Probing integrin selectivity: rational design of highly active and selective ligands for the α5β1 and αvβ3 integrin receptor

Article abstract of DOI:10.1002/anie.200700008

(Chemical Equation Presented) Try and fit in: A strategy for the rational design of α5β1 ligands for the purpose of lead generation and biochemical studies on integrin selectivity is based α5β1 homology modeling. Ligand 1 can bind α5β1 with activities in

Full text of DOI:10.1002/anie.200700008

Angewandte
Chemie
DOI: 10.1002/anie.200700008
Integrin Ligands
Probing Integrin Selectivity: Rational Design of Highly Active and
Selective Ligands for the a5b1 and avb3 Integrin Receptor**
Dominik Heckmann, Axel Meyer, Luciana Marinelli, Grit Zahn, Roland Stragies, and
Horst Kessler*
[
7]
Rational drug design relies on an iterative procedure of initial
protein–structure determination, followed by the design,
chemical synthesis, and subsequent biological evaluations of
specific compounds. However, there is still a large gap
between known protein sequences and 3D structures. To
date, the most successful theoretical approach to bridge this
gap is homology modeling. It is possible to construct an
approximate 3D model of the structural unknown protein if
the sequence homology to the known 3D structure of the
reference protein is higher than 40%. Such a homology-
clinical trials. As a result of our research, the cyclic N-
[
5b]
methylated pentapeptide c(-RGDf[NMe]V-),
cilengitide, has entered phase II trials for patients with
glioblastoma.
Recent knock-out experiments showed, however, that
genetically altered mice (lacking the av integrin) show
extensive angiogenesis in some cases, whereas other mice
(lacking the b3 or b5 integrins) show no significant effects,
and as such, the idea that these two integrins are proangio-
known as
[
8,9]
genic was seriously questioned.
proangiogenic function of the a5b1 receptor has been clearly
On the other hand, the
[
1]
modeled structure is suitable for rational drug design.
Herein we describe the successful use of our recently
[
10,11]
demonstrated
so that the a5b1 integrin moved into the
[
2]
published homology model of the integrin a5b1 to design
potent (with activities up to the subnanomolar range) and
selective ligands for the two highly similar integrin receptors
a5b1 and avb3. Structural considerations were used to trigger
potency and selectivity in both directions. These ligands could
allow functional studies in vivo of the role of these two
integrin subtypes and might be used as lead structures for
antiangiogenic cancer therapy.
focus of research. Although crystal structures of the extra-
cellular domains of the avb3 and aIIbb3 integrins have been
solved and provided a deep insight into the ligand bind-
ing, very little detailed structural information about the
a5b1 receptor itself or about ligand–receptor interactions
have been obtained until now.
Furthermore, there are only a few small-molecule ligands
known to bind a5b1,
[
12,13]
[
14]
[
15,16]
which prompted us to focus our
Integrins constitute an important class of heterodimeric
cell-adhesion receptors that are involved in many severe
pathological processes, such as tumor metastasis, thrombosis,
research on this integrin subtype. A first hint for the design of
new a5b1 ligands came from our homology model of the a5b1
integrin in complex with
a
recently reported ligand
[
3]
[2,16]
inflammation, and osteoporosis. Therefore, they have been
(SJ749).
The high sequence similarity between the avb3
[
4]
attractive therapeutic targets for several years. Since Brooks
et al. reported that various low-molecular-weight ligands (for
example, our synthesized cyclopentapeptide cyclo(-Arg-Gly-
and a5b1 receptors (av:a5, 53% identity; b3:b1, 55%
identity in the integrinꢀs headgroup) makes it possible to
test our hypothetical model by synthesizing a series of
rationally designed compounds. Like other ligands targeting
the RGD-binding site of integrins, our compounds possess a
free a carboxylate function as well as a basic moiety at a
distance of about 13 Š. There are two “hot spots” in the
binding pockets in which mutations suitable for achieving
selectivity between the a5b1 and avb3 integrin can be found:
in the b subunit, (b3)-Arg214 is replaced by (b1)-Gly217 and
additionally (b3)-Arg216 is mutated into (b1)-Leu219
(Figure 1 highlights the important mutations). The substitu-
tion of both arginine residues expands this site of the a5b1
binding pocket, which, in comparison with the avb3 integrin,
allows the introduction of bulky moieties into the ligandꢀs
core structure. Secondly, the a5 subunit turned out to be less
acidic owing to the mutation of (av)-Asp150 to (a5)-Ala159.
