Development of an early-phase bulk enabling route to sodium-dependent glucose cotransporter 2 inhibitor ertugliflozin

Article abstract of DOI:10.1021/op400289z

The development and optimization of a scalable synthesis of sodium-dependent glucose cotransporter 2 inhibitor, ertugliflozin, for the treatment of type-2 diabetes is described. Highlights of the chemistry are a concise, four-step synthesis of a structurally complex API from known intermediate 4 via persilylation-selective monodesilylation, primary alcohol oxidation, aldol-crossed-Cannizzaro reaction, and solid-phase acid-catalyzed bicyclic ketal formation. The final API was isolated as the Lpyroglutamic acid cocrystal.

Full text of DOI:10.1021/op400289z

Article
pubs.acs.org/OPRD
Development of an Early-Phase Bulk Enabling Route to Sodium-
Dependent Glucose Cotransporter 2 Inhibitor Ertugliflozin
David Bernhardson,‡ Thomas A. Brandt,*,† Catherine A. Hulford,‡ Richard S. Lehner,† Brian R. Preston,§
Kristin Price,† John F. Sagal,† Michael J. St. Pierre,† Peter H. Thompson,† and Benjamin Thuma‡
‡
§
†Chemical Research and Development, Medicinal Chemistry, Research Analytical, Pfizer Worldwide Research and Development,
Eastern Point Road, Groton, Connecticut 06340 United States
S
* Supporting Information
ABSTRACT: The development and optimization of a scalable synthesis of sodium-dependent glucose cotransporter 2 inhibitor,
ertugliflozin, for the treatment of type-2 diabetes is described. Highlights of the chemistry are a concise, four-step synthesis of a
structurally complex API from known intermediate 4 via persilylation−selective monodesilylation, primary alcohol oxidation,
aldol-crossed-Cannizzaro reaction, and solid-phase acid-catalyzed bicyclic ketal formation. The final API was isolated as the L-
pyroglutamic acid cocrystal.
efficient route to ertugliflozin starting from D-mannose.4 Also,
INTRODUCTION
Ertugliflozin 1, a glucose-derived C-glycoside that contains a
novel bridged bicyclic ketal motif, is a selective sodium-
■
the strategy of arylating an acyclic Weinreb amide intermediate
was revisited for development of the commercial route.5
Synthesis of Known Compound 4. As a starting point for
a scalable synthesis of ertugliflozin (1), intermediate 4 can be
dependent glucose cotransporter (SGLT) 2 inhibitor for the
treatment of diabetes (Figure 1).1 SGLT2 inhibition provides
made from TMS-gluconolactone 2 and the requisite aryl
bromide 3 (Scheme 2) as reported in the literature.6,7 To
obtain 4, we modified the published procedure, which gave an
isolated product C1 anomer ratio of 85:15, by stirring the final
reaction mixture at room temperature until the ratio was ≥98:2.
This facilitated tracking downstream intermediates via HPLC
(i.e., only had to monitor a single diastereomer vs two
diastereomers). We also made modifications to the reaction
workup procedure to reduce solvent volumes and isolated tetrol
4 as the crystalline methanol solvate.8 The solvated methanol
Figure 1. Structure of SGLT2 inhibitor candidate ertugliflozin (1).
was removed via vacuum oven drying to obtain amorphous
product in 65−71% isolated yield and >98% purity.
an insulin-independent mechanism for regulation of glucose
Evaluation of C6 Oxidation without Protecting
homeostasis in type-2 diabetes mellitus (T2DM) patients.
Groups. Our strategy to introduce the requisite hydroxy-
SGLT2 is expressed exclusively in the renal proximal tubule
methylene functionality at the C5 position of the sugar required
cells of the kidney, and inhibition of this transporter promotes
the oxidation of the C6 hydroxy group to an aldehyde. A
urinary glucose excretion that helps to avert hyperglycemia in
number of attempts were made to selectively oxidize tetrol 4 to
diabetics.2 Manufacture of multikilogram quantities of 1 was
C6 aldehyde 6 with no secondary hydroxyl group protection.
required for toxicological and clinical evaluation, and this paper
Although enzymatic C6 oxidation is known for galactose using
describes the enabling work that resulted in the delivery of early
galactose oxidase (opposite C4 stereochemistry compared to
active pharmaceutical ingredient (API) campaigns.
glucose), there was no enzyme known to catalyze the analogous
Medicinal Chemistry Synthesis. The synthetic route
employed by our medicinal chemistry colleagues for analoging
efforts (Scheme 1) relied on late-stage introduction of the aryl
substituent via addition of a metalated aryl species to the
oxidation for glucose.9 Of the various conditions tried, the
cryogenic Swern protocol (DMSO, oxalyl chloride, NEt3)
showed promise with ∼50% conversion, but this initial result
Weinreb amide of an acyclic version of the glucose backbone.3
could not be improved upon as overoxidation became apparent
with increased amounts of oxidant. TEMPO-based oxidations
This route provided an acetophenone intermediate that
were tried with numerous stoichiometric oxidants with no
underwent cyclization followed by global deprotection to
positive results despite literature reports of the successful
provide final analogue targets. From a throughput perspective,
this route was long and linear with a low overall yield (13 steps
oxidation of sugars without C1 aryl substitution.10,11 Either no
reaction or overoxidation to the carboxylic acid was observed.
with 0.3% yield) and as such was not appropriate for
consideration as a route amenable to manufacture kilogram
quantities of API in a short time frame. More recently, our
medicinal chemistry colleagues have reported a much more
Received: October 14, 2013
© XXXX American Chemical Society
A
dx.doi.org/10.1021/op400289z | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 1. Medicinal chemistry route for late-stage introduction of structural diversity
Scheme 2. C-glycoside formation
Scheme 3. Kornblum oxidation of arylsulfonate esters
Another approach that showed some promise was a Kornblum-
type oxidation, which was carried out by selectively activating
the C6 primary hydroxyl group as a sulfonate ester and heating
in DMSO with base (Scheme 3).12 The p-toluenesulfonate
ester formed the desired oxidation product, but required high
reaction temperature and gave incomplete conversion along
with various byproducts, most notably sulfonate ester
hydrolysis to give back 4.
