Synthesis of conformationally restricted carbocyclic nucleosides: The role of the O(4')-atom in the key hydration step of adenosine deaminase

Article abstract of DOI:10.1002/(SICI)1522-2675(19991215)82:12<2119::AID-HLCA2119>3.0.CO;2-Y

Conformationally restricted carbocyclic nucleosides with either a northern(N)-type conformation, i.e., N-type 2'-deoxy-methanocarba-adenosine 8 ((N)MCdAdo) or a southern(S)-type conformation, i.e. S-type 2'-deoxy- methanocarba-adenosine 9, ((S)MCdAdo), were used as substrates for adenosine deaminase (ADA) to assess the enzyme's preference for a fixed conformation relative to the flexible conformation represented by the carbocyclic nucleoside aristeromycin (10). Further comparison between the rates of deamination of these compounds with those of the two natural substrates adenosine (Ado: 1) and 2'-deoxyadenosine (dAdo; 2), as well as with that of the conformationally locked nucleoside LNA-Ado (11), which, like the natural substrates, has a furanose O(4') atom, helped differentiate between the roles of the O(4') anomeric effect and sugar conformation in controlling the rates of deamination by ADA. Differences in rates of deamination as large as 10000 can be attributed to the combined effect of the Q(4') atom and the enzyme's preference for an N-type conformation. The hypothesis proposed is that ADA's preference for N-type substrates is not arbitrary; it is rather the direct consequence of the conformationally dependent O(4') anomeric effect, which is more efficient in N-type conformers in promoting the formation of a covalent hydrate at the active site of the enzyme. The formation of a covalent hydrate at the active site of ADA precedes deamination. A new and efficient synthesis of the important carbobicyclic template 14a, a useful intermediate for the synthesis of (N)MCdAdo (8) and other conformationally restricted nucleosides is also reported.

Full text of DOI:10.1002/(SICI)1522-2675(19991215)82:12<2119::AID-HLCA2119>3.0.CO;2-Y

Helvetica Chimica Acta ± Vol. 82 (1999)
2119
Synthesis of Conformationally Restricted Carbocyclic Nucleosides: The Role
of the O(4')-Atom in the Key Hydration Step of Adenosine Deaminase
by Victor E. Marqueza)*, Pamela Russa), Randolph Alonsoa), Maqbool A. Siddiquia), Susana Hernandeza),
Clifford Georgeb), Marc C. Nicklausa), Fang Daia), and Harry Ford, Jr.a)
a) Laboratory of Medicinal Chemistry, Division of Basic Sciences, National Cancer Institute, National Institutes
of Health, Bethesda, MD 20892, USA
b) Laboratory for the Structure of Matter, Naval Research Laboratory, Washington, DC 20375, USA
Dedicated to Prof. Dr. Frank Seela on the occasion of his 60th birthday
Conformationally restricted carbocyclic nucleosides with either a northern(N)-type conformation, i.e., N-
type 2'-deoxy-methanocarba-adenosine 8 ((N)MCdAdo), or a southern(S)-type conformation, i.e. S-type 2'-
deoxy-methanocarba-adenosine 9, ((S)MCdAdo), were used as substrates for adenosine deaminase (ADA) to
assess the enzymeꢀs preference for a fixed conformation relative to the flexible conformation represented by the
carbocyclic nucleoside aristeromycin (10). Further comparison between the rates of deamination of these
compounds with those of the two natural substrates adenosine (Ado; 1) and 2'-deoxyadenosine (dAdo; 2), as
well as with that of the conformationally locked nucleoside LNA-Ado (11), which, like the natural substrates,
has a furanose O(4') atom, helped differentiate between the roles of the O(4') anomeric effect and sugar
conformation in controlling the rates of deamination by ADA. Differences in rates of deamination as large as
10000 can be attributed to the combined effect of the O(4') atom and the enzymeꢀs preference for an N-type
conformation. The hypothesis proposed is that ADAꢀs preference for N-type substrates is not arbitrary; it is
rather the direct consequence of the conformationally dependent O(4') anomeric effect, which is more efficient
in N-type conformers in promoting the formation of a covalent hydrate at the active site of the enzyme. The
formation of a covalent hydrate at the active site of ADA precedes deamination. A new and efficient synthesis
of the important carbobicyclic template 14a, a useful intermediate for the synthesis of (N)MCdAdo (8) and
other conformationally restricted nucleosides, is also reported.
