2,6-Bis(3,4,5-trihydroxybenzylydene) derivatives of cyclohexanone: Novel potent HIV-1 integrase inhibitors that prevent HIV-1 multiplication in cell-based assays

Article abstract of DOI:10.1016/j.bmc.2003.10.005

A number of 2,6-bisbenzylidenecyclohexane-1-one derivatives have been synthesized and tested as HIV-1 integrase (IN) inhibitors with the aim of obtaining compounds capable to elicit antiviral activity at non-cytotoxic concentrations in cell-based assays.

Full text of DOI:10.1016/j.bmc.2003.10.005

Bioorganic & Medicinal Chemistry 12 (2004) 199–215
2,6-Bis(3,4,5-trihydroxybenzylydene) derivatives of cyclohexanone:
novel potent HIV-1 integrase inhibitors that prevent HIV-1
multiplication in cell-based assays
Roberta Costi,a Roberto Di Santo,a Marino Artico,a,* Silvio Massa,b Rino Ragno,c
Roberta Loddo,d Massimiliano La Colla,d Enzo Tramontano,d Paolo La Collad,* and
Alessandra Panid
a
`
Istituto Pasteur-Fondazione Cenci Bolognetti, Dipartimento di Studi Farmaceutici, Universita degli Studi di Roma ‘La Sapienza’,
P.le A. Moro 5, I-00185 Rome, Italy
b
`
Dipartimento Farmaco Chimico Tecnologico, Universita degli Studi di Siena, Via A. Moro 5, San Miniato, I-53100 Siena, Italy
Dipartimento di Studi di Chimica e Tecnologia delle Sostanze Biologicamente Attive, Universita degli Studi di Roma ‘La Sapienza’,
c
`
P.le Aldo Moro 5, I-00185 Rome, Italy
`
d
Dipartimento di Biologia Sperimentale, Sezione di Microbiologia, Universita degli Studi di Cagliari, Cittadella Universitaria,
I-09042 Monserrato, Cagliari, Italy
Received 4 April 2003; accepted 7 October 2003
Abstract—A number of 2,6-bisbenzylidenecyclohexane-1-one derivatives have been synthesized and tested as HIV-1 integrase (IN)
inhibitors with the aim of obtaining compounds capable to elicit antiviral activity at non-cytotoxic concentrations in cell-based
assays. 3,5-Bis(3,4,5-trihydroxybenzylidene)-4-oxocyclohexaneacetic acid (20d) resulted one of the most potent and selective deri-
vatives in acutely infected MT-4 cells (EC50 and CC50 values of 2 and 40 mM, respectively). In enzyme assays with recombinant
HIV-1 integrase (rIN), this compound proved able to inhibit both 30-processing and disintegration with IC50 values of 0.2 and 0.5
mM, respectively. In order to develop a model capable to predict the anti HIV-IN activity and useful to design novel derivatives, we
performed a comparative molecular field analysis (CoMFA) like 3-D-QSAR. In our model the ligands were described quantitatively
in the GRID program, and the model was optimized by selecting only the most informative variables in the GOLPE program. We
found the predictive ability of the model to increase significantly when the number of variables was reduced from 20,925 to 1327. A
Q2 of 0.73 was obtained with the final model, confirming the predictive ability of the model. By studying the PLS coefficients in
informative 3-D contour plots, ideas for the synthesis of new compounds could be generated.
# 2003 Elsevier Ltd. All rights reserved.
1. Introduction
ends of host DNA strand breaks. Removal of mispaired
nucleotides and gap repair lead to provirus formation.
Due to its peculiar properties and to the absence of cel-
lular counterparts, IN is an attractive target for selective
chemotherapeutic intervention.
Among the virus-coded enzymes that are essential for
HIV-1 replication, integrase (IN) plays a fundamental
role by inserting the retro-transcribed viral DNA into
the host chromosome. Initially, IN recognizes the LTR
termini of the linear double-stranded viral DNA mol-
ecule, of which it removes two nucleotides from each 30
end (30 processing reaction), thus leaving recessed 30-OH
termini. Then, IN catalyzes joining of the latter to the 50
Studies performed so far have led to the identification of
a great number of HIV-1 IN inhibitors. Most of them
have been described in enzyme assays with purified
recombinant IN (rIN) and 21mer duplex oligonucleo-
tides reproducing the U5 end of HIV-1 LTRs. However,
only a few compounds are able to inhibit rIN at con-
centrations of 1 mM or lower1À5 and also to prevent the
HIV-1 multiplication in cell-based assays at non-cyto-
toxic concentrations. Therefore, the search for novel
Keywords: Polyhydroxylated aromatics; Cyclohexanone derivatives;
Anti-HIV-1-IN agents; QSAR studies.
* Corresponding authors. Tel./fax: +39-06-446-2731; e-mail: marino.
artico@uniroma1.it (M. Artico); Tel.: +39-070-6754147; fax:
+39-070-6754210; e-mail: placolla@unica.it (P. La Colla).
0968-0896/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2003.10.005

200
R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
anti-IN agents active in cell-based assays, other than in
enzyme assays, continues.
hydrogen-bonds to specific chemical functions of the
enzyme’s catalytic core domain.
Natural products and related synthetic analogues6 are
among the compounds investigated as potential IN inhi-
bitors. One of the most thoroughly investigated class7,8 is
represented by hydroxylated aromatics, such as CAPE
(1),9 flavones (2),9 curcumin (3),10À13 tyrphostins (4),14 bis-
catechols (5),15 dicaffeoylquinic acids (6),16,17 l(À)chicoric
acid (7),17,18 and digalloyl-l-tartaric acid (8).19 With the
exception of 5 and 8, the above compounds share a 3,4-
dihydroxycinnamoyl moiety, sometimes incorporated into
a ring structure, which is the likely pharmacophore since
its integrity is crucial for maintaining the anti-IN activity.
Nevertheless, the lack of this moiety in several potent IN
inhibitors, such as styrylquinoline (9),20 aryl dioxobuta-
noic acids (10)21 and the flavone derivative baicalein (11)9
suggests that additional pharmacophore groups (and,
therefore, different modes of interaction with the target
enzyme) may exist among hydroxylated aromatics (Fig. 1).
This prevents definitive conclusions on their mode of
action; in particular, it is still unclear whether the catechol
Recently, we have been engaged in the design, synthesis
and biological evaluation of novel curcumin-related
derivatives such as 12.12 Although they lacked activity
in cell-based assays and showed the characteristic cyto-
toxicity of the catechol system, many derivatives turned
out to be potent inhibitors of the rIN in enzyme assays.
Interestingly, potent anti-rIN activity correlated with
the presence of two styryl moieties bearing an unsub-
stituted 3,4-dihydroxy group (catechol system) linked
to: (i) a ketoalkane; (ii) a cycloalkanone, eventually
containing an heteroatom; (iii) a benzene ring (Fig. 2).
CoMFA and CoMSIA 3-D QSAR analyses and dock-
ing simulations have been recently performed on the
above compounds22 and the importance of hydrogen-
bonding interactions in determining binding at the
active site has been documented.
In this study, we present novel cinnamoyl derivatives
synthesized to establish whether improvement in both
cytotoxicity and anti-HIV-1 activity could be obtained
maintaining the focus on the catechol system (Fig. 3).
hydroxyls act by chelating the divalent cations (Mg++
,
Mn++) required for IN catalysis9 or by donating
Figure 1. HIV-1 integrase inhibitors.

R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
201
Figure 2. Previously synthesized curcumin-like derivatives.
Figure 3. Newly synthesized curcumin-like derivatives.

202
R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
The rationale for the synthesis was based on the obser-
vation that compounds 6–10 are hydroxylated aro-
matics bearing one or two carboxyl groups and/or 3,4,5-
trihydroxycinnamoyl moieties which are capable to
inhibit rIN in enzyme assays as well as to prevent the
HIV-1 multiplication in cell-based assays.
2. Chemistry
Condensation of 3,4-dichlorobenzaldehyde with acetone
or appropriate heterocycloalkanones in alkaline med-
ium afforded the related bis-benzylidene derivatives 13a,
14a,c,e, which were then reduced to the corresponding
alcohols 13b, 14b,d,f by treatment with sodium borohy-
dride (Scheme 1).
In particular, the activity shown by styrylquinolines
(9) led us to hypothesize that compounds containing
3,4,5-trihydroxycinnamoyl moieties could retain the
capability to chelate divalent metal ions, or to donate
H-bonds, while losing the cytotoxicity peculiar of the
catechol system. In addition, the correlation between
the presence of carboxyl groups and the activity
against both rIN in enzyme assays and the HIV-1
multiplication in acutely infected cells led us to intro-
duce carboxyl groups into both 3,4-dihydroxy-
cinnamoyl and 3,4,5-trihydroxycinnamoyl bearing
compounds.
3,4,5-Trihydroxy, 3(4)-OH, 4(3)-NO2 and 3,4-dihydroxy-
benzylidene derivatives 14g–q, 15–17 and 19–20 were
synthesized by reacting the corresponding arylaldehyde
with the appropriate compound containing active
methylene groups, in glacial acetic acid under a stream
of gaseous hydrochloric acid, as depicted in Schemes 2–
4. Barbituric and thiobarbituric derivatives 18a,b were
obtained by condensation of 3,4,5 - trihydroxy-
benzaldehyde with barbituric or thiobarbituric acid,
respectively, in boiling water (Scheme 3).
The new bis-2,6-benzylidene derivatives (13–20) (Fig. 3)
were prepared as outlined in Schemes 1–4 and tested in
enzyme and cell-based assays according to previously
reported procedures.12
Ethyl 4-oxocyclohexaneacetate and the corresponding
acid, used for the synthesis of derivatives 20a–d, were
synthesized according to literature.28,29
Based on previously reported 3-D QSARs22,23 studies,
our aim was to generate a model capable to help in the
design of new active compounds starting from the ana-
lysis of cinnamoyl derivatives reported in a previous12
and in this paper. Traditionally, SYBYL/CoMFA24,25
(comparative molecular field analysis) is the method
used to create this kind of models, but additional 3-D
quantitative structure–activity relationship (QSAR)
methods are available.26,27 In order to generate molec-
ular descriptors and the GOLPE program26 for the
multivariate regression analyses, the GRID program27
was used.
3. Results and discussion
3.1. Anti-HIV-1 cell-based assays and anti-rIN assays
In a previous work, we described the synthesis and bio-
logical activity of a series of geometrically restrained
cinnamoyl compounds including the 2,6-bis-(3,4-dihy-
droxybenzylidene)cyclohexanone (12a).12 Although
endowed with potent inhibitory activity against rIN in
enzyme assays, these compounds proved highly cyto-
toxic and totally ineffective in preventing the HIV-1
multiplication in acutely infected cells.
Scheme 1.
Scheme 2.

