Discovery of AMP mimetics that exhibit high inhibitory potency and specificity for AMP deaminase

Article abstract of DOI:10.1021/ja983153j

The first potent, specific, and cell-penetrable AMP deaminase (AMPDA) inhibitors were discovered through an investigation of 3-substituted 3,6,7,8-tetrahydroimidazo[4,5-d][1,3]diazepin-8-ol analogues. Inhibition constants for the most potent inhibitors we

Full text of DOI:10.1021/ja983153j

308
J. Am. Chem. Soc. 1999, 121, 308-319
Discovery of AMP Mimetics that Exhibit High Inhibitory Potency
and Specificity for AMP Deaminase
Mark D. Erion,* Srinivas Rao Kasibhatla, Brett C. Bookser, Paul D. van Poelje,
M. Rami Reddy, Harry E. Gruber, and James R. Appleman
Contribution from Metabasis Therapeutics, Inc., 9390 Towne Centre DriVe,
San Diego, California 92121
ReceiVed September 2, 1998
Abstract: The first potent, specific, and cell-penetrable AMP deaminase (AMPDA) inhibitors were discovered
through an investigation of 3-substituted 3,6,7,8-tetrahydroimidazo[4,5-d][1,3]diazepin-8-ol analogues. Inhibition
constants for the most potent inhibitors were 105-fold lower than the KM for the substrate AMP. High affinity
required the presence of both the 8-hydroxyl and the 3-substituent and is postulated to arise from a cooperative
interaction that reduces binding entropy costs and enables the diazepine base to adopt a binding conformation
that mimics the transition-state (TS) structure. The high specificity of the inhibitor series for AMPDA relative
to other AMP-binding enzymes (>105) is attributed in part to the diazepine base which favors interactions
with residues used to stabilize the TS structure and precludes interactions typically used by AMP-binding
enzymes to bind AMP. In contrast, discrimination between AMPDA and adenosine deaminase (ADA), two
enzymes postulated to stabilize a similar TS structure, is highly dependent on the 3-substituent. Replacement
of the ribose group in the potent ADA inhibitor coformycin (Ki (ADA) ) 10-11 M vs Ki (AMPDA) ) 3 ×
10-6 M) with 3-carboxy-4-bromo-5,6,7,8-tetrahydronaphthylethyl led to a >1010-fold change in specificity
(Ki (ADA) > 10-3 M vs Ki (AMPDA) ) 2 × 10-9 M). Inhibitors from the series readily penetrate cells and
inhibit intracellular AMPDA activity. Incubation of isolated rat hepatocytes with AMPDA inhibitors had no
effect on secondary metabolite levels during normoxic conditions but led to increased adenosine production
and adenylate sparing under conditions that induce net ATP breakdown. These results suggest that inhibitors
of AMPDA may represent site- and event-specific drugs that could prevent or attenuate ischemic tissue damage
resulting from a stroke or a heart attack.
Proteins that bind nucleotides represent drug targets with
Despite the therapeutic potential of these drug targets, few
ligands are known that possess suitable potency, specificity, and
bioavailability for clinical development. Discovery of potent and
bioavailable ligands is hindered by the hydrophilic nature of
nucleotide binding site cavities and the large proportion of the
total binding affinity attributed to electrostatic interactions
between positively charged amino acid residues and the
negatively charged phosphate group. For example, the binding
affinity of adenosine for adenylate binding sites is usually 4-6
orders of magnitude less than the affinity of the phosphorylated
compound, e.g., ATP or AMP.6 The dilemma in the design of
inhibitors for intracellular enzymes, therefore, stems from the
importance of the phosphate or a highly charged phosphate
surrogate for binding affinity and the inability of charged
molecules to diffuse across cell membranes and penetrate cells.
Enzyme specificity also represents a considerable challenge,
since many enzyme targets utilize the same ligand. This is
especially true for enzymes that bind ATP or AMP, given
nature’s widespread use of these nucleotides as phosphate donors
and allosteric regulators. Consequently, suitable ATP- and AMP-
binding site ligands require extraordinarily high ligand specifici-
enormous untapped potential. Current targets under investigation
within the pharmaceutical industry include P2X and P2Y recep-
tors, which bind nucleotides such as ATP and UTP,1 GTP-
binding proteins,2 transcription factors, polymerases and other
DNA- and RNA-binding proteins, and enzymes within cell-
signaling pathways which utilize ATP as a phosphate donor,
e.g., tyrosine kinases,3 cell cycle kinases, and mitogen-activated
protein kinase.4 Other targets include enzymes that control flux
through biosynthetic pathways by using nucleotide binding at
an allosteric site as an on/off switch. The nucleotide most
commonly used for this purpose is AMP since the intracellular
AMP level is a good indicator of the cell energy status.5
* To whom correspondence should be addressed. Phone: 619-622-5520.
FAX: 619-458-3504. E-mail: erion@mbasis.com.
(1) (a) Jacobson, K. A.; Kim, Y. C.; Wildman, S. S.; Mohanram, A.;
Harden, T. K.; Boyer, J. L.; King, B. F.; Burnstock, G. J. Med. Chem.
1998, 41, 2201-2206. (b) Jianguo, G. G.; MacDermott, A. B. Nature 1997,
389, 749-753.
(2) Mittal, R.; Ahmadian, M. R.; Goody, R. S.; Wittinghofer, A. Science
1996, 273, 115-117.
(3) (a) Escalante, C. R.; Yie, L.; Thanos, D.; Aggarwal, A. K. Nature
1998, 391, 103-106. (b) Levitzki, A.; Gazit, A. Science 1995, 267, 1782-
1788.
(4) (a) Norman, T. C.; Gray, N. S.; Koh, J. T.; Schultz, P. G. J. Am.
Chem. Soc. 1996, 118, 7430-7431. (b) Tong, L.; Pav, S.; White, D. M.;
Rogers, S.; Crane, K. M.; Cywin, C. L.; Brown, M. L.; Pargellis, C. A.
Nat. Struct. Biol. 1997, 4, 311-316.
(5) (a) Hardie, D. G. Nature 1994, 370, 599-600. (b) Corton, J. M.;
Gillespie, J. G.; Hardie, D. G. Curr. Biol. 1994, 4, 315-324. (c) Moore,
F.; Weekes, J.; Hardie, D. G. Eur. J. Biochem. 1991, 199, 691-697.
(6) In addition to having decreased electrostatic interactions, most
nucleotide mimetics with decreased negative charge use groups that are
structurally dissimilar to phosphate based on geometry, anionic charge, and
molecular size. In a few cases, modest binding affinity is retained. For
example, a series of cyclic dicarboxylic acids have replaced the phosphate
of cyclic guanosine 3′,5′-monophosphate with micromolar inhibitor poten-
cies: Tulshian, D.; Czarniecki, M.; Doll, R. J.; Ahn, H.-S. J. Med. Chem.
1993, 36, 1210-1220.
10.1021/ja983153j CCC: $18.00 © 1999 American Chemical Society
Published on Web 12/31/1998

