Identification and synthesis of N′-(2-oxoindolin-3-ylidene)hydrazide derivatives against c-Met kinase

Article abstract of DOI:10.1016/j.bmcl.2011.04.064

A series of N′-(2-oxoindolin-3-ylidene)hydrazide derivatives were identified as moderately potent inhibitors against c-Met kinase by pharmacophore-based virtual screening and chemical synthesis methods. The structure-activity relationship (SAR) at various

Full text of DOI:10.1016/j.bmcl.2011.04.064

Bioorganic & Medicinal Chemistry Letters 21 (2011) 3749–3754
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Identification and synthesis of N0-(2-oxoindolin-3-ylidene)hydrazide
derivatives against c-Met kinase
Zhongjie Liang a, , Dengyou Zhang a, , Jing Ai b, , Limin Chen a, Hengshuai Wang a, Xiangqian Kong a,
a,c,
b,
Mingyue Zheng a, Hong Liu a, , Cheng Luo
, Meiyu Geng , Hualiang Jiang a,d, Kaixian Chen a
⇑
⇑
⇑
a Drug Discovery and Design Center, State Key Laboratory of Drug Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai 201203, China
b Division of Anti-Tumor Pharmacology, State Key Laboratory of Drug Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai 201203, China
c Center for Systems Biology, Soochow University, Suzhou 215006, China
d School of Pharmacy, East China University of Science and Technology, Shanghai, China
a r t i c l e i n f o
a b s t r a c t
A series of N0-(2-oxoindolin-3-ylidene)hydrazide derivatives were identified as moderately potent inhib-
itors against c-Met kinase by pharmacophore-based virtual screening and chemical synthesis methods.
The structure–activity relationship (SAR) at various positions of the scaffold was investigated and its
binding mode with c-Met kinase was analyzed by molecular modeling studies. In this study, two potent
Article history:
Received 27 February 2011
Revised 9 April 2011
Accepted 13 April 2011
Available online 22 April 2011
compounds D2 and D25, with IC50 value at 1.3 lM and 2.2 lM against c-Met kinase respectively, were
identified. Finally, based on the clues extracted from this study, future development for the optimization
of this scaffold was discussed.
Keywords:
c-Met kinase
N0-(2-Oxoindolin-3-ylidene)hydrazide
Pharmacophore-based virtual screening
Ó 2011 Elsevier Ltd. All rights reserved.
MET (mesenchymal–epithelial transition factor) is a proto-
oncogene that encodes tyrosine kinase receptor c-Met kinase, also
known as hepatocyte growth factor receptor (HGFR). c-Met kinase,
together with its ligand hepatocyte growth factor (HGF), is in-
volved in cell proliferation, migration and invasion as well as is
essential for normal embryonic development.1 c-Met kinase can
be activated via a variety of mechanisms, including ligand-depen-
dent activation (autocrine or paracrine), gene amplification, gene
mutation and cross-talk with other receptors (heterodimeriza-
tion).2 The activation gives rise to receptor dimerization and
recruitment of several SH2 domain-containing signal transducers
that activate a number of pathways including the Ras–MAPK,
PI3k-Akt-mTOR cascades and so on.3 When deregulated, the HGF/
c-Met kinase pathway has been implicated in tumorigenesis and
metastasis,4 which implies the kinase as a potential target for can-
cer treatment. Given a growing number of drug candidates inves-
tigated in the clinical phase, c-Met kinase has risen as a kinase
target of priority.5
several potential therapeutic strategies have been tried by target-
ing HGF/c-Met kinase pathway, including HGF antagonists, anti-
bodies against HGF or c-Met kinase, and small molecular
inhibitors against phosphorylation of the tyrosine residues in the
tyrosine kinase domain. Among these strategies, inhibition of the
tyrosine kinase activity mediated by an ATP-competitive small
molecule (type I inhibitor) or non-ATP-competitive small molecule
(type II inhibitor) is a pharmacologically attractive method, which
has been demonstrated for other tyrosine kinases.6
To date, numerous efforts have been made toward the develop-
ment of small molecular inhibitors against c-Met tyrosine kinase.7–
13
However, due to poor pharmacokinetic properties, quite a few
small molecular inhibitors have been limited to in vitro assay or
brief in vivo studies, and were not applicable for the clinical pur-
poses. To identify more potent and novel scaffolds against c-Met
kinase, in this study, pharmacophore-based virtual screening and
molecular docking methods were performed to discover new c-
Met kinase inhibitor leads. The procedure is shown in Figure 1,
firstly, 24 reported compounds with a wide range of bioactivities
and structural diversity were chosen to construct selective phar-
macophore models.7–13 Then, the 3D-QSAR pharmacophore gener-
ation implemented in Accelrys Discovery Sudio2.1 (Accelrys Inc.,
San Diego, CA) was performed to set up pharmacophore models.