With the aim of lowering the reaction temperature required
for DMSO to displace the sulfonate ester and undergo clean
oxidation to give aldehyde 6, we then looked at sulfonate
esters13 with electron-withdrawing groups as shown in Table 1.
The p-nitro analogue (entry 1) worked reasonably well but had
B
dx.doi.org/10.1021/op400289z | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Revised Approach Using Protecting Groups. With
attempts at a scalable, selective, direct oxidation of the C6
hydroxy group exhausted, we turned to a global protection,
selective monodeprotection, and oxidation strategy. We looked
at peracetylation of tetrol 4 followed by selective C6 enzymatic
monodeacetylation, but after reaction optimization, the
conversion for the monodeacetylation reaction was only 20%.
We then looked at persilylation of 4 with TMSCl to give 7,
followed by monodesilylation using catalytic K2CO3 in MeOH
at 0 °C to provide alcohol 8 (Scheme 4).14 These conditions
worked fairly well but had the disadvantage of multiple solvent
exchanges: from the dichloromethane solvent for the
persilylation reaction, to methanol for the monodesilylation
reaction, and then to a nonalcoholic solvent for the subsequent
oxidation reaction. Also, upon prolonged exposure of the
monodesilylated product (8) to methanol during distillation,
overdesilylation products started to appear. Once over-
desilylation starts to occur, a statistical mixture of all silylated
products results. Since dichloromethane was identified as the
optimal solvent for the subsequent oxidation step, we sought to
identify conditions that would selectively desilylate the C6
TMS-ether without changing the solvent. A hint of how to
proceed came from a pilot of the persilylation reaction where
pyridine was used as solvent and base. After workup and
chromatography, there was 75% monodesilylated product 8
present. Attempts to use aqueous pyridine-HCl gave ∼65%
desired monodesilylated product but also produced a significant
amount of overdesilylated products. However, aqueous
pyridinium p-toluenesulfonate (PPTs) was found to cleanly
provide the desired monodesilylation product 8.
Table 1. Screen of arylsulfonate esters in the Kornblum
oxidation
b
a
b
5 equiv iPr2NEt used as base. Note: All substrates were screened
across time (5−30 min) and temperature (90−145 °C). For
comparison purposes, the 15 min residence time at 130 °C results
are shown. These reactions were run using 5 equiv collidine as base in
10 volumes DMSO as solvent unless otherwise noted.
The optimized conditions employed 5 equiv of aqueous
PPTs in 2 volumes of water and gave >90% conversion to
alcohol 8 within 8 h. Interestingly, the monodesilylated product
appears to be in equilibrium with persilylated starting material 7
as well as very minor amounts of overdesilylated products.
Another subtle observation was that when commercially
available PPTs was dissolved in water and then aged, the
profile for the monodesilylation reaction deteriorated over time
in that large amounts of overdesilylated products began to
appear. A likely culprit was that pyridine in the concentrated
aqueous PPTs solution was evaporating off, thus leaving free p-
TsOH available to indiscriminately promote silyl ether
hydrolysis. This was easily remedied by freshly making the
aqueous PPTs by mixing 5 equiv of p-TsOH with 5.5 equiv of
pyridine (0.5 equiv extra pyridine as a buffer). So the final
protocol was to make persilylated sugar 7, wash with water to
remove the imidazole−HCl byproduct, and then stir the crude
dichloromethane solution with concentrated aqueous PPTs for
8 h to provide monodesilylated product, 8. The phases were
simply separated, and the dichloromethane layer was washed
with pH 7 buffer, followed by azeotropic drying of the solution,
which was then ready for the subsequent oxidation step.
higher thermal potential compared to the chloro-substituted
sulfonate esters. The p-chloro analogue (entry 2) was less
reactive under the desired flow conditions, but did give
complete conversion in lab-scale batch pilots where the reaction
went to completion in 2 h at 135 °C. The 2,6-dichloro and
2,4,6-trichloro analogues (entries 3 and 4) gave the best results
and comparable reaction profiles, with the 2,6-dichloro
substrate giving slightly less hydrolysis byproduct 4.
When the synthesis of the 2,6-dichloroarylsulfonate ester
starting material (5) was scaled up to ∼100 g for a proof-of-
concept run on the flow system, it decomposed during silica gel
chromatography and during solvent removal in vacuo at 40−50
°C. The instability of our starting material, therefore, made the
Kornblum oxidation route nonviable for scale-up. Additionally,
the aldehyde product (6) was not isolable as a single compound
from the successful experiments. However, it was demonstrated
on small scale that the crude aldehyde solution could be carried
directly into the next step chemistry with success, although this
concept was not pursued further. Project timelines for API
delivery did not allow for further follow-up on the Kornblum
oxidation approach.
