ARTICLE

pubs.acs.org/OPRD

An Efficient Process for the Manufacture of Carmegliptin§

Stefan Abrecht,† Jean-Michel Adam,† Ulrike Bromberger,‡ Ralph Diodone,‡ Alec Fettes,*,† Rolf Fischer,‡
Volker Goeckel,‡ Stefan Hildbrand,‡ G
erard Moine,† and Martin Weber‡
F. Hoffmann-La Roche Ltd., Pharmaceuticals Division, Grenzacherstrasse 124, CH-4070 Basel, Switzerland
†
pRED, Pharma Research and Early Development, Process Research and Synthesis
‡
Pharma Technical Development Actives

ABSTRACT: A short and high-yielding synthesis of carmegliptin (1) suitable for large-scale production is reported. The tricyclic
core was assembled efficiently by a decarboxylative Mannich addition
Mannich cyclization sequence. Subsequent crystallization-
induced dynamic resolution of enamine 7 using (S,S)-dibenzoyltartaric acid was followed by diastereoselective enamine reduction to
give the fully functionalized tricyclic core with its three stereogenic centers. The C-3 nitrogen was introduced by Hofmann
rearrangement of amide 28, and the resulting amine 10 was coupled with (S)-fluoromethyl lactone 31. Following cyclization to
lactam 13 and amine deprotection, 1 was obtained in 27
31% overall yield with six isolated intermediates.

1. INTRODUCTION overall yield. Besides the long linear sequence and the very low
Type 2 diabetes (T2D) is a disease characterized by high levels overall yield, several issues were identified in this synthesis: (i)
of blood glucose in the context of insulin resistance and relative selectivity in the Dieckmann cyclization (step 4); (ii) nonscalable
insulin deficiency. The disease has been classified by the Center hydride reduction (step 7); (iii) safety issues in the Curtius
for Disease Control and Prevention (CDC) as an epidemic, with rearrangement (step 10); (iv) optical resolution by preparative
the prevalence rates doubling over the past 15 years and expected HPLC on chiral stationary phase; (v) coupling of amine 10 with
to increase significantly worldwide, even in developing countries, rac-11 (step 10). This route was deemed unsuitable for large-
in the years to come.1 It is estimated that worldwide 285 million scale supply since it did not meet our criteria for safety,
people (6.6% of the population) suffer from diabetes.2 Inhibition robustness, and economical as well as ecological viability.
of dipeptidyl peptidase IV (DPP-IV) is a proven approach for the In this contribution, the efforts from Process Research and
treatment of T2D.3 DPP-IV is responsible for the degradation/ Process Development to devise a scalable route that addresses
deactivation of glucagon-like peptide 1 (GLP-1) by cleaving the the aforementioned issues and would allow for the production of
N-terminal two amino acids, resulting in reduced insulin carmegliptin on ton scale are described.8 Route exploration and a
secretion. scalable synthesis of the optically active fluoromethyl lactone 31
Carmegliptin4 (1) (Figure 1) is a new generation DPP-IV needed for the lactam side chain are described elsewhere.9
inhibitor developed at Roche, structurally distinct from earlier
drug candidates.5 The structure is characterized by a 2-amino-
2. RESULTS AND DISCUSSION
hexahydro-benzo[a]quinolizine core, featuring a densely func-
tionalized six-membered azacycle with three stereogenic centers 2.1. Decarboxylative Mannich Addition
Mannich Cycli-
and all the substituents adopting an equatorial orientation.6 The zation
Enamine Formation Sequence. In the Discovery
C-3 nitrogen is incorporated in a chiral lactam side chain Chemistry synthesis, the assembly of the tricyclic core was
possessing a fluoromethyl substituent in β-position which poses achieved by a Michael addition
Dieckmann cyclization seque-
a significant synthetic challenge. nce from rac-4. The Michael addition was found to exhibit a
The Discovery Chemistry route (Scheme 1) consists of a 21- high potential for a runaway behavior as a result of ethyl acrylate
step synthesis with a 16-step longest linear sequence.4b The (used in large excess) polymerization. For the first API delivery,
synthesis started with 2-(3,4-dimethoxyphenyl)ethylamine and the problem was solved by performing the Michael addition
its conversion to imine 3 by Bischler
Napieralski reaction. in acetonitrile under copper(I) catalysis.10 Additionally, the
Decarboxylative malonate addition was followed by a 1,4-addi- Dieckmann cyclization suffered from poor regioselectivity (4:3
tion and a Dieckmann reaction7 to give the tricyclic enol ester mixture of desired/undesired regioisomers) necessitating tedi-
rac-6. The enamine formation, diastereoselective hydride reduc- ous isomer separation and resulted in moderate isolated yield. A
tion (establishing the relative configuration of all three stereo- more efficient access to (()-enamine 7 was required.
centers), and Curtius rearrangement (introducing the C-3 The synthesis of the enaminoester rac-7 by double Mannich
nitrogen functionality) provided the bis-protected diamine rac- reaction had already been reported starting from acetone dicar-
9. Teoc (trimethylsilylethyl carbamate) cleavage and chiral boxylic acid (Scheme 2).11 This approach was briefly evaluated
HPLC separation afforded enantiopure amine 10 in ∼5% yield but not further pursued as the thermal stability of the amino acid
from amine 2. The lactam moiety was introduced by coupling of
amine 10 with acid chloride rac-11 and cyclization. Boc depro- Received: January 25, 2011
tection and crystallization finally led to carmegliptin in about 2% Published: May 05, 2011

r 2011 American Chemical Society 503 dx.doi.org/10.1021/op2000207 | Org. Process Res. Dev. 2011, 15, 503–514
Organic Process Research & Development ARTICLE

promising results were found with enamine rac-7. A screening

of chiral acids revealed that dibenzoyltartaric acid (DBTA) and
dibenzoyltartaric acid monodimethylamide give 1:1 salts of very
high diastereomeric purity (dr > 99.5:0.5) from ethanol in
reasonable yield. These hits could be reproduced and scaled
without difficulty to multigram quantities to afford the desired
salt in up to 45% yield after filtration.
Intriguingly, upon close inspection, these high yields were
Figure 1. Structure of carmegliptin. accompanied by virtually racemic mother liquors, hinting at a
concomitant racemization process of the enamine under acidic
intermediate rac-14 was questionable.12 Control of double conditions. Crystallization-induced dynamic resolutions15
Mannich addition on acetone dicarboxylic acid was an additional (CIDR) of related substrates with different electronic properties
liability. using camphorsulfonic acid have been reported previously.16
The solution consisted in first reacting imine hydrochloride 3 On the basis of these findings, a process was developed which
with 3-oxo-glutaric acid mono ethyl ester (16) to afford ester rac- utilizes the commercially available nonnatural enantiomer (S,S)-
15 (Scheme 3) and then perform a Mannich cyclization. These dibenzoyl-D-tartaric acid ((þ)-DBTA) as preferred resolving
two steps could potentially be telescoped; however, ester rac-15 agent17 and relies upon the in situ racemization of the dissolved
offered itself as a stable crystalline intermediate allowing easy fraction of enamine and concomitant crystallization of the
purification and was therefore isolated. On the other hand, the desired diastereomeric salt as thermodynamic sink. Since race-
Mannich cyclization and enamine formation were best per- mization turned out to be rather slow at ambient temperature,
formed in a one-pot sequence. prolonged heating of the mixture was required to achieve good
Cyclic anhydride 1813 (Scheme 3) was prepared by addition of conversion. A reaction temperature of 60 °C in EtOH was found
acetone dicarboxylic acid (17) to acetic anhydride in acetic acid14 to be a good compromise between rate of racemization and
at 10
15 °C. The product crystallized from the reaction mixture yield.18 Under optimized conditions, the dibenzoyltartaric acid
and was isolated by filtration in ∼80% yield. salt of the enamine was obtained in 93% yield and diastereomeric
The 3-oxo-glutaric acid mono ethyl ester (16) was prepared by excesses greater than 99.5% (Scheme 4). As the racemization

