Synthesis and biological evaluation of novel curcumin analogs as anti-cancer and anti-angiogenesis agents

Article abstract of DOI:10.1016/j.bmc.2004.05.006

A series of novel curcumin analogs were synthesized and screened for anti-cancer and anti-angiogenesis activities at Emory University and at the National Cancer Institute (NCI). These compounds are symmetrical α,β-unsaturated and saturated ketones. The majority of the analogs demonstrated a moderate degree of anti-cancer activity. Compounds 10, 11, and 14 exhibited a high degree of cytotoxicity in the NCI in vitro anti-cancer cell line screen. In addition, this screen revealed that these compounds inhibit tumor cell growth with a higher potency than the commonly used chemotherapeutic drug, cisplatin. In independent in vitro screens conducted at Emory, the same compounds plus 4, 5, 8, 9, and 13 exhibited a high degree of cytotoxicity to tumor cells. Analogs that were effective in the anti-cancer screens were also effective in in vitro anti-angiogenesis assays. Compounds 4, 9, 11, and 14 were most effective in the anti-angiogenesis assays run at Emory. In the assays conducted by the NCI, compound 14 was almost as potent as the anti-angiogenic drug TNP-470, which has undergone clinical trials. Based on the favorable in vitro anti-cancer and anti-angiogenesis results with 14, further in vivo tests were conducted. This compound effectively reduced the size of human breast tumors grown in female athymic nude mice and showed little toxicity. This data, coupled with the remarkable in vitro data, suggests that compound 14 may potentially be an effective chemotherapeutic agent. As a follow-up, a 3D quantitative structure relationship based on 14 has been developed. It shows a cross-validated r²(q²)=0.83 and a predictive r ²(p²)=0.71. COMPARE analysis suggests the compound to be a possible RNA/DNA antimetabolite, but also implies that the compound's cytotoxicity may arise from a presently unknown mechanism.

Full text of DOI:10.1016/j.bmc.2004.05.006

Bioorganic& Medicinal Chemistry 12 (2004) 3871–3883
Synthesis and biological evaluation of novel curcumin analogs
as anti-cancer and anti-angiogenesis agents
Brian K. Adams,a,b,e, Eva M. Ferstl,b, Matthew C. Davis,b Marike Herold,b
Serdar Kurtkaya,b Richard F. Camalier,c Melinda G. Hollingshead,c Gurmeet Kaur,c
Edward A. Sausville,c Frederick R. Rickles,d James P. Snyder,b,* Dennis C. Liottab
and Mamoru Shojia,*
aWinship Cancer Institute, Emory University, Atlanta, GA 30322, USA
bDepartment of Chemistry, Emory University, Atlanta, GA 30322, USA
cDevelopmental Therapeutics Program, DCTD, NCI-Frederick, Frederick, MD 21702, USA
dDepartments of Medicine and Pediatrics, The George Washington University School of Medicine and Health Sciences,
Washington, DC 20037, USA
eProgram in Molecular and Systems Pharmacology, Emory University, USA
Received 28 January 2004; accepted 3 May 2004
Available online 5 June 2004
Abstract—A series of novel curcumin analogs were synthesized and screened for anti-cancer and anti-angiogenesis activities at
Emory University and at the National Cancer Institute (NCI). These compounds are symmetrical a,b-unsaturated and saturated
ketones. The majority of the analogs demonstrated a moderate degree of anti-cancer activity. Compounds 10, 11, and 14 exhibited a
high degree of cytotoxicity in the NCI in vitro anti-cancer cell line screen. In addition, this screen revealed that these compounds
inhibit tumor cell growth with a higher potency than the commonly used chemotherapeutic drug, cisplatin. In independent in vitro
screens conducted at Emory, the same compounds plus 4, 5, 8, 9, and 13 exhibited a high degree of cytotoxicity to tumor cells.
Analogs that were effective in the anti-cancer screens were also effective in in vitro anti-angiogenesis assays. Compounds 4, 9, 11, and
14 were most effective in the anti-angiogenesis assays run at Emory. In the assays conducted by the NCI, compound 14 was almost
as potent as the anti-angiogenic drug TNP-470, which has undergone clinical trials. Based on the favorable in vitro anti-cancer and
anti-angiogenesis results with 14, further in vivo tests were conducted. This compound effectively reduced the size of human breast
tumors grown in female athymic nude mice and showed little toxicity. This data, coupled with the remarkable in vitro data, suggests
that compound 14 may potentially be an effective chemotherapeutic agent. As a follow-up, a 3D quantitative structure relationship
based on 14 has been developed. It shows a cross-validated r2ðq2Þ ¼ 0:83 and a predictive r2ðp2Þ ¼ 0:71. COMPARE analysis
suggests the compound to be a possible RNA/DNA antimetabolite, but also implies that the compound’s cytotoxicity may arise
from a presently unknown mechanism.
ꢀ 2004 Elsevier Ltd. All rights reserved.
1. Introduction
which is consumed daily by millions of people. Curcu-
min has been used by traditional medicine for liver
Curcumin (diferuloylmethane) 1 (Chart 1) is a b-dike-
tone constituent of the turmeric that is obtained from
the powdered root of Curcuma longa Linn. It is used as a
spice to give a specific flavor and yellow color to curry,
OH
O
MeO
HO
OMe
OH
1
Curcumin
O
O
OMe
OMe
MeO
MeO
OMe
OH
MeO
HO
Keywords: Synthetic curcumin analogs; Anti-cancer agent in vitro and
in vivo; Anti-angiogenesis properties; 3D quantitative structure rela-
tionship (QSAR).
2
3
* Corresponding authors. Tel.: +1-404-727-2415; fax: +1-404-727-65-
86; e-mail: snyder@euch4e.chem.emory.edu
These authors contributed equally in the writing of this manuscript.
BDMPP
BMHPC
Chart 1.
0968-0896/$ - see front matter ꢀ 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2004.05.006

3872
B. K. Adams et al. / Bioorg. Med. Chem. 12 (2004) 3871–3883
disease (jaundice), indigestion, urinary tract diseases,
rheumatoid arthritis, and insect bites. This phytochem-
ical has also demonstrated both anti-cancer and anti-
angiogenicproperties. Its anti-tumor properties include
growth inhibition and apoptosis induction in a variety
of cancer cell lines in vitro, as well as the ability to in-
hibit tumorigenesis in vivo.1–7 Curcumin’s anti-angio-
genesis effects include the inhibition of vascular
endothelial cell (VEC) proliferation in vitro and capil-
lary tube formation and growth in vivo.8;9 Because of its
anti-cancer and anti-angiogenic properties, low molec-
ular weight and lack of toxicity, curcumin could be an
ideal candidate for a chemotherapeutic agent.10 How-
ever, studies have demonstrated that its use is limited
in vivo due to low potency and poor absorption char-
acteristics.11 This notwithstanding, curcumin remains a
good lead compound for the design of analogs with a
similar safety profile, but increased activity and solu-
bility.
superoxide radicals and stimulate Phase 2 enzyme
transcription.15 We tested 2 and 3 in a number of in vitro
cell viability screens, and found the compounds to be
more active than curcumin. As a follow-up, a number of
novel mono-ketone lead compounds were synthesized
and biologically evaluated as potential anti-cancer and
anti-angiogenesis agent(s). In addition, compounds
similar to those described herein have been probed re-
cently as integrase inhibitors16 and blockers of herpes
simplex virus-1.17 During the course of our work, a
series of informative reports from the Dimmock–De
Clercq team18–21 and from El-Subbagh et al.17 examined
the broad-spectrum cell cytotoxicity of mono-ketones
closely related to those under study here. As illuminated
below, an important distinction between the latter
studies and our own is the location of the aryl substit-
utents in the ortho-position. This and the utilization of
an ortho-F instead of hydroxy or alkoxy substitutents
appears to impart an unexpected boost in cell-kill
activity. Accordingly, the main objective of the present
investigation is the synthesis and testing of a number of
symmetrical, but truncated curcumin analogs in order to
develop novel, active and selective anti-cancer and anti-
angiogenesis agent(s). In this paper, we describe the
synthesis of curcumin analogs, which were tested in
in vitro anti-cancer activity by us and the NCI. Fur-
thermore, the best compound 14 was tested for in vivo
antitumor effect.
Biologic analyses of various curcumin derivatives have
been conducted.5;8;12;13 These studies focused mainly on
changes in the b-diketone structure and aryl substitution
pattern of the molecule. Our intent was to explore
mono-carbonyl derivatives, diarylpentanoids, to see if
they possess increased anti-cancer and anti-angiogenic
activity and to pursue the mechanism of action of these
derivatives. Before the compounds were synthesized,
database searches were performed to identify commer-
cially available mono-ketone analogs with potential for
biological activity. A Unity/Tripos 2D similarity search
of the Available Chemicals Database (ACD) produced
two compounds with high similarity to curcumin: the
mono-ketones 1,5-bis(3,4-dimethoxyphenyl)-1,4-penta-
diene-3-one (BDMPP), 2, and 2,6-bis((3-methoxy-4-
hydroxyphenyl)-methylene)-cyclohexanone (BMHPC),
3 (Chart 1). BMHPC has been demonstrated to bind to
nuclear type II sites on breast cancer cells with high
affinity (Kd 1–7 nM) and to inhibit cell proliferation.14
In the same study, a close relative of BMHPC, 2,6-
bis((3,4-dihydroxyphenyl)-methylene)-cyclohexanone
(BDHPC) was shown to inhibit mouse mammary tumor
growth in vivo.14 Both curcumin and its analogs have
been shown to induce quinone reduction, quench
2. Results and discussion
2.1. Chemistry
The target compounds depicted in Scheme 1 were ob-
tained by allowing ketones to react with a variety of
aromaticaldehydes under basicaldol ocndensation
conditions to give the a,b-unsaturated ketones 4–8 and
10–13 (Table 1). Starting materials were selected to
allow an assessment of the biological consequences of
varying the substituents on the aromaticring and the
a-carbons of the ketone. All three mono-hydroxy
Scheme 1.

