Stereochemistry of C-6 nucleophilic displacements on 1,1- difluorocyclopropyldibenzosuberanyl substrates. An improved synthesis of multidrug resistance modulator LY335979 trihydrochloride

Article abstract of DOI:10.1021/jo049051v

Studies of the displacement chemistry of 1,1- difluorocyclopropyldibenzosuberanyl alcohol 4 and its activated bromide derivative 6 have led to an improved approach to anti-2, a key precursor to LY335979 3HCl (1). Bromination of either syn-4 or anti-4 gave

Full text of DOI:10.1021/jo049051v

Ster eoch em istr y of C-6 Nu cleop h ilic Disp la cem en ts on
1,1-Diflu or ocyclop r op yld iben zosu ber a n yl Su bstr a tes. An Im p r oved
Syn th esis of Mu ltid r u g Resista n ce Mod u la tor LY335979
Tr ih yd r och lor id e
Charles J . Barnett,*,† Bret Huff,* Michael E. Kobierski, Michael Letourneau, and
Thomas M. Wilson
Chemical Product Research and Development Division, Lilly Research Laboratories,
A Division of Eli Lilly and Company, Indianapolis, Indiana, 46285
cjbarnett@comcast.net; bhuff@lilly.com
Received J une 6, 2004
Studies of the displacement chemistry of 1,1-difluorocyclopropyldibenzosuberanyl alcohol 4 and
its activated bromide derivative 6 have led to an improved approach to a n ti-2, a key precursor to
LY335979 3HCl (1). Bromination of either syn -4 or a n ti-4 gave anti-oriented 6, indicating
thermodynamically controlled product stereochemistry via a stabilized 1,1-difluorohomotropylium
ion intermediate. Reaction of 6 with piperazine proceeded irreversibly to provide an isomeric mixture
of piperazine products, with the syn:anti product ratio increased by solvent effects. Reaction of 6
with pyridine and pyrazine, on the other hand, gave anti-pyridinium and pyrazinium salts,
respectively, apparently via equilibration of initially formed syn products. Reduction of pyrazinium
salt 11 with lithium borohydride/TFA provided a n ti-2 unaccompanied by its syn isomer. A practical
and expeditious approach to 1 was derived from these new results.
SCHEME 1
In tr od u ction
The development of multiple drug resistance is a
serious limitation on the effectiveness of many oncolytic
drugs. New findings on the biological basis of multiple
drug resistance have permitted the rational search for
modulators of specific resistance pathways.1 Compound
LY335979 3HCl (1) has been found to be a clinically
useful modulator of P-glycoprotein-mediated multiple
drug resistance in research beginning at the former
Syntex Corporation and continuing2 at Lilly Research
Laboratories.3 The expansion of clinical trials prompted
us to embark on a program to create a practical and cost-
effective synthetic approach to 1.
similar structure saturated in place of the difluorocyclo-
propyl group in 1.
The Syntex strategy5 for the synthesis of 1 is indicated
LY335979‚3HCl was first prepared at Syntex as part
retrosynthetically in Scheme 1.
of an effort to improve upon the activity of MS-073,4 a
One key disconnection was at the C-N bond joining
the substituted piperazinyl moiety to the chiral 2-hy-
droxypropoxyquinoline linker via the substituted pipera-
zine 2 possessing the required anti stereochemistry6 and
the known7 (R)-epoxide 3. Although other approaches
could be envisioned, this one provided the greatest
† Retired J une 30, 2001. Current address: 1305 Brickmill East,
Mount Pleasant, SC 29466-7919.
(1) Norman, B. H. Drugs Future 1998, 23, 1001-1013. Ford J M.;
Hait, W. N. Pharmacological Reviews 1990 42, 155-99.
(2) LY335979‚3HCl was licensed from Syntex Corporation in No-
vember, 1994. The Syntex serial number was RS-33295-198.
10.1021/jo049051v CCC: $27.50 © 2004 American Chemical Society
Published on Web 10/02/2004
J . Org. Chem. 2004, 69, 7653-7660
7653

Barnett et al.
SCHEME 2
SCHEME 3
convergence and relied upon the opening of an epoxide
by an amine, a well-known and stereochemically reliable
reaction.
On the other hand, the synthesis of a n ti-2 as described
by the Syntex group was encumbered with relative
stereochemical problems as indicated in Scheme 2.
Chlorination of syn -4 (SOCl2) provided a mixture of
chlorides 5 in an anti:syn ratio of about 90:10.8 Reaction
of the mixture with formylpiperazine, however, unfortu-
nately gave rise to a nearly equal mixture of piperazinyl
isomers a n ti-7 and syn -7 (anti:syn 55:45).9 The mixture
could be separated by chromatography, providing purified
a n ti-7, which gave a n ti-2 upon deprotection with strong
base.
Although it was possible to utilize the original ap-
proach to a n ti-2 for the preparation of relatively small
quantities of LY335979, the unfavorable impact of am-
biguous stereochemistry of the piperazine displacement
step on efficiency of the synthesis quickly became appar-
ent. We describe herein studies leading to successful
resolution of the stereochemistry issue, leading to an
expeditious approach to 1 suitable for commercial scale
use.
geous. Reaction of 4 with either phosphorus tribromide
in halogenated hydrocarbon solvents or, preferably, 48%
aqueous HBr, however, gave crystalline dibenzosuberanyl
bromide 6 (Scheme 2) as a single, anti stereoisomer in
high yield.