Introduction. ± Adenosine deaminase (ADA, EC 3.5.4.4) is a critical enzyme of the
purine metabolic pathway that catalyzes the irreversible hydrolytic deamination of
adenosine (Ado; 1) and 2'-deoxyadenosine (dAdo; 2) ± as well as of other exogenous
substrates ± to their hypoxanthine derivatives 4 and 5, respectively, and ammonia
(Scheme 1) [1][2]. Aberrations in the expression and function of ADA have been
associated with impaired B- and T-cell-based immunity, whereas elevated levels of
ADA have been observed in certain lymphomas and leukemias [3 ± 5]. Thus, a
complete understanding of the ADA-catalyzed mechanism of deamination should
provide additional means of intervening therapeutically in ADA-related disorders. The
rational design of the first transition-state analogue inhibitor of ADA by Wolfenden
and Evans in 1970 [6] was followed four years later by the discovery of the potent
fermentation-product inhibitors coformycin (6) [7][8] and 2'-deoxycoformycin (7) [9],
which were shown to function as stable, high-affinity transition-state-analogue
inhibitors of ADA [10][11]. Further confirmation of the transition-state mimicry of
these analogues came from several crystal structures of ADA with the ligands bound at
the active site of the enzyme [12 ± 14]. In the case of the simple purine riboside 3

2120
Helvetica Chimica Acta ± Vol. 82 (1999)
(nebularine), covalent hydration at the active site of ADA is able to generate the
potent transition-state inhibitor (6S)-1,6-dihydro-9H-purin-6-ol riboside 3a [11]
(Scheme 1). However, since the equilibrium constant for this critical hydration step
is very small, the apparent inhibition constant of nebularine is only in the micromolar
range [11]. On the other hand, the transition-state inhibitors coformycin (6) and 2'-
deoxycoformycin (7), where the 8-hydroxy substituent of the diazepine ring provides a
stable mimic of a hydrated ring, are able to inhibit ADA at picomolar levels [10][11].
From these observations, it was concluded that all ADA substrates must form a
covalent hydrate (i.e. 1a and 2a in Scheme 1) as a first critical step before deamination.
Calculation of relative hydration free energy differences in heteroaromatic bases have
revealed the importance of electronic and steric factors in facilitating hydration of the
aglycon as a key step in the deamination reaction [15]. For example, a comparison
between nebularine (3) and its 8-aza analogue showed a 400-fold greater ADA
inhibitor potency for the latter, which is attributable to the enhancing effect of
hydration caused by the N-atom at the 8-position [15]. The role of the sugar moiety in
this key hydration step, on the other hand, has for the most part been ignored and even
considered negligible [15]. In the present work, we wish to address this issue by
comparing the deamination rates of substrates that differ from Ado and dAdo in that
the furanose ring has been replaced by a cyclopentane ring.
Scheme 1
In addition, since ADA-inhibitor complexes [12 ± 14] have revealed a preferred
northern(N)-type conformation for the bound inhibitor at the active site ± and by
extension one can assume that substrates will bind with the same preferred
conformation ± we have chosen to study the deamination rates of carbocyclic
adenosine nucleosides locked in antipodal conformations (northern(N)- and southern-

Helvetica Chimica Acta ± Vol. 82 (1999)
2121
(S)-type) similar to those of the furanose ring, as defined by the pseudorotation phase
angle P [16]. Normally, conventional nucleosides in solution undergo rapid equilibra-
tion between ranges of P defined by two main furanose puckering domains centered
around a 3T2 (N-type) and a 2T3 (S-type) conformation (Fig. 1). These conformational
modes, in turn, favor particular orientations of the glycosyl torsion angle c (syn or anti)
and the 4-(hydroxymethyl) chain defined by g (‡sc,
sc, and ap). Indeed, the range
of values of c and g are not uniformly populated, and their preferred distribution is
coupled to a particular sugar pucker [17]. The conformationally locked carbocyclic
adenosine substrates selected for this study were (N)MCdAdo (8), as a rigid N-type
conformer, and (S)MCdAdo (9) as a rigid S-type conformer. The fermentation product
aristeromycin (Ari; 10) was used as a base-line reference for a flexible, carbocyclic
nucleoside substrate. Indeed, aristeromycin has been shown to undergo rapid
equilibration in solution between a southwestern-type and an N-type conformer, with
a bias towards the former [18]. Comparison between the rates of deamination of these
compounds with those of the two natural substrates Ado (1) and dAdo (2), as well as
with that of the conformationally locked nucleoside LNA-Ado (11) [19], helped
differentiate the roles of the O(4') atomꢀs anomeric effect and sugar conformation in
controlling the rates of deamination by ADA. As in the natural substrates, the
conformationally locked nucleoside LNA-Ado has a O(4') furanose atom, and its
enhanced rate of deamination relative to the conformationally locked carbocyclic
nucleosides highlights the importance of the anomeric effect contributed by the O(4')
atom.
In this work, we also wish to present a new and more efficient six-step synthesis of
the critical precursor 14a, which is an extremely useful intermediate for the convergent
synthesis of (N)MCdAdo and other N-type purine analogues. The overall yield of 35%
obtained for 14a compares favorably with our previous seven-step synthesis reported
earlier (25% yield) [20].
Results. ± Synthesis and Structural Analysis of Carbocyclic Nucleosides. The
syntheses of the conformationally locked N-type nucleoside (N)MCdAdo (8), as well

2122
Helvetica Chimica Acta ± Vol. 82 (1999)
Fig. 1. Pseudorotational cycle displaying all the puckering modes (P) and maximum puckering amplitude (nmax
)
as that of the S-type analogue (S)MCdAdo (9), have been reported [20][21]. While the
X-ray structure of (N)MCdAdo (8) was published in conjunction with its synthesis
[20]1), we now report for the first time the crystal structure of (S)MCdAdo (9)2). From
comparing the two structures, it is clear that these carbocyclic nucleosides adopt fixed
conformations that mimic the ring pucker of a furanose ring (Figs. 2 and 3). As
discussed in a preliminary communication [22], the crystallographic pseudorotational
parameters for the two molecules (A and B) in the unit cell of (N)MCdAdo are in
complete agreement with the expected values for a N-type (2E) conformation (P ˆ
342.788 (A) and P ˆ 339.258 (B)), whereas the equivalent parameters for the antipodal
(S)MCdAdo correlate with a pure S-type (3E) conformation (P ˆ 199.108 (A) and
P ˆ 200.208 (B)). Except for the molecule A of (S)MCdAdo, all torsion angles c are in
the anti range, while the less encumbered angle g alternates between the ap and ‡ sc
conformations as it is normally seen in conventional nucleosides [22].