B. K. Adams et al. / Bioorg. Med. Chem. 12 (2004) 3871–3883
Table 1. Physical properties of synthesized curcumin analogs
3873
No
Mp ꢁC (solv)a
Formula
C, H, N calculated
C, H, N found
4
155 (A)
C17H12O3
C17H12O3
C17H12O3
C17H12F2O
C, 76.68; H, 5.30
C, 76.68; H, 5.30
C, 76.68; H, 5.30
C, 75.55; H, 4.48
C, 77.53; H, 6.16
C, 71.99; H, 5.18
C, 78.41; H, 5.92
C, 74.01; H, 5.23
C, 73.07; H, 4.52
C, 74.75; H, 5.96; N, 4.36
C, 67.92; H, 5.16; N, 3.77
C, 74.97; H, 7.40
C, 76.56; H, 5.32
C, 76.41; H, 5.48
C, 76.48; H, 5.01
C, 75.30; H, 4.55
C, 77.26; H, 6.17
C, 71.95; H, 5.46
C, 78.15; H, 6.03
C, 73.23; H, 5.23
C, 73.07; H, 4.47
5
198–200 (A)
244–246 (A)
68–70 (B)
6
7
8
122–123 (C)
128–130 (C)
148–149 (A)
148–150 (A)
157–158 (B)
168–171 (D)
186–189 (C)
68–70 (C)
C
C
19H18O3
21H18O5
9
10
11
12
13
14
15
C20H18O3
C19H16O4
C19H14F2O2
C20H17NO2
C, 75.11; H, 5.75; N, 4.68
C, 68.63; H, 5.81; N, 3.52
C, 75.13; H, 7.44
C21H19F2NO3
C
17H20O3
a Solvents: A, acetone–H2O; B, EtOAc–hexanes; C, EtOH; D, MeOH–H2O.
substitution patterns on the aromaticrings were ex-
plored because of the reported increase in anti-carcino-
genicactivity for ortho-hydroxy groups.12 In the present
work, since ortho-substituents consistently exhibit
better activity than their meta and para-counterparts
(vide infra), we also explored replacement of the
hydroxy groups with methoxy groups or fluorine.
Annulation of the a-carbons allowed us to examine the
preferred stericenvironment for this region of the
structure, as well as to study the bio-effects of modifying
hydrophilicity.
and median lethal concentration (LC50).30;31 GI50 refers
to the concentration at which the drug inhibits tumor
cell growth by 50%. TGI is defined as the concentration
at which the drug inhibits tumor cell growth by 100%.
LC50 is the concentration at which the drug decreases
the original tumor cell number by 50% (50% tumor cell
death). The NCI anti-cancer drug screen was designed
to distinguish between compounds that have a broad
spectrum of activity and those that are tumor or sub-
panel selective.32 In the assays performed at Emory,
human breast cancer (MDA-MB-231) and melanoma
(RPMI-7951) cell lines were treated for 72 h with at least
four different concentrations of compounds ranging
from 0.1 to 50 lM. A Neutral Red cell viability assay33
was used to determine a GI50 value for each compound
(Table 4).
In the series of hydroxy-substituted benzaldehydes (4–6,
10, 11, 13), chalcone derivatives 4,22;23 6,16;24–26 and
10,13;27 were previously reported. The aldol condensa-
tions proceeded using an excess of base in aqueous
ethanol. Yields of isolated products decreased signifi-
cantly when the hydroxy group was moved from the
ortho- to the meta- or para-positions (4, 5, 6). Prolonged
reaction times (up to two weeks) or gentle warming did
not positively affect the outcome of the reactions. By
contrast, aldol reactions involving salicylaldehyde were
considerably cleaner, producing 4 (75%), 10 (70%), 11
(60%), and 13 (75%) without complication. When flu-
oro- and methoxy-substituted aromaticaldehydes were
allowed to react with ketones in the presence of base
(NaOH), compounds 7, 8,28 and 12 were obtained in
analytically pure form and in moderate to good yields.
Analog 14 was obtained using an acid promoted aldol
condensation (glacial acetic acid), following the proce-
dure described by McElvain and McMahon.29 In
addition, analogs 4 and 7, both of which contain ortho-
substituents on the aromaticrings, were either aecty-
lated or reduced to the resulting derivatives, 9 and 15.
In the NCI anti-cancer cell line screen, the tested analogs
showed a broad spectrum of activity, as well as dis-
tinctive patterns of selectivity. With regard to the
broad-spectrum activity, Table 2 demonstrates that
compounds 10, 11, and 14 are effective against all of the
cell lines, as shown by their full panel mean-graph
midpoint (MG-MID) values. The MG-MID is the cal-
culated mean full panel or individual subpanel GI50,
TGI, or LC50. The former MG-MID (lM) values are
Table 2. Full panel mean-graph midpoint (MG-MID) GI50, TGI, LC50
values (lM)a of compounds in the NCI in vitro anti-cancer cell line
screenb
Full panel MG-MID
Compounds
GI50
TGI
LC50
10
11
2.3
1.2
0.7
9.5
7.3
7.3
4.7
37.0
25.2
16.0
90.6
67.2
14
Cisplatinc
Curcuminc
2.4
55.8
23.5
2.2. Anticancer cell line screening
a Calculated mean panel GI50, TGI, or LC50, which includes values at
either limit (> or <).
Curcumin analogs 10, 11, and 14 were subjected to the
NCI’s in vitro anti-cancer cell line screen, while the en-
tire series 1–15 was examined in a variety of in vitro
screens at Emory. In the NCI screen, 60 human tumor
cell lines were treated for 48 h with 10-fold dilutions of
compounds at a minimum of five concentrations (0.01–
100 lM). A sulforhodamine B (SRB) endpoint assay was
used to calculate three response parameters: median
growth inhibition (GI50), total growth inhibition (TGI),
b Human cancer cell lines were treated for 48 h with 10-fold dilutions of
compounds at a minimum of five concentrations (0.01–100 lM).
Sulforhodamine B (SRB) assay was used to calculate GI50, TGI, or
LC50 values. Data represents the mean of at least three experiments
(except for compound 10, n ¼ 1).
c The values for cisplatin and curcumin are based on previous screens
conducted by the NCI where the highest concentration tested was
also 100 lM.