Furthermore, the replacement of (av)-Thr212 by (a5)-
Gln221 results in a different shape of this binding region,
which offers the opportunity to gain selectivity by modifica-
tion of the basic moieties.
[
5a]
Asp-D-Phe-Val-) = c(-RGDfV-) ), which are recognized by
the avb3 and avb5 integrins, block angiogenesis in response
[
5b]
to growth factors in tumors, many selective avb3-ligands
have been developed and some compounds have reached
[*] Dipl.-Chem. D. Heckmann, Dr. A. Meyer, Prof. Dr. H. Kessler
Department Chemie
TU München
Lichtenbergstrasse 4, 85747 Garching (Germany)
Fax: (+49)89-2891-3210
E-mail: kessler@ch.tum.de
Dr. L. Marinelli
Dipartimento di Chimica Farmaceutica e Tossicologica
Università di Napoli “Federico II”
Via D. Montesano, 49-80131 Napoli (Italy)
Dr. G. Zahn, Dr. R. Stragies
Jerini AG
Invalidenstrasse 130, 10115 Berlin (Germany)
[
**] The authors gratefullyacknowledge financial support bythe
Deutsche Forschungsgemeinschaft (SFB 563) and technical assis-
tance byM. Wolff, B. Cordes, M. Kranawetter, and G. Clever.
Guided by these observations, we synthesized a series of
compounds based on a tyrosine scaffold that has already been
[
17]
successfully employed in the integrin field. Putative binding
modes were determined by using the AutoDock program (for
Supporting information for this article is available on the WWW
under http://www.angewandte.org or from the author.
Angew. Chem. Int. Ed. 2007, 46, 3571 –3574
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
3571

Communications
Figure 1. Ribbon representation of the a5b1 integrin binding pocket
a5 blue, b1 red) with the predicted binding pose of 3b (gray). The
(
side chains of important residues are highlighted and the correspond-
ing residues of the avb3 integrin are shown in yellow and labeled in
parentheses. The MIDAS metal is represented as a magenta sphere.
Figure 2. Superposition of the a5b1 (gray) and avb3 (transparent red)
receptors represented as Connollysurfaces. Compounds 3b and 3c
were docked in the a5b1 integrin. The mesitylene function of the
a5b1-selective compound 3b would clash with (b3)-Arg214, which is
not present in the a5b1 receptor. Compound 3c shows no selectivity
because of the different position of its bulkymesit yl ene function.
[
18]
computational details see the Supporting Information)
whose reliability at predicting ligand–receptor complexes
has been demonstrated in prior studies on the avb3 inte-
[
19]
grin.
Figure 1 shows compound 3b in the a5b1 binding pocket
together with the mutated residues of avb3 (yellow). The
selectivity (3b, 3d, and 3 f; Table 1). The former could be
explained by an increasing lipophilicity, the latter by a
restricted flexibility of the aromatic ring. Among all of the
synthesized a-amino acids, 3 f exhibits the best affinity for
a5b1 (I C= 0.7 nm) and good selectivity against avb3
(ꢀ300-fold). The para-isopropyloxy group is placed well for
further interaction with the (b1)-SDL (specificity-determin-
ing loop), presenting an additional hydrogen-bond acceptor
to the serine residue (b1)-Ser171.
2
+
2+
carboxylate group of 3b coordinates the metal (Ca or Mn )
in the MIDAS (metal-ion-dependent adhesion site) region,
which resides in the b1 subunit, while the basic amino
pyridine inserts into a narrow groove at the top of the (a5)-
b-propeller domain, forming hydrogen bonds to the highly
conserved (a5)-Asp227. The tyrosine scaffold enables p–p
interaction with (a5)-Phe187 while the mesitylene function
interacts with (b1)-Tyr127 in the same manner. The a5b1
selectivity of 3b is determined by the bulky mesitylene group,
which cannot be placed at this position in the avb3 pocket
because of a steric clash with (b3)-Arg214. The conforma-
tional difference between the sulfonamide (substituents are
5
0
In addition to the a-tyrosine ligands, we synthesized a
compound series based on a b-amino acid scaffold. Com-
pound 6e shows moderate affinity towards a5b1, but high
affinity towards the avb3 receptor. Considering that various
known selective avb3 ligands are substituted in some way on
[
20,21]
9
08 twisted about the SO ÀN bond) in comparison with the
the b-position to the carboxylate,
it is not surprising that
2
planar amide bond causes the mesitylene group in compound
this substitution pattern shows high affinity for the avb3
integrin and only average affinity for a5b1. In general, all
synthesized b-amino acid derivatives exhibit lower a5b1
affinity than the corresponding a-amino acids. As the bulky
side chain is more flexible when attached at the b-position,
this selectivity might not be caused (at least not only) by steric
interaction with (b3)-Arg214 or adjacent residues. A closer
look at the a5 subunit reveals two major differences when
compared with the av subunit (see Figure 1): first, the (a5)-
Ala159 is replaced by the (av)-Asp150, which favors an extra
hydrogen donor on the basic group opposite to the one
interacting with (av)-Asp218. Secondly, the mutation of (av)-
Thr212 to (a5)-Gln221, which shortens this region of the
binding pocket, causes the slightly longer b-amino acids to
better fit into the avb3 binding site.
