Development of a scalable synthesis of (S)-3-fluoromethyl-γ- butyrolactone, building block for Carmegliptin's Lactam moiety

Article abstract of DOI:10.1021/op200019k

Several new routes are reported for the synthesis of (S)-3-fluoromethyl- γ-butyrolactone. An asymmetric hydrogenation-based synthesis was chosen as the enabling route to produce the lactone on a 10-kg scale. A superior stereoselective route starting from (S)-tert-butyl glycidyl ether which afforded the desired lactone in three steps with ～50% overall yield was finally selected for further development and production.

Full text of DOI:10.1021/op200019k

ARTICLE
pubs.acs.org/OPRD
Development of a Scalable Synthesis of (S)-3-Fluoromethyl-γ-
butyrolactone, Building Block for Carmegliptin’s Lactam Moiety
Jean-Michel Adam,† Joseph Foricher,† Steven Hanlon,† Bruno Lohri,*,† Gꢀerard Moine,† Rudolf Schmid,†
Helmut Stahr,‡ Martin Weber,‡ Beat Wirz,† and Ulrich Zutter*,†
F. Hoffmann-La Roche Ltd., Pharmaceuticals Division, †pRED, Pharma Research and Early Development, Process Research and
Synthesis, and ‡Pharma Technical Development Actives, Grenzacherstrasse 124, CH-4070 Basel, Switzerland
ABSTRACT: Several new routes are reported for the synthesis of (S)-3-fluoromethyl-γ-butyrolactone. An asymmetric hydro-
genation-based synthesis was chosen as the enabling route to produce the lactone on a 10-kg scale. A superior stereoselective route
starting from (S)-tert-butyl glycidyl ether which afforded the desired lactone in three steps with ∼50% overall yield was finally
selected for further development and production.
Scheme 1. Assembly of carmegliptin
1. INTRODUCTION
Carmegliptin (1) is a potent and long-acting DPP-IV inhibitor
which in clinical studies has proven its suitability as a safe oral
antidiabetic agent with once-daily administration.1 Carmegliptin
belongs to a new, structurally distinct class of DPP-IV inhibitors
featuring a tricyclic 2-amino-hexahydro-benzo[a]quinolizine
core bearing heteroaromatic, aromatic, or nonaromatic substit-
uents in the 3-position.2 Specifically, the substituent consists of a
five-membered fluoromethyl-substituted lactam in the case of 1.
The preceding paper in this issue describes process research and
development work for the synthesis of the Boc-protected 2,3-
diamino-hexahydro-benzo[a]quinolizine core 2 and its conver-
sion into the active pharmaceutical ingredient (API) by coupling
with (S)-3-fluoromethyl-γ-butyrolactone ((S)-3, Scheme 1).3 In
the present contribution the process chemistry for the synthesis
of fluoromethyl lactone (S)-3 is reported.4
Scheme 2. Synthesis of fluoromethyl lactone rac-3
In the initial discovery chemistry work, fluoromethyl lactone
rac-3 was prepared from 1,3-diacetoxyacetone (4) by a sequence
of HornerꢀWadsworthꢀEmmons olefination with triethyl
phosphonoacetate, acid-catalyzed transesterificationꢀlactoniza-
tion to provide the hydroxymethyl butenolide 6,5 double bond
hydrogenation, and final deoxyfluorination with diethylamino-
sulfur trifluoride (DAST, Scheme 2). Fluoromethyl lactone rac-3
was used in the coupling with enantiomerically pure tricyclic core
2, an approach requiring diastereomer separation on the way to
the API.3
In subsequent discovery chemistry work, fluoromethyl lactone
(S)-3 was prepared from (S)-paraconic acid ((S)-11) which was
obtained by a literature-known route involving Evans’ alkylation
chemistry as key step (Scheme 3).6 Borane reduction of (S)-11
followed by deoxyfluorination of (R)-7 with DAST produced
(S)-3, ready for use in the coupling with 2.3
Limitations of this synthesis of (S)-3 towards scale-up were its
high number of steps, the low overall yield, and the known pro-
pensity of the intermediate (R)-7 to racemize.7 The limited thermal
stability and the high costs of DAST used in the deoxyfluorina-
tion reaction were also of some concern. A new route to (S)-3
should preferably be shorter, avoid (R)-7 as intermediate, and
use an inexpensive fluorination reagent. Various approaches
towards these goals were evaluated in the course of the project,
among them approaches involving asymmetric catalysis, enzy-
matic desymmetrization, and syntheses from readily available
enantiopure precursors.
2. ASYMMETRIC HYDROGENATION-BASED SYNTHESIS
Asymmetric reduction of fluoromethyl butenolide 13 (Scheme 4)
could provide an attractive and straightforward access to (S)-3
using asymmetric hydrogenation,8 hydrosilylation,9 conjugate
hydride addition,10 or yeast reduction11 that have been reported
for the reduction of butenolides, specifically also for 3-alkyl-,
3-acetoxymethyl-, and 3-benzyloxymethyl-butenolides.12 Asym-
metric reductions of 3-halomethyl-butenolides, though, have not
yet been described to the best of our knowledge. Process research
work commenced with the search for an efficient synthesis of
Received: January 25, 2011
Published: April 12, 2011
r
2011 American Chemical Society
515
dx.doi.org/10.1021/op200019k Org. Process Res. Dev. 2011, 15, 515–526
|

Organic Process Research & Development
ARTICLE
Scheme 3. Synthesis of fluoromethyl lactone (S)-3
Scheme 5. tert-Butyl glycidyl ether route to 13
Scheme 4. Wittig route to fluoromethyl butenolide 13
Deoxyfluorination of 6 was performed with bis(2-methox-
yethyl)aminosulfur trifluoride (Deoxofluor, 1.1 equiv), a reagent
reported to be thermally more stable than DAST.17 Fluoro-
methyl butenolide 13 was obtained in 60ꢀ81% yield after dis-
tillation. Crystallization of the low-melting 13 (mp 28ꢀ29 °C)
from MTBE or MTBE/heptane was used to ensure a quality
suitable for asymmetric hydrogenation. Some efforts were under-
taken to replace the expensive Deoxofluor reagent. Reaction of 6
with inexpensive perfluoro-1-butanesulfonyl fluoride (PBSF,
2 equiv) in the presence of diisopropylethylamine (DIPEA,
6 equiv) and DIPEA 3HF (2 equiv)18 afforded 13 in 76% yield
3
on small scale after silica gel chromatography and distillation.
However, on larger scale yields dropped despite extensive
parameter variation (stoichiometry, order of reagent addition,
solvent), and chromatographic purification remained unavoid-
able, leading to discontinuation of this deoxyfluorination protocol.
2.2. tert-Butyl Glycidyl Ether Route to Fluoromethyl Bute-
nolide 13. The approaches reported above relied on DAST or
Deoxofluor as nucleophilic fluorine source and an activated
alcohol as electrophilic reaction partner. In our search for more
economical fluoride sources we shifted our focus to epoxides as
electrophiles. Their propensity to undergo ring-opening with
various HF-delivering reagents is well documented.19 tert-Butyl
glycidyl ether (rac-14, Scheme 5) was selected as the starting
epoxide because of its ready availability as a cheap bulk chemical
and the expected stability of the tert-butyl ether under ring-
opening conditions. In the event, the previously unreported fluo-
rinative ring-opening of rac-14 was accomplished with potassium
hydrogen difluoride (KHF2) in triethylene glycol (TEG) analo-
gous to a protocol devised for other epoxides.20 Thus, heating
rac-14 in a suspension of KHF2 in TEG (130 °C, mol ratio rac-
14/KHF2/TEG = 1:2:1.5) provided the regioisomeric fluorohy-
drins rac-15 and rac-16 (∼70%) in a 95:5 ratio. Major bypro-
ducts were diol 20 (∼25%), originating from competing epoxide
opening with TEG, and ether 21 (∼3%), the secondary product
from epoxide opening with rac-15. Distillation afforded rac-15/
rac-16 mixtures with slightly improved ratios of up to 98:2 in
yields of 61ꢀ65% based on rac-14.21 Replacing the TEG solvent
by diethylene glycol (DEG) or ethylene glycol (EG) led to lower
yields due to increased formation of solvolysis products, while
tetraethylene glycol (TEEG) required a higher reaction temperature
fluoromethyl butenolide 13. Several routes towards 13 were
investigated (Schemes 4 and 5).
2.1. Wittig Route to Fluoromethyl Butenolide 13. In
analogy to the first steps of the initial discovery chemistry route
to rac-3, 1,3-diacetoxyacetone (4) was subjected to Wittig re-
action with Ph3PdCHCO2Et (1.2 equiv)13 in methyl tert-butyl
ether (MTBE) to afford the R,β-unsaturated ester 5 in 98% yield
after filtration over silica gel (Scheme 4). The Wittig reaction
replaced the less selective, chromatography-requiring, and lower-
yielding HornerꢀWadsworthꢀEmmons reaction previously
used. The requisite diacetate 4 was produced by acetylation
(3 equiv Ac2O, pyridine) of inexpensive 1,3-dihydroxyacetone
(12) in nonoptimized yields of 78ꢀ85% after crystallization.14
Treatment of 5 with dry HCl in methanol resulted in clean
deacetylationꢀlactonization to afford the hydroxymethyl bute-
nolide 6 in 96ꢀ98% yield after crystallization from heptane/
dichloromethane (DCM).15 Proof of concept was also demon-
strated for a one-step Wittig access to 6. Reaction of 12 with
Ph3PdCHCO2Et (1.2 equiv) in DCM followed by spontaneous
cyclization provided 6 in an isolated yield of 70ꢀ78%.14 Workup
consisted in separation of the water-soluble 6 from the DCM-
soluble triphenylphosphine oxide (TPPO) and crystallization.
Product purity, however, was unsatisfactory and further work
would have been required to optimize this one-step access to 6.16
516
dx.doi.org/10.1021/op200019k |Org. Process Res. Dev. 2011, 15, 515–526