3874
B. K. Adams et al. / Bioorg. Med. Chem. 12 (2004) 3871–3883
less than those for both curcumin and cisplatin when
tested in the same manner. For example, the full panel
MG-MID (lM) GI50 value for 14 (0.7) is 13.6- and 10.4-
fold better than cisplatin (9.5) and curcumin (7.3),
respectively. Table 2 also shows a TGI value for 14 (2.4)
that is 23.3- and 9.8-fold more efficacious than cisplatin
(55.8) and curcumin (23.5), respectively. Finally, this
table illustrates that the LC50 value for 14 (16.0) is
5.7- and 4.2-fold improved over cisplatin (90.6) and
curcumin (67.2), respectively. Cisplatin is a DNA cross-
linking agent that is commonly used in the treatment of
a number of different cancers, including testis, ovary,
bladder, head and neck, and lung. While the exact tar-
get(s) of our compounds is not well understood, it is
likely that the hydrophobicity and flexible molecular
shapes of these agents allow them to enter the nuclear
compartment where interaction with DNA (e.g., minor
groove binding or strand cross-linking) could possibly
occur. COMPARE analysis (see below) supports this
hypothesis. Although compound 14 seems to be more
effective than cisplatin in vitro, a direct comparison
between these two agents against human tumor xeno-
grafts in nude mice could yield useful information as to
compound 14’s potential as an anti-cancer agent. The
ratio obtained by dividing a compound’s full panel MG-
MID concentration by its individual subpanel MG-
MID is considered as a measure of selectivity.32 Ratios
greater than 6 indicate high selectivity toward the cor-
responding cell line subpanel, while ratios between 3 and
6 refer to moderate selectivity. Compounds 11 and 14
both showed moderate selectivity toward leukemia cell
lines based on GI50 values (Table 3).
Both Figure 1 and Table 3 demonstrate that compound
14 was similar to, but more active than both curcumin
and cisplatin against each of the individual cancer sub-
types. This drug was especially effective against leuke-
mia, colon, CNS, prostate, and breast cancer cells with
GI50 values less than 1.0 lM. Furthermore, Figure 1
shows that compound 14 was at least 10-fold more ac-
tive than either curcumin or cisplatin against these
cancer types. Table 3 demonstrates that analog 11 was
effective against leukemia and prostate cancer cells (GI50
values less than 1.0 lM). Derivative 10 did not show
GI50 values less than 1.0 lM for any tumor cell type, but
did show effectiveness against individual cell lines such
as leukemia RPMI-8226 and renal RXF-393 with GI50
values of 0.6 and 0.5 lM, respectively (data not shown).
Studies carried out at Emory focused on two highly
malignant human cancer cell lines, RPMI 7951 (mela-
noma) and MDA-MB-231 (breast). The MDA-MB-231
cell line was also tested as part of the NCI screen, but
the RPMI 7951 cell line was not. The GI50 values for 11
and 14 calculated against the breast cancer cell line were
very similar to those calculated by the NCI. Both
Table 3. Mean growth inhibitory concentration (GI50, lM)a of compounds in the NCI in vitro anti-cancer cell line screenb
Cancer cell subpanel
Compounds
Leukemia
NSCLe
Colon
CNS
Melanoma Ovarian
Renal
Prostate
Breast
10
11
1.6 (1.4)c
0.4 (3.0)c
0.2 (3.5)c
6.3
3.1
1.6
1.1
9.4
9.2
2.1
1.1
2.5
1.4
0.5
4.7
5.8
2.2
1.5
1.4
8.5
7.1
3.3
1.4
0.8
6.3
8.9
2.0
1.3
1.7
0.8
2.5
1.6
14
Cisplatind
Curcumind
0.3
0.7
0.4
5.6
0.6
21.0
4.7
10.2
8.9
13.3
5.9
3.7
11.2
a 50% Growth inhibition (micro-molar concentration of compounds, which causes a 50% inhibition of cell growth).
b See footnote b from Table 2.
c GI50 selectivity ratios obtained by dividing the compound’s full panel GI50 MG-MID (lM) (Table 2) by its leukemia subpanel MG-MID are shown
in parentheses. A selectivity ratio greater than 3.0 suggests that the compound is potentially more active against a particular type of cancer cell.32
d See footnote cfrom Table 2.
e Nonsmall cell lung cancer.
24
22
20
18
16
14
Compound14
Cisplatin
Curcumin
12
10
8
6
4
2
0
Leukemiaa Colona Renala Prostatea Breasta NSCL
Cancer Subpanel
CNS Melanoma Ovarian
Figure 1. Mean growth inhibitory concentration (GI50) of compound 14 compared to cisplatin and curcumin in a subset of cells in the NCI in vitro
anti-cancer screen. Compound 14 is more active than both of the latter agents against a wide variety of cancer typesa and at least 10-fold more potent
than either curcumin or cisplatin against the nine cancer types presented.

B. K. Adams et al. / Bioorg. Med. Chem. 12 (2004) 3871–3883
3875
Table 4. Mean growth inhibitory concentration (GI50, lM)a of com-
pounds in the Emory University in vitro anti-cancer cell line screenb
Table 5. Mean inhibitory concentration (IC50, lM)a of compounds in
the NCI in vitro anti-angiogenesis screenb
Cancer cell line
Assay
Compounds
Melanoma RPMI 7951 Breast MDA-MB-231
Compound
Cord formation
Cell migration
Curcumin (1)
BDMPP (2)
4.5 2.1
3.3 2.5
3.4 2.2
1.0 0.5
2.8 0.2
5.7 0.9
8.5 0.3
1.8 0.5
1.9 1.5
3.0 0.9
0.8 0.0
1.9 0.0
0.9 0.0
0.7 0.0
11.1 1.1
5.4 0.8
5.0 0.0
2.3 0.5
3.9 0.2
7.5 0.7
12.6 0.5
3.1 0.5
2.7 0.7
3.6 0.0
0.8 0.0
6.5 0.0
1.9 0.1
0.8 0.4
Curcumin (1)
10
11
14c
TNP-470
>10
>30
1.8 0.1
5.0 + 0.0
7.1 0.1
0.8 0.1
0.6 0.0
BMHPC (3)
4
6.5 0.7
1.5 0.4
0.7 0.2
5
6
a 50% Inhibitory concentration (molar concentration of compounds,
which causes a 50% inhibition as compared to control).
7
8
b Human Umbilical Vein Endothelial Cells (HUVECs) were used to
examine the effects of the compounds on cord formation and cell
migration. Data represents the mean of at least two experiments
performed in replicates of 2–8.
9
10
11
12
13
14
c Compound 14 was equipotent to TNP-470 in the Cord Formation
assay and only 2-fold less potent in the cell migration assay.
a See footnote a from Table 3.
b Human melanoma and breast cancer cell lines were treated for 72 h
with at least four different concentrations of compounds ranging
from 0.1 to 50 lM. Neutral Red cell viability assay33 was used to
calculate a GI50 value for each compound. Data represents the
mean SD of 2–5 experiments performed in duplicate.
analogs were subjected to in vitro proliferation assays at
Emory. The NCI analysis consists of three assays,
growth inhibition, cord formation, and cell migration.
Only the results of the cord formation and cell migration
assays are presented (Table 5) since growth inhibition
was studied at Emory (Table 6). In the cord formation
assay, HUVECs were allowed to form cords on matri-
gel-coated plates. Drug effect after 24 h compared to
untreated controls was assessed by measuring the length
of cords formed and the number of junctions. The cell
migration assay was conducted using a Boyden chamber
system and collagen-coated polycarbonate filters. The
ability of the drugs to inhibit HUVEC migration toward
a chemo-attractive source (bFGF, VEGF, or Swiss 3T3
conditioned medium) was determined by examining the
number of cells that migrated onto the filter. Negative,
unstimulated control values were subtracted from
stimulated control and drug treated values, and the data
was plotted as mean migrated cell SD. The IC50 was
compounds produced a value of 0.8 lM in our study
(Table 4), while 14 produced a value of 1.0 lM and 11
produced a value of 1.1 lM in the NCI study (data not
shown). The two compounds were the most effective
against both cell lines tested in this study, in agreement
with the results of the NCI screen.
Among the bis-benzylidene-acetone analogs (4–9),
compound 4 showed the highest activity based on GI50
(Table 4). The presence of ortho-OH was previously
shown to be important for activity in other curcumin
analogs.12 This effect is confirmed by our assays. Ana-
lyzing the data for the hydroxy-substituted derivatives
(4–6) revealed that moving the OH from the ortho- to
the para-position (as in curcumin) decreased cytotoxicity
(4 > 5 > 6). Replacing the hydroxyl group in 4 with
methoxy, 8, or fluorine, 7, led to a slight decrease in
activity. Also, esterifying 4 to compound 9 decreased
activity only very slightly, implying that under the assay
conditions the esters are probably hydrolyzed. Among
the unsaturated analogs derived from cyclic ketones (10,
11, 12, 13, 14), all of the analogs with ortho-hydroxyl
groups (10, 11, and 13) exhibited good activity. The
introduction of heteroatoms in the six-membered ring
(tetrahydropyran and piperidine) produced derivatives
11 and 13 that were more active than compound 4. The
oxygen-analog (11) was slightly more active than the
N-methyl-piperidone derivative 13. The ortho fluoro-
substituted piperidone analog 14 was the most active
compound in the NCI screen and in the assays per-
formed at Emory. The arylethyl alcohol 15 was devoid
of any cytotoxic activity at concentrations up to 50 lM
(data not shown).
Table 6. Mean growth inhibitory concentration (GI50, lM)a of com-
pounds in the Emory University in vitro anti-angiogenesis screenb
Cell line
Compounds
HUVEC
MS1
SVR
Curcumin (1)
BDMPP (2)
16.6 5.6
3.6 1.4
12.0 3.8
1.8 1.2
3.2 0.3
7.6 2.4
6.1 2.9
6.8 1.9
2.3 1.4
8.0 4.5
2.9 3.4
4.1 0.7
4.0 2.2
1.1 0.5
5.9 1.6
1.9 1.5
4.4 2.2
1.8 1.4
4.1 1.6
6.0 1.1
9.0 2.4
1.6 1.0
1.1 0.9
2.0 2.1
0.8 0.3
8.0 0.7
2.1 0.1
1.1 0.2
9.8 2.0
3.3 2.9
5.4 3.2
2.2 1.4
4.4 0.6
5.5 1.4
10.3 0.2
3.1 0.9
2.9 2.9
3.5 2.3
1.5 1.6
8.5 1.6
1.6 0.6
1.2 0.5
BMHPC (3)
4
5
6
7
8
9
10
11
12
13
14
a See footnote a from Table 3.
b Human and murine endothelial cell lines were treated for 72 h with
compounds at a minimum of four concentrations between 0.1 and
50 lM. Neutral Red assay33 was used to calculate a GI50 value for
each compound. Data represents the mean SD of 2–7 experiments
performed in duplicate.
2.3. Anti-angiogenesis screening
Curcumin analogs 10, 11, and 14 were subjected to the
NCI’s in vitro anti-angiogenesis screen. In addition, all

