A R T I C L E
Integrin ?4 signaling promotes tumor angiogenesis
Sotiris N. Nikolopoulos,1Pamela Blaikie,1Toshiaki Yoshioka, Wenjun Guo, and Filippo G. Giancotti*
Cell Biology Program, Memorial Sloan-Kettering Cancer Center, New York, New York 10021
1These authors contributed equally to this work.
Mice carrying a targeted deletion of the signaling portion of the integrin ?4 subunit display drastically reduced angiogenesis
in response to bFGF in the Matrigel plug assay and to hypoxia in the retinal neovascularization model. Molecular cytology
indicates that ?6?4 signaling promotes branching of ?4?medium- and small-size vessels into ?4?microvessels without
exerting a direct effect on endothelial cell proliferation or survival. Signaling studies reveal that ?6?4 signaling induces
endothelial cell migration and invasion by promoting nuclear translocation of P-ERK and NF-?B. Upon subcutaneous
implantation of various cancer cells, the mutant mice develop smaller and significantly less vascularized tumors than wild-
type controls. These results provide genetic evidence that ?6?4 signaling promotes the onset of the invasive phase of
pathological angiogenesis and hence identify a novel target for antiangiogenic therapy.
the last phase of the process, the endothelial cells acquire a
quiescent, differentiated phenotype: they deposit a basement
membrane and acquire polarity, coincident with the formation
of a lumen. Pericytes and smooth muscle cells are finally re-
cruited to ensheathe the newly formed vessels. These steps are
repeated in an iterative manner, as mature vessels become
locally destabilized and groups of endothelial cells reacquire an
in Risau, 1997).
Multiple integrins are likely to contribute to tumor angiogen-
esis. The integrins mediate adhesion to the extracellular matrix
and regulate cell survival, proliferation, and migration (Giancotti
and Ruoslahti, 1999; Miranti and Brugge, 2002). Known angio-
genic factors, such as bFGF and VEGF, enhance the expression
and activity of endothelial integrins (Byzova et al., 2000; Klein
et al., 1993), whereas negative regulators of angiogenesis, such
as class 3 semaphorins, promote vascular remodeling by inhib-
iting integrin function (Serini et al., 2003). Studies with adhesion
blocking reagents and knockout mice have implicated ?5?1
and ?v integrins in angiogenesis (Eliceiri and Cheresh, 1999;
Hynes, 2002). However, the mechanisms by which these and
possibly other integrins function in angiogenesis are not clear
(Sheppard, 2002). In addition to playing an adhesive role, the
integrins may play a signaling role during tumor angiogenesis.
Integrin-specific signals impart a stringent control to the action
of receptor tyrosine kinases (RTKs), determining whether cells
proliferate or undergo growth arrest, migrate or remain station-
ary, and live or undergo apoptosis when adhering to a specific
matrix (Giancotti and Tarone, 2003). Hence, integrin signals can
The possibility of ameliorating or even suppressing the progres-
sion of cancer with antiangiogenic drugs has attracted vivid
interest (Hanahan and Folkman, 1996). Studies on transgenic
mouse models of multistage carcinogenesis have revealed the
existence of a discrete angiogenic step. In RIP-Tag and K14-
HPV16 mice, which develop islet cell and epidermal squamous
enhanced angiogenesis precedes thetransition from carcinoma
in situ to invasive carcinoma (Arbeit et al., 1996; Folkman et al.,
1989), and mutations that impair angiogenesis inhibit disease
progression (Bergers et al., 2000; Coussens et al., 2000; Inoue
et al., 2002). Thus, the angiogenic step precedes and is poten-
tially rate limiting for tumor invasion and growth. Small meta-
static lesions co-opt existing host vessels rather than eliciting
angiogenesis, but these vessels eventually regress, and subse-
quent tumor expansion requires robust neoangiogenesis (Ho-
lashet al.,1999).These observationssuggest thatangiogenesis
is required both during initial tumor invasion and growth and
during metastatic spread.
Tumor cells elicit angiogenesis through both enhanced pro-
duction of proangiogenic factors, generally VEGF and bFGF,
and decreased generation of angiogenesis inhibitors (Hanahan
and Folkman, 1996). As a result, host vessels in the vicinity of
an invasive phenotype. Upon detaching from adjacent cells and
penetrating the underlying basement membrane, these cells
proliferate and migrate as cords in the interstitial matrix. During
S I G N I F I C A N C E
To analyze the physiological role of ?6?4 signaling in the absence of the potentially confounding effect of loss of adhesion, we have
generated mice carrying a targeted deletion of the C-terminal signaling portion of the integrin ?4 subunit. Our analysis of these
mice provides genetic evidence that ?6?4 signaling controls pathological angiogenesis by promoting the acquisition of an invasive
phenotype by angiogenic endothelial cells. Since it is known that ?6?4 signaling also promotes carcinoma cell invasion, its inhibition
may be especially beneficial for cancer therapy.
CANCER CELL : NOVEMBER 2004 · VOL. 6 · COPYRIGHT 2004 CELL PRESS471
A R T I C L E
with this hypothesis, studies on signaling molecules that func-
tion downstream of integrins and RTKs, such as focal adhesion
kinase (FAK), Src, Shc, and ILK, have documented a general
role for joint integrin-RTK signaling in angiogenesis (Hood et
al., 2003; Lai and Pawson, 2000; Tan et al., 2004).
ied predominantly in the context of epithelial and tumor biology
studies. ?6?4 signaling proceeds through Src family kinase-
mediated phosphorylation of the large cytoplasmic tail of ?4,
recruitment of Shc, and activation of Ras (Dans et al., 2001;
Gagnoux-Palacios et al., 2003; Mainiero et al., 1995) and PI-
3K (Shaw et al., 1997). In stratified and transitional epithelia,
?6?4 mediates, upon cessation of signaling, assembly of hemi-
desmosomes (Dans et al., 2001; Murgia et al., 1998; Spinardi
et al., 1993). Activation of the EGF-R and Ron RTKs enhances
phosphorylation of ?4, causing disruption of hemidesmosomes
et al., 2003; Trusolino et al., 2001), suggesting that these RTKs
decrease the ability of ?6?4 to mediate stable adhesion but
naling contributes to carcinoma invasion and growth (Gambaletta
et al., 2000; Trusolino et al., 2001). Although it is conceivable
that similar mechanisms underlie the invasive phase of angio-
genesis, the observation that ?4 null embryos do not display
defectivevasculogenesis ordevelopmental angiogenesis(Dow-
ling et al., 1996; van der Neut et al., 1996) has discouraged an
examination of the role of ?6?4 during angiogenesis.
In this study, we have used a genetic approach to examine
the role of ?6?4 signaling in postnatal angiogenesis. Prior stud-
ies had shown that mice carrying a targeted deletion of the
entire cytoplasmic domain of ?4 lack hemidesmosomes and,
like ?4 null mice, die at birth due to extensive blistering of the
skin and upper gastrointestinal tract (Murgia et al., 1998). To
analyze the role of ?6?4 signaling in the absence of the effect
of loss of adhesion strengthening, we have generated mice
carrying a deletion of the C-terminal signaling segment of the
?4 tail. These mice are viable and fertile and do not display
signs of epidermal fragility. Through an analysis of these mice,
we provide evidence that ?6?4 signaling promotes pathological
and tumor angiogenesis.
such a mutation in mice. To construct the vector, we cloned
the sequences encoding the cytoplasmic domain of ?4 up to
amino acid 1355, a stop codon, a SV40 polyadenylation signal,
and a neomycin resistance gene, immediately downstream of
the exon encoding the transmembrane segment of the protein
(Figure 1A). Southern blotting and PCR analysis indicated suc-
cessful introduction of the mutation in mice (Figures 1B and
1C). Analysis of the intercrosses between heterozygous mice
carrying the targeted deletion revealed that the mutation was
transmitted with the expected Mendelian frequency. Both ho-
mozygous and heterozygous ?4 mutant mice were found to be
viable and fertile and to not manifest skin fragility. Histological
analysis of the skin did not reveal any defect in epidermal adhe-
sion to the basement membrane (data not shown). Thus, dele-
tion of the signaling domain of ?4 has no obvious effect on
embryonic and postnatal development.