Organic Process Research & Development
ARTICLE
Table 1. Fluorinative ring-opening of epoxide rac-14
#
reagent (equiv) solv. cond. °C/h rac-15/rac-16a yield dist. (%)
1
2
3
4
5
6
7
8
KHF2 (2.5)
KHF2 (2)b
KHF2 (2)
KHF2 (2)
KHF2 (2)
TEG
TEG
DEG
EG
130/10
130/6
130/5
130/4
150/7
110/2
110/2
110/4
95:5
64
95:5
63
95:5
54
92:8
30
TEEG
neat
95:5
58
Et3N 3HF (2)
79:21
82:18
86:14
53
3
DIPEA 3HF (2) neatc
ndd
ndd
3
Et3N 2HF (2.7) neat
3
a By GC analysis; conversion >90% unless otherwise noted. b KHF2 was
finely ground. c Conversion 76%. d nd = not distilled.
of 150 °C (Table 1, entries 1ꢀ5).22 HF-delivering agents other
than KHF2 such as Et3N 3HF, DIPEA 3HF, or Et3N 2HF23
3
3
3
proved inferior in terms of regio- and/or chemoselectivity
(Table 1, entries 6ꢀ8). TEMPO-catalyzed bleach oxidation of
the mixture rac-15/rac-16 provided ketone 17 in 84ꢀ95%
crude yield, whereby the minor fluorohydrin rac-16 was
oxidized to the corresponding water-soluble acid which was
washed out in the workup.
Addition of the Li-enolate of tert-butyl acetate to 17 (THF,
ꢀ70 °C) produced hydroxy ester 18 which upon heating in 1,2-
dimethoxyethane (DME) in the presence of catalytic amounts of
conc. H2SO4 underwent ester and ether cleavage and subsequent
lactonization providing hydroxy lactone 19 in virtually quantita-
tive yield.24 Dehydration of 19 was accomplished preferably with
acetic anhydride/triethylamine in the presence of a catalytic
amount of DMAP to give 13 in 72ꢀ75% yield after distillation
and crystallization from MTBE.25
2.3. Asymmetric Hydrogenation of Fluoromethyl Buteno-
lide 13. Microbial reduction, hydrosilylation, and hydrogenation
were evaluated for the asymmetric reduction of 13. Microbial
reduction was abandoned early on because only 1 out of 1200
strains provided the desired (S)-configuration, and moreover,
enantioselectivity (89% ee) and rate were unsatisfactory.26 In
Cu-catalyzed hydrosilylations of 13 using Buchwald conditions9
Figure 1. Bisphosphine ligands evaluated in the asymmetric hydro-
genation of 13.
Scheme 6. Asymmetric hydrogenation of fluoromethyl
butenolide 13
(1 mol % bisphosphine ligand, 2.5 mol % CuCl2 H2O, 0.2 equiv
3
NaOt-Bu, 4 equiv polymethylhydrosiloxane) the atropisomeric
bisphosphine (R)-BIPHEMP (22a, Figure 1) was identified to be
the best ligand leading to (S)-3 with up to 91% ee. However,
preparative runs pointed to difficulties in the workup and
purification at larger-scale, prompting us to search for a viable
asymmetric hydrogenation protocol.
In the asymmetric hydrogenation of 13, Rh, Ir, and Ru
catalysts derived from a variety of bisphosphine ligands 22aꢀt
(Figure 1) were evaluated, and two effective catalysts were identi-
fied, one Rh- the other Ru-based (Scheme 6).27 In the Rh domain
(Table 2, left column) the highest ee of 93% was achieved with
the catalyst derived from the bithiopheneꢀbisphosphine ligand
(S)-TMBTP (22l).28 This catalyst was moderately active and
substrate/catalyst ratios (S/C) of up to 1000 were reached
(50 bar H2, DCM, 40 °C).29 Among the Ru catalysts30 tested
(Table 2, right column), the complex [Ru(OAc)2((R)-3,5-t-Bu-
MeOBIPHEP)] (23) derived from the sterically demanding
bisphosphine 22f was highly active (S/C up to 15,000). It proved
most suitable, affording (S)-3 in 92% ee under screening condi-
tions and in 96ꢀ97% ee under optimized conditions (5ꢀ10 bar
H2, MeOH, 30 °C, substrate conc. 24% w/w).31 Solvent effects
observed in the hydrogenation of 13 with catalyst 23 are shown
in Table 3. Although the highest enantioselectivity was obtained
in 2,2,2-trifluoroethanol (TFE, >99% ee, entry 1), methanol was
selected as solvent for cost reasons.
517
dx.doi.org/10.1021/op200019k |Org. Process Res. Dev. 2011, 15, 515–526

Organic Process Research & Development
ARTICLE
Table 2. Influence of bisphosphine ligand in the Rh- and
Ru-catalyzed hydrogenation of 13 to (S)-3
Scheme 7. Enzymatic desymmetrization route
Rh-cat.a
Ru-cat.b
#
P^P
% ee
P^P
22a
% ee
1
ent-22a
ent-22b
ent-22c
52
55
45
76
2
22b
68
3
22c
81
4
22d
92
91c
5
ent-22e
22f
6
ent-22f
ent-22g
43
43
92
7
ent-22g
22h
87c
95
8
9
22i
65
82c
10
11
12
13
14
15
16
17
18
19
20
ent-22j
22k
with Lipase PS Amano was established under technically relevant
conditions (20% w/v substrate concentration in vinyl acetate/
MTBE 1:1) affording monoacetate (R)-25 with 95ꢀ97% ee and
in yields of 90ꢀ95%. The major byproduct (∼3%) was the
corresponding diacetate. Tosylation of (R)-25 (catalyzed by
ent-22k
22l
59
77
37c,e
92d
55c
47
22l
22m
ent-22n
22n
22o
69
58
0.1 equiv Me3N HCl)33 followed by HCl/MeOH-mediated
3
ester cleavageꢀlactonization furnished the tosyloxymethyl lac-
ent-22p
ent-22q
56
54
tone (S)-2734 which was readily transformed into (S)-3 by
reaction with Et3N 2HF (6 equiv, in situ prepared from 4 equiv
3
Et3N 3HF and 2 equiv Et3N).23 Configurational integrity was
22r
22s
82
63
3
maintained over the three steps and the overall yield of (S)-3
amounted to 71% after final chromatography and distillation.
Both the acyclic and the cyclic tosylates (S)-26 and (S)-27
proved to be crystalline compounds and the former one allowed
for upgrading of enantiomeric purity by crystallization, albeit
with considerable loss of yield. Attempts to perform fluorine
introduction at the stages of monoacetate (R)-25 or the acyclic
tosylate (S)-26 failed.
22t
70
a [RhCl(COD)]2 þ P^P, S/C 100, 50 bar H2, 40 °C, DCM, 10% w/v,
16ꢀ20 h, 0.1-g scale; conversion >99%. b Catalyst [Ru(OAc)2(P^P)]
preformed, or in situ formed ([Ru(OAc)2(COD)], P^P, 40 °C, 2 h), S/C
100, 40ꢀ50 bar H2, 40 °C, MeOH, 7ꢀ10% w/v, 18ꢀ20 h, 0.1ꢀ0.2 g
scale; conversion >99%. c (R)-3 was formed. d 93% ee at S/C 1000 and
1-g scale. e Experiment in DCM, S/C 50.
Overall this enzyme-based route to (S)-3, which is attractive
due to the ease and efficiency of the introduction of the stereo-
genic center, afforded ∼35% yield based on 12 over six steps.
Shortcomings of the route are the unsatisfactory synthesis of diol
24 and the required chromatographic purifications. Consequently,
this route could not compete with other routes described herein.
Table 3. Influence of solvent in the hydrogenation of 13 to
(S)-3 with Ru catalyst 23a
#
solvent
S/C
c (% w/w)
% ee
1
TFE
1000
1000
1000
1000
100
15
24
24
20
11
11
7
99.5
93
93
93
92
90
90
51
51
47
2
EtOH
PrOH
MeOH
MeOH
i-PrOH
DCM
EtOAc
toluene
THF
4. SYNTHESES OF (S)-3 USING STARTING MATERIALS
FROM THE CHIRAL POOL
3
4
5
4.1. The Malonate Route. Inspired by the regioselective
epoxide ring-opening with KHF2 in the asymmetric hydrogena-
tion route, a stereoselective synthesis starting from optically
active tert-butyl glycidyl ether (S)-14 was designed (Scheme 8).
In contrast to rac-14, the enantiomer (S)-14 was expensive and
commercially available only in small quantities (10 g).35 Thus at
the outset of our investigation (S)-14 was secured in sufficient
amounts (100 g) by hydrolytic kinetic resolution (HKR) of rac-
14 using the method of Jacobsen.36 Enantioselective epoxide
ring-opening with 0.55 equiv water in presence of 0.5 mol % (R,
R)-(Salen)CoIIIacetato complex (rt, 20 h) provided the enantio-
pure (S)-14 with 99.8% ee in 44% yield after distillation.37 Bulk
quantities of (S)-14 with ee g98% were later offered by external
suppliers. The formation of the fluorohydrin (R)-15 was accom-
plished as described above for the racemic series (Scheme 5)
by epoxide opening with KHF2 and TEG. Reducing the amount
of TEG (mol ratio (S)-14/KHF2/TEG = 1:2:0.8) and adding
a catalytic amount of tetrabutylammonium hydrogensulfate
6
100
7
100
8
100
10
10
10
9
100
10
100
a 50 bar H2, 40 °C; experiments with S/C 100 at 0.1-g scale, with S/C
1000 at 1-g scale; conversion >99%.
3. ENZYMATIC DESYMMETRIZATION ROUTE
Proof of concept was established for a desymmetrization
approach based on the literature-known enzymatic mono-
acetylation of diol 24 (Scheme 7).32 Diol 24 was synthesized
from 1,3-dihydroxyacetone (12) by Wittig olefination with
Ph3PdCHCO2t-Bu followed by Pd-catalyzed double bond hydro-
genation in the presence of Et3N (0.2 equiv) to avoid hydro-
genolytic deoxygenation. The enzymatic monoacetylation of 24
518
dx.doi.org/10.1021/op200019k |Org. Process Res. Dev. 2011, 15, 515–526

