Targeting integrins: Insights into structure and activity of cyclic RGD pentapeptide mimics containing azabicycloalkane amino acids

Article abstract of DOI:10.1016/j.bmc.2005.08.048

A small library of cyclic RGD pentapeptide mimics incorporating stereoisomeric 5,6- and 5,7-fused bicyclic lactams was synthesized. This library was found to contain high-affinity ligands for the α_vβ ₃ integrin. The aim of this study

Full text of DOI:10.1016/j.bmc.2005.08.048

Bioorganic & Medicinal Chemistry 14 (2006) 169–180
Targeting integrins: Insights into structure and activity of cyclic
RGD pentapeptide mimics containing azabicycloalkane amino acids
Laura Belvisi,a Anna Bernardi,a Matteo Colombo,a, Leonardo Manzoni,b
Donatella Potenza,a Carlo Scolastico,a,* Giuseppe Giannini,c Marcella Marcellini,c
Teresa Riccioni,c Massimo Castorina,c Pietro LoGiudicec and Claudio Pisanoc,*
aDipartimento di Chimica Organica e Industriale and Centro Interdisciplinare Studi bio-molecolari e applicazioni Industriali, (CISI),
`
Universita degli Studi di Milano, via G. Venezian 21, I-20133 Milan, Italy
bCNR-Istituto di Scienze e Tecnologie Molecolari, via C. Golgi 19, I-20133 Milan, Italy
cResearch and Development, Sigma-Tau, via Pontina Km 30,400, I-00040 Pomezia, Italy
Received 15 February 2005; accepted 2 August 2005
Available online 7 October 2005
Abstract—A small library of cyclic RGD pentapeptide mimics incorporating stereoisomeric 5,6- and 5,7-fused bicyclic lactams was
synthesized. This library was found to contain high-affinity ligands for the avb3 integrin. The aim of this study was to investigate activ-
ity, selectivity, and structure of these ligands in order to identify new specific av-integrin antagonists that could be evaluated as tumor
angiogenesis inhibitors. In vitro screening, including receptor-binding assays to purified avb3, avb5, and a5b1 integrins, and platelet
aggregation assay, revealed ST1646 as a potent, highly selective avb3/avb5 integrin antagonist. Structure determination of the cyclic
RGD pentapeptide mimics performed by a combination of NMR spectroscopy, and molecular mechanics and dynamics calculations
showed a strong dependence of the RGD cyclopeptide conformation on lactam ring size and stereochemistry. ST1646 revealed the
highest ability within the library to adopt the proper RGD orientation required for binding to the avb3 integrin, as deduced from the
recently solved crystal structure of the extracellular segment of integrin avb3 in complex with a cyclic pentapeptide ligand.
ꢀ 2005 Elsevier Ltd. All rights reserved.
1. Introduction
units which combine to give a wide amount of heterodi-
mers. Besides cell adhesion to extracellular matrix,
integrins also mediate intracellular events that control
cell shape, migration, proliferation, and survival.5,8
Many integrins recognize polypeptide domains contain-
ing the Arg-Gly-Asp (RGD) aminoacid sequence present
in several matrix associated adhesive glycoproteins.9,10
Specificity and efficacy of these molecular recognition
processes are determined by the context of the RGD se-
quence, including flanking residues, conformational pre-
sentation of the triad, and individual features of the
integrin binding pockets.11,12
Cell adhesion is essential for the proper functionality of
many physiological processes such as embryogenesis,
cell differentiation, hemostasis, wound healing, and im-
mune response, but also for pathophysiological events
such as tumor cell invasion, formation of metastases,
and tumor-induced angiogenesis.1–4 These cell adhesion
processes are mediated by cell-surface receptors, among
which integrins represent the most diverse and promi-
nent class.5–8
Integrins are a family of membrane-spanning adhesion
receptors composed of noncovalently linked a and b sub-
Among such RGD-dependent integrins, the vitronectin
receptors aVb3 and aVb5 have recently received increas-
ing attention as interesting therapeutic targets, because
of their critical role in tumor-induced angiogenesis and
metastasis formation.13
Keywords: Integrin receptor antagonists; Peptidomimetics; Conforma-
tional analysis; Angiogenesis inhibitors.
*
Corresponding authors. Tel.: +39 02 50314090; fax: +39 02 50314072
(C.S.); tel.: +39 06 91393760; fax: +39 06 91393988 (C.P.); e-mail
addresses: carlo.scolastico@unimi.it; claudio.pisano@sigma-tau.it
Present address: Nikem Research, via Zambeletti 25, I-20021,
Baranzate di Bollate (MI), Italy.
avb3 is not generally expressed on epithelial cells and
normal endothelial cells (EC), but it is significantly
upregulated on activated EC and in metastatic tumor
0968-0896/$ - see front matter ꢀ 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2005.08.048

170
L. Belvisi et al. / Bioorg. Med. Chem. 14 (2006) 169–180
cells. Growth factors such as fibroblast growth factor-2
(FGF2) and tumor necrosis factor-a (TNFa) stimulate
avb3 expression in the developing chick chorioallantoic
membrane (CAM) and in the angiogenesis model of
the rabbit cornea.14 The importance of avb3 in tumor
angiogenesis is also demonstrated by the fact that avb3
antagonists, including cyclic RGD peptides and mono-
clonal antibodies, were successfully used to inhibit blood
vessel development and tumor growth in different mod-
els.15,16 It is noteworthy that avb3 antagonists have very
little effect on pre-existing blood vessels, indicating the
usefulness of targeting this receptor for therapeutic ben-
efit without adverse side effects.
O
( )n
HN
N
COOH
( )n
1N
Arg
Gly
2
3
O
O
NH2
Asp
n=1,2
II
I
Figure 1. Azabicycloalkane scaffolds of type I and general formula II
of the library cyclo(Arg-Gly-Asp-Lactam).
dipeptide analogs for the synthesis of a library of general
formula cyclo(Arg-Gly-Asp-Lactam) II (Fig. 1). This li-
brary was found to contain specific high-affinity ligands
Recent studies have implicated a related integrin, aVb5,
in angiogenesis, possibly via a signaling pathway distinct
from that of avb3 and activated by a different growth
factor.17 The existence of distinct angiogenic pathways
can be explained with the prevalence of specific growth
factors and/or cell-adhesive proteins in different
conditions.
O
N
H
X1
Arg
Gly
a : X1= Pmc; X2 = tBu
b : X1 = X2 = H
O
H
HN
Asp
Furthermore, many steps that characterize the acquisi-
tion of invasive and metastatic potential by tumor cells,
such as intravasation, adhesion to the vessel wall,
extravasation, infiltration, and proliferation into target
tissue, involve integrins. This implies that migratory tu-
mor cells express the integrin receptors and the enzymat-
ic machinery for recognition, adhesion, matrix
degradation, and penetration involved in cell–cell and
cell–ECM interactions. For example, integrin avb3
expression is upregulated on certain invasive tumors
including metastatic melanoma18 and late stage glioblas-
toma.19 Thus, inhibition of avb3 receptor, in addition to
blocking tumor-induced angiogenesis, has impact on a
number of mechanisms involved in tumor progression.
X2
1a (5,6 cis S)
1b (5,6 cis S)
H
H
O
O
N
N
X1
X1
Arg
Gly
Arg
O
O
H
H
HN
HN
Gly
Asp
Asp
X2
X2
2a (5,6 trans R)
2b (5,6 trans R)
3a (5,6 trans S)
3b (5,6 trans S)
Taken together, these results suggest that selective avb3
and/or avb5 antagonists may represent a promising ap-
proach for the inhibition of tumor angiogenesis and tu-
mor growth. Various RGD-containing cyclic peptides
have been developed by different groups as active and
selective integrin antagonists that compete with matrix
molecules for specific integrin receptors.20–22
H
H
O
O
N
N
X1
X1
Arg
Gly
Arg
Gly
O
O
H
H
HN
HN
Asp
Asp
X2
X2
An efficient procedure of spatial screening, performed by
Kessler et al. and based on the synthesis of stereoisomer-
ic cyclic peptide libraries, led to the highly active aVb3-se-
lective first-generation cyclic pentapeptide cyclo(Arg-
Gly-Asp-D-Phe-Val).23–25 Extensive modifications of
this lead structure with different peptidomimetics and
carbohydrate scaffolds have been performed, and new
potent antagonists have been identified.26–28 The system-
atic derivatization of the lead peptide resulted in the N-
4a (5,7 trans S)
4b (5,7 trans S)
5a (5,7 trans R)
5b (5,7 trans R) ST1646
O
O
N
N
X1
X1
Arg
O
Arg
O
H
H
Gly
Gly
HN
HN
alkylated
cyclopeptide
cyclo(Arg-Gly-Asp-D-Phe-
Asp
Asp
X2
[NMe]Val),29 which has entered clinical phase II studies
as anticancer drug (cilengitide, EMD121974).
X2
6a (5,7 cis S)
6b (5,7 cis S)
7a (5,7 cis R)
7b (5,7 cis R)
Recently, our group has reported a library of cyclic
RGD pentapeptide mimics, based on azabicycloalkane
amino acid scaffolds of type I (Fig. 1).30 Stereoisomeric
5,6- and 5,7-fused bicyclic lactams showing different re-
verse-turn mimetic properties31–33 were exploited as
Figure 2. Cyclic RGD pentapeptide mimics cyclo(Arg-Gly-Asp-Lac-
tam) 1–7. The ring size and the stereochemistry at the bridgehead (cis
or trans) and at the C3 (S or R) carbon of the bicyclic lactam are
reported in parentheses.