Immunoprecipitation and FACS analysis on primary kera-
tinocytes from wild-type and mutant mice indicated that the ?4-
1355T subunit associates with ?6 and is expressed at the cell
surface as well as wild-type ?4 (Figures 1D and 1E). To test the
adhesive ability of the mutant integrin, wild-type and mutant
keratinocytes were plated on laminin-5 at 4?C. At this tempera-
ture, the function of ?3?1, which also binds to laminin-5, is
inactivated, and adhesion proceeds only through ?6?4 (Gag-
cytes attached to laminin-5 at 4?C as efficiently as wild-type
keratinocytes, suggesting that the mutant integrin retains intact
ligand binding capacity (Figure 1F). In accordance with the ab-
sence of a skin fragility phenotype, transmission electron mi-
croscopy (EM) revealed that the skin of mutant mice contained
well-structured hemidesmosomes (C. Puri, C. Tacchetti, and
F.G.G., unpublished data). Thus, deletion of the C-terminal sig-
naling domain of ?4 does not affect the ability of ?6?4 to estab-
lish a transmembrane connection between laminin-5 and the
adhesion in vivo.
To examine the effect ofdeletion of the ?4 substrate domain
on signaling, primary keratinocytes isolated from wild-type and
mutant mice were plated on laminin-5 or, as a control, on colla-
gen I in the presence of serum and subjected to immunoblotting
with anti-phospho-ERK and anti-phospho-AKT antibodies. As
shown in Figure 1G, adhesion to laminin-5 induced significant
phosphorylation of ERK in wild-type but not in mutant keratino-
cytes, whereas adhesion to collagen I caused similarly high
activation of ERK in both types of cells. This result is consistent
with the role of the ?4 substrate domain recruitment of Shc and
activation of Ras to ERK signaling (Dans et al., 2001; Mainiero
et al., 1997). In addition, adhesion to laminin-5 led to significant
phosphorylation of AKT in wild-type keratinocytes, but it in-
duced a much more limited effect in mutant keratinocytes (Fig-
ure 1G), in agreement with the hypothesis that the ?4 substrate
domain activates PI-3K to AKT signaling (Shaw et al., 1997).
We concluded that targeted deletion of the C-terminal segment
of the ?4 tail impairs ?6?4-dependent signaling through ERK
bly of hemidesmosomes.
Targeted deletion of the integrin ?4 substrate domain
impairs signaling to ERK and AKT
Two developments made it possible to address the role of ?4
ing effects of loss of adhesion. First, it became clear that the
N-terminal part of the cytoplasmic domain of ?4 to amino acid
1355 is sufficient for interaction with the plakin HD-1/plectin
and hence for association with the keratin cytoskeleton (Schaap-
veld et al., 1998). Second, mapping studies revealed that the
five major tyrosine phosphorylation sites of ?4, including those
involved in the recruitment of Shc and PI-3K, are located in the
C-terminal portion of the ?4 tail, downstream of amino acid
1355 (Dans et al., 2001). We thus reasoned that a deletion of
referred to as “substrate domain”) would suppress ?6?4 signal-
ing without interfering with adhesion strengthening.
?6?4 and its ligand, laminin-5, are expressed
in tumor vasculature
The mutant mice did not display any macroscopic defect sug-
CANCER CELL : NOVEMBER 2004
A R T I C L E
Figure 1. Targeted deletion of the ?4 substrate domain
A: Replacement vector, wild-type locus, and mutant locus are shown above. Solid boxes, exons; TM, exon encoding the transmembrane segment; open
boxes, cDNA sequences; solid asterisk, stop codon; polyA, SV40 polyadenylation signal; neo, neomycin resistance cassette; TK, thymidine kinase; E, EcoRI;
N, NcoI; p5? and p3?, probes for Southern blotting. Wild-type protein (?4 WT) and truncated mutant (?4 1355T) are shown below. White boxes, fibronectin
type-III repeats; open asterisks, tyrosine phosphorylation sites.
B: Southern blotting on genomic DNA from wild-type (?/?), homozygous (?/?), and heterozygous mutant (?/?) mice. Samples were digested with NcoI
and probed with a 500 bp radioactive cDNA probe complementary to sequences in the extracellular domain of ?4.
C: PCR analysis on intercrosses between heterozygous mutant (?/?) mice. The 0.7 kb band originates from the homozygous mutant allele, and the 0.3 kb
band originates from the wild-type allele.
D: Wild-type (?/?) and homozygous mutant (?/?) keratinocytes were immunoprecipitated with the anti-?6 mAb GoH3 and probed with rabbit anti-?4-
exo. Equal amounts of total lysates were directly probed with anti-?4-exo. Arrows point to the expected electrophoretic mobilities of wild-type and mu-
E: Wild-type and mutant keratinocytes were subjected to FACS analysis with mAb 346-11A, which binds to the extracellular domain of mouse ?4.
F: Wild-type (WT) and mutant (1355T) keratinocytes were plated for 1 hr on microtiter plates coated with the indicated amounts of laminin-5 at 4?C. Cell
adhesion to fibronectin at 4?C was negligible (data not shown).
G: Wild-type (WT) and mutant (1355T) keratinocytes were deprived of growth factors, detached, and plated for 2 hr on laminin-5 (Ln-5) or collagen I (Col I)
(top), or they were kept in suspension (S) or plated on laminin-5 for the indicated minutes (bottom). Equal amounts of total proteins were probed with
antibodies to activated ERK (p-ERK) and keratin-5 (top) or to activated AKT (p-AKT) and vinculin (bottom).
?6?4 signaling does not play an essential role during embryonic
vasculogenesis and angiogenesis. This conclusion is consistent
with the observation that ?6?4 is expressed in blood vessels
only after completion of developmental angiogenesis (Hiran et
al., 2003). To examine the potential role of ?6?4 in tumor angio-
genesis, we first studied the expression of ?6?4 in paraffin-
embedded sections of human papillary thyroid carcinoma,
breast adenocarcinoma, prostate carcinoma, and glioblastoma
CANCER CELL : NOVEMBER 2004 473
A R T I C L E
Figure 2. Expression of ?6?4 in tumor vasculature
A: Consecutive paraffin-embedded sections of
the indicated human tumors were subjected to
immunohistochemistry with goat anti-?4 and rat
anti-PECAM-1 or stained with anti-?4 alone.
Scale bar, 10 ?m.
B: Frozen sections of B16F0 melanoma tumors
from wild-type mice were doubly stained with
rat anti-?4 (green) and goat anti-PECAM-1,
mouse anti-smooth muscle ?-actin, or rabbit
anti-Laminin-5 (red) (far left and two center col-
umns) or doubly stained with goat anti-PECAM-1
(green) and rabbit anti-Laminin-5 (red) (far right
column). In the far left column, the arrows point
points to a presumptive lymphatic vessel. In the
center left column, the arrows point to ?4-posi-
tive vessels ensheathed by mural cells, and the
arrowheads point to ?4-positive vessels lacking
mural cells. Asterisks indicate ?4?peripheral
nerves. Scale bar, 40 ?m.
multiforme.Significant levelsof ?6?4were detectedin medium-
and small-size vessels in all these tumors (Figure 2A). Since
tumor cells in breast and prostate cancer samples expressed
high levels of ?6?4, these samples were subjected to anti-
PECAM-1 staining to unequivocally identify tumor vessels (Fig-
To further characterize the expression of ?6?4 during tumor
xenografts.Double stainingwithantibodiesto ?4andPECAM-1
showed that ?6?4 is expressed in these tumors in medium-
(arrows) and small-size vessels, but not in microvessels (Figure
2B). The anti-?4 antibodies also reacted with structures resem-
bling peripheral nerves (Figure 2B, asterisks). Double staining
with antibodies to ?4 and to the neurofilament protein S-100
confirmed the identification of these structures as peripheral
nerves (data not shown). This observation is consistent with the
known expression of ?6?4 in Schwann cells (Einheber et al.,
1993) and the increasing evidence that tumors, including mela-
noma, are innervated (Seifert and Spitznas, 2002). Notably, the
anti-?4 antibodies also stained vessel-like structures that re-
acted withanti-PECAM-1 very weakly(Figure 2B,double arrow-
gesting that ?6?4 is also expressed in tumor lymphatics.
