Design, synthesis, and functional evaluation of leukocyte function associated antigen-1 antagonists in early and late stages of cancer development

Article abstract of DOI:10.1021/jm3016848

The integrin leukocyte function associated antigen 1 (LFA-1) binds the intercellular adhesion molecule 1 (ICAM-1) by its α_L-chain inserted domain (I-domain). This interaction plays a key role in cancer and other diseases. We report the structure-based design, small-scale synthesis, and biological activity evaluation of a novel family of LFA-1 antagonists. The design led to the synthesis of a family of highly substituted homochiral pyrrolidines with antiproliferative and antimetastatic activity in a murine model of colon carcinoma, as well as potent antiadhesive properties in several cancer cell lines in the low micromolar range. NMR analysis of their binding to the isolated I-domain shows that they bind to the I-domain allosteric site (IDAS), the binding site of other allosteric LFA-1 inhibitors. These results provide evidence of the potential therapeutic value of a new set of LFA-1 inhibitors, whose further development is facilitated by a synthetic strategy that is versatile and fully stereocontrolled.

Full text of DOI:10.1021/jm3016848

Article
pubs.acs.org/jmc
Design, Synthesis, and Functional Evaluation of Leukocyte Function
Associated Antigen‑1 Antagonists in Early and Late Stages of Cancer
Development
Eider San Sebastian,† Tahl Zimmerman,‡ Aizpea Zubia,†,§ Yosu Vara,† Elyette Martin,⊥ Finton Sirockin,∥
́
Annick Dejaegere,⊥ Roland H. Stote,⊥ Xabier Lopez,§ David Pantoja-Uceda,# María Valcarcel,∞,●
́
Lorea Mendoza,∞ Fernando Vidal-Vanaclocha,∞,△ Fernando P. Cossío,*,§ and Francisco J. Blanco*,‡,×
†Ikerchem Ltd., 20009 San Sebastian
‡Structural Biology Unit, CIC bioGUNE, 48160 Derio, Spain
§Departamento de Química Organ
ica I, Facultad de Química, Universidad del País Vasco/Euskal Herriko Unibertsitatea, 20018 San
Sebastian, Spain
∥Novartis Institute for Biomedical Research, CH-4002 Basel, Switzerland
⊥Integrated Structural Biology Department, Institut de Gen
etique et de Biologie Molec
INSERM U964, 10412 Illkirch, Strasbourg, France
́
, Spain
́
́
́
́
́
ulaire et Cellulaire, CNRS UMR7104,
#Instituto de Química-Física Rocasolano, CSIC, 28006 Madrid, Spain
∞Dominion Pharmakine Ltd., 48160 Derio, Spain
×IKERBASQUE, Basque Foundation for Science, 48011 Bilbao, Spain
S
* Supporting Information
ABSTRACT: The integrin leukocyte function associated antigen 1 (LFA-1) binds the
intercellular adhesion molecule 1 (ICAM-1) by its αL-chain inserted domain (I-domain). This
interaction plays a key role in cancer and other diseases. We report the structure-based design,
small-scale synthesis, and biological activity evaluation of a novel family of LFA-1 antagonists.
The design led to the synthesis of a family of highly substituted homochiral pyrrolidines with
antiproliferative and antimetastatic activity in a murine model of colon carcinoma, as well as
potent antiadhesive properties in several cancer cell lines in the low micromolar range. NMR
analysis of their binding to the isolated I-domain shows that they bind to the I-domain
allosteric site (IDAS), the binding site of other allosteric LFA-1 inhibitors. These results
provide evidence of the potential therapeutic value of a new set of LFA-1 inhibitors, whose
further development is facilitated by a synthetic strategy that is versatile and fully
stereocontrolled.
for inhibitors designed to disrupt the LFA-1/ICAM-1
INTRODUCTION
■
interaction. In this respect, there is strong interest in the
design and discovery of low-molecular-weight compounds that
could be developed into orally available therapeutics with
highly favorable pharmacological properties.
LFA-1 (also known as αLβ2 or CD11a/CD18) is a
heterodimeric protein of the integrin family expressed on the
surfaces of all leukocytes and is critical for their antigen-specific
responses and homing. The most important natural ligand of
LFA-1 is the intercellular adhesion molecule 1 (ICAM-1),
which is found on the surfaces of endothelial and epithelial
cells, as well as on leukocytes and fibroblasts, and is up-
regulated at sites of inflammation.1 Domain 1 of ICAM-1
interacts with the metal ion dependent adhesion site (MIDAS)
of the so-called inserted domain or I-domain of the αL chain of
the integrin (Figure 1). The interaction between ICAM-1 and
LFA-1 plays a key role in the pathogenesis of inflammatory and
autoimmune disease conditions such as psoriasis, rheumatoid
arthritis, and transplant rejection2 as well as in the development
of various types of cancer metastasis such as gastrointestinal
carcinoma,3,4 melanoma,5,6 lymphoma,7,8 and colon carcino-
ma.9 LFA-1 therefore constitutes an attractive therapeutic target
Studies of low-molecular-weight inhibitors of LFA-1 reported
to date have revealed the existence of two types of antagonists,
one that is allosteric in nature and the other that supposedly
acts as a competitive inhibitor, but there is some controversy
surrounding the exact binding site.10 Allosteric inhibitors bind
to a hydrophobic cleft between the C-terminal helix and the
central β-sheet of the I-domain known as the I-domain
allosteric site (IDAS), also known as the lovastatin site (L-
site), since this cholesterol-lowering drug was the first allosteric
inhibitor of LFA-1 identified.11 Binding to the IDAS is thought
Received: July 5, 2012
Published: January 22, 2013
© 2013 American Chemical Society
735
dx.doi.org/10.1021/jm3016848 | J. Med. Chem. 2013, 56, 735−747

Journal of Medicinal Chemistry
Article
resonance (NMR) of the mechanism of action of a novel family
of LFA-1 antagonists.
RESULTS
■
Design and Chemical Synthesis of LFA-1/ICAM-1
Inhibitors. We have previously described the design of
inhibitors of the very late antigen 1/vascular cell adhesion
molecule 1 (VLA-4/VCAM-1) interaction on the basis of the
geometric and electronic properties of VCAM-1, which led to
the generation of highly substituted homochiral pyrrolidines
with potent antimetastatic activity in vivo.17 Because of the high
similarity between the three-dimensional structures of VCAM-1
and ICAM-1, we thought that the same approach could be used
to generate inhibitors of the LFA-1/ICAM-1 interaction, for
which not only the crystal structure of the CAM moiety is
known, but also the structure of its complex with the I-domain
of LFA-1.18 Therefore, the design of the first family of LFA-1
antagonists was carried out based on the electronic and
geometric features of those residues in ICAM-1 that establish
key interactions with the I-domain: Glu34, whose carboxyl
group is coordinated with the Mg2+ cation at the MIDAS, and
Lys39, which forms an ion pair with residue Glu241 in the I-
domain (Figure S1A in Supporting Information). In the
solvated crystal structure we found that the side chains of
residues Glu34 and Lys39 have a high degree of conformational
freedom and are surrounded by residues Ile33, Thr35, and
Leu54.
Figure 1. Cartoon representation of the crystal structure (PDB entry
1MQ8) of the complex between the αL I-domain of human LFA-1
(green) and the first three extracellular domains of ICAM-1 (domain 2
is omitted for clarity, and domain 3 was not resolved).18 The side
chains of the residues in both proteins involved in the coordination of
the Mg2+ cation (represented by a sphere) are shown in sticks with
oxygens and nitrogen atoms in red and blue, respectively. Two water
molecules coordinating the Mg2+ cation are shown as small red
crosses. The inset displays the details of the coordination of the Mg2+
cation. The side chains of ICAM-1 residue K39 and I-domain residue
E241 that form a salt bridge are also shown in sticks. The I-domain
present in this structure is a mutant with an engineered disulfide
bridge that locks the protein in a conformation with intermediate
affinity for ICAM-1.
Therefore, molecular mimetics of some of these residues
were generated (Scheme 1), using highly substituted chiral
Scheme 1. List of Prepared LFA-1 Inhibitors 1a−l and
Retrosynthetic Analysis Leading to Intermediates 2a−c, 3a,b,
to stabilize the low-affinity conformation of the I-domain (with
a closed MIDAS), preventing its conversion into its high affinity
form (open MIDAS) and thereby blocking efficient binding to
ICAM-1. Members of the second group were originally
described as competitive inhibitors of the αL I-domain/
ICAM interaction12 but were later shown to bind both the
αL13 and the β2 integrin chains. These compounds are named
α/β I-like allosteric inhibitors.14 Binding of these inhibitors to
the β2 chain is thought to block the interaction of the αL
residue Glu310 with the I-like domain of the β2 chain, which is
structurally similar to the I-domain and contains another
MIDAS. The coordination of the αL Glu310 residue with the
MIDAS of the β2 I-like domain is believed to pull the C-
terminus of the I-domain and trigger the conformational
change into its high affinity form. An inhibitor that blocks this
interaction causes the I-domain to remain in its low-affinity
conformation.15 Allosteric inhibitors bind to the IDAS primarily
through hydrophobic interactions, while competitive antago-
nists binding the MIDAS presumably establish additional
specific interactions. Highly selective inhibitors are desirable for
therapeutic applications to avoid deleterious side effects, but the
design of small molecules that compete with ICAM-1 for
binding to the MIDAS of the αL I-domain remains a
challenging goal.
a
and 4a−c
a
See the Experimental Section and the Supporting Information for
further details and compound characterization. Synthesis of VLA-4
inhibitor 1m was previously described.17
With the aim of finding novel competitive LFA-1 antagonists
we have previously analyzed the geometric and electronic
features of the ICAM-1 interaction with the I-domain in both
the open and closed conformations of the MIDAS region16 and
studied the most promising chemical motifs for MIDAS
binding. Here we describe the rationale for the design, strategy
for synthesis, probing of inhibitory activity in vitro and in vivo,
in silico optimization, and characterization by nuclear magnetic
pyrrolidines as scaffolds. As relevant examples, Figure S1B,C
shows the stereoelectronic features of two representative
inhibitors. The amino groups of the pyrrolidine ring (1b) or
the 1-acylamino groups (1j) were designed to mimic the amino
group of Lys39 of ICAM-1, whereas the carboxyl groups
reproduce the negative charge of the Glu34 residue. Since our
previous study on the affinity of aliphatic carboxylic acids for
736
dx.doi.org/10.1021/jm3016848 | J. Med. Chem. 2013, 56, 735−747

