A practical asymmetric synthesis of the antiviral agent lobucavir, BMS-180194

Article abstract of DOI:10.1021/op970214+

A practical synthesis of the antiviral agent lobucavir, [1R-(1α,2β,3α)]-2-amino-9-[2,3-bis(hydroxymethyl)cyclobutyl]- 6H-purin-6-one (BMS-180194), is described. The key chiral intermediate, [1S-(1α,2β,3α)]-3-hydroxy-1,2-cyclobutanedimethanol, dibenzoate e

Full text of DOI:10.1021/op970214+

Organic Process Research & Development 1998, 2, 393−399
A Practical Asymmetric Synthesis of the Antiviral Agent Lobucavir,
BMS-180194
Janak Singh,*,† Gregory S. Bisacchi,‡ Saleem Ahmad,‡ Jollie D. Godfrey, Jr.,† Thomas P. Kissick,† Toomas Mitt,‡
Octavian Kocy,‡ Truc Vu,† Chris G. Papaioannou,† Michael K. Wong,† James E. Heikes,† Robert Zahler,‡ and
Richard H. Mueller†
The Bristol-Myers Squibb Pharmaceutical Research Institute, P.O. Box 4000, Princeton, New Jersey 08543-4000
Abstract:
A practical synthesis of the antiviral agent lobucavir, [1R-
(1r,2â,3r)]-2-amino-9-[2,3-bis(hydroxymethyl)cyclobutyl]-6H-
purin-6-one (BMS-180194), is described. The key chiral inter-
mediate, [1S-(1r,2â,3r)]-3-hydroxy-1,2-cyclobutanedimethanol,
dibenzoate ester, was made by an asymmetric [2 + 2] cyclo-
addition of dimenthyl fumarate with ketene dimethyl acetal
followed by sequential diester reduction, benzoylation, deket-
alization, and stereoselective ketone reduction. Regioselective
N9-alkylation of the tetra-n-butylammonium salt of 2-amino-
6-iodopurine with the derived cyclobutyltriflate furnished the
purinecyclobutyl dibenzoate. Methanolysis followed by acid
hydrolysis produced lobucavir in a 35% overall yield with an
ee > 99%.
Figure 1. BMS 180194 (lobucavir).
Scheme 1
Introduction
Lobucavir, BMS-180194 (Figure 1),1-7 has potent activity
against a variety of herpes family viruses (herpes simplex
virus, human cyctomegalovirus, varicella-zoster virus), as
well as hepatitis B virus and human immunodeficiency virus.
Lobucavir presently is undergoing clinical trials, and we have
developed a practical synthesis to produce multikilogram
quantities of this potential medicinal agent as a single
enantiomer.
lengthy or use environmentally unfriendly reagents, e.g.,
dithioketene acetal or dinitrogen tetraoxide. In some cases,
the reactions are difficult to scale-up (e.g., ozonolysis,
photolysis). In Ichikawa’s group’s4 synthesis of lobucavir,
the key step is an asymmetric [2 + 2] cycloaddition of
thioketene dimethyl acetal with the fumarate ester 1a using
a chiral titanium catalyst (11% yield in 12 steps, Scheme
1). The asymmetric synthesis of lobucavir reported by Pariza
et al.5 employed dimenthyl fumarate 1b and thioketene
dimethyl acetal to produce cyclobutanone 3a in 27% overall
yield. Intermediate 3a was then converted to a cyclobuty-
lamine for the stepwise construction of the guanine moiety.
Cotterill and Roberts9 prepared mesylate 6 via a photochemi-
cal rearrangement of bicyclic epoxycyclobutanone 4 in 3%
overall yield over 11 steps (Scheme 2). Hsiao and Hannick10
obtained chiral cyclobutanone 3b from R,R-(+)-diethyl
tartrate via the key oxirane intermediate 7 in 32% overall
yield over 10 steps. The corresponding bis-(t-BuPh2Si) ether
Literature methods for the synthesis of substituted cyclo-
butanes are not suitable for the multikilogram scale prepara-
tion of lobucavir.1-13 The synthetic sequences are often
† Chemical Process Research
‡ Medicinal Chemistry Division
(1) (a) Slusarchyk, W. A.; Young, M. G.; Bisacchi, G. S.; Hockstein, D. R.;
Zahler, R. Tetrahedron Lett. 1989, 30, 6453. (b) Norbeck, D. W.; Kern, E.;
Hayashi, S.; Rosenbrook, W.; Sham, H.; Herrin, T.; Plattner, J. J.; Erickson,
J.; Clement, J.; Swanson, R.; Shipkowitz,, N.; Hardy, D.; Marsh, K.; Arnett,
G.; Shannon, W.; Broder, S.; Mitsuya, H. J. Med. Chem. 1990, 33, 1281.
(2) Bisacchi, G. S.; Braitman, A.; Cianci, C. W.; Clark, J. M.; Field, A. K.;
Hagen, M. E.; Hockstein, D. R.; Malley, M. F.; Mitt, T.; Slusarchyk, W.