addition of ethanol to a suspension of cyclic anhydride 18 in crystallization process went on, the mixture became very thick,
heptane and was directly used in the next step. The solution of making efficient stirring indispensable. Slight variation of various
the monoethyl ester was slowly added to a turbid solution reaction parameters (e.g., temperature, concentration, extended
consisting of imine hydrochloride 3 and catalytic amounts of reaction times, or cooling rate) did not reflect negatively on the
sodium acetate in an ethanol/water mixture. Water was added in chemical or optical purity of the product with diastereomeric
order to solubilize imine hydrochloride 3 which otherwise gave excesses of all batches greater than 99%.
an extremely thick suspension in ethanol. CO2 release was Up to 50 kg of (S)-20 were produced in-house. For larger
controlled by the addition rate as well as the amount of NaOAc. quantities, the synthesis of (S)-20 was outsourced to contract
A catalytic amount of base was required to trigger the reaction. manufacturers that were able to supply the material on ton scale.
The product was obtained in 81
86% yield (from imine hydro- In order to better understand the resolution process, the
chloride 3). The product crystallized spontaneously from the kinetics of the underlying racemization and its mechanism were
reaction mixture as the hydrochloride salt. Hydrochloric acid was investigated: (S)-Enamine 7 was submitted to acidic conditions,
added in order to compensate for the catalytic NaOAc and and the extent of racemization was followed as a function of time
heptane was additionally added to ensure maximum recovery of and temperature. A homogeneous reaction mixture was needed
the product. to guarantee reliable analytical sample preparation. This was
The Mannich cyclization was subsequently performed in achieved during the early phases of the reaction by using (R,R)-
methanol by reacting hydrochloride rac-15 with formaldehyde dibenzoyl-L-tartaric acid ((
)-DBTA), which gave the more
in the presence of NaOAc to buffer the reaction. The cyclic ester soluble diastereomer of the salt. The temperature interval studied
rac-6 was not isolated and was directly reacted with ammonium spanned from 40 to 70 °C, a range which could potentially be
acetate leading to enamine rac-7 which was isolated in ∼70% useful for the dynamic resolution process. The results are
yield (from ketoester 15) after aqueous workup and crystal- summarized in Figure 2.19
lization from methanol. Under the optimized conditions used for CIDR (EtOH, 60 °C,
2.2. Crystallization-Induced Dynamic Resolution. In the DBTA), the rate constant of racemization was determined to be
Discovery Chemistry synthesis, chirality was introduced as early k = 0.89 s
1. As expected, k increased linearly with the tempera-
as in the third step by nucleophilic addition to dihydroisoquino- ture (Figure 2) and correlated well with the pKa of the acid used
line 3 to give rac-4. Optical resolution of amine 10 was performed (ksulfonic acids > kTFA > kcarboxylic acids).
only at step 12 by preparative HPLC on chiral stationary phase Moreover, the racemization was studied in various solvents at
(Scheme 1). While this approach was viable for the production of 50 °C using (
)-DBTA. Figure 3 shows the extent of racemiza-
initial amounts of drug substance, an alternative access to tion as a function of time with the six solvents investigated.20
enantiomerically pure API avoiding preparative HPLC was These results clearly indicated that the rate of racemization
sought. As a result of time pressure, classical resolution by increases with polarity (MeOH > EtOH > 2-PrOH). Polar protic
diastereomeric salt formation was investigated first at the expense solvents seem favored over aprotic polar solvents (alcohols >
of an enantioselective synthesis which was anticipated to take
much longer to develop. DMSO ≈ 1,4-dioxane). Given the solvent dependency of the
Several intermediates deemed suitable for such a resolution pKa, these findings are in agreement with the observed increase in
process were submitted to a salt screening, and the most reaction rate with the pKa of the acid used.
504 dx.doi.org/10.1021/op2000207 |Org. Process Res. Dev. 2011, 15, 503–514
Organic Process Research & Development ARTICLE

Scheme 1. Discovery Chemistry synthesis of carmegliptina

a
Reagents and conditions: a) HCO2Me, Δ; b) POCl3, MeCN; c) HO2CCH2CO2Et, neat, 120 °C; d) ethyl acrylate, neat; e) t-BuOK, neat (5 steps); f)
NH4OAc, MeOH; g) NaBH4, TFA, THF; h) Boc2O, CH2Cl2; i) KOH, aq THF; j) DPPA, Et3N, TMSCH2CH2OH, PhMe, 80 °C; k) Et4NF, MeCN; l)
chiral HPLC; m) Et3N, CH2Cl2; n) NaH, DMF; o) HCl, dioxane; p) HCl, 2-PrOH.

To investigate the mechanism of the underlying racemization Scheme 2. Initial double-Mannich route to enol ester rac-6
process, trapping experiments were designed based on the
hypothesis that ring-opened intermediates are involved and
may be trapped using a suitable reducing agent. Indeed, earlier
experiments showed a slight erosion of enantiomeric purity
during the subsequent hydride-mediated reduction of enamine
(S)-7 to the corresponding amine, suggesting that scrambling of
the stereogenic center is operative under these conditions. Thus,
(S)-7 was treated with trifluoroacetic acid (TFA) followed by
addition of NaBH4. The resulting mixture was treated with
Boc2O in dichloromethane to facilitate isolation of byproducts
formed in trace amounts. The results of these experiments, which
led to the isolation of ring-opened 22, clearly favor the mecha-
nisms depicted in Scheme 5.
2.3. Synthesis of Boc-Protected Amine 25. The subsequent
reduction of the dibenzoyltartaric acid salt (S)-20 with NaBH4
and TFA establishes two additional stereogenic centers, making
it one of the key transformations of the overall synthesis. The
reaction conditions developed by Discovery Chemistry (TFA,
THF, rt, then NaBH4) led to 25 with good diastereoselectivity
following Boc protection, and the undesired diastereomers were
depleted by crystallization from heptane. However, these condi- To avoid these drawbacks, a scalable protocol was devised for
tions prescribed the addition of NaBH4 in one portion to a the first supply campaigns (up to several kg), which consisted in
solution of the enamine and TFA (40 equiv) to avoid the the preformation of a solution of enamine salt (S)-20 and TFA
formation of byproducts that were observed if the addition was (10 equiv) in THF followed by addition of this solution to a
performed portionwise. The resulting uncontrollable foaming suspension of NaBH4 in THF below
10 °C. It is worth noting
and vigorous hydrogen evolution limited this procedure to that the reaction proceeded equally well using the dibenzoyltar-
laboratory scale. taric acid salt (S)-20 instead of the free base (S)-7. Excess TFA
505 dx.doi.org/10.1021/op2000207 |Org. Process Res. Dev. 2011, 15, 503–514
Organic Process Research & Development ARTICLE

Scheme 3. Scale-up route to enaminoester rac-7a

a
Reagents and conditions: a) Ac2O, AcOH, 10
15 °C; b) EtOH, heptane, rt; c) NaOAc (0.1 equiv), MeOH/H2O, rt; d) aq CH2O, MeOH, rt; e)
NH4OAc, MeOH, 50 °C.

Scheme 4. Crystallization-induced dynamic resolution

was removed by concentration to dryness. Following basic

aqueous workup, the crude mixture was treated with Boc2O in
dichloromethane to give ester 25 in yields of up to 70% after
crystallization by solvent exchange to heptane. In this protocol
the addition of TFA to enamine salt (S)-20 and the subsequent
dosing of this solution to NaBH4 in THF were very exothermic,
and hence both required cryogenic reactors. To overcome this
drawback for subsequent larger campaigns, the reducing agent
was preformed in THF by slow addition of TFA to NaBH4
below
5 °C, followed by addition of enamine salt (S)-20.
Optimized amounts for the formation of the trifluoroacetoxy
borohydride reagent were found to be 5.0 equiv of TFA and 1.35
equiv of NaBH4. Under such conditions, the formal reducing
agent was determined to be NaBH1.7(O2CCF3)2.3 by measuring Figure 2. Extent of racemization as a function of temperature and time.
the amount of evolved hydrogen gas. This reagent, which has
only limited stability at temperatures above
10 °C, was prepa- ratio of approximately 13:1:1 (Scheme 6). The diastereomeric
red just prior to use and directly combined with enamine salt (S)- ratio was rather insensitive to the reaction temperature, the mode
20. After the addition the temperature was successively increased of addition, and the NaBH4/TFA ratio.
within 10 h to ambient temperature. The reaction mixture was 2.4. Introduction of the C-3 Nitrogen. In the Discovery
quenched by the addition of water and the pH adjusted with 50% Chemistry synthesis, the C-3 amine was introduced in a
aq KOH to pH 11.0
11.5. The need for dichloromethane can be three-step sequence consisting of ester saponification, Curtius

removed if the protection is carried out prior to extraction.21 Yamada rearrangement22 mediated by diphenylphosphoryl azide
Thus, the quenched solution was treated at 5
10 °C with a (DPPA)23 using 2-trimethylsilyl ethanol as nucleophile to give
solution of Boc2O (1.25 equiv) in toluene, and the resulting the Teoc-protected amine, and finally fluoride-induced depro-
mixture was allowed to warm up to room temperature. The pH tection. The upscaling of the Curtius rearrangement was not
had to be kept constant at approximately pH 10 to prevent ester possible from a safety perspective, due to its requirement of using
hydrolysis during the entire reaction period. The crude Boc- diphenylphosphoryl azide at a reaction temperature of 80 °C,
protected ester 25 was extracted with toluene and processed significantly above the maximum permissible safe process tem-
further. perature of 50 °C determined by in-depth studies in our safety
The reduction did not proceed with complete stereoselectiv- laboratory. To circumvent this safety issue, a Hofmann rearran-
ity. Besides the desired (2S,3S)-isomer 25, two other diastereo- gement, a well-established alternative to the Curtius reaction, was
mers, i.e. 26 (2S,3R) and 27 (2R,3S), were formed in a typical envisaged.24
506 dx.doi.org/10.1021/op2000207 |Org. Process Res. Dev. 2011, 15, 503–514
Organic Process Research & Development ARTICLE