Journal of Medicinal Chemistry
Article
the MIDAS showed that an increase in the length of the alkyl
chain beyond n = 1 has little effect on the binding,16 we
concluded that this part of the molecule should allow for the
elongation of the alkyl chain (n > 1; see Scheme 1), if necessary
for the better positioning of functional groups. Finally, the alkyl
and aryl groups attached to the pyrrolidine ring provide the
surrounding environment. The pyrrolidine ring and the four
chiral centers restrict conformational freedom. This, in turn,
preorganizes the putative inhibitors and therefore decrease the
entropy penalty paid for receptor binding.
chlorides 6a,b and the corresponding aldehyde. Given their
instability, these imines were used without further purification.
The silver cation mediated [3 + 2] cycloaddition reaction
between 4a−c and the nitroalkenes 3a,b (Scheme 2) yielded
the corresponding pyrrolidines 2a−c. It is remarkable that in
this reaction four stereogenic centers were generated in a single
preparative step with complete regio- and stereocontrol. Methyl
esters 2a−c were hydrolyzed and coupled with the correspond-
ing amino acid methyl ester hydrochlorides in the presence of
diethyl phosphorocyamidate (Scheme 3) to yield compounds
The chemical synthesis of the novel compounds 1a−l relies
on our previously described method for the synthesis of highly
substituted nitroprolines via [3 + 2] cycloadditions between
homochiral nitroalkenes 3 and imines 4,17 as shown in the
retrosynthetic analysis for compound 1 gathered in Scheme 1.
Compounds 3a,b were readily prepared in seven steps from L-
isoleucine. This preparative sequence involves substitution of
the amino group of the starting L-amino acid 5 (Scheme 2) by
a
Scheme 3. Synthesis of Inhibitors 1a−i and 1j−l
a
Scheme 2. Synthesis of Intermediates 2a−c, 3a,b, and 4a−c
a
The substitution patterns are those indicated in Scheme 1. Reagents
and conditions: (a) LiOH 1M, DME,
3 h, 0 °C; (b)
HCl·NH2(CH2)nCH2COOCH3, Et3N, diethyl phosphorocyanidate
(DEPC), DMF, 16 h, 0 °C to rt; (c) LiOH 1 M, DME, 25 min, 0
°C; (d) N-phthaloylglycine anhydride, DMF, Et3N, 48 h, rt; (e) EtOH,
hydrazine, 24 h, 78°C; (f) LiOH 1 M, DME, 25 min, 0 °C.
7a−i. Simple hydrolysis of these intermediates afforded
inhibitors 1a−i with good yields. In order to prepare analogues
of the inhibitors 1a−i with primary amino residues, compounds
7a−i were coupled with phthaloylglycine anhydride with
moderate yields. Hydrazine deprotection of phthalimidoyl
derivatives 8j−l followed by hydrolysis of the methyl ester
moiety resulted in the formation of the desired inhibitors 1j−l.
It is noteworthy that the above-described synthesis of inhibitors
is versatile and fully stereocontrolled and uses readily available
starting materials. It is therefore well suited for further
developments including extensive exploration of the structural
space.
Biological Assays with First-Generation Compounds.
In an initial series of cell-based adhesion experiments,
compound 1b emerged as a promising inhibitor of the LFA-
1/ICAM-1 interaction and therefore was selected for further
testing in in vitro and in vivo activity assays. As depicted in
Figure 2A, preincubation of murine CT26 colon carcinoma
a
Unless otherwise indicated, the substitution patterns are those
indicated in Scheme 1. Reagents and conditions: (a) H2SO4 1 N,
NaNO2, 24 h, 0°C; (b) 2,2-dimethoxypropane, TsOH·2H2O, MeOH,
24 h, 45°C; (c) NaH, THF, R1CH2Br, 72 h, rt; (d) LiAlH4, Et2O, 3 h,
rt; (e) TCIA, TEMPO, CH2Cl2, 15 min, rt; (f) CH3NO2, Et3N, 16 h,
rt; (g) MsCl, DIPEA, CH2Cl2, 2 h, −78 °C; (h) R2CHO, Et3N,
MgSO4, CH2Cl2, 16 h, rt; (i) AgOAc, Et3N, CH3CN, 5 h, rt.
the hydroxy group with retention of configuration, Williamson
coupling with the appropriately substituted benzyl bromide,
and nitroaldol reaction before the corresponding elimination
reaction.
The starting imines 4a−c (Schemes 1 and 2) were prepared
in one step by condensation between methyl ester hydro-
737
dx.doi.org/10.1021/jm3016848 | J. Med. Chem. 2013, 56, 735−747

Journal of Medicinal Chemistry
Article
in vivo. Because the CT26 cell line is the most aggressive of the
two (displaying a faster growth in vitro and a larger metastatic
spread in vivo), further biological assays were conducted using
the CT26-CC cells.
Compound 1b abrogated the PMA-induced CT26 murine
colon carcinoma (CT26-CC) cell adhesion to HSE cells, as did
the aLFA-1 monoclonal antibody (Figure S2B). Preincubation
of B16 melanoma (B16M) cells with either compound 1b or
aLFA-1 abolished the increase in B16M cell adhesion to HSE
cells observed in response to the integrin overexpression and
activation in HSE cells treated with murine recombinant
vascular endothelial growth factor (VEGF; see Figure S2C).
Compound 1b displayed a strong antimetastatic activity in vivo,
as seen by the 25-fold reduction in both density and volume of
the metastasis produced in mice by the injection of CT26-CC
treated with compound 1b compared to mice receiving
untreated cells (Figure 2C,D).
Preincubation of human peripheral blood leukocytes (PBLs)
with compound 1b significantly decreased the PMA-induced
PBL adhesion to immobilized human recombinant soluble
ICAM-1 (sICAM-1) (Figure 3B), which indicates that the
antiadhesive effect of compound 1b is due to the specific
blockade of the interaction between LFA-1 and ICAM-1. In
addition, both aLFA-1 and compound 1b abrogated the
increased production of VEGF by B16M cells treated with
sICAM-1 and H202 (Figure S2D).
Figure 2. (A) PMA-treated and untreated CT26-CC cell adhesion to
HSE cells in the absence and presence of first-generation compounds
1a−c and the VLA-4 inhibitor 1m. Statistically significant (p < 0.05)
differences with respect to untreated (∗) or PMA-treated cells (∗∗)
are indicated. (B) PMA-treated and untreated Lim51b-CC cell
adhesion to HSE cells in the presence and absence of monoclonal
antibody aLFA-1, compound 1b, or compound 1a. Statistically
significant (p < 0.01) differences with respect to untreated (∗) or
PMA-treated cells (∗∗) are indicated. (C) Decrease of metastasis
density in mice injected with untreated or compound 1b-treated
CT26-CC cells. Statistically significant (p < 0.05) differences with
respect to control mice (∗) are indicated. (D) Decrease of total
metastasis volume in mice injected with untreated or compound 1b-
treated CT26-CC cells. Statistically significant (p < 0.05) differences
with respect to control mice (∗) are indicated.
Structure−Activity Relationship Analysis and Design
of Second-Generation Compounds. The cell adhesion
results show that compound 1b is the most potent inhibitor of
the LFA-1/ICAM-1 interaction among the first-generation
compounds. Compound 1a is less potent, and compound 1c is
not active. The structure of compound 1a is very similar to the
inactive VLA4 inhibitor 1m except for a longer chain containing
the carboxyl group (n = 1; see Scheme 1). This chain
lengthening was introduced in the three first-generation
compounds to facilitate the coordination of the magnesium
cation at the MIDAS, and our results indicate that it is a critical
structural feature for their activity. However, compound 1c
does not inhibit CT26-CC cell adhesion, indicating that R1 and
R2 substituents of a more polar nature than phenyl ones are
unfavorable and can override the favorable length of the
carboxylic chain. The increase in potency of 1b relative to 1a
should be due to the increase in size and hydrophobicity
conferred by the methyl group as substituent R3. A more
detailed investigation in 3D space of these issues was assessed
by computational methods. Binding of compound 1b to the
MIDAS motif of the high affinity form of the αL I-domain was
assessed by virtual docking followed by molecular dynamics
(MD) simulation to sample the conformational space of the
complex (see the methods section and the Supporting
Information for the detailed procedure). The presence or
absence of interactions mimicking the coordination of ICAM-1
residue Glu34 to the Mg2+ cation and the salt bridge between
ICAM-1 residue Lys39 and I-domain residue Glu241 was
inspected in snapshots taken at regular time intervals along the
MD simulation. As indicated in Figure S4, the interaction of the
carboxyl group of the docked 1b with the Mg2+ ion is
established and remains stable along the simulation. The
distance between the carboxylic oxygen coordinating the cation
and the cation is in the range of 3.5−4.2 Å, indicating that the
electrostatic interaction is not optimal and suggesting that it
could be strengthened by an increase in the chain length.
However, the distance between the amino group in compound
(CT26-CC) cells with 1a or 1b abrogated the increase in
CT26-CC cell adhesion to hepatic sinusoidal endothelial
(HSE) cells in response to integrin activation by the treatment
of the CT26-CC cells with phorbol miristate acetate (PMA),
while incubation with compound 1c did not. Interestingly, the
chemically related VLA-4 inhibitor 1m did not show any
inhibitory activity in this assay, indicating that the new family of
compounds is specific for the disruption of the LFA-1/ICAM-1
interaction. This conclusion is supported by a reverse
experiment (Figure S2A) showing that compound 1b did not
inhibit B16 M cell adhesion to immobilized VCAM-1 (the
ligand of VLA-4), providing a cleavage point between the two
family of compounds.
We also observed that preincubation of murine Lim51b
colon carcinoma (Lim51b-CC) cells with compound 1b or
anti-LFA-1 monoclonal antibody (aLFA-1) abolished the
increase in PMA-induced Lim51b-CC cell adhesion to HSE
cells and to a larger extent than did incubation with compound
1a (Figure 2B). Using dose−response data and a simple Hill
type equation (Figure S3), the IC50 of 1b was estimated to be
in the micromolar range (IC50 ≈ 27.6 μM/105 cells). The two
colon carcinoma cell lines (CT26 and Lim51b) differ in their
rates of growth in vitro and in their ability for metastatic spread
738
dx.doi.org/10.1021/jm3016848 | J. Med. Chem. 2013, 56, 735−747

Journal of Medicinal Chemistry
Article
Figure 3. (A) PMA-treated and untreated CT26-CC cell adhesion to HSE cells in the presence and absence of compounds 1d−l or 1b. Statistically
significant (p < 0.01) differences with respect to untreated (∗) or PMA-treated cells (∗∗) are indicated. (B) PMA-treated and untreated PBL cell
adhesion to immobilized human recombinant ICAM-1 in the presence and absence of compounds 1d−i or 1b. Statistically significant (p < 0.01)
differences with respect to untreated (∗) or PMA-treated cells (∗∗) are indicated. (C) Primary tumors induced in mice 15 days after injection of
CT26-CC cells and daily treatment with 1h or 1j. (D) Volume of the primary tumors shown in (C). (E) Expression levels of protein Ki67 in the
same mice primary tumors. Statistically significant (p < 0.05) differences with respect to control mice (∗) are indicated (see Supporting Information
for further details).
1b and Glu241 in the I-domain fluctuates between 9 and 13 Å,
distances that are too large for a stable salt bridge. Despite the
long-range nature of the electrostatic interactions, it can be
assumed that no salt bridge is established between these two
groups of atoms. These results guided the design of a second
generation of inhibitors (Scheme 1, 1d−l) in which the relative
positions of the charged groups were modified to favor the
simultaneous formation of the two interactions. This was
achieved by elongation of the carboxylic arm of the compounds
(1d−i in Scheme 1) to provide higher flexibility and therefore
greater freedom to bend as needed and by addition of a
substituent with a terminal amino group attached to the amino
group of the pyrrolidine (1j−l in Scheme 1). Elongation of the
carboxylic arm in the antagonist is also suggested by the
observation that compound 1m, originally described as a VLA-4
antagonist,17 does not inhibit LFA-1 (Figure 2A). This
compound has an aliphatic carboxylic acid that is a single
methylene shorter than 1b (see Scheme 1). As a first approach
to predict the quality of the new structures proposed, the
electrostatic potential generated by the two key residues in
ICAM-1 (Figure S1A) was compared to the one generated by
first-generation compound 1b (Figure S1B) and second-
generation compound 1j (Figure S1C). Both compounds
effectively mimic the spatial distribution and the magnitude of
the electrostatic potential generated by the two key residues in
ICAM-1. In addition, compound 1j displays the crevice-like
space between the two electrostatic centers present in the
natural ligand (Figure S1A,C).
Biological Assays with Second-Generation Com-
pounds. CT26-CC cell adhesion to HSE cells shows that
compounds 1h and 1j are stronger inhibitors than first-
generation compound 1b (Figure 3A). Compound 1h reduced
PBL adhesion to immobilized sICAM-1 to basal levels (Figure
3B), indicating that the inhibition is due to the specific
blockade of the sICAM-1 dependent adhesion. Compounds 1h
and 1j were chosen for in vivo assays as representative of the
subset of active second-generation compounds with an
elongated carboxyl group arm and with an elongated amino
group arm, respectively. Compounds 1h and 1j decreased the
volume of primary tumors in tissues of mice injected with
CT26-CC cells by an average of 85% and 77% with respect to
the control (Figure 3C,D). They also decreased the expression
of the Ki67 antigen by approximately 40% (Figure 3E), a
marker of cell proliferation, indicating a reduction of the
malignancy or invasive character of primary tumors.19
Experimental Characterization of Compound Binding
and Competition with ICAM-1. To further analyze whether
our antagonists inhibit ICAM-1 binding to the LFA-1 I-domain
in a competitive manner, their binding site onto the isolated I-
domain was mapped by measuring the NMR chemical shift
perturbations (CSPs) experienced by the individual I-domain
1
residues upon compound addition. For this purpose, H−15N
HSQC spectra of 15N enriched I-domain were acquired in the
739
dx.doi.org/10.1021/jm3016848 | J. Med. Chem. 2013, 56, 735−747

