Design, synthesis and in vitro evaluation of novel bivalent S-adenosylmethionine analogues

Article abstract of DOI:10.1016/j.bmcl.2011.11.017

In optimal cases, bivalent ligands can bind with exceptionally high affinity to their protein targets. However, designing optimised linkers, that orient the two binding groups perfectly, is challenging, and yet crucial in both fragment-based ligand design

Full text of DOI:10.1016/j.bmcl.2011.11.017

Bioorganic & Medicinal Chemistry Letters 22 (2012) 278–284
Contents lists available at SciVerse ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Design, synthesis and in vitro evaluation of novel bivalent
S-adenosylmethionine analogues
Catherine Joce a,b, Rebecca White a,b, Peter G. Stockley b, Stuart Warriner a,b
,
W. Bruce Turnbull a,b, Adam Nelson a,b,
⇑
a School of Chemistry, University of Leeds, Leeds LS2 9JT, UK
b Astbury Centre for Structural Molecular Biology, University of Leeds, Leeds LS2 9JT, UK
a r t i c l e i n f o
a b s t r a c t
Article history:
In optimal cases, bivalent ligands can bind with exceptionally high affinity to their protein targets. How-
ever, designing optimised linkers, that orient the two binding groups perfectly, is challenging, and yet
crucial in both fragment-based ligand design and in the discovery of bisubstrate enzyme inhibitors. To
further our understanding of linker design, a series of novel bivalent S-adenosylmethionine (SAM) ana-
logues were designed with the aim of interacting with the MetJ dimer in a bivalent sense (1:1 ligand/MetJ
dimer). A range of ligands was synthesised and analyzed for ability to promote binding of the Escherichia
coli methionine repressor, MetJ, to its operator DNA. Binding of bivalent SAM analogues to the MetJ
homodimer in the presence of operator DNA was evaluated by fluorescence anisotropy and the effect
of linker length and structure was investigated. The most effective bivalent ligand identified had a flexible
linker, and promoted the DNA–protein interaction at 21-times lower concentration than the correspond-
ing monovalent control compound.
Received 28 September 2011
Revised 3 November 2011
Accepted 6 November 2011
Available online 11 November 2011
Keywords:
MetJ
Methionine repressor
S-Adenosylmethionine analogues
Bivalent ligands
Ó 2011 Elsevier Ltd. All rights reserved.
Bivalent ligands, in which two identical binding groups are
linked by a spacer unit, can have exceptionally high binding affin-
ity, and can, consequently, be useful modulators of biological func-
tion.1 However, the optimisation of linkers between the binding
groups can be challenging, and, yet, is crucial in both fragment-
based ligand (and drug) design2 and in the discovery of bisubstrate
inhibitors.3 In this paper, we describe a series of symmetrical biva-
lent ligands that was prepared to extend our understanding of the
effects of linker length and flexibility on biological activity.
The design of bivalent SAM analogues started by examination of
the SAM–MetJ dimer–DNA complex crystal structure (Fig. 1).4 In
this structure, a pair of co-repressor molecules is arranged sym-
metrically with the terminal carboxyl groups ꢀ5 Å apart. When
designing the structure of bivalent SAM ligands, consideration
was given to the structures of previously identified SAM analogues
(Panel A, Fig. 2). Aza-SAM, 2, is a stable nitrogen analogue of the
(unstable9) natural co-repressor SAM, 1, which has been shown
to bind to the protein in an identical conformation.10 Previously,
we have shown that both amide formation at the carboxy terminus
S-Adenosylmethionine (SAM, 1)4,5 was chosen as the binding
unit due to its well studied interaction with the Escherichia coli
methionine repressor, MetJ (Fig. 1). MetJ is a homodimeric DNA-
binding protein which functions in complex with two molecules
of the co-repressor, SAM. Binding of positively charged SAM mole-
cules is believed to promote binding of the repressor to its target
DNA by a unique mechanism based primarily on long-range elec-
trostatics transferred through the protein structure.6,7 The dissoci-
ation constant of the MetJ-consensus, minimum operator DNA
complex in the presence of saturating (1 mM) levels of SAM 1
has been determined to be 4 ( 3) nM by filter binding studies.