AMP Deaminase Inhibitors
J. Am. Chem. Soc., Vol. 121, No. 2, 1999 309
Scheme 1. Purine Catabolic Pathways
ties or risk generation of dose-limiting toxicities. Unfortunately,
discovery of highly specific ligands is hampered by the close
similarity of the binding-site architectures of these enzymes as
revealed by their X-ray structures.7
adenosine levels is expected in nonischemic tissues since these
tissues are associated with no net ATP breakdown and therefore
have low levels of intracellular AMP.14 The postulated site- and
event-specific action of AMPDA inhibitors is of prime impor-
tance since earlier attempts to directly stimulate adenosine
receptors using nonspecific and receptor subtype-specific ago-
nists consistently failed to produce a successful clinical candidate
because of an unacceptably narrow therapeutic index presumably
arising from simultaneous stimulation of adenosine receptors
in ischemic as well as nonischemic tissues.12 Evidence sup-
porting AMPDA as an attractive alternative target for cardio-
protective drugs is reported in a recent study linking partial
AMPDA deficiency to improved prognosis in heart failure
patients.15
One well-recognized drug target with promising therapeutic
potential is the cytosolic enzyme AMP deaminase (AMPDA,
EC 3.5.4.6). AMPDA catalyzes the deamination of AMP to IMP
as part of the purine catabolic pathway (Scheme 1).8 Like many
other AMP-binding enzymes, no AMPDA inhibitors have been
discovered that simultaneously exhibit high affinity, high
specificity, and good cell penetration.9 Consequently, the
physiological role of AMPDA and the therapeutic utility of
AMPDA inhibitors are poorly defined. Initial interest in
AMPDA as a drug target arose from studies indicating that the
human intraerythrocytic malaria parasite, unlike most mam-
malian cells, lacks the de novo purine biosynthesis pathway and
therefore is highly dependent on salvage pathways for the
synthesis of purine nucleotides, especially guanylates.10 Ac-
cordingly, AMPDA inhibitors were expected to decrease IMP
levels in parasite-infected erythrocytes and thereby inhibit
parasite growth by guanylate starvation.
More recently, AMPDA has attracted considerable interest
as a potential target for drugs that prevent or limit the tissue
damage caused by prolonged ischemic events such as myocar-
dial infarction or stroke.11 Cells in ischemic tissues naturally
attempt to limit cellular damage by breaking down intracellular
ATP to adenosine which then diffuses out of the cell and
activates nearby membrane receptors. Activation of one or more
of the four adenosine receptors (A1, A2A, A2B, or A3) leads to
a variety of tissue-protective pharmacologies, including vasodi-
lation, decreased oxygen consumption, and decreased inflam-
mation.12 Inhibitors of AMPDA are expected to further enhance
the surge of adenosine production and thereby gain additional
pharmacological benefits since a significant portion of the ATP
breakdown product and adenosine precursor, AMP, is consumed
through deamination to IMP.13 Little or no elevation of
Herein, we describe the first compounds with potencies and
specificities sufficient for the delineation of the physiological
role of AMPDA and its potential as a drug target.
Inhibitor Design
Transition-state (TS) inhibitors frequently exhibit extraordi-
narily high binding affinities and enzyme specificities.16 High
affinity is achieved because TS inhibitors engage in the full
complement of interactions used by the enzyme during catalysis
to preferentially stabilize the TS and thereby lower the reaction
energy barrier. High specificity is achieved because the set of
interactions used to stabilize a TS are highly dependent on both
the specific reaction catalyzed by the enzyme and its unique
set of small molecule substrates. Both properties are important
in the design of AMPDA inhibitors, since a large enhancement
in binding affinity would enable removal of the highly charged
phosphate group of AMP and thereby yield compounds that
are both cell penetrable and potent inhibitors. TS inhibitors of
AMPDA would also exhibit high AMPDA specificity since they
differ structurally from the ground-state structure, i.e., AMP,
and therefore likely bind with low affinity at other AMP-binding
sites.
The TS structure recognized and stabilized by AMPDA is
not known in great detail but is expected to resemble the well-
characterized TS structure utilized by the related enzyme,
adenosine deaminase (ADA),17 based on the requirement of zinc
for efficient catalysis by both enzymes and the similarity in pKa
values and solvent isotope effects.18 In the case of ADA,
deamination is postulated to proceed via the tetrahedral inter-
mediate 1, which is produced by a rate-limiting, zinc-assisted
addition of a hydroxyl ion to C6 of the N1-protonated adenine
(7) For example, see structures of: (a) glycogen phosphorylase b: Sprang,
S. R.; Withers, S. G.; Goldsmith, E. J.; Fletterick, R. J.; Madsen, N. B.
Science 1991, 254, 1367-1371. (b) Phosphofructokinase: Schirmer, T.;
Evans, P. R. Nature 1990, 343, 140-145. (c) Fructose 1,6-bisphosphatase:
Ke, H.; Liang, J.; Zhang, Y.; Lipscomb, W. N. Biochemistry 1991, 30,
4412-4420. (d) Adenylate kinase: Schultz, G. E.; Elzinga, M.; Marx. F.;
Schirmer, R. H. Nature 1974, 250, 120-123.
(8) (a) Zielke, C. L.; Suelter, C. H. J. Biol. Chem. 1971, 246, 1313-
1317. (b) Lowenstein, J. M. Physiol. ReV. 1972, 52, 382-414. (c) Bontemps,
F.; Van den Berghe, G.; Hers, H. G. J. Clin. InVest. 1986, 77, 824-830.
(9) The first AMPDA inhibitors were neither potent nor cell penetrable.
Hampton, A.; Sasaki, T.; Perini, F.; Slotin, L. A.; Kappler, F. J. Med. Chem.
1976, 19, 1029-1033.
(10) (a) Webster, H. K.; Whaun, J. M.; Walker, M. D.; Bean, T. L.
Biochem. Pharmacol. 1984, 33, 1555-1557. (b) Zielke, C. L.; Suelter, C.
H. In The Enzymes, 3rd ed.; Boyer, P. D., Ed.; Academic Press: New York,
1971; Vol. IV, pp 72-73.
(11) (a) Gruber, H. E. U.S. Patent 4,912,092, 1990. (b) Erion, M. D.;
Bookser, B. C., Kasibhatla, S. R. U.S. Patent 5,731,432, 1998.
(12) (a) Jacobson, K. A.; van Galen, P. J. M.; Williams, M. J. Med.
Chem. 1992, 35, 407-422. (b) Erion, M. D. Annu. Rep. Med. Chem. 1993,
28, 295-304.
(13) (a) Chen, W.; Hoerter, J.; Gueron, M. J. Mol. Cell. Cardiol. 1996,
28, 2163-2174. (b) Chen, W.; Gueron, M. J. Mol. Cell. Cardiol. 1996, 28,
2175-2182.
(14) Tagetmeyer, H. J. Mol. Cell. Cardiol. 1985, 17, 1013-1018.
(15) Loh, E.; Rebbeck, R. R.; Mahoney, P. D.; Stolle, C. A.; Swain, J.
L.; Holmes, E. W. Circulation 1996, 94, (Suppl. 1) 2534.
(16) Wolfenden, R. Acc. Chem. Res. 1972, 5, 10-18.
(17) (a) Wolfenden, R.; Wentworth, D. F.; Mitchell, G. N. Biochemistry
1969, 16, 5071-5077. (b) Evans, B. E.; Wolfenden, R. Biochemistry 1973,
12, 392-397. (c) Kurz, L. C.; Frieden, C. Biochemistry 1983, 22, 382-
389. (d) Kurz, L. C.; Frieden, C. Biochemistry 1987, 26, 8450-8457. (e)
Weiss, P. M.; Cook, P. F.; Hermes, J. D.; Cleland, W. W. Biochemistry
1987, 26, 6, 7378-7384.
(18) (a) Merkler, D. J.; Schramm, V. L. Biochemistry 1993, 32, 5792-
5799. (b) Merkler, D. J.; Kline, P. C.; Weiss, P.; Schramm, V. L.
Biochemistry 1993, 32, 12993-13001. (c) Bagdassarian, C. K.; Schramm,
V. L.; Schwartz, S. D. J. Am. Chem. Soc. 1996, 118, 8825-8836.

310 J. Am. Chem. Soc., Vol. 121, No. 2, 1999
Erion et al.
Scheme 2
Scheme 3. Hydration of Purine Riboside
base (Scheme 2). Irreversible breakdown of the intermediate
produces the 6-oxo purine product inosine and ammonia.
Evidence for the tetrahedral intermediate is derived from
pseudosubstrate kinetic data,17b inhibitory constants for mol-
ecules bearing close resemblance to 1,17a UV-difference17c and
13C NMR spectra,17d inverse solvent isotope effects,17e and more
recently, crystallographic data for ADA-inhibitor complexes.19
Most telling were studies with purine riboside (2), which
hydrates to an extremely limited extent in aqueous solution (Keq
) 10-7), but was nevertheless found in the hydrated form (3)
in the ADA-inhibitor complex (Scheme 3).19a The affinity of
3 was estimated to be 10-13 M,20 which presumably reflects its
close structural resemblance to the ADA TS structure. The
molecular basis for the high affinity is apparent from the X-ray
structure of the ADA complex, which shows multiple interac-
tions between the 6-hydroxyl group and both the nearby active-
site residues and the zinc ion.
Figure 1. Murine adenosine deaminase interactions with 3 based on
the reported X-ray structure.19 Active-site residues in bold represent
residues found in ADA. Residues and associated residue numbers listed
underneath represent corresponding residues for the L-, E-, and
M-isoenzymes of AMP deaminase, respectively, after sequence align-
ment.
a short-lived, TS-like, tetrahedral intermediate similar to 1,
generated by addition of a hydroxyl ion to N1-protonated
AMP.18b Last, the ADA TS inhibitor coformycin (4)24 is a highly
potent inhibitor of AMPDA as its corresponding 5′-monophos-
phate (5).25 Coformycin inhibits human erythrocytic ADA with
AMPDA is expected to stabilize a similar TS structure on
the basis of several findings. First, the aligned sequences for
the E-, L-, and M-AMPDA isoenzymes21 with ADA22 show
that the zinc-binding residues of ADA, namely, His15, His17,
His214, and Asp295, and the catalytic residues, namely, His238,
Asp296, and Glu217, are completely conserved (Figure 1).23,18a
Second, 14C and 15N heavy-atom kinetic isotope effects support
(19) (a) Wilson, D. K.; Rudolph, F. B.; Quiocho, F. A. Science 1991,
252, 1278-1284. (b) Sharff, A. J.; Wilson, D. K.; Chang, Z.; Quiocho, F.
A. J. Mol. Biol. 1992, 226, 917-921. (c) Wilson, D. K.; Quiocho, F. A.
Biochemistry 1993, 32, 1689-1694. (d) Wilson, D. K.; Quiocho, F. A.
Struct. Biol. 1994, 1, 691-694.
(20) (a) Kati, W. M.; Wolfenden, R. Science 1989, 243, 1591-1593.
(b) Kati, W. M.; Wolfenden, R. Biochemistry 1989, 28, 7919-7927. (c)
Jones, W.; Kurz, L. C.; Wolfenden, R. Biochemistry 1989, 28, 1242-1247.
(d) Jones, W.; Wolfenden, R. J. Am. Chem. Soc. 1986, 108, 7444-7445.
(21) (a) Sabina, R. L.; Morisaki, T.; Clarke, P.; Eddy, R.; Shows, T. B.;
Morton, C. C.; Holmes, E. W. J. Biol. Chem. 1990, 265, 9423-9433. (b)
Morisaki, T.; Sabina, R. L.; Holmes, E. W. J. Biol. Chem. 1990, 265,
11482-11486. (c) Mahnke-Zizelman, D. K.; Sabina, R. L.; J. Biol. Chem.
1992, 267, 20866-20877.
a Ki ) 10 pM,24a whereas coformycin 5′-monophosphate inhibits
rabbit muscle AMPDA25 and yeast AMPDA26 with Kis of 60
and 10 pM, respectively. Inhibition was postulated to be by TS
mimicry since in each case only the 8R-hydroxyl stereoisomer
exhibited high affinity.24b,26 In addition to the 8-hydroxyl, the
diazepine base and the ribose moieties x were also essential for
high ADA affinity with the absence of any one component
(24) (a) Agarwal, R. P.; Spector, T.; Parks, R. E., Jr. Biochem. Pharmacol.
1977, 26, 359-367. (b) Radzicka, A.; Wolfenden, R. Methods Enzymol.
1995, 249, 284-312. (c) Bzowska, A.; Lassota, P.; Shugar, D. Z.
Naturforsch. 1985, 40c, 710-714.
(25) Frieden, C.; Kurz, L. C.; Gilbert, H. R. Biochemistry 1980, 19,
5303-5309.
(26) Merkler, D. J.; Brenowitz, M.; Schramm, V. L. Biochemistry 1990,
29, 8358-8364.
(22) Yeung, C. Y.; Ingolia, D. E.; Roth, D. B.; Shoemaker, C.; Al-Ubaidi,
M. R.; Yen, J. Y.; Ching, C.; Bobonis, C.; Kaufman, R. J.; Kellems, R. E.
J. Biol. Chem. 1985, 260, 10299-10307.
(23) Sequence alignment compared murine ADA with human E-, L-,
and M-AMPDA isoenzymes.