L. Belvisi et al. / Bioorg. Med. Chem. 14 (2006) 169–180
171
of aVb3 integrin, that are presently being evaluated as
very promising antiangiogenic drugs. Among the pep-
tides tested, compound ST1646 (Fig. 2 and Table 1,
compound 5b) showed the highest affinity to aVb3,
inhibiting echistatin binding to aVb3 with an IC50 of
3.8 0.9 nM.
c(RGDfV)23–25 or EMD121974,29 respectively, with aza-
bicycloalkane scaffolds showing different reverse-turn
mimetic properties could constrain the RGD sequence
into different conformations and possibly provide the re-
quired activity and selectivity for integrin antagonism.
To test this hypothesis the compounds were screened
in a solid-phase assay for their ability to compete with
radiolabeled echistatin for the binding to purified hu-
man avb3 and avb5 and to compete with radiolabeled
fibrinogen for the binding to a5b1 integrin. The well-
characterized integrin antagonists cyclic pentapeptides
c(RGDfV) and EMD121974, as well as the natural inte-
grin ligands vitronectin, fibronectin, and fibrinogen,
were used as positive controls (Table 1).
Herein, we report the in vitro screening of activity and
selectivity, and the structure determination by spectro-
scopic and computational means of the small library
of cyclic RGD pentapeptide mimics cyclo(Arg-Gly-
Asp-Lactam). Effects of the structural constraint intro-
duced by the bicyclic template on the conformation of
the RGD sequence have been especially investigated
by examining the dependence of the cyclopeptide con-
formation on lactam ring size and stereochemistry.
Moreover, efforts to evaluate the ability of the library
members to adopt the proper RGD orientation required
for binding to the avb3 integrin have been carried out,
using the recently solved crystal structure of the cyclic
pentapeptide ligand EMD121974 in complex with the
extracellular segment of avb3 integrin.34
Among the seven peptides tested, compounds 2b and 5b
showed the highest affinities to aVb3, and inhibited radi-
olabeled echistatin binding to aVb3 with an IC50 of
14.3 4.7 and 3.8 0.9 nM, respectively. Interestingly,
the affinity of 5b for the aVb3 integrin in this kind of as-
say was almost 50 and 5 times higher than the affinity of
the lead structures c(RGDfV) and EMD121974, used as
reference compounds.
2. Results
The most active compound 5b (ST1646) inhibited radio-
labeled echistatin binding to avb5 with IC50 equal to
1.4 0.2 nM. Like echistatin, compounds 2b and 5b
have similar potency against aVb3 and aVb5 whereas
c(RGDfV) and EMD121974 are, respectively, over
1000- and 100-fold more selective for aVb5 in this kind
of assay (Table 1).35
2.1. Screening in the solid-phase receptor-binding assay
The small library of seven cyclic RGD pentapeptide
mimics incorporating stereoisomeric 5,6- and 5,7-fused
bicyclic lactams was screened in vitro for activity and
selectivity. The replacement of the D-Phe-Val or the D-
Phe-[NMe]Val dipeptide present in the lead structures
Selectivity of compound 5b was demonstrated by its
inability to compete with radiolabeled fibrinogen for
a5b1 interaction (IC50 > 10ꢀ5 M). Similar to c(RGDfV),
ST1646 was 500 times less active than echistatin in pre-
venting in vitro platelet aggregation in response to
thrombin receptor-activating peptide (Table 2). Thus,
the inability of ST1646 to inhibit aIIbb3-mediated plate-
let aggregation confirmed the selectivity of this
compound.
Table 1. In vitro binding of RGD peptides to purified integrin
receptorsa
Compound
IC50 SD (nM)
125I]Echistatin
binding to avb3 binding to avb5 binding to a5b1
[
125I]Echistatin
[
[
125I]Fibrinogen
Echistatin
c(RGDfV)
EMD121974 18.9 3.1
0.29 0.08
196 17
0.29 0.02
0.11 0.03
0.13 0.01
11.8 2.7
162 45
>10,000
Vitronectin
Fibronectin
44 17
835 287
17408 2966
154 21
Table 2. Effect of RGD peptides on TRAP-induced guinea pig platelet
aggregationa
Fibrinogen
1b
Compound
IC50 SD (lM)
>1000
9.02 1.9
>1000
2b
3b
14.3 4.7
202 23
Echistatin
c(RGDfV)
EMD121974
2b
0.018 0.0001
10.5 1.7
>10
4b
5b (ST1646)
461 122
3.8 0.9
1.3 0.4
1.39 0.2
>1000
>10,000
>10
6b
7b
491 131
3343 230
5b (ST1646)
12 2.2
>1000
a Peptide-mediated inhibition of platelet aggregation was determined
by incubating platelet-rich plasma (PRP) with different concentra-
tions of the peptides or with vehicle and quantifying the extent of
platelet aggregation by a turbidimetric method within 4 min after the
addition of the agonist 11-mer thrombin receptor activating peptide
(TRAP). The results are plotted (Allfit program) and expressed as the
antagonist concentration that inhibited 50% of platelet aggregation.
Values are means standard deviation of the determinations from
three or four independent experiments. As shown, ST1646 was 500-
fold less active than echistatin in preventing in vitro platelet aggre-
gation, demonstrating a low affinity for the aIIbb3 receptor.
a The compounds were tested in a solid-phase receptor assay for their
ability to compete for the binding of [125I]echistatin to either purified
avb3 or avb5 and for the binding of [125I]fibrinogen to purified a5b1.
Integrin-coated 96-well plates were incubated with radiolabeled
ligand in the presence of serially diluted competing compounds. After
incubation, plates were washed and radioactivity was measured with
a c-counter. The IC50 was calculated as the concentration of com-
pound required for 50% inhibition of ligand binding as estimated by
the Allfit program. Values are means standard deviation of the
determinations from three or four independent experiments.