However, in the absence of SAM, 1, saturation binding of the pro-
tein to the DNA was never achieved and the Kd was estimated to
of aza-SAM (to give 3) and removal of the a-NH2 group (to give 4)
do not greatly affect function.11 In contrast, quaternisation of the
50-position amine of 4, to give the charged analogue 5, significantly
improved activity.11 A range of bivalent SAM derivatives was,
therefore, designed in which secondary amides were used to link
two aza-SAM analogues (Panel B, Fig. 2). The optimal separation
of the terminal carbons in the linkers of such bivalent ligands
may be estimated by analysis of the structure of the SAM–MetJ
dimer–DNA complex. Provided that the bivalent ligands interact
with MetJ analogously to SAM, and that the secondary amides
adopt the preferred12 syn conformation, the optimal separation of
the terminal carbons of the linker is ꢀ5.3 Å. The proposed interac-
tion of an exemplar bivalent ligand to the MetJ dimer, in which a
simple linker bridges between the binding units, is shown in Panel
C, Figure 2. It was proposed to optimise the bivalent analogues by
varying the length and flexibility of the linker and by investigating
the effect of quaternisation.
be 10 l
M.8
⇑
Corresponding author. Tel.: +44 113 343 6502; fax: +44 113 343 6565.
E-mail address: a.s.nelson@leeds.ac.uk (A. Nelson).
0960-894X/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2011.11.017

C. Joce et al. / Bioorg. Med. Chem. Lett. 22 (2012) 278–284
279
NH2
N
O
H2N
OH
N
S
N
O
N
HO
OH
SAM, 1
Figure 1. Co-crystal structure of two molecules of co-repressor SAM, 1, bound on the surface of the MetJ dimer and the chemical structure of SAM.
Initial studies focused on bivalent analogues incorporating rigid
linkers, for example linkers based on 7a and 7b (Scheme 1) in
which the benzylic carbon atoms are separated by 5.8 and 5.0 Å
respectively. Accordingly, coupling of the carboxylic acid 6 with
the diamines 7a and 7b gave the bivalent derivatives 8a and 8b
in 68% and 66% yield respectively; unfortunately, partial (ꢀ20%)
corresponding unquaternised ligands 9a,b; alternatively, methyla-
tion and deprotection afforded the quaternised derivatives 10a,b
(Scheme 2). The bivalent ligands 9c–e (Fig. 3 and Supplementary
data) and the monovalent benzyl amides 11 and 12 were also pre-
pared (Fig. 4 and Supplementary data).
It was also decided to prepare a series of bivalent ligands with
more flexible linkers in which the length of the linker was varied
epimerization,
a to the carbonyl group, occurred under the reac-
tion conditions (Scheme 1).
The intermediates 8 were exploited in the synthesis of two
different analogue classes. Acid-catalyzed deprotection gave the
more widely (see Scheme 3). The a-NH2 group was omitted for this
series, both because this group has only a small11 effect on the
function of monovalent amides (e.g., 3) and because of the epimer-
A
NH2
N
NH2
N
H2N
O
O
H2N
N
N
O
NHPr
NHPr
N
OH
N
N
N
N
N
N
O
N
O
HO
OH
3
HO
OH
aza-SAM, 2
N
N
I
H2N
H2N
O
N
NHPr
N
N
N
O
N
O
N
HO
OH
HO
N
OH
5
4
HO
OH
B
N
O
N
N
N
O
N
NH2
H
H
N
H2N
N
O
N
N
5.3 Å
N
N
O
HO
OH
C
Figure 2. Design of bivalent ligands. (A) Monovalent ligands for the MetJ dimer. (B) Design of symmetrical bivalent SAM analogues; the optimal separation of the terminal
carbon atoms in the linker was expected to be ꢀ5.3 Å. (C) Illustration of the possible interaction between an exemplar bivalent ligand and the MetJ homodimer. In this
example, a simple linker joins the terminal carboxyls of two binding units analogues via secondary amides.

280
C. Joce et al. / Bioorg. Med. Chem. Lett. 22 (2012) 278–284
NH2
H2N
H2N
NHBoc
CO2H
N
8a
8b
H2N
7a
N
O
N
or
N
N
O
O
NH2
6
7b
i
Scheme 1. The intermediates 8a and 8b were prepared by coupling the carboxylic acid 6 and the diamines 7 (PyBOP, Pr2NEt, DMF); partial (ca. 20%) epimerization was
observed and all bivalent derivatives were consequently isolated as mixtures of diastereoisomers.