Journal of Medicinal Chemistry
Article
absence and in the presence of first- and second-generation
LFA-1 antagonists.
All the tested LFA-1 antagonists affected the chemical shifts
of the same set of I-domain residues (Figure 4A and Figure
type. This observation suggests that the compounds bind to the
same region but with a smaller affinity. This is consistent with
the absence of a large pocket in the crystal structure of the
mutant (PDB code 1MQA).
In all the complexes that were studied by NMR, the CSPs of
the MIDAS residues were very small and did not change when
the complexes were examined in the presence of the cation
chelating agent ethylenediaminetetraacetic acid (EDTA, data
not shown). These results confirm that the compounds do not
interact with the MIDAS in either the closed or the open
conformation of the isolated I-domain. Binding was observed to
be highly selective with respect to the αL I-domain subclass,
since neither compound 1e nor lovastatin were found to bind
to the I-domain of the integrin α2 chain (Figure S6). Nor did
compound 1e bind to the unrelated B1 domain of protein G
(data not shown).
Compound 1b was docked to the IDAS site of the I-domain
using the measured CSPs. The model shows that compound 1b
binds the IDAS in the same place and with an orientation
equivalent to that of lovastatin (Figure 4B). A pharmacophore
analysis revealed common features in the two structures (Figure
S7). In the one of lovastatin bound to the I-domain, there are
two hydrophobic features and two hydrogen bond acceptor
features. In the structure of compound 1b docked on the I-
domain, these features are also present, although they are not
exactly superposable. One of the acceptors in 1b exhibits a
doubled-headed acceptor feature (corresponding to the
carboxyl group), although it interacts with the same lysine of
the I-domain as lovastatin (K160 and K287 as numbered in
1ZOP and 1CQP, respectively). Compound 1b also exhibits an
additional hydrophobic feature formed by the aromatic group at
position R2.
The affinity of the binding of selected compounds to the
isolated I-domain was determined by titrating the protein with
each compound and fitting the measured CSP to a single site
binding equilibrium equation (Figure 4C). The values of KD
( fitting error) were 10.1 1.3 and 7.4 0.8 μM for LFA-1
Figure 4. (A) Overlay of the 1H−15N HSQC spectrum of the I-
domain (50 μM) in the presence (black) or absence (red) of
compound 1b (200 μM) in sodium phosphate buffer pH 7.4 at 22 °C.
The numbers beside the signals indicate the corresponding I-domain
residue. (B) Cartoon representation of the crystal structure (PDB code
1CQP) of the I-domain (cyan ribbon) in complex with allosteric
inhibitor lovastatin (yellow sticks), overlaid with the NMR data guided
docked compound 1b (black sticks). The I-domain in this structure
(1ZOP) is the wild type one, with low affinity for ICAM-1. (C)
Representation of the CSP experienced by the signal of Thr291 as a
function of compound 1b concentration. The signal of residue Thr291
is located in the upper right corner of the HSQC spectrum region
shown in panel A. Error bars depict the experimental error of the
measurements, and the error in the KD is the fitting error.
antagonists 1b and 1e, respectively, and 12.9
4.4 μM for
lovastatin binding to the wild type I-domain. Compound 1b
bound the mutant I-domain locked in the open MIDAS form
with KD = 81.1
bound to the IDAS of the I-domain, although with a 2-fold
reduced affinity (KD =19.3 2.3 μM). Furthermore, two
12.6 μM. The VLA-4 inhibitor 1m also
unrelated compounds (chloramphenicol and ethyl 4,6-dime-
thoxy-1H-indole-2-carboxylate) used as negative controls also
bound to the I-domain and to the same IDAS region as
lovastatin and the novel compounds (Figure S8), although they
caused much smaller CSPs and thus presumably bind with
much smaller affinity.
We next studied the effect of compound binding to the IDAS
on the ICAM-1 binding to the I-domain. It has been shown by
surface plasmon resonance (SPR) measurements that ICAM-1
binds with very different affinities to the I-domain in the closed
or open MIDAS forms (KD ≈ 1500 μM and KD ≈ 0.15 μM for
the wild type and mutant I-domains, respectively).18 SPR
measurements show that binding of the mutant I-domain to
immobilized ICAM-1 was not abrogated in the presence of
saturating amounts of compound 1b (Figure S9). The binding
to the wild type I-domain is too weak to reliably measure the
possible inhibition by the compound using SPR, but this can be
investigated by NMR. Binding of ICAM-1 reduced the intensity
of several I-domain residue signals both in the complex
interface (including the MIDAS) and in the IDAS (Figure 5),
S5A). The residues experiencing the largest CSPs mapped the
binding of the compounds to a region encompassing the IDAS
(Figure S5B,C), which is the binding site of lovastatin and
other allosteric inhibitors.11 To confirm that our compounds
bound to the same site as lovastatin, we measured the CSPs
caused by lovastatin under our experimental conditions and
found the same result (Figure S5D). To examine if the
compounds bound to the MIDAS in its open conformation, the
high affinity form of the I-domain, we performed the same
measurements on a mutant protein locked in the open MIDAS
conformation by an engineered disulfide bridge.20 We observed
again that residues in the IDAS region were perturbed while the
MIDAS residues were not (Figure S5E). The CSPs measured
on the mutant I-domain were smaller than those on the wild
740
dx.doi.org/10.1021/jm3016848 | J. Med. Chem. 2013, 56, 735−747

Journal of Medicinal Chemistry
Article
Figure 5. (Left) Representation of the I-domain surface that interacts with ICAM-1 (PDB code 1MQ8). Residues that make direct contacts with
ICAM-1 are depicted with yellow sticks and labeled. Residues whose labels are red are those whose cross-peaks recover the least with the addition of
compound 1b. The Mg2+ ion and MIDAS site waters are represented by cyan and red spheres, respectively. (Right) Bar graphs of the intensity of the
1H−15N HSQC signals of the red-labeled residues in the indicated conditions.
which is consistent with the two sites being structurally linked
as previously reported.21,22 The presence of saturating amounts
of the compound provoked a partial recovery in the intensity of
some signals in the IDAS region, but little or no recovery was
observed for signals of residues known to make direct contacts
with ICAM-1, indicating that ICAM-1 binding to the I-domain
is not abolished by compound 1b. These results suggest that
ICAM-1 and compound 1b do not compete for binding to the
MIDAS of the isolated I-domain and that compound binding to
the IDAS does not block the low-affinity interaction with
ICAM-1 in solution. The same conclusion cannot be made with
respect to the mutant, since it is locked in its high affinity form
by a disulfide bridge.
Binding of our compounds to the I-domain produced CSPs
in residues Thr267 and Asp291 (in the β5−α6 and β6−α7
loops, respectively) that, given the large distance between these
regions and the IDAS, are likely a consequence of a
conformational change induced by compound binding and
not of compound binding directly. Indeed, in a previous
theoretical study of αL I-domain dynamics, we showed that the
dynamics of these two loops were correlated with the IDAS and
the α7 helix on one side and the MIDAS on the other.22
Because the two forms of the I-domain must be in equilibrium
in solution (with the closed form being the predominant one),
these perturbations may report a shift in the equilibrium toward
the closed form. Consistent with this hypothesis, we have
observed CSPs in the β5−α6 and β6−α7 loop residues of the
disulfide bridged mutant on addition of 1,4-dithiothreitol
(DTT, which reduces the cystine and unlocks the open
conformation of the I-domain presumably shifting the
equilibrium toward the closed form; data not shown). We
also observed a significant signal broadening in the NMR
spectrum of the αL I-domain in residues of these loops and the
C-terminus in the presence of ICAM-1 (presumably the
consequence of a shift of the conformational equilibrium
toward the open form; data not shown). To further characterize
the effect of compound binding to the IDAS, we measured the
backbone dynamics of the I-domain free and bound to
compound 1b by means of 15N relaxation measurements. The
residues in the α2−α3 (Asp191) and β5−α6 (Thr267) loops,
as well as in the C-terminal α7 helix (Glu301, Gln303), are
more dynamic in the picosecond to nanosecond time scale than
the rest of the molecule, as is also the N-terminal residue
(Figure S10). The relatively flexible character of the α2−α3
linker and the C-terminal helix in the absence of compound, is
consistent with heteronuclear Overhauser effect (NOE) data of
the I-domain R189W mutant.23 In the presence of saturating
amounts of compound 1b, signal broadening made impossible
an unambiguous assignment of certain signals and caused an
increase in the error of the order parameters in others.
Consequently, a comparison of the two sets of parameters for
the C-terminal helix was inconclusive. Nevertheless, the average
15N{1H} NOE values in this helix change from 0.761 0.016
to 0.822
0.032 (Figure S11), which suggests an increased
rigidity in the presence of compound 1b and that the I-domain
is stabilized in its low-affinity closed form. Interestingly, a
rigidification of the C-terminal residues of the I-domain upon
binding to lovastatin can also be observed by examining the B-
factors in the corresponding crystallographic structures (1ZOP
and 1CQP, respectively).
DISCUSSION
■
The structure-based rational design presented here, and our
previous experience with the generation of substituted
nitroprolines via [3 + 2] cycloadditions between homochiral
nitroalkenes and imines, has allowed us to efficiently synthesize
a family of novel and potent LFA-1 antagonists based on highly
substituted homochiral pyrrolidines. The synthesis of these
inhibitors is versatile, fully stereocontrolled, and uses readily
available starting materials. In addition, the molecular size and
the features of these inhibitors are compatible with the
acceptable values for oral bioavailability.24
The in vitro cell adhesion assays show that the compounds
inhibit the intercellular adhesion of different cancer cell types
mediated by LFA-1. The most potent compounds display little
effect on the basal cell adhesion levels but abolish the increase
in adhesion when the cells are treated with PMA or VEGF. The
compounds display a strong antiproliferative activity in vivo and
provoke a reduction in the expression of the Ki67 antigen by
the tumor cells, which indicates a reduction in their invasive
character and is consistent with the strong antimetastatic
activity shown by compound 1b. Since this compound reduced
741
dx.doi.org/10.1021/jm3016848 | J. Med. Chem. 2013, 56, 735−747