AMP Deaminase Inhibitors
J. Am. Chem. Soc., Vol. 121, No. 2, 1999 311
Table 1. Inhibition Constants
compd AMPDA Ki (µM) ADA Ki (µM) Ki (ADA)/Ki (AMPDA)
producing a >106-fold decrease in potency (e.g., the deribosy-
lated analogue 6). These results were interpreted as indicating
that reduced entropy costs contribute significantly to the high
binding affinity.27
4
5
6
8
9
10
13
14
22
3
0.00006a,c
>1000
4
0.00001a,b
>1000a,d
50e
∼10-5
>107
<10-1
∼10-3
10-1
Two TS inhibitor design strategies were explored during the
course of our studies. One strategy focused on purine riboside
analogues since the hydrated species of purine riboside (3) is
thought to gain its extraordinarily high affinity for ADA through
TS mimicry and therefore should be readily recognized by the
AMPDA purine binding site. The major complication of this
approach is that the hydrated form is virtually nonexistent in
aqueous solutions (Keq ) 10-7) which in turn results in the
overall modest inhibitory potency associated with 3 (Ki ) 10-5
M). Accordingly, we focused on the identification of structural
modifications of purine riboside that enhance hydration of the
1,6-double bond without compromising the affinity of the
hydrated species for the deaminase binding site. Our initial work
entailed the development of a method for calculating relative
hydration free energies using a combination of quantum
mechanical calculations and free energy perturbation methodol-
ogy28 and incorporating the results into predictions of the relative
binding affinities of ADA inhibitors.29 The accuracy of these
predictions and the similarity of the ADA and AMPDA purine
binding sites suggest that a similar strategy could be used in
the design of AMPDA inhibitors.
The second strategy, and the subject of this manuscript,
entailed the design of coformycin analogues. Coformycin (4)
is a 50000-fold weaker inhibitor of AMPDA (Ki ) 3 µM)30
relative to coformycin monophosphate (Ki ) 0.00006 µM),25
which is consistent with the relative affinity decreases observed
for other nucleotide-binding enzymes with the corresponding
nucleoside analogue. Nevertheless, coformycin represents a good
inhibitor of AMPDA and a compound that does not suffer from
the limitations of coformycin monophosphate, namely instability
to plasma phosphatases and lack of cell membrane permeability.
Coformycin itself, however, is not suitable for development as
an AMPDA inhibitor because of its high ADA inhibitory
potency (Ki ) 10-11 M) and the strong association between
chronic ADA inhibition and severe immunodeficiency.31 Ac-
cordingly, the use of coformycin as a lead TS inhibitor of
AMPDA requires modification of the structure in a manner that
retains or improves the AMPDA binding affinity while simul-
taneously reversing the 105-fold preference of coformycin for
ADA.
0.006
6.1
>1000
>1000
>1000
45
0.4
0.015
0.002
>1000
>103
>105
>106
a Ki’s determined for the 8R-stereoisomer; all other Ki’s, except Ki
for compound 22, were determined for the 8R/S stereoisomeric mixture
and therefore are likely to be 2-fold higher than the value for the 8R-
stereoisomer. b Reference 24a. c Reference 25. d Reference 24c. e A
value of 40 µM was determined for the 8R-stereoisomer, ref 27a.
Scheme 4. General Synthesis of N3-Substituted Coformycin
Aglycon Analogues
AMPDA active site which likely contains numerous positively
charged residues positioned to accommodate the negatively
charged phosphate group, whereas the corresponding residues
in the ADA site are hydrophobic. Exploitation of these differ-
ences, however, requires the N3-substituent to also maintain
optimal alignment of the base or risk compromising the TS
mimicry and therefore the high binding affinity.
Results
N3-substituted coformycin aglycon analogues were synthe-
sized according to the general reaction sequence shown in
Scheme 4. Treatment of the known heterocycle 6,7-dihydroimi-
dazo-[4,5-d][1,3]diazepin-8(3H)-one (7),32 which was prepared
in eight steps from 4-methylimidazole, with NaH in DMF
followed by the electrophile and NaI gave the N3-alkylation
product in modest yield.33 Reduction of the 8-keto group to the
8-R/S-alcohol with sodium borohydride followed, when neces-
sary, by treatment with NaOH to hydrolyze the carboxylate ester
produced the final product in good overall yield. Two analogues
are highlighted here, namely compound 8, which was prepared
Our approach focused on N3-substituted coformycin aglycon
analogues since substituents in this region of the active site can
clearly discriminate between ADA and AMPDA as illustrated
by the >105-fold difference in specificity exhibited by 4 and 5
(Table 1). Presumably, the large difference in binding affinity
arises from the presence of a phosphate binding site in the
(27) (a) Kati, W. M.; Acheson, S. A.; Wolfenden, R. Biochemistry 1992,
31, 7356-7366. (b) Wolfenden, R.; Wentworth, D. F.; Mitchell, G. N.
Biochemistry 1977, 16, 5071-5077. Similarly, interactions outside of the
base binding region were important for cytidine deaminase, see: Carlow,
D. C.; Short, S. A.; Wolfenden, R. Biochemistry 1998, 37, 1199-1203.
(28) Erion, M. D.; Reddy, M. R. J. Comput. Chem. 1995, 12, 1513-
1521.
using 4-propoxyphenylpropyl methanesulfonate as the electro-
phile, and compound 10, which was prepared using 6-mesyloxy-
2-benzylhexanoic acid ethyl ester.
Compound 8 inhibited AMPDA and ADA with a Ki of 4
µM and 6 nM, respectively (Table 1). The ADA inhibitory
potency represented a 104-fold improvement in binding affinity
relative to the base alone (6) suggesting that the hydrophobic
N3-substituent formed highly favorable interactions with the
ADA binding site. To assess the binding interactions responsible
(29) Erion, M. D.; Reddy, M. R. J. Am. Chem. Soc. 1998, 120, 3295-
3304.
(30) Coformycin (4) is reported to inhibit rabbit muscle AMPDA30a with
Ki ) 50 nM and yeast AMPDA30b with Ki ) 2 µM. In our studies, the R/S
racemic mixture of 4 inhibited both porcine heart AMPDA and recombinant
human L-type AMPDA with Ki ) 3 µM (Table 1). (a) Agarwal, R. P.;
Parks, R. E., Jr. Biochem. Pharmacol. 1977, 26, 663-666. (b) Merkler, D.
J.; Schramm, V. L. J. Biol. Chem. 1990, 265, 4420-4426.
(31) Kredich, N. M.; Hershfield, M. S. In The Metabolic Basis of
Inherited Disease, 6th ed.; Scrivner, C. R., Eds.; McGraw-Hill: New York,
1989; pp 1045-1075.
(32) Chan, E.; Putt, S. R.; Showalter, H. D. H.; Baker, D. C. J. Org.
Chem. 1982, 47, 3457-3464.
(33) Bookser, B. C.; Kasibhatla, S. R.; Appleman, J. R.; Erion, M. D.,
unpublished results.

312 J. Am. Chem. Soc., Vol. 121, No. 2, 1999
Erion et al.
for the high ADA preference, 8 was evaluated by using the
murine ADA X-ray structure19 and a method for conformation
analysis that combines a Monte Carlo searching strategy with
energy minimization.34 The lowest energy conformations showed
the N3-substituent to reside in a different region of the binding
site cavity than that occupied by the ribose moieties of 3 and 4.
Analysis of the binding interactions depicted in Figure 2
indicated that the low-energy binding conformations were
stabilized through van der Waals contacts between the alkylaryl
substituent and a hydrophobic cavity produced by the Gly184
R-methylene and the side chains of Met155, Leu106, Leu62,
and Ile188 (Leu62 is faint and Ile188 is not shown in Figure
2). In addition, the ether oxygen, which enhances binding
affinity 10-fold relative to the corresponding alkyl analogue,35
formed a hydrogen bond with His157 and a good electrostatic
interaction with the main chain NH of Asp185. In contrast, the
ribosyl group binds preferentially to the other face of the binding
site due presumably to hydrogen bonds formed between the 5′-
hydroxyl group and both Asp19 and His17.
Figure 2. Overlay of purine riboside hydrate (3) (red), coformycin
(4) (white), and compound 8 (yellow) in ADA active site following
inhibitor docking and energy minimization (see Experimental Section).
Zinc ion is represented by the green sphere. Atomic distances are
denoted for the electrostatic interactions between the ether oxygen of
8 and both the imidazole of His157 (2.7 Å) and the backbone amido
hydrogen of Asp185 (4.3 Å).
Compound 8 also showed an improvement, albeit only 250-
fold, over the base 6 in AMPDA inhibitory potency. This gain
in potency suggested that, like ADA, the AMPDA binding site
has significant hydrophobic character. The larger increase in
ADA inhibitory potency, however, indicated that hydrophobic
groups at N3 were unlikely to yield compounds with high
AMPDA specificity. In contrast, compounds with a negatively
charged carboxylate group attached to the hydrophobic N3-
substituent at the appropriate position exhibited high AMPDA
specificity. For example, compound 10 showed no ADA
inhibition even at concentrations of 1 mM. In general, inclusion
of a carboxylic acid in the N3-substituent resulted in a 10- to
100-fold increase in AMPDA binding affinity (e.g., 10 vs 9)
and complete specificity for AMPDA.
Synthesis of over 100 other carboxylate-containing com-
pounds led to the discovery of a series of highly potent and
selective AMPDA inhibitors containing a 3-carboxy-5,6,7,8-
tetrahydronaphthylethyl group at N3.35 The electrophiles for this
series of compounds were synthesized from compound 11,
which was prepared from cyclohexanone in one step by using
the procedure reported by Boger and co-workers36 with some
minor modifications (Scheme 5). Conversion of 11 to the desired
electrophile 12 was accomplished in two steps and 54% overall
yield using a Diels-Alder reaction with dihydrofuran followed
by reaction of the crude tricyclic intermediate with BBr3.
Reaction of the electrophile with 7 under the conditions
described earlier gave 13. The brominated analogue 14 was
prepared in a similar manner using bromide 19, which was
synthesized by using the route shown in Scheme 5. Intermediate
15 was prepared in one step from 11, following the reported
procedure.36 Reaction of 15 with Br2 in acetic acid gave a single
brominated product whose structure was assigned as 16 on the
basis of NOE experiments. The styryl derivative 17 was
prepared via a Stille coupling on the intermediate triflate.37
Hydroboration of 17 followed by oxidation gave the primary
alcohol 18, which was converted to the bromide 19 with CBr4
in the presence of PPh3.
Compounds 13 and 14 inhibited AMPDA with Kis of 15 and
2 nM, respectively, which represents a 25- to 200-fold improve-
ment in inhibition relative to that by the initial inhibitor 10
(Table 1). No appreciable ADA inhibition was observed at a
concentration of 1000 µM. These results indicate that 14 is an
AMPDA inhibitor with a potency 1000-fold greater than that
of coformycin and only 33-fold lower than that of coformycin
monophosphate. Furthermore, the high AMPDA specificity of
14 corresponds to a >1010-fold reversal in the AMPDA/ADA
specificity relative to that of coformycin.
To assess whether the increased affinities of 13 and 14 were
due solely to increased interactions with active-site residues or
whether the side chain, like ribose-5-phosphate, resulted in good
affinity through a synergistic interaction with the 8-hydroxy
tetrahydroimidazol-diazepine base, we prepared the 8-deoxy
analogue of 14 (Scheme 6). 4(5)-Nitrohistamine hydrochloride
20, available from histamine in three steps,38 was hydrogenated
in moderate yield (32%) using triethyl orthoformate to trap the
unstable diamine hydrogenation product as the diazepine 21.
(34) Guida, W. C.; Bohacek, R. S.; Erion, M. D. J. Comput. Chem. 1992,
13, 214-228.
(37) Echavarren, A. M.; Stille, J. K. J. Am. Chem. Soc. 1987, 109, 5478-
(35) Kasibhatla, S. R.; Bookser, B. C.; Appleman, J. R.; Probst, G.; Xiao,
W.; Erion, M. D., unpublished results.
5486.
(38) Tautz, W.; Teitel, S.; Brossi, A. J. Med. Chem. 1973, 16, 705-
707.
(36) Boger, D. L.; Mullican, M. D. Org. Synth. 1987, 65, 98-107.