R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
203
were obtained with N1-alkyl (14i, 14l) or N1-benzyl
(14o) derivatives of 3,5-bis(3,4-dihydroxybenzyl-
idene)piperidin-4-one. In fact, although fairly cytotoxic
and ineffective in preventing the HIV-1 multiplication in
cell-based assays, they were found to inhibit rIN at sub-
micromolar concentrations in 30-processing and strand
transfer assays. Nevertheless, 14i and 14l, which were
substituted at position 1 of the 4-piperidinone moiety
with methyl and ethyl groups, respectively, were 2.5-
fold less potent than their parent compound 12d
reported previously12 (IC50=0.2 mM), and 14o, which
was substituted with a benzyl group, was 8-fold less
potent.
Differently from catechols (14i, 14l and 14o) which
shared inactive in cell-based assays but active in enzyme
assays, nitrophenols (14g, 14h, 14j, 14k, 14m, 14n, 14p,
14q) were inactive either in cellular or in enzyme assays.
Therefore, replacement of one of catechol hydroxyls
with a nitro group led to compounds deprived of anti-
IN activity.
Scheme 3.
Then, based on the observation that the addition of
carboxyl groups and/or a third hydroxyl group to the
catechol moiety of CAPE-like analogues favours the
anti-IN activity,17,18,20 with the aim of optimizing the
antiviral activity/cytotoxicity profile of our cynnamoyl
compounds, we designed novel analogues (Fig. 3) char-
acterized by one or more of the following features: (i)
the presence of a third hydroxyl group at the 50-position
of both phenyl rings connected to a piperidinone (15a–
d), a cyclohexanone (15e), a pyranone (15f), a thiopyr-
anone (15g) or a cyclopentanone (16) ring; (ii) alkyl-
ation or benzylation of the 4-piperidinone moiety (15b–
d); (iii) the presence of carboxyl or acetic groups (free or
esterified with ethanol) at position 4 of the cyclohex-
anone nucleus (19a–d; 20a–d); (iv) monobenzylidene
substitution on indanone (17a), indandione (17b), bar-
bituric (18a) or thiobarbituric acid (18b).
Introduction of a third hydroxyl and carboxylic group
in the structure of 12a led to a progressive increase of
potency in enzymatic tests (IC50) and gave rise to anti-
HIV-1 activity in cell-based assays (EC50) (compare 12a
with 15e, 19d, 20b and 20d). As expected, unlike their
catechol counterparts 19a,b and 20a,b, the trihydroxy-
derivatives bearing carboxyl groups (19d and 20d) in the
cyclohehanone ring proved the most potent HIV-1
inhibitors in cell-based assays. Noteworthy, esterifica-
tion with ethanol (19c and 20c) diminished their
potency.
Scheme 4.
Therefore, to investigate whether the catechol moiety
could be replaced by 3,4-disubstituted systems compris-
ing or not OH groups, we first prepared and tested for
biological activity a series of 3,4-dichlorophenyl, 4-
hydroxy-3-nitrophenyl and 3-hydroxy-4-nitrophenyl
derivatives together with 9, 10 and 12a used as reference
drugs (Table 1). No matter whether ketones (13a, 14a,
14c, 14e) or alcohols (13b, 14b, 14d, 14f), all 3,4-
dichlorobenzylidene derivatives were ineffective as inhi-
bitors of the HIV-1 multiplication in cell-based assays
as well as of the HIV-1 rIN in enzyme assays. Also
inactive were all hydroxy-nitro derivatives (14g, 14h,
14j, 14k, 14m, 14n, 14p, 14q). Partially positive results
Compounds 15a–g, 16, 17a–b, 18a–b, 19a–d, 20a–d
turned out to be potent inhibitors of the rIN 30-proces-
sing, strand transfer activities, showing IC50 values as
low as 0.2 mM . In addition, they also inhibited the dis-
integration reaction (Table 1), which is the reversal of
the strand transfer reaction.30,31 Since the occurrence of
disintegration requires only the IN core domain, it has
been used to probe binding of the drugs to the enzyme.6
Thus, since the above compounds inhibited the disin-
tegration activity in the low micromolar range, they
very likely bind to the rIN core region.

204
R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
Table 1. Cytotoxicity, antiviral and anti-integrase activities of derivatives 13–20
a
b
d
Compd
CC50
EC50
SIc
IC50
30-Processing
Strand transfer
Disintegration
13a
13b
14a
14b
14c
14d
14e
14f
14g
14h
14i
>200
121
120
102
200
165
31
119
11
1.2
5
25
1.9
5
15
4
4.7
19
3.5
35
46
40
117
50
115
100
180630
16
>200
>121
—
—
>100
>100
>100
>100
>100
>100
nd
nd
nd
nd
nd
nd
nd
nd
nd
nd
nd
nd
nd
nd
nd
nd
>120—
>102
>200
>165
>31
>119
>11
>1.2
>5
>25
>1.9
>5
>15
>4
>4.7
>19
>3.5
22
>46
ꢀ40— .5
303.9 .7
>50—
34
20
1.6
>16
>19
>50—
1.0
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
1.6
—
>100
9Æ2
>100
9Æ2
>100
>100
>100
>100
>100
>100
>100
>100
>100
0.5Æ0.2
>100
0.9Æ0.4
14j
14k
14l
>100
>100
>100
>100
0.5Æ0.2
1.1Æ0.3
14m
14n
14o
14p
14q
15a
15b
15c
15d
15e
15f
15g
16
17a
17b
18a
18b
19a
19b
19c
19d
20a
20b
20c
20d
9
>100
>100
>100
>100
1.7Æ0.3
2.2Æ0.3
nd
nd
nd
>100
>100
>100
>100
0.3Æ0.1
0.4Æ0.1
Æ0.1
0.5Æ0.2
0.6Æ0.2
0.4Æ0.1
1.1Æ0.2
2.3Æ0.6
Æ3.010
0.5Æ0.1
0.9Æ0.3
0.7Æ0.2
0.5Æ0.2
4.0Æ1.05.3
1.6Æ0.2
0.4Æ0.15
1.0Æ0.7
0.3Æ0.1
0.3Æ0.1
4.3Æ1.0nd
1.2Æ0.4
1.9Æ0.5
0.2Æ0.1
—
0.5Æ0.2
1.0Æ0.3
0.9Æ0.3
2.0Æ0.4
1.6Æ0.3
Æ2.0
1.9Æ1.4
0.2Æ0.1
nd
0
0
1.4
Æ0.2
Æ0.4
3.4
5
6.0Æ2.09.0
0.7Æ0.2
Æ0.3
.0
—
—
0.3Æ0.1
0.2Æ0.1
Æ0.5
19
50
70230.5
5
42
67
704
5
nd
3.0
Æ1.0
2.1Æ0.6
nd
0.1Æ0.05
0.3Æ0.2
1.2Æ0.5
Æ0.3
>5
—
—
5.4
0
—
—
0.2Æ0.1
1.2Æ0.3
0.2Æ0.1
Æ0.1
>42
12.5
17.2.5
>5
>41
2.6
200
1.2
—
>7.4
2.8Æ0.4
0.7Æ0.3
Æ0.2
41
40310.3
4.202
nd
1.6Æ0.6
0.5Æ0.2
1.0
0.05
nd
Æ0.1
>100
—
83.3
—
—
2.4
60
10
12a
—
—
7.4
0.9Æ0.3
nd, not determined.
a Compound concentration (mM) required to reduce the exponential growth of MT-4 cells by 50%, as determined by MTT method.
b Compound concentration (mM) required to achieve 50% protection of MT-4 cells from the HIV-1 induced cytopathicity, as determined by MTT
method.
c Selectivity index: ratio CC50/EC50
.
d Compound concentration (mM) required to reduce HIV-1 IN 30-processing, strand transfer or disintegration activity by 50%. Data represent mean
intracellular proteins;32 (ii) the nucleophilic attack of
3.2. Anti-rRT assays
the a,b-double bond by thiols (such as glutathion),
Numerous IN inhibitors have been reported to affect
additional viral enzymes in cell-free assays. Therefore,
in order to exclude the possibility that title compounds
act non-specifically as inhibitors of the various rIN
activities, we tested selected representatives of the new
cynnamoyl derivatives (14g, 14h, 19d and 20d) against
the HIV-1 rRT in enzyme assays (see Experimental).
None of the compounds was active at 100 mM (data not
shown).
which is promoted by electron-withdrawing sub-
stituents on the phenyl ring and by the overall lipo-
philicity. Although the above studies32 have
highlighted the physico-chemical features relevant to
the cytotoxicity of cinnamoyl compounds, the bio-
chemical complexity of the cell machinery is so huge
that approaches to improve the selectivity index of
catechol derivatives (by lowering the cytotoxicity) are
still largely empirical. Nevertheless, in order to make a
more realistic correlation between the structure and the
cytotoxicity of our cynnamoyl derivatives, we tested
representative compounds against several cell lines
derived from haemathological and solid human tumors
and against a reference ‘normal’ foreskin fibroblast cell
line.
3.3. Antiproliferative activity
The cytotoxicity of 3,4-dihydroxycinnamoyl compounds
is known to be mainly related to: (i) oxidation to ortho-
quinone reactive species, which leads to cross-linking of