Scheme 4. Persilylation/selective monodesilylation
C
dx.doi.org/10.1021/op400289z | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 5. Oxidation of C6 hydroxy group to aldehyde
Parikh−Doering Oxidation. The oxidation reaction to
give aldehyde 9 used sulfur trioxide pyridine complex (Scheme
5). Initially, the reaction was run by dissolving SO3-pyridine in
DMSO, then adding this solution to the alcohol−triethylamine
solution at room temperature per standard protocol.15 This
performed well on lab scale but gave poor results in the first
Kilolab run due to longer processing times.
Scheme 6. Aldol crossed-Cannizzaro reaction
Follow-up experiments confirmed that as SO3-pyridine was
aged in DMSO at room temperature over 2 h, followed by
addition of alcohol 8, the conversion to aldehyde 9 dropped off
significantly (<50% conversion). Furthermore, react-IR data
showed that the DMSO−SO3 adduct degraded over time in
DMSO. Therefore, the procedure was modified so that alcohol
8, NEt3, and DMSO were dissolved in dichloromethane and
cooled to 10 °C, then SO3-pyridine was added portionwise as a
solid or as a slurry in dichloromethane. After stirring 3 h at 10
°C, the reaction was typically complete.
During a brief solvent screen (CH2Cl2, EtOAc, THF, 2-
MeTHF) of this reaction using 5 equiv of DMSO, impurity 10
(Figure 2) was observed at levels of 15% (CH2Cl2) to 100%
Figure 2. SO3-adduct impurity.
(THF, 2-MeTHF). By increasing the amount of DMSO to 40
equiv, this impurity was significantly diminished with EtOAc or
CH2Cl2 as solvent.
Aldol Crossed-Cannizzaro Reaction. In the next step,
aldehyde 9 was treated with base in the presence of
formaldehyde, promoting three sequential reactions: global
desilylation and enolization to form sodium enolate 11, an aldol
reaction with formaldehyde to give aldol product 12, and
crossed-Cannizzaro reduction16 of aldehyde 12 to give pentol
product 13 (Scheme 6).
The main issue with this reaction is formation of undesired
byproducts. Figure 3 shows the main impurities that have been
identified by LCMS. Compound 4 is the most significant
impurity and results from Cannizzaro reduction of 9 as well as
from desilylation of 7 (present in starting material 9). Initial
reaction conditions used aqueous K3PO4 or Cs2CO3 as base
and gave a ∼50:50 ratio of 13:4. This ratio was improved to
∼80:20 by switching to anhydrous reaction conditions and
using NaOEt as base. The elimination product 14 results from a
dehydration reaction occurring in place of aldehyde enolate
Figure 3. Aldol crossed-Cannizzaro reaction major byproducts.
addition to the formaldehyde electrophile. The carboxylic acid
byproduct 15 forms when the starting material aldehyde acts as
the hydride donor in the Cannizzaro reaction instead of
formaldehyde, thus becoming oxidized in the process. The
formation of self-aldol dimer 16 is surprising, especially at levels
of 10% or higher. This finding makes more sense after realizing
that the self-aldol dimer formation is likely not reversible (can
form stable intramolecular hemiacetal under basic conditions)
whereas the aldolization with formaldehyde has been shown to
be readily reversible by HPLC. Carboxylic acid 15 and self-aldol
D
dx.doi.org/10.1021/op400289z | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 7. Aldol crossed-Cannizzaro reaction conditions
Scheme 8. Bicyclic ketal formation
dimer 16 formation were suppressed by increasing to 20 equiv
of paraformaldehyde.
oven drying of the 13-MeOH solvate drives off the methanol to
leave behind amorphous material.
Intermediates 7, 8, and 9 were all oils or foams, and thus, the
steps to make these compounds were telescoped together with
no isolations. The main challenge of the telescoped sequence in
converting tetrol 4 to pentol 13 was to avoid premature
desilylation of reaction intermediates. Therefore, processing in
the Kilolab was carried out continuously until product 13 was
isolated as a solid in 41% overall yield.
The final process (Scheme 7) was to take the crude DCM
solution of aldehyde 9 and exchange the solvent to ethanol, add
paraformaldehyde (20 equiv), and warm the resultant slurry to
55 °C (paraformaldehyde not in solution). Then NaOEt (2
equiv) was added as a 21 wt % solution in ethanol to give a
clear solution (NaOEt depolymerizes the paraformaldehyde).
When the reaction was complete, 18 equiv of aqueous sodium
bisulfite (NaHSO3) was added to the 55 °C solution and stirred
for 30 min. The NaHSO3 reacts with excess formaldehyde to
give the formaldehyde−sodium bisulfite addition compound.17
This renders the excess formaldehyde nonvolatile during
further batch processing. The remainder of the workup
consisted of ethanol removal via distillation and extraction of
product 13 into MTBE. Process safety testing of the aqueous
bisulfite side stream solution, by warming to 60 °C and holding
for 24 h, indicated the potential for the liberation of a
noncondensable gas (formaldehyde and/or sulfur dioxide).
After adjusting the pH of the side stream from its initial value of
5−6 up to 7, the pressure buildup in the test cell was <0.2 bar
upon re-evaluation. Thus, the pH of waste streams generated
during scale-up were adjusted to 7 before drumming up for
disposal.
During early pilots, pentol 13 was purified by flash
chromatography to remove the numerous reaction byproducts.
The purified product spontaneously crystallized from a
MeOH−toluene solution as the MeOH solvate. This material
was subsequently used to seed a crude methanol solution of 13
post workup and afforded isolated pentol product in 97−98%
purity. Interestingly, major impurity 4, which was present at a
∼20% level in the crude methanol solution, purged very well.
