Process research and development and scale-up of a 4,4-difluoro-3,3- dimethylproline derivative

Article abstract of DOI:10.1021/op7001112

The multikilogram production of the proline derivative 1, a key intermediate of a HIV protease inhibitor, required the design of a synthetic route able to be safely, effectively, and easily scaled up. Synthesis of the proline skeleton began with construction of racemic glycine derivative 4, via an ester enolate Claisen rearrangement of Boc-glycine 3-methyl-but-2-enyl ester (3) in the absence of a Lewis acid. After a classical resolution of 4 with (S)- phenylglycinol, (S)-4 was transformed into bromo-lactone 6b with NBS. The bromo-lactone was transformed to proline alcohol 8 via a base-promoted rearrangement involving lactone solvolysis. An NMR study suggested that a bicyclic lactone was initially formed, which subsequently opened by the methanol solvent to form 8. The requisite ketone for fluorination was prepared via oxidation of the enantiomerically pure 8, using NaClO and catalytic TEMPO. gem-Difluoro proline 1 was then prepared from the ketone via fluorination with Deoxo-Fluor. During this study it was discovered that SiO₂ promoted fluorination by Deoxo-Fluor. This study allowed the production of 7.5 kg of 1 after 10 steps, in 4.5% molar yield and high purity (94-99% HPLC assay).

Full text of DOI:10.1021/op7001112

Organic Process Research & Development 2008, 12, 322–338
Full Papers
Process Research and Development and Scale-up of a
4,4-Difluoro-3,3-dimethylproline Derivative
Francesco Rossi,† Francesco Corcella,† Francesco Saverio Caldarelli,† Franco Heidempergher,† Chiara Marchionni,†
Miriam Auguadro,† Marco Cattaneo,† Lucio Ceriani,† Giuseppina Visentin,†,‡ Giambattista Ventrella,†,‡ Vittorio Pinciroli,†,‡
Giuliano Ramella,† Ilaria Candiani,† Angelo Bedeschi,† Attilio Tomasi,*,† Billie J. Kline,§ Carlos A. Martinez,§
Daniel Yazbeck,§ and David J. Kucera*,§
ActiVe Pharmaceutical Ingredients and Analytical DeVelopment & Quality Control, NerViano Medical Sciences, Pharmaceutical
Sciences Unit Viale Pasteur 10, 20014 NerViano, Italy, and Chemical Research and DeVelopment, La Jolla Laboratories, Pfizer
Inc., 10578 Science Center DriVe, San Diego, California 92121-1111, U.S.A.
Abstract:
During scale-up of the original synthetic pathway, many
liabilities were noted: (1) in the absence of an enantiomer
recycling method, the bulk of the synthesis required twice the
quantity of materials to be processed, (2) the CLEC subtilisin
Carlsberg enzyme used in the optical resolution was no longer
commercially available, (3) the overall yield of isolated 1 was
about 1.5%, (4) the calculated sum of all materials (including
solvents and reagents) needed to produce 2 kg of 1 was
estimated to be about 13 t, (5) a large quantity of flammable
solvents were employed (about 8,000 L for 2 kg of 1), and (6)
very low space-time yields were calculated (estimated frame
time to produce 2 kg of 1 was of about 1 year). In order to
furnish enough material, greater than 2 kg of 1 was required
along with a more suitable synthetic pathway. New route
exploration demonstrated the possibility to optically resolve
intermediate 4 and use a single enantiomer throughout the
synthesis. In this explorative study some racemization was
observed during the basic rearrangement and the oxidation steps
(6 to 7, 8 to 9), affording the final product 1 with 62% ee from
4 with 96% ee. Thus, in light of these results and the liability
of the original route, this new route was developed. The ensuing
process development employed the following strategy: (1)
quickly transform several hundred grams of material according
to the unoptimized conditions to test the proposed procedure,
which would allow establishment of analytical controls and
provide material for examination of various steps, (2) identify
potential issues for the scale-up in the available plant, (3) verify
changes that may improve or simplify the process, (4) isolate
and characterize all intermediates and their main by-product,
(5) carry out the Process Safety Reviews (PSRs), (6) where
necessary carry out a preliminary run at reduced scale, and (7)
implement the process at full scale.
The multikilogram production of the proline derivative 1, a key
intermediate of a HIV protease inhibitor, required the design of
a synthetic route able to be safely, effectively, and easily scaled
up. Synthesis of the proline skeleton began with construction of
racemic glycine derivative 4, via an ester enolate Claisen rear-
rangement of Boc-glycine 3-methyl-but-2-enyl ester (3) in the
absence of a Lewis acid. After a classical resolution of 4 with (S)-
phenylglycinol, (S)-4 was transformed into bromo-lactone 6b with
NBS. The bromo-lactone was transformed to proline alcohol 8 via
a base-promoted rearrangement involving lactone solvolysis. An
NMR study suggested that a bicyclic lactone was initially formed,
which subsequently opened by the methanol solvent to form 8.
The requisite ketone for fluorination was prepared via oxidation
of the enantiomerically pure 8, using NaClO and catalytic
TEMPO. gem-Difluoro proline 1 was then prepared from the
ketone via fluorination with Deoxo-Fluor. During this study it was
discovered that SiO2 promoted fluorination by Deoxo-Fluor. This
study allowed the production of 7.5 kg of 1 after 10 steps, in 4.5%
molar yield and high purity (94–99% HPLC assay).
Introduction
Proline derivative 1 is an intermediate for the synthesis of a
HIV protease inhibitor.1,2 Compound 1 Was originally synthe-
sized by transformation of Boc-glycine 2 to iodo-lactone 6a,
which under basic conditions underwent a rearrangement to
alcohol 8, which was then oxidized to ketone 9. gem-Difluoro
derivative 10 was prepared by nucleophilic fluorination of 9,
and the title compound 1 was obtained enantiomerically pure
by enzymatic hydrolysis of the ester group of 9, using CLEC
(cross-linked enzymes chrystals) subtilisin Carlsberg.
Results and Discussion
* To whom correspondence should be addressed. Tel: +39.0331.58.1312.
Fax: +39.0331.58.1348. E-mail: attilio.tomasi@nervianoms.com.
† Active Pharmaceutical Ingredients, Nerviano Medical Sciences.
‡ Analytical Development & Quality Control, Nerviano Medical Sciences.
§ Pfizer Inc.
(1) Hammond, J. L.; Patick, A. K. U.S. Pat. Appl. Publ. 2005171038,
2005.
Esterification of Boc-Glycine. In the process originally
developed, a solution of Boc-Gly-OH (2), 3-methyl-2-buten-
1-ol (prenol), and N,N-dimethylaminopyridine (DMAP) in
MTBE was treated with a solution of DCC. The resulting
suspension was stirred at room temperature until disappearance
of 2 was evident by HPLC. The suspension was then filtered,
(2) Kucera, D. J.; Saeed, N. L.; Scott, R. W. PCT Int. Appl. 2005054187,
2005.
322
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development
10.1021/op7001112 CCC: $40.75
2008 American Chemical Society
Published on Web 02/21/2008

Scheme 1. Original synthetic approach to 1; (S)-1 secured
by enzymatic resolution of rac-10
Scheme 2. New route after development; (S)-4 obtained by
a classical crystallization, thereby halving the total amount
of material processed from intermediate 4 to 1
These improvements allowed scale-up of the process to 30
kg of 2 in a 250-L reactor. Average results on this scale were
82% yield, 93.3% HPLC purity, and 93.0% HPLC assay.
Ireland-Claisen Rearrangement. The Claisen rearrange-
ment of ester 3 to the glycine 4 derivative implied the use of
ZnCl2 as a Lewis acid, as this transformation is well-prece-
dented. The yield of the original method for the sigmatropic
rearrangement was high (about 80%). Nevertheless, there were
concerns about scaling up this process to convert 120 kg of 3
in a 170-L cryogenic reactor, because of (1) high cost of the
commercial ZnCl2/THF solution; (2) toxicity of ZnCl2; besides
its high toxicity in an aqueous environment, it is also considered
a potential mutagen and carcinogen; (3) low throughput; the
dilute ZnCl2 solution is about 50% of the reaction volume;
processing of 120 kg of 3 with the original conditions in our
plant would require 12 batches, 3 days per batch, 36 working
days, 108 man-days, and 8 working weeks; (4) waste produc-
tion; about 100 L of waste ZnCl2 solution is produced per batch;
moreover, ZnCl2 must be destroyed separately from other
process waste streams, according to the Italian and European
laws.
and after an extractive workup, 3 was isolated almost quanti-
tatively as pale yellow oil.
Limiting factors that were identified in the original procedure
were large quantities of MTBE used (low flash point) and
the filtration and drying of large amounts of DCU by-product.
The filtration represented a major constraint for scale-up,
because of the filter capacity in our plant. Another drawback
was the residual prenol (about 15 mol %) present in isolated 3,
which would entail the use of additional LDA in the next step.
Substitution of DCC with N-(3-dimethylaminopropyl)-N′-
ethylcarbodiimide (EDC) was examined, since EDC and its
corresponding urea are water-soluble. CH2Cl2 was chosen as
the solvent in order to permit the reaction to be carried out in
a very concentrated manner (about 30 wt % Boc-Gly-OH) with
a very short cycle time. These conditions resulted in only a
small quantity of residual prenol in isolated ester 3 (about 1.8%
by HPLC assay). The urea by-product remained in the aqueous
phase after the extractive workup, thus avoiding the filtration
step. The product was isolated simply by concentration to a
fluid oil of high purity.
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
323