The formation of 6 represented apparent inversion of
the configuration of the hydroxyl-bearing center, but the
cyclopropyldibenzosuberane-based structures of this se-
ries suggested the possibility of stabilized cationic inter-
mediates instead of an SN2 mechanism. The reaction
sequence shown in Scheme 3 was carried out in order to
clarify the mechanism of the formation of 6. Reaction of
syn -4 with benzoic acid under Mitsunobu10 conditions
gave, after purification, a n ti-8. The assignment of anti
stereochemistry to the product was supported by NMR
data that showed no NOE at the cyclopropyl hydrogens
upon irradiation of the C-6 proton. In contrast, the
isomeric benzoate syn -8, prepared by esterification of
syn -4 with benzoic anhydride, clearly showed a NOE
between the cyclopropyl and C-6 protons. Hydrolysis of
a n ti-8 with sodium hydroxide in methanol gave the
previously unreported a n ti-4. Reaction of a n ti-4 with
phosphorus tribromide under the conditions described for
conversion of syn -4 to 6 gave anti-bromide 6 as the only
product with apparent retention of configuration. These
results clearly indicated that the stereochemical outcome
of bromination of syn -6 was not caused by Walden
inversion and was best explained by an SN1 mechanism
involving the fluorinated dibenzohomotropylium ion11
intermediate 9 (Scheme 3).
Ster eoch em ica l Stu d ies
Syn th esis of Br om id e 6 fr om syn -4. We found the
bromination of dibenzosuberanol syn -4 to be the method
of choice for activating the molecule for amine displace-
ment. Attempted preparation of sulfonate esters of syn -4
failed to give useful products, apparently because of the
instability of these intermediates. As mentioned previ-
ously, chlorination with SOCl2 provided a mixture of
stereoisomeric chlorides and was not otherwise advanta-
(3) For an excellent review of the biological research on LY335979,
see: Dantzig, A. H.; Law, K. L.; Cao, J .; Starling, J . J . Curr. Med.
Chem. 2001 8, 39-50.
(4) (a) Sato, W.; Fukazawa, N.; Suzuki, T.; Yusa, K.; Tsuruo, T.
Cancer Res. 1991, 51, 2420-4. (b) Suzuki, T.; Fukazawa, N.; San-nohe,
K.; Sato, W.; Yano, O.; Tsuruo, T. J . Med. Chem. 1997, 40, 2047-2052.
(5) (a) Pfister, J . R.; Makra, F.; Muehldorf, A. V.; Wu, H.; Nelson,
J . T.; Cheung, P.; Bruno, N. A.; Casey, S. M.; Zutshi, N.; Slate, D. L.
Bioorg. Med Chem Lett. 1995, 5, 2473-2476. (b) Pfister, J . R.; Slate,
D. L. U.S. Patent 5,643,909, 1997.
(6) For this article we define the relative stereochemistry in all
structures related to 2 as anti on the basis of the relatively opposite
positions of the cyclopropyl and C-6 substituent groups with respect
to the dibenzocycloheptane ring. Syn refers to the orientation of these
groups on the same side of the dibenzocyloheptane ring.
(7) Fukazawa, N.; Suzuki, T.; Kawauchi, N.; Komatsu, H.; Otsuka,
K.; Nakajima Y. U.S. Patent 5,463,61, 1995.
Thus formation of 6 occurred under thermodynamic
control by (reversible) attack of bromide ion on 9, and
(8) The anti:syn ratio of chlorides 5 was determined by repeating
the reported experiment from ref 5b.
(9) The anti:syn ratio for the isomeric mixture of 7 was determined
by HPLC of the crude product as a part of this work. A ratio of these
isomers of anti:syn 52:48 after chromatographic separation was
reported by Syntex (ref 5b).
(10) Hughes, D. L. Org. React. 1992, 42, 335-656.
(11) For a definitive paper on the chemistry of the dibenzohomo-
tropylium ion, see: Childs, R. F.; Brown, M. A.; Anet, F. A. L.; Winstein,
S. J . Am. Chem. Soc. 1972, 94, 2175-2183 and references therein.
7654 J . Org. Chem., Vol. 69, No. 22, 2004

Improved Synthesis of LY335979 Trihydrochloride
TABLE 1. Solven t Effects on th e Ster eoch em ica l
ou tcom e of th e Rea ction of 6 w ith P ip er a zin e
any of the kinetically formed syn-bromide was apparently
rapidly equilibrated to 6 via reversion to 9. These results
provided a framework for understanding the displace-
ments of amine nucleophiles on 6.
Disp la cem en ts on 6 by Nitr ogen Nu cleop h iles. A
3-fold excess of piperazine was successfully substituted
for the much more expensive formylpiperazine in the
preparation of a n ti-2 without reduction in yield or
formation of bis-substituted piperazine byproducts,12
albeit with no effect on the ratio of piperazine products.
These results shortened the synthesis of a n ti-2 by
obviating the protection and deprotection of the essential
reagent, piperazine. The resulting isomeric mixture of
piperazine products 2 were, however, not easily separated
by preparative HPLC. We have previously reported an
alternative, nonchromatographic separation scheme in
which the gross amount of unwanted syn -2 was removed
by crystallization of the free base mixture from acetoni-
trile and the resulting enriched mixture was converted
to the hydrochloride or hydrobromide salts and crystal-
lized from dichloromethane.13 Recrystallization from
dichloromethane provided purified a n ti-2 as its salt (anti:
syn 99.5:0.5) in about 32% overall yield from bromide 6.
Formation of the isomeric piperazines was not revers-
ible. Exposure of each of the isomers of 2 separately to
the conditions of piperazine displacement in acetonitrile
returned the compounds unchanged. We surmised that
the reaction of 6 with piperazine resulted in almost
indiscriminate, irreversible attack on the homotropylium
ion 9 in view of the results of the foregoing study of the
stereochemistry of formation of bromide 6.
We hoped to influence the relative rates of the compet-
ing reactions by varying the reaction medium. The results
of a study of solvent effects on the displacement of
bromide 6 by piperazine are shown in Table 1.