This finding was even more surprising, given the fact that tetrol
4 was also isolated as the crystalline MeOH solvate in the first
step of the synthesis as mentioned above (Scheme 2). Vacuum
The initially isolated pentol product 13 typically contained
2−3% of tetrol 4. If this impurity is carried forward, it forms a
dimer with final API that cannot be purged during
crystallization of API, so a recrystallization of 13 was
performed. Dried amorphous 13 containing impurity 4 quickly
dissolved in MeOH and then gave a very thick suspension.
When this mixture was warmed to 50 °C and held for 1 h, it
became free-flowing, likely due to a form conversion. The
isolated recrystallized product was obtained in 94% yield and
≥99.8% purity. Once again, the solvated methanol was
removed via vacuum oven drying to give amorphous 13.
Thorough drying was required as residual MeOH levels of >5
wt % lead to incomplete reaction in the subsequent step.
Formation of Final API. The conversion of pentol 13 to
bridged bicyclic ketal 1 was originally carried out in the
presence of stoichiometric trifluoroacetic acid (Scheme 8).
Although the cyclization readily occurred with 10 mol % TFA,
using a solid-phase acid catalyst was more convenient for post
reaction removal. Using a 1−2 mol % loading of SiliaBond tosic
acid in dichloromethane gave clean conversion to 1.18
When the bicyclic ketal formation was carried out in acetate
solvents (EtOAc, iPrOAc), small amounts of acetylated
impurities were observed. The reaction in dichloromethane
typically goes to >98% completion after 1 h but needs to stir for
∼18 h to completely consume starting material. After reaction
completion, the solid-phase acid catalyst was simply filtered off,
and the solution of 1 was taken directly into the
E
dx.doi.org/10.1021/op400289z | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 9. L-Pyroglutamic acid co-crystal formation
Scheme 10. Overall synthetic scheme for ertugliflozin (1) ·L-PGA
cocrystallization step. The ability to start with very pure pentol
13 was a major advantage since the bicyclic ketal formation was
quantitative and gave off one equivalent of methanol as the only
byproduct. This was important since the cocrystallization
process did not purge organic impurities very well due to the
mostly aqueous nature (80% v/v) of the solvent mixture.
The final API exists as an amorphous foam that was
inappropriate for clinical development; however, treatment of 1
with L-pyroglutamic acid (L-PGA) provides a clinically viable
cocrystalline complex.19 A concentrated solution of 1 in 1:1 2-
propanol:water was warmed to 55 °C and treated with an
aqueous solution of L-pyroglutamic acid followed by cooling,
seeding, and drying to provide excellent yield of 1·L-PGA in a
typical ratio of 1.0:1.1 of 1:L-PGA (Scheme 9). Final API purity
was typically ≥99.9% by HPLC. The properties of the cocrystal
are interesting in that bicyclic ketal 1 will selectively dissolve in
most organic solvents except hydrocarbons and water, while L-
pyroglutamic acid is highly soluble in water. These divergent
properties made it challenging to identify a good solvent to
rinse solids from the tank to the filter and to rinse the filter
cake. In the end, since the solution contained only high-purity
API, the chilled mother liquor was used for rinsing operations.
The anhydrous 1·L-PGA cocrystal has poor solubility in
MTBE, so a high-yielding reslurry operation was possible. This
was employed for some lots to deagglomerate the initially
isolated solids as an alternative to milling, while other lots were
milled with a Fitz mill. The MTBE reslurry was also
demonstrated to remove non-API-related organic impurities.
The one-pot conversion of pentol 13 to 1·L-PGA was also
explored using L-PGA as acid to promote the bicyclic ketal
formation; however, the two step sequence described above
gave better yield and purity of API.
CONCLUSIONS
■
The overall synthesis for the manufacture of ertugliflozin (1) is
shown below in Scheme 10. The yield for the five-step process
was 26%. This synthesis was employed for three campaigns that
delivered seven discrete batches of API (from 1.1 kgA to 10.8
kgA) for a total of 43 kgA of API, enough to cover all project
needs from regulatory toxicology studies through the start of
phase 3 clinical trials. A few key highlights are the discovery of
novel selective monodesilylation conditions for the primary silyl
ether hydrolysis of the persilylated sugar, the efficient aldol
crossed-Cannizzaro reaction to provide the requisite quaternary
center at C5, and the identification of the MeOH solvate of
pentol 13 as a key purification point in the synthesis to set the
final API purity. This process is not commercially viable,5 so
this synthesis represents a “fit-for-purpose” route that facilitated
toxicology and clinical testing to demonstrate the viability of
ertugliflozin as a new treatment for diabetes.
F
dx.doi.org/10.1021/op400289z | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
solvate. 1H NMR (MeOH-d4, 600 MHz): δ 7.56 (d, J = 1.8 Hz,
1H), 7.47 (dd, J = 8.5, 2.1 Hz, 1H), 7.38 (d, J = 8.2 Hz, 1H),
7.11 (d, J = 8.2 Hz, 2H), 6.81 (d, J = 8.8 Hz, 2H), 4.10 (d, J =
15.3 Hz, 1H), 4.01 (d, J = 15.3 Hz, 1H), 4.00 (q, J = 7.0 Hz,
2H), 3.95 (dd, J = 12.0, 2.1 Hz, 1H), 3.83 (dd, J = 12.0, 5.6 Hz,
1H), 3.77 (t, J = 9.1 Hz, 1H), 3.61 (ddd, J = 10.0, 5.6, 2.0 Hz,
1H), 3.44 (t, J = 9.4 Hz, 1H), 3.11 (d, J = 9.4 Hz, 1H), 3.09 (s,
3H), 1.37 (t, J = 7.0 Hz, 3H). 13C NMR (MeOH-d4, 150
MHz): δ 159.0, 139.8, 139.3, 135.1, 133.2, 132.1, 130.9, 130.1,
128.5, 115.7, 102.6, 78.8, 76.2, 75.2, 71.9, 64.7, 63.0, 49.8, 39.5,
15.4. HRMS: (ESI−) Calcd for C22H27Cl1O7 (M − H)−:
437.1364, Found: 437.1373.