Organic Process Research & Development
ARTICLE
Scheme 8. Malonate route to fluoromethyl lactone (S)-3
Scheme 9. HF and HCl addition to cyclopropanolactone 34
substitution followed by an intramolecular epoxideꢀcyclopro-
pane ring rearrangement.44
The HF-addition to the cyclopropane ring in 34 proved to be
much more difficult than expected. Thus, treatment of 34 with
(TBAHS) led to a slightly improved yield of a 95:5 mixture of the
two regioisomeric fluorohydrins. The major regioisomer (R)-15
was isolated by distillation over a Fenske ring-packed column
with ∼99% regioisomeric purity (GC) in ∼65% yield. Subse-
quent treatment of (R)-15 with triflic anhydride and pyridine in
DCM gave the crude triflate (R)-28 with negligible racemization
(e0.3%). In order to avoid racemization, the triflate formation as
well as the hydrolysis were performed at ꢀ10 °C.38 Since the neat
triflate (R)-28 was only marginally stable at rt, it was either used
immediately after isolation or stored at ꢀ20 °C.39
HF delivering reagents such as Et3N 3HF, 70% HF pyridine, and
3
even 100% HF under various conditions (<100 °C) did not
provide any of the desired fluoromethyl lactone 30. Finally, and
surprisingly, treatment of 34 with neat Et3N 3HF (12 equiv)45
3
and FeF3 (2 equiv) at 140 °C for 7ꢀ9 days directly afforded (S)-
3. Apparently, under these drastic reaction conditions 30, which
was detected only in traces (GC and DC, 1ꢀ2%),46 was
converted into (S)-3 by an unexpected demethoxycarbonylation
reaction. After workup and distillation, (S)-3 was isolated in up to
65% yield (GC 96.8%, 97.4% ee). Despite the shortness and good
overall yield of 35ꢀ40%, this two-step cyclopropanolactone
route was not considered for scale-up due to the harsh reaction
conditions in the second step and the corrosive nature of the
The ensuing substitution of (R)-28 with sodium dimethyl
malonate in DME proceeded with full inversion of configuration
providing the intermediate (S)-29 which was transformed in situ
into lactone (S)-3 by treatment with 2 M H2SO4 (85 °C,
48 h).40,41 While the tert-butyl ether cleavage and the lactoniza-
tion ((S)-29 f 30) was effected at reflux temperature in <0.5 h,
the hydrolysis of the methyl ester and the subsequent decarb-
oxylation (30 f 31 f (S)-3) required up to 2 days. Extraction
with DCM followed by distillation delivered (S)-3 in 80ꢀ84%
yield with g99% chemical and enantiomeric purity (GC).
Starting from commercially available (S)-14 with 98.0% ee,
(S)-3 with 97.8% ee was obtained in three steps with 50ꢀ55%
overall yield. Consequently this short and efficient malonate
route was considered most suitable for further development.
A somewhat less efficient alternative involved the substitution
of the triflate (R)-28 with the Li-enolate of tert-butyl acetate
providing the tert-butyl ester (S)-32. In order to obtain high
selectivity and yield, this SN2 substitution had to be performed at
low temperature (THF, ꢀ75 °C) and in the presence of an
additive like 1,3-dimethyl-2-imidazolidinone (DMI, 3 equiv).42
Lactonization of (S)-32 was effected after deprotection with TFA
and provided (S)-3 in 94% yield (GC 99.3%, 99.4% ee).
Et3N 3HF reagent.47
3
On the other hand, HCl added smoothly to 34 in presence of a
catalytic amount of AlCl3 (2 equiv HCl, 2 mol % AlCl3, DME,
50 °C)48 furnishing the chloromethyl lactone 35. Subsequent
demethoxycarbonylation using TsOH H2O in DMSO (2 equiv,
3
120 °C, 3 h) gave (S)-36 which was converted into the
iodomethyl lactone (S)-37 by a Finkelstein reaction with NaI.
While (S)-37 was successfully converted into (S)-3 with AgF, the
substitution of (S)-36 and (S)-37 with less expensive fluoride
salts such as CsF in polar solvents like N-methylformamide
(MFA), DMSO or MeCN failed.
5. SCALE-UP OF THE ASYMMETRIC HYDROGENATION-
BASED SYNTHESIS
The asymmetric hydrogenation route, being the earliest one
available, was used as enabling route for the production of API on
the 10-kg scale. Initially, the Wittig route (Scheme 4) served to
prepare the hydrogenation substrate 13 and first batches of (S)-3
were obtained by the Rh process which subsequently was
replaced by the Ru process. Later, the glycidyl ether route to
13 (Scheme 5) and the Ru process were used. Such rapid changes
of routes and processes are not atypical for project situations in
which API supply is on the critical path.
4.2. The Cyclopropanolactone Route. In a further approach
we have envisaged to establish the fluoromethyl group by HF-
addition to cyclopropanolactone 34 to provide the fluoromethyl
lactone 30 (Scheme 9) which can be converted into (S)-3 by
hydrolysis and decarboxylation as was already demonstrated in
the malonate route (cf. Scheme 8). The literature-known cyclo-
propanolactone 34 was synthesized from commercially avail-
able glycidyl nosylate (R)-3343 in a single step by malonate
519
dx.doi.org/10.1021/op200019k |Org. Process Res. Dev. 2011, 15, 515–526

Organic Process Research & Development
ARTICLE
The Wittig route was scaled without optimization and pro-
vided 13 in an overall yield of 40% after distillation and crystal-
lization. The Deoxofluor-mediated transformation 6 f 13 was
performed at temperatures not exceeding 25 °C without further
safety measures. The glycidyl ether route to 13 was readily scaled
up to 20 kg. For the fluorinative epoxide opening a mode
involving addition of rac-14 to the KHF2/TEG mixture at
125 °C over 1 h (mol ratio rac-14/KHF2/TEG = 1:2:1.5) was
shown to be unproblematic by a calorimetric investigation
(ΔH = ꢀ46 kJ/mol, ΔadiaTmax = 43 °C, ΔadiaTaccu = 36 °C).
In the event the reaction was performed by addition of rac-14
over 15 min at 140 °C and a reaction time at 140 °C of 5 h to
provide a yield of 64% for the distilled rac-15/rac-16 97:3
mixture. In the subsequent steps yields of 84% for crude ketone
17 and 67% for distilled and crystallized fluoromethyl butenolide
13 were obtained. The final crystallization was necessary to
obtain a high-quality material for the asymmetric reduction. An
overall yield of 36% thus was achieved for the tert-butyl glycidyl
ether route to the hydrogenation substrate 13.
Not unexpectedly, the quality of 13 played an important role
in the development of efficient hydrogenation processes. In the
Rh process S/C ratios of 1000 were reached in small-scale
experiments, but only 400ꢀ500 were realized upon upscaling,
even after 2-fold distillation of 13. Nevertheless, more than 0.5 kg
of (S)-3 with an ee of 93% were prepared in 94% yield after
distillation. The Ru-catalyzed hydrogenation of substrate 13,
prepared by the Wittig route and purified by distillation followed
by crystallization, proceeded smoothly at the kg-scale at S/C ratios
of 12,500 affording (S)-3 with 95% ee in 98ꢀ99% yield after
distillation. More than 5 kg of (S)-3 were prepared with an overall
yield of 39% over the five steps starting from dihydroxyacetone
(12). The quality of 13 prepared by the tert-butyl glycidyl ether
route was even more critical. A purification protocol involving
decolorization with charcoal, filtration over silica gel, distillation,
and final crystallization was necessary to produce material of
adequate quality for the hydrogenation. Preparative 5-kg batch
hydrogenations of 13 purified in this way proceeded efficiently at
S/C ratios of 9000 providing (S)-3 with ee values in the range of
96.3ꢀ96.8% in average 93% yield after distillation corresponding
to an overall yield of 33% over the six steps starting from rac-14.
Occasionally, catalyst inhibition was observed, for which the
reasons remained unclear. Acetic acid, a potential contaminant,
was demonstrated to have no inhibitory effect up to 2 mol %
(∼240 equiv based on Ru catalyst). Triethylammonium acetate
which might have been entrained from step 19 f 13 was found
to stop the hydrogenation reaction completely at the 0.2 mol %
level. However, there was no clear evidence for the presence of
this contaminant in substrate batches showing reduced hydro-
genation rates. Substrate batches of suboptimal quality some-
times showed a tendency to trigger hydrodefluorination leading
to 3-methyl-γ-butyrolactone as byproduct accompanied by
acidic vapors most probably from HF.
starting from commercial epoxide (S)-14 of 99.4% ee, a produc-
tion process was developed for the synthesis sequence via (S)-29
(Scheme 8), and the process was scaled to a batch size of >100 kg
of (S)-14 or 100 kg of isolated (S)-3. The conversion of (S)-14 to
the fluorohydrin (R)-15 was accomplished within 24 h at
135ꢀ140 °C. A Hastelloy vessel had to be employed due to
the corrosive nature of the KHF2 reagent and the harsh reaction
conditions. The extended reaction time was required due to the
low solubility of KHF2. A reaction protocol involving mixing all
components at rt (mol ratio (S)-14/KHF2/TEG = 1:2:0.8) and
subsequent heating to 135 °C was demonstrated to be safe by a
calorimetric investigation (ΔH = ꢀ50 kJ/mol, ΔadiaTmax
=
88 °C). The bulk amount of insoluble inorganic salts was filtered
off, and the crude step product was rectified using a distillation
column giving (R)-15 in 59% yield. Major focus for process
development was to circumvent the isolation of the unstable
triflate (R)-28. This was achieved by developing a telescoped
process in which (R)-15 was reacted with triflic anhydride at ꢀ10
to ꢀ20 °C in the presence of pyridine, filtering off the pyridinium
triflate precipitation formed at ꢀ15 °C, and dosing the solution
of (R)-28 in DCM at ꢀ5ꢀ0 °C directly into a solution of sodium
dimethyl malonate in DME. For the subsequent H2SO4-pro-
moted deprotectionꢀlactonizationꢀhydrolysisꢀdecarboxyla-
tion sequence the reaction time could be reduced from 48 h to
18ꢀ20 h by working at 92ꢀ94 °C and at higher concentration.
Also this step had to be conducted using a Hastelloy vessel as the
reaction mixture was highly corrosive due to traces of fluoride
formed combined with the acidic reaction conditions at ele-
vated temperatures.49 The crude product was purified by
distillation to give (S)-3 in a yield of 73%, an assay of 98ꢀ
100% (w/w), and an ee of 99.3%. Using this process a total of
>500 kg of (S)-3 was produced in an overall yield of 43% based
on (S)-14.
7. CONCLUSION
In summary, route exploration for optically active fluoro-
methyl lactone (S)-3 has resulted in the identification of four
new pathways to synthesize this small molecule: (a) the five- or
six-step Ru-catalyzed asymmetric hydrogenation route—in
which the butenolide substrate 13 was prepared either from
dihydroxyacetone or, more favorably, from tert-butyl glycidyl
ether (rac-14)—leading to (S)-3 with 96ꢀ97% ee in 33ꢀ39%
overall yield; (b) a six-step enzymatic desymmetrization route
starting from 1,3-dihydroxyacetone (95ꢀ97% ee, 35% overall);
(c) the three-step malonate route starting from optically active
tert-butyl glycidyl ether (S)-14 (98ꢀ99% ee, overall 50ꢀ55% lab
scale, and 43% technical scale); and (d) the two-step cyclopro-
panolactone route starting from optically active glycidyl nosylate
(R)-33 (97.5% ee, 35ꢀ40% overall). In the routes selected for
scale-up the stereogenic center was established either by asym-
metric hydrogenation or by using an enantiopure starting
material, i.e. (S)-14. In both scaled routes the very inexpen-
sive fluorine-introducing reagent KHF2 was used to establish
the primary fluoride functionality, thus replacing the initially
used expensive DAST and Deoxofluor reagents, respectively,
and greatly contributing to lower chemical costs. Finally, the
three-step malonate route was selected for technical devel-
opment and for API production at larger scale on the basis
of its shortness, its highest yield, the commercial availability
of enantiopure starting material, and its estimated lowest
process costs.
The enantiomeric purity of the (S)-3 materials produced in the
asymmetric hydrogenations (93% ee Rh-catalyzed, 96ꢀ97% ee Ru-
catalyzed) was sufficient for use in the API synthesis. The minor
diastereomer formed in the coupling with aminobenzoquinolizine 2
was readily removed in the crystallizations towards the API.
6. SCALE-UP OF THE MALONATE ROUTE
The malonate route to (S)-3, once available, was considered
the most suitable for development to technical scale. Accordingly,
520
dx.doi.org/10.1021/op200019k |Org. Process Res. Dev. 2011, 15, 515–526