3
c to fold back towards the a subunit (see Figure 2). This
position is allowed for both integrins and, hence, the
selectivity for a5b1 is lost, although 3c still shows nanomolar
affinity for both integrins.
It could be argued that there is no need to put the bulky
mesitylene group exactly in this position. Docking calcula-
tions show alternative binding modes in which the hydro-
phobic residue sticks out of the integrin. This, however, would
expose the mesitylene to the surrounding water and thus
result in a decreased affinity. Attachment of a hydrophobic
moiety adjacent to the carboxylate function is of particular
importance for integrin binding, which has been demon-
[
20]
strated in prior work. In the case of 3b, the mesitylene
moiety enables a p–p interaction with (b1)-Tyr127 and forms
a hydrogen bond with the backbone carbonyl group of (b1)-
Asn218.
To clearly demonstrate that the smaller a5 binding pocket
can be used to gain avb3 selectivity, we attached a methyl
group to all possible positions of the 2-aminopyridine ring
(Table 1). As expected, the 4-methylaminopyridine in 6c
The ortho substitution pattern of the attached aromatic
residue seems to be important for affinity as well as for
3
572
www.angewandte.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2007, 46, 3571 –3574

Angewandte
Chemie
Table 1: IC50 values of integrin ligands on a5b1 and avb3.
towards the platelet integrin aIIbb3, which is crucial for
developing leads for antiangiogenic cancer therapy.
The small compound library (Table 1) was produced by a
Mitsunobu reaction of protected of a- or b-tyrosine esters
with several basic aminopyridinyl alcohols. The aminopyr-
idinyl alcohol 1 was synthesized by nucleophilic substitution
of the commercially available 2-bromopyridine in neat 3-
[
a]
[a]
EntryCompound
R
IC
a5b1
[nm]
0
IC50 [nm]
avb3
5
[23]
aminopropanol at 1508C in 95% yield (Scheme 1). The
1
2
3
4
5
3a
3b
3c
3d
3e
243
190
703
3.4
2.5
46
3.1
1624
318
416
6
3 f
0.7
279
7
8
6a
6b
3946
215
13
2.5
9
6c
67
0.9
1
0
1
6d
6e
6969
264
245
1.2
Scheme 1. Synthesis of ligands 3a–f: a) 3-aminopropanol, 1508C,
1
2 h; b) N-Boc-tyrosine methyl ester, PBu , ADDP, THF, 08C, 8 h;
1
3
c) aqueous HCl, dioxane, 1 h; d) benzoyl chloride, NaHCO , dioxane,
3
water, 0.5 h (a); aromatic acid, HATU, DIPEA, DMF, 3 h (b,d,e,f) or
mesitylenesulfonyl chloride, DIPEA, DMF, 8 h (b); e) LiOH, methanol,
[
a] IC50 values are derived from a competitive ELISA test byusing the
immobilized natural integrin ligands fibronectin and vitronectin and the
soluble integrins a5b1 and avb3, respectively(for details, see the
Supporting Information).
H O, HPLC purification. ADDP=azodicarboxylic dipiperidide, Boc=
2
tert-butoxycarbonyl, DIPEA=N,N-diisopropylethylamine, DMF=N,N-
dimethylformamide, HATU=O-(7-azabenzotriazol-1-yl)-N,N,N’,N’-tet-
ramethyluronium hexafluorophosphate.
shows no influence on selectivity because the methyl group
sticks out of the binding pocket. In contrast, the methyl group
at position six of the pyridine ring (6a) has a massive impact
on a5b1 binding affinity. Compound 6d shows decreased
affinity to both the a5b1 and the avb3 integrin. Intra-
molecular steric hindrance causes the pyridine ring to twist
out of the plane, which hampers the formation of a bidentate
salt bridge to the conserved (a5)-Asp227 (or Asp218 in av).