R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
205
Among the catechol derivatives reported previously,12
12c was the most potent, followed by 12a and 12b. The
above compounds resulted equally potent against hae-
mathological and solid tumor cell lines and against the
reference ‘normal’ foreskin fibroblast cell line. Interest-
ingly, the substitution of either of the OH groups of the
catechol moiety with nitro groups yielded very cytotoxic
derivatives. The 3-hydroxy-4-nitrophenyl (14h) emerged
as the most cytotoxic compound, followed by the 3-
nitrophenyl-4-hydroxy derivative. It is noteworthy that
introduction of a third hydroxyl group at position 5 of
the phenyl ring makes the catechol derivative less cyto-
toxic (compare 12d with 15a). Likewise, the introduc-
tion of carboxyl or carboxyethyl groups at position 4 of
the cyclohexanone also leads to a significant decrease of
cytotoxicity [more pronounced with the carboxyl (19b)
than with the carboxyethyl group (19a)], which is more
evident in the case of solid tumour-derived cell lines and
the reference fibroblast cell line. As far as cytotoxicity is
concerned, the addition to 19b of a third hydroxyl
group at position 5 of the phenyl ring (19d) is without
further effects.
compartment. Etoposide and camptothecin were used
as reference drugs. None of the compounds significantly
increased the amount of protein-linked DNA breaks
(PLDB) in KB cells (data not shown) with respect to
those of etoposide 10and 100 mM (which gave a 12.5-
fold and 21-fold increase in PLDB with respect to
untreated controls, respectively) and camptothecin 1
and 10 mM (which gave a 4-fold and 7.4-fold increase in
PLDB with respect to untreated controls, respectively).
3.5. Molecular modeling and 3-D QSAR studies
To gain insight into the main binding features of title
compounds into the HIV-1 rIN, either derivatives
described by Buolamwini et al.,22 or the newly synthe-
sized IN inhibitors were analyzed by means of 3-D
QSAR studies. Due to the insufficient number of com-
pounds, no effort was done to develop 3-D QSAR
models for anti-HIV activity in cell-based assays. Due
to the homogeneous results obtained in 30-processing
and strand transfer assays, and to the fact that the pre-
viously reported derivatives12 were tested only in 30-
processing assays, only the IC50s obtained in the latter
assays were used for the 3-D QSAR studies.
3.4. Protein-linked DNA breaks
Due to the fact that polyhydroxylated compounds have
been reported to be topoisomerase inhibitors,33 the
most cytotoxic derivatives (14g and 14h) were investi-
gated in an intact cell assay (see Experimental) for anti-
topoisomerase activity (Table 2).
3-D QSAR approaches rely on advanced statistical
methods to correlate a dependent variable (usually the
biological activity in enzyme assays) with 3-D chemical
descriptors, such as molecular interaction potentials,
which are interaction energies of the considered mol-
ecules with proper molecular probes. In the past, most
of such studies have been carried out by means of the
comparative molecular field analysis (CoMFA) devel-
This assay was preferred to an enzyme assay to take
into account the capability of title compounds to pene-
trate the plasma membrane and to reach the nuclear
25
oped by Cramer more than 10years ago. In the past
Table 2. Antiproliferative activity of compounds 12a–d, 14g,h, 15a, 19a,b and 19d
Cell linesb
aIC50 (mM)
12a
12b
12c
12d
14g
14h
15a
19a
19b
19d
Leukemia/lymphoma
Wil2-NS
CCRF-SB
Raji
CCRF-CEM
MOLT-4
MT-4
25.7
6.7
8.6
5.4
9.3
7.4
30.7
12.8
15.3
8.4
16.2
10.0
5.5
2.6
2.5
1.8
1.9
4.0
50.9
2.7
2.5
3.5
1.4
2.3
11.0
0.97
0.8
0.8
0.3
0.3
1.2
38.4
19.0
37.4
54.2
43.6
62.7
13.3
84.2
21.0
30.6
60.2
22.9
29.5
35.0
13.7
15.2
4.3
3.7
5.0
55.3
55.8
25.4
23.2
42.0
62.5
43.7
32.4
38.6
70.0
Carcinoma
SK-MEL-28
MCF7
SKMES-1
HepG2
DU145
HT-29
HeLa
ACHN
5637
8.2
10.3
5.9
9021.3
7.012.4
25.3
20.0
24.047.2
7.9
12.2
22.2
12.3
3.2
4.6
3.7
8.9
3.9
16.4
4.6
1.9
2.4
1.4
>100
>100
>100
>100
>100
63.1
7.5
61.1
75.6
>100
>100
>100
87.3
>100
>100
>100
>100
>100
>100
>100
90.5
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
87.2
>100
>100
6.8
66.5
.9 10 2.4
6.6 55.9
46.2
3.7
1.9
3.3
1.3
35.6
27.5
3.6
4.2
0.9
1.0
7.6
78.4
2.9
6.0
1.7
17.7
32.6
2.9
0.57
Normal cells
CRL7065
85
16.9
5.7
72
8.8
0.9
30.2
>100
>100
>100
a Compound concentration required to reduce cell proliferation by 50%, as determined by the MTT method, under conditions allowing untreated
controls to undergo at least three consecutive rounds of multiplication. Data represent mean values (ÆSD) for three independent determinations.
b Wil2-NS, human splenic B-lymphoblastoid cells; CCRF-SB, human acute B-lymphoblastic leukemia; Raji, human Burkitt lymphoma; CCRF-
CEM and MOLT-4, human acute T-lymphoblastic leukemia; MT-4, human CD4+ T-cells containing an integrated HTLV-1 genome; SK-MEL-
28, human skin melanoma; MCF7, human breast adenocarcinoma; SKMES-1, human lung squamous carcinoma; HepG2, human epathocellular
carcinoma; DU145, human prostate carcinoma; HT-29, human colon adenocarcinoma; HeLa, human cervix carcinoma; ACHN, human renal
adenocarcinoma; 5637, human bladder carcinoma; CRL7065, normal foreskin fibroblast.

206
R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
Table 3. Statistical results of the GRID/GOLPE analyses
few years, the so-called GRID/GOLPE34 approach has
also been used. Although CoMFA and GRID/GOLPE
rely on the same principles and similar statistical algo-
rithms, GRID offers an excellent set of molecular
probes for computation of interaction potentials, and
GOLPE provides, besides a classical partial least-
squares (PLS) method,35 different useful tools for data
pretreatment, selection of variables and interpretation
of results. This often leads to more meaningful results
and limits the risk of deriving over fitted models.36
Training_Set
PCa
Objectsb
N c
R2
Q2
SDEPd
A
B
C
2
3
2
24
41
64e
913
1293
1327
0.91
0.97
0.84
0.70
0.90
0.73
0.59
0.33
0.55
a Number of optimal principal components from cross-validated
analysis using five random groups.
b Number of compounds in the model.
c Number GRID variables selected from the fractional factorial
selection.
d Standard error of prediction.
In a recent study reported by Buolamwini22 on curcumin-
like derivatives12 and other hydroxylated IN inhibitors,8
a CoMFA was performed leading to the identification, at
3-D level, of the main molecular determinants which were
likely responsible for inhibition of the 30-processing
activity of rIN. In that study, removal of some com-
pounds from the training set was determinant to get sta-
tistically improved results for the 3-D QSAR.
e Compound 7b was in common for the A and B sets.
In the present 3-D QSAR studies, we selected the
GRID/GOLPE approach because of the aforemen-
tioned advantages over the classical CoMFA approach.
Indeed, the atom probes available in GRID27 for the
calculation of molecular interaction fields are much
more numerous than those implemented in CoMFA,
and most of them better resemble the functional amino
acids constituting the putative binding sites of enzymes
and receptors.
Another advantage of the GRID/GOLPE method is
that it includes advanced mathematical tools such as
variable pre-treatment, D-Optimal, and FFD variable
selection methods as well as smart region definition
(SRD).37 This allows to obtain more robust and easily
interpretable 3-D-QSAR models.
A GRID/GOPLE analysis was performed on three sets
of derivatives selected as follows: (i) derivatives reported
by Buolamwini paper22 (Set A); (ii) newly synthesized
derivatives described in the present work (Set B); (iii)
both foregoing groups (Set C). The results of GRID/
GOLPE analyses of the three molecular series depicted
as Model A, Model B and Model C, respectively (car-
ried out using the OH2 probe) are summarized in Table
3 and Figure 4.
Models with high statistical coefficients were obtained
for all the three training sets. It is noteworthy that the
set B gave PLS models with better statistical figures (see
Table 3), although the three models are comparable
based on the fact that the A and C sets show a lower
value of principal components. These findings support
the hypothesis that A and B series can be considered as
a whole set. In fact, the use of the B set as test set for the
A set, and vice versa, showed low value of prediction
standard error (SDEP) (Tables 4 and 5).
Figure 4. Experimental versus recalculated plots for the three 3-D
QSAR models. Model A (A), Model B (B) and Model C (C).
To assess the real predictive ability of the models
obtained, the same external test set used by Buo-
lamwini22 was used. Our Model A displayed a SDEP of
0.60 on the external Test_Set, performing slightly better
than the previous CoMFA model22 (SDEP 0.64). The
higher SDEP value of Model B against the external
Test_Set is mainly due to large structural differences
between the two sets. In fact, the joined Set C per-
formed a satisfactory prediction with a SDEP value of
only 0.69.