AMP Deaminase Inhibitors
J. Am. Chem. Soc., Vol. 121, No. 2, 1999 313
Scheme 5a
a Reagents: (a) 11,36 toluene, sealed tube, 140 °C, dihydrofuran. (b) BBr3, CH2Cl2; EtOH, 54% (after two steps). (c) Vinylene carbonate, toluene,
sealed tube, see ref 36. (d) Br2, AcOH. (e) Tf2O, pyr. (f) CH2dCHSn(nBu)3, (PPh3)2PdCl2, LiCl, DMF. (g) 9-BBN, THF; 30% H2O2, 3 N NaOH.
(h) CBr4, PPh3, CH2Cl2.
Scheme 6a
a Reagents: (a) 20,38 HC(OMe)3, H2, 10% Pd/C, EtOH. (b) CH2Cl2, Et3N. (c) NaH, DMF, NaI, 23.39 (d) 10% Pd/C, H2 (1 atm), MeOH, (C6H11)2NH.
Alkylation of 21 with the corresponding benzyl ester of 19, 23,39
by using NaH and NaI in DMF provided the N3-alkylation
product, which upon hydrogenolysis in the presence of dicy-
clohexylamine gave 22.40 The large decrease in AMPDA
inhibitor potency of 22 (>105-fold) supported the postulate that
both the base and the N3-substituent were required for potent
inhibition.
The specificity of 13 for the AMP site of AMPDA relative
to other AMP binding enzymes was analyzed using five AMP-
binding enzymes, namely, adenylate kinase, which uses AMP
as a substrate, adenosine kinase, which produces AMP as a
product, and glycogen phosphorylase, fructose 1,6-bisphos-
phatase, and phosphofructokinase, which use AMP as an
allosteric regulator. No activity was observed at concentrations
>100000-fold the AMPDA Ki.
In addition to being evaluated for potency and specificity,
the compound series was evaluated in cell assays to determine
the degree of cell penetration and the effect of AMPDA
inhibition on intracellular metabolites. Cell uptake studies
showed that 13 rapidly distributed across cell membranes and
into rabbit erythrocytes to achieve an intracellular concentration
70% of the extracellular concentration in <5 min. Similar results
were found by using bovine endothelial cells (∼50%) and rat
(39) The benzyl ester of bromide 19, compound 23, was prepared from
18 in three steps. Hydrolysis of 18 with 1 N NaOH in dioxane at 70 °C,
followed by alkylation of the carboxylic acid with BnBr in the presence of
K2CO3 in DMF at room temperature yielded the benzyl ester, which was
subsequently treated with CBr4 in the presence of PPh3.
(40) 1H NMR showed showed trace amounts of the debrominated product
(<5%).
hepatocytes (∼100%). Intracellular drug levels were confirmed
by analysis of cell lysate AMPDA activity, which closely
matched the inhibitory activity predicted from the intracellular
drug levels measured by HPLC and the IC50 curve for the
isolated enzyme.
To assess the functional consequences of AMPDA inhibition,
we incubated bovine endothelial cells and isolated rat hepato-
cytes with 13 and monitored the intracellular levels of key
intermediary metabolites. No changes were observed in the
levels of ATP, ADP, IMP, or adenosine (Ado), suggesting that
flux through AMPDA under basal conditions is minimal. Since
flux through AMPDA is expected to increase during net ATP
breakdown,41 we studied isolated rat hepatocytes treated with
fructose, which is known to reproducibly produce a dose-
dependent decrease in ATP levels as a consequence of the ATP
utilization associated with the phosphorylation of fructose by
fructose kinase.42 Similar to results reported in the literature,
fructose at a concentration of 0.5 mg/mL in the cell medium
depleted cellular ATP levels by 50% in 5-10 min under normal
aerobic conditions and elevated downstream metabolites AMP
and IMP (Figure 3A). Identically treated cells in the presence
of 20 µM 13 exhibited a 33-fold increase in the AMP/IMP ratio,
thereby confirming that 13 penetrated the cells and substantially
inhibited AMPDA. Ado was elevated 7-fold over baseline levels
(41) (a) Van den Berghe, G.; Bontemps, M. F.; Van den Berghe, F. Prog.
Neurobiol. 1992, 39, 547-561. (b) Skladanowski, A. C. In Myocardial
Energy Metabolism; de Jong, J. W., Ed.; Martinus Nijhoff Publishers:
Boston, 1988; pp 53-65.
(42) Van den Berghe, G.; Bontemps, F.; Hers, G.-G. Biochem. J. 1980,
188, 913-920.

314 J. Am. Chem. Soc., Vol. 121, No. 2, 1999
Erion et al.
Figure 3. Freshly isolated rat hepatocytes treated with fructose (0.5 mg/mL). (A) AMP and IMP intracellular concentration in hepatocytes preincubated
with either vehicle (open bar) or 13 (20 µM) (shaded bar) for 5 min followed by fructose for 5 min relative to non-fructose-treated hepatocytes
(control, dashed line). (B) Adenosine concentration vs concentration of 13 at 10 min post-fructose addition. (EC50 ) 0.4 µM).
Figure 5. Freshly isolated rat hepatocytes preincubated with vehicle
or 13 (20 µM) for 5 min and then subjected to 15 min of anoxia. Figure
indicates percent adenylate content relative to t ) 0 min. Total adenylate
as well as AMP (p ) 0.007), ADP (p ) 0.017), and ATP (p ) 0.003)
were significantly elevated relative to adenylate levels in the vehicle-
treated cells.
Figure 4. Freshly isolated rat hepatocytes preincubated for 5 min with
vehicle or 13 (20 µM) and then treated with fructose (0.5 mg/mL at t
) 0 min). AMP (squares, solid lines) and Ado (circles, dashed lines)
levels associated with vehicle-treated (open symbols) and drug-treated
(shaded symbols) cells were monitored for 35 min.
phosphate group. Evidence supporting these compounds as TS
inhibitors is two-fold. First, the 8-deoxy analogue of 14
(compound 22) exhibited a 105-fold decrease in binding affinity,
which is consistent with the loss in potency observed in earlier
studies with the 8S-stereoisomer of 2′-deoxycoformycin and
ADA24b and the 8S-isomer of coformycin monophosphate and
AMPDA.26 In the case of ADA, the large contribution of a single
hydroxyl group is attributed to its interactions with the active-
site zinc ion and nearby residues.19 Since the residues used by
ADA to coordinate the zinc ion and bind the TS inhibitor 3 are
completely conserved in the three AMPDA isoenzymes, the
large decrease in inhibitory potency for 22 suggests that the
8-hydroxyl of 14 forms the same hydrogen bond pattern and is
in a position suitable for TS mimicry.
A second observation indicating that the inhibitor series
achieves high potency through TS mimicry is the poor AMPDA
inhibitory potency of the base, i.e., 6, and therefore the high
dependence of inhibition on the N3-substituent. A similar finding
was reported for ADA by Wolfenden in that the ribose group
of coformycin proved to be essential for high affinity presumably
because it reduced entropy losses through assisting the alignment
of the base and the 8-hydroxyl group in a manner most optimal
for TS mimicry.27a Analysis of coformycin monophosphate (5)
binding to AMPDA shows a similar dependency since removal
of the ribose 5-phosphate group led to a 108-fold loss in binding
affinity. Not all of the lost binding affinity is attributed to
decreased TS mimicry since some of the decreased affinity is
associated with removal of the phosphate group (∼104) which
is likely to be reflected in both the ground state and TS.
following exposure to fructose, suggesting that AMPDA inhibi-
tion had the desired effect of diverting ATP breakdown toward
Ado. Most of the Ado (75%) was secreted into the medium as
determined in a separate experiment. In addition, the elevation
of Ado levels was dose-dependent with an EC50 for 13 of 0.4
µM (Figure 3B). Analysis of the time course showed that no
changes in AMP or Ado levels occurred during incubation of
the cells with 13 prior to the addition of fructose, whereas large
increases in both were observed following fructose treatment
(Figure 4).
Similar but less pronounced effects were observed in hepa-
tocytes subjected to anoxia rather than fructose treatment. Under
anoxic conditions, ATP depletion was slower and less extensive
with a 40% reduction observed over 15 min. ADP and AMP
levels rose 30% and 160%, respectively. Treatment with 13 (20
µM) led to significantly higher levels of total adenine nucleotides
relative to that for vehicle-treated cells and to an adenylate pool
equal to that for the normoxic control cells (Figure 5). Consistent
with the high degree of adenylate sparing, Ado was elevated a
modest 1.2-fold in drug- vs vehicle-treated cells (p ) 0.03).
Discussion
Compounds 13 and 14 represent the first potent AMPDA
inhibitors described that are not monophosphates. Both com-
pounds also exhibit high enzyme specificity and good cell
penetration. Success was achieved using a TS inhibitor strategy
as a means to regain a large proportion of the >104-fold loss in
binding affinity associated with removal of the highly charged