1
)
)
CCDC entry NAKZUU.
2
Crystallographic data have been deposited with the Cambridge Crystallographic Data Centre as deposition
No. CCDC-133070. Copies of the data can be obtained, free of charge, by applying to the CCDC, 12 Union
Road, Cambridge CB21EZ, UK (fax: ‡ 44(0)1223-336033; e-mail: deposit@ccdc.com.oc.uk).

Helvetica Chimica Acta ± Vol. 82 (1999)
2123
Fig. 2. Molecular structure of (N)MCdAdo (8): a) molecule A (c ˆ 154.88 (anti), g ˆ ‡54.88 (‡sc)) and
b) molecule B (c ˆ 167.68 (anti), g ˆ 177.68 (ap))
Fig. 3. Molecular structure of (S)MCdAdo (9): a) molecule A (c ˆ ‡50.98 (syn), g ˆ ‡54.08 (‡sc)) and
b) molecule B (c ˆ 129.88 (anti), g ˆ 171.98 (ap))
The new approach to the synthesis of (N)MCdAdo (8) was based on the easily
available chiral template (3R,4S)-4-(phenylmethoxy)-3-[(phenylmethoxy)methyl]cy-
clopent-1-ene (12) [23 ± 25]. This approach compared favorably with respect to the one
devised earlier from chiral cyclopentenone 13 [20], particularly in the generation of the
key carbobicyclic intermediates 14a,b (Scheme 2). In the same manner as reported
previously with the PhSeCl/NaN3 system [25], the double bond in 12 reacted regio- and
stereoselectively with PhSeCl/AgCF3CO2 to give the trans-2-(phenylseleno)cyclopen-
tan-1-ol intermediate 15 after hydrolysis of the trifluoroacetate ester (Scheme 3).
Following oxidation of the (phenylseleno) group, unidirectional elimination of
PhSeOH gave exclusively the corresponding allylic alcohol 16. This allylic alcohol
was inverted following a Mitsunobu esterification which gave the intermediate
benzoate 17 and the new allylic alcohol 18 after ester hydrolysis. Finally, a hydroxy-

2124
Helvetica Chimica Acta ± Vol. 82 (1999)
Scheme 2
Scheme 3
directed cyclopropanation under Simmons-Smith conditions gave the key carbobicyclic
intermediate 14a in 35% overall yield. This process required one step less than the
previously reported approach [20] and provided a 10% increase in the overall yield of
the final product. The synthesis of (N)MCdAdo (8) from 14a proceeded essentially as
reported before from intermediate 14b [20], i.e., via 19 and 20 (Scheme 4).
The conformationally locked, pentofuranose-containing LNA-Ado (11), the syn-
thesis of which has been reported earlier [19], was a kind gift from Professor Wengel.
On conformational grounds, however, the locked 2,5-dioxabicyclo[2.2.1]heptane
system does not have the same type of ring pucker as the bicyclo[3.1.0]hexane. The
reported X-ray structure of the corresponding LNA-uracil analogue [26] shows that the
2,5-dioxabicyclo[2.2.1]heptane system has a P value of 17.48, which is ca. 34 ± 388 away
from the pseudorotational angle of the locked N-type bicyclo[3.1.0]hexane template of
carbocyclic nucleoside 8. These different envelope conformations, 2E vs. 3E (see Fig. 1),
could account for some differences in the rates of deamination discussed below.

Helvetica Chimica Acta ± Vol. 82 (1999)
2125
Scheme 4
Adenosine Deaminase Activity. The compounds listed in the Table can be grouped
in two categories depending on their substrate properties. The first three substrates 1, 2,
and 11 fall in the category of good substrates, as indicated by the various kinetic
parameters such as Km, Kcat, and the catalytic efficiency (Kcat/Km). The difference
between these substrates and the carbocyclic analogues as a group appears substantial
in view of the relative hydrolysis rates measured at 50 mm substrate concentration.