A.; Sundeen, J. E.; Terry, B. J.; Toumari, A. V.; Weaver, E. R.; Young, M.
G.; Zahler, R. J. Med. Chem. 1991, 34, 1415 and cited references.
(3) Bisacchi, G. S.; Singh, J.; Godfrey, J. D., Jr.; Kissick, T. P.; Mitt, T.; Malley,
M. F.; Di Marco, J. D.; Gougoutas, J. Z.; Mueller, R. H.; Zahler, R. J. Org.
Chem. 1995, 60, 2902.
(4) Ichikawa, Y.-I.; Narita, A.; Shiozawa, A.; Hayashi, Y.; Narasaka, K. J. Chem.
Soc., Chem. Commun. 1989, 1919.
(5) Pariza, R. J.; Hannick, S. M.; Sowin, T. J.; Doherty, E. M. EP 452729
October 23, 1991.
(6) Ahmad, S. Tetrahedron Lett. 1991, 32, 6997 and references therein.
(7) Flaherty, J. F., Jr. Am. J. Health-Syst. Pharm. 1996, 53 (Suppl. 2), S4.
(8) (a) Agrofoglio, L.; Suhas, E.; Farese, A.; Condom, R.; Challand, R.; Earl,
R. A.; Guedj, R. Tetrahedron 1994, 50, 10611 and references therein. (b)
Borthwick, A. D.; Biggadike, K. Tetrahedron 1992, 48, 571. (c) Huryn, D.
M.; Okabe, M. Chem. ReV. 1992, 92, 1745.
(9) Cotterill, I. C.; Roberts, S. M. J. Chem. Soc., Perkin Trans. 1 1992, 2585.
(10) (a) Hsiao, C.-N.; Hannick, S. M. Tetrahedron Lett. 1990, 31, 6609. (b)
Doering, W. von E.; Roth, H. D. Tetrahedron, 1970, 26, 2825-35. (c) For
an improved method of resolution of the Feist’s acid refer patent, see:
Godfrey, J. D., Jr.; Mueller, R. H.; Kissick, T. P.; Singh, J. US 5,525,726,
June 11, 1996.
(11) Hanzawa, Y.; Ito, H.; Taguchi, T. Synlett 1995, 299.
(12) Chen, X.; Siddiqi, S. M.; Schneller, S. W. Tetrahedron Lett. 1992, 33, 2249.
(13) Jung, M. E.; Sledeski, A. W. J. Chem. Soc., Chem. Commun. 1993, 589.
10.1021/op970214+ CCC: $15.00 © 1998 American Chemical Society and Royal Society of Chemistry
Published on Web 10/07/1998
Vol. 2, No. 6, 1998 / Organic Process Research & Development
•
393

Scheme 2
Scheme 6
Scheme 3
Scheme 7
Scheme 4
Scheme 5
3c was also made by them via Feist’s acid route employing
a cumbersome chemical resolution (Scheme 3). Taguchi and
co-workers11 reported a nine-step synthesis of the cyclobu-
tanol 10a (20% overall yield) from a sugar derivative
involving a rearrangement of the vinylfuranoside 8 to the
cyclobutanol intermediate 9 (Scheme 4). Schneller and co-
workers12 made intermediate 10a and its antipode by the
enzymatic hydrolysis of the corresponding racemic acetate.
Jung et al.13 reported a chiral synthesis of cyclobutene 13
(Scheme 5) from the photocycloadduct 11 employing an
enzymatic resolution followed by epimerization of a half acid
ester intermediate. Adenine was introduced by the opening
of an epoxide, and the extra hydroxyl group on the
cyclobutane ring was removed by the free radical decom-
position of a thiocarbonate derivative.
At present, the two most promising routes for the
construction of the cyclobutane ring of lobucavir are ring
expansion of resolved Feist’s acid and asymmetric [2 + 2]
cycloaddition of a ketene acetal with a fumaric acid deriva-
tive. Slusarchyk et al.1a reported an initial Bristol-Myers
Squibb synthesis (Scheme 6) of racemic lobucavir (5.4%
overall yield in 11 steps). Racemic cyclobutanone 3,
prepared via an achiral [2 + 2] cycloaddition, was converted
to the tosylate 10 and then coupled with 2-amino-6-
benzyloxypurine 14. Bissachi et al.2 subsequently described
the synthesis of optically pure lobucavir (Scheme 7) (1.5%
overall yield in 13 steps) that utilized a chemical resolution.