EXPERIMENTAL SECTION
■
General. TMS-gluconolactone 2 was purchased from SAFC
and Asymchem. 4-Bromo-1-chloro-2-(4-ethoxybenzyl)benzene
(3) was purchased from Asymchem.
Reactions were monitored by HPLC analysis using an
Agilent 1100 series HPLC equipped with a Waters XBridge
C18 column (4.6 mm × 75 mm), 3 min hold with 0.05%/60%/
40% (v/v/v) DEA/methanol/water then gradient elution from
0.05%/60%/40% (v/v/v) DEA/methanol/water to 0.05% (v/
v) DEA/methanol over 7 min, hold for 3 min, then from 0.05%
(v/v) DEA/methanol to 0.05%/60%/40% (v/v/v) DEA/
methanol/water with 2 min equilibration time; flow rate =
2.0 mL/min, 25 °C, 225 nm detection. Key retention times: 3
(3.42 min), 4 (1.28 min), 6 (4.10 min), 7 (4.53 min), 9 (4.14
min, 4.17 min − split peak), 13 (1.10 min), 1 (1.00 min).
HRMS data obtained on an Agilent 1290 with Agilent 6230
TOF LC/MS.
Preparation of ((2R,3R,4S,5R,6S)-6-(4-Chloro-3-(4-
e t h o x y b e n z y l ) p h e n y l ) - 6 - m e t h o x y - 3 , 4 , 5 - t r i s -
(trimethylsilyloxy)tetrahydro-2H-pyran-2-yl)methanol (6). A
solution of imidazole (18.3 kg, 267.9 mol) and 4 (24 kg, 54.7
mol) in DCM (240 L, 10 L/kg) was cooled to 5 °C (Note:
Some imidazole precipitates upon cooling.) and chlorotrime-
thylsilane (28.0 kg, 257.0 mol) was added over 40 min at ≤12
°C. The reaction was warmed to room temperature and stirred
for 1 h at which point HPLC confirmed reaction completion.
Note: The IPC sample was pulled and washed with water
before analyzing to show 98% conversion. Water (120 L, 5 L/
kg) was added followed by stirring for 20 min and then phase
separation. A solution of pyridinium p-toluene sulfonate was
made by dissolving p-toluenesulfonic acid monohydrate (51.5
kg, 270.7 mol) and pyridine (24.0 kg, 303.4 mol) in water (64
L, 2.7 L/kg). This PPTs solution was added to the crude DCM
solution of 7 and the biphasic mixture stirred for 8 h at room
temperature to give predominantly monodesilylation product 6.
Note: The ratio of monodesilylated product 6 to persilylated
starting material 7 was 92:8 (equilibrium ratio). The layers were
separated, and the DCM phase was washed with pH 7
phosphate buffer (120 L, 5 L/kg) . Note: buffer made by
dissolving NaH2PO4 (3.06 kg) and Na2HPO4 (5.38 kg) in
water (120 L). The DCM phase was azeotropically dried by
concentrating under partial vacuum to 38 L. The crude DCM
solution of 6 was taken directly into the next step with an
assumed yield of 90% (32.25 kg) based on HPLC assay.
Preparation of (2S,3R,4S,5R,6S)-6-(4-Chloro-3-(4-
e t h o x y b e n z y l ) p h e n y l ) - 6 - m e t h o x y - 3 , 4 , 5 - t r i s -
(trimethylsilyloxy)tetrahydro-2H-pyran-2-carbaldehyde (9).
Note: At the beginning of processing, an in-line aqueous
oxone scrubber was set up to oxidize the dimethylsulfide
liberated during processing. To the 38 L DCM solution
containing 6 (∼32.25 kg, 49.2 mol) was added DCM (124 L, 5
L/kg total DCM) followed by cooling to 10 °C. Dimethyl
sulfoxide (138.4 kg, 1771.3 mol) was added (exothermic)
followed by NEt3 (18.0 kg, 177.9 mol). Sulfur trioxide−
pyridine complex (22.0 kg, 138.2 mol) was slurried in DCM
(324 L, 10 L/kg) and added to the reactor over 45 min at <12
°C and stirred for 3 h at 10 °C. Note: For smaller kilolab runs
the SO3−pyridine was added in three portions as a solid and
gave slightly better conversion. Once reaction completion
(≤5% of 6) was confirmed by HPLC, water (149 L) was added
followed by warming to room temperature and separating the
phases. The DCM phase was washed with saturated aqueous
NH4Cl (149 L) and concentrated under partial vacuum to a
final volume of 57 L. To this solution was added EtOH (127 L)
followed by concentrating in vacuo to a final volume of 38 L.
GC headspace confirmed removal of DCM (<1%). The crude
EtOH solution of 9 was taken directly into the next step with
an assumed yield of 85% (27.33 kg) based on HPLC assay.