172
L. Belvisi et al. / Bioorg. Med. Chem. 14 (2006) 169–180
Table 3. Effect of RGD peptides on BMEC adhesion to vitronectin
and fibronectina
to either vitronectin or fibronectin in a dose-dependent
manner with IC50 values equal to 0.9 and 37.5 lM,
respectively. For both cell-adhesion proteins, the activi-
ty of this compound was higher, on a molar basis, than
those of c(RGDfV) and of EMD121974. In contrast, all
the other RGD-containing cyclic pseudopeptides did
not exert a significant antagonist activity. Similar results
were obtained when the compounds were tested on hu-
man microvascular endothelial cells (data not shown).
Compound
IC50 SD (lM)
EC adhesion to
vitronectin
EC adhesion to
fibronectin
c(RGDfV)
EMD121974
10.3 2.6
2.7 0.3
>100
>100
66.8 14
>100
1b
2b
53.1 6.1
>100
53.9 13.9
>100
—
3b
4b
Taken together, the in vitro screening results indicate
that 5b (ST1646) represents a potent and selective
avb3/avb5 integrin antagonist to be tested for anti-angio-
genic activity in vivo. To this aim, a multigram scale
synthesis of 5b was developed by optimizing both the
bicyclic lactam preparation and the peptide synthesis.36
43.2 5.6
0.9 0.2
>100
5b (ST1646)
37.5 2.6
>100
>100
6b
7b
>100
a Bovine microvascular endothelial cell (BMEC) adhesion was per-
formed allowing the cells to adhere to 96-well plates coated with
either fibronectin or vitronectin for 1–3 h at 37 ꢁC, in the presence or
absence of various concentrations of RGD peptides. Toluidine-
stained adherent cells were quantified on a microtiter plate reader at
600 nm. Results, expressed as mean compound concentration SD
that inhibited 50% of cell adhesion (from three independent experi-
ments), show that ST1646 was the most active among the compounds
tested.
2.3. Computational studies
Monte Carlo/energy minimization (MC/EM) conforma-
tional searches designed to investigate the effects of the
bicyclic lactam on the cyclopeptide conformation were
performed on the cyclo(Ala-Gly-Ala-Lactam) pentapep-
tide analogs 1c–7c (Table 4). The principal types of
backbone geometries calculated for these compounds
and their relative stabilities are given in Table 4. The
minimum energy conformations of these cyclic penta-
peptide mimics are characterized by the formation of
peptide secondary structures, in particular c- and b-
turns, that may be stabilized by intramolecular hydro-
gen bonds. In c-turns, the C@O of the first residue (i)
may be hydrogen bonded to the NH of the third residue
(i + 2), giving rise to a seven-membered ring. In b-turns,
the C@O of the first residue (i) may be hydrogen bonded
2.2. Inhibition of cellular adhesion to vitronectin and
fibronectin
To assess the ability of the cyclic RGD pentapeptide
mimics to behave as integrin antagonists for ECs,
BMECs were allowed to adhere to immobilized vitro-
nectin or fibronectin in the presence of increasing con-
centrations of the synthetic pseudopeptides (ranging
between 0.1 and 100 lM). As shown in Table 3, com-
pound 5b (ST1646) significantly inhibited cell adhesion
Table 4. Characteristics of low-energy conformers (MC/EM, AMBER*, and H2O GB/SA) calculated for cyclic pentapeptide mimics 1c–7c
cyclo(Ala-Gly-Ala-Lactam)a
˚
Compound (no. of conf. <3 kcal/mol)
DE (kcal/mol)
Conformation
CbArg–CbAsp distance (A)
1c cyclo(Ala-Gly-Ala-5,6 cis S) (6 conformers)
0.0
2.4
2.9
Type SII—c(Gly)/bII0(Gly-Asp)b
Type SIII—Inv.c(Asp)/bII0(Gly-Asp)b
Type SI—bII0(lactam)/c(Gly)
8.1
8.6
7.4
2c cyclo(Ala-Gly-Ala-5,6 trans R) (3 conformers)
3c cyclo(Ala-Gly-Ala-5,6 trans S) (4 conformers)
0.0
0.8
0.0
0.5
0.6
0.9
Type SIV—bI(Pro-Arg)/Inv.c(Asp)
Type SIII—Inv.c(Asp)/bII0(Gly-Asp)b
Type SI—bII0(lactam)/c(Gly)
9.4
8.6
7.7
9.4
8.2
8.8
Type SIV—bI(Pro-Arg)c/Inv.c(Asp)
Type SII—c(Gly)/bII0(Gly-Asp)b
Type SIII—Inv.c(Asp)/bII0(Gly-Asp)b
4c cyclo(Ala-Gly-Ala-5,7 trans S) (10 conformers)
5c cyclo(Ala-Gly-Ala-5,7 trans R) (5 conformers)
0.0
0.4
0.9
1.4
Type SII—c(Gly)/bII0(Gly-Asp)b
Type SIII—Inv.c(Asp)/bII0(Gly-Asp)b
Type SIV—bI(Pro-Arg)c/Inv.c(Asp)
Type SI—bII0(lactam)/c(Gly)
8.3
8.7
9.4
7.7
0.0
0.2
1.0
Type SIII—Inv.c(Asp)/bII0(Gly-Asp)b
Type SII—c(Gly)/bII0(Gly-Asp)b
Type SIV—bI(Pro-Arg)/Inv.c(Asp)
8.5
8.2
9.3
6c cyclo(Ala-Gly-Ala-5,7 cis S) (3 conformers)
7c cyclo(Ala-Gly-Ala-5,7 cis R) (5 conformers)
0.0
0.1
Type SIV—bI(Pro-Arg)/Inv.c(Asp)
Type SIII—Inv.c(Asp)/bII0(Gly-Asp)b
9.4
8.7
0.0
0.7
Type SIII—Inv.c(Asp)/bII0(Gly-Asp)b
Type SIV—bI(Pro-Arg)/Inv.c(Asp)
8.6
9.3
a The lowest energy conformer of each conformational family within 3 kcal/mol of the global minimum is described.
b Distorted type II0 b-turn.
c Distorted type I b-turn.

L. Belvisi et al. / Bioorg. Med. Chem. 14 (2006) 169–180
173
to the NH of the fourth residue (i + 3), forming a ten-
membered ring. The classification into specific b-turn
or c-turn structural types was based on the geometry
of the peptide backbone, as described by the / and w
torsion angles in residues i + 1 and i + 2 (b-turn) or in
residue i + 1 (c-turn). Accordingly, assignment of a
low-energy conformation to a particular turn type was
made, where possible, on the basis of the ideal / and
w torsion angles ( 30ꢁ) reported by Rose et al.37
site, the most extended conformation of the AGA se-
quence and the highest Cb(Ala)–Cb(Ala) distance values
can be observed in SIV geometry, featuring a bI-turn/in-
verse c-turn arrangement with the Pro residue of the lac-
tam in the i + 1 position of the bI-turn and the Ala
replacing the Asp residue in the i + 1 position of the in-
verse c-turn. Intermediate situations in terms of kink
and distance are introduced by the SII and SIII structural
types, featuring a c-turn at Gly or an inverse c-turn at Ala
(Asp), respectively, and a common distorted bII0-turn
with Gly at the i + 1 position. As an example, the con-
formers calculated for compound 3c are reported in Fig-
ure 3 to show the four cyclopeptide conformations.
Moreover, a strong dependence of the preferred cyclopep-
tide conformations on lactam ring size and stereochemis-
try can be observed in Table 4. For instance, the SI
structural type can be detected only among the conform-
ers of cyclopentapeptides incorporating b-turn inducer
bicyclic lactams, like 1c, 3c, and 4c, with a relative stability
that increases as the turn mimetic properties of the scaf-
fold improve.38
The structural constraints introduced by the cyclic penta-
peptide system and the azabicycloalkane amino acid force
the compounds to assume only few conformations. As
shown in Table 4, four significant geometries (denoted
as SI, SII, SIII, and SIV) can be detected among the con-
formers within 3 kcal/mol of the global minimum calcu-
lated for compounds 1c–7c in water, as implicitly
represented by the GB/SA solvation model.47 Each cyclo-
peptide geometry shows a specific b/c-turn arrangement
that induces a differently kinked conformation of the
AGA sequence with a characteristic Cb(Ala)–Cb(Ala)
distance value. The strongest kink and the shortest Cb–
Cb distances can be observed in SI structural type, featur-
ing a bII0/c-turn arrangement with the bicyclic template in
the i + 1 and i + 2 positions of the bII0-turn and the Gly
residue in the i + 1 position of the c-turn. On the opposite
MC/SD simulations of the cyclic RGD pentapeptide
mimics 1b–7b were then performed at 300 K, starting
from the cyclopeptide backbone geometries located by
the previous MC/EM step. For each compound, conver-
A Type SI
B Type SII
C Type SIII
D Type SIV
Figure 3. Minimum energy conformations of compound 3c cyclo(Ala-Gly-Ala-5,6 trans S). (A) Type SI, bII0(lactam)/c(Gly). (B) Type SII, c(Gly)/
distorted bII0(Gly-Asp). (C) Type SIII, inverse c(Asp)/distorted bII0(Gly-Asp). (D) Type SIV, bI(Pro-Arg)/inverse c(Asp). The formation of
˚
hydrogen bonds in which HO distance < 2.5 A, N–HO bond angle > 120ꢁ, and HO@C angle > 90ꢁ is indicated by a violet line.