Chart 1. Influence of maturation time and temperature
during the Ireland-Claisen rearrangement of 3a
Figure 1. Major impurities isolated during the synthesis of 4.
In the original procedure, the reaction mixture was held for
1 h at -65 ( 5 °C after LDA addition and was then warmed
to 25 °C and allowed to react for about 6 h. Interestingly, the
reaction required that prescribed time at 25 °C to reach the
completion. When the reaction was quenched directly after LDA
addition, only a 52% yield of 4 was obtained. The maximum
conversion was obtained after about 4–5 h at 25 °C.
A literature survey indicated that ZnCl2 can be replaced by
Lewis acids commonly used for the Claisen rearrangement, such
as BCl3, MgCl2, AlCl3, etc.3 Ireland and co-workers 4,5 first and
Bartlett and co-workers6 later showed the ability of trimethylsilyl
chloride to promote the Ireland-Claisen rearrangement in the
absence of the Lewis acid. These conditions6 and modifications
thereof were unsuccessful for the rearrangement of 3.
a Product 4 slowly decomposed under isothermal conditions at 20 °C (graph
a). Heating did not increase the extent of conversion, but rather decreased the
quality and yield (graph b).
Although most Claisen rearrangements are performed in the
absence of a catalyst, the Ireland-Claisen rearrangement seems
to require either a Lewis acid or an enolate-quenching agent
(such as trimethylsilyl chloride) to depress side reactions of the
formed enolate. Kazmaier and co-workers7 claimed that, in the
absence of the stabilizing effect of a chelating metal (i.e., ZnCl2),
the corresponding lithium enolates of sterically demanding
R-alkylated amino acids do not undergo the rearrangement but
rather decompose during warming.
Table 1. Effect of variables on the yield of 4 and the
formation of prenola
LDA
Despite these concerns, the aim was to develop a simple,
safe, and productive process, and thus preparation of 4 in the
absence of a Lewis acid was examined. Surprisingly, the yields
of 4 produced in the first experiments were about 55%. Two
major impurities were identified: urea 11 and carbamate 12
(although by-product 11 was isolated and characterized (see
Experimental Section), the structure of 12 was inferred from
its MS spectra [C13H20NO4]- 254; [C8H10NO3]- 168). The
presence of 11 and 12 suggested that an excess of both LDA
and prenol was detrimental to the quality and the yield of 4
and that optimization of the synthesis of 3 could play an
important role by reduction of residual prenol and thus of LDA.
The new scenario outlined from these exploratory experi-
ments was considered a promising starting point and an
acceptable compromise between avoiding ZnCl2 and obtaining
slightly lower yields. Consequently this new approach was
developed and optimized.
temp
LDA addition
4
3
prenol
entry (°C) (equiv)
(h)
(mol %) (mol %) (mol %)
1
2
3
4
5
6
7
-30
2.2
2.8
3.4
2.8
3.4
2.8
2.2
0.5
1.8
3
1.8
0.5
1.8
3
54.6
78.0
51.2
73.1
81.6
76.0
78.6
1.8
1.0
2.5
0.9
0.8
0.9
0.9
17.8
9.1
19.6
10.6
7.7
8.6
9.2
-47.5
-30
-47.5
-65
-47.5
-65
a Temperature during LDA addition was an important variable. The yield and
quantity of 4 decreased above -30 °C.
an increased maturation period (beyond 6 h) was detrimental
to the yield and quality of 4; a maturation period of 24 h reduced
the yield of 4 from 70% to about 50%. Similar results were
obtained for an increase in the maturation temperature; heating
the reaction mixture from 20 to 60 °C dramatically increased
impurities and decreased yields to about 30%. Additional studies
indicated that the quality and yield of 4 were not dependent on
the warming rate from -65 to 20 °C.
To better define the process, other variables were also
examined: (1) reaction temperature (-65 to -30 °C), (2) LDA
addition rate (0.5 to 3 h), and (3) LDA quantity (2.2 to 3.4
equiv). A set of experiments was designed using a fractional
factorial design, three center points, and no replication. The data
were fitted by linear regression analysis, and the regression
explains 83% of the data (R2 ) 82.7).
Once LDA was added to a solution of 3, the mixture could
be warmed to room temperature and then quenched. Conversely,
after warming, the reaction mixture could be stirred for an
additional period of time (maturation). A study aimed to
establish the best operating procedure, including warming rate
and maturation, period was undertaken. The results showed that
(3) Lutz, R. P. Chem. ReV. 1984, 84, 205–247.
(4) Ireland, R. E.; Mueller, R. H. J. Am. Chem. Soc. 1972, 94, 5897–
5898.
(5) Ireland, R. E.; Mueller, R. H.; Willard, A. K. J. Am. Chem. Soc. 1976,
98, 2868–2877.
From these experiments, only the reaction temperature
appeared to be a critical variable. Maximum yields of 4 were
(6) Bartlett, P. A.; Barstow, J. F. J. Org. Chem. 1982, 47, 3933–3941.
(7) Kazmaier, U.; Maier, S. Tetrahedron 1996, 52, 941–954.
324
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development

Table 2. Physical parameters of the salts between rac-4 and (S)-phenylethylamine and (S)-phenylglycinol determined by DSCa
entry
salt
∆Hp(kJ/mol)
T50 (K)
Teu (K)
1/C
xeu
Rmax (%)
1
2
(S)-phenylethylamine/rac-4
(S)-phenylglycinol/rac-4
5.9
16.2
424
435
409
410
18
3.6
0.37
0.23
20.3
35.3
a ∆Hp is the melting enthalpy of the salt between (S)-4 and the (S)-enatiomer of the chiral base. T50 is the melting temperature of the salt between rac-4 and the resolving
base. Teu is the temperature at the eutectic composition. Xeu is the eutectic composition, and Rmax is the maximum obtainable yield. C is given by the following formula:
1 - 2xeu
2 - 2xeu
∆Hp
1
1
C )
ln 2xeu
)
-
(
)
R
T50 Teu
.
obtained at -65 °C (78–80%), but only a small decrease in
yield was observed at -48 °C (about 75%). LDA addition at
temperatures above -48 °C rapidly decreased yields, and at
-30 °C yields were around 50% (Table 2, entries 1 and 3).
The quantity of prenol formed during the reaction was tem-
perature-dependent; it increased from 8-10% below -50 °C
to 18–20% at -30 °C. No correlation was detected for residual
starting material 3, which was constant at 1.0–2.5%.
The reaction was further investigated, using in situ ATR
FTIR analysis, with the aim to understand better the role of
temperature and the stability of the species formed during the
rearrangement. The reaction was run several times under the
optimized conditions, and IR analysis evidenced two major
changes occurring during the transformation from 3 to 4: (1)
the decrease of the absorption bands at 1752 and 1714 cm-1,
and (2) the formation of a new absorption band at 1620 cm-1.
From the comparison of IR spectra of Boc-Gly-OH 2, ester
3, and acid 4, the absorption bands at 1752 and 1714 cm-1
were assigned to the CdO stretching of the Boc group and of
the ester and acid moiety present in 3 and 4, respectively. The
signal at 1620 cm-1 was assigned to the stretching of the aza-
enolate and of the conjugated enolate formed upon addition of
1 and 2 equiv of LDA, respectively. This assumption was
confirmed by reaction of the lithium salt of 2 with LDA; the
addition of LDA caused the appearance of a signal at 1620
cm-1, whose intensity doubled after addition of the first
equivalent of LDA. Since this substrate cannot undergo sig-
matropic rearrangement, this can be explained with the sequen-
tial formation of the aza- and conjugated enolates.
The collection of the IR spectra recorded during the reaction
was analyzed by chemometric software. During LDA addition
at -65 °C, the absorption band at 1620 cm-1 increased, while
the absorptions of the starting material (1752 and 1714 cm-1)
correspondingly decreased. The formation of the new species
was fast8 at cryogenic temperature; the transformation time was
approximately 1 min, since IR spectra were collected every min.
Nevertheless, once the intermediate was formed, it remained
unchanged for at least 1 h at - 65 °C. As soon as the
temperature was raised to 0 °C, this intermediate species
suddenly rearranged.
These experimental results can be explained by the follow-
ing: (a) Ester 3 reacts quickly with LDA at -65 °C, forming
sequentially the aza- and the conjugated enolate 3a and 3b. (b)
Enolate 3a and 3b is quite stable at cryogenic temperatures but
becomes highly reactive when the temperature is warmed to 0
°C. (c) At around 0 °C, enolate 3b undergoes rearrangement,
eventually forming 4a. Parallel to this mechanism, another
pathway can occur, leading to the formation of by-product. For
example, either the aza-enolate 3a, conjugated enolate 3b, or
product 4a can decompose to isocyanate species 3c, 3d and
4b, respectively, which in turn can react with prenol or LDA
to yield impurities 11 and 12.
From this investigation, the cryogenic temperature appears
necessary to prevent decomposition of aza-enolate 3a to
isocyanate 3c. Once the conjugated enolate 3b is formed, it is
reasonably stable and must be warmed to effect the sigmatropic
rearrangement. When Boc-Gly-OLi was reacted with only 1
equiv of LDA at -65 °C, followed by warming to 0 °C, in
situ ATR FTIR spectroscopy revealed disappearance of the Boc
signals. The resulting mixture was then quenched with EtOH,
to yield the ethyl carbamate of glycine.
Figure 2. (a) ART FTIR analysis of the rearrangement of 3.
The addition of LDA caused formation of a new absorption
band around 1620 cm-1. The absorbance of this new signal was
proportional to the quantity of added LDA. It remained
constant after the end of addition and for an aging time of about
1 h. (b) Selected IR spectra before addition of LDA (black),
after addition of 2 equiv of LDA (red), and after raising the
temperature to 0 °C (blue).
In the original process, 4 was isolated as an oil after phase
separations and concentration using MTBE as main solvent.
(8) Gaul, C.; Arvidsson, P. I.; Bauer, W.; Gawley, R. E.; Seebach, D.
Chem.-Eur. J. 2001, 7, 4117–4125.
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
325

was developed in which the reaction mixture was quenched
into water, and the rich organic phase was then extracted with
aqueous NaOH. Some hexane was added to the rich aqueous
layer, and the pH was then adjusted to 2–3 with aqueous HCl,
which afforded 4 as a filterable solid. Hexane was found useful
for removal of lipophilic impurities, which otherwise caused
oiling out of product 4. The isolation of intermediate 4 as a
solid improved the quality. Obtaining this racemic mixture with
a consistently high purity was a prerequisite for ensuring a
successful optical resolution.
The optimized process was employed for the campaign
production of 4 from 117 kg of 3, using the following
conditions: 2.5 equiv of LDA at -65 °C, with a 2–3 h LDA
addition time. Residual 3 was about 0.8%, formed prenol was
about 8.0 mol %, and the yield of 4 was about 84.7% before
quenching (HPLC assay) and 72% isolated. Interestingly, after
optimization, the performance of the ZnCl2-free process was
comparable to that of the original process that included ZnCl2.
Optical Resolution. Early studies indicated that enantio-
merically pure phenylglycinol or phenylethylamine were suit-
able optical resolution agents for 4. To accelerate the choice
between these two chiral bases, the DSC-based approach
9,10
developed by Kozma and Bruggink
was implemented.
According to this method, DSC analysis of the salt formed by
racemic 4 and the resolving agent gives the eutectic composition
(xeu), which indicates the maximum obtainable yield11 and
therefore the effectiveness of the studied resolving agent. [The
maximum obtainable yield is always lower than 50%, according
to the equation Rmax ) 100 ·(0.5 - xeu)/(1 - xeu)]. DSC analysis
of racemic 4 and phenylethylamine gave a maximum obtainable
yield of 20%, while the value calculated for the phenylglycinol
system was greater (35%). On the basis of these results,
phenylglycinol was chosen as a resolving agent.
Figure 3. (a) Concentration profiles of the main species present
during the reaction. The first intermediate (blue profile) formed
upon addition of LDA, while the concentration of starting 3
decreased (red profile). When the temperature was raised to 0
°C (around 130 min) the conjugated enolate disappeared,
eventually forming the product 4a (gray profile). (b) The
calculated IR spectra relative to the main reaction species are
in good agreement with the expected spectra. The red profile
corresponds to 3, and the blue profile corresponds to 3b. The
gray curve has characteristic signals at about 1754 and 1712
cm-1, as expected for 4 or 4a.
Using enantiomerically enriched 4, the entire cascade from
intermediate 4 to final product 1 was run to confirm that (+)-
(S)-phenylglycinol was the enantiomer to be used to isolate
desired (S)-4 and eventually 1.
Scheme 3. Rationalization of the [3,3]-sigmatropic
rearrangement of Boc-glycine ester 3
Early solvent screening indicated that the resolution per-
formed well with a mixture of CH3CN/MeOH. With the aim
to find a less toxic and less flammable solvent system, other
solvents were explored. Among other solvents, attention was
focused on acetone or CH3CN and H2O. Acetone/H2O (9:1)
dissolved salt n 5-fold more than salt p (the desired diastere-
omeric salt) and afforded salt p in 100% ee, but unfortunately
in low yields (28%). [According to the nomenclature proposed
by Wilen and co-workers,11 salt p refers to the salt pair for which
the sign of the optical rotation for each component of the salt
is the same (for instance (+), (+) or (-), (-)]. Conversely,
salt n was approximately 10-fold more soluble than salt p in
CH3CN/MeOH (9:1), which in principle should lead to high
purity and higher yields. It was found that salt p can be resolved
up to 86% ee after one crystallization and that the quality can
be improved up to the target 96% ee by suspension of the salt
(9) Kozma, D.; Pokol, G.; Acs, M. J Chem. Soc., Perkin Trans. 2 1992,
435–439.
(10) Ebbers, E.; Ariaans, G. J. A.; Zwanenburg, B.; Bruggink, A.
Tetrahedron: Asymmetry 1998, 9, 2745–2753.
To simplify the workup, isolation of 4 directly from water at
acidic pH in the absence of MTBE was envisioned. A procedure
(11) Jacques, J.; Collet, A.; Wilen, S. H. Enantiomers, Racemates, and
Resolutions; Krieger Publishing Company: Malabar, FL, 1994.
326
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development