To examine if the expression of ?4 in endothelial cells corre-
lated with the presence of vascular smooth muscle cells, we
subjected the tumor sections to double staining with antibodies
to ?4 and to smooth muscle ?-actin. As shown in Figure 2B,
approximately half of the ?4?vessels were found to be en-
sheathed by smooth muscle cells (arrows), whereas the remain-
der were not (arrowheads), suggesting that endothelial cells do
not express ?4 in response to a signal generated by mural cells.
Significant amounts of laminin-5 were detected in the basement
membrane of both ?4?medium- and small-size vessels and
CANCER CELL : NOVEMBER 2004
A R T I C L E
Figure 3. The ?4 substrate domain promotes an-
giogenesis in response to bFGF in the Matrigel
plug assay and to hypoxia in the retinal neovas-
A: Wild-type (WT) and mutant (1355T) mice were
injected s.c. with Matrigel containing PBS or
bFGF. After 7 days, the plugs were removed and
photographed. The picture shows representa-
tive bFGF-containing plugs excised from wild-
type and mutant mice.
B: Confocal images of representative bFGF-con-
taining plugs excised from FITC-Lectin-injected
wild-type and mutant mice.
injected wild-type and mutant mice were lysed
and subjected to fluorimetry. The graph shows
the mean (?SD) from three experiments (*p ?
D: PBS- and bFGF-containing plugs from wild-
to immunoprecipitation and immunoblotting
with anti-VEGF-R. Scale bar, 100 ?m.
E: P7 wild-type (WT) and mutant (1355T) mice
were exposed to hyperoxia and returned to nor-
with anti-PECAM-1 and counterstained with he-
matoxylin. Scale bar, 50 ?m.
F: Quantification of vascular glomeruli abutting
the limitingmembrane in wild-type (WT)and mu-
tant (1355T) retinas (n ? 5 mice per genotype)
(*p ? 0.004).
binding integrin in these smaller vessels. In fact, the staining
patterns generated by anti-laminin-5 and anti-PECAM-1 anti-
bodies were virtually identical (Figure 2B). Antibodies to ?6 dec-
orated all PECAM-1?vessels, irrespective of ?4 expression,
indicating that the ?4?microvessels express ?6?1 (data not
shown).It ispossiblethat ?6?1oranotherlaminin bindinginteg-
rin, such as ?1?1, mediates endothelial cell adhesion to lami-
nin-5 inmicrovessels. These resultsindicate thatthe endothelial
cells of tumor vessels deposit and organize a laminin-5-rich
basement membrane and, as they mature, attach to it through
than that in wild-type plugs. The medium-size vessels penetrat-
ing into these plugs generated significantly fewer branches than
expected, and these secondary branches only occasionally
formed tertiary ramifications (Figure 3B). Fluorimetry indicated
that the mutant plugs had incorporated approximately 5-fold
less FITC-Lectin than wild-type controls (Figure 3C). In addition,
immunoblotting showed that the mutant plugs contained a
much smaller amount of VEGF-R and, by inference, of angio-
genic endothelial cells than wild-type plugs (Figure 3D). These
observations indicate that loss of ?4 signaling impairs bFGF-
induced angiogenesis to a significant extent.
We examined if ?6?4 signaling is required for angiogenesis
in the retinal neovascularization model. In this model, angiogen-
esis is driven by hypoxia-induced production of VEGF (Shweiki
et al., 1992). P7 mice were maintained in 75% oxygen for 5
days to induce central avascularization in the retina and then
returned to normoxicconditions for 5 additionaldays. Histologi-
cal analysis indicated that numerous vascular glomeruli pene-
trated the inner limiting membrane and abutted in the vitreous
in wild-type mice, whereas the development of these abnormal
vessels was significantly blunted in mutant mice (Figure 3F).
Quantification of the results confirmed that mutant mice have
a significantly reduced angiogenic response to retinal hypoxia
(Figure 3G). Taken together, these results indicate that ?6?4
The ?4 substrate domain promotes
bFGF- and VEGF-mediated angiogenesis
To examine if ?6?4 signaling plays a role in bFGF-induced
angiogenesis, Matrigel plugs containing bFGF were implanted
scopic analysis revealed that the plugs from mutant mice were
the development of vascular ramifications in the plugs, the mice
were injected with an endothelial-specific FITC-labeled Lectin
tree was in mutant plugs much less developed and complex
CANCER CELL : NOVEMBER 2004 475
A R T I C L E
signaling promotes both bFGF- and VEGF-induced angiogen-
?6?4 signaling is not required for endothelial
proliferation or survival
To examine the cellular mechanism by which ?6?4 signaling
regulates angiogenesis, we conducted immunohistochemical
studies on Matrigel plugs from wild-type and mutant mice. Anti-
PECAM-1 staining of frozen sections showed that the angio-
genic vessels of mutant mice penetrated significantly less into
(Figure 4A). The wild-type plugs contained two types of vessels:
small-size vessels, which were detected predominantly at the
the plug. By contrast, the mutant plugs contained almost exclu-
sively peripheral small-size vessels, and these were somewhat
reduced in number as compared to those of wild-type plugs.
those of microvessels did not express the integrin (Figure 4A).
These observations suggest that deletion of the ?4 substrate
domain interferes with the sprouting of ?4?small-size vessels
We next evaluated endothelial cell survival and proliferation
in Matrigel plugs from wild-type and mutant mice. Anti-BrdU
staining revealed that the number of endothelial cells in S phase
was significantly reduced in the plugs from mutant mice. How-
ever, the number of BrdU?nuclei per PECAM-1?vessel was
reduction of BrdU staining in the plugs of mutant mice was
secondary to reduced sprouting, and it was not due to an intrin-
sic proliferative defect (Figure 4B). In addition, the small-size
vessels, which express ?4, displayed very few BrdU?nuclei as
compared to the smaller PECAM-1??4?capillaries, indicating
that ?6?4 is expressed in quiescent vessels (Figure 4C). These
observations suggest that signaling by the ?4 substrate domain
is not required for endothelial proliferation during angiogenesis.
wild-type or mutant plugs (data not shown), suggesting that
?4 signaling is not required for endothelial cell survival during
angiogenesis. Together with the pattern of expression of ?6?4
during angiogenesis, these results suggest that ?6?4 signaling
observations are consistent with the hypothesis that ?6?4 func-
tions at a step of angiogenesis that precedes overt endothelial
cell proliferation and migration in the interstitial matrix.
Figure 4. The ?4 substrate domain promotes branching of quiescent small-
size vessels into proliferative microvessels without exerting a direct effect
on endothelial cell proliferation
A: Sections of bFGF-containing plugs from wild-type and mutant mice were
stained with rat anti-PECAM-1 (left panels). Sections of bFGF-containing
plugs from wild-type mice were subjected to double staining with goat anti-
PECAM-1 (green) and rat anti-?4 (red) (right panels). The margins of the
plugs are marked with a dotted white line. The endothelial cells of small
vessels express ?6?4 while those of capillaries do not. Scale bar, 100 ?m.
B: Sections of bFGF-containing plugs from wild-type and mutant mice in-
jected with BrdU were subjected to double staining with anti-PECAM-1 (red)
and anti-BrdU (green) (left). The graph shows the mean (?SD) number of
BrdU?cells per 100 vessel cross-sections examined (right).
C: Sectionsof bFGF-containingplugs fromwild-type miceinjected withBrdU
were subjected to double staining with anti-?4 (red) and anti-BrdU (green).
The ?4 substrate domain promotes endothelial
migration and invasion
and invasion, we isolated endothelial cells from the lungs of
wild-type and mutant mice. However, both types of cells lost
expression of ?6?4 upon plating in culture (data not shown).