Despite the poor deamination rates among the locked carbocyclic analogues 8 ± 10,
there appears to be a clear preference for the N-type conformation. This observation is
in agreement with the published X-ray structures of ADA-inhibitor complexes where
the sugar moieties of the nucleosides bound at the active site appear puckered as N-
type sugars [12 ± 14]. Interestingly, the rate of deamination of the conformationally
flexible aristeromycin (10) lies between the rates of deamination of the rigid N-type
(N)MCdAdo (8) and the rigid S-type (S)MCdAdo (9) analogues. Obviously, the
carbocyclic nucleosides are in a class by themselves as poor substrates of ADA, which
appears to be a direct consequence of lacking the furanose O-atom. It has been clearly
demonstrated by Chattopadhyaya and co-workers that the electronic nature of the
aglycon is effectively transmitted to the pentofuranose moiety through the anomeric
effect [27][28]. Indeed, the direct correlation of the protonation/deprotonation
equilibrium of the aglycon with the two-state N/S pseudorotational equilibrium of
the sugar moiety permits the facile calculation of the pKa from the pD-dependent
change of DG8 in the N/S equilibrium [27][28]. On the other hand, a similar pKa
determination is not possible in carbocyclic nucleosides, and specifically for aristero-
mycin (10), because of the lack of the O(4') atom and its associated anomeric effect
[18][28]. Even though exact kinetic parameters between the two groups could not be
compared, the differences in relative rates of deamination revealed some interesting
trends worth discussing. For example, a direct comparison between the best ADA
substrate, dAdo (2), and the best equivalent carbocyclic substrate, (N)MCdAdo (8),
reveals a 122-fold difference in the rate of deamination in favor of the former that can
be attributed to a combined effect of the O(4') atom and the enzymeꢀs preference for
binding N-type substrates (Table). For the other two substrates, Ari (10) and

2126
Helvetica Chimica Acta ± Vol. 82 (1999)
(S)MCdAdo (9), this difference can be even greater (200 and 10000, resp.). These
larger differences are probably the result of higher energy penalties incurred for having
a conformation far removed from the ideal N-type conformation. To gauge the
conformational factor alone, a comparison between the rates of deamination of
(N)MCdAdo (8) and (S)MCdAdo (9) would show the enzymeꢀs capacity to effectively
discriminate between the two rigid antipodal conformations. The nearly 100-fold
difference between the rates of deamination of (N)MDdAdo (8) and (S)MCdAdo (9)
demonstrates the enzymeꢀs clear preference for an N-type conformer. In the real world,
however, the effect of conformation is not as dramatic because of the nucleosideꢀs
inherent flexibility and capacity to adapt to the conformation demanded by the
enzyme, something that would not be possible for (N)MDdAdo (8) and (S)MCdAdo
(9). Indeed, Ado (1) and dAdo (2), which are known to exist in solution preferentially
as S-type conformers [27], are expected to bind to the enzyme as N-type conformers
[12 ± 14]. From our study, it can be approximated that the combined anomeric effect of
the O(4') atom and the preference for an N-type conformation could account for
differences in deamination rates as large as 10000. Since for nucleosides such as Ado (1)
and dAdo (2), the anomeric effect is greater when the conformation of the sugar is of
the N-type, the preference shown by ADA for N-type conformers is not a capricious
preference; it results from the fact that in an N-type conformation, the anomeric effect
stabilizes the protonated base more efficiently, a prerequisite for the formation of the
covalent hydrate at the active site of ADA prior to deamination (Scheme 1) [11]. In
these experiments, the dramatic effect of the O(4') atom was revealed by its capacity to
increase the rate of deamination of the conformationally locked nucleoside LNA-Ado
(11) to ca. 20-fold relative to the best carbocyclic substrate (N)MCdAdo (8). This
makes LNA-Ado only a 4-fold less efficient substrate than Ado (1). Such an increase in
rate is clearly a direct consequence of the O(4') atomꢀs role in the critical hydration step
prior to deamination. However, why did the rate of deamination of LNA-Ado (11) not
match the level of deamination of Ado (1)? We have already discussed in the previous
section that the X-ray structures of the locked 2,5-dioxabicyclo[2.2.1]heptane and the
bicyclo[3.1.0]hexane systems used as N-type templates differ by 34 ± 388 in pseudor-
otation (P). This 34 ± 388 difference in P could account for the partial substrate
recovery observed for LNA-Ado (11), as well as for the fact that there could be
possible steric clashes at the receptor site between the structural elements added to
constrain the rings and some critical amino-acid side chains. It must be remembered
that, while there is a favorable entropic advantage in binding a constrained substrate,
the enthalpy penalties could be rather severe when the fit is not perfect. It is possible,
therefore, that while LNA-Ado (11) exists as an N-type conformer, it does not have the
ideal conformation for a perfect fit. Despite this shortcoming, if one assumes that both
conformationally rigid compounds, (N)MCdAdo (8) and LNA-Ado (11), approximate
the preferred N-type conformation and are able to bind at the active site of ADA, the
role of the O(4') atom appears conclusive. Further confirmation of the importance of
the O(4') atom anomeric effect, as a critical determinant for the hydration step
catalyzed by ADA and related enzymes, is found in the recent work of Lindell and
co-workers who report that the 5'-monophosphate of the carbocyclic analogue of
nebularine is a very poor inhibitor of the mechanistically equivalent adenosine 5'-
monophosphate deaminase (AMPDA) [29].