Organic Process Research & Development
ARTICLE
8. EXPERIMENTAL SECTION
silica gel 60 using heptane/ethyl acetate 3:1 as the eluent.
Seventeen fractions of 5 L were collected and evaporated (45 °C/
g2 mbar) to give 5 (6.20 kg, 98%) as colorless oil, GC 99.9%.5 IR
8.1. General Remarks. Unless otherwise noted, reagents and
solvents were used as received from commercial suppliers. All
reactions were carried out under argon or nitrogen atmosphere.
Column chromatography on silica gel 60 (mesh 70-230, Merck)
was used to provide reference samples for analytical purposes of
products and byproducts. Yields are weight-based and not
corrected for assays unless otherwise noted. The chemical purity
and conversions were determined by gas chromatography (GC,
area %). The enantiomeric excess (ee) was determined by chiral
GC using a cyclodextrin-based BGB-176 column (30 m ꢁ 0.25
mm; BGB-Analytik Ltd.) for (S)-14, (R)-15, and (R)-28, and a
cyclodextrin-based BGB-175 column (30 m ꢁ 0.25 mm, BGB-
Analytik Ltd.) for (S)-3. NMR spectra were recorded using
tetramethylsilane (TMS) as an internal standard. Chemical shifts
(δ) and coupling constants (J)—determined assuming first-
order behavior—are given in ppm and hertz (Hz). 19F NMR
chemical shifts are calculated on the basis of CFCl3 as standard.
Low-resolution electron impact mass spectra (EI-MS) were
obtained at an ionization voltage of 70 eV or by positive or
negative ion spray ionization (ESI-MS). Data are reported in the
form of m/z (intensity relative to base = 100).
1
(film): ν 2982, 1741, 1713, 1370, 1207, 1145, 1028 cmꢀ1. H
NMR (CDCl3, 400 MHz): δ 1.30 (t, J = 7 Hz, 3H, CO2CH2-
CH3), 2.08 and 2.12 (s, 3H each, C(O)CH3), 4.20 (q, J = 7 Hz,
2H, CO2CH2CH3), 4.72 and 5.26 (s, 2H each, CH2OAc), 6.00
(quint, J = 1.6 Hz, 1H, dCHꢀ). EI-MS (m/z) 185.2 (48), 171.1
(12), 157.2 (48), 142.1 (100), 129.1 (19), 113 (15), 97.1 (39).
8.2.3. 4-Hydroxymethyl-5H-furan-2-one (6). In a 60-L reactor
ester 5 (6.10 kg, 25.0 mol) was dissolved in methanol (28 L), and
acetyl chloride (177 mL, 2.5 mol) was added over 10 min under
cooling, maintaining the temperature at 22ꢀ23 °C. After stirring
at 22 °C for 18 h and at 50 °C for 2 h, the reaction mixture was
concentrated (45 °C/g10 mbar), and the residual yellowish oil
(2.80 kg) was azeotroped with toluene (3 ꢁ 19 L, 45 °C/g10
mbar). The crude yellow oil was dissolved in DCM (6 L), and
crystallization was effected by slowly cooling to ꢀ5 °C followed
by addition of heptane (22 L) over 30 min. After additional
stirring at ꢀ5 °C for 30 min, the suspension was filtered. The
filter cake was washed with cold heptane (4 L) and dried (rt/10
mbar/16 h) to give hydroxymethyl butenolide 6 (2.75 kg, 96.5%)
as white powder, GC 98.6%.5 IR (Nujol): ν 3419, 3115, 2954,
1
1743, 1450, 1370, 1254, 1140, 1066, 1012 cmꢀ1. H NMR
8.2. Wittig Route to 13. 8.2.1. Acetic Acid 3-Acetoxy-2-oxo-
propyl Ester (4). A 500-L reactor was charged with 1,3-dihydrox-
yacetone (12, 25 kg, 277.5 mol) and pyridine (81 L). To the
suspension obtained was added acetic anhydride (81 L, 862 mol)
over 40 min (exothermic!), maintaining the jacket temperature
(Tj) between 15 and 22 °C. The resultant reddish solution was
stirred at 20 °C for 3 h and then concentrated at 50ꢀ55 °C/g10
mbar. The oily residue was dissolved in DCM (250 L) and the
solution washed with 25% hydrochloric acid (2 ꢁ 125 L) and
water (125 L). The organic layer was concentrated (40 °C/g10
mbar), and the dark-red, oily residue (120 L) was dissolved in
toluene (144 L) at ∼30 °C. Crystallization was induced by
addition of heptane (125 L) over 15 min followed by seeding
with pure diacetate 4. After stirring for 1 h, additional heptane
(125 L) was added to improve stirrability, and the suspension
was stirred at 20 °C overnight and at 0 °C for 2 h. The suspension
was filtered, and the filter cake was washed with precooled
heptane (150 L) and dried (30ꢀ35 °C/g10 mbar), affording
diacetate 4 (37.5 kg, 77.6%) as white powder, GC 100%. An
analogous smaller-scale reaction starting from 4.0 kg (44.4 mol)
12 gave 6.26 kg (81%) 4, GC 99.7%. 1H NMR (CDCl3,
400 MHz): δ 2.17 (s, 6H, CH3CO2), 4.75 (s, 4H, CH2CO).
EI-MS (m/z) 102.2 (7), 101.1 (100), 73.1 (34), 43.3 (93). Anal.
Calcd for C7H10O5 (174.15): C 48.28; H 5.79; O 45.93. Found:
C 48.28; H 5.78.
8.2.2. 4-Acetoxy-3-acetoxymethyl-but-2-enoic Acid Ethyl
Ester (5). A 60-L reactor was charged with diacetate 4 (2.25 kg,
12.9 mol), MTBE (35 L), and Ph3PdCHCO2Et (5.45 kg, 15.6
mol). The solution was stirred under reflux for 5 h and allowed to
cool to rt during 14 h. MTBE was exchanged with heptane (40 L,
45 °C/450ꢀ170 mbar), and the resulting suspension was stirred
at rt overnight. After addition of toluene (7 L) and additional
stirring at rt for 1 h and at 0ꢀ4 °C for 2 h, the suspension was
filtered and the filter cake (TPPO) washed with precooled
toluene (10 L). The filtrate was concentrated (45 °C/110 mbar)
to give 3.25 kg crude R,β-unsaturated ester 5 as reddish oil. This
material, combined with 3.25 kg 5 from a duplicate reaction, was
dissolved in a 3:1 mixture of heptane/ethyl acetate (8 L) and
chromatographically filtered over a column containing 15 kg
(CDCl3, 400 MHz): δ 2.55 (t, J = 5 Hz, 1H, OH), 4.60 (d, J = 5
Hz, 2H, CH2OH), 4.88 (s, 2H, CO2CH2), 6.04 (t with fine
structure, 1H, dCHꢀ). EI-MS (m/z) 96 (10), 85 (100), 84
(61), 70 (25), 68 (13), 67 (10), 57.1 (16), 56.1 (13), 55 (49),
39.1 (32), 38.1 (11), 31.2 (24), 29.3 (44). Anal. Calcd for
C5H6O3 (114.10): C 52.63; H 5.30; O 42.07. Found: C 52.39;
H 5.21; H2O 0.23.
8.2.4. 4-Fluoromethyl-5H-furan-2-one (13). A 160-L reactor
was charged with hydroxymethyl butenolide 6 (10.35 kg, 90.7
mol) and DCM (42 L), and the solution was cooled to ꢀ10 °C.
Bis-(2-methoxyethyl)aminosulfur trifluoride (23.5 kg, 95%, 101
mol, Deoxofluor, Matrix Scientific) was added over 50 min,
maintaining the temperature strictly between ꢀ5 and ꢀ10 °C.
During the addition, a yellowish emulsion was obtained which
turned into an orange-red solution after completion of the
addition. After stirring at 15ꢀ20 °C for 2 h, a solution of water
(5 L) in ethanol (25 L) was added over 30 min, maintaining the
temperature between ꢀ5 and ꢀ10 °C, and the mixture was
allowed to reach rt. After concentration (40 °C/650ꢀ120 mbar)
to a volume of ∼30 L, the residue was dissolved in DCM (42 L),
and the solution was washed with 1.0 M hydrochloric acid (3 ꢁ
83 L). The aqueous layers were back-extracted with DCM (3 ꢁ
30 L), and the combined organic layers were evaporated to
dryness. The resulting dark-brown liquid (8.32 kg) was distilled
in four portions (4-L pear-shaped flask, 20-cm Vigreux column)
providing fluoromethyl butenolide 13 (6.30 kg) as colorless
liquid, bp 73ꢀ80 °C/∼0.1 mbar, GC 97.7%. IR (film): ν 1778,
1738, 1448, 1133, 1040, 1003 cmꢀ1. The distillate was dissolved
in MTBE (7 L) and crystallized under stirring at 0 °C (2 h)
and ꢀ20 °C (3 h). Filtering over a precooled filter, washing with
precooled MTBE (4 L, ꢀ20 °C), and drying (rt/g1 mbar/24 h)
provided fluoromethyl butenolide 13 (5.69 kg, 54%) as white
crystalline powder, mp 30ꢀ40 °C, GC 99.8%. 1H NMR (CDCl3,
400 MHz): δ 4.89 (s with fine structure, 2H, CH2O), 5.33
(d with fine structure, 2JHF = 46 Hz, 2H, CH2F), 6.11 (s with fine
structure, dCHꢀ); 1.60 (s, ∼0.2 H, ∼10 mol % H2O,). EI-MS
(m/z) 116.2 (7), 87.2 (100), 86.3 (7), 59.2 (25), 58.3 (12), 57.1
(17), 39.3 (11), 29.4 (5). An analytical sample of 13 obtained
521
dx.doi.org/10.1021/op200019k |Org. Process Res. Dev. 2011, 15, 515–526

