N-Hydroxy-N′-aminoguanidines as anti-cancer lead molecule: QSAR, synthesis and biological evaluation

Article abstract of DOI:10.1016/j.bmcl.2011.04.009

The intrinsic pharmacophore model (rpred2 and rm2 of 0.858 and 0.725) has been developed and used as a query to screen in-house built library based on N-hydroxy-N'-aminoguanidine (HAG) analogs. The pharmacophoric modeled based HITs were synthesized and ev

Full text of DOI:10.1016/j.bmcl.2011.04.009

Bioorganic & Medicinal Chemistry Letters 21 (2011) 3324–3328
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
N-Hydroxy-N0-aminoguanidines as anti-cancer lead molecule: QSAR, synthesis
and biological evaluation
Arijit Basu a, , Barij N. Sinha a, Philipp Saiko b, Geraldine Graser b, Thomas Szekeres b
⇑
a Department of Pharmaceutical Sciences, Birla Institute of Technology, Mesra, Ranchi 835215, India
b Department of Medical and Chemical Laboratory Diagnostics, Medical University of Vienna, Waehringer Guertel 18-20, A-1090 Vienna, Austria
a r t i c l e i n f o
a b s t r a c t
The intrinsic pharmacophore model (r2pred and r2m of 0.858 and 0.725) has been developed and used as a
query to screen in-house built library based on N-hydroxy-N’-aminoguanidine (HAG) analogs. The phar-
macophoric modeled based HITs were synthesized and evaluated for anticancer activity and cytotoxicity.
Article history:
Received 6 February 2011
Revised 17 March 2011
Accepted 4 April 2011
Available online 9 April 2011
One of the compounds (15) appeared as promising lead candidate with an IC50 value of 11
HL-60 promyelocytic leukemia cells. Compound 15 reveals significantly lower cytotoxicity against HeLa
and Vero cell with CC50 values of more than 100 M.
lM yielded in
l
Keywords:
QSAR
Ó 2011 Elsevier Ltd. All rights reserved.
Pharmacophore modeling
N-Hydroxy-N0-aminoguanidines
N-Hydroxysemicarbazides
Ribonucleotide Reductase
Hydroxyaminoguanidines (HAGs),1–3 hydroxamic acids,4
hydroxysemicarbazides (HSs),5 and amidoximes6 were extensively
explored as anticancer agents. These compounds bear similar phar-
macophoric features (-C (@X) NHOH; X = O, NH), which was iden-
tified as the basic pharmacophore for anti-tumor activity.7–9
Amongst these compounds, HAGs and HSs were explored for over
three decades primarily against L1210 murine leukemia.2,3,5 They
were found to inhibit DNA synthesis as a consequence of specifi-
cally inhibiting the R2 subunit of the enzyme Ribonucleotide
Reductase (RR).10 RR is one of the most widely accepted anticancer
targets11 and considered for further exploration through computer
aided design. The current work is therefore initiated primarily with
the objective of building a ligand based model by exploring various
pharmacophore modeling options. The method employed here
consists of 3D Pharmacophore space modeling (PHASE, version
3.0, Schrödinger, LLC, New York, NY, 2008).
In the current study, 3D-QSAR strategies were explored to de-
velop a robust QSAR model, which reveals significant external pre-
dictability. Here, we present lead design, identification, synthesis
and anti-cancer evaluation based on developed QSAR model.
Structural datasets2,3,5 (Table 1) were chosen and divided into
training (75%) and test (25%) set. The compounds were clustered
as per our earlier methodology12,13 according to increasing order
of biological activity. Sequentially, every fourth molecule was cho-
sen as a test set molecule resulting to a 25% sorted test set.
Pharmacophoric sites were created using the default settings in
PHASE (Schrodinger, LLC)14 with hydrogen bond acceptor (A),
hydrogen bond donor (D), hydrophobic (H), positive (P) and ring
aromatic (R). A few four point hypotheses were identified, consist-
ing of acceptor (A), one donor (D), and one ring aromatic (R) in dif-
ferent combinations (AAAR, ADDR and AADR). The AADR
combination survived during the scoring process, their survival
scores were in the range of 2.6–3.2 (Table 2a).This combination
yielded six hypotheses. The best hypothesis (1) consists of four
pharmacophoric features, with two acceptor (A1 and A2), one do-
nor (D), and one ring aromatic (R) (Fig. 1 a and b). The pharmaco-
phoric distances are shown in Table 2b. The training set comprises
two types of analogs, HSs and HAGs. The average fitness for HS and
HAGs were 2.74 and 2.65, respectively. The most active compound
T36 (HAG analog) showed superior fitness of 2.84.