NH2
HN
N
H2N
O
O
N
N
HO
OH
N
NH
N
N
N
HO
OH
N
O
N
O
NH2
N
NH2
9a
9b
(p-substitution), 80%
(m-substitution), 89%
5M aq. HCl
NHBoc
N
H2N
O
O
N
N
O
O
N
HN
NH
N
N
N
O
O
N
O
N
O
NH2
N
NHBoc
8a
(p-substitution)
(m-substitution)
8b
1. MeI, MeOH
−
2. TFA H2O
NH2
N
I
H2N
O
O
N
N
HO
OH
N
HN
NH
N
N
N
HO
OH
N
N
O
O
NH2
I
N
NH2
.4 TFA
10a
10b
(p-substitution), >98%
(m-substitution), 68%
Scheme 2. Synthesis of the bivalent ligands 9a,b and 10a,b.
ization observed in couplings of
a
-NHBoc acids. Strain-promoted
Titration of MetJ into a solution containing fixed concentrations
of F-metC and ligand allows the change in anisotropy related to ter-
nary complex formation to be measured (Fig. 5).
‘click’ reaction15 between the azides 13 (n = 211 and 4) and the cyc-
looctyne 1411 and acetonide removal gave the corresponding biva-
lent analogues 15 as mixtures of regioisomers. Furthermore,
The half-maximal concentration of MetJ, EC50, required to
promote complex formation was determined for each ligand. The
concentration of the ligands used was comparable to the low
millimolar SAM concentrations used in previous in vitro binding
assays, and to the estimated SAM concentration in vivo.8 The
monovalent ligands 11 and 12 (2 mM) and the bivalent ligands 9
and 10 (1 mM) all promoted the formation of the MetJ–DNA com-
plex (Table 1). The EC50 was found to be 120 10 nM and
36 6 nM in the presence of the unquaternised and quaternised
monovalent control ligands 11 and 12 respectively. The unquatern-
ised bivalent ligands 9a–e (1 mM) promoted complex formation up
to about threefold more effectively than the corresponding mono-
reaction between the carboxylic acid11 17 and a series of
a,x-dia-
mines, and deprotection, gave the bivalent analogues 18 (n = 2, 3, 4
and 6). Finally, treatment of the bivalent analogues 15 and 18 with
methyl iodide gave the corresponding quaternised derivatives 16
and 19.
A fluorescence anisotropy-based binding assay was used to
compare the relative ability of SAM analogues to promote MetJ di-
mer–DNA complex formation. Since SAM has low affinity for MetJ
in the absence of DNA,8,13 measurements were made in the pres-
ence of the DNA fragment F-metC, a fluorescently-labeled analogue
of the shortest naturally occurring operator sequence, metC.14

C. Joce et al. / Bioorg. Med. Chem. Lett. 22 (2012) 278–284
281
N
NH2
H2N
O
O
N
N
N
HN
N
HO OH
HO OH
O
NH
N
N
N
N
O
N
NH2
9c
9d
9e
NH2
N
NH2
HN
H2N
O
N
O
N
HO OH
O
N
N
NH
N
N
N
HO
OH
N
O
N
NH2
NH2
N
NH2
HO OH
H2N
NH
N
O
O
N
N
N
N
N
O
N
HN
O
N
NH2
HO OH
N
NH2
Figure 3. Structures of the bivalent ligands 9c–e; partial epimerization occurred during the amide formation and the subsequent derivatives were isolated as mixtures of
diastereoisomers.
NH2
NH2
N
N
I
H2N
H2N
O
O
N
N
O
O
N
N
N
N
N
N
NHBn
NHBn
HO
HO
OH
12
O1H1
Figure 4. Structures of the monovalent ligands 11 and 12; partial epimerization occurred during the amide formation and the subsequent derivatives were isolated as
mixtures of diastereoisomers
valent analogue 11. Similarly, the quaternised bivalent ligands 10a
and 10b promoted, respectively, complex formation 2.4- and 3.6-
fold more effectively than the quaternised monovalent analogue
12. It is notable that the EC50 value of the ligand 10b was signifi-
cantly lower than that of the (unstable9) co-repressor, SAM. For
these bivalent analogues with rigid linkers, the nature of the linker
did not have a profound effect on function: in each case, the EC50
values were comparable to those of the corresponding monovalent
controls (either 11 or 12) and, consequently, it is unlikely that the
linkers allow the binding groups to engage with the MetJ dimer in
a bivalent sense.