After the clustering and validation of these models, 4 rational
models were extracted to screen the SPECS database. The SPECS
database containing 190,000 chemicals (http://www.specs.net)
was firstly filtered by the druglikeness and then adopted for
As HGF/c-Met kinase signaling plays a role of significance in
tumorigenesis and metastasis, several different strategies have
been explored to inhibit c-Met kinase, each focusing on one of
the serial steps that regulate c-Met kinase activation.3 Up to now,
⇑
Corresponding authors. Tel.: +86 21 50806600; fax: +86 21 50807088 (C.L.).
E-mail addresses: hliu@mail.shcnc.ac.cn (H. Liu), cluo@mail.shcnc.ac.cn (C. Luo),
mygeng@mail.shcnc.ac.cn (M. Geng).
These authors contributed equally to this study.
0960-894X/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2011.04.064

3750
Z. Liang et al. / Bioorg. Med. Chem. Lett. 21 (2011) 3749–3754
Figure 1. Flow chart of virtual and experimental screening strategy combined with chemical synthesis method for discovering c-Met kinase inhibitors.
Table 1
Molecular structures of 5 hits that show inhibitory activity against c-Met kinase (The compounds named with D⁄ were purchased from the SPECS database, while the compounds
named with S⁄ were synthesized in this study.)
Code
Compound
Inhibition (%)
IC50 (lM)
N
O
O
O
D1
51.3
10.3
N
N
N
HN
N
H
O
H
N
O
D2
D3
56.4
55.1
1.3
3.7
N
O
HN
N
O
H
O
O
O
O
N
O
O
O
Cl
O
N
O
D4
61.4
3.3
N
O
Cl
N
N
O
N
N
O
N NH
O
H
N
H
N
O
O
O
N
O
N
O
D5
54.5
0.78
O
O
O
N
O
N
O

Z. Liang et al. / Bioorg. Med. Chem. Lett. 21 (2011) 3749–3754
3751
Table 2
pharmacophore-based virtual screening. During the screening pro-
c-Met kinase inhibition activity for compounds D6–D11 of general structure
cess, the BEST method was used, and the maximum number of
conformers generated was set to 250. The pharmacophore models
were employed one by one with the flexible search method to
screen the database with Discovery Studio2.1. The molecular dock-
ing method via software DOCK4.014 and AUTODOCK3.0515 was
employed to evaluate the top 5000 small molecules from the phar-
macophore-based database searching method. The active and inac-
tive conformation of c-Met kinase crystal structure (PDB entries:
2RFS10 and 3C1X8) were retrieved from the Protein Data Bank
(http://www.rcsb.org/pdb) as the receptor models. Furthermore,
3000 compounds with the highest score by DOCK4.014 were scored
by AUTODOCK3.0515 program. Among the top 1000 molecules
scored by AUTODOCK, the conformation with the highest score of
each molecule was evaluated by CScore method implemented in
Sybyl software package for predicting the binding score of the gi-
ven protein-ligand complex structures. Taking account to the score
and scaffold diversity classified by Pipeline program, 124 com-
pounds among 50 scaffolds were purchased and evaluated for their
ability to inhibit enzymatic activity of the c-Met kinase by in vitro
assay which is described in the Supplementary data.
