Stefano J.Mandriota
1
, Lotta Jussila
2
,
Michael Jeltsch
2
, Amelia Compagni
3
,
Danielle Baetens, Remko Prevo
4
,
Suneale Banerji
4
, Joachim Huarte,
Roberto Montesano, David G.Jackson
4
,
Lelio Orci, Kari Alitalo
2
, Gerhard Christofori
3
and Michael S.Pepper
5
Department of Morphology, University Medical Centre, 1 rue Michel
Servet, 1211 Geneva 4, Switzerland,
2
Molecular/Cancer Biology
Laboratory and Ludwig Institute for Cancer Research, Haartman
Institute, University of Helsinki, Finland,
3
Institute of Molecular
Pathology, Vienna, Austria and
4
MRC Human Immunology Unit,
Institute of Molecular Medicine, John Radcliffe Hospital, Oxford, UK
1
Present address: The Wellcome Trust Centre for Human Genetics,
Roosevelt Drive, Oxford, UK
5
Corresponding author
e-mail: michael.pepper@medecine.unige.ch
Metastasis is a frequent and lethal complication
of cancer. Vascular endothelial growth factor-C
(VEGF-C) is a recently described lymphangiogenic
factor. Increased expression of VEGF-C in primary
tumours correlates with dissemination of tumour cells
to regional lymph nodes. However, a direct role for
VEGF-C in tumour lymphangiogenesis and subse-
quent metastasis has yet to be demonstrated. Here we
report the establishment of transgenic mice in which
VEGF-C expression, driven by the rat insulin pro-
moter (Rip), is targeted to b-cells of the endocrine
pancreas. In contrast to wild-type mice, which lack
peri-insular lymphatics, RipVEGF-C transgenics
develop an extensive network of lymphatics around
the islets of Langerhans. These mice were crossed
with Rip1Tag2 mice, which develop pancreatic b-cell
tumours that are neither lymphangiogenic nor meta-
static. Double-transgenic mice formed tumours
surrounded by well developed lymphatics, which
frequently contained tumour cell masses of b-cell
origin. These mice frequently developed pancreatic
lymph node metastases. Our ®ndings demonstrate
that VEGF-C-induced lymphangiogenesis mediates
tumour cell dissemination and the formation of lymph
node metastases.
Keywords: islet of Langerhans/lymphangiogenesis/
tumour metastasis/VEGF-C
Introduction
The metastatic spread of tumour cells is responsible for the
majority of cancer deaths. Tumour cell dissemination is
mediated by a number of mechanisms, including local
tissue invasion, lymphatic spread, haematogenous spread,
or direct seeding of body cavities or surfaces. Clinical and
pathological observations have long suggested that, for
many tumours, the most common pathway of initial
dissemination is via lymphatics, with patterns of spread
via afferent vessels following routes of natural drainage
(reviewed by Fidler, 1997; Cotran et al., 1999; Sleeman,
2000). However, the lymphatic system has traditionally
been overshadowed by the greater emphasis placed on the
blood vascular system. This has been due in part to the
absence of suitable markers that distinguish lymphatic
from blood vascular endothelium, and to the lack of
identi®cation of lymphatic-speci®c growth factors.
In recent years, these limitations have been relieved by
the discovery of a small number of potential lymphatic-
speci®c markers (reviewed by Jackson, 2001). These
include: LYVE-1, a lymphatic endothelial receptor for
the extracellular matrix/lymphatic ¯uid mucopolysac-
charide hyaluronan (Banerji et al., 1999); Prox-1, a
homeobox gene product involved in regulating early
lymphatic development (Wigle and Oliver, 1999); podo-
planin, a glomerular podocyte membrane mucoprotein
(Breiteneder-Geleff et al., 1999); and the vascular
endothelial growth factor receptor-3 (VEGFR-3), a
transmembrane tyrosine kinase receptor for the lymphatic
endothelial growth factors vascular endothelial growth
factor-C (VEGF-C) and VEGF-D (reviewed in Veikkola
et al., 2000). Targeted inactivation of the VEGFR-3 gene
has revealed that it also plays an important role in the
development of the early blood vascular system, prior to
the emergence of lymphatic vessels (Dumont et al., 1998).
VEGF-C, the ®rst ligand to be discovered for VEGFR-3
(Joukov et al., 1996; Lee et al., 1996), is a member of the
VEGF family of polypeptide growth factors, which
comprises VEGF-A, -B, -C, -D and orf virus VEGFs (or
VEGF-E) (reviewed in Eriksson and Alitalo, 1999;
Ferrara, 1999). VEGF-C is produced in a pre-pro-peptide
form that is proteolytically processed to a mature
homodimer of ~40 kDa (Joukov et al., 1997). Proteolytic
processing increases the af®nity of VEGF-C for VEGFR-3
some 400-fold, and also enables it to bind to and activate
VEGFR-2 (Joukov et al., 1997). Based on its expression
pro®le and its binding to VEGFR-3, VEGF-C has been
implicated in the development of the lymphatic system
(Kukk et al., 1996; Lymboussaki et al., 1999). In addition,
transgenic overexpression of VEGF-C using the keratin 14
promoter induces lymphatic vessel enlargement/dilatation
in the skin (Jeltsch et al., 1997), and recombinant VEGF-C
induces lymphangiogenesis in the chick chorioallantoic
membrane (Oh et al., 1997). The capacity of VEGF-C to
bind to and activate VEGFR-2 may partially explain why
it also stimulates angiogenesis under certain experimental
conditions (Cao et al., 1998; Witzenbichler et al., 1998).