We used solely statistical metrics for model validation due to
the unavailability of any co-crystal structure depicting the active
site of RR. Further, it precluded the use of receptor based informa-
tion and overlay study during pharmacophore-hypothesis genera-
tion for validation. Thus, a satisfactory hypothesis has been used
for the development of pharmacophore based QSAR model and val-
idated through various statistical metrics. The QSAR models were
subjected for external validation employing various statistical met-
rics,12,15,16 which provided significant model (r2 = 0.989, F = 220.14,
q2 = 0.889 r2Pred ¼ 0:858, rm2 ¼ 0:725, cR2p ¼ 0:742, Table 2a). The
⇑
Corresponding author.
E-mail address: arijit4uin@gmail.com (A. Basu).
0960-894X/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2011.04.009

A. Basu et al. / Bioorg. Med. Chem. Lett. 21 (2011) 3324–3328
3325
Table 1
Structures of derivatives of N-hydroxy-N0-aminoguanidine and N-hydroxysemicarbazide used for QSAR model building
Code
R1
R2
R3
R4
X
Y
PI50
T1
-OH
-OH
-OH
-OH
-H
-OH
-H
-H
-H
-H
-OCH3
-OH
-H
-H
-OH
-H
-H
-H
-H
-H
-H
-H
-H
-OH
-H
-H
-OCH3
-H
-H
-H
-H
-CH3
-CH3
-CH3
-H
-Cl
-H
-Br
-OCH3
I
-H
-H
-H
-H
-H
-H
-CN
-N(CH3)2
-H
-H
-H
OCH3
-OCH2Ph
-Ph
-OH
-H
-H
-OCH3
-H
-H
-OSO2CH3
-H
-CN
-H
-Cl
-OCH2C6H5
-H
-CF3
-O(CH2)5CH3
-O(CH2)3CH3
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
-Cl
Br
Br
Br
-H
I
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
NH
O
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
5.187
4.996
5.143
4.424
4.095
5.328
3.681
3.350
3.399
3.433
4.521
4.220
4.318
4.370
3.883
4.902
4.403
3.499
4.821
4.796
4.830
5.119
5.027
4.783
4.896
4.597
5.149
5.036
5.108
5.069
4.962
4.376
3.050
4.479
4.467
5.569
5.357
5.268
4.550
4.492
4.975
3.749
4.636
3.217
5.481
5.167
4.315
4.039
3.322
4.673
3.900
T2
T3*
T4
T5
T6
T7
T8
T9
I
-H
-H
-NO2
-OCH3
-H
-H
-H
-H
-H
I
-CF3
-H
-CF3
-OCH3
-H
NO2
-
-Br
-NO2
-H
-H
-H
-H
-H
—
-H
-H
-H
-H
-OCH3
-H
-H
-H
-H
-H
-H
-H
-H
-H
-H
-H
-H
-Br
-H
-H
-OCH3
-H
-H
-H
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
T10
T11
T12
T13*
T14
T15
T16
T17
T18*
T19
T20
T21⁄
T22
T23
T24
T25
T26
T27
T28
T29*
T30
T31
T32
T33
T34
T35*
T36
T37*
T38
T39
T40*
T41*
T42
T43
T44
T45*
T46
T47
T48*
T49*
T50*
T51
O
NH
NH
NH
NH
NH
NH
NH
NH
NH
NH
NH
NH
NH
NH
O
—
—
—
—
—
—
—
—
—
—
—
S
—
—
—
—
—
—
—
S
-H
NCH3
—
—
—
—
-CH3
-Cl
-Cl
-H
—
—
—
—
—
—
-Cl
-C(CH3)2
—
—
—
—
—
—
—
—
—
—
—
—
—
—
NH
O
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
*
Test set molecules, PI50 = ꢀlog10IC50
QSAR model was shown in Figure 2 which demonstrates favorable
(blue) and unfavorable (red) regions.
of the developed model. The scatter plot for the test set (Fig. 3b)
indicates a reasonably good correlation (r2Pred ¼ 0:858) between
the predicted and experimental activities. Further, it reveals the
way test and training set compounds were chosen. Both sets were
constituted with compounds covering a wide range of actives,
resulting in a classic ‘cigar’ shaped plot rather than a ‘dumbbell’
plot.
