Optimizing the binding of fullerene inhibitors of the HIV-1 protease through predicted increases in hydrophobic desolvation

Article abstract of DOI:10.1021/jm970689r

We have developed and applied a computational strategy to increase the affinity of fullerene-based inhibitors of the HIV protease. The result is a ~50-fold increase in affinity from previously tested fullerene compounds. The strategy is based on the design of derivatives which may potentially increase hydrophobic desolvation upon complex formation, followed by the docking of the hypothetical derivatives into the HIV protease active site and assessment of the model complexes so formed. The model complexes are generated by the program DOCK and then analyzed for desolvated hydrophobic surface. The amount of hydrophobic surface desolvated was compared with a previously tested compound, and if this amount was significantly greater, it was selected as a target. Using this approach, two targets were identified and synthesized, using two different synthetic approaches: a diphenyl C₆₀ alcohol (5) based on a cyclopropyl derivative of Bingel (Chem. Ber. 1993, 126, 1957-1959) and a diisopropyl cyclohexyl C⁶⁰ alcohol (4a) as synthesized by Ganapathi et al. (J. Org. Chem. 1995, 60, 2954-2955). Both showed tighter binding than the originally tested compound (diphenethylaminosuccinate methano-C⁶⁰, K(i) = 5 μM) with K(i) values of 103 and 150 nM, respectively. In addition to demonstrating the utility of this approach, it shows that simple modification of fullerenes can result in high-affinity ligands of the HIV protease, for which they are highly complementary in structure and chemical nature.

Full text of DOI:10.1021/jm970689r

2
424
J . Med. Chem. 1998, 41, 2424-2429
Notes
Op tim izin g th e Bin d in g of F u ller en e In h ibitor s of th e HIV-1 P r otea se th r ou gh
|
P r ed icted In cr ea ses in Hyd r op h obic Desolva tion
Simon H. Friedman,*,†,§ Padma S. Ganapathi, Yves Rubin,* and George L. Kenyon*
‡
,‡
,†
Department of Pharmaceutical Chemistry, University of California, San Francisco, California 94143-0446, and Department of
Chemistry and Biochemistry, University of California, Los Angeles, California 90095-1569
Received October 10, 1997
We have developed and applied a computational strategy to increase the affinity of fullerene-
based inhibitors of the HIV protease. The result is a ∼50-fold increase in affinity from
previously tested fullerene compounds. The strategy is based on the design of derivatives which
may potentially increase hydrophobic desolvation upon complex formation, followed by the
docking of the hypothetical derivatives into the HIV protease active site and assessment of
the model complexes so formed. The model complexes are generated by the program DOCK
and then analyzed for desolvated hydrophobic surface. The amount of hydrophobic surface
desolvated was compared with a previously tested compound, and if this amount was
significantly greater, it was selected as a target. Using this approach, two targets were
identified and synthesized, using two different synthetic approaches: a diphenyl C60 alcohol
(
5) based on a cyclopropyl derivative of Bingel (Chem. Ber. 1993, 126, 1957-1959) and a
diisopropyl cyclohexyl C60 alcohol (4a ) as synthesized by Ganapathi et al. (J . Org. Chem. 1995,
0, 2954-2955). Both showed tighter binding than the originally tested compound (diphen-
ethylaminosuccinate methano-C60, K ) 5 µM) with K values of 103 and 150 nM, respectively.
6
i
i
In addition to demonstrating the utility of this approach, it shows that simple modification of
fullerenes can result in high-affinity ligands of the HIV protease, for which they are highly
complementary in structure and chemical nature.
In tr od u ction
firmed with experiment, and compound 1 (Table 1) was
shown to bind with a Ki value of 5 µM. The modeled
complex of this tested compound, generated using the
program DOCK, showed the core fullerene moiety fitting
snugly into the active site, with the water-solubilizing
2
For approximately the last 10 years, the human
immunodeficiency virus protease (HIVP) has been a
target for anti-HIV drug development.1 We were drawn
to the fullerene core as the basis for inhibitors of this
enzyme because the fullerene structure of C60 is highly
complementary, both sterically and chemically, to HIVP,
thus potentially leading to its binding to the HIVP active
“arms” extending out of the active site into the solvent
(Figure 1a,b). This modeled complex was supported in
several ways: (1) The binding kinetics were competitive.
(2) The binding affinity of a closely related derivative,
the diamine parent of compound 1, had a nearly
2
,3
site.
The active site of uncomplexed HIVP is ap-
proximately cylindrical and lined predominantly with
hydrophobic amino acids.4 Our original modeling sug-
identical K value; this insensitivity of the binding
i
affinity to the nature of the solubilizing arms supported
gested that an appropriately derivatized C60 would fit
tightly in the active site.2 Furthermore, the modeled
minimal interaction of them with the active site. (3)
The estimated binding affinity was consistent with the
amount of desolvated hydrophobic surface.