174
L. Belvisi et al. / Bioorg. Med. Chem. 14 (2006) 169–180
Table 5. Population of H-bonds during the 10 ns MC/SD simulations (AMBER*, H2O GB/SA) of cyclic RGD pentapeptide mimics 1b–7b
cyclo(Arg-Gly-Asp-Lactam)
Donor
Acceptor
Turn
% H-bonda
1b
2b
3b
4b
5b
6b
7b
Asp NH
Lact. NH
Lact. NH
Arg NH
Gly NH
Arg CO
Gly CO
Arg CO
Asp CO
Lact. CO
c(Gly)
31
3
45
24
11
3
5
Inv. c(Asp)
bII0(Gly-Asp)
bII0(lactam)
bI(Pro-Arg)
23
40
16
23
33
57
32
9 + 47b
16 + 7b
6 + 21b
0 + 20b
0+1b
26 + 34b
1 + 14b
20 + 44b
5 + 26b
6+48b
9 + 15b
5 + 18b
0 + 50b
0 + 2b
1 + 12b
0 + 28b
0 + 2b
13 + 55b
0 + 42b
0 + 0b
4 + 15b
a
˚
% H-bond is the percentage of conformations sampled during the simulation in which HO distance <2.5 A
b
˚
˚
Percentage of conformations featuring a distorted b-turn in which 2.5 A < HO distance < 4 A.
gence of the simulations was obtained and a nice agree-
ment with the results provided by the MC/EM searches
was observed, in terms of backbone geometries, intra-
molecular H-bonds, and Cb–Cb distances. Population
of H-bonds during the 10 ns MC/SD simulations in
GB/SA water of 1b–7b is reported in Table 5. According
to the corresponding Arg-NHꢁ ꢁ ꢁO@C-Asp H-bond, the
bII0-turn at bicyclic lactam is significantly populated
only during the MC/SD simulations of 1b, 3b, and 4b.
(chloroform), which does not offer strong hydrogen
bond competition.
1
We first performed H NMR analysis to determine the
lowest concentration at which aggregation was not sig-
nificant. Then, the internal hydrogen-bonding equilibria
in these compounds were evaluated by measuring the
chemical shift of the N–H protons and their temperature
coefficients (Dd/DT). NOESY spectra were recorded to
investigate both sequential and long-range nOeÕs that
provide evidences of preferred conformations and give
insight into stable reverse turn conformations.
2.4. NMR spectroscopy
1
In the H NMR spectrum of compound 1a (5,6 cis 3S)
The conformational behavior of protected RGD-penta-
peptides 1a–5a was elucidated by their H NMR fea-
1
the chemical shifts of amide protons Arg-NH, Gly-
NH, and Asp-NH (Table 6)39 suggest that they are all
involved in a hydrogen bond interaction, but each to a
different extent. The temperature dependence (DdNH/
DT) (Table 6) also reflects this behavior and suggests
that the participation of the hydrogen bond of Arg-
tures in CDCl3 solution. The intrinsic folding
propensity of peptides is determined by several factors,
including hydrogen-bond-driven processes. So, we
undertook a conformational study of protected RGD-
pentapeptides 1a–5a, in a relatively nonpolar solvent
a
Table 6. Chemical shifts and temperature coefficients of NH protons of compounds 1a–5a in CDCl3
NH
1a (5,6 cis S)
1H d
DdNH/DT
2a (5,6 trans R)
1H d
DdNH/DT
3a (5,6 trans S)
1H d
DdNH/DT
4a (5,7 trans S)
1H d
DdNH/DT
5a (5,7 trans R)
1H d
DdNH/DT
(ppm)
(ppb/K)
(ppm)
(ppb/K)
(ppm)
(ppb/K)
(ppm)
(ppb/K)
(ppm)
(ppb/K)
Arg
Gly
7.80
8.83
7.62
6.79
ꢀ5.6
ꢀ6.5
ꢀ0.8
+1.8
7.18
7.93
7.56
7.35
ꢀ0.5
ꢀ2.0
ꢀ8.0
ꢀ0.7
7.06
8.06
7.02
7.59
ꢀ1.6
ꢀ12.8
ꢀ9.5
7.19
7.78
7.94
7.64
ꢀ3.0
ꢀ8.0
6.99
7.61
7.62
7.49
ꢀ12.0
ꢀ4.2
Asp
Lactam
ꢀ12.0
ꢀ6.0
ꢀ12.0
ꢀ1.2
ꢀ5.7
a The temperature coefficients for 5,6 bicyclic derivatives were obtained in the temperature range (250–300 K); for the 5,7 bicyclic derivatives the
range was 290–320 K.
NH
A
B
NH
NHPmc
HN
HN
NHPmc
O
O
H
O
H
N
N
H
O
N
N
H7
H7
H
NH
H
O
O
H N
O
H5
H5
O
H
O
H
N
N
H
N
H
H
H
H
N
O
H
H
COOtBu
COOtBu
Figure 4. Significant contacts NOE and b- and c-turn conformations preferred by compound 1a These conformations are not exclusive and can
occur simultaneously; for simplicity we depicted them separately. The dotted lines indicate hydrogen bonds.

L. Belvisi et al. / Bioorg. Med. Chem. 14 (2006) 169–180
175
NH and Gly-NH in 1a becomes much more significant
in the low temperature region. At the same time, the po-
sitive value of DdNH/DT (+1.8 ppb/K, 6.79 ppm) for
Lactam-NH indicates that at low temperature, also the
small amount of hydrogen bond experienced by Lac-
tam-NH was broken.
while the Lactam-NH bond is oriented below the plane.
Hence, the Arg-NH cannot form a 10-membered ring
hydrogen bond (b-turn) with C@O Asp (below the
medium plane) but, if any, a c-turn with lactamic
C@O. In fact, Arg-NH exhibits a medium NOE cross-
peak with H6 (Fig. 5). The chemical shifts and the tem-
perature dependence of amide protons Gly-NH and
Lactam-NH (respectively, 7.93 ppm and ꢀ2 ppb/K for
Gly-NH; 7.35 ppm and ꢀ0.7 ppb/K for Lactam-NH)
(Table 6) are typical of protons locked in a hydrogen
bond. NOESY spectra of compound 2a show NOEs be-
tween Gly-NH and Arg-NH, and between Gly-NH and
Arg-Ha (Fig. 5). These are indicative of a b-turn confor-
mation stabilized by a hydrogen bond between Gly-NH
and lactamic C@O. Moreover, Gly-NH shows a NOE
cross-peak with Lactam-NH as well as the amidic pro-
ton Lactam-NH that is inside to the cyclic pentapeptide
and can experience a hydrogen bond with C@O of Gly.
These data indicate that 2a can fold into a b-hairpin-like
conformation where the b-turn is stabilized by a second
hydrogen bond (Fig. 5).
3
The NOE patterns and the spin–spin coupling JHNH
a
expected for a b-turn conformation are observed in
compound 1a. In fact, the long range NOEs of Gly-
NH with Arg-NH and H7 and the extreme value of J
3
(ArgNH-H8 = 9 Hz) are indicative of a b-turn confor-
mation (Fig. 4B). This b-turn was stabilized by a hydro-
gen bond between Gly-NH and C@O of the bicyclic
lactam. Moreover, Arg-NH forms an additional hydro-
gen bond with C@O of Asp, stabilizing a b-turn with the
proline residue at the i + 2 position (Fig. 4A). The long
range NOE between Arg-NH and Asp-Ha confirms this
conformation.
At the same time, the presence of a c-turn (Fig. 4A), ob-
served by modeling, was confirmed by the temperature
dependence and the value of chemical shift of the amide
proton Asp-NH. The contact NOE between Lactam-
NH and H5 suggests a conformation where Lactam-
NH is not involved in an internal hydrogen bond. An
indication of the rigidity of this part of the cycle comes
from the observation that the side-chain protons of the
Asp residue have restricted rotation: the chemical shifts
for the two b protons are very different (0.5 ppm). In
Figures 4A and B, the significant contacts NOE, and
b- and c-turn conformations preferred by compound
1a are reported. These conformations are not exclusive
and can occur simultaneously; for simplicity, we depict-
ed them separately. Thus, in CDCl3 solution, the
pseudopeptide 1a assumes a very compact folding stabi-
lized by three hydrogen bonds.
1
The H NMR spectrum of compound 3a (5,6 trans S)
shows broad signals, also in the CHa region, indicative
of a slow equilibrium between more equivalent confor-
mations. In fact, the spectra of the same compound
recorded at different temperature (CDCl3, 330 K) or in
DMSO solution show sharp signals. The 5,6 trans 3S-
fused bicyclic lactam used as external constraints for
the RGD tripeptide in the mimic 3a is a very good
inductor of b-turn.32 Indeed, the chemical shift of ami-
dic Arg-NH of compound 3a was at 7.06 ppm (DdNH/
DT = ꢀ1.6 ppb/K) and it represents the behavior of a
amide proton in the hydrogen-bonded state. But, the
temperature dependence (DdNH/DT) and the chemical
shift value (Table 6) of amide proton Lactam-NH hint
toward involvement in a hydrogen bond. Most likely,
Lactam-NH is inside to the cyclopeptide and, in agree-
ment with modeling, experiences a seven-membered
hydrogen-bonded ring with C@O of Gly. The formation
of this c-turn is exclusive with the presence of b-turn
where Arg-NH experiences a hydrogen bond. Compila-
tion of NOE data (Fig. 6) and NH chemical shift
temperature dependence for protons Gly-NH and Asp-
NH (respectively, ꢀ12.8 and ꢀ9.5 ppb/K) in compound
3a provide further evidence for multiple folded confor-
mations in equilibrium. So, the NMR data obtained
for 3a do not give evidence of a unique solution
conformation.
The peptidomimetic 2a is more prone to aggregation so
the spectra were performed below 2 mM CDCl3 solu-
tion. In this compound, the configuration of the scaffold
(5,6 trans R) arranges the peptidic cycle so that the Arg
residue is found above the plane of the bicyclic lactam,
NH
NHPmc
N
H
O
H
H6
N
H
N
H7
O
NH
O
H
N
O
N
H
NHPmc
H
H6
H
H
O
H
N
H
N
N
H
O
H
H
N
O
N
H
H
O
H
O
H
H
N
N
H
H
COOtBu
H
O
Figure 5. Significant contacts NOE and conformation preferred by
compound 2a. Pseudopeptide 2a can fold into b-hairpin-like
conformation where the b-turn is stabilized by a second hydrogen
bond to form (10 + 14) hydrogen bonded rings.
a
COOtBu
Figure 6. Key NOE connectivities found for mimetic 3a.