AMP Deaminase Inhibitors
J. Am. Chem. Soc., Vol. 121, No. 2, 1999 315
accepting residues near the 6-amino group. Recently published
protein engineering studies suggest that this hydrogen bond
pattern accounts for the high binding affinity and specificity of
these enzymes for adenine-containing ligands (e.g., AMP)
relative to the corresponding 6-oxopurine analogues (e.g.,
GMP).44 Consistent with these reports was our finding that the
TS inhibitor 13 showed no appreciable binding to five other
AMP-binding enzymes. The molecular basis for the high
selectivity is suggested by the ADA-3 X-ray structure which
showed the 6-hydroxyl of 3 to form strong interactions with
active-site residues out of the plane of the base. In contrast,
AMP binding sites typically interact with the 6-amino group
through in-plane interactions and therefore contain residues
poorly positioned to interact with the 8-hydroxyl of coformycin
analogues. Furthermore, like 3, coformycin has a hydrogen bond
donor near the corresponding N1 position of adenine that likely
donates a hydrogen bond to the nearby negatively charged
Glu217, whereas most other AMP binding sites utilize the
opposite hydrogen bond pattern.
Figure 6. Structure overlay of AMP and compound 13 (yellow). The
AMP structure is derived from the Cambridge crystal structure database
(FEXROP, r ) 0.0580). Compound 13 represents the energy-minimized
structure (3.2 kcal/mol above the global minimum) most closely
matching the structure of AMP.
Discovery of compounds with N3-substituents that retain most
of the potency of coformycin monophosphate suggests that
groups other than ribose 5-phosphate can position the diazepine
base in a manner analogous to the TS structure and thereby
gain the synergistic effect of the base and N3-substituent. The
modest 17-fold decrease in potency (assumes only the 8R-
stereoisomer is active) of 14 relative to coformycin monophos-
phate (5) is remarkable, given that the substituent not only had
to maintain good TS mimicry but also had to regain the binding
energy associated with the phosphate moiety. In general,
carboxylic acids are considered relatively poor mimics of
phosphates, given the large difference in geometry (planar vs
tetrahedral), molecular charge, and number of heteroatoms.
Some gain in affinity is possible with a carboxylic acid,
however, if the carboxylic acid is positioned near the positively
charged phosphate binding site. In this case, favorable through-
space electrostatic interactions are possible as was concluded
in earlier studies with purine nucleoside phosphorylase and a
carboxylic acid-containing inhibitor.43 The possibility that the
carboxylate increases binding affinity 10-100 fold33,35 through
interaction with the phosphate binding site is supported by the
finding that low-energy conformations of 13 position the
carboxylate in the same region of conformational space as that
of the phosphate group of AMP (Figure 6).
The high specificity of the inhibitor series for AMPDA over
ADA was dependent on the N3-substituent rather than the
diazepine base since the residues used by ADA to bind the base
were completely conserved in AMPDA. Hydrophobic N3-
substituents led to compounds with deaminase specificities that
favored ADA by as much as 1000-fold and potencies in the
low nanomolar range (e.g., 8, Ki ) 6 nM). In contrast, analogues
with hydrophobic N3-substituents containing a carboxylic acid
achieved extraordinarily high AMPDA specificities. Since the
carboxylic acid provided only a modest gain in AMPDA
inhibitory potency, the large increase in specificity arises
predominantly from the inability of the ADA binding site to
accommodate a negatively charged group. Consistent with this
conclusion are molecular modeling studies showing that inhibi-
tors containing a carboxylic acid had much lower interaction
energies than the corresponding descarboxy analogues. Experi-
mentally, compounds 13 and 14 favored AMPDA binding by
at least 105-fold on the basis of the ratio Ki(ADA)/Ki(AMPDA).
Thus, replacement of the ribosyl moiety of coformycin with
the 3-carboxytetrahydronaphthylethyl group led to a 1010-fold
change in deaminase specificity.
In addition to high AMPDA binding affinity and enzyme
specificity, the compound series showed good cell penetration.
Analysis of intracellular inhibitor concentration over time by
HPLC showed that compound 13 rapidly equilibrated across
cell membranes (e5 min) to achieve an intracellular concentra-
tion approximately 50% of the extracellular concentration. The
intracellular inhibitor concentration was confirmed by determin-
ing AMPDA activity in cell lysates and showing that the degree
of AMPDA inhibition closely corresponded to the inhibition
In addition to inhibitory potency, the diazepine base and N3-
substituent were also essential for the high specificity of the
inhibitor series for AMPDA relative to other AMP-binding
enzymes and adenosine deaminase. Crystallographic studies
indicate that AMP-binding enzymes utilize very similar binding
site cavities and interactions to bind AMP. One feature common
to nearly all AMP binding sites is the presence of a hydrogen
bond-donating residue near N1 and two hydrogen bond-
(44) (a) Erion, M. D.; Takabayshi, K.; Smith, H. B.; Kessi, J.; Wagner,
S.; Honig, S.; Shames, S. L.; Ealick, S. E. Biochemistry 1997, 36, 11725-
11734. (b) Erion, M. D.; Guida, W. C.; Stoeckler, J. D.; Ealick, S. E.
Biochemistry 1997, 36, 11735-11748. (c) Stoeckler, J. D.; Poirot, A. F.;
Smith, R. M.; Parks, R. E., Jr.; Ealick, S. E.; Takabayashi, K.; Erion, M.
D. Biochemistry 1997, 36, 11749-11756.
(43) (a) Erion, M. D.; Niwas, S.; Rose, J. D.; Ananthan, S.; Allen, M.;
Secrist, J. A., III; Babu, Y. S.; Bugg, C. E.; Guida, W. C.; Ealick, S. E.;
Montgomery, J. A. J. Med. Chem. 1993, 36, 3771-3783. (b) Ealick, S. E.;
Babu, Y. S.; Bugg, C. S.; Erion, M. D.; Guida, W. C.; Montgomery J. A.;
Secrist, J. A., III. Proc. Natl. Acad. Sci. U.S.A. 1991, 88, 11540-11544.