R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
207
Table 4. Predictivity of Models A, B and C
Model
SDEPa
Set Ab
Set Bc
External Test_Setd
A
B
C
—
0.95
—
0.92
—
—
0.60
1.03
0.69
a Standard deviation of errors of prediction.
b Derivatives reported in literature.22
c Derivatives reported in the present work.
d Test set as reported by Buolamwini.22
The GOLPE program allows also a graphical repre-
sentation of the 3-D QSAR models by means of PLS
coefficients. In order to identify which part of the mol-
ecules correlated with the variation of activity, we ana-
lyzed the PLS coefficient plots for the three models (Fig.
5). As can be seen, there are regions around the ligands
where the different substituents of the derivatives could
generate favourable/unfavourable interactions with the
water probe. However in this plot the signs of the coef-
ficients can induce to errors: in fact, coefficients have
opposite meaning depending on the fact that the com-
pound produces positive or negative field values in this
areas. In Figure 5A–C, cyan polyhedra (negative PLS
coefficients) delimitate regions where a favourable
(negative) interaction leads to increased activity,
whereas an unfavourable (positive) interaction leads to
decreased activity. On the other hand, yellow polyhedra
(positive PLS coefficients) represent regions where
unfavourable (negative) interactions lead to decreased
activity, whereas favourable (positive) interactions lead
to increased activity.
In Figure 5A (contour plot for the Model A), cyan
polyhedra appear in the region surrounding the meta
hydroxyls groups of the highly active compound 7f,
suggesting a possible favourable hydrogen bond inter-
action for this derivative with those groups. However,
the same polyhedra overlap the methoxyl groups of the
less active compounds 8b and 9b, suggesting that either
the loss of hydrogen bonding interactions, or the
enhancement of steric hindrance exerted by the methyl
groups, is detrimental for biological activity. Yellow
polyhedra represent zones where the presence of even
small groups is poorly tolerated. In fact, yellow poly-
hedra are found in the vicinity of the methoxyl group of
8b and 9b (right side of Fig. 5A) and in the region
occupied by the methoxyl group of 11a in the left side of
Figure 5A.
Figure 5. GRID/GOLPE PLS coefficients contour maps for the three
3-D QSAR models. Model A (A), Model B (B) and Model C (C)
(Contour levels 0.0040 yellow, À0.0040 cyan, for colour code, see
text). To aid interpretation, high (green), medium (orange) and low
active derivative structures are displayed. In (A) derivatives 7f, 8b and
9b, in (B) 13a, 14c and 20d and in (C) 14c and 20d.
contrary, the inactive derivative 14p (not shown) places
in the same region a benzyl group, which interacts
negatively, probably because of the enhanced steric
hindrance. Interestingly, the negative contribution of
the chlorine substituents should also be noted. Figure
5B reports compounds 13a and 14c in which chlorine
atoms replace the hydroxyl groups of 20d. In this case,
the presence of either yellow or cyan polyhedra high-
lights the loss of hydrogen bond interactions. In fact,
the favourable interaction mediated by the presence of
OHs groups becomes unfavourable when chlorine atoms
are introduced. In particular, negative cyan spots sur-
round the positive hydrogens of the para hydroxyls of
20d (favourable interactions) and, at the same time, they
surround the partially negative charged and bigger chlor-
ine atoms of 13a and 14c (unfavourable interactions).
In Figure 5B (contour plot for the Model B), yellow
polyhedra have a more enlarged area than in Figure 5A.
In fact, in the Set B the molecular diversity is mainly
focused in the central ring bridging the two hydroxyl-
ated benzenes. A new yellow polyhedra can be observed
near the region occupied by the carboxymethyl group in
position 4 of the cyclohexanone moiety of the IN inhi-
bitor 20d (green structure in Fig. 5B). In fact, in this
region the negative carboxyl group exerts a favourable
interaction with the positive PLS coefficient. On the

208
R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
Table 5. Residual of the predictions by the GRID/GOLPE models
Present work Set B
Exp. pIC50
JMC199812 Set A
pIC50
JMC200222 External Test Set
Compd
Aa
Compd
B
Compd
pIC50
A
B
C
12
6.05
4.00
4.00
4.00
4.00
5.05
4.00
4.00
4.00
4.00
4.00
6.300
4.00
4.00
6.301.52
4.00
4.00
5.77
4.00
4.00
6.52
6.400
6.300
6.15
5.85
5.22
6.15
5.80
6.52
6.701.57
5.52
6.00
6.701.63
5.92
6.700
6.700
5.55
6.15
5.59
0.69
À0.43
À0.74
À1.17
À1.35
À0.48
0.34
À1.33
À0.86
À1.43
À1.91
10ab
10bb
10cb
10db
11ab
11bb
11cb
11db
12ab
12bb
13b
5.89
4.70
4.00
5.64
4.00
6.15
4.00
4.00
6.101.12
4.52
4.00
4.00
6.05
0.09
À1.30
À1.41
À0.12
À0.20
1.71
26c
5.15
4.22
4.00
4.00
4.00
4.00
5.70
5.70
4.26
4.00
4.00
5.05
5.100
4.36
4.82
4.17
4.74
5.100
4.00
4.85
4.00
4.12
4.40
À0.05
À0.28
À0.61
À0.70
À0.67
À1.30
1.13
À0.24
À1.21
À1.12
À0.87
À1.34
À1.33
0.28
À0.09
À0.42
À0.85
À0.45
À1.03
À1.23
0.55
13a
13b
14a
14b
14c
14d
14e
14f
14g
14h
14i
27c
28c
29c
30c
31c
À0.34
32c
À0.19
33c
0.48
0.21
0.40
34c
À0.30
À0.33
À0.12
1.22
À0.94
À1.56
À1.52
À0.35
À0.34
À0.95
À0.53
À1.19
À0.67
À0.46
À1.49
À0.70
À1.45
À1.55
À0.95
À0.87
À1.05
À0.93
0.33
À0.71
À0.79
À1.79
0.01
35c
36c
.85
7ab
37c
14j
14k
14l
À0.75
À1.08
7bb
.9438c
39c
0.21
7cb
6.00
0.19
À0.28
0.23
À0.68
À0.15
À0.67
À0.35
À0.06
À0.92
À0.34
À0.77
À0.95
À0.71
7db
.70 6.700
.79 6.700
.66 6.700
5.52
40c
14m
14n
14o
14p
14q
15a
15b
15c
15d
15e
15f
15g
16
17a
17b
18a
18b
19a
19b
19c
19d
20a
20b
20c
20d
9
À0.59
À1.00
0.71
À0.74
À1.02
À0.23
7eb
41c
À0.25
7fb
42c
À0.17
8ab
À0.52
À1.92
0.18
.6243c
44c
8bb
4.07
4.00
4.43
6.22
4.00
4.52
À0.05
0.26
0.00
9ab
45c
9bb
À0.51
46c
.30
.24
9cb
0.69
47c
À0.23
9db
À1.35
À0.43
48c
À0.65
0.27
À0.53
À1.31
À0.29
À0.01
1.50
3b
À0.10
0.38
0.67
.46
.44
0.44
0.99
0.03
.85
6.700
5.62
1.02
a A=Model A, B=Model B, C=Model C.
b Numeration of compounds as in the original paper.12
c External test set compounds numbered as in the CoMFA work.22
Figure 5C shows the contour plot of the PLS coeffi-
cient for the Model C, which summarize the previous
plots.
cells and of exploring whether additional mechanisms of
action, for instance fusion, are involved in the inhibition
activity of title compounds.
Using molecular dynamics generated conformations, the
GRID interaction fields and the statistical tools of the
program GOLPE, we developed significant and predictive
models, indicated by the high cross-validation coefficients
and the low SDEP values for the external test set.
4. Conclusions
The overall evaluation of the biological activity
revealed that, in order to show anti-HIV-1 activity in
cell-based assays, title compounds require the presence
of a trihydroxyl moiety coupled with a carboxyl group
at position 4 of a cyclohexanone ring. The assemblage
of the above chemical requisites on the curcuminoid
cyclic structure of 12a led to obtain compound 20d, a
potent anti-IN derivatives as active as 9 and 10 in
infected cells.
From the PLS coefficient map the models could be used
as potential generator of design ideas for new anti-IN
compounds.
Moreover, it is interesting to observe that the present
statistical results are highly comparable with those
found in previous 3-D-QSAR study conducted with
CoMFA and CoMSIA on the Set A of compounds.22
As an important complement to previous CoMFA
results, GRID/GOLPE allowed us to elucidate the role
This interesting result encourage further synthetic and
biological studies on cyclohexanone analogues of 20d,
with the aim of increasing anti-HIV-activity in infected