This synthesis was later modified considerably to address
several problems. The synthesis of 1S,2S-cyclobutanone 3e
via the chemical resolution of cyclobutanedicarboxylic acid
diamide 16 required a prohibitively expensive reagent, (R)-
2-phenylglycinol, and used the highly toxic chemicals
dinitrogen tetraoxide and carbon tetrachloride. The highly
stereoselective reduction of the cyclobutanone 3e to the
1S,2S,3S-cyclobutanol 10a was done with the expensive
reagent LS-Selectride, and this reaction also produced the
rearranged side product 10b. Coupling of the purine
derivative 14 with the cyclobutyltosylate 10c was the most
problematic step in this synthesis due to formation of a
cyclobutene elimination product and the corresponding N7-
alkylation product. Intermediate 10a and the final product
lobucavir required column chromatography for their purifica-
tion.
Ahmad6 subsequently disclosed in a preliminary com-
munication that the asymmetric [2 + 2] cycloaddition of
ketene dimethyl acetal with dimenthyl fumarate 1b in the
presence of a Lewis acid produced the cycloadduct 17 with
high diastereoselectivity (Scheme 8). Cycloadduct 17 was
394
•
Vol. 2, No. 6, 1998 / Organic Process Research & Development

applicable to different systems.8,16-18 We reported in a
preliminary communication the reaction of the tetra-n-
butylammonium salt of 2-amino-6-iodopurine (19a) with the
cyclobutyl triflate 10e to form the dibenzoate 20b in high
yield with 93% N9-regioselectivity and its subsequent
conversion to the final product.3 In this paper we provide
the full description of this synthesis of lobucavir, obtained
in 35-40% overall yield in 10 steps.
Scheme 8
Results and Discussion
A practical synthesis of lobucavir (BMS 180194) was
achieved via a convergent approach (Scheme 8) in four
phases. The optically pure 2S,3S-cyclobutanone 3e was
prepared via the diastereoselective [2 + 2] cycloaddition of
ketene dimethyl acetal with dimenthyl fumarate 1b followed
by reduction, benzoylation, and hydrolysis of the ketal. The
second stage involved stereoselective reduction of cyclo-
butanone 3e to the 1S,2S,3S-cyclobutanol 10a. In the next
phase, purine cyclobutyldibenzoate 20b was prepared by the
regioselective N9-alkylation of the n-tetrabutylammonium
salt of 2-amino-6-iodopurine (19a) with the cyclobutyltriflate
10e. Finally, we developed a single-step conversion of the
intermediate 20b to lobucavir.
Synthesis of Cyclobutanone 3e via Diastereoselective
[2 + 2] Cycloaddition. Diastereoselective [2 + 2] cyclo-
addition of l-dimenthyl fumarate 1b with ketene dimethyl
acetal catalyzed by diisobutylaluminum chloride (DIBAC)
(toluene, -70 °C) produced the cycloadduct 17 in 98% yield
and 90% diastereomeric excess.6 Recrystallization from
MeOH furnished pure 17 in 83% yield (de > 99%). Thus,
we have replaced the less desirable dithioketene acetal by a
diketene acetal in the diastereoselective [2 + 2] cycloaddition
reaction. Two mole equivalents of the Lewis acid was
essential for this reaction. The catalyst was precomplexed
with dimenthyl fumarate, and then ketene dimethyl acetal was
introduced at below -70 °C to minimize polymerization of
the latter. The catalyst considerably enhanced the rate of
the reaction. In the absence of the catalyst, no cycloaddition
occurred at ambient temperature, and a complex mixture of
products was obtained at 80 °C. The reaction is inefficient
at temperatures above -60 °C due to the polymerization of
the ketene dimethyl acetal, although facial selectivity in this
reaction is maintained even at -40 °C. The synergistic
stereodifferentiating influence of the two chiral auxiliaries
in diester 1b resulted in this remarkable facial selectivity,
which is reminiscent of some [4 + 2] Diels-Alder cyclo-
addition reactions.19-21
converted to the key cyclobutanone intermediate 3e via the
reduction of ester groups, dibenzoylation, and deacetalization.
The methods generally used for the transformation of
cyclobutanones to the purine derivatives proceed via either
cyclobutylamine or a cyclobutanol intermediate.8 Since the
synthesis via cyclobutylamine requires a stepwise construc-
tion of the guanine moiety, the more convergent approach
involving N9-alkylation of a purine derivative with cyclo-
butanol or an activated derivative is often preferred. 2,3-
Disubstituted cyclobutanones are generally reduced to the
desired 1,2-cis cyclobutanols either by enzymes14 or with
sterically hindered reducing agents, e.g., LS-Selectride.1 The
coupling of cyclobutanols with purine derivatives by the
Mitsunobu method is generally inefficient and fraught with
purification difficulties.15 The regioselective N9-alkylation
of the 6-substituted purines with the O-activated cyclo-
butanols is also inefficient due to the competing N7-
alkylation. Although several approaches have been adopted
to tackle this historical problem, there is no general method
Several other Lewis acids were examined for the cyclo-
addition reaction. Et2AlCl was the only other catalyst that
formed the cycloadduct 17 in yield and diastereoselectivity
(16) (a) Trost, B. M.; Madsen, R.; Guile, S. G.; Elia, A. E. H. Angew. Chem.,
Int. Ed. Engl. 1996, 35, 1569. (b) Robins, M. J.; Zou, R.; Guo, Z.; Wnuk,
S. F. J. Org. Chem. 1996, 61, 9207.
(17) Clausen, F. P.; Juhl-Christensen, J. Org. Prep. Proced. Int. 1993, 25, 373.