complex indicated that the majority of the surface that
becomes desolvated upon complexation would be hydro-
phobic. An estimate of the binding affinity of a fullerene
based on the calculated hydrophobic desolvation indi-
cated a potential affinity constant approximately in the
low-micromolar to nanomolar range. This was con-
Because of the success of the initial model design and
consistency of the model with experimental data, we
have sought to improve on the binding affinity of these
compounds in a rational manner by the use of modeling
techniques to introduce addends on the C60 surface that
are able to exploit regions of the HIVP which did not
interact with the initial lead compounds. The method
involves identification of regions unexploited in model
complexes of the previously tested compounds. This is
followed by design of hypothetical derivatives which are
potentially able to interact with these regions. The
hypothetical derivatives are then fitted into the active
|
Abbreviations: HIV (human immunodeficiency virus), HIVP (hu-
man immunodeficiency virus protease), NOESY (nuclear Overhauser
and exchange spectroscopy), TROESY (transverse rotating frame
Overhauser enhancement spectroscopy), DIBAL-H (diisobutylalumi-
num hydride), NMP (N-methylpyrolidone), DMSO (dimethyl sulfoxide),
EDTA (ethylenediamine tetraacetate), HMPA (hexamethylphosphorus
triamide), LDA (lithium diisopropylamide), TFA (trifluoroacetic acid).
†
University of California, San Francisco.
University of California, Los Angeles.
Current address: Department of Chemistry, California Institute
‡
§
5
of Technology, 139-74, Pasadena, CA 91125.
site, using the program DOCK. The modeled com-
S0022-2623(97)00689-4 CCC: $15.00 © 1998 American Chemical Society
Published on Web 05/30/1998

Notes
J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 13 2425
goal was to design C60 derivatives that could fill one or
both of these regions with a nonpolar group and, in so
doing, increase the amount of hydrophobic desolvation
relative to compound 1 and therefore increase the
binding affinity.
The structure of hypothetical derivatives were mod-
6
eled using the SYBYL package. The program MIND-
OCK was subsequently used to generate hundreds of
possible orientations of the hypothetical derivative with
the active site. MINDOCK is a version of DOCK which
uses rigid body minimization to relieve unfavorable
7
contacts within the complexes. The complexes were
then scored within MINDOCK using a previously
described approach which sums “good” and “bad” van
5
der Waals interactions and electrostatics.
Following the initial screen of the compounds using
DOCK’s scoring, top scoring complexes from each run
were further analyzed. Specifically, the amount of
nonpolar surface that was desolvated during complex
formation was analyzed. Molecular surfaces for the
inhibitor alone, the uncomplexed HIVP, and the inhibi-
tor/protease complex were generated using the program
8
MS. MS generates surfaces which form at the contact
point between a probe water sphere and the van der
Waals surface of the molecule being analyzed. The files
that are generated contain atom types for each surface
element. Using these files and an ′awk′ script, the total
surface for each atom type N, O, and C was separately
summed. The total of C atom areas was used as an
estimate of nonpolar surface. Using the difference
between the final complex surface and the sum of the
uncomplexed surfaces, it was possible to determine the
amount of surface desolvated and the proportion of it
due to nonpolar desolvation. Typically, the desolvation
due to nonpolar surface was in the 90% range. The
speed with which MINDOCK generated complexes
allowed the procedure to be used iteratively, adding or
removing atoms to a derivative to optimize the increase
in hydrophobic desolvation.
F igu r e 1. (a) DOCK-generated complex of compound 1 with
HIVP in “top” view. The thin line in panel a indicates the plane
of the cross section depicted in panel c. (b) This same complex
viewed from the “front”. (c) Identical front view of the
compound 1 complex, but with only a thin cross section of the
complex showing. The cross section contains the C -axis and
2
both catalytic aspartates. The dots are the points of a solvent-
accessible molecular surface generated from the complex.
plexes are then analyzed for the amount of hydrophobic
surface desolvation brought about through complex
formation. If this amount is a significant increase over
the hydrophobic surface desolvation achieved by the
lead compound 1, it is selected as a synthetic target.
Because the shape of C60 is highly complementary to
the HIVP binding site while being a completely confor-
mationally restricted scaffold, the modeling of hypo-
thetical compounds and their interactions with the
HIVP active site are simplified.
When we analyzed the originally tested compound 1,2
we found that it formed a complex that could desolvate
2
3
33-352 Å in nonpolar surface depending on slight
variations in the geometry of the complex (Table 1). The
diamine parent of compound 1 created similar com-
plexes (i.e., with the solubilizing arms extending into
2
solvent) and desolvated 381 Å of nonpolar surface. The
2
value of 381 Å became a floor upon which hypothetical
Mod elin g Str a tegy
compounds would have to improve substantially in order
to be selected as synthetic candidates. Table 1 shows
some of the 40 compounds analyzed using this method,
together with their desolvated surface area breakdowns.
The amount of increase in hydrophobic desolvation
relative to compound 1 was used to estimate the
potential improvement in binding relative to this com-
pound.
The first step in the design strategy was the identi-
fication of regions of the HIVP that were unable to
interact with the previously tested compound 1. Ex-
amination of modeled complexes of compound 1 with the
HIVP and their associated molecular surfaces indicated
large apolar regions in the middle of the active site with
which compound 1 was unable to interact. Figure 1c
shows a thin cross section through the middle of the
DOCK-generated model complex (i.e., containing the C2-
axis and catalytic aspartates) of compound 1 and the
protease. It shows two symmetrical water-exposed
channels lined predominantly with nonpolar surfaces.