One of the significant observations while visualizing this QSAR
model was the large clustering of the blue boxes near the ring
aromatic (R) feature, suggesting larger rings/multiple rings are
favorable for biological activity. The scatter plot of predicted
PI50 versus observed PI50 (Fig. 3a and b) confirms the superiority

3326
A. Basu et al. / Bioorg. Med. Chem. Lett. 21 (2011) 3324–3328
Table 2a
Pharmacophoric hypothesis developed for the particular data seta
Hyp ID
Survival
Site
Volume
Selectivity
1b
2
3
4
5
3.178
3.078
3.075
2.998
2.571
2.566
0.72
0.67
0.68
0.59
0.51
0.37
0.651
0.623
0.612
0.606
0.38
0.984
0.99
0.991
0.991
0.988
1.019
6
0.51
a
No of ligand matched = 39
QSAR analysis of HYP ID
Figure 2. QSAR model visualized in the context of the most active compound T36.
Blue regions show favorable regions, red disfavored regions.
b
1
resulted in the following statistics r
2 = 0.989,
F = 220.14, q2 = 0.889 rp2red ¼ 0:858, rm2 ¼ 0:725, cRp2 ¼ 0:742.
Supplementary data. The synthesized compounds (3-15) were
screened against the HL-60 human promyelocytic leukemia cell
line19,20 and also screened for their cytotoxic potential against
HeLa and Vero cell cultures. Five compounds (6, 10, 11, 14 and
15) were found to be active against HL-60 human promyelocytic
Table 2b
Inter-pharmacophoric distances for hypothesis 1
Site1
Site2
Distance
A1
A1
A1
A2
A2
D
A2
D
R
D
R
4.711
3.892
3.753
2.023
8.463
7.475
leukemia cells yielding IC50 values between 11–95
Compound 15 showed promise with an IC50 value of 11
relatively high cytotoxic concentration (CC50) of >100 M (Supple-
lM (Table 3).
lM and
l
mentary Table). The pharmacophoric fitness (Fig. 1a) of compound
15 was 2.26. This was overall satisfactory and comparable to com-
pounds with similar activity.
R
The experimental and predicted values are in congruence (Table
3). Moreover, identification of five true negatives (3, 4, 8, 9 and 13)
by the developed QSAR model is trulyoutstanding. The fitness score
for these ligands were less than two; the fitness poses were shown
in Fig. 1d.
Therefore, the obtained predictive QSAR model can be utilized
as a query tool for 3D databases screening to identify possible lead
as anticancer agent. An in-house library was constructed with the
synthetically feasible fragments using SMILIB.17The constructed li-
brary was then predicted against the developed model. Eight mol-
ecules (1, 2, 5, 6, 7, 10, 11, 12, 14, and 15) were identified as
significant HITs for synthesis. However, five non-significant mole-
cules (3, 4, 8, 9 and 13) were also chosen for synthesis to evaluate
the true negatives.
These observations further consolidate the model, enabling it to
be further explored for screening additional databases in future.
Compound 15 with an IC50 of 11
lM and relatively high cytotoxic
concentration >100 M have been identified as the lead molecule.
l
The study was focused on exploring possible ligand based mod-
eling options to develop robust QSAR model(s) for the particular
kind of analogs. These models were validated computationally as
well as experimentally. The resulting robust and reproducible
Synthesis18of compounds (3-15) were outlined in Scheme 1.
Structures of the synthesized compounds are given in Table 3
physicochemical properties and spectral details are given in
Figure 1. (a) Pharmacophore maps of the generated model. The distance between the various pharmacophoric features are given. Hydrogen bond acceptor (A), hydrogen
bond donor (D), hydrophobic (H), positive (P) and ring aromatic (R). (b) Most active molecule T36 is embedded in the map. (c) Newly synthesized compound 15 is embedded
in the map. (d) Fitness of five true negatives on the developed hypothesis.

A. Basu et al. / Bioorg. Med. Chem. Lett. 21 (2011) 3324–3328
3327
6
5
4
3
a
6
b
5
4
3
4
5
6
3
4
5
6
observed
observed
Figure 3. Red line indicates the ‘best fit’ amongst the predicted and experimental PI50. (a) Training set compounds, (b) test set compounds.
H
NH
NHOH
R
S
N
NH
i.
iii.
N
H2N
H2N
N
N
NH2
NHOH
ii.
H
H
3-15
2
1
Scheme 1. Reagents and conditions: (i). CH3I/MeOH, 12 h, rt, (ii) NH2OH, MeOH, rt, 4 h (iii) Ar-CHO, MeOH, reflux 3 h.