(18) (a) Jones, M. F.; Roberts, S. M. J. Chem. Soc., Perkin Trans. 1 1988, 2927.
(b) Geen, G. R.; Grinter, T. J.; Kincey, P. M.; Jarvest, R. L. Tetrahedron
1990, 46, 6903. (c) Rasmussen, M.; Chan, H.-S. Aust. J. Chem. 1975, 28,
1031.
(14) Ichikawa, J.; Shimada, K.; Konakawa, O.; Shiozawa, A. JP 05103679,
October, 9, 1991; Chem. Abstr. 1993, 119, 137552b.
(15) (a) Wachtmeister, J.; Classon, B.; Samuelsson, B. Tetrahedron 1997, 53,
1861. (b) Jenny, T. F.; Schneider, K. C.; Benner, S. A. Nucleosides
Nucleotides 1992, 11, 1257.
(19) Corey, E. J.; Wei-guo, S. Tetrahedron Lett. 1988, 28, 3423.
(20) Walborsky, H. M.; Barash, L.; Davis, T. C. Tetrahedron 1963, 19, 2333.
(21) Furuta, K.; Iwanaga, K.; Yamamoto, H. Tetrahedron Lett. 1986, 27, 4507.
Vol. 2, No. 6, 1998 / Organic Process Research & Development
•
395

Scheme 9
comparable to that of DIBAC. The homogeneous complexes
of Et2AlCl and DIBAC with dimenthyl fumarate produced
the best results. Similar observations were made by Wal-
borsky et al.20 and Yamamoto et al.21 for Lewis acid-
catalyzed Diels-Alder reactions. Catalysts that did not form
homogeneous complexes with dimenthyl fumarate, e.g., BF3,
TiCl4, SnCl4, or AlCl3, led to extensive decomposition of
the ketene dimethyl acetal.
Reduction of the cycloadduct 17 with LAH formed the
diol 18a in quantitative yield. The chiral auxiliary l-menthol
was easily recovered by partitioning the quenched reaction
mixture between heptane and water. Benzoylation of the
diol 18a to 18b followed by acid hydrolysis of the ketal
furnished crystalline cyclobutanone 3e in 90% overall yield.
The alternate protecting groups, t-BuMe2Si or benzyl,
produced the corresponding cyclobutanone diethers 3a and
3b as oils, and these were considered unsuitable for purifica-
tion on a large scale.
selectivity. Reduction of the tert-butyldimethylsilyl-protected
cyclobutanone 3a and the dipivaloyl ester 3f with DIBAC
formed the corresponding 1,2-cis cyclobutanols with 91 and
95% diastereoselectivity, respectively. Reduction of 3e with
isobutylaluminum dichloride formed 10a with high stereo-
selectivity, but, presumably because of enolization, 10% of
the unreacted cyclobutanone 3e was obtained. Cyclo-
butanone 3e also was reduced with IrCl4 [H3PO3, HCl, EtOH,
reflux]24 with high stereoselectivity (98% 10a), but the yield
was very low. As expected, in analogy with the substituted
cyclohexanones, the reduction of 3e with sterically less
demanding reducing agents, e.g., LAH or NaBH4, produced
predominantly the more stable 1,2-trans epimer 10d.
Cis-2,3-disubstituted cyclobutanone 23 was prepared for
a comparison of the stereochemistry of reduction with
DIBAC (Scheme 9).25 The differing stereoselectivities were
striking. Reduction of 23 with DIBAC was almost stereo-
specific, producing the less stable all-cis epimer 24 (99%
diastereoselectivity).26 All-cis stereoreoselectivity also was
obtained during the reduction of 23 with LiAl(OtBu)3H to
24 (83% diastereoselectivity).27a Reduction of trans-2,3-
disubstituted cyclobutanone 3e with LiAl(OtBu)3H formed
the more stable epimer 10d with 93% diastereoselectivity.
Stereoselective Reduction of Cyclobutanone 3e to the
Cyclobutanol 10a. Slusarchyk et al.1a and Bisacchi et al.2
reported the reduction of 2,3-trans disubstituted cyclobu-
tanone 3e to 1S,2S,3S-cyclobutanol 10a with LS-Selectride
in excellent diastereoselectivity. However, LS-Selectride is
available only as a solution in THF and is very expensive.
The process typically produced 5-10% of the undesired
rearrangement product 10b (Scheme 7), and the workup of
the reaction was cumbersome for large scale synthesis.