Owing to the high sequence identity between avb3 and avb5
methyl substituted aminopyridinyl alcohols 4a–d were syn-
thesized from the corresponding 2-chloromethylpyridines by
oxidation to the pyridine-N-oxide with meta-chloroperben-
[
24]
zoic acid (MCPBA; 78–89%),
nucleophilic substitution
(> 95%), and reduction under hydrogen atmosphere with Pd
on carbon (60–82%) (Scheme 2). The aminopyridyl alcohols
were coupled to the corresponding N-Boc-protected a- or b-
tyrosine methyl ester through a Mitsunobu reaction with
tributyl phosphine and azodicarboxylic dipiperidid (ADDP)
to give the tyrosine ethers 2 and 5a–e in poor to moderate
[
22]
(b3:b5, 65% identity in the integrinꢀs headgroup),
we
assume that the described modifications might have similar
effects on binding the avb5 integrin. Compound 6a is a potent
ligand for avb3 and has considerably lower activity towards
a5b1. Compounds 3 f and 6a offer the opportunity to
investigate the role of both integrins, a5b1 and avb3, in
biological processes. As expected, both compounds exhibit
low binding affinity (IC > 10 mm for 3e and ꢀ 1 mm for 6a)
[
25]
yields (15–68%).
The yield of the Mitsunobu reaction could be increased by
employing an N-Boc-protected aminopyridyl alcohol; 5c
could actually only be prepared through this method. The
fully protected ligand precursors were Boc-deprotected with
5
0
Angew. Chem. Int. Ed. 2007, 46, 3571 –3574
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
3573

Communications
[
1] A. Hillisch, L. F. Pineda, R. Hilgenfeld, Drug Discovery Today
004, 9, 659.
2
[
2] L. Marinelli, A. Meyer, D. Heckmann, A. Lavecchia, E.
Novellino, H. Kessler, J. Med. Chem. 2005, 48, 4166 – 4204.
3] R. O. Hynes, Cell 2002, 110, 673.
[
[
[
4] S. A. Mousa, Curr. Opin. Chem. Biol. 2002, 6, 534.
5] a) M. Aumailley, M. Gurrath, G. Müller, J. Calvete, R. Timpl, H.
Kessler, FEBS Lett. 1991, 291, 50; b) M. A. Dechantsreiter, E.
Planker, B. Mathä, E. Lohof, G. Hölzemann, A. Jonczyk, S. L.
Goodman, H. Kessler, J. Med. Chem. 1999, 42, 3033.
[
6] P. C. Brooks, A. M. P. Montgomery, M. Rosenfeld, R. A. Reis-
feld, T. Hu, G. Klier, D. A. Cheresh, Cell 1994, 79, 1157.
7] M. Shimaoka, T. A. Springer, Nat. Rev. Drug Discovery 2003, 2,
[
703.
[
8] R. O. Hynes, Nat. Med. 2002, 8, 918.
[
9] B. L. Bader, H. Rayburn, D. Crowley, R. O. Hynes, Cell 1998, 95,
507.
[
[
[
10] E. L. George, E. N. Georges-Labouesse, R. S. Patel-King, H.
Rayburn, R. O. Hynes, Development 1993, 119, 1079.
11] S. Kim, K. Bell, S. A. Mousa, J. A. Varner, Am. J. Pathol. 2000,
156, 1345.
12] a) J. P. Xiong, T. Stehle, R. Zhang, A. Joachimiak, M. Frech, S. L.
Goodman, M. A. Arnaout, Science 2002, 296, 151; b) K.-E.
Gottschalk, H. Kessler Angew. Chem. 2002, 41, 3767; Angew.
Chem. Int. Ed. 2002, 41, 3767.
Scheme 2. Synthesis of ligands 6a–e: a) 4a–e, N-Boc-b-tyrosine methyl
ester, PBu , ADDP, THF, 08C, 8 h; b) aqueous HCl, dioxane, 1 h;
[13] T. Xiao, J. Takagi, B. S. Coller, J. H. Wang, T. A. Springer, Nature
2004, 432, 59.
[14] J. Takagi, K. Strokovich, T. A. Springer, T. Walz, Embo J. 2003,
3
c) PhCOCl, dioxane, H O, NaHCO ; d) LiOH, methanol, H O.
2
3
2
[a] Yield: 4a,b,d: 66–48% over three steps; 4c: 30% over five steps.
22, 4607.