R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
209
of hydrogen bonding in the receptor binding of our
ligands, by using the OH2 probe provided by GRID.
product, which was chromatographed on silica gel col-
umn (chloroform as eluent) to obtain 0.91 g of pure 14d
(91% yield): mp 147–150 ꢁC (from ethanol).
Yield (%), reaction time (h), mp (ꢁC), recrystallization
solvent; chromatographic system are reported for each
compound:
5. Experimental
5.1. Chemistry
Melting points were determined on a Buchi 530melting
point apparatus and are uncorrected. Infrared (IR)
spectra (Nujol mulls) were recorded on a Perkin-Elmer
297 spectrophotometer. 1H NMR spectra were recorded
at 200 MHz on a Bruker AC 200 spectrometer using
tetramethylsilane (Me4Si) as internal reference standard.
13b: 39%, 0.5 h, 71–73 ꢁC, benzene, silica gel/chloro-
form;
14b: 32%, 1.5 h, oil, silica gel/chloroform;
14f: 100%, 0.5 h, 122–124 ꢁC, benzene/cyclohexane.
1
All compounds were routinely checked by TLC and H
NMR. TLC was performed by using aluminum-baked
silica gel plates (Fluka DC-Alufolien Kieselgel 60F 254).
Developed plates were visualized by UV light. Solvents
were reagent grade and, when necessary, were purified
and dried by standard methods. Concentration of solu-
tions after reactions and extractions involved the use of
a rotary evaporator (Buchi) operating at a reduced
pressure (ca. 20Torr). Organic solutions were dried
over anhydrous sodium sulfate. Analytical results
agreed to within Æ0.40% of the theoretical values. All
compounds were analyzed for C, H, N, and, when pre-
sent, S and Cl.
5.2.3. General procedure for the preparation of com-
pounds 14g–q, 15, 16, 19 and 20. Example: 3,5-bis(3,4,5-
trihydroxybenzylidene)piperidin - 4 - one hydrochloride
(15a). A solution of 4-piperidinone hydrate hydrochlo-
ride (0.5 g, 3.2 mmol) in glacial acetic acid (20 mL),
previously saturated with anhydrous hydrogen chloride,
was treated with 3,4,5-trihydroxybenzaldehyde (1.68 g,
9.75 mmol). After stirring at room temperature for 48 h
the precipitate which formed was filtered, washed in
turn with water, ethanol and light petroleum ether to
obtain 0.57 g of 15a (46% yield): mp >300 ꢁC (from
ethanol); IR cmÀ1 3150, 3300 (OH and NH) and 1580
(CO); 1H NMR (DMSO-d6) d 4.37 (s, 4H, CH2), 6.49 (s,
4H, benzene H), 7.57 (s, 2H, ¼CH–), 9.36 (br s, 8H, OH
and NH).
5.2. Syntheses
5.2.1. General procedure for the preparation of com-
pounds 13a, 14a, 14c, 14e. Example: 2,6-bis(3,4-dichloro-
By this procedure were prepared compounds 14g–q, 15,
16, 19 and 20 starting from the appropriate arylalde-
hyde. Yield (%), reaction time (h), mp (ꢁC), recrys-
tallization solvent; chromatographic system if
benzylidene)pyran-4-one (14c).
A solution of 3,4-
dichlorobenzaldehyde (1.75 g, 9.9 mmol) and pyran-4-
one (0.5 g, 4.9 mmol) in ethanol (15 mL) was added
onto a well stirred solution of sodium hydroxide (0.8 g,
20mmol) in water (15 mL). The mixture was stirred at
room temperature for 1 h then treated with water (50
mL). Crystalline precipitate was filtered and recrys-
tallized from N,N-dimethylformamide to obtain 1.74 g
of 14c (51% yield); mp 225–227 ꢁC; IR cmÀ1 1610(CO);
1
necessary, IR (cmÀ1); H NMR data are reported for
each compound:
14g: 13%, 120h, >300 ꢁC, dimethylsulfoxide; IR cmÀ1
3100 (OH) and 1600 (CO); 1H NMR (DMSO-d6) d 4.46
(s, 4H, CH2), 7.27 (s, 2H, benzene C5-H), 7.71 (s, 2H,
benzene C6-H), 7.78 (s, 2H, ¼CH–), 8.04 (s, 2H, ben-
zene C2-H), 10.53 (br s, 4H, OH and NH).
1H NMR (DMSO-d6) d 4.90(s, 4H, CH ), 7.42–7.81 (s,
2
8H, benzene H and ¼CH–).
Yield (%), reaction time (h), mp (ꢁC), recrystallization
solvent are reported for each already known compound:
14h: 21%, 72 h, >300 ꢁC, dimethylsulfoxide/water; IR
cmÀ1 3300(OH) and 1610(CO); 1H NMR (DMSO-d6)
d 4.46 (s, 4H, CH2), 7.11 (d, Jo=8.4 Hz, 2H, benzene
C6-H), 7.32 (s, 2H, benzene C2-H), 7.80(s, 2H, ¼CH–),
7.99 (d, Jo=8.4 Hz, 2H, benzene C5-H), 10.71 (br s, 4H,
OH and NH).
13a:38 100%, 1 h, 175–177 ꢁC, ethanol;
14a:39 30%, 0.5 h, 147–149 ꢁC, tetrahydrofuran/water;
14e:40 21%, 0.5 h, 145–147 ꢁC, ethanol.
14i: 61%, 72 h, >300 ꢁC, dimethylsulfoxide; IR cmÀ1
1
3500 (OH and NH) and 1570 (CO); H NMR (DMSO-
5.2.2. General procedure for the preparation of com-
pounds 13b, 14b, 14d, 14f. Example: 2,6-bis(3,4-dichloro-
benzylidene) - 4 - hydroxypyrane (14d). Sodium
borohydride (0.05 g, 1.4 mmol) was added into a well
stirred solution of 14c (1.0g, 2.9 mmol) in anhydrous
tetrahydrofuran (100 mL) and water (0.1 mL). The
mixture was stirred at room temperature for 0.5 h. The
solvent was removed and the residue was dissolved in
ethyl acetate. The organic solution was washed with
brine and dried. Evaporation of the solvent gave crude a
d6) d 3.00 (s, 3H, CH3), 4.60(s, 4H, CH 2), 6.89–6.96 (m,
6H, benzene H), 7.70(s, 2H, ¼CH–), 9.44 and 9.91 (2 br
s, 4H, OH), 11.18 (br s, 1H, NH).
14j: 76%, 120h, >300 ꢁC, dimethylsulfoxide; IR cmÀ1
3220(OH) and 1610(CO); 1H NMR (DMSO-d6) d 2.97
(s, 3H, CH3), 4.64 (s, 4H, CH2), 7.32 (d, Jo=8.2 Hz,
2H, benzene C5-H), 7.72 (d, Jo=8.2 Hz, 2H, benzene
C6-H), 7.83 (s, 2H, ¼CH–), 8.04 (s, 2H, benzene C2-H),
11.68 (br s, 3H, OH and NH).

210
R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
14k: 82%, 72 h, 248–250 ꢁC, dimethylsulfoxide; IR cmÀ1
3300(OH) and 1610(CO); 1H NMR (DMSO-d6) d 2.94
(s, 3H, CH3), 4.59 (s, 4H, CH2), 7.11 (d, Jo=8.6 Hz,
2H, benzene C6-H), 7.32 (s, 2H, benzene C2-H), 7.81 (s,
2H, ¼CH–), 7.97 (d, Jo=8.6 Hz, 2H, benzene C5-H),
11.43 (br s, 3H, OH and NH).
15e: 30%, 1.5 h, >300 ꢁC, dimethylsulfoxide/water,
silica gel chloroform/methanol/formic acid 10:1:0.1; IR
cmÀ1 3300(OH) and 1580(CO); 1H NMR (DMSO-d6)
d
1.71 (m, 2H, CH2CH2CH2), 2.83 (m, 4H,
CH2CH2CH2), 6.51 (s, 4H, benzene H), 7.35 (s, 2H,
¼CH–), 9.53 (br s, 6H, OH).15f: 55%, 48 h, >300 ꢁC,
dimethylsulfoxide; IR cmÀ1 3300(OH) and 1590(CO);
1H NMR (DMSO-d6) d 4.84 (s, 4H, CH2), 6.38 (s, 4H,
benzene H), 7.38 (s, 2H, ¼CH–), 8.21 (br s, 6H, OH).
15g: 60%, 2 h, >300 ꢁC, dimethylsulfoxide/water; IR
cmÀ1 3300(OH) and 1580(CO); 1H NMR (DMSO-d6)
d 4.00 (m, 4H, CH2), 6.40(s, 4H, benzene H), 7.30(s,
2H, ¼CH–), 9.38 (br s, 6H, OH).
14l: 32%, 72 h, 270–272 ꢁC, dimethylsulfoxide/water; IR
cmÀ1 330(OH and NH) and 1560(CO);
1H NMR
(DMSO-d6) d 1.26 (t, 3H, CH3), 3.35 (m, 2H, CH2CH3),
4.56 (s, 4H, CH2), 6.89–6.99 (m, 6H, benzene H), 7.71
(s, 2H, ¼CH–), 9.42 and 9.83 (2 br s, 4H, OH), 11.36 (br
s, 1H, NH).
14m: 33%, 72 h, >300 ꢁC, dimethylsulfoxide; IR cmÀ1
3220(OH) and 1610(CO); 1H NMR (DMSO-d6) d 1.26
(t, 3H, CH2CH3), 3.37 (q, 2H, CH2CH3), 4.60(s, 4H,
CH2), 7.32 (d, Jo=8.7 Hz, 2H, benzene C5-H), 7.74 (d,
Jo=8.7 Hz, 2H, benzene C6-H), 7.84 (s, 2H, ¼CH–), 8.07
(s, 2H, benzene C2-H), 11.80(br s, 3H, OH and NH).
16: 36%, 2 h, >300 ꢁC, dimethylsulfoxide/water, silica
gel chloroform/methanol/formic acid 10:1:0.1; IR cmÀ1
1
3150, 3200 (OH) and 1580 (CO); H NMR (DMSO-d6)
d 4.75 (s, 4H, CH2), 6.61 (s, 4H, benzene H), 7.12 (s, 2H,
¼CH–), 9.38 (br s, 6H, OH).
14n: 26%, 48 h, 245–247 ꢁC, dimethylsulfoxide/water;
IR cmÀ1 3300(OH) and 1610(CO); 1H NMR (DMSO-
d6) d 1.25 (t, 3H, CH2CH3), 3.32 (q, 2H, CH2CH3), 4.58
(s, 4H, CH2), 7.14 (d, Jo=8.6 Hz, 2H, benzene C6-H),
7.34 (s, 2H, benzene C2-H), 7.83 (s, 2H, ¼CH–), 8.02 (d,
Jo=8.6 Hz, 2H, benzene C5-H), 11.41 and 11.59 (2br s,
3H, OH and NH).
19a: 18%, 3 h, 102–104 ꢁC, methanol, silica gel chloro-
form/methanol 20:1; IR cmÀ1 3250(OH), 1680(CO
ester) and 1580(CO ketone); 1H NMR (DMSO-d6) d
1.08 (t, 3H, CH3), 2.82–3.16 (m, 5H, CH and CH2), 4.03
(q, 2H, CH2CH3), 6.79–6.97 (s, 6H, benzene H), 7.49 (s,
2H, ¼CH–), 9.40(br s, 4H, OH).
19b: 92%, 2 h, >270 ꢁC, methanol/water, silica gel
14o: 63%, 72 h, 180–182 ꢁC, methanol; IR cmÀ1 3200
(OH and NH) and 1570(CO); 1H NMR (DMSO-d6) d
4.51 (s, 6H, CH2), 6.78–7.57 (m, 11H, benzene H), 7.73
(s, 2H, ¼CH–), 9.41 and 9.80(2 br s, 4H, OH), 11.40(br
s, 1H, NH).14p: 60%, 120 h, 227–230 ꢁC, N,N-dime-
thylformamide/water; IR cmÀ1 3220(OH) and 1610
(CO); 1H NMR (DMSO-d6) d 4.46 (s, 2H, CH2Ph), 4.53
(s, 4H, CH2), 7.27–7.54 (m, 7H, benzene C5-H and
benzyl H), 7.65 (d, Jo=7.8 Hz, 2H, benzene C6-H), 7.84
(s, 2H, ¼CH–), 7.97 (s, 2H, benzene C2-H), 11.87 (br s,
2H, OH).
chloroform/methanol/formic acid 10:2:0.1; IR cmÀ1
3260(OH), 1680(CO acid) and 1620(CO ketone);
1H
NMR (DMSO-d6) d 2.69–3.17 (m, 5H, CH2 and CH),
6.79–6.98 (m, 6H, benzene H), 7.49 (s, 2H, ¼CH–), 9.22
(br s, 4H, OH), 12.50(br s, 1H, COOH).
19c: 37%, 2.5 h, >300 ꢁC, methanol, silica gel chloro-
form/methanol/formic acid 10:1:0.1; IR cmÀ1 3250
(OH), 1700(CO ester) and 1580(CO ketone); 1H NMR
(DMSO-d6) d 1.11 (t, 3H, CH3), 2.51 (m, 1H, CH),
2.80–3.16 (m, 4H, CH2), 4.05 (q, 2H, CH2CH3), 6.51 (s,
4H, benzene H), 7.39 (s, 2H, ¼CH–), 9.31 (br s, 6H,
OH).
14q: 65%, 120h, 211–213 ꢁC, N,N-dimethylformamide/
1
water; IR cmÀ1 3300 (OH) and 1610 (CO); H NMR
(DMSO-d6) d 4.15 (br s, 6H, CH2), 7.02 (d, Jo=8.5 Hz,
2H, benzene C6-H), 7.20–7.36 (m, 7H, benzene C2-H
and benzyl H), 7.71 (s, 2H, ¼CH–), 7.92 (d, Jo=8.5 Hz,
2H, benzene C5-H), 11.26 (br s, 2H, OH).
19d: 29%, 12 h, >300 ꢁC, methanol, silica gel chloro-
form/methanol/formic acid 10:2:0.1; IR cmÀ1 3320
1
(OH), 1600 (CO acid) and 1580 (CO ketone); H NMR
(DMSO-d6) d 2.73–3.21 (2 m, 5H, CH and CH2), 6.53
(s, 4H, benzene H), 7.38 (s, 2H, ¼CH–), 9.33 (br s, 7H,
OH).
15b: 30%, 72 h, >300 ꢁC, ethanol; IR cmÀ1 3300 (OH
and NH) and 1580(CO); 1H NMR (DMSO-d6) d 3.00
(s, 3H, CH3), 4.61 (s, 4H, CH2), 6.52 (s, 4H, benzene H),
6.86 (s, 2H, ¼CH–), 9.61 (br s, 7H, OH and NH).
15c: 57%, 48 h, >300 ꢁC, ethanol; IR cmÀ1 3300 (OH
and NH) and 1580(CO); 1H NMR (DMSO-d6) d 1.26
(t, 3H, CH3), 3.33 (m, 2H, CH2CH3), 4.57 (br s, 4H,
CH2), 6.54 (s, 4H, benzene H), 6.86 (s, 2H, ¼CH–), 9.31
(br s, 7H, OH and NH).
20a: 10%, 1.5 h, 143–145 ꢁC, methanol/water, silica gel
chloroform/methanol/formic acid 10:1:0.1; IR cmÀ1
3250(OH), 1630(CO ester) and 1580(CO ketone); 1H
NMR (DMSO-d6) d 1.08 (t, 3H, CH3), 2.60–3.06 (m,
7H, CH, CH2 and OH), 3.98 (q, 2H, CH2CH3),
6.77–6.95 (s, 6H, benzene H), 7.48 (s, 2H, ¼CH–), 9.40
(br s, 4H, OH).
20b: 55%, 1.5 h, 230–232 ꢁC, methanol, silica gel
15d: 68%, 120h, >300 ꢁC, dimethylsulfoxide; IR cmÀ1
chloroform/methanol/formic acid 10:1:0.1; IR cmÀ1
1
3300 (OH and NH) and 1580 (CO); H NMR (DMSO-
d6) d 4.41 (s, 6H, CH2), 7.43–7.52 (s, 9H, benzene H),
7.66 (s, 2H, ¼CH–), 9.61 (br s, 7H, OH and NH).
3250(OH), 1680(CO acid) and 1570(CO ketone);
1H
NMR (DMSO-d6) d 2.07 (m, 1H, CH), 2.36 (d, 2H,
CH2COOH), 2.60–3.06 (m, 4H, CH2), 6.78–6.97 (s, 6H,