Preparation of (2S,3R,4S,5S,6R)-2-(4-Chloro-3-(4-
ethoxybenzyl)phenyl)-6-(hydroxymethyl)-2-methoxytetrahy-
dro-2H-pyran-3,4,5-triol (4). A solution of TMS-gluconolac-
tone (2) (14.09 kg, 30.2 mol) in toluene (17 L, 1.2 L/kg) was
cooled to −15 °C and held for later use.
A solution of 4-bromo-1-chloro-2-(4-ethoxybenzyl)benzene
(3) (8.00 kg, 24.6 mol) in toluene (48 L, 6 L/kg) and THF (16
L, 2 L/kg) was cooled to −84 °C followed by addition of
hexyllithium (2.3 M in hexane, 8.04 kg, 26.5 mol) over 20 min
at ≤−74 °C followed by stirring for 40 min. Once the
halogen−metal exchange reaction was deemed complete by
HPLC or React IR, the −15 °C toluene solution of 2 prepared
above was added over 1 h at ≤−74 °C and then stirred 1 h.
Next, a solution of MsOH (3.24 kg, 33.7 mol) in MeOH (40 L)
was added as a 10 °C solution in 40 L over 50 min at ≤−71 °C.
The reaction mixture was warmed to room temperature over 5
h and then stirred for 20 h to ensure >98% single anomer
product present by HPLC (actual anomer ratio 99.3:0.7). The
reaction was then basified by adding 5 M aqueous NaOH (3.1
L) and stirred for 1 h. The pH was checked by pulling a sample,
stripping the organic solvent, and then reconstituting in 2-
MeTHF/water and checking the pH of the aqueous phase
which was found to be pH 9−10. Precipitated inorganic salts
were filtered off and rinsed with toluene (24 L). The filtrate was
concentrated to 80 L in vacuo, then diluted with toluene (48 L),
and concentrated to 80 L total volume. This was done to
remove THF and MeOH (<1% each) before aqueous workup.
The crude toluene solution of 4 was diluted with 2-MeTHF (80
L) and washed with water (40 L) followed by concentrating the
organic phase in vacuo to 32 L. To this solution was added
toluene (48 L) followed by concentrating to 32 L in vacuo to
remove 2-MeTHF solvent (<1%). Note: 2-MeTHF cosolvent
was needed to give good phase separation during water wash.
The solvent was exchanged from toluene to MeOH by doing a
constant volume distillation in vacuo while adding MeOH (120
L) over 8 h at 30 °C. GC headspace results showed 0.05%
remaining toluene. After cooling to room temperature, the
MeOH solution of crude 4 was seeded with a MeOH slurry of
crystalline MeOH-4 solvate, stirred for 5 h at room temper-
ature, and then cooled to −15 °C and stirred 2 h before
isolating solids via filtration. The filter cake was washed with
heptane (8 L), and the solids were dried in a vacuum oven at 40
°C for 47 h to provide 7.69 kg (71%) of 4 as an off-white solid.
Note: Oven drying drives off MeOH such that the isolated
product is amorphous 4 and not the crystalline 4-MeOH
G
dx.doi.org/10.1021/op400289z | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Preparation of (2S,3R,4S,5S)-2-(4-chloro-3-(4-
ethoxybenzyl)phenyl)-6,6-bis(hydroxymethyl)-2-methoxyte-
trahydro-2H-pyran-3,4,5-triol (13). Note: Reactor vented
through an aqueous MeOH scrubber to sequester form-
aldehyde and an oxone scrubber to oxidize dimethylsulfide. To
the 38 L EtOH solution containing 9 (∼27.33 kg, 41.8 mol)
was added EtOH (194 L, 8.5 L/kg total EtOH). Paraformalde-
hyde (25.0 kg, 832.5 mol) was slurried in EtOH (76 L) and
added to the reactor followed by heating to 55 °C. To the
warm reaction mixture was added 20 wt % NaOEt in EtOH
(27.11 kg, 83.6 mol) over 5 min and the reaction stirred for 4 h.
Reaction completion (<5% of intermediate 12) was confirmed
by HPLC analysis. While still at 55 °C, a solution of sodium
bisulfite (78.3 kg, 752.4 mol) in water (272 L) was added over
20 min followed by stirring another 30 min before cooling to
35 °C and stripping off most of the EtOH solvent under partial
vacuum. The remaining aqueous phase was extracted with
MTBE (273 L) followed by phase separation. The MTBE
phase was washed with of water (164 L), then the water wash
was back extracted with MTBE (150 L). The two MTBE
extracts were combined and concentrated under partial vacuum
to a final volume of 40 L followed by addition of MeOH (109
L). The solution was concentrated under partial vacuum to a
volume of 40 L followed by addition of MeOH (109 L) and
concentrating to a volume of 91 L. GC analysis indicated no
residual MTBE was present. The crude MeOH product
solution was seeded with crystalline 13-MeOH solvate and
then granulated for 8 h at room temperature before cooling to 5
°C, stirring for another 2 h, collecting solids via filtration, and
rinsing with cold (−15 °C) MeOH (21 L) and room
temperature heptane (55 L). After vacuum oven drying for
three days at 40 °C, pentol 13 (10.63 kg, 41% from compound
4) was isolated as a off-white solid with 98% purity. Note: Oven
drying drives off MeOH such that the isolated product is
amorphous 13 and not the crystalline 13-MeOH solvate.