Scheme 5. Isolation of ring-opened products from CIDRa

a
Reagents and conditions: a) i. TFA, THF, rt; ii. NaBH4, rt; iii. Boc2O,
CH2Cl2, rt.

bistrifluoroacetate (PIFA)29 offer advantages with respect to

Figure 3. Extent of racemization as a function of solvent and time. selectivity and reaction rate.
Initial attempts using PIDA or PIFA failed using the standard
The requisite amide 28 was found to be easily accessible from conditions reported in the literature (1.2 equiv PIDA or PIFA,
ester 25 by a mild amidation reaction with formamide (10 equiv) MeCN/H2O, rt), giving practically no conversion. However, in
and sodium methoxide (3 equiv) in THF.25 This reaction the presence of sodium hydroxide, isolated yields of >50% were
presumably proceeds by nucleophilic attack of the formamide obtained with PIDA (Scheme 8), along with significant amounts
anion to the ester to give a mixed imide, which is cleaved by of unsymmetrical urea 2930 and symmetrical urea 30 (up to 20%
sodium methoxide to generate the desired amide 28. combined) (Figure 4). These byproducts were only sparingly
For large campaigns, isolation and purification of ester 25 was soluble in organic solvents and turned out to be difficult to
avoided by telescoping the reduction
protection
amidation remove in the subsequent workup.
sequence (Scheme 7): This required that residual water be For the first supply campaigns, the importance of slow dosing
azeotropically removed after Boc protection to avoid ester of PIDA was recognized, resulting in decreased amounts of urea
hydrolysis under the basic amidation conditions. Both undesired byproducts 29 and 30 (8
10% combined). After reaction
cis diastereomers 26 and 27 (which were the main contaminants completion, acetonitrile and iodobenzene were distilled off,31
present in crude ester 25), partially epimerized to the thermo- the pH of the aqueous residue was adjusted to pH 9
10, and the
dynamically more stable trans isomers under the basic amidation product was extracted with dichloromethane. Crystallization was
conditions, affording crude amide mixtures containing all three achieved by solvent exchange to toluene. To remove 30 to
undesired amide stereoisomers in the range of 2
4% each. acceptable levels,32 a polishing filtration of both the dichloro-
Amide 28, which precipitated from the reaction mixture, could methane and hot toluene product solutions was required.
conveniently be isolated by filtration on kilogram scale. However, An extensive screening of the relevant reaction parameters
scale-up of this protocol was thwarted by bad filtration properties (amount of base, temperature, concentration, H2O/MeCN ratio,
of the crystals and hence associated long filtration times. The and rate of addition) revealed that urea byproducts can be
filterability of the crystals was improved by diluting the suspension minimized to <2% by adding a solution of PIDA (1.15 equiv)
with water followed by trituration at 60
65 °C for several hours. in H2O/MeCN within 3
4 h at 24
28 °C to a suspension of the
Moreover, this treatment dissolved all precipitated inorganic sodium amide and KOH (5 equiv) in water (8.3 kg/kg amide 28) and
salts. The impurity resulting from methoxide-induced transester- acetonitrile (4.3 kg/kg amide 28). The improved impurity profile
ification of the ethyl ester 25 with methanol26 was mostly hydrolyzed allowed considerable simplification of the workup by using
to the corresponding sodium carboxylate, which was completely toluene/THF 5:1 (w/w) at 70
75 °C for the extraction instead
removed in the mother liquor in the subsequent solid
liquid of dichloromethane: This high temperature and the use of THF
separation. Isolated yields of this telescoped three-step process as cosolvent ensured a clear phase separation by completely
were typically around 80%. All undesired isomers were comple- dissolving amine 10 as well as urea 30. Azeotropic removal of
tely separated into the mother liquor, yielding isolated amide 28 THF and water with toluene led to the precipitation of 30 which
in purities of >99.6% (HPLC). The only organic byproduct was filtered off at 70
80 °C. Crystallization of amine 10 from the
present in >0.10% was the aforementioned methyl ester with filtrate occurred upon cooling to
10 °C. Using this optimized
∼0.2% by area (HPLC). process, enantiomerically pure amine 10 (ee >99.8%) was obtai-
Many oxidizing agents have been reported for the Hofmann ned in isolated yields of 85
90% and purities >99% (w/w), while
rearrangement, including hypochlorite, hypobromite, and eliminating dichloromethane as solvent. The only organic by-
NBS as the most prominent examples.27 Modern variations product present in >0.10% was the urea derivative 30 with
based on iodosobenzene diacetate (PIDA)28 or iodosobenzene 0.15
0.20% by area (HPLC).
507 dx.doi.org/10.1021/op2000207 |Org. Process Res. Dev. 2011, 15, 503–514
Organic Process Research & Development ARTICLE

Scheme 6. Stereoisomers obtained by hydride-mediated enamine reduction following treatment with Boc2O

Scheme 7. Telescoped three-step process to amide 28 Scheme 9. Telescoped three-step process to fluoromethyl
lactam 13

Scheme 8. Hofmann rearrangement to amine 10

Reagents and conditions: a) 0.10 equiv of 6-chloro-2-pyridinol, PhMe,

110
85 °C; b) MsCl, Et3N, THF, 22
30 °C; c) t-BuOLi in THF,
10
to 0 °C, then aq workup and crystallization from MeOH (77
80%
overall yield).

Alternatively, the direct coupling of amine 10 with (S)-

fluoromethyl lactone 3133 to form the hydroxybutyramide 32
as a single diastereomer was investigated (Scheme 9). Subse-
quent activation of the primary alcohol was followed by treat-
ment with a strong base to induce intramolecular cyclization to
lactam 13.
The coupling step could be performed in good yield in the
Figure 4. Byproducts obtained in the Hofmann rearrangement. presence of catalytic amounts of 2-hydroxypyridine (0.20 equiv)
in refluxing toluene to provide hydroxybutyramide 32 in ∼93%
From a safety perspective, the reaction was unproblematic: The conversion.34 Owing to its low solubility in many solvents,
Hofmann rearrangement was dose-controlled with an enthalpy of product isolation was effected by simple filtration of the pre-
ΔrH =
415 kJ mol
1 and an associated total adiabatic temperature cipitated hydroxybutyramide 32. Subsequent activation of the
rise (ΔadiaTmax) of only 14 °C. primary alcohol as mesylate was carried out under standard
2.5. Formation of the Fluoromethyl Lactam. Originally, the conditions (MsCl, Et3N, THF) to give 33. Cyclization of the
lactam side chain was introduced by a two-step reaction sequence mesylate was achieved by treatment with LHMDS (3 equiv) at
consisting of coupling amine 10 with acid chloride rac-11, followed by
10 to 0 °C.35 To circumvent the drawbacks associated with the
intramolecular chloride displacement under basic conditions (NaH, use of this silicon-containing base on large scale,36 a screening
DMF, rt). Following workup, a 1:1 mixture of diastereomers was was initiated with the aim of minimizing byproduct formation.
obtained which were separated by silica gel chromatography to give With bases like sodium tert-butoxide, significant amounts
lactam 13 in 38% yield over two steps. The preparation of the acid (>10%) of the O-alkylated cyclized product were formed. Only
chloride from the corresponding lactone (SOCl2, ZnCl2, 80 °C, 4 d) lithium tert-butoxide37 showed a performance similar to that of
turned out to be very challenging, resulting in low yields (60%) and LHMDS (<2% O-alkylated byproduct) in the cyclization reac-
extensive byproduct formation, and was accompanied by a nauseat- tion and was therefore chosen as the preferred base on
ing stench that discouraged us from scaling up this reaction. larger scale.
508 dx.doi.org/10.1021/op2000207 |Org. Process Res. Dev. 2011, 15, 503–514
Organic Process Research & Development ARTICLE