3876
B. K. Adams et al. / Bioorg. Med. Chem. 12 (2004) 3871–3883
calculated from the plotted data. The Emory anti-angi-
ogenesis screen focused on the ability of the compounds
to inhibit endothelial cell growth in vitro. HUVECs, and
two SV40 large T antigen-immortalized murine endo-
thelial cell lines, MS1 and SVR34, were treated for 72 h
with at least four different concentrations of compounds
ranging from 0.1 to 50 lM (Table 6). The SVR cells also
contain a mutant H-ras gene. A Neutral Red cell via-
bility assay33 was used to calculate the GI50 value for
each compound.
150
100
50
*
0
Control
2
20
100
Compound 14 (mg/kg)
The NCI data demonstrated that compounds 10, 11,
and 14 were effective against angiogenesis in both of the
assays described above (Table 5). Compound 14 was
almost as potent as the anti-angiogenesis drug, TNP-
470, in the cell migration assay, with IC50 values of
0.6 lM (TNP-470) and 0.8 lM (14). TNP-470 was
slightly more potent than 14 in the cord formation
assay, with IC50 values of 0.7 lM (TNP-470) and 1.5 lM
(14). TNP-470 has proven to be effective in vitro and in
animal model studies and is currently being tested in
Phase III clinical trials.35 Furthermore, compound 14
was more active than curcumin in both of the assays.
This is seen especially in the cord formation assay where
the IC50 values were >10 lM (curcumin) and 1.5 lM (14).
Figure 2. Regression of breast tumors in nude mice after treatment
with compound 14. 1,000,000 MDA-MB-231 breast cancer cells were
inoculated in the flanks of nude mice and tumors were allowed to grow
for three weeks. Compound 14 was administered by injection near the
base of the tumors three times a week for two weeks. The graph
demonstrates a dose dependent decrease in tumor weight after treat-
ment (Ãp < 0:05; n ¼ 5 mice/group).
treatment with compound 14. While there was no dif-
ference between the average tumor weight of the 2 mg/kg
group and control group, the average tumor weight in
the 20 mg/kg group was decreased to about 70% of the
control value, and that of the 100 mg/kg group de-
creased to about 45%. Although the latter dose appears
to be slightly above the practical treatment range, it was
necessary to effectively treat the extremely large tumors.
Remarkably, the 100 mg/kg dose demonstrated no
harmful side effects. No liver, kidney, or spleen toxicity
was seen, and all of the treated mice demonstrated
normal weight gain (data not shown). Furthermore, this
dose was well below the maximum tolerated dose
(MTD) of 200 mg/kg iv (400 mg/kg ip) determined by
the NCI. The MTD is calculated as the dose above
which the animals die from a single bolus injection.
Compound 14 appears to be much safer than Cisplatin,
which shows an MTD of 10 mg/kg ip. Thus, 14 seems to
have potential as a chemotherapeutic agent since it
demonstrates better activity but lower toxicity than a
commonly used drug.
In the experiments performed at Emory, the same gen-
eral pattern observed for the anti-cancer testing was also
seen in the anti-angiogenesis screen (Table 6). Com-
pound 4 showed the highest activity among the bis-
benzylidene-acetone analogs (4, 5, 6, 7, 8). In line with
the observations mentioned above, moving the OH from
the ortho- to para-position in the hydroxy-substituted
derivatives (4, 5, 6) decreases cytotoxicity (4 > 5 > 6).
Furthermore, replacement of the hydroxyl group in 4
with a methoxy (as in 8) or a fluorine group (as in 7)
leads to a slight decrease in activity. The unsaturated
analogs derived from cyclic ketones that contain ortho-
hydroxyl groups (10, 11, and 13) demonstrated good
activity, while the introduction of heteroatoms into
carbocyclic ketones (cf., tetrahydropyran derivative 11)
leads to compounds that are slightly more active than 4.
As was seen in the anti-cancer screen, the oxygen-analog
(11) is slightly more active than the N-methyl-piperidone
derivative (13). The ortho-fluoro substituted analog 14
was the most active compound in the NCI anti-angio-
genesis screen and in the Emory assay. Again, the
arylethyl alcohol (15) lacked any anti-angiogenic activ-
ity (data not shown).
2.5. Structure–activity correlation
In a first attempt to develop a quantitative structure
activity correlation, we have employed the quasi-atom-
isticreecptor model embedded in Quasar 3.5.
36
The
method requires a chemically meaningful superposition
of ligands accompanied by realistic atomic charges and
a reliable set of biological responses. The ligands are
divided into training and test sets. Subsequently, a
genetic algorithm is employed to iteratively construct a
three-dimensional envelope of complementary atomistic
probes about each ligand in the training set until
such envelopes provide the best ‘prediction’ of the bio-
logical data. In the present instance, the similarity of
structures presented no difficulty in the development of a
suitable molecular alignment (see Experimental). The
responses of the breast cancer cell line MDA-MB-231 to
the curcuminoids (Table 4) were taken as a representa-
tive sample of the biological data. Nineteen compounds
2.4. In vivo antitumor activity and toxicity
Since compound 14 proved to be the most active com-
pound in the in vitro anti-cancer and anti-angiogenesis
screens, further tests were performed to measure the
activity of this compound in vivo. The ability of com-
pound 14 to induce tumor regression was examined
by growing solid breast tumors in the flanks of
female athymicnude (nu/nu) mice. After three weeks of
tumor growth, drug was administered subcutaneously
near the base of the tumor for two weeks. Figure 2
shows a dose dependent decrease in tumor weight after

B. K. Adams et al. / Bioorg. Med. Chem. 12 (2004) 3871–3883
3877
Figure 3. Average quasi-atomisticreceptor model (Quasar 3.5) for the
curcuminoids derived by using compound 14 (pictured) as the phar-
macophore template. The Quasar receptor probes are colored as fol-
lows: positively-charged salt bridge (red), negatively-charged salt
bridge (blue), H-bond donor (green), H-bond acceptor (yellow), pos-
itively charged hydrophobic (saddle brown), negatively charged
hydrophobic (chocolate brown), H-bond flip-flop (purple), solvent
(cyan).
Figure 4. Quasar correlation of the curcuminoid quasi-atomistic
model shown in Figure 3; r2 0.85, cross-validated r2ðq2Þ 0.83, predictive
r2ðp2Þ 0.71; training set, green; test set, red. For error bars, see
Experimental.
with GI50’s ranging from 0.8 to 100 lM were assem-
bled.37 The same structures were encoded as a total of
114 conformers. For convenience in processing by
Quasar, the GI50’s were expressed as DG’s (kcal/mol).
For the training set, the resulting correlation gave an
r2 ¼ 0:85, a cross-validated r2ðq2Þ ¼ 0:83, an rms devi-
ation of 0.24 kcal/mol and a maximum deviation of
0.64 kcal/mol. Similar statistics arise for the test set (see
Experimental). Figure 3 displays the average quasi-
atomistic receptor model for the entire class of curc-
uminoids surrounding the highly active prototype 14.
While there is a fair amount of scatter in the probe types
from the model averaging process (see Experimental),
the central carbonyl is appropriately surrounded by a
cluster of red probes corresponding to a positively
charged salt-bridge. Likewise, the piperidone nitrogen is
associated with a yellow probe, namely a hydrogen-
bond acceptor.
rical analogs. Compounds with substitutions in the
meta- or para-positions (5 and 6) are not as active.
(3) The introduction of heteroatoms in the cyclic ketone
portion (11, 12, 13, 14) yields compounds with im-
proved anti-cancer and anti-angiogenesis activity.
(4) Saturation of the olefinicbond of one of the a,b-
unsaturated ketones produced an arylethyl alcohol
(15) that possessed little anti-cancer and direct
anti-angiogenic activity. While the unsaturated com-
pounds are likely acting as Michael acceptors, which
would normally result in unacceptable toxicity lev-
els, the high MTD suggests that the Michael addi-
tions may be occurring selectively in tumor cells.
Consistent with this hypothesis, treatment of
tumor-bearing mice with compound 14 represses
the tumors but shows no evidence of toxicity in
the animals.
Figure 4 depicts the overall correlation of training and
test sets, which represent GI50’s in terms of DGo over a
3.5 kcal/mol range. Scramble tests applied to the data
demonstrate the correlation to be robust (see Experi-
mental). Active and inactive compounds can be distin-
guished quite clearly. The substantial predictive
r2ðp2Þ ¼ 0:71 provides a useful tool for future synthetic
planning.
2.6. Possible mechanisms of action; COMPARE
In order to gain insight into the possible biological basis
for the cytotoxic action of compound 14, we performed
a COMPARE analysis48 against compounds exposed to
the NCI’s 60-cell line assay panel. Three indicators are
provided by the database: GI50 (growth inhibition),
LC50 (lethal concentration), and TGI (total growth
inhibition). Averaging over the top 25 hits in each cat-
egory, the dominant suggested mechanism is an RNA/
DNA antimetabolite (36% of total). This was followed
by topoisomerase II inhibition and antimitoticaction,
each of which registered at 8%. Interestingly, the largest
category for the compound was ‘unknown’ (40%).
Consequently, the cytotoxic behavior of 14 may repre-
sent a new mechanism for controlling tumor cell pro-
liferation. This question is under active investigation.
Efforts to relate structural modifications of the mono
keto-curcumin analogs with the observed anti-cancer
and anti-angiogenesis data support the following gen-
eral conclusions:
(1) The symmetrical a,b-unsaturated ketone structure
found in the novel analogs shows increased in vitro
anti-cancer and anti-angiogenesis activity compared
to the b-diketone structure of curcumin.
(2) ortho-Substitution on the aromaticrings (as for 4,
10, 11, 13, 14) enhances the activity of the symmet-