A very slight concentration effect was noted in aceto-
nitrile (entries 1-3), with the product ratio only slightly
more favorable at lower concentrations. Acetonitrile, the
standard solvent choice, proved to be the best for favoring
a n ti-2. Other solvents examined gave less favorable anti:
syn ratios, and tetrahydrofuran provided a good (relative)
yield of the undesired syn isomer.
entry
solventa
2, anti:syn
1
2
3
4
5
6
7
8
9
10
11
12
13
acetonitrile, [6]init 0.125 M
acetonitrile, [6]init 0.208 M
acetonitrile, [6]init 0.62 M
acetone
isopropyl alcohol
ethyl acetate
55:45
54:46
42:58
41:59
40:60
39:61
36:64
35:65
34:66
34:66
26:74
26:74
15:85
methanol
ethanol
dimethyl formamide
chloroform
toluene
tert-butyl methyl ether
tetrahydrofuran
a
Initial concentration of 6 was 0.156 M except as noted. For
details see Experimental Section.
tion failed to provide useful amounts of desired products,
but a single pyridinium tosylate was isolated in 18%
yield. The analogous pyridinium bromide (a n ti-10, below)
was obtained in quantitative yield, as a single stereoiso-
mer, when bromide 6 was reacted with pyridine in
chloroform at 57 °C.
The following NMR experiment allowed assignment of
anti stereochemistry to the product of the foregoing
reaction. Bromide 6 was dissolved in pyridine-d5 and the
resulting process was followed, initially at 27 °C. The
relative configuration of the initially predominant prod-
uct (syn -10) was assigned from observation of a NOE
enhancement between the cyclopropyl fusion and C-6
protons. As the temperature was increased from 27 to
47 °C; however, the initial product was replaced by a new
component that showed a similar proton NMR spectrum
but lacked the NOE observed in the early product. Thus,
structure a n ti-10 was assigned to the final, more stable
product.14 These results are in accord with observations
reported by Winstein et al.15 of the solvolyses of the
acetate esters of didesfluoro-4 in which a syn-oriented
product predominated (under kinetic control) regardless
of the relative stereochemistry of the starting materials.
This is, to our knowledge, the first example of a reaction
involving a dibenzohomotropylium species in which an
initially formed syn displacement product is equilibrated
to the (isolated) anti isomer.
A search of the literature did not uncover any reports
of reactions of other nitrogen heterocycles with dibenzo-
cyclopropylsuberanyl halides. The following serendipitous
experiment proved, however, to be very illuminating.
Compound syn -4 was reacted with piperazine in toluene
in the presence of pyridinium tosylate in an effort to
prepare piperazines 2 by direct displacement on 4, thus
precluding the activating step of bromination. The reac-
(12) A sample of the potential side product, N,N-dialkylated pip-
erazine i (stereochemistry not determined), was prepared by reaction
of 2 and 6 and determined not to be detectable in the purified product,
2, by HPLC analysis.
(13) Astleford, B. A.; Barnett, C. J .; Kobierski, M. E.; Wilson, T. M.
U.S. Patent 6,570,016, 2003. Astleford, B. A.; Barnett, C. J .; Kobierski,
M. E.; Wilson, T. M. U.S. Patent 6,624,304, 2003.
(14) After 2 days at 57 °C, none of the initial product (syn -10) could
be observed by NMR.
J . Org. Chem, Vol. 69, No. 22, 2004 7655

Barnett et al.
TABLE 2. NMR Stu d y of th e Effect of Solven ts on th e
Rea ction of Br om id e 6 w ith P yr a zin e
12 required the high-temperature conditions of chlorodi-
fluoroacetate pyrolysis.
We made several improvements to the Syntex difluo-
rocyclopropanation protocol that made it more efficient
and amenable to large-scale use. For example, large
stoichiometric excesses (15-25 equiv) of alkali salts of
chlorodifluoroacetic acid had been required in the original
process to drive the reaction substantially to completion
at the reported reaction temperature (160 °C). Although
the yield as calculated on the consumption of 12 was
acceptable at 60-70%, the very poor conversion of the
(anhydrous) chlorodifluoroacetate salt caused the process
to become unacceptably expensive. Additionally, the non-
product-forming decomposition of excess carbene reagent
generated copious amounts of gaseous and solid waste.
We considered this problem to be one of competition
between product-forming difluorocarbene insertion to the
target olefin and many potential side reactions involving
difluorocarbene: oligomerization to tetrafluoroethylene
and its insertion products, higher polymers, and reactions
with solvent.18
pyrazine
temp reaction yield
entry
solvent
molar excess (°C) time (h) (%)
1
2
3
4
5
6
7
none
DMSO
110
37
30
70
75
25
55
0.5
4
25
4.5
24
91a
93a
88a
60
3.0
3.0
3.0
6.0
3.0
6.0
DMSO-CH2Cl2 (1:9)
DMF
CH3CN
CH2Cl2
THF
25
4
23
18a
0
a
Isolated yield. Experiments 4, 5, and 7 were carried out in a
NMR tube in the indicated deuterated solvent. Yields for 4, 5, and
7 are given as ratios of starting material to product at the indicated
time.
As expected, reaction of bromide 6 with neat pyrazine
also provided the desired anti-difluorodibenzosubera-
nylpyrazinium bromide16 (11), further confirming the
stereochemical course of the reaction. In contrast to the
pyridine experiment, however, the corresponding syn-
oriented product was not observable by NMR, suggesting
a relatively rapid equilibration of the isomeric products.
Importantly, no bis-pyrazinium products could be de-
tected in the isolated product. A number of solvents were
screened for the reaction, and the results are summarized
in Table 2. DMSO clearly provided best results for the
pyrazine displacement, but addition of cosolvents to the
DMSO reactions did not have a deleterious effect on the
rate or yield.
In preliminary experiments we observed that the
amount of excess lithium19 chlorodifluoracetate required
for complete consumption of 12 was inversely propor-
tional to reaction temperature and directly proportional
to the rate of addition of the solution of chlorodifluoro-
acetate salt. Thus, the rates of product formation (k2) and
other carbene processes (k1) converged as temperature
was increased, improving the competitiveness of difluo-
rocyclopropanation. Slower addition rates increased the
instantaneous ratio of difluorocarbene to available olefin
substrate, thus favoring product formation (k2). There
were, of course, practical boundaries on both of these
improvements. Accelerated rate calorimetry experiments
indicated that 13 begins to undergo decomposition at 220
°C. Additionally, the concentration effect attributable to
slow addition rates diminished as the concentration of
12 was reduced by the progress of the reaction. For pilot
plant scale processing we obtained good results at tem-
Syn th esis of LY335979‚3HCl
As a result of the preceding studies we have developed
a convenient, stereoselective, and scalable approach to 1
by incorporating pyrazine in place of piperazine for the
incorporation of the piperazine ring as depicted in
Scheme 4.