Helvetica Chimica Acta ± Vol. 82 (1999)
2127
Table. Kinetic Constants for Various Substrates of ADAa)
Relative rateb)
1
1
Km [mm]
Kcat [s
]
Kcat/Km [mm 1s
]
Ado (1)
dAdo (2)
LNA-Ado (11)
(N)MCdAdo (8)
Ari (10)
33.1 Æ 3.7
188
245
59.4
±
±
±
5.7
10.6
0.49
±
±
±
100
121
19
0.99
0.58
0.01
23.0 Æ 2.3
122 Æ 12
±
±
±
(S)MCdAdo (9)
a) Kinetic constants were determined by UV at 378 and pH 7.4 with calf-intestine ADA. b) Relative rates were
measured spectrophotometrically at 378 with 50 mm substrate concentration.
Experimental Part
1. General. All chemical reagents were commercially available. M.p.: MelTemp-II apparatus, Laboratory
Devices, USA; uncorrected. Column flash chromatography (FC): silica gel 60 (230 ± 400 mesh; E. Merck). 1H-
and 13C-NMR Spectra: Bruker-AC-250 instrument at 250 and 62.9 MHz, resp.; d in ppm rel. to the solvent in
which they were run (7.24 ppm for CDCl3). Elemental analyses were performed by Atlantic Microlab, Inc.,
Atlanta, GA.
2. X-Ray Crystal Structure of (1S,3S,4R,5S)-1-(6-Amino-9H-purin-9-yl)-4-(hydroxymethyl)bicyclo[3.1.0]-
2
hexan-3-ol (9). Crystal Data ): 2(C12H15N5O2) ´ 3(H2O), Mr 576.63; crystal size: 0.40 Â 0.08 Â 0.02 mm3;
monoclinic P21, a ˆ 11.460(2) Š, b ˆ 7.991(2) Š, c ˆ 15.524(3) Š, a ˆ 908, b ˆ 95.55(1)8, g ˆ 908; Vˆ
1377.9(5) Š3, Z ˆ 2, Dc ˆ 1.390 mg/m3, l(CuKa) ˆ 1.54178 Š, m ˆ 0.877 mm 1, F(000) ˆ 612, Tˆ 293(2) K. Final
R indices (I > 2s(I0)) R1 ˆ 0.0658, wR2 ˆ 0.1395 for 2136 reflections. Tables of atomic coordinates, bond
2
distances, and angles, and anisotropic thermal parameters have been deposited ).
3. Kinetic Analysis and Adenosine Deaminase Assays. Adenosine deaminase (calf intestine) activity was
measured spectrophotometrically with a Shimazdu-UV-PC-2101 spectrophotometer by following the disap-
pearance of adenosine at 265 nm in 1-ml cuvettes with 0.1m phosphate (pH 7.4) at 258. For the substrates listed in
the Table, initial hydrolysis rates were determined by following the disappearance of substrate at 265 nm in 1-ml
cuvettes with 0.1m phosphate (pH 7.4) at 378. A similar molar extinction coefficient and a De ˆ 7900 was
assumed for all adenine analogues examined. Kinetic constants were determined from Lineweaver-Burk plots
with Graph-Pad Prism V 2.01 (Graphpad Software, Inc. San Diego, CA), a personal-computer-based curve-
fitting program. A minimum of three determinations were performed. Relative-rate experiments were
performed in duplicate using 50 mm substrate concentration at 378.
4. Syntheses. (3R,4S)-4-(Phenylmethoxy)-3-[(phenylmethoxy)methyl]cyclopent-1-ene (12) was prepared in
95% yield as described in [25].
(1S,4S)-4-(Phenylmethoxy)-3-[(phenylmethoxy)methyl]cyclopent-2-en-1-ol (16). A mixture of 12 (2.36 g,
8.61 mmol) and phenylselenium chloride (1.98 g, 10.3 mmol) in dry DMSO (8 ml) was stirred under Ar at r.t.
until solution occurred. Silver trifluoroacetate (2.28 g, 10.3 mmol) was added, and the mixture was stirred for
1 h. The mixture was then cooled over ice, and hydrolysis of the trifluoroacetate ester was accomplished with 5%
KOH/EtOH (15 ml) while stirring for 0.25 h. The mixture was poured into ice-cold H2O (100 ml) and extracted
with Et2O (3 Â 100 ml), the combined Et2O extract washed with sat. NH4Cl soln. (35 ml), dried (MgSO4), and
evaporated, and the residue purified by FC (silica gel, step gradient of 10%, 20%, and 50% AcOEt/hexanes):
reasonably pure (1S,2S,3R,4S)-4-(phenylmethoxy)-3-[(phenylmethoxy)methyl]-2-(phenylseleno)cyclopentan-
1-ol (15; 2.98 g, 74%).