3878
B. K. Adams et al. / Bioorg. Med. Chem. 12 (2004) 3871–3883
3. Conclusions
distd H2O added, and the resulting purple solution was
neutralized by gently bubbling CO2 through it. The
yellow precipitate was filtered off, washed with distd
H2O, and dried under vacuum. The products were
purified by recrystallization from the indicated sol-
vent(s).
The synthetic curcumin analogs prepared in this study
were effective in anti-cancer and anti-angiogenesis
screens performed both at Emory University and at the
NCI. The compounds were more efficacious than curc-
umin and the commonly used chemotherapeutic drug,
cisplatin against a variety of tumor cell lines. In the NCI
anti-cancer screen, 10, 11, and 14 all show remarkable
cytotoxic activity, while 11 and 14 demonstrate selec-
tivity toward leukemia cell lines. The studies also dem-
onstrate the compounds to exert impressive blockade of
endothelial cell proliferation. In the NCI anti-angio-
genesis assays, 10, 11, and 14 were potent inhibitors of
cord formation and cell migration, and compound 14
was just as potent as TNP-470 in both of these assays. 14
also effectively induced regression of large human breast
tumors in mice and demonstrated little toxicity during
the treatment. Finally, it is important to mention that
this compound has been selected by the NCI for further
evaluation. Depending upon the results of these and
other studies, compound 14 may represent a promising
lead for clinical development. In the search for backup
compounds, we expect that the 3D-QSAR based on 14
will provide insights and syntheticguidance.
4.1.1.
1,5-Bis-(2-hydroxyphenyl)-penta-1,4-dien-3-one
(4). The crude product was recrystallized from ace-
tone/H2O to give 4 as a yellow solid (75%): mp 155 ꢁC
1
[lit.15 159 ꢁC (EtOH)]. H NMR (400 MHz, CD3OD) d
8.09 (d, J ¼ 16 Hz, 2H), 7.63 (dd, J ¼ 8:4, 1.6 Hz, 2H),
7.31 (d, J ¼ 16 Hz, 2H), 7.24 (td, J ¼ 7:6, 1.6 Hz, 2H),
6.88 (t, J ¼ 7:2 Hz, 4H). 13C NMR (100 MHz, CD3OD)
192.80, 158.90, 140.97, 133.10, 130.21, 126.33, 123.23,
121.00, 117.23. HREIMS: m=z 248.0845 ((M)H2O)þ,
C17H12O2 requires 248.0837).
4.1.2.
1,5-Bis-(3-hydroxyphenyl)-penta-1,4-dien-3-one
(5). The crude product was recrystallized from ace-
tone/H2O to produce 5 as a yellow solid (15%): mp 198–
200 ꢁC. 1H NMR (400 MHz, CD3OD) d 7.70 (d,
J ¼ 16 Hz, 2H), 7.24 (t, J ¼ 7:6 Hz, 2H), 7.17 (d,
J ¼ 16 Hz, 2H), 7.17 (d, J ¼ 8 Hz, 2H), 7.11 (s, 2H),
6.73 (dd, J ¼ 8, 2.4 Hz, 2H). 13C NMR (100 MHz,
CD3OD) d 191.61, 159.25, 145.47, 137.61, 131.20,
126.36, 121.41, 119.09, 115.88. HREIMS: m=z 266.0946
(Mþ, C17H14O3 requires 266.0943).
4. Experimental
All reagents were obtained from commercial suppliers
and used without further purification. Reaction progress
was monitored through thin layer chromatography
(TLC) on pre-coated glass plates (silica gel 60 F254
,
4.1.3.
1,5-Bis-(4-hydroxyphenyl)-penta-1,4-dien-3-one
(6). The crude product was recrystallized from ace-
tone/H2O to produce 6 as a yellow solid (6%): mp 244–
246 ꢁC [lit.17 248–250 ꢁC]. 1H NMR (400 MHz, CD3OD)
d 7.71 (2H, d, J ¼ 16 Hz), 7.58 (4H, d, J ¼ 8:8 Hz), 7.07
(2H, d, J ¼ 16 Hz), 6.84 (4H, d, J ¼ 8:4 Hz). 13C NMR
(100 MHz, CD3OD) d 191.86, 161.79, 145.34, 131.84,
127.84, 123.57, 117.07. HREIMS: m=z 266.0948 (Mþ,
C17H14O3 requires 266.0943).
0.25 mm thickness) purchased from EM Science. Flash
chromatography was carried out with silica gel 60 (230–
1
400 mesh ASTM) from EM Science. H NMR and 13C
NMR spectra were recorded on a Mercury 300 or
Varian 400 spectrometer. Unless otherwise specified, all
NMR spectra were obtained in deuterated chloroform
(CDCl3) and referenced to the residual solvent peak;
chemical shifts are reported in parts per million, and
coupling constants in hertz (Hz). Melting points were
determined on a Thomas Hoover capillary melting point
apparatus and are uncorrected. Mass spectra were ob-
tained on either a VG 70-S Nier Johnson or JEOL Mass
Spectrometer. Elemental analyses were performed by
Atlantic Microlab (Norcross, GA) for C, H, and N and
agreed with the proposed structures within 0.4% of the
theoretical values. The syntheses of compounds 4,15;16
6,16–19 8,23 9,38 and 1013;20 have been previously reported.
4.1.4.
2,6-Bis-(2-hydroxybenzylidene)-cyclohexanone
(10). The crude product was recrystallized from ace-
tone/H2O to produce 10 as a yellow solid (70%): mp
1
148–149 ꢁC [lit.21 150 ꢁC (CHCl3)]. H NMR (400 MHz,
CD3OD) d 7.98 (2H, s,), 7.32 (2H, dd, J ¼ 7:6; 1:2 Hz),
7.19 (2H, td, J ¼ 7:6; 1:2 Hz), 6.86 (4H, m), 2.86 (4H,
m), 1.75 (2H, m). 13C (100 MHz, CD3OD) d 192.88,
158.18, 137.26, 134.48, 131.55, 124.45, 120.21, 116.62,
29.93, 24.78. HREIMS: m=z 306.1263 (Mþ, C20H18O3
requires 306.1256).
4.1. General procedure for preparation of hydroxy-
substituted 3,5-bis-(arylidene)-ketones 4, 5, 6, 10, 11,
and 13
Aqueous NaOH (20 wt %, 15 mL, 75 mmol) was added
dropwise to a vigorously stirred solution of the appro-
priate hydroxy-benzaldehyde (51 mmol) and ketone
(25 mmol) in abs EtOH (20 mL). The reaction was stir-
red at room temperature (RT) for 48 h, then 100 mL
4.1.5.
3,5-Bis-(2-hydroxybenzylidene)-tetrahydro-4-H-
pyran-4-one (11). The crude product was recrystallized
from acetone/H2O to yield 11 as a yellow solid (60%):
mp 148–150 ꢁC. H NMR (400 MHz, CD3OD) d 8.08
(2H, s), 7.24 (2H, td, J ¼ 8:4; 1:6 Hz), 7.09 (2H, dd,
1