Reduction of 3e with L-Selectride was unsatisfactory, as it
produced 10a with only 77% diastereoselectivity.1a We
identified DIBAC as an alternative reagent for a more
practical diastereoselective synthesis of cyclobutanol 10a.
DIBAC has been used only rarely for the reduction of
cyclic ketones. Ashby and Noding22 studied the addition
and reduction reactions of alkylaluminum halides, including
Et2AlCl, with 4-tert-butyl cyclohexanone. Giacomelli and
co-workers23 described the reduction of acyclic and a few
cyclic ketones with DIBAC and chiral alkylaluminum
reagents. We have discovered that the reduction of trans-
dibenzoyl cyclobutanone 3e with DIBAC in CH2Cl2 at -35
°C followed by a quench with MeOH and aqueous NH4Cl
at low temperature furnished carbinol 10a with 92% dia-
stereoselection. It is noteworthy that quenching of the
reaction at low temperature completely suppressed re-
arrangement of the benzoyl group from its primary position
in 10a to a secondary site in 10b. Simple workup followed
by recrystallization gave product 10a with high purity (ee
> 99%) and in 70-80% yield. Furthermore, the wrong
isomer 10d can be recycled by Swern oxidation back to
cyclobutanone 3e. This reduction method was used to
prepare several kilograms of 10a. The reduction of 3e with
DIBAC in toluene produced 10a with 90% diastereoselec-
tivity. The reduction was not efficient in THF. Toluene is
always preferred over CH2Cl2 as a solvent for enviromental
safety considerations. However, CH2Cl2 is the solvent of
choice for this reduction as the reactions could be easily
scaled-up to produce 10a with reproducibly high stereo-
(24) Eliel, E. L.; Doyle, T. W.; Hutchins, R. O.; Gilbert, E. C. Organic Syntheses;
Wiley: New York, 1988; Collect. Vol. VI, p 215 and cited references.
(25) (a) Racemic 2,3-cis cyclobutanone dibenzoate 23 was prepared from the
Feist’s acid (i) via cis-anhydride (ii)25b by modified procedures. Esterification
of the cis-anhydride (ii) [in situ, (MeO)3CH, MeOH, H2SO4, yield 64%]
followed by reduction (LiAlH4, THF) to diol iiib25c and benzoylation
[(PhCO)2O, Et3N, DMAP, EtOAc] produced iiic in 78% yield. Olefin iiic
was epoxidized (MCPBA, CH2Cl2, yield 99%) and then rearranged (LiI,
EtOAc, yield 99%) to cyclobutanone 23.8 (b) Ettlinger, M. G.; Kennedy,
F. Chem. Ind., 1957, 891. (c) Gajewski, J. J. J. Am. Chem. Soc. 1971, 93,
4450.
(26) (a) Haubenstock, H.; Eliel, E. L. J. Am. Chem. Soc., 1962, 84, 2363. (b)
Haubenstock, H. J. Org. Chem. 1973, 38, 1765.
(27) (a) Dehmlow, E. V.; Buker, S. Chem. Ber. 1993, 126, 2759. (b) Gung, B.
W. Tetrahedron 1996, 52, 5263 and cited references.
(22) Ashby, E. C.; Noding, S. A. J. Org. Chem. 1979, 44, 4792.
(23) Giacomelli, G.; Lardicci, L.; Palla, F.; Caporusso, A. M. J. Org. Chem.
1984, 49, 1725.
396
•
Vol. 2, No. 6, 1998 / Organic Process Research & Development

In cis-isomer 23, the R-face of the carbonyl is severely
blocked by a bulky pseudoaxial group at the C3 position;
therefore, attack from the â-face is kinetically preferred.
Besides this overwhelming steric effect, a conformational
change of the carbonyl plane in cyclobutanone 23 also may
be favorable for hydride attack from the â-face. The
observed stereoselectivity is consistent with the classical
steric approach control model of Dauben.27,28
We also explored catalytic hydrogenation of cyclobu-
tanone 3e under a variety of conditions as a more viable
alternative to DIBAC.29 Reduction of 3e with H2 and
commercial ruthenium black in alcohol solvents formed the
R-epimer 10a with ∼85% selectivity.30 Selectivity for 10a
was also high (82%) with tris(triphenylphosphine)rhodium(I)
chloride in the presence of diphenylsilane.31 Moderate
selectivity (∼75% 10a) was obtained with freshly made Ru
black or with Ru/alumina. However, in general, catalytic
hydrogenations of 3e were often problematic due to incom-
plete reduction or overreduction of the phenyl rings.
A plausible mechanism for the rearrangement of 10a to
10b is shown in Scheme 10. During reduction of 3e, the
hydride transfer in the aluminum complex 26 occurs from
the less hindered â-face via a six-center transition state37 to
produce the alanate 27, which is stable at low temperature.