In contrast to VEGF-A, whose crucial role in tumour
angiogenesis is well established (Ferrara, 1999), very little
is known about the function of VEGF-C in tumour
Vascular endothelial growth factor-C-mediated
lymphangiogenesis promotes tumour metastasis
The EMBO Journal Vol. 20 No. 4 pp. 672±682, 2001
672 ã European Molecular Biology Organization

formation and/or progression. Nonetheless, a correlation
between VEGF-C expression, tumour lymphangiogenesis
and the formation of metastases in regional lymph nodes
has recently been described. Thus, levels of VEGF-C in
primary tumours are signi®cantly correlated with lymph
node metastases in a variety of cancers, including thyroid,
prostate, gastric, colorectal and lung (Bunone et al., 1999;
Tsurusaki et al., 1999; Yonemura et al., 1999; Akagi et al.,
2000; Niki et al., 2000; Ohta et al., 2000). One study has
described a strong correlation between lymphatic vessel
density and VEGF-C expression (Ohta et al., 1999).
However, in this study, no correlation was observed
between lymphatic vessel density and lymph node
metastases. Despite the continuing accumulation of
correlative clinical data, a functional role for VEGF-C in
tumour lymphangiogenesis and/or lymphatic enlargement,
and its role in tumour cell dissemination, have yet to be
demonstrated directly.
In order to test the hypothesis that VEGF-C-induced
lymphangiogenesis can promote tumour metastasis, we
generated transgenic mouse lines in which VEGF-C
expression, driven by the rat insulin promoter, is targeted
to b-cells of the islets of Langerhans, and tested their
capacity to form metastases by crossing them with the
Rip1Tag2 transgenic line, a transgenic mouse model of
non-metastatic b-cell carcinogenesis (Hanahan, 1985).
Results
A full-length human VEGF-C cDNA was cloned between
the rat insulin II gene promoter (Rip) and the SV40 small
T antigen intron and polyadenylation signal (Figure 1A).
Two founder (F0) mice (designated nos 23 and 24) were
identi®ed, which were capable of germline transmission.
RipVEGF-C transgenic mice were viable, similar in size to
wild-type littermates, normoglycaemic and fertile. Mice
derived from F0 no. 23 had a single copy of the
transgene, whereas those derived from F0 no. 24 had
four to six copies integrated in a head-to-tail array
(Figure 1B and data not shown). Pancreas-speci®c
expression of the VEGF-C transgene was con®rmed by
reverse transcription±polymerase chain reaction (RT±
PCR) screening of several organs, using oligonucleotide
primers designed to amplify VEGF-C cDNA of human but
not of mouse origin (Figure 1C and data not shown). In
contrast to wild-type littermates (Figure 2A), immuno-
histochemical analysis of the pancreata of RipVEGF-C
transgenic mice revealed strong VEGF-C staining in the
majority of islet cells (Figure 2B), as would be expected
from the relative abundance of insulin-expressing b-cells
in islets (~80%).
Striking morphological differences were observed
between transgenic and wild-type pancreata. Clearly
demarcated spaces, lined by a single layer of ¯attened
cells, which frequently contained lymphocytes rather than
red blood cells, were observed around the majority of
transgenic islets (Figure 3C and E). These morphological
features were observed in all RipVEGF-C mice killed
between E15.5 and 18 months. These structures, which
were never observed around islets of wild-type mice
(Figure 3A), were identi®ed as lymphatic vessels on the
basis of three separate criteria. First, they were lined by
endothelial cells that were immunoreactive for VEGFR-3
(Figure 2C). Weak VEGFR-3 immunostaining was
also observed within some islets of both wild-type and
transgenic mice (Figure 2C and data not shown).
Secondly, as revealed by transmission electron micro-
scopy (TEM), the basement membrane of these vessels
was either absent or discontinuous, and the endothelial
cells were devoid of the characteristic fenestrations found
in contiguous capillary blood vesselsÐboth hallmarks of
lymphatic endothelium (Figure 4A and B). Thirdly, as
revealed by immunohistochemistry and immuno-EM, they
stained with an antibody to the novel lymphatic endothe-
lial marker LYVE-1 (Banerji et al., 1999; Prevo et al.,
2001) on both their luminal and abluminal surfaces
(Figure 3D, F and 4C), similar to lymphatic vessels in
normal wild-type mouse tissues (Prevo et al., 2001).
Fig. 1. Molecular characterization of RipVEGF-C transgenic mice.
(A) The transgene was constructed by cloning the complete human
VEGF-C cDNA (nucleotides 1±1997; DDBJ/EMBL/GenBank
accession No. X94216) between the ~695 bp BamHI±XbaI fragment
of Rip (Hanahan, 1985) and the SV40 small T antigen intron and
polyadenylation signal. The L-shaped arrow indicates insulin gene
transcription initiation. E, EcoRI restriction sites. (B) Ten micrograms
of genomic DNA from two RipVEGF-C transgenic mice from families
23 and 24 or from a wild-type littermate (wt) (the latter spiked with the
RipVEGF-C transcriptional unit as indicated in picograms) were
digested with EcoRI and analysed by Southern blotting using the SV40
moiety of the transgene as a probe. Single or double asterisks indicate
the 3¢ end of the insertion site in family 23 and 24, respectively.
Markers on the left indicate kilobases. (C) Reverse transcription
products from oligo dT-primed total RNAs from pancreata of
RipVEGF-C transgenic mice from families 23 and 24 or from wild-
type mice (wt) were analysed by PCR using hVEGF-C speci®c primers
or acidic ribosomal phosphoprotein P0 (P0) primers. Where indicated,
RT was omitted, or PCR mix alone (mix) was analysed. Markers on the
left indicate base pairs.
VEGF-C-mediated lymphangiogenesis and metastasis
673