[
15] D. Zimmermann, E. W. Guthöhrlein, M. Malesevic, K. Sewald,
L. Wobbe, C. Heggemann, N. Sewald, ChemBioChem 2005, 6,
aqueous HCl in dioxane, acylated with an activated aromatic
acid (HATU, DIPEA in DMF) in the case of the ligands
272.
[
16] J. M. Smallheer, C. A. Weigelt, F. J. Woerner, J. S. Wells, W. F.
Daneker, S. A. Mousa, R. R. Wexler, P. K. Jadhav, Bioorg. Med.
Chem. Lett. 2004, 14, 383.
[17] G. P. Curley, H. Blum, M. J. Humphries, Cell. Mol. Life Sci. 1999,
56, 427.
18] G. M. Morris, D. S. Goodsell, R. S. Halliday, R. Huey, W. E.
Hart, R. K. Belew, A. J. Olson, J. Comput. Chem. 1998, 19, 1639.
19] L. Marinelli, A. Lavecchia, K. E. Gottschalk, E. Novellino, H.
Kessler, J. Med. Chem. 2003, 46, 4393.
20] C. Gibson, G. A. Sulyok, D. Hahn, S. L. Goodman, G. Hölze-
mann, H. Kessler, Angew. Chem. 2001, 113, 169; Angew. Chem.
Int. Ed. 2001, 40, 165.
[21] J. H. Hutchinson, W. Halczenko, K. M. Brashear, M. J. Breslin,
P. J. Coleman, L. T. Duong, C. Fernandez-Metzler, M. A.
Gentile, J. E. Fisher, G. D. Hartman, J. R. Huff, D. B. Kimmel,
C. T. Leu, R. S. Meissner, K. Merkle, R. Nagy, B. Pennypacker,
J. J. Perkins, T. Prueksaritanont, G. A. Rodan, S. L. Varga, G. A.
Wesolowski, A. E. Zartman, S. B. Rodan, M. E. Duggan, J. Med.
Chem. 2003, 46, 4790.
3
b,d,e,f or acylated with benzoyl chloride and NaHCO in
3
dioxane/water in the case of 6a–e and 3a. The sulfonamide 3c
was produced with mesitylenesulfonyl chloride and DIPEA in
DMF. In the last step, the methyl ester was finally cleaved
with 5 equivalents of LiOH in methanol/water and the
resulting ligands purified by using reverse-phase HPLC
techniques (for details regarding synthesis and compound
characterization see the Supporting Information).
[
[
[
Taking into account the re-evaluated role of the a5b1
integrin in the development of antiangiogenic drugs for
cancer therapy, we herein present the small non-peptidic
molecule 3 f, which selectively binds to the a5b1 integrin in
the subnanomolar range (IC = 0.7 nm). Minor modifications
5
0
of the compounds allow the design of highly active ligands
with good selectivity for avb3 over the a5b1 receptor,
whereas very low affinity towards the platelet integrin
aIIbb3 has been observed. On the basis of a5b1 homology
modeling, analysis of the ligand binding mode, and extensive
data on structure–activity relationships, we proposed a model
suitable for the rational design of selective a5b1 ligands for
the purpose of lead generation and biochemical studies on
integrin selectivity.
[22] L. Marinelli, K. E. Gottschalk, A. Meyer, E. Novellino, H.
Kessler, J. Med. Chem. 2004, 47, 4166.
[
23] A. Trejo, H. Arzeno, M. Browner, S. Chanda, S. Chen, D. D.
Corner, S. A. Dalrymple, P. Dunten, J. Lafague, B. Lovejoy, J.
Freire-Moar, J. Lim, J. Mcintosh, J. Miller, E. Papp, D. Reuter, R.
Roberts, F. Sanpablo, J. Saunders, K. Song, A. Villasenor, S. D.
Warren, M. Welch, P. Weller, P. E. Whiteley, L. Zeng, D. M.
Goldstein, J. Med. Chem. 2003, 46, 4702.
[24] K. C. Lee, D. Y. Chi, J. Org. Chem. 1999, 64, 8576.
[25] J. J. Marugan, C. Manthey, B. Anaclerio, L. Lafrance, T. Lu, T.
Markotan, K. A. Leonard, C. Crysler, S. Eisennagel, M. Das-
gupta, B. Tomczuk, J. Med. Chem. 2005, 48, 926.
Received: January 2, 2007
Published online: March 30, 2007
Keywords: antitumor agents · drug design · integrin ligands ·
.
receptors · structure–activityrelationships
3
574
www.angewandte.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2007, 46, 3571 –3574