Scheme 10. Deprotection and salt formation

(∼93% yield). However, this procedure requires an additional

processing step with associated loss in overall yield.
Figure 5. Conversion as a function of time, volume, and temperature in
The formation of alkyl chlorides was avoided by using acetone
the formation of amide 32.
as organic solvent:41 13 was suspended in a water/acetone
mixture, and concentrated aqueous hydrochloric acid (3.5 equiv)
A limiting factor in this three-step sequence was the filtration was added at ∼50 °C (Scheme 10). Gas evolution (isobutene
of the hydroxybutyramide 32, which turned out to be extremely and carbon dioxide) did not start until after dosing of the first
slow and could not be improved despite extensive efforts. It equivalent of acid. The maximum rate of gas formation was
became clear that isolation of this compound is very impractical observed just before the end of dosing. Subsequent dilution with
on large scale and should be avoided. A precondition for a acetone initiated the crystallization of carmegliptin (1), which
telescoped process was to achieve a higher conversion in the was obtained in 92
95% yield and excellent quality with no
coupling step with lactone 31, since residual amine 10 would be impurities >0.15% by area (HPLC) present.
mesylated, leading to contamination of the API with the corre-
sponding sulfonamide. The conversion in this equilibrium 3. CONCLUSION
reaction38 could be increased by ensuring as complete a pre-
In summary, a safe and robust process for the synthesis of the
cipitation of the hydroxybutyramide from the reaction mixture as DPP-IV inhibitor carmegliptin was developed (Scheme 11),
possible. This was achieved by adhering to a strict concentration which is suitable for large-scale production and uses readily
and temperature profile: The reaction was started by adding available starting materials. The synthesis of dibenzoyltartaric
lactone 31 (1.3 equiv) to a solution of amine 10 and 6-chloro-2- acid salt (S)-7 was successfully outsourced to several contract
hydroxypyridine39 (0.1 equiv) in toluene (18 mL/g amine) at manufacturers, which were able to provide this building block in
105
110 °C. After 5
7 h, conversion levels of 80
85% were excellent quality on ton scale.
typically reached, whereupon 50% of the toluene was distilled off, The salient features of our route include the efficient assembly
and at the same time, the temperature was decreased (to 85 °C) of the tricyclic skeleton by double Mannich reaction to give rac-7,
to induce the further precipitation of the product (Figure 5).40 its high-yielding crystallization-induced dynamic resolution
The reaction was usually judged complete (<1.5% amine) after using a cheap, commercially available tartaric acid derivative as
another 7
9 h at this temperature. resolving agent, and a stereocontrolled reduction of the resolved
Upon complete conversion, toluene was partly distilled off and enamine (S)-7 to provide the core of carmegliptin with its three
replaced by THF. Following mesylation and cyclization, aqueous stereogenic centers in high yield and excellent control of absolute
work-up, solvent exchange to methanol and subsequent crystal- and relative configuration. Moreover, the lactam side chain was
lization afforded fluoromethyl lactam 13 in 77
80% yield and introduced by coupling amine 10 with fluoromethyl lactone 31,
purities of >99.5% by area (HPLC). followed by intramolecular cyclization.
2.6. Deprotection and Salt Formation. Deprotection and Intensive process development (downstream from out-
hydrochloride formation was accomplished in a single step by sourced (S)-7) led to significant simplification and optimization
treatment of fluoromethyl lactam 13 with hydrochloric acid in a of the overall process. The overall yield of this eight-step
mixture of water and a suitable organic solvent. The choice of the sequence could be increased to over 50% with only three isolated
solvent proved to be crucial. With alcoholic solvents (MeOH, intermediates as a result of extensive telescoping. All undesired
EtOH, or 2-PrOH), good yields of 1 were obtained; however, solvents (e.g., dichloromethane) and genotoxic impurities were
considerable amounts of the corresponding genotoxic alkyl eliminated in the final process, with less than 200 kg of solvent
chlorides were formed, which could not be removed to levels being needed to produce 1 kg of API.42 Using the route outlined
acceptable for chronic treatment. With THF, ring-opening to above, nearly 1000 kg of API have been produced thus far.
afford 4-chloro-1-butanol and 1,4-dichlorobutane was observed
to some extent. Additionally ∼0.3% (w/w) of THF always 4. EXPERIMENTAL SECTION
remained in the isolated crystals, an amount which is far above
our internal acceptable limit for THF in drug substances. 4.1. General Remarks. Unless noted otherwise, reagents and
In order to obtain 1 devoid of genotoxic alkyl chlorides, two solvents were used as received from commercial suppliers.
approaches have been evaluated. First, removal of alkyl chlorides Technical grade solvents were used for all experiments. All
and, second, avoiding their formation. Removal of the alkyl reactions were carried out under a positive pressure of either
chlorides was successfully achieved by an additional recrystalliza- argon or nitrogen. Yields are weight-based and not corrected for
tion procedure from aqueous acetonitrile or aqueous acetone assays unless otherwise noted. The chemical purity and
509 dx.doi.org/10.1021/op2000207 |Org. Process Res. Dev. 2011, 15, 503–514
Organic Process Research & Development ARTICLE

Scheme 11. Manufacturing process of carmegliptin

Reagents and conditions: a) 16, NaOAc (0.1 equiv), MeOH, H2O; b) aq CH2O, MeOH rt then NH4OAc, 50 °C; c) (þ)-DBTA, EtOH, 60 °C, 15 h; d)
i) NaBH4, F3CCO2H, THF,
10 °C to rt; ii) Boc2O, PhMe/H2O, 5 °C to rt, (pH 10); iii) NaOMe, HCONH2, THF, 35 °C, then crystallization from
H2O/MeOH; e) PIDA, KOH, MeCN/H2O, 24
28 °C, then crystallization from PhMe; f) i) 31 (1.3 equiv), 6-chloro-2-pyridinol (0.1 equiv), PhMe,
110
85 °C; ii) MsCl, Et3N, THF, 22
30 °C; iii) t-BuOLi, THF,
10 to 0 °C, then crystallization from MeOH; g) HCl, acetone/H2O, 50 °C.
conversions were determined by HPLC (% by area). Enantio- cake was dried at 45 °C/20
30 mbar until constant weight to
meric excesses (ee) were determined by HPLC on chiral give 42 kg of the title compound (81% yield, 96% by area
stationary phase using the indicated conditions. 1H NMR spectra (HPLC) and 0.2% (w/w) residual EtOH). The product may
were recorded at 400 MHz using tetramethylsilane (TMS) as contain up to 1.8% (w/w) NaCl.
internal reference. Data are presented as follows: chemical shift 1
H NMR (400 MHz, DMSO-d6) δ 1.20 (t, J = 7.3 Hz, 3H),
(δ) in ppm, multiplicity (s = singlet, d = doublet, t = triplet, q = 2.83
3.01 (m, 2H), 3.18
3.35 (m, 2H), 3.34 (dd, J = 19 and 4
quartet, m = multiplet, br = broad), coupling constants (J) in Hz, Hz, 1H), 3.46 (dd, J = 19 and 8 Hz, 1H), 3.68 (d, J = 19 Hz, 1H),
and integration. 13C NMR spectra were recorded at 100 MHz. 3.71 (s, 3H), 3.73 (s, 3H), 3.77 (d, J = 19 Hz, 1H), 4.12 (t, J = 7
Absorptions in IR spectra are recorded in wavenumbers (cm
1). Hz, 2H), 4.76 (br dd, J = 8 and 4 Hz, 1H), 6.77 (s, 3H), 6.78 (s,
Low-resolution mass spectra were obtained by positive or 3H). MS ESI: m/z 322.2 ([M þ H]þ)
negative ion spray ionization (ESI). Data are reported in the 4.4. Preparation of (()-2-Amino-9,10-dimethoxy-1,6,7,11b-
form of m/z. tetrahydro-4H-pyrido[2,1-a]isoquinoline-3-carboxylic Acid
4.2. Preparation of Pyran-2,4,6-trione (18). Acetone dicar- Ethyl Ester (rac-7). Ketoester hydrochloride 15 (2.60 kg, 7.19
boxylic acid (32.5 kg, 222 mol, 1.00 equiv) was charged in the mol, 1.00 equiv) was charged in the reactor and dissolved by
reactor followed by AcOH (46 L). The suspension was cooled to addition of MeOH (15 L). A solution of NaOAc (0.590 kg, 7.19
15
20 °C, and Ac2O (35 L) was added. The reaction mixture mol, 1.00 equiv) in MeOH (6 L) was added. The addition system
was stirred for 4
5 h at 8
10 °C. A solution was obtained during was rinsed with MeOH (3 L). The obtained solution was
the course of the reaction after which the product started to transferred in a feed tank, and the reactor and transfer lines were
crystallize. The resulting suspension was filtered, and the filter rinsed with MeOH (5 L). The resulting solution was added at
cake was washed with PhMe (65 L) and dried at 45 °C/30 mbar 20
25 °C over 25
30 min to a solution of 37% aq CH2O (0.58
to give 22.5 kg of the title compound (79% yield). L, 7.55 mol, 1.05 equiv) in MeOH (13 L). The feed tank and
1
H NMR (400 MHz, DMSO-d6) δ 3.68 (s, 2H), 5.23 (s, 1H), transfer lines were rinsed with MeOH (10 L). The reaction
12.5 (br s, 1H). mixture was stirred for 4
5 h at 20
25 °C (IPC indicated
4.3. Preparation of (Z)-4-(6,7-Dimethoxy-1,2,3,4-tetrahy- conversion to intermediate rac-6), whereupon NH4OAc (1.7 kg,
dro-isoquinolin-1-yl)-3-hydroxy-but-2-enoic Acid Ethyl Es- 21.6 mol, 3.0 equiv) was added. The reaction mixture was heated
ter Hydrochloride (15). Cyclic anhydride 18 (22.3 kg, 174 mol, to 45
50 °C overnight and then cooled to 20
25 °C. The
1.20 equiv) was suspended in heptane (83 L) at room tempera- resulting red reaction mixture was concentrated under pressure
ture. EtOH (58 L) was added dropwise over 15
20 min while at ∼25
30 °C to an oily residue which was taken in CH2Cl2 (22
maintaining the temperature between 20
25 °C. After 1 h L) and H2O (11 L). The resulting pH 5
6 was adjusted to pH >
reaction, the resulting clear solution was transferred to a feed 7 by addition of aq NaHCO3 (1.5 kg in 15 L water). The organic
tank. The reactor and the lines were rinsed with EtOH. The phase was separated and washed with aq NaCl (1.7 kg in 15 L
resulting solution was added over 1.5 h to a turbid solution water). The aqueous phases were re-extracted with CH2Cl2 (20
consisting of imine hydrochloride 3 (33.0 kg, 145 mol, 1.0 equiv), L). The organic phases were combined, and solvent was ex-
NaOAc (1.20 kg, 14.5 mol, 0.10 equiv), H2O (8.5 L), and EtOH changed to MeOH (7 L final MeOH volume). The resulting
(250 L) while keeping the temperature between 20
25 °C. suspension was dissolved at reflux. The solution was cooled over
Crystallization of the product was usually observed during the 15 h to
25 °C to
20 °C and stirred for 5 h. The suspension
reaction. After 1.5 h, 37% aq HCl (1.5 L, 17 mol, 0.12 equiv) was was filtered, and the filter cake was washed with cold (
25 °C)
added to the resulting suspension. Heptane (240 L) was added MeOH (4 L) and cold (
25 °C) heptane (1.6 L). The crystals
over 30 min. The yellow suspension was cooled and stirred for 2 were dried under reduced pressure at 35 °C to give 1.83 kg of the
h at 0
5 °C. The suspension was filtered and washed with a cold title compound (71%, 97% by area (HPLC), 7.4% (w/w)
(0 °C) mixture of EtOH (55 L) and heptane (110 L). The filter residual MeOH).
510 dx.doi.org/10.1021/op2000207 |Org. Process Res. Dev. 2011, 15, 503–514
Organic Process Research & Development ARTICLE