We note that endothelial cells migrating out of human saphe-
nous vein explants also lose expression of ?6?4 (Hiran et al.,
2003). We thus used transient transfection to introduce wild-
type or mutant ?6?4 in human umbilical vein endothelial cells
(HUVECs), as reported previously (Dans et al., 2001). HUVECs,
which expressalmost undetectable levels ofendogenous ?6?4,
were electroporated with plasmids encoding ?6 and either wild-
type ?4 or mutant ?4-1355T and panned on anti-?4 coated
plates to isolate cells expressing comparable levels of recombi-
nant wild-type or mutant ?6?4, respectively (Figure 5A).
To examine the effect of ?6?4 signaling on endothelial cell
migration, parental HUVECs and their derivatives expressing
CANCER CELL : NOVEMBER 2004
A R T I C L E
Figure 5. The ?4 substrate domain promotes en-
dothelial cell migration and invasion in vitro
A: HUVECs (Ctrl) and HUVECs transfected with
?6 in combination with either wild-type ?4 or ?4-
1355T were subjected to immunoblotting with
anti-?4-exo and anti-?-actin.
B: HUVECs (Ctrl) and HUVECs transfected with ?6
in combination with either wild-type ?4 or ?4-
1355T were plated on laminin-5 and induced to
migrate across an artificial wound in response
to bFGF. When indicated, migrating cells were
treated with the NF-?B inhibitor BAY 11-7082
(12.5 ?M) or the MAPK inhibitor PD98059 (50 ?M).
The graph shows the mean percentage of
wound closure at 8 hr (?SD) from three experi-
ments (*p ? 0.001 versus ?4-1355T; p ? 0.002 ver-
sus Ctrl or BAY 11-7082 inhibitor; and p ? 0.003
versus PD98059 inhibitor).
C: HUVECs and the indicated derivatives were
grown on Cytodex-3 beads and placed in colla-
gen gels containing bFGF for 72 hr. The graph
indicates the average number (?SD) of cord-
like structures emanating from each bead (*p ?
0.002 versus Ctrl, and p ? 0.001 versus ?4-1355).
D: HUVECs transfected with ?6 in combination
with either wild-type ?4 or ?4-1355T were in-
duced to migrate across an artificial wound in
response to bFGF for 30 min, fixed, and stained
with antibodies to p65 or P-ERK. Nuclei were
stainedwith DAPI.Arrows pointto nucleishowing
significant nuclear accumulation of P-ERK or
either wild-type or mutant ?6?4 were plated on laminin-5 at
confluency and subjected to in vitro wound assay. Expression
of wild-type ?6?4 increased endothelial cell migration in re-
sponse to bFGF. By contrast, the mutant integrin did not cause
this effect (Figure 5B), indicating that the ?4 substrate domain
promotes endothelial cell migration.
To evaluate the effect of ?6?4 signaling on endothelial cell
invasion, HUVECs expressing wild-type or mutant ?4 were
grown on Cytodex-3 beads and then incubated in collagen gels
containing bFGF. Over a 3 day period, the endothelial cells
expressing wild-type ?4 migrated radially out of the beads and
assembled into cords invading the collagen gel. By contrast,
both control cells and cells expressing mutant ?4 invaded the
collagen gel only to a limited extent (Figure 5C). Taken together,
these observations suggest that signaling by the ?4 substrate
domain promotes endothelial cell migration and invasion, in
accordance with the hypothesis that it controls the onset of the
invasive phase of angiogenesis.
Zahir et al., 2003). To examine the mechanism by which ?6?4
promotes endothelial cell migration,we compared ERK and NF-
?B signaling in HUVECs expressing wild-type or mutant ?6?4
during migration on laminin-5. The cells were plated at conflu-
ency on laminin-5 and, 30 min after wounding, were subjected
to immunofluorescent staining with antibodies to P-ERK and
the p65 subunit of NF-?B. The cells expressing wild-type ?6?4
displayed significant nuclear accumulation of P-ERK and NF-
mutant ?6?4 did not show significant nuclear accumulation of
P-ERK or NF-?B under the same conditions (Figure 5D). These
results suggest that ?6?4 signaling promotes both ERK and
NF-?B signaling in migrating endothelial cells.
To examine the role of ERK and NF-?B signaling in endothe-
lial cell migration, HUVECs expressing wild-type ?6?4 were
subjected to in vitro wound closure assay in the presence of
the MEK inhibitor PD98059 or the NF-?B inhibitor BAY11-072.
These compounds reduced the migration of ?6?4-expressing
HUVECs to levels similar to those displayed by control HUVECs
the ?4 substrate domain promotes endothelial cell migration by
inducing NF-?B and ERK signaling.
To examine if deletion of the ?4 substrate domain impairs
The ?4 substrate domain induces nuclear
accumulation of ERK and NF-?B during endothelial cell
migration in vitro and angiogenesis in vivo
Prior studies had provided evidence that ?6?4 controls ERK
and NF-?B signaling (Mainiero et al., 1997; Santoro et al., 2003;
CANCER CELL : NOVEMBER 2004477
A R T I C L E
signaling in endothelial cells in vivo, sections of Matrigel plugs
of NF-?B. We observed significant levels of P-ERK and p65 in
the nuclei of many endothelial cells of small- and intermediate-
size vessels from wild-type plugs. By contrast, both signaling
molecules were predominantly confined to the cytoplasm in
endothelial cells of similar vessels from mutant plugs (Figure
6A). To confirm this observation, the samples were subjected
to double staining with antibodies to ?4 and to P-ERK or p65
lated in the nuclei of endothelial cells expressing wild-type ?4
but remained largely confined to the cytoplasm in endothelial
cells expressing the mutantintegrin (Figure 6B). Taken together,
these results suggest that the ?4 substrate domain promotes
nuclear translocation of ERK and NF-?B during angiogenesis.
The ?4 substrate domain promotes
To test the role of the ?4 substrate domain in tumor angiogen-
esis,we injectedB16F0melanoma cells,LLC1Lewis lungcarci-
tumors grew in mutant mice to a size significantly smaller than
they did in wild-type mice. Although the 60.5 tumors also ex-
panded less rapidly in mutant mice, the reduction in tumor
growth was in this case smaller (Figure 7A). To compare the
density of microvessels in the tumors grown in wild-type and
mutant mice, we used anti-PECAM-1 staining. The density of
microvessels in each of the four tumors grown subcutaneously
in mutant mice was significantly reduced as compared to that
of tumors grown under identical conditions in wild-type mice
(Figure 7B). This was also true for the 60.5 tumors, which grew
relatively well in mutant mice, suggesting that these tumors are
somewhat less dependent on angiogenesis for growth. Finally,
in the context of other studies, we also examined the effect of
loss of ?4 signaling on angiogenesis in an orthotopic model
of mammary carcinogenesis. In this case, the tumors became
mice (Figures 7A and 7B), suggesting that ?6?4 signaling does
not contribute to tumor angiogenesis in this specific system.
nism, injection protocol, and specific genetic background of
mice—may have influenced the outcome of this specific experi-
ment. Since loss of ?4 signaling inhibited tumor angiogenesis
to a significant extent in four out of five xenotransplantation
models tested, we concluded that ?6?4 signaling plays a sig-
nificant and broad, but perhaps not universal, role in tumor
lature, we injected wild-type and mutant mice bearing B16F0
xenografts with FITC-Lectin. Confocal analysis and 3D recon-
struction confirmed that the defective angiogenic response of
mutant mice to tumors was due to reduced branching (Figure
7C). Two arguments rule out the possibility that immunological
factors contribute to the tumor angiogenesis defect of mutant
mice. First, ?6?4 is not expressed in the immune system. Sec-
were injected in wild-type and mutant mice of pure syngeneic
background, making an immunological response unlikely. In
addition, the reduced angiogenesis in tumors of mutant mice
Figure 6. The ?4 substrate domain promotes nuclear translocation of ERK
and NF-?B in endothelial cells
A: Confocal analysis of nuclear translocation of P-ERK and NF-?B in vivo.