Organic Process Research & Development
ARTICLE
by three-fold crystallization from MTBE showed a mp of 28ꢀ
with diisopropylamine (34.4 kg, 47.7 L, 340 mol) and dry THF
(162 kg). The mixture was cooled to ꢀ60 °C, and butyl lithium
(1.6 M in hexane, 138 kg, 203 L, 325 mol) was slowly added.
After stirring at ꢀ60 °C for 45 min, the light-yellow solution was
cooled to ꢀ70 °C, and tert-butyl acetate (39 kg, 45 L, 335.5 mol)
was added within 5 min at below ꢀ65 °C. Stirring was continued
at ꢀ70 °C for 45 min, and the above THF solution of ketone 17
(theoretically 302 mol corresponding to 44.8 kg) was added
within 5 min. The reaction mixture was allowed to warm to 0 °C
and was treated with saturated ammonium chloride solution
(480 L), and the phases were separated. The organic layer was
washed successively with 10% ammonium chloride solution (2 ꢁ
180 L) and 10% sodium chloride solution (90 L). The slightly
yellow organic layer was dried by passage through a column filled
with sodium sulfate (95 kg) and evaporated to dryness (40 °C/
g50 mbar). The oily residue was dissolved in ethyl acetate
(100 L) and again evaporated to dryness, providing crude
hydroxy ester 18 as slightly yellow oil which was left in the
reactor for the next step, GC 98.4%. IR (film): ν 3476 (br), 2975,
1
29 °C, colorless needles. According to H NMR this material
contained about 5 mol % H2O.
8.3. tert-Butyl Glycidyl Ether Route to 13. 8.3.1. 1-tert-
Butoxy-3-fluoro-propan-2-ol (rac-15). A 630-L Hastelloy reactor
was charged with triethylene glycol (115 L, 129 kg, 859 mol) and
potassium hydrogen difluoride (89 kg, 1140 mol), and the
resulting suspension was heated to 140 °C. tert-Butyl glycidyl
ether (rac-14, 75 kg, GC 98%, 565 mol) was added over 45 min,
and the suspension was stirred at 140 °C for 5 h, cooled to rt, and
treated with water (330 L). After stirring for 30 min, the pH was
adjusted to 7ꢀ7.5 by addition of 28% sodium hydroxide (58 L),
and the reaction mixture was extracted with MTBE (3 ꢁ 400 L).
The organic phases were washed with brine (80 kg sodium
chloride in 185 L water), combined, and concentrated (45 °C/
g10 mbar) to a volume of ∼100 L. After removal of insoluble
particles by filtration through a Teflon membrane filter, the
residual oil was purified by short path distillation to yield 54.1 kg
(64%) of a colorless liquid, bp 68 °C/20 mbar, GC 95.2% rac-15
and 3.7% rac-16. IR (film): ν 3422 (br), 2975, 1474, 1390, 1365,
1195, 1085, 1015 cmꢀ1. 1H NMR (CDCl3, 400 MHz): δ 1.20 (s,
9H, C(CH3)3), 2.43 (d, J = 5.3 Hz, 1H, OH), 3.45 (m, 2H,
1
1727, 1706, 1366, 1155, 1084, 1020 cmꢀ1. H NMR (CDCl3,
400 MHz): δ 1.18 and 1.47 (s, 9H each, C(CH3)3), 2.54 (m, 2H,
CH2CO2), 3.37 (d, J = 2.3 Hz, 2H, CH2O), 3.87 (s, 1H, OH),
4.38 (dm, 2JHF = 47 Hz, 2H, CH2F). EI-MS (m/z) 177.2 (24),
153.3 (10), 135.2 (65), 134.2 (30), 121.2 (39), 119.1 (11), 102.2
(34), 59.4 (10), 57.3 (100).
CH2O), 3.94 (dm, 3JHF = 17.7 Hz, 1H, CHOH), 4.45 (dm, 2JHF
=
47.3 Hz, 2H, CH2F). EI-MS (m/z) 135.2 (11), 100.3 (10), 59.3
(63), 57.3 (100), 41.4 (19), 29.5 (14).
8.3.2. 3-tert-Butoxy-2-fluoro-propan-1-ol (rac-16). A refer-
ence sample of fluorohydrin rac-16 was obtained by chromatog-
raphy as colorless oil. IR (film): ν 3402 (br), 2975, 1475, 1364,
8.3.5. 4-Fluoromethyl-4-hydroxy-dihydro-furan-2-one (19).
The 630-L reactor containing crude hydroxy ester 18 from the
previous step was charged with DME (60 L) and 96% sulfuric
acid (0.79 kg, 430 mL, 7.7 mol). The two-phase mixture was
heated under stirring at 55ꢀ60 °C (Tj 80 °C; caution: very
strong gas evolution!) for 4 h. After cooling to rt, sodium acetate
trihydrate (4.0 kg, 29.4 mol) was added to the dark-brown
solution and the mixture was vigorously stirred for 30 min. Ethyl
acetate (64 L) was added and the resulting suspension passed
through a column filled with sodium sulfate (20 kg). The column
was rinsed with ethyl acetate (30 L) and the combined eluents
were evaporated to dryness (45 °C/g40 mbar) to provide crude
lactone 19 (41.6 kg, 102% by weight) as dark-brown oil, GC 98%,
1.4% w/w residual DME, 0.2% w/w residual ethyl acetate. An
analytical sample was obtained by bulb-to-bulb distillation pro-
viding a yellow oil, bp ∼150 °C/0.1 mbar. IR (film): ν 3420 (br),
1
1195, 1076, 1050 cmꢀ1. H NMR (CDCl3, 400 MHz): δ 1.21
(s, 9H, C(CH3)3), 2.23 (t, J = 6.3 Hz, 1H, OH), 3.62 and 3.85
(m, 2H each, CH2O), 4.63 (dm, 2JHF = 48 Hz, 1H, CHF). EI-MS
(m/z) 135 (70), 59 (100), 57.1 (69), 41.1 (13).
8.3.3. 1-tert-Butoxy-3-fluoro-propan-2-one (17). A 630-L
reactor was charged with DCM (145 L), water (145 L), sodium
bicarbonate (11.5 kg), potassium bromide (4 kg), rac-15/rac-16
mixture (54.1 kg, GC 95.2% rac-15 and 3.7% rac-16, 360 mol),
and after cooling to 0 °C, with TEMPO (275 g). Bleach (245 L,
301 kg, ∼10% active Cl, ∼400 mol) was added to the two-phase
mixture under vigorous stirring over a period of 1.5 h while
maintaining the temperature below 10 °C. After stirring for an
additional 15 min, the reaction mixture remained dark-red,
indicating that the oxidation was complete (potassium iodi-
deꢀstarch test positive). A solution of saturated sodium bisulfite
solution (∼1 L) was added until the reaction mixture was fully
discolored (potassium iodideꢀstarch test negative). The two
phases were allowed to separate, and the aqueous phase was
extracted with DCM (2 ꢁ 95 L). The combined organic phases
were washed with brine (31 kg sodium chloride dissolved in 73 L
water) and concentrated (25 °C/g50 mbar) to minimum stir
volume. THF (46 L) was added, and the reactor content was
again concentrated to minimum stir volume. The THF charge/
concentration process was repeated another time to ensure
complete azeotropic removal of DCM. The residual crude ketone
17 was dissolved in THF (30 L) and the solution carried on into
the next step. A yield of 84% was determined for crude 17 based on
evaporation of an aliquot of the solution, GC purity 97.6%. IR
1
2968, 1760, 1295, 1184, 1021, 1000 cmꢀ1. H NMR (CDCl3,
400 MHz): δ 2.63 and 2.74 (AB with fine structure, J = 18 Hz, 1H
each, CH2CO2), 2.85 (br s, 1H, OH), 4.27 and 4.34 (AB with fine
structure, J = 10.2 Hz, 1H each, CH2O), 4.47 (d, 2JHF = 47 Hz,
2H, CH2F). EI-MS (m/z) 135.2 (15), 106.2 (16), 101.2 (6), 76.2
(100), 43.3 (7).
8.3.6. 4-Fluoromethyl-5H-furan-2-one (13). A 630-L Hastel-
loy reactor was charged with crude hydroxy lactone 19 (39.9 kg,
derived from theoretically 290 mol 17) and DCM (190 L), and
the solution was cooled to ꢀ10 °C. Acetic anhydride (36.0 kg,
33.3 L, 352 mol) and triethylamine (35.4 kg, 48.8 L, 350 mol)
were added, each over 15 min, followed by a solution of
4-dimethylaminopyridine (0.73 kg, 6.0 mol) in DCM (10 L)
over 10 min (exothermic!) resulting in a temperature increase to
0 °C. The brownish-black reaction mixture was stirred at 0 °C for
1 h and at rt for 8 h. After quenching with ethanol (4 L) and
stirring for 15 min, the mixture was diluted with DCM (100 L),
washed in two portions with a ∼1 N HCl solution saturated with
sodium chloride (prepared from 122 L water, 40 kg sodium
chloride and 13.5 kg 37% HCl), and with brine (3 ꢁ 80 L).
The combined aqueous phases were back-extracted with DCM
1
(film): ν 2976, 1746, 1367, 1193, 1099, 1063 cmꢀ1. H NMR
(CDCl3, 400 MHz): δ 1.22 (s, 9H, C(CH3)3), 4.15 (d, 4JHF = 1.3
2
Hz, 2H, CH2O), 5.13 (d, JHF = 47.5 Hz, 2H, CH2F). EI-MS
(m/z) 133.1 (6), 115.3 (5), 87.3 (30), 59.3 (13), 57.3 (100).
8.3.4. 3-tert-Butoxymethyl-4-fluoro-3-hydroxybutyric Acid
tert-Butyl Ester (18). A 630-L Hastelloy reactor was charged
522
dx.doi.org/10.1021/op200019k |Org. Process Res. Dev. 2011, 15, 515–526