316 J. Am. Chem. Soc., Vol. 121, No. 2, 1999
Erion et al.
expected for that inhibitor concentration with the isolated
enzyme. Last, endothelial cells treated with hydrophobic
proesters of 1335 showed intracellular levels of 13 similar to
those for nonproester treated cells, indicating that the generation
of 13 intracellularly through the action of esterases did not result
in a trapping and accumulation of 13 inside the cell. These
studies therefore suggest that 13 enters cells, most likely via
passive diffusion, and that the levels achieved are relatively high
despite the presence of a carboxylic acid and a log P ≈ -0.7.
Previous studies with other carboxylic acids (e.g., NSAIDs,
purine nucleoside phosphorylase inhibitors,43a angiotensin con-
verting enzyme inhibitors, etc.) suggest that carboxylic acids
exhibit good cell penetration properties relative to phosphonic
acids but are less cell penetrable than highly lipophilic com-
pounds.
reason for the difference in the metabolic fate of AMP under
anoxic conditions relative to fructose treatment is unclear but
may relate to differences in adenosine kinase activity (unpub-
lished results). This adenylate sparing activity suggests that
AMPDA inhibition might prevent the tissue damage associated
with conditions such as acute myocardial infarction, stroke, and
heart failure all of which involve net ATP breakdown, decreased
adenylate pools, and irreversible tissue damage.45 Agents that
prevent loss of the adenylate pool may therefore increase the
probability of tissue survival after reperfusion since the ade-
nylates can serve to assist in the repletion of ATP reserves
following reoxygenation of the tissue. Recent studies in heart
failure patients indicate that partial AMPDA deficiency is
associated with improved prognosis based on the observation
that these patients tolerate a longer duration of heart failure
symptoms before referral for transplant evaluation than patients
with normal AMPDA levels.15
Inhibition of AMPDA in Cells
Extracellular adenosine concentration is tightly regulated in
vivo by enzymes controlling adenosine metabolism and produc-
tion. Adenosine production is highly dependent on ATP
breakdown and the resulting increase in intracellular AMP levels
as well as the enzymes that utilize AMP, namely 5′-nucleotidase,
which catalyzes the dephosphorylation of AMP to adenosine,
and AMPDA, which catalyzes the deamination of AMP and
thereby effectively limits adenosine production by diverting
AMP metabolism toward IMP. At the start of our work, flux
through both enzymes was expected to dramatically increase
during hypoxia on the basis of the increase in AMP concentra-
tion (∼1 µM to 1 mM) and the relatively high KM’s of both
enzymes for AMP. Unknown, however, was the proportion of
AMP breakdown catalyzed by AMPDA and therefore the
maximal increase in adenosine production expected for an
AMPDA inhibitor.13
To assess the effect of AMPDA inhibition on secondary
metabolite levels during normoxic conditions and during condi-
tions that result in net ATP breakdown, we analyzed freshly
isolated rat hepatocytes treated with an AMPDA inhibitor and
subjected to either vehicle, fructose, or anoxia. Fructose alone
induced approximately 50% ATP breakdown over a 10 min
period and elevations in the levels of the downstream metabolites
AMP, IMP, and Ado of 4.7-, 6- and 2-fold, respectively. In the
presence of 13 (20 µM), the metabolite profile changed
dramatically as indicated by the 33-fold increase in the AMP/
IMP ratio and the 7-fold rise in total adenosine. Importantly,
AMPDA inhibition had no effect on either AMP or adenosine
levels in the absence of fructose. These results suggest that
AMPDA inhibitors might elevate adenosine levels with high
event specificity and therefore enhance local adenosine levels
only at ischemic sites and not in nonischemic tissues. Accord-
ingly, AMPDA inhibitors might raise adenosine in a manner
that provides pharmacological benefit to the ischemic tissue
while minimizing side effects and unrelated pharmacology.
Another possible benefit of AMPDA inhibition is apparent
from the results obtained with hepatocytes subjected to anoxia.
Similar to fructose, anoxia led to ATP breakdown, an elevation
of AMP levels (160%) and, albeit less pronounced, an elevation
of adenosine levels. In contrast to results obtained with fructose,
cells subjected to anoxia in the presence of 13 showed only a
slight further enhancement of adenosine levels. Instead, there
was a significant increase in the adenylate pool with total
adenylates (ATP + ADP + AMP) achieving a level equal to
normoxic cells. These results suggest that the AMP preserved
through AMPDA inhibition is predominantly reconverted to
ADP and ATP by adenylate kinase and not to adenosine. The
Concluding Remarks
Despite nearly two decades of pharmaceutical research and
impressive pharmacological activity in numerous animal models
of human diseases, no adenosine receptor agonist, with the
exception of adenosine itself, has successfully completed human
clinical trials. The lack of success is attributed to the simulta-
neous activation of adenosine receptors at sites unrelated to the
disease and thereby the production of dose-limiting side effects
and an overall narrow therapeutic window.12,46 As presented in
this paper, AMPDA inhibitors represent an alternative strategy
for activation of adenosine receptors. Our discovery and
characterization of the first potent, specific, and cell penetrable
inhibitor of AMPDA led to the observation that AMPDA
inhibitors elevate adenosine levels only in cells undergoing net
ATP breakdown. Accordingly, AMPDA inhibitors are expected
to elevate adenosine levels in the whole animal only at sites
undergoing an ischemic episode and thereby protect tissues from
ischemic damage incurred during a stroke, heart attack, or related
conditions without simultaneously producing adenosine-medi-
ated side effects.
Experimental Section
General Procedures. Glassware for moisture-sensitive reactions was
flame-dried and cooled to rt in a desiccator. All reactions were carried
out under an atmosphere of nitrogen. Anhydrous solvents were
purchased from Aldrich and stored over 4 Å molecular sieves. THF
was freshly distilled from Na/benzophenone ketyl under nitrogen. Flash
chromatography was performed on 230-400 mesh EM Science silica
gel 60. Melting points were determined in open capillary tubes with a
Thomas-Hoover capillary melting point apparatus and are uncorrected.
1H and 13C NMR were obtained on a Varian Gemini-200 operating at
200 and 50 MHz, respectively. 1H and 13C NMR spectra were recorded
in units δ with tetramethylsilane (δ 0.00) and CDCl3 (δ 77.0) or DMSO-
d6 (δ 39.9) as reference line internal standards, respectively. UV spectra
(200-350 nm) of methanolic solutions were recorded on a Kontron
Uvikon 860. Microanalyses were carried out at Robertson Laboratory,
Inc. Madison, NJ.
3,6,7,8-Tetrahydroimidazo[4,5-d][1,3]diazepin-8-ol (6).47 A mix-
ture of ketone 7 (150 mg, 1 mmol) and NaBH4 (40 mg, 1 mmol) in 5
(45) (a) Ingwall, J. S. Circulation 1993, 87, (Suppl. VII), 58-62. (b)
Nascimben, L.; Friedrich, J.; Liao, R.; Pauletto, P.; Pessina, A.; Ingwall, J.
S. Circulation 1995, 91, 1824-1833. (c) Scheuer, J. Circulation 1993, 87,
(Suppl. VII), 54-57.
(46) (a) Appleman, J. R.; Erion, M. D. Expert Opin. InVest. Drugs 1998,
7, 225-243. (b) Williams, M. Drug DeV. Res. 1993, 28, 438-444.
(47) The 8R-coformycin aglycon analogue was prepared previously with
minimal characterization: (a) Smal, E., Ph.D. Thesis, University of Alabama,
1985. (b) Al-Razzak, L. A.; Bendetti, A. E.; Waugh, W. N.; Stella, V. J.
Pharm. Res. 1990, 7, 452.