These channels are present because the active site is
not a perfect cylinder but instead is elliptical in cross
section, and the bulkiness of the side chains of 1
prevents them from gaining access to these areas. The
Many of the analyzed compounds are chiral and are
most readily synthesized as racemates. The modeling
required analysis of each enantiomer separately, be-
cause the different enantiomers should interact differ-
ently with the chiral environment of the protease active
site. In Table 1, therefore, the surface desolvation listed
is for the enantiomer that gave the largest amount of
desolvation, which is the actual enantiomer pictured.
Figure 2c shows the cross section of HIVP, now binding

2
426 J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 13
Notes
Ta ble 1. Fullerene Derivatives That Were Analyzed for Their
Ability To Desolvate Nonpolar Surfaces in Model Complexes
Formed with the HIV-1 Proteasea
F igu r e 2. (a) DOCK-generated complex of compound 5 with
HIVP in “top” view. The thin line in panel a indicates the plane
of the cross section depicted in panel c. (b) This same complex
viewed from the “front”. (c) Identical front view of the
compound 5 complex, but with only a thin cross section of the
a
Molecules 2 and 3 were not synthesized but are shown to
2
complex showing. The cross section contains the C -axis and
represent some of the range of derivatives that were analyzed and
rejected as possible synthetic targets. s.e.: standard error. b Several
conformations of this compound were originally used for docking
in selecting it as a synthetic target, all of them chair conformations.
both catalytic aspartates. The dots are the points of a solvent-
accessible molecular surface generated from the complex.
increases in hydrophobic solvation they produced, as
well as for their relative ease of synthesis. Our initial
interest was in bifunctional compounds such as com-
pound 3 which potentially could interact with both
solvent-exposed channels simultaneously. For synthetic
The major experimentally observed conformation is
a boat,
however. The hydrophobic desolvation indicated is the value of
desolvation for this boat structure, a value that is somewhat
smaller than the originally determined value but still a large
enough improvement over compound 1 to warrant testing.
1
0
ease, however, monofunctional compounds were first
examined closely and were found to satisfy the design
criteria, i.e., a large increase in hydrophobic desolvation
relative to compound 1.
compound 5 in a DOCK-generated complex. The previ-
ously water-exposed channel (Figure 1c) is now partially
occluded by the substituent on the fullerene framework,
producing a decrease in nonpolar surface exposure to
water. This is supported by the total amount of non-
polar surface desolvation for this compound indicated
The preparation of compound 4b was attempted first
as described in a preliminary communication because
of its large hydrophobic desolvation determined by
2
2
in Table 1, which is 472 Å , an improvement of ∼90 Å
9
calculation (Table 1). A Diels-Alder approach to
over compound 1. A substantial increase in hydrophobic
prepare the ketone 8b with cis geometry of the isopropyl
substituents was investigated (Scheme 1). Not unex-
pectedly, the Diels-Alder reactivity of the diene 7b was
found to be too low to undergo cycloaddition with C60.
Accordingly, the desired cycloadduct 8b could not be
obtained under various conditions (toluene, 110 °C; 1,2-
dichlorobenzene, up to 180 °C; toluene, sonication, 50
desolvation (445 Å2 total) is also calculated for the
9
cyclohexane-fused C60 derivative 4a . Its diastereomer
4
b with an all-cis configuration of the substituents on
the cyclohexane ring fits even better based on these
criteria (Table 1), but its synthesis has so far proven
challenging (see below). One of a number of possible
bis-adduct derivatives,10 diammonium salt 3, was also
calculated to give a large increase in hydrophobic
1
3
°
C). To circumvent this impass, we utilized a stepwise
cyclization reaction of the lithium dienolate 7a (6, LDA,
THF, -78 °C) to C60 which proceeds via a double-
2
desolvation (458 Å total). This example illustrates the
potential of multiple adducts of fullerenes, although
their synthesis in a straightforward fashion needs to be
1
4
Michael addition mechanism. The diastereomerically
pure ketone obtained in this reaction was subsequently
identified as the trans diastereomer (8a ) by unequivocal
stereochemical assignment of its alcohol reduction
1
1,12
developed, possibly through a combinatorial approach.
Resu lts a n d Discu ssion
Compounds 4a ,b and 5 were selected as targets for
synthesis and testing because of the relatively large
1
product 4a from 2D NOESY and 2D T-ROESY H NMR
1
5
experiments (Supporting Information).
Both sets of

Notes
J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 13 2427
Sch em e 1
4a and 5 could produce a molecule which binds in a
different orientation, but with equal or even increased
hydrophobic desolvation and affinity. We are currently
exploring this possibility.
Fortunately, both of the compounds 4a and 5 were
soluble enough to assay in the presence of 1% (v/v)
organic solvent. N-Methylpyrrolidone (NMP) was used
in the case of 4a and dimethyl sulfoxide (DMSO) in the
case of 5. The results are shown in Table 1. Compound
4
a has a Ki of 150 nM and compound 5 a Ki of 103 nM,
representing a significant increase in affinity of ∼50-
fold relative to the lead compound 1. The assay condi-
tions used were identical to those originally used for
2
compound 1 (50 mM NaOAc, pH 5.5, 1.0 M NaCl, 5%
glycerol, 1% NMP or DMSO, respectively, 2 mM EDTA),
except that the relatively stable mutant enzyme Q7K
1
7
was used. A control experiment showed that the Ki of
compound 1 with the mutant enzyme was identical with
its originally determined Ki. The use of standard high-
salt conditions for the assay undoubtedly contributes
to the affinity of the tested compounds, which is
primarily hydrophobically driven. However, of primary
interest to this work is the relative change in affinity,
in comparison to the compounds that were tested in an
identical manner, i.e., compound 1. As before, the
inhibitors were preincubated with the enzyme for 5 min.