Journal of Medicinal Chemistry
Article
the secretion of VEGF by B16M cells, and VEGF promotes the
generation of new blood vessels,25,26 this observation suggests
that the reduction in metastasis volume is due to an
antiangiogenic effect of LFA-1 inhibition by compound 1b.
Interestingly a proangiogenic phenotype of CT-26-CC cells has
been reported to be activated via LFA-1-dependent VEGF.27
The compounds, especially 1b, 1h, and 1j, reduce the
adhesion of cells to immobilized sICAM-1, indicating that they
block the interaction between LFA-1 and ICAM-1, as expected.
However, they do not block ICAM-1 binding to the isolated I-
domain. Even though they were designed to bind to the
MIDAS of the LFA-1 I-domain, they do not when the isolated
I-domain is studied in solution. Instead they bind to the IDAS,
the site of allosteric inhibitors. These apparently contradictory
results could be explained by considering one of the following
hypotheses: (i) the compounds are competitive inhibitors and
bind the MIDAS of the I-domain in the context of the intact
integrin heterodimer but not in the isolated I-domain; (ii) the
compounds bind to the IDAS site of the integrin and are
allosteric inhibitors of LFA-1; (iii) the compounds act
allosterically via binding to a site other than the αL I-domain,
possibly the MIDAS of the β2 I-like domain.
The first hypothesis is supported by the success of the design
strategy: the compounds were designed to mimic key features
in ICAM-1 for its binding to the MIDAS, and in effect these
compounds exhibit LFA-1 antagonist activity in cell based in
vitro assays and in vivo. Furthermore, a rational optimization of
the first-generation inhibitors resulted in a second generation
with more potent compounds. In opposition to this hypothesis
is the observation that the compounds do not compete for
ICAM-1 binding to the isolated I-domain (neither in its low- or
high-affinity forms) and thus argue against a competitive mode
of action. Still it is possible that the I-domain exhibits different
behavior when isolated than when in the context of the whole
integrin. To clarify this issue, it would be necessary to study the
inhibitory activity of our compounds against full-length active
LFA-1.
The second hypothesis is supported by a mode of binding of
our compounds to the I-domain similar to that of the allosteric
inhibitor lovastatin. We could not detect inhibition of ICAM-1
weak binding to the isolated I-domain in the presence of our
compounds, but this fact does not rule out that they act
allosterically. There is evidence that when the I-domain is
trapped in its closed form (as IDAS binding inhibitors are
thought to do), it retains a basal low affinity for ICAM-1, as
seen in a mutant I-domain locked in its closed form by a
disulfide bridge.18 As a consequence, an inhibitory effect would
only be observed when a given signal that provokes the
downward shift of the C-terminal α7 helix is blocked by an
IDAS site binding compound. Rolling adhesion of cells
expressing the isolated wild type I-domain on their membranes
to immobilized ICAM-1 is inhibited in the presence of allosteric
inhibitors,28,29 supporting the model whereby IDAS binding
inhibitors block the conformational changes leading to the
high-affinity form. Further support for the second hypothesis is
found in computational studies showing an increased rigidity in
the IDAS and C-terminal helix in the presence of an IDAS
binding inhibitor,22,30 and our backbone dynamics data show
that the C-terminus is rigidified in the presence of the IDAS
binding compound 1b. In addition, the full length LFA-1 with
its I-domain locked in its open form by a disulfide bridge is
resistant to lovastatin31 and the same occurs with the isolated
and locked-open I-domain expressed on the cell surface.32
Consistent with these results, our compound 1b does not
disrupt the binding of ICAM-1 to the I-domain locked in the
open form by a disulfide bridge. However, it looks as if any
small hydrophobic molecules may bind the I-domain IDAS in
solution, which raises the question of whether binding of a
compound to the promiscuous IDAS really means that it blocks
the conformational change to the high-affinity I-domain form.
Two considerations suggest that this is not the case for any
compound. First, the binding affinity is much smaller for the
two unrelated compounds studied here than for our
compounds and lovastatin. Second, even minor differences in
compound structure affect compound activity, as evidenced by
pravastatin, a far less effective inhibitor and weaker binder of
the I-domain than lovastatin. The only difference between
pravastatin and lovastatin is a hydroxyl group in place of a
methyl.33 So while a variety of compounds have the potential to
bind the IDAS, not all have the correct structure for high-
affinity binding to the IDAS or for inhibition of the LFA-1/
ICAM-1 interaction. To confirm that the mechanism of action
of our compounds is purely allosteric, we would need to show
that binding to the IDAS leads to inhibition of ICAM-1
binding. However, we cannot observe inhibition of the I-
domain in its closed form, and it is impossible to observe
allosteric inhibition with the disulfide-bridged active form.
Again, the isolated I-domain is not a good model for binding
studies with ICAM-1.
The third hypothesis is suggested by the structural similarity
between the MIDAS of the I-domain and of the I-like-domain,
as well as by the observation that while some of the α/β I-like
allosteric inhibitors14 were designed to be competitive ones,12
they did not compete with ICAM-1 binding to the isolated I-
domain.13 Further research on their mechanism of action
showed mixed results regarding binding to the I-like-domain14
or the I-domain13 in assays with soluble extracellular domains
or with deletion mutants of the integrin heterodimer,
respectively. We, however, have no evidence that our
compounds act in this way; rather they appear to bind to the
IDAS of the isolated I-domain. This is in contrast to what is
observed with the α/β I-like allosteric inhibitor XVA143;34
therefore, this third possibility is highly unlikely.
The binding sites of known LFA-1 antagonists have generally
been determined by high resolution studies (such as NMR) on
the isolated I-domain or by low resolution ones (such as
epitope mapping with antibodies) with deletion mutants or the
full length integrin. Both experimental approaches have their
limitations and may, in the end, yield apparently contradictory
results for a system as complex as LFA-1, which contains two
multidomain chains, many points of conformational change,
and many ion binding sites. Therefore, it may be imprudent to
rely only on the isolated I-domain for the study and design of
LFA-1 inhibitors, as we and others have done.
CONCLUSION
■
We have designed a novel family of LFA-1 antagonists that can
be synthesized from readily available starting materials. Some of
these compounds display potent antiadhesive, antimetastatic,
and antiproliferative properties in the low micromolar range in
in vitro and in vivo assays. Although these inhibitors were
designed to be competitive, they bind to the allosteric site of
the isolated αL I-domain. However, neither an allosteric nor a
competitive mechanism of action can yet be discerned in the
context of the full integrin. The synthetic strategy is versatile
and fully stereocontrolled and therefore well suited for further
742
dx.doi.org/10.1021/jm3016848 | J. Med. Chem. 2013, 56, 735−747