On quenching with acid at low-temperature intermediate 27
furnished the desired product 10a. At higher temperature
(>-10 °C), intermediate 27 is in equilibrium with isomeric
alanate 29, with the bulkier group (alanate) at the primary
position. Presumably, rearrangement occurs via the bicyclic
intermediate 28. It was interesting to find that isolated
carbinol 10a on treatment with 1 mol equiv of DIBAC in
THF-CH2Cl2 (1:1) at room temperature gave a mixture of
10a and 10b in a 1:4 ratio. When this mixture was treated
with DBU in THF at room temperature for 70 h, the ratio of
10a:10b changed to 3:2. Thus, the equilibrium is favorable
for 10a as the free alcohol. However, in the presence of
the strongly oxophilic aluminum counterion, the bulkier
alkoxyaluminum moiety prefers to reside at the sterically
less encumbered primary position.
Regioselective N9-Alkylation of 2-Amino-6-iodopurine
19. Alkylation of 2-amino-6-benzyloxypurine 14 with
cyclobutyltosylate 10c (or the corresponding mesylate) in
the presence of K2CO3 and 18-crown-6 in DMF at 110 °C
produced the N9-alkylation product 20a in 50% chromato-
graphed yield.2 The ratio of N9 to N7 isomers (20a and
21a) was 76:24 (Table 1, entry 1). Cyclobutene 22 also was
formed (7%) in this reaction. Problematically, this reaction
was irreproducible even on a 1-10 g scale. The change of
alkali metal counterions to Na or Li (made from 14 and NaH
or LiH) resulted in no improvement in the yield or regio-
selectivity in this reaction.
Table 1. N9-Regioselective alkylation of the substituted
2-aminopurines
ratio
N9:N7
N9-isomer
(yield, %)
entry
purine
electrophile
Scheme 10
1
2
3
4
5
6
7
14
10c
10c
10f
10e
10e
10e
10e
76:24
75:25
89:11
93:7
93:7
80:20
80:20
20a (50)a,b
20a (44)c,d
20b (64)e
20b (76)f
20b (70)f
20c (74)f
20d (54)f
19e
19a
19a
19b
19c
20d
a K2CO3, 18-crown-6, DMF, 110 °C, 20 h. b 7% 22. c K2CO3, DMF, 110 °C,
10-20 h. d 15% 22. e CH3CN, 60-90 °C, 5-10 h. f CH2Cl2, 25 °C, 3-5 h.
Jones and Roberts18a observed increased N9-regioselec-
tivity in reactions with 6-chloropurines compared to 6-alkoxy-
purines. Geen et al.18b have reported a further enhancement
in N9-regioselectivity during alkylations with 6-iodopurines.
We prepared 2-amino-6-iodopurine 19 from 2-amino-6-
chloropurine and aqueous HI in 90% yield by improvement
of a literature method (yield 35%).3,33 Unfortunately, the
reaction of 2-amino-6-iodopurine 19 with tosylate 10c
(K2CO3, DMF, 110 °C) resulted in extensive degradation
under these harsh reaction conditions. Evidently, the limited
solubility of purine alkali metal salts and the low reactivity
of the cyclobutyl tosylate (or mesylate) caused problems in
this reaction.
(28) Dauben, W. G.; Fonken, G. J.; Noyce, D. S. J. Am. Chem. Soc. 1956, 78,
2579.
(29) Mitsui, S.; Saito, H.; Yamashita, Y.; Kaminaga, M.; Senda, Y. Tetrahedron
1973, 29, 1531.
(30) Bonnet, M.; Geneste, P.; Rodriguez, M. J. Org. Chem. 1980, 45, 40.
(31) Semmelhack, M. F.; Misra, R. N. J. Org. Chem. 1982, 47, 2469.
(32) Haubenstock, H. In Topics in Stereochemistry; Allinger, N. L., Eliel, E. L.,
Wilen, S. H., Eds.; John Wiley: New York, 1983; Vol. 14, p 282.
(33) Koda, R. T.; Biles, J. A.; Wolf, W. J. Pharm. Sci. 1968, 57, 2056.
(34) (a) Hagberg, C.-E.; Johansson, K. N.-G.; Kovacs, Z. M. I.; Stening, G. B.
Eur. Patent Appl. 55239, 1981. (b) Hakimelahi, G. H.; Khalafi-Nezhad, A.
HelV. Chim. Acta 1989, 72, 1495. (c) Seela, F.; Gumbiowski, R. Heterocycles
1989, 29, 795.
There are a few reports on the use of the more soluble
tetraalkylammonium salts of purines, mostly used in situ, in
the alkylation reactions.34 We prepared the tetra-n-butyl-
ammonium salts 19a,c-e as isolated, pure solids by reacting
tetra-n-butylammonium hydroxide in stoichiometric amounts
with the corresponding purines. Unlike the corresponding
alkali metal salts, these salts are very soluble in DMF. Salt
(35) Wolfe, M. S.; Anderson, B. L.; Borcherding, D. R.; Borchardt, R. T. J.
Org. Chem. 1990, 55, 4712.