We found evidence for a slow-on kinetics without
preincubation; however, with a 5-min preincubation,
time courses for the evolution of complex were linear
with respect to time, with 0, or near 0, intercepts.
The data were best fit to a competitive model of
inhibition, although the standard errors for alcohols 4a
and 5 shown in Table 1 are high. Typical standard
deviation in repeated determinations of the inhibited
velocities was 30% of the mean with these compounds,
whereas the identical assay with a control inhibitor
Sch em e 2
(acetylpepstatin) showed standard deviations of 8% from
the mean. This may be due to solubility problems and
possible surface absorption effects leading to some
scatter in the assay results.
Con clu sion s
The purpose of the work described in this paper was
to improve the affinity of fullerene inhibitors of the
HIVP and to explore the use of docking as an iterative
design tool. Compounds 4a and 5 satisfied the design
criteria of increased hydrophobic desolvation, relative
to previously tested compounds, and showed signifi-
cantly increased binding affinity. In their current form,
these compounds are probably too nonpolar to be drugs.
However, because large regions of their surfaces are
both unmodified and in solvent contact in modeled
complexes, there is an opportunity to add new elements
to the unmodified regions that can potentially add
binding energy as well as modulate solubility, toxic-
2
D experiments showed key correlations between the
flagpole hydrogen Hf and Ha, Me(a), and Me(c) and
between Hc and Hb, Hd, which taken with the H-H
coupling data demonstrate that alcohol 4a exists largely
in the boat conformation in solution.
Compound 5 was prepared by DIBAL-H reduction of
the ketone 9 whose synthesis has been reported by
Bingel (Scheme 2). The original intent behind the OH
groups on compounds 4a ,b and 5 was to use them as
branch points to introduce additional functionalities,
primarily to aid in solubilization and possibly aid in
binding. Top scoring complexes of compounds 4a and
1
6
3
,18
ity,
and bioavailability. The total surface area of
compounds 4a and 5 (molecular surfaces of 466 and 485
2
5
, however, show the OH group partially buried in the
Å , respectively) is directly comparable to that of typical
active site, so that additional groups could not be linked
through the OH without introducing severe steric
clashes within the active site. The modeled complex of
the alcohol 4b, however, positions the OH group in
contact with solvent and should allow further deriva-
tization without abrogating the specific binding mode
modeled. Further derivatization of the OH groups in
clinically used HIVP inhibitors (e.g., Indinavir, 544. Å2
total ms). Despite having a relatively high molecular
weight, the high density of packing of the atoms in the
fullerene core means that the actual surface presented
to solvent, which is what determines a molecule’s
interactions with solvent and protein, is comparable to
clinically relevant compounds.

2
428 J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 13
Notes
In general the analysis of the interaction of fullerene
1619 (w), 1463 (m), 1360 (m), 1251 (s), 997 (m); HR-MS (CI,
+
NH
3
) calcd for C16
H
32SiO (MH ) 269.2301, found 269.2538.
derivatives with the HIVP active site is simplified, in
part due to the nature of the fullerene sphere and its
complementarity to the HIVP active site. The starting
point for design is an active site in large contact with a
completely conformationally restricted C60 sphere. Ad-
ditional elements that are introduced to the C60 surface
are positioned by the sphere, which can orient itself in
a limited number of ways, thereby simplifying the
prediction of the binding mode. The ability of the
fullerene surface to position multiple elements required
for binding to a target, and to do so in a conformationally
restricted manner, may make them of utility in phar-
maceutical applications as generic, conformationally
restricted pharmacophore scaffolds.19
Keton e 8a , [(tr a n s-3′,6′-Diisop r op yl-4′-oxo)-1,2-cyclo-
h exa n o]bu ck m in ster fu ller en e. LDA was generated by
addition of 0.68 mL (1.0 mmol) of 1.47 M n-BuLi to a solution
2
of 111 mg of i-Pr NH (0.15 mL, 1.1 mmol) in 1 mL of THF
cooled to 0 °C, and the reaction was stirred for 10 min at 0 °C.
A solution of enone 3a (184.8 mg, 1.2 mmol) in 7 mL of THF
was added to the LDA solution at -78 °C, and the reaction
mixture was stirred further for 15-20 min. A total of 1.6 mL
(
0
∼8 equiv) of the resulting solution was added to C60 (16.0 mg,
.0222 mmol) in 20 mL of toluene at -50 °C, and the reaction
mixture was stirred for 10 min and quenched with 3 mL of
.5 M HCl. The toluene layer was separated, the aqueous
0
layer was extracted with toluene, and the combined toluene
layers were washed with water and brine and dried over
anhydrous Na SO . The solvent was evaporated in vacuo, and
2 4
the residue was purified by flash chromatography on silica gel.