O
R3
NH
O
N
R1
N
R2
Code
Compound
R3
Inhibition (%)
IC50 (lM)
R1
H
R2
H
O
D6
D7
11.2
33.3
NA
HN
H
H
H
H
28.3
O2N
In vitro assay indicated that compounds D1 ꢀ D5 listed in Ta-
ble 1 stand out as the most potent c-Met kinase inhibitors. At the
concentration of 10
>50.0%. Their IC50 values were also determined, ranging from
0.78 M to 10.3 M. Among the five hit compounds shown in Ta-
lM, their inhibition against c-Met kinase are
D8
75.2
4.3
OH
l
l
ble 1, the compound D2 with isatin derivative scaffold stands out
for further study because of its good druglikeness and moderate
potency. Interestingly, we noticed that quite a few studies demon-
strated these isatin derivatives were to be potential kinase inhibi-
D9
H
H
H
H
H
H
6.3
29.4
12.8
NA
NA
NA
D10
D11
tors.16–18 Therefore, the compound D2 with IC50 value at 1.3
lM
against c-Met kinase, was chosen as the lead compound for further
studies. The sub-structure similarity searching method against
SPECS database and chemical synthesis were performed to obtain
more analogues for the exploring their SAR. The synthesis method
and spectrum data were applied in the Supplementary data. Final-
ly, 37 non-ATP competitive compounds with moderate inhibitory
activity against c-Met kinase were obtained in this study.
In the chemical synthesis procedure, we made our efforts to ap-
pend diverse substitutents on the various positions of the N0-(2-
oxoindolin-3-ylidene)hydrazide scaffold, influencing the c-Met ki-
nase activity. Initially, we screened a wide range of hydrazide
groups analogues listed in Table 2. We observed the inhibitory
activity apparently abolished when the m-methylbenzoamine ben-
analogues D20 and D21 inhibited c-Met kinase activity with an
IC50 value of 8.1 M and 15.8 M, respectively. Compound D22
and D23 weakly inhibited the activity of c-Met kinase.
l
l
Table 3
c-Met kinase inhibition activity for compounds S1–S6 of general structure
Code
Compound
Inhibition (%)
IC50 (lM)
zohydrazide (D2, IC50 = 1.3 lM) was changed to p-cyclopropane
carbonyl amine benzohydrazide (D6). And the p-nitrobenzohyd-
R1
R2
R3
razide analogue (D7) showed weak c-Met kinase activity with
IC50 28.3
displayed potent c-Met kinase activity with an IC50 value of
4.3
M. Alkyl hydrazides analogues (D9 ꢀ D11) are not beneficial
lM. The 3-hydroxyl-2-naphthohydrazide analogue D8
S1
S2
5-F
H
H
5.7
0
NA
NA
l
for c-Met kinase activity. Then, we turned our attention to those
substituents on the benzene ring of isatin. The in vitro assay data
indicated that the introduction of halogen in the 5-position was
not favorable for the inhibitor’s potency when the N-1-position
was not substituted (Table 3). Moreover, a pool of substituents in
the N-1-position were also attempted to identify more potent
inhibitors (Table 4). It was observed that when N-1-position
beared methyl group, p-nitro benzohydrazide analogue D12 re-
5-F
O
S3
S4
S5
S6
5-F
H
H
H
H
21.1
24.1
6.7
NA
NA
NA
NA
HO
5-Cl
5-Cl
5-F
vealed the inhibitory potency with an IC50 value of 12.3
lM,
HO
O
whereas m-nitro benzohydrazide analogue D13 abolished its
inhibitory activity. In addition, the halogen analogues D14 and
D15 totally abolished the inhibition against c-Met kinase. After
p-nitro benzohydrazide group was kept, the introduction of more
hydrophobic or hydrophilic substituents (D16, D17, D18 and
D19) in the N-1-position abolished their inhibitory activity. In
comparison of compound D17, it was very interesting that
13.3
F

3752
Z. Liang et al. / Bioorg. Med. Chem. Lett. 21 (2011) 3749–3754
To explore the SAR among these compounds, we further synthe-
with the hinge region of the c-Met kinase. The NH atom in the
backbone and the carbonyl of Met1160 H-bond with the oxygen
atom of carbonyl and N1H atom of D2, respectively. The central
phenyl ring of D2 which is flanked on one side by the gatekeeper
residue (Leu1157) and on the opposite side by Phe1223 (DFG mo-
sized a series of isatins containing N-morpholinomethyl group in
the N-1-position (Table 5). We found 5-Br and benzohydrazide
substituted analogue D24 displayed potent c-Met kinase inhibition
activity with an IC50 value of 3.5 lM. The corresponding analogue
D25, appending an p-methoxy group on the benzene, improved
activity slightly. The inhibitory activity of 5-F substituted ana-
logues S7, S8 and S11 are significantly reduced. It was also ob-
served that 5-Cl substituted analogues S9, S10 and S12 are weak
inhibitors against c-Met kinase. These results indicate that the less
negatively-charged halogen is beneficial for the potency of
inhibitor.