Organic Process Research & Development
ARTICLE
(2 ꢁ 100 L). The combined organic phases were evaporated, and
the residue was azeotropically dried with toluene (2 ꢁ 110 L,
40ꢀ45 °C/g25 mbar). The residual oily suspension was taken
up in MTBE (160 L), the suspension was stirred at rt for 30 min
and centrifuged, and the cake was washed with MTBE (100 L).
The combined filtrate and wash solution were evaporated, and
the residue was azeotroped with toluene (3 ꢁ 30 L) to ensure
complete removal of acetic acid. The dark-brown residue was
dissolved in DCM (120 L), the solution filtered through a
pressure filter filled with Norit SA II charcoal (1.7 kg) and
Dicalite (5 kg), the filter washed with DCM (35 L), and the
filtrate evaporated to dryness (25ꢀ50 °C/g4 mbar). The
material, after dissolution in ethyl acetate (60 L), was chromato-
graphically filtered over silica gel (100 kg) using heptane/ethyl
acetate 2:1 (total 1600 L), and the fractions containing 13 were
collected and evaporated (45 °C/g12 mbar). The residue was
dissolved in DCM (100 L) and the solution again filtered
through a pressure filter filled with Norit SA II charcoal (1.7
kg) and Dicalite (5 kg). The filter cake was washed with DCM
(70 L) and the filtrate evaporated (45 °C/g15 mbar) affording
crude 13 (28.0 kg, GC 97.3%, assay 93.6%) as oil (caution:
tendency for crystallization!). The material was distilled in four
portions (B€uchi rotavapor R-220) to yield 13 (24.7 kg, 73%) as
colorless oil, bp 78ꢀ83 °C/0.2ꢀ0.1 mbar. For further purifica-
tion the distilled product that meanwhile had solidified was
dissolved in MTBE (75 L), the solution cooled to 16ꢀ18 °C,
seeded with 10 g of 13, and stirred for 4 h. Seeding was repeated
with 10 g of 13 and the mixture cooled to 5ꢀ0 °C over a period of
4 h in which crystallization occurred exothermically. The suspen-
sion was stirred at 0 °C for 12 h, cooled to ꢀ15 °C over a period
of 3 h, and stirred for 3 h. Filtering over a precooled suction-filter,
washing with cold MTBE (25 L, ꢀ15 °C), and drying (rt/<30
mbar/24 h) provided fluoromethyl butenolide 13 (22.6 kg, 67%
based on 19, 36% overall based on rac-14) as white crystalline
powder, GC 100%, assay 98.9%, 0.2% w/w residual MTBE.
8.4. Asymmetric Hydrogenation of 13. 8.4.1. (S)-4-Fluoro-
methyl-dihydro-3H-furan-2-one ((S)-3) by Ru-Catalyzed Hydroge-
nation. In a glovebox (O2 content < 2 ppm), an Erlenmeyer flask
was charged with [Ru(OAc)2((R)-3,5-t-Bu-MeOBIPHEP)] (23,
6.50 g, 5.2 mmol) and methanol (250 mL). The mixture was
stirred for 20 min at rt and transferred into a stainless steel
catalyst addition device. The device was sealed, pressurized with
argon (7 bar), and connected to a 50-L Hastelloy C22 autoclave.
Under an argon atmosphere the autoclave was charged with a
solution of fluoromethyl butenolide 13 (5.40 kg, 46.5 mol,
material prepared by tert-butyl glycidyl ether route) in methanol
(11.5 L). Methanol (10.2 L) was added, and the autoclave was
sealed, evacuated, and pressurized with argon at 1 bar. This
procedure was repeated seven times, and then the catalyst
solution was introduced into the autoclave. Argon was replaced
with hydrogen (15 bar), and the reaction mixture was hydro-
genated under vigorous stirring (600 rpm) at 30 °C for 18 h;
hydrogen uptake g99% after 4 h, conversion by GC 100% after
18 h. The hydrogen atmosphere was replaced with argon, and the
orange reaction solution was transferred to a 20-L flask (using
methanol (8 L) for rinsing) and concentrated (40 °C/g20
mbar). The residue (5.18 kg) was distilled (B€uchi rotavapor
R-220, pressure at vacuum pump ∼0.5 mbar, bp 85ꢀ90 °C) to
provide fluoromethyl lactone (S)-3 (5.10 kg, 93%) as colorless
liquid, GC 99.4%, 96.8% ee, Ru not detectable by X-ray
fluorescence (<10 ppm). Yields of 92ꢀ93% and ee values of
96.3ꢀ96.8% were obtained in four further hydrogenation runs at
the 4ꢀ5 kg scale. IR (film): ν 1766, 1169, 1020, 946 cmꢀ1. 1H
NMR (CDCl3, 600 MHz): δ 2.43 and 2.68 (ABC, JAB = 17.8 Hz,
JAC = 6.5 Hz, JBC = 9.1 Hz, 1H each, CH2CO2), 2.96 (m, 1H,
CHCH2F), 4.24 and 4.45 (ABX, JAB = 9.4 Hz, JAX = 5.7 Hz, JBX
=
7.7 Hz, 1H each, CH2O), 4.47 (dm, 2JHF = 47 Hz, 2H, CH2F).
13C NMR (CDCl3, 150 MHz, 1H-decoupled): δ 29.73 (d, 3JCF
=
2
6.7 Hz, CH2CO2), 35.66 (d, JCF = 20.3 Hz,CHCH2F), 69.18
3
1
(d, JCF = 5.8 Hz, CH2O), 82.67 (d, JCF = 172.2 Hz, CH2F),
175.85 (s, CdO). 19F NMR (CDCl3, 377 MHz, H-coupled):
1
δ ꢀ224.98 (td, 2JFH = 46.8 Hz, 3JFH = 18.7 Hz). EI-MS (m/z)
118.3 (Mþ•, 22), 74.3 (33), 73.3 (11), 60.3 (67), 59.3 (100), 41.4
(43), 39.2 (25), 32.4 (55), 29.4 (18).
8.4.2. (S)-4-Fluoromethyl-dihydro-3H-furan-2-one ((S)-3) by
Rh-Catalyzed Hydrogenation. In a glovebox (O2 content < 2
ppm), an Erlenmeyer flask was charged with [RhCl(COD)]2
(716.7 mg, 1.454 mmol), (S)-TMBTP (22l, 1.889 g, 3.198
mmol), and DCM (135 mL). The mixture was stirred at rt for
15 min and transferred into a stainless steel catalyst addition
device. The device was sealed, pressurized with argon (7 bar),
and connected to a 2-L Hastelloy C4 autoclave, previously
charged with fluoromethyl butenolide 13 (135 g, 1.163 mol,
material prepared by the Wittig route) and DCM (877 mL). The
autoclave was sealed and purged with argon, and the catalyst
solution was introduced into the autoclave. Argon was replaced
with hydrogen (50 bar), and the reaction mixture was vigorously
stirred (800 rpm) at 40 °C for 18 h; conversion 100%. The
hydrogen atmosphere was replaced with argon, and the orange
reaction solution was combined with the solution of an identical
second batch and concentrated (40 °C/g20 mbar). The residue
was distilled through a 15-cm Vigreux column affording fluor-
omethyl lactone (S)-3 (260 g, 94%) as colorless liquid, bp
70ꢀ72 °C/0.04 mbar, GC 96.6%, 93.2% ee.
8.5. Malonate Route. 8.5.1. (S)-tert-Butyl Glycidyl Ether (S)-
14. A 500-mL round-bottomed flask equipped with an air
balloon, a magnetic stirrer, and an ice bath was charged with
(1R,2R)-(ꢀ)-1,2-cyclohexanediamino-N,N0-bis-(3,5-di-tert-butyl-
salicylidene) cobalt(II) (6.04 g, 10 mmol), tert-butyl glycidyl
ether rac-14 (260.4 g, 2000 mmol), acetic acid (2.40 g, 40 mmol),
and THF (20 mL). After cooling to 0 °C, deionized water (19.8 g,
1100 mmol) was added all at once under stirring and cooling in
an ice bath. The dark-red reaction mixture was stirred under an
air atmosphere at 0 °C for 0.5 h and then at rt for 20 h. The
round-bottomed flask was equipped with a Fenske ring-packed
column (20 cm ꢁ 2 cm), and the product was separated from the
higher-boiling (R)-tert-butyl glyceryl ether by distillation at
reduced pressure affording (S)-14 (113.2 g, 43.5%) as colorless
liquid, bp 84ꢀ86 °C/100 mbar, GC 99.8%, >99.9% ee.
[R]20365 = ꢀ23.0 (c 1.0, CHCl3). 1H NMR (CDCl3, 400 MHz):
δ 1.21 (s, 9H, C(CH3)3), 2.62 and 2.80 (ABC, JAB = 5.1 Hz, JAC
=
2.7 Hz, JBC = 4.3 Hz, 1H each, CH2 epoxide), 3.10 (m, 1H,
CHOCH2), 3.42 and 3.53 (ABC, JAB = 10.7 Hz, JAC = 5.3 Hz,
JBC = 3.9 Hz, 1H each, CH2Ot-Bu).
8.5.2. (R)-1-tert-Butoxy-3-fluoro-propan-2-ol ((R)-15). A
white suspension of tert-butyl glycidyl ether (S)-14 (195.3 g,
1500 mmol, 98.0% ee, commercial source), triethylene glycol
(150 mL), potassium hydrogen difluoride (234.3 g, 3000
mmol), and tetrabutylammonium hydrogen sulfate (2.6 g, 7.5
mmol) was stirred at 130 °C for 20 h. After cooling to rt the
brown suspension was diluted with DCM (300 mL) and
washed with water (1.0 L), 5% NaHCO3 (300 mL), and
10% brine (200 mL). All aqueous layers were extracted
sequentially with DCM (150 mL), and the combined organic
523
dx.doi.org/10.1021/op200019k |Org. Process Res. Dev. 2011, 15, 515–526