Scheme 4. cis-Lactone isomer favored from
halolactonization (9:1)
EtOAc was chosen as reaction solvent because it could also
be used for the recovery of (S)-4 from salt p and for the Boc-
deprotection of 5. An additional advantage of EtOAc was that
6b could be recovered quantitatively as its hydrochloride salt
by performing the deprotection with HCl (g).
Examination of reaction variables for the preparation of 5b
focused on the reaction temperature and the influence of water
(possibly carried in from the isolation of (S)-4). The presence
of dissolved H2O did not influence the yield or diastereomeric
ratio (Table 3, entries 1 vs 2, and 3 vs 4,), but a lower
temperature resulted in a slightly greater cis/trans ratio (Table
3, entries 1 and 3, or 2 and 4). The optimal conditions for the
production of 5b were set at 0 °C, with 1.1 equiv of NBS.
For removal of the Boc group of 5b, the following conditions
were explored: TFA/EtOAc, HCl(aq)/EtOAc, and HCl (g)/
EtOAc. The goal was to obtain 6b as a filterable solid directly
after the deprotection, avoiding distillation or other manipula-
tions. TFA was eliminated because of its cost and the formation
of by-product. HCl(aq) was eliminated because a substantial
amount of 6b remained dissolved and thus low yields were
obtained. HCl(g)/EtOAc proved to be the best solution. After
some trials the optimal concentration of cis/trans-5b in EtOAc
and the stoichiometry of HCl were defined (Table 4, entry 3).
Complete deprotection of 5b required about 14 h, but most of
the conversion occurred in the first hour, which was followed
by sudden crystallization of 6b. To avoid the spontaneous
crystallization, HCl was charged in two portions 1 h apart.
This process was implemented in the plant, and 6b was
obtained in 91% yield and 93:7 cis/trans ratio, with a somewhat
long filtration and drying time (1 day).
in the same solvent mixture (15 L of solvent for 1 kg of salt p).
The whole process, from rac-4 to purified salt p, was executed
on a plant scale with 10 kg of rac-4 and resulted in salt p with
g96% ee and 42% yield (this value corresponds to a 40% yield
(based on (2S)-4, 100% ee), which agreed well with the DSC-
based yield estimation of 35%). It was observed that the
composition of the mother liquors of the second treatment have
the same composition as the material isolated from the first
crystallization. Therefore by recrystallization from the CH3CN/
MeOH (9:1) mixture, an additional quantity of purified salt p
was obtained. The overall yield in terms of purified salt p was
44%.
Racemization of the undesired enantiomer (R)-4 was also
examined. While no success was obtained in the direct base-
mediated racemization of (R)-4, the corresponding methyl ester
was racemized after refluxing in CH3ONa/CH3OH for 8 h. After
ester saponification with NaOH, the resolution with (S)-phenyl
glycinol provided the salt p in overall yields of about 30% (from
recycled salt n to obtained salt p) and 96% ee.
Rearrangement to Proline Skeleton and Boc-Protection.
The transformation of bromolactone 6b to proline derivative 8
requires lactone methanolysis followed by displacement of the
halogen atom by the amine group. During early development,
various quantities of NaHCO3 and K2CO3 were investigated
under various conditions. The major issues were low conversion
of 6b to 8 and epimerization of the R-carbon (about 7%). [Since
the two stereocenters of 6b are not involved in the basic
rearrangement, their configuration should be preserved. Nev-
ertheless the acidity of the proton at the R-carbon might in
principle lead to partial epimerization, with the consequent
formation of diastereomers.] Two opposite results were obtained
combining different reagents and different temperatures: K2CO3
(2–3 equiv) and methanol at reflux resulted in complete
conversion but also epimerization, whereas NaHCO3 (2–3
equiv) in methanol at 25 °C resulted in no epimerization but
low conversion. One approach to obtain good conversion and
low epimerization would be to decrease the strength of the base
and increase the reaction temperature, or to reduce the stoichi-
ometry of the base. Reduction of base from 2 to 1 equiv would
hypothetically lead to a reaction mechanism in which the role
of the base consists only of deprotonation of amonium salt 6b.
The free amine would in turn displace the bromine atom to
form a bicyclic lactone, which would then be opened by
methanolysis. Our hypothesis was supported by the isolation
of bridged-amino lactone 13, after Boc-protection of the reaction
mixture.
Bromolactonization and Boc-Deprotection. The transfor-
mation of acid (S)-4 to proline 8 was accomplished by
cyclization to reactive halo lactone 5, followed by rearrangement
to proline 8. During these transformations a new stereocenter
is created but is of no consequence since it is destroyed after
oxidation to ketone 9. In the preliminary process studies, these
transformations had been previously performed with rac-4, and
thus preservation of the chirality from (S)-4 to 1 had to be
proven. In the first evaluation it was discovered that the
stereocenter R to the carboxyl was substantially preserved;
partial racemization was observed during the basic rearrange-
ment and oxidation step with Py.SO3 (the ee decreased about
12% over these two steps). As far as the bromo-lactonization
of 4 and the Boc-deprotection of 5 were concerned, no
racemization was observed. During the cyclization of 4, cis and
trans diasteroisomers were obtained, and therefore substrates
with potentially different reactivities were generated for the basic
rearrangement to 8. It was indeed found that the cis isomer
reacted more quickly and more cleanly than the trans isomer
(vide infra), and fortunately, lactonization was selective for cis-5
(cis/trans ratio about 9:1).
Halo-lactonization could be effected with I2 to give cis/trans-
5a or with NBS to give cis/trans-5b. NBS was chosen as the
preferential reagent because (1) only 1.1 equiv was needed,
compared to 3 equiv of I2 (greater atom economy); (2)
consequently, less sulfite would be required for reduction of
the unreacted reagent; and (3) NBS is cheaper than I2.
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
327

Table 3. Optimization of the bromo-lactonizationa
chiral HPLC (A%)
cis/trans(A%
entry
conditions
yield (%)
HPLC)
(3S,5S)-cis-5b (3 S,5R)-trans-5d (3R,5R)-cis-5b (3R,5S)-trans-5d
1
2
3
4
salt pb; -20 °C; no stripping
salt pb; -20 °C; stripping
salt pb ; 30 °C; no stripping
salt pb ; 30 °C; stripping
98
98
96:4
96:4
91:9
92:8
94.3
94.4
89.0
90.2
4.3
3.9
8.7
8.2
1.4
1.7
1.9
1.6
ndc
nd
nd
nd
94
a No epimerization was observed in all cases. b Diastereomeric ratio ) 97.8:2.2. c nd ) not detected.
Table 4. Optimization of the concentration and quantity of
HCl
Scheme 5. Reaction of trans-(3S,5R)-6d with 1 equiv of
NaHCO3
yield
quality
HPLC
entry (M) (equiv) (mol %) (mol %) purity (A%) cis/trans
5b HCl(g)
5
6
6
1
2
3
4
0.5
0.6
0.5 2 + 2
0.6 2 + 2
2
2
21
17
73
99.5
99.5
99.5
99.1
90:10
91:9
91:9
91:9
78
93.6
89.4
(NaHCO3 in MeOH) was not stirred.] The concentration of
cyclic intermediate 14 reached a maximum early in the reaction
and evolved to product, which supported the proposed mech-
anism. After completion, the reaction mixture was treated with
NEt3 and Boc2O. HPLC analysis then showed that epimerization
had not occurred.
The same experiment was performed with (3S,5R)-6d
(containing about 1% of diastereomer (3R,5R)-6c). In this
experiment a certain degree of epimerization of the starting
material was noted, as the detected amounts of (3R,5R)-7-d4
and 14 were greater than expected. [The expected quantity of
14 was about 0.04%, which is 40% of the starting (3R,5R)-6c
(1%).]
Since the conversion was still not complete after 36 h (an
indication that the trans isomer reacted more slowly than the
cis), the reaction was continued for a total of 52 h. At this point,
NMR analysis showed almost complete disappearrance of all
starting material and a mixture of two products in 73:27 ratio.
The minor product was expected to be the diastereomer of 7a-
d4, namely, 7-d4. In order to unambiguously identify the NMR
signals of these two products, a sample of 8 with a cis/trans
ratio of 87:13 was treated with HCl in EtOAc to give the
corresponding diastereomeric mixture of 7 and 7a. This allowed
interpretation of the secondary product present after 52 h, which
was confirmed to be (2R,4R)-7-d4. The end-reaction mixture
was then treated with Boc2O, and HPLC analysis revealed a
To verify the hypothesis, an experiment was performed with
only 1 equiv of NaHCO3 in refluxing methanol. Thus, pure
enatiomer (3R,5R)-6c [because of the availability of pure
intermediates in early studies, the undesired enantiomer was
used for this evaluation] and 1 equiv of NaHCO3 were mixed
with deuterated methanol in a NMR tube, and reaction progress
was followed by 1H NMR. Graphically, the results of the first
experiment showed a decreasing amount of 6c and an increasing
amount of transient intermediate 14 (as observed by Kazmaier
and co-workers12), which reached a maximum after about 6 h
of heating. [The bicycle is easily identified from the very small
coupling constants of the two geminal hydrogens with the
bridgehead hydrogen, which is about 1 Hz for the exoproton
and nearly zero (not measurable) for the endoproton.]
The NMR data further showed complete disappearance of
the cis isomer after 36 h, and no epimerization since only 7-d4
was formed. [It must be considered that the reaction rate
observed in the NMR experiments could be less than the rate
observed for a standard reaction as the result of an isotope effect,
as deuterated methanol takes part in the reaction as a reagent,
and also because of the fact that the heterogeneous mixture
Figure 4. Cyclic intermediate 13 isolated after Boc-protection
of proline derivative (2R,4R)-7.
(12) Mues, H.; Kazmaier, U. Synthesis 2001, 487–498.
328
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development