Op tim iza tion of Cyclop r op a n a tion . The implemen-
tation of this synthetic plan relied on an efficient and
reliable method for the difluorocylopropanation of diben-
zosuberanone (12). Alternative methods to the previously
disclosed chlorodifluoroacetate pyrolysis5 were examined.
We found that lower-temperature methods, such as
dehalogenation of dibromodifluoromethane with zinc or
triphenylphosphine, and the reaction of methyltrifluo-
romethylstannane with sodium iodide17 all failed to effect
difluorocyclopropanation on the relatively electron-poor
olefinic bond of 12 while reported examples were easily
repeated. We concluded that difluorocyclopropanation of
(18) Detailed analysis of the byproducts was not carried out. The
volatile fraction was presumed to consist of tetrafluoroethylene and,
perhaps, hexafluorocyclopropane, and perfluorocyclobutane. A solid
residue remained in the vessel with empirical formula CF2, indicating
that it was a polymer of tetrafluoroethylene. The various reactions of
difluorocarbene that form these compounds have been documented.
See: Birchall, J . M.; Fields, F.; Haszeldine, R. N.; McLean, R. J . J .
Fluorine Chem. 1980, 15, 487.
(19) Sodium chlorodifluoroacetate was used in earlier examples of
this reaction (ref 5; Beard, C.; Berkoz, B.; Dyson, N. H.; Harrison, I.
T.; Hodge, P.; Kirkham, L. H.; Lewis, G. S.; Giannini, D.; Lewis, B.;
Edwards, J . A.; Fried, J . H. Tetrahedron 1969, 25, 1219 and references
therein). Our preference for the lithium salt was based on its increased
solubility in the reaction medium. The resulting increased reaction
concentration favored product formation and afforded better yields
based on ClF2CCO2Li. (See Experimental Section.)
(15) See ref 11.
(16) Huff, B. E.;, LeTourneau, M. E.; Wilson, T. M.; Bush, J . K,;
Reutzel-Edens, S. M. U.S. Patent 6,521,755, 2003.
(17) For
a review of these and other difluorocyclopropanation
methods, see: Brahms, D. L. S.; Dailey, W. P. Chem. Rev. 1996, 96,
1585.
7656 J . Org. Chem., Vol. 69, No. 22, 2004

Improved Synthesis of LY335979 Trihydrochloride
SCHEME 4
peratures of about 180 °C20 and using a 5.6-fold molar
excess of lithium chlorodifluoroacetate.
reagent for the glycidylation might be questioned because
of its relatively high price. The alkylation of glycidyl
derivatives by arylalkoxide ions is, however, subject to
racemization due to competing Payne-type epoxide open-
ing (followed by reclosure of the epoxide) by intramo-
lecular displacement of the leaving group. As the classic
studies by McClure23 and Sharpless24 elegantly showed,
the amount of racemization observed is inversely de-
pendent upon leaving group reactivity in displacements
of glycidol derivatives. Our own studies25 of the reactions
of chiral glycidol derivatives with 5-hydroxyquinoline
indicated that glycidyl nosylate was the best reagent for
this purpose, in agreement with the Syntex results.
The synthesis was easily completed from the key
intermediates 2 and 3 by a modified version of the Syntex
procedure. Piperazine 2 hydrochloride was converted to
the free base with sodium carbonate in ethanol and
reacted with epoxide 3 in acetonitrile without isolation.
The reaction mixture was filtered through silica gel and
then concentrated, and 1 was obtained as a crystalline
salt by addition of ethanolic HCl in 74% yield.
We were able to reduce the intermediate ketone 13 to
syn -4 without isolation, effecting a substantial improve-
ment in cycle time. Thus, addition of 5.6 equiv of lithium
chlorodifluoroacetate over 11.5 h at 180 °C provided the
desired ketone 13, which was converted to syn -4 in situ
with aqueous sodium borohydride in 70% overall yield
from 4.
Syn th esis of a n ti-2. Reaction of syn -4 with 48% HBr
in dichloromethane with azeotropic removal of water
conveniently provided a solution of bromide 6 in dichlo-
romethane. Addition of pyrazine and DMSO and further
reaction gave crystalline pyrazinium bromide 11, pre-
cipitated from the reaction mixture after dilution with
ethyl acetate in 88% overall yield from syn -4.
Reduction of the pyrazinium ring of 11 was required
to assimilate this molecule into the synthesis of 1.
Catalytic hydrogenation conditions, even in the presence
of added acid, were unsatisfactory as poor yields and
difficult purification of the product were encountered.21
Fortunately, lithium borohydride reduction of 11 in the
presence of trifluoroacetic acid (TFA) gave much better
results. The progress of the reduction and evolution of
hydrogen could be conveniently controlled by the rate of
introduction of borohydride into the reaction mixture. It
was therefore possible to carry out the reaction at pilot
plant scales with adequate control. The crystalline hy-
drochloride of a n ti-2 was obtained by HCl treatment of
the product in ethyl acetate in 70% yield from 11.
F in a l Step s. Epoxide 3 was readily prepared by
reaction of commercially available 5-hydroxyquinoline
(14)22 with glycidyl nosylate by a modified procedure
utilizing potassium tert-butoxide in THF as base. As a
practical matter, the choice of (R)-glycidyl nosylate as a
Su m m a r y
We have established that nucleophilic displacement
reactions on 1,1-difluorodibenzosuberanyl substrates rel-
evant to the synthesis of LY335979 proceed via the
corresponding homotropylium ion intermediate 9. Halo-
genations give predominately or exclusively anti-halide
derivatives under thermodynamic control. Displacements
by cyclic amines provide different stereochemical out-
comes, depending on oxidation state. Piperazine displace-
ments lead to irreversibly formed mixtures of isomers,
(23) McClure, D. E.; Arison, B. H.; Baldwin J . J . J . Am. Chem. Soc.