176
L. Belvisi et al. / Bioorg. Med. Chem. 14 (2006) 169–180
H
N
NH
NHR
NHPmc
O
NH
H
NH
O
H
H
N
H
H6
H
O
N
N
N
H
H7
O
N
N
H7
O
H N
H
O
H N
H
H
O
H
H
O
H
H
N
O
H
N
O
H
H
5a R=Pmc, R'=tBu
5b R=H, R'=H
COOtBu
H
COOR'
Figure 7. Key NOE connectivities found for mimetic 4a.
Figure 8. Key NOE connectivities found for mimetics 5a and 5b.
In previous studies,32 we observed that the (5,7 trans 3S)
lactam included in the pentapeptide 4a is considered by
modeling a very good b-turn inducer.
ondary or tertiary structural element (a-helix, b-sheet,
and reverse turns). So, we analyzed the deprotected
cyclopentapeptide mimics 1b–5b (Fig. 2) in D2O solu-
tion. In this solvent, the amide protons Arg-NH and
Gly-NH of mimic 1b exchange very slowly and this
behavior suggests their involvement in strong intramo-
lecular hydrogen bonds. The exchange is slow also for
Lactam-NH of 2b and 5b, and for Arg-NH of com-
pound 3b.
The 1H NMR spectrum of 4a in 3 mM CDCl3 solutions
exhibited amide protons in the range 7.19–7.94 ppm.
These chemical shifts were characteristic of peptide
backbone NHs strongly involved in hydrogen bonding.
Since these hydrogen bonds cannot occur simultaneous-
ly, we propose that 4a equilibrates between different
folded conformations, similar to compound 3a.
After these results, we decided to study the conforma-
tional features of the most active compound 5b in
H2O/D2O (9:1) solution in order to observe the behavior
of amide protons (Table 7). It is remarkable that the
amide proton Gly-NH (7.65 ppm) is locked in an intra-
molecularly H-bonded state also in a competitive sol-
vent like H2O. The NOESY spectrum shows NOEs
between Gly-NH and Arg-NH, and between Gly-NH
and Arg-Ha. These are indicative of a b-turn conforma-
tion stabilized by a hydrogen bond between Gly-NH
and lactamic C@O. Moreover, the chemical shift and
the slow exchange rate of Lactam-NH are indicative of
a hydrogen-bonded proton that can form a c-turn cen-
tered on aspartic residue. Protons Arg-NH and Asp-
NH are solvent exposed; other significant NOE contacts
are between Asp-NH/Lactam-NH, Asp-NH/Gly-Ha1,
and Asp-NH/Gly-Ha2 (Fig. 8). It is interesting to ob-
serve that the NOE contacts and the conformational
preferences of compound 5b are similar to those of ana-
log pseudopeptide 5a.
The middle- and long-range NOE cross-peaks observed
in NOESY spectrum are summarized in Figure 7. A
single conformation cannot account for all of these
long-range NOEs; therefore, the NOESY data provide
further evidence of multiple folded conformations. The
NOE between Arg-NH and Gly-NH could arise from
conformation SIV (Fig. 3). The NOE between Asp-
NH and Lactam-NH is consistent with the c-turn-like
folding pattern SII (Fig. 3).
Compound 5a, like compound 2a, is more prone to
aggregation so the spectra are recorded in 2 mM CDCl3
solution. In this compound, the Arg residue is above the
plane of the bicyclic lactam, while the R configuration of
the stereocenter in position 3 orients the aspartic residue
below the plane. So, Arg-NH cannot form a 10-mem-
bered ring hydrogen bonding with C@O(Asp). The
chemical shift value of the amide protons Gly-NH
(7.61 ppm) and Lactam-NH (7.49 ppm), and the tem-
perature coefficients (DdNH/DT = ꢀ4.2 and ꢀ1.2 ppb/
K, respectively) indicate that these protons are locked
in an intramolecularly H-bonded state. Lactam-NH,
inside to the cyclic pentapeptide, can form a c-turn cen-
tered on aspartic residue. Protons Arg-NH and Asp-NH
(dNH = 6.99 and 7.62 ppm, respectively; DdNH/
3. Discussion and conclusions
The crystal structure of the extracellular segment of inte-
grin aVb3 in complex with the cyclic pentapeptide ligand
EMD121974 c(Arg-Gly-Asp-D-Phe-[NMe]Val) has
been reported.34 This structure provides the exact con-
DT = ꢀ12.0 ppb/K) are in equilibrium between
a
nonH-bonded and a H-bonded state. The NOESY spec-
trum shows the following cross-peaks: Asp-NH/Lactam-
NH, Gly-NH/Arg-NH, Arg-NH/H6 (weak), Arg-NH/
H7 and Arg-NH/Arg-Ha strong (Fig. 8). These data
provide further evidence for multiple folded conforma-
tions in equilibrium (type SII, SIII, and SIV geometries).
Table 7. Chemical shifts (ppm) of NH protons of compounds 5a in
CDCl3 and 5b in H2O/D2O (9:1) solution
NH
5a
1H (CDCl3)
5b
1H (H2O)
Down-field shift (ppm)
Arg
Gly
6.99
7.61
7.62
7.49
8.22
7.65
8.92
7.95
1.23
0.04
1.3
Amide proton–deuterium exchange rates provide infor-
mation regarding the possible participation of an amide
proton in a stable intramolecular H-bond within a sec-
Asp
Lactam
0.54