The main impurity 4 was purged to low levels from 13 by
doing a recrystallization from MeOH. To MeOH (55 L, 6.3 L/
kg) was added 13 (8.77 kg, 18.7 mol). The resulting slurry was
warmed to 50 °C and stirred for 1 h (initially thick slurry
becomes free-flowing) before cooling to room temperature and
granulating for 2 h. The slurry was cooled to −15 °C and
stirred for 2 h before collecting the solids via filtration followed
by rinsing with cold (−15 °C) MeOH (9 L) and room
temperature heptane (38 L). After vacuum oven drying for 23 h
at 50 °C, pentol 13 (8.23 kg, 94%) was isolated as a pure white
solid with 99.8% purity, and compound 4, at <0.2% level. Note:
Oven drying drives off MeOH such that the isolated product is
amorphous 13 and not the crystalline 13-MeOH solvate.
1H NMR (MeOH-d4, 600 MHz): δ 7.55 (d, J = 2.3 Hz, 1H),
7.50 (dd, J = 8.2, 2.3 Hz, 1H), 7.36 (d, J = 8.2 Hz, 1H), 7.11 (d,
J = 8.8 Hz, 2H), 6.81 (d, J = 8.8 Hz, 2H), 4.18 (d, J = 11.2 Hz,
1H), 4.10 (d, J = 15.2 Hz, 1H), 4.03−3.93 (m, 4H), 4.01 (q, J =
7.0 Hz, 2H), 3.85 (d, J = 11.7 Hz, 1H), 3.79 (d, J = 9.4 Hz,
1H), 3.13 (s, 3H), 3.08 (d, J = 9.4 Hz, 1H), 1.37 (t, J = 7.0 Hz,
3H). 13C NMR (MeOH-d4, 150 MHz): δ 159.0, 140.2, 139.7,
135.0, 133.2, 132.2, 131.0, 130.0, 128.6, 115.7, 103.7, 81.7, 78.9,
74.0, 71.9, 65.8, 64.6, 63.7, 51.4, 39.5, 15.4. HRMS: (ESI−)
Calcd for C23H29Cl1O8 (M − H)−: 467.1470, Found: 467.1478.
Preparation of (1R,2S,3S,4R,5R)-5-(4-Chloro-3-(4-
ethoxybenzyl)phenyl)-1-(hydroxymethyl)-6,8-dioxabicyclo-
[3.2.1]octane-2,3,4-triol (1). To a solution of 13 (8.23 kg, 17.5
mol) in DCM (41 L, 5 L/kg) was added SiliaBond tosic acid
(516 g, 0.35 mol) followed by stirring for 18 h at room
temperature. HPLC confirmed reaction completion (<0.2% 13
remaining). The silica-bound acid catalyst was removed via
filtration through a 0.5 μm cartridge. The crude DCM solution
of 1 was taken directly into the next step with an assumed yield
of 100% (7.67 kg) based on HPLC assay.
Preparation of 1·L-PGA. The DCM solution of 1 (∼7.67 kg,
17.56 mol) was concentrated atmospherically to 19 L followed
by addition of iPrOH (40 L) and then concentrated in vacuo
down to 18.5 L (2.4 L/kg) at 35 °C. GC analysis confirmed
DCM removal (<0.5%). The iPrOH solution of 1 was partially
cooled to 30 °C, and water (10.7 L, 1.4 L/kg) was added,
followed by weighing the solution (26.0 kg). More iPrOH (0.9
kg) was added to bring total solution weight to 26.9 kg and was
heated to 55 °C. In a separate container was added water (31.7
L) and L-pyroglutamic acid (6.80 kg, 52.7 mol) to give a
solution that was then added to the iPrOH solution of 1 over 1
h followed by cooling to room temperature over 2 h. Seeds of
1·L-PGA were added to induce crystallization followed by
stirring for 18 h at room temperature, cooling to 3 °C, and
stirring another 2 h before collecting solids via filtration. The
first 5 L of mother liquor was collected and used to rinse the
remaining solids in the tank onto the filter followed by rinsing
the filter cake with room temperature heptane (2 × 23 L). The
isolated solids were dried in the vacuum oven at 35 °C for 6 h
and then at 55 °C for 9 h to give 1·L-PGA (9.59 kg, 96%) as a
pure white solid with HPLC purity of 99.9%. The ratio of 1:L-
PGA was 1.0:1.10.
1H NMR (MeOH-d4, 600 MHz): δ 7.46 (d, J = 1.8 Hz, 1H),
7.39 (dd, J = 8.2, 2.3 Hz, 1H), 7.36 (d, J = 8.2 Hz, 1H), 7.09 (d,
J = 8.8 Hz, 2H), 6.80 (d, J = 8.8 Hz, 2H), 4.26 (dd, J = 9.4, 4.7
Hz, 1H), 4.15 (d, J = 7.0 Hz, 1H), 4.04 (s, 2H), 3.99 (q, J = 7.0
Hz, 2H), 3.85 (d, J = 12.3 Hz, 1H), 3.79 (dd, J = 8.2, 1.2 Hz,
1H), 3.69 (d, J = 12.3 Hz, 1H), 3.66 (t, J = 7.9 Hz, 1H), 3.62−
3.58 (m, 1H), 3.56 (d, J = 8.2 Hz, 1H), 2.53−2.46 (m, 1H),
2.41−2.29 (m, 2H), 2.20−2.14 (m, 1H), 1.36 (t, J = 7.0 Hz,
3H). 13C NMR (MeOH-d4, 150 MHz): δ 181.3, 175.9, 159.0,
139.9, 138.7, 135.2, 133.0, 131.0, 130.7, 130.0, 127.3, 115.7,
109.8, 86.4, 79.5, 77.9, 73.3, 68.2, 64.7, 62.2, 57.1, 39.5, 30.6,
26.2, 15.4. HRMS: (ESI−) Calcd for C22H25Cl1N1O7 (M −
H)−: 435.1210, Found: 435.1216.