(36) It was interesting to find that the reaction of purine salt 19b with the triflate
of the isomeric cyclobutanol 10d produced exclusively the corresponding
diastereomeric N9-alkylation product. This observation was made by Dr.
Y. Pendri. Evidently, the steric hindrance of the R-side chain next to the
leaving group further blocked the attack at the N7 site.
(37) (a) Geen, G. R.; Kincey, P. M.; Choudary, B. M. Tetrahedron Lett. 1992,
33, 4609. (b) Lin, T.-S.; Liu, M.-C. Tetrahedron Lett. 1984, 25, 905.
Vol. 2, No. 6, 1998 / Organic Process Research & Development
•
397

19a, although sparingly soluble in CH2Cl2 at ambient
temperature, readily dissolved during reactions with the
electrophiles. In the absence of a reactive electrophile, 19a
was alkylated by CH2Cl2 to produce a mixture of the
corresponding bis-methylene products. As expected, at high
temperature these salts undergo “self-alkylations” to form
several byproducts.
Alkylation of the tetra-n-butylammonium salt 19e with
tosylate 10c (DMF, 110 °C) gave 20a in 44% chromato-
graphed yield (Table 1, entry 2). The N9-regioselectivity
was the same (75%) as with the alkali metal salts. The
elimination side product 22 was formed in 15% yield. The
tetra-n-butylammonium salts of 6-chloro and 6-iodopurines
19c and 19b, respectively, decomposed extensively during
reaction with the tosylate 10c due to the lability of the
halogeno purines and the high reaction temperature. Reac-
tion of the tetra-n-butylammonium salt of 2-amino-6-
iodopurine (19a) with the more reactive cyclobutylnosylate
10f could be performed at 60-90 °C in acetonitrile.
Although high (89%) N9-regioselectivity was achieved, the
yield was improved only moderately (to 64%), and product
isolation required chromatography (Table 1, entry 3). Con-
sequently, we next explored the possible use of trifluoro-
methanesulfonate (triflate) as a leaving group.
Triflates are known not only for their high reactivity but
also for their instability. There are only a few reports of
their use in the alkylation of purines.18a,35 We prepared
cyclobutyltriflate 10e from cyclobutylcarbinol 10a and triflic
anhydride in quantitative yield. While triflate 10e is a white
solid, it was not isolated because of stability concerns. After
workup, the solution of triflate in CH2Cl2 was concentrated
to an optimal volume for further reaction with the purine
salts. We were encouraged to find that reaction of 10e with
the tetra-n-butylammonium salt of 2-amino-6-iodopurine
(19a) at ambient temperature furnished 20b with 93% N9-
regioselectivity in 76% crystallized yield (Table 1, entry 4).36
The minor N7-isomer 21b was easily separated by fractional
crystallization from hot EtOH. In addition, the reaction
conditions were sufficiently mild to entirely prevent elimina-
tion to the cyclobutene 22. The tetra-n-butylammonium
triflate byproduct from the alkylation reaction could not be
completely removed during workup. Some of it remained
in the organic phase even after extensive aqueous washes.
An efficient crystallization protocol subsequently was de-
veloped to remove the low levels of tetra-n-butylammonium
triflate in the product. At the same time, after screening of
several phase-transfer counterions, we found that benzyltri-
ethylammoniun triflate has greater solubility in water and is
completely washed away from the organic phase during
workup. Reaction of benzyltriethylammoniun salt 19b with
triflate 10e gave 20b with 93% regioselection in 70%
crystallized yield (Table 1, entry 5). Thus, either of these
phase-transfer counterions can be used for the large scale
synthesis of 20b. The phase-transfer agent cations may be
recovered from the large scale reactions and recycled.
For comparison, alkylation of the tetra-n-butylammonium
salt of 2-amino-6-chloropurine (19c) with triflate 10e gave
20c in 74% chromatographed yield with 80% N9-regio-
selectivity (Table 1, entry 6). We further found that N2-
acylation of the purine did not significantly influence the
N9-regioselectivity. Reaction of the tetra-n-butylammonium
salt of N2-isobutyryl-6-chloropurine (20d) with the triflate
10e in CH2Cl2 formed 20d with 80% N9-regioselectivity in
54% yield upon chromatography (Table 1, entry 7).
In situations involving reversible reactions, e.g., with sugar
derivatives, R-haloalkyl ethers, and, in a few instances,
Michael acceptors, extremely high N9-regioselectivity has
been obtained during reactions with purines or persilylated
purines.8,16,17,37 The reaction of tris-trimethylsilyl guanine
with triflate 10e in the presence of TMSOTf produced
exclusively the corresponding N7-regioisomer. In the pres-
ence of fluoride ion, a mixture of N9 and N7 products was
obtained. Therefore, this simpler approach did not work to
our advantage.