L. Belvisi et al. / Bioorg. Med. Chem. 14 (2006) 169–180
177
formation of EMD121974 bound to aVb3 integrin and
can serve as a basis for understanding the general mode
of interaction of integrins with other RGD-containing
ligands.
According to the results provided by spectroscopic and
computational studies, a strong dependence of the cyclo-
peptide conformations on lactam ring size and stereo-
chemistry has been observed. The effects of the
structural constraint introduced by the bicyclic template
on the conformation of the RGD sequence are mainly
dictated by the turn mimetic properties of the scaffold.32
Examination of the three-dimensional structure of the
cyclic pentapeptide antagonist EMD121974 bound to
the integrin aVb3 (Protein Data Bank, entry 1L5G) re-
veals a conformation characterized by an inverse c-turn
with Asp at position (i + 1) and by a distorted bII0-turn
with Gly and Asp at the (i + 1) and (i + 2) positions
(Fig. 9). A distance between the Cb atoms of Asp and
Compounds 2b and 5b, containing the poor b-turn
inducers bicyclic lactams 5,6 trans 3R and 5,7 trans
3R, respectively, show preferred cyclopeptide geometries
featuring a lightly kinked or an almost extended confor-
mation of the RGD sequence. Only type SIII (inverse c-
turn at Asp and distorted bII0-turn at Gly-Asp) and type
SIV (bI-turn at Pro-Arg and inverse c-turn at Asp)
geometries can be detected among the conformers with-
in 3 kcal/mol of the global minimum calculated for the
simplified AGA cyclopeptide 2c (Table 4). According
to the H-bond analysis, these turns are populated by
40% (inverse c-turn at Asp), 20% (distorted bII0-turn
at Gly-Asp), and 26% (bI-turn at Pro-Arg) during the
10 ns MC/SD simulation of 2b (Table 5). The average
˚
Arg of 8.9 A is observed in this pentapeptide bound con-
formation. Contrary to what had been assumed previ-
ously,23 the inhibition of aVb3 integrin does not
require a strong kink of the RGD sequence. The type
SIII geometry obtained by computational methods for
the cyclopentapeptide mimics (see above, Fig. 3 and Ta-
ble 4) is very similar to the X-ray binding conformation
of EMD121974, in terms of backbone arrangement,
intramolecular H-bonds, and Cb–Cb distance.
˚
With the aim of identifying highly selective integrin
antagonists, we synthesized and tested a small library
of cyclic pseudopeptides in which the triad arginine, gly-
cine, and aspartic acid (RGD) was attached to (5,6)- or
(5,7)-fused bicyclic lactams with different configuration
at two stereocenters. The use of rigid peptidomimetic
scaffolds and conformationally constrained cyclic pep-
tides that match biologically active conformation might
enhance ligand binding for entropic reasons.
of the Cb(Arg)–Cb(Asp) distance is 8.8 A over the same
trajectory. The chemical shifts and the temperature
dependence of amide protons Gly-NH and Lactam-
NH of the protected cyclopeptide 2a (Table 6), as well
as NOE data (Fig. 5) and the slow exchange rate of Lac-
tam-NH of 2b in D2O, provide evidence for the impor-
tance of the b-hairpin-like conformation of type SIV.
The lowest energy conformer of the simplified AGA
cyclopeptide 5c is characterized by an inverse c-turn at
Asp and by a distorted bII0-turn at Gly-Asp (type SIII
geometry). Type SII and SIV geometries can be also
detected among the conformers within 3 kcal/mol of
the global minimum (Table 4). NMR data of compound
5a provide evidence for different folded conformations
in equilibrium, showing, in particular, the involvement
of Lactam-NH in strong intramolecular hydrogen bonds
(type SIII and SIV geometries). Compound 5b in H2O
solution keeps the conformational preferences of the
protected analog 5a. Again Gly-NH and Lactam-NH
are inner to the pentapeptide ring and one of the
preferred conformations is stabilized by a b-turn and
an inverse c-turn (type SIV geometry). According to
Figure 9. X-ray, aVb3-bound conformation of EMD121974 from
1L5G.34
A
B
Type SIV
Type SIII
Figure 10. Conformations of 5b sampled during the 10 ns MC/SD simulation after energy minimization. (A) Type SIII, inverse c (Asp)/distorted
bII0(Gly-Asp). (B) Type SIV, distorted bI(Pro-Arg)/inverse c(Asp).

178
L. Belvisi et al. / Bioorg. Med. Chem. 14 (2006) 169–180
the H-bond analysis, the inverse c-turn at Asp (33%)
and the distorted bII0-turn at Gly-Asp (50%) and bI-
turn at Pro-Arg (13%) are mainly populated during
the 10 ns MC/SD simulation of 5b (Table 5). The aver-
1). Docking studies40 are currently in progress to gain
deeper insights into ligand–receptor interactions of cyc-
lic RGD pentapeptide mimics containing azabicycloal-
kane amino acids.
˚
age of the Cb(Arg)–Cb(Asp) distance is 8.5 A over the
same trajectory. Remarkably, the 5,7 trans 3R bicyclic
scaffold of compound 5b appears to force the cyclopep-
tide to assume preferred conformations very similar to
the X-ray aVb3-bound conformation of EMD121974.
Energy-minimized conformations of 5b obtained from
frames featuring the binding requirements (type SIII
geometry) or the SIV geometry and corresponding to
the most populated geometries of the 10 ns MC/SD tra-
jectory are shown in Figure 10. The root-mean-square
(RMS) deviation in the rigid superimposition between
the type SIII conformation of 5b (Fig. 10A) and the
4. Materials and methods
4.1. Cell cultures
Primary cultures of bovine microvascular endothelial
cells (BMECs) were obtained from bovine adrenal
glands as described by Folkman.41 BMECs were main-
tained in DMEM supplemented with 20% fetal calf ser-
um (FCS), 50 U/ml heparin (Sigma, St. Louis, MO),
50 lg/ml bovine brain extract, and 100 U/ml gentamicin.
Human umbilical vein endothelial cells (HUVECs)
and human microvascular dermal endothelial cells
(HMECs) were obtained from BioWhittaker (Walkers-
ville, MD) and grown in EGM-2 (BioWhittaker).
˚
X-ray structure of bound EMD121974 is 0.16 A for
the backbone atoms of the RGD sequence. Binding of
5b might actually be enhanced by the high structural
preorganization.
Compounds 2b and 5b showed the highest affinities to
aVb3 and inhibited radiolabeled echistatin binding to
aVb3 with an IC50 of 14.3 4.7 and 3.8 0.9 nM,
respectively. The conformational studies described
above suggested that the type SIII and SIV geometries
are the conformations mainly contributing to the con-
formational equilibria of cyclopeptides 2b and 5b. Both
these geometries feature a minimal kink of the RGD
sequence and an orientation of the Asp NH group
suitable to maintain the same electrostatic and hydrogen
bond ligand–receptor interactions observed in the
crystalline complex of EMD121974 with aVb3.
4.2. Solid-phase receptor-binding assay
The receptor-binding assays were performed as de-
scribed previously.42,43 Purified receptors avb3 and
avb5 (Chemicon International Inc., Temecula, CA) were
diluted, respectively, to 500 ng/ml and 1000 ng/well in
coating buffer [20 mM Tris–HCl (pH 7.4), 150 mM
NaCl, 2 mM CaCl2, 1 mM MgCl2, and 1 mM MnCl2],
whereas a5b1 was diluted to 1000 ng/ml in 20 mM Tris–
HCl (pH 7.4), 150 mM NaCl, and 1 mM MnCl2. An ali-
quot of the diluted receptors (100 ll/well) was added to
96-well microtiter plates and incubated overnight at
4 ꢁC. The coating solution was removed by aspiration,
and 200 ll of blocking solution (coating buffer contain-
ing 1% bovine serum albumin (BSA)) was added to the
wells, which were incubated for an additional 2 h at
room temperature. After incubation, the plates were
rinsed with 200 ll of blocking solution (3·) and incubat-
ed with appropriate radiolabeled ligands for 3 h at room
temperature. 0.05 and 0.1 nM [125I]echistatin (Amer-
sham Pharmacia) were used, respectively, for avb3 and
avb5, whereas 20 nM [125I]fibrinogen (Amersham Phar-
macia) was used for a5b1 receptor-binding assay. After
incubation, the plates were sealed and counted in the
c-counter (Packard). Each data point is the result of
the average of triplicate well, and was analyzed by non-
linear regression analysis with the Allfit program.
Conformational studies of compounds 1b, 3b, and 4b, or
related derivatives, containing efficient reverse-turn and
b-turn inducer bicyclic lactams,38 provide evidence for
multiple folded conformations in equilibrium. In partic-
ular, the contribution of the SI structural type to these
conformational equilibria is worth noting. This geome-
try features a strong kink of the RGD motif as a conse-
quence of the bII0/c-turn arrangement with the bicyclic
template in the i + 1 and i + 2 positions of the bII0-turn
and the Gly residue in the i + 1 position of the c-turn at
the opposite side. According to the corresponding
Arg-NHꢁ ꢁ ꢁO@C-Asp H-bond, the bII0-turn at bicyclic
lactam is significantly populated during the MC/SD
simulations of 1b, 3b, and 4b (Table 5). The presence
in conformational equilibria of geometries featuring
kinked backbone conformations of the RGD sequence,
short Cb(Arg)–Cb(Asp) distances, and inside orienta-
tion of the amide proton Asp-NH (c-turn at Gly) might
be responsible for reduced structural preorganization
for binding and therefore for lower affinities to aVb3
(Table 1).
4.3. Platelet aggregation assay
Blood samples were collected in guinea pigs via cardiac
puncture in 3.8% sodium citrate at a final dilution of 1–
10. Platelet-rich plasma (PRP) was prepared by centrifu-
gation at 120g for 10 min at room temperature. Platelet
aggregation was assayed following the addition of
TRAP 42-55 (Sigma), 25–100 lM, to PRP at 37 ꢁC by
light transmission (4 channel aggregometer, PACKS-4,
Helena Laboratories, Beaumont, Texas).
Computational studies of 6b–7b and 6c–7c derivatives
suggest that both the poor turn inducer bicylic scaffolds
5,7 cis force the cyclopeptide to assume mainly the type
SIII and SIV geometries. In spite of the proper orienta-
tion thus achieved by the pharmacophoric groups re-
quired for binding to the aVb3 integrin, compounds 6b
and 7b showed low affinities toward this receptor (Table
Vehicle or peptide solutions at different concentrations
were added to aliquots of the same PRP, two minutes
prior to TRAP addition. The extent of platelet aggrega-