1979, 101, 3666.
(24) (a) Klunder, J . M.; Onami, T.; Sharpless, K. B. J . Org. Chem.
1989, 54, 4(6), 1295-1304. (b) Sharpless, K. B.; Onami, T. H. PCT
Int. App. WO 8800190 A1, 1988.
(25) Reaction of 5-hydroxyquinoline with (R)-glycidyl nosylate gave
about 2% racemization in our hands as found by chiral HPLC analysis
of the crude product. Optical purity was restored, however, by
crystallization of the product (3) during isolation. Overall stereochem-
ical outcomes for this reaction were in agreement with the Syntex
results.5
(20) We were constrained to reaction temperatures of 180 °C or less
at pilot plant scale because of reaction safety and vessel heating system
limitations. For laboratory experiments a reaction temperature of 210
°C was preferred.
(21) There was some evidence (from analysis of side products) for
self-condensation of partially reduced pyrazine intermediates, but none
of these products were completely characterized.
(22) Purchased from Orgasynth (France).
J . Org. Chem, Vol. 69, No. 22, 2004 7657

Barnett et al.
The mixture was allowed to warm, stirred at ambient tem-
perature for 1 h, and then heated to 60 °C for 4 h. The reaction
was cooled to ambient temperature and concentrated. The
resulting solid was chromatographed over silica gel (heptane-
1,2-dichloroethane 3:2) to provide 5.00 g of a mixture of anti-
and syn-benzoates (anti-syn 2:1, NMR). A 1.5-g portion of the
mixture was combined with 5 mL of 5 N sodium hydroxide
and 10 mL of tetrahydrofuran. The resulting mixture was
heated to 65 °C for 7 h and then cooled to ambient tempera-
ture. The mixture was partitioned by addition of 10 mL of tert-
butyl methyl ether, and the organic phase was dried with
magnesium sulfate and concentrated, affording 1.14 g of white
solid containing the desired anti-benzoate and a mixture of
anti- and syn-alcohols. Chromatography of the mixture (silica
gel, heptane-1,2-dichloroethane 3:2) afforded 302 mg of a n ti-8
as a white solid, mp 162-164 °C: 1H NMR (500 MHz, CDCl3)
δ 8.11 (d, J ) 7.2 Hz, 2H), 7.59 (t, J ) 7.5 Hz, 1H), 7.47 (t, J
) 7.5 Hz, 2H), 7.44 (d, J ) 6.8 Hz, 2H), 7.39 (d, J ) 7.6 Hz,
2H), 7.33 (d, J ) 7.5 Hz, 2H), 7.26 (d, J ) 7.7 Hz, 2H), 6.72 (s,
1H), 3.49 (d, J ) 12.5 Hz, 2H). Anal. Calcd for C23H16F2O2: C,
76.23; H, 4.45. Found: C, 75.92; H, 4.44.
whereas pyridine and pyrazine, on the other hand,
afforded the anti heteroarenium salts with demonstrated
equilibration of the initially predominant syn products.
We have exploited the novel pyrazine displacement
process (6 f 11) and several other innovative improve-
ments to establish a highly efficient and practical route
to LY335995 from dibenzosuberenone (12).
Exp er im en ta l Section
(1a R,6R,10bR)-1,1-Diflu or o-1,1a ,6,10b-tetr a h yd r od iben -
zo[a ,e]cyclop r op a [c]-cycloh ep ten -6-ol (syn -4). A mixture
of dibenzosuberenone (12,5 10.0 Kg, 48.5 mol) and 50 L of
triglyme was heated to 180 °C. To this mixture was added a
50 wt % solution of 65 L (272 mol, 5.6 equiv) of lithium
chlorodifluoroacetate in glyme at a rate of 0.5 equiv/h until
less than 2.0% of the dibenzosuberenone was present by GC
analysis (J W Scientific DB-1 methyl silicate 30 m × 0.25 mm
column, 20 min run time, initial temp 150 °C, final temp 250
°C). The mixture was cooled to 35-45 °C over 2 h and was
treated with 8.0 Kg of Hyflo26 followed by 20 L of ethyl acetate.
After stirring for 30 min, the suspended solids were removed
by filtration through a prewetted (EtOAc) 7.0 Kg Hyflo plug
and washed with 80 L of ethyl acetate. The combined filtrate
and washes were distilled at 25-30 °C and 100 mmHg to a
volume of approximately 60 L. To this solution of ketone 14
was added a 12 wt % solution of sodium borohydride in
aqueous 14 M sodium hydroxide (5.3 Kg, 16.8 mol) over 1 h
while maintaining the temperature below 40 °C. The reaction
mixture was stirred at 20-25 °C for 3 h, at which time GC
analysis (see above for conditions) showed less than 1% of 14.
A mixture of 50 L of water, 35 L of methanol, and 6.2 Kg of
concentrated HCl was added over 40 min while maintaining
the reactor temperature between 25 and 35 °C. The neutralized
mixture was stirred for 45 min, and the precipitate was
filtered, washed twice with 50 L of water followed by 30 L of
methanol, and dried in a filter dryer at 45-50 °C. The crude
product was reslurried in 65 L of methylene chloride for 35
min at 20-25 °C and 1 h at 0-5 °C on the filter dryer. The
solid was filtered, washed with 25 L of cold methylene chloride
followed by 2 × 20 L of cold heptane, and dried in a filter dryer
at 45-50 °C to give 9.5 Kg (75%) of syn -4, mp 233 °C (lit.5b
230.1-230.6 °C): 1H NMR (500 MHz, DMSO-d6) δ 7.61 (d, J
) 7.3 Hz, 2H), 7.24 (m, 4H), 7.20 (m, 2H), 6.72 (br s, 1H), 3.24
(d, J ) 13.4 Hz, 2H), 2.07 (br s, 1H).