Initial elution with toluene/cyclohexane (2:8) gave 2 mg of
Exp er im en ta l Section
unreacted C60. Further elution with toluene afforded ketone
1
5
a (11.4 mg, 59%, 67% based on recovered C60): H NMR (500
Syn th eses. 2,7-Dim eth yl-5-octen -4-on e (6). To a solu-
tion of diisopropylamine (5.05 g, 7 mL, 50 mmol) in 5 mL of
THF cooled to 0 °C was added 25 mL (50 mmol) of a 2.0 M
solution of n-BuLi in hexanes, and the mixture was stirred
for 10 min at 25 °C. The resulting LDA solution was cooled
to -78 °C, 4-methyl-2-pentanone (5.0 g, 50 mmol) in 8 mL of
THF was added dropwise, and the reaction was stirred for 20
min at -78 °C. Then, isobutyraldehyde (3.6 g, 50 mmol) in
MHz, CDCl
3
/CS
.40 (d, J ) 6.4 Hz, 3H, CH
.43 (d, J ) 6.9 Hz, 3H, CH
2
, 1:4) δ (ppm) 1.11 (d, J ) 6.7 Hz, 3H, CH
3
),
),
1
1
3
), 1.41 (d, J ) 6.8 Hz, 3H, CH
3
3
), 3.16 (dd, J ) 20.0, 3.0 Hz, 1H,
), 3.27 (d of sept, J ) 6.7, 1.5 Hz, 1H, H ), 3.35 (d of sept, J
6.4, 6.0 Hz, 1H, H ), 3.40 (dd, J ) 20.0, 14.5 Hz, 1H, H ),
), 3.98 (ddd, J ) 14.5, 3.0, 1.5 Hz,
); C NMR (125.7 MHz, C Cl /CS , 1:3) δ (ppm) 18.6,
H
c
f
)
a
d
3
1
.83 (d, J ) 6.0 Hz, 1H, H
b
1
3
H, H
e
2
D
2
4
2
1
-
0 mL of THF was added dropwise over a period of 5 min at
78 °C. After the addition, the reaction mixture was allowed
20.4, 24.1, 27.1, 27.2, 29.8, 32.8, 37.9, 46.7, 67.8, 68.1, 134.6,
134.7, 135.1, 135.8, 138.7, 138.8, 139.8, 140.4, 141.2 (2C’s),
141.3, 141.46, 141.53, 141.54, 141.7, 141.8, 141.94 (2C’s),
141.99, 142.01, 142.45, 142.46, 142.5, 142.6, 142.8, 142.9,
143.0, 143.9, 144.2, 144.3, 144.4, 144.5, 144.7, 144.84, 144.85,
145.1, 145.20 (2C’s), 145.21 (2C’s), 145.25, 145.46, 145.51,
145.53, 145.9, 146.05, 146.06, 146.10, 146.16, 146.21, 146.30,
146.32, 147.29, 147.33, 147.40, 153.9, 155.3, 156.3, 209.3; FT-
to reach 25 °C. The reaction was quenched with water and
acidified with 3 M HCl and the aqueous layer extracted with
ether (3 × 30 mL). The combined organic layer was washed
3 2
with water, NaHCO , and brine and dried over anhydrous Na -
SO
ketone.
The crude â-hydroxy ketone was heated at reflux overnight
in a mixture of 20 mL of CH Cl and 4 mL of trifluoroacetic
acid. The reaction mixture was diluted with 100 mL of ether,
washed with water, NaHCO , and brine, and dried over
anhydrous Na SO The solvent was evaporated and the
4
. The solvent was evaporated to give the crude â-hydroxy
-
1
IR (KBr) ν (cm ) 2957 (m), 1719 (s, CdO), 1460 (m), 527 (s).
Alcoh ol 4a , [(cis,tr a n s-3′,6′-Diisop r op yl-4′-h yd r oxy)-
1,2-cycloh exa n o]bu ck m in ster fu ller en e. A solution of ke-
tone 8a (55 mg, 0.063 mmol) in 40 mL of toluene was cooled
to 0 °C, 65 µL (0.065 mmol) of a 1 M solution of DIBAL-H was
added, and the reaction stirred for 20 min. TLC of the reaction
mixture indicated the presence of large amounts of starting
material. A total of 140 µL (0.140 mmol) of DIBAL-H was
added in two portions, and the reaction was stirred further
for 30 min. The reaction was quenched with 10 mL of water,
and the toluene layer was separated. The aqueous layer was
extracted with toluene (10 mL), and the combined organic
phases were washed with water and brine and dried over
2
2
3
2
4
.
residue distilled in a Kugelrohr apparatus (50 °C, ∼0.5 Torr)
to obtain pure enone 6 as a colorless oil (6.0 g, 78%): 1H NMR
(
(
360 MHz, CDCl
d, J ) 6.9 Hz, 6H, CH
CHCH ), 2.39 (d, J ) 7.0 Hz, 2H, CH
6.7, 1.4 Hz, 1H, Me CH-CHd), 6.02 (dd, J ) 16.0, 1.4 Hz,
H, HCdCHCdO), 6.76 (dd, J ) 16.0, 6.7 Hz, 1H, HCdCHCd
3
) δ (ppm) 0.91 (d, J ) 6.7 Hz, 6H, CH
), 2.13 (d of sept, J ) 7.0, 6.9 Hz, 1H,
), 2.42 (sept of d, J
3
), 1.05
3
Me
)
1
2
2
2
2
1
3
O); C NMR (50.3 MHz, CDCl ) δ (ppm) 21.0, 22.4, 24.8, 30.8,
4
3
8.8, 127.6, 153.0, 200.4; FT-IR (film) ν (cm-1) 2958 (vs), 2859
2 4
anhydrous Na SO . Removal of the solvent and flash chro-
matography over silica gel with cyclohexane/toluene (1:1) gave
unreacted ketone 8a (8 mg, 15%). Further elution with toluene
gave the alcohol 4a (38 mg, 69%, 81% based on recovered 8a ):
(m), 1685 (s, CdO), 1668 (s, CdO), 1630 (s), 1463 (m), 1358
(m).