tif) involves in a p–p stacking interaction. The amide portion of D2
occupies the urea site and forms iterative H-bonds with c-Met ki-
nase. The amide and carbonyl of D2 H-bond with the backbone of
Asp1222 and the terminal of Lys1110, respectively. Finally, the ter-
minal benzyl ring mostly resides in a hydrophobic site and inter-
acts with a few aromatic residues including Ile1130, Phe1134
and Phe1200. To further investigate molecular basis of the deriva-
tives with N-morpholinomethyl substitent interacting with recep-
tor, the binding mode of the complex structure of D25 with c-Met
kinase was analyzed. In the hinge region, the carbonyls of isatin
core and hydrazide H-bond with the hydroxyl group of Tyr1159
and the NH atom in the backbone of Met1160. The bromine in
To investigate the molecular basis of N0-(2-oxoindolin-3-yli-
dene)hydrazide derivatives interacting with c-Met kinase, molecu-
lar docking approach was performed to analyze the complex
structures of compounds D2 and D25 with the kinase. As shown
in Figure 2, the isatin core engages in key H-bonding interactions
Table 4
c-Met kinase inhibition activity for compounds D12–D23 of general structure
Code
D12
D13
Compound
Inhibition (%)
57.8
IC50
12.3
NA
(lM)
R1
H
R2
R3
CH3
O2N
H
CH3
12.3
NO2
D14
D15
H
H
CH3
14.4
25.9
NA
NA
Cl
CH3
Br
D16
D17
H
H
CH3CH2
2.7
NA
NA
O2N
O2N
O
12.4
N
H
D18
D19
H
H
3.9
NA
NA
O2N
O2N
N
48.1
O
O
OH
D20
D21
H
H
55.5
58.6
8.1
N
H
Br
O
15.8
N
H
H2N
O
D22
D23
H
H
37.6
22.6
NA
NA
N
H
NO2
O
N
H

Z. Liang et al. / Bioorg. Med. Chem. Lett. 21 (2011) 3749–3754
3753
A
B
Table 5
c-Met kinase inhibition activity for compounds D24–S12 of general structure
Code
Compound
Inhibition (%) IC50 (lM)
R1
R2
R3
N
D24
D25
S7
5-Br
85.4
73.6
1.4
3.5
2.2
NA
NA
O
O
N
5-Br
5-F
O
Figure 3. Future directions for designing c-Met kinase inhibitors with greater
potency. (A) Base scaffold with three possible positions for modifications. (B) The
scaffold in the ATP-binding site of the c-Met kinase. Regarding the whole pocket,
the residues in 4 Å with the compound D2 are shown as surface. The hinge region,
benzene ring in the DFG motif, urea site and hydrophobic site are colored red, cyan,
yellow and blue respectively.
N
N
O
O
S8
5-F
0.0
O
O
N
Leu1140, Leu1157, Ala1221 and Phe1223. The N-morpholinometh-
yl group orients the solvent out of the binding pocket.
S9
5-Cl
48
NA
O
Based on the SAR and binding modes analysis aforementioned,
the modifications on N0-(2-oxoindolin-3-ylidene)hydrazide scaf-
fold were proposed to attempt to obtain more potent compounds.