(1a R,6â,10bR)-r el-1,1-Diflu or o-1,1a ,6,10b-tetr a h yd r od i-
ben zo[a ,e]cyclop r op a [c]-cycloh ep ten -6-yl Ben zoa te (syn -
8). A solution of 4.19 g (16.2 mmol) of syn -4 in 100 mL of
tetrahydrofuran was treated with 2.0 g (50 mmol, 60% mineral
oil dispersion) of sodium hydride followed by 4.5 g (20 mmol)
of benzoic anhydride. The mixture was heated to 60 °C for 1
h. The reaction mixture was poured into 70 mL of ethyl acetate
and quenched with 70 mL of 1 M hydrochloric acid. The
organic phase was washed with saturated aqueous sodium
carbonate, dried with magnesium sulfate, and concentrated
to a solid. The crude solid was purified by silica gel chroma-
tography using a 3:2 mixture of heptane-1,2-dichloroethane
as eluent to provide 3.04 g (55%) of syn -8 as a white solid, mp
230-231 °C: 1H NMR (500 MHz, CDCl3) δ 8.35 (d, J ) 7.3
Hz, 2H), 7.88 (s, 1H), 7.71 (t, J ) 7.3 Hz, 1H), 7.61 (t, J ) 7.7
Hz, 2H), 7.53 (dd, J ) 5.5, 4.7 Hz, 2H), 7.34 (dd, J ) 6.5, 4.2
Hz, 2H), 7.25 (m, 4H), 3.40 (d, J ) 13.2 Hz, 2H). Anal. Calcd
for C23H16F2O2: C, 76.23; H, 4.45. Found: C, 75.93; H, 4.45.
(1a R,6R,10bR)-r el-1,1-Diflu or o-1,1a ,6,10b-tetr a h yd r od i-
ben zo[a ,e]cyclopr opa[c]-cycloh epten -6-yl Ben zoate (a n ti-
8). A mixture containing 5.16 g (20 mmol) of syn -4, 3.17 g (26
mmol) of benzoic acid, and 6.55 g (25 mmol) of triphenylphos-
phine in 150 mL of tetrahydrofuran was cooled to -30 °C and
treated with 5.05 g (25 mmol) of diisopropyl azodicarboxylate.
(1a R,6â,10bR)-r el-1,1-Diflu or o-1,1a ,6,10b-tetr a h yd r od i-
ben zo[a ,e]cyclop r op a [c]-cycloh ep ten -6-ol (a n ti-4). A mix-
ture of 220 mg (0.66 mmol) of a n ti-8, 2.0 mL of 1 N sodium
hydroxide, and 2.0 mL of methanol was heated to 65 °C for 28
h. The reaction was partitioned by the addition of 8 mL of tert-
butyl methyl ether. The organic layer was dried with sodium
sulfate and concentrated, affording 157 mg (92%) of a n ti-4 as
1
a white solid, mp 160 °C: H NMR (500 MHz, CDCl3) δ 7.36
(d, J ) 7.5 Hz, 2H), 7.28 (m, 2H), 7.22 (m, 4H), 5.51 (br s, 1H),
3.41 (d, J ) 12.6 Hz, 2H), 2.60 (br s, 1H). Anal. Calcd for
C
16H12F2O: C, 74.41; H, 4.68. Found: C, 74.12; H, 4.72.
(1a R,6â,10bR)-6-Br om o-1,1-d iflu or o-1,1a ,6,10b-tetr a h y-
d r o-d ib en zo[a ,e]cyclop r op a [c]cycloh ep t en e (6). PBr3
method. To a mechanically stirred suspension of syn -4 (18.4
g, 71.2 mmol) in 150 mL of dichloromethane that had been
cooled to 10-17 °C was added phosphorus tribromide (9.6 g,
35.6 mmol) dropwise over 15 min. The cooling bath was
removed, and the reaction mixture was stirred for 2 h at
ambient temperature. Analysis by gas chromatography indi-
cated complete consumption of 4. Cold water (92 mL) and
activated carbon (1.84 g) were added, and the resulting
biphasic mixture was stirred for 30 min. The activated carbon
was removed by filtration through Hyflo, and the layers were
allowed to separate. The organic layer was washed twice with
184 mL of water and once with 184 mL of brine, dried over
magnesium sulfate, and concentrated in vacuo to give 21.7 g
(94.8%) of compound 6 as a light yellow foam. A 2.0-g portion
was recrystallized from heptane, mp 120-122 °C: 1H NMR
(300 MHz, CDCl3) δ 3.36 (s, 1H), 3.40 (s, 1H), 5.77 (s, 1H),
7.16-7.38 (m, 8H). Anal. Calcd for C16H11BrF2: C, 59.84; H,
3.45; Br, 24.88. Found: C, 59.96; H, 3.62; Br, 24.63.
Gen er a l P r oced u r e for Stu d ies of th e Rea ction of 6
w ith P ip er a zin e. A mixture of compound 6 (50.0 mg, 0.156
mmol), piperazine (44.0 mg, 0.511 mmol), and the appropriate
solvent (1.0 mL) was heated to 50 °C (thermostated oil bath)
for 18 h. The mixture was cooled to room temperature,
sampled, and analyzed by HPLC (Zorbax SB-CN column, λ )
225 nm, 1.0 mL/min, 40 °C, 25 mM NaH2PO4, 0.1% TEA, pH
2.5) to determine isomer ratios. Results are shown in Table 1.
(1a R,6R,10bR)-1-(1,1-Diflu or o-1,1a ,6,10b-tetr a h yd r od i-
b en zo[a ,e]cyclop r op a [c]-cyclo-h ep t en -6-yl)p yr id in iu m
Br om id e (a n ti-10). A mixture of 200 mg (0.623 mmol) of 6
and 3 mL of pyridine was heated at 50 °C for 5 h. The reaction
was allowed to cool, effecting a (partial) crystallization of the
product. The reaction was diluted with 5 mL of tert-butyl
methyl ether (MTBE), and the crystals were collected by
filtration, washed with MTBE, and dried to yield 260 mg
(100%) of a n ti-10 (monohydrate) as a tan solid, mp 207-208
°C: 1H NMR (500 MHz, CDCl3) δ 8.65 (d, J ) 6.6 Hz, 2H),
8.54 (t, J ) 7.6 Hz, 1H), 8.09 (t, J ) 7.2 Hz, 2H), 8.00 (m, 3H),
7.45 (m, 2H), 7.40 (m, 4H), 2.65 (d, J ) 12.6 Hz, 2H). Anal.