R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
211
benzene H), 7.48 (s, 2H, ¼CH–), 9.84 and 12.30(2 br s,
NaCl, 20mM Tris–HCl pH 7.9), Washing Buffer (60
5H, OH and COOH).
mM imidazole, 0.5 M NaCl, 20 mM Tris–HCl pH 7.9)
and Elute Buffer (1 M imidazole, 0.5 M NaCl, 50 mM
Tris–HCl pH 7.9, 5 mM b-mercaptoethanol).
20c: 46%, 2.5 h, 135–137 ꢁC, methanol, silica gel
chloroform/methanol/formic acid 10:1:0.1; IR cmÀ1
3320(OH), 1680(CO ester) and 1580(CO ketone); 1H
NMR (DMSO-d6) d 1.13 (t, 3H, CH3), 2.29–3.00 (2 m,
7H, CH and CH2), 4.00 (q, 2H, CH2CH3), 6.49 (s, 4H,
benzene H), 7.38 (s, 2H, ¼CH–), 9.60(br s, 6H, OH).
The following oligonucleotides representing the term-
inal 21 nucleotides of the HIV-1 U5 LTR were used: B:
50-ACT GCT AGA GAT TTT CCA CAC-30 (minus
strand); C: 50 GTG TGG AAA ATC TCT AGC A-30
(plus strand); D (dumbbell): 50-TGC TAG TTC TAG
CAG GCC CTT GGG CCG GCG CTT GCG CC-30.
For standard 30-processing assays, B was annealed with
C in 0.1 M NaCl by heating at 80 ꢁC and slowly cooling
to room temperature overnight. This double stranded
substrate was labeled by introducing at the 30 end of C
the two missing nucleotides using a-[32P]-dGTP, cold
dTTP and Klenow polymerase. Unincorporated a-[32P]-
dGTP was separated from the duplex substrate by two
consecutive runs through G-25 Sephadex quick spin
columns. Dumbbell oligonucleotide was labelled at the
50 end using g-[32P]ATP and T4 polynucleotide kinase,
purified through a G-25 Sephadex quick spin column
and self annealed.
20d: 8%, 2 h, >300 ꢁC, methanol/water, silica gel
chloroform/methanol/formic acid 10:2:0.1; IR cmÀ1
3300(OH), 1680(CO acid) and 1560(CO ketone);
1H
NMR (DMSO-d6) d 2.06 (m, 1H, CH), 2.32 (d, 2H,
CH2COOH), 2.60and 3.06 (2m, 4H, CH 2), 6.51 (s, 4H,
benzene H), 7.37 (s, 2H, ¼CH–), 8.30and 15.30(2 br s,
7H, OH).Derivatives 17a and 17b were obtained by
reacting 1-indanone or 1,3-indandione, respectively,
with trihydroxybenzaldehyde, in the 1:1 ratio.
17a: 41%, 2.5 h, 290–292 ꢁC, dimethylformamide/water;
IR cmÀ1 3500, 3410 and 3120 (OH), 1600 (CO); 1H
NMR (DMSO-d6) d 4.00 (s, 2H, CH2), 6.78 (s, 2H,
benzene H), 7.30(s, 1H, ¼CH–), 7.43–7.70(m, 3H,
indan C4-H, C5-H and C6-H), 7.76 (d, Jo=7.4 Hz, 1H,
indan C7-H), 9.05 (br s, 3H, OH).
5.3.2. rIN assays. Standard 30-processing and strand
transfer reaction conditions were: 10mM Hepes pH 7.5,
10mM MnCl , 1 mM DTT, 40mM NaCl, 5 nM of
2
17b: 27%, 2 h, >300 ꢁC, dimethylformamide/water; IR
40mer oligo B/C and 100 nM rIN (considered as
monomer). Incubation was carried out at 37 ꢁC for 1 h
in a volume of 15 mL. Reactions were stopped by adding
7.5 mL of sample buffer (96% formamide, 20mM EDTA,
0.08% bromophenol blue and 0.25% xylene cyanol),
samples were layered into a denaturing 15% poly-
acrylamide gel (7 M urea, 0.09 M Tris borate, pH 8.3,
EDTA 2 mM, 15% acrylamide) and run for 1 h at 80
Watt. Reaction products were monitored and quantified
with a Bio-Rad Personal FX Phosphoimager. For disin-
tegration activity, the standard reaction conditions were
the same as described for the 30-processing assay with the
exception of the presence of 3 nM oligo D (dumbbell sub-
strate). Reactions were incubated at 37ꢁC for 1 h in a
volume of 15 mL, reactions were stopped with sample buf-
fer and analyzed by denaturing PAGE as described above.
cmÀ1 3330and 3100(OH), 1640(CO);
1H NMR
(DMSO-d6) d 7.57 (s, 1H, ¼CH–), 7.72 (s, 2H, benzene
H), 7.92 (m, 4H, indan H), 9.50and 9.73 (2br s, 3H, OH).
5.2.4. General procedure for the preparation of com-
pounds 18. Example: 5-(3,4,5-trihydroxybenzylidene)-
barbituric acid (18a).
A
mixture of 3,4,5 -
trihydroxybenzaldehyde (0.67 g, 3.9 mmol) and barbit-
uric acid (0.50 g, 3.9 mmol) in water (10 mL) was
refluxed for 0.5 h. The precipitate which formed was
filtered and washed with water, ethanol, and light pet-
roleum ether, in turn, to obtain 0.91 g of pure 18a (88%
yield): mp >300 ꢁC (from dimethylsulfoxide/water); IR
cmÀ1 3500, 3480, 3310 and 3210, (OH and NH), 1730,
1680and 1650(CO); 1H NMR (DMSO-d6) d 7.54 (s,
2H, benzene H), 8.02 (s, 1H, ¼CH–), 9.50(br s, 3H,
OH), 11.10(br s, 2H, NH).
5.3.3. Biological assays. Compounds were solubilized in
DMSO at 200 mM and then further diluted. MT-4 cells
were grown at 37 ꢁC in a 5% CO2 atmosphere in RPMI
1640medium supplemented with 10% fetal calf serum
(FCS), 100 IU/mL penicillin G, and 100 mg/mL strep-
tomycin. Cell cultures were checked periodically for the
absence of mycoplasma contamination with a Myco-
Tect Kit (Gibco). HIV-1 stock solutions had titers of
4.5ꢂ106 50% cell culture infectious dose (CCID50)/mL.
Cytotoxicity evaluation was based on the viability of
mock-infected cells, as monitored by the MTT method.
Activity of the compounds against the HIV-1 multi-
plication in acutely infected cells was based on inhibi-
tion of virus-induced cytopathic effect in MT-4 cells and
was determined by the 3-(4,5-dimethylthiazol-1-yl)-2,5-
diphenyltetrazolium bromide (MTT) method.41
Yield (%), reaction time (h), mp (ꢁC), recrystallization
1
solvent; IR (cmÀ1); H NMR data are reported for:
18b: 77%, 0.5 h, >300 ꢁC, dimethylsulfoxide/water; IR
cmÀ1 3300(OH and NH), 1640(CO);
1H NMR
(DMSO-d6) d 7.60(s, 2H, benzene H), 8.30 (s, 1H,
¼CH–), 9.62 (br s, 3H, OH), 12.22 (br s, 2H, NH).
5.3. Microbiology
5.3.1. Integrase and DNA substrates. Expression of the
rIN protein with an amino-terminal polyhistidine tag
was obtained by IPTG induction of the Escherichia coli
strain BL21(DE3) containing the pINSD.His vector.
Protein purification was carried out following the
Novagen procedure, except for the presence of 5 mM
CHAPS in Binding Buffer (5 mM imidazole, 0.5 M
5.3.4. rRT assays. Purified rRT was assayed for its
RNA-dependent DNA polymerase associated activity in