ASSOCIATED CONTENT
■
S
* Supporting Information
1
Copies of H and 13C NMR spectra for compounds 4, 13, and
1·L-PGA. This material is available free of charge via the
Internet at http://pubs.acs.org.
AUTHOR INFORMATION
■
Corresponding Author
*E-mail: thomas.brandt@pfizer.com
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
The authors gratefully acknowledge the Pfizer Kilo Lab and
B185 staff for their valuable contributions to this work, the
Pfizer Separation Sciences Group, Steven Boucher of the Pfizer
Process Safety Laboratory, Mike Burns from the Biocatalysis
Center of Emphasis, Jared VanHaitsma from Research
Analytical for HRMS data, and Alicia Gutierrez for help with
manuscript editing.
H
dx.doi.org/10.1021/op400289z | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
REFERENCES
■
(1) Mascitti, V.; Thuma, B. A.; Smith, A. C.; Robinson, R. P.; Brandt,
T.; Kalgutkar, A. S.; Maurer, T. S.; Samas, B.; Sharma, R.
MedChemComm 2013, 4, 101.
(2) Washburn, W. N. J. Med. Chem. 2009, 52, 1785.
(3) Mascitti, V.; Maurer, T. S.; Robinson, R. P.; Bian, J.; Boustany-
Kari, C. M.; Brandt, T.; Collman, B. M.; Kalgutkar, A. S.; Klenotic, M.
K.; Leininger, M. T.; Lowe, A.; Maguire, R. J.; Masterson, V. M.; Miao,
Z.; Mukaiyama, E.; Patel, J. D.; Pettersen, J. C.; Preville, C.; Samas, B.;
She, L.; Sobol, Z.; Steppan, C. M.; Stevens, B. D.; Thuma, B. A.;
Tugnait, M.; Zeng, D.; Zhu, T. J. Med. Chem. 2011, 54, 2952.
(4) Mascitti, V.; Preville, C. Org. Lett. 2010, 12, 2940.
(5) See paper immediately following this paper for details of the
commercial route optimization for ertugliflozin.
(6) Meng, W.; Ellsworth, B. A.; Nirschl, A. A.; McCann, P. J.; Patel,
M.; Girotra, R. N.; Wu, G.; Sher, P. M.; Morrison, E. P.; Biller, S. A.;
Zahler, R.; Deshpande, P. P.; Pullockaran, A.; Hagan, D. L.; Morgan,
N.; Taylor, J. R.; Obermeirer, M. T.; Humphreys, W. G.; Khanna, A.;
Discenza, L.; Robertson, J. G.; Wang, A.; Han, S.; Wetterau, J. R.;
Janovitz, E. B.; Flint, O. P.; Whaley, J. M.; Washburn, W. N. J. Med.
Chem. 2008, 51, 1145.
(7) Deshpande, P. P.; Singh, J.; Pullockaran, A.; Kissick, T.;
Ellsworth, B. A.; Gougoutas, J. Z.; Dimarco, J.; Fakes, M.; Reyes,
M.; Lai, C.; Lobinger, H.; Denzel, T.; Ermann, P.; Crispino, G.;
Randazzo, M.; Gao, Z.; Randazzo, R.; Lindrud, M.; Rosso, V.; Buono,
F.; Doubleday, W. W.; Leung, S.; Richberg, P.; Hughes, D.; Washburn,
W. N.; Meng, W.; Volk, K. J.; Mueller, R. H. Org. Process Res. Dev.
2012, 16, 577.
(8) Gougoutas, J. Z.; Lobinger, H.; Ramakrishnan, S.; Deshpande, P.
P.; Bien, J. T.; Lai, C.; Wang, C.; Riebel, P.; Grosso, J. A.; Nirschl, A.
A.; Singh, J.; Dimarco, J. D. PCT Int. WO/2008/002824, 2008.
(9) Sun, L.; Butler, T.; Alcalde, M.; Petrounia, I. P.; Arnold, F. H.
ChemBioChem 2002, 3, 781.
(10) Angelin, M.; Hermansson, M.; Dong, H.; Ramstrom, O. Eur. J.
̈
Org. Chem. 2006, 4323.
(11) Gunnars, S. PCT Int. Appl. WO/2001/034657, 2001.
(12) Kornblum, N.; Jones, W. J.; Anderson, G. J. J. Am. Chem. Soc.
1959, 81, 4113.
(13) Yamamura, H.; Kawasaki, J.; Saito, H.; Araki, S.; Kawai, M.
Chem. Lett. 2001, 30, 706.
(14) Hurst, D. T.; McInnes, A. G. Can. J. Chem. 1965, 43, 2004.
(15) Parikh, J. R.; Doering, W. E. J. Am. Chem. Soc. 1967, 89, 5505.
(16) Mazur, A. W.; Hiler, G. D. J. Org. Chem. 1997, 62, 4471.
(17) Viste, A.; Hogan, R. E. Proc. South Dakota Acad. Sci. 1985, 64,
70.
(18) Badley, R. D.; Ford, W. T. J. Org. Chem. 1989, 54, 5437.
(19) Mascitti, V.; Collman, B. M. PCT Int. Appl. WO/2010/023594,
2010.
I
dx.doi.org/10.1021/op400289z | Org. Process Res. Dev. XXXX, XXX, XXX−XXX