Sections of bFGF-containing plugs from wild-type and mutant mice were
stained with anti-PECAM-1 (green) and anti-P-ERK (red) (left) or anti-
PECAM-1 (green) and anti-p65 (red) (right). Nuclei were stained with DAPI
(blue). Arrowheads point to endothelial cells of wild-type vessels containing
nuclear P-ERK or NF-?B or to endothelial cells of mutant vessels containing
cytoplasmic P-ERK or NF-?B. Scale bar, 4 ?m.
B: Sections from the same plugs were subjected to double staining with
anti-?4 (green) and anti-P-ERK (red) or anti-p65 (red) and counterstaining
with DAPI. Arrowheads are as in A. Scale bar, 10 ?m.
CANCER CELL : NOVEMBER 2004
A R T I C L E
Figure 7. The ?4 substrate domain promotes tu-
A: The B16F0 melanoma, LLC1 Lewis Lung Carci-
noma, B6RV2 lymphoma, and 60.5 fibrosarcoma
cells were injected s.c. in wild-type or ?4 mutant
mice. The YD-Neu mammary carcinoma cells
were injected orthotopically in MMTV-Neu mice
expressing wild-type or mutant ?4 to avoid an
immune response to rat Neu. The graphs show
mean tumor volumes (?SD) after 10 days (60.5),
12 days (B16F0 and LLC1), 13 days (B6RV2), or 20
days (YD-Neu) (*p ? 0.004 in B16F0; p ? 0.09 in
LLC1; p ? 0.01 in B6RV2).
B: Sections of the indicated tumor xenografts
from wild-type and mutant mice were stained
with anti-PECAM-1 antibodies. The graphs show
the average microvessel densities (?SD) in each
tumor. Ten random high-power fields per tumor
section were evaluated. Scalebar, 200 ?m (*p ?
0.01 inB16F0; p? 0.004in LLC1;p ?0.02 inB6RV2;
and p ? 0.005 in 60.5).
C: Confocal images of B16F0 melanoma tumors
excised from FITC-Lectin-injected wild-type and
mutant mice. The graph shows the average
number of branches (?SD) per high-power field
for each tumor (*p ? 0.02). Scale bar, 100 ?m.
does notappear tobe a consequenceof reducedtumor growth,
because the 60.5 tumors grew relatively well but evoked re-
duced angiogenesis in mutant mice. Taken together, these re-
sults identify a role for ?6?4 signaling in tumor angiogenesis.
1999). However, genetic studies suggest more complex roles
(Hynes, 2002). In particular, it is possible that the ?v integrins
may have both positive and negative signaling roles during tu-
matrix during the invasive phase of tumor angiogenesis, but
they induce an active, negative signal at the end of the process,
either upon becoming unligated or upon binding to a known
negative regulator of angiogenesis, such as Thrombospondin,
Tumstatin (a fragment of the ?3 chain of type IV collagen), or
PEX (a fragment of MMP2) (Sheppard, 2002). In this model, the
blocking agents interfere with positive signaling while allowing
negative signaling to occur. Among other integrins involved in
angiogenesis, ?5?1 has attracted considerable interest. Both
knockout studies and antibody-blockingexperiments have indi-
cated that ?5?1 and its ligand, fibronectin, are required for
developmental and pathological angiogenesis (Hynes, 2002).
However, it is not known at what step of angiogenesis ?5?1
nism. Our results indicate that ?6?4 signaling specifically con-
Our results provide genetic evidence that the ?6?4 integrin
promotes tumor angiogenesis—and, presumably, other forms
of pathological angiogenesis—by a signaling mechanism. Im-
?6?4 promotes nuclear translocation of P-ERK and NF-?B and
acquisition of an invasive phenotype at the onset of the invasive
phase of angiogenesis. These results suggest the intriguing
possibility that ?6?4 performs a similar signaling function in
cancer cells and in angiogenic endothelial cells.
In order to design effective anti-integrin drugs for antiangio-
genesis, it is important to understand the mechanisms by which
specific integrins participate in this process. Prior studies with
adhesion-blocking antibodies and RGD-containing peptides
have led to the hypothesis that ?v?3 and av?5 promote tumor
angiogenesis by a signaling mechanism (Eliceiri and Cheresh,
CANCER CELL : NOVEMBER 2004479
A R T I C L E
genesis, these results provide a novel potential target for thera-
The role of ?6?4 in angiogenesis described here is unex-
pected. Prior studies have shown that neither ?6?4 nor its sig-
naling functions are required during developmental angiogen-
esis (Dowling et al., 1996; Murgia et al., 1998; van der Neut et
al., 1996). In addition, basedon the observation that ?6?4 levels
increase during vessel maturation, La Flamme and colleagues
In retrospect, it is not surprising that ?6?4 does not play a
role during developmental angiogenesis, as it is expressed in
endothelial cells only after completion of this process (Hiran et
al., 2003). In addition, our studies do not rule out the possibility
that ?6?4 also contributes to the maturation of adult vessels.
They simply show that that its signaling function contributes to
initiate the invasive phase of angiogenesis. We have demon-
strated this role of ?6?4 signaling in several systems: the Matri-
gel plug assay, the retinal neovascularization model, and four
xenograft models of tumor angiogenesis. This said, increasing
evidence indicates that angiogenesis is driven by different
growth factors and cytokines and, hence, proceeds by partially
distinct mechanisms, depending on developmental stage, tis-
sue, and disease state (LeCouter et al., 2002; Risau, 1997). In
particular, the two major angiogenic growth factors, bFGF and
VEGF, cooperate with distinct ?v integrins to induce angiogen-
esis (Friedlander et al., 1995; Hood et al., 2003). We have ob-
servedthat lossof ?4 signalingdoesnot suppressangiogenesis
in a specific orthotopic model of mammary carcinogenesis.
Thus, although our results suggest that ?6?4 signaling partici-
pates in both bFGF- and VEGF-induced angiogenesis, future
studies will be necessary to examine how general the require-
ment for ?6?4 signaling is during tumor angiogenesis. Since
the angiogenic Id transcription factors induce expression of the
genes encoding ?6?4 and its ligand, laminin-5 (Ruzinova et al.,
2003), it is possible that ?6?4 signaling is especially important
when angiogenesis is driven by Id.
What is the mechanism by which ?6?4 signaling controls
angiogenesis? ?6?4 is expressed during angiogenesis in rela-
tively mature vessels. The endothelial cells of ?4?vessels dis-
play a very low proliferative index, making it unlikely that ?6?4
signaling promotes endothelial proliferation. In addition, the an-
giogenic endothelium of mutant mice does not display evidence
of increased apoptosis, excluding the hypothesis that ?6?4
signaling plays a necessary role in endothelial cell survival. The
plugs suggests that ?6?4 signaling is necessary for the genera-
tion of ?4?sprouts from ?4?vessels, i.e., during the initial step
of the invasive phase of angiogenesis. As expression of wild-
type, but not mutant, ?6?4 promotes endothelial cell migration
and invasion in vitro, we propose that ?6?4 plays a similar role
in vivo (Figure 8). In fact, ?6?4 may play a general role during
branching morphogenesis, as it has been shown that anti-?6?4
antibodies suppress branching of the ureteric bud in the devel-
oping kidney (Zent et al., 2001) and the formation of epithelial
cords by breast epithelial cells embedded in Matrigel (Stahl et
Sprouting angiogenesis is thought to commence with the
acquisition of an invasive phenotype by specific endothelial
cells. The basement membrane underlying these cells is de-
graded as they migrate into the underlying interstitial matrix.
We have observed a defect in nuclear accumulation of ERK and
Figure 8. Hypothetical model of ?6?4 function in angiogenesis
NF-?B in the endothelial cells of small vessels in mutant plugs.
This finding suggests that the ?4 substrate domain regulates
sprouting angiogenesis by promoting nuclear translocation of
key transcription factors and that this event precedes and may
by sprouting endothelial cells. In agreement with this model,
prior studies have shown that angiogenesis requires integrin
signaling to both ERK and NF-?B (Hood et al., 2003; Klein et
al., 2002). Although these transcriptional regulators may play
that they play specific roles in the acquisition of the invasive
phenotype. In particular, we have shown that ?4 signaling pro-
motes nuclear translocation of P-ERK and NF-?B as endothelial
cells commence to migrate on laminin-5 and that these signals
are necessary to promote endothelial cell migration in vitro. In
agreement with this model, it is known that AP-1 and NF-?B
coordinately control the expression of genes involved in cell
migration and invasion (Vincenti and Brinckerhoff, 2002).