L. Belvisi et al. / Bioorg. Med. Chem. 14 (2006) 169–180
179
tion was quantified as the maximum change in light
transmission within 4 min after the addiction of agonist.
RGD side-chain dihedral angles were defined as internal
coordinate degrees of freedom in the Monte Carlo part
of the algorithm. A time step of 1 fs was used for the sto-
chastic dynamics (SD) part of the algorithm. At least
two 10 ns simulations were run for each RGD cyclopep-
tide starting from different conformations to test the
convergence. Samples were taken at 2 ps intervals dur-
ing each simulation, yielding 5000 conformations for
analysis.
The results were plotted (Allfit program) and expressed
as the antagonist concentration that inhibited 50% of
platelet aggregation.
4.4. Cell-adhesion assay
Ninety-six-well plates were coated with either fibronec-
tin (Sigma, St. Louis, MO) or vitronectin (Calbiochem,
San Diego, CA) (both at 5 lg/ml in phosphate-buffered
saline) overnight at 4 ꢁC. Approximately 4–5 · 104 cells/
100 ll were seeded in each well and allowed to adhere
for 1–3 h at 37 ꢁC in the presence of various concentra-
tions of RGD peptides. Nonadherent cells were re-
moved with PBS and the remaining cells were fixed
with 4% paraformaldehyde for 10 min. Adherent cells
were stained with 1% toluidine blue for 10 min and
rinsed with water. Stained cells were solubilized with
1% SDS and quantified on a microtiter plate reader at
600 nm (Victor2, Wallac). Experiments were performed
in quadruplicate and repeated at least three times. Re-
sults are expressed as mean compound concentra-
tion SEM that inhibited 50% of cell adhesion.
4.6. NMR spectroscopy
All the spectra were acquired at a 400 MHz Bruker spec-
trometer equipped with pulsed field gradients. The fol-
lowing experiments were carried out (at 300 K):
TOCSY, NOESY (mixing time 200, 400, and 600 ms),
and HSQC.
The protected RGD-pentapeptides 1a–5a (Fig. 2) were
analyzed by NMR spectroscopy in CDCl3 solution.
1
We first performed H NMR analysis to determine the
lowest concentration at which intermolecular hydrogen
bonding occurs.51 The data reported in this paper were
obtained from samples at 300 K and 1–5 mM concentra-
tions, at which aggregation was not significant. Traces
of acid were removed from CDCl3 by passing it through
an alumina column. Amide hydrogen temperature coef-
ficients were measured from 1D experiments carried out
in the range 240–320 K. The cyclopentapeptide mimics
1b–5b (Fig. 2) were analyzed by NMR spectroscopy in
D2O solution; compound 5b was also studied in H2O/
D2O (9:1) solution.
4.5. Computational studies
Conformational preferences of the RGD cyclopeptides
have been investigated by molecular mechanics44 calcu-
lations within the framework of MacroModel45 version
5.5, using the MacroModel implementation of the Am-
ber all-atom force field46 (denoted AMBER*) and the
implicit water GB/SA solvation model of Still et al.47
A two-step protocol was used.
Acknowledgments
Monte Carlo/energy minimization (MC/EM) conforma-
tional searches48 of the AGA (Ala-Gly-Ala) cyclopep-
tide analogs containing methyl groups instead of the
Arg and Asp side chains were performed as the first step.
The torsional space of each AGA cyclopeptide was
randomly varied with the usage-directed Monte Carlo
conformational search of Chang, Guida, and Still.48
Ring-closure bonds were defined in the six- and seven-
membered rings of the 5,6- and 5,7-fused bicyclic lac-
tams, respectively, and in the cyclopeptide ring. Amide
bonds were included among the rotatable bonds. For
each search, at least 1000 starting structures for each
variable torsion angle were generated and minimized un-
The authors thank CNR and MIUR (COFIN and
FIRB research programs) for financial support and
CILEA for computing facilities.
Supplementary data
Description of the synthesis and experimental data for
the preparation of compound 5b. Tables S1–S6 listing
the 1H and 13C NMR chemical shifts d (ppm) of pseudo-
pentapeptides 1a–7a and 1b–7b. Supplementary data
associated with this article can be found in the online
version at doi:10.1016/j.bmc.2005.08.048.
˚
til the gradient was less than 0.05 kJ/Amol using the
truncated Newton–Raphson method49 implemented in
MacroModel. Duplicate conformations and those with
an energy greater than 6 kcal/mol above the global min-
imum were discarded. The nature of the stationary
points individuated was tested by computing the eigen-
values of the second-derivative matrix.
References and notes
1. Gumbiner, B. M. Cell 1996, 84, 345.
2. Aplin, A. E.; Howe, A.; Alahari, S. K.; Juliano, R. L.
Pharmacol. Rev. 1998, 50, 197.
3. Eliceiri, B. P.; Cheresh, D. A. Curr. Opin. Cell Biol. 2001,
13, 563.
4. Hood, J. D.; Cheresh, D. A. Nat. Rev. Cancer 2002, 2, 91.
5. Howe, A.; Aplin, A. E.; Alahari, S. K.; Juliano, R. L.
Curr. Opin. Cell Biol. 1998, 10, 220.
6. Hynes, R. O. Cell 2002, 110, 673.
7. Hynes, R. O. Cell 1987, 48, 549.
8. Giancotti, F. G.; Ruoslahti, E. Science 1999, 285, 1028.
Free simulations of the complete RGD cyclic peptides
(Asp and Arg side chains were considered ionized) were
then performed at 300 K using the metropolis Monte
Carlo/stochastic dynamics (MC/SD) hybrid simulation
algorithm,50 starting from the cyclopeptide backbone
geometries located by the previous MC/EM step.