212
R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
a volume of 50 mL containing 50mM Tris–HCl pH 7.8,
MacroModel version 7.1.42 Each molecule has been
energy minimized to a low gradient. The nonbonded
cutoff distances were set to 20A for both van der Waals
and electrostatic interactions. An initial random velo-
city to all atoms corresponding to 300 K was applied.
Three subsequent molecular dynamics run were then
performed. The first was carried out for 10ps with a 1.5
fs time-step at a constant temperature of 300 K for
equilibration purposes. The next molecular dynamics
was carried out for 20ps, during which the system is
coupled to a 150K thermal bath with a time constant of
5 ps. The time constant represents approximately the
half-life for equilibration with the bath; consequently
the second molecular dynamics command caused the
molecule to slowly cool to approximately 150K. The
third and last dynamics cooled the molecule to 50K
over 20ps. A final energy minimization was then carried
out for 250iterations using conjugate gradient. The
minimizations and the molecular dynamics were in all
cases performed for all the compounds in aqueous
solution. The obtained conformer from the final mini-
mization of each compound was used for the molecular
alignment and used to calculate the 3-D-QSAR model.
For the molecular alignment the atom by atom proce-
dure was used using as fitting points a common sub-
structure (Fig. 6).
80mM KCl, 6 mM MgCl , 1 mM DTT, 0.1 mg/mL
2
BSA, 0.5 OD260 unit/mL template:primer [poly(rC)-oli-
go(dG)12-18] and 10 mM [3H]dGTP (1 Ci/mmole). After
a 30min incubation at 37 ꢁC samples were spotted on
glass fiber filter (whatman GF/A) and the acid-insoluble
radioactivity was determined.
5.3.5. Antitumor activities. Compounds were dissolved
in DMSO at 100 mM and diluted in culture medium.
Leukemia cells were grown in RPMI-1640medium
supplemented with 10% FCS, 100 units/mL penicillin G
and 100 mg/mL streptomycin. Solid tumor-derived cells
were grown in their specific media supplemented with
10% FCS and antibiotics. Cell cultures were incubated at
37 ꢁC in a humidified, 5% CO2 atmosphere. The absence
of mycoplasma contamination was checked periodically
by the Hoechst staining method. Exponentially growing
human leukemia and solid tumor-derived cells were
resuspended at a density of 1ꢂ105 cells/mL in growth
medium containing serial dilutions of the drugs. Cell
viability was determined after 96 h at 37 ꢁC by the 3-(4,5-
dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide
(MTT) method. Cell growth at each drug concentration
was expressed as percentage of untreated controls. The
concentration resulting in 50% growth inhibition (IC50)
was determined by linear regression analysis.
5.4.2. GRID Calculations. The interaction energies were
calculated by using GRID (version 20)43 with a grid
5.3.6. Quantitation of protein-linked DNA breaks. A
modified in vivo K-SDS co-precipitation assay33 was
used to quantify protein-linked DNA breaks (PLDB).
Briefly, KB cells were seeded at a density of 2.5ꢂ105/mL
and labeled with [14C]thymidine (0.75 mCi/mL) for 48 h.
Monolayers were washed twice and, after trypsiniza-
tion, cells were divided into aliquots of 3ꢂ105 cells/mL
and incubated for 1 h at 37 ꢁC. Duplicate samples were
treated with test drugs (at concentrations 10-fold higher
than respective CC50 values) and further incubated for 1
h. Then, cells were collected by centrifugation at 2000g
for 10min and resuspended in 1 mL of warm lysis buf-
fer (1.5% SDS, 5 mM EDTA, 0.4 mg/mL salmon sperm
DNA), and the viscous cell lysates were sheared by
passage through a 22-gauge needle five times. After a 10
min incubation at 65 ꢁC, KCl was added to a final con-
centration of 100 mM, samples were chilled in ice bath
for 10min and centrifuged at 3500 g for 10min at 4 ꢁC.
Pellets were resuspended in 1 mL of warm washing
buffer (10mM Tris–HCl, pH 8.0, 100mM KCl, 1 mM
EDTA, 0.1 mg/mL salmon sperm DNA) and incubated
for 10min at 65 ꢁC, chilled in ice bath and centrifuged
as above. After a second washing step, pellets were
solubilized at 65 ꢁC in 40 0mL of water and the radio-
activity was determined in a scintillation counter.
spacing of 1 A and the grid dimensions (A): Xmin/Xmax
,
À24.0/6.0; Ymin/Ymax, À2.0/24.0; and Zmin/Zmax, 20.0/
44.0.
5.4.3. GOLPE analyses. PLS models were calculated
with GOLPE 4.5.12.44
5.4.4. Probe selection. In this study we started with 10
probes (BOTH, Csp3, DRY, OH2, Mg2+, NHamide, Nsp2
,
N+sp3, Osp2, OCO ) according to the nature of the rIN
2
active site. For 3-D QSAR, water probes were selected
on the basis of the PCA and PLS tools available in
GOLPE. Plotting the PCA and PLS scores, we evinced
that the water, NHamide, Nsp2, N+sp3, Osp2 and OCO
,
2
probes were those better describing the molecular
diversity of the Set A (Figs. 7A and B). Selection of the
water probe (OH2) was mainly based on the experi-
mental/recalculated plot (Fig. 7C) of a preliminary PLS
analysis where the OH2 probe visually performed a better
correlation other than a good discrimination between
active and inactive/less active derivatives (Fig. 7D).
5.4.5. Variable preselection. The resulting probe-target
interaction energies for each compound were unfolded
to produce one-dimensional vector variables for each
compound, which were assembled in the so-called X
matrix. This matrix was pretreated by first using a cutoff
5.4. Molecular modeling and 3-D QSAR
5.4.1. Molecular modeling and molecular alignment. The
compounds were built in the most extended conforma-
tion using standard bond distances and angles as defined
in the Macromodel41 graphical interface Maestro. Selec-
tion of the compound conformations was performed by
using the AMBER force field and using molecular
dynamics with simulated annealing as implemented in
Figure 6. Common scaffold used for molecular alignment.

R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
213
Figure 7. PCA and PLS score plots on the Set A using 10probes (A and B). Experimental versus recalculated pIC 50 from the preliminary PLS (C).
PLS score plot on only the OH2 probe (D), red means less active and green more active. GRID probes are color coded as following: BOTH white,
Csp3 red, DRY green, OH2 blu, Mg2+ cyan, NHamide yellow, Nsp2 magenta, N+sp3 orange, Osp2 ivory, OCO violet. For the plot the first two principal
components were use.
2
of 5 kcal/mol to produce a more symmetrical distribu-
tion of energy values and then zeroing small variable
values and removing variables with small standard
deviation, using appropriate cutoffs. In addition, vari-
ables taking only two and three distribution were also
removed. From an initial number of 20,925 the pre-
treatment selected 3051 active variable.
until the R2 and Q2 value did not increase significantly.
In the FFD selection the cross validation was conducted
using five random groups for 20times and a maximum
of tree principal components.
5.4.8. Cross-validation. The models were validated using
random groups. Molecules were assigned in a random
way to five groups of equal size. Reduced models were
built keeping out one group at a time. The formation of
the groups was repeated 20times and using a maximum
model dimensionality of three components.
5.4.6. Smart region definition (SRD). A number of seeds
(1000) were selected using a D-optimal design criterion
in the weight space. Structural differences between dif-
ferent molecules in the series will be reflected in groups
of variables, and therefore groups were generated
around each seed in the 3-D-space. Variables with a
distance of no more than 2 A to the seeds were included
in the groups. If two neighbouring groups (with a dis-
tance smaller than 10A ) contained the same informa-
tion the groups were collapsed. The groups were used in
the variable selection procedure replacing the original
variables. The effect of the groups on the predictivity
was evaluated and groups instead of individual vari-
ables were removed from the data file.
6. Analytical results
Compd
Elemental analysis calculated/found
C
H
N
S
Cl
13b
14b
14c
14d
14f
54.58
54.63
58.00
58.103.93
55.11
55.21
54.84
54.71
52.803.27
52.71
3.23
3.41
3.89
—
—
—
—
—
—
37.91
37.73
34.24
5.4.7. Region selection. The effect of the grouped vari-
ables on the predictivity was evaluated using a frac-
tional factorial design (FFD) procedure. A number of
reduced models (twice the number of variables) were
built removing some of the variables according to the
FFD design. The effect of dummy variables (20%) on
the predictivity was calculated and only if a variable had
a positive effect on the predictivity larger than the effect
of the average dummy variable was the variable inclu-
ded in the final model. The FFD selection was repeated
—
—
34.14
2.92
2.91
3.39
3.45
—
—
—
—
—
—
—
—
34.24
34.28
34.08
34.21
—
7.42
32.81
32.63
3.35
—
7.54
(continued on next page)