A mixture of crude 15 (2.20 g, 4.71 mmol) and NaIO4 (2.02 g, 9.42 mmol) was stirred in MeOH/H2O 9 :1
(100 ml) for 1 h. After evaporation, the residue was stirred with AcOEt (100 ml), and the insoluble solid was
removed by filtration. The filtrate was evaporated and the residue purified by FC (silica gel, step gradient of
25% and 50% AcOEt/hexanes): 16 (1.06 g, 73%). Clear oil. 1H-NMR (CDCl3): 7.44 ± 7.30 (m, 2 Ph); 6.03 (br. s,
H
C(2)); 5.10 ± 5.04 (symmetrical m, H C(1)); 4.90 ± 4.84 (symmetrical m, H C(4)); 4.67 (AB(ꢁdꢀ, J ˆ 11.9,
1 H, PhCH2); 4.64 (AB(ꢁdꢀ), J ˆ 11.7, 1 H, PhCH2); 4.61 (AB(ꢁdꢀ), J ˆ 11.9, 1 H, PhCH2); 4.52 (AB(ꢁdꢀ), J ˆ
11.9, 1 H, PhCH2); 4.26 (br. s, PhCH2OCH2); 2.38 (ddd, J ˆ 14.4, 6.8, 3.2, Ha C(5)); 2.09 (ddd, J ˆ 14.1, 6.6, 3.2,
Hb C(5)); 1.91 (br. s, OH). 13C-NMR (CDCl3): 145.54; 138.25; 138.03; 132.89; 128.25; 128.24; 127.57; 127.52;

2128
Helvetica Chimica Acta ± Vol. 82 (1999)
127.48; 82.10; 75.28; 72.70; 71.28; 66.51; 41.16. Anal. calc. for C20H22O3 ´ 0.25 H2O (314.89): C 76.27, H 7.21;
found: C 76.40; H 7.28.
(1R,4S)-4-(Phenylmethoxy)-3-[(phenylmethoxy)methyl]cyclopent-2-enyl Benzoate (17). To a stirred soln.
of 16 (1.06 g, 3.32 mmol) in dry benzene (50 ml), benzoic acid (0.63 g, 5.13 mmol) and Ph3P (1.79 g, 6.84 mmol)
were added. Diethyl azodicarboxylate (DEAD; 1.08 ml, 6.84 mmol) was added dropwise, and the reaction was
allowed to continue at r.t. for 0.5 h. After all the volatiles were removed under reduced pressure, the residue was
purified by FC (silica gel, 10% AcOEt/hexanes): 17 (1.27 g, 85%). Thick oil. 1H-NMR (CDCl3): 8.16 ± 8.12,
7.67 ± 7.34 (2m, 3 Ph); 6.17 (br. s, H C(2)); 5.92 ± 5.82 (symmetrical m, H C(1)); 4.75 ± 4.60 (m, 2 PhCH2,
H
C(4)); 4.35 (br. s, 2 H, PhCH2OCH2); 3.01 (dt, J ˆ 14.2, 7.3, Ha C(5)); 2.08 (dt, J ˆ 14.4, 4.2, Hb C(5)).
13C-NMR: 166.46; 147.62; 138.48; 138.22; 133.07; 130.44; 129.82; 128.57; 128.55; 128.51; 128.47; 127.87; 127.85;
127.80; 80.80; 76.78; 73.11; 71.37; 66.73; 38.22. Anal. calc. for C27H26O4 (414.50): C 78.24, H 6.32; found: C 78.40,
H 6.72.
(1R,4S)-4-(Phenylmethoxy)-3-[(phenylmethoxy)methyl]cyclopent-2-en-1-ol (18). A stirred soln. of 17
(1.55 g, 3.59 mmol) in MeOH (50 ml) was treated with K2CO3 (1.24 g, 8.98 mmol). After 3 h, the solvent was
evaporated and the residue purified by FC (silica gel, 5% MeOH/CHCl3): 18 (0.874 g, 80%). Oil. 1H-NMR
(CDCl3): 7.54 ± 7.38 (m, 2 Ph); 6.03 (br. s, H C(2)); 4.75 ± 4.55 (m,2  PhCH2O); 4.49 ± 4.54 (dd, J ˆ 6.8, 3.9,
H
C(4)); 4.30 (br. s, PhCH2OCH2); 2.74 (dt, J ˆ 14.2, 7.1, Ha C(5)); 2.68 (br. s, OH); 1.82 (dt, J ˆ 14.2, 3.9,
Hb C(5)); 13C-NMR (CDCl3): 144.86; 138.39; 138.20; 133.01; 128.44; 127.82; 127.78; 127.71; 80.99; 73.98;
72.84; 71.48; 66.62; 41.16. Anal. calc. for C20H22O3 ´ 0.5 H2O (319.39): C 75.21, H 7.26; found: C 75.46, H 7.51.
(1S,2R,4S,5R)-4-(Phenylmethoxy)-5-[(phenylmethoxy)methyl]bicyclo[3.1.0]hexan-2-ol (14a). A stirred
soln. of alcohol 18 (0.66 g, 2.15 mmol) in dry CH2Cl2 (30 ml) was cooled over an ice-salt bath. Then 1m diethyl
zinc in hexane (2.4 ml) was added dropwise, and the reaction was allowed to continue for 0.25 h after the
addition. Separately, a soln. of diiodomethane (0.4 ml, 4.74 mmol) in dry CH2Cl2 (10 ml) was prepared, and
5.2 ml of this soln. was added to the mixture. After 5 min stirring, an equal amount of diethylzinc was added
dropwise, followed by the remaining soln. of diiodomethane (4.8 ml). The mixture was stirred for a total of 6 h,
while the cooling bath gradually warmed to r.t. Then, the mixture was poured into an aq. sat. NH4Cl soln.
