Cancer Cell
Article
Effective Suppression of Vascular Network Formation
by Combination of Antibodies Blocking VEGFR
o
at
,
m
C
In
cu
a
hibition of endothelial sprouting and vascular networkAntiangiogenic therapeutics are nowwidely used in the treatmnot all patients respond, whereas others become refractory to therapy. Thus there is high demand for a second generation of
improved antiangiogenic therapeutics. Here we report a monoclonal antibody that acts by inhibiting the formation ofVEGFR-3 homodimers and VEGFR-3/VEGFR-2 heterodimers, which arewell-established targets for antiangiogenic therapy.
Our data suggest that employing a combination of ligand binding and dimerization inhibitors provides more effective block-
ing of VEGFR activation for enhanced inhibition of tumor angiogenesis and lymphangiogenesis in vivo.formation in vivo. These results suggest that receptor dimerization inhibitors could be used to enhance anti-
angiogenic activity of antibodies blocking ligand binding in tumor therapy.
INTRODUCTION
Angiogenesis is the growth of new blood vessels from preexist-
ing vasculature. The importance of angiogenesis for the growth
of tumors was realized decades ago (Folkman, 1971), and the
first antiangiogenic agents have recently been approved for clin-
ical use (Jain et al., 2006). Although these treatments have been
highly successful in the treatment of many types of solid tumors,
most patients are either refractory or eventually acquire resis-
tance to antiangiogenic therapy (Jain et al., 2009; Crawford
and Ferrara, 2009). Therefore novel antiangiogenic therapeutics
are needed to complement existing therapies.
Vascular endothelial growth factors (VEGFs) stimulate angio-
genesis and lymphangiogenesis by activating VEGF receptor
(VEGFR) tyrosine kinases in endothelial cells (Tammela et al.,
2005). VEGFR-3 (also known as Flt4) belongs to this family
that, in addition, comprises VEGFR-1/Flt-1 and VEGFR-2/KDR/
Flk-1 (Alitalo et al., 2005; Shibuya and Claesson-Welsh, 2006).
Mice deficient in the Vegfr3 gene die in utero due to abnormal
development of the blood vasculature resulting in cardiovascular
failure (Dumont et al., 1998). On the other hand, loss of the
VEGFR-3 ligand Vegfc results in embryonic lethality due to lack
of lymphatic vessel formation (Ka¨rkka¨inen et al., 2004).
The Vegfr3 gene is expressed in the entire vasculature of the
developing embryo, but expression becomes restricted to the
lymphatic system and a few specialized fenestrated blood vessel
endothelia in adults (Partanen et al., 2000; Kaipainen et al., 1995).
However, expression of VEGFR-3 is again induced in the
Significance
ent of solidmalignancies, but according to clinical experienceimproves VEGFR inhibition and results in stronger inSUMMARY
Antibodies that block vascular endothelial growth factor (VEGF) have become an integral part of antiangio-
genic tumor therapy, and antibodies targeting other VEGFs and receptors (VEGFRs) are in clinical trials.
Typically receptor-blocking antibodies are targeted to the VEGFR ligand-binding site. Here we describe
a monoclonal antibody that inhibits VEGFR-3 homodimer and VEGFR-3/VEGFR-2 heterodimer formation,
signal transduction, as well as ligand-induced migration and sprouting of microvascular endothelial cells.