Chart 2. NMR study of the rearrangement of cis-6ca
Scheme 6. By-product 15 found during the synthesis of 8;
15 could be converted to amine 16 and then extracted by
acidic water
Table 5. Exploration of alternate oxidation reagentsa
time
racemization
(%)
a Cyclic intermediate 14 was formed early in the reaction, and subsequently
transformed into 7-d4.
entry
oxidant
(h) conversionb
1
2
3
4
5
6
Pyr ·SO3/DMSOc
3
20
2
complete
10
40
33
n.a.d
9
PCC
complete
Chart 3. NMR study of the rearrangement of trans-6d
NMO/TPAP
incomplete
incomplete
incomplete
complete
NMO/RuCl2(PPh3)3
7% NaOCl/AcOH
14% NaOCl/AcOH
2
20
20
11
a The NaOCl/AcOH conditions produced as much racemization as the original
Pyr·SO3/DMSO conditions but reduced the hazard/pollution issues related and the
costs. b Disappearance of (2S)-8 (TLC). c Original conditions. d Coeluted with an
uncharacterized impurity during HPLC analysis.
(LC-MS analysis performed on the aqueous phase after
purification showed MW ) 442, giving evidence of the
proposed structure).
A comparison between the early and optimized processes
showed that the yield from 4 to 8 was improved from 33% to
about 73%, which allowed for the conversion of 35 kg of (2S)-4
into 29 kg of 8.
Oxidation. The initial procedure for oxidation of racemic
alcohol 8 to ketone 9, using Pyr ·SO3/DMSO, was performed
successfully on the racemic series in reasonable yields (80%
after chromatography). When (2S)-8 was oxidized under these
conditions, about 10% racemization at the R-carbon occurred.
With these conditions, there were also concerns about waste
streams (Me2S) and cost of Pyr ·SO3.
Of the alternate oxidation conditions explored (Table 5), it
was discovered that NaOCl/CH3CO2H resulted in a similar
extent of racemization but satisfied the waste stream and cost
issues. Milder conditions with NaOCl were permitted by
inclusion of 2,2,6,6-tetramethylpiperidine 1-oxyl (TEMPO).13
The NaOCl/TEMPO conditions were then optimized to
minimize racemization, increase yields, and establish a safe and
cis/trans ratio of 26:74, confirming the epimerization observed
by NMR analysis.
Before scaling-up the new conditions, alternate bases were
briefly investigated. Epimerization was not observed with 1
equiv of NEt3 or K2CO3, but some epimerization was observed
when g1.5 equiv of base was employed. Consequently, the
optimized process was defined as the use of no more than 1.2
equiv of NaHCO3 and a reaction time of g17 h at 65 °C and
ambient pressure.
During scale-up, an impurity was observed in Boc-protected
proline 8. On the basis of its molecular weight (LC-MS), it
was inferred to be dimer 15, formally resulting from the addition
of 6b to 8. This impurity suggested that concentration may be
a critical factor. We hypothesised that the bromine atom might
be substituted by a nucleophile. We therefore envisaged a
method to purge the impurity, provided that it would be
transformed into a basic species, which can be then extracted
by acidic water. Treatment of the 8/15 mixture with 1,2-
ethylenediamine produced a mixture of 8 and 16, from which
16 was selectively washed out with HCl(aq) to afford pure 8
(13) Lucio Anelli, P.; Biffi, C.; Montanari, F.; Quici, S. J. Org. Chem.
1987, 52, 2559–2562.
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
329

Table 6. Optimization of the oxidation conditions
a
entry NaOCl (equiv) TEMPO (mol %) KBr (mol %)
NaHCO3
purity of 9 (A%)
assay of 9 (%) yield of 9 (%)
1
2
3
4
5
6
7
2.8
2.8
2.8
1.1
1.1
1.1
1.1
10.4
10.4
13.7
8% aq in NaOCl 63.0
8% aq in NaOCl incomplete conversion
8% aq in NaOCl
80.0
84.0
13.7
1.0
1.0
3.3
1.0
1.0
1.0
1.0
1.0
solid in NaOCl
b
solid in NaOCl
solid in CH2Cl2
96.9
96.2
99.2
97.0
93.9
88.6
96.8
99.2
90.0
85.0
98.0
c
104.0
a Unless otherwise specified, the amount of NaHCO3 was adjusted to maintain pH 8. b No reaction was observed until NaHCO3 was added (after 24 h). c The quantity of
NaHCO3 was optimized to 0.7 mol % relative to NaOCl.
environmentally friendly process (Table 6). The influence of
some variables, such as the concentration of NaOCl and the
ratio of the reagents of the catalytic cycle described by
Montanari and co-workers,13 were screened. NaOCl can be used
in almost equimolar ratio to 8 (Table 6, entry 4) regardless of
its concentration (5% and 13% NaOCl solutions), while
TEMPO and KBr can be dosed in catalytic amounts (1 mol %;
Table 6, entry 4). NaHCO3 was used to buffer the reaction
mixture at pH 8; without the pH buffer control, oxidation was
not observed and degradation of 8 resulted (24 h). The reaction
began when NaHCO3 was added, and high conversion was
achieved after 24 h (Table 6, entry 5). Under the optimized
conditions, NaHCO3 was added to a CH2Cl2 solution of 8,
TEMPO, and KBr, which was then charged with a 5% NaOCl
solution. The conversion was complete within 1 h, and after
workup, a quantitative yield of 9 was obtained with complete
retention of chirality. After the quench of residual NaOCl with
a 5% Na2S2O5 wash, the oily product could be used directly in
the next step. The oxidation was scaled up to 12.9 kg of 8 in a
250-L reactor, which afforded 9 in an average yield of 95.5%
and 100% ee.
DAST and DF possess similar reactivity towards ketones,
but DF is more thermally stable than DAST,19,20 and this was
particularly relevant since the reaction required high tempera-
ture. To determine the maximum reaction temperature that could
be safely used, a preliminary hazard study was performed. To
determine whether the thermal history of a sample of DF could
influence its stability or not, samples of DF in toluene were
aged for 8 h at different temperatures (DSC analyses, from 40
to 90 °C), and then the samples were heated to 300 at 10 °C/
min. In all experiments degradation began around 120 °C, but
the energy released by degradation slightly decreased as the
aging temperature increased. The decomposition that occurred
during the aging at 80 and 90 °C was approximately 7% and
16% less than the expected, which indicated partial degradation
during the isothermal aging. Further experiments with a Calvet
calorimeter (C80) showed the onset of the decomposition to
be around 70 °C, a value much lower than that detected by
DSC (120 °C). Nevertheless the decomposition at 70 °C was
negligible and became significant only around 100 °C (5 bar).
This study indicated that, for safety reasons, the maximum
reaction temperature should not exceed 70–80 °C.
Fluorination. gem-Difluoro derivative 10 could be obtained
by nucleophilic fluorination of ketone 9. In principle, a variety
of fluorinating reagents, such as SF4, DAST, Deoxo-Fluor (DF),
2,2-difluoro-1.3-dimethylimidazolidine14,15 (DFI), and N,N-
diethyl-R,R-difluoro(m-methylbenzil)amine16 (DFMBA), could
effect this transformation. However, only DAST and DF could
be used in our plant because of the toxicity of SF4 and the
dehydrating characteristics of DFMBA and DFI,17 which would
potentially increase the quantity of by-product such as fluoro-
vinyl derivative 17. Recently Hu and co-workers18 have
demonstrated the possibility to fluorinate the ketone 9 changing
the Boc-protective group with the benzyl group, and substituting
DF with SF4, obtaining benzyl-protected 10 in higer hields.
In the first attempt to scale up the fluorination process, DF
in 1,2-dichloroethane (DCE) was used in a glass reactor at 55
°C and gave 10 in 46% yield after chromatography. Although
this process was acceptable at this stage, several major issues
surfaced: (1) long reaction time (4 days), (2) requisite large
quantity of DF (4.4 equiv), (3) use of toxic DCE, (4) low
space-time yields, (5) extensive formation of by-product
(especially 17 and 19), and (6) requirement of chromatographic
purification, poorly effective for removal of fluoro-vinyl im-
purity 17.
The immediate goal with these fluorination conditions was
to reduce the quantity of DF and to increase the reaction rate
(the use of catalysts such as BF3 ·Et2O, EtOH, and CF3COOH
did not improve the conversion). It was subsequently found that
the reaction rate could be increased by an increase in reaction
temperature (70 °C), while the quantity of DF could be
(14) Hayashi, H.; Sonoda, H.; Fukumura, K.; Nagata, T. Chem. Commun.
2002, 1618–1619.
(15) Sonoda, H. Fain Kemikaru 2002, 31, 5–12.
(16) Kobayashi, S.; Yoneda, A.; Fukuhara, T.; Hara, S. Tetrahedron 2004,
60, 6923–6930.
(17) Umetani, H.; Sonoda, H.; Komatsu, H. Nucleosides Nucleotides Nucleic
Acids 2003, 22, 667–669.
(19) Lal, G. S.; Pez, G. P.; Pesaresi, R. J.; Prozonic, F. M.; Cheng, H. J.
Org. Chem. 1999, 64, 7048–7054.
(18) Hu, S.; Martinez, C. A.; Kline, B.; Yazbeck, D.; Tao, J.; Kucera, D. J.
Org Process Res. DeV. 2006, 10, 650–654.
(20) Messina, P. A.; Mange, K. C.; Middleton, W. J. J. Fluorine Chem.
1989, 42, 137–143.
330
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development

result of steric and electronic effects, such as ꢀ-elimination to
fluoro-vinyl 17. Differential reactivities of the stereoisomers of
20 explain that the maximum observed yield of 10 observed
was about 45%, with about 40% of unreacted 9 and about
10–12% of 17.
To improve the selectivity of the fluorination and to simplify
the purification, the effects of solvents, additives, and catalysts
were screened. One goal was to explore whether the polarity
of the medium could stabilize intermediate 2022 and prevent
formation of 17 via ꢀ-elimination. Also, the presence of catalysts
or additives could conceivably promote formation of 20 or
convert by-product 17 via addition of a fluorine atom, both of
which would lead to higher yields of 10.25,26 The first series of
experiments employed H2O, NaHCO3, spray-dried KF (KFsd),
amorphous KF (KF98%), and Aliquat 336 (PTC) as catalyst and
additives. [Other reagents used in the full factorial experimental
design were AlCl3, 18-crown-6, and TBAF. These reagents led
to partial or complete degradation of ketone 9. KF can be found
commercially spray-dried form (KFsd) or as an amorphous form
(KF98%). In principle these forms could display different
reactivities, because of their different solubility and adsorption
properties. Nevertheless, KFsd is more expensive and less
available than KF98%. Aliquat 336 (trioctylmethylammonium
chloride), a phase transfer catalyst, and 18-crown-6 were
envisioned to promote addition of F- to intermediate 20.] For
this evaluation, the following variables and ranges were selected:
DF from 1.6 to 3 mol/mol of 9; DCE from 0 to 0.68 mL/mmol
of 9; H2O from 0.1 to 0.3 equiv; KFsd from 1 to 2 equiv, and
KF98% from 1 to 2 equiv. [The quantity of DF exceeds the
minimum quantity actually needed to convert the ketone 9 (1.3
mol/mol of 9, as elsewhere stated in this article). The rationale
is to supply that part of DF destroyed by the addition of catalysts
and additives, which can be reactive or contain appreciable
quantity of water.] To prevent extensive degradation of 9 and
10 when KF is used, NaHCO3 must be included to reduce the
acidity of the reaction medium. The screening set of 18
experiments (fractional factorial design resolution 5, 2 central
points, no replications) was performed with an automated
multiple reactor and the model validity was ascertained by
statistical analysis of the results (ANOVA, R2, Q2, etc.).
The fitted model indicated that H2O could range from 10%
to 30% without any influence on the yield of 10. Likewise, the
quantity and quality of KF and the quantity of DCE had no
influence on the results. The negative effect of decreasing DF
is understood by considering that when water and/or KF are
set at their upper value and DF at its lower one, most of the
DF reacts with water and/or additives instead of with 9, resulting
in low conversions to 10.
Figure 5. Determination of the maximum safe fluorination
temperature with DF. The DSC analyses established the
decomposition onset at about 120 °C (black curve, graph a),
while the C80 calorimeter detected the onset at about 75 °C
(graph b). Nevertheless, the decomposition was modest and slow
(blue curve, graph b), as shown by the slight pressure increase
(red curve, graph b).
Figure 6. The major by-products formed during the fluorina-
tion of 9 with DF. One characteristic of DF is the tendency to
form compounds such as 19. These by-product are UV–visible
and possess a molar extinction coefficient greater than that of
10 at 210 nm.
Scheme 7. Reaction mechanism proposed by Middleton22
minimized to 1.3 equiv. With these optimized conditions, the
reaction reached the maximum yield (40–45%) in only 8 h.
Although not supported by conclusive experimental proof,
the following fluorination mechanism has been postulated: trace
moisture reacts with DF to provide HF, which then adds to the
carbonyl bond to form a fluorohydrin. This reacts with DF,
generating HF and 20. Further reaction of 20 with DF generates
10, while creating additional HF catalyst.21–24
(21) Singh, R. P.; Shreeve, J. Synthesis 2002, 2561–2578.
(22) Middleton, W. J. J. Org. Chem. 1975, 40, 574–578.
(23) Milos, H. In Organic Reactions; Milos, H., Ed.; Wiley: New York,
1988; Vol. 35, pp 513–637..
(24) Yin, J.; Zarkowsky, D. S.; Thomas, D. W.; Zhao, M. M.; Huffman,
M. A. Org. Lett. 2004, 6, 1465–1468.
(25) Feiring, A. E. J. Org. Chem. 1980, 45, 1958–1961.
(26) Mase, T.; Houpis, I. N.; Akao, A.; Dorziotis, I.; Emerson, K.; Hoang,
T.; Iida, T.; Itoh, T.; Kamei, K.; Kato, S.; Kato, Y.; Kawasaki, M.;
Lang, F.; Lee, J.; Lynch, J.; Maligres, P.; Molina, A.; Nemoto, T.;
Okada, S.; Reamer, R.; Song, J. Z.; Tschaen, D.; Wada, T.; Zewge,
D.; Volante, R. P.; Reider, P. J.; Tomimoto, K. J. Org. Chem. 2001,
66, 6775–6786.
As DF reacts with the ketone 9, two diastereomeric monof-
luoroderivatives (i.e., 20) are formed. Different reaction path-
ways from each diastereomeric intermediate are possible as a
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
331