Journal of Medicinal Chemistry
Article
0.99 (m, 1H), 0.91 (t, 3H, J = 6.9), 0.74 (d, 3H, J = 6.8); 13C NMR (δ
ppm, CDCl3) 175.9, 138.3, 135.7, 128.7, 128.4, 127.6, 127.3, 126.9,
91.5, 79.7, 70.3, 66.2, 65.0, 50.2, 35.9, 34.7, 34.3, 26.2, 19.5, 13.3, 12.0;
developments including extensive exploration of the structural
space.
mp 104−105 °C; [α]D +13.23 (c 0.94, CH2Cl2). Anal. Calcd for
25
EXPERIMENTAL SECTION
Chemical Reagents and Compound Characterization Meth-
ods. All starting materials and reagents were purchased from Sigma-
Aldrich and used without further purification. Melting points of solid
■
C27H35O6N3: C, 65.16; H, 7.10; N, 8.45. Found: C, 64.59; H, 7.12; N,
8.56.
{(2S,3R,4S,5S)-3-[1-(S)-(2-Fluorobenzyloxy)-2-methylbutyl]-4-
nitro-5-(3-thienyl)}prolyl-β-alanine, 1c. 95% yield; IR (KBr, cm−1)
compounds were determined on a Buchi B-540 apparatus and are
̈
1
3372, 3332, 1726, 1673, 1552, 1358, 1232; H NMR (δ ppm, J Hz,
uncorrected. No melting points are reported for oily compounds.
Determination of the purity of tested compounds was performed by
combustion analysis (C, H, N) carried out on a Leco CHNS-932
elemental analyzer calibrated with sulfamethazine. The results
confirmed a ≥95% purity. Merck silica gel 60F-254 plates were used
for analytical TLC. NMR spectra were measured in CDCl3 and
DMSO-d6 on Varian Gemini 200, Bruker Avance 300, and Bruker
Avance 500 spectrometers. Chemical shifts (δ) are expressed in ppm
and coupling constants (J) in hertz. IR spectra were recorded in a
Perkin-Elmer 1600 series FTIR spectrometer connected to a PC using
KBr disks. Optical rotations were measured on a Perkin-Elmer 243B
polarimeter.
Synthesis of Compounds 7a−i. The corresponding pyrrolidine
methyl ester 2a−c (5.0 mmol) was dissolved in DME (25 mL) and
cooled to 0 °C. LiOH, 1 N aqueous solution (15 mL), was added
dropwise, and the progress of the reaction was monitored by TLC.
After completion of the reaction, citric acid, 10% aqueous solution (15
mL, pH ≈ 6), was added. The resulting solution was extracted with
CH2Cl2 (3 × 20 mL), and the combined organic fractions were dried
and evaporated. The crude product was triturated in Et2O, yielding the
corresponding pyrrolidine acid as a white solid. To a round-bottom
flask under argon atmosphere, the obtained pyrrolidine 2-carboxylic
acid (1 mmol) and the appropriate aliphatic ε-amino ester (1 mmol)
in 2.5 mL of anhydrous DMF were introduced, and the mixture was
cooled with an ice/water bath. DECP (0.18 mL, 1.2 mmol) in 0.5 mL
of DMF and TEA (0.29 mL, 2.05 mmol) were added dropwise, and
the resulting mixture was stirred at room temperature for 16 h. Then
AcOEt (100 mL) and toluene (100 mL) were added, and the organic
solution was washed with 50 mL fractions of H2O, Na2S2O3, 1 N
aqueous solution, H2O, NaHCO3 saturated aqueous solution, and
NaCl saturated aqueous solution, dried (Na2SO4), and evaporated.
The crude mixture was purified by flash chromatography (ethyl
acetate/hexanes). The characterization of these intermediate com-
pounds can be found in the Supporting Information.
CDCl3) 7.83 (tb, 1H, J = 5.3), 7.39−6.90 (m, 7H), 6.31 (sb, 2H), 5.29
(db, 1H, J = 6.0), 4.69 (d, 1H, J = 11.1), 4.59−4.54 (m, 2H), 3.77 (d,
1H, J = 7.3), 3.70−3.46 (m, 3H), 3.18 (db, 1H, J = 6.5), 2.58 (tb, 2H, J
= 5.8), 1.94−1.71 (m, 1H), 1.63−1.38 (m, 1H), 1.31−1.06 (m, 1H),
0.96−0.83 (m, 6H); 13C NMR (δ ppm, CDCl3) 176.2, 171.6, 163.3,
158.4, 135.6, 130.3, 130.2, 130.0, 129.8, 126.2, 125.8, 124.2, 124.1,
122.7, 115.6, 115.2, 90.3, 82.8, 67.2, 67.1, 63.8, 63.7, 50.4, 37.3, 34.7,
33.7, 25.8, 14.4, 11.6; mp 113−114 °C; [α]D25 +63.55 (c 1.1, CH2Cl2).
Anal. Calcd for C24H30O6N3FS: C, 56.78; H, 5.97; N, 8.28. Found: C,
56.36; H, 5.97; N, 8.21.
4-{(2S,3R,4S,5S)-3-[(1S,2S)-1-(Benzyloxy)-2-methylbutyl]-4-nitro-
5-phenylpyrrolidine-2-carboxamido}butanoic Acid, 1d. 68% yield;
IR (KBr, cm−1) 3397, 1733, 1653, 1552, 1362; 1H NMR (δ ppm, J Hz,
CDCl3) 7.43−7.35 (m, 5H), 7.34−7.29 (m, 6H), 5.36 (dd, 1H, J = 6.6,
J′ = 2.2), 4.79 (d, 1H, J = 11.4), 4.60 (d, 1H, J = 6.6), 4.55 (d, 1H, J =
11.4), 3.78 (d, 1H, J = 5.9), 3.71 (d, 1H, J = 7.4), 3.51−3.44 (m, 1H),
3.40−3.34 (m, 1H), 3.07 (db, 1H, J = 7.3), 2.42 (dt, 2H, J = 7.1, J′ =
2.4), 1.95−1.89 (m, 2H), 1.88−1.81 (m, 1H), 1.54−1.46 (m, 1H),
1.36−1.30 (m, 1H), 1.22−1.13 (m, 1H), 0.94 (t, 3H, J = 7.4), 0.87 (d,
3H, J = 6.9); 13C NMR (δ ppm, CDCl3) 176.7, 173.1, 138.3, 135.1,
128.8, 128.6, 128.4, 127.9, 127.6, 126.7, 91.0, 82.8, 73.2, 67.1, 63.9,
50.5, 38.4, 37.3, 31.3, 26.1, 24.9, 14.5, 11.7; [α]D +25.0 (c 1.10,
25
CH2Cl2). Anal. Calcd for C27H35O6N3: C, 65.17; H, 7.09; N, 8.44.
Found: C, 65.03; H, 7.11; N, 8.40.
4-{(2S,3R,4S,5S)-3-[(2S)-1-(Benzyloxy)-2-methylbutyl]-2-methyl-
4-nitro-5-phenylpyrrolidine-2-carboxamido}butanoic Acid, 1e. 71%
1
yield; IR (KBr, cm−1) 3387, 3337, 1743, 1663, 1552, 1364; H NMR
(δ ppm, J Hz, CDCl3) 7.98 (tb, 1H, J = 5.9), 7.42−7.29 (m, 10H), 5.47
(dd, 1H, J = 6.7, J′ = 2.7), 4.77 (d, 1H, J = 6.7), 4.72 (d, 1H, J = 11.1),
4.44 (d, 1H, J = 11.1), 3.87 (d, 1H, J = 4.1), 3.57−3.50 (m, 1H), 3.37−
3.31 (m, 1H), 3.15 (d, 1H, J = 2.6), 2.50−2.43 (m, 2H), 2.01−1.93
(m, 2H), 1.88 (d, 1H, J = 8.6), 1.60 (s, 3H), 1.37−1.29 (m, 1H),
1.13−1.04 (m, 1H), 0.95 (t, 3H, J = 7.3), 0.79 (d, 3H, J = 7.0); 13C
NMR (δ ppm, CDCl3) 176.8, 176.1, 138.3, 135.3, 128.9, 128.6, 128.5,
127.7, 127.3, 126.8, 91.4, 79.8, 70.5, 66.4, 65.2, 50.5, 38.5, 36.0, 29.7,
Synthesis and Characterization of Compounds 1a−i. The
corresponding compound 7 (1.0 mmol) was dissolved in DME (5
mL) and cooled to 0 °C. LiOH, 1 N aqueous solution (3 mL), was
added dropwise, and the progress of the reaction was monitored by
TLC. After completion of the reaction, citric acid, 10% aqueous
solution (3 mL, pH ≈ 6), was added. The resulting solution was
extracted with CH2Cl2 (3 × 4 mL), and the combined organic
fractions were dried and evaporated. The crude product was triturated
in Et2O, yielding the corresponding product 1 as a white solid.
{(2S,3R,4S,5S)-3-[(1S)-1-Benzyloxy-2-methylbutyl]-4-nitro-5-
phenyl}prolyl-β-alanine, 1a. 98% yield; IR (KBr, cm−1) 3372, 3331,
1728, 1635, 1549, 1365; 1H NMR (δ ppm, J Hz, CDCl3) 7.73 (tb, 1H,
J = 4.9), 7.38−7.19 (m, 10H), 6.11 (sb, 2H), 5.35 (dt, 1H, J = 5.9, J′ =
1.4), 4.74 (d, 1H, J = 11.1), 4.49 (d, 1H, J = 6.6), 4.47 (d, 1H, J =
11.1), 3.73 (d, 1H, J = 7.4), 3.64 (d, 1H, J = 5.9), 3.59−3.42 (m, 2H),
3.15 (db, 1H, J = 5.1), 2.60 (t, 2H, J = 5.1), 1.92−1.78 (m, 1H), 1.59−
1.37 (m, 1H), 1.30−1.06 (m, 1H), 0.99−0.83 (m, 6H); 13C NMR (δ
ppm, CDCl3) 175.8, 172.3, 138.0, 134.7, 128.4, 127.7, 126.4, 90.8,
82.5, 72.9, 66.9, 63.6, 50.2, 37.1, 34.6, 33.6, 25.8, 14.3, 11.6; mp 140−
141 °C; [α]D25 +63.23 (c 0.99, CH2Cl2). Anal. Calcd for C26H33O6N3:
C, 64.57; H, 6.89; N, 8.69. Found: C, 64.05; H, 6.97; N, 8.62.
{(2S,3R,4S,5S)-3-[(1S)-1-Benzyloxy-2-methylbutyl]-2-methyl-4-
nitro-5-phenyl}prolyl-β-alanine, 1b. 93% yield; IR (KBr, cm−1) 3372,
26.4, 25.0, 19.6, 13.4, 12.0; [α]D +4.8 (c 1.10, CH2Cl2). Anal. Calcd
25
for C28H37O6N3: C, 65.73; H, 7.29; N, 8.21. Found: C, 65.67; H, 7.23;
N, 8.26.
5-{(2S,3R,4S,5S)-3-[(2S)-1-(Benzyloxy)-2-methylbutyl]-4-nitro-5-
phenylpyrrolidine-2-carboxamido}pentanoic Acid, 1f. 70% yield; IR
1
(KBr, cm−1) 3377, 1723, 1663, 1552, 1377; H NMR (δ ppm, J Hz,
CDCl3) 7.41−7.35 (m, 5H), 7.34−7.28 (m, 6H), 5.36 (dd, 1H, J = 6.6,
J′ = 2.3), 4.78 (d, 1H, J = 11.4), 4.58 (d, 1H, J = 6.6), 4.56 (d, 1H, J =
11.4), 3.77 (d, 1H, J = 5.4), 3.71 (d, 1H, J = 7.5), 3.45−3.38 (m, 1H),
3.35−3.29 (m, 1H), 3.08 (d, 1H, J = 7.4), 2.40 (t, 2H, J = 7.1), 1.89−
1.81 (m, 1H), 1.76−1.70 (m, 2H), 1.68−1.62 (m, 2H), 1.55−1.48 (m,
1H), 1.22−1.13 (m, 1H), 0.95 (t, 3H, J = 7.4), 0.87 (d, 3H, J = 6.9);
13C NMR (δ ppm, CDCl3) 177.7, 172.6, 138.3, 135.1, 128.8, 128.6,
128.1, 127.9, 127.6, 126.7, 91.0, 82.8, 73.2, 67.1, 64.0, 50.5, 38.6, 37.3,
33.4, 29.0, 26.1, 21.9, 14.5, 11.7; [α]D25 +18.3 (c 0.70, CH2Cl2). Anal.
Calcd for C28H37O6N3: C, 65.73; H, 7.29; N, 8.21. Found: C, 65.70;
H, 7.33; N, 8.20.
5-{(2S,3S,4S,5S)-3-[(2S)-1-(Benzyloxy)-2-methylbutyl]-2-methyl-4-
nitro-5-phenylpyrrolidine-2-carboxamido}pentanoic Acid, 1g. 75%
1
yield; IR (KBr, cm−1) 3387, 3322, 1733, 1668, 1552, 1372; H NMR
(δ ppm, J Hz, CDCl3) 7.85 (tb, 1H, J = 5.9), 7.43−7.29 (m, 10H), 5.46
(dd, 1H, J = 6.8, J′ = 2.9), 4.75 (d, 1H, J = 6.8), 4.71 (d, 1H, J = 11.1),
4.44 (d, 1H, J = 11.1), 3.86 (d, 1H, J = 4.1), 3.49−3.41 (m, 1H), 3.30−
3.23 (m, 1H), 3.13 (d, 1H, J = 2.7), 2.42 (t, 2H, J = 6.9), 1.99−1.92
(m, 1H), 1.79−1.72 (m, 2H), 1.70−1.63 (m, 2H), 1.58 (s, 3H), 1.36−
1.28 (m, 1H), 1.11−1.02 (m, 1H), 0.95 (t, 3H, J = 7.3), 0.78 (d, 3H, J
1
3316, 1729, 1663, 1555, 1372; H NMR (δ ppm, J Hz, CDCl3) 8.25
(tb, 1H, J = 6.2), 7.38−7.25 (m, 12H), 5.40 (dd, 1H, J = 6.5, J′ = 2.5),
4.69 (d, 1H, J = 6.5), 4.67 (d, 1H, J = 11.1), 4.39 (d, 1H, J = 11.3),
3.81 (d, 1H, J = 3.9), 3.65−3.49 (m, 2H), 3.07 (s, 1H), 2.59 (tb, 2H, J
= 5.5), 2.08−1.81 (m, 1H), 1.52 (s, 3H), 1.38−1.20 (m, 1H), 1.13−
743
dx.doi.org/10.1021/jm3016848 | J. Med. Chem. 2013, 56, 735−747