(26) A diatomaceous silica filter aid.
7658 J . Org. Chem., Vol. 69, No. 22, 2004

Improved Synthesis of LY335979 Trihydrochloride
Calcd for C21H16BrF2N‚H2O: C, 60.30 H, 4.34, N, 3.35. Found
were added, and the mixture was stirred for 30 min at 20-25
°C. The resulting inorganic precipitate was filtered and washed
with 200 L of toluene. The combined toluene fractions were
washed with a solution of 5.1 Kg (63.7 mol) of 50% aqueous
NaOH in 330 L of water, and the aqueous layer was back-
extracted with 130 L of toluene. The combined toluene layers
were washed with 130 L of water, and approximately 160 L
of toluene was distilled off to effect azeotropic drying of the
solution. Basic alumina (16.5 Kg) was added to the dried
yellow-orange toluene solution, and the mixture was stirred
for 30 min at 20-25 °C. The alumina was filtered and washed
with 200 L of toluene. The combined filtrate and wash were
concentrated (50 mmHg) to about 100 L and heated to 45-55
°C. Heptane (500 L) was then added at such a rate to maintain
the temperature at >45 °C. The solution was allowed to cool
to 40 °C and was seeded. The crystallizing mixture was cooled
to 0 °C over 2 h and then stirred at 0-5 °C for an additional
2 h. The crystals were filtered, washed with 100 L of heptane,
and dried in vacuo (25 °C) to provide 19.1 Kg (74%) of 3 as a
crystalline solid, mp 79-81 °C, 98.6% ee (chiral HPLC):27 1H
NMR (500 MHz, CDCl3) δ 2.83 (dd, J ) 4.8, 2.7 Hz, 1H), 2.97
(m, 1H), 3.48 (m, 1H), 4.10 (dd, J ) 11.0, 6.0 Hz, 1H), 4.43
(dd, J ) 11.0, 2.7 Hz, 1H), 6.85 (d, J ) 7.8 Hz, 1H), 7.38 (dd,
J ) 8.5 Hz, 4.1 Hz, 1H), 7.59 (m, 1H), 7.71 (d, J ) 8.5 Hz,
1H), 8.61 (m, 1H), 8.90 (m, 1H); 13C NMR (CDCl3, 75.5 MHz)
δ 153.9, 150.8, 149.1, 130.8, 129.2, 122.2, 120.8, 120.3, 105.3,
69.3, 50.0, 44.6.
[6(R)-(1a R,6R,10bR)]-4-(1,1-Diflu or o-1,1a ,6,10b)tetr a h y-
dr odiben zo[a ,e]cyclopr opa[c]cycloh epten -6-yl)-R-[(5-qu in -
olin yloxy)m eth yl]-p ip er a zin eeth a n ol, Tr ih yd r och lor id e
Sa lt (1). A suspension of 20.3 Kg of 2 hydrochloride (87% as
2, 48.9 mol) and 10.3 Kg (99.7 mol) of powdered sodium
carbonate in 340 L of ethanol was stirred at 20-25 °C for 1.5
h. Addition of 9.83 Kg (48.9 mol) of 3 was completed in one
portion, and the reaction mixture was heated to 60-70 °C for
36 h. The mixture was cooled to 20-25 °C, filtered through a
46-Kg plug of silica gel, and eluted with an additional 400 L
of ethanol. The combined filtrate and washes were distilled
at 45-55 °C (170-180 mmHg) to a volume of approximately
265 L and then heated to 60-70 °C with stirring. An 88-Kg
solution of HCl (3.28 N in ethanol) was added over 35 min
and transferred with 10 L of anhydrous ethanol. The solution
was seeded, causing the trihydrochloride salt to precipitate.
The mixture was cooled to 20-25 °C over 2 h and stirred at
that temperature for 15 h. The precipitate was filtered onto a
filter dryer, washed with 105 and 30 L of ethanol, and partially
dried on the filter. Methanol (200 L) was heated to 50-60 °C
and added to the wet cake to dissolve the crude product on
the filter and transfer the solution to another vessel. An
additional 170 L portion of methanol was used to complete
the transfer. The solution was distilled at 35-45 °C and 200-
300 mmHg to a volume of approximately 200 L. The mixture
was again heated to 60-65 °C for 3 h to dissolve all solids
and then was cooled to 20-25 °C over 30 min. The suspension
was stirred for 3 h, and then 590 L of ethyl acetate were added
to complete the crystallization. After stirring for 14 h at 20-
25 °C, the precipitate was filtered, washed with 210 L of ethyl
acetate, and dried in vacuo at 45 °C to provide 26.0 Kg (74.6%)
of 1 as a light yellow crystalline methanol solvate. The solvate
was converted to a crystalline hydrate (xH2O) by equilibration
in moist nitrogen. The hydrated crystals were correlated by
physical and pharmacological data with material prepared at
the Syntex laboratories.
C, 60.68; H, 4.35, N, 3.37.
(1a R,6R,10bR)-1-(1,1-Diflu or o-1,1a ,6,10b-tetr a h yd r od i-
b en zo[a ,e]cyclop r op a [c]-cycloh ep t en -6-yl)p yr a zin iu m
Br om id e (11). To a stirred slurry of 27.3 Kg (105.7 mol) of
syn -4 in 220 L of dichloromethane was added 24.1 Kg (143.0
mol) of 48% aqueous HBr. The resulting slurry was heated to
reflux for 22 h with azeotropic removal of water via a Dean-
Stark separator. After the mixture cooled to 20-25 °C, 14.0
Kg of anhydrous sodium sulfate and 7.0 Kg of activated carbon
were added. The mixture was stirred at 20-25 °C for 1 h and
filtered with Hyflo, and the filter cake was washed with 2 ×
60 L of dichloromethane. The combined filtrate and washes
were distilled (40-45 °C, atmospheric pressure) to provide a
solution of 6 in 150 L of dichloromethane. To this solution was
added 32.4 Kg (404.5 mol) of pyrazine followed by 16.5 L of
dimethyl sulfoxide, and the resulting solution was heated to
30 °C for 25 h. The resulting yellow-orange suspension was
diluted with 180 L of ethyl acetate, and the mixture was stirred
at 20-25 °C for 2 h to complete crystallization of the product.