Organic Process Research & Development
ARTICLE
layers were dried over K2CO3. After removal of the main part
of the solvent by distillation at normal pressure, the residue
(∼300 mL) was distilled over a Fenske ring-packed column
(20 cm ꢁ 2 cm) affording fluorohydrin (R)-15 (149.2 g, 66%)
as colorless liquid, bp 104ꢀ105 °C/100 mbar, GC 99.2%,
98.0% ee. [R]20365 = þ12.0 (c 1.0, CHCl3). IR (film): ν 3420
(br), 1475, 1365, 1197, 1089, 1019 cmꢀ1. 1H NMR (CDCl3,
400 MHz): δ 1.20 (s, 9H, C(CH3)3), 2.47 (d, J = 5.4 Hz, 1H,
the reactor and the filter cake were washed with DCM (2 ꢁ
50 mL). The cold (ꢀ15 °C) filtrate and wash solutions were
added to the 2-L reactor containing the deprotonated malonate
at 0ꢀ5 °C over a period of 1 h. The reaction mixture, initially an
orange solution and, transiently, towards the end of the addition,
an orange gel-like suspension, was stirred at 0 °C for 19 h.
Sulfuric acid (2.0 M, 300 mL) was added to the now clear, orange
solution at <10 °C (exothermic) and the solvent (DCM) was
distilled off (Tj 80 °C, reaction temperature up to 94 °C, strong
isobutylene and CO2 gas evolution). The remaining reaction
mixture (∼400 mL) was heated under reflux at 92ꢀ94 °C
(Tj 110 °C) for 20 h upon which the remaining solvent (DME)
and excess dimethyl malonate were removed by distillation
(bp ∼85 °C/630 mbar). The residue was treated with water
(100 mL) and DCM (700 mL), the organic layer was separated
and washed with 8% NaHCO3 (100 mL), and the aqueous layers
were back-extracted with DCM (400 mL each). The combined
organic layers were evaporated (50 °C/200 mbar), and the
residue was distilled to provide fluoromethyl lactone (S)-3
(32.0 g, 81% based on (R)-15) as a colorless liquid, bp
72ꢀ74 °C/0.5 mbar, GC 98.6%, 99.3% ee.
3
OH), 3.45 (m, 2H, CH2O), 3.95 (dm, JHF = 17.7 Hz, 1H,
2
CHOH), 4.45 (dm, JHF = 47.3 Hz, 2H, CH2F). ESI-MS
(m/z) 151 ([M þ H]þ, 20), 95 ([M þ H ꢀ C4H8]þ, 35). Anal.
Calcd for C7H15FO2 (150.19): C 55.98; H 10.07; F 12.65; O
21.31. Found: C 55.72; H 9.89; F 12.44; O 21.45.
8.5.3. Trifluoromethanesulfonic Acid (R)-1-tert-butoxy-
methyl-2-fluoro-ethyl ester ((R)-28). To a colorless solution of
fluorohydrin (R)-15 (30.0 g, 200 mmol) and pyridine (32 mL,
400 mmol) in DCM (300 mL) at ꢀ10 °C was added trifluor-
omethanesulfonic anhydride (34.6 mL, 210 mmol) over 1 h. Stirring
at ꢀ10 °C was continued for 1 h, and the yellow, cold suspension
was washed with 1.0 M HCl (200 mL), 10% brine (120 mL), and
10% Na2CO3 (120 mL). The aqueous layers were extracted with
DCM (80 mL), and the combined organic layers were dried over
K2CO3 and filtered. After the addition of K2CO3 (1.4 g, 10 mmol),
the solvent was evaporated (20 °C/g10 mbar) affording crude
triflate (R)-28 (56.9 g) as yellow oil (GC 98.5%, 97.8% ee) which
was used without purification in the next step. [R]20D = ꢀ22.8 (c 1.0,
CHCl3). 1H NMR (CDCl3, 400 MHz): δ 1.20 (s, 9H, C(CH3)3),
3.66 (d, J = 5.5 Hz, 2H, CH2O), 4.67 (dd, 2JHF = 47.0 Hz, J = 4.2 Hz,
2H, CH2F), 5.05 (dm, 3JHF = 19.0 Hz, 1H, CHOTf).
8.5.4. (S)-4-Fluoromethyl-dihydro-3H-furan-2-one ((S)-3).
Sodium tert-butoxide (21.1 g, 220 mmol) was dissolved in
DME (200 mL) and dimethyl malonate (31.7 g, 240 mmol)
was added at 20 °C over 15 min. After cooling to 0 °C, the triflate
(R)-28 (56.9 g, ∼195 mmol) was added at 0 °C over 15 min, and
the orange solution was stirred at 0 °C for 7 h. The cooling bath
was removed, 2.0 M H2SO4 (200 mL) was added all at once, and
the yellow reaction mixture was heated under reflux for 48 h
(isobutylene and CO2 gas evolution at e85 °C). After cooling to
rt, the reaction mixture was extracted with DCM (3 ꢁ 200 mL),
and the organic layers were washed with 2% NaHCO3 (100 mL).
The combined organic layers were dried (Na2SO4) and filtered,
and the solvent was removed by rotary evaporation (35 °C/g10
mbar) affording 22.1 g brown oil. Purification by distillation over
a Vigreux column provided fluoromethyl lactone (S)-3 (19.8 g,
84% based on (R)-15) as colorless oil, bp 68ꢀ69 °C/0.5 mbar,
GC 99.2%, 97.8% ee. [R]20D = ꢀ37.3 (c 1.0, CHCl3). Anal. Calcd
for C5H7FO2 (118.11): C 50.85; H 5.97; F 16.09. Found: C
50.42; H 5.90; F 15.92.
’ AUTHOR INFORMATION
Corresponding Author
*ulrich.zutter@roche.com; bruno.lohri@roche.com (for asym-
metric hydrogenations)
Notes
See the accompanying paper in this issue by Abrecht et al. (DOI:
10.1021/op2000207).
’ ACKNOWLEDGMENT
We thank J.-C. Jordan, Dr. N. Burki, Dr. A. St€ampfli and M.
Deichmann and their teams for GC analytical support; Drs. C.
Bartelmus, W. Huber and J. Schneider and their teams for
spectroscopic measurements; and Drs. C. Lautz and T. Glarner
and teams for calorimetric determinations and thermal safety
evaluations. The skillful synthetical assistance of M. Baumann, K.
Bolliger, J. Gallert, F. Gantz, T. Gasser, D. Guthauser, M. Hohler,
M. Lalonde, A. Meili, B. Muller, G. Schaffner, K. Sch€onb€achler,
M. Steiner, E. Stihle, P. Stocker, F. Tixeront, and H.-R. Wetzel is
gratefully acknowledged. We thank also Dr. K. P€untener, P. Dott
and their teams for the preparation of ruthenium catalysts; and
Drs. A. Fettes, M. Karpf, K. P€untener, M. Scalone and G.
Wuitschik for helpful discussions.
8.5.5. Preparation of (S)-3 without Isolation of Triflate (R)-28.
A 2-L jacketed glass reactor was charged with sodium tert-
butoxide (38.4 g, 400 mmol) and DME (335 mL). To the
yellowish solution at 3ꢀ10 °C was added dimethyl malonate
(57.2 g, 433 mmol) over 15 min (exothermic). Separately, a 0.5-L
jacketed glass reactor was charged with fluorohydrin (R)-15
(50.0 g, 333 mmol, obtained from (S)-14 of 99.4% ee), DCM
(350 mL) and, at 5 °C, with pyridine (31.0 mL, 30.5 g, 383
mmol). To the clear colorless solution was added at ꢀ17
to ꢀ7 °C trifluoromethanesulfonic anhydride (56.8 mL, 97.7 g,
346 mmol) over 45 min using an infusion pump. The resultant
suspension was stirred at ꢀ20 to ꢀ15 °C until a conversion
of >99.7% was reached (∼1 h). The cold suspension was passed
through a pressure filter (charged with a 9 cm paper filter) and
’ REFERENCES
(1) Mattei, P.; B€ohringer, M.; Di Giorgio, P.; Fischer, H.; Hennig,
M.; Huwyler, J.; Koc-er, B.; Kuhn, B.; Loeffler, B. M.; MacDonald, A.;
Narquizian, R.; Rauber, E.; Sebokova, E.; Sprecher, U. Bioorg. Med.
Chem. Lett. 2010, 20, 1109.
(2) B€ohringer, M.; Fischer, H.; Hennig, M.; Hunziker, D.; Huwyler, J.;
Kuhn, B.; Loeffler, B. M.;Luebbers, T.;Mattei, M.;Narquizian, R.;Sebokova,
E.; Sprecher, U.; Wessel, H. P. Bioorg. Med. Chem. Lett. 2010, 20, 1106.
(3) Abrecht, S.; Adam, J.-M.; Bromberger, U.; Diodone, R.; Fettes,
A.; Fischer, R.; G€ockel, V.; Hildbrand, S.; Moine, G.; Weber, M. Org.
Process Res. Dev. 2011, DOI: 10.1021/op2000207.
(4) See also: (a) Abrecht, S.; Adam, J.-M.; Fettes, A.; Foricher, J.;
Lohri, B.; Mattei, P.; Moine, G.; Schmid, R.; Zutter, U. (F. Hoffmann-La
Roche AG). PCT Int. Appl. WO 2006/125728 A1, 2006. (b) Zutter, U.
524
dx.doi.org/10.1021/op200019k |Org. Process Res. Dev. 2011, 15, 515–526