Greater quantities of SiO2 produced a negative effect, due to
the fast reaction between DF and SiO2, and resulted in
decomposition of the DF. Also, in this case small quantity of
water improved the reaction performance. Thus, 5 wt % SiO2
(with respect to 9) was included in the fluorination conditions.
This modified fluorination process was successfully imple-
mented in a 35-L Halar-coated reactor and provided yields
consistent with the laboratory trials (∼47% yield of 10, 100%
ee, and 0.1 mol % of residual 17 after the KF treatment). Crude
10 prepared in this manner could be purified by chromatography
or, preferably, by chemical purification (vide infra).
Figure 7. Effect of variables in the fluorination of 9. The effect
is negligible when the value is around zero or when the error
bar crosses the zero line. In this case only the quantity of DF
had an influence on the system.
Ester Hydrolysis. Partial racemization was observed in early
studies of the basic hydrolysis of ester 10 with aqueous
hydroxide and MeOH. Another issue observed was the low
reaction rate (several days), presumably due to the lipophilic
character of 10. Development of this hydrolysis was aimed at
establishing an acceptable balance between racemization and
reaction rate, and two main variables were examined: the
composition of the aqueous alkali solution and the quantity of
methanol used as the cosolvent.
Analysis of the experimental data (Table 7) shows that the
KF treatment did not react with 10 and only slightly decom-
posed 9 but was successful in the decomposition of 17 when
DF was at its upper level of 3 equiv (Table 7, entries 2, 4, 8,
10, 12, 14). This finding was unexpected; the aim was to
transform 17 into 10 or avoid its formation. Regardless, an
effective method for elimination of 17 was discovered, and the
goal to simplify the chromatographic purification of 10 was
achieved.
It was found that a high concentration of NaOH increased
the reaction rate, but racemization was problematic (Table 8,
entry 2 vs entry 1). Substitution of NaOH with LiOH, at the
same concentration, depressed the reaction rate (Table 8, entry
3). The same negative effect was observed by a decrease in
the quantity of NaOH (Table 8, entries 4 and 5). It was
demonstrated that methanol was necessary for the reaction to
proceed at a significant rate (Table 8, entry 6), although the
volume could be reduced without affecting the reaction rate
(Table 8, entries 7 and 8). The conditions in entry 5 (Table 8)
were chosen for the production campaign, since a minimum
quantity of base was used. Although the yield of these chosen
conditions was lower than other conditions in Table 9, the yield
increased up to 90% (from 10) upon scale, due to easier
manipulation of the solids on scale.
Purification of gem-Difluoro-ester 10 and gem-Difluoro-
acid 1. The use of preparative MPLC (medium pressure liquid
chromatography, stainless steel columns, 30 cm internal diam-
eter, 1.5 m high, with a hydraulic-driven piston for compression
of the silica gel (4 bar)) is demanding for production applications
in terms of cycle time, use of solvents/silica, and waste
generation Nevertheless, MPLC was used for the first scaled
batches since extractions did not remove residual 9, 18, and
the DF-related by-product.
During the saponification studies, a structure-dependent
hydrolysis rate was observed; unreacted ketone 9 hydrolyzed
much faster than 10 (a few hours for 9 vs 4 days for 10).
Moreover, it was discovered that NaClO decolorized the target
product 1 and decomposed fluoro-vinyl 17 into water-soluble
compounds (the double bond fluoro-vinyl 17 could be decom-
posed oxidatively by NaClO, forming polar water-soluble
products). These two findings paved the way to a chemical
purification of 10 and 1. The goal was not to decompose 17
with KF but instead eliminate unreacted 9 immediately after
the fluorination by means of selective hydrolysis with NaOH
in H2O/n-hexane. Keto-acid 21 was successively extracted into
H2O, while the other compounds remained in the n-hexane
phase. After solvent exchange and hydrolysis with NaOH/
The best conditions found were DF 3 mol/mol of 9; H2O
0.1 mol/mol of 9, NaHCO3 0.2 equiv of 9; KF98% 1 equiv of 9
and PTC 0.1 mol/mol of 9, 70 °C, 5 h for the fluorination, plus
other 5 h for the decomposition. The results were (averange of
4 trials) the following: 10 about 40 mol %, the formed fluoro-
vinyl 17 about 9 mol %, and the residual 9 about 37%. After
decomposition with KF 9 and 10 remained unaffected (31 mol
%, and 40 mol %, respectively), while 17 disappeared.
Accordingly to the best laboratory conditions, the process
was scaled up in a 35-L Halar-coated reactor (Halar is a
polifluorinated polymer resistant to HF). Unexpectedly, after
5 h of reaction, only 10% conversion to 10 had occurred. After
an additional day of reaction and several doses of water, the
conversion to 10 reached only 30%.
The root cause investigation revealed that the reactor material
indeed influences the extent of fluorination. In particular, all of
the experiments conducted in glass reactors were faster and
performed better than those done in Halar or Teflon materials.
This observation was contrary to the expectation that the
inertness of Halar would allow for better performance because
prevented reaction of DF with the reactor lining.
Since pitting of glass reactors occurred in the laboratory
experiments, it was surmised that reaction of the silica glass
with DF or generated HF produced a compound that had a
positive effect on the fluorination. Reaction of a strong
fluorinating reagent with silicates or SiO2 would be expected
to produce a fluorosilicate. Indeed, Padma and co-workers
described the preparation of pyridinium hexafluorosilicate from
SiO2 and polypyridinium hydrofluoride (PPHF).27,28 To confirm
this hypothesis, the fluorination was tested in the presence of
SiO2 (1.5-10 wt % with respect to 9) in a Teflon-lined
multireactor system (a full factorial experimental design of 52
experiments was employed, using an Argonaut Quest 210
system, equipped with 20 10-mL Teflon reactors). It was found
that small amounts of SiO2 did promote the fluorination of 9.
(27) Radhamani, K. N.; Padma, D. K. J. Fluorine Chem. 1993, 64, 95–99.
(28) Kalbandkeri, R. G.; Mohamed, K. S.; Padma, D. K.; Murthy, A. R. V.
Polyhedron 1985, 4, 787–789.
332
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development

Table 7. Raw data from the fractional factorial experimental designa
a The column “addition order” indicates the operation order required for the best results: NaHCO3, KF, and PTC are added to the reaction mixture after 5 h of fluorination.
Addition of KF and PTC in the fluorination step led to complete decomposition of 9. In the last two columns, the results after fluorination (5 h, 70 °C) and selective
decomposition of residual impurities (5 h, 70 °C), respectively, are collated. While the mass balance for the fluorination was near 100% by HPLC assay, the mass balance of
the decomposition phase decreased as a result of the conversion of fluorovinyl 17 into more polar by-products that were extracted with water. Entry 10 shows the best
operating conditions.
MeOH (according to Table 9, entry 5), 18 and the DF-related
by-product were extracted into toluene, while product 1 and
fluoro-vinyl impurity 22 remained in the aqueous phase, which
in turn was treated with NaClO. After oxidation of the fluoro-
vinyl 17 impurity, final product 1 was isolated by filtration from
water, after adjustment to pH 2.5, in the presence of n-hexane
to avoid gumming. This new procedure was implemented in
production; a significant contribution to simplification of the
overall process. The overall yields from 9 to 1 ranged from
39% to 42%, and the quality of 1 was high (HPLC assay
94–99%, 100% ee).
DF at 70 °C for 4 days); and the necessity of chromatographic
purification of 10. Moreover, the overall yield was very low
(1.5%), despite careful manipulation of materials (estimated at
about 7 t/kg of 1). A new synthetic strategy, based on an early
classical optical resolution, was evaluated and enabled. The early
resolution halved the material to be processed in the later steps
of the route. Toxic and hazardous reagents (ZnCl2, MeI,
Pyr·SO3/DMSO) were avoided wherever possible, and in the
case of DF, safer and faster conditions were established (the
reaction time was decreased from 4 days to 5–10 h, thus
reducing the risk associated with the heating a thermally
sensitive reagent). The chromatographic purification of the crude
10 was substituted with a practical chemical purification, which
was based upon different reaction rates of 9 with respect to 10,
17, and 18, during the final step, and the ability of NaClO to
selectively decompose impurity 22 into water-soluble by-
product. MTBE was eliminated from the process and was
substituted with toluene for reactions and with water for the
Conclusion
The original process for the production of 1 consisted of
nine steps, which included an enzymatic resolution as the last
transformation. The main drawbacks of this route were the late
optical resolution; the use of toxic and hazardous reagents,
solvents, and conditions (ZnCl2, CH3I, Pyr ·SO3/DMSO, MTBE,
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
333