As shown in Figure 3, the hydrazide portion seems to be a linker
without any interaction with c-Met kinase, which could be re-
placed with a few more drug-like groups. In the R3 position, the
terminal 3-methyl-benzene of compound D2 doesn’t enter the
hydrophobic site completely. Another amide group might be in-
serted before the 3-methyl-benzene in the urea site, which would
N
N
N
S10
S11
S12
5-Cl
5-F
28.7
26.6
39.9
NA
NA
NA
O
O
O
HO
HO
F
5-Cl
afford the H-bond with the Glu1127 in a-helix and make the ben-
zene deeper in the hydrophobic site. Meanwhile, different H-bond
donors, acceptors and hydrophobic substituents could be at-
tempted in the urea and hydrophobic sites. Considering the N-mor-
pholinomethyl group into the R2 position of the isatin derivatives,
compound D23 seems to be a good lead compound for future
inhibitor optimization. Meanwhile, the clues extracted from SAR
the isatin probably matches the space of the binding pocket quite
well compare with other halogens. The methoxy-benzene locates
in the hydrophobic pocket surrounded by Val1092, Ala1108,
Figure 2. Binding mode analysis of compounds D2 and D25 with c-Met kinase at inactive conformation. The N lobe and C lobe of the c-Met kinase domain are red and blue,
respectively. Compounds D2 and D25 are represented in green and orange sticks, respectively. Meanwhile, key residues in the pocket forming interactions with compounds
are represented in gray sticks on the right.

3754
Z. Liang et al. / Bioorg. Med. Chem. Lett. 21 (2011) 3749–3754
of halogen analogues in R1 position indicate that bromine is fa-
vored here. With the N-morpholinomethyl group into the nitrogen
of the isatin kept in the hinge region, different H-bond donors,
acceptors and hydrophobic substituents in the R3 region could be
beneficial for the inhibitor’s potency as aforementioned. Taken to-
gether, our study demonstrates that an efficient and cost-effective
virtual screening procedure can be used to identify a series of N0-
(2-oxoindolin-3-ylidene)hydrazide derivatives against c-Met ki-
nase. These studies gave rise to the discovery of potent compound
2. Dharmawardana, P. G.; Giubellino, A.; Bottaro, D. P. Curr. Mol. Med. 2004, 4,
855.
3. Comoglio, P. M.; Giordano, S.; Trusolino, L. Nat. Rev. Drug Discov. 2008, 7, 504.
4. Birchmeier, C.; Birchmeier, W.; Gherardi, E.; Vande Woude, G. F. Nat. Rev. Mol.
Cell Biol. 2003, 4, 915.
5. Underiner, T. L.; Herbertz, T.; Miknyoczki, S. J. Anticancer Agents Med. Chem.
2010, 10, 7.
6. Baselga, J. Science 2006, 312, 1175.
7. Cai, Z. W.; Wei, D.; Schroeder, G. M.; Cornelius, L. A.; Kim, K.; Chen, X. T.;
Schmidt, R. J.; Williams, D. K.; Tokarski, J. S.; An, Y.; Sack, J. S.; Manne, V.;
Kamath, A.; Zhang, Y.; Marathe, P.; Hunt, J. T.; Lombardo, L. J.; Fargnoli, J.;
Borzilleri, R. M. Bioorg. Med. Chem. Lett. 2008, 18, 3224.
D2 and D25, with the value of IC50 at 1.3 lM and 2.2 lM, respec-
8. Schroeder, G. M.; Chen, X. T.; Williams, D. K.; Nirschl, D. S.; Cai, Z. W.; Wei, D.;
Tokarski, J. S.; An, Y.; Sack, J.; Chen, Z.; Huynh, T.; Vaccaro, W.; Poss, M.;
Wautlet, B.; Gullo-Brown, J.; Kellar, K.; Manne, V.; Hunt, J. T.; Wong, T. W.;
Lombardo, L. J.; Fargnoli, J.; Borzilleri, R. M. Bioorg. Med. Chem. Lett. 2008, 18,
1945.
9. Porter, J.; Lumb, S.; Lecomte, F.; Reuberson, J.; Foley, A.; Calmiano, M.; le Riche,
K.; Edwards, H.; Delgado, J.; Franklin, R. J.; Gascon-Simorte, J. M.; Maloney, A.;
Meier, C.; Batchelor, M. Bioorg. Med. Chem. Lett. 2009, 19, 397.
10. Bellon, S. F.; Kaplan-Lefko, P.; Yang, Y.; Zhang, Y.; Moriguchi, J.; Rex, K.;
Johnson, C. W.; Rose, P. E.; Long, A. M.; O’Connor, A. B.; Gu, Y.; Coxon, A.; Kim, T.
S.; Tasker, A.; Burgess, T. L.; Dussault, I. J. Biol. Chem. 2008, 283, 2675.