1
H NMR (400 MHz, CDCl3) δ 1.28 (t, J = 7.3 Hz, 3H), 20
25 °C within 30 min and stirred at this temperature for 100 min.
2.40
2.72 (m, 4 H), 3.03
3.16 (m, 3 H), 3.43
3.52 (m, 3 H), During the entire reaction time the pH was kept constant at pH
3.73 (d, J = 13.7 Hz, 1 H), 3.86 (s, 6 H), 4.18 (m, 2 H), 6.59 (s, 1 ∼10 by the careful addition of 50% aq KOH (16 kg in total).
H), 6.60 (s, 1 H). 13C NMR (150 MHz, CDCl3) δ 168.56, Upon complete conversion (<0.5% by area (HPLC) of free amine),
154.24, 147.68, 147.46, 128.68, 126.92, 111.47, 108.25, 91.05, the stirrer was stopped, and the biphasic mixture was allowed to
59.05, 58.04, 56.10, 55.86, 53.67, 50.99, 37.56, 28.94, 14.62. separate for 20 min. The lower aqueous layer was separated and
HRMS (ESI): m/z 333.1 ([M þ H]þ); Expected mass: extracted with PhMe (240 kg). The combined organic layers were
332.1736; Found: 332.1734 (calculated for M). washed with H2O (180 kg) and concentrated under reduced
4.5. Preparation of (S)-2-Amino-9,10-dimethoxy-1,6,7,11b- pressure at 30
50 °C to almost dryness. The residue was treated
tetrahydro-4H-pyrido[2,1-a]isoquinoline-3-carboxylic Acid with THF (400 kg), and the resulting mixture was concentrated
Ethyl Ester, (S,S)-2,3-Bisbenzoyloxy-succinic Acid Salt ((S)- at 30
50 °C under reduced pressure to almost dryness. The residue
20). A 160-L Hastelloy reactor was charged with enamine rac-7 was diluted with THF (1500 kg), and the mixture was concentrated
(27.7 kg, 74.5 mol, containing ∼8% (w/w) MeOH) and EtOH at 30
50 °C under reduced pressure to a residual volume of 1400

(97 L). (S,S)-(þ)-Dibenzoyl-D-tartaric acid (28.5 kg, 79.5 mol, 1500 L. In the case of a complete solvent exchange (<0.10% (w/
1.06 equiv) was added under stirring followed by EtOH (41 L) to w) H2O, <8.0% (w/w) PhMe) the residue was treated at
give an off-white suspension. The mixture was heated to 60 °C 32
38 °C first within 10 min with formamide (131 kg, 2902
over 45 min and stirred at this temperature for 24 h. The thick mol, 10.0 equiv) and subsequently within 1 h with 30% NaOMe
yellowish suspension was cooled to ambient temperature over 8 in MeOH (156 kg, 868 mol, 3.0 equiv). The mixture was stirred
h and filtered. The crystals obtained were washed with EtOH (82 at 35 °C for 12 h. Upon complete conversion (<0.5% by area
L, 0 °C) and dried under reduced pressure at 45 °C for 24 h to (HPLC) of ethyl ester 25 and <3.0% of the corresponding
give 48.1 kg (93%, ee >99.5%) of the title compound as white methyl ester), the suspension was treated with H2O (740 kg).
crystals. The mixture was concentrated at 30
55 °C under reduced
1
H NMR (400 MHz, DMSO-d6) δ 1.20 (t, J = 7.2 Hz, 3H), pressure to a residual volume of 1200
1250 L. The suspension
2.41
2.29 (m, 1H), 2.68
2.59 (m, 1H) 2.86
2.72 (m, 1H), was diluted with MeOH (550 kg), and the resulting mixture was
3.02
2.90 (m, 1H), 3.15
3.05 (m, 1H), 3.40
3.27 (m, 2H), heated at 60 °C for 10 h. The suspension was cooled to 30 °C and
3.71 (s, 3H), 3.72 (s, 3H), 3.85
3.75 (m, 1H), 3.96
3.85 (m, subsequently centrifuged immediately. The crystals were washed
1H), 4.08 (q, J = 7.2 Hz, 2H), 5.74 (s, 2H), 6.62 (s, 1H), 6.81 (s, in two portions with a mixture of H2O (86 kg) and THF (66 kg)
1H), 7.58
7.53 (m, 4H), 7.73
7.66 (m, 2H), 8.03
7.96 (m, and dried at 70 °C and <30 mbar for 10 h to afford 100.4 kg
4H), 14.1 (br. s, 2H). 13C NMR (100 MHz, DMSO-d6) δ 14.5, (85%) of the title amide 28 as colorless crystals with an assay
26.8, 34.7, 49.5, 52.0, 55.4, 55.7, 57.1, 58.4, 71.7, 85.0, 109.0, (HPLC) of 97.8% (w/w).43
111.6, 125.1, 126.0, 128.8, 129.0, 129.3, 133.7, 147.4, 147.8, 1
H NMR (400 MHz, DMSO-d6) δ 1.17
1.25 (m, 1H), 1.38
155.6, 164.7, 167.0, 167.5. IR (cm
1): 3440, 3320, 2924, 2854, (s, 9H), 2.30
2.49 (m, 4H), 2.52
2.62 (m, 1H), 2.76
2.97 (m,
1734, 1618, 1376, 1260, 1237, 1129, 1027, 769, 714. MS (ESI): 3H), 3.07
3.12 (m, 1H), 3.70 (s, 3H), 3.71 (s, 3H), 3.65
3.75
m/z 333.1 ([M þ H]þ free amine). Anal. Calcd for (m, 1H), 6.61
6.71 (m, 3H), 6.84 (br s, 1H), 7.04 (br s, 1H). IR
C36H38N2O12: C, 62.6; H, 5.55; N, 4.06. Found: C, 62.45; H, (cm
1): 3446, 3346, 2923, 1684, 1663, 1535. MS (ESI): m/z
5.82; N, 3.77. [R]20D =
133.7° (c 1.025, DMSO). Chiral 406.2 ([M þ H]þ). Anal. Calcd for C21H31N3O5: C, 62.20; H,
HPLC: Chiralpak OD-RH column (5 μm, 250  4.6 mm), 7.71; N, 10.36. Found: C, 62.09; H, 7.76; N, 10.39.
MeCN/H2O: aq NaClO4 (0.5 M) 0.714:0.026:0.26, flow rate: 4.7. Preparation of (2S,3S,11βS)-3-(Amino-9,10-dimethoxy-
1.0 mL min
1, oven temperature: 40 °C, detection: 210 nm, Rt: 1,3,4,6,7,11β-hexahydro-2H-pyrido[2,1-a]isoquinolin-2-yl]-
14.0 min (R), 15.6 min (S). carbamic Acid tert-Butyl Ester (10). A suspension of amide 28
4.6. Preparation of (2S,3S,11βS)-3-(Carbamoyl-9,10-di- (120 kg, 296 mol) in a mixture of H2O (1000 kg) and MeCN
methoxy-1,3,4,6,7,11β-hexahydro-2H-pyrido[2,1-a]isoqui- (522 kg) was treated within 30 min at 15
28 °C with 50% aq
nolin-2-yl]-carbamic Acid tert-Butyl Ester (28). A 1000-L KOH (166.0 kg, 1479 mol, 5 equiv) and the resulting suspension
reactor was charged under a nitrogen atmosphere with NaBH4 stirred at 24
28 °C for 30 min. To the suspension was added
(14.8 kg, 391 mol, 1.35 equiv) and THF (550 kg). The suspen- within 3
4 h at 24
28 °C a solution of iodosobenzene diacetate
sion was treated at
17 to
10 °C within 4 h with TFA (162 kg, (106 kg, 329 mol, 1.10 equiv) in a mixture of H2O (270 kg) and
1421 mol, 5.0 equiv). (S)-20 (200 kg, 290 mol) was added as a MeCN (448 kg). After the addition, the suspension was stirred at
solid to the suspension within 5 h at
15 to
10 °C. The starting 24
28 °C for 1 h. Upon complete conversion (<0.1% by area
material container and the transfer lines were rinsed with THF (HPLC) of starting material), the suspension was concentrated
(100 kg). The reaction mixture was heated to 0 °C within 3 h and under reduced pressure and at a maximum internal temperature
stirred at this temperature for 3 h. The reaction mixture was of 45 °C (70 °C jacket temperature) to a residual volume of
heated to 20 °C within 3 h and stirred at this temperature for 5 h. approximately 1200 L. The pH of the mixture was adjusted to pH
Upon complete conversion (<1.0% by area (HPLC) of starting 9.5 by treatment with 37% aq HCl (33.8 kg) at 20
40 °C. THF
material), the reaction was transferred in a 2500-L vessel and (210 kg) and PhMe (1040 kg) were added at 20
40 °C, and the
quenched by the addition of H2O (658 kg). The mixture was resulting mixture was heated to 70
75 °C and stirred at this
stirred at 20
25 °C for 30 min and then cooled to 5
10 °C. The temperature for 30
60 min. The agitator was stopped, and the
pH of the mixture was adjusted at this temperature from biphasic mixture was allowed to separate for 30 min. The lower
approximately 1.8 to 11.0
11.5 by the careful addition of 50% aqueous layer was discharged and the organic layer washed with
aq KOH (187 kg). The mixture was treated at 5
10 °C within H2O (180 kg) at 70
75 °C. From the organic layer, THF and
30 min with a solution of Boc2O (80.5 kg, 369 mol, 1.27 equiv) in H2O were removed by azeotropic distillation with PhMe at a
PhMe (40.0 kg). The receiver and the transfer lines were rinsed maximum internal temperature of 70 °C. At the end of the
with PhMe (73 kg). The reaction mixture was allowed to warm to distillation the THF content should be <0.5%, and the volume of
511 dx.doi.org/10.1021/op2000207 |Org. Process Res. Dev. 2011, 15, 503–514
Organic Process Research & Development ARTICLE