QSAR model can be used as a query tool for future drug design for
this particular class of compounds. We have also identified a lead
molecule which is not only outstanding, but can also be a good
Due to these promising results, it deserves further preclinical and
in vivo testing and might be good candidate molecule for further
lead optimization.
starting point with an IC50 of 11 lM, CC50 > 100 lM i.e. a selectivity
index of more than 10. Detailed biochemical studies for compound
15 were recently reported by our group.21It proved to be a potent
inhibitor of RR and arabinofuranosylcytosine (Ara-C) synergist.
Acknowledgments
This work was supported by University Grants Commission, In-
dia. We are thankful to Dr. Ashoke Sharon, Dr. Venkatesan J. and
Nibha Mishra who has helped to improve the structure and lan-
guage of our manuscript. We thank to Dr. Kunal Roy for helping
us out with cR2p determination. Vlife Sciences, for providing the trial
version of MDS and Schrodinger, LLC for helping in extraction of
QSAR descriptors.
Table 3
Structure, biological activity (growth inhibition of HL-60 human promyelocytic
leukemia cells) and cytotoxicity of the synthesized compounds.
R3
NH
NH
NH
C
N
C
N
C
R2
C
H
N
H
NHOH
O
N
C
H
N
H
NHOH
Supplementary data
C
H
N
H
NHOH
R1
3, 5-14
4
15
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmcl.2011.04.009.
Code R1
R2
R3
HL-60 IC50
M)
CC50
(lM)
(
l
References and notes
Pred
-N(CH3)2 285
Exp
HeLa Cells VERO Cells
3
4
b
6
7
8
9
10
11
12
13
14
15
-H
—
-H
-OH
-H
—
-H
-H
>100
>100
>100
95
>100
>100
>100
67
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
1. Koneru, P. B.; Lien, E. J.; Avramis, V. I. Pharm. Res. 1993, 10, 515.
2. Tai, A. W.; Lien, E. J.; Lai, M. M.; Khwaja, T. A. J. Med. Chem. 1984, 27, 236.
3. T’Ang, A.; Lien, E. J.; Lai, M. M. J. Med. Chem. 1985, 28, 1103.
4. van’t Riet, B.; Wampler, G. L.; Elford, H. L. J. Med. Chem. 1979, 22, 589.
5. Ren, S.; Wang, R.; Komatsu, K.; Bonaz-Krause, P.; Zyrianov, Y.; McKenna, C. E.;
Csipke, C.; Tokes, Z. A.; Lien, E. J. J. Med. Chem. 2002, 45, 410.
6. Fylaktakidou, K. C.; Hadjipavlou-Litina, D. J.; Litinas, K. E.; Varella, E. A.;
Nicolaides, D. N. Curr. Pharm. Des. 2008, 14, 1001.
7. Nocentini, G. Crit. Rev. Oncol. Hematol. 1996, 22, 89.
8. Tsimberidou, A. M.; Alvarado, Y.; Giles, F. J. Expert. Rev. Anticancer Ther. 2002, 2,
437.
9. Cerqueira, N. M.; Pereira, S.; Fernandes, P. A.; Ramos, M. J. Curr. Med. Chem.
2005, 12, 1283.
10. Shao, J.; Zhou, B.; Zhu, L.; Bilio, A. J.; Su, L.; Yuan, Y. C.; Ren, S.; Lien, E. J.; Shih, J.;
Yen, Y. Biochem. Pharmacol. 2005, 69, 627.
—
-OH
-H
156
54
78
-OCH3 -H
-H
-OCH3
-H
-CH3
-H
230
246
156
72
-H
-CH3
-H
-Cl
-H
-H
-H
—
-H
-H
-H
-H
-H
98
60
-Cl
154
254
56
>100
>100
62
-OH -OCH3
-H
—
-H
—
9
11
IC50 = 50% growth inhibition of HL-60 cells, CC50 = 50% growth inhibition of HeLa
and VERO cell line.
11. Shao, J.; Zhou, B.; Chu, B.; Yen, Y. Curr. Cancer Drug Targets 2006, 6, 409.

3328
A. Basu et al. / Bioorg. Med. Chem. Lett. 21 (2011) 3324–3328
12. Basu, A.; Jasu, K.; Jayaprakash, V.; Mishra, N.; Ojha, P.; Bhattacharya, S. Eur. J.
Med. Chem. 2009, 44, 2400.