Conversion of 6-Iododibenzoate 20b to Lobucavir
(BMS-180194). In the final phase of the synthesis, an
efficient single-step process was developed for the conversion
of the 6-iododibenzoate 20b to optically pure lobucavir. The
direct hydrolysis8,17,38 of 20b with aqueous HCl or aqueous
NaOH gave lower quality product in moderate yields.
Methanolysis of 20b with NaOMe formed the 6-methoxy
diol 20e, which on refluxing with aqueous HCl furnished
lobucavir in 93% overall crystallized yield and ee > 99%.
The transesterification procedure is superior due to easy
extractive removal of the neutral byproduct methylbenzoate
from the aqueous solution of the HCl salt of the intermediate
6-methoxypurinediol 20e and, therefore, is the method of
choice. Alternatively, treatment of 6-iododibenzoate 20b
with a weaker acid (aqueous AcOH, heat) formed the guanine
cylobutyldibenzoate 20f in 77% yield. Further deprotection
of 20f to lobucavir was achieved either by transesterification
with catalytic NaOMe (79% yield) or by saponification with
aqueous NaOH (91% yield).
In summary, we have developed a practical asymmetric
synthesis of lobucavir as a potentially viable commercial
synthesis. The asymmetric cycloaddition reaction of dimen-
thyl fumarate with ketene dimethyl acetal furnished cyclo-
butane diester 17. Thus, chemical resolution or the use of
dithioketene acetal was circumvented. Reduction of diester
17 followed by benzoylation and deprotection gave optically
pure cyclobutanone 3e in high yield. Stereoselective reduc-
tion of cyclobutanone with DIBAC, instead of the expensive
reagent LS-Selectride, formed cyclobutanol 10a. The syn-
thesis of final intermediate 20b by an efficient regioselective
alkylation of the tetra-n-butylammonium salt of 2-amino-6-
iodopurine (19a) with cyclobutyltriflate 10e made this
convergent approach practical. Finally, dibenzoate 20b was
deprotected to give lobucavir in 35-40% overall yield (ee
> 99%) in 10 steps. The synthesis has been implemented
to produce over 100 kg of lobucavir for clinical studies.
Experimental Section
Experimental details for the preparation of cyclobutanone
3e were previously described in a preliminary communica-
(38) Linn, J. A.; McLean, E. W.; Kelly, J. L. J. Chem. Soc., Chem. Commun.
1994, 1913.
398
•
Vol. 2, No. 6, 1998 / Organic Process Research & Development

tion.6 Details for the alkylation of the cyclobutanol 10a, via
its triflate 10e, with tetra-n-butylammonium salt 19a to give
dibenzoate 20b and subsequent conversion to lobucavir were
previously reported in a note.3 These reactions have been
scaled-up further by our colleagues in the Kilo lab and Pilot
plant.
saturated aqueous solution of ammonium chloride (502 mL)
was added. After being stirred for 18 h, the mixture was
filtered through anhydrous magnesium sulfate (502 g), and
the filter cake was washed well with dichloromethane. The
filtrate was concentrated at reduced pressure, and the residue
was dried under pump vacuum at 35 °C to give 287.3 g of
crude product. The slightly wet solid was crystallized from
2 L of MeOH and 400 mL of H2O to give 160.6 g (yield
68%) of 10a; HPLC HI (215 nm, Zorbax-cyano column,
H2O-CH3CN gradient) 99.95%. This substance was identi-
cal to the literature compound.2,6 Anal. Calcd for
C20H20O5: C, 70.38; H, 5.94; H2O, 0.27. Found: C, 69.92;
H, 5.87; H2O, 0.27 (KF).
Preparation of [1S-(1r,2â,3â)]-3-Hydroxy-1,2-cyclo-
butanedimethanol, Dibenzoate Ester (10a). A 3-L, three-
necked flask equipped with a mechanical stirrer, internal
digital thermometer, addition funnel, and nitrogen inlet was
charged with 2385 mL of anhydrous dichloromethane. After
cooling to -40 °C, diisobutylaluminum chloride (156.6 g,
886 mmol, 1.27 mol equiv) was added. To this cold solution
was added, dropwise via addition funnel over 63 min, a
solution of cyclobutanone 3e (235 g, 699 mmol) in dichlo-
romethane (600 mL). The reaction was maintained at -40
°C for an additional 70 min. TLC analysis (silica gel,
toluene:ether, 1:1, Rf of 3e ) 0.64, 10a ) 0.36, and â-isomer
10b ) 0.27, visualized by UV and p-anisaldehyde reagent)
indicated that the reduction was essentially complete. The
reaction was quenched by the slow addition of methanol (502
mL). During the quench, the temperature rose from ∼-40
to -32 °C over 1 h. The cooling was removed, and a
Acknowledgment
We thank the Science Information Department, Bristol-
Myers Squibb, for helpful literature searches and Analytical
R & D for their valuable support during the course of this
work.
Received for review March 25, 1997.
OP970214+
Vol. 2, No. 6, 1998 / Organic Process Research & Development
•
399