Journal of Medicinal Chemistry
Article
= 6.9); 13C NMR (δ ppm, CDCl3) 177.9, 175.6, 138.4, 135.6, 128.9,
1.50−1.43 (m, 1H), 1.41−1.35 (m, 1H), 1.02−0.94 (m, 1H), 0.84 (d,
3H, J = 6.8), 0.78 (t, 3H, J = 7.3); 13C NMR (δ ppm, DMSO-d6)
173.0, 171.4, 137.9, 137.0, 132.5, 128.6, 128.0, 127.6, 127.3, 127.0,
125.0, 87.7, 80.7, 74.2, 67.0, 60.1, 58.8, 48.7, 38.1, 35.7, 33.4, 22.4,
15.9, 11.5; mp 115−116 °C; [α]D25 +7.9 (c 0.6, CH2Cl2). Anal. Calcd
for C29H38N4O7: C, 62.80; H, 6.91; N, 10.10. Found: C, 62.66; H,
6.93; N, 10.10.
128.5, 128.4, 127.6, 127.3, 126.9, 91.5, 79.9, 70.4, 66.3, 65.2, 50.5, 38.7,
36.1, 33.4, 29.0, 26.3, 21.9, 19.5, 13.4, 12.0; [α]D −3.7 (c 1.60,
25
CH2Cl2). Anal. Calcd for C29H39O6N3: C, 66.26; H, 7.48; N, 7.99.
Found: C, 66.35; H, 7.43; N, 7.97.
6-{(2S,3R,4S,5S)-3-[(2S)-1-(Benzyloxy)-2-methylbutyl]-4-nitro-5-
phenylpyrrolidine-2-carboxamido}hexanoic Acid, 1h. 57% yield; IR
1
(KBr, cm−1) 3356, 1698, 1658, 1562, 1380; H NMR (δ ppm, J Hz,
5-[(2S,3R,4S,5S)-1-(2-Aminoacetyl)-3-[(2S)-1-(benzyloxy)-2-meth-
ylbutyl]-2-methyl-4-nitro-5-phenylprolylamino]butanoic Acid, 1k.
80% yield; IR (KBr, cm−1) 3417, 1733, 1668, 1562, 1372; 1H NMR (δ
ppm, J Hz, DMSO-d6) 8.00 (t, 1H, J = 4.4), 7.96−7.90 (m, 1H), 7.39−
7.15 (m, 9H), 5.83 (t, 1H, J = 11.4), 5.61 (d, 1H, J = 9.5), 4.65 (d, 1H,
J = 11.3), 4.51 (d, 1H, J = 11.3), 3.65 (dd, 1H, J = 11.9, J′ = 6.8), 3.53
(t, 1H, J = 6.4), 3.46 (d, 1H, J = 15.6), 3.26−3.17 (m, 1H), 3.11−3.04
(m, 1H), 2.46 (d, 1H, J = 16.1), 2.33−2.21 (m, 2H), 1.80−1.68 (m,
2H), 1.61 (s, 3H), 1.50−1.44 (m, 1H), 1.41−1.36 (m, 1H), 1.03−0.94
(m, 1H), 0.84 (d, 3H, J = 6.7), 0.78 (t, 3H, J = 7.4); 13C NMR (δ ppm,
DMSO-d6) 174.5, 171.1, 137.9, 136.6, 132.4, 128.6, 128.3, 128.0,
127.6, 127.3, 127.0, 125.0, 87.6, 80.5, 74.2, 67.2, 60.1, 48.8, 41.7, 38.1,
CDCl3) 7.42−7.36 (m, 5H), 7.35−7.31 (m, 5H), 7.25 (t, 1H, J = 5.6),
5.37 (dd, 1H, J = 6.6, J′ = 2.4), 4.80 (d, 1H, J = 11.4), 4.60 (d, 1H, J =
6.6), 4.58 (d, 1H, J = 11.4), 3.80 (d, 1H, J = 6.5), 3.71 (d, 1H, J = 7.5),
3.47−3.41 (m, 1H), 3.32−3.25 (m, 1H), 3.08 (d, 1H, J = 8.0), 2.37 (t,
2H, J = 7.4), 1.90−1.82 (m, 1H), 1.73−1.67 (m, 2H), 1.65−1.59 (m,
2H), 1.54−1.49 (m, 1H), 1.48−1.42 (m, 2H), 1.23−1.14 (m, 1H),
0.96 (t, 3H, J = 7.4), 0.87 (d, 3H, J = 6.9); 13C NMR (δ ppm, CDCl3)
177.9, 172.5, 138.4, 135.3, 128.8, 128.6, 127.8, 127.5, 126.7, 91.0, 82.8,
73.2, 67.1, 64.0, 50.5, 38.8, 37.3, 33.7, 29.2, 26.2, 26.1, 24.3, 14.5, 11.7;
mp 105−106 °C; [α]D +22.5 (c 1.10, CH2Cl2). Anal. Calcd for
25
C29H39N3O6: C, 66.26; H, 7.48; N, 7.99. Found: C, 66.19; H, 7.50; N,
7.98.
31.6, 24.0, 22.4, 15.8, 11.4; mp 132−133 °C; [α]D −12.3 (c 0.75,
25
CH2Cl2). Anal. Calcd for C30H40N4O7: C, 63.36; H, 7.09; N, 9.85.
Found: C, 63.19; H, 7.12; N, 9.84.
6-{(2S,3R,4S,5S)-3-[(2S)-1-(Benzyloxy)-2-methylbutyl]-2-methyl-
4-nitro-5-phenylpyrrolidine-2-carboxamido}hexanoic Acid, 1i. 77%
1
yield; IR (KBr, cm−1) 3366, 3332, 1733, 1653, 1542, 1367; H NMR
6-[(2S,3R,4S,5S)-1-(2-Aminoacetyl)-3-[(2S)-1-(benzyloxy)-2-meth-
ylbutyl]-2-methyl-4-nitro-5-phenylprolylamino]pentanoic Acid, 1l.
83% yield; IR (KBr, cm−1) 3446, 1715, 1662, 1559, 1374; 1H NMR (δ
ppm, J Hz, DMSO-d6) 7.85−7−68 (m, 2H), 7.53 (m, 1H), 7.36−7.19
(m, 9H), 5.71 (m, 1H), 5.53 (d, 1H, J = 9.4), 4.63 (d, 1H, J = 11.4),
4.54 (d, 1H, J = 11.4), 3.70 (dd, 1H, J = 12.1, J′ = 6.4), 3.53 (t, 1H, J =
6.2), 3.42 (m, 1H), 3.26−3.17 (m, 1H), 3.11−3.04 (m, 1H), 2.45 (s,
1H), 2.29−2.21 (m, 2H), 1.61 (s, 3H), 1.61−153 (m, 4H), 1.52−1.39
(m, 2H), 1.08−0.96 (m, 1H), 0.85 (d, 3H, J = 6.9), 0.80 (t, 3H, J =
7.3); 13C NMR (δ ppm, DMSO-d6) 172.0, 168.3, 137.8, 129.6, 129.1,
128.3, 127.6, 127.4, 127.2, 93.6, 80.7, 73.7, 71.3, 66.8, 48.3, 39.4, 37.6,
(δ ppm, J Hz, CDCl3) 7.83 (tb, 1H, J = 5.9), 7.49−7.28 (m, 10H), 5.46
(dd, 1H, J = 6.8, J′ = 2.9), 4.75 (d, 1H, J = 6.5), 4.71 (d, 1H, J = 11.0),
4.44 (d, 1H, J = 11.1), 3.86 (d, 1H, J = 4.0), 3.56−3.39 (m, 1H), 3.29−
3.14 (m, 1H), 3.12 (d, 1H, J = 2.7), 2.37 (t, 2H, J = 7.1), 2.03−1.86
(m, 1H), 1.77−1.60 (m, 4H), 1.58 (s, 3H), 1.53−1.41 (m, 2H), 1.38−
1.24 (m, 2H), 0.94 (t, 3H, J = 6.8), 0.77 (d, 3H, J = 6.9); 13C NMR (δ
ppm, CDCl3) 178.3, 175.4, 138.4, 135.7, 128.9, 128.5, 128.4, 127.6,
127.3, 126.9, 91.5, 79.8, 70.5, 66.3, 65.1, 50.6, 39.0, 36.0, 33.7, 29.3,
26.3, 26.2, 24.3, 19.5, 13.4, 12.0; [α]D +4.9 (c 0.60, CH2Cl2). Anal.
25
Calcd. for C30H41N3O6: C, 66.77; H, 7.66; N, 7.79. Found: C, 66.56;
H, 7.60; N, 7.75.
35.3, 28.6, 23.6, 22.4, 18.1, 16.1, 12.1; mp 112−113 °C; [α]D +31.8
25
(c 0.65, acetone). Anal. Calcd for C31H42N4O7: C, 63.90; H, 7.27; N,
9.62. Found: C, 63.80; H, 7.20; N, 9.50.
Synthesis of Compounds 8j−l. The corresponding compound 7
(1.0 mmol) was dissolved in dry DMF (10 mL), and trifluoroacetic
acid (0.77 mL, 10.0 mmol) was added dropwise. The mixture was
stirred at room temperature for 5 h. Then, solvents were evaporated
under reduced pressure and the resulting salt was dissolved in dry
DMF (10 mL) under argon atmosphere. Et3N (0.22 mL, 1.6 mmol)
and N-phthaloylglycine anhydride (1.26 g, 3.0 mmol) were added, and
the mixture was stirred at room temperature for 48 h. After completion
of the reaction, the mixture was partitioned between H2O (5 mL) and
CH2Cl2 (5 mL). The organic phase was washed with 5 mL fractions of
HCl, 1 N aqueous solution, NaHCO3 saturated aqueous solution,
H2O, and NaCl saturated aqueous solution, dried (Na2SO4), and
evaporated. The crude mixture was purified by flash chromatography
(ethyl acetate/hexanes, 1:2), yielding the corresponding compound 8.
The characterization of these intermediate compounds can be found in
the Supporting Information.
Computational Simulations and Docking. The crystal structure
of the complex between the I-domain of LFA-1 (disulfide bridged
mutant L161C, F299C) and domains 1−3 of ICAM-1 (PDB entry
1MQ8) was used as the starting structure. Only the first domain of
ICAM1, which is bound directly to the I-domain, was retained. This
reduced complex (see Figure 1) was prepared for computational study
by the following procedure. The energy minimized complex was
solvated in a 74 Å × 74 Å × 74 Å box of TIP3P water molecules. Bond
lengths involving bonds between heavy atoms and hydrogen atoms
were constrained using the SHAKE algorithm during the whole
process. Periodic boundary conditions were set with the complex fixed,
and the water molecules were equilibrated for 20 ps at 300 K. The
constraints were removed, and the entire system was again equilibrated
for 20 ps. Van der Waals interactions were truncated at a cutoff
distance of 12 Å using a switch function. Electrostatic interactions were
truncated at 12.5 Å using a shift function. All calculations were done
using the CHARMM35 program. Docking of compound 1b to the
MIDAS site of the mutant I-domain was done by a Monte Carlo
search procedure followed by molecular dynamics simulations (MD)
using the CHARMM program. The free energy decomposition
analysis was done by the molecular mechanics Poisson−Boltzmann
surface area method36 following protocols previously described.37 See
Supporting Information for further details.
The model of compound 1b bound to the IDAS site of the I-
domain was made with the program HADDOCK.38 The residues of
the wild type αL I-domain whose signals had changed in the HSQC
spectra in the presence of compound 1b above a threshold of the
average plus the experimental error were defined as “active” residues
for the docking with this program. One-thousand structures were
generated of the compound 1b/I-domain complex using the PDB file
1ZOP as an initial structure for the I-domain. The most energetically
favorable docked structure was minimized and subjected to a 3 ns
molecular dynamics simulation using the CHARMM 27 program and
following the same protocol as that explained above. Then 250 ps of a
Synthesis of Compounds 1j−l. To the corresponding compound
8 (1.0 mmol) dissolved in absolute EtOH (15 mL) hydrazine (0.1
mL) was added dropwise, and the mixture was heated at reflux for 24
h. Then the reaction mixture was allowed to reach room temperature
and the precipitate was removed by filtration. The filtrate was
evaporated under reduced pressure, and the residue was dissolved in
DME (5 mL) and cooled to 0 °C. LiOH, 1 N aqueous solution (3
mL), was added dropwise, and the mixture was stirred at room
temperature for 3 h. After completion of the reaction, citric acid, 10%
aqueous solution (3 mL, pH ≈ 6), was added. The resulting solution
was extracted with CH2Cl2 (5 × 4 mL), and the combined organic
fractions were dried and evaporated. The crude product was triturated
in Et2O, yielding the corresponding product 1 as a white solid.
[(2S,3R,4S,5S)-1-(2-Aminoacetyl)-3-[(2S)-1-(benzyloxy)-2-methyl-
butyl]-2-methyl-4-nitro-5-phenylprolylamino]-β-alanine, 1j. 72%
1
yield; IR (KBr, cm−1) 3407, 1718, 1673, 1567, 1382; H NMR (δ
ppm, J Hz, DMSO-d6) 7.96 (t, 1H, J = 5.5), 7.93−7.86 (m, 1H), 7.38−
7.17 (m, 9H), 5.84 (t, 1H, J = 10.7), 5.57 (d, 1H, J = 7.9), 4.64 (d, 1H,
J = 11.3), 4.50 (d, 1H, J = 11.4), 3.66 (dd, 1H, J = 11.7, J′ = 6.7), 3.51
(t, 1H, J = 5.3), 3.41−3.24 (m, 3H), 2.48−2.33 (m, 3H), 1.59 (s, 3H),
744
dx.doi.org/10.1021/jm3016848 | J. Med. Chem. 2013, 56, 735−747