4
-(ter t-Bu tyld im eth ylsilyloxy)-2,7-d im eth yl-3,7-octa d i-
1
H NMR (400 MHz, CDCl
3
/CS
), 1.39 (d, J ) 6.5 Hz, 3H, CH
), 1.72 (d, J ) 6.9 Hz, 3H, CH
2
, 1:2) δ (ppm) 1.19 (d, J ) 6.7
), 1.41 (d, J ) 6.3
), 2.04 (br s, 1H,
), 3.13 (ddd, J )
), 3.20 (d of sept, J ) 6.7, 2.4 Hz,
), 3.25 (d of sept, J ) 6.5, 5.7 Hz, 1H, H ), 3.36 (dd, J )
.6, 5.7 Hz, 1H, H ), 3.71 (ddd, J ) 11.8, 5.6, 2.4 Hz, 1H, H ),
.73 (ddd, J ) 10.6, 6.3, 4.6 Hz, 1H, H ); C NMR (125.7 MHz,
/CS , 1:2) δ (ppm) 19.4, 21.5, 24.8, 28.7, 28.8, 31.2, 33.2,
en e (7b). LDA was generated as above at 0 °C from diiso-
propylamine (216 mg, 0.3 mL, 2.13 mmol) and n-BuLi (1.6 M,
Hz, 3H, CH
Hz, 3H, CH
OH), 2.63 (ddd, J ) 14.6, 6.3, 5.6 Hz, 1H, H
1
1
4
5
3
3
3
3
0
.9 mL, 1.44 mmol) in 3 mL of THF. After addition of 1.5 mL
e
of HMPA, the enone 6 (200 mg, 1.30 mmol) in 2 mL of THF
was added dropwise at -78 °C, and the reaction was stirred
for 20 min. tert-Butyldimethylsilyl triflate (0.35 mL, 1.55
mmol) was added, and the reaction mixture was slowly
warmed to 25 °C and stirred further for 2 h. The reaction was
quenched with 10 mL of water and extracted with pentane (3
4.6, 11.8, 10.6 Hz, 1H, H
H, H
d
g
a
b
f
1
3
c
C
6
D
6
2
4
1
1
1
9.1, 58.1, 68.3, 69.4, 70.0, 135.1, 135.3 (2C), 135.4, 138.8,
38.9, 139.7, 140.1, 141.4, 141.5, 141.6, 141.74, 141.77, 141.8,
41.9, 142.1, 142.2, 142.6, 142.7, 143.2, 144.5, 144.59, 144.66,
44.7, 145.0, 145.1, 145.32, 145.37, 145.4, 145.42, 145.48,
×
15 mL). The combined pentane extracts were washed with
water and brine and dried over anhydrous Na SO The
2
4
.
solvent was evaporated and the residue purified by flash
chromatography on silica gel with hexane to give enol ether
1
145.78, 145.91, 146.0, 146.1, 146.2, 146.31, 146.36, 146.4,
47.0, 147.4, 147.5, 155.9, 156.5, 157.6, 157.8; HR-MS (FAB)
7
b as a colorless oil (245 mg, 70%): H NMR (360 MHz, CDCl
3
)
1
δ (ppm) 0.12 (s, 6H, SiCH
3
), 0.95 (d, J ) 6.7 Hz, 6H, CH(CH
), 1.003 (d, J ) 6.7 Hz, 6H, CH(CH
.32 (sept of d, J ) 6.7, 1.8 Hz, 1H, Me
9.7, 6.7 Hz, 1H, Me CH), 4.52 (d, J ) 9.7 Hz, 1H, HCdC-
OTBS)), 5.7-5.8 (AB-m, 2H, HCdCH); C NMR (90.5 MHz,
CDCl ) δ (ppm) -3.7, 18.5, 22.4, 23.1, 24.9, 26.0, 30.8, 121.0,
25.9, 136.2, 146.3; FT-IR (film) ν (cm-1) 2948 (vs), 2865 (vs),
3
)
)
2
2
),
),
+
calcd for C70H20O (M ) 876.1514, found 876.1523.