AMP Deaminase Inhibitors
J. Am. Chem. Soc., Vol. 121, No. 2, 1999 317
mL of CH2Cl2 and 5 mL of methanol was stirred at rt for 1 h and then
slurried with 2 g of SiO2. The solvent was evaporated and the powder
loaded onto a SiO2 column and eluted with CH3CN/EtOH/NEt3 mixtures
of 10:1:0.1, 5:1:0.06, 3:1:0.04, and 1:1:0.02 to yield 111 mg (73%) of
6 as a brown solid (deliquescent and blackened on standing in air):
mp 124-127 °C; Rf 0.23 (3:1 CH3CN/0.2 M NH4Cl); UV (MeOH)
3-(5-Carboxy-6-phenylhexyl)coformycin Aglycon (10). The ester
9 (4.13 g, 10.7 mmol) was dissolved in 60 mL of dioxane and treated
with 0.25 M NaOH (64 mL, 16 mmol). The mixture was stirred at 55
°C for 14 h, cooled, diluted with 20 mL of water, and extracted with
CH2Cl2. The aqueous layer was slurried with Dowex-1 acetate (ca. 25
mL volume) for 1 h and then filtered. The resin was washed with water
and then batchwise slurried with 0.1 N AcOH (6 × 100 mL) for 20
min each at 5 °C. After filtration, the combined filtrates were frozen
and lyophilized to provide 2.9 g (70%) of the carboxylate 10 as a white
1
205 nm (ꢀ 15700), 283 (9060); H NMR (DMSO-d6) δ 3.24 (s, 2H),
4.84 (br s, 1H), 5.23 (br s, 1H), 7.18 (br s, 1H), 7.30 (s, 1H), 8.06 (br
s, 1H), 11.93 (br s, 1H); 13C NMR (DMSO-d6) δ 47.9 (C-7), 65.9 (C-
8), 127.3 (C-8a), 129.9 (C-5), 136.9 (C-3a), 146.6 (C-2); MS (ES) m/z
153 ([M + H]+), 151 ([M - H]-). An analytical sample was prepared
by adding a small amount of ethanolic HCl to a methanolic (5 mL)
solution of 6 (53 mg). The solvent was evaporated and the residue
triturated with hot MeOH. The resulting white solid was dried under
high vacuum to provide 31 mg of the HCl salt of 6 (deliquescent and
blackened on standing in air): mp 214 °C (dec); 1H NMR (DMSO-d6)
δ 3.47 (dd, 1H, J ) 12, 5 Hz), 3.60 (br d, 1H, J ) 12 Hz), 5.02 (dd,
1H, J ) 5, 2 Hz), 7.65 (s, 1H), 7.95 (d, 1H, J ) 7 Hz). Anal. Calcd for
C6H8N4‚HCl‚0.33H2O: C, 37.04; H, 5.00; N, 28.80. Found: C, 37.51;
H, 5.07; N, 28.13.
6,7-Dihydroimidazo[4,5-d][1,3]diazepin-8(3H)-one (7). A suspen-
sion of 10 g of 6,7-dihydroimidazo-[4,5-d][1,3]diazepin-8(3H)-one
hydrochloride DMSO solvate32 in 50 mL of triethylamine was stirred
for 30 min before adding 500 mL of CH2Cl2. The mixture was then
stirred for 10 min and filtered. The solid was collected and washed
with CH2Cl2 to provide 5.65 g (91%) of 7 as a light brown solid: mp
260 °C; 1H NMR (DMSO-d6) δ 3.79 (br s, 2H), 7.33 (s, 1H), 7.63 (br
s, 1H), 8.3 (br s, 1H), 12.2 (br s, 1H); 13C NMR (DMSO-d6) δ 52.8
(C-7), 115.4 (C-8a), 136.2 (C-5), 149.6 (C-2), 161.1 (C-3a), 180.2 (C-
8). Anal. Calcd for C6H6N4O‚0.33H2O: C, 46.17; H, 4.30; N, 35.89.
Found: C, 46.32; H, 4.08; N, 35.76.
3-(3-(4-Propoxyphenyl)propyl)coformycin Aglycon (8). Prepared
from 4-propoxyphenylpropyl methanesulfonate and ketone 7 as de-
scribed for the preparation of compound 9: mp 134-135 °C; 1H NMR
(DMSO-d6) δ 0.96 (t, 3H, J ) 7 Hz), 1.71 (hex, 2H, J ) 7 Hz), 1.93
(quin, 2H, J ) 7 Hz), 2.49 (t, 2H, J ) 7 Hz), 3.16 (br s, 2H), 3.84 (t,
2H, J ) 7 Hz), 3.88 (t, 2H, J ) 7 Hz), 4.62 (br s, 1H), 4.90 (d, 1H, J
) 5 Hz), 6.83 (d, 2H, J ) 8 Hz), 6.96 (d, 1H, J ) 4 Hz), 7.10 (d, 2H,
J ) 8 Hz), 7.29 (s, 1H), 7.4-7.5 (m, 1H). Anal. Calcd for
C18H24N4O2: C, 65.83; H, 7.37; N, 17.06. Found: C, 65.49; H, 6.99;
N, 16.81.
1
solid: mp 58-61 °C; Rf 0.39 (3:1 CH3CN/0.2 N NH4Cl); H NMR
(DMSO-d6) δ 1.1-1.7 (m, 6H), 2.6-2.9 (m, 3H), 3.15 (br s, 2H), 3.81
(t, 2H, J ) 7 Hz), 4.80 (br s, 1H), 6.96 (d, 1H, J ) 4 Hz), 7.1-7.3 (m,
5H), 7.19 (s, 1H), 7.47 (m, 1H). Anal. Calcd for C19H24N4O3‚0.2CH3-
CO2H‚1H2O: C, 60.26; H, 6.99; N, 14.48. Found: C, 60.21; H, 6.80;
N, 14.33.
Ethyl 1-(2-Bromoethyl)-5,6,7,8-tetrahydronaphthyl-3-carboxylate
(12). A solution of ethyl 2-oxo-5,6,7,8-tetrahydro-2H-1-benzopyran-
3-carboxylate 1136 (20.0 g, 0.09 mol) and distilled 2,3-dihydrofuran
(63.0 g, 0.9 mmol) in a dry 18 × 4 cm resealable glass tube was flushed
with Ar and sealed with a Teflon plug. The reaction was heated at 140
°C for 14 h. The resulting dark reaction mixture was then cooled and
the excess 2,3-dihydrofuran evaporated under reduced pressure to afford
22.9 g of a dark residue presumed to be the tricyclic Diels-Alder
adduct: 1H NMR (CDCl3) δ 1.28 (t, 3H, J ) 7.7 Hz), 1.3-3.0 (series
of m, 10H), 3.6 (m, 2H), 4.2 (m, 3H), 4.7 (d, 1H, J ) 9.4 Hz), 6.9 (s,
1H). The crude product was dissolved in 200 mL of CH2Cl2, cooled to
-78 °C, and treated with BBr3 (67.6 g, 0.27 mol). The reaction mixture
was then warmed slowly to rt, stirred 4 h, cooled again to -78 °C,
and quenched with 60 mL of EtOH. After the mixture stirred 16 h, the
solvent was evaporated, and the residue was diluted with hexane,
washed with water and 10% bicarbonate, and dried (MgSO4). The
solvent was removed and the residue purified by chromatography (2%
EtOAc in hexane) to give 15.1 g (54%) of 12 as an oil: 1H NMR
(DMSO-d6) δ 1.3 (t, 3H, J ) 7 Hz), 1.7 (m, 4H), 2.7 (m, 4H), 3.1 (t,
2H, J ) 7 Hz), 3.68 (t, 2H, J ) 7 Hz), 4.32 (q, 2H, J ) 7 Hz), 7.56
(s, 1H), 7.62 (s, 1H).
3-[2-(3-Carboxy-5,6,7,8-tetrahydronaphthyl)ethyl]coformycin Ag-
lycon (13). 13 was prepared from compound 12 and ketone 7 by using
the procedures described for the preparation of compounds 9 and 10:
mp >220 °C dec; UV (MeOH) 214 nm (ꢀ 29700), 243 (12400), 282
(8190); 1H NMR (DMSO-d6) δ 1.74 (m, 4H), 2.77 (m, 4H), 2.9 (t, 2H,
J ) 7 Hz), 3.1 (br s, 2H), 4.1 (t, 2H, J ) 7 Hz), 4.8 (br s, 1H), 7.0 (d,
J ) 4 Hz, 1H), 7.3 (s, 1H), 7.5 (br s, 3H). Anal. Calcd for C19H22N4O3‚
0.5H2O‚0.3CH3CO2H: C, 61.71; H, 6.39; N, 14.68. Found: C, 61.32;
H, 6.23; N, 14.55.
Ethyl 1-Hydroxy-4-bromo-5,6,7,8-tetrahydronaphthyl-3-carboxy-
late (16). To a solution of 1536 (3.2 g, 14.5 mmol) in acetic acid (30
mL) was added a solution of bromine (0.75 mL, 14.5 mmol) in acetic
acid (30 mL) over 1.5 h. After an additional 1 h, the solvent was
evaporated, and the residue was diluted with ice-water, extracted with
ether, dried (MgSO4), and purified on SiO2 (30% EtOAc/hexane) to
give 3.0 g (69%) of 16: 1H NMR (DMSO-d6) δ 1.30 (t, 3H, J ) 7
Hz), 1.7 (m, 4H), 2.6 (m, 4H), 4.30 (q, 2H, J ) 7 Hz), 6.9 (s, 1H), 9.9
(s, 1H).
Ethyl 1-Vinyl-4-bromo-5,6,7,8-tetrahydronaphthyl-3-carboxylate
(17). A solution of 16 (2.0 g, 6.7 mmol) in pyridine (15 mL) was cooled
to 0 °C and slowly treated with trifluoromethanesulfonic anhydride (1.35
mL, 8.05 mmol). After warming to rt, the mixture was stirred for 3 h
and the solvent evaporated. The residue was diluted with 100 mL of
ice-water and extracted with ether. The combined extracts were dried
(MgSO4) and evaporated to afford 2.4 g of a brown syrup. The crude
triflate was dissolved in DMF (30 mL), treated with (PPh3)2PdCl2 (0.2
g, 0.33 mmol), LiCl (0.85 g, 20.1 mmol), and vinyltributyltin (2.35
mL, 8.05 mmol), and heated at 80 °C. After 4 h, the solvent was
evaporated, and the residue was diluted with ether, washed with water
and saturated NaF, dried (MgSO4), and purified by chromatography
(5% EtOAc in hexane) to give 1.6 g (77%) of 17: 1H NMR (DMSO-
d6) δ 1.33 (t, 3H, J ) 7 Hz), 1.7 (m, 4H), 2.7 (m, 4H), 4.32 (q, 2H, J
) 7 Hz), 5.3-5.8 (m, 2H), 6.8-7.0 (m, 1H), 7.5 (s, 1H).
3-(5-Carboethoxy-6-phenylhexyl)coformycin Aglycon (9). Sodium
hydride (479 mg of a 60% oil dispersion, 12 mmol) was washed with
hexanes, suspended in DMF (55 mL), and treated with ketone 7 (1.71
g, 11.4 mmol). The mixture was heated to 60 °C for 30 min under N2,
cooled to 22 °C, and treated with a solution of ethyl 2-(benzyl)-6-
methanesulfonyloxyhexanoate (3.74 g, 11.4 mmol) in 12 mL of DMF.
NaI (427 mg, 2.85 mmol) was then added and the resulting mixture
stirred at 90 °C for 90 min to produce a homogeneous brown solution.
The solvent was removed and the residue chromatographed on 220 g
of SiO2. Elution with CH2Cl2/methanol mixtures of 20:1, 18:1, and
16:1 provided 1.44 g (33%) of 3-(5-carboethoxy-6-phenylhexyl)-6,7-
dihydroimidazo[4,5-d]-[1,3]diazepin-8(3H)-one as a light yellow
gum: 1H NMR (DMSO-d6) δ 1.01 (t, 3H, J ) 7 Hz), 1.20 (quin, 2H,
J ) 7 Hz), 1.4-1.8 (m, 4H), 2.62 (m, 1H), 2.7-2.8 (m, 2H), 3.73 (d,
2H, J ) 4 Hz), 3.92 (t, 2H, J ) 7 Hz), 3.94 (q, 2H, J ) 7 Hz), 7.1-
7.3 (m, 5H), 7.40 (d, 1H, J ) 4 Hz), 7.61 (s, 1H), 8.35 (br m, 1H).
The ketone (1.43 g, 3.75 mmol) in CH2Cl2/MeOH (1:1, 36 mL) was
treated with NaBH4 (146 mg, 3.75 mmol). After stirring the mixture at
rt for 45 min, SiO2 (4.5 g) was added and the solvent evaporated. The
powder was loaded onto a 45 g SiO2 column and eluted with CH2Cl2/
methanol/triethylamine mixtures of 25:1:0.25 and 20:1:0.2. The fraction
containing the product was reduced to a minimal volume and diluted
with ether to yield after filtration and drying 1.07 g (74%) of compound
1
9 as a white solid: mp 121-124 °C; H NMR (DMSO-d6) δ 1.03 (t,
3H, J ) 7 Hz), 1.22 (q, 2H, J ) 7 Hz), 1.4-1.7 (m, 4H), 2.5-2.7 (m,
1H), 2.7-2.8 (m, 2H), 3.15 (m, 2H), 3.81 (t, 2H, J ) 7 Hz), 3.95 (q,
2H, J ) 7 Hz), 4.81 (m, 1H), 4.91 (d, 1H, J ) 5 Hz), 6.95 (d, 1H, J
) 4 Hz), 7.1-7.3 (m, 5H), 7.25 (s, 1H), 7.45 (m, 1H). Anal. Calcd for
C21H28N4O3: C, 65.60; H, 7.34; N, 14.57. Found: C, 65.39; H, 7.34;
N, 14.52.
Ethyl 1-(2-Hydroxyethyl)-4-bromo-5,6,7,8-tetrahydronaphthyl-
3-carboxylate (18). To a solution of 9-BBN dimer (3.1 g, 12.8 mmol)