Organic Process Research & Development
ARTICLE
(F. Hoffmann-La Roche Inc.). U.S. Pat. Appl. Publ. U.S. 2008/108816
A1, 2008.
(25) Other conditions such as SOCl2/pyridine, POCl3/pyridine,
and the Vilsmeier and Burgess reagents provided also good conversions
(5) The HornerꢀWadsworthꢀEmmons route to hydroxymethyl
butenolide 6 is known; see: (a) Gadir, S. A.; Smith, Y.; Taha, A. A.;
Thaller, V. J. Chem. Res., Synop. 1986, 222. (b) Harrison, T.; Myers, P. L.;
Pattenden, G. Tetrahedron 1989, 45, 5247. (c) Lattmann, E.; Hoffmann,
H. M. R. Synthesis 1996, 155. (d) Balasubramaniam, R. P.; Moss, D. K.;
Wyatt, J. K.; Spence, J. D.; Gee, A.; Nantz, M. H. Tetrahedron 1997,
53, 7429.
and yields. MsCl/Et3N, DEAD/PPh3, and SO3 NEt3 led to unsatisfac-
tory conversions and/or yields.
3
(26) More than 250 strains produced the undesired enantiomer (R)-
3, many with 100% conversion and ee values >98%. Screenings were
performed in a 24 deep-well format at 0.1% w/v substrate concentration.
(27) Iridium catalysts in a short screening afforded only low
conversions and enantioselectivities.
(6) Crawforth, J. M.; Fawcett, J.; Rawlings, B. J. J. Chem. Soc., Perkin
Trans. 1 1998, 1721.
(28) (a) Benincori, T.; Cesarotti, E.; Piccolo, O.; Sannicolꢀo, F. J. Org.
Chem. 2000, 65, 2043.(b) Antognazza, P.; Benincori, T.; Brenna, E.;
Cesarotti, E.; Sannicolꢀo, F.; Trimarco, L. (Italfarmaco Sud S.P.A.). PCT
Int. Appl. WO 96/01831 A1, 1996.
(29) A solvent screening in the Rh-catalyzed hydrogenation with 22l
as the ligand showed DCM (92ꢀ93% ee) and trifluoromethylbenzene
(91% ee) to be preferred solvents. Toluene (73% ee) and the group of
methanol, isopropanol, THF and ethyl acetate (e33% ee) turned out to
be vastly inferior. Whether traces of HX (X = halogen) released from the
solvent contributed to the higher ee’s in the halogenated solvents—as
suggested by a reviewer—was not investigated.
(7) See e.g.: (a) Mori, K.; Yamane, K. Tetrahedron 1982, 38, 2919.
(b) Mori, K.; Chiba, N. Liebigs Ann. Chem 1989, 957. A sample of (R)-7
prepared in-house by the Evans route suffered from ∼90% racemization
upon standing at ambient temperature for three weeks; Dr. P. Mattei,
personal communication.
(8) (a) Ohta, T.; Miyake, T.; Seido, N.; Kumobayashi, H.; Takaya,
H. J. Org. Chem. 1995, 60, 357. (b) Kamlage, S.; Sefkow, M.; Zimmer-
mann, N.; Peter, M. G. Synlett 2002, 77. (c) Bronze-Uhle, E. S.; de Sairre,
M. I.; Donate, P. M.; Frederico, D. J. Mol. Catal. A: Chem. 2006, 259, 103.
(9) Hughes, G.; Kimura, M.; Buchwald, S. L. J. Am. Chem. Soc. 2003,
125, 11253.
(30) Preformed Ru catalysts were prepared as described; cf.:
(a) Heiser, B.; Broger, E. A.; Crameri, Y. Tetrahedron: Asymmetry
1991, 2, 51.(b) Cereghetti, M.; Foricher, J.; Heiser, B.; Schmid, R. (F.
Hoffmann-La Roche Inc.). U.S. Patent 5,488,172, 1996. (c) Chan,
A. S. C.; Laneman, S. Inorg. Chim. Acta 1994, 223, 165. Other Ru
catalysts were made in situ from [Ru(OAc)2(COD)] and the bispho-
sphine ligand in the presence of the starting material 13 and MeOH
(40 °C, 2 h). Only marginal ee differences were observed in cases where
both the preformed and the in situ prepared catalyst were evaluated.
(31) At this point, it may be of interest to note that hydrogenation of
the analogous nonfluorinated compound 3-methyl-2-buten-4-olide with
Ru catalyst 23 provided (S)-3-methyl-γ-butyrolactone with ee values of
only 77% and 59% ee at S/C 100 and 5000, respectively (50 bar H2, 40
°C, 20 h, MeOH). On the other hand, the Rh-catalyzed hydrogenation
(10) Geiger, C.; Kreitmeier, P.; Reiser, O. Adv. Synth. Catal. 2005,
347, 249.
(11) See e.g. Takabe, K.; Tanaka, M.; Sugimoto, M.; Yamada, T.;
Yoda, H. Tetrahedron: Asymmetry 1992, 3, 1385.
(12) Asymmetric hydrogenation of hydroxymethyl butenolide 6 was
reported to lead to material of very low ee, likely due to the propensity of
the optically active product 7 to racemize (cf. reference 8c).
(13) Commercial material was used, or the ylide was prepared and
isolated from [(ethoxycarbonyl)methyl]triphenylphosphonium bro-
mide in a NaOH/DCM two-phase system.
(14) 1,3-Dihydroxyacetone (12) in substance exists mainly in the
dimeric hemiacetal form unless freshly distilled; see: (a) Fischer,
H. O. L.; Mildbrand, H. Ber. Dtsch. Chem. Ges. B 1924, 57B, 707.
(b) Davies, L. Bioorg. Chem. 1973, 2, 197. (c) Yoda, H.; Mizutani, M.;
Takabe, K. Synlett 1998, 855. The monomer/dimer ratio had no
influence on the reaction.
of 3-methyl-2-buten-4-olide ([RhCl(COD)]2
50 bar H2, 40 °C, 20 h, DCM) provided 92% ee, equal to the ee
obtained for 13.
þ
22l, S/C 100,
(32) Banfi, L.; Basso, A.; Guanti, G.; Zannetti, M. T. Tetrahedron:
Asymmetry 1997, 8, 4079.
(33) Yoshida, Y.; Sakakura, Y.; Aso, N.; Okada, S.; Tanabe, Y.
Tetrahedron 1999, 55, 2183.
(15) The formerly used aqueous protocol (10% H2SO4 in MeOH)
was less selective and led to material that was difficult to purify by
distillation or crystallization.
(16) An alternative two-step access to 6 involved reaction of 12 with
Ph3PdCHCO2t-Bu and ester cleavageꢀcyclization with 0.25 M HCl in
methanol. Hydroxymethyl butenolide 6 was obtained in 73% yield over
two steps after chromatography which proved inevitable.
(17) Singh, R. P.; Shreeve, J. M. Synthesis 2002, 2561 and references
cited therein.
(34) For an independent synthesis based on a microbial
BayerꢀVilliger oxidation of 3-benzyloxy-cyclobutanone see: Mazzini,
C.; Lebetron, J.; Alphand, V.; Furstoss., R. J. Org. Chem. 1997, 62, 5215.
For a synthesis of (R)-27 starting from (R)-paraconic acid ((R)-11) see:
Mori, K. Tetrahedron 1983, 39, 3107.
(35) For the preparation of optically active (S)-tert-butyl glycidyl
ether see: (a) Thakur, S. S.; Li, W.-J.; Shin, C.-K.; Kim, G.-J. Chirality
2005 (Volume Date 2006), 18, 37. (b) Kim, M.-J.; Lim, I. T.; Choi, G.-
B.; Whang, S.-Y.; Ku, B.-C.; Choi, J.-Y. Bioorg. Med. Chem. Lett. 1996,
6, 71. (c) Kotik, M.; Brichac, J.; Kyslik, P. J. Biotechnol. 2005, 120, 364.
(36) Schaus, S. E.; Brandes, B. D.; Larrow, J. F.; Tokunaga, M.;
Hansen, K. B.; Gould, A. E.; Furrow, M. E.; Jacobsen, E. N. J. Am. Chem.
Soc. 2002, 124, 1307 and references cited therein.
(18) Yin, J.; Zarkowsky, D. S.; Thomas, D. W.; Zhao, M. M.;
Huffman, M. A. Org. Lett. 2004, 6, 1465.
(19) For a review see: Haufe, G. J. Fluorine Chem. 2004, 125, 875.
(20) Van Hijfte, L.; Heydt, V.; Kolb, M. Tetrahedron Lett. 1993,
34, 4793.
(21) The primary alcohol rac-16 is slightly less volatile than the
secondary alcohol rac-15. Pure samples of rac-15 and rac-16 were
obtained by chromatography.
(37) The lower boiling (S)-14 was separated from the higher-boiling
(22) Further polar solvents such as DMF, DMA, NMP, DMPU, and
DMSO proved poor in terms of chemoselectivity and yields and/
or rates.
(R)-tert-butyl glyceryl ether by distillation.
(38) When the reaction mixture was warmed up and stirred at
ambient temperature, increasing racemization was observed (1 h: 0.75%;
24 h: 23% (S)-28 formation).
(39) At ambient temperature the stability of the neat triflate (R)-28
could be extended to 1ꢀ2 days by the addition of solid K2CO3 or DIPEA
(5 mol%). As solution in DCM (R)-28 was stable at ambient tempera-
ture for two days without any additive. Although the corresponding
mesylate, tosylate and nosylate derivatives were stable compounds, only
the triflate (R)-28 was sufficiently reactive to effect a selective and
efficient malonate substitution.
(23) Et3N 2HF has been reported to be more nucleophilic than
3
Et3N 3HF, see: (a) Giudicelli, M. B.; Picq, D.; Veyron, B. Tetrahedron
3
Lett. 1990, 31, 6527. (b) Chou, T. S.; Becke, L. M.; O’Toole, J. C.; Carr,
M. A.; Parker, B. E. Tetrahedron Lett. 1996, 37, 17.
(24) The use of catalytic amounts of conc. H2SO4 allowed for a
nonaqueous workup (quenching with solid NaOAc followed by filtra-
tion over Na2SO4), hence, avoiding extractive isolation of the rather
water-soluble hydroxy lactone 19. Use of excess of conc. H2SO4
necessitated an aqueous workup and exhaustive extraction of 19 with
ethyl acetate leading to high volume factors.
(40) In order to minimize HF elimination to the cyclopropane side
product (S)-29a, the excess of sodium dimethyl malonate (1.1 equiv)
525
dx.doi.org/10.1021/op200019k |Org. Process Res. Dev. 2011, 15, 515–526

Organic Process Research & Development
ARTICLE
was kept as low as possible. With g2 equiv sodium dimethyl malonate,
the intermediate (S)-29 was completely converted into (S)-29a.
(41) The triflate substitution with sodium dimethyl malonate
(generated either with NaH or preferably with NaOt-Bu) was tested
in the solvents DME (0 °C, ∼7 h) and THF (0 °C, ∼20 h). DME was
preferred over THF due to the higher boiling point (86 vs 66 °C)
allowing a higher reflux temperature and accelerating the hydrolysis of
the malonic ester and the subsequent decarboxylation with aq H2SO4.
(42) The additive DMI was superior to DMPU and preferred over
the highly toxic HMPA; cf.: Al Abed, Y.; Zaman, F.; Shekhani, M. S.;
Fatima, A.; Voelter, W. Tetrahedron Lett. 1992, 33, 3305.
(43) (R)-Glycidyl 3-nitrobenzenesulfonate (R)-33 was offered in
bulk quantity by an external supplier.
(44) (a) Pirrung, M. C.; Dunlap, S. E.; Trinks, U. P. Helv. Chim. Acta
1989, 72, 1301. (b) Kitaori, K.; Mikami, M.; Furukawa, Y.; Yoshimoto,
H.; Otera, J. Synlett 1998, 499. (c) Mikami, M.; Furukawa, Y.; Oodera, J.;
Yoshimoto, P. H. (Daiso Co., Ltd.). Jpn. Pat. JP 10059955, 1998.
(45) Et3N 3HF (37% HF in Et3N), Honeywell No. 1790. For HF
3
addition to activated cyclopropane rings see: Cotterill, I. C.; Finch, H.;
Highcock, R. M.; Holt, R. A.; Mahon, M. F.; Molloy, K. C.; Morris, J. G.;
Roberts, S. M.; Short, K. M.; Sik, V. J. Chem. Soc., Perkin Trans. 1
1990, 1353.
(46) It is noteworthy that under basic (Na2CO3) or thermal condi-
tions (GC, g200 °C) the lactones 30 and 35 partially eliminate HX
providing cyclopropanolactone 34.
(47) Experiments with corrosive HF reagents such as Et3N 3HF
3
were performed in HF-resistant perfluoro-alkoxy-copolymer PFA-flasks
(Supplier: Bohlender GmbH, Waltersberg 8, D-97947 Gr€unsfeld).
(48) Cf. Steffen K.-D. (H€uls AG). U.S. Patent 5,463,111, 1995.
(49) The main source of fluoride is the formation of cyclopropane
(S)-29a in the malonate substitution reaction (cf. reference40). In the
ensuing treatment with 2 M H2SO4 (S)-29a is further converted into
cyclopropanolactone 34 and hydroxymethyl lactone 7 which were
observed as trace byproducts.
526
dx.doi.org/10.1021/op200019k |Org. Process Res. Dev. 2011, 15, 515–526