B. K. Adams et al. / Bioorg. Med. Chem. 12 (2004) 3871–3883
3879
J ¼ 7:6; 1:6 Hz), 7.90–7.86 (4H, m), 4.84 (4H, d,
J ¼ 1:6 Hz). 13C NMR (100 MHz, CD3OD) d 187.82,
158.48, 133.93, 132.46, 131.88, 123.29, 120.47, 116.80,
70.02. HREIMS: m=z 290.0933 (Mþ, C19H16O4 requires
290.0943).
161.04 (d, J ¼ 251 Hz), 134.78, 131.54 (d, J ¼ 9 Hz),
131.08 (d, J ¼ 2 Hz), 129.56 (d, J ¼ 4 Hz), 124.30 (d,
J ¼ 4 Hz), 122.84 (d, J ¼ 14 Hz), 116.23 (d, J ¼ 22 Hz),
68.91 (d, J ¼ 5 Hz). HREIMS: m=z 312.0950 (Mþ,
C19H14F2O2 requires 312.0962).
4.3. 1,5-Bis-(2-methoxyphenyl)-penta-1,4-dien-3-one (8)
4.1.6. 3,5-Bis-(2-hydroxybenzylidene)-1-methyl-4-piper-
idone (13). The crude product was recrystallized from
methanol/H2O to produce 13 as a yellow solid (75%):
mp 168–171 ꢁC. H NMR (400 MHz, CD3OD) d 8.11
(2H, s), 7.23 (4H, t, J ¼ 7:6 Hz), 6.88 (4H, t,
J ¼ 8:0 Hz), 3.76 (4H, d, J ¼ 1:2 Hz), 2.42 (3H, s). 13C
NMR (100 MHz, CD3OD) d 188.40, 158.58, 135.09,
133.09, 132.27, 131.64, 123.59, 120.38, 116.87. HRE-
IMS: m=z 303.1259 (Mþ)H2O, C20H17NO2 requires
303.1259).
NaOH (0.10 mmol) was added as a solid to a stirred
solution of 2-methoxybenzaldehyde (2.50 mmol) and
acetone (1.20 mmol) in EtOH abs (5 mL). A yellow solid
started forming within one hour, stirring was continued
at room temperature for 20 h, the product filtered off,
washed with cold EtOH abs and H2O distd, and dried
under vacuum yielding 8 as a yellow solid (60%): mp
1
1
122–123 ꢁC (EtOH), [lit.21;39;40 120–121 ꢁC (EtOH)]. H
NMR (400 MHz, CDCl3) d 8.07 (2H, d, J ¼ 16 Hz), 7.63
(2H, dd, J ¼ 7:6, 1:6 Hz), 7.37 (2H, ddd, J ¼ 7:2 Hz,
J ¼ 1:6 Hz, J ¼ 1:2 Hz), 7.18 (2H, d, J ¼ 16 Hz), 6.99
(2H, t, J ¼ 7:6 Hz), 6.93 (2H, d, J ¼ 8:4 Hz). 13C NMR
(100 MHz, CDCl3) d 190.21, 158.74, 138.39, 131.78,
128.89, 126.42, 124.12, 120.92, 111.35, 55.70. HREIMS:
m=z 294.1256 (Mþ, C19H18O3 requires 294.1256).
4.2. General procedure for preparation of fluoro-substi-
tuted 3,5-bis-(arylidene)-ketones 7 and 12
A
solution of substituted fluorobenzaldehyde
(5.00 mmol) in EtOH abs (1 mL) was added at room
temperature over a period of 5 min, with stirring, to a
solution of NaOH (0.75 mmol) and ketone (2.50 mmol)
in a mixture of EtOH abs (7 mL) and H2O distd (7 mL).
The solution turns yellow immediately, and within
10 min, 7 produced a yellow oil, while 12 produced a
yellow precipitate. Stirring was continued at room
temperature for 3 h.
4.4. 1,5-Bis-(2-acetylphenyl)-penta-1,4-diene-3-one (9)
This was obtained by acetylation of 4 using acetylchlor-
ide (2.4 equiv) and K2CO3 (3 equiv) in THF. The crude
product was recrystallized from EtOH and obtained as a
yellow solid (80%): mp 128–130 ꢁC, [lit.38 129 ꢁC
(EtOH)].
4.5. 3,5-Bis-(2-fluorobenzylidene)-piperidin-4-one, acetic
acid salt (14)
4.2.1. 1,5-Bis-(2-fluorophenyl)-penta-1,4-dien-3-one (7).
The product was produced as a yellow oil and
isolated from the reaction mixture after diluting with
H2O dist by extraction with CH2Cl2 (3·). The
combined organic layers were dried over MgSO4 and
concentrated. The crude product was purified by column
chromatography on silica gel (15% EtOAc/hexanes)
yielding the product as a yellow solid (50%): mp 68–
70 ꢁC. 1H NMR (400 MHz, CDCl3,) d 7.86 (2H, d,
J ¼ 16 Hz), 7.63 (2H, td, J ¼ 7:6; 1:6 Hz), 7.42–
7.35 (2H, m), 7.18 (2H, d, J ¼ 16 Hz), 7.26–7.10 (4H,
m). 13C NMR (100 MHz, CDCl3) d 189.17, 161.84 (d,
J ¼ 253 Hz), 136.30 (d, J ¼ 2 Hz), 132.13 (d, J ¼ 9 Hz),
129.56 (d, J ¼ 2 Hz), 127.78 (d, J ¼ 6 Hz), 124.73 (d,
J ¼ 4 Hz), 123.03 (d, J ¼ 11 Hz), 116.47 (d, J ¼ 22 Hz).
HREIMS: m=z 270.0865 (Mþ, C17H12F2O requires
270.0856).
4-Piperidone hydrochloride monohydrate (307 mg,
2.00 mmol) was suspended in glacial acetic acid (8 mL)
and saturated with HCl gas at room temperature. To the
resulting clear solution 2-fluorobenzaldehyde (0.59 mL,
5.60 mmol) was added and the reaction allowed to stand
at room temperature for 48 h. The forming yellow
crystals were filtered off, washed with EtOH abs, and
dried under vacuum yielding 14 as yellow plates (91%):
mp 186–189 ꢁC. 1H NMR (400 MHz, DMSO-d6) d 10.11
(1H, s-br), 7.90 (2H, s), 7.57 (2H, qd, J ¼ 7:6; 1:6 Hz),
7.51 (2H, td, J ¼ 8:0; 1:2 Hz), 7.37 (4H, q, J ¼ 10:0 Hz),
4.37 (4H, s), 3.60–3.20 (1H, s-br), 1.91 (3H, s). 13C NMR
(100 MHz, DMSO-d6) d 181.95, 172.04, 160.33 (d,
J ¼ 249 Hz), 132.57 (d, J ¼ 9 Hz), 131.81 (d, J ¼ 4 Hz),
131.05, 129.86, 124.96 (d, J ¼ 3 Hz), 121.49 (d,
J ¼ 13 Hz), 116.1 (d, J ¼ 21 Hz), 43.79, 21.11. HRE-
IMS: m=z 311.1123 (Mþ)HOAc, C19H15NOF2 requires
311.1122).
4.2.2. 3,5-Bis-(2-fluorobenzylidene)-tetrahydro-4-H-pyr-
an-4-one (12). The obtained yellow solid, was filtered
off, washed with water and hexanes, and dried under
vacuum. The crude product was recrystallized from
EtOH yielding 12 (84%): mp 157–158 ꢁC. 1H NMR
(400 MHz, CDCl3) d 7.92 (2H, d, J ¼ 1:2 Hz), 7.40–7.36
(2H, m), 7.20–7.18 (4H, m), 7.13 (2H, t, J ¼ 9:6 Hz),
4.80 (4H, s). 13C NMR (100 MHz, CDCl3) d 185.09,
4.6. General procedure for preparation of substituted
saturated alcohols 15
Alcohols 15 were obtained by hydrogenation of 4 or 7 in
EtOH abs at 45 psi for 4 h using Raney Nickel as the

3880
B. K. Adams et al. / Bioorg. Med. Chem. 12 (2004) 3871–3883
catalyst. Filtration through Celite and concentration
under vacuum yielded the crude product, which was
purified by chromatography on silica gel (25% EtOAc/
hexanes). Reaction progress can be detected by disap-
pearance of the yellow color of the starting materials.
Excessive hydrogenation yielded alcohols (15) exclu-
sively (white solids, 60–80% yield: R ¼ 2-OH; mp 79–
81 ꢁC, R ¼ 2-F; mp 68–70 ꢁC (Scheme 1)).
were then transferred to a 96-well plate and the
absorbances were read on a plate reader (BIO-
TEK Instruments, Winooski, VT) at a wavelength of
570 nm.
4.9. NCIin vitro anti-cancer cell line screen
The human tumor cell lines used in the cancer screening
panel were grown in RPMI 1640 medium containing 5%
fetal bovine serum and 2 mM L-glutamine. The cells
4.7. Cell culture
were inoculated into 96-well microtiter plates in 100 lL
at plating densities ranging from 5000 to 40,000 cells/
well depending on the doubling time of individual cell
lines. After cell inoculation, the microtiter plates were
incubated at 37 ꢁC, 5% CO2, 95% air in a humid atmo-
sphere for 24 h prior to addition of experimental drugs.
After 24 h, two plates of each cell line were fixed in situ
by the gentle addition of 50 lL of cold 50% (w/v) TCA
(final concentration, 10% TCA) and incubated for
60 min at 4 ꢁC, to represent a measurement of the cell
population for each cell line at the time of drug addition
(Tz). Experimental drugs were solubilized in DMSO at
400-fold the desired final maximum test concentration
and stored frozen prior to use. At the time of drug
addition, an aliquot of frozen concentrate was thawed
and diluted to twice the desired final maximum test
concentration with complete medium containing 50 lg/
mL gentamicin. An additional 4-fold, 10-fold, or one
half log serial dilutions were made to provide a total of
five drug concentrations plus control. Aliquots of
100 lL of these different drug dilutions were added to
the appropriate microtiter wells already containing
100 lL of medium, resulting in the required final drug
concentrations. Following drug addition, the plates were
incubated for an additional 48 h. For adherent cells, the
assay was terminated by the addition of cold TCA. The
supernatant was discarded, and the plates were washed
five times with tap water and air-dried. Sulforhodamine
B (SRB) solution (100 lL) at 0.4% (w/v) in 1% acetic
acid was added to each well, and plates were incubated
for 10 min at room temperature. After staining, un-
bound dye was removed by washing five times with 1%
acetic acid and the plates were air dried. Bound stain
was subsequently solubilized with 10 mM trizma base,
and the absorbance was read on an automated plate
reader at a wavelength of 515 nm. For suspension cells,
the methodology is the same except that the assay was
terminated by fixing settled cells at the bottom of the
wells by gently adding 50 lL of 80% TCA (final con-
centration, 16% TCA). Using the eight absorbance
measurements [time zero, (Tz), control growth in the
absence of drug, (C), and test growth in the presence of
drug at the five concentration levels (Ti)], the percentage
growth was calculated at each of the drug concentra-
tions levels.
RPMI-7951 human melanoma and MDA-MB-231
human breast cancer cells were purchased from Ameri-
can Type Cell Collection (ATCC), Rockville, MD.
Human Umbilical Vein Endothelial Cells (HUVECs)
were obtained from the Department of Dermatology,
Emory University. Murine endothelial cells infected
with either simian virus 40 (SV40) large T antigen (MS1)
or SV40 large T antigen and activated H-ras (SVR) were
a kind gift from Dr. Jack Arbiser at Emory. RPMI 7951
and MDA-MB-231 cell lines were maintained in MEM-
alpha medium (GIBCO-BRL, Long Island, NY) con-
taining 10% fetal bovine serum (FBS) (Mediatech,
Herndon, VA), penicillin (100 units/mL) and strepto-
mycin (100 lg/mL) and 2 mM
L-glutamine. Cells were
incubated at 37 ꢁC in 5% CO2, 95% air in a humid
atmosphere. MS1 and SVR cell lines were cultured in
DMEM (Mediatech, Herndon, VA) containing 10%
fetal bovine serum (FBS) (Mediatech, Herndon, VA),
penicillin (100 units/mL), streptomycin (100 lg/mL) and
2 mM
L-glutamine. HUVECs were cultured in M199
medium (GIBCO-BRL) containing 20% FBS, penicillin
(100 units/mL), streptomycin (100 lg/mL), amphotericin
B as Fungizone (0.25 lg/mL), 2 mM L-glutamine, hep-
arin (10,000 units/mL), and endothelial mitogen. In
order to avoid mutational changes of cells lines, cells
have been expanded in culture, aliquoted, and stored in
a liquid nitrogen freezer when purchased from the
ATCC.
4.8. Neutral red assay for cell viability
This assay was carried out according to the methods
published by Zhang et al.33 Tumor cells (MDA-MB-231,
RPMI 7951) for the anti-cancer screen, or endothelial
cells (HUVEC, MS1, SVR) for the anti-angiogenesis
screen were plated at a concentration of 20,000 cells/well
in a 48-well plate and incubated overnight. Compounds
or vehicle, dimethyl sulfoxide (DMSO 0.1% (Sigma,
St. Louis, MO)) were then added and the plates were
incubated for 72 h. Supernatant from each well was
aspirated and warm (37 ꢁC) media containing Neutral
Red (stock-3.33 lg/lL) solution (GIBCO-BRL, Long
Island, NY) was added at a concentration of 50 lg/mL
(approximately 15 lL/mL medium). The plates were
incubated at 37 ꢁC for 30 min. In principle, only viable
cells take up the Neutral Red dye. Next, the cells were
washed twice with PBS and 400 lL of alcoholic-
HCl (0.5 N HCl, 35% ethanol) was added to each well.
The plates were placed on a plate shaker until all
residues were solubilized. The solubilized mixtures
Percentage growth inhibition was calculated as:
½ðTi ꢀ TzÞ=ðC ꢀ TzÞꢁ ꢂ 100 for concentrations for
which Ti P Tz,
½ðTi ꢀ TzÞ=Tzꢁ ꢂ 100 for concentrations for which
Ti < Tz.