The possibility of treatingchronic diseases, such as diabetic
retinopathy, rheumatoid arthritis, and cancer, with antiangio-
genic compounds has attracted considerable interest. Because
?6?4 signaling is not required during development and normal
adult life, compounds blocking ?6?4 signaling may curb patho-
logical angiogenesis without exerting significant toxic effects.
nent oftumor invasion(Hanahan andFolkman, 1996).As cancer
cells invade through the extracellular matrix, they are met by
cords of angiogenic endothelial cells, bringing them nourish-
ment. Since ?6?4 signaling appears to play key roles in both
tumor invasion and tumor angiogenesis, its inhibition may be
especially beneficial for cancer therapy. If validated, this model
would provide a rational basis to future efforts to develop ?6?4
inhibitors for cancer therapy.
Targeted deletion of the ?4 substrate domain
The ClaI/XbaI fragment of mouse ?4 gene was isolated from a 129 Sv library
(Murgia et al., 1998) and subcloned in pBluescript to generate pB/S-m?4-
CANCER CELL : NOVEMBER 2004
A R T I C L E
ClaI/Xba. Site-directed mutagenesis was used to introduce a NheI site within
the sequences encoding the transmembrane domain of pB/S-m?4-ClaI/
frames. PCR was used to introduce a stop codon followed by an XbaI and
an EcoRI site in pcDNA3-?4, thereby generating pcDNA3-?4Cyto-1355T. To
insertthe cDNAfragmentencoding theN-terminalportionof thecytoplasmic
domain of ?4 (amino acids 741–1355) downstream of and in frame with
the exon encoding the transmembrane domain of the protein, a NheI/XbaI
fragment of pcDNA3-?4Cyto-1355T was subcloned in pB/Sm?4-ClaI/XbaI,
and a ClaI/EcoRI fragment of the resulting plasmid was inserted in the
targeting vector previously used to delete the entire the cytoplasmic domain
of ?4 (Murgia et al., 1998). The resulting replacement vector, which carried
a left arm of 5 kb and a right arm of 3.8 kb, was linearized and electroporated
in ES cells. Positively transfected cells that had undergone homologous
recombination were selected in 0.5 mg/ml G418 and 0.2 mM Gancyclovir
and identified by Southern blotting. Two distinct ES cell lines were found to
carry the expected mutation, and both were injected into blastocyst-stage
C57BL/6 mouse embryos. The embryos were then transplanted into the
uteri of pseudopregnant C57BL/6 mice. Extensively chimeric mice derived
from both lines were crossed to C57BL/6 females. Heterozygous offspring
were used to generate mice homozygous for the targeted deletion. Mice
were genotyped by PCR using tail genomic DNA. The following primers were
used for amplification: 5?-GGAAATAGCAGAGCAGGATAC-3? (wild-type),
5?-CTCGTGCTTTACGGTATCGC-3? (recombinant), 5?-CTCGGTTGCAGCA
AGGAAG-3? (common). For Southern blotting, tail genomic DNA was di-
gested with NcoI and, after agarose gel electrophoresis and transfer to
a nylon membrane, was hybridized to a 500 bp radioactive cDNA probe
complementary to sequences in the extracellular domain of ?4, as described
previously (Murgia et al., 1998). Except when indicated, the experiments
were conducted on mice of mixed genetic background.
mutant ?4-1355T were plated on dishes coated with laminin-5, grown until
confluent, and starved. Monolayers were scratched with a P200 pipette tip
and incubated in the presence of serum and 20 ng/ml bFGF for 18 hr. Wound
closure was monitored by digital photography. To monitor ERK and NF-?B
signaling during migration, control and transfected HUVECs were subjected
to invitro woundingfor 30min, fixedwith 3.7% formaldehyde, andsubjected
control and transfected HUVECs were grown on Biosolin Cytodex-3 micro-
gels (3D Collagen Cell Culture Kit; Chemicon). The gels were overlaid with
DMEM with 10% fetal bovine serum, 2 mmol glutamine, and 10 ng/ml bFGF.
HUVEC invasion was quantified 72 hr later by counting the average number
of capillary-like structures per microcarrier bead.
Immunofluorescence microscopy and immunohistochemistry
Tissues and plugs were embedded in paraffin or snap frozen in OCT com-
pound (Tissue-Tek). Paraffin-embedded sections were stained with hema-
toxylin and eosin or subjected to immunoperoxidase staining with the indi-
cated antibodies using the ABC Staining Kit (Vector Laboratories). Frozen
sections were subjected to immunofluorescent staining with the indicated
antibodies. To measure cell proliferation in vivo, mice were injected i.v. with
5 ?M BrdU/100 g body weight and sacrificed 1 hr later. Cryostat as well
as paraffin-embedded sections of Matrigel or tumors were subjected to
immunofluorescent or immunohistochemical staining with anti-BrdU anti-
bodies (BrdU Labeling and Detection Kit I; Roche). To estimate cell death
in vivo, TUNEL assays were performed on paraffin-embedded sections (In
Situ Cell Death Detection Kit; Roche).
Matrigel plug assay
Eight-week-old micewere injecteds.c. with 400 ?l ofgrowth factor-depleted
Matrigel (BD Biosciences) supplemented with 400 ng/ml bFGF and 1 ?g/ml
heparin sulfate and sacrificed7 days later (Passaniti et al.,1992). To visualize
angiogenesis, the mice were injected intravenously with 20 ?g of FITC-
Isolectin B4 (Vector Labs) 30 min before harvesting the plugs. Fluorescently
labeled vessels were examined by confocal microscopy. To quantify angio-
genesis, FITC-Lectin-containing plugs were homogenized in RIPA buffer
analysis. Alternatively, plug lysates were immunoprecipitated and subjected
to immunoblotting with anti-VEGF-R2 antibodies. Each experimental group
consisted of five mice. Each mouse was injected with Matrigel alone or
Matrigel supplemented with bFGF and heparin sulfate. Experiments were
repeated three times.
Cells, antibodies, and other reagents
Keratinocytes were isolated from the skin of newborn mice and grown on
collagen I-coated plates in EMEM.06 with 8% Chelex-treated FBS, 2 ng/ml
EGF, and0.06 mM CaCl2(Hager et al.,1999). Primary HUVECswere cultured
on gelatin-coated dishes (Klein et al., 2002). Rat mAbs to ?4 (346-11A), ?6
(GoH3), and PECAM-1 (MEC 13.3) were from Pharmingen. Goat anti-
PECAM-1 (M-20) and -?4 (C-20) and rabbit anti-VEGF-R2 (C-20) and -NF-
?B p65 (C-20) were from Santa Cruz. Rabbit anti-P-ERK and -P-AKT were
from Cell Signaling, and anti-keratin-5 (AF 138) was from Babco. Mouse
mAbs to smooth muscle ?-actin (clone 1A4) and to ?-actin (clone AC-74)
were fromSigma, and those toNF-?B p65 (clone 2A12A7)were from Zymed.
Affinity-purified rabbit antibodies to the LE4-6 modules of the mouse laminin
?2 chain (Sasaki et al., 2001), mouse mAb 3E1 to ?4, and rabbit anti-?4-
exo serum to a GST fusion protein comprising the N-terminal domain of ?4
(Mainiero et al., 1997) were described previously. FITC- and Cy3-conjugated
inin-5 matrices were prepared as previously described (Spinardi et al., 1995).
Purified laminin-5 was from Chemicon. Human fibronectin and rat tail colla-
gen type I were from Collaborative Research. FITC-Lectin (isolectin B4)
was from Vector Laboratories. The MEK inhibitor PD98059 and the NF-?B
inhibitor BAY 11-7082 were from Calbiochem.