The ability of the unquaternised ligands 4, 15, and 18 (2 mM) to
promote the formation of the MetJ–DNA complex was also investi-
gated using fluorescence anisotropy (Table 2). In the presence of
the monovalent ligand 4, the EC50 was 1000 100 nM.11 Mixtures
of the regioisomeric triazoles 15a and 15b promoted complex for-
mation at 9- and 13-fold lower concentration than the monovalent
ligand 4 respectively. In contrast, he activity of the bivalent ligands
18—which had shorter and more flexible linkers than the triazoles
15—depended critically on the length and nature of the linker; the
compound with the shortest linker—18a in which n = 2—had simi-
lar activity to the monovalent derivative 4. However, as the linker
length increased, the EC50 improved from 700 50 nM (with 18a,
n = 2) to 47 1 nM (with 18d, n = 6). Unfortunately, derivatives
with longer linkers (n = 9 and 12) were not soluble under the con-
ditions of the assay. The clear dependence of the activity of the li-
gands 18a–d on the length of the linker is consistent with
interaction with the MetJ dimer in a bivalent sense.
generating only modest improvements in affinity of up to 3.5-fold.
The EC50 improved from 150 10 nM with the quaternised mono-
valent analogue 5 to 42 3 nM with the most active quaternised
bivalent derivative 16b.
The results obtained in this study highlight the challenges asso-
ciated with designing effective linkers in bivalent ligands. The
activity of the bivalent ligands 9 and 10, which have rather rigid
linkers, was disappointing because the ligand concentrations re-
quired to promote DNA–protein interaction were comparable to
those of the corresponding monovalent controls (11 or 12). For
effective interaction, a rigid linker must control both the spacing
and the relative orientation of the pair of binding groups: with 9
and 10, it is likely that the linker design did not allow these ligands
to engage with the MetJ dimer in a bivalent sense. However, in
sharp contrast, the activity of the bivalent ligands, 18a–d (n = 2,
3, 4, or 6) did depend critically on the length of the flexible linker.
The effective length of flexible linkers in bivalent ligands has been
estimated using alternative approaches including molecular
dynamics simulations16 and random walk models.17 The optimal
separation of the terminal carbons in the linkers of effective biva-
lent ligands 18 was estimated to be ꢀ5.3 Å (Fig. 1). However, the
most effective unquaternised bivalent ligand 18 had n = 6 (i.e.,
18d), reinforcing that the effective lengths of flexible linkers are
much shorter than their extended lengths.
Designing symmetrical ligands which interact effectively with
their target in a bivalent sense is a difficult task, as observed previ-
ously,18–20 and highlighted again by this study. Here, modest affin-
ity enhancements due to bivalency were only observed when
flexible linkers were exploited: in the case of 18d, a 21-fold in-
crease in activity was observed compared to its monovalent ana-
logue, 4. We, therefore, find that simple, flexible linkers are a
useful starting point in the design of bivalent ligands; once
The ability of the quaternised ligands 16 and 19 (2 mM) to pro-
mote the formation of the MetJ–DNA complex was also investi-
gated using fluorescence anisotropy (Table 3). The effect of the
linker length was less profound with the quaternised analogues,

282
C. Joce et al. / Bioorg. Med. Chem. Lett. 22 (2012) 278–284
N
H
N
H2N
O
N
N
N3
n-1
O
N
O
N
O
O
13a
13b
(n=2)
(n=4)
O
O
N
N
N
O
N
N
H
N
O
NH2
N
14
N
1. MeCN
2. 5M aq. HCl
H2N
N
O
HO
OH
N
N
N
N
N
HO
OH
H
N
O
NH2
N
N
O
n-1
N
O
N
N
HN
O
15a
(n = 2) 34% over two steps
15b (n = 4) step 1: 40%, step 2: 81%
N
MeI, H2O-MeCN
H2N
HO
OH
O
N
N
I
N
N
N
N
N
O
HO
OH
NH2
H
N
N
N
O
n-1
N
N
O
I
N
HN
O
16a
16b
(n = 2) 73%
(n = 4) 88%
N
H2N
OH
N
O
N
N
O
N
O
O
17
1. H2N(CH2)nNH2, PyBOP, DIPEA, DMF
2. 5M aq. HCl
HO
OH
N
N
O
N
H
H2N
O
N
N
N
N
N
H
N
O
NH2
n-1
N
O
N
N
18a
(n = 2) 18% over two steps
(n = 3) step 1: 96%, step 2: 74%
(n = 4) step 1: 99%, step 2: 73%
(n = 6) step 1: 96%, step 2: 71%
HO
OH
18b
18c
18d
MeI, H2O-MeCN
HO
OH
N
N
O
N
I
H
H2N
O
N
N
N
N
N
N
H
O
NH2
n-1
N
N
O
N
I
HO
OH
19b
19d
(n = 3) 78%
(n = 6) 79%
Scheme 3. Synthesis of bivalent ligands; the triazoles 15 and 16 were prepared as mixtures of regioisomers.