the mixture was adjusted to 1700
1800 L. The mixture was layer was concentrated at <40 °C under reduced pressure to a
heated to 70
80 °C to obtain a dimmish solution. To remove residual volume of 150 L. MeOH (950 kg) was added and the
the urea byproduct 30, the solution was filtered at 70
80 °C. mixture concentrated at <40 °C under reduced pressure to a
The first reactor, the filter, and the transfer lines were rinsed with residual volume of 500
530 L. The resulting suspension was
hot PhMe (400 kg). The filtrate was concentrated at a maximum heated to reflux temperature (∼65 °C) and stirred at this
internal temperature of 80 °C under reduced pressure to a temperature for 30
60 min to obtain a clear (or dimmish)
residual volume of 1000
1100 L, whereby the product partly solution. The mixture was allowed to cool to
10 °C within 5 h
precipitated. The suspension was heated to 90 °C to obtain a and the resulting suspension stirred for 2 h at
10 °C. The
clear solution. The solution was cooled to
8 °C within 7 h and crystals were centrifuged, subsequently washed in portions with a
subsequently stirred at this temperature for 2 h. The crystals were cooled mixture (
5 °C) of MeOH (81 kg) and H2O (51 kg), and
filtered off using a centrifuge, washed in two portions with cooled dried at 45
50 °C/< 30 mbar to afford 88.8 kg (78% yield) of the
PhMe (<0 °C; 300 L), and dried at 60 °C/<30 mbar for 7 h to title compound as colorless to slightly yellow crystals with an
afford 94.6 kg (85%) of amine 10 as slightly yellow crystals with assay (HPLC) of 99.1% (w/w). 1H NMR (400 MHz, DMSO-d6)
an assay (HPLC) of 99.8% (w/w) with <0.10% of the corre- δ 1.37 (s, 9H); 1.37
1.49 (m, 1H); 2.07
2.15 (m, 1H);
sponding enantiomer (by chiral HPLC). 2.22
2.32 (m, 1H); 2.33
2.65 (m, 5H); 2.72
2.78 (m, 1H);
1
H NMR (400 MHz, CDCl3) δ 1.26 (br s), 1.34 (q, J = 12 Hz, 2.81
2.92 (m, 2H); 3.02
3.17 (m, 2H); 3.63
3.76 (m, 2H);
1H), 1.49 (s, 9H), 2.19 (t, J = 12 Hz, 1H), 2.48
2.66 (m, 3H), 3.70 (s, 3H); 3.71 (s, 3H); 3.85
3.93 (m, 1H); 4.33
4.42 (m,
2.67
2.75 (m, 1H), 2.92
2.99 (m, 1H), 3.01
3.15 (m, 2H), 1H); 4.45
4.55 (m, 1H); 6.64 (s, 1H); 6.78 (s, 1H); 6.92 (d, J =
3.18
3.24 (m, 1H), 3.39
3.51 (m, 1H), 3.837 (s, 3H), 3.844 (s, 8 Hz, NH). IR (cm
1): 3401, 2926, 2854, 1696, 1512, 1466,
3H), 4.55 (br s), 6.57 (s, 1H), 6.66 (s, 1H). IR (cm
1): 3355, 1010, 982. MS (ESI): m/z 478.1 ([M þ H]þ). Anal. Calcd for
2923, 2854, 1679, 1523, 1463, 1251. MS (Ion Spray) m/z 378.4 C25H36FN3O5: C, 62.87; H, 7.60; N, 8.80. Found: C, 62.49; H,
(MHþ). Anal. Calcd for C20H31N3O4: C, 63.64; H, 8.28; N, 7.56; N, 8.67.
11.13. Found: C, 63.77; H, 7.97; N, 11.15. Chiral HPLC: 4.8.2. Using t-BuOLi as Base for Cyclization. A 1500-mL
Chiralpak AD-H column (5 μm, 250  4.6 mm), n-heptane/ reactor equipped with a mechanical stirrer, a Pt-100 thermo-
isopropanol (with 0.1% NHEt2) 60:40, flow rate: 1.0 mL min
1, meter, a dropping funnel, and a nitrogen inlet was charged with
oven temperature: 30 °C, pressure: 76 bar, detection: 210 nm, Rt: amine 10 (30.0 g, 79.5 mmol) and 6-chloro-2-pyridinol (1.20 g,
4.8 min ((R)-10), 6.4 min ((S)-10). 9.10 mmol) in PhMe (480 g). The mixture was heated to
4.8. Preparation of (2S,3S,11βS)-3-((4S)-Fluoromethyl-2- 85
90 °C, and (S)-fluoromethyl lactone 31 (12.2 g, 103 mmol)
oxo-pyrrolidin-1-yl)-9,10-dimethoxy-1,3,4,6,7,11β-hexahy- was added within 60 min. After the addition, the mixture was
dro-2H-pyrido[2,1-a]isoquinolin-2-yl]-carbamic Acid tert- heated to 105 °C and stirred at this temperature for 6 h.
Butyl Ester (13). 4.8.1. Using LHMDS as Base for Cyclization. A
Approximately 300 mL of PhMe were distilled off, and the
solution of amine 10 (90.0 kg, 238 mol) and 6-chloro-2-pyridinol
resulting thick suspension (250
300 mL) was stirred at 85 °C
(3.60 kg, 27.8 mol, 0.120 equiv) in anhydrous PhMe (1180 kg)
for another 13 h. One hundred milliliters of PhMe were distilled
was heated to 85
90 °C and treated at this temperature within
off and replaced by 400 g of THF. At the end of the distillation a
60 min with (S)-fluoromethyl lactone 319 (36.6 kg, 310 mol, 1.30
equiv, 99.0
99.5% ee). After the addition, the mixture was reaction volume of 500 mL was adjusted. The mixture was cooled
heated to 105 °C and stirred at this temperature for 7 h. to 23 °C and treated at 23
30 °C with methanesulfonyl chloride
Approximately 570 kg of PhMe were distilled off at 70
90 °C (13.8 g, 0.120 mol) followed by Et3N (14.0 g, 0.140 mol). The
under reduced pressure and the resulting suspension (800
850 resulting thick suspension was allowed to stir at 25
30 °C for 75
L) was subsequently stirred at 85 °C for another 12 h. Upon min, cooled to
10 to 0 °C, and subsequently treated at this
complete conversion (<1.0% starting material) the reaction temperature within 2 h with t-BuOLi (20% in THF, 95.0 g, 0.240
mixture is concentrated to a residual volume of ∼500 L. The mol). After addition completion, the suspension was stirred for 2
residue is diluted with THF (1200 kg) and subsequently con- h at
5 °C. The mixture was treated at <0 °C within 30 min with
centrated to a residual volume of ∼1600 L. The water-free H2O (70 g). The layers were allowed to separate at 0
5 °C for 30
suspension was cooled to 23
25 °C and treated at a maximum min, and the aqueous layer was subsequently discharged. The pH
temperature of 30 °C with methanesulfonyl chloride (40.8 kg, of the organic layer was adjusted to pH 10.2 with 3% aq H2SO4
356.0 mol, 1.50 equiv) followed by Et3N (41.0 kg, 405 mol, 1.70 (∼32 g). H2O (38 g) was added, and the layers were allowed to
equiv) within 45 min. The resulting thick suspension was stirred separate at 0
5 °C for 20 min. The aqueous layer was removed,
at 25
30 °C for 75 min. Upon complete mesylation (<1.5% and the organic layer was washed at room temperature with H2O
hydroxybutyramide 32), the reaction mixture was cooled to
10 (70 g). The two aqueous phases were combined and extracted
to 0 °C and subsequently treated at this temperature within 2 h with PhMe (65 g). From the combined organic layers THF and
with LHMDS (23.8% in THF, 503 kg, 715 mol, 3.00 equiv). After PhMe were completely distilled off and replaced by MeOH (300
complete addition, the resulting dimmish solution was stirred for g in total). At the end of the solvent exchange a volume of
additional 2 h at
5 °C. Upon reaction completion (<0.2% ∼200 mL was adjusted in the reactor. The resulting suspension
mesylate 33) the reaction mixture was treated at
5 to 5 °C was heated to reflux temperature and stirred at this temperature
within 1 h with H2O (220 kg). The layers were allowed to for 30
60 min to obtain a clear solution. The solution was
separate at 0
5 °C for 30 min and the aqueous layer was allowed to cool to
10 °C within 5
7 h and the resulting
subsequently discharged. The organic layer was washed at suspension stirred for 2 h at
10 °C. The crystals were filtered
0
5 °C twice with 5% aq H2SO4 (2  240 kg). Both combined off, washed with a cooled mixture (
5 °C) of MeOH (30 g) and
aqueous layers were back extracted with PhMe (195 kg). The H2O (10 g) and dried at 45 °C/< 30 mbar to afford 30.0 g (79%
combined organic layers were concentrated to a residual volume yield) of the title product as colorless crystals with an assay
of 700
800 L and washed once with H2O (95 kg). The organic (HPLC) of 99.2% (w/w).
512 dx.doi.org/10.1021/op2000207 |Org. Process Res. Dev. 2011, 15, 503–514
Organic Process Research & Development ARTICLE

