect on both the response variables.
2
Table 11. The 2 full factorial design for optimization of the reduction reaction
Factors (process parameters)
Response variables
Impurity 14 (%) Purity (%)
S. No.
BF ·Et O equiv
DCM volume
3
2
1
2
3
4
5
6
7
2
8
2.07
2.68
2.38
13.1
13.83
2.46
2.15
96.67
96.02
96.25
85.58
84.48
96.37
96.71
1.5
1.5
1
10
10
8
1
12
10
12
1.5
2
3 2
Effect of BF ·Et O equiv on furanose impurity 14
The effect tool in the DoE analysis such as halfꢀnormal plot and the Pareto chart (Figure 11)
show the significant parameters affecting the response variables and these indicated that the
3 2
amount of furanose impurity 14 is affected by the amount of BF ꢁEt O. This fact was
augmented by ANOVA analysis (Table 12). The contour plot (Figure 12) indicated that
amount of furanose impurity 14 tends to decrease by increasing amount of BF ꢁEt O.
3
2
2
2
ACS Paragon Plus Environment

Page 23 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 11. Halfꢀnormal plot and Pareto chart for impurity 14.
3 2
Figure 12. Contour plot for effect of BF ꢁEt O equiv and DCM volume on formation of
furanose impurity 14. Orange region indicates furanose impurity content >10%, while
moving towards green and blue region (more equiv of BF ꢁEt O equiv) shows that amount of
3
2
furanose is decreasing.
2
3
ACS Paragon Plus Environment

Organic Process Research & Development
Page 24 of 38
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Table 12. ANOVA analysis for furanose impurity 14
Degrees
Sum of
squares
Mean
p-value prob >
Source
of
F-value
squares
F
freedom
Model
128.936
128.936
1
1
4
6
128.936
128.936
1622.261
1622.261
< 0.0001
< 0.0001
significant
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3 2
A (BF ꢁEt O equiv)
Residual
Cor total
0.317916666
177.0606
0.079479166
Effect of BF ·Et O equiv on purity of 1
3
2
It is also apparent from the Pareto chart and halfꢀnormal plot (Figure 13), contour plot
Figure 14) and ANOVA analysis (Table 13) that the purity has direct relationship with the
equivalency of BF ꢁEt O. Purity tends to increase by increasing amount of BF ꢁEt O as seen
in contour plot (Figure 14).
(
3
2
3
2
2
4
ACS Paragon Plus Environment

Page 25 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 13. Halfꢀnormal plot and Pareto chart for purity of 1.
Figure 14. Contour plot for effect of BF
is shown to be increasing from 88% to 98% while proceeding towards red region of plot
more equiv of BF ꢁEt O equiv).
3 2
ꢁEt O equiv and DCM volume on purity of 1. Purity
(
3
2
2
5
ACS Paragon Plus Environment

Organic Process Research & Development
Page 26 of 38
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Table 13. ANOVA analysis for purity of 1
Degrees
of
freedom
Sum of
squares
Mean
squares
p-value prob >
Source
F-value
F
Model
135.9556
135.9556
1
1
4
6
135.9556
135.9556
812.8075
812.8075
< 0.0001
< 0.0001
significant
3 2
A (BF ꢁEt O equiv)
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Residual
Cor total
0.669066667
185.7529714
0.167266667
Finally, a design space was generated from the contour plot and the model was validated by
conducting three validation experiments with 1.7 and 2.0 equiv of BF ꢁEt O and at ꢀ45 ºC
3
2
temperature. The results of the validation were as expected, and results were within the 95%
CI.
Identification of CPPs for the reductive elimination reaction
Amongst all the process parameters studied in the reduction reaction during the screening and
optimization DoE it was found that reaction temperature and BF ꢁEt O equiv were found to
3
2
be CPPs with respect to response variables such as purity and furanose impurity 14 level. The
reaction mass temperature should be maintained between ꢀ45 to ꢀ35 ºC and amount of
3 2
BF ꢁEt O should be maintained between 1.75−2.0 equiv in order to obtain purity >96% and
furanose impurity 14 <3%. As a result, we have developed an efficient process and
successfully scaled up with 80 kg scale. After optimizing the reaction conditions, we then
focused on the isolation of crude and subsequent purification to obtain ICH quality
canagliflozin 1. In this step we are aiming to achieve >80% yield and >98% purity at crude
stage of canagliflozin 1. We screened different solvents and mixtures of solvents for the
purification and we found that combination of DCM and isopropyl acetate was suitable to
attain purity >99.80% and all other impurities below the specified levels. The crude isolation
and purification process for final stage is highly efficient and capable enough to wash up to
10% of furanose impurity 14 to the extent of less than 0.1%.
In our endeavors, with various screening techniques, process optimization employing DoE
and considering above modifications, the process was telescoped by avoiding the isolation of
2
6
ACS Paragon Plus Environment