Table 8. Screening of base, concentration, and quantity of
methanol for the hydrolysis of 10
aqueous
base
base
MeOH
time
(h)
yield
(%)
entry
(equiv) (mL/g 10)
1
2
3
4
5
6
7
8
8% NaOH
20% NaOH
8% LiOH
8% NaOH
8% NaOH
8% NaOH
8% NaOH
8% NaOH
6
6
6
2
4
6
6
6
17
17
17
17
17
0
96
24
96
96
96
96
96
96
100
100a
<50
46
87
11
100
100
4
10
a 7% Racemization occurred.
Table 9. Summary of the manufacturing campaign of 1
reactor
batch
av batch volume no. of quantity
product
size (kg) size (kg)
(L)
batches
(kg)
Figure 8. Diagram for the chemical purification of 10 and 1.
Decomposition of 17 could be postponed until after the
saponification of 10 and 17 to 1 and 22, respectively, and was
thus performed with NaClO instead of KF and PTC. The main
advantage of such a modification is the reduction of residence
time in the Halar reactor; as a result of the small reactor volume
(maximum 35 L), 10 fluorination batches were required to
convert all of ketone 9 to 10. While the decomposition of 17
with KF must be done in the Halar reactor, decomposition of
22 with NaClO could be accomplished in larger glass or glass-
lined reactors. The latter option permitted the combination of
several batches, provided that the residual DF was quenched
with aqueous NH4OH before the NaClO decomposition process.
This simplified procedure allowed isolation of pure 1 without
chromatographic purification and with no decrease in yield.
2
3
4
n.a.
30
13.5
10
14
27
17.5
12.9
3.5
2.7
n.a.
20.6
11.7
8.7
12.4
13.5
17.5
9.7
n.a.
250
170a
250
250
250
250
250
35b
n.a.
5
103
129
87
11
8
salt p
62
salt p purified
5
54
6
4
35
8
2
29
9
3
26
10
1
2.6
2.2
10
5
11
100
7.5
a Cryogenic reactor. b Halar-coated reactor.
crystallization of 4 and 1. It was discovered that the Ireland-
Claisen rearrangement of 3 to 4 could be efficiently executed
in the absence of Lewis acid. The stereochemical pathway for
the basic rearrengement of 6b was elucidated, which led to
control of the configuration of the R-carbon. A safe, environ-
ment-friendly, and reliable oxidation of 8 to 9 with NaOCl/
TEMPO was employed. Finally it was discovered that silica
gel promoted the DF-mediated fluorination of ketone 9.
These improvements resulted in a reliable and safe process.
The overall yield from 2 to 1 increased about 300% (from 1.5%
to 4.5%), while the major improvements were obtained for the
transformation of 4 to 8 (from 33% to 73%) and for the
oxidation of 8 to 9 (from 64% to 95%). These yield improve-
ments resulted in the reduction of all materials used in the
process from 7 t/kg of 1 to less than 2 t/kg of 1.
in ppm based on tetramethylsilane (δ ) 0 ppm), coupling
constants (J) (H, H) are given in Hz, spectral splitting pattern
are designated as singlet (s), doublet (d), triplet (t), quadruplet
(q), multiplet or overlapping signals (m), broad signal (br),
solvent is given in parentheses.
13C NMR. Varian 100 or 125 MHz spectrometer chemical
shifts (δ values) are reported in ppm based on tetramethylsilane
(δ ) 0 ppm), all spectra are proton broadband decoupled,
multiplicities of the signals were determined by DEPT measure-
ments, solvent is given in parentheses.
MS: LC-MS. HPLC-IONTRAP-MS: Agilent 1100 Series
LC/MSD Trap VL.
ATR FTIR in Situ Spectroscopy. ReactIR 4000, Mettler
Toledo, equipped with chemometric software, ConcIR from
Mettler Toledo.
FC. Silica gel 60 (Fluka, 40–63 µm), air or mechanic
pressure of ca. 0.2-3 bar, eluents are given in parentheses.
pH measurements. The solutions were pH tested at 25 °C
with a digital Metrohm 632 pH meter that had been referenced
to pH 7.00 and 4.00 with buffer standard solutions (CER).
Techniques. Reactions involving air- or moisture-sensitive
reagents or intermediates were performed under nitrogen.
The successful scaling of this process in the plant gave 7.5
kg of 1. The campaign was conducted with glass-lined 250-L,
cryogenic 170-L, and Halar-coated 35-L reactors (Table 9). The
number of batches per single step ranged from 3 to 11, while
the entire campaign lasted about 6 months.
Experimental Section
Instrumentation and Materials. 1H NMR. Varian 400 or
500 MHz spectrometer chemical shifts (δ values) are reported
334
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development

SolVents. Toluene, MeOH, THF, CH2Cl2, EtOH, n-hexane,
methyl alcohol, EtOAc, and 1,2-dichloroethane [CAUTION:
carcinogenic substance. R45, 11, 22, 36/37/38; S53, 45] were
purchased in pure grade quality (>99%) from Carlo Erba
Reagenti.
4 nm bandwidth. Phase A: n-heptane/EtOH 94:6. Elution:
isocratic n-heptane/EtOH (94:6) for 60 min.
HPLC Method 4. Used for optical purity of 6b. Diacel
Chiracel OJ-RH 5 µm column (4.6 × 250 mm), 40 °C chamber,
0.5 mL/min flow rate, 5 µL injection vol, 205 nm detection, 4
nm bandwidth. Phase A: n-heptane. Phase B: 2-propanol.
Elution: isocratic 40% B for 130 min.
HPLC Method 5. Used for optical purity of 8. Diacel
Chiracel OJ-RH 5-µm column (4.6 × 250 mm), 25 °C chamber,
0.5 mL/min flow rate, 10 µL injection vol, 205 nm detection,
4 nm bandwidth. Phase A: n-heptane/EtOH 93:7. Elution:
isocratic 100% A for 35 min.
Preparation of tert-Butoxycarbonylamino-acetic Acid
3-Methyl-but-2-enyl Ester (3). EDC (715 g, 3.7 mol) was
suspended in CH2Cl2 (1.2 L), which was then followed by
addition of prenol (309 g, 3.6 mol) and DMAP (22.8 g, 0.19
mol). This suspension was stirred at 4-8 °C, and 2 (650 g,
3.71 mol) was then added over 2 h. The temperature was raised
to 20 °C, and the solution was then washed sequentially with
5% aqueous NaHCO3 (2 L), 0.5 N HCl (2 L), and H2O (2 L).
After distillation of the solvent, 841 g (95%) of 3 was obtained
as an oily residue: 94.3% HPLC purity (tR ) 12.5 min), 96.6%
HPLC assay, and 0.8% residual prenol (HPLC Method 1). Other
analytical data are in accordance with the literature.6
Reagents and Substrates. LDA was purchased from Aldrich
Chemical Co. or Chemetall as a 1.8 M THF solution; THF
(NMT 0.05% H2O). NaOH, H2SO4, TMSCl, NaHCO3, Na2CO3
and K2CO3, and silica gel 60 230–400 mesh were purchased
in RPE quality from Carlo Erba Reagenti. TEA, TBD, NaOMe
and Boc2O were purchased from Aldrich Chemical Co. Deoxo-
Fluor [CAUTION: this reactive is potentially explosive, ex-
tremely flammable, reacts violently with water, is toxic if
swallowed and causes severe burns. R35, 25, 2, 14, 11, 20; S9,
26, 33, 35, 36/37/39, 45] was purchased from Air Products
(http://www.airproducts.com) as a 50 wt % solution in toluene,
was stored at 4 °C, and was handled in dry conditions under
nitrogen. Two types of KF were used: spray-dried material of
99% purity (abbreviated KFsd) and nonspray-dried material of
98% purity (abbreviated KF98%) [CAUTION: mutagenic and
teratogen substance, if decomposes may generate HF. Avoid
strong acids. R23/24/25; S26, 45]. Aliquat 336 was purchased
from Fluka, silica gel 60 (230–400 mesh) was purchased from
CER, and all other reagents were obtained from Aldrich
Chemical Co.
Preparation of (()-2-(2,2-Dimethyl-propionylamino)-3,3-
dimethyl-pent-4-enoic Acid (rac-4). Compound 3 (50 g, 205.5
mmol) was dissolved in THF (170 mL) and was then cooled
to -65 ( 5 °C under N2. LDA (281 mL of a 1.8 M solution
in THF, 514 mmol) was added to the solution of 3, while
maintaining the reaction temperature at -65 ( 5 °C (30 min
required). After reacting for 1 h at -65 ( 5 °C, the mixture
was warmed to 20 °C over about 20 min and was then
maintained at 20 °C for about 1 h. Toluene (50 mL) was added,
and the mixture was then cooled to 0 °C. The pH was adjusted
to 2–3 with 10% aqueous H2SO4, and the phases were then
separated. rac-4 was extracted from the organic phase with 1
M NaOH (190 mL). n-Hexane (50 mL) was added to the
resulting aqueous phase, and the mixture was stirred vigorously
for 15 min and then cooled to -10 °C. The pH was adjusted
to 2–3 with 20% aqueous H2SO4. Precipitated rac-4 was filtered,
washed with n-hexane, and then dried under house vacuum at
35–40 °C for 15 h to give 37 g (72%) of rac-4 as a white solid:
93.7% HPLC purity (tR ) 9.2 min); 94.5% HPLC assay (HPLC
Method 1); mp ) 109.5 °C. Other analytical data are in
accordance with the literature.6
Statistical Analysis. Data was evaluated with MODDE,
7.0.0.1 software (Umetrics AB, Sweden) and Microsoft Excel
9.0 sofware.
Parallel Multireactor. Automated HEL Duet 317, manual
Argonaut Quest 210.
Differential Scanning Calorimetry (DSC). Mettler Toledo
DSc822 plus sample robot TSO801RO.
CalVet Calorimeter. Setaram C80D.
HPLC Analytical Methods. The following HPLC methods
were developed for in-process control (ICP) and determination
of chemical and optical purity of each product, intermediate,
and by-product. Analysis of prepared compounds was per-
formed on an Agilent HP1100 system equipped with a
quaternary pump and a diode array detector (UV-PDA cell 10-
mm path).
HPLC Method 1. Used for chemical purity and ICP of 1, 3,
4, 5b, 6b, 8, 9, and 10. Merck Symmetry C18 5 µm column
(4.6 × 250 mm); 27 °C chamber, 1.0 mL/min flow rate, 5 µL
injection vol, 210 nm detection, 4 nm bandwidth. Phase A: 25
nM KH2PO4 adjusted to pH 2.5. Phase B: CH3CN. Elution:
isocratic 45% B from 0 to 4 min; linear gradient from 45% to
70% B from 4 to 15 min; isocratic 70% B from 15 to 25 min;
linear gradient from 70% to 45% B from 25 to 30 min.
HPLC Method 2. Used for optical purity of 1, 4, salt p, 9,
and 10. Chiralcel OJ-RH 5 µm column (0.46 × 15 cm) fitted
with a Chiralcel OJ-RH precolumn (0.4 × 1 cm), 30 °C
chamber, 0.6 mL/min flow rate, 10 µL injection vol, 205 nm
detection, 4 nm bandwidth. Phase A: H2O + 0.1% CF3CO2H.
Phase B: CH3CN+ 0.1% CF3CO2H. Elution: isocratic 75% A,
25% B.
Isolation of 2-(3,3-Diisopropyl-ureido)-3,3-dimethyl-pent-
4-enoic Acid (11). This by-product of rac-4 was isolated from
the organic fraction and was purified by flash chromatography
on silica gel (hexane/EtOAc 1:1, Rf ) 0.3): tR ) 9.3 min (HPLC
1
Method 1); H NMR (500 MHz, DMSO-d6) δ 1.09 (s, 6H,
(CH3)2-C-CH2), 1.18 (d, J ) 2.7 Hz, 6H, (CH3)2-CH-N), 3.90
(m, J ) 6.9 Hz, 2H, (CH3)2-CH), 4.12 (d, J ) 8.8 Hz, 1H,
NH-CH), 4.87 (d, J ) 8.8 Hz, 1H, NH-CO), 5.08 (dd, J )
17.5, 10.8 Hz, 2H, CH2dCH), 5.92 (dd, J ) 10.7, 6.7 Hz, 1H,
CH2dCH), 12.6 (s, 1H, COOH). MS (ion trap): [C14H26N2O3
+ H]+ 271.2, [C14H26N2O3 + Na]+ 563.3.
HPLC Method 3. Used for optical purity of 5b. Diacel
Chiracel OJ-RH 5 µm column (4.6 × 250 mm), 35 °C chamber,
0.5 mL/min flow rate, 10 µL injection vol, 205 nm detection,
Resolution of (()-2-(2,2-Dimethyl-propionylamino)-3,3-
dimethyl-pent-4-enoic Acid (rac-4) with (+)-(S)-Phenylgly-
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
335