The crystals were filtered, washed with 180 L of ethyl acetate,
and dried in vacuo at 40 °C to provide 35.5 Kg (88%) of 11 as
a light yellow crystalline solid, mp 165 °C: 1H NMR (500 MHz,
DMSO-d6) δ 9.46 (d, J ) 2.7 Hz, 2H), 8.77 (t, J ) 1.7 Hz, 2H),
7.74 (d, J ) 7.5 Hz, 2H), 7.42-7.55 (m, 6H), 7.27 (s, 1H), 3.19
(d, J ) 12.6 Hz, 2H); 13C NMR (126 MHz DMSO-d6) δ 152.5,
135.8, 135.4, 134.4, 133.2, 132.1, 129.9, 129.6, 112.9, 110.6,
108.3, 77.4, 28.7, 28.6; FD MS m/e 321 (M - Br). Anal. Calcd
For C20H15BrF2N2: C, 59.87; H, 3.77; N, 6.98. Found: C, 59.84;
H, 3.66; N, 6.83.
(1a R,6R,10bR)-1-(1,1-Diflu or o-1,1a ,6,10b-tetr a h yd r od i-
ben zo[a ,e]cyclopr opa[c]-cycloh epten -6-yl)-piper azin e Hy-
d r och lor id e (a n ti-2 HCl). To stirred slurry of 35.4 Kg of 13
(88.2 mol) in 185 L of EtOAc was added trifluoroacetic acid
(34.6 Kg, 303.5 mol). To this mixture was added a 2.0 M
solution of lithium borohydride in THF (59.6 Kg, 66.5 L, 133.0
mol) over 1.5 h while maintaining the reactor temperature
between 10 and 20 °C. The reaction mixture was stirred at
20-25 °C for 2 h. The reaction solution was added carefully
to 351 Kg of 10% (w/w) aqueous potassium carbonate solution
over 15 min while maintaining the vessel temperature below
25 °C. The mixture was stirred at 20 °C for 30 min, and the
layers were separated. The organic phase was washed with 2
× 300 Kg of 10% (w/w) aqueous potassium carbonate solution
and once with 185 Kg of 15% (w/w) aqueous sodium chloride
solution and dried over 53.4 Kg of anhydrous sodium sulfate.
After filtering the drying agent, 8.8 Kg (1.0 equiv) of 37%
aqueous HCl was added (15-20 °C) over 35 min, and the
resulting slurry was stirred at 20 °C for 9 h. The precipitate
was filtered, washed with 100 L of ethyl acetate, and dried in
vacuo (45 °C) to provide 24.7 Kg of anti-2 HCl (85.6%, or 21.1
Kg as anti-2 (HPLC assay), 66% yield), mp 274-278 °C (dec):
1H NMR (500 MHz, DMSO-d6) δ 9.41 (br s, 2H), 7.17-7.31
(m, 8H), 4.17 (s, 1H), 3.52 (d, J ) 12.4 Hz, 2H), 3.11 (br s,
4H), 2.48-2.51 (m, 4H); 13C NMR (126 MHz, DMSO-d6) δ
142.3, 133.4, 130.5, 129.6, 129.0, 128.4, 115.9, 113.6, 111.3,
76.2, 49.0, 43.6, 29.2, 29.1, 29.0; FD MS m/e 326 (M+). Anal.
Calcd for C20H21ClF2N2: C, 66.20; H, 5.83; N, 7.72. Found: C,
66.08; H, 5.90; N, 7.72.
5-[(2R)-Oxir a n ylm eth oxy]-qu in olin e (3). To a 0 °C sus-
pension of 18.5 Kg (127.3 mol) of 5-hydroxyquinoline in 231 L
of THF was added 15.1 Kg (127.3 mol) of potassium tert-
butoxide (95% potency) in three portions, and the resulting
thick slurry was stirred for 1 h at 20-25 °C. A solution of 33.0
Kg (127.3 mol) of (R)-glycidyl nosylate in 83 L of THF was
then added over 1 h. The reaction mixture was stirred for 16
h at 23-26 °C, at which time HPLC analysis indicated that
the reaction was complete. The reaction mixture was concen-
trated (50-100 mmHg) to about 200 L. A 440-L quantity of
toluene was added, and the mixture was concentrated (50
mmHg) to a volume of about 200 L. Another 130 L of toluene
Ack n ow led gm en t. We gratefully acknowledge the
contributions made by J ason Cronin, Ch.E., J ennifer
Flynn, Ch.E., Brittany Long, Ch.E., Mark Pittman,
Ch.E., Bruce Parker, Ch.E., and J eff Vicenzi, Ch.E.
(27) A sample prepared under similar conditions showed an ee of
99.2% with an optical rotation of [R]25D -36.4 (c 2.10, EtOH), lit.7 [R]23
-35.6.
D
J . Org. Chem, Vol. 69, No. 22, 2004 7659

Barnett et al.
Su p p or tin g In for m a tion Ava ila ble: Table of 1H NMR
data for 1,1-difluoro-1,1a,6,10b-tetrahydrodibenzo-[a,e]cyclo-
propa[c]-cyclohepten-6-yl derivatives, 1H NMR spectra for syn
and anti compounds 4 and 8 and NOE spectra for syn and
anti compounds 8 and 10, and a table relating efficiency for
the conversion lithium chlorodifluoroacetate to temperature
in difluorocyclopropanation. This material is available free of
charge via the Internet at http://pubs.acs.org.
J O049051V
7660 J . Org. Chem., Vol. 69, No. 22, 2004