1
2
.002 (s, 9H, C(CH)
3
3
2
CH), 2.70 (d of sept, J
Alcoh ol 5, [3′-P h en yl-3′-(r-h yd r oxyben zyl)-1,2-cyclo-
p r op a n o]bu ck m in ster fu ller en e. Compound 5 was obtained
by DIBAL-H reduction of ketone 9 which was prepared by the
)
(
2
1
3
1
6
3
method of Bingel. To a solution of ketone 9 (52 mg, 5.7 ×
-5
1
10 mol) in 30 mL of dry toluene was added DIBAL-H (87 µL

Notes
J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 13 2429
HIV-Protease Inhibition. Bull. Chem. Soc. J pn. 1996, 69, 2143-
of 1 M solution in toluene) with stirring. The reaction was
performed in a foil-wrapped vessel, under argon. After 3.5 h,
an additional 40 mL of toluene and 240 µL of DIBAL-H were
added. After 2.5 h an additional 580 µL of 1 M DIBAL-H was
added. After 24 h, TLC showed quantitative conversion to a
2
151. (b) Schuster, D. I.; Wilson, S. R.; Schinazi, R. F. Anti-
Human Immunodeficiency Virus Activity and Cytotoxicity of
Derivatized Buckminsterfullerenes. Bioorg. Med. Chem. Lett.
1
996, 6, 1253-1256. (c) J ensen, A. W.; Wilson, S. R.; Schuster,
D. I. Biological Applications of Fullerenes. Bioorg. Med. Chem.
1996, 4, 767-779. (d) Schinazi, R. F.; Sijbesma, R.; Srdanov,
G.; Hill, C. L.; Wudl, F. Synthesis and Virucidal Activity of a
Water-Soluble, Configurationally Stable, Derivatized C60 Ful-
lerene. Antimicrob. Agents Chemother. 1993, 37, 1707-1710.
4) Wlodawer, A.; Miller, M.; J askolski, M.; Sathyanarayana, B. K.;
Baldwin, E.; Weber, I. T.; Selk, L. M.; Clawson, L.; Schneider,
J .; Kent, S. B. X-Ray Structure of the Empty HIV-1 Protease.
Science 1989, 245, 616-621.
lower R
f
product (R
f
) 0.65, toluene, SiO
2
). The reaction
, filtered,
mixture was extracted with water, dried with MgSO
and evaporated to dryness in vacuo, yielding 45 mg of crude
product (87%). This material was purified by chromatography
on silica gel with toluene/hexanes (3:1), affording 6.5 mg of
analytically pure product: H NMR (300 MHz, CDCl
4
(
1
3
) δ (ppm)
2
.67 (d, 1 H, J ) 5.0 Hz), 6.44 (d, 1 H, J ) 5.0 Hz), 7.5 (m, 10
13
H); C NMR (75.0 MHz, CDCl
3
) d (ppm) 4 resonances < 100
(5) Meng, E. C.; Shoichet, B. K.; Kuntz, I. D. Automated Docking
ppm at 57.49, 72.35, 78.42, 79.10 corresponding to the 4
nonaromatic carbons in the structure, a cluster of 56 resolvable
peaks was observed at ∼122-151 ppm corresponding to most
of the peaks expected from the 70 aromatic carbons in the
molecule; MS (LSIMS) 915.9 (expected 916.1).
En zym e Assa ys. Compounds were assayed at 25 °C using
recombinant HIV protease (mutant Q7K).
performed in 50-µL volume under final conditions of 50 mM
NaOAc, pH 5.5, 1.0 M NaCl, 5% glycerol, 1% organic solvent
With Grid-Based Energy Evaluation. J . Comput. Chem. 1992,
1
3, 505-524.
(
6) Molecular Modeling System SYBYL, Version 6.2; Tripos Associ-
ates, Inc.
(
7) Gschwend, D. A.; Kuntz, I. D. Orientational Sampling and Rigid-
Body Minimization in Molecular Docking Revisited - On-The-
Fly Optimization and Degeneracy Removal. J . Comp. Mol. Des.
1996, 10, 123-132.
1
7
Assays were
(
(
8) Michael Connolly, University of California, San Francisco.
9) Ganapathi, P. S.; Friedman, S. H.; Kenyon, G. L.; Rubin, Y.
Sequential “Double-Michael” Additions of Dienolates with C60
:
(inhibitor dissolved in either NMP or DMSO as noted in text),
Rapid Access to Sterically Congested Buckminsterfullerene
Derivatives with Defined Stereochemistry. J . Org. Chem. 1995,
60, 2954-2955.
2
mM EDTA. The inhibitor was preincubated with the enzyme
for 5 min at which time the complexation was initiated by
addition of substrate. The complexation was quenched at
(10) By addition of two C2v-symmetric groups, such as the carbenoid
leading to 3, at junctions between two six-membered rings of
<
15% product formation by the addition of 10 µL of 10% TFA.
The cleavage products of the substrate peptide H-Lys-Ala-Arg-
Val-Tyr-p-nitro-Phe-Glu-Ala-Nle-NH (Bachem) were assayed
by HPLC using a 10-40% (acetonitrile, 0.1% TFA)/(water,
.1% TFA) gradient over 30 min at 1 mL/min. The product
C
60, up to eight regioisomers can result. By decreasing the
symmetry of the added group, the number of possible isomers
is greatly multiplied because of the inherent stereoisomerism
in the corresponding products.
2
0
(
11) Franz, A.; An, Y.-Z.; Ganapathi, P. S.; Neier, R.; Rubin, Y.