B. K. Adams et al. / Bioorg. Med. Chem. 12 (2004) 3871–3883
3881
Three dose response parameters were calculated for each
4.11. In vivo anti-tumor activity and toxicity
experimental agent. Growth inhibition of 50% (GI50)
was calculated from ½ðTi ꢀ TzÞ=ðC ꢀ TzÞꢁ ꢂ 100 ¼ 50,
which is the drug concentration resulting in a 50%
lower net protein increase in the treated cells
(measured by SRB staining) as compared to the net
protein increase seen in the control cells. The drug
concentration resulting in total growth inhibition (TGI)
was calculated from Ti ¼ Tz. The LC50 (concentration of
drug resulting in a 50% reduction in the measured pro-
tein at the end of the drug treatment as compared to that
at the beginning) indicating a net loss of cells following
treatment was calculated from ½ðTi ꢀ TzÞ=Tzꢁ ꢂ 100 ¼
ꢀ50. Values were calculated for each of these
three parameters if the level of activity is reached;
however, if the effect was not reached or was exceeded,
the value for that parameter was expressed as greater
or less than the maximum or minimum concentration
tested.
Human tumor xenografts: Twenty female, athymicnude
(nu/nu) mice were purchased from Harlan Laboratories.
MDA-MB-231 human breast cancer cells (106 cells/
0.1 mL) were injected subcutaneously (sc) into the left
flank of each mouse. The cells were given as a suspen-
sion in sterile MEM-alpha medium (GIBCO-BRL,
Long Island, NY) containing 10% fetal bovine serum
(FBS) (Mediatech, Herndon, VA), penicillin (100 units/
mL) and streptomycin (100 lg/mL) and 2 mM L-gluta-
mine. The tumors were allowed to grow for three weeks
and then treated with either 5% dextrose in water
(vehicle), 2, 20, or 100 mg/kg compound 14 for two
weeks. The drugs were given as a scinjection near the
base of the tumors. After treatment, the mice were
sacrificed and the tumors were excised and weighed. The
weights of the tumors from the vehicle-treated mice were
compared to that of the drug treated mice and statistical
analysis was performed using ANOVA. Data are pre-
sented as mean SE.
4.10. NCIanti-angiogenesis screen
Toxicity: The weight of each mouse was recorded before
and after treatment. At the end of the study, liver,
spleen, and kidney samples were submitted to Dirk
Dillehay, DVM, Emory University, for pathological
examination. The maximum tolerated dose (MTD) for
compound 14 was determined by the NCI as previously
described.41
Cord formation assay: Matrigel (60 lL of 10 mg/mL;
Collaborative Laboratories, StonyBrook, NY) was
placed in each well of an ice-cold 96-well plate. The plate
was allowed to sit at room temperature for 15 min then
incubated at 37 ꢁC for 30 min to permit the matrigel to
polymerize. In the meantime, HUVECs were prepared
in EGM-2 medium at a concentration of 2 · 105 cells/
mL. The test compounds were prepared at 2· the de-
sired concentration (5 concentration levels) in the same
medium. Cells (500 lL) and 2· drug (500 lL) were
mixed and 200 lL of this suspension was placed in
duplicate on the polymerized matrigel. After 24 h incu-
bation, triplicate pictures were taken for each concen-
tration using a Bioquant Image Analysis system. Drug
effect (IC50) was assessed compared to untreated con-
trols by measuring the length of cords formed and
number of junctions.
4.12. QSAR development
The 20 curcumin analogs were constructed in Macro-
model 7.1,42 optimized with the MMFF force field43
supplemented by the GBSA/H2O solvent model and
subjected to a conformational search with the Monte
Carlo method44 to give a total of 114 conformers.
Electrostatic potential atomic partial charges were ap-
plied to each of the structures using the MNDO/ESP
procedure in MOPAC 6.0.45 The conformers were
superimposed by means of FlexS in Sybyl 6.746 with the
structure of 14 as template to give the pharmacophore.
A single pdb file containing all the conformers was
constructed with PrGen 2.147 and then submitted to
Quasar 3.5 for QSAR analysis.36 For the training set of
14 compounds, the iterative Quasar procedure evolved
to the quasi-atomisticbinding site depicted in Figure 3
with r2 ¼ 0:85 and a cross-validated r2 (i.e., q2) ¼ 0.83.
The RMS deviation is 0.24 kcal/mol with a maximum
deviation of 0.64 kcal/mol. These values arise by aver-
aging over the best 300 models provided by Quasar’s
genetic search algorithm. The average receptor model is
depicted in Figure 3. When applied to a test set of five
additional analogs, a predictive r2 (i.e., p2) ¼ 0.71 was
obtained with an RMS deviation of 0.44 kcal/mol and a
maximum deviation of 0.72 kcal/mol. The QSAR cor-
relation combining both training and test sets is depicted
in Figure 4. The individual error bars arise by averaging
the maximum and minimum deviations in kcal/mol for
each ligand over the 300 best models. To assure that the
correlation is not fortuitous, we performed three
scramble tests on the data. The average for the three is
Cell migration assay: Migration was assessed using a
48-well Boyden chamber and 8 lm pore size collagen-
coated (10 lg/mL rat tail collagen; Collaborative
Laboratories, StonyBrook, NY) polycarbonate filters
(Osmonics, Inc., Minnetonka, MN). The bottom
chamber wells received 27–29 lL of DMEM medium
alone (baseline) or medium containing chemo-attractant
(bFGF, VEGF, or Swiss 3T3 cell conditioned medium).
The top chambers received 45 lL of the HUVEC cell
suspension (1 · 106 cells/mL) prepared in DMEM con-
taining 1% BSA with or without test compounds. After
5 h incubation at 37 ꢁC, the membrane was rinsed in
PBS, fixed, and stained in Diff-Quick solutions (Baxter
Healthcare Co., Miami, FL). The filter was placed on a
glass slide with the migrated cells facing down and cells
on top were removed using a Kimwipe. The testing was
performed in 4–6 replicates and five fields were counted
from each well. Negative unstimulated control values
were subtracted from stimulated control and drug
treated values and data was plotted as mean migrated
cell SD. IC50 values were calculated from the plotted
data.