(500 mL) was stirred at rt for 12 h to form S-methylisothiosemicarbazide. The
base was liberated using equimolar amount of sodium bicarbonate. The
precipitate was filtered and washed with cold water and finally with cold
methanol: Yield 90%, mp 108–110 °C. C2H6N2S2, C, 19.71%; H, 4.95%; N, 23.02%;
S, 52.59%. The resulting S-methylisothiosemicarbazide (0.4 mol) was reacted
with hydroxylamine (0.5 mol) solution (A cold solution of potassium
13. Dewangan, P.; Verma, S.; Basu, A. Lett. Drug Des. Discovery 2008, 5, 494.
14. PHASE procedure. Six built-in types of pharmacophoric features: hydrogen bond
acceptor (A), hydrogen bond donor (D), hydrophobic (H), negative ionizable
(N), positive ionizable (P), and aromatic ring (R). Ligands were processed with
the LigPrep program to assign protonation states appropriate for pH 7.0.
Conformer generation was carried out with the MacroModel. Potentials were
computed using the OPLS2005 force field. The default pharmacophore feature
definitions were used in site generation. After the identification of the sites
hypotheses were generated by a systematic variation of the number of sites.
The number of matching active compounds was kept default of 39 (entire
training set). All the molecules were considered as active no active/inactive
threshold were assigned. The scoring was done using: Survival score
hydroxide (0.5 mol) in 250 mL of methanol was added to
a cold
hydroxyamine hydrochloride) at room temperature and stirred for 4 h. The
solution was concentrated under reduced pressure on cooling crude N-
Hydroxy-N0-amino guanidine separates out, which is then re-crystallized
from ethanol. Yield 70%, mp 119–121 °C. CH6N4O, C, 13.43%; H, 6.68%; N,
62.06%.
(b)General method for synthesis of compounds (3–15). N-hydroxy-N0-amino
guanidine (0.5 mol) and different aromatic aldehydes (0.5 mol) were refluxed
for 3 h to obtain different compounds 3–15. The resulting compounds were re-
crystallized from 95% ethanol. Compound 15: Yield 78%, mp 160 °C, FAB-MS
278(M+), 279 (M+1), 1H NMR (DMSO, 400 MHz) d 7.5–8.6 (m, 5H, C14H9), d
8.67(s, 1H, CH), d 9.25(s, 1H, NH), d 10.52s, 1H, NH), d 11.32 (s, 1H, OH).
19. Saiko, P.; Ozsvar-Kozma, M.; Bernhaus, A.; Jaschke, M.; Graser, G.; Lackner, A.;
Grusch, M.; Horvath, Z.; Madlener, S.; Krupitza, G.; Handler, N.; Erker, T.;
Jaeger, W.; Fritzer-Szekeres, M.; Szekeres, T. Int. J. Oncol. 2007, 31, 1261.
20. Madlener, S.; Illmer, C.; Horvath, Z.; Saiko, P.; Losert, A.; Herbacek, I.; Grusch,
M.; Elford, H. L.; Krupitza, G.; Bernhaus, A.; Fritzer-Szekeres, M.; Szekeres, T.
Cancer Lett. 2007, 245, 156.
formula = (Vector
score) + (Site
score) + (Volume
score) + (Selectivity
score) + (Number of matches-1)ꢀ(Reference ligand relative conformational
energy) + (Ref ligand activity). Max intersite distance=2.00Å, initial and final
box size is 20 Å and 1 Å, respectively. The best hypothesis was chosen based on
the survival score by PHASE. The top scored hypothesis was then used to build
pharmacophore-based 3D QSAR models with
spacing = 1 Å.
3 PLS factors and Grid
15. Roy, P.; Roy, K. Q. S. A. R. Comb. Sci. 2008, 27, 302.
16. Mitra, I.; Saha, A.; Roy, K. Molecular Simulation 2010, 36, 1067.
17. Schüller, A.; Schneider, G.; Byvatov, E. Q. S. A. R. Comb. Sci. 2003, 22, 719.
18. (a) General method for synthesis of N-Hydroxy-N‘-amino guanidine
(2).Thiosemicarbazide (0.5 mol) and Methyl iodide (0.5 mol) in methanol
21. Saiko, P.; Graser, G.; Giessrigl, B.; Lackner, A.; Grusch, M.; Krupitza, G.; Basu, A.;
Sinha, B.; Jayaprakash, V.; Jaeger, W. Biochem. Pharmacol. 2011, 81, 50.