Retinal hypoxia model
P7 mice were exposed to 75% oxygen for 5 days and then returned to
normoxic conditions for 5 days. Mice of the same age kept in normal air
were used as controls. Eyes were fixed in 4% paraformaldehyde, embedded
bycounting thenumberof PECAM-1-positiveglomerulipenetrating theinner
For immunoprecipitation and immunoblotting analyses, keratinocytes were
lysed in RIPA buffer with 10 mM EDTA and protease inhibitors. Equivalent
amounts of total proteins were immunoprecipitated with mAb GoH3 and
subjected to immunoblotting with anti-?4-exo or directly subjected to immu-
noblotting. FACS analysis and adhesion assays were performed as pre-
viously described (Murgia et al., 1998). For signaling studies, the keratino-
cytes were plated on laminin-5 or collagen I for the indicated times, lysed,
and subjected to immunoblotting with anti-phospho-ERK and anti-phos-
Six-month-old mice were injected s.c. with 106tumor cells per flank. B6RV2
human lymphoma cells, B16F0 mouse melanoma cells, and LLC1 Lewis
Lung carcinoma cells were injected in wild-type and mutant mice of mixed
genetic background. The 60.5 fibrosarcoma cells, which are derived from
129 Sv mice (Pozzi et al., 2000), were injected in syngeneic wild-type and
mutant mice of pure background. The YD-Neu mouse mammary carcinoma
cells, which were generated by introducing rat Neu in YD cells (Dankort et
al., 2001), were implanted orthotopically at 5 ? 106in Matrigel diluted 1:1
in PBS. To avoid an immune response to rat Neu, the cells were injected in
MMTV-Neu transgenic mice expressing either wild-type or mutant ?4, as
these mice are tolerant to rat Neu. These mice had been backcrossed into
an FVB/n background (W.G. and F.G.G., unpublished data). The tumors
were excised after the indicated number of days. Final tumor dimensions
were measured by caliper.
Endothelial cell studies
HUVECs were electroporated with equimolar amounts of plasmids encoding
?6 and either ?4 or ?4-1355T (Dans et al., 2001), deprived of growth factors
cells were washed with PBS and recovered by trypsin-EDTA treatment. For
in vitro wound assays, equal numbers of cells expressing wild-type ?4 or
CANCER CELL : NOVEMBER 2004481
A R T I C L E
Giancotti, F.G., and Ruoslahti, E. (1999). Integrin signaling. Science 285,
and Knockout Mouse Facility and the Molecular Cytology Facility of the
Memorial Sloan-Kettering Cancer Center (MSKCC) for help; and M. Dans,
S. Klein, D. Lyden, and A. Petridis for help and discussions. This work was
and P30 CA08748 (to MSKCC).
Giancotti, F.G., and Tarone, G. (2003). Positional control of cell fate through
joint integrin-receptor protein tyrosine kinase signaling. Annu. Rev. Cell Dev.
Biol. 19, 173–206.
Hager, B., Bickenbach, J.R., and Fleckman, P. (1999). Long-term culture of
murine epidermal keratinocytes. J. Invest. Dermatol. 112, 971–976.
Hanahan, D. (1985). Heritable formation of pancreatic ?-cell tumours in
transgenic mice expressing recombinant insulin/simian virus 40 oncogenes.
Nature 315, 115–122.
Received: April 16, 2004
Revised: July 14, 2004
Accepted: September 10, 2004
Published: November 15, 2004
Hanahan, D., and Folkman, J. (1996). Patterns and emerging mechanisms
of the angiogenic switch during tumorigenesis. Cell 86, 353–364.
Hiran, T.S., Mazurkiewicz, J.E., Kreienberg, P., Rice, F.L., and LaFlamme,
S.E. (2003). Endothelial expression of the ?6?4 integrin is negatively regu-
lated during angiogenesis. J. Cell Sci. 116, 3771–3781.
Holash, J., Maisonpierre, P.C., Compton, D., Boland, P., Alexander, C.R.,
Zagzag, D., Yancopoulos, G.D., and Wiegand, S.J. (1999). Vessel cooption,
regression, and growth in tumors mediated by angiopoietins and VEGF.
Science 284, 1994–1998.
Arbeit, J.M., Munger, K., Howley, P.M., and Hanahan, D. (1994). Progressive
squamous epithelial neoplasia in K14–HPV16 transgenic mice. J. Virol. 68,
Arbeit, J.M., Olson, D.C., and Hanahan, D. (1996). Upregulation of FGFs
and their receptors during multi-stage epidermal carcinogenesis in K14-
HPV16 transgenic mice. Oncogene 13, 1847–1857.
Hood, J.D., Frausto, R., Kiosses, W.B., Schwartz, M.A., and Cheresh, D.A.
(2003). Differential ?v integrin-mediated Ras-ERK signaling during two path-
ways of angiogenesis. J. Cell Biol. 162, 933–943.
Bergers, G., Brekken, R., McMahon, G., Vu, T.H., Itoh, T., Tamaki, K., Tan-
zawa, K., Thorpe, P., Itohara, S., Werb, Z., and Hanahan, D. (2000). MMP-9
triggers the angiogenic switch during carcinogenesis. Nat. Cell Biol. 2, 737–
Hynes, R.O. (2002). A reevaluation of integrins as regulators of angiogenesis.
Nat. Med. 8, 918–921.
Inoue, M., Hager, J.H., Ferrara, N., Gerber, H.P., and Hanahan, D. (2002).
VEGF-A has a critical, nonredundant role in angiogenic switching and pan-
creatic ? cell carcinogenesis. Cancer Cell 1, 193–202.
Byzova, T.V., Goldman, C.K., Pampori, N., Thomas, K.A., Bett, A., Shattil,
to VEGF: activation of the integrins. Mol. Cell 6, 851–860.
Klein, S., Giancotti, F.G., Presta, M., Albelda, S.M., Buck, C.A., and Rifkin,
D.B. (1993). Basic fibroblast growth factor modulates integrin expression in
microvascular endothelial cells. Mol. Biol. Cell 4, 973–982.
Coussens, L.M., Tinkle, C.L., Hanahan, D., and Werb, Z. (2000). MMP-9
supplied by bone marrow-derived cells contributes to skin carcinogenesis.
Cell 103, 481–490.
Klein, S., de Fougerolles, A.R., Blaikie, P., Khan, L., Pepe, A., Green, C.D.,
Koteliansky, V., and Giancotti, F.G. (2002). ?5?1 integrin activates an NF-
?B-dependent program of gene expression important for angiogenesis and
inflammation. Mol. Cell. Biol. 22, 5912–5922.
Dankort, D., Maslikowski, B., Warner, N., Kanno, N., Kim, H., Wang, Z.,
Moran, M.F., Oshima, R.G., Cardiff, R.D., and Muller, W.J. (2001). Grb2 and
Shc adapter proteins play distinct roles in Neu (ErbB2)-induced mammary
tumorigenesis: implications for human breast cancer. Mol. Cell. Biol. 21,
sensitizes cardiovascular signaling in the mouse embryo. Genes Dev. 14,
Dans, M., Gagnoux-Palacios, L., Blaikie, P., Klein, S., Mariotti, A., and Gian-
cotti, F.G. (2001). Tyrosine phosphorylation of the ?4 integrin cytoplasmic
domain mediates Shc signaling to ERK and antagonizes formation of hemi-
desmosomes. J. Biol. Chem. 276, 1494–1502.
LeCouter, J., Lin, R., and Ferrara, N. (2002). Endocrine gland-derived VEGF
and the emerging hypothesis of organ-specific regulation of angiogenesis.
Nat. Med. 8, 913–917.
Dowling, J., Yu, Q.C., and Fuchs, E. (1996). ?4 integrin is required for hemi-
desmosome formation, cell adhesion and cell survival. J. Cell Biol. 134,
Mainiero, F., Pepe, A., Wary, K.K., Spinardi, L., Mohammadi, M., Schles-
singer, J., and Giancotti, F.G. (1995). Signal transduction by the ?6?4 integ-
with the cytoskeleton of hemidesmosomes. EMBO J. 14, 4470–4481.
tion of Schwann cell integrin expression suggests a role for ?6?4 in myelina-
tion. J. Cell Biol. 123, 1223–1236.