(50 ml), and extracted with CH2Cl2 (3 Â 50 ml). The combined extract was washed with sat. NH4Cl soln. (2 Â
50 ml), dried (MgSO4), and evaporated, and the residue purified by FC (silica gel, step gradient of 25% and
50% AcOEt/hexanes): 14a (0.560 g, 80%). Oil. 1H-NMR (CDCl3): 7.45 ± 7.30 (m, 2 Ph); 4.65 ± 4.40
(m, 2 PhCH2O, H C(4)); 4.31 (t, J ˆ 8.0, H C(2)); 3.98 (d, J ˆ 10.5, 1 H, PhCH2OCH2); 3.21 (d, J ˆ 10.5,
1 H, PhCH2OCH2); 2.37 (dt, J ˆ 12.9, 7.4, Ha C(3)); 1.70 ± 1.60 (m, OH, H C(3)); 1.40 ± 1.20 (m, H C(1),
Ha C(6)); 0.63 (dd, J ˆ 7.4, 5.6, Hb C(6)). 13C-NMR (CDCl3): 138.84; 138.53; 128.49; 128.47; 127.88; 127.78;
127.72; 127.70; 73.01; 72.13; 71.87; 70.02; 35.59; 32.99; 28.20; 7.00. Anal. calc. for C21H24O3 (324.41): C 77.74,
H 7.46; found: C 77.51, H 7.40.
(1R,2S,4S,5S)-4-(6-Amino-9H-purin-9-yl)-2-hydroxybicyclo[3.1.0]hexan-1-methanol (8). Diethyl azodicar-
boxylate (DEAD; 0.63 ml, 3.98 mmol) was added dropwise to a soln. of Ph3P (1.04 g, 3.98 mmol) in dry THF
(8.5 ml) previously cooled in an ice-salt bath. After 20 min, this soln. was added dropwise to a soln. of 14a
(0.560 g, 1.73 mmol) and 6-chloro-1H-purine (0.431 g, 2.77 mmol) in dry THF (8.5 ml), which was also cooled in
an ice-salt bath. The mixture was stirred cold for 0.5 h and then at r.t. for 3 h. The volatiles were removed and the
residue purified by FC (silica gel, 25% AcOEt/hexanes): 6-chloro-9-{(1S,2S,4S,5R)-4-(phenylmethoxy)-5-
[(phenylmethoxy)methyl]bicyclo[3.1.0]hex-2-yl}-9H-purine (19; 0.741 g, 93%), slightly contaminated with
DEAD. 1H-NMR (CDCl3): 9.20 (s, H C(2)); 8.80 (s, H C(8)); 7.50 ± 7.30 (m, 2 Ph); 5.35 (d, J ˆ 5.2, H C(2'));
4.75 (t, J ˆ 8.5, H C(4')); 4.65 (br. s, PhCH2O); 4.55 (AB (m), PhCH2O); 4.30 ± 4.50 (DEAD contamination);
4.25 (d, J ˆ 9.9, 1 H, PhCH2OCH2); 3.25 (d, J ˆ 9.9, 1 H, PhCH2OCH2); 2.15 (m, Ha C(3')); 1.95
(m, Hb C(3')); 1.80 (m, H C(1')); 1.45 (DEAD contamination); 1.25 (m, Ha C(6')); 0.95 (m, Hb C(6')).
Since the DEAD contamination was successfuly removed only after the final deprotection step, 19 (0.481 g,
1.04 mmol) was dissolved in a minimum amount of MeOH and added to a sat. NH3/MeOH soln. (90 ml) in a
sealed pressure steel bomb which was heated at 708 overnight. After cooling, the soln. was evaporated and co-
evaporated twice with MeOH (100 ml) before purification by FC (silica gel, 2% MeOH/CHCl3): 9-
{(1S,2S,4S,5R)-4-(phenylmethoxy)-5-[(phenylmethoxy)methyl]bicyclo[3.1.0]hex-2-yl}-9H-purin-6-amine (20;
0.347 g, 75%). Heavy syrup. 1H-NMR (CDCl3): 8.65 (s, H C(2)); 8.40 (s, H C(8)); 7.50 ± 7.30 (m, 2 Ph);
6.75 (br. s, NH2); 5.25 (d, J ˆ 6.1, H C(2')); 4.75 (t, J ˆ 8.2, H C(4')); 4.65 (br. s, PhCH2O); 4.55 (AB (m),
PhCH2O); 4.15 ± 4.35 (m, 1 H of PhCH2OCH2, DEAD contamination); 3.25 (d, J ˆ 9.9, 1 H, PhCH2OCH2);
2.15 (m, Ha C(3')); 1.95 (m, Hb C(3')); 1.70 (m,
(m, Ha C(6')); 0.90 (m, Hb C(6')).
H C(1')); 1.40 (DEAD contamination); 1.20

Helvetica Chimica Acta ± Vol. 82 (1999)
2129
Compound 20 was used directly for the final deprotection step: A soln. of 20 (0.347 g, 0.787 mmol) in
MeOH (96 ml) was treated with 96% formic acid (4 ml) and stirred overnight in the presence of Pd black
(0.5 g). The insolubles were removed by filtration with the aid of Celite, and the filtrate was evaporated to give a
white solid. To this solid, CHCl3 (50 ml) was added and the mixture heated, cooled, and filtered. The remaining
solid was recrystallized from MeOH/Et2O: pure 8 (0.160 g, 78%). Solid. M.p. 259 ± 2618 (dec.), identical to that
of an authentic sample.