318 J. Am. Chem. Soc., Vol. 121, No. 2, 1999
Erion et al.
in THF (80 mL) was added a solution of 17 (5.3 g, 17.1 mmol) in
THF (20 mL). After stirring for 16 h, the solution was cooled to -30
°C and treated with 30% H2O2 (4.3 mL, 40.0 mmol) followed by 3 N
NaOH (13.3 mL, 40.0 mmol). The mixture was warmed to 0-10 °C
and stirred for 3 h, diluted with water, and extracted with ether. The
combined extracts were washed with water and brine, dried (MgSO4),
and evaporated. The residue was purified by chromatography (30%
EtOAc in hexane) to give 3.2 g (57%) of 18 as an oil: 1H NMR
(DMSO-d6) δ 1.31 (t, 3H, J ) 7 Hz), 1.73 (m, 4H), 2.6-2.9 (m, 6H),
3.6 (q, 2H, J ) 5 Hz, collapsible to t with D2O), 4.32 (q, 2H, J ) 7
Hz), 4.7 (t, 1H, J ) 5 Hz, collapsible with D2O), 7.27 (s, 1H).
Ethyl 1-(2-Bromoethyl)-4-bromo-5,6,7,8-tetrahydronaphthyl-3-
carboxylate (19). To a 0 °C solution of 18 (3.2 g, 9.7 mmol) in dry
CH2Cl2 (60 mL) was added triphenylphosphine (3.8 g, 14.6 mmol)
followed by carbon tetrabromide (4.8 g, 14.6 mmol). After the resulting
mixture was stirred for 1 h, the solvent was evaporated and the residue
purified by chromatography (5% EtOAc in hexane) to give 2.9 g (76%)
of 19 as an oil: 1H NMR (DMSO-d6) δ 1.31 (t, 3H, J ) 7 Hz), 1.7 (m,
4H), 2.7 (m, 4H), 3.1 (t, 2H, J ) 7 Hz), 3.69 (t, 2H, J ) 7 Hz), 4.32
(q, 2H, J ) 7 Hz), 7.3 (s, 1H).
Adenosine Monophosphate Deaminase. Porcine heart AMPDA was
purified essentially as described by Smiley et al. through the phospho-
cellulose step.48 Inhibition of AMPDA activity was determined at 37
°C in a 0.1-mL assay mixture containing inhibitor, ∼0.005 units of
porcine heart or recombinant human E-type AMPDA, 0.1% bovine
serum albumin, 10 mM ATP, 250 mM KCl, and 50 mM MOPS at pH
6.5. The concentration of the substrate AMP was varied from 0.125 to
10.0 mM. The reaction was initiated by the addition of AMPDA to the
otherwise complete reaction mixture and terminated after 5 min by
injection into an HPLC system. Activities were determined from the
amount of IMP formed over 5 min. IMP was separated from AMP by
HPLC using a Beckman Ultrasil-SAX anion-exchange column (4.6 mm
× 25 cm) with an isocratic buffer system (12.5 mM potassium
phosphate, 30 mM KCl, pH 3.5) and detected spectrophotometrically
at 254 nm.
Adenosine Deaminase. Inhibition of ADA from calf intestinal
mucosa was determined spectrophotometrically at pH 7.0 and 37 °C.
The reaction mixture (1 mL) contained inhibitor, 0.001 units ADA,
and 40 mM potassium phosphate. The concentration of adenosine was
varied from 20 to 100 µM. The reaction was initiated by addition of
ADA and monitored continuously at 265 nm for 15 min. Decreases in
absorbance reflected conversion of adenosine to inosine.
Hepatocyte Uptake Studies. Isolated rat hepatocytes were prepared
from fed rats by the collagenase digestion technique.49 Cells were
suspended at a density of 75 mg wet weight/ml of Krebs bicarbonate
buffer supplemented with 10 mM glucose. Incubations were performed
in sealed plastic tubes under a 95% oxygen/5% carbon dioxide
atmosphere. The tubes were kept at 37 °C and continuously agitated
in a shaking water bath. Cells were given a 15 min equilibration period
prior to the start of an experiment. Typical energy charges of cell
preparations ranged from 0.8 to 0.85, while ATP concentrations ranged
from 1.8 to 2.4 µmol/g of cells. These values are comparable to those
measured in freeze-clamped rat liver biopsies [0.81 ( 0.03 and 2.21
( 0.4 for energy charge and ATP content, respectively]. Cells were
incubated with AMPDA inhibitors at the indicated concentrations, and
aliquots were removed at the specified time intervals over the course
of 1 h. The removal of extracellular drug, and the extraction of cells
was achieved by centrifugation of the aliquots through an oil layer into
10% perchloric acid. The acid layer was neutralized, clarified, and
analyzed for drug content by HPLC. A C18 reverse phase column and
an acetonitrile gradient were used to identify and quantify peaks relative
to known standards.
Nucleotide Determinations in Rat Hepatocytes Subjected to
Fructose or Anoxia. Isolated rat hepatocytes were incubated in the
presence of vehicle or an AMPDA inhibitor for 5 min. Cellular stress
was induced by the addition of fructose (0.5 mg/mL) to the cell medium
as described by Woods et al.50 or by subjecting the cells to 15 min of
anoxia by replacement of the gaseous phase with a 95% nitrogen/5%
carbon dioxide gas mixture. At appropriate time intervals prior to and
after stress induction, aliquots of the cell suspensions were removed,
and the cells were extracted by acidification with perchloric acid. After
neutralization and clarification of the cell extracts, nucleotide content
was determined by HPLC. Separation of nucleotides was performed
on an anion-exchange column eluted with a potassium phosphate/
potassium chloride gradient. Identification and quantitation of nucle-
otides was achieved by comparison to the elution profile and peak areas
of authentic standards (Sigma, St. Louis, MO) of known concentration.
Statistical analysis was performed with the unpaired Student’s t test.
A value of p < 0.05 was considered statistically significant.
3-[2-(3-Carboxy-4-bromo-5,6,7,8-tetrahydronaphthyl)ethyl]cofor-
mycin Aglycon (14). 14 was prepared from compound 19 and ketone
7 by using the procedures described for compounds 9 and 10. mp >200
1
°C dec; H NMR (DMSO-d6) δ 1.69 (m, 4H), 2.65 (m, 4H), 2.85 (t,
2H, J ) 7 Hz), 3.18 (br s, 2H), 4.0 (t, 2H, J ) 7 Hz), 4.8 (br s, 1H),
4.95 (br s, 1H), 6.86 (s, 1H), 7.0 (d, 1H, J ) 4 Hz), 7.27 (s, 1H), 7.5
(br s, 1H). Anal. Calcd for C19H21N4O3Br‚2.5H2O‚1.2CH3CO2H: C,
46.70; H, 5.59; N, 10.18. Found: C, 46.13; H, 4.94; N, 10.17.
3,6,7,8-Tetrahydroimidazo[4,5-d]-[1,3]diazepine Hydrochloride
(21). A mixture of 4(5)-nitrohistamine hydrochloride salt 2038 (4.36 g,
22.6 mmol), triethyl orthoformate (150 mL), and 10% Pd/C (3.3 g) in
EtOH (200 mL) was stirred under 1 atm H2 at rt. After 2 h, the reaction
mixture was warmed to 50 °C and stirred 4 h and then warmed to 70
°C for 48 h. The solvent and excess triethyl orthoformate were removed
under reduced pressure. The resulting dark hygroscopic solid was
triturated with EtOH to give 1.25 g (32%) of 21 as a brown solid: 1H
NMR (DMSO-d6) δ 3.01 (t, 2H, J ) 4.3 Hz), 3.6 (m, 2H), 7.57 (s,
1H) 7.87 (d, 1H, J ) 6.9 Hz), 10.2 (br s, 1H); MS (ES) m/z 137 (MH+).
3-[2-(3-Carboxy-4-bromo-5,6,7,8-tetrahydronaphthyl)ethyl]-3,6,7,8-
tetrahydroimidazo[4,5-d]-[1,3]diazepine Dicyclohexylammonium Salt
(22). Compound 21 was free-based as described for compound 7 and
alkylated with 2339 as described for the preparation of 9. The resulting
benzyl ester (110 mg, 0.22 mmol) was subjected to hydrogenolysis by
stirring with dicyclohexylamine (77 mg, 0.44 mmol) and 30 mg of
10% Pd/C in MeOH (20 mL) under 1 atm H2. After 16 h, the solvent
was removed under reduced pressure and the residue triturated with
ether to give 80 mg of 2240 as light yellow solid: 1H NMR (D2O) δ
1.0-2.1 (series of m, 24H), 2.3 (t, 2H, J ) 6 Hz), 2.6 (t, 2H, J ) 6
Hz), 2.8 (t, 2H, J ) 4 Hz), 2.9 (t, 2H, J ) 6 Hz), 3.0-3.3 (m, 2H), 3.3
(t, 2H, J ) 4 Hz), 4.1 (t, 2H, J ) 6 Hz), 6.7 (s, 1H), 6.92 (s, 1H), 6.96
(s, 1H); MS (ES) m/z 418 (MH+).
ADA Binding Conformation of 8. The X-ray structure of murine
adenosine deaminase complexed with (6R)-6-hydroxy-1,6-dihydropurine
riboside (3) (pdb file name: 2ADA) provided the initial atomic
coordinates used to generate the computer model29 and conduct the
energy calculations. Compound 8 was docked in the ADA active site
by superpositioning the base with the base of 3 and applying the
previously described atomic constraints. The torsion angle for each
rotatable bond was randomly varied between 0 and 360 degrees using
the Monte Carlo (MC) search method within the MacroModel program.
All conformations generated (500) were subjected to a short (100-step)
energy minimization procedure by using the conjugate gradient
minimizer and the AMBER force field. Conformations were compared
for uniqueness relative to previously generated conformations. The five
lowest energy conformations were selected and further minimized using
the all-atom force field of AMBER.
Adenosine Determinations in Rat Hepatocytes Subjected to
Fructose or Anoxia. The experimental conditions described above were
also used in studies evaluating the effect of drug treatment on total
adenosine. Adenosine levels were quantitated by reverse phase HPLC.
In some instances, generation of adenosine was confirmed by its
conversion to ethenoadenosine with chloroacetaldehyde. By using the
oil-separation technique described above, the amount of adenosine
secreted into the cell medium was determined. The concentration
(48) Smiley, K. L., Jr.; Berry, A. J.; Suelter, C. H. J. Biol. Chem. 1967,
242, 2502-2506.
(49) Berry, M. N.; Friend, D. S. J. Cell. Biol. 1969, 43, 506-520.
(50) Woods, H. F. Biochem. J. 1970, 119, 501-510.
Conformational Analysis of 13. Low energy conformations of 13
were identified following gas-phase energy minimization of conforma-
tions generated by using the MC conformational search method.

AMP Deaminase Inhibitors
J. Am. Chem. Soc., Vol. 121, No. 2, 1999 319
required to elicit a half-maximal increase in total adenosine levels 10
min post-fructose exposure was defined as the ED50.
Supporting Information Available: Experimental proce-
dures and data for ethyl 2-(phenylmethyl)-6-methanesulfonoxy-
hexonate and 4-propoxyphenylpropyl methanesulfonate. Assay
procedures for phosphofructokinase, fructose 1,6-bisphosphatase,
glycogen phosphorylase, adenylate kinase and adenosine kinase
(5 pages, print/PDF). See any current masthead page for ordering
information and Web access instructions.
Acknowledgment. We thank Mr. Gary Probst, Ms. Wei
Xiao, Ms. Maureen Cottrell, Dr. Christen Anderson, Mr. Colin
Ingraham, Mr. Patrick McCurley, Dr. William Sun, and Dr. Ken
Takabayashi for their scientific contributions to this work. We
would also like to thank Ms. Lisa Weston for her assistance in
preparing this manuscript.
JA983153J