Importantly, we show that combined use of antibodies blocking ligand binding and receptor dimerizationDOI 10.1016/j.ccr.2010.11.001*Correspondence: kari.alitalo@helsinki.fiLigand Binding and Recept
Denis Tvorogov,
1
Andrey Anisimov,
1,6
Wei Zheng,
1,6
Veli-M
Wolfgang Holnthoner,
1,7
Hanna Helotera¨,
1
Tanja Holopainen
Pa¨ivi M. Ojala,
4,5
and Kari Alitalo
1,
*
1
Molecular/Cancer Biology Laboratory, Research Programs Unit, Bio
Department of Pathology, Haartman Institute and Helsinki University
2
Protein Chemistry Research Group, Institute of Biotechnology
3
Peptide and Protein Laboratory, Department of Virology, Haartman
4
Genome-Scale Biology Program, Research Programs Unit, Biomedi
University of Helsinki, Helsinki, Finland
5
Foundation for the Finnish Cancer Institute
6
These authors contributed equally to this work
7
Present address: Ludwig Boltzmann Institute, A-1200 Vienna, Austri630 Cancer Cell 18, 630–640, December 14, 2010 ª2010 Elsevier Incr Dimerization
ti Leppa¨nen,
1
Tuomas Tammela,
1
Simonas Laurinavicius,
4
1
Michael Jeltsch,
1
Nisse Kalkkinen,
2
Hilkka Lankinen,
3
edicum Helsinki, Institute for Molecular Medicine Finland,
entral Hospital
stitute
m Helsinki and Institute of Biomedicine.

angiogenic blood vascular endothelium in tumors (Valtola et al.,
1999; Partanen et al., 2000; Tammela et al., 2008). Several
studies have shown that interference with VEGFR-3 function
inhibits tumor lymphangiogenesis and metastasis in mice (He
et al., 2002; Alitalo et al., 2005). Specific targeting of VEGFR-3
can be achieved by small molecular weight tyrosine kinase inhib-
itors (Heckman et al., 2008), by trapping the VEGFR-3 ligands
VEGF-C and VEGF-D with soluble extracellular domain of
VEGFR-3 (Burton et al., 2008; Jeon et al., 2008; Ma¨kinen et al.,
2001) or by VEGFR-3 blocking monoclonal antibodies (Persaud
et al., 2004). Recent evidence has also suggested that blocking
VEGFR-3 can improve the inhibition of tumor growth obtained
with other antiangiogenic therapies (Tammela et al., 2008).
Current VEGFR-2 and VEGFR-3 blocking antibodies are
directed against the ligand binding domains of these receptors
(Hicklin et al., 2001; Witte et al., 1998; Pytowski et al., 2005;
Persaud et al., 2004). Thus far other types of function-blocking
antibodies against the VEGFR tyrosine kinases have not been
described. Analogies to alternative modes of receptor tyrosine
kinase inhibition can be derived from published work on the
epidermal growth factor receptor (EGFR) family. For example,
trastuzumab, a potent anti-ErbB2 antibody, inhibits the activa-
tion of this ligand-less receptor via mechanisms apparently
involving antibody-dependent cellular cytotoxicity or inhibition
of the proteolytic shedding of ErbB2 (Valabrega et al., 2007;
Hynes and Lane, 2005).
Here we report on antibodies directed against the VEGFR-3
extracellular domain that inhibit receptor dimerization, and
show that their combination with antibodies blocking ligand
binding results in more effective inhibition of VEGFR-3 activation
and vascular network formation in vitro and in vivo.
RESULTS
Characterization of the 2E11 Anti-VEGFR-3 Antibody
Monoclonal antibodies against the extracellular domain of
VEGFR-3 were tested for blocking of VEGFR-3 activation and
survival/proliferation of BaF3 cells expressing a VEGFR-3/eryth-
ropoietin (Epo) receptor chimera. In the absence of IL-3 these
cells survive only in the presence of a VEGFR-3 ligand in the
culture medium (Ma¨kinen et al., 2001). Figure 1A shows
a comparison of four antibodies in this assay. As can be seen
from the figure, the addition of increasing amounts of the 2E11
antibody, but not of 9D9 or AFL4 antibodies to the medium con-
taining 25 ng/ml human recombinant VEGF-C inhibited the
survival of the cells. The previously publishedmonoclonal human
anti-human VEGFR-3 antibody 3C5 (Persaud et al., 2004) and
the previously published VEGFR-3-Ig soluble receptor (Ma¨kinen
et al., 2001) were used as positive controls for VEGFR-3 inhibi-
tion. The inhibition of VEGFR-3 activation by the 2E11 and 3C5
antibodies was confirmed by using VEGF-C induced VEGFR-3
phosphorylation in endothelial cells (Figure 1B).