180
L. Belvisi et al. / Bioorg. Med. Chem. 14 (2006) 169–180
9. Ruoslahti, E.; Pierschbacher, M. D. Science 1987, 238,
491.
10. Ruoslahti, E. Ann. Rev. Cell Dev. Biol. 1996, 12, 697.
11. Plow, E. F.; Haas, T. A.; Zhang, L.; Loftus, J.; Smith, J.
W. J. Biol. Chem. 2000, 275, 21785.
12. Suehiro, K.; Smith, J. W.; Plow, E. F. J. Biol. Chem. 1996,
271, 10365.
13. Eliceiri, B. P.; Cheresh, D. A. J. Clin. Invest. 1999, 103,
1227.
carbon C3. They are classified according to the C3
configuration as (3S) or (3R) bicycles. For convenience,
to indicate the bridgehead configuration, the cis and trans
descriptors are used, depending on the relative position of
the hydrogen atoms at the bridgehead and the Pro Ca
carbon.
34. Xiong, J.-P.; Stehle, T.; Zhang, R.; Joachimiak, A.; Frech,
M.; Goodman, S. L.; Arnout, M. A. Science 2002, 296,
151.
14. Brooks, P. C.; Clark, R. A. F.; Cheresh, D. A. Science
1994, 264, 569.
15. Brooks, P. C.; Stromblad, S.; Klemke, R.; Visscher, D.;
Sarkar, F. H.; Cheresh, D. A. J. Clin. Invest. 1995, 96,
1815.
16. Kumar, C. C.; Malkowski, M.; Yin, Z.; Tanghetti, E.;
Yaremko, B.; Nechuta, T.; Varner, J.; Liu, M.; Smith, E.
M.; Neustadt, B.; Presta, M.; Armstrong, L. Cancer Res.
2001, 61, 2232.
17. Friedlander, M.; Brooks, P. C.; Shaffer, R. W.; Kincaid, C.
M.; Varner, J. A.; Cheresh, D. A. Science 1995, 270, 1500.
18. Mitjans, F.; Meyer, T.; Fittschen, C.; Goodman, S.;
Jonczyk, A.; Marshall, J. F.; Reyes, G.; Piulats, J. Int. J.
Cancer 2000, 87, 716.
19. MacDonald, T. J.; Taga, T.; Shimada, H.; Tabrizi, P.;
Zlokovic, B. V.; Cheresh, D. A.; Laug, W. E. Neurosur-
gery 2001, 48, 151.
20. Scarborough, R. M.; Naughton, M. A.; Teng, W. E.; Rose,
J. W.; Phillips, D. R.; Nannizzi, L.; Arfsten, A.; Campbell,
A. M.; Charo, I. F. J. Biol. Chem. 1993, 268, 1066.
21. Bach, A. C., II; Espina, J. R.; Jackson, S. A.; Stouten,
P. F. W.; Duke, J. L.; Mousa, S. A.; DeGrado, W. F.
J. Am. Chem. Soc. 1996, 118, 293.
35. The use of [125I]echistatin as radiolabeled ligand in
integrin receptor-binding assays toward aVb3 and aVb5
has been reported in previous literature Ref. 16,42
showing data consistent with our results. However,
binding data and selectivity may differ significantly when
alternate ligand and/or binding assay is used. For exam-
ple, EMD121974 is reported to be 10-fold more selective
for aVb3 in binding assay using vitronectin as the ligand.52
36. The large-scale synthetic approach to 5b is reported in the
supporting material.
37. Rose, G. D.; Gierasch, L. M.; Smith, J. A. Adv. Prot.
Chem. 1985, 37, 1–109.
38. Previous computational studies32 suggested the following
trend in b-turn inducing properties of bicyclic lactams: 5,6
trans S > 5,7 trans S > 5,6 cis S.
39. In solvent such as CDCl3, NH amide protons that are
involved in hydrogen bonding resonate around d 6.5–
8 ppm.
40. Docking studies of selected peptidic, pseudopeptidic, and
nonpeptidic ligands into the avb3 and avb5 integrin
binding sites have been recently reported. (a) Marinelli,
L.; Lavecchia, A.; Gottschalk, K.-E.; Novellino, E.;
Kessler, H. J. Med. Chem. 2003, 46, 4393; (b) Marinelli,
L.; Gottschalk, K.-E.; Meyer, A.; Novellino, E.; Kessler,
H. J. Med. Chem. 2004, 47, 4166; (c) Moitessier, N.;
Henry, C.; Maigret, B.; Chapleur, Y. J. Med. Chem. 2004,
47, 4178.
22. Muller, G.; Gurrath, M.; Kessler, H. J. Comput.-Aided
¨
Mol. Des. 1994, 8, 709.
23. Haubner, R.; Gratias, R.; Diefenbach, B.; Goodman, S. L.;
Jonczyk, A.; Kessler, H. J. Am. Chem. Soc. 1996, 118, 7461.
24. Haubner, R.; Finsinger, D.; Kessler, H. Angew. Chem. Int.
Ed. Engl. 1997, 36, 1374.
41. Folkman, J.; Haudenschild, C. C.; Zetter, B. R. Proc.
Natl. Acad. Sci. U.S.A. 1979, 76, 5217.
25. Wermuth, J.; Goodman, S. L.; Jonczyk, A.; Kessler, H.
J. Am. Chem. Soc. 1997, 119, 1328.
26. Lohof, E.; Planker, E.; Mang, C.; Burkhart, F.; Dechants-
reiter, M. A.; Haubner, R.; Wester, H.-J.; Schwaiger, M.;
Ho¨lzemann, G.; Goodman, S. L.; Kessler, H. Angew.
Chem. Int. Ed. Engl. 2000, 39, 2761.
42. Kumar, C. C.; Nie, H.; Rogers, C. P.; Malkowski, M.;
Maxwell, E.; Catino, J. J.; Armstrong, L. J. Pharmacol.
Exp. Ther. 1997, 283, 843.
43. Suehiro, K.; Gailit, J.; Plow, E. F. J. Biol. Chem. 1997,
272, 5360.
44. Burkert U.; Allinger N. L. Molecular Mechanics. ACS
Monograph 177, American Chemical Society, Washing-
ton, DC, 1982.
45. Mohamadi, F.; Richards, N. G. J.; Guida, W. C.;
Liskamp, R.; Lipton, M.; Caufield, C.; Chang, G.;
Hendrickson, T.; Still, W. C. J. Comput. Chem. 1990, 11,
440.
27. Schumann, F.; Muller, A.; Koksch, M.; Muller, G.;
¨
¨
Sewald, N. J. Am. Chem. Soc. 2000, 122, 12009.
28. Haubner, R.; Schmitt, W.; Ho¨lzemann, G.; Goodman, S.
L.; Jonczyk, A.; Kessler, H. J. Am. Chem. Soc. 1996, 118,
7881.
29. Dechantsreiter, M. A.; Planker, E.; Matha¨, B.; Lohof, E.;
Ho¨lzemann, G.; Jonczyk, A.; Goodman, S. L.; Kessler, H.
J. Med. Chem. 1999, 42, 3033.
46. Weiner, S. J.; Kollman, P. A.; Nguyen, D. T.; Case, D. A.
J. Comput. Chem. 1986, 7, 230.
30. Belvisi, L.; Bernardi, A.; Checchia, A.; Manzoni, L.;
Potenza, D.; Scolastico, C.; Castorina, M.; Cupelli, A.;
Giannini, G.; Carminati, P.; Pisano, C. Org. Lett. 2001, 3,
1001.
47. Still, W. C.; Tempczyk, A.; Hawley, R. C.; Hendrickson,
T. J. Am. Chem. Soc. 1990, 112, 6127.
48. Chang, G.; Guida, W. C.; Still, W. C. J. Am. Chem. Soc.
1989, 111, 4379.
31. Angiolini, M.; Araneo, S.; Belvisi, L.; Cesarotti, E.;
Checchia, A.; Crippa, L.; Manzoni, L.; Scolastico, C.
Eur. J. Org. Chem. 2000, 2571.
32. Belvisi, L.; Bernardi, A.; Manzoni, L.; Potenza, D.;
Scolastico, C. Eur. J. Org. Chem. 2000, 2563.
33. The bicyclic lactams incorporate a natural Ca(S) proline
residue, but vary in the lactam ring size (6 or 7), and in the
stereochemistry at the bridgehead and at the N-bearing
49. Ponder, J. W.; Richards, F. M. J. Comput. Chem. 1987, 8,
1016.
50. Guarnieri, F.; Still, W. C. J. Comput. Chem. 1994, 15,
1302.
51. Gardner, R. R.; Liang, G.-B.; Gellman, S. H. J. Am.
Chem. Soc. 1999, 121, 1806.
52. Goodman, S. L.; Ho¨lzemann, G.; Sulyok, G. A. G.;
Kessler, H. J. Med. Chem. 2002, 45, 1045.