Journal of Medicinal Chemistry
Article
stable part of the simulation was subsequently used to calculate an
average structure that was used as a model of the complex.
the mean standard deviation (SD) of three separate experiments,
each in sextuplicate (n = 18).
Cloning, Expression, and Purification of the LFA-1 I-Domain.
The clone of the I-domain of the human integrin αL chain (Uniprot
code P20701) used here codes for residues 128−307 of the αL chain
plus an initial methionine (residue numbering refers to the chain after
processing of the N-terminal 25-residue long signal peptide). A
synthetic gene of this domain with codons optimized for expression in
E. coli (Entelechon GmbH) was subcloned into the pET14d
expression vector. The high affinity I-domain mutant was produced
by introducing mutations K287C and K294C by site-directed
mutagenesis, as previously described.20 Both clones were expressed
at 37 °C in BL21 (DE3) cells, and purification of the wild type and
active forms of the LFA-1 I-domain was performed as described18,20
except that a single size-exclusion chromatography on a Superdex75
column (GE Health Care) was used after refolding. Protein purity was
evaluated by SDS−PAGE. Identity was confirmed by mass
spectrometry, and concentration was measured by absorbance at 280
nm using the extinction coefficients calculated from the amino acid
composition. Single (15N) and double (13C,15N) uniformly labeled
samples of the wild type and mutant I-domains were produced by
expression in minimal medium containing 15N-NH4Cl and both 15N-
NH4Cl and 13C-D-glucose, respectively.
NMR Spectroscopy and Data Analysis. NMR spectra were
recorded at 22 °C on a 600 MHz Bruker Avance spectrometer with a
TXI cryoprobe, of samples in 10 mM sodium phosphate, pH 7.4, 150
mM NaCl, 1 mM MgCl2 3% (v/v), 2H2O, and 80 μM 2,2-dimethyl-2-
silapentane-5-sufonic acid (DSS) as an internal chemical shift
reference. The assignment of the backbone resonances of the LFA-1
I-domain using triple resonance experiments recorded on 13C,15N
labeled I-domain protein samples has already been reported39 and
deposited (BMRB code 18941), and the same strategy was used here
for the assignment of the mutant I-domain. The solubility in the NMR
sample buffer was determined at 22 °C for all the compounds and
lovastatin in the absence or presence of 0−10% perdeuterated DMSO,
and this information was used in the design of the binding and the
titration experiments. The binding site of the compounds was mapped
analyzing the CSP of the signals assigned in the 1H−15N HSQC
spectrum of the I-domain in the presence and absence of compound.
All signals whose CSP was larger than the average plus the
experimental error were considered to be significantly perturbed by
the presence of the added compound. Likewise, those peaks that had
broadened to the point of disappearance were considered significantly
affected. The dissociation constant KD of the I-domain complexes with
compounds 1b, 1e, 1m, and lovastatin was calculated analyzing the
CSPs resulting from changes in the amide signal of the I-domain
residue Thr291 upon titration with the corresponding compound.
Backbone 15N relaxation measurements and analysis were performed
essentially as described.40 Compound 1b disruption of the LFA-1 I-
domain/ICAM-1 interaction was investigated on a 100 μM 15N-
labeled I-domain sample in PBS plus 10 mM MgSO4, 3 mM HEPES,
0.01% sodium azide, pH 7.4, prepared with or without 100 μM D1D2-
ICAM-1. This molecule contains the first two extracellular soluble
domains of ICAM-1 produced in CHO cells as described.41 NMR
spectra of these samples with or without 320 μM compound 1b were
recorded (4 h long), and the changes in the intensity of the cross-
peaks were measured by volume integration.
For the hepatic metastasis assay with compound 1b, hepatic
metastases were produced by the intrasplenic injection of 1.5 × 105
viable CT26-CC cells suspended in 0.1 mL of Hank’s balanced salt
solution into 6- to 8-week-old anesthetized male BALB/c mice (n = 10
per group; three independent experiments). Prior to their inoculation
the cells were preincubated for 30 min with or without 10 μg/mL (20
μg per 5 × 105 cells) of compound 1b solubilized with
poly(amidoamine) dendrimer generation 3.5 (PAMAM G3.5) at a
4:1 ratio (compound/dendrimer). The rationale for this preincubation
time is to evaluate the inhibition of the adhesion of the injected cells to
the endothelial cells of the mice as the early event in the process of
metastasis. In this way, we do something analogous to the in vitro cell
adhesion assays where the CT26-CC cells are incubated for 30 min
with the compounds before the adhesion to HSE cells. The duration of
30 min is the same as the preincubation of cells with the anti-LFA1
antibody in the in vitro adhesion assays and comes from established
procedures used in analogous cell adhesion assays (with B16 M and
HSE cells) as previously described.46 The same protocol (30 min
preincubation) was used previously to evaluate the antimetastatic
activity of VLA-4 inhibitors.17 Mice were killed by cervical dislocation
on the 14th day after the injection of cancer cells. Livers were
removed, fixed, and processed for histological analyses. An integrated
image analysis system (Olympus Microimage 4.0 capture kit)
connected to an Olympus BX51TF microscope was used to quantify
the number of foci and their average diameters in serially cut and
hematoxylin/eosin stained hepatic tissue sections. Densitometric
analysis of digitalized microscopic images was used to distinguish
metastatic tissue from normal hepatic tissue. Previously described
stereological procedures were employed,42 and the following
parameters were calculated: the liver metastasis density, which was
the number of metastases per 100 mm3 of liver (based on the mean
number of foci detected in fifteen 10 × 10 mm2 sections per liver) and
the liver metastasis volume (mean percentage of liver volume occupied
by metastases).
For the antiproliferative assays of compounds 1h and 1j, CT26-CC
cells (5 × 105) were subcutaneously injected into BALB/c mice. Mice
received daily subcutaneous injections of 2.5 mg/kg compound 1h/
PAMAMG3.5 or compound 1j/PAMAM G3.5. Primary tumors were
removed on day 15, and tumor volume was measured. Expression of
Ki67 was scored by immunohistochemistry, counting the number of
positively stained cells, and was expressed as a percentage of the total
tumor cells counted across 10 randomly selected fields of the section
examined under the microscope (average standard deviation).
Differences in average values were evaluated for statistical
significance by ANOVA and Bonferroni’s or Tamhane’s post hoc
test for in vitro cell adhesion in vivo assays, respectively.
Further details on methods and materials are given as Supporting
Information.
ASSOCIATED CONTENT
■
S
* Supporting Information
Complete chemical characterization of intermediates 2a,b, 7a−
i, 8j−l, NMR spectra of the final products 1a−l, detailed
experimental and computational procedures, figures of addi-
tional in vitro assays for validation of several of the compounds,
figures with additional NMR data on compound binding to the
I-domain, and tables with data on the docking studies. This
material is available free of charge via the Internet at http://
pubs.acs.org.
Cell Adhesion Assays and in Vivo Tumor Growth Experi-
ments. Quantitative evaluation of in vitro intercellular adhesion was
carried out based on previously validated bioassays.42 In the absence of
signaling events, the integrin in the cell surface is in a state of low
affinity for its ICAM-1 ligand (bent heterodimer and closed I-
domain).43 Therefore, in the assays, cell adhesion is stimulated by cell
treatment with VEGF or with PMA. VEGF increases the expression of
the two LFA-1 chains in monocytes.44 PMA does not affect LFA-1
expression, conformation, or affinity for soluble ICAM-1 but facilitates
the ligand-dependent clustering of the integrins on the cell membrane,
thus increasing the avidity of the cells for ICAM-1.45 For PBL
adhesion to immobilized ICAM-1, the commercial form containing the
five extracellular domains (R&D Systems) was used. All the results are
AUTHOR INFORMATION
■
Corresponding Author
*For F.P.C.: phone, 34943015442; e-mail, fp.cossio@ehu.es.
For F.J.B.: phone, 34946572521; e-mail, fblanco@cicbiogune.
es.
745
dx.doi.org/10.1021/jm3016848 | J. Med. Chem. 2013, 56, 735−747

Journal of Medicinal Chemistry
Article
metastasize in the liver of nude mice. Jpn. J. Cancer Res. 1996, 87, 153−
160.
Present Addresses
●Innoprot S.L., 48160 Derio, Spain.
△Institute of Applied Molecular Medicine, CEU-San Pablo
University, 28668 Boadilla del Monte, Spain.
(4) Gulubova, M. V. Expression of cell adhesion molecules, their
ligands and tumour necrosis factor alpha in the liver of patients with
metastatic gastrointestinal carcinomas. Histochem. J. 2002, 34, 67−77.
(5) Miele, M. E.; Bennett, C. F.; Miller, B. E.; Welch, D. R. Enhanced
metastatic ability of TNF-alpha-treated malignant melanoma cells is
reduced by intercellular adhesion molecule-1 (ICAM-1, CD54)
antisense oligonucleotides. Exp. Cell Res. 1994, 214, 231−241.
(6) Wang, H. H.; McIntosh, A. R.; Hasinoff, B. B.; MacNeil, B.;
Rector, E.; Nance, D. M.; Orr, F. W. Regulation of B16F1 melanoma
cell metastasis by inducible functions of the hepatic microvasculature.
Eur. J. Cancer. 2002, 38, 1261−1270.
(7) Harning, R.; Myers, C.; Merluzzi, V. J. Monoclonal antibodies to
lymphocyte function-associated antigen-1 inhibit invasion of human
lymphoma and metastasis of murine lymphoma. Clin. Exp. Metastasis
1993, 11, 337−342.
(8) St-Pierre, Y.; Aoudjit, F.; Lalancette, M.; Potworowski, E. F.
Dissemination of T cell lymphoma to target organs: a post-homing
event implicating ICAM-1 and matrix metalloproteinases. Leuk.
Lymphoma 1999, 34, 53−61.
(9) Fujisaki, T.; Tanaka, Y.; Fujii, K.; Mine, S.; Saito, K.; Yamada, S.;
Yamashita, U.; Irimura, T.; Eto, S. CD44 stimulation induces integrin-
mediated adhesion of colon cancer cell lines to endothelial cells by up-
regulation of integrins and c-Met and activation of integrins. Cancer
Res. 1999, 59, 4427−4434.
(10) Zimmerman, T.; Blanco, F. J. Inhibitors targeting the LFA-1/
ICAM-1 cell-adhesion interaction: design and mechanism of action.
Curr. Pharm. Des. 2008, 14, 2128−2139.
(11) Kallen, J.; Welzenbach, K.; Ramage, P.; Geyl, D.; Kriwacki, R.;
Legge, G.; Cottens, S.; Weitz-Schmidt, G.; Hommel, U. Structural
basis for LFA-1 inhibition upon lovastatin binding to the CD11a I-
domain. J. Mol. Biol. 1999, 292, 1−9.
(12) Gadek, T. R.; Burdick, D. J.; McDowell, R. S.; Stanley, M. S.;
Marsters, J. C., Jr.; Paris, K. J.; Oare, D. A.; Reynolds, M. E.; Ladner,
C.; Zioncheck, K. A.; Lee, W. P.; Gribling, P.; Dennis, M. S.; Skelton,
N. J.; Tumas, D. B.; Clark, K. R.; Keating, S. M.; Beresini, M. H.;
Tilley, J. W.; Presta, L. G.; Bodary, S. C. Generation of an LFA-1
antagonist by the transfer of the ICAM-1 immunoregulatory epitope
to a small molecule. Science 2002, 295, 1086−1089.
(13) Keating, S. M.; Clark, K. R.; Stefanich, L. D.; Arellano, F.;
Edwards, C. P.; Bodary, S. C.; Spencer, S. A.; Gadek, T. R.; Marsters, J.
C., Jr.; Beresini, M. H. Competition between intercellular adhesion
molecule-1 and a small-molecule antagonist for a common binding site
on the alphal subunit of lymphocyte function-associated antigen-1.
Protein Sci. 2006, 15, 290−303.
(14) Shimaoka, M.; Salas, A.; Yang, W.; Weitz-Schmidt, G.; Springer,
T. A. Small molecule integrin antagonists that bind to the beta2
subunit I-like domain and activate signals in one direction and block
them in the other. Immunity 2003, 19, 391−402.
(15) Yang, W.; Shimaoka, M.; Chen, J.; Springer, T. A. Activation of
integrin beta-subunit I-like domains by one-turn C-terminal alpha-helix
deletions. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 2333−2338.
(16) San Sebastian, E.; Mercero, J. M.; Stote, R. H.; Dejaegere, A.;
Cossio, F. P.; Lopez, X. On the affinity regulation of the metal-ion-
dependent adhesion sites in integrins. J. Am. Chem. Soc. 2006, 128,
3554−3563.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
This work was financially supported by the Ministerio de
Ciencia y Competitividad (MINECO) Grant CTQ2011-28680
to F.J.B., Grant CTQ2010-16959 and Consolider-Ingenio 2010
CSD2007-00006 to F.P.C., and Grant IT-324-07 from the
UPV/EHU (UFI QOSYC 11/22) to F.P.C. We thank J. M.
Casasnovas, at the CNB-CSIC, Madrid, Spain, for the generous
́
gift of the D1-D2-ICAM-1 sample, and we thank C. Gonzalez,
at the IQFR-CSIC, Madrid, Spain, for help with puckering
analysis. The SGI/IZO-SGIker UPV/EHU (supported by the
National Program for the Promotion of Human Resources
within the National Plan of Scientific Research, Development
and Innovations, Fondo Social Europeo MCyT) is gratefully
acknowledged for generous allocation of computational
resources in Spain. The authors acknowledge support from
the University of Strasbourg, the Centre National de la
Recherche Scientifique (CNRS), and INSERM. The authors
also acknowledge the Institut du Developpement et des
Ressources en Informatique Scientifique (IDRIS), the Centre
Informatique National de l’Enseignement Super
and the Centre d’Etude du Calcul Parallele de Strasbourg
(Universite de Strasbourg) for generous allocations of
́
ieur (CINES),
̀
́
computer time in France. E.M. was supported by the Ligue
Nationale Contre le Cancer. We gratefully acknowledge EGIDE
and the Minister
support.
̀ ̀
e des Affaires Etrangeres, France for their
DEDICATION
■
́
́
This work is dedicated to Jesus Marıa Rico Rodrıguez.
́
ABBREVIATIONS USED
■
LFA-1, leukocyte function-associated antigen 1; ICAM-1,
intercellular adhesion molecule 1; sICAM-1, soluble intercel-
lular adhesion molecule 1; MIDAS, metal ion dependent
adhesion site; IDAS, I-domain allosteric site; NMR, nuclear
magnetic resonance; DTT, 1,4-dithiothreitol; NOE, nuclear
Overhauser effect; TLC, thin layer chromatography; PDB,
Protein Data Bank; MD, molecular dynamics; CHO, Chinese
hamster ovary; PMA, phorbol miristate acetate; VEGF, vascular
endothelial growth factor; CSP, chemical shift perturbation;
PBS, phosphate buffered saline; CC, colon carcinoma; VLA-4,
very late antigen 1; VCAM-1, vascular cell adhesion molecule 1;
PBL, peripheral blood leukocyte; HSE, hepatic sinusoidal
endothelial; BMRB, Biological Magnetic Resonance Data Bank
(17) Zubia, A.; Mendoza, L.; Vivanco, S.; Aldaba, E.; Carrascal, T.;
Lecea, B.; Arrieta, A.; Zimmerman, T.; Vidal-Vanaclocha, F.; Cossio, F.
P. Application of stereocontrolled stepwise [3+2] cycloadditions to the
preparation of inhibitors of alpha4beta1-integrin-mediated hepatic
melanoma metastasis. Angew. Chem., Int. Ed. 2005, 44, 2903−2907.
(18) Shimaoka, M.; Xiao, T.; Liu, J. H.; Yang, Y.; Dong, Y.; Jun, C.
D.; McCormack, A.; Zhang, R.; Joachimiak, A.; Takagi, J.; Wang, J. H.;
Springer, T. A. Structures of the alpha L I domain and its complex with
ICAM-1 reveal a shape-shifting pathway for integrin regulation. Cell
2003, 112, 99−111.
REFERENCES
■
(1) Gahmberg, C. G. Leukocyte adhesion: CD11/CD18 integrins
and intercellular adhesion molecules. Curr. Opin. Cell Biol. 1997, 9,
643−650.
(2) Yonekawa, K.; Harlan, J. M. Targeting leukocyte integrins in
human diseases. J. Leukocyte Biol. 2005, 77, 129−140.
(3) Yasoshima, T.; Denno, R.; Kawaguchi, S.; Sato, N.; Okada, Y.;
Ura, H.; Kikuchi, K.; Hirata, K. Establishment and characterization of
human gastric carcinoma lines with high metastatic potential in the
liver: changes in integrin expression associated with the ability to
(19) Nagasako, Y.; Misawa, K.; Kohashi, S.; Hasegawa, K.; Okawa, Y.;
Sano, H.; Takada, A.; Sato, H. Evaluation of malignancy using Ki-67
746
dx.doi.org/10.1021/jm3016848 | J. Med. Chem. 2013, 56, 735−747