Mainiero, F., Murgia, C., Wary, K.K., Curatola, A.M., Pepe, A., Blumemberg,
M., Westwick, J.K., Der, C.J., and Giancotti, F.G. (1997). The coupling of
?6?4 integrin to Ras-MAP kinase pathways mediated by Shc controls kera-
tinocyte proliferation. EMBO J. 16, 2365–2375.
Eliceiri, B.P., and Cheresh, D.A. (1999). The role of ?v integrins during angio-
genesis: insights into potential mechanisms of action and clinical develop-
ment. J. Clin. Invest. 103, 1227–1230.
Miranti, C.K., and Brugge, J.S. (2002). Sensing the environment: a historical
perspective on integrin signal transduction. Nat. Cell Biol. 4, E83–E90.
Folkman, J., Watson, K., Ingber, D., and Hanahan, D. (1989). Induction of
angiogenesis during the transition from hyperplasia to neoplasia. Nature
Murgia, C., Blaikie, P., Kim, N., Dans, M., Petrie, H.T., and Giancotti, F.G.
(1998). Cell cycle and adhesion defects in mice carrying a targeted deletion
of the integrin ?4 cytoplasmic domain. EMBO J. 17, 3940–3951.
Friedlander, M., Brooks, P.C., Shaffer, R.W., Kincaid, C.M., Varner, J.A., and
Cheresh, D.A. (1995). Definition of two angiogenic pathways by distinct ?v
integrins. Science 270, 1500–1502.
Passaniti, A., Taylor, R.M., Pili, R., Guo, Y., Long, P.V., Haney, J.A., Pauly,
R.R., Grant, D.S., and Martin, G.R. (1992). A simple, quantitative method
for assessing angiogenesis and antiangiogenic agents using reconstituted
basement membrane, heparin, and fibroblast growth factor. Lab. Invest. 67,
Gagnoux-Palacios, L., Dans, M., van’t Hof, W., Mariotti, A., Pepe, A., Mene-
guzzi, G., Resh, M.D., and Giancotti, F.G. (2003). Compartmentalization of
integrin ?6?4 signaling in lipid rafts. J. Cell Biol. 162, 1189–1196.
Gambaletta, D., Marchetti, A., Benedetti, L., Mercurio, A.M., Sacchi, A., and
Falcioni, R. (2000). Cooperative signaling between ?6?4 integrin and ErbB-2
receptor is required to promote PI-3K-dependent invasion. J. Biol. Chem.
Pozzi, A., Moberg, P.E., Miles, L.A., Wagner, S., Soloway, P., and Gardner,
?1 knockout mice cause reduced tumor vascularization. Proc. Natl. Acad.
Sci. USA 97, 2202–2207.
CANCER CELL : NOVEMBER 2004
A R T I C L E
Risau, W. (1997). Mechanisms of angiogenesis. Nature 386, 671–674. Spinardi, L., Ren, Y.L., Sanders, R., and Giancotti, F.G. (1993). The ?4
subunit cytoplasmic domain mediates the interaction of ?6?4 integrin with
the cytoskeleton of hemidesmosomes. Mol. Biol. Cell 4, 871–884.
Ruzinova, M.B., Schoer, R.A., Gerald, W., Egan, J.E., Pandolfi, P.P., Rafii,
S., Manova, K., Mittal, V., and Benezra, R. (2003). Effect of angiogenesis
inhibition by Idloss and the contributionof bone-marrow-derived endothelial
cells in spontaneous murine tumors. Cancer Cell 4, 277–289.
Spinardi, L., Einheber, S., Cullen, T., Milner, T.A., and Giancotti, F.G. (1995).
A recombinant tail-less integrin ?4 subunit disrupts hemidesmosomes but
does not suppress ?6?4-mediated cell adhesion to laminins. J. Cell Biol.
Santoro, M.M., Gaudino, G., and Marchisio, P.C. (2003). The MSP receptor
regulates ?6?4 and ?3?1 integrins via 14-3-3 proteins in keratinocyte migra-
tion. Dev. Cell 5, 257–271.
Stahl, S., Weitzman, S., and Jones, J.C. (1997). The role of laminin-5 and
its receptors in mammary epithelial cell branching morphogenesis. J. Cell
Sci. 110, 55–63.
Sasaki, T., Gohring, W., Mann, K., Brakebusch, C., Yamada, Y., Fassler, R.,
and Timpl, R. (2001). Short arm region of laminin-5 ?2 chain: structure,
mechanism of processing and binding to heparin and proteins. J. Mol. Biol.
Tan, C., Cruet-Hennequart, S., Troussard, A., Fazli, L., Costello, P., Sutton,
K., Wheeler, J., Gleave, M., Sanghera, J., and Dedhar, S. (2004). Regulation
of tumor angiogenesis by integrin-linked kinase (ILK). Cancer Cell 5, 79–90.
Schaapveld, R.Q., Borradori, L., Geerts, D., van Leusden, M.R., Kuikman,
I., Nievers, M.G., Niessen, C.M., Steenbergen, R.D., Snijders, P.J., and Son-
nenberg, A. (1998). Hemidesmosome formation is initiated by the ?4 integrin
subunit, requires complex formation of ?4 and HD1/plectin, and involves a
direct interaction between ?4 and the bullous pemphigoid antigen 180. J.
Cell Biol. 142, 271–284.
Trusolino, L., Bertotti, A., and Comoglio, P.M. (2001). A signaling adapter
function for ?6?4 integrin in the control of HGF-dependent invasive growth.
Cell 107, 643–654.
A. (1996). Epithelial detachment due to absence of hemidesmosomes in
integrin ?4 null mice. Nat. Genet. 13, 366–369.
Seifert, P., and Spitznas, M. (2002). Axons in human choroidal melanoma
suggest the participation of nerves in the control of these tumors. Am. J.
Ophthalmol. 133, 711–713.
Vincenti, M.P., and Brinckerhoff, C.E. (2002). Transcriptional regulation of
collagenase (MMP-1, MMP-13) genes in arthritis: integration of complex
signaling pathways for the recruitment of gene-specific transcription factors.
Arthritis Res. 4, 157–164.
Serini, G., Valdembri, D., Zanivan, S., Morterra, G., Burkhardt, C., Caccavari,
F., Zammataro, L., Primo, L., Tamagnone, L.,Logan, M., et al. (2003). Class 3
semaphorins control vascular morphogenesis by inhibiting integrin function.
Nature 424, 391–397.
Xia, Y., Gil, S.G., and Carter, W.G. (1996). Anchorage mediated by integrin
?6?4 to laminin 5 (epiligrin) regulates tyrosine phosphorylation of a mem-
brane-associated 80-kD protein. J. Cell Biol. 132, 727–740.
Shaw, L.M., Rabinovitz, I., Wang, H.H., Toker, A., and Mercurio, A.M. (1997).
Activation of PI-3K by the ?6?4 integrin promotes carcinoma invasion. Cell
Zahir, N., Lakins, J.N., Russell, A., Ming, W., Chatterjee, C., Rozenberg, G.I.,
Marinkovich, M.P., and Weaver, V.M. (2003). Autocrine laminin-5 ligates
?6?4 integrin and activates RAC and NF?B to mediate anchorage-indepen-
dent survival of mammary tumors. J. Cell Biol. 163, 1397–1407.
Sheppard, D. (2002). Endothelial integrins and angiogenesis: not so simple
anymore. J. Clin. Invest. 110, 913–914.
Zent, R., Bush, K.T., Pohl, M.L., Quaranta, V., Koshikawa, N., Wang, Z.,
Kreidberg, J.A., Sakurai, H., Stuart, R.O., and Nigam, S.K. (2001). Involve-
ment of laminin binding integrins and laminin-5 in branching morphogenesis
of the ureteric bud during kidney development. Dev. Biol. 238, 289–302.
Shweiki, D., Itin, A., Soffer, D., and Keshet, E. (1992). Vascular endothelial
growth factor induced by hypoxia may mediate hypoxia-initiated angiogen-
esis. Nature 359, 843–845.
CANCER CELL : NOVEMBER 2004 483