214
R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
Compd
Elemental analysis calculated/found
Compd
Elemental analysis calculated/found
C
H
N
S
Cl
8.17
C
H
N
S
Cl
14g
14h
14i
52.61
52.51
52.61
52.57
61.62
61.75
53.64
53.71
53.64
53.71
62.45
62.21
54.61
54.75
54.61
54.58
67.02
67.18
59.604.23
59.71
50.60
50.73
55.96
55.98
55.95
56.78
57.87
57.93
62.72
62.85
64.86
65.00
61.29
61.35
58.76
58.81
64.04
64.14
71.64
71.77
68.09
68.21
50.10
50.21
47.14
47.16
67.31
67.28
65.97
65.91
62.44
62.31
60.87
60.91
67.91
67.98
66.66
3.72
3.65
3.72
3.81
5.17
5.28
4.05
4.16
4.05
4.15
5.49
5.63
4.36
4.21
4.36
4.31
5.19
5.25
9.69
—
—
—
—
—
—
—
—
—
—
—
—
7.68
—
7.68
—
—
66.71
63.15
63.21
60.58
60.49
5.38
—
—
—
9.78
9.69
9.71
3.59
3.31
9.38
9.28
9.38
9.21
3.47
3.44
9.10—
9.31
9.10—
9.28
3.01
3.10—
—
8.09
8.02
8.10
3.43
3.21
3.32
3.37
3.21
3.11
2.81
2.56
—
8.28
8.17
8.23
9.09
9.10
7.92
7.93
7.92
7.83
8.78
8.56
20c
20d
5.30—
5.43
4.84
—
—
—
—
—
—
—
—
4.91
—
—
—
14j
Acknowledgements
14k
14l
Authors thank the Italian Ministero della Sanita, Isti-
tuto Superiore di Sanita- XI Progetto AIDS 1999
(grants no. 40C.8 and no. 40C.47) for financial support.
Acknowledgements are also due to Italian MURST
(40%).
14m
14n
14o
14p
14q
15a
15b
15c
15d
15e
15f
15g
16
7.53
7.51
7.61
References and notes
7.35
2 8.0
6.77
—
—
—
—
—
—
—
—
—
—
—
1. Hong, H.; Neamati, N.; Wang, S.; Nicklaus, M. C.;
Mazumder, A.; Zhao, H.; Burke, T. R., Jr.; Pommier, Y.;
Milne, G. W. A. J. Med. Chem. 1997, 40, 930.
2. Neamati, N.; Hong, H.; Mazumder, A.; Wang, S.;
Sunder, S.; Nicklaus, M. C.; Milne, G. W. A.; Proksa, B.;
Pommier, Y. J. Med. Chem. 1997, 40, 942.
3. Nicklaus, M. C.; Neamati, N.; Hong, H.; Mazumder, A.;
Sunder, S.; Chen, J.; Milne, G. W. A.; Pommier, Y. J.
Med. Chem. 1997, 40, 920.
4. Zhao, H.; Neamati, N.; Hong, H.; Mazumder, A.; Wang,
S.; Sunder, S.; Milne, G. W. A.; Pommier, Y.; Burke,
T. R., Jr. J. Med. Chem. 1997, 40, 242.
5. Zhao, H.; Neamati, N.; Mazumder, A.; Sunder, S.; Pom-
mier, Y.; Burke, T. R., Jr. J. Med. Chem. 1997, 40, 1186.
6. Pommier, Y.; Pilon, A. A.; Bajaj, K.; Mazumder, A.;
Neamati, N. Antiviral Chem. Chemother. 1997, 8, 463.
7. Fesen, M. K.; Kohn, K. W.; Leteurtre, F.; Pommier, Y.
Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 2399.
8. Burke, T. R., Jr.; Fesen, M. R.; Mazumder, A.; Wang, Y.;
Carothers, A. M.; Grunberger, D.; Driscoll, J.; Khon, K.;
Pommier, Y. J. Med. Chem. 1995, 38, 4171.
9. Fesen, M. K.; Pommier, Y.; Leteurtre, F.; Hiroguchi, S.;
Young, J.; Kohn, K. W. Biochem. Pharmacol. 1994, 48,
595.
10. Mazumder, A.; Raghavan, K.; Weistein, J.; Kohn, K. W.;
Pommier, Y. Biochem. Pharmacol. 1995, 49, 1165.
11. Mazumder, A.; Neamati, N.; Sunder, S.; Schultz, J.;
Pertze, H.; Eich, E.; Pommier, Y. J. Med. Chem. 1997, 40,
3057.
12. Artico, M.; Di Santo, R.; Costi, R.; Novellino, E.; Greco,
G.; Massa, S.; Tramontano, E.; Marongiu, M. E.; De
Montis, A.; La Colla, P. J. Med. Chem. 1998, 41, 3948.
13. Li, C. J.; Zhang, L. J.; Dezube, B. J.; Crumpacker, C. S.;
Pardee, A. B. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 1839.
14. Mazumder, A.; Gazit, A.; Levitzki, A.; Nicklaus, M.;
Yung, J.; Kohlhagen, G.; Pommier, Y. Biochemistry 1995,
34, 15111.
15. LaFemina, R. L.; Graham, P. L.; LeGrow, K.; Hastings,
J. C.; Wolfe, A.; Young, S. D.; Emini, E. A.; Hazuda,
D. J. Antimicrob. Agents Chemother. 1995, 39, 320.
16. McDougall, B. R.; King, P. J.; Wu, B. W.; Hostomsky,
Z.; Reinacke, M. G.; Robinson, W. E., Jr. Antimicrob.
Agents. Chemother. 1998, 42, 140.
4.14
4.23
4.18
4.45
4.31
4.78
4.83
5.09
5.21
4.86
4.97
4.90—
4.71
4.33
4.21
4.15
4.21
4.53
4.35
4.51
4.50—
3.57
3.41
3.05
3.15
2.88
2.93
5.40—
5.53
4.74
4.83
5.01
5.07
4.38
4.40
5.70—
5.64
5.09
6.81
6.77
6.71
8.69
8.75
8.40
8.35
8.13
8.01
7.12
7.21
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
8.25
8.14
—
—
—
17a
17b
18a
18b
19a
19b
19c
19d
20a
20b
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
10.60
10.43
10.00
1.05
—
11.44
11.47
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—

R. Costi et al. / Bioorg. Med. Chem. 12 (2004) 199–215
215
17. Robinson, W. E., Jr.; Reinecke, M. G.; Abdel-Malek, S.;
Jia, Q.; Chow, S. A. Proc. Natl. Acad. Sci. U.S.A. 1996,
93, 6326.
30. Chow, S. A.; Vincent, K. A.; Ellison, V.; Brown, P. O.
Science 1992, 255, 723.
31. Chow, A.; Brown, P. O. J. Virol. 1994, 68, 3896.
32. Stanwell, C.; Ye, B.; Yuspa, S. H.; Burke, T. R., Jr.
Biochem. Pharmacol. 1996, 52, 475.
33. Lee, K. H.; Imakura, Y.; Haurna, M; Beers, S. A.;
Thurston, L. S.; Dai, H. J.; Chen, C. H. J. Nat. Prod.
1989, 52, 606.
34. Baroni, M.; Constantino, G.; Cruciani, G.; Valigi, R.;
Clementi, S. Quant. Struct.-Act. Relats. 1993, 12, 9.
35. Dunn, W. J., III; Wold, S.; Edlund, U.; Hellberg,
S.; Gasteiger, J. Quant. Struct.-Act. Relats. 1984, 3,
131.
18. Lin, Z.; Neamati, N.; Zhao, H.; Kiryu, Y.; Turpin, J. A.;
Aberham, C.; Strebel, K.; Kohn, K.; Wiotvrouw, M.;
Pannecouque, C.; Debyser, Z.; De Clercq, E.; Rice, W. G.;
Pommier, Y.; Burke, T. R., Jr. J. Med. Chem. 1999, 42, 1401.
19. King, P. J.; Ma, G.; Miao, W.; Jia, Q.; McDougall, B. R.;
Reinecke, M. G.; Cornell, C.; Kuan, J.; Kim, T. R.;
Robinson, W. E., Jr. J. Med. Chem. 1999, 42, 497.
20. Mekouar, K.; Mouscardet, J.-F.; Desmaele, D.; Subra, F.;
Leh, H.; Savoure, D.; Auclair, C.; d’Angelo, J. J. Med.
´
Chem. 1998, 41, 2846, and references cited therein.
21. Hazuda, D. J.; Felock, P.; Witmer, M.; Wolfe, A.; Still-
mock, K.; Grobler, J. A.; Espeseth, A.; Gabryelski, L.;
Schleif, W.; Blau, C.; Miller, M. D. Science 2000, 287, 646.
22. Buolamwini, J. K.; Assefa, H. J. Med. Chem. 2002, 45, 841.
23. Raghavan, K.; Buolamwini, J. K.; Fesen, M. R.;
Pommier, Y.; Kohn, K. W.; Weinstein, J. N. J. Med.
Chem. 1995, 38, 890.
36. Matter, H.; Schwab, W. J. Med. Chem. 1999, 42, 4506.
37. Pastor, M.; Cruciani, G.; Clementi, S. J. Med. Chem.
1997, 40, 1455.
38. Beaver, D. J.; Roman, D. P.; Stoffel, P. J. J. Am. Chem.
Soc. 1957, 79, 1236.
39. Dibella E. P. US 3,389,986 (Cl. 71-123,) 25 June 1968,
Chem. Abstr. 1968, 69, 51812y.
24. SYBYL, Molecular Modeling Software, version 6.2;
Tripos Inc.: St. Louis, MO, 1995.
40. Rovnyak, G. C.; Millonig, R. C.; Schwartz, J.; Shu, V. J.
Med. Chem. 1982, 25, 1482.
25. Cramer, R. D., III; Patterson, D. E.; Bunce, J. D. J. Am.
Chem. Soc. 1988, 110, 5959.
26. Clementi S. GOLPE 3.0; Multivariate Infometric
Analyses (MIA): Perugia, Italy, 1995; SGI.
27. Goodford, P. J. J. Med. Chem. 1985, 28, 849.
28. Alonso, F.; Mico, I.; Najera, C.; Sansano, J. M.; Yus, M.;
Ezquerra J. Yruretagoyena, B.; Gracia, I. Tetrahedron
1995, 51, 10259.
29. Ungnade, H. E.; Morris, F. V. J. Am. Chem. Soc. 1948,
70, 1898.
41. Pauwels, R.; Balzarini, J.; Baba, M.; Snoeck, R.; Schols,
D.; Herdewijn, P.; Desmyster, J.; De Clerq, E. J. Virol.
Methods 1988, 20, 309.
42. Mohamadi, F.; Richards, N. G. J.; Guida, W. C.;
Liskamp, R.; Lipton, M.; Caufield, C.; Chang, G.;
Hendrickson, T.; Still, W. C. J. Comput. Chem. 1990,
11, 440.
43. GRID 20; Molecular Discovery Ltd.: Oxford, UK, 2001.
44. GOLPE 4.5.12; Multivariate Infometric Analyses: Viale
del Castagni, 16 Perugia, Italy, 1999.