4.9. Preparation of (2S,3S,11βS)-1-(2-Amino-9,10-dimeth- Mattei, P.; Narquizian, R.; Wessel, H. P. (F. Hoffmann-La Roche AG).
oxy-1,3,4,6,7,11β-hexahydro-2H-pyrido[2,1-a]isoquinolin- U.S. Pat. Appl. 2004/0259902, 2004.
3-yl)-(4S)-fluoromethyl-pyrrolidin-2-one Dihydrochloride (5) For an overview on DPP-IV inhibitor syntheses, see: Mulakayala,
(1). A suspension of carbamate 13 (136 kg, 285 mol) in a mixture N.; Reddy, C. H. U.; Iqbal, J.; Pal, M. Tetrahedron 2010, 66, 4919.
of H2O (112 kg) and acetone (122 kg) was treated at 50 °C (6) The coupling constant between H2 and H3 is J = 11.5 Hz, which
within 60 min with 37% aq HCl (98.0 kg). After 90 min at is typical for a diaxial orientation of these protons.
(7) Brossi, A.; Lindlar, H.; Walter, M.; Schnider, O. Helv. Chim. Acta
47
52 °C the solution was polish filtered through a 5 μm filter.
1958, 41, 119.
The first reactor and the transfer lines were washed with a hot (8) Abrecht, S.; Adam, J.-M.; Fettes, A.; Hildbrand, S. (F. Hoffmann-La
(47
52 °C) mixture of H2O (13.0 kg) and acetone (116 kg). Roche AG). PCT Int. Appl. WO/2008/031749 A1, 2008. Bromberger, U.;
The filtrate was cooled to 25 °C and treated at this temperature Diodone, R.; Hildbrand, S.; Meier, R. (F. Hoffmann-La Roche AG). PCT
within 80 min with acetone (1600 kg) whereupon the product Int. Appl. WO/2009/027276 A1, 2009.
crystallized out. The resulting suspension was stirred for 1 h at (9) Details on the synthesis of the requisite (S)-fluoromethyl lactone
25 °C and subsequently centrifuged. The crystals were washed in 31 are described elsewhere: Adam, J.-M.; Foricher, J.; Hanlon, S.; Lohri,
two portions with acetone (391 kg) and dried at 50 °C and <30 B.; Moine, G.; Schmid, R.; Stahr, H.; Weber, M.; Wirz, B.; Zutter, U. Org.
mbar until constant weight to afford 122.4 kg (95%) of the title Process Res. Dev. 2011, 15, DOI: 10.1021/op200019k.
compound as colorless crystals with an assay (HPLC) of 98.8% (10) Xu, L. W.; Li, J.-W.; Xia, C.-G.; Hu, X.-X. Synlett 2003, 2425.
(11) (a) Chapman, J. H.; Holton, P. G.; Ritchie, A. C.; Walker, T.;
(w/w).
Webb, G. B.; Whiting, K. D. E. J. Chem. Soc. 1962, 2471. (b) Schneider,
1
H NMR (400 MHz, D2O) δ 2.11
2.22 (m, 1H); 2.45 (dd, W.; K€ammerer, E.; Schilken, K. Pharmazie 1966, 26.(c) Walker T.;
J = 17.6 Hz, 6.7 Hz; 1H); 2.76 (dd, J = 17.6 Hz, 9.55 Hz, 1H); England W.; Frederick K.; Rhondda M.; Wales G.; Ritchie A. C. U.S.
2.90
3.05 (m, 1H); 3.08
3.19 (m, 2H); 3.24
3.36 (m, 1H); Patent 3,105,835, 1963.
3.43 (dd, J = 9.8 Hz, 5.75 Hz, 1H); 3.49
3.58 (m, 1H); (12) Decomposition was observed at room temperature as also
3.70
3.84 (m, 4H); 3.87 (s, 3H); 3.88 (s, 3H); 4.12 (td, J = previously observed in the literature; see reference 9.
11.6 Hz, 4.5 Hz, 1H); 4.45
4.55 (m, 1H); 4.56
4.68 (m, 3H); (13) (a) The cyclic anhydride is mainly in the enol form in DMSO-
6.91 (s, 1H), 6.95 (s, 1H). IR (cm
1): 3237, 2925, 1682, 496, d6. For keto-enol form analytical data see also:Kiang, A. K.; Tan, S. F.;
478. MS (ESI): m/z 378.3 ([M þ H]þ (free amine)). Anal. Wong, W. S. J. Chem. Soc (C) 1971, 2721. (b) For earlier
Calcd for C20H30Cl2FN3O3: C, 53.34; H, 6.71; N, 9.33; Cl, preparations see: Findlay, S. P. J. Org. Chem. 1957, 22, 1385–1394.
15.74; F 4.22; O, 10.66. Found: C, 53.04; H, 6.43; N, 9.45; Cl, Ray, J. A.; Harris, T. M. Tetrahedron Lett. 1982, 23, 1971. Kozikowski,
A. P.; Schmiesing, R. Tetrahedron Lett. 1978, 19, 4241.Ruminski P. G.;
15.66; F, 4.29; O, 11.09. Dhingra O. P. U.S. Patent 4,747,871. Sharma, V. K.; Shahriari-Zavareth,
H.; Garratt, P. J.; Sondheimer, F. J. Org. Chem. 1983, 48, 2379. Meltzer,
’ AUTHOR INFORMATION P. C.; Wang, B.; Chen, Z.; Blundell, P.; Jayaraman, M.; Gonzales, M. D.;
George, C.; Madras, B. K. J. Med. Chem. 2001, 44, 2619 and references
Corresponding Author cited therein.
[email protected]. (14) Precharging of the diacid in acetic acid led to clump formation.
(15) For reviews on crystallization-induced dynamic transforma-
Notes tions, see: (a) Brands, K. M. J.; Davies, A. J. Chem. Rev. 2006, 106, 2711.
§
See the accompanying paper in this issue by Adam et al. (b) Kolarovic, A.; Berkes, D. A.; Baran, P.; Povazanec, F. Tetrahedron
(DOI: 10.1021/op200019k). Lett. 2005, 46, 975. (c) Berkes, D.; Jakubec, P.; Winklerova, D.;
Povazanec, F.; Daich, A. Org. Biomol. Chem. 2007, 5, 121.
(16) (a) Openshaw, H. T.; Whittaker, N. J. Chem. Soc. 1963, 1461.
’ ACKNOWLEDGMENT (b) Clark, R. D.; Kern, J. R.; Kurz, L. J.; Nelson, J. T. Heterocycles 1990,
31, 353.
We thank Dr. T. Glarner, Dr. C. Lautz and their teams for (17) Dibenzoyltartaric acid was preferred over dibenzoyltartaric acid
process calorimetry and safety evaluations; Dr. N. Burki and J.-C. monodimethylamide for cost reasons.
Jordan as well as their teams for analytical support; M. S. (18) Temperatures higher than 60 °C for extended periods of time
Baumgartner, R. Benz, M. Christ, T. Gasser, M. Hohler, M. led to decomposition of the resolving agent.
Lander, P. M€uhlethaler, T. Naber, A. Saladin, G. Schaffner, K. (19) The kinetic studies were performed on a Multimax ART
Sch€onb€achler, C. Weber, and D. Zurwerra for their skillful workstation (Mettler-Toledo) equipped with a 16  5 mL reactor
technical assistance; all central analytical laboratories for their block. Dosing of the solvents as well as sampling (quench and dilution)
support; and Drs. M. Karpf and R. Schmid for helpful discussions. was fully automated. HPLC analyses were performed off-line (the
quench and dilution protocol provided stable samples).
(20) Negative values of the enantiomeric excess indicate that crystal-
’ REFERENCES lization of the (R)-enamine (
)-DBTA salt has started, in which case the
(1) http://www.cdc.gov/chronicdisease/resources/publications/AAG/ (R)-enantiomer becomes predominant, and analytical sampling be-
ddt.htm. comes difficult because of the heterogeneous reaction mixture.
(2) http://www.diabetesatlas.org/content/diabetes-and-impaired- (21) Dichloromethane is the only solvent which can be used to
glucose-tolerance. extract the relatively polar unprotected ester.
(3) Gupta, R.; Walunj, S. S.; Tokala, R. K.; Parsa, K. V. L.; Singh, (22) Shioiri, T.; Ninomiya, K.; Yamada, S. J. Am. Chem. Soc. 1972,
S. K.; Pal, M. Current Drug Targets 2009, 10, 71. 102, 6203. Ninomiya, K.; Shioiri, T.; Yamada, S. Tetrahedron 1974, 30, 2151.
(4) (a) B€ohringer, M.; Fischer, H.; Hennig, M.; Hunziker, D.; (23) Liang, H. Synlett 2008, 2554.
Huwyler, J.; Kuhn, B.; L€offler, B. M.; L€ubbers, T.; Mattei, P.; Narquizian, (24) Hofmann rearrangements can also be problematic upon scale-
R.; Sebokova, E.; Sprecher, U.; Wessel, H. P. Bioorg. Med. Chem. Lett. up because the reactions tend to be exothermic.
2010, 20, 1106. (b) Mattei, P.; B€ohringer, M.; Di Giorgio, P.; Fischer, H.; (25) Allred, E. L.; Hurwitz, M. D. J. Org. Chem. 1965, 30, 2376.
Hennig, M.; Huwyler, J.; Kocer, B.; Kuhn, B.; L€offler, B. M.; MacDonald, (26) Approximately 2% of the methyl ester (which is formed
A.; Narquizian, R.; Rauber, E.; Sebokova, E.; Sprecher, U. Bioorg. Med. immediately when sodium methoxide is added to the reaction mixture)
Chem. Lett. 2010, 20, 1109.(c) B€ohringer, M.; Kuhn, B.; L€ubbers, T.; usually remained after the reaction.