C. Joce et al. / Bioorg. Med. Chem. Lett. 22 (2012) 278–284
283
SAM binding
promotes.....
MetJ
MetJ
MetJ
MetJ
F
.....F-metC binding.
Complex
F
LOW ANISOTROPY
HIGH ANISOTROPY
Figure 5. Cartoon illustrating the fluorescence anisotropy assay. The SAM molecules promote the formation of a SAM–F-metC–protein complex, with two MetJ dimers bound
to the 18 base-pair DNA duplex.
Table 1
Table 3
EC50 values, that is the concentration of MetJ monomer required
to promote half-maximal formation of its complex with the F-
metC DNA, in the presence of ligands (see Scheme 2 and Fig. 3);
improvements in the activity of bivalent ligands relative to a
monovalent control are shown
EC50 values, that is the concentration of MetJ monomer required to promote half-
maximal formation of its complex with the F-metC DNA, in the presence of ligands
(see Scheme 3 and Fig. 2); improvements in the activity of bivalent ligands relative to
a monovalent control are shown
Compound
EC50 (nM)
150 10
Fold-improvement over monovalent control 5
Compound
EC50 (nM)
Fold-improvement over
monovalent control
(11 or 12)
5
—
16a (n = 2)
16b (n = 4)
19b (n = 3)
19d (n = 6)
50
42
108
51
6
3
3
2
3.0-fold
3.6-fold
1.4-fold
2.9-fold
SAM, 1
17.3 0.3
120 10
—
11
9a
9b
9c
9d
9e
—
61
37
3
2
1
4
5
2.0-fold
3.2-fold
3.1-fold
1.6-fold
1.1-fold
The concentration of the ligands was 2 mM and the concentration of F-metC was
10 nM. EC50 values were determined based on an average of three titrations, fitted
to a sigmoidal growth logistic model (see Supplementary data).
39
75
111
12
10a
10b
36
15
10
6
5
1
—
2.4-fold
3.6-fold
Acknowledgments
The concentration of the monovalent ligands 11 and 12 was
2 mM, and the concentration of the bivalent ligands 9 and 10
was 1 mM. The concentration of F-metC was 10 nM, and the final
concentration of DMSO was 2%. EC50 values were determined
based on an average of three titrations, fitted to a sigmoidal
growth logistic model (see Supplementary data).
We thank the Wellcome Trust, EPSRC and BBSRC for funding.
Supplementary data
Supplementary data (supplementary schemes, experimental
details for the preparation of ligands and fluorescence anisotropy
experiments) associated with this article can be found, in the on-
line version, at doi:10.1016/j.bmcl.2011.11.017.
Table 2
EC50 values, that is the concentration of MetJ monomer required
to promote half-maximal formation of its complex with the F-
metC DNA, in the presence of ligands (see Scheme 3 and Fig. 2);
improvements in the activity of bivalent ligands relative to a
monovalent control are shown
References and notes
1. (a) Kramer, R. H.; Karpen, J. W. Nature 1998, 395, 710; (b) Pickens, J. C.; Mitchell,
D. D.; Liu, J.; Tan, X.; Zhang, Z.; Verlinde, C. L.; Hol, W. G.; Fan, E. Chem. Biol.
2004, 11, 1205; (c) LaFrate, A. L.; Carlson, K. E.; Katzenellenbogen, J. A. Bioorg.