REFERENCES
[1] R. Wolfenden, J. Kaufman, J. B. Macon, Biochemistry 1969, 8, 2412.
[2] B. E. Evans, R. Wolfenden, Biochemistry 1973, 12, 392.
[3] For a review of ADA deficiency, see M. S. Hershfield, B. S. Mitchell, in ꢁThe Metabolic and Molecular Basis
of Inherited Diseaseꢀ, Ed. C. R. Scriver, A. L. Beaudet, W. S. Sly, D. Valle, McGraw Hill, New York, 1995,
pp. 1725 ± 1768.
[4] J. K. Lowe, B. Gowans, L. Brox, Cancer Res. 1977, 37, 3013.
[5] J. M. Wilson, B. S. Mitchell, P. E. Daddona, W. N. Kelley, J. Clin. Invest. 1979, 64, 1475.
[6] B. Evans, R. Wolfenden, J. Am. Chem. Soc. 1970, 92, 4751.
[7] M. Ohno, N. Yagisawa, S. Shibahara, S. Kondo, K. Maeda, H. Umezawa, J. Am. Chem. Soc. 1974, 96, 4326.
[8] H. Nakamura, G. Koyama, Y. Iitaka, M. Ohno, N. Yagisawa, S. Kondo, K. Maeda, H. Umezawa, J. Am.
Chem. Soc. 1974, 96, 4327.
[9] P. W. K. Wood, H. W. Dion, S. M. Lange, L. F. Dahl, L. Durham, J. Heterocycl. Chem. 1974, 11, 641.
[10] S. Cha, R. P. Agarwal, R. E. Parks Jr., Biochem. Pharmacol. 1975, 24, 2187.
[11] W. M. Kati, S. A. Acheson, R. Wolfenden, Biochemistry 1992, 31, 7356.
[12] D. K. Wilson, F. B. Rudolph, F. A. Quiocho, Science (Washington, D.C.) 1991, 252, 1278.
[13] A. J. Sharf, D. K. Wilson, Z. Chang, F. A. Quiocho, J. Mol. Biol. 1992, 226, 917.
[14] Z. Wang, F. A. Quiocho, Biochemistry 1998, 37, 8314.
[15] M. D. Erion, M. R. Reddy, J. Am. Chem. Soc. 1998, 120, 3295.
[16] C. Altona, M. Sundaralingam, J. Am. Chem. Soc. 1972, 94, 8205.
[17] For a comprehensive review of these concepts, see W. Saenger, ꢁPrinciples of Nucleic Acid Structureꢀ,
Springer-Verlag, New York, 1984.
[18] C. Thibaudeau, A. Kumar, S. Bekiroglu, A. Matsuda, V. E. Marquez, J. Chattopadhyaya, J. Org. Chem.
1998, 63, 5447.
[19] A. A. Koshkin, S. K. Singh, P. Nielsen, V. K. Rajwanshi, R. Kumar, M. Meldgaard, C. E. Olsen, J. Wengel,
Tetrahedron 1998, 54, 3607.
[20] M. A. Siddiqui, H. Jr. Ford, C. George, V. E. Marquez, Nucleosides Nucleotides 1996, 15, 235.
[21] A. Ezzitouni, V. E. Marquez, J. Chem. Soc., Perkin Trans. 1 1997, 1073.
[22] V. E. Marquez, P. Russ, R. Alonso, M. A. Siddiqui, K.-J. Shin, C. George, M. C. Nicklaus, F. Dai, H. Ford
Jr., Nucleosides Nucleotides 1999, 18, 521.
[23] K. Biggadike, A. D. Borthwick, A. M. Exall, B. E. Kirk, S. M. Roberts, P. Youds, A. M. Z. Slawin, D. J.
Williams, J. Chem. Soc., Chem. Commun. 1987, 255.
[24] K. Biggadike, A. D. Borthwick, D. Evans, A. M. Exall, B. E. Kirk, S. M. Roberts, L. Stephenson, P. Youds, J.
Chem. Soc., Perkin Trans 1 1988, 549.
[25] A. Ezzitouni, P. Russ, V. E. Marquez, J. Org. Chem. 1997, 62, 4870.
[26] S. Obika, N. Daishu, Y. Hari, K.-I. Morio, Y. In, T. Ishida, T. Imanishi, Tetrahedron Lett. 1997, 38, 8735.
[27] C. Thibaudeau, J. Plavec, J. Chattopadhyaya, J. Org. Chem. 1996, 61, 266.
[28] For a comprehensive review of these concepts, see C. Thibaudeau, J. Chattopadhyaya, ꢁStereoelectronic
Effects in Nucleosides and Nucleotides and their Structural Implicationsꢀ, Uppsala University Press,
Uppsala, Sweden, 1999.
[29] S. D. Lindell, B. A. Moloney, B. D. Hewit, C. G. Earnshaw, P. J. Dudfeld, J. E. Dancer, Bioorg. Med. Chem.
Lett. 1999, 9, 1985.
Received August 23, 1999