513 dx.doi.org/10.1021/op2000207 |Org. Process Res. Dev. 2011, 15, 503–514

Organic Process Research & Development ARTICLE

(27) (a) Hofmann, A. W. v. Chem. Ber. 1881, 14, 2725. (b) Wallis,
E. S.; Lane, J. F. Org. React. 1949, 3, 267.
(28) Zhang, L. H.; Kauffman, G. S.; Pesti, J. A.; Yin, J. G. J. Org. Chem.
1997, 62, 6918.
(29) (a) Radhakrishna, A. S.; Parham, M. E.; Riggs, R. M.; Loudon,
G. M. J. Org. Chem. 1979, 44, 1746. (b) Loudon, G. M.; Radhakrishna,
A. S.; Almond, M. R.; Blodgett, J. K.; Boutin, R. H. J. Org. Chem. 1984,
49, 4272.
(30) This byproduct is presumably formed by intramolecular trap-
ping of the intermediate isocyanate, followed by intermolecular addition
to a second isocyanate.
(31) During this distillation 29 is quantitatively hydrolyzed to yield
30.
(32) The urea byproduct 30 is carried through to the final step where
it is deprotected. It could not be depleted by crystallization in any step
and hence must be limited to max. 0.25% in isolated amine 10.
(33) 31 was typically produced with 99.0
99.5% ee. See reference 8
for details.
(34) Openshaw, H. C.; Whittaker, N. J. Chem. Soc. C 1969, 89.
(35) The cyclization was also performed under Mitsunobu condi-
tions using diethyl azodicarboxylate (DEAD), di-tert-butyl azodicarbox-
ylate (DTAD) or diisopropyl azodicarboxylate (DIAD) in combination
with triphenylphosphine or tributylphosphine. However, the crude yield
usually was below 70%, the separation from the phosphine oxides was
difficult, and the cost of the reagents was an issue. This variant was
therefore abandoned in favor of above cheap and effective mesyla-
tion
cyclization method. For related cyclizations under Mitsunobu
conditions, see: Bell, I. M.; Beshore, D. C.; Gallicchio, S. N.; Williams,
T. M. Tetrahedron Lett. 2000, 41, 1141.
(36) During workup, silicon-containing byproducts (e.g., hexam-
ethyldisiloxane or trimethylsilanol) are formed, which are present in all
organic and aqueous waste streams and which cause plugging of the
pipes upon incineration if no appropriate incineration facilities are
in place.
(37) Lithium tert-butoxide is commercially available as a 20% (w/w)
solution in THF from e.g. Chemetall.
(38) If hydroxybutyramide 32 is heated over the weekend in toluene
(10 mL/g hydroxybutyramide) at 105 °C in the presence of 2-hydro-
xypyridine, 20% by area (HPLC) of amine 10 are formed, showing that it
is indeed a reversible reaction.
(39) This catalyst was found to be considerably more active than
2-hydroxypyridine.
(40) The reaction was originally conducted at higher concentration
(10 L/kg amine) and in a full batch mode. However, under such
conditions, the product crashed out already during the heating period,
leading to an extremely thick suspension and crust formation on the
reactor walls. Also, a constant reaction temperature of 85 °C was probed,
leading to an overall considerably lower and incomplete conversion.
(41) The deprotection of 13 with HCl in acetone revealed only
traces of the nongenotoxic aldol condensation products mesityloxide
and isomesityloxide. For discussion, see: Coffey, D. S.; Hawk, M. K. N.;
Pedersen, S. W.; Ghera, S. J.; Marler, P. G.; Dodson, P. N.; Lytle, M. L.
Org. Process Res. Dev. 2004, 8, 945.
(42) Although all distillates are recyclable, recycling of the solvent
was not taken into account.
(43) Typically, yields of 80% were obtained instead of the 85% in the
experiment described herein. This high yield is due to holdup from the
previous batches.

514 dx.doi.org/10.1021/op2000207 |Org. Process Res. Dev. 2011, 15, 503–514