A common mechanism for antibody inhibition of receptor
Cancer Cell
A More Effective VEGFR Blocking Mechanismactivation is to block ligand binding to the receptor. It has been
shown that the 3C5 antibody strongly inhibits the binding of
VEGF-C to VEGFR-3 and the VEGF-C-induced mitogenic
response in cells that expresses a chimeric human VEGFR-3-
FMS receptor (Persaudet al., 2004). However, unlike the3C5anti-
body, the 2E11 antibody did not block the binding of VEGFR-3
Canextracellular domain to immobilized VEGF-C (Figures 1C and
1D). These data indicated that although 2E11 and3C5both inhibit
VEGFR-3 activation, their inhibition mechanisms are different.
The binding epitopes of AFL4 and 9D9 were mapped to linear
peptide sequences in VEGFR-3 immunoglobulin homology
domain 5 (D5) and D6, respectively (see Figures S1A and S1B
available online). In contrast, the 2E11 binding site could not
be mapped to a linear epitope, nor was the epitope in the ligand
binding region (D1–D3), as this antibody recognized VEGFR-3
where this region had been deleted (Figure S1C). Furthermore,
2E11 bound to nonreduced but not to reduced VEGFR-3
polypeptides in western blotting analysis (Figure S2B), suggest-
ing that the epitope is conformational and sensitive to denatur-
ation of VEGFR-3. Figure S1D shows the K
d
values for 2E11,
9D9, and AFL4 obtained from surface plasmon resonance
analysis using monomeric VEGFR-3D1-7. Because the 2E11
antibody bound better to the nonreduced receptor, we searched
for the binding epitope in D5 that undergoes proteolytic cleavage
after receptor biosynthesis, rendering the remaining fragments
bound by a disulfide bridge (Pajusola et al., 1994).
A Polypeptide Loop Extending from the VEGFR-3 D5
Is Critical for 2E11 Antibody Binding and Receptor
Activation
Figure S2A shows the sequence comparison of D5 in human and
mouse VEGFR-3 and VEGFR-2. The proteolytic cleavage site in
VEGFR-3 (Lee et al., 1996) is marked with a red arrowhead and
the cysteine residues are marked red. Figures S2C and S2D
showacomputermodel of aVEGFR-3D5-relatedstructurebased
on the immunoglobulin-homologydomainofmyelin basicprotein-
C(KelleyandSternberg, 2009; Idowuet al., 2003). InVEGFR-3D5,
the extended loop (underlined in Figure S2A; containing the
SLRRRQQQ sequence) would contain the cleavage site between
R472 and S473 (red arrowhead in Figures S2A and S2C). In Fig-
ure S2D the surface of the immunoglobulin homology domain is
colored red for negative charge and blue for positive charge.
Although no actual data is available for a possible D5-D5 interac-
tion, this model suggested a possible scenario where the posi-
tively charged residues of the elongated loop ‘‘arm’’ could contact
the negatively charged surface of the ‘‘armpit,’’ thus contributing
to dimer stabilization and activation of the receptor.
Figure 2A schematically outlines themutagenesis strategy used
to interrogate the importanceofD5and itselongated,cleaved loop
structure for 2E11 antibody binding and receptor function. The
disulfidebonds in the figure are hypothetical andbasedondeduc-
tions fromtheD5model. TheeffectofD5cysteine toserine residue
replacements on VEGFR-3 expression, cleavage and autophos-
phorylation in transfected 293T cells in the absence and presence
ofVEGF-Careshown inFigure2B (left panel). The transfectedcells
were analyzed by VEGFR-3 immunoprecipitation and western
blotting using anti-phosphotyrosine (pY) or VEGFR-3 antibodies.
As can be seen from the results, the C445S and C534Smutations
and their combination prevented receptor autophosphorylationand processing. The C466S mutation decreased VEGFR-3
expression levels while retaining at least some phosphorylation,
and blocked cleavage of the receptor, whereas C486S allowed
both processing and ligand-induced phosphorylation.
A similar analysis was carried out with a chimeric VEGFR-3
receptor where the loop region was substituted with the
cer Cell 18, 630–640, December 14, 2010 ª2010 Elsevier Inc. 631