3882
B. K. Adams et al. / Bioorg. Med. Chem. 12 (2004) 3871–3883
p2 ¼ ꢀ0:30, demonstrating the robustness of the plot in
G. Y.; Balzarini, J.; De Clercq, E. J. Med. Chem. 1999, 42,
1358–1366.
Figure 4.
19. Dimmock, J. R.; Kandepu, N. M.; Nazarali, A. J.;
Kowalchuk, T. P.; Motaganahalli, N.; Quail, J. W.;
Mykytiuk, P. A.; Audette, G. F.; Prasad, L.;
ꢀ
Perjesi, P.; Allen, T. M.; Santos, C. L.; Szydlowski, J.;
Acknowledgements
De Clercq, E.; Balzarini, J. J. Med. Chem. 1998, 41, 4012–
4020.
This work was supported by grants from the Depart-
ment of Health and Human Services (CA82995) and the
US Department of Defense, the Division of US Army
(DAMD17-00-1-0241), and a contract with the George
Washington University Medical Center. A portion of
the studies described herein were also supported with
Federal funds from the National Cancer Institute,
National Institutes of Health, under Contract N01-CM-
5600. Mention of trade names or commercial products
in this publication does not imply endorsement by the
US Government.
20. Dimmock, J. R.; Kumar, P.; Nazarali, A. J.; Motagana-
halli, N.; Kowalchuk, T. P.; Beazeley, M. A.; Quail, J. W.;
Oloo, E. O.; Allen, T. M.; Szydlowski, J.; De Clercq, E.;
Balzarini, J. Eur. J. Med. Chem. 2000, 35, 967–
977.
21. Dimmock, J. R.; Padmanilayam, M. P.; Puthucode, R. N.;
Nazarali, A. J.; Motaganahalli, N.; Zello, G. A.; Quail, J.
W.; Oloo, E. O.; Kraatz, H.-B.; Prisciak, J. A.; Allen, T.
M.; Santos, C. L.; Balzarini, J.; De Clercq, E.; Manava-
thu, E. K. J. Med. Chem. 2001, 44, 586–593.
22. McGookin, A.; Heilbron, I. M. J. Chem. Soc. 1924, 125,
2105–2199.
23. Nauduri, D.; Reddy, G. B. S. Chem. Pharm. Bull. 1998, 46,
1254–1260.
24. Artico, M.; DiSanto, R.; Costi, R.; Novellino, E.; Greco,
G.; Massa, S.; Tramontano, E.; Marongiu, M. E.;
DeMontis, A.; LaColla, P. J. Med. Chem. 1998, 41,
3948–3960.
25. Jennings, W. P.; McGookin, A. J. Chem. Soc. 1934, 1741–
1742.
26. Sardjiman, S. S.; Reksohadiprodjo, M. S.; Hakim, L.; van
der Goot, H.; Trimmerman, H. Eur. J. Med. Chem. Chim.
Ther. 1997, 32, 625–630.
27. Borsche, W.; Geyer, A. Justus Liebigs Ann. Chem. 1912,
393, 29–60.
28. Kraft, D.; Cacciapaglia, R.; Boehmer, V.; El-Fadl, A. A.;
Haekema, S. J. Org. Chem. 1992, 57, 826–834.
29. McElvain, S. M.; McMahon, R. E. J. Am. Chem. Soc.
1949, 71, 901–906.
30. Grever, M.; Schepartz, S.; Chabner, B. Semin. Oncol.
1992, 19, 622–638.
31. Monks, A.; Scudiero, D.; Skehan, P., et al. J. Natl. Cancer
Inst. 1991, 83, 757–766.
32. Boyd, M.; Paull, K. Drug Dev. Res. 1995, 34, 91–
109.
References and notes
1. Mehta, K.; Pantazis, P.; McQueen, T.; Aggarwal, B. B.
Anti-Cancer Drugs 1997, 8, 470–481.
2. Kuo, M. L.; Huang, T. S.; Lin, J. K. Biochim. Biophys.
Acta 1996, 1317, 95–100.
3. Jee, S. H.; Shen, S. C.; Tseng, C. R.; Chiu, H. C.; Kuo,
M. L. J. Invest. Dermatol. 1998, 111, 656–661.
4. Kawamori, T.; Lubet, R.; Steele, V. E.; Kelloff, G. J.;
Kaskey, R. B.; Rao, C. V.; Reddy, B. S. Cancer Res. 1999,
597–601.
5. Singletary, K.; MacDonald, C.; Iovinelli, M.; Fisher, C.;
Wallig, M. Carcinogenesis 1998, 19, 1039–1043.
6. Ruby, A. J.; Kuttan, G.; Babu, K. D.; Rajasekeharan,
K. N.; Kuttan, R. Cancer Lett. 1995, 94, 79–83.
7. Kuttan, R.; Bhanumathy, P.; Nirmala, K.; George, M. C.
Cancer Lett. 1985, 29, 197–202.
8. Arbiser, J. A.; Klauber, N.; Rohan, R.; van Leeuwen, R.;
Huang, M. T.; Fisher, C.; Flynn, E.; Byers, H. R. Mol.
Med. 1998, 4, 376–383.
9. Thaloor, D.; Singh, A. K.; Sidhu, G. S.; Prasad, P. V.;
Kleinman, H. K.; Maheshwari, R. K. Cell Growth
Differentiation 1998, 9, 305–312.
10. (a) Kelloff, G.; Crowell, J.; Hawk, E. J. Cell. Biochem.
1996, 26, 72–85; (b) Aggarwal, B. B.; Kumar, A.; Bharti,
A. C. Anticancer Res. 2003, 23, 363–398.
33. Zhang, S. Z.; Lipsky, M. M.; Trump, B. F.; Hsu, I. C. Cell
Biol. Toxicicol. 1990, 6, 219–234.
34. Arbiser, J. L.; Moses, M. A.; Fernandez, C. A.; Ghiso, N.;
Cao, Y.; Klauber, N.; Frank, D.; Brownlee, M.; Flynn, E.;
Parangi, S.; Byers, H. R.; Folkman, J. Proc. Natl. Acad.
Sci. 1997, 94, 861–866.
11. Shoba, G.; Joy, D.; Joseph, T., et al. Planta Med. 1998, 64,
353–356.
35. Yanase, T.; Tamura, M.; Fujita, K.; Kodama, S.; Tanaka,
K. Cancer Res. 1993, 53, 2566–2570.
12. Anto, R. J.; George, J.; Dinesh Babu, K. V.; Rajasekeh-
aran, K. N.; Kuttan, R. Mutation Res. 1996, 370, 127–131.
13. Dinkova-Kostova, A. T.; Abeygunawardana, C.; Talalay,
P. J. Med. Chem. 1998, 41, 5287–5296.
14. Markaverich, B. M.; Schauweker, T. H.; Gregory, R. R.;
Varma, M.; Kittrell, F. S.; Medina, D.; Varma, R. S.
Cancer Res. 1992, 52, 2482–2488.
36. Vedani, A.; Dobler, M. 5D-QSAR: J. Med. Chem. 2002,
45, 2139–2149.
37. While only 14 curcuminoid compounds are listed in Tables
2–6, for the purpose of QSAR model-building the tabu-
lated values were supplemented by six additional data-
points from the same compound class; Lu, Y. Emory
University, private communication.
15. Dinkova-Kostova, A. T.; Abeygunawardana, C.; Talalay,
P. J. Med. Chem. 1998, 41, 5287–5296.
38. Mora, P. T.; Szeki, T. J. Am. Chem. Soc. 1950, 72, 3009–
3013.
16. Artico, M.; Di Santo, R.; Costi, R.; Novellino, E.; Greco,
G.; Massa, S.; Tramontano, E.; Marongui, M.; De Montis,
A.; La Colla, P. J. Med. Chem. 1998, 41, 3948–
3960.
39. Tanaka, T.; Miyaguchi, M.; Mochisuki, R. K.; Tanaka,
S.; Okamota, M. Heterocycles 1987, 25, 463–484.
40. Baeyer, A.; Villiger, V. Chem. Ber. 1902, 35, 3013–
3033.
17. El-Subbagh, E. I.; Abu-Zaid, S. M.; Mahran, M. A.;
Badria, F. A.; Al-Obaid, A. M. J. Med. Chem. 2000, 43,
2915–2921.
41. Hollingshead, M. G.; Alley, M. C.; Camalier, R. F.;
Abbott, B. J.; Mayo, J. G., et al. Life Sci. 1995, 57, 131–
141.
18. Dimmock, J. R.; Vashishtha, S. C.; Quail, J. W.; Puga-
zhenthi, U.; Zimpel, Z.; Sudom, A. M.; Allen, T. M.; Kao,
42. Mohamadi, F.; Richards, N. G. J.; Guida, W. C.;
Liskamp, R.; Lipton, M.; Caufield, C.; Chang, G.;

B. K. Adams et al. / Bioorg. Med. Chem. 12 (2004) 3871–3883
3883
Hendrickson, T.; Still, W. C. J. Comput. Chem. 1990, 11,
440–467.
47. Zbinden, P.; Dobler, M.; Folkers, G.; Vedani, A. Quant.
Struct.-Act. Relat. 1998, 17, 122–130.
43. Halgren, T. A. J. Comp. Chem. 1996, 17, 490–641.
44. Chang, G. G.; Guida, W. C.; Still, W. C. J. Am. Chem.
Soc. 1989, 111, 4379–4386.
45. Stewart, J. P. P. J. Comp.-Aided Mol. Des. 1990, 4, 1–5.
46. Lemmen, C.; Lengauer, T. J. Comp.-Aided Mol. Des. 1997,
11, 357–368.
48. (a) Paull, K. D.; Shoemaker, R. H.; Hodes, L.; Monks, A.;
Scudiero, D. A.; Rubenstein, L.; Plowman, J.; Boyd, M.
R. J. Natl. Cancer Inst. 1989, 81, 1088–1092; (b) Devel-
opment Therapeutics Program NCI/NIH. COMPARE
Methodology.http://dtp.nci.nih.gov/docs/compare/com-
pare_methodology.html.