Tandem Cycloadditions of N,O-Ketene N-1,3-Butadienyl-N-
alkyl-O-silylacetals with C60: A Straightforward Stereoselective
Synthesis of Bicyclic Derivatives of 1,2,3,4-Tetrahydrobuckmin-
sterfullerene. In Fullerenes Proceedings Series: Recent Advances
in the Chemistry and Physics of Fullerenes and Related Materi-
als; Kadish, K. M., Ruoff, R. S., Eds.; 1996; Vol. 3, pp 1326-
was quantitated by integration of peak areas followed by
comparison to product standard curves. Determination of
kinetic constants was done with the program KinetAsyst
(IntelliKinetics). No special precautions were taken during
the binding experiments to avoid ambient light and/or elimi-
nate oxygen from the reaction media. However, a control
assay using argon-sparged buffers and inhibitor solutions, as
well as foil-wrapped assay tubes and subdued lighting, showed
full inhibition by compound 5.
1
341.
(12) Czarnik, A. W.; Ellman, J . A. Special Issue on Combinatorial
Chemistry. Acc. Chem. Res. 1996, 29, 112-170.
(13) High-pressure Diels-Alder reactions will be attempted shortly.
(
(
(
14) J ung, M. E. In Comprehensive Organic Synthesis; Trost, B. M.,
Fleming, I., Eds.; Pergamon Press: New York, 1991; Vol. 4, pp
Ack n ow led gm en t . We thank Dr. Dan Gschwend
and Prof. Irwin Kuntz for access to MINDOCK in early
stages of development. This research was supported by
a grant from the National Institute of Health to G.L.K.
1
-67.
15) Hwang, T.-S.; Shaka, A. J . Cross Relaxation without TOCSY:
Transverse Rotating-Frame Overhauser Effect Spectroscopy (T-
ROESY). J . Am. Chem. Soc. 1992, 114, 3157-3159.
16) Bingel, C. Cyclopropylation of Fullerenes. Chem. Ber. 1993, 126,
(Grant GM 39552) and by a Beckman Young Investiga-
1
957-1959.
tor Award and a NSF-NYI Award (CHE-9457693) to
Y.R.
(17) Rose, J . R.; Salto, R.; Craik, C. S. Regulation of Autoproteolysis
of the HIV-1 and HIV-2 Proteases with Engineered Amino Acid
Substitutions. J . Biol. Chem. 1993, 268, 11939-11945.
(18) (a) Yamago, S.; Tokuyama, H.; Nakamura, E.; Kikuchi, K.;
Su p p or tin g In for m a tion Ava ila ble: 2D T-ROESY NMR
spectrum of cis,trans-alcohol 4a , solution-determined structure
of compound 8a , and space-filling representation of the docked
structure of all-cis-alcohol 4b (3 pages). Ordering information
can be found on any current masthead page.
Kananishi, S.; Sueki, K.; Nakahara, H.; Enomoto, S.; Ambe, F.
14
In vivo Biological Behavior of a Water-Miscible Fullerene:
C
Labeling Absorption, Distribution, Excretion and Acute Toxicity.
Chem. Biol. 1995, 2, 385-389. (b) Bullard-Dillard, R.; Creek,
1
4
K. E.; Scrivens, W. A.; Tour, J . M. Tissue Sites of Uptake of C-
Labeled C60. Biiorg. Chem. 1996, 24, 376-385. (c) Moussa, F.;
Trivin, F.; C e´ olin, R.; Hadchouel, M.; Sizaret, P. Y.; Greugny,
V.; Fabre, C.; Rassat, A.; Szwarc, H. Early Effects of C60
Administration in Swiss Mice: A Preliminary Account for in vivo
Refer en ces
(1) (a) Debouck, C. The HIV-1 Protease as a Therapeutic Target
for AIDS. AIDS Res. Human Retrovir. 1992, 8, 153-164. (b)
Hirsch, M. S.; D’Aquila, R. T. Therapy for Human Immunode-
ficiency Virus Infection. N. Engl. J . Med. 1993, 328, 1686-1695.
(2) Friedman, S. H.; DeCamp, D. L.; Sijbesma, R. P.; Srdanov, G.;
Wudl, F.; Kenyon, G. L. Inhibition of the HIV-1 Protease by
Fullerene Derivatives - Model Building Studies and Experimen-
tal Verification. J . Am. Chem. Soc. 1993, 115, 6506-6509.
(3) For related work in this area, see: (a) Nakamura, E.; Tokuyama,
H.; Yamago, S.; Shiraki, T.; Sugiura, Y. Biological Activity of
Water-Soluble Fullerenes. Structural Dependence of DNA Cleav-
age, Cytotoxicity, and Enzyme Inhibitory Activities Including
C
60 Toxicity. Fullerene Sci. Technol. 1996, 4, 21-29.
(19) For recent work on combinatorial libraries on molecular scaf-
folds, see: (a) Still, W. C. Discovery of Sequence-Selective
Peptide Binding by Synthetic Receptors Using Encoded Com-
binatorial Libraries. Acc. Chem. Res. 1996, 29, 155-163. (b)
Carell, T.; Wintner, E. A.; Sutherland, A. J .; Rebek, J .;
Dunayevskiy, Y. M.; Vouros, P. New Promise in Combinatorial
Chemistry: Synthesis, Characterization, and Screening of Small-
Molecule Libraries in Solution. Chem. Biol. 1995, 2, 171-183.
J M970689R