11. Albrecht, B. K.; Harmange, J. C.; Bauer, D.; Berry, L.; Bode, C.; Boezio, A. A.; Chen,
A.; Choquette, D.; Dussault, I.; Fridrich, C.; Hirai, S.; Hoffman, D.; Larrow, J. F.;
Kaplan-Lefko, P.; Lin, J.; Lohman, J.; Long, A. M.; Moriguchi, J.; O’Connor, A.;
Potashman, M. H.; Reese, M.; Rex, K.; Siegmund, A.; Shah, K.; Shimanovich, R.;
Springer, S. K.; Teffera, Y.; Yang, Y.; Zhang, Y.; Bellon, S. F. J. Med. Chem. 2008,
51, 2879.
12. D’Angelo, N. D.; Bellon, S. F.; Booker, S. K.; Cheng, Y.; Coxon, A.; Dominguez, C.;
Fellows, I.; Hoffman, D.; Hungate, R.; Kaplan-Lefko, P.; Lee, M. R.; Li, C.; Liu, L.;
Rainbeau, E.; Reider, P. J.; Rex, K.; Siegmund, A.; Sun, Y.; Tasker, A. S.; Xi, N.; Xu,
S.; Yang, Y.; Zhang, Y.; Burgess, T. L.; Dussault, I.; Kim, T. S. J. Med. Chem. 2008,
51, 5766.
13. Kim, K. S.; Zhang, L.; Schmidt, R.; Cai, Z. W.; Wei, D.; Williams, D. K.; Lombardo,
L. J.; Trainor, G. L.; Xie, D.; Zhang, Y.; An, Y.; Sack, J. S.; Tokarski, J. S.; Darienzo,
C.; Kamath, A.; Marathe, P.; Lippy, J.; Jeyaseelan, R., Sr.; Wautlet, B.; Henley, B.;
Gullo-Brown, J.; Manne, V.; Hunt, J. T.; Fargnoli, J.; Borzilleri, R. M. J. Med. Chem.
2008, 51, 5330.
tively. Finally, the potentially helpful clues obtained in the in silico
and chemical synthesis study on the isatin derivative scaffold will
light up the road upon the future development of c-Met kinase
inhibitors for the using as therapeutic agents against HGF/c-Met ki-
nase signaling related tumorigenesis and metastasis.
Acknowledgments
We gratefully thank Dr. Manqing Hou from National Institutes of
Health for her helpful comments. We also gratefully acknowledge
the financial support from the National Natural Science Foundation
of China (20972174, 81025017, 30725046, 91029704 and
21021063), the State Key Program of Basic Research of China grant
(2009CB918502), Shanghai Committee of Science and Technology
(10410703900), the Chinese Academy of Sciences (XDA01040305),
National Science & Technology Major Project on (2009ZX09301-
001). Guangdong S&T Dept. (2010A030100006) and Program of
Shanghai Subject Chief Scientist (10XD1405100).
Supplementary data
14. Ewing, T.; Kuntz, I. J. Comput. Chem. 1998, 18, 1175.
15. Buzko, O. V.; Bishop, A. C.; Shokat, K. M. J. Comput. Aided Mol. Des. 2002, 16, 113.
16. Chu, W.; Zhang, J.; Zeng, C.; Rothfuss, J.; Tu, Z.; Chu, Y.; Reichert, D. E.; Welch,
M. J.; Mach, R. H. J. Med. Chem. 2005, 48, 7637.
17. Lawrence, H. R.; Pireddu, R.; Chen, L.; Luo, Y.; Sung, S. S.; Szymanski, A. M.; Yip,
M. L.; Guida, W. C.; Sebti, S. M.; Wu, J.; Lawrence, N. J. J. Med. Chem. 2008, 51,
4948.
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmcl.2011.04.064.
References and notes
18. Zhou, L.; Liu, Y.; Zhang, W.; Wei, P.; Huang, C.; Pei, J.; Yuan, Y.; Lai, L. J. Med.
Chem. 2006, 49, 3440.
1. Giordano, S.; Ponzetto, C.; Di Renzo, M. F.; Cooper, C. S.; Comoglio, P. M. Nature
1989, 339, 155.