cinol. rac-4 (1.0 kg, 4.1 mol) was dissolved in CH3CN/MeOH
(9:1, 19 L), and (+)-(S)-phenylglycinol (560 g, 4.1 mol) was
then added. The mixture was heated at reflux for 1 h and was
then cooled to 20 °C over 20 h. The solid was filtered, washed
with CH3CN/MeOH (9:1, 0.8 L), and then dried at 40 °C for
20 h to give crude salt p (48% yield, 84% ee). Crude salt p
was slurried in CH3CN/MeOH (9:1, 15 L per kg of crude) at
reflux for 1 h and was then cooled to 20 °C over 20 h. The
solid was filtered, washed with CH3CN/MeOH (9:1, 0.8 L),
and dried at 40 °C for 20 h to give 670 g (43% overall yield)
of pure salt p: 95.2% HPLC assay (HPLC Method 1); 96% ee
[tR (S)-4 ) 14.6 min, (R)-4 ) 13.1 min, HPLC Method 2]; mp
185.5 °C.
Preparation of 2-(2,2-Dimethyl-propionylamino)-3,3-di-
methyl-pent-4-enoic Acid [(S)-4] by Hydrolysis of salt p. Salt
p (809 g, 2.1 mol) was suspended in H2O (3.8 L) and was then
cooled to 5 °C. EtOAc (2.2 L) was added, and the temperature
was readjusted to 5 °C. The pH was adjusted to 3.0 ( 0.5 with
10% aqueous HCl, while maintaining the internal temperature
at <10 °C. The phases were separated and the aqueous phase
was then extracted with EtOAc (485 mL). The EtOAc extracts
were combined to afford a solution of (S)-4 (quantitative yield)
that can be directly used for the synthesis of 5b.
6 was obtained as a white solid: 97.4% HPLC purity; 92%
HPLC assay; cis/trans ratio ) 93:7 (tR cis-6b ) 12.5 min, tR
trans-6b ) 12.2 min, HPLC Method 1); 96% ee [tR (3S,5S)-
6b ) 19.3 min, tR (3R,5R)-6b ) 11.0 min, tR (3S,5R)-6b )
1
91.2 min, tR (3R,5S)-6b ) 15.3 min, HPLC Method 4]; H
NMR (500 MHz, DMSO-d6) cis-6b δ 0.90 (s, 3H, CH3-C-CH),
1.34 (s, 3H, CH3-C-CH), 3.64 (dd, J ) 11.6, 9.9 Hz, 1H, CH2-
CBr), 3.99 (dd, J ) 11.6, 2.3 Hz, 1H, CH2-CBr), 4.39 (s, 1H,
CH-N), 4.65 (dd, J ) 10.0, 2.4 Hz, 1H, CH-CH2Br), 8.95 (s,
3H, NH3), trans-6b δ 1.13 (s, 3H, CH3-C-CH), 1.29 (s, 3H,
CH3-C-CH), 3.64 (dd, J ) 11.6, 9.9 Hz, 1H, CH2-CBr), 3.99
(dd, J ) 11.6, 2.3 Hz, 1H, CH2-CBr), 4.45 (s, 1H, CH-N), 4.61
(dd, J ) 9.4, 3.3 Hz, 1H, CH-CH2Br), 8.95 (s, 3H, NH3+); 13C
NMR (100.6 MHz, DMSO-d6) cis-6b δ 15.0 (CH3), 22.8 (CH3),
30.7 (CH2), 44.3 [(CH3)2-C], 58.7 (N-C), 86.8 (C-CHO), 172.6
(CO2CN), trans-6b δ 20.0 (CH3), 22.3 (CH3), 32.5 (CH2), 41.5
[(CH3)-C], 55.6 (N-C), 86.4 (C-CHO), 171.4 (CO2CN). MS
(ESI): [C7H13BrNO2]+ 222.0.
Preparation of (2S)-4-Hydroxy-3,3-dimethyl-pyrrolidine-
1,2-dicarboxylic Acid 1-tert-Butyl Ester 2-Methyl Ester (2S)-
8. NaHCO3 (16.2 g, 193 mmol) was suspended in MeOH (100
mL) at ambient temperature. Compound 6b (50.0 g, 193 mmol)
was dissolved in MeOH (150 mL), and the resulting solution
was then added to the NaHCO3 suspension while the internal
temperature was at 20 ( 5 °C. The mixture was heated at reflux
(65 ( 2 °C) for g17 h and was then cooled to 20 ( 5 °C. A
solution of Boc2O (50.7 g, 232 mmol) and NEt3 (26.9 mL, 193
mmol) was added, and the mixture was then stirred at 20 ( 5
°C for g1 h. The reaction mixture was concentrated to one-
third of the volume, toluene (210 mL) was added, and the
mixture was then washed with H2O (2 × 170 mL). The organic
phase was concentrated under vacuum (50-100 mbar) at 40
( 5 °C to give a 50 wt % solution of 8. n-Hexane (350 mL)
was added at 40 ( 5 °C, and the solution was then slowly
cooled to 20 ( 5 °C. After crystallization, the suspension was
stirred at -10 °C for 2 h and was then filtered. The solid was
washed with n-hexane (50 mL) and was then dried at 35 ( 5
°C for g18 h to give 42.2 g (80%) of 8 as a white solid: 93%
HPLC purity; cis/trans ratio ) 90:1 [tR cis-(2S)-8 ) 6.4 min,
tR trans-(2S)-8 ) 6.0 min, HPLC Method 1]; 100% ee [tR
(3S,5S)-8 14.7 min, tR (3R,5R)-8 13.7 min, tR (3S,5R)-8 17.8
min, tR (3R,5S)-8 16.1 min, HPLC Method 5]; mp 87.4 °C; 1H
NMR (500 MHz, DMSO-d6) cis-(2S)-8 δ 0.78 (s, 3H, CH3-
C-CH), 1.11 (s, 3H, CH3-C-CH), 1.33 and 1.41 rotamers [s,
9H, (CH3)3-CO], 3.03–3.10 (m, 1H, CH2-N), 3.57 – 3.83 (m,
1H, CH2-N), 3.65 (s, 3H, CH3O), 3.78–3.82 (m, 1H, CH-OH),
3.87 (s, 1H, CH-COOCH3), 5.18 (d, J ) 5.02 Hz, 1H, CH-
OH); 13C NMR (100.6 MHz, DMSO-d6) cis-(2S)-8 mixture of
rotamers δ 16.8 and 16.9 (CH3), 25.0 and 25.2 (CH3), 27.8 and
28.0 [(CH3)3-CO], 43.6 and 44.5 (CH-OH), 50.3 and 51.0
(CH2), 51.3 and 51.4 (CH3O), 67.0 and 67.4 (CH-CO2CH3),
74.2 and 74.9 (CH-OH), 78.9 and 79.0 [C-(CH3)2], 153.0 and
153.7 [CO-C(CH3)3], 170.3 and 170.9 (CO). MS (ESI)
[C13H23NO5 + H]+ 273.7, [C9H15NO5 + H]+ 217.8, [C8H15NO3
+ H]+ 174.1.
Preparation of cis/trans-[(3S)-5-Bromometlyl-4,4-dimeth-
yl-2-oxo-tetrahydro-furan-3-yl]-carbamic Acid tert-Butyl
Ester (5b). NBS (42 g, 234 mmol.) was suspended in EtOAc
(261 mL), and the mixture was then cooled to 0 °C. A solution
of (S)-4 (51.8 g, 213 mmol) was added at a rate to maintain
the internal temperature at 0 ( 5 °C. The reaction was quenched
by addition of H2O (170 mL), followed by Na2SO3 (6.5 g).
The mixture was washed with 2% aqueous NaHCO3 (170 mL),
which afforded a solution of 5 (quantitative yield) that was
directly used for the synthesis of 6b. An analytical sample was
prepared in 94% yield by crystallization from n-hexane: 90%
HPLC assay; cis/trans ratio ) 93:7 (tR cis-5b ) 12.5 min, tR
trans-5b ) 12.2 min, HPLC Method 1); 96% ee [tR (3S,5S)-
5b ) 19.3 min, tR (3R,5R)-5b ) 16.8 min, tR (3S,5R)-5b )
1
26.6 min, tR (3R,5S)-5b ) 24.0 min, HPLC Method 3]; H
NMR (400 MHz, DMSO-d6) δ 0.69 (s, 3H, CH3-C-CH), 1.07
(s, 3H, CH3-C-CH), 1.40 (s, 9H, (CH3)3CO), 3.51 (dd, J ) 11.5,
10.0 Hz, 1H, CH2-CBr), 3.86 (dd, J ) 11.5, 2.6 Hz, 1H, CH2-
CBr), 4.54 (dd, J ) 10.0, 2.6 Hz, 1H, CH-CH2Br), 4.60 (d, J
) 9.8 Hz, 1H, CH-N), 7.40 (d, J ) 9.8 Hz, 1H, NH); 13C NMR
(100.6 MHz, DMSO-d6) δ 15.1 (CH3), 22.2 (CH3), 28.1
[(CH3)3CO], 30.4 (CH2), 43.8 [(CH3)-C], 59.8 (N-C), 78.6
[C-(CH3)3], 84.3 (C-CHO), 155.8 [CO(CH3)3], 173.4 (CO2CN).
MS (ESI): [C12H20BrNO4 + Na]+ 345.9, [C8H12BrNO4 + H]+
265.9, [C7H12BrNO2 + H]+ 224.0.
Preparation of (+)-(S)-3-Amino-5-bromomethyl-4,4-di-
methyl-dihydro-furan-2-one (+)-(3S)-6b. The solution of 5b
in EtOAc prepared above was concentrated to 1 M and was
measured by KF to contain e0.2 wt % H2O. An equal volume
of 4 M HCl/EtOAc was added in two portions over a period of
at least 3 h, while maintaining the temperature at 20 ( 5 °C.
The mixture was allowed to react for g10 h at 20 ( 5 °C and
was then cooled to -10 °C for 2 h. The solid was filtered, and
the solid was washed with n-hexane without mixing. After
drying at 40 ( 5 °C under vacuum for g15 h, 50 g (91%) of
Isolation of 7,7-Dimethyl-3-oxo-2-oxa-5-aza-bicyclo[2.2.1]-
heptane-5carboxylic Acid tert-Butyl Ester (13). Isolated from
the organic phase in the synthesis of 8 and purified by flash
336
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development