Med. Chem. 2009, 17, 3528.
2. (a) Murray, C. W.; Rees, D. C. Nat. Chem. 2009, 1, 187; (b) Hadjuk, P. J.; Greer, J.
Nat. Rev. Drug Disc. 2007, 6, 211.
3. Lavogina, D.; Enkvist, E.; Uri, A. ChemMedChem 2010, 5, 23.
4. Rafferty, J. B.; Somers, W. S.; Saint-Girons, I.; Phillips, S. E. V. Nature 1989, 341,
705.
5. Phillips, S. E. V.; Manfield, I.; Parsons, I.; Davidson, B. E.; Rafferty, J. B.; Somers,
W. S.; Margarita, D.; Cohen, G. N.; Saint-Girons, I.; Stockley, P. G. Nature 1989,
341, 711.
6. Parsons, I. D.; Persson, B.; Mekhalfia, A.; Blackburn, G. M.; Stockley, P. G. Nucleic
Acids Res. 1995, 23, 211.
7. Cooper, A.; McAlpine, A.; Stockley, P. G. FEBS Lett. 1994, 348, 41.
8. Phillips, S. E. V.; Manfield, I.; Parsons, I.; Davidson, B. E.; Rafferty, J. B.; Somers,
W. S.; Margarita, D.; Cohen, G. N.; Saint-Girons, I.; Stockley, P. G. Nature
(London) 1989, 341, 711.
Compound
EC50 (nM)
Fold-improvement over
monovalent control 4
4
1000 100
—
15a (n = 2)
15b (n = 4)
18a (n = 2)
18b (n = 3)
18c (n = 4)
18d (n = 6)
110
76
700 50
320 40
220 20
6
2
9.1-fold
13-fold
1.4-fold
3.1-fold
4.5-fold
21-fold
47
1
The concentration of the ligands was 2 mM and the concentra-
tion of F-metC was 10 nM. EC50 values were determined based on
an average of three titrations, fitted to
logistic model (see Supplementary data).
a sigmoidal growth
9. Hoffman, J. L. Biochemistry 1986, 25, 4444.
10. Thompson, M. J.; Mekhalfia, A.; Jakeman, D. L.; Phillips, S. E. V.; Phillips, K.;
Porter, J.; Blackburn, G. M. Chem. Commun. 1996, 791.
11. Joce, C.; Caryl, J.; Stockley, P. G.; Warriner, S.; Nelson, A. Org. Biomol. Chem.
2009, 7, 635.
12. Chakrabarti, P.; Dunitz, J. D. Helv. Chim. Acta 1982, 65, 1555.
13. Stockley, P. G.; Baron, A. J.; Wild, C. M.; Parsons, I. D.; Miller, C. M.; Holtham, C.
A. M.; Baumberg, S. Biosens. Bioelectron. 1998, 13, 637.
14. Phillips, S. E. V. Curr. Opin. Struct. Biol. 1991, 1, 89.
15. Agard, N. J.; Prescher, J. A.; Bertozzi, C. R. J. Am. Chem. Soc. 2004, 126, 15046.
evidence of bivalent interaction with a target has been obtained
with such ligands, optimisation of the linker may subsequently
be possible. The development of strategies for designing effective
linkers remains an important challenge, however, because of the
large affinity enhancements that are possible with bivalent ligands
in optimal cases.

284
C. Joce et al. / Bioorg. Med. Chem. Lett. 22 (2012) 278–284
16. Kitov, P. I.; Shimizu, H.; Homans, S. W.; Bundle, D. R. J. Am. Chem. Soc. 2003, 125,
3284.
19. Lindhorst, T. K.; Bruegge, K.; Fuchs, A.; Sperling, O. Beilstein J. Org. Chem. 2010,
6, 801–809.
17. Knoll, D.; Hermans, J. J. Biol. Chem. 1983, 258, 5710.
18. Zhang, Y.; Gilliam, A.; Maitra, R.; Damaj, M. I.; Tajuba, J. M.; Seltzman, H. H.;
Thomas, B. F. J. Med. Chem. 2000, 53, 7048.
20. Bujotzek, A.; Shan, M.; Haag, R.; Weber, M. J. Comput. Aided Mol. Des. 2011, 25,
253.