Page 27 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
intermediates in the first three and last two steps. In the telescoped process, after completion
2
of the first step, reaction solvent was distilled to remove excess of SOCl . The obtained oily
residue was dissolved in fresh chlorobenzene and carried forward to the FriedelꢀCrafts
acylation step without isolation. After completion of the FriedelꢀCrafts acylation reaction, in
situ reduction was carried out by adding TMDS. The reaction mass was quenched into ice
cold water and the layers were separated. The organic layer was basified to pH 9−10 with
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
saturated NaHCO solution and was distilled up to 90%. Lastly the product 2b was
crystallized from MeOH at ꢀ5 °C with overall 83% yield for the three steps. The isolated
product showed >98% purity by HPLC. The fourth, Cꢀarylation step was carried out in the
mixture of THF and toluene using nꢀBuLi at ꢀ70 °C. After completion of reaction, the reaction
mass was quenched into a mixture of TFAꢀwater and this allowed TMS group deprotection at
about 0 °C for 3 h. The reaction mass was basified to pH 9−10 with saturated NaHCO
3
solution and the layers were separated. The organic layer, containing product, was washed
twice with water to remove inorganic salts and finally the product was crystallized by adding
toluene (as antiꢀsolvent) to furnish the intermediate 12 in 84% yield. The isolated product had
a purity by HPLC of >99%. The final two stages were telescoped by avoiding the isolation of
methyl glycoside 4. In the fifth step, after completion of methoxylation reaction to produce 4,
3
DCM was added and the reaction mass basified to pH 7−8 with aqueous NaHCO solution.
Thereafter, the organic and aqueous layers were separated. The organic layer was washed with
water and distilled under vacuum to obtain crude methyl glycoside 4 which was then
dissolved in fresh DCM and carried forward to the reduction stage without further isolation. In
the sixth and final stage, stereoselective reductive elimination was accomplished by using
3 2
BF ꢁEt O as Lewis acid and TES as hydride source at ꢀ45 °C. After completion of reaction,
the reaction mass was quenched with saturated NaHCO solution and the layers were
3
separated. The organic layer containing product was washed three times with a (1:4) mixture
2
7
ACS Paragon Plus Environment

Organic Process Research & Development
Page 28 of 38
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
of MeOH and water to remove inorganic salts and the mass was then distilled under vacuum.
The crude canagliflozin was crystallized from a mixture of MTBE and water in >80% yield
for two stages. Afterwards, the crude canagliflozin thus obtained was purified in a mixture of
isopropyl acetate and DCM to obtain final API. The product showed >99.80% purity by
HPLC. In the total process, the Cꢀarylation and reduction steps are critical; hence to identify
CPPs, DoE was employed during the optimization of these two steps. Finally, the redesigned
process furnished the canagliflozin with an overall yield of 47%, around 99.8% purity and
meeting other quality parameters (Table 14).
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Table 14. Final quality data of canagliflozin 1
Residual solvents (ppm)
Residual metal
Entry
Purity (%)
MTBE
ND
Isopropyl acetate
DCM
104
85
Chlorobenzene
content
1
2
3
99.80
99.85
99.87
899
858
674
ND
ND
ND
ND
ND
ND
ND
52
ND
ND = not detected
Impurity control strategy
The furanose impurity 14 forms during final reduction step. The rate of formation of this
impurity depends on the reaction temperature and the BF ꢁEt O equiv. Formation of this
impurity was controlled by optimizing the reaction temperature (ꢀ45 to ꢀ35 °C) and BF ꢁEt
3
2
3
2
O
equiv (1.75−2 equiv). We have developed highly efficient isolation and purification process
in the final stage which facilitates the ready removal of this impurity into the filtrate to an
extent that less than 0.10% remains in the isolated product by HPLC.
Impurity 15 is a regioꢀisomeric impurity of canagliflozin 1. The source of this impurity is the
starting material, 5ꢀiodoꢀ2ꢀmethylbenzoic acid 8b. The above starting material containing the
corresponding impurity undergoes similar transformations in the downstream process to form
2
8
ACS Paragon Plus Environment

Page 29 of 38
Organic Process Research & Development
1
2
3
4
5
6
7
8
9
impurity 15. This impurity is controlled by controlling the content of regioisomer in the
starting material. We have developed crystallization process for stage 3 (aglycon 2b) and
stage 4 (compound 12), which is efficient enough to purge this impurity into the filtrate
during their isolation.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Impurity 16 is the desfluoro analogue of canagliflozin 1. It is a process impurity and forms
during the synthesis of compound 12. The desfluorination reaction of aglycon 2b occurs in
the presence of strong base, nꢀBuLi. The desfluorinated compound reacts with TMS protected
gluconolactone 3 and is converted into impurity 16 during the course of reaction. The
selection of suitable reaction temperature is crucial to control the formation of this impurity.
By optimizing the reaction temperature (ꢀ75 to ꢀ50 °C), we are able to control the formation
of impurity 16 to a negligible level in the Cꢀarylation reaction. In addition it is further
washed out to an extent of less than 0.05% during the isolation of compound 12 from the
mixture of THF and toluene.
Impurity 17 is the αꢀisomer of canagliflozin and it forms in very low levels during final
reduction step due to steric hindrance of bulky silane used for the reduction. This impurity is
highly soluble in crystallization solvent, and gets washed out into the filtrate to an extent of
less than 0.05% in the isolated solid during isolation of crude and subsequent purification of
canagliflozin 1.
Impurity 18 is a dimeric impurity which was observed during scale up batches. This impurity
is highly soluble in crystallization solvent media and easily gets purged away during isolation
of the crude and subsequent purification of canagliflozin 1.
2
9
ACS Paragon Plus Environment