chromatography on silica gel (n-hexane/EtOAc 75:25 f 70:
30): tR 10.0 min (HPLC Method 1); H NMR (400 MHz,
20.8 (CH3), 27.5 [(CH3)3C], 45.1 [C(CH3)2], 50.3 (CH2), 51.4
(CH3O), 66.8 (CH-CO2CH3), 80.0 [C-(CH3)3], 126.3 (CF2),
152.7 [COC(CH3)3], 169.8 (COOCH3). MS (ESI): [C9H13F2NO4
+ H]+ 237.8, [C8H13F2NO2 + H]+ 194.0, [C6H11F2N + H]+
134.1.
Isolation of 4-Fluoro-3,3-dimethyl-2,3-dihydro-pyrrole-
1,2-dicarboxylic Acid 1-tert-Butyl Ester 2-Methylester (17).
Isolated in the synthesis rac-4 and purified by flash chroma-
tography on silica gel [hexane/EtOAc 4:0.1 f 3:7; Rf ) 0.7
hexane/EtOAc (1:2)] to give an oil: tR 16.9 min (HPLC Method
1); 1H NMR (500 MHz, 70 °C, DMSO-d6) δ 1.01 (s, 3H, CH3-
C-CH), 1.36 (s, 3H, CH3-C-CH), 1.42 [s, 9H, (CH3)3-CO], 3.75
(s, 3H, CH3O), 4.31 (s, 1H, CH-COOCH3), 6.51 (s, 1H, CH-
CF2); 13C NMR (100.6 MHz, DMSO-d6) δ 21.7 (CH3), 26.3
(CH3), 27.8 [(CH3)3C], 52.0 (CH3O), 69.2 (CH-CO2CH3), 108.9
(CH-CF2). MS (ESI) [C9H12FNO4 + H]+ 217.8; [C8H12FNO2
+ H]+ 173.9; [C6H8FN + H]+ 114.0.
Isolation of N-[N,N-Bis(2-methoxy-ethyl)-aminosulfanyl]-
4,5-difluoro-3,3-dimethyl-pyrrolidine-1,2-dicarboxylic Acid
1-tert-Butyl Ester 2-Methyl Ester (18). Isolated in the synthesis
rac-4 and purified by flash chromatography on silica gel
(hexane/EtOAc 1:0 f 2:1; Rf ) 0.3 hexane/EtOAc 1:2) to give
18 as a mixture of diastereomers: tR 21.0, 21.3 min (HPLC
Method 1); 1H NMR (500 MHz, DMSO-d6) mixture of
diastereomers δ 0.97 and 1.07 (s, 3H, CH3-C-CH), 1.1.8 and
1.25 (s, 3H, CH3-C-CH), 1.45 [s, 9H, (CH3)3-CO], 3.23–2.21
(m, 2H, CH3O-CH2-CH2-N), 3.29 (s, 3H, CH3O-CH2-CH2-N),
3.51 (t, J ) 5.97, 2H, CH3O-CH2-CH2-N), 3.75 (s, 3H, CH3O),
4.19 and 4.21 (s, 1H, CH-COOCH3), 5.98 (d, J ) 5.97, 1H,
F-CH); 13C NMR (125 MHz, 70 °C, DMSO-d6) mixture of
diastereomers δ 19.9 (CH3), 21.8 (CH3), 28.5 [(CH3)3C], 47.1
[C(CH3)2], 52.4 (CH3O), 58.7 (CH3O-CH2), 60.0 (N-CH2), 69.8
(CH-CO2CH3), 71.3 (CH2-O), 83.0 [C-(CH3)3], 153.8 [COC-
(CH3)3], 169.8 (COOCH3). MS (ESI) [C19H34F2N2O6S + Na]+
479.0; [C19H33FN2O6S + H]+ 436.9; [C14H25FN2O4S + H]+
336.9.
Preparation of 4,4-Difluoro-3,3-dimethyl-pyrrolidine-1,2-
dicarboxylic Acid 1-tert-Butyl Ester (1) without Chroma-
tography. CAUTION: Avoid the use of acid in the presence
of Deoxo-Fluor and KF to prevent evolution of HF (g). Ketone
9 (50 g, 184 mmol) was charged into a PTFE- or Halar-coated
reactor at 23 °C, which was then followed by the addition of
silica gel (2.5 g). Deoxo-Fluor (244.7 g of a 50 wt % solution
in toluene, 552.9 mmol) was slowly added. The resulting
mixture was heated to 70 ( 2 °C and was then stirred at this
temperature for 5–10 h. The mixture was cooled to 23 °C and
was then quenched in 10% aqueous NH4OH (425 mL) while
the pH was maintained around 8.5. The organic phase was
separated and was then washed with H2O (425 mL). The solvent
was distilled under vacuum to give a mixture of 9 and 10 as an
oil, which was then dissolved in n-hexane (133 mL). NaOH (2
N, 154 mL) was added, and the resulting biphasic mixture was
agitated at 23 °C for about 24 h until no residual 9 remained in
the organic phase. The layers were separated, and the organic
phase was concentrated under vacuum to give crude 10 as an
oil. The residue was dissolved in MeOH (255 mL), and 2 N
NaOH (255 mL) was then added. After stirring for 4 days at
23 °C, the solution was adjusted to ∼pH 7 with 10% aqueous
1
DMSO-d6) δ 1.01 (s, 3H, CH3), 1.07 (s, 3H, CH3), 1.39 [s,
9H, (CH3)3], 3.20 (d, J ) 11.4 Hz, 1H, CH2), 3.54 (br d, J )
11.4, 1H, CH2), 4.08 (br s, 1H, CH-N), 4.67 (s, 1H, CH-O);
13C NMR (100.6 MHz, DMSO-d6) δ 17.3 (CH3), 17.6 (CH3),
27.8 [(CH3)3C], 48.1 [C(CH3)2], 48.8 (CH2), 65.3 (CH-N), 80.1
[C(CH3)3)], 84.2 (CH-O), 154.0 [COC(CH3)3], 172.8 (CO). MS
(ion trap) [C12H19NO4]+ 241.28.
Preparation of 3,3-Dimethyl-4-oxo-pyrrolidine-1,2-dicar-
boxylic Acid 1-tert-Butyl Ester 2-Methyl Ester (9). A mixture
of 8 (300 g, 1.1 mol), TEMPO (1.7 g, 9.1 mmol), KBr (1.7 g,
17.2 mmol), NaHCO3 (83.4 g, 1.0 mol), and CH2Cl2 (1670 mL)
was cooled to 0 °C. NaClO (1.8 L of a 5% aqueous solution)
was added over 70 min, while maintaining the reaction
temperature at <7 °C. The solution was stirred vigorously at 3
°C for 60 min. The resulting phases were separated, and the
aqueous fraction was then extracted with CH2Cl2 (500 mL).
The combined CH2Cl2 phases were sequentially washed with
5% aqueous Na2S2O5 (1.8 L) and H2O (1.8 L). The resulting
CH2Cl2 solution was filtered to remove residual salts and was
then evaporated to give 299 g (100%) of 9 as an oil: 97% HPLC
purity; 99% HPLC assay (tR 12.1 min, HPLC Method 1); 100%
ee [tR (2S)-9 25.2 min, tR (2R)-9 22.8 min, HPLC Method 2];
1H NMR (500 MHz, DMSO-d6) mixture of rotamers δ 0.96
(s, 3H, CH3-C-CH), 1.24 and 1.25 (s, 3H, CH3-C-CH), 1.40
and 1.45 [s, 9H, (CH3)3-CO], 3.68 and 3.72 (s, 3H, CH3O),
3.82 (d, J ) 18.6 Hz, 1H, CH2-N), 4.09 (d, J ) 18.9 Hz, 1H,
CH2-N), 4.35 and 4.38 (s, 1H, CH-COOCH3); 13C NMR (100.6
MHz, DMSO-d6) δ 18.2 (CH3), 24.7 (CH3), 27.8 [(CH3)3C],
49.3 [C(CH3)2], 50.9 (CH2), 52.0 (CH3O), 67.4 (CH-CO2CH3),
81.0 [C-(CH3)3], 172.2 (COOCH3), 212.8 (CO). MS (ESI):
[C14H22NO5+ H]+ 271.6; [C10H14NO5 + H]+ 215.8.
Preparation of (S)-4,4-Difluoro-3,3-dimethyl-pyrrolidine-
1,2-dicarboxylic Acid 1-tert-Butyl Ester 2-Methyl Ester (10).
CAUTION: Avoid the use of acid in the presence of Deoxo-
Fluor and KF to prevent evolution of HF (g). Ketone 9 (50 g,
184 mmol) was charged into a PTFE- or Halar-coated reactor
at 23 °C and was then followed by the addition silica gel (2.5
g). Deoxo-Fluor (244.7 g of a 50 wt % solution in toluene,
552.9 mmol) was slowly added. The resulting mixture was
heated to 70 ( 2 °C and was then stirred at this temperature
for 5–10 h. NaHCO3 (3.1 g, 37 mmol), KF (10.7 g, 184 mmol),
and Aliquat 336 (8.4 mL, 18.4 mmol) were added at 70 °C,
and the mixture was allowed to stir at 70 °C for an additional
5–24 h. The mixture was cooled to 23 °C and was then
quenched in 10% aqueous NH4OH (425 mL) while the pH was
maintained around 8.5. The organic phase was separated and
was then washed with H2O (425 mL). Evaporation of the
organic phase, followed by flash chromatography on silica gel
(n-hexane/EtOAc 30:1 f 10:1.5), gave 25.4 g (47%) of 10 as
an oil: 98% HPLC purity; 97% HPLC assay (tR 16.1 min, HPLC
Method 1); 100% ee (tR (2S)-10 51.7 min, tR (2R)-10 43.6 min,
HPLC Method 2); 1H NMR (500 MHz, DMSO-d6) mixture of
rotamers δ 1.01 (s, 3H, CH3-C-CH), 1.25 (s, 3H, CH3-C-CH),
1.36 and 1.43 [s, 9H, (CH3)3-CO], 3.70 and 3.73 (s, 3H, CH3O),
3.87 (dd, J ) 26.0, 12.48 Hz, 2H, CH2-N), 4.10 (s, 1H, CH-
COOCH3); 13C NMR (100.6 MHz, DMSO-d6) δ 16.9 (CH3),
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
337

HCl. The organic solvent was distilled off under vacuum. The
resulting aqueous solution was diluted with H2O (100 mL) and
was then washed with toluene (2 × 175 mL). The pH was
adjusted to 12.5 with 10% aqueous NaOH, and NaOCl (31 mL
of a 3.5 wt % aqueous solution) was then added. The resulting
mixture was stirred at 23 °C until 17 no longer remained (2 h
required). n-Hexane (75 mL) was added to the mixture, and 1
was then precipitated by adjustment to pH 2.5 with 10%
aqueous HCl. The solid was filtered, washed with water, and
dried to give 1 in 38–45% yield from 9 (45–50% for the
fluorination and 85–95% for hydrolysis): 96% HPLC purity;
97% HPLC assay (tR 9.2 min, HPLC Method 1); 100% ee [tR
(2S)-1 14.1 min, tR (2R)-1 13.2 min, HPLC Method 2]; 1H NMR
(500 MHz, DMSO-d6) mixture of rotamers δ 1.06 (s, 3H,
CH3-C-CH), 1.24 (s, 3H, CH3-C-CH), 1.38 and 1.43 [s, 9H,
(CH3)3-CO], 3.80 (dd, J ) 25.7, 13.2, 2H, CH2-N), 3.98 (s,
1H, CH-CO2CH3), 12.97 (s, 1H, COOH); 13C NMR (75 MHz,
DMSO-d6) mixture of rotamers δ 17.7 and 17.9 (CH3), 21.4
and 21.6 (CH3), 28.1 and 28.3 [(CH3)3C], 44.9 and 45.6
[C(CH3)2], 51.0 and 51.5 (CH2), 67.3 and 67.8 (CH-CO2CH3),
80.3 and 80.8 [C-(CH3)3], 126.8 and 127.3 (CF2), 153.3 and
153.7 [COC(CH3)3], 170.3 and 170.8 (COOH). MS (ESI)
[C8H11F2NO4 + H]+ 223.9; [C7H11F2NO2 + H]+ 180.0;
[C6H9F2N + H]+ 134.1.
Acknowledgment
This work could not have been done without the important
contribution of many colleagues. We are thankful to C.
Brusamolino, I. Colombo, A. Re, L. Biagini, and M. De Pietro
for analytical support; G. Ottainano, G. Mandirola, S. Capezzera,
P. Leonetti, D. Mercuri, and Franco Rossi for production
operations; M. Ungari for hazard evaluation; and A. Adduci
and C. Lista for lab support.
Received for review May 22, 2007.
OP7001112
338
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development