Journal of Medicinal Chemistry
Article
labeling index for gastric stromal tumor. Gastric Cancer 2003, 6, 168−
172.
(36) Wang, W.; Kollman, P. A. Free energy calculations on dimer
stability of the HIV protease using molecular dynamics and a
continuum solvent model. J. Mol. Biol. 2000, 303, 567−582.
(37) Lafont, V.; Schaefer, M.; Stote, R. H.; Altschuh, D.; Dejaegere,
A. Protein−protein recognition and interaction hot spots in an
antigen−antibody complex: free energy decomposition identifies
“efficient amino acids”. Proteins 2007, 67, 418−434.
(20) Shimaoka, M.; Lu, C.; Palframan, R. T.; von Andrian, U. H.;
McCormack, A.; Takagi, J.; Springer, T. A. Reversibly locking a protein
fold in an active conformation with a disulfide bond: integrin alphaL I
domains with high affinity and antagonist activity in vivo. Proc. Natl.
Acad. Sci. U.S.A. 2001, 98, 6009−6014.
(38) Dominguez, C.; Boelens, R.; Bonvin, A. M. HADDOCK: a
protein−protein docking approach based on biochemical or
biophysical information. J. Am. Chem. Soc. 2003, 125, 1731−1737.
(39) Zimmerman, T.; Oyarzabal, J.; Sebastian, E. S.; Majumdar, S.;
Tejo, B. A.; Siahaan, T. J.; Blanco, F. J. ICAM-1 peptide inhibitors of
T-cell adhesion bind to the allosteric site of LFA-1. An NMR
characterization. Chem. Biol. Drug Des. 2007, 70, 347−53.
(40) Pantoja-Uceda, D.; Arolas, J. L.; Garcia, P.; Lopez-Hernandez,
E.; Padro, D.; Aviles, F. X.; Blanco, F. J. The NMR structure and
dynamics of the two-domain tick carboxypeptidase inhibitor reveal
flexibility in its free form and stiffness upon binding to human
carboxypeptidase B. Biochemistry 2008, 47, 7066−7078.
(41) Casasnovas, J. M.; Stehle, T.; Liu, J. H.; Wang, J. H.; Springer, T.
A. A dimeric crystal structure for the N-terminal two domains of
intercellular adhesion molecule-1. Proc. Natl. Acad. Sci. U.S.A. 1998, 95,
4134−4139.
(42) Vidal-Vanaclocha, F.; Amezaga, C.; Asumendi, A.; Kaplanski, G.;
Dinarello, C. A. Interleukin-1 receptor blockade reduces the number
and size of murine B16 melanoma hepatic metastases. Cancer Res.
1994, 54, 2667−2672.
(43) Carman, C. V.; Springer, T. A. Integrin avidity regulation: are
changes in affinity and conformation underemphasized? Curr. Opin.
Cell Biol. 2003, 15, 547−556.
(44) Heil, M.; Clauss, M.; Suzuki, K.; Buschmann, I. R.; Willuweit,
A.; Fischer, S.; Schaper, W. Vascular endothelial growth factor (VEGF)
stimulates monocyte migration through endothelial monolayers via
increased integrin expression. Eur. J. Cell Biol. 2000, 79, 850−857.
(45) Constantin, G.; Majeed, M.; Giagulli, C.; Piccio, L.; Kim, J. Y.;
Butcher, E. C.; Laudanna, C. Chemokines trigger immediate beta2
integrin affinity and mobility changes: differential regulation and roles
in lymphocyte arrest under flow. Immunity 2000, 13, 759−769.
(46) Mendoza, L.; Carrascal, T.; De Luca, M.; Fuentes, A. M.; Salado,
C.; Blanco, J.; Vidal-Vanaclocha, F. Hydrogen peroxide mediates
vascular cell adhesion molecule-1 expression from interleukin-18-
activated hepatic sinusoidal endothelium: implications for circulating
cancer cell arrest in the murine liver. Hepatology 2001, 34, 298−310.
(21) Huth, J. R.; Olejniczak, E. T.; Mendoza, R.; Liang, H.; Harris, E.
A.; Lupher, M. L., Jr.; Wilson, A. E.; Fesik, S. W.; Staunton, D. E. NMR
and mutagenesis evidence for an I domain allosteric site that regulates
lymphocyte function-associated antigen 1 ligand binding. Proc. Natl.
Acad. Sci. U.S.A. 2000, 97, 5231−5236.
(22) Gaillard, T.; Martin, E.; San Sebastian, E.; Cossio, F. P.; Lopez,
X.; Dejaegere, A.; Stote, R. H. Comparative normal mode analysis of
LFA-1 integrin I-domains. J. Mol. Biol. 2007, 374, 231−249.
(23) Legge, G. B.; Kriwacki, R. W.; Chung, J.; Hommel, U.; Ramage,
P.; Case, D. A.; Dyson, H. J.; Wright, P. E. NMR solution structure of
the inserted domain of human leukocyte function associated antigen-1.
J. Mol. Biol. 2000, 295, 1251−1264.
(24) Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J.
Experimental and computational approaches to estimate solubility and
permeability in drug discovery and development settings. Adv. Drug.
Delivery Rev. 2001, 46, 3−26.
(25) Ferrara, N. Vascular endothelial growth factor: basic science and
clinical progress. Endocr. Rev. 2004, 25, 581−611.
(26) Hicklin, D. J.; Ellis, L. M. Role of the vascular endothelial
growth factor pathway in tumor growth and angiogenesis. J. Clin. Oncol
2005, 23, 1011−1027.
(27) Valcarcel, M.; Arteta, B.; Jaureguibei.tia, A.; Lopategi, A.;
Martinez, I.; Mendoza, L.; Muruzabal, F. J.; Salado, C.; Vidal-
Vanaclocha, F. Three-dimensional growth as multicellular spheroid
activates the proangiogenic phenotype of colorectal carcinoma cells via
LFA-1-dependent VEGF: implications on hepatic micrometastasis. J.
Transl. Med. 2008, 6, 57.
(28) Salas, A.; Shimaoka, M.; Chen, S.; Carman, C. V.; Springer, T.
Transition from rolling to firm adhesion is regulated by the
conformation of the I domain of the integrin lymphocyte function-
associated antigen-1. J. Biol. Chem. 2002, 277, 50255−50262.
(29) Salas, A.; Shimaoka, M.; Kogan, A. N.; Harwood, C.; von
Andrian, U. H.; Springer, T. A. Rolling adhesion through an extended
conformation of integrin alphaLbeta2 and relation to alpha I and beta
I-like domain interaction. Immunity 2004, 20, 393−406.
(30) Nam, K.; Maiorov, V.; Feuston, B.; Kearsley, S. Dynamic control
of allosteric antagonism of leukocyte function antigen-1 and
intercellular adhesion molecule-1 interaction. Proteins 2006, 64,
376−384.
(31) Lu, C.; Shimaoka, M.; Zang, Q.; Takagi, J.; Springer, T. A.
Locking in alternate conformations of the integrin alphaLbeta2 I
domain with disulfide bonds reveals functional relationships among
integrin domains. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 2393−2398.
(32) Lu, C.; Shimaoka, M.; Ferzly, M.; Oxvig, C.; Takagi, J.; Springer,
T. A. An isolated, surface-expressed I domain of the integrin
alphaLbeta2 is sufficient for strong adhesive function when locked in
the open conformation with a disulfide bond. Proc. Natl. Acad. Sci.
U.S.A. 2001, 98, 2387−2392.
(33) Weitz-Schmidt, G.; Welzenbach, K.; Brinkmann, V.; Kamata, T.;
Kallen, J.; Bruns, C.; Cottens, S.; Takada, Y.; Hommel, U. Statins
selectively inhibit leukocyte function antigen-1 by binding to a novel
regulatory integrin site. Nat. Med. 2001, 7, 687−692.
(34) Welzenbach, K.; Hommel, U.; Weitz-Schmidt, G. Small
molecule inhibitors induce conformational changes in the I domain
and the I-like domain of lymphocyte function-associated antigen-1.
Molecular insights into integrin inhibition. J. Biol. Chem. 2002, 277,
10590−10598.
(35) Brooks, B. R.; Bruccoleri, R. E.; Olafson, B. D.; States, D. J.;
Swaminathan, S.; Karplus, M. CHARMM: a program for macro-
molecular energy, minimization, and dynamics calculations. J. Comput.
Chem. 1983, 4, 187−217.
747
dx.doi.org/10.1021/jm3016848 | J. Med. Chem. 2013, 56, 735−747